diff --git a/.gitattributes b/.gitattributes index 0ebb283a7d2612e49fcce1d59e9e2d5f5fdee86e..0df01b536e59269adbadbc0741f9e51406c20997 100644 --- a/.gitattributes +++ b/.gitattributes @@ -1325,3 +1325,130 @@ Networking/XML/O'Reilly[[:space:]]XSLT[[:space:]]Cookbook.chm filter=lfs diff=lf Networking/XML/OReilly.XSLT.2nd.Edition.Jul.2008.pdf filter=lfs diff=lfs merge=lfs -text Networking/XML/XML[[:space:]]for[[:space:]]Dummies[[:space:]]4th[[:space:]]Edition.pdf filter=lfs diff=lfs merge=lfs -text Networking/XML/xml_by_example.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/BSD/No[[:space:]]Starch[[:space:]]Press[[:space:]]-[[:space:]]Absolute[[:space:]]FreeBSD[[:space:]]The[[:space:]]Complete[[:space:]]Guide[[:space:]]to[[:space:]]FreeBSD[[:space:]]2nd[[:space:]]Edition.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/BSD/No[[:space:]]Starch[[:space:]]Press[[:space:]]-[[:space:]]Absolute[[:space:]]OpenBSD[[:space:]]UNIX[[:space:]]For[[:space:]]The[[:space:]]Practical[[:space:]]Paranoid.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/BSD/O'Reilly[[:space:]]Mastering[[:space:]]FreeBSD[[:space:]]and[[:space:]]OpenBSD[[:space:]]Security.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Assembly[[:space:]]Language[[:space:]]Step-by-Step[[:space:]]-[[:space:]]Programming[[:space:]]with[[:space:]]Linux,[[:space:]]3rd[[:space:]]Edition.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Cluster/Linux[[:space:]]Enterprise[[:space:]]Cluster.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Cluster/OReilly[[:space:]]High[[:space:]]Performance[[:space:]]Linux[[:space:]]Clusters[[:space:]]With[[:space:]]Oscar[[:space:]]Rocks[[:space:]]openmosix[[:space:]]And[[:space:]]Mpi.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Core[[:space:]][[:space:]]Linux[[:space:]]Concepts[[:space:]](Kernel,Networking,FineTuning,Device[[:space:]]Drivers,Thin[[:space:]]Client)/Linux[[:space:]]Thin[[:space:]]Client[[:space:]]Networks[[:space:]]Design[[:space:]]and[[:space:]]Deployment.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Core[[:space:]][[:space:]]Linux[[:space:]]Concepts[[:space:]](Kernel,Networking,FineTuning,Device[[:space:]]Drivers,Thin[[:space:]]Client)/O'Reilly[[:space:]]-[[:space:]]Linux[[:space:]]Device[[:space:]]Drivers[[:space:]]2nd[[:space:]]Edition.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Core[[:space:]][[:space:]]Linux[[:space:]]Concepts[[:space:]](Kernel,Networking,FineTuning,Device[[:space:]]Drivers,Thin[[:space:]]Client)/shell_scripting.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Core[[:space:]][[:space:]]Linux[[:space:]]Concepts[[:space:]](Kernel,Networking,FineTuning,Device[[:space:]]Drivers,Thin[[:space:]]Client)/Understanding[[:space:]]Linux[[:space:]]Network[[:space:]]Internals.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/DNS/DNS[[:space:]]and[[:space:]]BIND,[[:space:]]5th[[:space:]]ed.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/DNS/Dns[[:space:]]in[[:space:]]Action[[:space:]]A[[:space:]]Detailed[[:space:]]and[[:space:]]Practical[[:space:]]Guide[[:space:]]to[[:space:]]Dns.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/DNS/Pro[[:space:]]DNS[[:space:]]and[[:space:]]BIND.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Essentials[[:space:]]Books/Backup[[:space:]]And[[:space:]]Recovery.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Essentials[[:space:]]Books/Ebook[[:space:]]Teach[[:space:]]Yourself[[:space:]]Linux[[:space:]]In[[:space:]]24[[:space:]]Hours[[:space:]]Sharereactor.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Essentials[[:space:]]Books/Hacking[[:space:]]Linux[[:space:]]Exposed.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Essentials[[:space:]]Books/How[[:space:]]Linux[[:space:]]Works[[:space:]]-[[:space:]]What[[:space:]]Every[[:space:]]Super-User[[:space:]]Should[[:space:]]Know.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Essentials[[:space:]]Books/Linux[[:space:]]Annoyances[[:space:]]For[[:space:]]Geeks.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Essentials[[:space:]]Books/Linux[[:space:]]Bible[[:space:]]2008[[:space:]]Edition.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Essentials[[:space:]]Books/Linux[[:space:]]The[[:space:]]Complete[[:space:]]Reference.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Essentials[[:space:]]Books/Linux[[:space:]]Timesaving[[:space:]]Techniques[[:space:]]for[[:space:]]Dummies.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Essentials[[:space:]]Books/Linux[[:space:]]Troubleshooting[[:space:]]For[[:space:]]System[[:space:]]Administrators[[:space:]]And[[:space:]]Power[[:space:]]Users[[:space:]].chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Essentials[[:space:]]Books/Linux+[[:space:]]Certification_Bible.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Essentials[[:space:]]Books/O'Reilly[[:space:]]-[[:space:]]Essential[[:space:]]System[[:space:]]Administration[[:space:]]3rd[[:space:]]Edition.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Essentials[[:space:]]Books/O'Reilly[[:space:]]-[[:space:]]Learning[[:space:]]Red[[:space:]]Hat[[:space:]]Enterprise[[:space:]]Linux[[:space:]]&[[:space:]]Fedora[[:space:]]4th[[:space:]]Edition.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Essentials[[:space:]]Books/O'Reilly[[:space:]]-[[:space:]]Learning[[:space:]]the[[:space:]]Vi[[:space:]]Editor[[:space:]]6th[[:space:]]Edition.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Essentials[[:space:]]Books/O'Reilly[[:space:]]-[[:space:]]UNIX[[:space:]]in[[:space:]]a[[:space:]]Nutshell[[:space:]]3rd[[:space:]]Edition.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Essentials[[:space:]]Books/O'Reilly[[:space:]]-[[:space:]]UNIX[[:space:]]Power[[:space:]]Tools[[:space:]]3rd[[:space:]]Edition.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Essentials[[:space:]]Books/OReilly[[:space:]]Linux[[:space:]]System[[:space:]]Administration.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Essentials[[:space:]]Books/Red[[:space:]]Hat[[:space:]]Linux[[:space:]]Security[[:space:]]And[[:space:]]Optimization.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Essentials[[:space:]]Books/User[[:space:]]Mode[[:space:]]Linux.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Essentials[[:space:]]Networking[[:space:]](VPN,SNMP,Load[[:space:]]Balancing)/O'Reilly[[:space:]]-[[:space:]]Server[[:space:]]Load[[:space:]]Balancing.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Essentials[[:space:]]Networking[[:space:]](VPN,SNMP,Load[[:space:]]Balancing)/OpenVPN[[:space:]]-[[:space:]]Building[[:space:]]And[[:space:]]Integrating[[:space:]]Virtual[[:space:]]Private[[:space:]]Networks.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/For[[:space:]]Dummies[[:space:]]-[[:space:]]Linux[[:space:]]All[[:space:]]in[[:space:]]One[[:space:]]Desk[[:space:]]Reference.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/For[[:space:]]Dummies[[:space:]]-[[:space:]]Linux[[:space:]]Timesaving[[:space:]]Techniques.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/IDS[[:space:]]Firewall[[:space:]]&[[:space:]]Security/Configuring[[:space:]]IPCop[[:space:]]Firewalls[[:space:]]-[[:space:]]Closing[[:space:]]Borders[[:space:]]With[[:space:]]Open[[:space:]]Source.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/IDS[[:space:]]Firewall[[:space:]]&[[:space:]]Security/O'Reilly[[:space:]][[:space:]]Building[[:space:]]Internet[[:space:]]Firewalls[[:space:]]2nd[[:space:]]Edition.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/IDS[[:space:]]Firewall[[:space:]]&[[:space:]]Security/O'Reilly[[:space:]]-[[:space:]]Building[[:space:]]Secure[[:space:]]Servers[[:space:]]with[[:space:]]Linux.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/IDS[[:space:]]Firewall[[:space:]]&[[:space:]]Security/O'Reilly[[:space:]]-[[:space:]]Building[[:space:]]Secure[[:space:]]Servers[[:space:]]with[[:space:]]Linux.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/IDS[[:space:]]Firewall[[:space:]]&[[:space:]]Security/O'Reilly[[:space:]]-[[:space:]]Network[[:space:]]Security[[:space:]]with[[:space:]]OpenSSL.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/IDS[[:space:]]Firewall[[:space:]]&[[:space:]]Security/Security[[:space:]]Warrior.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/IDS[[:space:]]Firewall[[:space:]]&[[:space:]]Security/Snort[[:space:]]For[[:space:]]Dummies.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/IDS[[:space:]]Firewall[[:space:]]&[[:space:]]Security/The[[:space:]]Best[[:space:]]Damn[[:space:]]Firewall[[:space:]]Book[[:space:]]Period[[:space:]].pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/IDS[[:space:]]Firewall[[:space:]]&[[:space:]]Security/The[[:space:]]Tao[[:space:]]Of[[:space:]]Network[[:space:]]Security[[:space:]]Monitoring[[:space:]]-[[:space:]]Beyond[[:space:]]Intrusion[[:space:]]Detection.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/LDAP/O'Reilly[[:space:]]-[[:space:]]LDAP[[:space:]]System[[:space:]]Administration.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Linux[[:space:]]All-In-One[[:space:]]Desk[[:space:]]Reference[[:space:]]For[[:space:]]Dummies[[:space:]](Feb[[:space:]]2005).pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Linux[[:space:]]For[[:space:]]Dummies,[[:space:]]6th[[:space:]]Edition[[:space:]](Feb[[:space:]]2005).pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Mail[[:space:]]Servers/No[[:space:]]Starch[[:space:]]Press[[:space:]]-[[:space:]]The[[:space:]]Book[[:space:]]of[[:space:]]Postfix.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Mail[[:space:]]Servers/O'Reilly[[:space:]]-[[:space:]]Postfix[[:space:]]The[[:space:]]Definitive[[:space:]]Guide.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Mail[[:space:]]Servers/OReilly[[:space:]]sendmail[[:space:]]4th[[:space:]]Edition.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Mail[[:space:]]Servers/Packtpub.Linux.Email.Nov.2009.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/NFS[[:space:]]&[[:space:]]NIS/O'Reilly[[:space:]]-[[:space:]]Managing[[:space:]]NFS[[:space:]]and[[:space:]]NIS[[:space:]]2nd[[:space:]]Edt.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/No[[:space:]]Starch[[:space:]]Press[[:space:]]-[[:space:]]Linux[[:space:]]Appliance[[:space:]]Design.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/No[[:space:]]Starch[[:space:]]Press[[:space:]]-[[:space:]]Linux[[:space:]]Firewalls.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/No[[:space:]]Starch[[:space:]]Press[[:space:]]-[[:space:]]Linux[[:space:]]For[[:space:]]Non-Geeks[[:space:]](2004).chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/No[[:space:]]Starch[[:space:]]Press[[:space:]]-[[:space:]]Programming[[:space:]]Linux[[:space:]]Games.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/O'Reilly[[:space:]]-[[:space:]]Linux[[:space:]]Cookbook.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/O'Reilly[[:space:]]-[[:space:]]Linux[[:space:]]Networking[[:space:]]Cookbook.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/O'Reilly[[:space:]]Fedora[[:space:]]Linux.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/O'Reilly[[:space:]]Linux[[:space:]]Desktop[[:space:]]Hacks.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/O'Reilly[[:space:]]Linux[[:space:]]Device[[:space:]]Drivers[[:space:]](3rd[[:space:]]Edition).chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/O'Reilly[[:space:]]Linux[[:space:]]in[[:space:]]a[[:space:]]Nutshell[[:space:]](5th[[:space:]]Edition).chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/O'Reilly[[:space:]]Linux[[:space:]]Kernel[[:space:]]in[[:space:]]a[[:space:]]Nutshell.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/O'Reilly[[:space:]]Linux[[:space:]]Multimedia[[:space:]]Hacks.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/O'Reilly[[:space:]]Linux[[:space:]]Network[[:space:]]Administrator's[[:space:]]Guide[[:space:]](3rd[[:space:]]Edition).chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/O'Reilly[[:space:]]Linux[[:space:]]System[[:space:]]Administration.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/O'Reilly[[:space:]]Linux[[:space:]]System[[:space:]]Programming.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/O'Reilly[[:space:]]LPI[[:space:]]Linux[[:space:]]Certification[[:space:]]In[[:space:]]A[[:space:]]Nutshell[[:space:]](2nd[[:space:]]Edition).chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/O'Reilly[[:space:]]Running[[:space:]]Linux[[:space:]](5th[[:space:]]Edition).chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/O'Reilly[[:space:]]SUSE[[:space:]]Linux.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/O'Reilly[[:space:]]Understanding[[:space:]]The[[:space:]]Linux[[:space:]]Kernel[[:space:]](3rd[[:space:]]Edition).pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/O'Reilly_-_Linux_Command_Directory.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/OReilly[[:space:]]-[[:space:]]Linux[[:space:]]Networking[[:space:]]Cookbook.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/OReilly.Building.Embedded.Linux.Systems.Aug.2008.eBook-DDU.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Red[[:space:]]Hat[[:space:]]Enterprise[[:space:]]Linux[[:space:]]4[[:space:]]for[[:space:]]Dummies[[:space:]](Apr[[:space:]]2005).pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Red[[:space:]]Hat[[:space:]]Linux[[:space:]]Fedora[[:space:]]for[[:space:]]Dummies.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Red[[:space:]]Hat[[:space:]]Software[[:space:]]-[[:space:]]Linux[[:space:]]Complete[[:space:]]Command[[:space:]]Reference.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Samba/O'Reilly[[:space:]]Samba[[:space:]]2nd[[:space:]]Edition.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Sams[[:space:]]-[[:space:]]Teach[[:space:]]Yourself[[:space:]]Linux[[:space:]]in[[:space:]]24[[:space:]]Hours.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/SELinux/SELinux[[:space:]]By[[:space:]]Example[[:space:]]-[[:space:]]Using[[:space:]]Security[[:space:]]Enhanced[[:space:]]Linux[[:space:]].chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Squid/O'Reilly[[:space:]]-[[:space:]]Squid[[:space:]]The[[:space:]]Definitive[[:space:]]Guide.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Squid/O'Reilly[[:space:]]-[[:space:]]Squid[[:space:]]The[[:space:]]Definitive[[:space:]]Guide.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/SSH[[:space:]]&[[:space:]]VPN/IPsec[[:space:]]Virtual[[:space:]]Private[[:space:]]Network[[:space:]]Fundamentals.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/SSH[[:space:]]&[[:space:]]VPN/O'Reilly[[:space:]]-[[:space:]]SSH[[:space:]]The[[:space:]]Secure[[:space:]]Shell[[:space:]]The[[:space:]]Definitive[[:space:]]Guide.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/SSH[[:space:]]&[[:space:]]VPN/O'Reilly[[:space:]]-[[:space:]]The[[:space:]]Secure[[:space:]]Shell[[:space:]]The[[:space:]]Definitive[[:space:]]Guide.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/SUSE[[:space:]]Linux[[:space:]]10[[:space:]]for[[:space:]]Dummies.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/SUSE[[:space:]]Linux[[:space:]]9.3[[:space:]]For[[:space:]]Dummies[[:space:]](May[[:space:]]2005).pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Linux/Wiley.Linux.Bible.2009.Edition.Jan.2009.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Operating[[:space:]]Systems[[:space:]]-[[:space:]]Internals[[:space:]]and[[:space:]]Design[[:space:]]Principles[[:space:]]-[[:space:]]7th[[:space:]]E.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Operating[[:space:]]Systems[[:space:]]and[[:space:]]Middleware.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Operating.Systems.-.A.concept-based.approach.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/OSX/Mac[[:space:]]mini[[:space:]]Hacks[[:space:]]&[[:space:]]Mods[[:space:]]for[[:space:]]Dummies.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/OSX/Mac[[:space:]]OS[[:space:]]X[[:space:]]GUI[[:space:]]Customization[[:space:]]for[[:space:]]Dummies[[:space:]]v2.1.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/OSX/Mac[[:space:]]OS[[:space:]]X[[:space:]]Panther[[:space:]]Timesaving[[:space:]]Techniques[[:space:]]for[[:space:]]Dummies[[:space:]](2004).pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/OSX/Mac[[:space:]]OS[[:space:]]X[[:space:]]Tiger[[:space:]]Timesaving[[:space:]]Techniques[[:space:]]for[[:space:]]Dummies.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/OSX/Macs[[:space:]]for[[:space:]]Dummies,[[:space:]]8th[[:space:]]Edition[[:space:]](2004).pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/OSX/O'Reilly[[:space:]]Essential[[:space:]]Mac[[:space:]]OS[[:space:]]X[[:space:]]Panther[[:space:]]Server[[:space:]]Administration.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/OSX/O'Reilly[[:space:]]Learning[[:space:]]Unix[[:space:]]for[[:space:]]Mac[[:space:]]OS[[:space:]]X.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/OSX/O'Reilly[[:space:]]Mac[[:space:]]OS[[:space:]]X[[:space:]]Panther[[:space:]]In[[:space:]]a[[:space:]]Nutshell[[:space:]](2nd[[:space:]]Edition).chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/OSX/O'Reilly[[:space:]]Mac[[:space:]]OS[[:space:]]X[[:space:]]The[[:space:]]Missing[[:space:]]Manual[[:space:]](Panther[[:space:]]Edition).chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/OSX/O'Reilly[[:space:]]Mac[[:space:]]OS[[:space:]]X[[:space:]]Unwired.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/OSX/O'Reilly[[:space:]]Switching[[:space:]]to[[:space:]]the[[:space:]]Mac[[:space:]]The[[:space:]]Missing[[:space:]]Manual[[:space:]](Tiger[[:space:]]Edition).chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Plan9/Introduction_to_Operating_Systems_Abstractions_Using_Plan_9_from_Bell_Labs_-_Ballesteros.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Plan9/Notes_on_The_Plan_9_3rd_Edition_Kernel_Source_-_Ballesteros.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Unix/BSD[[:space:]]&[[:space:]]Unix/O'Reilly[[:space:]]-[[:space:]]Learning[[:space:]]the[[:space:]]UNIX[[:space:]]Operating[[:space:]]System.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Unix/BSD[[:space:]]&[[:space:]]Unix/O'Reilly[[:space:]]-[[:space:]]The[[:space:]]Complete[[:space:]]FreeBSD.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Unix/BSD[[:space:]]&[[:space:]]Unix/UNIX[[:space:]]Filesystems[[:space:]]-[[:space:]]Evolution,[[:space:]]Design,[[:space:]]&[[:space:]]Implementation.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Unix/Classic[[:space:]]Shell[[:space:]]Scripting.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Unix/Learning[[:space:]]the[[:space:]]vi[[:space:]]and[[:space:]]Vim[[:space:]]Editors,[[:space:]]7e.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Unix/Linux[[:space:]]Complete[[:space:]]Command[[:space:]]Reference.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Unix/linux-Addison.Wesley.The.Art.Of.Unix.Programming.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Unix/Sams[[:space:]]Publishing[[:space:]]-[[:space:]]Unix[[:space:]]Shell[[:space:]]Programming,[[:space:]]3rd[[:space:]]Edition.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Unix/Unix[[:space:]]For[[:space:]]Dummies.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Windows[[:space:]]Server/Active[[:space:]]Directory/O'Reilly[[:space:]]Active[[:space:]]Directory[[:space:]](2nd[[:space:]]Edition).chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Windows[[:space:]]Server/Active[[:space:]]Directory/O'Reilly[[:space:]]Active[[:space:]]Directory[[:space:]]Cookbook[[:space:]]for[[:space:]]Windows[[:space:]]Server[[:space:]]2003[[:space:]]&[[:space:]]Windows[[:space:]]2000.chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Windows[[:space:]]Server/Sybex.Mastering.System.Center.Configuration.Manager.2007.R2.Apr.2009.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Windows[[:space:]]Server/Sybex.Windows.Server.2008.Hyper-V.Insiders.Guide.to.Microsofts.Hypervisor.Apr.2009.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Windows[[:space:]]Server/Wiley.Windows.Server.2008.Bible.Jul.2008.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Windows[[:space:]]Server/Windows[[:space:]]Server[[:space:]]2003[[:space:]]for[[:space:]]Dummies[[:space:]](2003).chm filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Windows[[:space:]]Server/Windows[[:space:]]Server[[:space:]]2003[[:space:]]For[[:space:]]Dummies.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Windows[[:space:]]XP/Windows[[:space:]]XP[[:space:]]Digital[[:space:]]Music[[:space:]]For[[:space:]]Dummies[[:space:]](Nov[[:space:]]2004).pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Windows[[:space:]]XP/Windows[[:space:]]Xp[[:space:]]For[[:space:]]Dummies,[[:space:]]1st[[:space:]]Edition[[:space:]](2001).pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Windows[[:space:]]XP/Windows[[:space:]]XP[[:space:]]for[[:space:]]Dummies,[[:space:]]2nd[[:space:]]Edition.pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Windows[[:space:]]XP/Windows[[:space:]]XP[[:space:]]Gigabook[[:space:]]For[[:space:]]Dummies[[:space:]](2004).pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Windows[[:space:]]XP/Windows[[:space:]]XP[[:space:]]Hacks[[:space:]]&[[:space:]]Mods[[:space:]]For[[:space:]]Dummies[[:space:]](Sep[[:space:]]2005).pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Windows[[:space:]]XP/Windows[[:space:]]XP[[:space:]]Just[[:space:]]the[[:space:]]Steps[[:space:]]for[[:space:]]Dummies[[:space:]](2005).pdf filter=lfs diff=lfs merge=lfs -text +Operating[[:space:]]Systems/Windows[[:space:]]XP/Windows[[:space:]]XP[[:space:]]Timesaving[[:space:]]Techniques[[:space:]]for[[:space:]]Dummies,[[:space:]]2nd[[:space:]]Edition[[:space:]](Jan[[:space:]]2005).pdf filter=lfs diff=lfs merge=lfs -text diff --git a/Operating Systems/BSD/No Starch Press - Absolute FreeBSD The Complete Guide to FreeBSD 2nd Edition.pdf b/Operating Systems/BSD/No Starch Press - Absolute FreeBSD The Complete Guide to FreeBSD 2nd Edition.pdf new file mode 100644 index 0000000000000000000000000000000000000000..94c73e8eec09d1439f1210d30eb7aba9e245efe7 --- /dev/null +++ b/Operating Systems/BSD/No Starch Press - Absolute FreeBSD The Complete Guide to FreeBSD 2nd Edition.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:913ee76d2f44e5199935bc9ce07343a5545caf66fd099fc5ae093037cb37bb91 +size 7473566 diff --git a/Operating Systems/BSD/No Starch Press - Absolute OpenBSD UNIX For The Practical Paranoid.pdf b/Operating Systems/BSD/No Starch Press - Absolute OpenBSD UNIX For The Practical Paranoid.pdf new file mode 100644 index 0000000000000000000000000000000000000000..ae4b642a070a788ff6f0768860c38bf8784453cd --- /dev/null +++ b/Operating Systems/BSD/No Starch Press - Absolute OpenBSD UNIX For The Practical Paranoid.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f664b905117f0a70033dd7dbe61ca04b9a41dbee7401f44f3264ab41c75abcba +size 4445591 diff --git a/Operating Systems/BSD/No Starch Press - Building a Server with FreeBSD 7 Apr 2008.chm b/Operating Systems/BSD/No Starch Press - Building a Server with FreeBSD 7 Apr 2008.chm new file mode 100644 index 0000000000000000000000000000000000000000..2de0754fed95051f1d4e9244782ebae1a6f4d7f6 Binary files /dev/null and b/Operating Systems/BSD/No Starch Press - Building a Server with FreeBSD 7 Apr 2008.chm differ diff --git a/Operating Systems/BSD/O'Reilly Mastering FreeBSD and OpenBSD Security.chm b/Operating Systems/BSD/O'Reilly Mastering FreeBSD and OpenBSD Security.chm new file mode 100644 index 0000000000000000000000000000000000000000..50b2329e5bca1ddefc9d07029615897c2d3c5638 --- /dev/null +++ b/Operating Systems/BSD/O'Reilly Mastering FreeBSD and OpenBSD Security.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2b2bbbb4e6c44ebba82455856e243d6042882e33d29dca278f950cf46a8e7844 +size 1398253 diff --git a/Operating Systems/Linux/Assembly Language Step-by-Step - Programming with Linux, 3rd Edition.pdf b/Operating Systems/Linux/Assembly Language Step-by-Step - Programming with Linux, 3rd Edition.pdf new file mode 100644 index 0000000000000000000000000000000000000000..6e502d2d59363179cc72085d926093bf74df7bd8 --- /dev/null +++ b/Operating Systems/Linux/Assembly Language Step-by-Step - Programming with Linux, 3rd Edition.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8ad49991824cba67916393d77538b1652d5110b44186b2fc1ae57096481e321e +size 4208452 diff --git a/Operating Systems/Linux/Cluster/Linux Enterprise Cluster.chm b/Operating Systems/Linux/Cluster/Linux Enterprise Cluster.chm new file mode 100644 index 0000000000000000000000000000000000000000..c3ccbc555022528dc772ac8e555be209e6c5d4c1 --- /dev/null +++ b/Operating Systems/Linux/Cluster/Linux Enterprise Cluster.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:090e54141659dea75f523abccd74813d623456e7e4a008ef996b4a1771328293 +size 3907968 diff --git a/Operating Systems/Linux/Cluster/OReilly High Performance Linux Clusters With Oscar Rocks openmosix And Mpi.chm b/Operating Systems/Linux/Cluster/OReilly High Performance Linux Clusters With Oscar Rocks openmosix And Mpi.chm new file mode 100644 index 0000000000000000000000000000000000000000..4fcad0e8b7025df5b01ef7dff6c5c6cc0516b072 --- /dev/null +++ b/Operating Systems/Linux/Cluster/OReilly High Performance Linux Clusters With Oscar Rocks openmosix And Mpi.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0fcb8ac7ebb9be624d9e742a1d81ec2096bf6187b6b4852dbdd4308523d8574c +size 1816094 diff --git a/Operating Systems/Linux/Core Linux Concepts (Kernel,Networking,FineTuning,Device Drivers,Thin Client)/Linux Thin Client Networks Design and Deployment.pdf b/Operating Systems/Linux/Core Linux Concepts (Kernel,Networking,FineTuning,Device Drivers,Thin Client)/Linux Thin Client Networks Design and Deployment.pdf new file mode 100644 index 0000000000000000000000000000000000000000..525c299c86349b3ac549c32cfdbb3741c1614a37 --- /dev/null +++ b/Operating Systems/Linux/Core Linux Concepts (Kernel,Networking,FineTuning,Device Drivers,Thin Client)/Linux Thin Client Networks Design and Deployment.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f18a2ada7d87c5570d9d6e1bfba7bb70c03ae3d9e757001c8a0d5b8b41b51d1f +size 5513442 diff --git a/Operating Systems/Linux/Core Linux Concepts (Kernel,Networking,FineTuning,Device Drivers,Thin Client)/O'Reilly - Linux Device Drivers 2nd Edition.pdf b/Operating Systems/Linux/Core Linux Concepts (Kernel,Networking,FineTuning,Device Drivers,Thin Client)/O'Reilly - Linux Device Drivers 2nd Edition.pdf new file mode 100644 index 0000000000000000000000000000000000000000..580ac6adf4fe92ae16e11dc28710583dd982a781 --- /dev/null +++ b/Operating Systems/Linux/Core Linux Concepts (Kernel,Networking,FineTuning,Device Drivers,Thin Client)/O'Reilly - Linux Device Drivers 2nd Edition.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:31acd7351b2b97cc74e13f312ada794f1623888a17cece837b1a968cf562ea06 +size 7718811 diff --git a/Operating Systems/Linux/Core Linux Concepts (Kernel,Networking,FineTuning,Device Drivers,Thin Client)/OReilly Zero Configuration Networking The Definitive Guide.chm b/Operating Systems/Linux/Core Linux Concepts (Kernel,Networking,FineTuning,Device Drivers,Thin Client)/OReilly Zero Configuration Networking The Definitive Guide.chm new file mode 100644 index 0000000000000000000000000000000000000000..833370923736230a7848a00593f08a67c49b9582 Binary files /dev/null and b/Operating Systems/Linux/Core Linux Concepts (Kernel,Networking,FineTuning,Device Drivers,Thin Client)/OReilly Zero Configuration Networking The Definitive Guide.chm differ diff --git a/Operating Systems/Linux/Core Linux Concepts (Kernel,Networking,FineTuning,Device Drivers,Thin Client)/Understanding Linux Network Internals.chm b/Operating Systems/Linux/Core Linux Concepts (Kernel,Networking,FineTuning,Device Drivers,Thin Client)/Understanding Linux Network Internals.chm new file mode 100644 index 0000000000000000000000000000000000000000..5b6e06d7d167f0c26d6106a0268a1e4635bbe111 --- /dev/null +++ b/Operating Systems/Linux/Core Linux Concepts (Kernel,Networking,FineTuning,Device Drivers,Thin Client)/Understanding Linux Network Internals.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:32788f15ddf97b27f67c3e98d28f94e1a063286b64cd39d5bcae2cc63760726c +size 7868419 diff --git a/Operating Systems/Linux/Core Linux Concepts (Kernel,Networking,FineTuning,Device Drivers,Thin Client)/shell_scripting.pdf b/Operating Systems/Linux/Core Linux Concepts (Kernel,Networking,FineTuning,Device Drivers,Thin Client)/shell_scripting.pdf new file mode 100644 index 0000000000000000000000000000000000000000..f183ec0d07f0c8625373ac8dc44482e6afd4fd5b --- /dev/null +++ b/Operating Systems/Linux/Core Linux Concepts (Kernel,Networking,FineTuning,Device Drivers,Thin Client)/shell_scripting.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ed6ce7ff16656c34ca2b9fd8517e38ffdd9316e43f84156ab760cafb4662d1bf +size 14325554 diff --git a/Operating Systems/Linux/DNS/Bind9 Administration Reference Manual.pdf b/Operating Systems/Linux/DNS/Bind9 Administration Reference Manual.pdf new file mode 100644 index 0000000000000000000000000000000000000000..cc1da908a1321aa5279fcc6dfd84e20471ac4625 --- /dev/null +++ b/Operating Systems/Linux/DNS/Bind9 Administration Reference Manual.pdf @@ -0,0 +1,22737 @@ +%PDF-1.2 +2 0 obj +<< /S /GoTo /D (1.0) >> +endobj +4 0 obj +(Table of Contents) +endobj +6 0 obj +<< /S /GoTo /D (2.0) >> +endobj +8 0 obj +(Chapter 1. Introduction ) +endobj +10 0 obj +<< /S /GoTo /D (2.1.1) >> +endobj +12 0 obj +(1.1. Scope of Document) +endobj +14 0 obj +<< /S /GoTo /D (2.2.1) >> +endobj +16 0 obj +(1.2. Organization of This Document) +endobj +18 0 obj +<< /S /GoTo /D (2.3.1) >> +endobj +20 0 obj +(1.3. Conventions Used in This Document) +endobj +22 0 obj +<< /S /GoTo /D (2.4.1) >> +endobj +24 0 obj +(1.4. The Domain Name System \(DNS\)) +endobj +26 0 obj +<< /S /GoTo /D (2.4.1.2) >> +endobj +28 0 obj +(1.4.1. DNS Fundamentals) +endobj +30 0 obj +<< /S /GoTo /D (2.4.2.2) >> +endobj +32 0 obj +(1.4.2. Domains and Domain Names) +endobj +34 0 obj +<< /S /GoTo /D (2.4.3.2) >> +endobj +36 0 obj +(1.4.3. Zones) +endobj +38 0 obj +<< /S /GoTo /D (2.4.4.2) >> +endobj +40 0 obj +(1.4.4. Authoritative Name Servers) +endobj +42 0 obj +<< /S /GoTo /D (2.4.4.1.3) >> +endobj +44 0 obj +(1.4.4.1. The Primary Master) +endobj +46 0 obj +<< /S /GoTo /D (2.4.4.2.3) >> +endobj +48 0 obj +(1.4.4.2. Slave Servers) +endobj +50 0 obj +<< /S /GoTo /D (2.4.4.3.3) >> +endobj +52 0 obj +(1.4.4.3. Stealth Servers) +endobj +54 0 obj +<< /S /GoTo /D (2.4.5.2) >> +endobj +56 0 obj +(1.4.5. Caching Name Servers) +endobj +58 0 obj +<< /S /GoTo /D (2.4.5.4.3) >> +endobj +60 0 obj +(1.4.5.1. Forwarding) +endobj +62 0 obj +<< /S /GoTo /D (2.4.6.2) >> +endobj +64 0 obj +(1.4.6. Name Servers in Multiple Roles) +endobj +66 0 obj +<< /S /GoTo /D (3.0) >> +endobj +68 0 obj +(Chapter 2. BIND Resource Requirements) +endobj +70 0 obj +<< /S /GoTo /D (3.5.1) >> +endobj +72 0 obj +(2.1. Hardware requirements) +endobj +74 0 obj +<< /S /GoTo /D (3.6.1) >> +endobj +76 0 obj +(2.2. CPU Requirements) +endobj +78 0 obj +<< /S /GoTo /D (3.7.1) >> +endobj +80 0 obj +(2.3. Memory Requirements) +endobj +82 0 obj +<< /S /GoTo /D (3.8.1) >> +endobj +84 0 obj +(2.4. Nameserver Intensive Environment Issues) +endobj +86 0 obj +<< /S /GoTo /D (3.9.1) >> +endobj +88 0 obj +(2.5. Supported Operating Systems) +endobj +90 0 obj +<< /S /GoTo /D (4.0) >> +endobj +92 0 obj +(Chapter 3. Nameserver Configuration) +endobj +94 0 obj +<< /S /GoTo /D (4.10.1) >> +endobj +96 0 obj +(3.1. Sample Configurations) +endobj +98 0 obj +<< /S /GoTo /D (4.10.7.2) >> +endobj +100 0 obj +(3.1.1. A Caching-only Nameserver) +endobj +102 0 obj +<< /S /GoTo /D (4.10.8.2) >> +endobj +104 0 obj +(3.1.2. An Authoritative-only Nameserver) +endobj +106 0 obj +<< /S /GoTo /D (4.11.1) >> +endobj +108 0 obj +(3.2. Load Balancing) +endobj +110 0 obj +<< /S /GoTo /D (4.12.1) >> +endobj +112 0 obj +(3.3. Notify) +endobj +114 0 obj +<< /S /GoTo /D (4.13.1) >> +endobj +116 0 obj +(3.4. Nameserver Operations) +endobj +118 0 obj +<< /S /GoTo /D (4.13.9.2) >> +endobj +120 0 obj +(3.4.1. Tools for Use With the Nameserver Daemon) +endobj +122 0 obj +<< /S /GoTo /D (4.13.9.5.3) >> +endobj +124 0 obj +(3.4.1.1. Diagnostic Tools) +endobj +126 0 obj +<< /S /GoTo /D (4.13.9.6.3) >> +endobj +128 0 obj +(3.4.1.2. Administrative Tools) +endobj +130 0 obj +<< /S /GoTo /D (4.13.10.2) >> +endobj +132 0 obj +(3.4.2. Signals) +endobj +134 0 obj +<< /S /GoTo /D (5.0) >> +endobj +136 0 obj +(Chapter 4. Advanced Concepts) +endobj +138 0 obj +<< /S /GoTo /D (5.14.1) >> +endobj +140 0 obj +(4.1. Dynamic Update) +endobj +142 0 obj +<< /S /GoTo /D (5.14.11.2) >> +endobj +144 0 obj +(4.1.1. The journal file) +endobj +146 0 obj +<< /S /GoTo /D (5.15.1) >> +endobj +148 0 obj +(4.2. Incremental Zone Transfers \(IXFR\)) +endobj +150 0 obj +<< /S /GoTo /D (5.16.1) >> +endobj +152 0 obj +(4.3. Split DNS) +endobj +154 0 obj +<< /S /GoTo /D (5.17.1) >> +endobj +156 0 obj +(4.4. TSIG) +endobj +158 0 obj +<< /S /GoTo /D (5.17.12.2) >> +endobj +160 0 obj +(4.4.1. Generate Shared Keys for Each Pair of Hosts) +endobj +162 0 obj +<< /S /GoTo /D (5.17.12.7.3) >> +endobj +164 0 obj +(4.4.1.1. Automatic Generation) +endobj +166 0 obj +<< /S /GoTo /D (5.17.12.8.3) >> +endobj +168 0 obj +(4.4.1.2. Manual Generation) +endobj +170 0 obj +<< /S /GoTo /D (5.17.13.2) >> +endobj +172 0 obj +(4.4.2. Copying the Shared Secret to Both Machines) +endobj +174 0 obj +<< /S /GoTo /D (5.17.14.2) >> +endobj +176 0 obj +(4.4.3. Informing the Servers of the Key's Existence) +endobj +178 0 obj +<< /S /GoTo /D (5.17.15.2) >> +endobj +180 0 obj +(4.4.4. Instructing the Server to Use the Key) +endobj +182 0 obj +<< /S /GoTo /D (5.17.16.2) >> +endobj +184 0 obj +(4.4.5. TSIG Key Based Access Control) +endobj +186 0 obj +<< /S /GoTo /D (5.17.17.2) >> +endobj +188 0 obj +(4.4.6. Errors) +endobj +190 0 obj +<< /S /GoTo /D (5.18.1) >> +endobj +192 0 obj +(4.5. TKEY) +endobj +194 0 obj +<< /S /GoTo /D (5.19.1) >> +endobj +196 0 obj +(4.6. SIG\(0\)) +endobj +198 0 obj +<< /S /GoTo /D (5.20.1) >> +endobj +200 0 obj +(4.7. DNSSEC) +endobj +202 0 obj +<< /S /GoTo /D (5.20.18.2) >> +endobj +204 0 obj +(4.7.1. Generating Keys) +endobj +206 0 obj +<< /S /GoTo /D (5.20.19.2) >> +endobj +208 0 obj +(4.7.2. Creating a Keyset) +endobj +210 0 obj +<< /S /GoTo /D (5.20.20.2) >> +endobj +212 0 obj +(4.7.3. Signing the Child's Keyset) +endobj +214 0 obj +<< /S /GoTo /D (5.20.21.2) >> +endobj +216 0 obj +(4.7.4. Signing the Zone) +endobj +218 0 obj +<< /S /GoTo /D (5.20.22.2) >> +endobj +220 0 obj +(4.7.5. Configuring Servers) +endobj +222 0 obj +<< /S /GoTo /D (5.21.1) >> +endobj +224 0 obj +(4.8. IPv6 Support in BIND 9) +endobj +226 0 obj +<< /S /GoTo /D (5.21.23.2) >> +endobj +228 0 obj +(4.8.1. Address Lookups Using AAAA Records) +endobj +230 0 obj +<< /S /GoTo /D (5.21.24.2) >> +endobj +232 0 obj +(4.8.2. Address Lookups Using A6 Records) +endobj +234 0 obj +<< /S /GoTo /D (5.21.24.9.3) >> +endobj +236 0 obj +(4.8.2.1. A6 Chains) +endobj +238 0 obj +<< /S /GoTo /D (5.21.24.10.3) >> +endobj +240 0 obj +(4.8.2.2. A6 Records for DNS Servers) +endobj +242 0 obj +<< /S /GoTo /D (5.21.25.2) >> +endobj +244 0 obj +(4.8.3. Address to Name Lookups Using Nibble Format) +endobj +246 0 obj +<< /S /GoTo /D (5.21.26.2) >> +endobj +248 0 obj +(4.8.4. Address to Name Lookups Using Bitstring Format) +endobj +250 0 obj +<< /S /GoTo /D (5.21.27.2) >> +endobj +252 0 obj +(4.8.5. Using DNAME for Delegation of IPv6 Reverse Addresses) +endobj +254 0 obj +<< /S /GoTo /D (6.0) >> +endobj +256 0 obj +(Chapter 5. The BIND 9 Lightweight Resolver) +endobj +258 0 obj +<< /S /GoTo /D (6.22.1) >> +endobj +260 0 obj +(5.1. The Lightweight Resolver Library) +endobj +262 0 obj +<< /S /GoTo /D (6.23.1) >> +endobj +264 0 obj +(5.2. Running a Resolver Daemon) +endobj +266 0 obj +<< /S /GoTo /D (7.0) >> +endobj +268 0 obj +(Chapter 6. BIND 9 Configuration Reference) +endobj +270 0 obj +<< /S /GoTo /D (7.24.1) >> +endobj +272 0 obj +(6.1. Configuration File Elements) +endobj +274 0 obj +<< /S /GoTo /D (7.24.28.2) >> +endobj +276 0 obj +(6.1.1. Address Match Lists) +endobj +278 0 obj +<< /S /GoTo /D (7.24.28.11.3) >> +endobj +280 0 obj +(6.1.1.1. Syntax) +endobj +282 0 obj +<< /S /GoTo /D (7.24.28.12.3) >> +endobj +284 0 obj +(6.1.1.2. Definition and Usage) +endobj +286 0 obj +<< /S /GoTo /D (7.24.29.2) >> +endobj +288 0 obj +(6.1.2. Comment Syntax) +endobj +290 0 obj +<< /S /GoTo /D (7.24.29.13.3) >> +endobj +292 0 obj +(6.1.2.1. Syntax) +endobj +294 0 obj +<< /S /GoTo /D (7.24.29.14.3) >> +endobj +296 0 obj +(6.1.2.2. Definition and Usage) +endobj +298 0 obj +<< /S /GoTo /D (7.25.1) >> +endobj +300 0 obj +(6.2. Configuration File Grammar) +endobj +302 0 obj +<< /S /GoTo /D (7.25.30.2) >> +endobj +304 0 obj +(6.2.1. acl Statement Grammar) +endobj +306 0 obj +<< /S /GoTo /D (7.25.31.2) >> +endobj +308 0 obj +(6.2.2. acl Statement Definition and Usage) +endobj +310 0 obj +<< /S /GoTo /D (7.25.32.2) >> +endobj +312 0 obj +(6.2.3. controls Statement Grammar) +endobj +314 0 obj +<< /S /GoTo /D (7.25.33.2) >> +endobj +316 0 obj +(6.2.4. controls Statement Definition and Usage) +endobj +318 0 obj +<< /S /GoTo /D (7.25.34.2) >> +endobj +320 0 obj +(6.2.5. include Statement Grammar) +endobj +322 0 obj +<< /S /GoTo /D (7.25.35.2) >> +endobj +324 0 obj +(6.2.6. include Statement Definition and Usage) +endobj +326 0 obj +<< /S /GoTo /D (7.25.36.2) >> +endobj +328 0 obj +(6.2.7. key Statement Grammar) +endobj +330 0 obj +<< /S /GoTo /D (7.25.37.2) >> +endobj +332 0 obj +(6.2.8. key Statement Definition and Usage) +endobj +334 0 obj +<< /S /GoTo /D (7.25.38.2) >> +endobj +336 0 obj +(6.2.9. logging Statement Grammar) +endobj +338 0 obj +<< /S /GoTo /D (7.25.39.2) >> +endobj +340 0 obj +(6.2.10. logging Statement Definition and Usage) +endobj +342 0 obj +<< /S /GoTo /D (7.25.39.15.3) >> +endobj +344 0 obj +(6.2.10.1. The channel Phrase) +endobj +346 0 obj +<< /S /GoTo /D (7.25.39.16.3) >> +endobj +348 0 obj +(6.2.10.2. The category Phrase) +endobj +350 0 obj +<< /S /GoTo /D (7.25.40.2) >> +endobj +352 0 obj +(6.2.11. lwres Statement Grammar) +endobj +354 0 obj +<< /S /GoTo /D (7.25.41.2) >> +endobj +356 0 obj +(6.2.12. lwres Statement Definition and Usage) +endobj +358 0 obj +<< /S /GoTo /D (7.25.42.2) >> +endobj +360 0 obj +(6.2.13. options Statement Grammar) +endobj +362 0 obj +<< /S /GoTo /D (7.25.43.2) >> +endobj +364 0 obj +(6.2.14. options Statement Definition and Usage) +endobj +366 0 obj +<< /S /GoTo /D (7.25.43.17.3) >> +endobj +368 0 obj +(6.2.14.1. Boolean Options) +endobj +370 0 obj +<< /S /GoTo /D (7.25.43.18.3) >> +endobj +372 0 obj +(6.2.14.2. Forwarding) +endobj +374 0 obj +<< /S /GoTo /D (7.25.43.19.3) >> +endobj +376 0 obj +(6.2.14.3. Access Control) +endobj +378 0 obj +<< /S /GoTo /D (7.25.43.20.3) >> +endobj +380 0 obj +(6.2.14.4. Interfaces) +endobj +382 0 obj +<< /S /GoTo /D (7.25.43.21.3) >> +endobj +384 0 obj +(6.2.14.5. Query Address) +endobj +386 0 obj +<< /S /GoTo /D (7.25.43.22.3) >> +endobj +388 0 obj +(6.2.14.6. Zone Transfers) +endobj +390 0 obj +<< /S /GoTo /D (7.25.43.23.3) >> +endobj +392 0 obj +(6.2.14.7. Operating System Resource Limits) +endobj +394 0 obj +<< /S /GoTo /D (7.25.43.24.3) >> +endobj +396 0 obj +(6.2.14.8. Server Resource Limits) +endobj +398 0 obj +<< /S /GoTo /D (7.25.43.25.3) >> +endobj +400 0 obj +(6.2.14.9. Periodic Task Intervals) +endobj +402 0 obj +<< /S /GoTo /D (7.25.43.26.3) >> +endobj +404 0 obj +(6.2.14.10. Topology) +endobj +406 0 obj +<< /S /GoTo /D (7.25.43.27.3) >> +endobj +408 0 obj +(6.2.14.11. The sortlist Statement) +endobj +410 0 obj +<< /S /GoTo /D (7.25.43.28.3) >> +endobj +412 0 obj +(6.2.14.12. RRset Ordering) +endobj +414 0 obj +<< /S /GoTo /D (7.25.43.29.3) >> +endobj +416 0 obj +(6.2.14.13. Synthetic IPv6 responses) +endobj +418 0 obj +<< /S /GoTo /D (7.25.43.30.3) >> +endobj +420 0 obj +(6.2.14.14. Tuning) +endobj +422 0 obj +<< /S /GoTo /D (7.25.43.31.3) >> +endobj +424 0 obj +(6.2.14.15. The Statistics File) +endobj +426 0 obj +<< /S /GoTo /D (7.25.44.2) >> +endobj +428 0 obj +(6.2.15. server Statement Grammar) +endobj +430 0 obj +<< /S /GoTo /D (7.25.45.2) >> +endobj +432 0 obj +(6.2.16. server Statement Definition and Usage) +endobj +434 0 obj +<< /S /GoTo /D (7.25.46.2) >> +endobj +436 0 obj +(6.2.17. trusted-keys Statement Grammar) +endobj +438 0 obj +<< /S /GoTo /D (7.25.47.2) >> +endobj +440 0 obj +(6.2.18. trusted-keys Statement Definition and Usage) +endobj +442 0 obj +<< /S /GoTo /D (7.25.48.2) >> +endobj +444 0 obj +(6.2.19. view Statement Grammar) +endobj +446 0 obj +<< /S /GoTo /D (7.25.49.2) >> +endobj +448 0 obj +(6.2.20. view Statement Definition and Usage) +endobj +450 0 obj +<< /S /GoTo /D (7.25.50.2) >> +endobj +452 0 obj +(6.2.21. zone Statement Grammar) +endobj +454 0 obj +<< /S /GoTo /D (7.25.51.2) >> +endobj +456 0 obj +(6.2.22. zone Statement Definition and Usage) +endobj +458 0 obj +<< /S /GoTo /D (7.25.51.32.3) >> +endobj +460 0 obj +(6.2.22.1. Zone Types) +endobj +462 0 obj +<< /S /GoTo /D (7.25.51.33.3) >> +endobj +464 0 obj +(6.2.22.2. Class) +endobj +466 0 obj +<< /S /GoTo /D (7.25.51.34.3) >> +endobj +468 0 obj +(6.2.22.3. Zone Options) +endobj +470 0 obj +<< /S /GoTo /D (7.25.51.35.3) >> +endobj +472 0 obj +(6.2.22.4. Dynamic Update Policies) +endobj +474 0 obj +<< /S /GoTo /D (7.26.1) >> +endobj +476 0 obj +(6.3. Zone File) +endobj +478 0 obj +<< /S /GoTo /D (7.26.52.2) >> +endobj +480 0 obj +(6.3.1. Types of Resource Records and When to Use Them) +endobj +482 0 obj +<< /S /GoTo /D (7.26.52.36.3) >> +endobj +484 0 obj +(6.3.1.1. Resource Records) +endobj +486 0 obj +<< /S /GoTo /D (7.26.52.37.3) >> +endobj +488 0 obj +(6.3.1.2. Textual expression of RRs) +endobj +490 0 obj +<< /S /GoTo /D (7.26.53.2) >> +endobj +492 0 obj +(6.3.2. Discussion of MX Records) +endobj +494 0 obj +<< /S /GoTo /D (7.26.54.2) >> +endobj +496 0 obj +(6.3.3. Setting TTLs) +endobj +498 0 obj +<< /S /GoTo /D (7.26.55.2) >> +endobj +500 0 obj +(6.3.4. Inverse Mapping in IPv4) +endobj +502 0 obj +<< /S /GoTo /D (7.26.56.2) >> +endobj +504 0 obj +(6.3.5. Other Zone File Directives) +endobj +506 0 obj +<< /S /GoTo /D (7.26.56.38.3) >> +endobj +508 0 obj +(6.3.5.1. The Directive) +endobj +510 0 obj +<< /S /GoTo /D (7.26.56.39.3) >> +endobj +512 0 obj +(6.3.5.2. The Directive) +endobj +514 0 obj +<< /S /GoTo /D (7.26.56.40.3) >> +endobj +516 0 obj +(6.3.5.3. The Directive) +endobj +518 0 obj +<< /S /GoTo /D (7.26.57.2) >> +endobj +520 0 obj +(6.3.6. BIND Master File Extension: the Directive) +endobj +522 0 obj +<< /S /GoTo /D (8.0) >> +endobj +524 0 obj +(Chapter 7. BIND 9 Security Considerations) +endobj +526 0 obj +<< /S /GoTo /D (8.27.1) >> +endobj +528 0 obj +(7.1. Access Control Lists) +endobj +530 0 obj +<< /S /GoTo /D (8.28.1) >> +endobj +532 0 obj +(7.2. chroot and setuid \(for UNIX servers\)) +endobj +534 0 obj +<< /S /GoTo /D (8.28.58.2) >> +endobj +536 0 obj +(7.2.1. The chroot Environment) +endobj +538 0 obj +<< /S /GoTo /D (8.28.59.2) >> +endobj +540 0 obj +(7.2.2. Using the setuid Function) +endobj +542 0 obj +<< /S /GoTo /D (8.29.1) >> +endobj +544 0 obj +(7.3. Dynamic Update Security) +endobj +546 0 obj +<< /S /GoTo /D (9.0) >> +endobj +548 0 obj +(Chapter 8. Troubleshooting) +endobj +550 0 obj +<< /S /GoTo /D (9.30.1) >> +endobj +552 0 obj +(8.1. Common Problems) +endobj +554 0 obj +<< /S /GoTo /D (9.30.60.2) >> +endobj +556 0 obj +(8.1.1. It's not working; how can I figure out what's wrong?) +endobj +558 0 obj +<< /S /GoTo /D (9.31.1) >> +endobj +560 0 obj +(8.2. Incrementing and Changing the Serial Number) +endobj +562 0 obj +<< /S /GoTo /D (9.32.1) >> +endobj +564 0 obj +(8.3. Where Can I Get Help?) +endobj +566 0 obj +<< /S /GoTo /D (10.0) >> +endobj +568 0 obj +(Appendix A. Appendices) +endobj +570 0 obj +<< /S /GoTo /D (10.33.1) >> +endobj +572 0 obj +(A.1. Acknowledgements) +endobj +574 0 obj +<< /S /GoTo /D (10.33.61.2) >> +endobj +576 0 obj +(A.1.1. A Brief History of the DNS and BIND) +endobj +578 0 obj +<< /S /GoTo /D (10.34.1) >> +endobj +580 0 obj +(A.2. Historical DNS Information) +endobj +582 0 obj +<< /S /GoTo /D (10.34.62.2) >> +endobj +584 0 obj +(A.2.1. Classes of Resource Records) +endobj +586 0 obj +<< /S /GoTo /D (10.34.62.41.3) >> +endobj +588 0 obj +(A.2.1.1. HS = hesiod) +endobj +590 0 obj +<< /S /GoTo /D (10.34.62.42.3) >> +endobj +592 0 obj +(A.2.1.2. CH = chaos) +endobj +594 0 obj +<< /S /GoTo /D (10.35.1) >> +endobj +596 0 obj +(A.3. General DNS Reference Information) +endobj +598 0 obj +<< /S /GoTo /D (10.35.63.2) >> +endobj +600 0 obj +(A.3.1. IPv6 addresses \(A6\)) +endobj +602 0 obj +<< /S /GoTo /D (10.36.1) >> +endobj +604 0 obj +(A.4. Bibliography \(and Suggested Reading\)) +endobj +606 0 obj +<< /S /GoTo /D (10.36.64.2) >> +endobj +608 0 obj +(A.4.1. Request for Comments \(RFCs\)) +endobj +610 0 obj +<< /S /GoTo /D (10.36.65.2) >> +endobj +612 0 obj +(Bibliography) +endobj +614 0 obj +<< /S /GoTo /D (10.37.1) >> +endobj +616 0 obj +(Standards) +endobj +618 0 obj +<< /S /GoTo /D (10.38.1) >> +endobj +620 0 obj +(Proposed Standards) +endobj +622 0 obj +<< /S /GoTo /D (10.39.1) >> +endobj +624 0 obj +(Proposed Standards Still Under Development) +endobj +626 0 obj +<< /S /GoTo /D (10.40.1) >> +endobj +628 0 obj +(Other Important RFCs About DNS Implementation) +endobj +630 0 obj +<< /S /GoTo /D (10.41.1) >> +endobj +632 0 obj +(Resource Record Types) +endobj +634 0 obj +<< /S /GoTo /D (10.42.1) >> +endobj +636 0 obj +(DNS and the Internet) +endobj +638 0 obj +<< /S /GoTo /D (10.43.1) >> +endobj +640 0 obj +(DNS Operations) +endobj +642 0 obj +<< /S /GoTo /D (10.44.1) >> +endobj +644 0 obj +(Other DNS-related RFCs) +endobj +646 0 obj +<< /S /GoTo /D (10.45.1) >> +endobj +648 0 obj +(Obsolete and Unimplemented Experimental RRs) +endobj +650 0 obj +<< /S /GoTo /D (10.45.66.2) >> +endobj +652 0 obj +(A.4.2. Internet Drafts) +endobj +654 0 obj +<< /S /GoTo /D (10.45.67.2) >> +endobj +656 0 obj +(A.4.3. Other Documents About BIND) +endobj +658 0 obj +<< /S /GoTo /D (10.45.68.2) >> +endobj +660 0 obj +(Bibliography) +endobj +662 0 obj +<< /S /GoTo /D [661 0 R /Fit ] >> +endobj +667 0 obj << +/Length 668 0 R +/Filter /FlateDecode +>> +stream +xڥ +@<ŖIvϔT"$1Ba%a,AF6"wL+cu?Tq% ј 5[&`6WO.q;2fYUm5ϡ/G?Gg_'$b큔AAaddSӱFUG$ڸji!Z5?8qBendstream +endobj +668 0 obj +176 +endobj +661 0 obj << +/Type /Page +/Contents 667 0 R +/Resources 666 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 670 0 R +>> endobj +665 0 obj << +/D [661 0 R /XYZ 71.731 705.352 null] +>> endobj +663 0 obj << +/D [661 0 R /XYZ 71.731 657.532 null] +>> endobj +664 0 obj << +/D [661 0 R /XYZ 71.731 657.532 null] +>> endobj +666 0 obj << +/Font << /F21 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +674 0 obj << +/Length 675 0 R +/Filter /FlateDecode +>> +stream +x3T0BCsC=sc e`glje +07P L$NPP̌Rsšhendstream +endobj +675 0 obj +66 +endobj +673 0 obj << +/Type /Page +/Contents 674 0 R +/Resources 672 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 670 0 R +>> endobj +671 0 obj << +/D [673 0 R /XYZ 71.731 705.352 null] +>> endobj +672 0 obj << +/ProcSet [ /PDF ] +>> endobj +679 0 obj << +/Length 680 0 R +/Filter /FlateDecode +>> +stream +xڥN0> endobj +676 0 obj << +/D [678 0 R /XYZ 71.731 705.352 null] +>> endobj +677 0 obj << +/Font << /F24 669 0 R /F25 681 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +685 0 obj << +/Length 686 0 R +/Filter /FlateDecode +>> +stream +x3T0BCsC=sc e`glje +07P L$NPP̌Rsšhendstream +endobj +686 0 obj +66 +endobj +684 0 obj << +/Type /Page +/Contents 685 0 R +/Resources 683 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 670 0 R +>> endobj +682 0 obj << +/D [684 0 R /XYZ 71.731 705.352 null] +>> endobj +683 0 obj << +/ProcSet [ /PDF ] +>> endobj +728 0 obj << +/Length 729 0 R +/Filter /FlateDecode +>> +stream +x_SWહ̧R>2'38Sn^hm -VOy>HHrbD8XS@KO<0_2e:8>*ѿ-^[{bU +u|7G_˷9HY;Nkp41iToW__Ny/}_2FN/RRGϮ1{wzßV˳&4UŰ8v ^w0.uru>tVOތcأ"ߋ{*TPQ8*TPBvRB +*Th'*TPB{!SD.B߇1g'@эFub'4Օ}\DTPB©PB +*T, +*TPBE;PB +*T|YsU˪pX&wWgmE3aznϭ*ˢTeF +*Tƴ +*TPBE;PB +*TNoUy>D/b yq<>oZ7V~h/Wa,^iyͷ?6Pu/➏oɺqoN +*T +*TPBETPB +* +N*TPB +B3-eSAnMyrt6oOVOhg{y{m^"2K +*T +*TPBETPB +*ZI +*TP[ ?UX"-[!/J?yO[ cd}pr25O)E]ԫoSۊ{ǎ{{p*)B +*TPQ8*TPBµ +*TPBE;PB +*Tr͝9NGFm3oCL_zYŻcH<*TPBvRB +*T( +*TPBBK;PB +*T~'|L +PezGCqvztNw/هOnj9lskW&~FSFA;PB +*T +*TPBETPB +*Z)HDz5?&|2LrrnBEPDy[(Cg"T +(ߋP.b*TPQ8*TPBvRB +*Th'*3}qǮn>tdٸ;tnTga.)eN*TPB +SB +*TXhi'*TPBvRB +*T( +*V}[9=8R:f:p"E1icH+ +*TPTPB +* +B +*TPN*TPB +B +*TPQ8*vI\#$7~l%eQjq|Jd^'7?7Y@yqu):~ PBvRB +*T( +*TPBEI +*TPǣ諒NΌ{?`'W^N/[+!Ueu>ok*z>xq27Kːb'S^r} +*TP1@;PB +*T +*TPBETPB +*ZI +*TWQw Pe˜p +Mw oGNflgy*yyv%Q8N*TPB +B +*TPQ8*TPBvRB +*griDbze"K:;m7D: ţa^fm;\T4ee?a5ެRyyE1*Th'*TPB©PB +*TTPBʾ<Ӏe#<_*_9/&ooS֡/&ُ뒊vRB +*T( +*TPBBK;PB +*T +*TPBETPBʓPzo@QUaqe +eV!Uu5yܮfF"4uendstream +endobj +729 0 obj +3904 +endobj +727 0 obj << +/Type /Page +/Contents 728 0 R +/Resources 726 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 670 0 R +/Annots [ 730 0 R 732 0 R 733 0 R 734 0 R 735 0 R 736 0 R 737 0 R 738 0 R 739 0 R 740 0 R 741 0 R 742 0 R 743 0 R 744 0 R 745 0 R 746 0 R 747 0 R 748 0 R 749 0 R 750 0 R 751 0 R 752 0 R 753 0 R 754 0 R 755 0 R 756 0 R 757 0 R 758 0 R 759 0 R 760 0 R 761 0 R 762 0 R 763 0 R 764 0 R 765 0 R 766 0 R 767 0 R 768 0 R 769 0 R 770 0 R 771 0 R 772 0 R 773 0 R 774 0 R 775 0 R 776 0 R 777 0 R 778 0 R 779 0 R 780 0 R 781 0 R 782 0 R 783 0 R 784 0 R 785 0 R 786 0 R 787 0 R 788 0 R 789 0 R 790 0 R 791 0 R 792 0 R 793 0 R 794 0 R 795 0 R 796 0 R 797 0 R 798 0 R 799 0 R 800 0 R 801 0 R 802 0 R 803 0 R 804 0 R ] +>> endobj +730 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [119.551 619.061 183.58 626.045] +/Subtype /Link +/A << /S /GoTo /D (ch01) >> +>> endobj +732 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [532.999 619.061 537.981 626.045] +/Subtype /Link +/A << /S /GoTo /D (ch01) >> +>> endobj +733 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 601.682 240.028 610.593] +/Subtype /Link +/A << /S /GoTo /D (12) >> +>> endobj +734 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [532.999 601.682 537.981 610.593] +/Subtype /Link +/A << /S /GoTo /D (12) >> +>> endobj +735 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 588.73 287.669 597.642] +/Subtype /Link +/A << /S /GoTo /D (19) >> +>> endobj +736 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [532.999 588.73 537.981 597.642] +/Subtype /Link +/A << /S /GoTo /D (19) >> +>> endobj +737 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 577.717 308.122 584.69] +/Subtype /Link +/A << /S /GoTo /D (39) >> +>> endobj +738 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [532.999 577.717 537.981 584.69] +/Subtype /Link +/A << /S /GoTo /D (39) >> +>> endobj +739 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 562.828 298.14 571.739] +/Subtype /Link +/A << /S /GoTo /D (104) >> +>> endobj +740 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 562.828 537.981 571.739] +/Subtype /Link +/A << /S /GoTo /D (104) >> +>> endobj +741 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 551.933 270.593 558.788] +/Subtype /Link +/A << /S /GoTo /D (111) >> +>> endobj +742 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 551.933 537.981 558.788] +/Subtype /Link +/A << /S /GoTo /D (111) >> +>> endobj +743 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 538.982 309.876 545.836] +/Subtype /Link +/A << /S /GoTo /D (118) >> +>> endobj +744 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 538.982 537.981 545.836] +/Subtype /Link +/A << /S /GoTo /D (118) >> +>> endobj +745 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 526.031 216.627 532.885] +/Subtype /Link +/A << /S /GoTo /D (142) >> +>> endobj +746 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 526.031 537.981 532.885] +/Subtype /Link +/A << /S /GoTo /D (142) >> +>> endobj +747 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 513.079 303.51 519.933] +/Subtype /Link +/A << /S /GoTo /D (165) >> +>> endobj +748 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 513.079 537.981 519.933] +/Subtype /Link +/A << /S /GoTo /D (165) >> +>> endobj +749 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 498.071 303.898 506.982] +/Subtype /Link +/A << /S /GoTo /D (172) >> +>> endobj +750 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 498.071 537.981 506.982] +/Subtype /Link +/A << /S /GoTo /D (172) >> +>> endobj +751 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 487.177 277.667 494.031] +/Subtype /Link +/A << /S /GoTo /D (179) >> +>> endobj +752 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 487.177 537.981 494.031] +/Subtype /Link +/A << /S /GoTo /D (179) >> +>> endobj +753 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 474.225 283.555 481.079] +/Subtype /Link +/A << /S /GoTo /D (185) >> +>> endobj +754 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 474.225 537.981 481.079] +/Subtype /Link +/A << /S /GoTo /D (185) >> +>> endobj +755 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 459.216 283.983 468.128] +/Subtype /Link +/A << /S /GoTo /D (194) >> +>> endobj +756 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 459.216 537.981 468.128] +/Subtype /Link +/A << /S /GoTo /D (194) >> +>> endobj +757 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 446.265 269.897 455.176] +/Subtype /Link +/A << /S /GoTo /D (204) >> +>> endobj +758 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 446.265 537.981 455.176] +/Subtype /Link +/A << /S /GoTo /D (204) >> +>> endobj +759 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 433.314 320.525 442.225] +/Subtype /Link +/A << /S /GoTo /D (212) >> +>> endobj +760 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 433.314 537.981 442.225] +/Subtype /Link +/A << /S /GoTo /D (212) >> +>> endobj +761 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [119.551 418.106 258.09 426.992] +/Subtype /Link +/A << /S /GoTo /D (ch02) >> +>> endobj +762 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 418.106 537.981 426.992] +/Subtype /Link +/A << /S /GoTo /D (ch02) >> +>> endobj +763 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 402.629 254.573 411.54] +/Subtype /Link +/A << /S /GoTo /D (222) >> +>> endobj +764 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 402.629 537.981 411.54] +/Subtype /Link +/A << /S /GoTo /D (222) >> +>> endobj +765 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 389.678 238.105 398.589] +/Subtype /Link +/A << /S /GoTo /D (230) >> +>> endobj +766 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 389.678 537.981 398.589] +/Subtype /Link +/A << /S /GoTo /D (230) >> +>> endobj +767 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 376.726 253.039 385.637] +/Subtype /Link +/A << /S /GoTo /D (234) >> +>> endobj +768 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 376.726 537.981 385.637] +/Subtype /Link +/A << /S /GoTo /D (234) >> +>> endobj +769 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 365.832 328.445 372.686] +/Subtype /Link +/A << /S /GoTo /D (239) >> +>> endobj +770 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 365.832 537.981 372.686] +/Subtype /Link +/A << /S /GoTo /D (239) >> +>> endobj +771 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 350.823 279.888 359.735] +/Subtype /Link +/A << /S /GoTo /D (242) >> +>> endobj +772 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 350.823 537.981 359.735] +/Subtype /Link +/A << /S /GoTo /D (242) >> +>> endobj +773 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [119.551 335.615 243.037 344.502] +/Subtype /Link +/A << /S /GoTo /D (ch03) >> +>> endobj +774 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 335.615 537.981 344.502] +/Subtype /Link +/A << /S /GoTo /D (ch03) >> +>> endobj +775 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 320.139 252.501 329.05] +/Subtype /Link +/A << /S /GoTo /D (sample_configuration) >> +>> endobj +776 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 320.139 537.981 329.05] +/Subtype /Link +/A << /S /GoTo /D (sample_configuration) >> +>> endobj +777 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 307.187 306.668 316.099] +/Subtype /Link +/A << /S /GoTo /D (270) >> +>> endobj +778 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 307.187 537.981 316.099] +/Subtype /Link +/A << /S /GoTo /D (270) >> +>> endobj +779 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 294.236 331.176 303.147] +/Subtype /Link +/A << /S /GoTo /D (274) >> +>> endobj +780 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 294.236 537.981 303.147] +/Subtype /Link +/A << /S /GoTo /D (274) >> +>> endobj +781 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 281.284 224.258 290.196] +/Subtype /Link +/A << /S /GoTo /D (280) >> +>> endobj +782 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 281.284 537.981 290.196] +/Subtype /Link +/A << /S /GoTo /D (280) >> +>> endobj +783 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 268.333 186.908 277.244] +/Subtype /Link +/A << /S /GoTo /D (notify) >> +>> endobj +784 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 268.333 537.981 277.244] +/Subtype /Link +/A << /S /GoTo /D (notify) >> +>> endobj +785 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 255.382 255.082 264.293] +/Subtype /Link +/A << /S /GoTo /D (368) >> +>> endobj +786 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 255.382 537.981 264.293] +/Subtype /Link +/A << /S /GoTo /D (368) >> +>> endobj +787 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 244.368 369.661 251.342] +/Subtype /Link +/A << /S /GoTo /D (370) >> +>> endobj +788 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 244.368 537.981 251.342] +/Subtype /Link +/A << /S /GoTo /D (370) >> +>> endobj +789 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 229.479 291.216 238.39] +/Subtype /Link +/A << /S /GoTo /D (diagnostic_tools) >> +>> endobj +790 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 229.479 537.981 238.39] +/Subtype /Link +/A << /S /GoTo /D (diagnostic_tools) >> +>> endobj +791 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 218.585 307.425 225.439] +/Subtype /Link +/A << /S /GoTo /D (admin_tools) >> +>> endobj +792 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 218.585 537.981 225.439] +/Subtype /Link +/A << /S /GoTo /D (admin_tools) >> +>> endobj +793 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 203.576 221.618 212.487] +/Subtype /Link +/A << /S /GoTo /D (660) >> +>> endobj +794 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 203.576 537.981 212.487] +/Subtype /Link +/A << /S /GoTo /D (660) >> +>> endobj +795 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [119.551 188.368 213.817 197.255] +/Subtype /Link +/A << /S /GoTo /D (ch04) >> +>> endobj +796 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 188.368 537.981 197.255] +/Subtype /Link +/A << /S /GoTo /D (ch04) >> +>> endobj +797 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 172.891 228.681 181.803] +/Subtype /Link +/A << /S /GoTo /D (dynamic_update) >> +>> endobj +798 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 172.891 537.981 181.803] +/Subtype /Link +/A << /S /GoTo /D (dynamic_update) >> +>> endobj +799 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 159.94 253.707 168.851] +/Subtype /Link +/A << /S /GoTo /D (journal) >> +>> endobj +800 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 159.94 537.981 168.851] +/Subtype /Link +/A << /S /GoTo /D (journal) >> +>> endobj +801 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 147.362 303.021 155.9] +/Subtype /Link +/A << /S /GoTo /D (incremental_zone_transfers) >> +>> endobj +802 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 147.362 537.981 155.9] +/Subtype /Link +/A << /S /GoTo /D (incremental_zone_transfers) >> +>> endobj +803 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 134.037 202.141 142.948] +/Subtype /Link +/A << /S /GoTo /D (724) >> +>> endobj +804 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 134.037 537.981 142.948] +/Subtype /Link +/A << /S /GoTo /D (724) >> +>> endobj +725 0 obj << +/D [727 0 R /XYZ 71.731 705.352 null] +>> endobj +1 0 obj << +/D [727 0 R /XYZ 244.331 642.462 null] +>> endobj +726 0 obj << +/Font << /F29 669 0 R /F30 731 0 R /F25 681 0 R /F31 805 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +849 0 obj << +/Length 850 0 R +/Filter /FlateDecode +>> +stream +xKsW,+oԽ=L&T +żتR%3ߏ&Xƴ%0| +$#V]='$_qg0[!+ҝ;~Z|myt^!-?rܽC{0 ÝO,^u|/=CvE2x_//^A'>5P"p*TPB +K'*TPwuAggWMltcvG7^?`֍~s<7i_B +*TPtRB +*TH'/fc\;ټ=ݟ_wEB_g͇4u,{mi(ZQtRB +*T +*TPBBK:Pb.v\r^_y+ͬRp:{E#4:QyLETPB +*ZI +*TP"TPBmޏN?_g˲mPU#$>LPB +*TN +*TPb%TPB +*I +*TP"p*TPV}Z:E\E籟s; u]βœ"#*TP8*TPBtRB +*TH'*TPBPB +*T, +o$ +8"Iŕ}$EI{f;/PB +,N*TPB +B +*TP8*TPBtRB +*TH'^e}$C2^ uDpRB +*B +*TPВN*TPB +B +*TP8;L8߬J5JҮaoZW,BUwL ~N)*TPBETPB +*ZI +*TP"TPB +* kO^/jƇQU~-*pU^b(c{U{N3TPB +*B +*TPВN*TPB +BlT)RlRdnM"NزPd,bԤYJ +*B +*TPВN*TPB +B +*TP8*T`b;SUOaN;<'+yF?s*TP"TPB +*B +*TPВN*TPB +5l.]I[]K3>y5m=[vk׻4j +y֣t#8 + +*TPBBK:PB +*TN +*TL}ۋ7Ϯj?.|>:i.닖gG;]i{2*EէsTPBPB +*T,ʭU3Wu]fey{+iXR/b ;B>PB +*T +*TPrZl":Ew/=͊_>7gGtrv=2L_-kvbpޝygҭ&-!M>UZubRB +*TN +*!y !?۸]C(B_WȔ%{~QOϞ?Z>lzz7&]^[*}jZ{mv/.pB +*TPN*TPB +SB +*TXhI'*T|*}˸z<=&C˸!^.znήeVyV[;j\UĕU[ԥ +*T +*TPBBK:PB +*T +*TPT6$u!JP^N&݈G'FtR{ICjO|(PB +*H'*TPBtRB +*T +*TPbV6*n>-as2|`D|rWMA3oVj:3iD=*TPBtRB +*T +*TPBBK:PB +oy\Qk;KoXV^ [-4{2>=>nN|+ YѧW,2 ԉRB +*?J'*TPBtRB +*T +*TPB Ǹ~0 e8Ƴm3-<-c<^+,tKNl9PB +SB +*TXhI'*TPBtRB +*Tɂ[ e]u[Z[%[{o&џ#)y^ZiLrR + +*TPBBK:PB +*T +*TPBETPB +*_F芵ѽ-nܬ]7>jM~~tqHjS/.B +*TPN*TPB +SB +*TXh}}[yI3_C6Qð5yrQB +*I +*TP"TPB +*B +*TPВXngb\ Ǐ +*TPBETPB +*N*TPB +B +*TP8Q78;\Џ<*TPBETPB +*ZI +*TP"p*TSzCqu8w'wOvtr2jWtŤ*pRN*TPB +B +*TP8*TPBtRBQ5\hk2Xv|%ﾟNA}̻#ysLyIHa]T9PB +SB +*TXhI'*TPBtRB +*w}UU> B +*TH'*TPBPB +*T,{7ﰰXCRdj0=TPZ-iHcy'nB +*TP8*TPB +*TPBE:u W!3zM(yLu⑉endstream +endobj +850 0 obj +3858 +endobj +848 0 obj << +/Type /Page +/Contents 849 0 R +/Resources 847 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 670 0 R +/Annots [ 851 0 R 852 0 R 853 0 R 854 0 R 855 0 R 856 0 R 857 0 R 858 0 R 859 0 R 860 0 R 861 0 R 862 0 R 863 0 R 864 0 R 865 0 R 866 0 R 867 0 R 868 0 R 869 0 R 870 0 R 871 0 R 872 0 R 873 0 R 874 0 R 875 0 R 876 0 R 877 0 R 878 0 R 879 0 R 880 0 R 881 0 R 882 0 R 883 0 R 884 0 R 885 0 R 886 0 R 887 0 R 888 0 R 889 0 R 890 0 R 891 0 R 892 0 R 893 0 R 894 0 R 895 0 R 896 0 R 897 0 R 898 0 R 899 0 R 900 0 R 901 0 R 902 0 R 903 0 R 904 0 R 905 0 R 906 0 R 907 0 R 908 0 R 909 0 R 910 0 R 911 0 R 912 0 R 913 0 R 914 0 R 915 0 R 916 0 R 917 0 R 918 0 R 919 0 R 920 0 R 921 0 R 922 0 R 923 0 R 924 0 R 925 0 R 926 0 R 927 0 R 928 0 R 929 0 R 930 0 R ] +>> endobj +851 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 647.47 183.033 654.324] +/Subtype /Link +/A << /S /GoTo /D (tsig) >> +>> endobj +852 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 647.47 537.981 654.324] +/Subtype /Link +/A << /S /GoTo /D (tsig) >> +>> endobj +853 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 632.461 371.833 641.372] +/Subtype /Link +/A << /S /GoTo /D (815) >> +>> endobj +854 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 632.461 537.981 641.372] +/Subtype /Link +/A << /S /GoTo /D (815) >> +>> endobj +855 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 621.567 312.476 628.421] +/Subtype /Link +/A << /S /GoTo /D (820) >> +>> endobj +856 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 621.567 537.981 628.421] +/Subtype /Link +/A << /S /GoTo /D (820) >> +>> endobj +857 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 608.616 300.85 615.47] +/Subtype /Link +/A << /S /GoTo /D (831) >> +>> endobj +858 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 608.616 537.981 615.47] +/Subtype /Link +/A << /S /GoTo /D (831) >> +>> endobj +859 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 593.607 372.042 602.518] +/Subtype /Link +/A << /S /GoTo /D (836) >> +>> endobj +860 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 593.607 537.981 602.518] +/Subtype /Link +/A << /S /GoTo /D (836) >> +>> endobj +861 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 580.655 371.574 589.567] +/Subtype /Link +/A << /S /GoTo /D (839) >> +>> endobj +862 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 580.655 537.981 589.567] +/Subtype /Link +/A << /S /GoTo /D (839) >> +>> endobj +863 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 567.704 340.032 576.615] +/Subtype /Link +/A << /S /GoTo /D (851) >> +>> endobj +864 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 567.704 537.981 576.615] +/Subtype /Link +/A << /S /GoTo /D (851) >> +>> endobj +865 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 554.753 323.584 563.664] +/Subtype /Link +/A << /S /GoTo /D (867) >> +>> endobj +866 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 554.753 537.981 563.664] +/Subtype /Link +/A << /S /GoTo /D (867) >> +>> endobj +867 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 543.859 217.175 550.713] +/Subtype /Link +/A << /S /GoTo /D (880) >> +>> endobj +868 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 543.859 537.981 550.713] +/Subtype /Link +/A << /S /GoTo /D (880) >> +>> endobj +869 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 530.907 187.456 537.761] +/Subtype /Link +/A << /S /GoTo /D (884) >> +>> endobj +870 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 530.907 537.981 537.761] +/Subtype /Link +/A << /S /GoTo /D (884) >> +>> endobj +871 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 516.272 188.562 524.81] +/Subtype /Link +/A << /S /GoTo /D (899) >> +>> endobj +872 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 516.272 537.981 524.81] +/Subtype /Link +/A << /S /GoTo /D (899) >> +>> endobj +873 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 505.004 199.092 511.859] +/Subtype /Link +/A << /S /GoTo /D (DNSSEC) >> +>> endobj +874 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 505.004 537.981 511.859] +/Subtype /Link +/A << /S /GoTo /D (DNSSEC) >> +>> endobj +875 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 489.996 259.107 498.907] +/Subtype /Link +/A << /S /GoTo /D (916) >> +>> endobj +876 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 489.996 537.981 498.907] +/Subtype /Link +/A << /S /GoTo /D (916) >> +>> endobj +877 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 477.044 263.261 485.956] +/Subtype /Link +/A << /S /GoTo /D (936) >> +>> endobj +878 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 477.044 537.981 485.956] +/Subtype /Link +/A << /S /GoTo /D (936) >> +>> endobj +879 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 464.093 298.987 473.004] +/Subtype /Link +/A << /S /GoTo /D (948) >> +>> endobj +880 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 464.093 537.981 473.004] +/Subtype /Link +/A << /S /GoTo /D (948) >> +>> endobj +881 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 451.142 260.91 460.053] +/Subtype /Link +/A << /S /GoTo /D (961) >> +>> endobj +882 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 451.142 537.981 460.053] +/Subtype /Link +/A << /S /GoTo /D (961) >> +>> endobj +883 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 438.19 272.656 447.102] +/Subtype /Link +/A << /S /GoTo /D (977) >> +>> endobj +884 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 438.19 537.981 447.102] +/Subtype /Link +/A << /S /GoTo /D (977) >> +>> endobj +885 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 425.239 258.31 434.15] +/Subtype /Link +/A << /S /GoTo /D (984) >> +>> endobj +886 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 425.239 537.981 434.15] +/Subtype /Link +/A << /S /GoTo /D (984) >> +>> endobj +887 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 412.287 354.976 421.199] +/Subtype /Link +/A << /S /GoTo /D (1000) >> +>> endobj +888 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 412.287 537.981 421.199] +/Subtype /Link +/A << /S /GoTo /D (1000) >> +>> endobj +889 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 399.336 338.379 408.247] +/Subtype /Link +/A << /S /GoTo /D (1005) >> +>> endobj +890 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 399.336 537.981 408.247] +/Subtype /Link +/A << /S /GoTo /D (1005) >> +>> endobj +891 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 388.442 266.001 395.296] +/Subtype /Link +/A << /S /GoTo /D (1009) >> +>> endobj +892 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 388.442 537.981 395.296] +/Subtype /Link +/A << /S /GoTo /D (1009) >> +>> endobj +893 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 375.491 339.714 382.345] +/Subtype /Link +/A << /S /GoTo /D (1020) >> +>> endobj +894 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 375.491 537.981 382.345] +/Subtype /Link +/A << /S /GoTo /D (1020) >> +>> endobj +895 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 360.482 385.83 369.393] +/Subtype /Link +/A << /S /GoTo /D (1026) >> +>> endobj +896 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 360.482 537.981 369.393] +/Subtype /Link +/A << /S /GoTo /D (1026) >> +>> endobj +897 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 347.531 393.591 356.442] +/Subtype /Link +/A << /S /GoTo /D (1033) >> +>> endobj +898 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 347.531 537.981 356.442] +/Subtype /Link +/A << /S /GoTo /D (1033) >> +>> endobj +899 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 334.579 424.574 343.491] +/Subtype /Link +/A << /S /GoTo /D (1040) >> +>> endobj +900 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 334.579 537.981 343.491] +/Subtype /Link +/A << /S /GoTo /D (1040) >> +>> endobj +901 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [119.551 319.371 274.418 328.258] +/Subtype /Link +/A << /S /GoTo /D (ch05) >> +>> endobj +902 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 319.371 537.981 328.258] +/Subtype /Link +/A << /S /GoTo /D (ch05) >> +>> endobj +903 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 303.894 297.711 312.806] +/Subtype /Link +/A << /S /GoTo /D (1060) >> +>> endobj +904 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 303.894 537.981 312.806] +/Subtype /Link +/A << /S /GoTo /D (1060) >> +>> endobj +905 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 290.943 276.133 299.854] +/Subtype /Link +/A << /S /GoTo /D (lwresd) >> +>> endobj +906 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 290.943 537.981 299.854] +/Subtype /Link +/A << /S /GoTo /D (lwresd) >> +>> endobj +907 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [119.551 275.735 269.069 284.622] +/Subtype /Link +/A << /S /GoTo /D (ch06) >> +>> endobj +908 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 275.735 537.981 284.622] +/Subtype /Link +/A << /S /GoTo /D (ch06) >> +>> endobj +909 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 260.258 273.811 269.17] +/Subtype /Link +/A << /S /GoTo /D (configuration_file_elements) >> +>> endobj +910 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 260.258 537.981 269.17] +/Subtype /Link +/A << /S /GoTo /D (configuration_file_elements) >> +>> endobj +911 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 249.364 274.738 256.218] +/Subtype /Link +/A << /S /GoTo /D (address_match_lists) >> +>> endobj +912 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 249.364 537.981 256.218] +/Subtype /Link +/A << /S /GoTo /D (address_match_lists) >> +>> endobj +913 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 234.355 251.336 243.267] +/Subtype /Link +/A << /S /GoTo /D (1264) >> +>> endobj +914 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 234.355 537.981 243.267] +/Subtype /Link +/A << /S /GoTo /D (1264) >> +>> endobj +915 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 221.404 308.332 230.315] +/Subtype /Link +/A << /S /GoTo /D (1272) >> +>> endobj +916 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 221.404 537.981 230.315] +/Subtype /Link +/A << /S /GoTo /D (1272) >> +>> endobj +917 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 208.453 261.747 217.364] +/Subtype /Link +/A << /S /GoTo /D (1301) >> +>> endobj +918 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 208.453 537.981 217.364] +/Subtype /Link +/A << /S /GoTo /D (1301) >> +>> endobj +919 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 195.501 251.336 204.413] +/Subtype /Link +/A << /S /GoTo /D (1306) >> +>> endobj +920 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 195.501 537.981 204.413] +/Subtype /Link +/A << /S /GoTo /D (1306) >> +>> endobj +921 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 182.55 308.332 191.461] +/Subtype /Link +/A << /S /GoTo /D (1315) >> +>> endobj +922 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 182.55 537.981 191.461] +/Subtype /Link +/A << /S /GoTo /D (1315) >> +>> endobj +923 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 169.598 274.907 178.51] +/Subtype /Link +/A << /S /GoTo /D (Configuration_File_Grammar) >> +>> endobj +924 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 169.598 537.981 178.51] +/Subtype /Link +/A << /S /GoTo /D (Configuration_File_Grammar) >> +>> endobj +925 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 158.684 287.46 165.668] +/Subtype /Link +/A << /S /GoTo /D (1408) >> +>> endobj +926 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 158.684 537.981 165.668] +/Subtype /Link +/A << /S /GoTo /D (1408) >> +>> endobj +927 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 143.696 333.955 152.717] +/Subtype /Link +/A << /S /GoTo /D (acl) >> +>> endobj +928 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 143.696 537.981 152.717] +/Subtype /Link +/A << /S /GoTo /D (acl) >> +>> endobj +929 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 132.682 309.417 139.765] +/Subtype /Link +/A << /S /GoTo /D (1455) >> +>> endobj +930 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 132.682 537.981 139.765] +/Subtype /Link +/A << /S /GoTo /D (1455) >> +>> endobj +846 0 obj << +/D [848 0 R /XYZ 71.731 705.352 null] +>> endobj +847 0 obj << +/Font << /F25 681 0 R /F30 731 0 R /F31 805 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +974 0 obj << +/Length 975 0 R +/Filter /FlateDecode +>> +stream +x[o` z<>6I-`V^R62aHځKt횴D &!&\}g i1\sӬJB4'O*Tuvwrw˓ū|5)_ًuot5See(tqi^ӴHz? h8^/~/ۻz׳Xס׿4}_p֛^nGT|*I +*TP"p*TPB +KցoKf6(/7ePLCS- +Bź`x~]Q fo6_] iܿ>YXu)$*TPBPB +*TL +*TPBET.=EtK=Ul/gǴnPSB +*B +*TPtRB +*T>j]iLzr+V!ˊ͖,.U(XTP"p*TPB +-B +*TPN*TPB +C5<ξ}jY=olBZtxq^}B +*N*TPB +B +*TP8af]jty9^:HB +*TL +*TPBETPB +*N*Tg"nxLV5M4]&6['Bw;(?TPBE:PB +*TN +*TPqjP2Wܩüds]. EtjEaUaor'VL:s5[TPB +*&ZI +*TP"TPB +*34[pieW~ПS&$IpSC桃?65Ĭ%1H +*TP8*TPBtRB +*T;Y;Yi{d@"tCcm +*TP"p*TPB +-B +*TP8*TUevr?[m3A^P?-(X:W *TPpI +*TP"TPB +c7jw_+n]+dY6{fk,MzcboI +*TP"TPB +*B +*TP16{iG|ރl t$dMօ7{( ~}H'*TPBPB +*T.TIAՠ eӆya^~4j͋5+$y(!I +*TP8*TPB +*TPBE:PB.~ZV\.ZVƢ7~^|1^~VII}'ETPB +SB +*ThI'*TPBPB +*IeQ+ܨUﶷPhy;̯ R\u(KV#! #*TP8*TPBtRB +*TH'*TPB#4ߔ%)_445ETĕN*TPB +B +*TP8*TPBtRY{l{Sȶr{hίN7'U 8EuiI̱mtRB +*T +*TPBDK:PB +*T +*M-7 8/nq:^Lu^ѻ?t1'VUIΜ4ucTP8*TPB +*TP&v_/XqoҺ"SڦF-*TPBE:PB +*TN +*TPb%G[ڎ)mot(mlǃ&uO<Ӳ׿Zoun8F*TPN*TPB +SB +*ThI'*Tv:p-$|R.݌FGT'!;$/p*TPB +-B +*TPN*TPB +SB +ol*[7] ϞgǛ$dixz5L}Y u?mt!"MuqLbtLB +*TPN*TPB +SB +*TBWv:|1hǃR);)B,N6D1P"pB +*TPN*TPB +SB +*T$?lIh4y};iß_/Wvr3NRI"Nr}'>PBwI +*TP"TPB +*rfJhqؽ{ϟEpJ* eթYQ4|PBPB +*TL +*TPBE:PB +*TN1qSnۭG wo$t0'WҮ#6* ϐB +*TH'*TPBPB +*TL +*ǪSsE.w %.,xCڄwMiLz~]鳥u-nyWOuԘ5!.#UҘs +*TN +*TP"pB +*TPN*TPrL'Am;Qwb)?w`: Ë/VuSJж~ԩPBE:PB +*TN +*TPb%T=UǩfUYɴxu;YgZYYw-nSBPB +*TL +*TPBE:PbflFʳ^ʳ-7Lc u7CUJTPtRB +*TH'*TPqF.k-:z?h_q^ uQmx`nY8B +*I +*TP"p*TPB +-B +*TyHY&;̛/Xա*o6 I[ihRBtRB +*T +*TPBEܪ8GGq]?aYUYUܥ: +*TPBE:PB +*TN +*TPb%TCn븟Cnj$Iry#F~˸ MXTPB +*I +*TP"p*TPr[&yfMü8]Nz7,UšN*2È +*TP{B +*TP8*TPB +*TPvQkmEWRkVQ;SYF<*B +*TP1ђN*TPB +B +*TP8*TP
LwNxT݋㡖(*j4PB +*I +*TP"p*TPB +-B +*T~G;r3_#x;_~OI/GW1kBN JQfRB +*B +*TP1ђN*TPB +T-efCU5n$,cr,Y\!+]#endstream +endobj +975 0 obj +3483 +endobj +973 0 obj << +/Type /Page +/Contents 974 0 R +/Resources 972 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 1056 0 R +/Annots [ 976 0 R 977 0 R 978 0 R 979 0 R 980 0 R 981 0 R 982 0 R 983 0 R 984 0 R 985 0 R 986 0 R 987 0 R 988 0 R 989 0 R 990 0 R 991 0 R 992 0 R 993 0 R 994 0 R 995 0 R 996 0 R 997 0 R 998 0 R 999 0 R 1000 0 R 1001 0 R 1002 0 R 1003 0 R 1004 0 R 1005 0 R 1006 0 R 1007 0 R 1008 0 R 1009 0 R 1010 0 R 1011 0 R 1012 0 R 1013 0 R 1014 0 R 1015 0 R 1016 0 R 1017 0 R 1018 0 R 1019 0 R 1020 0 R 1021 0 R 1022 0 R 1023 0 R 1024 0 R 1025 0 R 1026 0 R 1027 0 R 1028 0 R 1029 0 R 1030 0 R 1031 0 R 1032 0 R 1033 0 R 1034 0 R 1035 0 R 1036 0 R 1037 0 R 1038 0 R 1039 0 R 1040 0 R 1041 0 R 1042 0 R 1043 0 R 1044 0 R 1045 0 R 1046 0 R 1047 0 R 1048 0 R 1049 0 R 1050 0 R 1051 0 R 1052 0 R 1053 0 R 1054 0 R 1055 0 R ] +>> endobj +976 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 645.412 355.912 654.433] +/Subtype /Link +/A << /S /GoTo /D (controls_statement_definition_and_usage) >> +>> endobj +977 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 645.412 537.981 654.433] +/Subtype /Link +/A << /S /GoTo /D (controls_statement_definition_and_usage) >> +>> endobj +978 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 634.498 306.289 641.482] +/Subtype /Link +/A << /S /GoTo /D (1528) >> +>> endobj +979 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 634.498 537.981 641.482] +/Subtype /Link +/A << /S /GoTo /D (1528) >> +>> endobj +980 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 619.51 352.784 628.531] +/Subtype /Link +/A << /S /GoTo /D (1533) >> +>> endobj +981 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 619.51 537.981 628.531] +/Subtype /Link +/A << /S /GoTo /D (1533) >> +>> endobj +982 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 606.693 290.13 615.579] +/Subtype /Link +/A << /S /GoTo /D (1540) >> +>> endobj +983 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 606.693 537.981 615.579] +/Subtype /Link +/A << /S /GoTo /D (1540) >> +>> endobj +984 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 593.607 336.625 602.628] +/Subtype /Link +/A << /S /GoTo /D (1547) >> +>> endobj +985 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 593.607 537.981 602.628] +/Subtype /Link +/A << /S /GoTo /D (1547) >> +>> endobj +986 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 580.79 306.289 589.677] +/Subtype /Link +/A << /S /GoTo /D (1565) >> +>> endobj +987 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 580.79 537.981 589.677] +/Subtype /Link +/A << /S /GoTo /D (1565) >> +>> endobj +988 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 567.704 357.765 576.725] +/Subtype /Link +/A << /S /GoTo /D (1605) >> +>> endobj +989 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 567.704 537.981 576.725] +/Subtype /Link +/A << /S /GoTo /D (1605) >> +>> endobj +990 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 556.79 308.89 563.774] +/Subtype /Link +/A << /S /GoTo /D (1621) >> +>> endobj +991 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 556.79 537.981 563.774] +/Subtype /Link +/A << /S /GoTo /D (1621) >> +>> endobj +992 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 541.936 312.188 550.713] +/Subtype /Link +/A << /S /GoTo /D (the_category_phrase) >> +>> endobj +993 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 541.936 537.981 550.713] +/Subtype /Link +/A << /S /GoTo /D (the_category_phrase) >> +>> endobj +994 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 530.887 302.772 537.871] +/Subtype /Link +/A << /S /GoTo /D (1836) >> +>> endobj +995 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 530.887 537.981 537.871] +/Subtype /Link +/A << /S /GoTo /D (1836) >> +>> endobj +996 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 515.899 349.267 524.92] +/Subtype /Link +/A << /S /GoTo /D (1860) >> +>> endobj +997 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 515.899 537.981 524.92] +/Subtype /Link +/A << /S /GoTo /D (1860) >> +>> endobj +998 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 503.082 311.27 511.968] +/Subtype /Link +/A << /S /GoTo /D (1879) >> +>> endobj +999 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 503.082 537.981 511.968] +/Subtype /Link +/A << /S /GoTo /D (1879) >> +>> endobj +1000 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 489.996 357.765 499.017] +/Subtype /Link +/A << /S /GoTo /D (2097) >> +>> endobj +1001 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 489.996 537.981 499.017] +/Subtype /Link +/A << /S /GoTo /D (2097) >> +>> endobj +1002 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 477.044 295.889 485.956] +/Subtype /Link +/A << /S /GoTo /D (boolean_options) >> +>> endobj +1003 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 477.044 537.981 485.956] +/Subtype /Link +/A << /S /GoTo /D (boolean_options) >> +>> endobj +1004 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 464.093 274.878 473.004] +/Subtype /Link +/A << /S /GoTo /D (2397) >> +>> endobj +1005 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 464.093 537.981 473.004] +/Subtype /Link +/A << /S /GoTo /D (2397) >> +>> endobj +1006 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 453.199 289.791 460.053] +/Subtype /Link +/A << /S /GoTo /D (access_control) >> +>> endobj +1007 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 453.199 537.981 460.053] +/Subtype /Link +/A << /S /GoTo /D (access_control) >> +>> endobj +1008 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 440.248 267.814 447.102] +/Subtype /Link +/A << /S /GoTo /D (2462) >> +>> endobj +1009 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 440.248 537.981 447.102] +/Subtype /Link +/A << /S /GoTo /D (2462) >> +>> endobj +1010 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 425.239 288.676 434.15] +/Subtype /Link +/A << /S /GoTo /D (2488) >> +>> endobj +1011 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 425.239 537.981 434.15] +/Subtype /Link +/A << /S /GoTo /D (2488) >> +>> endobj +1012 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 414.345 288.875 421.199] +/Subtype /Link +/A << /S /GoTo /D (zone_transfers) >> +>> endobj +1013 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 414.345 537.981 421.199] +/Subtype /Link +/A << /S /GoTo /D (zone_transfers) >> +>> endobj +1014 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 399.336 368.385 408.247] +/Subtype /Link +/A << /S /GoTo /D (2627) >> +>> endobj +1015 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 399.336 537.981 408.247] +/Subtype /Link +/A << /S /GoTo /D (2627) >> +>> endobj +1016 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 388.442 322.568 395.296] +/Subtype /Link +/A << /S /GoTo /D (2664) >> +>> endobj +1017 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 388.442 537.981 395.296] +/Subtype /Link +/A << /S /GoTo /D (2664) >> +>> endobj +1018 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 375.371 320.008 382.345] +/Subtype /Link +/A << /S /GoTo /D (2692) >> +>> endobj +1019 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 375.371 537.981 382.345] +/Subtype /Link +/A << /S /GoTo /D (2692) >> +>> endobj +1020 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 360.482 271.571 369.393] +/Subtype /Link +/A << /S /GoTo /D (topology) >> +>> endobj +1021 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 360.482 537.981 369.393] +/Subtype /Link +/A << /S /GoTo /D (topology) >> +>> endobj +1022 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 349.568 323.275 356.552] +/Subtype /Link +/A << /S /GoTo /D (the_sortlist_statement) >> +>> endobj +1023 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 349.568 537.981 356.552] +/Subtype /Link +/A << /S /GoTo /D (the_sortlist_statement) >> +>> endobj +1024 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 334.579 296.437 343.491] +/Subtype /Link +/A << /S /GoTo /D (rrset_ordering) >> +>> endobj +1025 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 334.579 537.981 343.491] +/Subtype /Link +/A << /S /GoTo /D (rrset_ordering) >> +>> endobj +1026 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 321.628 333.796 330.539] +/Subtype /Link +/A << /S /GoTo /D (synthesis) >> +>> endobj +1027 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 321.628 537.981 330.539] +/Subtype /Link +/A << /S /GoTo /D (synthesis) >> +>> endobj +1028 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 308.676 261.957 317.588] +/Subtype /Link +/A << /S /GoTo /D (tuning) >> +>> endobj +1029 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 308.676 537.981 317.588] +/Subtype /Link +/A << /S /GoTo /D (tuning) >> +>> endobj +1030 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 297.782 305.582 304.636] +/Subtype /Link +/A << /S /GoTo /D (statsfile) >> +>> endobj +1031 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 297.782 537.981 304.636] +/Subtype /Link +/A << /S /GoTo /D (statsfile) >> +>> endobj +1032 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 284.811 306.618 291.685] +/Subtype /Link +/A << /S /GoTo /D (server_statement_grammar) >> +>> endobj +1033 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 284.811 537.981 291.685] +/Subtype /Link +/A << /S /GoTo /D (server_statement_grammar) >> +>> endobj +1034 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 269.822 353.113 278.734] +/Subtype /Link +/A << /S /GoTo /D (server_statement_definition_and_usage) >> +>> endobj +1035 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 269.822 537.981 278.734] +/Subtype /Link +/A << /S /GoTo /D (server_statement_definition_and_usage) >> +>> endobj +1036 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 257.005 332.739 265.892] +/Subtype /Link +/A << /S /GoTo /D (3010) >> +>> endobj +1037 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 257.005 537.981 265.892] +/Subtype /Link +/A << /S /GoTo /D (3010) >> +>> endobj +1038 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 243.919 379.234 252.941] +/Subtype /Link +/A << /S /GoTo /D (3026) >> +>> endobj +1039 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 243.919 537.981 252.941] +/Subtype /Link +/A << /S /GoTo /D (3026) >> +>> endobj +1040 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 232.906 299.634 239.989] +/Subtype /Link +/A << /S /GoTo /D (3034) >> +>> endobj +1041 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 232.906 537.981 239.989] +/Subtype /Link +/A << /S /GoTo /D (3034) >> +>> endobj +1042 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 218.017 346.129 227.038] +/Subtype /Link +/A << /S /GoTo /D (3050) >> +>> endobj +1043 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 218.017 537.981 227.038] +/Subtype /Link +/A << /S /GoTo /D (3050) >> +>> endobj +1044 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 207.103 299.634 213.977] +/Subtype /Link +/A << /S /GoTo /D (zone_statement_grammar) >> +>> endobj +1045 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 207.103 537.981 213.977] +/Subtype /Link +/A << /S /GoTo /D (zone_statement_grammar) >> +>> endobj +1046 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 192.114 346.129 201.025] +/Subtype /Link +/A << /S /GoTo /D (3201) >> +>> endobj +1047 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 192.114 537.981 201.025] +/Subtype /Link +/A << /S /GoTo /D (3201) >> +>> endobj +1048 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 179.163 275.157 188.074] +/Subtype /Link +/A << /S /GoTo /D (3204) >> +>> endobj +1049 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 179.163 537.981 188.074] +/Subtype /Link +/A << /S /GoTo /D (3204) >> +>> endobj +1050 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 168.268 250.23 175.123] +/Subtype /Link +/A << /S /GoTo /D (3261) >> +>> endobj +1051 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 168.268 537.981 175.123] +/Subtype /Link +/A << /S /GoTo /D (3261) >> +>> endobj +1052 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 153.26 283.157 162.171] +/Subtype /Link +/A << /S /GoTo /D (3271) >> +>> endobj +1053 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 153.26 537.981 162.171] +/Subtype /Link +/A << /S /GoTo /D (3271) >> +>> endobj +1054 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 140.308 330.468 149.22] +/Subtype /Link +/A << /S /GoTo /D (dynamic_update_policies) >> +>> endobj +1055 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 140.308 537.981 149.22] +/Subtype /Link +/A << /S /GoTo /D (dynamic_update_policies) >> +>> endobj +971 0 obj << +/D [973 0 R /XYZ 71.731 705.352 null] +>> endobj +972 0 obj << +/Font << /F25 681 0 R /F30 731 0 R /F31 805 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1097 0 obj << +/Length 1098 0 R +/Filter /FlateDecode +>> +stream +x[oG``+Jev3J +M%"CR/njSZǶ&n}XN;OSơL_&:sYQ$_rޣB;2Uzy~two}ա.ҝWnҰ(ɣݿӦ;z19k迿y~t2y=8TUnLB{q*_tRB +*T +*TPBEI +*TP"TPB +*B +*_JwVnM" EuIA,;s/F +*TP"p*TPB +-B +*Tɧk7}6/V?ܩ MCT٠OTP8*TPB +*TPb%TPB +p#:Ne&{V}efLOG߯Հe(jA@ܝTH'*TPBPB +*T +*TPBE:PB +*_NOuNO +ӽ,rcf6o`7o B'}Y7NP4Ó TP"TPB +*B +*TPtRB +*T${Un?aq̺ÿӾrYm7kƋ^R57P<: TP8*TPB +*TPb%TPB +D"}C׉6aE< +itg{lyI$y_O^N8Kbbq/RB +*I +*TP"p*TPB +-2|º҄u2xºW[ |z|}:OC6|:[*eyR>BE:PB +*TN +*TPb%TPvf7[K_K|tlR,*OYAqu'RBPB +*T +*TPBE:PB׮A(Q}E_dYw|̺fyջhI;o7Upy^p|:=a}4MI^,,/-wSN*TPB +S9~GqCK4}Gue3MoGO|rFh^{z]g~S[{}SBPB +*T.TPB +-B +*TP8*HehEM/Vpu_Җޱ q_xn.ַ*HGeٟ8p"Tg̛˔гFyY`Ѽu 3YR(Lu{SBPB +*T +*TPBE:PBf)iڪBJ JVͭN, UUߵt/vy3]$R)R_V?C +*TP8*TPBtRB +*TH'*TPbJEkY*4HjIwpCu^/.ŤW!š,mUxSSBߩ +*TPBE:PB +*TN +*TP>Jڗ|<{;O}ǛѢ_6dvr# +u(ATnxSB +*T +*TPBETPB +*ZI +*TP1ZmUҝGqw^Z텵WffWΚi.. TY4U %pn|y8*TPBPB +*T +*TPBETPB +*B +*T< mcoz8J66**}-v?Wg/U VXaPBtRB +*T +*TPB@K:PB +*T +*T<=QZZF.KKP盔KѴ]tn;y2=v/ɗ_=.Qzw{/';l/V8ڿ Y9QZmrmD +*B +*TP1ВNklkwZ1}V|珵ՇZ9oW[]=3OGӓ]ZlfYw|pqEYHd[Za톦B +*T +*TPB@K:PB?֫9=S:mVOG}c߾ۦ9{V,񠙻8DUVmKPB +*T +*TPBE:PB +*TN +*T:,B\!re7odC\ j‘-|7ELv`/N +*TPN*TPB +SB +*ThI'*TPBtRB +*T +/^e*^,-t\q}֩kU~=kOM][RW}:{}-#B +*TN +*TPb%TPB +*I +*TP"p*T|*#~ +2 +EtY:ߤBFd6i^wMv{O.NAV>;x=3t7R!Iۿ$D:yTPB +B +*TP8*TPlGFH񝧑d<:_OO_bNׂL㐥QQjᖥB +*T +*TPBETPB +*ZI +*TPLUWTli&OUOFy3f^fΎ^dCxꢲRޝK +*TP8*TPBtRB +*Tg;_om?2['4osݗf>iߏJC/A9fB +*TPBETPB +*B +*TP1ВN*_'srUt>'~]iRšAe +B +*TP8*TPBtRB +*TH'*_'ǛWS}%շʹκ^6z̚jרvv>ZLZElPU( +*B +*TPtRB +*T +*T|;G}2 +ETosTzoxͼ{(kQFe|KQUfݲTPBPB +*T +*TPBE:PB +*Zo|l$닟L^Mړ&'Uhzˋfh(x2=$BjJaPB +*T.TPB +*I +*TP"?]}^uo|=xκy3]̻I O7rw$C2>B +*TN +*TPb%TPB +*IUdYpIorB|$$Eb +:*TP"p*TPB +-B +*TPN*TPB +Su|@=vI@E36}gg> endobj +1099 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 647.35 199.361 654.324] +/Subtype /Link +/A << /S /GoTo /D (3535) >> +>> endobj +1100 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 647.35 537.981 654.324] +/Subtype /Link +/A << /S /GoTo /D (3535) >> +>> endobj +1101 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 632.461 398.433 641.372] +/Subtype /Link +/A << /S /GoTo /D (types_of_resource_records_and_when_to_use_them) >> +>> endobj +1102 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 632.461 537.981 641.372] +/Subtype /Link +/A << /S /GoTo /D (types_of_resource_records_and_when_to_use_them) >> +>> endobj +1103 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 621.447 295.869 628.421] +/Subtype /Link +/A << /S /GoTo /D (3540) >> +>> endobj +1104 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 621.447 537.981 628.421] +/Subtype /Link +/A << /S /GoTo /D (3540) >> +>> endobj +1105 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 606.558 328.645 615.47] +/Subtype /Link +/A << /S /GoTo /D (3789) >> +>> endobj +1106 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 606.558 537.981 615.47] +/Subtype /Link +/A << /S /GoTo /D (3789) >> +>> endobj +1107 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 595.545 300.471 602.518] +/Subtype /Link +/A << /S /GoTo /D (3890) >> +>> endobj +1108 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 595.545 537.981 602.518] +/Subtype /Link +/A << /S /GoTo /D (3890) >> +>> endobj +1109 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 580.655 245.139 589.567] +/Subtype /Link +/A << /S /GoTo /D (Setting_TTLs) >> +>> endobj +1110 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 580.655 537.981 589.567] +/Subtype /Link +/A << /S /GoTo /D (Setting_TTLs) >> +>> endobj +1111 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 567.704 291.067 576.615] +/Subtype /Link +/A << /S /GoTo /D (4011) >> +>> endobj +1112 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 567.704 537.981 576.615] +/Subtype /Link +/A << /S /GoTo /D (4011) >> +>> endobj +1113 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 556.69 298.957 563.664] +/Subtype /Link +/A << /S /GoTo /D (4038) >> +>> endobj +1114 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 556.69 537.981 563.664] +/Subtype /Link +/A << /S /GoTo /D (4038) >> +>> endobj +1115 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 543.007 323.425 551.43] +/Subtype /Link +/A << /S /GoTo /D (4045) >> +>> endobj +1116 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 543.007 537.981 551.43] +/Subtype /Link +/A << /S /GoTo /D (4045) >> +>> endobj +1117 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 530.055 331.724 538.479] +/Subtype /Link +/A << /S /GoTo /D (4065) >> +>> endobj +1118 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 530.055 537.981 538.479] +/Subtype /Link +/A << /S /GoTo /D (4065) >> +>> endobj +1119 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 517.104 305.722 525.527] +/Subtype /Link +/A << /S /GoTo /D (4085) >> +>> endobj +1120 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 517.104 537.981 525.527] +/Subtype /Link +/A << /S /GoTo /D (4085) >> +>> endobj +1121 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 504.153 426.148 512.576] +/Subtype /Link +/A << /S /GoTo /D (4096) >> +>> endobj +1122 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 504.153 537.981 512.576] +/Subtype /Link +/A << /S /GoTo /D (4096) >> +>> endobj +1123 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [119.551 487.739 266.498 496.626] +/Subtype /Link +/A << /S /GoTo /D (ch07) >> +>> endobj +1124 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 487.739 537.981 496.626] +/Subtype /Link +/A << /S /GoTo /D (ch07) >> +>> endobj +1125 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 474.2 243.913 481.174] +/Subtype /Link +/A << /S /GoTo /D (Access_Control_Lists) >> +>> endobj +1126 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 474.2 537.981 481.174] +/Subtype /Link +/A << /S /GoTo /D (Access_Control_Lists) >> +>> endobj +1127 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 459.685 311.898 468.332] +/Subtype /Link +/A << /S /GoTo /D (4188) >> +>> endobj +1128 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [528.018 459.685 537.981 468.332] +/Subtype /Link +/A << /S /GoTo /D (4188) >> +>> endobj +1129 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 448.397 291.864 455.381] +/Subtype /Link +/A << /S /GoTo /D (4211) >> +>> endobj +1130 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 448.397 537.981 455.381] +/Subtype /Link +/A << /S /GoTo /D (4211) >> +>> endobj +1131 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 433.408 296.615 442.429] +/Subtype /Link +/A << /S /GoTo /D (4229) >> +>> endobj +1132 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 433.408 537.981 442.429] +/Subtype /Link +/A << /S /GoTo /D (4229) >> +>> endobj +1133 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 420.457 264.377 429.368] +/Subtype /Link +/A << /S /GoTo /D (dynamic_update_security) >> +>> endobj +1134 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 420.457 537.981 429.368] +/Subtype /Link +/A << /S /GoTo /D (dynamic_update_security) >> +>> endobj +1135 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [119.551 405.249 198.903 414.135] +/Subtype /Link +/A << /S /GoTo /D (ch08) >> +>> endobj +1136 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 405.249 537.981 414.135] +/Subtype /Link +/A << /S /GoTo /D (ch08) >> +>> endobj +1137 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 391.829 238.115 398.683] +/Subtype /Link +/A << /S /GoTo /D (4250) >> +>> endobj +1138 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 391.829 537.981 398.683] +/Subtype /Link +/A << /S /GoTo /D (4250) >> +>> endobj +1139 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 376.821 401.95 385.732] +/Subtype /Link +/A << /S /GoTo /D (4252) >> +>> endobj +1140 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 376.821 537.981 385.732] +/Subtype /Link +/A << /S /GoTo /D (4252) >> +>> endobj +1141 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 363.869 347.663 372.781] +/Subtype /Link +/A << /S /GoTo /D (4255) >> +>> endobj +1142 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 363.869 537.981 372.781] +/Subtype /Link +/A << /S /GoTo /D (4255) >> +>> endobj +1143 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 350.918 254.952 359.829] +/Subtype /Link +/A << /S /GoTo /D (4260) >> +>> endobj +1144 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 350.918 537.981 359.829] +/Subtype /Link +/A << /S /GoTo /D (4260) >> +>> endobj +1145 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [119.551 335.71 180.741 344.596] +/Subtype /Link +/A << /S /GoTo /D (ch09) >> +>> endobj +1146 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 335.71 537.981 344.596] +/Subtype /Link +/A << /S /GoTo /D (ch09) >> +>> endobj +1147 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 320.233 241.991 329.144] +/Subtype /Link +/A << /S /GoTo /D (4276) >> +>> endobj +1148 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 320.233 537.981 329.144] +/Subtype /Link +/A << /S /GoTo /D (4276) >> +>> endobj +1149 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 307.282 348.61 316.193] +/Subtype /Link +/A << /S /GoTo /D (4278) >> +>> endobj +1150 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 307.282 537.981 316.193] +/Subtype /Link +/A << /S /GoTo /D (4278) >> +>> endobj +1151 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 296.388 274.897 303.242] +/Subtype /Link +/A << /S /GoTo /D (historical_dns_information) >> +>> endobj +1152 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 296.388 537.981 303.242] +/Subtype /Link +/A << /S /GoTo /D (historical_dns_information) >> +>> endobj +1153 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 283.436 309.865 290.29] +/Subtype /Link +/A << /S /GoTo /D (classes_of_resource_records) >> +>> endobj +1154 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 283.436 537.981 290.29] +/Subtype /Link +/A << /S /GoTo /D (classes_of_resource_records) >> +>> endobj +1155 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 270.485 275.216 277.339] +/Subtype /Link +/A << /S /GoTo /D (4308) >> +>> endobj +1156 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 270.485 537.981 277.339] +/Subtype /Link +/A << /S /GoTo /D (4308) >> +>> endobj +1157 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 257.533 272.995 264.388] +/Subtype /Link +/A << /S /GoTo /D (4313) >> +>> endobj +1158 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 257.533 537.981 264.388] +/Subtype /Link +/A << /S /GoTo /D (4313) >> +>> endobj +1159 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 244.582 310.005 251.436] +/Subtype /Link +/A << /S /GoTo /D (4317) >> +>> endobj +1160 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 244.582 537.981 251.436] +/Subtype /Link +/A << /S /GoTo /D (4317) >> +>> endobj +1161 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 229.947 275.276 238.485] +/Subtype /Link +/A << /S /GoTo /D (ipv6addresses) >> +>> endobj +1162 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 229.947 537.981 238.485] +/Subtype /Link +/A << /S /GoTo /D (ipv6addresses) >> +>> endobj +1163 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [143.461 216.622 318.294 225.533] +/Subtype /Link +/A << /S /GoTo /D (bibliography) >> +>> endobj +1164 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 216.622 537.981 225.533] +/Subtype /Link +/A << /S /GoTo /D (bibliography) >> +>> endobj +1165 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 203.671 318.194 212.582] +/Subtype /Link +/A << /S /GoTo /D (rfcs) >> +>> endobj +1166 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 203.671 537.981 212.582] +/Subtype /Link +/A << /S /GoTo /D (rfcs) >> +>> endobj +1167 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 190.719 243.814 199.631] +/Subtype /Link +/A << /S /GoTo /D (4496) >> +>> endobj +1168 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 190.719 537.981 199.631] +/Subtype /Link +/A << /S /GoTo /D (4496) >> +>> endobj +1169 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 179.825 252.86 186.679] +/Subtype /Link +/A << /S /GoTo /D (internet_drafts) >> +>> endobj +1170 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 179.825 537.981 186.679] +/Subtype /Link +/A << /S /GoTo /D (internet_drafts) >> +>> endobj +1171 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [167.372 166.874 319.28 173.728] +/Subtype /Link +/A << /S /GoTo /D (4926) >> +>> endobj +1172 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 166.874 537.981 173.728] +/Subtype /Link +/A << /S /GoTo /D (4926) >> +>> endobj +1173 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.282 151.865 243.814 160.776] +/Subtype /Link +/A << /S /GoTo /D (4930) >> +>> endobj +1174 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [523.037 151.865 537.981 160.776] +/Subtype /Link +/A << /S /GoTo /D (4930) >> +>> endobj +1094 0 obj << +/D [1096 0 R /XYZ 71.731 705.352 null] +>> endobj +1095 0 obj << +/Font << /F25 681 0 R /F30 731 0 R /F31 805 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1208 0 obj << +/Length 1209 0 R +/Filter /FlateDecode +>> +stream +xڵXKo6z̯֬Hzl@b 4Y@KM&UJڬ;,%="Ycfo +"ë ,KDQLɪ<{Z_)"nu)AyaF>|+}?s͖d#{ސhltU"$C9YmIgBGid&s-ykݩ3dg3&)gFwRF4mc~"[=TYKG.ZV?"d *(=:2{9 +.d'̤@&xbݝ]+{WT7`,kֵGE3Y̹ʶc.{fu3lilXUi4A72h|.UXZuulIbgٵ~9f +!֑7$I,F9;Uv&f/gF,ipR؆|gɇwCB\ YfIkJB,ο% @{.2yhʇ֖T҃ FGc;յWi}~Ԙ%|oeڷ0=(ҭvCM˶&\hY*؍W8GuJ[JgcGپ3590 +ef=rEhSǂ8ij̸!~PJ^AIFFs:L̹d؍D|HzđAla:P%WEWh:a@ZNN=PP%% eO"'dFrDŽI>̓/`doݐ|Ъonԡ- ‹hoJTlF@CQG~YHq uzs}}q1Wq,c]sY SmCQ+E8pw9ݨçl /?o؝2:@P|Z͡!sӞU Q'X8vկbAw F>t'{;!6W5t``˞rvV XΑ֙7?=ބCaj$Q1Dq4_~@FP7Bj3Jhό?> endobj +1205 0 obj << +/D [1207 0 R /XYZ 71.731 705.352 null] +>> endobj +687 0 obj << +/D [1207 0 R /XYZ 71.731 657.532 null] +>> endobj +5 0 obj << +/D [1207 0 R /XYZ 304.322 642.462 null] +>> endobj +1175 0 obj << +/D [1207 0 R /XYZ 71.731 621.4 null] +>> endobj +1176 0 obj << +/D [1207 0 R /XYZ 267.026 612.725 null] +>> endobj +1177 0 obj << +/D [1207 0 R /XYZ 374.391 586.822 null] +>> endobj +688 0 obj << +/D [1207 0 R /XYZ 71.731 563.809 null] +>> endobj +9 0 obj << +/D [1207 0 R /XYZ 266.882 528.617 null] +>> endobj +1178 0 obj << +/D [1207 0 R /XYZ 71.731 516.446 null] +>> endobj +1179 0 obj << +/D [1207 0 R /XYZ 273.402 507.058 null] +>> endobj +1180 0 obj << +/D [1207 0 R /XYZ 244.242 481.155 null] +>> endobj +1181 0 obj << +/D [1207 0 R /XYZ 265.552 481.155 null] +>> endobj +1182 0 obj << +/D [1207 0 R /XYZ 71.731 474.017 null] +>> endobj +689 0 obj << +/D [1207 0 R /XYZ 71.731 446.122 null] +>> endobj +13 0 obj << +/D [1207 0 R /XYZ 360.361 412.987 null] +>> endobj +1183 0 obj << +/D [1207 0 R /XYZ 71.731 400.549 null] +>> endobj +1184 0 obj << +/D [1207 0 R /XYZ 191.5 391.428 null] +>> endobj +1185 0 obj << +/D [1207 0 R /XYZ 312.415 391.428 null] +>> endobj +1186 0 obj << +/D [1207 0 R /XYZ 351.708 391.428 null] +>> endobj +1187 0 obj << +/D [1207 0 R /XYZ 418.387 391.428 null] +>> endobj +1188 0 obj << +/D [1207 0 R /XYZ 221.647 378.476 null] +>> endobj +1189 0 obj << +/D [1207 0 R /XYZ 409.44 378.476 null] +>> endobj +1190 0 obj << +/D [1207 0 R /XYZ 457.868 378.476 null] +>> endobj +1191 0 obj << +/D [1207 0 R /XYZ 437.992 365.525 null] +>> endobj +1192 0 obj << +/D [1207 0 R /XYZ 264.017 352.574 null] +>> endobj +1193 0 obj << +/D [1207 0 R /XYZ 351.986 339.622 null] +>> endobj +1194 0 obj << +/D [1207 0 R /XYZ 445.504 339.622 null] +>> endobj +1195 0 obj << +/D [1207 0 R /XYZ 220.392 326.671 null] +>> endobj +1196 0 obj << +/D [1207 0 R /XYZ 237.029 313.72 null] +>> endobj +1197 0 obj << +/D [1207 0 R /XYZ 430.431 313.72 null] +>> endobj +1198 0 obj << +/D [1207 0 R /XYZ 426.327 300.768 null] +>> endobj +1199 0 obj << +/D [1207 0 R /XYZ 276.411 287.817 null] +>> endobj +1200 0 obj << +/D [1207 0 R /XYZ 469.862 287.817 null] +>> endobj +690 0 obj << +/D [1207 0 R /XYZ 71.731 257.765 null] +>> endobj +17 0 obj << +/D [1207 0 R /XYZ 404.897 224.63 null] +>> endobj +1201 0 obj << +/D [1207 0 R /XYZ 71.731 215.807 null] +>> endobj +1202 0 obj << +/D [1207 0 R /XYZ 71.731 190.951 null] +>> endobj +1203 0 obj << +/D [1207 0 R /XYZ 71.731 185.97 null] +>> endobj +1204 0 obj << +/D [1207 0 R /XYZ 71.731 185.97 null] +>> endobj +1206 0 obj << +/Font << /F29 669 0 R /F25 681 0 R /F36 669 0 R /F31 805 0 R /F37 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1240 0 obj << +/Length 1241 0 R +/Filter /FlateDecode +>> +stream +xXˮ6Wx)W(RH @iEBdQ{{>%۲oBEr8gΜ! H2Dt9< 𭟠ώ&zx-B~z_*'֌[{+.q%FSrIHǍjDUXqX~n/1%dg*J Ey K R qHAP gv7ӤDI^- 1K +M*[I}$EsyI:BQbqgwͻyۑ:Q+EWon|^_M|$(0.P e;ZS%y붭`"pβ$C`PS; ]6OTg)(0?lVg6CIpu`!ɃX u +.Qazm0rhNq)9(H:O(]mХEZ|E +) +mZUgv=6WZ(Ofʦ\ѵnsӭbm{5&5-OJ?fHrm?NaRendstream +endobj +1241 0 obj +1632 +endobj +1239 0 obj << +/Type /Page +/Contents 1240 0 R +/Resources 1238 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 1056 0 R +>> endobj +1237 0 obj << +/D [1239 0 R /XYZ 71.731 705.352 null] +>> endobj +1211 0 obj << +/D [1239 0 R /XYZ 71.731 604.929 null] +>> endobj +1212 0 obj << +/D [1239 0 R /XYZ 350.353 591.978 null] +>> endobj +1213 0 obj << +/D [1239 0 R /XYZ 71.731 579.859 null] +>> endobj +1214 0 obj << +/D [1239 0 R /XYZ 71.731 569.896 null] +>> endobj +1215 0 obj << +/D [1239 0 R /XYZ 71.731 569.896 null] +>> endobj +691 0 obj << +/D [1239 0 R /XYZ 71.731 466.285 null] +>> endobj +21 0 obj << +/D [1239 0 R /XYZ 374.029 430.993 null] +>> endobj +1216 0 obj << +/D [1239 0 R /XYZ 71.731 418.822 null] +>> endobj +1217 0 obj << +/D [1239 0 R /XYZ 427.004 409.434 null] +>> endobj +1218 0 obj << +/D [1239 0 R /XYZ 420.658 396.483 null] +>> endobj +1219 0 obj << +/D [1239 0 R /XYZ 511.142 396.483 null] +>> endobj +692 0 obj << +/D [1239 0 R /XYZ 71.731 384.363 null] +>> endobj +25 0 obj << +/D [1239 0 R /XYZ 243.136 357.111 null] +>> endobj +1220 0 obj << +/D [1239 0 R /XYZ 71.731 349.758 null] +>> endobj +1221 0 obj << +/D [1239 0 R /XYZ 71.731 303.945 null] +>> endobj +1222 0 obj << +/D [1239 0 R /XYZ 328.764 293.151 null] +>> endobj +1223 0 obj << +/D [1239 0 R /XYZ 119.551 280.199 null] +>> endobj +1224 0 obj << +/D [1239 0 R /XYZ 307.92 280.199 null] +>> endobj +693 0 obj << +/D [1239 0 R /XYZ 71.731 250.147 null] +>> endobj +29 0 obj << +/D [1239 0 R /XYZ 309.314 222.894 null] +>> endobj +1225 0 obj << +/D [1239 0 R /XYZ 71.731 215.542 null] +>> endobj +1226 0 obj << +/D [1239 0 R /XYZ 292.5 202.77 null] +>> endobj +1227 0 obj << +/D [1239 0 R /XYZ 420.22 189.819 null] +>> endobj +1228 0 obj << +/D [1239 0 R /XYZ 497.357 176.867 null] +>> endobj +1238 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F39 1242 0 R /F37 1210 0 R /F36 669 0 R /F40 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1277 0 obj << +/Length 1278 0 R +/Filter /FlateDecode +>> +stream +xڥYI8_aZH-tw2 f9m EDϣG]AH%Q۾-wwyL.#)Kv]3o̬NRŠ\/OY{:M1a>&[e{Yjdw_}xrۗPm,4.Mw,'y:i!S4B~pZ")d? +[)y37`iv%Tq؞"^u㈒$Ig>'^Vć>)97A%*%k@$S@unUP%eɪ#p%6ly2즔Z̀p^rq9 d#ckk]M"Xg)/rDp珿C(M-] cڑb K-ztZ6jlA`ES4rlE?W߆v ͏`!UV / b 28pYKu+O3]y5m:7S(D~N?h;٧.{5e$a/?&bx0/L4jy JYҘ0>|53rGȥfj;'1c ]?$m灡AxӋm,%QF:W' ;7պ՛ ilT,c㋉ +b̆`` b'qT'{&O*pE=/ѿkt!<xrXxnqm|CCn9g)a]hhYa>fg4*MHs+@*~734endstream +endobj +1278 0 obj +2318 +endobj +1276 0 obj << +/Type /Page +/Contents 1277 0 R +/Resources 1275 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 1056 0 R +/Annots [ 1279 0 R 1280 0 R ] +>> endobj +1279 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [314.587 544.79 366.891 553.701] +/Subtype /Link +/A << /S /GoTo /D (types_of_resource_records_and_when_to_use_them) >> +>> endobj +1280 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [239.649 515.963 294.164 522.817] +/Subtype /Link +/A << /S /GoTo /D (rfcs) >> +>> endobj +1274 0 obj << +/D [1276 0 R /XYZ 71.731 705.352 null] +>> endobj +1229 0 obj << +/D [1276 0 R /XYZ 71.731 657.532 null] +>> endobj +1230 0 obj << +/D [1276 0 R /XYZ 341.008 647.569 null] +>> endobj +1231 0 obj << +/D [1276 0 R /XYZ 434.975 647.569 null] +>> endobj +1232 0 obj << +/D [1276 0 R /XYZ 146.38 634.618 null] +>> endobj +1233 0 obj << +/D [1276 0 R /XYZ 294.951 634.618 null] +>> endobj +1234 0 obj << +/D [1276 0 R /XYZ 435.633 634.618 null] +>> endobj +1235 0 obj << +/D [1276 0 R /XYZ 130.341 621.666 null] +>> endobj +1236 0 obj << +/D [1276 0 R /XYZ 168.338 621.666 null] +>> endobj +1243 0 obj << +/D [1276 0 R /XYZ 71.731 615.246 null] +>> endobj +1244 0 obj << +/D [1276 0 R /XYZ 420.598 603.734 null] +>> endobj +1245 0 obj << +/D [1276 0 R /XYZ 163.546 577.831 null] +>> endobj +1246 0 obj << +/D [1276 0 R /XYZ 409.739 577.831 null] +>> endobj +1247 0 obj << +/D [1276 0 R /XYZ 71.731 570.693 null] +>> endobj +1248 0 obj << +/D [1276 0 R /XYZ 390.979 559.898 null] +>> endobj +1249 0 obj << +/D [1276 0 R /XYZ 463.098 559.898 null] +>> endobj +1250 0 obj << +/D [1276 0 R /XYZ 71.731 539.809 null] +>> endobj +694 0 obj << +/D [1276 0 R /XYZ 71.731 501.019 null] +>> endobj +33 0 obj << +/D [1276 0 R /XYZ 153.861 471.709 null] +>> endobj +1251 0 obj << +/D [1276 0 R /XYZ 71.731 464.357 null] +>> endobj +1252 0 obj << +/D [1276 0 R /XYZ 474.854 451.585 null] +>> endobj +1253 0 obj << +/D [1276 0 R /XYZ 119.551 438.634 null] +>> endobj +1254 0 obj << +/D [1276 0 R /XYZ 71.731 433.533 null] +>> endobj +1255 0 obj << +/D [1276 0 R /XYZ 365.824 420.701 null] +>> endobj +1256 0 obj << +/D [1276 0 R /XYZ 466.845 381.847 null] +>> endobj +1257 0 obj << +/D [1276 0 R /XYZ 71.731 361.757 null] +>> endobj +1258 0 obj << +/D [1276 0 R /XYZ 224.267 350.963 null] +>> endobj +1259 0 obj << +/D [1276 0 R /XYZ 119.551 338.011 null] +>> endobj +1260 0 obj << +/D [1276 0 R /XYZ 246.383 338.011 null] +>> endobj +1261 0 obj << +/D [1276 0 R /XYZ 421.939 338.011 null] +>> endobj +1262 0 obj << +/D [1276 0 R /XYZ 216.198 325.06 null] +>> endobj +1263 0 obj << +/D [1276 0 R /XYZ 316.262 325.06 null] +>> endobj +1264 0 obj << +/D [1276 0 R /XYZ 274.737 299.157 null] +>> endobj +1265 0 obj << +/D [1276 0 R /XYZ 330.627 299.157 null] +>> endobj +1266 0 obj << +/D [1276 0 R /XYZ 412.14 299.157 null] +>> endobj +1267 0 obj << +/D [1276 0 R /XYZ 119.551 286.206 null] +>> endobj +1268 0 obj << +/D [1276 0 R /XYZ 71.731 253.165 null] +>> endobj +1269 0 obj << +/D [1276 0 R /XYZ 153.036 242.37 null] +>> endobj +1270 0 obj << +/D [1276 0 R /XYZ 217.574 229.419 null] +>> endobj +1271 0 obj << +/D [1276 0 R /XYZ 272.615 216.467 null] +>> endobj +695 0 obj << +/D [1276 0 R /XYZ 71.731 188.473 null] +>> endobj +37 0 obj << +/D [1276 0 R /XYZ 299.257 159.163 null] +>> endobj +1272 0 obj << +/D [1276 0 R /XYZ 71.731 151.81 null] +>> endobj +1273 0 obj << +/D [1276 0 R /XYZ 260.501 139.038 null] +>> endobj +1275 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F37 1210 0 R /F40 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1306 0 obj << +/Length 1307 0 R +/Filter /FlateDecode +>> +stream +xڕXn68_a6v=NWEL҇c6YD*Z +}ysν$7FYEI9\>y٘ZYPgrWO_"6OaS3iv0 ?Ԣɶ)`M>YEboqhO2<˃m/e-UY?|ˣ~T#Oqm-'[KPS&5݋zqeÅz>qx&0*zg0MgQ2jCT$xtLva]~3CE Dyܸq욋5\u$koW +GLS-VOlw*)qܚ(WSueIkۙjJ,#O'V&94ݩD7W{nA(QjV%*pX(38S"?R@`CwڨO߯3*F(F XcqYh E2F((ґa)HCzc^՛ z3KF=/wTҢv5owI8D -FV GKe߼?hHđ((u:!ׁ {Ԅʣ\q탯<]zß> endobj +1308 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [122.869 588.999 182.644 597.537] +/Subtype /Link +/A << /S /GoTo /D (diagnostic_tools) >> +>> endobj +1303 0 obj << +/D [1305 0 R /XYZ 71.731 705.352 null] +>> endobj +1281 0 obj << +/D [1305 0 R /XYZ 71.731 629.537 null] +>> endobj +1282 0 obj << +/D [1305 0 R /XYZ 502.311 603.734 null] +>> endobj +696 0 obj << +/D [1305 0 R /XYZ 71.731 579.037 null] +>> endobj +41 0 obj << +/D [1305 0 R /XYZ 273.83 556.312 null] +>> endobj +1283 0 obj << +/D [1305 0 R /XYZ 71.731 547.86 null] +>> endobj +1284 0 obj << +/D [1305 0 R /XYZ 480.791 537.383 null] +>> endobj +1285 0 obj << +/D [1305 0 R /XYZ 233.233 524.431 null] +>> endobj +1286 0 obj << +/D [1305 0 R /XYZ 134.216 498.529 null] +>> endobj +1287 0 obj << +/D [1305 0 R /XYZ 180.422 498.529 null] +>> endobj +697 0 obj << +/D [1305 0 R /XYZ 71.731 481.547 null] +>> endobj +45 0 obj << +/D [1305 0 R /XYZ 240.942 458.081 null] +>> endobj +1288 0 obj << +/D [1305 0 R /XYZ 71.731 451.954 null] +>> endobj +1289 0 obj << +/D [1305 0 R /XYZ 260.66 439.152 null] +>> endobj +1290 0 obj << +/D [1305 0 R /XYZ 376.185 439.152 null] +>> endobj +1291 0 obj << +/D [1305 0 R /XYZ 391.677 426.2 null] +>> endobj +698 0 obj << +/D [1305 0 R /XYZ 71.731 383.197 null] +>> endobj +49 0 obj << +/D [1305 0 R /XYZ 249.74 359.849 null] +>> endobj +1292 0 obj << +/D [1305 0 R /XYZ 71.731 353.722 null] +>> endobj +1293 0 obj << +/D [1305 0 R /XYZ 199.51 327.969 null] +>> endobj +1294 0 obj << +/D [1305 0 R /XYZ 205.328 315.018 null] +>> endobj +1295 0 obj << +/D [1305 0 R /XYZ 252.5 315.018 null] +>> endobj +1296 0 obj << +/D [1305 0 R /XYZ 71.731 281.977 null] +>> endobj +1297 0 obj << +/D [1305 0 R /XYZ 129.235 271.182 null] +>> endobj +1298 0 obj << +/D [1305 0 R /XYZ 71.731 240.198 null] +>> endobj +699 0 obj << +/D [1305 0 R /XYZ 71.731 174.38 null] +>> endobj +1304 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F30 731 0 R /F42 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1327 0 obj << +/Length 1328 0 R +/Filter /FlateDecode +>> +stream +xڍYIsc +"UhKԳtIbfr)b5I(iGy rO@޾|wɉ @C#:,wTiL>#qmS䵦㗓G#᭛L\Ъ̼0Mtp +kt4$ý _/B7kGH Q8yk9< J}mpd#~pΰ.~Ů% C; mcpڛZ Xid CNH3 +'-+ ŀلd} 컑} k2(@ĻaƺQ-ak +4ӝͼjkĕ|,X+6gE~ij;bұd(?߹qm`n ՟7eg0`<0!ͭWH,.L`{I)V]rf@\)`ao >5wrM͟nXR~m9DosޱvZd>I2|y]%FSr5p-0֗wK~Y:X g^O|,WKc/Q#NX_ֵ/3 6 uפYGqHҌE7z?a#d͞endstream +endobj +1328 0 obj +2043 +endobj +1326 0 obj << +/Type /Page +/Contents 1327 0 R +/Resources 1325 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 1329 0 R +>> endobj +1324 0 obj << +/D [1326 0 R /XYZ 71.731 705.352 null] +>> endobj +53 0 obj << +/D [1326 0 R /XYZ 267.668 647.066 null] +>> endobj +1299 0 obj << +/D [1326 0 R /XYZ 71.731 636.701 null] +>> endobj +1300 0 obj << +/D [1326 0 R /XYZ 367.996 626.942 null] +>> endobj +1301 0 obj << +/D [1326 0 R /XYZ 217.891 588.088 null] +>> endobj +1302 0 obj << +/D [1326 0 R /XYZ 356.987 588.088 null] +>> endobj +1309 0 obj << +/D [1326 0 R /XYZ 71.731 580.95 null] +>> endobj +1310 0 obj << +/D [1326 0 R /XYZ 403.134 557.204 null] +>> endobj +1311 0 obj << +/D [1326 0 R /XYZ 486.719 557.204 null] +>> endobj +1312 0 obj << +/D [1326 0 R /XYZ 71.731 537.114 null] +>> endobj +700 0 obj << +/D [1326 0 R /XYZ 71.731 501.249 null] +>> endobj +57 0 obj << +/D [1326 0 R /XYZ 227.612 478.898 null] +>> endobj +1313 0 obj << +/D [1326 0 R /XYZ 71.731 470.26 null] +>> endobj +1314 0 obj << +/D [1326 0 R /XYZ 135.87 447.017 null] +>> endobj +1315 0 obj << +/D [1326 0 R /XYZ 255.39 434.066 null] +>> endobj +1316 0 obj << +/D [1326 0 R /XYZ 71.731 426.928 null] +>> endobj +1317 0 obj << +/D [1326 0 R /XYZ 513.658 390.23 null] +>> endobj +1318 0 obj << +/D [1326 0 R /XYZ 364.356 364.328 null] +>> endobj +701 0 obj << +/D [1326 0 R /XYZ 71.731 311.362 null] +>> endobj +61 0 obj << +/D [1326 0 R /XYZ 325.209 284.109 null] +>> endobj +1319 0 obj << +/D [1326 0 R /XYZ 71.731 273.966 null] +>> endobj +1320 0 obj << +/D [1326 0 R /XYZ 137.435 263.984 null] +>> endobj +1321 0 obj << +/D [1326 0 R /XYZ 71.731 243.895 null] +>> endobj +1322 0 obj << +/D [1326 0 R /XYZ 278.642 207.197 null] +>> endobj +1323 0 obj << +/D [1326 0 R /XYZ 259.693 181.295 null] +>> endobj +1325 0 obj << +/Font << /F31 805 0 R /F40 669 0 R /F25 681 0 R /F42 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1333 0 obj << +/Length 1334 0 R +/Filter /FlateDecode +>> +stream +xڅ +@ >Evh]rvU6q(b+i]J?/?!iFGB0>No:<08yELfk/sO=7Yn٤Cn> endobj +1330 0 obj << +/D [1332 0 R /XYZ 71.731 705.352 null] +>> endobj +1331 0 obj << +/Font << /F31 805 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1350 0 obj << +/Length 1351 0 R +/Filter /FlateDecode +>> +stream +xڥXnFУXH8 }*"G¼ewiEng%M\Μ93d_؅n8"-߄ ~k]Q_snMI7o?DE xx:u'?9k:Q`>=~|on*,SW\@ ~ya,]`u֣ $?J + 2&&0CpH.Guq|o7˜ +7Kc + L_G:˸uŊj.* oWt3*0?>hJy%bڤ5^%RHr3sЄQp.it}( TMBf~֧U/OIԥR?|Yĩʲ v@0 +A%waឌTdKBSIMYKq4^e-@JÖBix*J诋fyK)փD]bXp> endobj +1347 0 obj << +/D [1349 0 R /XYZ 71.731 705.352 null] +>> endobj +702 0 obj << +/D [1349 0 R /XYZ 71.731 657.532 null] +>> endobj +65 0 obj << +/D [1349 0 R /XYZ 471.943 642.462 null] +>> endobj +703 0 obj << +/D [1349 0 R /XYZ 71.731 621.4 null] +>> endobj +69 0 obj << +/D [1349 0 R /XYZ 296.922 590.385 null] +>> endobj +1335 0 obj << +/D [1349 0 R /XYZ 71.731 578.214 null] +>> endobj +1336 0 obj << +/D [1349 0 R /XYZ 71.731 578.214 null] +>> endobj +1337 0 obj << +/D [1349 0 R /XYZ 405.107 555.875 null] +>> endobj +1338 0 obj << +/D [1349 0 R /XYZ 71.731 548.736 null] +>> endobj +1339 0 obj << +/D [1349 0 R /XYZ 261.218 537.942 null] +>> endobj +1340 0 obj << +/D [1349 0 R /XYZ 172.681 512.039 null] +>> endobj +704 0 obj << +/D [1349 0 R /XYZ 71.731 484.044 null] +>> endobj +73 0 obj << +/D [1349 0 R /XYZ 261.149 448.852 null] +>> endobj +1341 0 obj << +/D [1349 0 R /XYZ 71.731 436.681 null] +>> endobj +1342 0 obj << +/D [1349 0 R /XYZ 210.031 427.293 null] +>> endobj +705 0 obj << +/D [1349 0 R /XYZ 71.731 384.29 null] +>> endobj +77 0 obj << +/D [1349 0 R /XYZ 290.983 351.155 null] +>> endobj +1343 0 obj << +/D [1349 0 R /XYZ 71.731 338.984 null] +>> endobj +1344 0 obj << +/D [1349 0 R /XYZ 119.551 316.645 null] +>> endobj +1345 0 obj << +/D [1349 0 R /XYZ 289.083 303.693 null] +>> endobj +706 0 obj << +/D [1349 0 R /XYZ 71.731 234.787 null] +>> endobj +81 0 obj << +/D [1349 0 R /XYZ 450.723 201.652 null] +>> endobj +1346 0 obj << +/D [1349 0 R /XYZ 71.731 192.829 null] +>> endobj +1348 0 obj << +/Font << /F29 669 0 R /F36 669 0 R /F25 681 0 R /F30 731 0 R /F31 805 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1376 0 obj << +/Length 1377 0 R +/Filter /FlateDecode +>> +stream +xWn0s[6&MWٍ탻80&oml6lG03xhyVAZRϝesS־'̓#i<]('uGEسP4^LOMN{L\ѧ Gx#L$䣛GE( _ode ͷ,endstream +endobj +1377 0 obj +817 +endobj +1375 0 obj << +/Type /Page +/Contents 1376 0 R +/Resources 1374 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 1329 0 R +>> endobj +1373 0 obj << +/D [1375 0 R /XYZ 71.731 705.352 null] +>> endobj +707 0 obj << +/D [1375 0 R /XYZ 71.731 604.566 null] +>> endobj +85 0 obj << +/D [1375 0 R /XYZ 353.336 571.431 null] +>> endobj +1352 0 obj << +/D [1375 0 R /XYZ 71.731 558.993 null] +>> endobj +1353 0 obj << +/D [1375 0 R /XYZ 137.544 549.872 null] +>> endobj +1354 0 obj << +/D [1375 0 R /XYZ 71.731 542.734 null] +>> endobj +1355 0 obj << +/D [1375 0 R /XYZ 71.731 537.753 null] +>> endobj +1356 0 obj << +/D [1375 0 R /XYZ 129.514 521.977 null] +>> endobj +1357 0 obj << +/D [1375 0 R /XYZ 71.731 521.877 null] +>> endobj +1358 0 obj << +/D [1375 0 R /XYZ 129.514 504.044 null] +>> endobj +1359 0 obj << +/D [1375 0 R /XYZ 71.731 501.887 null] +>> endobj +1360 0 obj << +/D [1375 0 R /XYZ 129.514 486.111 null] +>> endobj +1361 0 obj << +/D [1375 0 R /XYZ 71.731 483.954 null] +>> endobj +1362 0 obj << +/D [1375 0 R /XYZ 129.514 468.179 null] +>> endobj +1363 0 obj << +/D [1375 0 R /XYZ 71.731 468.079 null] +>> endobj +1364 0 obj << +/D [1375 0 R /XYZ 129.514 450.246 null] +>> endobj +1365 0 obj << +/D [1375 0 R /XYZ 71.731 450.146 null] +>> endobj +1366 0 obj << +/D [1375 0 R /XYZ 129.514 432.313 null] +>> endobj +1367 0 obj << +/D [1375 0 R /XYZ 71.731 430.874 null] +>> endobj +1368 0 obj << +/D [1375 0 R /XYZ 129.514 414.381 null] +>> endobj +1369 0 obj << +/D [1375 0 R /XYZ 71.731 412.224 null] +>> endobj +1370 0 obj << +/D [1375 0 R /XYZ 129.514 396.448 null] +>> endobj +1371 0 obj << +/D [1375 0 R /XYZ 71.731 395.008 null] +>> endobj +1372 0 obj << +/D [1375 0 R /XYZ 129.514 378.515 null] +>> endobj +1374 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F36 669 0 R /F45 681 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1388 0 obj << +/Length 1389 0 R +/Filter /FlateDecode +>> +stream +xW;6R\Ʌhzc%!1pEhXYTH'P$%˫ݽ@G# qy +3i#κ-rʬvTQv?6L#Ra =1e&q^ǏpTsѭBU{iJDeq\YIzP^{)8 /qJV:LiM"mEwp>Zp[-fBHwY1d27I 0AULhHh ގQ%:k]ULk.\ 3]K3Z:Lv] 5BS-jT()цjz@"ѵr_-p01n={ѶbM2QP^rGK>ԩz3=>ay. +;SuZ-Twm}E΅ ʇ yf3^냳Vc{֤dE_۠J`jlYGyga 똱xKf򙫣ӢPg+13ݴn킕gByfɁ٪K` +!. ٭uYȧˤuϛh[j0`X, +с5}u I|d@u9 1q*  'ESGk&y8\I̧reYTT>t$;wag s+D3w= ~;a ,]d=]Ywv܄#piVԴ= +Z];^h|<˓lr⹼飐\K]נ"ڝrtoKHHj+nwjK6_$U> endobj +1385 0 obj << +/D [1387 0 R /XYZ 71.731 705.352 null] +>> endobj +708 0 obj << +/D [1387 0 R /XYZ 71.731 657.532 null] +>> endobj +89 0 obj << +/D [1387 0 R /XYZ 438.125 642.462 null] +>> endobj +1378 0 obj << +/D [1387 0 R /XYZ 71.731 621.08 null] +>> endobj +709 0 obj << +/D [1387 0 R /XYZ 71.731 587.654 null] +>> endobj +93 0 obj << +/D [1387 0 R /XYZ 293.669 554.52 null] +>> endobj +710 0 obj << +/D [1387 0 R /XYZ 71.731 540.727 null] +>> endobj +97 0 obj << +/D [1387 0 R /XYZ 302.729 515.147 null] +>> endobj +1379 0 obj << +/D [1387 0 R /XYZ 71.731 504.782 null] +>> endobj +1380 0 obj << +/D [1387 0 R /XYZ 71.731 469.952 null] +>> endobj +711 0 obj << +/D [1387 0 R /XYZ 71.731 276.291 null] +>> endobj +101 0 obj << +/D [1387 0 R /XYZ 343.085 247.931 null] +>> endobj +1381 0 obj << +/D [1387 0 R /XYZ 71.731 237.788 null] +>> endobj +1382 0 obj << +/D [1387 0 R /XYZ 123.616 214.855 null] +>> endobj +1383 0 obj << +/D [1387 0 R /XYZ 314.807 214.855 null] +>> endobj +1384 0 obj << +/D [1387 0 R /XYZ 71.731 203.484 null] +>> endobj +1386 0 obj << +/Font << /F29 669 0 R /F25 681 0 R /F36 669 0 R /F40 669 0 R /F37 1210 0 R /F31 805 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1399 0 obj << +/Length 1400 0 R +/Filter /FlateDecode +>> +stream +xڝVKo6z̯|IQ洛m Zc!hHrtC %[#sDC~ԉ)9,AHl}A=;=;!7!ItW_A<%iĝc7+iUL.'ɵjT퇩{G+BW.kTP"6iBB""qNCpWWuTЫj}JIa1RJ!nj̐KQ{k$EmRwVbfuq)K[ZfSTO|-veK7 =[y^bNױkQFz#gh{'"H|l 'L!dMK٘9+Wd4r' 8ecZE Qg ޔdz=ơ,)?A:Ll_GV6zi=,勭Ft֞˲;hֲjCqcwݧ)#4kJhxʙ\&Oׂ<{P=Ø85s0VNϳ)jb\^u>o!S3eרl7G 4g?I?i 3/GZH}2:L!P}]p_+*QPA[ +_٪P{q+UTr m3Xo˶ؔV.t71U٩]$Ccr|>|`.-%zJz,yڮj#v,+ڕڝGHgOm-'CCGrpTP=vyfY`,54> +P&ެ,yέeiuUEXv@ΐvUs[]0G!IȘ#aI8r8`B`eaJ @aD$&ëO>f|R[7BfIdj#mY W V !=\0t33m69v; C'2nwa].}Ş9#aIo ! zHɗaԵNh|gd4S;Mqd S" endstream +endobj +1400 0 obj +1074 +endobj +1398 0 obj << +/Type /Page +/Contents 1399 0 R +/Resources 1397 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 1329 0 R +>> endobj +1396 0 obj << +/D [1398 0 R /XYZ 71.731 704.356 null] +>> endobj +712 0 obj << +/D [1398 0 R /XYZ 71.731 325.226 null] +>> endobj +105 0 obj << +/D [1398 0 R /XYZ 233.416 290.984 null] +>> endobj +1390 0 obj << +/D [1398 0 R /XYZ 71.731 278.546 null] +>> endobj +1391 0 obj << +/D [1398 0 R /XYZ 295.559 269.425 null] +>> endobj +1392 0 obj << +/D [1398 0 R /XYZ 71.731 262.287 null] +>> endobj +1393 0 obj << +/D [1398 0 R /XYZ 71.731 215.527 null] +>> endobj +1394 0 obj << +/D [1398 0 R /XYZ 71.731 210.546 null] +>> endobj +1395 0 obj << +/D [1398 0 R /XYZ 71.731 210.546 null] +>> endobj +1397 0 obj << +/Font << /F31 805 0 R /F37 1210 0 R /F36 669 0 R /F25 681 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1442 0 obj << +/Length 1443 0 R +/Filter /FlateDecode +>> +stream +xڭX[o6_@ĉ6ː KC1l{e&*$7~>]%^K0J|zy{gG?3=%#f;X,&ޮmޞعi!H>?[Y8^- Yo)F󝷵(" ^HQ^$Q6ğx)㏃b `s>RF&jûe~Bc@h_;aF£4B܆'^3VTyRH5 v.esho7z$̐lXXY kͰgvT !Kghw懙߃hso+lz\: ʈNNH+٘1aB>j*1?-`bDl4-|]~-ۙe*8WG[I2v<+k6Gփ5o2i9+Ԍ(Ѳ?\GHS[}lI1A\" ++zUO v'a`R-󦖝޸r>˦$!J֖WSh,(aفy#`^7ecil c#ñS%( aGX=-I7r͚:5MKfhBjVMTTa>k]{M[ժގtVosFҤ>daBi>]˚=X,^16/4R=F_W*O,YM+:*Ԑv3zIhk,]OB +e'ŔNށ\o + *4CcZЁ9&QmOj מMU:h_maR?RZ~d~QF-ޚf {[/Vr&qSk>.C ('~}-TMNSdgVYAk} @ {I(+=VpҭQ/EaJk5 #8&:A+S, "AWBOGVn} ;z~QED. $tۦqYktG4 #Mk12^Y>u %:݉[,ƩplmXi3[MXIR[*SONNN3I+͚ީ: J`^A.뮑eNi8e]MaXԾM$kozw\2` j`PX,\nŽ!/ +SfA,> R1V (K/]]_?3=RtzBŽ<۟~hf$jf1u +33գY.ʡDk93a>2@]7$ Пlz6(e0mVŠ o#Ɠ–|Xg*߰\1=0]-q+Z}f~߳.?>}[ox~t{F?#H8Kg?o9c#j[;r~L8UZmS:KchA8AI$q#Ei2ڭgs7zUHendstream +endobj +1443 0 obj +1690 +endobj +1441 0 obj << +/Type /Page +/Contents 1442 0 R +/Resources 1440 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 1447 0 R +/Annots [ 1444 0 R 1445 0 R 1446 0 R ] +>> endobj +1444 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [119.551 539.41 181.996 548.322] +/Subtype /Link +/A << /S /GoTo /D (rrset_ordering_title) >> +>> endobj +1445 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [134.764 398.012 199.52 406.899] +/Subtype /Link +/A << /S /GoTo /D (zone_transfers) >> +>> endobj +1446 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [362.906 398.012 427.662 406.899] +/Subtype /Link +/A << /S /GoTo /D (boolean_options) >> +>> endobj +1438 0 obj << +/D [1441 0 R /XYZ 71.731 705.352 null] +>> endobj +1439 0 obj << +/D [1441 0 R /XYZ 71.731 717.307 null] +>> endobj +1401 0 obj << +/D [1441 0 R /XYZ 71.731 611.305 null] +>> endobj +1402 0 obj << +/D [1441 0 R /XYZ 289.811 598.354 null] +>> endobj +1403 0 obj << +/D [1441 0 R /XYZ 71.731 566.03 null] +>> endobj +1404 0 obj << +/D [1441 0 R /XYZ 316.122 554.519 null] +>> endobj +1405 0 obj << +/D [1441 0 R /XYZ 445.604 554.519 null] +>> endobj +1407 0 obj << +/D [1441 0 R /XYZ 338.078 541.567 null] +>> endobj +713 0 obj << +/D [1441 0 R /XYZ 71.731 513.572 null] +>> endobj +109 0 obj << +/D [1441 0 R /XYZ 154.002 478.38 null] +>> endobj +1408 0 obj << +/D [1441 0 R /XYZ 71.731 466.209 null] +>> endobj +1409 0 obj << +/D [1441 0 R /XYZ 71.731 466.209 null] +>> endobj +1410 0 obj << +/D [1441 0 R /XYZ 234.369 443.87 null] +>> endobj +1411 0 obj << +/D [1441 0 R /XYZ 71.731 424.498 null] +>> endobj +1412 0 obj << +/D [1441 0 R /XYZ 71.731 424.498 null] +>> endobj +1413 0 obj << +/D [1441 0 R /XYZ 480.124 412.986 null] +>> endobj +1414 0 obj << +/D [1441 0 R /XYZ 317.805 400.034 null] +>> endobj +714 0 obj << +/D [1441 0 R /XYZ 71.731 383.068 null] +>> endobj +113 0 obj << +/D [1441 0 R /XYZ 299.625 349.799 null] +>> endobj +715 0 obj << +/D [1441 0 R /XYZ 71.731 336.273 null] +>> endobj +117 0 obj << +/D [1441 0 R /XYZ 409.262 310.427 null] +>> endobj +1415 0 obj << +/D [1441 0 R /XYZ 71.731 303.074 null] +>> endobj +716 0 obj << +/D [1441 0 R /XYZ 71.731 265.231 null] +>> endobj +121 0 obj << +/D [1441 0 R /XYZ 258.002 242.88 null] +>> endobj +1416 0 obj << +/D [1441 0 R /XYZ 71.731 234.243 null] +>> endobj +1417 0 obj << +/D [1441 0 R /XYZ 71.731 224.28 null] +>> endobj +1418 0 obj << +/D [1441 0 R /XYZ 71.731 224.28 null] +>> endobj +1419 0 obj << +/D [1441 0 R /XYZ 71.731 211.966 null] +>> endobj +1420 0 obj << +/D [1441 0 R /XYZ 139.476 196.056 null] +>> endobj +1421 0 obj << +/D [1441 0 R /XYZ 271.181 196.056 null] +>> endobj +1422 0 obj << +/D [1441 0 R /XYZ 71.731 145.082 null] +>> endobj +1440 0 obj << +/Font << /F31 805 0 R /F37 1210 0 R /F25 681 0 R /F30 731 0 R /F36 669 0 R /F40 669 0 R /F42 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1501 0 obj << +/Length 1502 0 R +/Filter /FlateDecode +>> +stream +xXIo6r̯𥨍8")ji0hȡuCEjdՒ4@|W-cߊ = +/|EQDg2oB}+)P8׷x.;)wz/)R7~*o]lyU?0È-<=<Ĉ "ŴwO#-I @ Z@h'؏1KË|DpƄae Eꯎo^v.KPI3&T%p-ҦqB5@'x j%`#dgĘG^.ZQaNX GG G8\.-cOY?T;-Pvk^h6}^a[2l\xPjZ + /I7.6<t܋n4汪}y),T],HL?"oZ80@E1AYFerB}Pñ} +9rTB( 04|"ϊ]K(f($/;~NW-,#;QK!U# bȝ? +HOڏ+5]~PFn\ >oxy٫PORѤ怒4j4ܔFE'3 [OAѵʭt`)6 nypL{CH, L@%M.+*/Է. rzJeX_)sk^3XHp ( +|-C۝$D>0MuzVeFM،6:KM6}2 s{={05035L]Y@*RhyX#E%}m\~ƯR(ڇim}G[u^uM?=ی*;Zt6 &*I`$R0 ;Ltm`؟HtIgߦ58NN{saS.]B}A S;ИV.kZBl<ppOq!?$#ïf.Dr0U'^;qe3b~3`՛clӷS8ئN6I<5U*o'u>`?o{},ra.a[Vd=!S%1cKd;㹮> endobj +1497 0 obj << +/D [1500 0 R /XYZ 71.731 705.352 null] +>> endobj +1498 0 obj << +/D [1500 0 R /XYZ 71.731 717.307 null] +>> endobj +1423 0 obj << +/D [1500 0 R /XYZ 71.731 657.532 null] +>> endobj +1424 0 obj << +/D [1500 0 R /XYZ 155.257 647.569 null] +>> endobj +1425 0 obj << +/D [1500 0 R /XYZ 170.241 647.569 null] +>> endobj +1426 0 obj << +/D [1500 0 R /XYZ 209.782 647.569 null] +>> endobj +1427 0 obj << +/D [1500 0 R /XYZ 212.273 647.569 null] +>> endobj +1428 0 obj << +/D [1500 0 R /XYZ 139.476 634.618 null] +>> endobj +1429 0 obj << +/D [1500 0 R /XYZ 145.285 634.618 null] +>> endobj +1430 0 obj << +/D [1500 0 R /XYZ 208.378 634.618 null] +>> endobj +1431 0 obj << +/D [1500 0 R /XYZ 214.186 634.618 null] +>> endobj +1432 0 obj << +/D [1500 0 R /XYZ 283.257 634.618 null] +>> endobj +1433 0 obj << +/D [1500 0 R /XYZ 294.684 634.618 null] +>> endobj +1434 0 obj << +/D [1500 0 R /XYZ 369.732 634.618 null] +>> endobj +1435 0 obj << +/D [1500 0 R /XYZ 378.857 634.618 null] +>> endobj +1436 0 obj << +/D [1500 0 R /XYZ 441.951 634.618 null] +>> endobj +1437 0 obj << +/D [1500 0 R /XYZ 456.058 634.618 null] +>> endobj +1448 0 obj << +/D [1500 0 R /XYZ 71.731 604.566 null] +>> endobj +1449 0 obj << +/D [1500 0 R /XYZ 71.731 604.566 null] +>> endobj +1450 0 obj << +/D [1500 0 R /XYZ 71.731 586.768 null] +>> endobj +1451 0 obj << +/D [1500 0 R /XYZ 442.147 575.838 null] +>> endobj +1452 0 obj << +/D [1500 0 R /XYZ 71.731 547.943 null] +>> endobj +1453 0 obj << +/D [1500 0 R /XYZ 71.731 547.943 null] +>> endobj +1454 0 obj << +/D [1500 0 R /XYZ 71.731 534.872 null] +>> endobj +1455 0 obj << +/D [1500 0 R /XYZ 139.476 517.059 null] +>> endobj +1456 0 obj << +/D [1500 0 R /XYZ 157.459 517.059 null] +>> endobj +1457 0 obj << +/D [1500 0 R /XYZ 276.59 517.059 null] +>> endobj +1458 0 obj << +/D [1500 0 R /XYZ 71.731 479.037 null] +>> endobj +1459 0 obj << +/D [1500 0 R /XYZ 71.731 479.037 null] +>> endobj +1460 0 obj << +/D [1500 0 R /XYZ 159.68 468.242 null] +>> endobj +1461 0 obj << +/D [1500 0 R /XYZ 211.725 468.242 null] +>> endobj +1462 0 obj << +/D [1500 0 R /XYZ 227.764 468.242 null] +>> endobj +1463 0 obj << +/D [1500 0 R /XYZ 260.97 468.242 null] +>> endobj +1464 0 obj << +/D [1500 0 R /XYZ 279.779 468.242 null] +>> endobj +1465 0 obj << +/D [1500 0 R /XYZ 312.984 468.242 null] +>> endobj +1466 0 obj << +/D [1500 0 R /XYZ 327.37 468.242 null] +>> endobj +1467 0 obj << +/D [1500 0 R /XYZ 354.598 468.242 null] +>> endobj +1468 0 obj << +/D [1500 0 R /XYZ 375.619 468.242 null] +>> endobj +1469 0 obj << +/D [1500 0 R /XYZ 420.779 468.242 null] +>> endobj +1470 0 obj << +/D [1500 0 R /XYZ 439.04 468.242 null] +>> endobj +1471 0 obj << +/D [1500 0 R /XYZ 484.201 468.242 null] +>> endobj +1472 0 obj << +/D [1500 0 R /XYZ 486.692 468.242 null] +>> endobj +1473 0 obj << +/D [1500 0 R /XYZ 139.476 455.291 null] +>> endobj +1474 0 obj << +/D [1500 0 R /XYZ 145.285 455.291 null] +>> endobj +1475 0 obj << +/D [1500 0 R /XYZ 71.731 443.729 null] +>> endobj +1476 0 obj << +/D [1500 0 R /XYZ 442.147 432.377 null] +>> endobj +1477 0 obj << +/D [1500 0 R /XYZ 71.731 404.482 null] +>> endobj +1478 0 obj << +/D [1500 0 R /XYZ 71.731 404.482 null] +>> endobj +1479 0 obj << +/D [1500 0 R /XYZ 71.731 389.508 null] +>> endobj +1480 0 obj << +/D [1500 0 R /XYZ 139.476 373.598 null] +>> endobj +1481 0 obj << +/D [1500 0 R /XYZ 139.476 373.598 null] +>> endobj +1482 0 obj << +/D [1500 0 R /XYZ 409.67 373.598 null] +>> endobj +1483 0 obj << +/D [1500 0 R /XYZ 71.731 322.624 null] +>> endobj +1484 0 obj << +/D [1500 0 R /XYZ 71.731 322.624 null] +>> endobj +1485 0 obj << +/D [1500 0 R /XYZ 180.731 311.829 null] +>> endobj +1486 0 obj << +/D [1500 0 R /XYZ 226.11 311.829 null] +>> endobj +1487 0 obj << +/D [1500 0 R /XYZ 231.919 311.829 null] +>> endobj +1488 0 obj << +/D [1500 0 R /XYZ 231.919 311.829 null] +>> endobj +1489 0 obj << +/D [1500 0 R /XYZ 304.048 311.829 null] +>> endobj +1490 0 obj << +/D [1500 0 R /XYZ 316.83 311.829 null] +>> endobj +1491 0 obj << +/D [1500 0 R /XYZ 71.731 299.71 null] +>> endobj +1492 0 obj << +/D [1500 0 R /XYZ 71.731 257.932 null] +>> endobj +1493 0 obj << +/D [1500 0 R /XYZ 71.731 214.096 null] +>> endobj +1494 0 obj << +/D [1500 0 R /XYZ 150.266 188.293 null] +>> endobj +1495 0 obj << +/D [1500 0 R /XYZ 211.994 188.293 null] +>> endobj +717 0 obj << +/D [1500 0 R /XYZ 71.731 155.417 null] +>> endobj +1499 0 obj << +/Font << /F31 805 0 R /F30 731 0 R /F25 681 0 R /F48 1503 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1588 0 obj << +/Length 1589 0 R +/Filter /FlateDecode +>> +stream +xXK6_ TH"[4@[ַtDBdr+ɻ_%sh4#f#&?a )E%oߧ_ +ts&C3-y"Fl|ؾ;Uww 6C[uܾSxwnUrwOi7L1d9˒sg2x6 ƨ4.C9ˆhМ[cF nVn6}Ohb[4wn7;fyUJ7=Ud~9'\(B:u>tP;-6gf/YB>Sav.t_9AyOK7V26u{ɕ{e 9On1I +2pq4 r>qIt6 ( @ (?- t"/Ay J:JThh%r +XE=1bb +nՕORke9 IYYa +|J5F & /PoUB9RR+#,PFq%Ez"< ,M)be1aˣ(qQӡpUCjnutRs$ěe}4S@[Dq1Zal2׋ڄƬ +^ Nc:?SF🧈0_aXC\U؍ȰSz>\wluy>R\1X)R, hDoJlJGk l~=5HHh +ԋ>RzhǓmZ.$xh; @.\jp!av!!@|cndyր~Xdr8! Z@9*Fȕ6X=~8U=꓈d0k:u.O+'Pf8E <0GzZ\6UV=`s/dTD.Hi4^B[>)=h0Co@xS,+r4H\0Ϯhhm|PrƯGyB>#(JT +.ôygw18B9N)ՐUʾƬtbΝ/+Yy<E+a b7>SÛ?0yq$qO3Y D"{vKsԚs%3_ňTi+‹TɣhMܟ3Ok:΍'k|.r(c3۠ͯR94$^uQ牡?KgWZꧨ cI_ QX'k'endstream +endobj +1589 0 obj +1375 +endobj +1587 0 obj << +/Type /Page +/Contents 1588 0 R +/Resources 1586 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 1447 0 R +>> endobj +1584 0 obj << +/D [1587 0 R /XYZ 71.731 705.352 null] +>> endobj +1585 0 obj << +/D [1587 0 R /XYZ 71.731 717.307 null] +>> endobj +125 0 obj << +/D [1587 0 R /XYZ 280.046 647.569 null] +>> endobj +1496 0 obj << +/D [1587 0 R /XYZ 71.731 641.442 null] +>> endobj +1504 0 obj << +/D [1587 0 R /XYZ 71.731 611.54 null] +>> endobj +1505 0 obj << +/D [1587 0 R /XYZ 71.731 611.54 null] +>> endobj +1506 0 obj << +/D [1587 0 R /XYZ 71.731 600.626 null] +>> endobj +1507 0 obj << +/D [1587 0 R /XYZ 139.476 582.812 null] +>> endobj +1508 0 obj << +/D [1587 0 R /XYZ 157.459 582.812 null] +>> endobj +1509 0 obj << +/D [1587 0 R /XYZ 361.372 582.812 null] +>> endobj +1510 0 obj << +/D [1587 0 R /XYZ 71.731 570.693 null] +>> endobj +1511 0 obj << +/D [1587 0 R /XYZ 71.731 570.693 null] +>> endobj +1512 0 obj << +/D [1587 0 R /XYZ 216.676 559.898 null] +>> endobj +1513 0 obj << +/D [1587 0 R /XYZ 231.062 559.898 null] +>> endobj +1514 0 obj << +/D [1587 0 R /XYZ 289.004 559.898 null] +>> endobj +1515 0 obj << +/D [1587 0 R /XYZ 294.812 559.898 null] +>> endobj +1516 0 obj << +/D [1587 0 R /XYZ 71.731 527.022 null] +>> endobj +1517 0 obj << +/D [1587 0 R /XYZ 71.731 527.022 null] +>> endobj +1518 0 obj << +/D [1587 0 R /XYZ 71.731 513.951 null] +>> endobj +1519 0 obj << +/D [1587 0 R /XYZ 139.476 496.138 null] +>> endobj +1520 0 obj << +/D [1587 0 R /XYZ 157.459 496.138 null] +>> endobj +1521 0 obj << +/D [1587 0 R /XYZ 71.731 484.018 null] +>> endobj +1522 0 obj << +/D [1587 0 R /XYZ 71.731 484.018 null] +>> endobj +1523 0 obj << +/D [1587 0 R /XYZ 217.782 473.224 null] +>> endobj +1524 0 obj << +/D [1587 0 R /XYZ 240.188 473.224 null] +>> endobj +1525 0 obj << +/D [1587 0 R /XYZ 256.227 473.224 null] +>> endobj +1526 0 obj << +/D [1587 0 R /XYZ 289.432 473.224 null] +>> endobj +1527 0 obj << +/D [1587 0 R /XYZ 291.923 473.224 null] +>> endobj +1528 0 obj << +/D [1587 0 R /XYZ 318.324 473.224 null] +>> endobj +1529 0 obj << +/D [1587 0 R /XYZ 324.132 473.224 null] +>> endobj +1530 0 obj << +/D [1587 0 R /XYZ 71.731 440.347 null] +>> endobj +1531 0 obj << +/D [1587 0 R /XYZ 71.731 440.347 null] +>> endobj +1532 0 obj << +/D [1587 0 R /XYZ 71.731 427.276 null] +>> endobj +1533 0 obj << +/D [1587 0 R /XYZ 139.476 409.463 null] +>> endobj +1534 0 obj << +/D [1587 0 R /XYZ 279.749 409.463 null] +>> endobj +1535 0 obj << +/D [1587 0 R /XYZ 289.831 396.512 null] +>> endobj +1536 0 obj << +/D [1587 0 R /XYZ 71.731 373.498 null] +>> endobj +1537 0 obj << +/D [1587 0 R /XYZ 71.731 373.498 null] +>> endobj +1538 0 obj << +/D [1587 0 R /XYZ 161.743 360.646 null] +>> endobj +1539 0 obj << +/D [1587 0 R /XYZ 177.782 360.646 null] +>> endobj +1540 0 obj << +/D [1587 0 R /XYZ 217.533 360.646 null] +>> endobj +1541 0 obj << +/D [1587 0 R /XYZ 233.025 360.646 null] +>> endobj +1542 0 obj << +/D [1587 0 R /XYZ 272.566 360.646 null] +>> endobj +1543 0 obj << +/D [1587 0 R /XYZ 289.164 360.646 null] +>> endobj +1544 0 obj << +/D [1587 0 R /XYZ 316.392 360.646 null] +>> endobj +1545 0 obj << +/D [1587 0 R /XYZ 332.989 360.646 null] +>> endobj +1546 0 obj << +/D [1587 0 R /XYZ 355.066 360.646 null] +>> endobj +1547 0 obj << +/D [1587 0 R /XYZ 357.557 360.646 null] +>> endobj +1548 0 obj << +/D [1587 0 R /XYZ 402.289 360.646 null] +>> endobj +1549 0 obj << +/D [1587 0 R /XYZ 408.097 360.646 null] +>> endobj +1550 0 obj << +/D [1587 0 R /XYZ 71.731 348.527 null] +>> endobj +1551 0 obj << +/D [1587 0 R /XYZ 71.731 348.527 null] +>> endobj +1552 0 obj << +/D [1587 0 R /XYZ 71.731 330.594 null] +>> endobj +1553 0 obj << +/D [1587 0 R /XYZ 71.731 320.632 null] +>> endobj +1554 0 obj << +/D [1587 0 R /XYZ 71.731 320.632 null] +>> endobj +1555 0 obj << +/D [1587 0 R /XYZ 71.731 309.638 null] +>> endobj +1556 0 obj << +/D [1587 0 R /XYZ 159.402 291.904 null] +>> endobj +1557 0 obj << +/D [1587 0 R /XYZ 71.731 279.785 null] +>> endobj +1558 0 obj << +/D [1587 0 R /XYZ 71.731 279.785 null] +>> endobj +1559 0 obj << +/D [1587 0 R /XYZ 181.319 268.99 null] +>> endobj +1560 0 obj << +/D [1587 0 R /XYZ 211.247 268.99 null] +>> endobj +1561 0 obj << +/D [1587 0 R /XYZ 217.225 268.99 null] +>> endobj +1562 0 obj << +/D [1587 0 R /XYZ 253.16 268.99 null] +>> endobj +1563 0 obj << +/D [1587 0 R /XYZ 259.137 268.99 null] +>> endobj +1564 0 obj << +/D [1587 0 R /XYZ 71.731 267.929 null] +>> endobj +1565 0 obj << +/D [1587 0 R /XYZ 159.402 251.058 null] +>> endobj +1566 0 obj << +/D [1587 0 R /XYZ 71.731 238.938 null] +>> endobj +1567 0 obj << +/D [1587 0 R /XYZ 71.731 238.938 null] +>> endobj +1568 0 obj << +/D [1587 0 R /XYZ 187.297 228.144 null] +>> endobj +1569 0 obj << +/D [1587 0 R /XYZ 217.225 228.144 null] +>> endobj +1570 0 obj << +/D [1587 0 R /XYZ 223.202 228.144 null] +>> endobj +1571 0 obj << +/D [1587 0 R /XYZ 259.137 228.144 null] +>> endobj +1572 0 obj << +/D [1587 0 R /XYZ 265.115 228.144 null] +>> endobj +1573 0 obj << +/D [1587 0 R /XYZ 71.731 227.083 null] +>> endobj +1574 0 obj << +/D [1587 0 R /XYZ 159.402 210.211 null] +>> endobj +1575 0 obj << +/D [1587 0 R /XYZ 71.731 198.092 null] +>> endobj +1576 0 obj << +/D [1587 0 R /XYZ 71.731 198.092 null] +>> endobj +1577 0 obj << +/D [1587 0 R /XYZ 71.731 185.947 null] +>> endobj +1578 0 obj << +/D [1587 0 R /XYZ 159.402 169.364 null] +>> endobj +1579 0 obj << +/D [1587 0 R /XYZ 334.772 156.413 null] +>> endobj +1586 0 obj << +/Font << /F31 805 0 R /F42 669 0 R /F25 681 0 R /F30 731 0 R /F37 1210 0 R /F48 1503 0 R /F39 1242 0 R /F50 1590 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1650 0 obj << +/Length 1651 0 R +/Filter /FlateDecode +>> +stream +xXKo6r̯63zQcwi5EDKJf}")[u>F3oE$?eV%x"&>PͶ'.?VdX-Cd?J*([~lUىD|fczvNu{@^lTDcib'( BLO]gO(Ć_q.E̡äb4쫊UM> ^e~KQR1Tm4J"`U-۲-QT2^Zվ nGˋrXnZT]rUۭ +Q b`aB/Zk-]<ھZ`MME<̐煖2["zFɠ{ " )| rQi-vky&cMA{C(P-QpΩ`EǓ>myAŏc{H?0^JyEObry"ݽ$V`[ޘ͆|^v edʕ}[P'C]YY_oYn]AڴgH{ݯ$YhmĨ ʕU0bkLfL "zsg@za +d8^ڪ5w୚mXYjz\D[Ukۊ畯qmvz/2(yL`t + 9=sq({>sA{f@@ /[coTx(ɔѩ\on}?n5ߓݥCPޅmNVa"lZo*/.n;;&w5yϞݩnĀ!3xh3ά2LGVE>`HZ<9蘷MCxa*3VBcfY])a<1b5Sy2;5sQta-"߸N8FqzW7,GAs3gB'gasGEJMy#Q+7IĎQI/Ƽ4gg*ҘfH Wr_˂Lzc%B9q_AYYZPuX12tƳm2N;#=}s0kwn#˸_mn{Zm+\@i.g:l^ pI2-U'.%B?ܨ5'u@6HL> endobj +1647 0 obj << +/D [1649 0 R /XYZ 71.731 705.352 null] +>> endobj +1580 0 obj << +/D [1649 0 R /XYZ 71.731 657.532 null] +>> endobj +1581 0 obj << +/D [1649 0 R /XYZ 71.731 657.532 null] +>> endobj +1582 0 obj << +/D [1649 0 R /XYZ 71.731 647.37 null] +>> endobj +1583 0 obj << +/D [1649 0 R /XYZ 159.402 629.637 null] +>> endobj +1591 0 obj << +/D [1649 0 R /XYZ 71.731 605.373 null] +>> endobj +1592 0 obj << +/D [1649 0 R /XYZ 159.402 588.79 null] +>> endobj +1593 0 obj << +/D [1649 0 R /XYZ 492.708 588.79 null] +>> endobj +1594 0 obj << +/D [1649 0 R /XYZ 159.402 575.838 null] +>> endobj +1595 0 obj << +/D [1649 0 R /XYZ 215.56 575.838 null] +>> endobj +1596 0 obj << +/D [1649 0 R /XYZ 276.172 575.838 null] +>> endobj +1597 0 obj << +/D [1649 0 R /XYZ 351.17 575.838 null] +>> endobj +1598 0 obj << +/D [1649 0 R /XYZ 71.731 563.854 null] +>> endobj +1599 0 obj << +/D [1649 0 R /XYZ 71.731 563.854 null] +>> endobj +1600 0 obj << +/D [1649 0 R /XYZ 71.731 551.575 null] +>> endobj +1601 0 obj << +/D [1649 0 R /XYZ 159.402 534.992 null] +>> endobj +1602 0 obj << +/D [1649 0 R /XYZ 71.731 522.872 null] +>> endobj +1603 0 obj << +/D [1649 0 R /XYZ 71.731 522.872 null] +>> endobj +1604 0 obj << +/D [1649 0 R /XYZ 71.731 510.728 null] +>> endobj +1605 0 obj << +/D [1649 0 R /XYZ 159.402 494.145 null] +>> endobj +1606 0 obj << +/D [1649 0 R /XYZ 71.731 469.074 null] +>> endobj +1607 0 obj << +/D [1649 0 R /XYZ 71.731 469.074 null] +>> endobj +1608 0 obj << +/D [1649 0 R /XYZ 71.731 458.081 null] +>> endobj +1609 0 obj << +/D [1649 0 R /XYZ 159.402 440.347 null] +>> endobj +1610 0 obj << +/D [1649 0 R /XYZ 71.731 404.382 null] +>> endobj +1611 0 obj << +/D [1649 0 R /XYZ 71.731 404.382 null] +>> endobj +1612 0 obj << +/D [1649 0 R /XYZ 71.731 391.331 null] +>> endobj +1613 0 obj << +/D [1649 0 R /XYZ 159.402 373.598 null] +>> endobj +1614 0 obj << +/D [1649 0 R /XYZ 71.731 361.478 null] +>> endobj +1615 0 obj << +/D [1649 0 R /XYZ 71.731 361.478 null] +>> endobj +1616 0 obj << +/D [1649 0 R /XYZ 175.342 350.684 null] +>> endobj +1617 0 obj << +/D [1649 0 R /XYZ 71.731 350.484 null] +>> endobj +1618 0 obj << +/D [1649 0 R /XYZ 159.402 332.751 null] +>> endobj +1619 0 obj << +/D [1649 0 R /XYZ 71.731 320.632 null] +>> endobj +1620 0 obj << +/D [1649 0 R /XYZ 71.731 320.632 null] +>> endobj +1621 0 obj << +/D [1649 0 R /XYZ 71.731 309.638 null] +>> endobj +1622 0 obj << +/D [1649 0 R /XYZ 159.402 291.904 null] +>> endobj +1623 0 obj << +/D [1649 0 R /XYZ 71.731 279.785 null] +>> endobj +1624 0 obj << +/D [1649 0 R /XYZ 71.731 279.785 null] +>> endobj +1625 0 obj << +/D [1649 0 R /XYZ 71.731 268.791 null] +>> endobj +1626 0 obj << +/D [1649 0 R /XYZ 159.402 251.058 null] +>> endobj +1627 0 obj << +/D [1649 0 R /XYZ 71.731 240.995 null] +>> endobj +1628 0 obj << +/D [1649 0 R /XYZ 71.731 240.995 null] +>> endobj +1629 0 obj << +/D [1649 0 R /XYZ 71.731 227.944 null] +>> endobj +1630 0 obj << +/D [1649 0 R /XYZ 159.402 210.211 null] +>> endobj +1631 0 obj << +/D [1649 0 R /XYZ 71.731 203.073 null] +>> endobj +1632 0 obj << +/D [1649 0 R /XYZ 150.266 192.278 null] +>> endobj +1633 0 obj << +/D [1649 0 R /XYZ 194.54 192.278 null] +>> endobj +1634 0 obj << +/D [1649 0 R /XYZ 385.601 192.278 null] +>> endobj +1635 0 obj << +/D [1649 0 R /XYZ 458.237 192.278 null] +>> endobj +1636 0 obj << +/D [1649 0 R /XYZ 243.146 179.327 null] +>> endobj +1637 0 obj << +/D [1649 0 R /XYZ 71.731 172.189 null] +>> endobj +1638 0 obj << +/D [1649 0 R /XYZ 324.211 135.491 null] +>> endobj +1639 0 obj << +/D [1649 0 R /XYZ 426.456 135.491 null] +>> endobj +1648 0 obj << +/Font << /F31 805 0 R /F39 1242 0 R /F25 681 0 R /F30 731 0 R /F37 1210 0 R /F50 1590 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1696 0 obj << +/Length 1697 0 R +/Filter /FlateDecode +>> +stream +xڥYY6q~1ȃhq$Rl  &bAK-GG7CYW_UQƇOab|ś|l)i A4fRwDds߶ԉhE0\Ǻ:u@?~ý-fIbD} X0]k ЭPzu\IȔҏywT׻&-"E0",y*OrcퟻRE9(_Vުԏ}WٍVȟl"*syF ((P󢭍oZjmzwK5&[Bld_k<el0>Hu.m(F,OYA#0*тX=j갡p?wy7p)Fϴza5I +q Mdκ)bc +Q @dCI,I<"J}ئM3W1i(.6 $1#!(\  tb84Ra{C !Ý{}ݔ\Ul2/sW&Kq̐1+x>@T/_f*,Q4,Dq2KfrEP5xaZS&jZ%DIlMf Pq␾l(.ra-`bhy9#B1cZ"Ңτ0$1[S:=ވ4m œ-Z6B?%]̊7?ez<養U6FR/l )5C~35K8U~շCMlsQ " R uF/xߊ.(RK4z͚g&w"$bStFєKyCS3 K0(l20tP: H5sC?O8܈5G,ixlnv:xj4;ZEL7C~Xg.Z1?4ÄX_l0 SwsZ!Xl4rL{ 3?5`8ْ=&Yݗd.3ki4 d:;*`#H";KzeF/O!-lgʦX¤MS‹BGTim[eмU$~99 ҭD ; +2AUaŶQ~~ 4)ٶj:.LpOA4śB8k =gcHgG̑_h"c4}T7~gVGUn[x﫩sU0-&V-KiBu`vϔw<zҋÛ?$Z|^PXfe9o[l3EQKMme g"Ҁ@b 7(r& t_g68z2yDѹlv*rξh9og|+RHqeJ1Pēڮ[+ *Df>$gQ: 9a}d)IP!܁2<3[D4r5EQjFOL2\ T|KH(nFt';3IV_ŀ*$юCkuYMyPE> endobj +1698 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [139.476 608.616 191.78 615.47] +/Subtype /Link +/A << /S /GoTo /D (controls_statement_definition_and_usage) >> +>> endobj +1693 0 obj << +/D [1695 0 R /XYZ 71.731 705.352 null] +>> endobj +1640 0 obj << +/D [1695 0 R /XYZ 311.847 647.569 null] +>> endobj +1641 0 obj << +/D [1695 0 R /XYZ 502.759 647.569 null] +>> endobj +1642 0 obj << +/D [1695 0 R /XYZ 206.454 634.618 null] +>> endobj +1643 0 obj << +/D [1695 0 R /XYZ 331.594 634.618 null] +>> endobj +1644 0 obj << +/D [1695 0 R /XYZ 476.32 634.618 null] +>> endobj +1645 0 obj << +/D [1695 0 R /XYZ 226.25 621.666 null] +>> endobj +1646 0 obj << +/D [1695 0 R /XYZ 383.29 621.666 null] +>> endobj +1652 0 obj << +/D [1695 0 R /XYZ 71.731 603.634 null] +>> endobj +1653 0 obj << +/D [1695 0 R /XYZ 360.574 590.782 null] +>> endobj +1654 0 obj << +/D [1695 0 R /XYZ 201.184 577.831 null] +>> endobj +1655 0 obj << +/D [1695 0 R /XYZ 237.168 577.831 null] +>> endobj +1656 0 obj << +/D [1695 0 R /XYZ 256.994 577.831 null] +>> endobj +1657 0 obj << +/D [1695 0 R /XYZ 302.712 577.831 null] +>> endobj +1658 0 obj << +/D [1695 0 R /XYZ 71.731 544.79 null] +>> endobj +1659 0 obj << +/D [1695 0 R /XYZ 157.459 533.996 null] +>> endobj +1660 0 obj << +/D [1695 0 R /XYZ 303.837 533.996 null] +>> endobj +1661 0 obj << +/D [1695 0 R /XYZ 368.374 533.996 null] +>> endobj +1662 0 obj << +/D [1695 0 R /XYZ 438.281 533.996 null] +>> endobj +1663 0 obj << +/D [1695 0 R /XYZ 139.476 521.044 null] +>> endobj +1664 0 obj << +/D [1695 0 R /XYZ 201.174 508.093 null] +>> endobj +1665 0 obj << +/D [1695 0 R /XYZ 378.099 508.093 null] +>> endobj +1666 0 obj << +/D [1695 0 R /XYZ 243.884 495.141 null] +>> endobj +1667 0 obj << +/D [1695 0 R /XYZ 304.386 495.141 null] +>> endobj +1668 0 obj << +/D [1695 0 R /XYZ 465.43 495.141 null] +>> endobj +1669 0 obj << +/D [1695 0 R /XYZ 360.166 482.19 null] +>> endobj +1670 0 obj << +/D [1695 0 R /XYZ 71.731 475.052 null] +>> endobj +1671 0 obj << +/D [1695 0 R /XYZ 157.459 464.257 null] +>> endobj +1672 0 obj << +/D [1695 0 R /XYZ 468.597 451.306 null] +>> endobj +1673 0 obj << +/D [1695 0 R /XYZ 206.185 438.355 null] +>> endobj +1674 0 obj << +/D [1695 0 R /XYZ 331.334 438.355 null] +>> endobj +1675 0 obj << +/D [1695 0 R /XYZ 392.763 438.355 null] +>> endobj +1676 0 obj << +/D [1695 0 R /XYZ 476.079 425.403 null] +>> endobj +1677 0 obj << +/D [1695 0 R /XYZ 71.731 405.314 null] +>> endobj +1678 0 obj << +/D [1695 0 R /XYZ 157.414 394.519 null] +>> endobj +1679 0 obj << +/D [1695 0 R /XYZ 360.467 394.519 null] +>> endobj +1680 0 obj << +/D [1695 0 R /XYZ 164.383 381.568 null] +>> endobj +1681 0 obj << +/D [1695 0 R /XYZ 368.076 368.616 null] +>> endobj +1682 0 obj << +/D [1695 0 R /XYZ 446.561 368.616 null] +>> endobj +1683 0 obj << +/D [1695 0 R /XYZ 272.018 355.665 null] +>> endobj +1684 0 obj << +/D [1695 0 R /XYZ 71.731 348.527 null] +>> endobj +1685 0 obj << +/D [1695 0 R /XYZ 71.731 330.594 null] +>> endobj +1686 0 obj << +/D [1695 0 R /XYZ 71.731 221.814 null] +>> endobj +1687 0 obj << +/D [1695 0 R /XYZ 233.014 209.912 null] +>> endobj +1688 0 obj << +/D [1695 0 R /XYZ 71.731 203.491 null] +>> endobj +1689 0 obj << +/D [1695 0 R /XYZ 71.731 203.491 null] +>> endobj +1690 0 obj << +/D [1695 0 R /XYZ 150.236 191.979 null] +>> endobj +1691 0 obj << +/D [1695 0 R /XYZ 71.731 185.949 null] +>> endobj +1692 0 obj << +/D [1695 0 R /XYZ 71.731 153.957 null] +>> endobj +1694 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F37 1210 0 R /F30 731 0 R /F39 1242 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1717 0 obj << +/Length 1718 0 R +/Filter /FlateDecode +>> +stream +xVMo6r̯QV)mZ4(v@(vȔ!ɻC@̛7o MA$pɒrwM{Lp-Xs*@Ʉ~~PL#Ϩ¸*`|0qnKӄ+JRS9~0ݡMuv!9S:W<@f'# +VD|\`cDr +.m-K\|ߵ/u3+Aޙu5fCegHۅ,)y8. P UzBǗc{x3fQӷ%N +ʠUs<:[WyQlDD@dOm/rƢxfqgm=8Sq S +TqTqa5r2} ("at:#-'wH gef2mj*ȀY:vչ~.7DX Zqmkzs\Q8WDwLOtOU6MT4T)Xqz%q dv#ßnFrq͊LlwA +ݙıʘW'v?u^ SپQj5ܴMz}9)s0ȓ6]?^L%zOuۘ岉 JG؂_gmvND TV (I:l Xt?oLq|a{rOYo/˜4:K0)'xua䑫eOd9_:۴&z`ԟyOLeP3qIc&Kz2s_BTJSPT>sw~/+;)ub~OCg(Y˄sJ.e 4/V_nOVSrf1^a:ax ttej0!Ϲ2endstream +endobj +1718 0 obj +1006 +endobj +1716 0 obj << +/Type /Page +/Contents 1717 0 R +/Resources 1715 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 1447 0 R +>> endobj +1714 0 obj << +/D [1716 0 R /XYZ 71.731 705.352 null] +>> endobj +1699 0 obj << +/D [1716 0 R /XYZ 71.731 641.539 null] +>> endobj +1700 0 obj << +/D [1716 0 R /XYZ 301.367 629.637 null] +>> endobj +1701 0 obj << +/D [1716 0 R /XYZ 71.731 622.498 null] +>> endobj +1702 0 obj << +/D [1716 0 R /XYZ 190.953 611.704 null] +>> endobj +1703 0 obj << +/D [1716 0 R /XYZ 390.572 611.704 null] +>> endobj +1704 0 obj << +/D [1716 0 R /XYZ 244.9 598.752 null] +>> endobj +1705 0 obj << +/D [1716 0 R /XYZ 415.727 598.752 null] +>> endobj +1706 0 obj << +/D [1716 0 R /XYZ 189 585.801 null] +>> endobj +1707 0 obj << +/D [1716 0 R /XYZ 301.746 585.801 null] +>> endobj +1708 0 obj << +/D [1716 0 R /XYZ 425.87 585.801 null] +>> endobj +718 0 obj << +/D [1716 0 R /XYZ 71.731 542.962 null] +>> endobj +129 0 obj << +/D [1716 0 R /XYZ 162.641 513.552 null] +>> endobj +1709 0 obj << +/D [1716 0 R /XYZ 71.731 503.187 null] +>> endobj +1710 0 obj << +/D [1716 0 R /XYZ 261.497 480.476 null] +>> endobj +1711 0 obj << +/D [1716 0 R /XYZ 71.731 468.357 null] +>> endobj +1712 0 obj << +/D [1716 0 R /XYZ 71.731 463.376 null] +>> endobj +1713 0 obj << +/D [1716 0 R /XYZ 71.731 463.376 null] +>> endobj +1715 0 obj << +/Font << /F31 805 0 R /F37 1210 0 R /F25 681 0 R /F30 731 0 R /F40 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1744 0 obj << +/Length 1745 0 R +/Filter /FlateDecode +>> +stream +xڕXK6_aRX3z-$EZ A +=VV\>zz7dr8o_yWO^L8X e~+(IHy[;qdٿ_zd#I[zs}wg6Hw눘bo[S|k`|bWd$c#mKS-3"i]󹠒ՠ]F$T6IgZ@,֝Yǥ0Oy@FQv2ʪ+NY+i͋8A!yËR Fl5)ʣ ;Y ~@Vbx-? +&7uۯ7 uұ*_Dޖml&|wg&ę, ,Cb+ZomRX.hx^uEZn ҄oKq]T4 +t@#if7B*\6"$l g6"P"$R3.M0)Ʉ쭳ȹ9ZkYwW#n>}L ^? 6BK6O(4ވ#%2J`z\N65S_-\=㬥C,a 6H JΦIषlxm3NM[~\Sy|@[eb1%FX M qt*ز!Da4?]MlX_yvGB6H V[উJ,hwPrqX`̂/ز:p +z||ϡ[7`e>(ň +LNG?6Ϻ[/"m"TnS ΫK?5L2"\1A,qᩓQ༥tfA@i2ܴ}ʇ3wJu&(= 63iYQ~%G h6uo`(Z." RWGκ+G-֓PQ+ReAZyJTd3ASS8+u5x]f1O؀aUhj}G18@,BϧV)SFL{hC/W:iuK1&ZKsjNJ^ E,opNұ>/rtη' !n֊K!XUab GSbRo'T=M +"Uٙ>I 8˦e'"|9 Ըě'?T%L6{;}0K_3ف=XGƿe uh"IꏂbR`sQ8GR ,'D8y4!nl_mϴ G-㿭s;No@@ d*;8&$e#B@B~V1lLQ~~d|W1-zJ];[WB:9<=u{Ks䋷{w3iI! kSs٬ `\Sendstream +endobj +1745 0 obj +1857 +endobj +1743 0 obj << +/Type /Page +/Contents 1744 0 R +/Resources 1742 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 1746 0 R +>> endobj +1741 0 obj << +/D [1743 0 R /XYZ 71.731 705.352 null] +>> endobj +719 0 obj << +/D [1743 0 R /XYZ 71.731 657.532 null] +>> endobj +133 0 obj << +/D [1743 0 R /XYZ 376.358 642.462 null] +>> endobj +720 0 obj << +/D [1743 0 R /XYZ 71.731 621.4 null] +>> endobj +137 0 obj << +/D [1743 0 R /XYZ 240.112 590.385 null] +>> endobj +1719 0 obj << +/D [1743 0 R /XYZ 71.731 578.214 null] +>> endobj +1720 0 obj << +/D [1743 0 R /XYZ 71.731 548.736 null] +>> endobj +1721 0 obj << +/D [1743 0 R /XYZ 395.682 537.942 null] +>> endobj +1722 0 obj << +/D [1743 0 R /XYZ 464.215 537.942 null] +>> endobj +1723 0 obj << +/D [1743 0 R /XYZ 171.844 524.99 null] +>> endobj +1724 0 obj << +/D [1743 0 R /XYZ 71.731 519.89 null] +>> endobj +721 0 obj << +/D [1743 0 R /XYZ 71.731 469.036 null] +>> endobj +141 0 obj << +/D [1743 0 R /XYZ 213.655 441.783 null] +>> endobj +1725 0 obj << +/D [1743 0 R /XYZ 71.731 431.64 null] +>> endobj +1726 0 obj << +/D [1743 0 R /XYZ 317.526 395.756 null] +>> endobj +1727 0 obj << +/D [1743 0 R /XYZ 71.731 375.666 null] +>> endobj +1728 0 obj << +/D [1743 0 R /XYZ 71.731 318.879 null] +>> endobj +1729 0 obj << +/D [1743 0 R /XYZ 71.731 287.995 null] +>> endobj +1730 0 obj << +/D [1743 0 R /XYZ 71.731 270.062 null] +>> endobj +1731 0 obj << +/D [1743 0 R /XYZ 332.61 233.365 null] +>> endobj +1732 0 obj << +/D [1743 0 R /XYZ 71.731 226.227 null] +>> endobj +1733 0 obj << +/D [1743 0 R /XYZ 184.975 202.481 null] +>> endobj +1734 0 obj << +/D [1743 0 R /XYZ 332.321 202.481 null] +>> endobj +1735 0 obj << +/D [1743 0 R /XYZ 382.82 202.481 null] +>> endobj +1736 0 obj << +/D [1743 0 R /XYZ 158.973 189.53 null] +>> endobj +1737 0 obj << +/D [1743 0 R /XYZ 232.925 189.53 null] +>> endobj +1738 0 obj << +/D [1743 0 R /XYZ 491.83 189.53 null] +>> endobj +722 0 obj << +/D [1743 0 R /XYZ 71.731 138.621 null] +>> endobj +1742 0 obj << +/Font << /F29 669 0 R /F36 669 0 R /F25 681 0 R /F30 731 0 R /F40 669 0 R /F37 1210 0 R /F31 805 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1776 0 obj << +/Length 1777 0 R +/Filter /FlateDecode +>> +stream +xڥY[o6q~:vߦ3M7m;h1}%&"KH%=$/ ȃd:| MH61\iȏoÅ{Z3:N"mELF.~GvYP?^GiL/iB}+~UF1HKkI$H.Y@- Է*!4e]yǎ45w`њ?\|%ׯ7J݀f{!hB;@pKQccad +z=.4D8}ꚶ&Hw 7[<|V>Q!_k.ԛ ݤCp)%˕~uvmNq-s/f2fXqNVჀ(hTM=o.iY}B٣W} =JMTckA} F9%NH։_hFze11{FV Ǿ̹ R $Eƚ4e̎Cb^aÈU 4P!رCC"tLVe^ +T\¯@%&Ǧc57&ZyU0we+TQ"-Ȑ9im쯞q.>tfF$B\Wj*EMNTD[Id (2yD) ӀDq:^})wwL [_A +ЏK&2gL'mSރ"G1\)R=t"ZqԶgĊYpW 촗8jq;er?Ͽ&Tcl8I])Rg`&N=y7RVo?s:l!d" #ZT"4'33 4v\#ڐ#smд&<;ؖiN?Xa8'+0lG6i2(/Uf@ |i'zS`Njf1;v7yw[ +)/_-i\!h .L}՘j;/F=L/#$ B|4 5z#Cf1|BCt6Vr DЁbeHsP#yl,1s+4r-쩗VvezܿԞ\\fLof=˻1S=%2Unn[N&yp[tvط|h–;Čۉ܊sr6Vf4R1gH.od> endobj +1778 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [119.551 595.355 198.415 604.266] +/Subtype /Link +/A << /S /GoTo /D (proposed_standards) >> +>> endobj +1773 0 obj << +/D [1775 0 R /XYZ 71.731 705.352 null] +>> endobj +145 0 obj << +/D [1775 0 R /XYZ 384.668 644.973 null] +>> endobj +1739 0 obj << +/D [1775 0 R /XYZ 71.731 632.802 null] +>> endobj +1747 0 obj << +/D [1775 0 R /XYZ 71.731 590.373 null] +>> endobj +1748 0 obj << +/D [1775 0 R /XYZ 221.517 579.579 null] +>> endobj +1749 0 obj << +/D [1775 0 R /XYZ 71.731 533.586 null] +>> endobj +1750 0 obj << +/D [1775 0 R /XYZ 215.48 522.792 null] +>> endobj +1751 0 obj << +/D [1775 0 R /XYZ 362.507 509.841 null] +>> endobj +1752 0 obj << +/D [1775 0 R /XYZ 468.009 509.841 null] +>> endobj +723 0 obj << +/D [1775 0 R /XYZ 71.731 492.74 null] +>> endobj +149 0 obj << +/D [1775 0 R /XYZ 183.663 459.605 null] +>> endobj +1753 0 obj << +/D [1775 0 R /XYZ 71.731 447.434 null] +>> endobj +1754 0 obj << +/D [1775 0 R /XYZ 147.496 425.095 null] +>> endobj +1755 0 obj << +/D [1775 0 R /XYZ 71.731 417.956 null] +>> endobj +1756 0 obj << +/D [1775 0 R /XYZ 71.731 361.17 null] +>> endobj +1757 0 obj << +/D [1775 0 R /XYZ 71.731 317.334 null] +>> endobj +1758 0 obj << +/D [1775 0 R /XYZ 71.731 299.401 null] +>> endobj +1759 0 obj << +/D [1775 0 R /XYZ 232.063 288.607 null] +>> endobj +1760 0 obj << +/D [1775 0 R /XYZ 71.731 255.566 null] +>> endobj +1761 0 obj << +/D [1775 0 R /XYZ 71.731 255.566 null] +>> endobj +1762 0 obj << +/D [1775 0 R /XYZ 71.731 211.731 null] +>> endobj +1763 0 obj << +/D [1775 0 R /XYZ 71.731 168.612 null] +>> endobj +1764 0 obj << +/D [1775 0 R /XYZ 440.085 157.101 null] +>> endobj +1765 0 obj << +/D [1775 0 R /XYZ 119.551 144.149 null] +>> endobj +1766 0 obj << +/D [1775 0 R /XYZ 199.85 144.149 null] +>> endobj +1767 0 obj << +/D [1775 0 R /XYZ 313.164 144.149 null] +>> endobj +1774 0 obj << +/Font << /F31 805 0 R /F36 669 0 R /F25 681 0 R /F30 731 0 R /F37 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1830 0 obj << +/Length 1831 0 R +/Filter /FlateDecode +>> +stream +xY[F_a%v2pI$mmVjR#avMe\q3̅^jm\ x/BBAqC-\O{(XxD!뇟} (Mu65Z],}$ϲIL=gԫ_yZo]Vk˜HkGk0P.Қ+BմzdIZ\8mr+~WІWQK߰>irV puCd +L.A=OJV ňKv/J$_ =f. ˺lYIMgpV|Rf#cX=wU/nkqO8wĦ54mIQ* +OdgSѠn ' zHYv_kN\3ju4 _m͛9-r_ec hYBsp)➜{M+=;f땍13@]u-V<ü| sٚ6qyٛ֘=j K`.N|;ZAd |")kw3*tioTulBTbꋥ[!`A6so65V#ս#K >O@Q_]<ׁGk؈onh}w7'Οƌendstream +endobj +1831 0 obj +1791 +endobj +1829 0 obj << +/Type /Page +/Contents 1830 0 R +/Resources 1828 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 1746 0 R +/Annots [ 1832 0 R ] +>> endobj +1832 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [452.18 143.596 497.011 152.507] +/Subtype /Link +/A << /S /GoTo /D (sample_configuration) >> +>> endobj +1826 0 obj << +/D [1829 0 R /XYZ 71.731 705.352 null] +>> endobj +1827 0 obj << +/D [1829 0 R /XYZ 71.731 717.307 null] +>> endobj +1768 0 obj << +/D [1829 0 R /XYZ 379.134 647.569 null] +>> endobj +1769 0 obj << +/D [1829 0 R /XYZ 119.551 634.618 null] +>> endobj +1770 0 obj << +/D [1829 0 R /XYZ 213.498 634.618 null] +>> endobj +1771 0 obj << +/D [1829 0 R /XYZ 215.989 634.618 null] +>> endobj +1772 0 obj << +/D [1829 0 R /XYZ 313.164 634.618 null] +>> endobj +1779 0 obj << +/D [1829 0 R /XYZ 71.731 628.197 null] +>> endobj +1780 0 obj << +/D [1829 0 R /XYZ 177.135 616.685 null] +>> endobj +1781 0 obj << +/D [1829 0 R /XYZ 271.819 616.685 null] +>> endobj +1782 0 obj << +/D [1829 0 R /XYZ 71.731 596.596 null] +>> endobj +1783 0 obj << +/D [1829 0 R /XYZ 134.216 572.85 null] +>> endobj +1784 0 obj << +/D [1829 0 R /XYZ 180.482 572.85 null] +>> endobj +1785 0 obj << +/D [1829 0 R /XYZ 180.821 559.898 null] +>> endobj +1786 0 obj << +/D [1829 0 R /XYZ 280.885 559.898 null] +>> endobj +1787 0 obj << +/D [1829 0 R /XYZ 122.869 546.947 null] +>> endobj +1788 0 obj << +/D [1829 0 R /XYZ 228.313 546.947 null] +>> endobj +1789 0 obj << +/D [1829 0 R /XYZ 71.731 540.182 null] +>> endobj +1790 0 obj << +/D [1829 0 R /XYZ 210.041 529.014 null] +>> endobj +1791 0 obj << +/D [1829 0 R /XYZ 256.307 529.014 null] +>> endobj +1792 0 obj << +/D [1829 0 R /XYZ 71.731 483.739 null] +>> endobj +1793 0 obj << +/D [1829 0 R /XYZ 71.731 460.108 null] +>> endobj +1794 0 obj << +/D [1829 0 R /XYZ 71.731 460.108 null] +>> endobj +1795 0 obj << +/D [1829 0 R /XYZ 71.731 439.379 null] +>> endobj +1796 0 obj << +/D [1829 0 R /XYZ 71.731 394.654 null] +>> endobj +1797 0 obj << +/D [1829 0 R /XYZ 71.731 365.827 null] +>> endobj +1798 0 obj << +/D [1829 0 R /XYZ 471.865 352.975 null] +>> endobj +1799 0 obj << +/D [1829 0 R /XYZ 71.731 332.886 null] +>> endobj +1800 0 obj << +/D [1829 0 R /XYZ 262.095 322.091 null] +>> endobj +1801 0 obj << +/D [1829 0 R /XYZ 71.731 314.953 null] +>> endobj +1802 0 obj << +/D [1829 0 R /XYZ 71.731 309.972 null] +>> endobj +1803 0 obj << +/D [1829 0 R /XYZ 129.514 294.196 null] +>> endobj +1804 0 obj << +/D [1829 0 R /XYZ 251.675 294.196 null] +>> endobj +1805 0 obj << +/D [1829 0 R /XYZ 297.941 294.196 null] +>> endobj +1806 0 obj << +/D [1829 0 R /XYZ 71.731 292.039 null] +>> endobj +1807 0 obj << +/D [1829 0 R /XYZ 129.514 276.263 null] +>> endobj +1808 0 obj << +/D [1829 0 R /XYZ 251.675 276.263 null] +>> endobj +1809 0 obj << +/D [1829 0 R /XYZ 346.359 276.263 null] +>> endobj +1810 0 obj << +/D [1829 0 R /XYZ 71.731 274.106 null] +>> endobj +1811 0 obj << +/D [1829 0 R /XYZ 129.514 258.33 null] +>> endobj +1812 0 obj << +/D [1829 0 R /XYZ 71.731 256.173 null] +>> endobj +1813 0 obj << +/D [1829 0 R /XYZ 129.514 240.398 null] +>> endobj +1814 0 obj << +/D [1829 0 R /XYZ 71.731 233.26 null] +>> endobj +1815 0 obj << +/D [1829 0 R /XYZ 71.731 217.384 null] +>> endobj +1816 0 obj << +/D [1829 0 R /XYZ 71.731 212.403 null] +>> endobj +1817 0 obj << +/D [1829 0 R /XYZ 129.514 194.57 null] +>> endobj +1818 0 obj << +/D [1829 0 R /XYZ 251.675 194.57 null] +>> endobj +1819 0 obj << +/D [1829 0 R /XYZ 297.941 194.57 null] +>> endobj +1820 0 obj << +/D [1829 0 R /XYZ 71.731 192.413 null] +>> endobj +1821 0 obj << +/D [1829 0 R /XYZ 129.514 176.637 null] +>> endobj +1822 0 obj << +/D [1829 0 R /XYZ 267.714 176.637 null] +>> endobj +1823 0 obj << +/D [1829 0 R /XYZ 313.98 176.637 null] +>> endobj +1824 0 obj << +/D [1829 0 R /XYZ 71.731 169.499 null] +>> endobj +1828 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F37 1210 0 R /F52 1210 0 R /F45 681 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1837 0 obj << +/Length 1838 0 R +/Filter /FlateDecode +>> +stream +xڵVM08zvz$@Bo!nD'g'i>{XTqF7of^? GeKs_^ 6MCEL +vs#qmvfmyuJe.}‚eں+ CY\|xi`B-69J~% +3-)Ƥ>@)3ݗnQ(slMt O w0>V X5j^a?Xpmx#kqmCYtHbGWJ@XVB2չ +Z7ʨqFQ}>\hK#1;F !q**NlγISa~Pz(wLol-!Ey~Qö$^SYy*(׹[aSƦE?dʭl} r_+}CX+]+ԓ|;d*J&{IbB&.Qɭ)gpBΆ;<Ѿuuc +-T=(#i%LL ? $xކI~5&5]Ēwna}3LeztI(P(pQ+:md$"X=+{3 +v\5S<5tC?b4Le(zBj?ũizWd{k07x1b٢!Iy8Eo]BfngD8`cendstream +endobj +1838 0 obj +749 +endobj +1836 0 obj << +/Type /Page +/Contents 1837 0 R +/Resources 1835 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 1746 0 R +>> endobj +1834 0 obj << +/D [1836 0 R /XYZ 71.731 705.352 null] +>> endobj +1825 0 obj << +/D [1836 0 R /XYZ 71.731 657.532 null] +>> endobj +1833 0 obj << +/D [1836 0 R /XYZ 162.59 497.732 null] +>> endobj +1835 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F37 1210 0 R /F52 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1847 0 obj << +/Length 1848 0 R +/Filter /FlateDecode +>> +stream +xVM0z̯@9 6ZV8T٤$@Rl&7oaAa#Jg7}67r~Yr}C1\64/YسMzG5q%$pZR$%#-֬ ۶SOW+_$|`šg2&ⲠT"GZ _cK=Ϙƙd6pfjML*yN$tdž8om5TID0D>Δ> endobj +1843 0 obj << +/D [1846 0 R /XYZ 71.731 705.352 null] +>> endobj +1844 0 obj << +/D [1846 0 R /XYZ 71.731 717.307 null] +>> endobj +1839 0 obj << +/D [1846 0 R /XYZ 71.731 496.683 null] +>> endobj +1840 0 obj << +/D [1846 0 R /XYZ 71.731 472.661 null] +>> endobj +1845 0 obj << +/Font << /F31 805 0 R /F37 1210 0 R /F25 681 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1874 0 obj << +/Length 1875 0 R +/Filter /FlateDecode +>> +stream +xڕXIs61B4 $㦉lĘBv_߇HM< Ax^\\R4 pRz ;{e"U]&Meʝt{ЋG5+KL`(YB̈4"$·} O:kD# .Pmҩ6Veg>X:ښJ3<} !k$XC{2Z[3Yn'0e+#/ -x=-D`c4kUAW|ʪɶU@sSK +ӅzGk`~SN,& @N2G4җ=[4Y5V Dwꛌ\LVKnoDl4TmSpeqHY8N ;6"6pk4Q0uj| ѢS@lV/[Wg5c*ǝhUu?K2n\eIuV q,8?f jU[ouu:o(VVF(s|h-gp=(7BLUfFFķf#}FXLo*U^n{X7LE( +C+Č`ePE?a7,櫃lti>DW`2|vH<8)K$0fjKp@.+,k+¸jƐcMrv +*Ťm]NgVCcm(O194̀Nug]6y +Cǀd{D)E9w Ӈ1`S`㥳1G5Rm3?|];_u{q@yZȃfЂ/\8Cg菛hUd + fAޅ|@ +E$' ˤJyԝn Ar$ƶm0Ƣ"Q%'#ۺ*YSUGOO֧sp ( G>7 +c| :=Ld2Lb\N#$}ؠS J%t <_| qsO-gNendstream +endobj +1875 0 obj +1752 +endobj +1873 0 obj << +/Type /Page +/Contents 1874 0 R +/Resources 1872 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 1746 0 R +/Annots [ 1876 0 R ] +>> endobj +1876 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [171.994 398.475 250.857 407.496] +/Subtype /Link +/A << /S /GoTo /D (proposed_standards) >> +>> endobj +1871 0 obj << +/D [1873 0 R /XYZ 71.731 705.352 null] +>> endobj +1841 0 obj << +/D [1873 0 R /XYZ 71.731 657.532 null] +>> endobj +1842 0 obj << +/D [1873 0 R /XYZ 145.005 647.569 null] +>> endobj +724 0 obj << +/D [1873 0 R /XYZ 71.731 570.868 null] +>> endobj +153 0 obj << +/D [1873 0 R /XYZ 145.411 535.765 null] +>> endobj +1849 0 obj << +/D [1873 0 R /XYZ 71.731 526.942 null] +>> endobj +1850 0 obj << +/D [1873 0 R /XYZ 498.006 514.206 null] +>> endobj +1851 0 obj << +/D [1873 0 R /XYZ 469.045 488.303 null] +>> endobj +1852 0 obj << +/D [1873 0 R /XYZ 71.731 481.165 null] +>> endobj +1853 0 obj << +/D [1873 0 R /XYZ 71.731 481.165 null] +>> endobj +1854 0 obj << +/D [1873 0 R /XYZ 394.995 457.419 null] +>> endobj +1855 0 obj << +/D [1873 0 R /XYZ 71.731 439.387 null] +>> endobj +1856 0 obj << +/D [1873 0 R /XYZ 273.821 400.632 null] +>> endobj +1857 0 obj << +/D [1873 0 R /XYZ 441.52 400.632 null] +>> endobj +1858 0 obj << +/D [1873 0 R /XYZ 471.647 400.632 null] +>> endobj +806 0 obj << +/D [1873 0 R /XYZ 71.731 375.561 null] +>> endobj +157 0 obj << +/D [1873 0 R /XYZ 417.267 348.309 null] +>> endobj +1859 0 obj << +/D [1873 0 R /XYZ 71.731 338.166 null] +>> endobj +1860 0 obj << +/D [1873 0 R /XYZ 320.116 328.184 null] +>> endobj +1861 0 obj << +/D [1873 0 R /XYZ 361.072 328.184 null] +>> endobj +807 0 obj << +/D [1873 0 R /XYZ 71.731 303.113 null] +>> endobj +161 0 obj << +/D [1873 0 R /XYZ 287.279 280.762 null] +>> endobj +1862 0 obj << +/D [1873 0 R /XYZ 71.731 274.635 null] +>> endobj +1863 0 obj << +/D [1873 0 R /XYZ 71.731 228.792 null] +>> endobj +1864 0 obj << +/D [1873 0 R /XYZ 71.731 228.792 null] +>> endobj +1865 0 obj << +/D [1873 0 R /XYZ 71.731 211.666 null] +>> endobj +1866 0 obj << +/D [1873 0 R /XYZ 203.426 200.065 null] +>> endobj +1867 0 obj << +/D [1873 0 R /XYZ 258.947 187.114 null] +>> endobj +1868 0 obj << +/D [1873 0 R /XYZ 71.731 174.994 null] +>> endobj +1869 0 obj << +/D [1873 0 R /XYZ 71.731 154.125 null] +>> endobj +1870 0 obj << +/D [1873 0 R /XYZ 166.783 142.581 null] +>> endobj +1872 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F37 1210 0 R /F36 669 0 R /F30 731 0 R /F40 669 0 R /F42 669 0 R /F39 1242 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1901 0 obj << +/Length 1902 0 R +/Filter /FlateDecode +>> +stream +xڕXK6r̯0rFH9l68ihDGw-;9MV5|C3,$ G$,әX3A[ƳDD Z?~$zɭ?{~hEXCb 3K V-^f,msL,5e[`\7E[x)s<{#jWU6EFf9/\׹!ڹd\cK/EkCձlC9rzݡ3{X}8*EL'Uh DNA(|6pۺ=51lq7L hH–Ue(0ed(&E@v]-Z9t3>rN1H"IcP)"/LȒA& +]a]ÂF;,f0iidG.e3 $|BHKB x"N&#tElj?"[HduT%u0XkZ}'=OatW{{׊RX6.@ Ъ.xTgBk9 WO t(B]* "4@ѩ:?I.3-4&$i13b?OsQeij<殞c. +MT>]Dec}2w!dڨAy$nb}CFGu  YNbjZ۵"u:JR$@æɦhX̫cEX=?]~yTW/˗ON Eh8(hճ^9跲S'48t~c6է〟F5bMHlqre]n6h'OFڡwȤ ' `>ZޭвM ;CZa[N\i Raƈ(qendstream +endobj +1902 0 obj +1504 +endobj +1900 0 obj << +/Type /Page +/Contents 1901 0 R +/Resources 1899 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 1903 0 R +>> endobj +1897 0 obj << +/D [1900 0 R /XYZ 71.731 705.352 null] +>> endobj +1898 0 obj << +/D [1900 0 R /XYZ 71.731 717.307 null] +>> endobj +808 0 obj << +/D [1900 0 R /XYZ 71.731 657.532 null] +>> endobj +165 0 obj << +/D [1900 0 R /XYZ 270.255 647.569 null] +>> endobj +1877 0 obj << +/D [1900 0 R /XYZ 71.731 641.442 null] +>> endobj +1878 0 obj << +/D [1900 0 R /XYZ 71.731 595.599 null] +>> endobj +1879 0 obj << +/D [1900 0 R /XYZ 282.279 584.805 null] +>> endobj +809 0 obj << +/D [1900 0 R /XYZ 71.731 546.847 null] +>> endobj +169 0 obj << +/D [1900 0 R /XYZ 418.157 517.537 null] +>> endobj +1880 0 obj << +/D [1900 0 R /XYZ 71.731 507.172 null] +>> endobj +810 0 obj << +/D [1900 0 R /XYZ 71.731 467.361 null] +>> endobj +173 0 obj << +/D [1900 0 R /XYZ 414.513 440.108 null] +>> endobj +1881 0 obj << +/D [1900 0 R /XYZ 71.731 429.743 null] +>> endobj +1882 0 obj << +/D [1900 0 R /XYZ 154.689 419.984 null] +>> endobj +1883 0 obj << +/D [1900 0 R /XYZ 195.645 419.984 null] +>> endobj +1884 0 obj << +/D [1900 0 R /XYZ 448.812 419.984 null] +>> endobj +1885 0 obj << +/D [1900 0 R /XYZ 71.731 407.864 null] +>> endobj +1886 0 obj << +/D [1900 0 R /XYZ 71.731 352.384 null] +>> endobj +1887 0 obj << +/D [1900 0 R /XYZ 343.967 340.482 null] +>> endobj +1888 0 obj << +/D [1900 0 R /XYZ 324.031 327.531 null] +>> endobj +1889 0 obj << +/D [1900 0 R /XYZ 383.01 314.579 null] +>> endobj +1890 0 obj << +/D [1900 0 R /XYZ 71.731 307.441 null] +>> endobj +811 0 obj << +/D [1900 0 R /XYZ 71.731 266.594 null] +>> endobj +177 0 obj << +/D [1900 0 R /XYZ 363.24 239.342 null] +>> endobj +1891 0 obj << +/D [1900 0 R /XYZ 71.731 228.977 null] +>> endobj +1892 0 obj << +/D [1900 0 R /XYZ 220.86 206.266 null] +>> endobj +1893 0 obj << +/D [1900 0 R /XYZ 306.478 206.266 null] +>> endobj +1894 0 obj << +/D [1900 0 R /XYZ 410.796 206.266 null] +>> endobj +1895 0 obj << +/D [1900 0 R /XYZ 71.731 194.146 null] +>> endobj +1899 0 obj << +/Font << /F31 805 0 R /F42 669 0 R /F25 681 0 R /F30 731 0 R /F40 669 0 R /F37 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1932 0 obj << +/Length 1933 0 R +/Filter /FlateDecode +>> +stream +xڝX[s6c~N^ ĥy'u$gi /Z-$GH:Â:|;7*^%%!gx~w+s0"*D4Mͳ_p6[vhy bgA&!AG8EQi4"U]WšJ_c=gZo94\=($V=RonoF臅><׶onb^4|kBOR~;PխZeU|VZm%` (vT)s,g=< h8.AY] akokbe 4{.P6Qjfy2 nξ/҆C7UOI8hogo'AB!"hSH& J* $cu% mz@%{`\ F<(qf%/Z=D֨3@N?JQ56kX[Ԛ'&Р@ycjX-ƕ@SfbJWE1B(b vjvGy#FcQ ѨQ||ު|) L˳UPFfۺt0@Q/߿ؗe-6'=@=Vi7:&yts6? +I,* a#{ !pXVx_)ۤ?;[ꥯ.PaHdSuSn0jZ&E)Njeiuu-L 4L*SW9nv `C.\v;`mwC*wH%cڌd s1GA}Vw*yaY+{Ţyp*dFC&ؾ+l׋BYQo!ش\?X5fHm E Ž:L>7E 2} +dQWqnuS k8mChSʼukDH+`uYl\CAL3eWU@t`HO +-r?knV)(ˈ#.tGO_,(\Gv +J EC#i"B yԲՍcvݚR]ca}k1rg}AMѐs^q+TIw7Z+YN-^zwqueyk #F׀f(``ZM12@H\F-,cvr&y}!=ީWVu aW&P1nf1k̩m>єޑ9 1d &\=S=3\8[hfOgv efy;) +]?A'd#9on,}`R̙?˛B, ,uU q`"seDJutiHԖտ9(endstream +endobj +1933 0 obj +1698 +endobj +1931 0 obj << +/Type /Page +/Contents 1932 0 R +/Resources 1930 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 1903 0 R +/Annots [ 1934 0 R ] +>> endobj +1934 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [421.375 399.914 486.132 408.935] +/Subtype /Link +/A << /S /GoTo /D (dynamic_update_policies) >> +>> endobj +1929 0 obj << +/D [1931 0 R /XYZ 71.731 705.352 null] +>> endobj +1896 0 obj << +/D [1931 0 R /XYZ 71.731 657.532 null] +>> endobj +1904 0 obj << +/D [1931 0 R /XYZ 138.022 603.734 null] +>> endobj +1905 0 obj << +/D [1931 0 R /XYZ 71.731 596.596 null] +>> endobj +1906 0 obj << +/D [1931 0 R /XYZ 275.344 585.801 null] +>> endobj +1907 0 obj << +/D [1931 0 R /XYZ 400.433 585.801 null] +>> endobj +1908 0 obj << +/D [1931 0 R /XYZ 480.95 585.801 null] +>> endobj +1909 0 obj << +/D [1931 0 R /XYZ 201.174 572.85 null] +>> endobj +812 0 obj << +/D [1931 0 R /XYZ 71.731 555.749 null] +>> endobj +181 0 obj << +/D [1931 0 R /XYZ 330.934 528.496 null] +>> endobj +1910 0 obj << +/D [1931 0 R /XYZ 71.731 518.354 null] +>> endobj +1911 0 obj << +/D [1931 0 R /XYZ 71.731 518.354 null] +>> endobj +1912 0 obj << +/D [1931 0 R /XYZ 425.769 508.372 null] +>> endobj +1913 0 obj << +/D [1931 0 R /XYZ 119.551 482.469 null] +>> endobj +1914 0 obj << +/D [1931 0 R /XYZ 71.731 475.465 null] +>> endobj +1915 0 obj << +/D [1931 0 R /XYZ 71.731 452.417 null] +>> endobj +1916 0 obj << +/D [1931 0 R /XYZ 71.731 431.547 null] +>> endobj +1917 0 obj << +/D [1931 0 R /XYZ 472.204 420.003 null] +>> endobj +1918 0 obj << +/D [1931 0 R /XYZ 71.731 412.865 null] +>> endobj +1919 0 obj << +/D [1931 0 R /XYZ 309.845 402.071 null] +>> endobj +813 0 obj << +/D [1931 0 R /XYZ 71.731 384.97 null] +>> endobj +185 0 obj << +/D [1931 0 R /XYZ 154.162 357.717 null] +>> endobj +1920 0 obj << +/D [1931 0 R /XYZ 71.731 350.365 null] +>> endobj +1921 0 obj << +/D [1931 0 R /XYZ 71.731 291.601 null] +>> endobj +814 0 obj << +/D [1931 0 R /XYZ 71.731 178.092 null] +>> endobj +1930 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F40 669 0 R /F30 731 0 R /F37 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1956 0 obj << +/Length 1957 0 R +/Filter /FlateDecode +>> +stream +xڕX[o6_10#C75kӝN?OpI +e_9YQ$DQ5U#I|nUFrcʱܟ~YFّ$ϫiǥ9³++/TpugAWyM^e +̠.wARLeL&NtJ+ zn|Ma8NTy<*؏y3@Y&!BxձE{'#q,i x#A=~wZveBR͝%"أpۅX~am{Z~ǀ9(gRϼ8$n ,eIXѩղǙmE  wB G&p(e kl%eE◀^˚)4:(\zpٳ@)c+|L|qt#ZO?_e cV?]I DKiA3OTqCKfJHJ)& l|WkAQAzG,mKa-l;8g}WKvxt堀 F:6H4Nm^/w;*\s8rSGB?lVWɊH_'njV\I ׻޸J 36q?P3f ;e-CI:f(j[zBӉVg__sN chM^gM4>=Ƚ=A1L`w0}@1E `Yc*S]MsMW~5J}7^#:/HyL_DLH~qjPӸ +pey$yUI^AEJgB#IX_7 d=I/N4JIo#u?Ρ9J%r_˸S%q~8<;`WS{|g9(c_SZtgo<wY45H3 noysA!z$.k׎~ u ܱԵK0Td0y= 97do^P$a%& `NۛoZŭ F5Tl^"aCzƴpNbcB ͺ7A6ea}V3[¢gN>xOfN#L=Z.o-o{OD9ANwotu;KUCV\xƦ}.qZvL}K #r) X촄.^Q a#;DAH+Oz?f,Y=\$%z:;RJkOy䘻ATYPKj&t^^ +ኴ^yZEsZ}T{Bwȗߣq5yמ0gU!4^Ig w[2g!z=m>`+>yeZ8x +Jk\b=QcsA4#XjdT+rkA//#yt:azEp(…N5cɹnta/1Q|˕Z +'hy''>ɉǩ^׳G,!x^sj +cWC,ꊯZx `i{Tj-uʿOCÄM'F}E"He?nབx4$CCCǎ0e^Ĥȿyh. 3p~.WC?FW4endstream +endobj +1957 0 obj +1914 +endobj +1955 0 obj << +/Type /Page +/Contents 1956 0 R +/Resources 1954 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 1903 0 R +>> endobj +1953 0 obj << +/D [1955 0 R /XYZ 71.731 705.352 null] +>> endobj +189 0 obj << +/D [1955 0 R /XYZ 151.144 645.343 null] +>> endobj +1922 0 obj << +/D [1955 0 R /XYZ 71.731 636.521 null] +>> endobj +1923 0 obj << +/D [1955 0 R /XYZ 71.731 636.521 null] +>> endobj +1924 0 obj << +/D [1955 0 R /XYZ 166.973 610.833 null] +>> endobj +1925 0 obj << +/D [1955 0 R /XYZ 397.515 610.833 null] +>> endobj +1926 0 obj << +/D [1955 0 R /XYZ 408.673 584.93 null] +>> endobj +1927 0 obj << +/D [1955 0 R /XYZ 349.755 571.979 null] +>> endobj +1928 0 obj << +/D [1955 0 R /XYZ 226.09 559.027 null] +>> endobj +1935 0 obj << +/D [1955 0 R /XYZ 71.731 538.938 null] +>> endobj +1936 0 obj << +/D [1955 0 R /XYZ 137.534 528.143 null] +>> endobj +1937 0 obj << +/D [1955 0 R /XYZ 407.727 528.143 null] +>> endobj +1938 0 obj << +/D [1955 0 R /XYZ 119.551 502.241 null] +>> endobj +1939 0 obj << +/D [1955 0 R /XYZ 376.424 489.289 null] +>> endobj +1940 0 obj << +/D [1955 0 R /XYZ 182.664 476.338 null] +>> endobj +815 0 obj << +/D [1955 0 R /XYZ 71.731 446.286 null] +>> endobj +193 0 obj << +/D [1955 0 R /XYZ 155.93 413.151 null] +>> endobj +1941 0 obj << +/D [1955 0 R /XYZ 71.731 400.98 null] +>> endobj +1942 0 obj << +/D [1955 0 R /XYZ 71.731 400.98 null] +>> endobj +1943 0 obj << +/D [1955 0 R /XYZ 71.731 358.551 null] +>> endobj +1944 0 obj << +/D [1955 0 R /XYZ 71.731 327.667 null] +>> endobj +1945 0 obj << +/D [1955 0 R /XYZ 71.731 309.734 null] +>> endobj +1946 0 obj << +/D [1955 0 R /XYZ 71.731 309.734 null] +>> endobj +816 0 obj << +/D [1955 0 R /XYZ 71.731 281.839 null] +>> endobj +197 0 obj << +/D [1955 0 R /XYZ 176.967 248.704 null] +>> endobj +1947 0 obj << +/D [1955 0 R /XYZ 71.731 239.882 null] +>> endobj +1948 0 obj << +/D [1955 0 R /XYZ 475.133 227.145 null] +>> endobj +1949 0 obj << +/D [1955 0 R /XYZ 71.731 207.056 null] +>> endobj +1950 0 obj << +/D [1955 0 R /XYZ 492.538 196.261 null] +>> endobj +1951 0 obj << +/D [1955 0 R /XYZ 164.283 170.358 null] +>> endobj +1954 0 obj << +/Font << /F31 805 0 R /F36 669 0 R /F30 731 0 R /F25 681 0 R /F37 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +1986 0 obj << +/Length 1987 0 R +/Filter /FlateDecode +>> +stream +xڭX[o6_af6+Rԭ{JtZ@a@=(m ;/PvS`ȃ\s#^8F ?\,^ȿz$K "pwp>B.n95gjC+L`9.{K:z&VUOS͌bd}(?|QJBD\H֓J`c>;A:v| v`ej01"*"{Is%n +Jӡ>5MgKm*iYQ d/ϔT>6t=I̥O׏\cI9_AVLĸǺS?u`'}m&wUW+AvR, +e`z2Pͅg*nXV$RE0:!$bAVb4%2N2ao|Ynbm<6^Iպ EمOIlѥUvP઎>*s]J:ԠdGqP0 ޏz8+h}@S[e1!ЁV-s `eE|(tL:- +]#Ms䳚4VLWr<us3գ&·1S̤^D1ٕ.ܝ~ۓ6ksJi#[FymNL>v6;ϊ2Z75DPbw ;.gPTO09$+hwW:u:: }͇ngZ;EcҴk-OSS(t8ܸhL(ڥ zmDm7 Lm=1oE,}ƸY01.z%e +»*&=͌9N02a1 ScKף<=+`r>mTۚG1+FObf #X.9RF:sB':<11 p)rwl7ؐUDXA0Re1xxxOºJҹxF?Ēd"&]kL.sඩat(UPek OhNbrwlm͒쪇U hˆsT7gÿ́ʔUŷڤiIxƑJh4}y]ʷKOԥtN$91(endstream +endobj +1987 0 obj +1857 +endobj +1985 0 obj << +/Type /Page +/Contents 1986 0 R +/Resources 1984 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 1903 0 R +>> endobj +1982 0 obj << +/D [1985 0 R /XYZ 71.731 705.352 null] +>> endobj +1983 0 obj << +/D [1985 0 R /XYZ 71.731 717.307 null] +>> endobj +1952 0 obj << +/D [1985 0 R /XYZ 71.731 657.532 null] +>> endobj +1958 0 obj << +/D [1985 0 R /XYZ 71.731 614.528 null] +>> endobj +817 0 obj << +/D [1985 0 R /XYZ 71.731 578.663 null] +>> endobj +201 0 obj << +/D [1985 0 R /XYZ 225.19 551.41 null] +>> endobj +1959 0 obj << +/D [1985 0 R /XYZ 71.731 541.045 null] +>> endobj +1960 0 obj << +/D [1985 0 R /XYZ 137.534 531.286 null] +>> endobj +1961 0 obj << +/D [1985 0 R /XYZ 71.731 524.148 null] +>> endobj +1962 0 obj << +/D [1985 0 R /XYZ 180.97 487.45 null] +>> endobj +1963 0 obj << +/D [1985 0 R /XYZ 71.731 454.409 null] +>> endobj +1964 0 obj << +/D [1985 0 R /XYZ 379.703 443.615 null] +>> endobj +1965 0 obj << +/D [1985 0 R /XYZ 71.731 436.477 null] +>> endobj +1966 0 obj << +/D [1985 0 R /XYZ 71.731 436.477 null] +>> endobj +1967 0 obj << +/D [1985 0 R /XYZ 71.731 419.351 null] +>> endobj +1968 0 obj << +/D [1985 0 R /XYZ 258.678 407.749 null] +>> endobj +1969 0 obj << +/D [1985 0 R /XYZ 119.551 394.798 null] +>> endobj +1970 0 obj << +/D [1985 0 R /XYZ 237.996 381.847 null] +>> endobj +1971 0 obj << +/D [1985 0 R /XYZ 295.33 368.895 null] +>> endobj +1972 0 obj << +/D [1985 0 R /XYZ 164.403 355.944 null] +>> endobj +1973 0 obj << +/D [1985 0 R /XYZ 71.731 348.806 null] +>> endobj +1974 0 obj << +/D [1985 0 R /XYZ 71.731 319.979 null] +>> endobj +1975 0 obj << +/D [1985 0 R /XYZ 349.805 307.127 null] +>> endobj +1976 0 obj << +/D [1985 0 R /XYZ 505.05 307.127 null] +>> endobj +818 0 obj << +/D [1985 0 R /XYZ 71.731 279.132 null] +>> endobj +205 0 obj << +/D [1985 0 R /XYZ 232.362 249.822 null] +>> endobj +1977 0 obj << +/D [1985 0 R /XYZ 71.731 239.457 null] +>> endobj +1978 0 obj << +/D [1985 0 R /XYZ 137.534 229.698 null] +>> endobj +1979 0 obj << +/D [1985 0 R /XYZ 71.731 222.56 null] +>> endobj +1984 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F40 669 0 R /F30 731 0 R /F37 1210 0 R /F39 1242 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2020 0 obj << +/Length 2021 0 R +/Filter /FlateDecode +>> +stream +xXK6z̯6fEzm&h@KܵhHrͯCҢ$٠bD4K(WEqeXH,YE$R?I}?l8X3'ħC{ˋwN #V[I[F3¢6NIEΨ0;T]WwņW׽9Sm<>X# f׽ҊTP~X;ыHD 3*x!cW +Z@Ӝmߋbp?)GqP<ࡍ۩^Z%סc-oDo#2q ^ nRG{YrUKl%BZ'R]ʹ[&njO yws5Lv|GIzé3~RauUV1(J8 J;غq$VV>8_XR[0{5W!V,aZTCUDÛc- ^Qx!K{d^ľYp}S ftcQ) 4+QU( QbfgQ+E}HffX$U})jm*5(Ww8hm( mIn@+9IX er3cD!KɥNq|3BDŽ؟8> Ҩ؜87,jIbUreZ7 =38rV=7W} on1.1ڲ_s%D9Ҵ4MC΋G̎P!>_A|UNgʥ)`=1H@ⵆ\CeWJծɞ5, +M8~vخkӬsAnɺd(&=FE/ļ6MBSO3>`6D>,]J5ΎxMx%w^v$3}[.ԉK7yM&9=E>jZD$Ï׾EB ڧ(-%nendstream +endobj +2021 0 obj +1626 +endobj +2019 0 obj << +/Type /Page +/Contents 2020 0 R +/Resources 2018 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 1903 0 R +>> endobj +2017 0 obj << +/D [2019 0 R /XYZ 71.731 705.352 null] +>> endobj +821 0 obj << +/D [2019 0 R /XYZ 71.731 717.307 null] +>> endobj +1980 0 obj << +/D [2019 0 R /XYZ 71.731 657.532 null] +>> endobj +1981 0 obj << +/D [2019 0 R /XYZ 143.342 634.618 null] +>> endobj +1988 0 obj << +/D [2019 0 R /XYZ 71.731 596.596 null] +>> endobj +1989 0 obj << +/D [2019 0 R /XYZ 71.731 596.596 null] +>> endobj +1990 0 obj << +/D [2019 0 R /XYZ 71.731 566.519 null] +>> endobj +1991 0 obj << +/D [2019 0 R /XYZ 233.283 554.917 null] +>> endobj +819 0 obj << +/D [2019 0 R /XYZ 71.731 511.914 null] +>> endobj +209 0 obj << +/D [2019 0 R /XYZ 291.281 484.661 null] +>> endobj +1992 0 obj << +/D [2019 0 R /XYZ 71.731 474.296 null] +>> endobj +1993 0 obj << +/D [2019 0 R /XYZ 137.534 464.536 null] +>> endobj +1994 0 obj << +/D [2019 0 R /XYZ 71.731 457.398 null] +>> endobj +1995 0 obj << +/D [2019 0 R /XYZ 143.342 446.604 null] +>> endobj +1996 0 obj << +/D [2019 0 R /XYZ 119.551 433.652 null] +>> endobj +1997 0 obj << +/D [2019 0 R /XYZ 241.414 433.652 null] +>> endobj +1998 0 obj << +/D [2019 0 R /XYZ 71.731 413.563 null] +>> endobj +1999 0 obj << +/D [2019 0 R /XYZ 71.731 395.63 null] +>> endobj +2000 0 obj << +/D [2019 0 R /XYZ 71.731 395.63 null] +>> endobj +2001 0 obj << +/D [2019 0 R /XYZ 71.731 365.553 null] +>> endobj +2002 0 obj << +/D [2019 0 R /XYZ 233.283 353.951 null] +>> endobj +820 0 obj << +/D [2019 0 R /XYZ 71.731 310.948 null] +>> endobj +213 0 obj << +/D [2019 0 R /XYZ 227.987 283.695 null] +>> endobj +2003 0 obj << +/D [2019 0 R /XYZ 71.731 273.33 null] +>> endobj +2004 0 obj << +/D [2019 0 R /XYZ 137.534 263.571 null] +>> endobj +2005 0 obj << +/D [2019 0 R /XYZ 71.731 256.433 null] +>> endobj +2006 0 obj << +/D [2019 0 R /XYZ 139.048 245.638 null] +>> endobj +2007 0 obj << +/D [2019 0 R /XYZ 468.578 245.638 null] +>> endobj +2008 0 obj << +/D [2019 0 R /XYZ 454.032 232.687 null] +>> endobj +2009 0 obj << +/D [2019 0 R /XYZ 489.539 232.687 null] +>> endobj +2010 0 obj << +/D [2019 0 R /XYZ 71.731 199.646 null] +>> endobj +2011 0 obj << +/D [2019 0 R /XYZ 397.684 188.851 null] +>> endobj +2012 0 obj << +/D [2019 0 R /XYZ 71.731 168.762 null] +>> endobj +2013 0 obj << +/D [2019 0 R /XYZ 71.731 168.762 null] +>> endobj +2018 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F30 731 0 R /F39 1242 0 R /F37 1210 0 R /F40 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2044 0 obj << +/Length 2045 0 R +/Filter /FlateDecode +>> +stream +xڝX[61bUꌴb\-Ih*PgA`Nm&Q9߹}ob.A7迧 A#?6"I޿-OQaX/ھ)i+Y *iBzᾕb -9-_t(N'kPU^7yX }m;^^M"%q򆡼/ S8QѸc_Z(y_dYcoak`"AoR5 =9x:”|"ehk>0ְvdY]>Z_j6Ry֛:}L^H,;&Xc^`ª%X#./ő܃=[W]ӈ(, v$ƜaN#G3L]YԑE1XSA a29LNS?#S}^ɠ&.^&'ik2ɚA4i]l;9T)Mc͆r~WY%-<=iQ@X>v鸹o&Ib^% -T7mJfMi +YiK!A8Cu ߪ:TN<)+f߮ƈ)44Z8-#]!\a px]sNflx qܐaL[f3X /W +ħEDY%Tˑ2S`ݢuЁ0ILѮk"DJ* 'si Cok''W/*~|?<2(ӵa>Lxn-H9:g^뾰N/DKybKc%YGt +ʎ?*/:ڶ0M9 +zХKŢ%?^AOBv8~j%9+\2)\gd;dcc[䥐kL!a77gف0p{4xUɛ 9*3X s5 ssxe߱kSJ)b; (;ϓXnfgs TED0^o.O{Aȇ8)}7b3վ*T3<]ӬYcsJgA29\hc4tVϧBVU8k7ZmrvZ2(4~6ڂ35 &p SÇ̷ AZ; +Xzx6ɪׇO<I݇=^HX,+k+8MC|9<9!R1bRwFZ~&E u38 +^AD ZG/=W5,P-&"f +@[B5NtQ_,DmPF n/Rendstream +endobj +2045 0 obj +1754 +endobj +2043 0 obj << +/Type /Page +/Contents 2044 0 R +/Resources 2042 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 1903 0 R +/Annots [ 2046 0 R 2047 0 R ] +>> endobj +2046 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [464.642 297.933 503.656 306.844] +/Subtype /Link +/A << /S /GoTo /D (ch05) >> +>> endobj +2047 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [385.839 280.717 440.354 288.911] +/Subtype /Link +/A << /S /GoTo /D (ipv6addresses) >> +>> endobj +2041 0 obj << +/D [2043 0 R /XYZ 71.731 705.352 null] +>> endobj +2014 0 obj << +/D [2043 0 R /XYZ 71.731 657.532 null] +>> endobj +2015 0 obj << +/D [2043 0 R /XYZ 233.283 647.569 null] +>> endobj +2016 0 obj << +/D [2043 0 R /XYZ 119.551 634.618 null] +>> endobj +217 0 obj << +/D [2043 0 R /XYZ 249.449 590.264 null] +>> endobj +2022 0 obj << +/D [2043 0 R /XYZ 71.731 579.899 null] +>> endobj +2023 0 obj << +/D [2043 0 R /XYZ 159.302 570.14 null] +>> endobj +2024 0 obj << +/D [2043 0 R /XYZ 314.667 570.14 null] +>> endobj +2025 0 obj << +/D [2043 0 R /XYZ 71.731 550.05 null] +>> endobj +2026 0 obj << +/D [2043 0 R /XYZ 431.656 539.256 null] +>> endobj +822 0 obj << +/D [2043 0 R /XYZ 71.731 501.298 null] +>> endobj +221 0 obj << +/D [2043 0 R /XYZ 293.066 466.106 null] +>> endobj +2027 0 obj << +/D [2043 0 R /XYZ 71.731 453.935 null] +>> endobj +2028 0 obj << +/D [2043 0 R /XYZ 71.731 453.935 null] +>> endobj +2029 0 obj << +/D [2043 0 R /XYZ 71.731 424.458 null] +>> endobj +2030 0 obj << +/D [2043 0 R /XYZ 206.186 413.663 null] +>> endobj +2031 0 obj << +/D [2043 0 R /XYZ 71.731 354.72 null] +>> endobj +2032 0 obj << +/D [2043 0 R /XYZ 224.427 343.925 null] +>> endobj +2033 0 obj << +/D [2043 0 R /XYZ 447.039 343.925 null] +>> endobj +2034 0 obj << +/D [2043 0 R /XYZ 345.58 330.974 null] +>> endobj +2035 0 obj << +/D [2043 0 R /XYZ 71.731 323.836 null] +>> endobj +2036 0 obj << +/D [2043 0 R /XYZ 71.731 323.836 null] +>> endobj +2037 0 obj << +/D [2043 0 R /XYZ 71.731 292.951 null] +>> endobj +823 0 obj << +/D [2043 0 R /XYZ 71.731 270.755 null] +>> endobj +225 0 obj << +/D [2043 0 R /XYZ 382.996 242.785 null] +>> endobj +2038 0 obj << +/D [2043 0 R /XYZ 71.731 232.42 null] +>> endobj +2039 0 obj << +/D [2043 0 R /XYZ 71.731 197.589 null] +>> endobj +2040 0 obj << +/D [2043 0 R /XYZ 71.731 166.283 null] +>> endobj +2042 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F37 1210 0 R /F40 669 0 R /F30 731 0 R /F36 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2064 0 obj << +/Length 2065 0 R +/Filter /FlateDecode +>> +stream +xn6B1`2"E.[lCW4큑Z,4CJoIQ۹Eԋ)CXAB|zop!g%ū7׀4 +Ei_R{L>'-U+ uYȄqo>ryJ9"„7m#'!Rq@F8af4Vu^fvUF~Tng po7I¼9 eZ,am ~u[Qjw/e3JҺălRЪIȄ^**HI)AAٜ2BX}r'zv6+ +o$.RVK".S|ZUևZ>gп]֏;QbYw>ՕZ]6Gg^Zkƙڅ?zV@ҺNYwTw++d/<rq@܈etGfa%.挒|yw~6@D$FF+" $f8ds1a7R ReI 4IdeE8t*)G$)O+'r1S iܗ9/TW}n0h xUkc<60EEתfS-D3Azq<tS? p͝iq@zkPsu ٭K;J{H k z + FGo9f ڽ~p +N + mU8;AwZکMǁA9o'dŗ4d#aO L@S# ?漽BD ÊWQlbꂪkA&3UI6 4 +UK^N +y&8@eeYud Fւ1b$I& |5 &Ňav<5(]oM uopCS`T+=Tޡ|#.WUPmyvh4J:t -uCaTwt3IC;FSXQ2C9y@8O&#(?ͭaJMU|fT(ܜ٤»鼃gg0_=..94)VςV61蘎< L ԉ(  ,n  wH($a'M>7 a Xendstream +endobj +2065 0 obj +1258 +endobj +2063 0 obj << +/Type /Page +/Contents 2064 0 R +/Resources 2062 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2066 0 R +>> endobj +2061 0 obj << +/D [2063 0 R /XYZ 71.731 705.352 null] +>> endobj +824 0 obj << +/D [2063 0 R /XYZ 71.731 657.532 null] +>> endobj +229 0 obj << +/D [2063 0 R /XYZ 359.898 647.066 null] +>> endobj +2048 0 obj << +/D [2063 0 R /XYZ 71.731 636.701 null] +>> endobj +2049 0 obj << +/D [2063 0 R /XYZ 71.731 575.968 null] +>> endobj +825 0 obj << +/D [2063 0 R /XYZ 71.731 544.662 null] +>> endobj +233 0 obj << +/D [2063 0 R /XYZ 221.216 519.346 null] +>> endobj +2050 0 obj << +/D [2063 0 R /XYZ 71.731 513.219 null] +>> endobj +2051 0 obj << +/D [2063 0 R /XYZ 71.731 454.425 null] +>> endobj +2052 0 obj << +/D [2063 0 R /XYZ 71.731 431.511 null] +>> endobj +2053 0 obj << +/D [2063 0 R /XYZ 71.731 387.328 null] +>> endobj +2054 0 obj << +/D [2063 0 R /XYZ 71.731 365.723 null] +>> endobj +2055 0 obj << +/D [2063 0 R /XYZ 71.731 331.139 null] +>> endobj +2056 0 obj << +/D [2063 0 R /XYZ 71.731 309.533 null] +>> endobj +2057 0 obj << +/D [2063 0 R /XYZ 71.731 274.95 null] +>> endobj +2058 0 obj << +/D [2063 0 R /XYZ 145.832 263.407 null] +>> endobj +826 0 obj << +/D [2063 0 R /XYZ 71.731 233.354 null] +>> endobj +237 0 obj << +/D [2063 0 R /XYZ 323.658 210.007 null] +>> endobj +2059 0 obj << +/D [2063 0 R /XYZ 71.731 203.88 null] +>> endobj +2060 0 obj << +/D [2063 0 R /XYZ 71.731 166.007 null] +>> endobj +2062 0 obj << +/Font << /F31 805 0 R /F40 669 0 R /F25 681 0 R /F37 1210 0 R /F42 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2088 0 obj << +/Length 2089 0 R +/Filter /FlateDecode +>> +stream +xڭX[6c~`,$$~\mͤnA6rL`IݦӮgG}dCVAQ(tu8?#+_Ofw"Ux/vϞ%"\ݾ]~$F՛rfȎ,*^0]BCVAT + WPR*1bT&bׄ1 s!0YmP1Qmcb%[F[b!Hp$'%o ED0)o(.%u"oL۱V|VEלrY&Pϲ>Z99EN[Rۣ;bSޓV,W^ꅹwIg`J^M񢯅{9kNK`̺!n}ڱHԅdU {<ֈ 6C1 +ǢRzC;oҎwMY޷{ΊOnGߛùwSgL{&X6 +qr51w|bXxhEwrG+wT֗I +%壕MRUUMk ^K?dL\d:v7S_2XWX&VUgPV6`%a\>+`d1r+=+r/ƄPEٹʿZٰFXjݻ#gc+y:HD>JhBqiޓc2)'m>D#cY2Kc> ]Fm^Zե93>X8N݉L'%}Cni%28'=En p\pb$C I_`ʃD`(c?F +q`'ut7K;d܄!^AGB,Aa{d_dn뷐&^溑u K:'es$c'}n_̮%vb"gx'cK3_ƑИ ,,#9 Nju9T,cu%.!ufΤˤ%6%8ڶ61x3idegc>>=vI8?bo떀ha(9a wr4ի_qCekr+{0 +wO 0h5Ҡ]bZtC{y\Qײ4礗=yi[[jYЯ?\OkjƍZlڜ̃z3hu+8ˬ{8$hayV^G(q +3N*<|0NўG4/v]p(! X͝p9)o}g~;h3'M4p]N6'2.}ϹB^S! +6>f,5l*}߰9Yap~+F.@h+DWvv + Da06/عlR<Ռ؇P, he[YCP>Y4> endobj +2085 0 obj << +/D [2087 0 R /XYZ 71.731 705.352 null] +>> endobj +2067 0 obj << +/D [2087 0 R /XYZ 71.731 625.762 null] +>> endobj +2068 0 obj << +/D [2087 0 R /XYZ 224.138 600.048 null] +>> endobj +827 0 obj << +/D [2087 0 R /XYZ 71.731 573.702 null] +>> endobj +241 0 obj << +/D [2087 0 R /XYZ 437.352 545.732 null] +>> endobj +2069 0 obj << +/D [2087 0 R /XYZ 71.731 535.367 null] +>> endobj +2070 0 obj << +/D [2087 0 R /XYZ 71.731 505.518 null] +>> endobj +2071 0 obj << +/D [2087 0 R /XYZ 160.228 481.772 null] +>> endobj +2072 0 obj << +/D [2087 0 R /XYZ 297.332 468.82 null] +>> endobj +2073 0 obj << +/D [2087 0 R /XYZ 71.731 456.701 null] +>> endobj +828 0 obj << +/D [2087 0 R /XYZ 71.731 414.212 null] +>> endobj +245 0 obj << +/D [2087 0 R /XYZ 452.631 386.21 null] +>> endobj +2074 0 obj << +/D [2087 0 R /XYZ 71.731 375.845 null] +>> endobj +2075 0 obj << +/D [2087 0 R /XYZ 208.218 353.134 null] +>> endobj +2076 0 obj << +/D [2087 0 R /XYZ 290.12 353.134 null] +>> endobj +2077 0 obj << +/D [2087 0 R /XYZ 71.731 345.996 null] +>> endobj +2078 0 obj << +/D [2087 0 R /XYZ 71.731 323.082 null] +>> endobj +829 0 obj << +/D [2087 0 R /XYZ 71.731 280.594 null] +>> endobj +249 0 obj << +/D [2087 0 R /XYZ 495.798 252.592 null] +>> endobj +2079 0 obj << +/D [2087 0 R /XYZ 71.731 242.227 null] +>> endobj +2080 0 obj << +/D [2087 0 R /XYZ 312.136 219.516 null] +>> endobj +2081 0 obj << +/D [2087 0 R /XYZ 71.731 199.427 null] +>> endobj +2082 0 obj << +/D [2087 0 R /XYZ 339.803 188.632 null] +>> endobj +2083 0 obj << +/D [2087 0 R /XYZ 418.348 188.632 null] +>> endobj +2084 0 obj << +/D [2087 0 R /XYZ 71.731 150.61 null] +>> endobj +2086 0 obj << +/Font << /F31 805 0 R /F37 1210 0 R /F25 681 0 R /F40 669 0 R /F30 731 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2102 0 obj << +/Length 2103 0 R +/Filter /FlateDecode +>> +stream +xڝUMs01C0SI'4d>`CjOlv?~}WHv1N'ڬ}ҀCM"sɂy +gys03*XmߧX }!ZdU[be6"wJ駋Bh"@s  6 Q}Ri#D!4WH)o($ +"J(]mR\ժŖ=P12@b%U6 +@HlgU!3:Gdh;ߎ;OIJxY : +x7 $,Kfd qN%@`GyѣGXbu6-+0o 袦e)Zom+~,j*|Vm#idHX/l +g oEd )ȷK[Э~Xb}T=;vIFx>q1Aw/BD)^V +gu:5Vi݇/{_.vkD(b#riLy\GZm1Bj Bl" PnV?͡Kڟ|2l[5a  JEendstream +endobj +2103 0 obj +709 +endobj +2101 0 obj << +/Type /Page +/Contents 2102 0 R +/Resources 2100 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2066 0 R +>> endobj +2099 0 obj << +/D [2101 0 R /XYZ 71.731 705.352 null] +>> endobj +2090 0 obj << +/D [2101 0 R /XYZ 71.731 554.605 null] +>> endobj +2091 0 obj << +/D [2101 0 R /XYZ 414.721 543.062 null] +>> endobj +2092 0 obj << +/D [2101 0 R /XYZ 71.731 530.942 null] +>> endobj +2093 0 obj << +/D [2101 0 R /XYZ 71.731 498.416 null] +>> endobj +2094 0 obj << +/D [2101 0 R /XYZ 136.428 486.872 null] +>> endobj +2095 0 obj << +/D [2101 0 R /XYZ 71.731 475.502 null] +>> endobj +2096 0 obj << +/D [2101 0 R /XYZ 71.731 442.227 null] +>> endobj +2097 0 obj << +/D [2101 0 R /XYZ 71.731 442.227 null] +>> endobj +2098 0 obj << +/D [2101 0 R /XYZ 71.731 418.564 null] +>> endobj +2100 0 obj << +/Font << /F31 805 0 R /F37 1210 0 R /F25 681 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2107 0 obj << +/Length 2108 0 R +/Filter /FlateDecode +>> +stream +xڅ0 E~vImG`5b*~ (`ni4@|"m! }UH/d sݧձ):u}ԗkV]e{vd&CVe.rHAf;/N%Dendstream +endobj +2108 0 obj +166 +endobj +2106 0 obj << +/Type /Page +/Contents 2107 0 R +/Resources 2105 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2066 0 R +>> endobj +2104 0 obj << +/D [2106 0 R /XYZ 71.731 705.352 null] +>> endobj +2105 0 obj << +/Font << /F31 805 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2134 0 obj << +/Length 2135 0 R +/Filter /FlateDecode +>> +stream +xڍXn8WY9@̈Z I tAIW3%:&"HϥxI땉 $q=]GW)%i?AL8\{O+7$!a 7mp& I~U$V~C;j^o4[>|rgGC<7d (Yє4&6qH4 ńNW}1|ki~CL9ɓȬMI6!%qxݲRh!VU/LX +l]kn8o5X䥽> }uΎ~ֽluGT/a;yVk €Yk=DQ*%vf!YsDIkR:Xlxxa(Y:Q1?A>ϘOvy<gN4켠K#y~ucI e!ҸwZ#0R(-B۵A&jT8 a(<pbwg Z4>+}5y;=oO9 +DQ@(ƽ:4[sUc 3l8&3?bEFzw tH 6pL(={z qNJڋ8aJpAQN\q=Czׅ +>hCmEY`h4 +~T8!*I_.6L庸`B6,+GqC;c}s`H֨)z=gB]POrwۥyZdF B6]6~zjz&K`! +%@\JrhZ.]S*72 Sabޙt +~޾ӝJhW/]0Sk 50Ӟ.5G Hg}E/if%ϳ<@+AT[G-!6Z`kр;>ȶϨp[,l(@WX`lvm(Ú]ؔm wh8/'4VzJ`}3^\u3?:q\FԾͻǮ+bMovXOP|Z|:,sGkS{#pL^%P}s.Tȴ1^W]3T{*%m +vd;ג]:YKChiWNޡLI K .T8)P1?exПaEj6D +00s0)R9q<{z0Ndg rp Kvvyh+|.imk` + K/hɔ% -y⡁=N/9.|~3q7.}'Oj?":T#k`KW? 8oendstream +endobj +2135 0 obj +1655 +endobj +2133 0 obj << +/Type /Page +/Contents 2134 0 R +/Resources 2132 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2066 0 R +>> endobj +2131 0 obj << +/D [2133 0 R /XYZ 71.731 705.352 null] +>> endobj +830 0 obj << +/D [2133 0 R /XYZ 71.731 657.532 null] +>> endobj +253 0 obj << +/D [2133 0 R /XYZ 503.343 642.462 null] +>> endobj +831 0 obj << +/D [2133 0 R /XYZ 71.731 621.08 null] +>> endobj +257 0 obj << +/D [2133 0 R /XYZ 378.04 590.385 null] +>> endobj +2109 0 obj << +/D [2133 0 R /XYZ 71.731 577.947 null] +>> endobj +2110 0 obj << +/D [2133 0 R /XYZ 71.731 548.736 null] +>> endobj +2111 0 obj << +/D [2133 0 R /XYZ 71.731 504.901 null] +>> endobj +2112 0 obj << +/D [2133 0 R /XYZ 153.304 494.106 null] +>> endobj +832 0 obj << +/D [2133 0 R /XYZ 71.731 438.152 null] +>> endobj +261 0 obj << +/D [2133 0 R /XYZ 339.599 405.017 null] +>> endobj +2113 0 obj << +/D [2133 0 R /XYZ 71.731 392.579 null] +>> endobj +2114 0 obj << +/D [2133 0 R /XYZ 450.386 383.458 null] +>> endobj +2115 0 obj << +/D [2133 0 R /XYZ 71.731 376.32 null] +>> endobj +2116 0 obj << +/D [2133 0 R /XYZ 424.704 352.574 null] +>> endobj +2117 0 obj << +/D [2133 0 R /XYZ 119.551 339.622 null] +>> endobj +2118 0 obj << +/D [2133 0 R /XYZ 152.756 326.671 null] +>> endobj +2119 0 obj << +/D [2133 0 R /XYZ 71.731 320.282 null] +>> endobj +2120 0 obj << +/D [2133 0 R /XYZ 119.551 295.787 null] +>> endobj +2121 0 obj << +/D [2133 0 R /XYZ 71.731 289.366 null] +>> endobj +2122 0 obj << +/D [2133 0 R /XYZ 137.444 277.854 null] +>> endobj +2123 0 obj << +/D [2133 0 R /XYZ 119.551 239 null] +>> endobj +2124 0 obj << +/D [2133 0 R /XYZ 203.186 239 null] +>> endobj +2125 0 obj << +/D [2133 0 R /XYZ 71.731 218.91 null] +>> endobj +2126 0 obj << +/D [2133 0 R /XYZ 137.534 208.116 null] +>> endobj +2127 0 obj << +/D [2133 0 R /XYZ 324.401 208.116 null] +>> endobj +2128 0 obj << +/D [2133 0 R /XYZ 119.551 195.165 null] +>> endobj +2129 0 obj << +/D [2133 0 R /XYZ 192.327 182.213 null] +>> endobj +2130 0 obj << +/D [2133 0 R /XYZ 268.241 182.213 null] +>> endobj +2132 0 obj << +/Font << /F29 669 0 R /F36 669 0 R /F25 681 0 R /F30 731 0 R /F37 1210 0 R /F31 805 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2139 0 obj << +/Length 2140 0 R +/Filter /FlateDecode +>> +stream +xڅK 0{٤Ik}> endobj +2136 0 obj << +/D [2138 0 R /XYZ 71.731 705.352 null] +>> endobj +2137 0 obj << +/Font << /F31 805 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2158 0 obj << +/Length 2159 0 R +/Filter /FlateDecode +>> +stream +xڥXmo619C̈HI٧]n@7 -QY$9nw$e%"%xs/4yGg%QtuIgiF;: g-p$N7w~2=Y%$F$g ?"sd t,ި>M-&a2oJ_>a1[ t +5,@U;>rek'LJ:չ6@[P$vMsMZfXF⢔-MVC^7O8S]o%ӈB -c`"WC)WRj@hۢv\z؃i!SC彋)/@T;A#H:/JkM=hݦqʎG#.JSYcu@edqB%F;1D)I`]7Xl{Sh)#41< W,|BJ,䄒cNҎ~l'JK;`]6Ă_2ʝaܲi{z; Ӎkj/Di(g^x$HY#gW=%ŏ!v+,&4!1 Z0fH)4q{:#<# `nZ|b_jԍXXZ- nIAlA׵3&O@ +AO95 ?#g]9<9f$j5l_p$_TvmQ:-BOht:d8d><,SCZN t_> ?GE<8- 3bX O`Pb$ I6dJUх4gSuRٶP*wȐ8y~`3]*o`4e›)BЙΫɶ#7GŽ aw`{dVc2P;Ķfr cPOc*,;:8D&<'dZ&E~ +S~4N;ws.wM|T2im1V-Wklb|D.s{qңISt^ &Ѕط!x-7Ӑ>k 4W+-c'. B+ +a9a]7kÞQ5v}yY-%el&h26_%\;|ˇ~̦97F g,`ؘԣg/QRz(eRQA֥ZiUdQoMӛ:b˖Xhk׺uZ5pKVӁIߛԋ iJ%eWX"E]fn+0vqu;F;{O.{ v=D,GM?,d7 ]Czu-'C <\_ ,Yi=6rq[)j}mQe. ݨ9NuuU+{nW}o*MWgʦnN[!GVCBܹD&mKw=:>:{&4׺˸JO`!⋶ڍ w n-oǧNmYfrZ/az"(ٟe9Kl! \̄u.9z&{Bendstream +endobj +2159 0 obj +1767 +endobj +2157 0 obj << +/Type /Page +/Contents 2158 0 R +/Resources 2156 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2161 0 R +/Annots [ 2160 0 R ] +>> endobj +2160 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [376.746 400.891 446.683 413.5] +/Subtype /Link +/A << /S /GoTo /D (address_match_lists) >> +>> endobj +2155 0 obj << +/D [2157 0 R /XYZ 71.731 705.352 null] +>> endobj +833 0 obj << +/D [2157 0 R /XYZ 71.731 657.532 null] +>> endobj +265 0 obj << +/D [2157 0 R /XYZ 492.787 642.462 null] +>> endobj +2141 0 obj << +/D [2157 0 R /XYZ 71.731 621.08 null] +>> endobj +2142 0 obj << +/D [2157 0 R /XYZ 71.731 621.08 null] +>> endobj +2143 0 obj << +/D [2157 0 R /XYZ 291.952 612.725 null] +>> endobj +2144 0 obj << +/D [2157 0 R /XYZ 237.179 599.774 null] +>> endobj +2145 0 obj << +/D [2157 0 R /XYZ 497.22 599.774 null] +>> endobj +2146 0 obj << +/D [2157 0 R /XYZ 301.646 573.871 null] +>> endobj +2147 0 obj << +/D [2157 0 R /XYZ 71.731 566.733 null] +>> endobj +2148 0 obj << +/D [2157 0 R /XYZ 71.731 566.733 null] +>> endobj +2149 0 obj << +/D [2157 0 R /XYZ 119.551 542.987 null] +>> endobj +834 0 obj << +/D [2157 0 R /XYZ 71.731 532.316 null] +>> endobj +269 0 obj << +/D [2157 0 R /XYZ 333.849 497.733 null] +>> endobj +2150 0 obj << +/D [2157 0 R /XYZ 71.731 485.295 null] +>> endobj +2151 0 obj << +/D [2157 0 R /XYZ 323.663 476.174 null] +>> endobj +2152 0 obj << +/D [2157 0 R /XYZ 71.731 464.054 null] +>> endobj +2153 0 obj << +/D [2157 0 R /XYZ 71.731 459.073 null] +>> endobj +2154 0 obj << +/D [2157 0 R /XYZ 71.731 459.073 null] +>> endobj +2156 0 obj << +/Font << /F29 669 0 R /F25 681 0 R /F37 1210 0 R /F36 669 0 R /F56 1242 0 R /F30 731 0 R /F31 805 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2167 0 obj << +/Length 2168 0 R +/Filter /FlateDecode +>> +stream +xڝXY6_G"E]}KҤH@mMGq~}y irK72bvJ uNI(n'!lE=S;W^Ɛ/C?\vGqnp`rWMy0]sDK\}Ma#iI21}Mi{^4\NxaM}m +H -0 +[; СQ$j7l,uSkӡ}t\k< gb::xEþfh>1QM7{[ 6j3YL)@LVoG'9'B'=R^A(* KCTƉ燑5\W >MpӘC8OV-lg|sq4S٣Qpxs-;/c%ۋ9f+a_v,~ ].<9Ә9k2A&B`Ap4>X&$z~Gs^$_3eRf}:=?J$yq‡ \(9n{&Mj@J:.=ލ + &a1>}^ `kyid=rqA cRVNKYhR qrJHq%F^"4 / _i-zAL2~~Ƞ'OUWendstream +endobj +2168 0 obj +1456 +endobj +2166 0 obj << +/Type /Page +/Contents 2167 0 R +/Resources 2165 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2161 0 R +>> endobj +2164 0 obj << +/D [2166 0 R /XYZ 71.731 705.352 null] +>> endobj +835 0 obj << +/D [2166 0 R /XYZ 71.731 188.891 null] +>> endobj +2165 0 obj << +/Font << /F31 805 0 R /F37 1210 0 R /F25 681 0 R /F39 1242 0 R /F56 1242 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2200 0 obj << +/Length 2201 0 R +/Filter /FlateDecode +>> +stream +xYK6r̯pX1zjC^mqCl&V]Q^CJ%(j8/~\y#I7"AcosO], 2_?OIյZ*>_m]CG<&x}y5Rգ^xϥ'(V>~Ms@F&8KXKM3 <˾AZ'fOuݍ\g% x2U,Ko=KV6?//lK{s nW7 ! #·QtmV8Ud2 p뜢dHc9N|a|\i(ѼTIf`WmV-AZP8[VYk?{TQtyPX:[[hn׽\VˣA6c$@,i´xs+Ny?v5#C aPP 1 ?,pVMԼma6g΀- EnWFRɿP02ݴ%͊wymMEh#δ9u^(.oy*eUpdlg諶Ap1𬂊f;kaigH$,NH$Dp/4fCDaj!ag|KaI 3Qzs ҂ns A=0Lm ѮOp|FLrɊkZV c|MAD /2ph\!R'_ijK+igWr~'N6MKW-~MCY|/Cm aZ3񊪟b`s*bۀ?0nlsJ ѐׅi:xjF^*=$_ A,(VaCFaW.mwlϻb܊n1x`Bm KLvkDNs:F,-1yYjS6]5#9i%3+4iXtkb 9SCමp+mt4le6*-&Uએ^Hk!1ϊ\R|,/$r,M508ɕ_GbWG9(wNXevTXL2o6$H"MKې1PP ?-+tSscFynm*Y[˞z,d$NŮ}y9AbPE N3I18z0P7}<+;U% P*q;ܣ3|{Sl$nܳ>¢k@Ǻ9R+6PF;YjUsM^Lt_گbBuen9m۩]|{T7Gp2y)Gm* N_$ sF/w' ?Xendstream +endobj +2201 0 obj +1940 +endobj +2199 0 obj << +/Type /Page +/Contents 2200 0 R +/Resources 2198 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2161 0 R +>> endobj +2197 0 obj << +/D [2199 0 R /XYZ 71.731 705.352 null] +>> endobj +273 0 obj << +/D [2199 0 R /XYZ 251.629 647.066 null] +>> endobj +836 0 obj << +/D [2199 0 R /XYZ 71.731 636.925 null] +>> endobj +277 0 obj << +/D [2199 0 R /XYZ 201.957 612.596 null] +>> endobj +2162 0 obj << +/D [2199 0 R /XYZ 71.731 600.158 null] +>> endobj +2163 0 obj << +/D [2199 0 R /XYZ 71.731 600.158 null] +>> endobj +2169 0 obj << +/D [2199 0 R /XYZ 130.311 579.321 null] +>> endobj +2170 0 obj << +/D [2199 0 R /XYZ 71.731 578.272 null] +>> endobj +2171 0 obj << +/D [2199 0 R /XYZ 261.578 567.664 null] +>> endobj +2172 0 obj << +/D [2199 0 R /XYZ 358.415 567.664 null] +>> endobj +837 0 obj << +/D [2199 0 R /XYZ 71.731 533.841 null] +>> endobj +281 0 obj << +/D [2199 0 R /XYZ 281.661 510.579 null] +>> endobj +2173 0 obj << +/D [2199 0 R /XYZ 71.731 501.941 null] +>> endobj +2174 0 obj << +/D [2199 0 R /XYZ 496.801 478.698 null] +>> endobj +2175 0 obj << +/D [2199 0 R /XYZ 71.731 458.609 null] +>> endobj +2176 0 obj << +/D [2199 0 R /XYZ 71.731 453.628 null] +>> endobj +2177 0 obj << +/D [2199 0 R /XYZ 129.514 437.852 null] +>> endobj +2178 0 obj << +/D [2199 0 R /XYZ 71.731 436.068 null] +>> endobj +2179 0 obj << +/D [2199 0 R /XYZ 129.514 419.919 null] +>> endobj +2180 0 obj << +/D [2199 0 R /XYZ 71.731 417.762 null] +>> endobj +2181 0 obj << +/D [2199 0 R /XYZ 129.514 401.986 null] +>> endobj +2182 0 obj << +/D [2199 0 R /XYZ 71.731 399.829 null] +>> endobj +2183 0 obj << +/D [2199 0 R /XYZ 129.514 384.054 null] +>> endobj +2184 0 obj << +/D [2199 0 R /XYZ 375.807 384.054 null] +>> endobj +2185 0 obj << +/D [2199 0 R /XYZ 71.731 381.897 null] +>> endobj +2186 0 obj << +/D [2199 0 R /XYZ 129.514 366.121 null] +>> endobj +2187 0 obj << +/D [2199 0 R /XYZ 71.731 361.04 null] +>> endobj +2188 0 obj << +/D [2199 0 R /XYZ 71.731 315.147 null] +>> endobj +2189 0 obj << +/D [2199 0 R /XYZ 71.731 271.312 null] +>> endobj +2190 0 obj << +/D [2199 0 R /XYZ 71.731 227.476 null] +>> endobj +2191 0 obj << +/D [2199 0 R /XYZ 350.282 203.73 null] +>> endobj +2192 0 obj << +/D [2199 0 R /XYZ 406.082 203.73 null] +>> endobj +2193 0 obj << +/D [2199 0 R /XYZ 461.882 203.73 null] +>> endobj +2194 0 obj << +/D [2199 0 R /XYZ 119.551 190.779 null] +>> endobj +2195 0 obj << +/D [2199 0 R /XYZ 194.012 190.779 null] +>> endobj +2196 0 obj << +/D [2199 0 R /XYZ 71.731 170.69 null] +>> endobj +2198 0 obj << +/Font << /F31 805 0 R /F40 669 0 R /F42 669 0 R /F52 1210 0 R /F37 1210 0 R /F25 681 0 R /F30 731 0 R /F45 681 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2225 0 obj << +/Length 2226 0 R +/Filter /FlateDecode +>> +stream +xڭWK6E/vvE-N$rh@-Ӗhtv;Ԑzy|3~tQEp 6+t;7[16Hg6"4m6뒟hKyJ|{VxU}w87 ιT=pF\ԅXt3f) Y> :^38iNh$3%&?WRax*m}z.Dpƫܘ{kJ& >Ԫja/܈eL$k!Qs(ҭ`-,&zUFVQԾz' +~+ֻFM+h$ +?syPMecgȫ.0ډZ> ^ZLZavQե:KJutb/weΛEI5f\UO {<}xǓCߠauF%"/xFo}X9#qJ]w{[]0#rq!h8RpjXK~mg0R >RcsAA ]ٷ>\G*cWAV1߽X|M~jLqs[ _M$`ViA!KFxJ|*86U1QCI01jL8 nhښ+8* +Ԇ\Gщ`!pmElKWFXORKzR3CkN7#(6n4y>qG~SydKλfCv(UҡN'(3T;E?" +[?àCQ}x0AѥryҦ9Fa3JC(&q6JK"f9DXXF,@SZP?9^Kuw7K$+ԱI:FV?G}ǿj٭mrn"HFdf}#ru Es E`mm@r=귛2i8k 6AIw䶻V:"-E5dQǥC$Ԓ wt Oz6{ ùbZ)~n7$oW0Y+(=1Tvct+w;GW v=T')zv}k}%,h-nŭ+ȎB +jsY԰zktGX;HJkÒ6_-L CN[b00IPԦ [( r;jjӥ5qw,ry86~ıZ)U}y}NgnZ`wnWqGy^}Mڕ1d%<&endstream +endobj +2226 0 obj +1358 +endobj +2224 0 obj << +/Type /Page +/Contents 2225 0 R +/Resources 2223 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2161 0 R +>> endobj +2221 0 obj << +/D [2224 0 R /XYZ 71.731 705.352 null] +>> endobj +2222 0 obj << +/D [2224 0 R /XYZ 71.731 717.307 null] +>> endobj +2202 0 obj << +/D [2224 0 R /XYZ 71.731 629.537 null] +>> endobj +2203 0 obj << +/D [2224 0 R /XYZ 119.551 590.782 null] +>> endobj +2204 0 obj << +/D [2224 0 R /XYZ 321.562 577.831 null] +>> endobj +838 0 obj << +/D [2224 0 R /XYZ 71.731 537.816 null] +>> endobj +285 0 obj << +/D [2224 0 R /XYZ 228.79 510.564 null] +>> endobj +2205 0 obj << +/D [2224 0 R /XYZ 71.731 500.421 null] +>> endobj +2206 0 obj << +/D [2224 0 R /XYZ 137.534 490.439 null] +>> endobj +2207 0 obj << +/D [2224 0 R /XYZ 119.551 477.488 null] +>> endobj +839 0 obj << +/D [2224 0 R /XYZ 71.731 452.417 null] +>> endobj +289 0 obj << +/D [2224 0 R /XYZ 201.957 430.066 null] +>> endobj +2208 0 obj << +/D [2224 0 R /XYZ 71.731 421.613 null] +>> endobj +2209 0 obj << +/D [2224 0 R /XYZ 71.731 411.651 null] +>> endobj +2210 0 obj << +/D [2224 0 R /XYZ 189.489 402.469 null] +>> endobj +2211 0 obj << +/D [2224 0 R /XYZ 71.731 391.798 null] +>> endobj +2212 0 obj << +/D [2224 0 R /XYZ 189.489 380.85 null] +>> endobj +2213 0 obj << +/D [2224 0 R /XYZ 71.731 370.18 null] +>> endobj +2214 0 obj << +/D [2224 0 R /XYZ 184.109 359.232 null] +>> endobj +840 0 obj << +/D [2224 0 R /XYZ 71.731 321.374 null] +>> endobj +293 0 obj << +/D [2224 0 R /XYZ 281.661 295.869 null] +>> endobj +2215 0 obj << +/D [2224 0 R /XYZ 71.731 287.232 null] +>> endobj +2216 0 obj << +/D [2224 0 R /XYZ 385.002 276.94 null] +>> endobj +2217 0 obj << +/D [2224 0 R /XYZ 71.731 269.802 null] +>> endobj +2218 0 obj << +/D [2224 0 R /XYZ 71.731 225.967 null] +>> endobj +2219 0 obj << +/D [2224 0 R /XYZ 71.731 197.14 null] +>> endobj +2220 0 obj << +/D [2224 0 R /XYZ 71.731 187.177 null] +>> endobj +2223 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F30 731 0 R /F40 669 0 R /F42 669 0 R /F37 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2246 0 obj << +/Length 2247 0 R +/Filter /FlateDecode +>> +stream +xڵWݏ6x䯈Cbl'<-*XmKDwlݻRUc{E~,*)RIEc(P?Þ%(%e_m?{4^9u#7 4ΉA? c9HEt6nzv19ņ N2V,"#O3v[Q}dLr̪95;ll(kpE=PAM'MsWZ6_J YympKdܚ 0N*o]MB[`g"+Ra'v~)q0Ƴs7 OKü|jhU2FQ֚O^jO-=bG#dg3yj68PӐlǍ1Y/H7"Y`s /{%<'WB[[dIœX%FŸ%ο`,#Xk dY]|W^ uOwWO<9r?O0ho]}IFIyG>4뎫M^xvb3ruj^iF1RUi4jaI%.3X1Vl + +[,?om2Ϩ[}(Cf' woP04b2:=I zL^|`%[8(d]!%J,3~jsVۡ]ُ:xW{i 2)6_@ +=699疃ӓZaA)lJЌȚoAI&M]pZ^B\DpaN - ™8FKN8h};oq}/AGnr X;5iC n9k>ճM!Ň/:V<5wLSKtTJ4Sc ){?NCdG Id˃$n6<egDRcj٭ǽ=muÆ wf8f4>+9),JY@IZ_ +R /kGB#"ΫgةGZURC('ݨ FS,18.B>=7}n \Chendstream +endobj +2247 0 obj +1330 +endobj +2245 0 obj << +/Type /Page +/Contents 2246 0 R +/Resources 2244 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2161 0 R +>> endobj +2243 0 obj << +/D [2245 0 R /XYZ 71.731 704.356 null] +>> endobj +2227 0 obj << +/D [2245 0 R /XYZ 71.731 596.596 null] +>> endobj +2228 0 obj << +/D [2245 0 R /XYZ 71.731 578.663 null] +>> endobj +2229 0 obj << +/D [2245 0 R /XYZ 71.731 568.7 null] +>> endobj +2230 0 obj << +/D [2245 0 R /XYZ 71.731 507.993 null] +>> endobj +2231 0 obj << +/D [2245 0 R /XYZ 413.932 495.042 null] +>> endobj +2232 0 obj << +/D [2245 0 R /XYZ 71.731 474.952 null] +>> endobj +2233 0 obj << +/D [2245 0 R /XYZ 71.731 457.02 null] +>> endobj +2234 0 obj << +/D [2245 0 R /XYZ 71.731 447.057 null] +>> endobj +2235 0 obj << +/D [2245 0 R /XYZ 71.731 382.664 null] +>> endobj +2236 0 obj << +/D [2245 0 R /XYZ 194.271 354.062 null] +>> endobj +841 0 obj << +/D [2245 0 R /XYZ 71.731 272.369 null] +>> endobj +297 0 obj << +/D [2245 0 R /XYZ 333.849 242.162 null] +>> endobj +2237 0 obj << +/D [2245 0 R /XYZ 71.731 229.724 null] +>> endobj +2238 0 obj << +/D [2245 0 R /XYZ 129.235 220.603 null] +>> endobj +2239 0 obj << +/D [2245 0 R /XYZ 71.731 188.28 null] +>> endobj +2240 0 obj << +/D [2245 0 R /XYZ 71.731 164.648 null] +>> endobj +2241 0 obj << +/D [2245 0 R /XYZ 71.731 159.667 null] +>> endobj +2242 0 obj << +/D [2245 0 R /XYZ 71.731 159.667 null] +>> endobj +2244 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F37 1210 0 R /F40 669 0 R /F58 2248 0 R /F36 669 0 R /F30 731 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2265 0 obj << +/Length 2266 0 R +/Filter /FlateDecode +>> +stream +xXo6Bo%)ڧ6il]0,6LBe1}Grw?ǻw3 0H 18\n'xDA)Hr0DfPUS֫8qQS4CYN/#C>aeu6k4f&ۥf8`x'D+8ڦATDfPX˶:^.p$G.'8tvvϵ-`(M% f9SY ͼ0: bU;hQc. mwR]щ%%ܶnhkOwmWw1!8>)T^(vJubټu$]4bإ`^ʭJ<= X5i8bKUWo0GA8b@3%kfP-ڗ +H(d0y4}fy v^0y}z9uU0 QBiS]%fn6s__*7z5krzC )b "(rvڌߖ'-d%t}R0&G Jy+;Ϭ7;h4 }BPc>9Wu -'>xݭt&Tv29^ȽxgIUTiu-Gbٱc3 Syqݪj^dB3;0QΆG fWÄ"Ůh4JA8u#FB-7aSnF7?}ZE-8rl_aӄO/`./MDkzo5.6;>@d *}9:%ӿ>_7]endstream +endobj +2266 0 obj +1352 +endobj +2264 0 obj << +/Type /Page +/Contents 2265 0 R +/Resources 2263 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2161 0 R +>> endobj +2262 0 obj << +/D [2264 0 R /XYZ 71.731 704.356 null] +>> endobj +2249 0 obj << +/D [2264 0 R /XYZ 71.731 454.892 null] +>> endobj +2250 0 obj << +/D [2264 0 R /XYZ 137.534 441.941 null] +>> endobj +2251 0 obj << +/D [2264 0 R /XYZ 187.904 441.941 null] +>> endobj +842 0 obj << +/D [2264 0 R /XYZ 71.731 429.822 null] +>> endobj +301 0 obj << +/D [2264 0 R /XYZ 272.373 402.569 null] +>> endobj +2252 0 obj << +/D [2264 0 R /XYZ 71.731 392.427 null] +>> endobj +2253 0 obj << +/D [2264 0 R /XYZ 71.731 392.427 null] +>> endobj +843 0 obj << +/D [2264 0 R /XYZ 71.731 336.663 null] +>> endobj +305 0 obj << +/D [2264 0 R /XYZ 351.276 308.303 null] +>> endobj +2254 0 obj << +/D [2264 0 R /XYZ 71.731 297.938 null] +>> endobj +2255 0 obj << +/D [2264 0 R /XYZ 137.534 288.178 null] +>> endobj +2256 0 obj << +/D [2264 0 R /XYZ 71.731 268.462 null] +>> endobj +2257 0 obj << +/D [2264 0 R /XYZ 361.122 257.294 null] +>> endobj +2258 0 obj << +/D [2264 0 R /XYZ 71.731 239.262 null] +>> endobj +2259 0 obj << +/D [2264 0 R /XYZ 71.731 214.291 null] +>> endobj +2260 0 obj << +/D [2264 0 R /XYZ 71.731 209.309 null] +>> endobj +2261 0 obj << +/D [2264 0 R /XYZ 71.731 209.309 null] +>> endobj +2263 0 obj << +/Font << /F31 805 0 R /F30 731 0 R /F25 681 0 R /F40 669 0 R /F56 1242 0 R /F37 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2317 0 obj << +/Length 2318 0 R +/Filter /FlateDecode +>> +stream +xڝY[۶_Śҧ6i=Ay)DȒ!.z;/:-XI6578G F K$ʏ/qd>E̮&Y!EȓVvGx^uS?[,6RMm?}&M6۟_pT@1I]aM)b4ڥE1YZ탼S, R VK&UtgG‡%)F1nz>N;ot׷}fpDz lŰd7Α2?[}P{::r4vF=0Aǽje a2W{rp2C)gEL +%p+]20MN=(qDp{v#L2NŊ${"0O[ݖA]4Z7.mU+|:5х8 bt͚fO$8}nPJ <$;oNYY:Z [ey|ZP:cݤuT>V2wug03 e[85>4Exne]VQO|$)=}wE'ȺA]1D0ݱ. D ')hRvK`荒|Pr>QV2Z +;#^n/ȱِ/t>@ȵd>V/M&yokitl']P/3f$%'$nyb$B[lw)o2D)yî͛7E]N$vҔ$TW+28݄ 00iٗh^הK`[WAֵ)Vܹ=wU”}[s$^zœ! 4'ՎLJQulp&GG}g&.O2^S\"9*ԣGȓ$j"M)п*ݖ8QAQ޼tJ Ƈh6fa_H1jz%!\F_o yr6љ}ht7 2(L7'/J3&L>XP*sWx8xN{n\3 +M&113b_u3lJI-v_j'>|;JRO^ +$ ''H[_e?<0 g~`( h}+N&fsuҞ)a.Yd3RfqNw[)J_v#ed4$_&zR= oM퉿Q~t{, ]xr  +@aWB9)EOÞ-[MyҢʮ;Yimb{4,O=xW0O,Ʃzd%*-0ik~=tƮfd%=xD2H{{]6 +D,P5?;I( d̏e=1PMxn K\ct4\3Cx -U͌B5,HX@p-[{W w28񣐭ڃae$9l 9O` +Ց &Fa m_*8l,JFߗ(҇|GsuKE; O2ocRQޗLrQ BQ5ҝW+Nr Ӝ[bTeLk'i)47j#{˧ձ^~c{#:R&OLgi +i2St2l]T\ag³pP&d0ϷQQ6:`ōMt3i?WIymj5Rw1V~/ GM dɄ:Ϝ7+ ?i|JR)`/Lzt8#kf58jO.p 0$d +5sL<Ti|N7)%0IOxHΡ7ͻhZǿ?q*b%]|]ӵ?ZY0>+s+ބݬqP Iizu> endobj +2319 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [306.976 206.918 480.194 215.83] +/Subtype /Link +/A << /S /GoTo /D (rndc) >> +>> endobj +2320 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [492.926 206.918 537.981 215.83] +/Subtype /Link +/A << /S /GoTo /D (admin_tools) >> +>> endobj +2321 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [71.731 193.967 146.948 202.878] +/Subtype /Link +/A << /S /GoTo /D (admin_tools) >> +>> endobj +2313 0 obj << +/D [2316 0 R /XYZ 71.731 705.352 null] +>> endobj +2314 0 obj << +/D [2316 0 R /XYZ 71.731 717.307 null] +>> endobj +2267 0 obj << +/D [2316 0 R /XYZ 71.731 608.317 null] +>> endobj +2268 0 obj << +/D [2316 0 R /XYZ 137.534 595.365 null] +>> endobj +2269 0 obj << +/D [2316 0 R /XYZ 194.539 595.365 null] +>> endobj +2270 0 obj << +/D [2316 0 R /XYZ 412.2 595.365 null] +>> endobj +2271 0 obj << +/D [2316 0 R /XYZ 226.638 582.414 null] +>> endobj +844 0 obj << +/D [2316 0 R /XYZ 71.731 552.362 null] +>> endobj +309 0 obj << +/D [2316 0 R /XYZ 309.027 525.109 null] +>> endobj +2272 0 obj << +/D [2316 0 R /XYZ 71.731 514.967 null] +>> endobj +2273 0 obj << +/D [2316 0 R /XYZ 71.731 514.967 null] +>> endobj +2274 0 obj << +/D [2316 0 R /XYZ 248.666 491.834 null] +>> endobj +2275 0 obj << +/D [2316 0 R /XYZ 383.161 491.834 null] +>> endobj +2276 0 obj << +/D [2316 0 R /XYZ 243.287 480.178 null] +>> endobj +2277 0 obj << +/D [2316 0 R /XYZ 135.691 468.521 null] +>> endobj +845 0 obj << +/D [2316 0 R /XYZ 71.731 435.891 null] +>> endobj +313 0 obj << +/D [2316 0 R /XYZ 387.93 407.53 null] +>> endobj +2278 0 obj << +/D [2316 0 R /XYZ 71.731 397.165 null] +>> endobj +2279 0 obj << +/D [2316 0 R /XYZ 137.534 387.406 null] +>> endobj +2280 0 obj << +/D [2316 0 R /XYZ 414.053 374.454 null] +>> endobj +2281 0 obj << +/D [2316 0 R /XYZ 71.731 356.422 null] +>> endobj +2282 0 obj << +/D [2316 0 R /XYZ 134.216 343.57 null] +>> endobj +2283 0 obj << +/D [2316 0 R /XYZ 380.668 343.57 null] +>> endobj +2284 0 obj << +/D [2316 0 R /XYZ 479.735 343.57 null] +>> endobj +2285 0 obj << +/D [2316 0 R /XYZ 289.432 330.619 null] +>> endobj +2286 0 obj << +/D [2316 0 R /XYZ 336.485 330.619 null] +>> endobj +2287 0 obj << +/D [2316 0 R /XYZ 182.076 304.716 null] +>> endobj +2288 0 obj << +/D [2316 0 R /XYZ 229.129 304.716 null] +>> endobj +2289 0 obj << +/D [2316 0 R /XYZ 325.397 304.716 null] +>> endobj +2290 0 obj << +/D [2316 0 R /XYZ 122.869 291.765 null] +>> endobj +2291 0 obj << +/D [2316 0 R /XYZ 184.567 291.765 null] +>> endobj +2292 0 obj << +/D [2316 0 R /XYZ 71.731 284.627 null] +>> endobj +2293 0 obj << +/D [2316 0 R /XYZ 418.366 273.832 null] +>> endobj +2294 0 obj << +/D [2316 0 R /XYZ 460.329 273.832 null] +>> endobj +2295 0 obj << +/D [2316 0 R /XYZ 119.551 247.929 null] +>> endobj +2296 0 obj << +/D [2316 0 R /XYZ 206.265 247.929 null] +>> endobj +2297 0 obj << +/D [2316 0 R /XYZ 301.357 247.929 null] +>> endobj +2298 0 obj << +/D [2316 0 R /XYZ 263.978 234.978 null] +>> endobj +2299 0 obj << +/D [2316 0 R /XYZ 323.922 234.978 null] +>> endobj +2300 0 obj << +/D [2316 0 R /XYZ 379.453 234.978 null] +>> endobj +2301 0 obj << +/D [2316 0 R /XYZ 71.731 188.986 null] +>> endobj +2302 0 obj << +/D [2316 0 R /XYZ 141.104 178.191 null] +>> endobj +2303 0 obj << +/D [2316 0 R /XYZ 261.238 178.191 null] +>> endobj +2304 0 obj << +/D [2316 0 R /XYZ 429.176 165.24 null] +>> endobj +2305 0 obj << +/D [2316 0 R /XYZ 229.677 152.288 null] +>> endobj +2306 0 obj << +/D [2316 0 R /XYZ 280.765 152.288 null] +>> endobj +2315 0 obj << +/Font << /F31 805 0 R /F30 731 0 R /F25 681 0 R /F40 669 0 R /F56 1242 0 R /F37 1210 0 R /F60 1503 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2358 0 obj << +/Length 2359 0 R +/Filter /FlateDecode +>> +stream +xڵYn6~% fC ZbwVK Qooߢxi1X(VTթSC6AEpn;?O{W +#J&AYY>{ ޯP'ü|i9ҿG}U_qv$߲Q?I]w=Rrw>>%%J>+PQ&e-]kgcpbS\=$81+ei2Ze[KYr"lW3Ol;mHdgWU}whİ4刖[&yᵸ}>.]Z=(Nab+zӥ{8 EUFނ2^o1m=X>Oy̥dG^Nx\Rmd*Wʏ j˯n$ڙ闥=$@HEb0^J0~^TB;jDYfn+XuGGXՆFakjd +fQlN~xq3)%I3.]HɌzո*5/d^j$~0f@DĜ#Ph1yVj*Hf\]հYf+lv +*B5O?!_%|%\WȻ&f-"'qW2Ҙ*b}VE[9 Mo+U^ђ.ǁG^ZE+.϶E#*C'|@|кu1Mg2n^\-YJ=Zh:nmB{A +ZQY5J1v֫YRcmi0g-x\W8nz;l~6kZ|Wd(Cgv*D1m i9LJՒ%GR!)JR=Q'@Dɢe)Qr(!jUap}g/|dH [@Zt6{K'EgQ-aLewZ _y1s.~L̄˞nΊ4q8?@#H GQAaIˤc<$GI} /yW(V>%y_aZ|/#Q˚`X'4V-%Jp@{mrtbs9/gc.fbVF`nqշnj1 BET3 |/y P7/.H,[+Si3Y SON)r')xxG4+Cr0Jw=c |nfȟK+Y3DyPٱ4've?6of̶klD&*9v,*__Οd@c3z ~i?xV4endstream +endobj +2359 0 obj +1947 +endobj +2357 0 obj << +/Type /Page +/Contents 2358 0 R +/Resources 2356 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2322 0 R +>> endobj +2355 0 obj << +/D [2357 0 R /XYZ 71.731 705.352 null] +>> endobj +934 0 obj << +/D [2357 0 R /XYZ 71.731 717.307 null] +>> endobj +2307 0 obj << +/D [2357 0 R /XYZ 134.774 647.569 null] +>> endobj +2308 0 obj << +/D [2357 0 R /XYZ 190.544 647.569 null] +>> endobj +2309 0 obj << +/D [2357 0 R /XYZ 267.266 647.569 null] +>> endobj +2310 0 obj << +/D [2357 0 R /XYZ 414.204 647.569 null] +>> endobj +2311 0 obj << +/D [2357 0 R /XYZ 119.551 634.618 null] +>> endobj +2312 0 obj << +/D [2357 0 R /XYZ 198.564 634.618 null] +>> endobj +2323 0 obj << +/D [2357 0 R /XYZ 413.704 616.685 null] +>> endobj +2324 0 obj << +/D [2357 0 R /XYZ 424.414 603.734 null] +>> endobj +2325 0 obj << +/D [2357 0 R /XYZ 226.499 590.782 null] +>> endobj +2326 0 obj << +/D [2357 0 R /XYZ 341.894 590.782 null] +>> endobj +2327 0 obj << +/D [2357 0 R /XYZ 479.617 590.782 null] +>> endobj +2328 0 obj << +/D [2357 0 R /XYZ 161.503 577.831 null] +>> endobj +2329 0 obj << +/D [2357 0 R /XYZ 398.153 577.831 null] +>> endobj +2330 0 obj << +/D [2357 0 R /XYZ 71.731 559.679 null] +>> endobj +2331 0 obj << +/D [2357 0 R /XYZ 158.843 546.947 null] +>> endobj +2332 0 obj << +/D [2357 0 R /XYZ 474.615 546.947 null] +>> endobj +2333 0 obj << +/D [2357 0 R /XYZ 360.744 521.044 null] +>> endobj +2334 0 obj << +/D [2357 0 R /XYZ 231.341 508.093 null] +>> endobj +2335 0 obj << +/D [2357 0 R /XYZ 189.279 495.141 null] +>> endobj +2336 0 obj << +/D [2357 0 R /XYZ 180.691 482.19 null] +>> endobj +2337 0 obj << +/D [2357 0 R /XYZ 355.404 482.19 null] +>> endobj +2338 0 obj << +/D [2357 0 R /XYZ 71.731 462.101 null] +>> endobj +2339 0 obj << +/D [2357 0 R /XYZ 259.545 451.306 null] +>> endobj +2340 0 obj << +/D [2357 0 R /XYZ 370.238 451.306 null] +>> endobj +2341 0 obj << +/D [2357 0 R /XYZ 379.084 438.355 null] +>> endobj +2342 0 obj << +/D [2357 0 R /XYZ 71.731 418.265 null] +>> endobj +2343 0 obj << +/D [2357 0 R /XYZ 310.503 407.47 null] +>> endobj +2344 0 obj << +/D [2357 0 R /XYZ 390.571 407.47 null] +>> endobj +931 0 obj << +/D [2357 0 R /XYZ 71.731 390.37 null] +>> endobj +317 0 obj << +/D [2357 0 R /XYZ 302.657 363.117 null] +>> endobj +2345 0 obj << +/D [2357 0 R /XYZ 71.731 352.975 null] +>> endobj +2346 0 obj << +/D [2357 0 R /XYZ 162.59 341.498 null] +>> endobj +932 0 obj << +/D [2357 0 R /XYZ 71.731 320.524 null] +>> endobj +321 0 obj << +/D [2357 0 R /XYZ 381.561 292.163 null] +>> endobj +2347 0 obj << +/D [2357 0 R /XYZ 71.731 281.798 null] +>> endobj +2348 0 obj << +/D [2357 0 R /XYZ 137.534 272.039 null] +>> endobj +2349 0 obj << +/D [2357 0 R /XYZ 387.422 272.039 null] +>> endobj +2350 0 obj << +/D [2357 0 R /XYZ 137.534 259.087 null] +>> endobj +933 0 obj << +/D [2357 0 R /XYZ 71.731 216.084 null] +>> endobj +325 0 obj << +/D [2357 0 R /XYZ 276.433 188.831 null] +>> endobj +2351 0 obj << +/D [2357 0 R /XYZ 71.731 175.899 null] +>> endobj +2352 0 obj << +/D [2357 0 R /XYZ 141.07 167.212 null] +>> endobj +2353 0 obj << +/D [2357 0 R /XYZ 194.868 155.556 null] +>> endobj +2354 0 obj << +/D [2357 0 R /XYZ 178.729 143.9 null] +>> endobj +2356 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F37 1210 0 R /F39 1242 0 R /F30 731 0 R /F40 669 0 R /F60 1503 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2417 0 obj << +/Length 2418 0 R +/Filter /FlateDecode +>> +stream +xXM6_ad1#~JBoI C`h[XYJr7PCj]'EQ %q4|pl.BJB$\ErxBwWZ8 '2 +͓Ċ/6n~\T5R|Wx˲UKdeOߛ-å)ϛ  XXwJ&kEAP}`I" 5fE7ݘ)}e[]x[W4XVk!bBhwhH;L֣9MY=@` +M?꽮)em8Sc%X5{|xg&LG]u.U^0ɽcB3<񳎉rateD($9:2u7{GjSEnVL.? ܽu682U|RS>E |E2JE:l ϒ/K=.0 0&2?Fy=tuj@t-/\h{cG2 2LH\-Zy=( ͉@3fphpBbmoRn(7LX7äȱ" a꥔髗 WDD9]gw+_ K{EEʣ[]냎 +>g9ciyЭV>=G+):۞bFJ. .mpn[,)UCU! +im*$z }فASD2n  jg*;f-C +{e;\nG8gBBIn}}0HyVѢЀr/@/ŬHʃC܏z^Pg*QǝV++ptE:[9Fqu;H9i6$SWߙ$8gas}Qʓ !p]x@Y䧑 :fDl@>ݕU3ͅ\f." i A'C*gf:NwFʄ~ګkiP#F߼o;98wG:\ SJetǧ*Ff^v'&E Q{f`;/w +No*}8 +Ku.?ZVm嗻]Gn?}Eژc# Q֮|nB!-ҿuQiFg;O(HP\П<?hlLNjPA + ~ϦfGDD~"TS7ܶ,?~s`űȳC֘tL,`<,|r|(+#:1[v$Lm>H+j4֧ږ FF[d9t̻t%MjjNEh=Q> endobj +2419 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [396.429 624.785 441.26 633.806] +/Subtype /Link +/A << /S /GoTo /D (tsig) >> +>> endobj +2420 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [178.758 580.95 231.061 589.861] +/Subtype /Link +/A << /S /GoTo /D (controls_statement_definition_and_usage) >> +>> endobj +2414 0 obj << +/D [2416 0 R /XYZ 71.731 705.352 null] +>> endobj +329 0 obj << +/D [2416 0 R /XYZ 355.336 647.066 null] +>> endobj +2360 0 obj << +/D [2416 0 R /XYZ 71.731 636.701 null] +>> endobj +2361 0 obj << +/D [2416 0 R /XYZ 137.534 626.942 null] +>> endobj +2362 0 obj << +/D [2416 0 R /XYZ 71.731 619.804 null] +>> endobj +2363 0 obj << +/D [2416 0 R /XYZ 137.534 609.009 null] +>> endobj +2364 0 obj << +/D [2416 0 R /XYZ 437.803 609.009 null] +>> endobj +2365 0 obj << +/D [2416 0 R /XYZ 199.122 596.058 null] +>> endobj +2366 0 obj << +/D [2416 0 R /XYZ 467.043 596.058 null] +>> endobj +2367 0 obj << +/D [2416 0 R /XYZ 71.731 575.968 null] +>> endobj +2368 0 obj << +/D [2416 0 R /XYZ 137.449 565.174 null] +>> endobj +2369 0 obj << +/D [2416 0 R /XYZ 71.731 521.239 null] +>> endobj +2370 0 obj << +/D [2416 0 R /XYZ 137.534 508.387 null] +>> endobj +2371 0 obj << +/D [2416 0 R /XYZ 312.694 495.436 null] +>> endobj +2372 0 obj << +/D [2416 0 R /XYZ 378.696 495.436 null] +>> endobj +935 0 obj << +/D [2416 0 R /XYZ 71.731 465.383 null] +>> endobj +333 0 obj << +/D [2416 0 R /XYZ 304.379 438.131 null] +>> endobj +2373 0 obj << +/D [2416 0 R /XYZ 71.731 424.976 null] +>> endobj +2374 0 obj << +/D [2416 0 R /XYZ 71.731 424.976 null] +>> endobj +2375 0 obj << +/D [2416 0 R /XYZ 146.45 404.856 null] +>> endobj +2376 0 obj << +/D [2416 0 R /XYZ 189.489 404.856 null] +>> endobj +2377 0 obj << +/D [2416 0 R /XYZ 157.21 393.199 null] +>> endobj +2378 0 obj << +/D [2416 0 R /XYZ 184.109 393.199 null] +>> endobj +2379 0 obj << +/D [2416 0 R /XYZ 178.729 381.543 null] +>> endobj +2380 0 obj << +/D [2416 0 R /XYZ 237.907 381.543 null] +>> endobj +2381 0 obj << +/D [2416 0 R /XYZ 286.648 381.543 null] +>> endobj +2382 0 obj << +/D [2416 0 R /XYZ 178.729 369.887 null] +>> endobj +2383 0 obj << +/D [2416 0 R /XYZ 205.628 369.887 null] +>> endobj +2384 0 obj << +/D [2416 0 R /XYZ 167.969 358.231 null] +>> endobj +2385 0 obj << +/D [2416 0 R /XYZ 205.628 358.231 null] +>> endobj +2386 0 obj << +/D [2416 0 R /XYZ 167.969 346.574 null] +>> endobj +2387 0 obj << +/D [2416 0 R /XYZ 167.969 334.918 null] +>> endobj +2388 0 obj << +/D [2416 0 R /XYZ 157.21 323.262 null] +>> endobj +2389 0 obj << +/D [2416 0 R /XYZ 211.008 323.262 null] +>> endobj +2390 0 obj << +/D [2416 0 R /XYZ 265.882 323.262 null] +>> endobj +2391 0 obj << +/D [2416 0 R /XYZ 306.23 323.262 null] +>> endobj +2392 0 obj << +/D [2416 0 R /XYZ 356.263 323.262 null] +>> endobj +2393 0 obj << +/D [2416 0 R /XYZ 211.008 311.606 null] +>> endobj +2394 0 obj << +/D [2416 0 R /XYZ 246.515 311.606 null] +>> endobj +2395 0 obj << +/D [2416 0 R /XYZ 286.863 311.606 null] +>> endobj +2396 0 obj << +/D [2416 0 R /XYZ 340.661 311.606 null] +>> endobj +2397 0 obj << +/D [2416 0 R /XYZ 157.21 299.949 null] +>> endobj +2398 0 obj << +/D [2416 0 R /XYZ 237.907 299.949 null] +>> endobj +2399 0 obj << +/D [2416 0 R /XYZ 273.952 299.949 null] +>> endobj +2400 0 obj << +/D [2416 0 R /XYZ 157.21 288.293 null] +>> endobj +2401 0 obj << +/D [2416 0 R /XYZ 237.907 288.293 null] +>> endobj +2402 0 obj << +/D [2416 0 R /XYZ 273.952 288.293 null] +>> endobj +2403 0 obj << +/D [2416 0 R /XYZ 157.21 276.637 null] +>> endobj +2404 0 obj << +/D [2416 0 R /XYZ 216.388 276.637 null] +>> endobj +2405 0 obj << +/D [2416 0 R /XYZ 252.432 276.637 null] +>> endobj +2406 0 obj << +/D [2416 0 R /XYZ 146.45 253.324 null] +>> endobj +2407 0 obj << +/D [2416 0 R /XYZ 194.868 253.324 null] +>> endobj +2408 0 obj << +/D [2416 0 R /XYZ 146.45 241.668 null] +>> endobj +2409 0 obj << +/D [2416 0 R /XYZ 237.907 241.668 null] +>> endobj +936 0 obj << +/D [2416 0 R /XYZ 71.731 185.725 null] +>> endobj +2415 0 obj << +/Font << /F31 805 0 R /F40 669 0 R /F25 681 0 R /F30 731 0 R /F48 1503 0 R /F37 1210 0 R /F56 1242 0 R /F60 1503 0 R /F52 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2459 0 obj << +/Length 2460 0 R +/Filter /FlateDecode +>> +stream +xڥYmh>%RҶmNUUmUp"LvnZvl`fHs|^ء*$qs +>a g""ULx=/HƫT}b)w~ +|waCu96}2֢O{Sf)X9^Sx)a|9ayTNeB$uFJ,_qC8Yt#> uˮƛRsPv,"%z٧Gi4#E@LXH,)^зBcp/Ј^j8xQ6t< ʳb| 1|mBisbn3ou*NsvqJIeT/fRʾjZW/WcY[44>)/c@c}zO/xTe\dF2Gè6.v\J8!yZX~-dBA﫱v2Jze3><ZsUi\5f8֚kmvuK*9705)ͤ3"j427Mc5 8{v?/Qٮtp1siHHǼ*',%@sӶ67c4V9ppR$B4Ն*R0?kWD`ڶꁮjE98vH17 Yyn*@LB/6 E?uk!Yڻr}ȓMgN(45N6;DRT̖Ҍ~j%5RM&;^,ɓ_S>-Ei\/ +j% & bPb7 BCm˟׆4rHcBrZ{:04&mlU=LV&<̱&2B蠥g.O^Dռ3 X[A!kԡ<f -/|FF '{ҊMo)~v{빾Kc~Iq^DmcJ Z Ks3<M + ̘ۗpg4 4M=A{$W9Z Me5Ј#!&fi *9@i¶/. +k[=uU0 Ź\_ ʳavQ|uGfAkK +fQJ=lU w% J\\HɂAhSй")hb[#V@[-,v.*$sI/TմF"<\R3}SSd2Jrgk߱&oqT&Mv pI\!u~ O'gƖyoUehea-,ueշmr#'upwfsfBUGNfBnptGP^W QɗMW7ğ );XymdqB%y?6+ Ӽr~_̀x 7l1IBs5w Ch$2NU nnQ)#e ~IIw|6|%{/89 ;wq85(; ]*S@N/fC~PZooq9)~B4&q> y+-gpJk&endstream +endobj +2460 0 obj +2298 +endobj +2458 0 obj << +/Type /Page +/Contents 2459 0 R +/Resources 2457 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2322 0 R +>> endobj +2456 0 obj << +/D [2458 0 R /XYZ 71.731 705.352 null] +>> endobj +337 0 obj << +/D [2458 0 R /XYZ 391.259 647.066 null] +>> endobj +2410 0 obj << +/D [2458 0 R /XYZ 71.731 636.701 null] +>> endobj +2411 0 obj << +/D [2458 0 R /XYZ 137.534 626.942 null] +>> endobj +2412 0 obj << +/D [2458 0 R /XYZ 481.36 626.942 null] +>> endobj +2413 0 obj << +/D [2458 0 R /XYZ 154.42 601.039 null] +>> endobj +2421 0 obj << +/D [2458 0 R /XYZ 71.731 593.901 null] +>> endobj +2422 0 obj << +/D [2458 0 R /XYZ 158.844 583.106 null] +>> endobj +2423 0 obj << +/D [2458 0 R /XYZ 132.004 570.155 null] +>> endobj +2424 0 obj << +/D [2458 0 R /XYZ 71.731 558.036 null] +>> endobj +2425 0 obj << +/D [2458 0 R /XYZ 71.731 502.556 null] +>> endobj +2426 0 obj << +/D [2458 0 R /XYZ 130.341 490.653 null] +>> endobj +2427 0 obj << +/D [2458 0 R /XYZ 161.324 477.702 null] +>> endobj +2428 0 obj << +/D [2458 0 R /XYZ 328.355 477.702 null] +>> endobj +2429 0 obj << +/D [2458 0 R /XYZ 265.821 451.799 null] +>> endobj +937 0 obj << +/D [2458 0 R /XYZ 71.731 439.68 null] +>> endobj +341 0 obj << +/D [2458 0 R /XYZ 282.234 417.329 null] +>> endobj +2430 0 obj << +/D [2458 0 R /XYZ 71.731 411.202 null] +>> endobj +2431 0 obj << +/D [2458 0 R /XYZ 259.574 398.4 null] +>> endobj +2432 0 obj << +/D [2458 0 R /XYZ 71.731 391.262 null] +>> endobj +2433 0 obj << +/D [2458 0 R /XYZ 510.738 354.564 null] +>> endobj +2434 0 obj << +/D [2458 0 R /XYZ 219.983 341.613 null] +>> endobj +2435 0 obj << +/D [2458 0 R /XYZ 71.731 321.523 null] +>> endobj +2436 0 obj << +/D [2458 0 R /XYZ 137.534 310.729 null] +>> endobj +2437 0 obj << +/D [2458 0 R /XYZ 71.731 290.639 null] +>> endobj +2438 0 obj << +/D [2458 0 R /XYZ 137.534 279.845 null] +>> endobj +2439 0 obj << +/D [2458 0 R /XYZ 71.731 259.755 null] +>> endobj +2440 0 obj << +/D [2458 0 R /XYZ 176.547 248.961 null] +>> endobj +2441 0 obj << +/D [2458 0 R /XYZ 294.343 248.961 null] +>> endobj +2442 0 obj << +/D [2458 0 R /XYZ 484.183 236.009 null] +>> endobj +2443 0 obj << +/D [2458 0 R /XYZ 232.446 223.058 null] +>> endobj +2444 0 obj << +/D [2458 0 R /XYZ 355.663 223.058 null] +>> endobj +2445 0 obj << +/D [2458 0 R /XYZ 425.202 223.058 null] +>> endobj +2446 0 obj << +/D [2458 0 R /XYZ 129.793 210.107 null] +>> endobj +2447 0 obj << +/D [2458 0 R /XYZ 216.208 210.107 null] +>> endobj +2448 0 obj << +/D [2458 0 R /XYZ 328.665 210.107 null] +>> endobj +2449 0 obj << +/D [2458 0 R /XYZ 448.844 210.107 null] +>> endobj +2450 0 obj << +/D [2458 0 R /XYZ 278.792 197.155 null] +>> endobj +2451 0 obj << +/D [2458 0 R /XYZ 71.731 164.114 null] +>> endobj +2452 0 obj << +/D [2458 0 R /XYZ 137.481 153.32 null] +>> endobj +2453 0 obj << +/D [2458 0 R /XYZ 472.213 153.32 null] +>> endobj +2454 0 obj << +/D [2458 0 R /XYZ 249.063 140.368 null] +>> endobj +2457 0 obj << +/Font << /F31 805 0 R /F40 669 0 R /F25 681 0 R /F30 731 0 R /F37 1210 0 R /F42 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2510 0 obj << +/Length 2511 0 R +/Filter /FlateDecode +>> +stream +xڭY[۸ҷ +8fu.S7EӢ}([[LD$srvU7q "SpoMM"KQMݾ7ϗ<3*BIlRB ~+T=x&I]mO#ޝ1ηD2ޙGx3-goB\(Jp](K7;\L ^LzjyЃ'\T={GU[ҹ̥ ޾νY|:SKE.<#S/7, S/ZyRRX$T&N(.8Cy{E5Me&3W9U `ïgV9`̫6ou2tźyؤݚi:|Uku +d v`#8YlGN9L-a| `NS,jUaNz/Q[UYP] r@u=zAY+LRo@LcT<=G5L&uP@NxspYDJ"dݫГ5=aB0/"JΔ tvEp %g&Ycҹ),"#ꙫ$TZɯ1s@kYG;L<Ɂ4.DxѲx|8-PR TւɭS4H Țp}H8DKrJ@/Gu^!e dFJ.MU8/]fvE #SE!*.4.k; C y)QUw;5aِyٰ _.Z)|? 览NٌHKL()\L]{Rȑy&+,ɜ~ƷY45F([ЖRU^nkY142Ag-c&yױCDو=Wbzކk̜Q@>߾T. YAzS0E +%?ab.-;P,}]J-lw:MH}$:Q"j4siU< Q #icJLp6A{lo fuWgiAhOo`zZztO +(kwMAA&(ZOMs)_=Źյ:PTXv; 45|AZ?|4%P}صКA#WA7K;eXϰpЗ['ChZ> Da"%'u2s6@0 sRF|2X:Ǯ委7 njs϶ƞjÒc{b@WЈe[K; Dx; z*f4n +-.+P {ϷTحo8].cDצLendstream +endobj +2511 0 obj +2230 +endobj +2509 0 obj << +/Type /Page +/Contents 2510 0 R +/Resources 2508 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2322 0 R +>> endobj +2507 0 obj << +/D [2509 0 R /XYZ 71.731 705.352 null] +>> endobj +2455 0 obj << +/D [2509 0 R /XYZ 379.185 647.569 null] +>> endobj +2461 0 obj << +/D [2509 0 R /XYZ 71.731 616.586 null] +>> endobj +2462 0 obj << +/D [2509 0 R /XYZ 208.088 603.734 null] +>> endobj +2463 0 obj << +/D [2509 0 R /XYZ 242.947 603.734 null] +>> endobj +2464 0 obj << +/D [2509 0 R /XYZ 71.731 591.614 null] +>> endobj +2465 0 obj << +/D [2509 0 R /XYZ 71.731 524.478 null] +>> endobj +2466 0 obj << +/D [2509 0 R /XYZ 137.534 512.576 null] +>> endobj +2467 0 obj << +/D [2509 0 R /XYZ 185.124 499.625 null] +>> endobj +2468 0 obj << +/D [2509 0 R /XYZ 278.125 499.625 null] +>> endobj +2469 0 obj << +/D [2509 0 R /XYZ 134.216 486.673 null] +>> endobj +2470 0 obj << +/D [2509 0 R /XYZ 441.929 486.673 null] +>> endobj +2471 0 obj << +/D [2509 0 R /XYZ 205.607 473.722 null] +>> endobj +2472 0 obj << +/D [2509 0 R /XYZ 71.731 466.584 null] +>> endobj +2473 0 obj << +/D [2509 0 R /XYZ 137.534 455.789 null] +>> endobj +2474 0 obj << +/D [2509 0 R /XYZ 244.431 455.789 null] +>> endobj +2475 0 obj << +/D [2509 0 R /XYZ 253.756 442.838 null] +>> endobj +2476 0 obj << +/D [2509 0 R /XYZ 71.731 422.748 null] +>> endobj +2477 0 obj << +/D [2509 0 R /XYZ 184.846 411.954 null] +>> endobj +2478 0 obj << +/D [2509 0 R /XYZ 249.602 411.954 null] +>> endobj +2479 0 obj << +/D [2509 0 R /XYZ 377.74 399.002 null] +>> endobj +2480 0 obj << +/D [2509 0 R /XYZ 430.87 399.002 null] +>> endobj +2481 0 obj << +/D [2509 0 R /XYZ 119.551 386.051 null] +>> endobj +2482 0 obj << +/D [2509 0 R /XYZ 208.238 386.051 null] +>> endobj +2483 0 obj << +/D [2509 0 R /XYZ 383.946 386.051 null] +>> endobj +2484 0 obj << +/D [2509 0 R /XYZ 419.672 386.051 null] +>> endobj +2485 0 obj << +/D [2509 0 R /XYZ 236.74 373.1 null] +>> endobj +2486 0 obj << +/D [2509 0 R /XYZ 370.397 373.1 null] +>> endobj +2487 0 obj << +/D [2509 0 R /XYZ 468.638 373.1 null] +>> endobj +2488 0 obj << +/D [2509 0 R /XYZ 71.731 353.01 null] +>> endobj +2489 0 obj << +/D [2509 0 R /XYZ 137.534 342.215 null] +>> endobj +2490 0 obj << +/D [2509 0 R /XYZ 71.731 322.126 null] +>> endobj +2491 0 obj << +/D [2509 0 R /XYZ 204.222 285.429 null] +>> endobj +2492 0 obj << +/D [2509 0 R /XYZ 297.043 285.429 null] +>> endobj +2493 0 obj << +/D [2509 0 R /XYZ 507.223 285.429 null] +>> endobj +2494 0 obj << +/D [2509 0 R /XYZ 489.809 272.477 null] +>> endobj +2495 0 obj << +/D [2509 0 R /XYZ 71.731 234.455 null] +>> endobj +2496 0 obj << +/D [2509 0 R /XYZ 71.731 178.975 null] +>> endobj +2497 0 obj << +/D [2509 0 R /XYZ 275.026 154.122 null] +>> endobj +2508 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F30 731 0 R /F37 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2529 0 obj << +/Length 2530 0 R +/Filter /FlateDecode +>> +stream +xXK6_aem bH꽹%mڴ@oEQp-"S(eCz؛M9|3.[P)#i0[_~39iS9_$R#n؟< 뭕[.ű*1]&}}VwYd)ڲVn0]J?~C`iB(A bE$2?n.0rxǦTmЖ.ψ© yMjFk'^ ʗh +ŭB;z.UɂLqGq*G)R +J,~#, [ݺr#G)e|;HQIh/<ۺk|e!+,v/N$ڹxM3R%qaWb8^ L\azsfƺyf 2#SBa%8S).X=22Jd 6#}KnI"7nW mH"ƅWZֱ(&q 2g$aeXtN7]v[$DZyUB{޼qZόg}*`yH4D يMYaI-b6,NcTpۋX߻/wN(n]o$d p-M!l2H{c%.|s彴rt`bl؋2Χ5IPԉvYf,$qiGސ.#Z6_HnEʡҡel-[YXyOSQg0?IF*nكGj%[%FdE: % շ3fލ]2hݣZ#bL/Mik=SsKl!9tJzzB{P/D|pn_]̼DRsQck7y8|; 2\|ll\|.(Yt]{P=h: \0;0`g 7퀪_OMh>|^߆}*9Y +o~3~Xcmh$Al5lA MD-d/$iɰ7'8q"W;Cⶵendstream +endobj +2530 0 obj +1510 +endobj +2528 0 obj << +/Type /Page +/Contents 2529 0 R +/Resources 2527 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2322 0 R +/Annots [ 2531 0 R ] +>> endobj +2531 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [200.347 533.896 265.104 540.75] +/Subtype /Link +/A << /S /GoTo /D (the_category_phrase) >> +>> endobj +2526 0 obj << +/D [2528 0 R /XYZ 71.731 705.352 null] +>> endobj +2498 0 obj << +/D [2528 0 R /XYZ 71.731 657.532 null] +>> endobj +2499 0 obj << +/D [2528 0 R /XYZ 128.677 647.569 null] +>> endobj +2500 0 obj << +/D [2528 0 R /XYZ 407.308 647.569 null] +>> endobj +2501 0 obj << +/D [2528 0 R /XYZ 126.465 634.618 null] +>> endobj +2502 0 obj << +/D [2528 0 R /XYZ 306.697 634.618 null] +>> endobj +2503 0 obj << +/D [2528 0 R /XYZ 460.826 634.618 null] +>> endobj +2504 0 obj << +/D [2528 0 R /XYZ 428.927 621.666 null] +>> endobj +2505 0 obj << +/D [2528 0 R /XYZ 330.607 608.715 null] +>> endobj +2506 0 obj << +/D [2528 0 R /XYZ 119.551 582.812 null] +>> endobj +2512 0 obj << +/D [2528 0 R /XYZ 71.731 575.674 null] +>> endobj +2513 0 obj << +/D [2528 0 R /XYZ 71.731 575.674 null] +>> endobj +2514 0 obj << +/D [2528 0 R /XYZ 71.731 558.491 null] +>> endobj +2515 0 obj << +/D [2528 0 R /XYZ 327.858 546.947 null] +>> endobj +2516 0 obj << +/D [2528 0 R /XYZ 71.731 523.933 null] +>> endobj +2517 0 obj << +/D [2528 0 R /XYZ 71.731 186.647 null] +>> endobj +2518 0 obj << +/D [2528 0 R /XYZ 137.445 174.744 null] +>> endobj +2519 0 obj << +/D [2528 0 R /XYZ 293.476 161.793 null] +>> endobj +2527 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F30 731 0 R /F37 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2549 0 obj << +/Length 2550 0 R +/Filter /FlateDecode +>> +stream +xڍXK6_aXjNMi +PE@[-D& Q;Ԑޅ)rf7.3㟲CPQ<̘A|b v91"HY3JӢ $Ey#^>"ڮep`FP'vΉY'&Vw- `G' DaB80绵ﭚHS_ߏ\ө(. P2˷aT5BEħGU]a^#[ld gU>|^ecm*nQ"4FըIshū/jOE}d&VBő"EC^ +ΫHO+4$)ЄdAЪClJaAp, T{hvԨUt9D.zS-@걿E_̋u35sVi wfEcbJ(X=pӆ"dd +QE70mI| k_GvC*c A f_΃2ܟAZId▤UqS /KnX 8J7nz~W GW58Hƹj5:ްsvos32N'hCK1fR撚A +OYJhNm~.j"b6(xvwW8+nʍkK -lY!$vd,VlWr.4LiƼWEaMu녏#:zUf 3oneԳ͝> endobj +2546 0 obj << +/D [2548 0 R /XYZ 71.731 704.356 null] +>> endobj +2520 0 obj << +/D [2548 0 R /XYZ 71.731 657.532 null] +>> endobj +2521 0 obj << +/D [2548 0 R /XYZ 247.42 647.569 null] +>> endobj +2522 0 obj << +/D [2548 0 R /XYZ 398.95 647.569 null] +>> endobj +2523 0 obj << +/D [2548 0 R /XYZ 140.293 634.618 null] +>> endobj +2524 0 obj << +/D [2548 0 R /XYZ 443.493 634.618 null] +>> endobj +2525 0 obj << +/D [2548 0 R /XYZ 123.616 608.715 null] +>> endobj +2532 0 obj << +/D [2548 0 R /XYZ 71.731 601.577 null] +>> endobj +938 0 obj << +/D [2548 0 R /XYZ 71.731 560.73 null] +>> endobj +345 0 obj << +/D [2548 0 R /XYZ 287.124 537.383 null] +>> endobj +2533 0 obj << +/D [2548 0 R /XYZ 71.731 528.745 null] +>> endobj +2534 0 obj << +/D [2548 0 R /XYZ 229.527 492.551 null] +>> endobj +2535 0 obj << +/D [2548 0 R /XYZ 71.731 467.48 null] +>> endobj +2536 0 obj << +/D [2548 0 R /XYZ 71.731 446.611 null] +>> endobj +2537 0 obj << +/D [2548 0 R /XYZ 71.731 409.996 null] +>> endobj +2538 0 obj << +/D [2548 0 R /XYZ 71.731 296.235 null] +>> endobj +2539 0 obj << +/D [2548 0 R /XYZ 317.715 284.333 null] +>> endobj +2540 0 obj << +/D [2548 0 R /XYZ 71.731 272.213 null] +>> endobj +2541 0 obj << +/D [2548 0 R /XYZ 71.731 239.687 null] +>> endobj +2542 0 obj << +/D [2548 0 R /XYZ 280.695 215.192 null] +>> endobj +2543 0 obj << +/D [2548 0 R /XYZ 71.731 203.073 null] +>> endobj +2544 0 obj << +/D [2548 0 R /XYZ 71.731 198.092 null] +>> endobj +2545 0 obj << +/D [2548 0 R /XYZ 71.731 198.092 null] +>> endobj +2547 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F37 1210 0 R /F30 731 0 R /F42 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2568 0 obj << +/Length 2569 0 R +/Filter /FlateDecode +>> +stream +xXKs61J3!B'θM܇C&!c +T0?PDێGJ}}.{)Fi FaLb +{so;"x!T/~x +(OBo-_gkUL}[n XbUUv[|âW +Hc +Pz~$y e9y~d(HrOEުԪLZAQŚC +O9g;K+=gRUum^SWs͔=RV4?+-pǸ jDk{/<6N<:[bC!GcxhGf(W߽g}\REdN̤+&΂ 6{;ܙvu+d/Wv'N&`N76/ڣNTKl8t)hz9O1J(M:ZM8v+m96G*f{ˤA8P:̼=gXYe*<ֹmTTH%Vo'8ܪx=f M +L65.Р;o[ jC-#ȶX(ZQBcBV:u4fϖe#6Bd[zyW0ۨb;~4xrJ~'\KJP.a1uUc3lZ*켜Qr2~< +f뫟{*PM/Eo] +M(}OWcoi:PBI(y-ȓ+ZSfKs rWKu~Sh;4\=cl}_/1RD+Dơp,̬_*`ķVSjw*-ÞO H]vZ6W+% wq7OUdo.[y ԧ:>VXPQ 1aA(ْ=R261zxZbM9g#g(Fp< +|β S6q?J(]RT-P1j ¦_CT 0,nnڍ$/S#@j?YH/,&4X?yu)+|5n줸Th6}}שv$CMS5{=e߱ϙg5Nӽˣjq{~D391^Q9vިJT}e_hdPYb0I0׺l@-,> {W]7S 7w%qN.L6[$6Jt*ُd! Jt$D{`CC#S^Pp}#3B!AD:Zi6Y;y)7 Ϗ Q +A6خydž=^)u3|S|> endobj +2565 0 obj << +/D [2567 0 R /XYZ 71.731 705.352 null] +>> endobj +939 0 obj << +/D [2567 0 R /XYZ 71.731 219.775 null] +>> endobj +349 0 obj << +/D [2567 0 R /XYZ 297.163 190.365 null] +>> endobj +2552 0 obj << +/D [2567 0 R /XYZ 71.731 183.013 null] +>> endobj +2553 0 obj << +/D [2567 0 R /XYZ 227.465 170.241 null] +>> endobj +2554 0 obj << +/D [2567 0 R /XYZ 318.044 170.241 null] +>> endobj +2555 0 obj << +/D [2567 0 R /XYZ 71.731 158.122 null] +>> endobj +2556 0 obj << +/D [2567 0 R /XYZ 71.731 158.122 null] +>> endobj +2566 0 obj << +/Font << /F31 805 0 R /F30 731 0 R /F25 681 0 R /F40 669 0 R /F37 1210 0 R /F56 1242 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2619 0 obj << +/Length 2620 0 R +/Filter /FlateDecode +>> +stream +xڽXK6_ D\hnI -C=AHMDnCDM7H`poHM?I0J(\B(#~7ycG3-MB64Qo\3tsnǼx  1oxGhéxmӷ|_d˛>כ($&|f10 X>'|>qn+!& 1mQQLQׯpDbfQ'yYZji/~XkdiZj xe>huRw!=0}fou2&;yZB|l{&[\)*IdiR/2l\4ì;ɓ#>0C)6e;ȵ;ZwS{PC l F hh4.r1"Ivmz.!x͛A߾W0yS2p{oBHLX( P49HNPWe[.H,6rr"Zn`' Qs;>U9WҎ +,7Zr}8q(>W K pz pf(3_}It9qKx1faL=M=SB se$w/4כ@W=k}U ˆǦX:9Y^wb8Q);¶Ɛb,bG'x?.([+ߊgt -YvW̴ݛfV:F5 O-|#aEfB`r@C7:눗9bCSU*x76mc0Li ?@F+yDښU;O,BNgb&Kw\d_y?qp[S6WH$Rޓ%β<3B~;_R\-?5rQ6\B)YOΛk{}qHzSCqimϐnTYAR׆ÚpA_TЭaO=3Mpy߷s]-U%0J.ޚ[W=Dah爮=)"8V .4 adP\BZY6;Q}I(MB4J6'ezk( +s xSZFyIQE8p:SE螹X,;fd?<Ᏹcs . +0(9@baOIQ]Uw;,s5{n5Pum!Sܮx\ F5;SKl>uu߿aSg}0ZqBULבVEd_nqbsAr db⿬u$:։atyo<>®CZ7R^#a̻Y@1I|Kު}ݵ(x\ƨ_vXﵜc LuLC)v]ܬÍw5HI!$ȓ2<]v^Z)4jQ Ͽ˅#Ew`9r{]!$)Gy8KV'!|pendstream +endobj +2620 0 obj +1657 +endobj +2618 0 obj << +/Type /Page +/Contents 2619 0 R +/Resources 2617 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2551 0 R +/Annots [ 2621 0 R 2622 0 R ] +>> endobj +2621 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [499.501 517.672 537.981 526.693] +/Subtype /Link +/A << /S /GoTo /D (lwresd) >> +>> endobj +2622 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [71.731 504.721 132.004 513.742] +/Subtype /Link +/A << /S /GoTo /D (lwresd) >> +>> endobj +2616 0 obj << +/D [2618 0 R /XYZ 71.731 705.352 null] +>> endobj +2557 0 obj << +/D [2618 0 R /XYZ 141.07 647.569 null] +>> endobj +2558 0 obj << +/D [2618 0 R /XYZ 146.45 635.913 null] +>> endobj +2559 0 obj << +/D [2618 0 R /XYZ 189.811 635.913 null] +>> endobj +2560 0 obj << +/D [2618 0 R /XYZ 222.09 635.913 null] +>> endobj +2561 0 obj << +/D [2618 0 R /XYZ 281.268 635.913 null] +>> endobj +2562 0 obj << +/D [2618 0 R /XYZ 292.028 635.913 null] +>> endobj +2563 0 obj << +/D [2618 0 R /XYZ 335.389 635.913 null] +>> endobj +2564 0 obj << +/D [2618 0 R /XYZ 367.668 635.913 null] +>> endobj +2570 0 obj << +/D [2618 0 R /XYZ 141.07 624.257 null] +>> endobj +2571 0 obj << +/D [2618 0 R /XYZ 178.729 624.257 null] +>> endobj +2572 0 obj << +/D [2618 0 R /XYZ 141.07 612.601 null] +>> endobj +2573 0 obj << +/D [2618 0 R /XYZ 200.248 612.601 null] +>> endobj +2574 0 obj << +/D [2618 0 R /XYZ 275.565 612.601 null] +>> endobj +2575 0 obj << +/D [2618 0 R /XYZ 286.325 612.601 null] +>> endobj +2576 0 obj << +/D [2618 0 R /XYZ 141.07 600.944 null] +>> endobj +2577 0 obj << +/D [2618 0 R /XYZ 184.109 600.944 null] +>> endobj +940 0 obj << +/D [2618 0 R /XYZ 71.731 568.314 null] +>> endobj +353 0 obj << +/D [2618 0 R /XYZ 376.066 539.953 null] +>> endobj +2578 0 obj << +/D [2618 0 R /XYZ 71.731 529.588 null] +>> endobj +2579 0 obj << +/D [2618 0 R /XYZ 137.534 519.829 null] +>> endobj +2580 0 obj << +/D [2618 0 R /XYZ 241.851 506.877 null] +>> endobj +2581 0 obj << +/D [2618 0 R /XYZ 71.731 486.788 null] +>> endobj +2582 0 obj << +/D [2618 0 R /XYZ 137.534 475.993 null] +>> endobj +2583 0 obj << +/D [2618 0 R /XYZ 71.731 442.952 null] +>> endobj +2584 0 obj << +/D [2618 0 R /XYZ 137.534 432.158 null] +>> endobj +2585 0 obj << +/D [2618 0 R /XYZ 71.731 386.166 null] +>> endobj +2586 0 obj << +/D [2618 0 R /XYZ 137.534 375.371 null] +>> endobj +2587 0 obj << +/D [2618 0 R /XYZ 285.725 375.371 null] +>> endobj +2588 0 obj << +/D [2618 0 R /XYZ 366.621 375.371 null] +>> endobj +2589 0 obj << +/D [2618 0 R /XYZ 71.731 355.281 null] +>> endobj +2590 0 obj << +/D [2618 0 R /XYZ 137.534 344.487 null] +>> endobj +2591 0 obj << +/D [2618 0 R /XYZ 281.491 344.487 null] +>> endobj +2592 0 obj << +/D [2618 0 R /XYZ 358.152 344.487 null] +>> endobj +941 0 obj << +/D [2618 0 R /XYZ 71.731 301.483 null] +>> endobj +357 0 obj << +/D [2618 0 R /XYZ 312.499 274.231 null] +>> endobj +2593 0 obj << +/D [2618 0 R /XYZ 71.731 264.088 null] +>> endobj +2594 0 obj << +/D [2618 0 R /XYZ 227.465 254.106 null] +>> endobj +2595 0 obj << +/D [2618 0 R /XYZ 326.542 254.106 null] +>> endobj +2596 0 obj << +/D [2618 0 R /XYZ 71.731 241.987 null] +>> endobj +2597 0 obj << +/D [2618 0 R /XYZ 141.07 220.831 null] +>> endobj +2598 0 obj << +/D [2618 0 R /XYZ 194.868 220.831 null] +>> endobj +2599 0 obj << +/D [2618 0 R /XYZ 141.07 209.175 null] +>> endobj +2600 0 obj << +/D [2618 0 R /XYZ 205.628 209.175 null] +>> endobj +2601 0 obj << +/D [2618 0 R /XYZ 141.07 197.519 null] +>> endobj +2602 0 obj << +/D [2618 0 R /XYZ 211.008 197.519 null] +>> endobj +2603 0 obj << +/D [2618 0 R /XYZ 141.07 185.862 null] +>> endobj +2604 0 obj << +/D [2618 0 R /XYZ 216.388 185.862 null] +>> endobj +2605 0 obj << +/D [2618 0 R /XYZ 141.07 174.206 null] +>> endobj +2606 0 obj << +/D [2618 0 R /XYZ 211.008 174.206 null] +>> endobj +2607 0 obj << +/D [2618 0 R /XYZ 259.426 174.206 null] +>> endobj +2608 0 obj << +/D [2618 0 R /XYZ 141.07 162.55 null] +>> endobj +2609 0 obj << +/D [2618 0 R /XYZ 205.628 162.55 null] +>> endobj +2610 0 obj << +/D [2618 0 R /XYZ 141.07 150.894 null] +>> endobj +2611 0 obj << +/D [2618 0 R /XYZ 254.046 150.894 null] +>> endobj +2612 0 obj << +/D [2618 0 R /XYZ 141.07 139.237 null] +>> endobj +2613 0 obj << +/D [2618 0 R /XYZ 200.248 139.237 null] +>> endobj +2617 0 obj << +/Font << /F31 805 0 R /F37 1210 0 R /F60 1503 0 R /F40 669 0 R /F25 681 0 R /F30 731 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2754 0 obj << +/Length 2755 0 R +/Filter /FlateDecode +>> +stream +xYߏ6Ew+هmlc=ԇj߮%-7@pb rjAf曙o~ؐd" .njI!iTyR:aZGx_a%~Yw?ta.ޭ\ߩ!Kx\wD܆e?dV|/-"h0A-!). Yt1CnM XLGdGփޯ_r1Mv]W`u??W&S`EW?YÈ6Ӭ٥ԇrY Pyޗg*g>U\vt$Ye)үqyA3anz47=, i&$piOt>T/!?[cN42uj]F lsfh+Р,(uZ^̔1ٖq/Yj$X+7YKVIFQUî+iVH h@ih_$bgco]Ua +`%9?7i:R<(TsMP-+Uj,KP,T- }0ϊ?7,MևgômLDžqPQLeCYB9gP9$YwMe.LC`V4 +g:1A* +G*JpTkX[C?˼ +ccsUwI:zBu5`Tö +$!KM#iO_fYҀBGR 4 [4+5 "6/h &1ec4]x'ؽG7>00e_ 1JЮ#Ni#}Ȁ2\h릓^Mcũ CFO 6}X| +Ip7sJNQSwc[7^wmuB`1Fn7_3^ƀc77lUDڲqAz:cx0F~tU}7"Cbwo}\h.t_>X,׶$2^t|p$Cz7Zd+ө„~I{сϒ k8PFyAUO -TmGn7A6]aRHKՋ;*NǷ|M#TӡMۼWaÛ/ p>UpX3߉rN4zr} +;ΔsRjfS06Ꞧ7j}]90p^uէgoS JqډI{8IbGM>: gaRgGp*tgQ& 3 }K\9Ǻ 8P*?}Рj멗x}+endstream +endobj +2755 0 obj +1391 +endobj +2753 0 obj << +/Type /Page +/Contents 2754 0 R +/Resources 2752 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2551 0 R +>> endobj +2751 0 obj << +/D [2753 0 R /XYZ 71.731 705.352 null] +>> endobj +2614 0 obj << +/D [2753 0 R /XYZ 141.07 647.569 null] +>> endobj +2615 0 obj << +/D [2753 0 R /XYZ 237.907 647.569 null] +>> endobj +2623 0 obj << +/D [2753 0 R /XYZ 141.07 635.913 null] +>> endobj +2624 0 obj << +/D [2753 0 R /XYZ 237.907 635.913 null] +>> endobj +2625 0 obj << +/D [2753 0 R /XYZ 141.07 624.257 null] +>> endobj +2626 0 obj << +/D [2753 0 R /XYZ 227.147 624.257 null] +>> endobj +2627 0 obj << +/D [2753 0 R /XYZ 141.07 612.601 null] +>> endobj +2628 0 obj << +/D [2753 0 R /XYZ 254.046 612.601 null] +>> endobj +2629 0 obj << +/D [2753 0 R /XYZ 141.07 600.944 null] +>> endobj +2630 0 obj << +/D [2753 0 R /XYZ 189.489 600.944 null] +>> endobj +2631 0 obj << +/D [2753 0 R /XYZ 141.07 589.288 null] +>> endobj +2632 0 obj << +/D [2753 0 R /XYZ 216.388 589.288 null] +>> endobj +2633 0 obj << +/D [2753 0 R /XYZ 141.07 577.632 null] +>> endobj +2634 0 obj << +/D [2753 0 R /XYZ 211.008 577.632 null] +>> endobj +2635 0 obj << +/D [2753 0 R /XYZ 141.07 565.976 null] +>> endobj +2636 0 obj << +/D [2753 0 R /XYZ 237.907 565.976 null] +>> endobj +2637 0 obj << +/D [2753 0 R /XYZ 141.07 554.319 null] +>> endobj +2638 0 obj << +/D [2753 0 R /XYZ 237.907 554.319 null] +>> endobj +2639 0 obj << +/D [2753 0 R /XYZ 141.07 542.663 null] +>> endobj +2640 0 obj << +/D [2753 0 R /XYZ 248.666 542.663 null] +>> endobj +2641 0 obj << +/D [2753 0 R /XYZ 141.07 531.007 null] +>> endobj +2642 0 obj << +/D [2753 0 R /XYZ 237.907 531.007 null] +>> endobj +2643 0 obj << +/D [2753 0 R /XYZ 141.07 519.351 null] +>> endobj +2644 0 obj << +/D [2753 0 R /XYZ 189.489 519.351 null] +>> endobj +2645 0 obj << +/D [2753 0 R /XYZ 254.046 519.351 null] +>> endobj +2646 0 obj << +/D [2753 0 R /XYZ 141.07 507.694 null] +>> endobj +2647 0 obj << +/D [2753 0 R /XYZ 205.628 507.694 null] +>> endobj +2648 0 obj << +/D [2753 0 R /XYZ 141.07 496.038 null] +>> endobj +2649 0 obj << +/D [2753 0 R /XYZ 227.147 496.038 null] +>> endobj +2650 0 obj << +/D [2753 0 R /XYZ 141.07 484.382 null] +>> endobj +2651 0 obj << +/D [2753 0 R /XYZ 216.388 484.382 null] +>> endobj +2652 0 obj << +/D [2753 0 R /XYZ 141.07 472.726 null] +>> endobj +2653 0 obj << +/D [2753 0 R /XYZ 254.046 472.726 null] +>> endobj +2654 0 obj << +/D [2753 0 R /XYZ 141.07 461.069 null] +>> endobj +2655 0 obj << +/D [2753 0 R /XYZ 205.628 461.069 null] +>> endobj +2656 0 obj << +/D [2753 0 R /XYZ 244.031 461.069 null] +>> endobj +2657 0 obj << +/D [2753 0 R /XYZ 141.07 449.413 null] +>> endobj +2658 0 obj << +/D [2753 0 R /XYZ 173.349 437.757 null] +>> endobj +2659 0 obj << +/D [2753 0 R /XYZ 216.71 437.757 null] +>> endobj +2660 0 obj << +/D [2753 0 R /XYZ 248.989 437.757 null] +>> endobj +2661 0 obj << +/D [2753 0 R /XYZ 308.167 437.757 null] +>> endobj +2662 0 obj << +/D [2753 0 R /XYZ 318.927 437.757 null] +>> endobj +2663 0 obj << +/D [2753 0 R /XYZ 362.288 437.757 null] +>> endobj +2664 0 obj << +/D [2753 0 R /XYZ 394.567 437.757 null] +>> endobj +2665 0 obj << +/D [2753 0 R /XYZ 141.07 426.101 null] +>> endobj +2666 0 obj << +/D [2753 0 R /XYZ 227.147 426.101 null] +>> endobj +2667 0 obj << +/D [2753 0 R /XYZ 275.888 426.101 null] +>> endobj +2668 0 obj << +/D [2753 0 R /XYZ 318.927 426.101 null] +>> endobj +2669 0 obj << +/D [2753 0 R /XYZ 388.864 426.101 null] +>> endobj +2670 0 obj << +/D [2753 0 R /XYZ 426.523 426.101 null] +>> endobj +2671 0 obj << +/D [2753 0 R /XYZ 464.181 426.101 null] +>> endobj +2672 0 obj << +/D [2753 0 R /XYZ 141.07 414.444 null] +>> endobj +2673 0 obj << +/D [2753 0 R /XYZ 232.527 414.444 null] +>> endobj +2674 0 obj << +/D [2753 0 R /XYZ 141.07 402.788 null] +>> endobj +2675 0 obj << +/D [2753 0 R /XYZ 227.147 402.788 null] +>> endobj +2676 0 obj << +/D [2753 0 R /XYZ 141.07 391.132 null] +>> endobj +2677 0 obj << +/D [2753 0 R /XYZ 243.287 391.132 null] +>> endobj +2678 0 obj << +/D [2753 0 R /XYZ 141.07 379.476 null] +>> endobj +2679 0 obj << +/D [2753 0 R /XYZ 248.666 379.476 null] +>> endobj +2680 0 obj << +/D [2753 0 R /XYZ 141.07 367.819 null] +>> endobj +2681 0 obj << +/D [2753 0 R /XYZ 264.806 367.819 null] +>> endobj +2682 0 obj << +/D [2753 0 R /XYZ 141.07 356.163 null] +>> endobj +2683 0 obj << +/D [2753 0 R /XYZ 216.388 356.163 null] +>> endobj +2684 0 obj << +/D [2753 0 R /XYZ 141.07 344.507 null] +>> endobj +2685 0 obj << +/D [2753 0 R /XYZ 205.628 344.507 null] +>> endobj +2686 0 obj << +/D [2753 0 R /XYZ 243.287 344.507 null] +>> endobj +2687 0 obj << +/D [2753 0 R /XYZ 307.844 344.507 null] +>> endobj +2688 0 obj << +/D [2753 0 R /XYZ 141.07 332.851 null] +>> endobj +2689 0 obj << +/D [2753 0 R /XYZ 221.767 332.851 null] +>> endobj +2690 0 obj << +/D [2753 0 R /XYZ 259.426 332.851 null] +>> endobj +2691 0 obj << +/D [2753 0 R /XYZ 323.984 332.851 null] +>> endobj +2692 0 obj << +/D [2753 0 R /XYZ 141.07 321.194 null] +>> endobj +2693 0 obj << +/D [2753 0 R /XYZ 221.767 321.194 null] +>> endobj +2694 0 obj << +/D [2753 0 R /XYZ 286.325 321.194 null] +>> endobj +2695 0 obj << +/D [2753 0 R /XYZ 340.446 321.194 null] +>> endobj +2696 0 obj << +/D [2753 0 R /XYZ 372.725 321.194 null] +>> endobj +2697 0 obj << +/D [2753 0 R /XYZ 421.143 321.194 null] +>> endobj +2698 0 obj << +/D [2753 0 R /XYZ 474.941 321.194 null] +>> endobj +2699 0 obj << +/D [2753 0 R /XYZ 141.07 309.538 null] +>> endobj +2700 0 obj << +/D [2753 0 R /XYZ 264.806 309.538 null] +>> endobj +2701 0 obj << +/D [2753 0 R /XYZ 141.07 297.882 null] +>> endobj +2702 0 obj << +/D [2753 0 R /XYZ 270.186 297.882 null] +>> endobj +2703 0 obj << +/D [2753 0 R /XYZ 141.07 286.226 null] +>> endobj +2704 0 obj << +/D [2753 0 R /XYZ 264.806 286.226 null] +>> endobj +2705 0 obj << +/D [2753 0 R /XYZ 141.07 274.569 null] +>> endobj +2706 0 obj << +/D [2753 0 R /XYZ 270.186 274.569 null] +>> endobj +2707 0 obj << +/D [2753 0 R /XYZ 141.07 262.913 null] +>> endobj +2708 0 obj << +/D [2753 0 R /XYZ 216.388 262.913 null] +>> endobj +2709 0 obj << +/D [2753 0 R /XYZ 141.07 251.257 null] +>> endobj +2710 0 obj << +/D [2753 0 R /XYZ 248.666 251.257 null] +>> endobj +2711 0 obj << +/D [2753 0 R /XYZ 141.07 239.601 null] +>> endobj +2712 0 obj << +/D [2753 0 R /XYZ 248.666 239.601 null] +>> endobj +2713 0 obj << +/D [2753 0 R /XYZ 141.07 227.944 null] +>> endobj +2714 0 obj << +/D [2753 0 R /XYZ 232.527 227.944 null] +>> endobj +2715 0 obj << +/D [2753 0 R /XYZ 141.07 216.288 null] +>> endobj +2716 0 obj << +/D [2753 0 R /XYZ 237.907 216.288 null] +>> endobj +2717 0 obj << +/D [2753 0 R /XYZ 141.07 204.632 null] +>> endobj +2718 0 obj << +/D [2753 0 R /XYZ 227.147 204.632 null] +>> endobj +2719 0 obj << +/D [2753 0 R /XYZ 141.07 192.976 null] +>> endobj +2720 0 obj << +/D [2753 0 R /XYZ 227.147 192.976 null] +>> endobj +2721 0 obj << +/D [2753 0 R /XYZ 141.07 181.32 null] +>> endobj +2722 0 obj << +/D [2753 0 R /XYZ 243.287 181.32 null] +>> endobj +2723 0 obj << +/D [2753 0 R /XYZ 141.07 169.663 null] +>> endobj +2724 0 obj << +/D [2753 0 R /XYZ 243.287 169.663 null] +>> endobj +2725 0 obj << +/D [2753 0 R /XYZ 302.787 169.663 null] +>> endobj +2726 0 obj << +/D [2753 0 R /XYZ 318.389 169.663 null] +>> endobj +2727 0 obj << +/D [2753 0 R /XYZ 350.667 169.663 null] +>> endobj +2728 0 obj << +/D [2753 0 R /XYZ 141.07 158.007 null] +>> endobj +2729 0 obj << +/D [2753 0 R /XYZ 259.426 158.007 null] +>> endobj +2730 0 obj << +/D [2753 0 R /XYZ 318.927 158.007 null] +>> endobj +2731 0 obj << +/D [2753 0 R /XYZ 334.528 158.007 null] +>> endobj +2732 0 obj << +/D [2753 0 R /XYZ 366.807 158.007 null] +>> endobj +2733 0 obj << +/D [2753 0 R /XYZ 141.07 146.351 null] +>> endobj +2734 0 obj << +/D [2753 0 R /XYZ 232.527 146.351 null] +>> endobj +2735 0 obj << +/D [2753 0 R /XYZ 292.028 146.351 null] +>> endobj +2736 0 obj << +/D [2753 0 R /XYZ 307.629 146.351 null] +>> endobj +2737 0 obj << +/D [2753 0 R /XYZ 339.908 146.351 null] +>> endobj +2738 0 obj << +/D [2753 0 R /XYZ 141.07 134.695 null] +>> endobj +2739 0 obj << +/D [2753 0 R /XYZ 248.666 134.695 null] +>> endobj +2740 0 obj << +/D [2753 0 R /XYZ 308.167 134.695 null] +>> endobj +2741 0 obj << +/D [2753 0 R /XYZ 323.768 134.695 null] +>> endobj +2742 0 obj << +/D [2753 0 R /XYZ 356.047 134.695 null] +>> endobj +2752 0 obj << +/Font << /F31 805 0 R /F37 1210 0 R /F60 1503 0 R /F52 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2834 0 obj << +/Length 2835 0 R +/Filter /FlateDecode +>> +stream +xX[o6_GY.omW{* +h[,y6w$Rr[` wx<$IGEbHf&IK2~o~6)& L_5֒%0o^|[-"Io?ޏkծ%rf\v +:?q@)p0bZbM` V;pVI1OHT]zQ "K!Iu彟W?$ ۅ):F 3IGmӺ;1:0x'>{ZoĞ*v +v*vYߌ3'G1Q}٢Y>dGRmlDNAN O1YWy8И5OYt[? Fa )bRƙ]"f S(his&Ur,3eqVeBe 뫩s:{mMq Z= _`A\w dC@X]m Lw iNuqzQ{չ BO% '*} +2TQ\!p(%pQVXg$NA*P +_֨m=TktD׻#uoE.&ǔ&ۇ∬۶ZSYVc02Xl^۵}[d:mMheē3ͣ[]BP^#WnED/> "svA DףkuףSDq dl*l 2;R_L;W+_CbH+S]"KqP<؍18.>j툡*ƺΏv[\U2 plX"|?B[l.þg&j6ߺ  WL3NQ25 צ1Mv #eح?oE΅2Ζ|M!'ڟcr[o%/[YUZXLp ўw# >]6!vnB{HGsg1$o2Tv:Dyށo^jrn>4\.>Sҩٷeendstream +endobj +2835 0 obj +1356 +endobj +2833 0 obj << +/Type /Page +/Contents 2834 0 R +/Resources 2832 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2551 0 R +>> endobj +2831 0 obj << +/D [2833 0 R /XYZ 71.731 705.352 null] +>> endobj +2743 0 obj << +/D [2833 0 R /XYZ 141.07 647.569 null] +>> endobj +2744 0 obj << +/D [2833 0 R /XYZ 167.969 635.913 null] +>> endobj +2745 0 obj << +/D [2833 0 R /XYZ 211.331 635.913 null] +>> endobj +2746 0 obj << +/D [2833 0 R /XYZ 243.609 635.913 null] +>> endobj +2747 0 obj << +/D [2833 0 R /XYZ 302.787 635.913 null] +>> endobj +2748 0 obj << +/D [2833 0 R /XYZ 313.547 635.913 null] +>> endobj +2749 0 obj << +/D [2833 0 R /XYZ 356.908 635.913 null] +>> endobj +2750 0 obj << +/D [2833 0 R /XYZ 389.187 635.913 null] +>> endobj +2756 0 obj << +/D [2833 0 R /XYZ 141.07 624.257 null] +>> endobj +2757 0 obj << +/D [2833 0 R /XYZ 248.666 624.257 null] +>> endobj +2758 0 obj << +/D [2833 0 R /XYZ 141.07 612.601 null] +>> endobj +2759 0 obj << +/D [2833 0 R /XYZ 200.248 612.601 null] +>> endobj +2760 0 obj << +/D [2833 0 R /XYZ 141.07 600.944 null] +>> endobj +2761 0 obj << +/D [2833 0 R /XYZ 200.248 600.944 null] +>> endobj +2762 0 obj << +/D [2833 0 R /XYZ 141.07 589.288 null] +>> endobj +2763 0 obj << +/D [2833 0 R /XYZ 184.109 589.288 null] +>> endobj +2764 0 obj << +/D [2833 0 R /XYZ 141.07 577.632 null] +>> endobj +2765 0 obj << +/D [2833 0 R /XYZ 205.628 577.632 null] +>> endobj +2766 0 obj << +/D [2833 0 R /XYZ 141.07 565.976 null] +>> endobj +2767 0 obj << +/D [2833 0 R /XYZ 248.666 565.976 null] +>> endobj +2768 0 obj << +/D [2833 0 R /XYZ 141.07 554.319 null] +>> endobj +2769 0 obj << +/D [2833 0 R /XYZ 254.046 554.319 null] +>> endobj +2770 0 obj << +/D [2833 0 R /XYZ 141.07 542.663 null] +>> endobj +2771 0 obj << +/D [2833 0 R /XYZ 254.046 542.663 null] +>> endobj +2772 0 obj << +/D [2833 0 R /XYZ 141.07 531.007 null] +>> endobj +2773 0 obj << +/D [2833 0 R /XYZ 259.426 531.007 null] +>> endobj +2774 0 obj << +/D [2833 0 R /XYZ 141.07 519.351 null] +>> endobj +2775 0 obj << +/D [2833 0 R /XYZ 211.008 519.351 null] +>> endobj +2776 0 obj << +/D [2833 0 R /XYZ 141.07 507.694 null] +>> endobj +2777 0 obj << +/D [2833 0 R /XYZ 211.008 507.694 null] +>> endobj +2778 0 obj << +/D [2833 0 R /XYZ 141.07 496.038 null] +>> endobj +2779 0 obj << +/D [2833 0 R /XYZ 227.147 496.038 null] +>> endobj +2780 0 obj << +/D [2833 0 R /XYZ 297.156 496.038 null] +>> endobj +2781 0 obj << +/D [2833 0 R /XYZ 307.916 496.038 null] +>> endobj +2782 0 obj << +/D [2833 0 R /XYZ 141.07 484.382 null] +>> endobj +2783 0 obj << +/D [2833 0 R /XYZ 200.248 484.382 null] +>> endobj +2784 0 obj << +/D [2833 0 R /XYZ 141.07 472.726 null] +>> endobj +2785 0 obj << +/D [2833 0 R /XYZ 232.527 472.726 null] +>> endobj +2786 0 obj << +/D [2833 0 R /XYZ 141.07 461.069 null] +>> endobj +2787 0 obj << +/D [2833 0 R /XYZ 227.147 461.069 null] +>> endobj +2788 0 obj << +/D [2833 0 R /XYZ 141.07 449.413 null] +>> endobj +2789 0 obj << +/D [2833 0 R /XYZ 270.186 449.413 null] +>> endobj +2790 0 obj << +/D [2833 0 R /XYZ 141.07 437.757 null] +>> endobj +2791 0 obj << +/D [2833 0 R /XYZ 205.628 437.757 null] +>> endobj +2792 0 obj << +/D [2833 0 R /XYZ 141.07 426.101 null] +>> endobj +2793 0 obj << +/D [2833 0 R /XYZ 200.248 426.101 null] +>> endobj +2794 0 obj << +/D [2833 0 R /XYZ 141.07 414.444 null] +>> endobj +2795 0 obj << +/D [2833 0 R /XYZ 221.767 414.444 null] +>> endobj +2796 0 obj << +/D [2833 0 R /XYZ 141.07 402.788 null] +>> endobj +2797 0 obj << +/D [2833 0 R /XYZ 221.767 402.788 null] +>> endobj +2798 0 obj << +/D [2833 0 R /XYZ 141.07 391.132 null] +>> endobj +2799 0 obj << +/D [2833 0 R /XYZ 248.666 391.132 null] +>> endobj +2800 0 obj << +/D [2833 0 R /XYZ 141.07 379.476 null] +>> endobj +2801 0 obj << +/D [2833 0 R /XYZ 243.287 379.476 null] +>> endobj +2802 0 obj << +/D [2833 0 R /XYZ 141.07 367.819 null] +>> endobj +2803 0 obj << +/D [2833 0 R /XYZ 243.287 367.819 null] +>> endobj +2804 0 obj << +/D [2833 0 R /XYZ 141.07 356.163 null] +>> endobj +2805 0 obj << +/D [2833 0 R /XYZ 232.527 356.163 null] +>> endobj +2806 0 obj << +/D [2833 0 R /XYZ 141.07 344.507 null] +>> endobj +2807 0 obj << +/D [2833 0 R /XYZ 232.527 344.507 null] +>> endobj +2808 0 obj << +/D [2833 0 R /XYZ 141.07 332.851 null] +>> endobj +2809 0 obj << +/D [2833 0 R /XYZ 178.729 332.851 null] +>> endobj +2810 0 obj << +/D [2833 0 R /XYZ 141.07 321.194 null] +>> endobj +2811 0 obj << +/D [2833 0 R /XYZ 264.806 321.194 null] +>> endobj +2812 0 obj << +/D [2833 0 R /XYZ 141.07 309.538 null] +>> endobj +2813 0 obj << +/D [2833 0 R /XYZ 270.186 309.538 null] +>> endobj +2814 0 obj << +/D [2833 0 R /XYZ 141.07 297.882 null] +>> endobj +2815 0 obj << +/D [2833 0 R /XYZ 227.147 297.882 null] +>> endobj +2816 0 obj << +/D [2833 0 R /XYZ 141.07 286.226 null] +>> endobj +2817 0 obj << +/D [2833 0 R /XYZ 232.527 286.226 null] +>> endobj +2818 0 obj << +/D [2833 0 R /XYZ 141.07 274.569 null] +>> endobj +2819 0 obj << +/D [2833 0 R /XYZ 275.565 274.569 null] +>> endobj +942 0 obj << +/D [2833 0 R /XYZ 71.731 241.939 null] +>> endobj +361 0 obj << +/D [2833 0 R /XYZ 391.402 213.578 null] +>> endobj +2820 0 obj << +/D [2833 0 R /XYZ 71.731 203.213 null] +>> endobj +2821 0 obj << +/D [2833 0 R /XYZ 137.534 193.454 null] +>> endobj +2822 0 obj << +/D [2833 0 R /XYZ 356.161 193.454 null] +>> endobj +2823 0 obj << +/D [2833 0 R /XYZ 360.166 167.551 null] +>> endobj +2824 0 obj << +/D [2833 0 R /XYZ 71.731 147.462 null] +>> endobj +2832 0 obj << +/Font << /F31 805 0 R /F37 1210 0 R /F60 1503 0 R /F40 669 0 R /F25 681 0 R /F30 731 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2889 0 obj << +/Length 2890 0 R +/Filter /FlateDecode +>> +stream +xX]61b̪ >6R"#UU[ x|Ld؆103TJZic_{V.y+!%'xϼ[{Sq"G$drͳo#Q6oY|v0qoT+RuukcוKy?6?>{LE.W`9 Q dwnk]vLuWP_l>Y0>Pvc2{ql ¼iy_Y)hC=C~A5zVq0dl$hW{l7e B=4צ&Eܶz#Qd8u@(bffS(otwzy;qqת+=Edf# cfrLbE]k7f3ͽO LҚ> 8nL(+ CSD |3ӣQ20Xm팊NJ"D|(1|.BҋJZ_T*NZq^ J#O( p8*5\^ıMl*fōPaWi=dNj +"96:[ %ƜWB3N/Yo8h0+&$6v[dq7^F44ǚ|tL Pc o+@ +\c.Ö} =ӭ>ˆ}&Sw(SjpAacjAIp~:a~Yc΅ ^QTu"O OMjևF7@7?9(ALP[Ah<^ީYc82gp+<YӾX WG&̓ LX[pʫV4]".5g ȠFS.umDP;K^1\ȓ,pM.Ld;|h3Go%so5ÇyEXf|K{sA71|oBɜ( h?C-wj . +\OO Qyy(x OQ>C+B3QP)1"E*]t5$A">]endstream +endobj +2890 0 obj +1683 +endobj +2888 0 obj << +/Type /Page +/Contents 2889 0 R +/Resources 2887 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2551 0 R +>> endobj +2886 0 obj << +/D [2888 0 R /XYZ 71.731 705.352 null] +>> endobj +2825 0 obj << +/D [2888 0 R /XYZ 71.731 657.532 null] +>> endobj +2826 0 obj << +/D [2888 0 R /XYZ 71.731 657.532 null] +>> endobj +2827 0 obj << +/D [2888 0 R /XYZ 71.731 647.45 null] +>> endobj +2828 0 obj << +/D [2888 0 R /XYZ 139.476 629.637 null] +>> endobj +2829 0 obj << +/D [2888 0 R /XYZ 367.429 629.637 null] +>> endobj +2830 0 obj << +/D [2888 0 R /XYZ 466.576 629.637 null] +>> endobj +2836 0 obj << +/D [2888 0 R /XYZ 71.731 606.623 null] +>> endobj +2837 0 obj << +/D [2888 0 R /XYZ 71.731 606.623 null] +>> endobj +2838 0 obj << +/D [2888 0 R /XYZ 71.731 591.749 null] +>> endobj +2839 0 obj << +/D [2888 0 R /XYZ 139.476 575.838 null] +>> endobj +2840 0 obj << +/D [2888 0 R /XYZ 486.161 562.887 null] +>> endobj +2841 0 obj << +/D [2888 0 R /XYZ 461.255 549.936 null] +>> endobj +2842 0 obj << +/D [2888 0 R /XYZ 71.731 524.865 null] +>> endobj +2843 0 obj << +/D [2888 0 R /XYZ 71.731 524.865 null] +>> endobj +2844 0 obj << +/D [2888 0 R /XYZ 71.731 513.951 null] +>> endobj +2845 0 obj << +/D [2888 0 R /XYZ 139.476 496.138 null] +>> endobj +2846 0 obj << +/D [2888 0 R /XYZ 139.476 496.138 null] +>> endobj +2847 0 obj << +/D [2888 0 R /XYZ 291.713 496.138 null] +>> endobj +2848 0 obj << +/D [2888 0 R /XYZ 448.324 496.138 null] +>> endobj +2849 0 obj << +/D [2888 0 R /XYZ 150.266 483.186 null] +>> endobj +2850 0 obj << +/D [2888 0 R /XYZ 234.648 483.186 null] +>> endobj +2851 0 obj << +/D [2888 0 R /XYZ 71.731 460.173 null] +>> endobj +2852 0 obj << +/D [2888 0 R /XYZ 71.731 460.173 null] +>> endobj +2853 0 obj << +/D [2888 0 R /XYZ 71.731 445.299 null] +>> endobj +2854 0 obj << +/D [2888 0 R /XYZ 139.476 429.388 null] +>> endobj +2855 0 obj << +/D [2888 0 R /XYZ 417.012 429.388 null] +>> endobj +2856 0 obj << +/D [2888 0 R /XYZ 181.528 416.437 null] +>> endobj +2857 0 obj << +/D [2888 0 R /XYZ 268.073 403.485 null] +>> endobj +2858 0 obj << +/D [2888 0 R /XYZ 399.778 403.485 null] +>> endobj +2859 0 obj << +/D [2888 0 R /XYZ 268.073 390.534 null] +>> endobj +2860 0 obj << +/D [2888 0 R /XYZ 378.259 390.534 null] +>> endobj +2861 0 obj << +/D [2888 0 R /XYZ 139.476 377.583 null] +>> endobj +2862 0 obj << +/D [2888 0 R /XYZ 71.731 367.501 null] +>> endobj +2863 0 obj << +/D [2888 0 R /XYZ 71.731 367.501 null] +>> endobj +2864 0 obj << +/D [2888 0 R /XYZ 71.731 352.646 null] +>> endobj +2865 0 obj << +/D [2888 0 R /XYZ 139.476 336.736 null] +>> endobj +2866 0 obj << +/D [2888 0 R /XYZ 238.006 323.785 null] +>> endobj +2867 0 obj << +/D [2888 0 R /XYZ 71.731 298.714 null] +>> endobj +2868 0 obj << +/D [2888 0 R /XYZ 71.731 298.714 null] +>> endobj +2869 0 obj << +/D [2888 0 R /XYZ 71.731 285.897 null] +>> endobj +2870 0 obj << +/D [2888 0 R /XYZ 139.476 269.987 null] +>> endobj +2871 0 obj << +/D [2888 0 R /XYZ 491.561 269.987 null] +>> endobj +2872 0 obj << +/D [2888 0 R /XYZ 299.056 257.035 null] +>> endobj +2873 0 obj << +/D [2888 0 R /XYZ 71.731 244.916 null] +>> endobj +2874 0 obj << +/D [2888 0 R /XYZ 71.731 244.916 null] +>> endobj +2875 0 obj << +/D [2888 0 R /XYZ 71.731 234.002 null] +>> endobj +2876 0 obj << +/D [2888 0 R /XYZ 139.476 216.189 null] +>> endobj +2877 0 obj << +/D [2888 0 R /XYZ 178.669 203.237 null] +>> endobj +2878 0 obj << +/D [2888 0 R /XYZ 71.731 193.086 null] +>> endobj +2879 0 obj << +/D [2888 0 R /XYZ 71.731 188.105 null] +>> endobj +2880 0 obj << +/D [2888 0 R /XYZ 278.805 176.637 null] +>> endobj +2887 0 obj << +/Font << /F31 805 0 R /F30 731 0 R /F25 681 0 R /F37 1210 0 R /F62 669 0 R /F58 2248 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2933 0 obj << +/Length 2934 0 R +/Filter /FlateDecode +>> +stream +xYIo7|̯Э2`g_Sb h@PP%9v_uV9n/ Z0|‡`^O( U ;z0 䙜4ZoԼf0񗩧ooЏ+ڜm[vK$m׏x+lq㫛{'B拕9Wq{qX%Q'?@Y=9_+Nj 9>9߬ xS2T6; k޼э~ +ԫ%. YZ=1"0c"hy5#q9B W*9%jk3piepCAT +BJ.dms)<^7Bz9-ad biSϖJGᖡ=ׂ2>)&Q7f='|37ސf+E^Z@ +9 I@|m.W AU X- R9iepgtf:i Z.X[ +;f{jb6Eg^TDF &i;O^E + +Q^@JXxYIQ_Eԏͱ:xNinfftZg$>a˖1l" yNoCӬ{A +JBkyz $==?!>Un/lЃ?F'U drD1.Hl0\>[h(LQ&~R \&M0Y>ʠa,Dz}ɥkAKZge} +WJ}TEz&[l?C bJ>Z( +ө3^N]ٕ1^;Xy <Z\jr\I!%&>E趦kT;뿰<1pݯ*,5|$S&'A/UuN[V>FGǡI Z1ޫ7 c^dk0 j +x zLjn + oKedw<Ǒʈ.#Tbj]{4jzFT6 PcWtVO|5x -ROz2-P^P5[ʒ}"wԅ֘v?[t‰\tֱ)YXQZ->%,r +?^ +{KIRRl(Y'lr`@q1 &T^$姑&@&mêlEICA 0Bhė55i\11Qrhg!mC L=o|8@n{Ab/Δ'Ig-5Vu ^ޛ @^^hQ+v3M+A ]~MrZŢ׉B#ڊf^O[ 7of2h .mѦ^,hOon4 Al N +z mN.AMiQ eL@yeD6pqftsBBc[:c1m=W ke\aQ9֬gxB팑LuuZAnh`N7B2أ1Q- ./tQb z>08Jf +&;6/)Em{7/w|ppDnbF.Dx㓊~\2}E\KsdG/Jp.ʊC`j*< ˏSQ/)6nU2ǔ.{ T2(@Ƨ8\֤yCǺe T/$۝/vjȴ)J4Hؔ KLIچi gAE 93wUks*Kʺ=cԭk>I# 4*]?oK@QyifVi"+ꋰB!NY/cJyendstream +endobj +2934 0 obj +1896 +endobj +2932 0 obj << +/Type /Page +/Contents 2933 0 R +/Resources 2931 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2936 0 R +/Annots [ 2935 0 R ] +>> endobj +2935 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [199.52 536.885 269.258 543.739] +/Subtype /Link +/A << /S /GoTo /D (statsfile) >> +>> endobj +2930 0 obj << +/D [2932 0 R /XYZ 71.731 705.352 null] +>> endobj +2881 0 obj << +/D [2932 0 R /XYZ 71.731 657.532 null] +>> endobj +2882 0 obj << +/D [2932 0 R /XYZ 71.731 657.532 null] +>> endobj +2883 0 obj << +/D [2932 0 R /XYZ 71.731 645.547 null] +>> endobj +2884 0 obj << +/D [2932 0 R /XYZ 139.476 629.637 null] +>> endobj +2885 0 obj << +/D [2932 0 R /XYZ 139.476 616.685 null] +>> endobj +2891 0 obj << +/D [2932 0 R /XYZ 71.731 593.672 null] +>> endobj +2892 0 obj << +/D [2932 0 R /XYZ 71.731 593.672 null] +>> endobj +2893 0 obj << +/D [2932 0 R /XYZ 71.731 580.7 null] +>> endobj +2894 0 obj << +/D [2932 0 R /XYZ 139.476 562.887 null] +>> endobj +2895 0 obj << +/D [2932 0 R /XYZ 486.869 562.887 null] +>> endobj +2896 0 obj << +/D [2932 0 R /XYZ 258.08 549.936 null] +>> endobj +2897 0 obj << +/D [2932 0 R /XYZ 71.731 526.922 null] +>> endobj +2898 0 obj << +/D [2932 0 R /XYZ 71.731 526.922 null] +>> endobj +2899 0 obj << +/D [2932 0 R /XYZ 71.731 512.048 null] +>> endobj +2900 0 obj << +/D [2932 0 R /XYZ 139.476 496.138 null] +>> endobj +2901 0 obj << +/D [2932 0 R /XYZ 71.731 458.115 null] +>> endobj +2902 0 obj << +/D [2932 0 R /XYZ 71.731 458.115 null] +>> endobj +2903 0 obj << +/D [2932 0 R /XYZ 71.731 447.201 null] +>> endobj +2904 0 obj << +/D [2932 0 R /XYZ 139.476 429.388 null] +>> endobj +2905 0 obj << +/D [2932 0 R /XYZ 417.6 390.534 null] +>> endobj +2906 0 obj << +/D [2932 0 R /XYZ 297.731 377.583 null] +>> endobj +943 0 obj << +/D [2932 0 R /XYZ 71.731 352.512 null] +>> endobj +365 0 obj << +/D [2932 0 R /XYZ 265.055 330.161 null] +>> endobj +2907 0 obj << +/D [2932 0 R /XYZ 71.731 321.708 null] +>> endobj +2908 0 obj << +/D [2932 0 R /XYZ 71.731 311.746 null] +>> endobj +2909 0 obj << +/D [2932 0 R /XYZ 71.731 311.746 null] +>> endobj +2910 0 obj << +/D [2932 0 R /XYZ 71.731 301.15 null] +>> endobj +2911 0 obj << +/D [2932 0 R /XYZ 139.476 283.336 null] +>> endobj +2912 0 obj << +/D [2932 0 R /XYZ 148.602 283.336 null] +>> endobj +2913 0 obj << +/D [2932 0 R /XYZ 205.827 283.336 null] +>> endobj +2914 0 obj << +/D [2932 0 R /XYZ 251.594 270.385 null] +>> endobj +2915 0 obj << +/D [2932 0 R /XYZ 354.308 270.385 null] +>> endobj +2916 0 obj << +/D [2932 0 R /XYZ 279.111 257.434 null] +>> endobj +2917 0 obj << +/D [2932 0 R /XYZ 71.731 245.314 null] +>> endobj +2918 0 obj << +/D [2932 0 R /XYZ 71.731 245.314 null] +>> endobj +2919 0 obj << +/D [2932 0 R /XYZ 71.731 234.4 null] +>> endobj +2920 0 obj << +/D [2932 0 R /XYZ 139.476 216.587 null] +>> endobj +2921 0 obj << +/D [2932 0 R /XYZ 236.511 216.587 null] +>> endobj +2922 0 obj << +/D [2932 0 R /XYZ 453.007 216.587 null] +>> endobj +2923 0 obj << +/D [2932 0 R /XYZ 71.731 191.516 null] +>> endobj +2924 0 obj << +/D [2932 0 R /XYZ 71.731 191.516 null] +>> endobj +2925 0 obj << +/D [2932 0 R /XYZ 71.731 178.699 null] +>> endobj +2926 0 obj << +/D [2932 0 R /XYZ 139.476 162.789 null] +>> endobj +2927 0 obj << +/D [2932 0 R /XYZ 148.602 162.789 null] +>> endobj +2931 0 obj << +/Font << /F31 805 0 R /F30 731 0 R /F25 681 0 R /F37 1210 0 R /F42 669 0 R /F39 1242 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +2996 0 obj << +/Length 2997 0 R +/Filter /FlateDecode +>> +stream +xYI6r̯0r D]lMt9mEU2>V;395 L\>xޤ!|1 +`S᝞> +`8KWw|s'ϻ+ܾ>FvM|usFrώ]ɖHk==0Һ~zAi x5^g((B9Y ?xK +WU?p=2%Got{JoþZöz4Gib֑ԇxC]U=DzkY}cyD Ue|#'#5MKNDˏgs8ʓVA,\y,д\mEfo ܭLkZ 6ŧ|ZA͍Z7=[R5UĬx10,i"lŊZ[zO-1^ۈ߉leO9WN.(fa k-@^RlPR8T Ē.AwvkʺzIqbĀ_' iq,pN*/fY7?rՃzԕŊ'g`9R=G( p4!BO+%KN%K54 Ǯy*dNXvkLttq(Ba\a p% p?flkO6Jt-! +v +m`>6Ig}5"E$ؓQI:KkXj$͎{b;2 pqv|}xUzET]wHF&@UcEBSB]gӳr,vneTv{7Ap4@omN? A6XZe􌗣Ģ]ܐhU!V:aۦD'O^6E |ѵ?,"S Q(v]>!))-Q޸0!Yq~N0̃ξh. +Ġ#Qtݐ@0A} 2]mɘ-+#|BHڷ3c}۫;Xk?.X^:i%xK>K8=kCA$|ʖԂvŏ+j>6r ]cNEs&f.tOӪd&0ٓÁ8DqkTO1zEΎendstream +endobj +2997 0 obj +1891 +endobj +2995 0 obj << +/Type /Page +/Contents 2996 0 R +/Resources 2994 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2936 0 R +>> endobj +2993 0 obj << +/D [2995 0 R /XYZ 71.731 705.352 null] +>> endobj +2928 0 obj << +/D [2995 0 R /XYZ 220.163 647.569 null] +>> endobj +2929 0 obj << +/D [2995 0 R /XYZ 355.483 634.618 null] +>> endobj +2937 0 obj << +/D [2995 0 R /XYZ 71.731 629.437 null] +>> endobj +2938 0 obj << +/D [2995 0 R /XYZ 157.459 616.685 null] +>> endobj +2939 0 obj << +/D [2995 0 R /XYZ 327.937 616.685 null] +>> endobj +2940 0 obj << +/D [2995 0 R /XYZ 366.671 616.685 null] +>> endobj +2941 0 obj << +/D [2995 0 R /XYZ 181.538 603.734 null] +>> endobj +2942 0 obj << +/D [2995 0 R /XYZ 71.731 596.596 null] +>> endobj +2943 0 obj << +/D [2995 0 R /XYZ 71.731 552.76 null] +>> endobj +2944 0 obj << +/D [2995 0 R /XYZ 340.331 529.014 null] +>> endobj +2945 0 obj << +/D [2995 0 R /XYZ 71.731 508.925 null] +>> endobj +2946 0 obj << +/D [2995 0 R /XYZ 295.958 498.13 null] +>> endobj +2947 0 obj << +/D [2995 0 R /XYZ 139.476 485.179 null] +>> endobj +2948 0 obj << +/D [2995 0 R /XYZ 139.476 472.227 null] +>> endobj +2949 0 obj << +/D [2995 0 R /XYZ 139.476 459.276 null] +>> endobj +2950 0 obj << +/D [2995 0 R /XYZ 267.226 459.276 null] +>> endobj +2951 0 obj << +/D [2995 0 R /XYZ 71.731 431.381 null] +>> endobj +2952 0 obj << +/D [2995 0 R /XYZ 71.731 431.381 null] +>> endobj +2953 0 obj << +/D [2995 0 R /XYZ 71.731 416.407 null] +>> endobj +2954 0 obj << +/D [2995 0 R /XYZ 139.476 400.497 null] +>> endobj +2955 0 obj << +/D [2995 0 R /XYZ 150.266 400.497 null] +>> endobj +2956 0 obj << +/D [2995 0 R /XYZ 512.015 400.497 null] +>> endobj +2957 0 obj << +/D [2995 0 R /XYZ 71.731 375.799 null] +>> endobj +2958 0 obj << +/D [2995 0 R /XYZ 71.731 375.799 null] +>> endobj +2959 0 obj << +/D [2995 0 R /XYZ 71.731 362.609 null] +>> endobj +2960 0 obj << +/D [2995 0 R /XYZ 139.476 346.699 null] +>> endobj +2961 0 obj << +/D [2995 0 R /XYZ 282.548 346.699 null] +>> endobj +2962 0 obj << +/D [2995 0 R /XYZ 71.731 310.614 null] +>> endobj +2963 0 obj << +/D [2995 0 R /XYZ 71.731 310.614 null] +>> endobj +2964 0 obj << +/D [2995 0 R /XYZ 71.731 297.762 null] +>> endobj +2965 0 obj << +/D [2995 0 R /XYZ 139.476 279.949 null] +>> endobj +2966 0 obj << +/D [2995 0 R /XYZ 315.922 279.949 null] +>> endobj +2967 0 obj << +/D [2995 0 R /XYZ 424.115 279.949 null] +>> endobj +2968 0 obj << +/D [2995 0 R /XYZ 211.984 266.998 null] +>> endobj +2969 0 obj << +/D [2995 0 R /XYZ 275.913 266.998 null] +>> endobj +2970 0 obj << +/D [2995 0 R /XYZ 429.046 266.998 null] +>> endobj +2971 0 obj << +/D [2995 0 R /XYZ 496.861 266.998 null] +>> endobj +2972 0 obj << +/D [2995 0 R /XYZ 139.476 254.046 null] +>> endobj +2973 0 obj << +/D [2995 0 R /XYZ 200.617 254.046 null] +>> endobj +2974 0 obj << +/D [2995 0 R /XYZ 71.731 242.061 null] +>> endobj +2975 0 obj << +/D [2995 0 R /XYZ 71.731 242.061 null] +>> endobj +2976 0 obj << +/D [2995 0 R /XYZ 71.731 231.013 null] +>> endobj +2977 0 obj << +/D [2995 0 R /XYZ 139.476 213.2 null] +>> endobj +2978 0 obj << +/D [2995 0 R /XYZ 71.731 188.129 null] +>> endobj +2979 0 obj << +/D [2995 0 R /XYZ 71.731 188.129 null] +>> endobj +2980 0 obj << +/D [2995 0 R /XYZ 71.731 177.215 null] +>> endobj +2981 0 obj << +/D [2995 0 R /XYZ 139.476 159.402 null] +>> endobj +2982 0 obj << +/D [2995 0 R /XYZ 139.476 159.402 null] +>> endobj +2983 0 obj << +/D [2995 0 R /XYZ 291.863 159.402 null] +>> endobj +2984 0 obj << +/D [2995 0 R /XYZ 250.339 146.45 null] +>> endobj +2985 0 obj << +/D [2995 0 R /XYZ 333.984 133.499 null] +>> endobj +2986 0 obj << +/D [2995 0 R /XYZ 387.662 133.499 null] +>> endobj +2994 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F30 731 0 R /F39 1242 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3045 0 obj << +/Length 3046 0 R +/Filter /FlateDecode +>> +stream +xYK6_[m fEԣ95"-m\^+K(w_zwS䰲Dr^|3U*(OP*;+~~wv5IDY%^z?Eiv{n~su+͖`!)Tݩh[Ygae&6~}nש$Ym;.)8A4\mY$U*e)؂:Z*PSjcO +T??` [a[0g٘/mڃpxD)Ryg7z}D, +T>W3v\7X;h !"LDY-`~r' R +4Y,u!YɏbخDٷ^}誺gPcI>z3f3 /Ta6b;[{$Z2qU! +^:)C1Ny@^r;詾#z` 4a*GfNe%ȴx{[aߨZd.A"]D"S0xCs`þ#z݊#XҞ*:cƕSD͠lOuI(QXG0J{:з-E9OS763uR'WfT77~qBR[jGE@;WfFiBȥ컞.;擎I(1LVuul>}[Uݎёdxk'|%:El@z:׵L+O`ޜF Tm5Ã޸l\7O%IYۏzְ#Ф|\Wm:mKU Yy}RY0q\]C9!1ֱ-4IY!۾oOi˔-X=ZƳx5`Ob%Sɜ!,jϑ>P.xoh+Di(Dne" /zP{+\au`E7L"ǸWA 9V!s8A$Fԭ׭^Jԋ*[B$B$]Ȓ~!aѯ _ dD C4{auWKzF/tWlyI> endobj +3047 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [191.76 481.029 236.591 489.941] +/Subtype /Link +/A << /S /GoTo /D (notify) >> +>> endobj +3041 0 obj << +/D [3044 0 R /XYZ 71.731 705.352 null] +>> endobj +3042 0 obj << +/D [3044 0 R /XYZ 71.731 717.307 null] +>> endobj +2987 0 obj << +/D [3044 0 R /XYZ 71.731 657.532 null] +>> endobj +2988 0 obj << +/D [3044 0 R /XYZ 71.731 657.532 null] +>> endobj +2989 0 obj << +/D [3044 0 R /XYZ 71.731 645.547 null] +>> endobj +2990 0 obj << +/D [3044 0 R /XYZ 139.476 629.637 null] +>> endobj +2991 0 obj << +/D [3044 0 R /XYZ 148.602 629.637 null] +>> endobj +2992 0 obj << +/D [3044 0 R /XYZ 353.621 603.734 null] +>> endobj +2998 0 obj << +/D [3044 0 R /XYZ 71.731 591.614 null] +>> endobj +2999 0 obj << +/D [3044 0 R /XYZ 71.731 591.614 null] +>> endobj +3000 0 obj << +/D [3044 0 R /XYZ 71.731 578.797 null] +>> endobj +3001 0 obj << +/D [3044 0 R /XYZ 139.476 562.887 null] +>> endobj +3002 0 obj << +/D [3044 0 R /XYZ 236.511 562.887 null] +>> endobj +3003 0 obj << +/D [3044 0 R /XYZ 267.873 549.936 null] +>> endobj +3004 0 obj << +/D [3044 0 R /XYZ 71.731 524.865 null] +>> endobj +3005 0 obj << +/D [3044 0 R /XYZ 71.731 524.865 null] +>> endobj +3006 0 obj << +/D [3044 0 R /XYZ 71.731 512.048 null] +>> endobj +3007 0 obj << +/D [3044 0 R /XYZ 139.476 496.138 null] +>> endobj +3008 0 obj << +/D [3044 0 R /XYZ 148.602 496.138 null] +>> endobj +3009 0 obj << +/D [3044 0 R /XYZ 139.476 457.284 null] +>> endobj +3010 0 obj << +/D [3044 0 R /XYZ 71.731 450.145 null] +>> endobj +3011 0 obj << +/D [3044 0 R /XYZ 148.602 439.351 null] +>> endobj +3012 0 obj << +/D [3044 0 R /XYZ 415.567 439.351 null] +>> endobj +3013 0 obj << +/D [3044 0 R /XYZ 474.505 439.351 null] +>> endobj +3014 0 obj << +/D [3044 0 R /XYZ 71.731 421.319 null] +>> endobj +3015 0 obj << +/D [3044 0 R /XYZ 157.316 408.467 null] +>> endobj +3016 0 obj << +/D [3044 0 R /XYZ 324.421 408.467 null] +>> endobj +3017 0 obj << +/D [3044 0 R /XYZ 506.977 408.467 null] +>> endobj +3018 0 obj << +/D [3044 0 R /XYZ 71.731 367.62 null] +>> endobj +3019 0 obj << +/D [3044 0 R /XYZ 71.731 367.62 null] +>> endobj +3020 0 obj << +/D [3044 0 R /XYZ 71.731 354.549 null] +>> endobj +3021 0 obj << +/D [3044 0 R /XYZ 139.476 336.736 null] +>> endobj +3022 0 obj << +/D [3044 0 R /XYZ 148.602 336.736 null] +>> endobj +3023 0 obj << +/D [3044 0 R /XYZ 301.487 310.833 null] +>> endobj +3024 0 obj << +/D [3044 0 R /XYZ 392.753 310.833 null] +>> endobj +3025 0 obj << +/D [3044 0 R /XYZ 282.409 271.979 null] +>> endobj +3026 0 obj << +/D [3044 0 R /XYZ 71.731 259.86 null] +>> endobj +3027 0 obj << +/D [3044 0 R /XYZ 71.731 259.86 null] +>> endobj +3028 0 obj << +/D [3044 0 R /XYZ 71.731 247.043 null] +>> endobj +3029 0 obj << +/D [3044 0 R /XYZ 139.476 231.132 null] +>> endobj +3030 0 obj << +/D [3044 0 R /XYZ 197.328 231.132 null] +>> endobj +3031 0 obj << +/D [3044 0 R /XYZ 233.722 218.181 null] +>> endobj +3032 0 obj << +/D [3044 0 R /XYZ 71.731 208.019 null] +>> endobj +3033 0 obj << +/D [3044 0 R /XYZ 71.731 203.038 null] +>> endobj +3034 0 obj << +/D [3044 0 R /XYZ 278.805 191.581 null] +>> endobj +3043 0 obj << +/Font << /F31 805 0 R /F30 731 0 R /F25 681 0 R /F39 1242 0 R /F62 669 0 R /F58 2248 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3095 0 obj << +/Length 3096 0 R +/Filter /FlateDecode +>> +stream +xXn6iQŅ[ #q0 ܬ[L<""3hk LobG(m,xd82^ot71U (wsL&j%LM "#fM+!ؿ.ȡNr#{F TKm{ as؞R4.2}Aqx7ZrXhrl'm".T +V1ȉmVI~kpqٷT+;#[OTTnSx3~Hg5U<7fzω^[%_hz!9Doť>s-ă˙3pbx}Xmk}",S"UēzdWP6b;)dv_sAzAyd}djRy;|iҗPzEgakPrk xol 95TkbĚ@ƅ2H]֙ÞErfZArePi} +t-oO<* /n7azj´X-#NKݍ@j΀졻sq}pc͟GQ qmޗۑmsJK6uN 8RӠo)Ck=*_&:(X=L;:۱ꢪ\&K 5`2~'hilmoΗ`aMK,ܡ\K[Cʄ fto/Wg)Nj.73kbv1i)XVC1u1,KJ +d튶 ۾_r( +q рDXw./j<{ʣk)j\ev +\JATtY3{f1֞kA%J9rKʁ44: +-V.L ߘV{Y5LyE"?ȡl/^~^N1=<,O[Nm39Z Th:I|(qz͎owC+H1ⱯK!?EXk灰IZf! ĐjbFZ*cj)hR,J7>) TIƟ˾ +,Lg6)g9 +endstream +endobj +3096 0 obj +1941 +endobj +3094 0 obj << +/Type /Page +/Contents 3095 0 R +/Resources 3093 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2936 0 R +/Annots [ 3097 0 R 3098 0 R 3099 0 R 3100 0 R 3101 0 R ] +>> endobj +3097 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [158.017 562.788 227.754 569.642] +/Subtype /Link +/A << /S /GoTo /D (statsfile) >> +>> endobj +3098 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [308.002 506.932 365.287 515.953] +/Subtype /Link +/A << /S /GoTo /D (server_statement_definition_and_usage) >> +>> endobj +3099 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [405.685 506.932 450.516 515.953] +/Subtype /Link +/A << /S /GoTo /D (incremental_zone_transfers) >> +>> endobj +3100 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [294.981 466.086 352.266 475.107] +/Subtype /Link +/A << /S /GoTo /D (server_statement_definition_and_usage) >> +>> endobj +3101 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [293.965 425.239 351.249 434.26] +/Subtype /Link +/A << /S /GoTo /D (server_statement_definition_and_usage) >> +>> endobj +3091 0 obj << +/D [3094 0 R /XYZ 71.731 705.352 null] +>> endobj +3092 0 obj << +/D [3094 0 R /XYZ 71.731 717.307 null] +>> endobj +3035 0 obj << +/D [3094 0 R /XYZ 71.731 657.532 null] +>> endobj +3036 0 obj << +/D [3094 0 R /XYZ 71.731 657.532 null] +>> endobj +3037 0 obj << +/D [3094 0 R /XYZ 71.731 645.547 null] +>> endobj +3038 0 obj << +/D [3094 0 R /XYZ 139.476 629.637 null] +>> endobj +3039 0 obj << +/D [3094 0 R /XYZ 139.476 629.637 null] +>> endobj +3040 0 obj << +/D [3094 0 R /XYZ 234.409 629.637 null] +>> endobj +3048 0 obj << +/D [3094 0 R /XYZ 71.731 606.603 null] +>> endobj +3049 0 obj << +/D [3094 0 R /XYZ 139.476 588.79 null] +>> endobj +3050 0 obj << +/D [3094 0 R /XYZ 148.602 588.79 null] +>> endobj +3051 0 obj << +/D [3094 0 R /XYZ 232.437 575.838 null] +>> endobj +3052 0 obj << +/D [3094 0 R /XYZ 460.597 575.838 null] +>> endobj +3053 0 obj << +/D [3094 0 R /XYZ 71.731 552.825 null] +>> endobj +3054 0 obj << +/D [3094 0 R /XYZ 71.731 552.825 null] +>> endobj +3055 0 obj << +/D [3094 0 R /XYZ 71.731 539.854 null] +>> endobj +3056 0 obj << +/D [3094 0 R /XYZ 139.476 522.04 null] +>> endobj +3057 0 obj << +/D [3094 0 R /XYZ 139.476 522.04 null] +>> endobj +3058 0 obj << +/D [3094 0 R /XYZ 216.128 509.089 null] +>> endobj +3059 0 obj << +/D [3094 0 R /XYZ 71.731 496.97 null] +>> endobj +3060 0 obj << +/D [3094 0 R /XYZ 71.731 496.97 null] +>> endobj +3061 0 obj << +/D [3094 0 R /XYZ 71.731 484.153 null] +>> endobj +3062 0 obj << +/D [3094 0 R /XYZ 139.476 468.242 null] +>> endobj +3063 0 obj << +/D [3094 0 R /XYZ 231.062 468.242 null] +>> endobj +3064 0 obj << +/D [3094 0 R /XYZ 71.731 456.123 null] +>> endobj +3065 0 obj << +/D [3094 0 R /XYZ 71.731 456.123 null] +>> endobj +3066 0 obj << +/D [3094 0 R /XYZ 71.731 443.306 null] +>> endobj +3067 0 obj << +/D [3094 0 R /XYZ 139.476 427.396 null] +>> endobj +3068 0 obj << +/D [3094 0 R /XYZ 231.062 427.396 null] +>> endobj +3069 0 obj << +/D [3094 0 R /XYZ 71.731 415.276 null] +>> endobj +3070 0 obj << +/D [3094 0 R /XYZ 71.731 415.276 null] +>> endobj +3071 0 obj << +/D [3094 0 R /XYZ 71.731 402.459 null] +>> endobj +3072 0 obj << +/D [3094 0 R /XYZ 139.476 386.549 null] +>> endobj +3073 0 obj << +/D [3094 0 R /XYZ 236.511 386.549 null] +>> endobj +3074 0 obj << +/D [3094 0 R /XYZ 434.805 386.549 null] +>> endobj +3075 0 obj << +/D [3094 0 R /XYZ 303.461 360.646 null] +>> endobj +3076 0 obj << +/D [3094 0 R /XYZ 391.88 360.646 null] +>> endobj +3077 0 obj << +/D [3094 0 R /XYZ 466.121 360.646 null] +>> endobj +3078 0 obj << +/D [3094 0 R /XYZ 71.731 335.575 null] +>> endobj +3079 0 obj << +/D [3094 0 R /XYZ 71.731 335.575 null] +>> endobj +3080 0 obj << +/D [3094 0 R /XYZ 71.731 324.661 null] +>> endobj +3081 0 obj << +/D [3094 0 R /XYZ 71.731 324.661 null] +>> endobj +3082 0 obj << +/D [3094 0 R /XYZ 71.731 311.71 null] +>> endobj +3083 0 obj << +/D [3094 0 R /XYZ 139.476 293.897 null] +>> endobj +3084 0 obj << +/D [3094 0 R /XYZ 71.731 273.807 null] +>> endobj +3085 0 obj << +/D [3094 0 R /XYZ 290 263.013 null] +>> endobj +3086 0 obj << +/D [3094 0 R /XYZ 71.731 178.166 null] +>> endobj +3087 0 obj << +/D [3094 0 R /XYZ 360.236 167.372 null] +>> endobj +3088 0 obj << +/D [3094 0 R /XYZ 143.541 154.42 null] +>> endobj +3093 0 obj << +/Font << /F31 805 0 R /F30 731 0 R /F25 681 0 R /F39 1242 0 R /F37 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3129 0 obj << +/Length 3130 0 R +/Filter /FlateDecode +>> +stream +xYKo#78B%`1~ d2AF{hwSV-R!٣-s EV}UWxO)cT(͓M{~oƾ P$U)7|x=|FunGpj.>ɣmEJ=lF?[LZ΄,PT5i]<7DezM6ڜ/F yIఢ2ۯ0h~0헾Ý^i~ao#f]oĀ ѓ'8an9GPmTy@e^](cy8`{ TZ! X}ħfDe<"5=%T90af'ȑ2 8Q `拄)^3ƸWN('|鱂{?? 6 +9ʪ[Cp um|$X4_|w)C;E钭mNUq͊) KӰ.7&v$q?cΛ'<>+Hd9*ۛͰ?*Xy6ɘ=c3XsàqKh 98gF49<^2\0623!MLx[n p_0JZcR)GVR3Yΰ9#[nXW7dw>⵺F4`$N<__S*v85b~md->M_88G[ ({';kp (NS/&y;c25wj,ff-w9Q&q-Դ_h9D` ϋ + t{)1f\"S n`){#5e%;T=h7AYCyUfU^OV=b}pCO=LTf5p }u=htt?eQ9⡳ C,M%sٞ4^Y|^e[42G׀gWydM1 SX [퇰LI6!ㄔiO$0xxl:tUm/x$Ѝ[]-LdYc<,W,*LgPV{ʮ ڴs´fcΪ\[zbAu`̺Np#C7kASyV9UT.4Xsߵ+\J ϳVkJyfErЖ4fґ_քߤvt[`mh[3 a@Xܭ Wמ!bӹʁ!ڄX/\r}qI/7YZ J:Otyw=aX6;јϛ^>kCWչOZuwSu"'ëՇ,DW|b jkM^=,b)^> endobj +3126 0 obj << +/D [3128 0 R /XYZ 71.731 705.352 null] +>> endobj +3089 0 obj << +/D [3128 0 R /XYZ 139.476 647.569 null] +>> endobj +3090 0 obj << +/D [3128 0 R /XYZ 464.803 647.569 null] +>> endobj +3102 0 obj << +/D [3128 0 R /XYZ 343.438 603.734 null] +>> endobj +3103 0 obj << +/D [3128 0 R /XYZ 71.731 583.644 null] +>> endobj +3104 0 obj << +/D [3128 0 R /XYZ 185.135 572.85 null] +>> endobj +3105 0 obj << +/D [3128 0 R /XYZ 71.731 539.809 null] +>> endobj +3106 0 obj << +/D [3128 0 R /XYZ 369.262 490.16 null] +>> endobj +3107 0 obj << +/D [3128 0 R /XYZ 71.731 436.362 null] +>> endobj +3108 0 obj << +/D [3128 0 R /XYZ 71.731 436.362 null] +>> endobj +3109 0 obj << +/D [3128 0 R /XYZ 71.731 421.388 null] +>> endobj +3110 0 obj << +/D [3128 0 R /XYZ 139.476 405.478 null] +>> endobj +3111 0 obj << +/D [3128 0 R /XYZ 148.602 405.478 null] +>> endobj +944 0 obj << +/D [3128 0 R /XYZ 71.731 336.572 null] +>> endobj +369 0 obj << +/D [3128 0 R /XYZ 234.259 313.224 null] +>> endobj +3112 0 obj << +/D [3128 0 R /XYZ 71.731 304.587 null] +>> endobj +3113 0 obj << +/D [3128 0 R /XYZ 71.731 248.303 null] +>> endobj +3114 0 obj << +/D [3128 0 R /XYZ 71.731 238.34 null] +>> endobj +3115 0 obj << +/D [3128 0 R /XYZ 71.731 238.34 null] +>> endobj +3116 0 obj << +/D [3128 0 R /XYZ 71.731 227.426 null] +>> endobj +3117 0 obj << +/D [3128 0 R /XYZ 139.476 209.613 null] +>> endobj +3118 0 obj << +/D [3128 0 R /XYZ 446.192 209.613 null] +>> endobj +3119 0 obj << +/D [3128 0 R /XYZ 272.834 183.71 null] +>> endobj +3120 0 obj << +/D [3128 0 R /XYZ 71.731 171.591 null] +>> endobj +3121 0 obj << +/D [3128 0 R /XYZ 71.731 171.591 null] +>> endobj +3122 0 obj << +/D [3128 0 R /XYZ 71.731 160.677 null] +>> endobj +3123 0 obj << +/D [3128 0 R /XYZ 139.476 142.864 null] +>> endobj +3127 0 obj << +/Font << /F31 805 0 R /F37 1210 0 R /F25 681 0 R /F30 731 0 R /F39 1242 0 R /F42 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3174 0 obj << +/Length 3175 0 R +/Filter /FlateDecode +>> +stream +xYK6r_[e`͈EJ&mm MR=m +K(r%hI|KW>+(*9r?y>"F\;]} NC/f۲x?Wk`(n7oHgEJٍ}pM&n7nU&4-ʺj\45&+9; P7Q7G~e['taA@`r!FJV~ XGF͋Z.(sHnb̧(k_F(&qsJG6J,GtlVRU2mP~v%MVJ{CM j]T-68>MRDMGn$(e7"L)ZFlQ*kj$KI&b0E2T4݄(Mnhy@G4D/,Dc| . Cd`Ť*'GH5ވ6qW*D잦YGF8es(ՂzeeAC4t,PeJ{Z&` bltH%Ead_Q;DD(#zP:5b ZiB (A=:x*wVLu:M2u +n yܝ*QK宩`lxl}K}J76eW Z 7fH$ r vH$-y܄"ԃM1 b.gB|S4$|Q( tXPK(DAY^hƏpaqBly29WJc qwul3r瀆hUy^2{f8&snW?pGhc6M~_vW}VjUqZT'hfMKF_E=*r9JԌ0=M䝊Zbb"H/|33Z"|v|=p +_|t!'c@ą:AwP.eʶsN6IX!7C,32&>uڷF53JJkvhQӉ-`KC'JGn~3v@ǞD`$=ۤ4U}2X}6EURBwjuV˰ O+Y}j7NB.uf؂="y8|>iF +AN$<3knݘ|v$ue>ܞsY3գy!8Yf~KFzhSK|DЄRR0IG 160'\ '͘+endstream +endobj +3175 0 obj +1640 +endobj +3173 0 obj << +/Type /Page +/Contents 3174 0 R +/Resources 3172 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 2936 0 R +/Annots [ 3176 0 R 3177 0 R 3178 0 R 3179 0 R ] +>> endobj +3176 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [372.898 619.644 430.182 628.531] +/Subtype /Link +/A << /S /GoTo /D (zone_statement_grammar) >> +>> endobj +3177 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [484.607 560.133 536.91 569.044] +/Subtype /Link +/A << /S /GoTo /D (address_match_lists) >> +>> endobj +3178 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [487.705 288.153 537.981 297.065] +/Subtype /Link +/A << /S /GoTo /D (synthesis) >> +>> endobj +3179 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [71.731 277.259 176.836 284.113] +/Subtype /Link +/A << /S /GoTo /D (synthesis) >> +>> endobj +3171 0 obj << +/D [3173 0 R /XYZ 71.731 705.352 null] +>> endobj +3124 0 obj << +/D [3173 0 R /XYZ 71.731 657.532 null] +>> endobj +3125 0 obj << +/D [3173 0 R /XYZ 156.094 621.666 null] +>> endobj +945 0 obj << +/D [3173 0 R /XYZ 71.731 604.7 null] +>> endobj +373 0 obj << +/D [3173 0 R /XYZ 256.854 581.218 null] +>> endobj +3131 0 obj << +/D [3173 0 R /XYZ 71.731 575.091 null] +>> endobj +3132 0 obj << +/D [3173 0 R /XYZ 71.731 542.2 null] +>> endobj +3133 0 obj << +/D [3173 0 R /XYZ 71.731 532.237 null] +>> endobj +3134 0 obj << +/D [3173 0 R /XYZ 71.731 532.237 null] +>> endobj +3135 0 obj << +/D [3173 0 R /XYZ 71.731 519.42 null] +>> endobj +3136 0 obj << +/D [3173 0 R /XYZ 139.476 503.51 null] +>> endobj +3137 0 obj << +/D [3173 0 R /XYZ 139.476 490.559 null] +>> endobj +3138 0 obj << +/D [3173 0 R /XYZ 305.681 490.559 null] +>> endobj +3139 0 obj << +/D [3173 0 R /XYZ 489.379 490.559 null] +>> endobj +3140 0 obj << +/D [3173 0 R /XYZ 71.731 452.536 null] +>> endobj +3141 0 obj << +/D [3173 0 R /XYZ 71.731 452.536 null] +>> endobj +3142 0 obj << +/D [3173 0 R /XYZ 71.731 439.719 null] +>> endobj +3143 0 obj << +/D [3173 0 R /XYZ 139.476 423.809 null] +>> endobj +3144 0 obj << +/D [3173 0 R /XYZ 381.067 423.809 null] +>> endobj +3145 0 obj << +/D [3173 0 R /XYZ 154.141 410.858 null] +>> endobj +3146 0 obj << +/D [3173 0 R /XYZ 337.84 410.858 null] +>> endobj +3147 0 obj << +/D [3173 0 R /XYZ 71.731 385.787 null] +>> endobj +3148 0 obj << +/D [3173 0 R /XYZ 71.731 385.787 null] +>> endobj +3149 0 obj << +/D [3173 0 R /XYZ 71.731 374.873 null] +>> endobj +3150 0 obj << +/D [3173 0 R /XYZ 139.476 357.06 null] +>> endobj +3151 0 obj << +/D [3173 0 R /XYZ 71.731 319.038 null] +>> endobj +3152 0 obj << +/D [3173 0 R /XYZ 71.731 319.038 null] +>> endobj +3153 0 obj << +/D [3173 0 R /XYZ 71.731 306.221 null] +>> endobj +3154 0 obj << +/D [3173 0 R /XYZ 139.476 290.31 null] +>> endobj +3155 0 obj << +/D [3173 0 R /XYZ 71.731 267.297 null] +>> endobj +3156 0 obj << +/D [3173 0 R /XYZ 71.731 267.297 null] +>> endobj +3157 0 obj << +/D [3173 0 R /XYZ 71.731 254.325 null] +>> endobj +3158 0 obj << +/D [3173 0 R /XYZ 139.476 236.512 null] +>> endobj +3159 0 obj << +/D [3173 0 R /XYZ 439.716 236.512 null] +>> endobj +3160 0 obj << +/D [3173 0 R /XYZ 214.185 223.561 null] +>> endobj +3161 0 obj << +/D [3173 0 R /XYZ 397.884 223.561 null] +>> endobj +3162 0 obj << +/D [3173 0 R /XYZ 71.731 198.49 null] +>> endobj +3163 0 obj << +/D [3173 0 R /XYZ 71.731 198.49 null] +>> endobj +3164 0 obj << +/D [3173 0 R /XYZ 71.731 187.576 null] +>> endobj +3165 0 obj << +/D [3173 0 R /XYZ 139.476 169.763 null] +>> endobj +3166 0 obj << +/D [3173 0 R /XYZ 416.892 156.811 null] +>> endobj +946 0 obj << +/D [3173 0 R /XYZ 71.731 139.711 null] +>> endobj +3172 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F30 731 0 R /F42 669 0 R /F39 1242 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3211 0 obj << +/Length 3212 0 R +/Filter /FlateDecode +>> +stream +xڝXMo6_*#"%5Ң·X02 (WE{"DI>Lg̼G +;>a8iq{tݯNADH JqvvnO6^@}!}x'}˧R~x8W\doI"WTlrsG Ax98F!q<(&a`J>3 Py<60=0I;qw󬝲Nkq7j̻ 3SkQ=m0u-fY[}YT|BoHb9י|+  hbCՍdcgGz8p;<ԡ> +Cn$Qb5QףzPnBX&aLY8A|D]+x)$ `AO0e[ϑθ*C'w tby ,L_:m _8{w0C{LmUdٌkd;I^ϖչ6mD(!sd|P;knx# +!mJ2ު|vXylErB>0b8x1wP|"-M;pd; BPZ.<]c62=:}U%`itPEe5eZX ;᯹]kߡk¹=;pOꄠ1]#ʜ2,j6nA҄2QҴOgE,mDcAF-["vL!zOl% };Tn&r2Y2U>.*6iYhŬ `D'و]ku'}w_yHRL9>_ujjH nWN,$vkf{o:PXhi_hܖc֌M =+^! }:tVK>~g/ĴJ{Rҁ3͘{Lo8 f>*l^wC,D|۲4j5נ3Fl$BF(XYXSK(͍Re{ bҝZY#ƙ~ `  bځ<[/XH=fpUw:n5kPdBK`02e,hGK}_5endstream +endobj +3212 0 obj +1319 +endobj +3210 0 obj << +/Type /Page +/Contents 3211 0 R +/Resources 3209 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 3213 0 R +>> endobj +3208 0 obj << +/D [3210 0 R /XYZ 71.731 705.352 null] +>> endobj +377 0 obj << +/D [3210 0 R /XYZ 225.878 647.569 null] +>> endobj +3167 0 obj << +/D [3210 0 R /XYZ 71.731 641.442 null] +>> endobj +3168 0 obj << +/D [3210 0 R /XYZ 486.729 628.64 null] +>> endobj +3169 0 obj << +/D [3210 0 R /XYZ 149.997 615.689 null] +>> endobj +3170 0 obj << +/D [3210 0 R /XYZ 308.729 615.689 null] +>> endobj +3180 0 obj << +/D [3210 0 R /XYZ 71.731 595.599 null] +>> endobj +3181 0 obj << +/D [3210 0 R /XYZ 156.362 584.805 null] +>> endobj +3182 0 obj << +/D [3210 0 R /XYZ 71.731 572.685 null] +>> endobj +3183 0 obj << +/D [3210 0 R /XYZ 71.731 540.159 null] +>> endobj +3184 0 obj << +/D [3210 0 R /XYZ 71.731 510.584 null] +>> endobj +3185 0 obj << +/D [3210 0 R /XYZ 141.13 497.732 null] +>> endobj +3186 0 obj << +/D [3210 0 R /XYZ 71.731 490.594 null] +>> endobj +3187 0 obj << +/D [3210 0 R /XYZ 137.534 479.799 null] +>> endobj +3188 0 obj << +/D [3210 0 R /XYZ 71.731 459.709 null] +>> endobj +3189 0 obj << +/D [3210 0 R /XYZ 119.551 423.012 null] +>> endobj +3190 0 obj << +/D [3210 0 R /XYZ 283.724 423.012 null] +>> endobj +3191 0 obj << +/D [3210 0 R /XYZ 71.731 410.893 null] +>> endobj +3192 0 obj << +/D [3210 0 R /XYZ 71.731 368.417 null] +>> endobj +3193 0 obj << +/D [3210 0 R /XYZ 71.731 328.965 null] +>> endobj +3194 0 obj << +/D [3210 0 R /XYZ 156.362 316.014 null] +>> endobj +3195 0 obj << +/D [3210 0 R /XYZ 71.731 303.894 null] +>> endobj +3196 0 obj << +/D [3210 0 R /XYZ 71.731 271.368 null] +>> endobj +3197 0 obj << +/D [3210 0 R /XYZ 71.731 247.705 null] +>> endobj +3198 0 obj << +/D [3210 0 R /XYZ 71.731 227.194 null] +>> endobj +3199 0 obj << +/D [3210 0 R /XYZ 141.13 215.292 null] +>> endobj +947 0 obj << +/D [3210 0 R /XYZ 71.731 198.191 null] +>> endobj +381 0 obj << +/D [3210 0 R /XYZ 255.108 174.844 null] +>> endobj +3200 0 obj << +/D [3210 0 R /XYZ 71.731 166.392 null] +>> endobj +3201 0 obj << +/D [3210 0 R /XYZ 456.095 155.915 null] +>> endobj +3209 0 obj << +/Font << /F31 805 0 R /F42 669 0 R /F25 681 0 R /F30 731 0 R /F37 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3255 0 obj << +/Length 3256 0 R +/Filter /FlateDecode +>> +stream +xXێ61_AĈIISmmM( +MD$ёwx}) A=\ΜExcGYlGxa-Gس$ !d!Zӗ}R-;#|qȎVa#|~s{mW}PT0_fT}#ve/E??<~ݚcB,ֆK'F(NpjA=r#\+NEPGmsfgGG)bζ[sU$e==*E8b3v!o ;cM#O[fed456j! `us"?ݾ9"[]Θ " S-KPĽ!GU5S$D1#ưVYU]c%W/?\0[ 1vUb 7NyV LHZ xb (Kc>A<'3CX!A|dI'I Bw8%'GW&&KlbF/ڣ*h:CnBѝ*gɮv(/w9ׯݥ\p;։}958  Qkq +\d^T]0ȐZt,؁zԂ9GeRӥޞ~ڌl)`d mn@@9g寪tz׫6ALI) p˃2ǡBlY)½$"x̳# !<-62Mָi)c<7KGZ8jEmO] +u*ڹTbY2+#='ےSyVA*K ȩ03iE;IԦl*(U#wq'ׂQXD;M^O#ԞgaX_Rԃھ]{Dd, + +>m3U. l^g}=- ӝ=K(;Fub{KYnz_; e9ӌ6JnrUlThO>]q#gMq{~wp𡆡n?bxaGÛ/3c 3"0Y/ۇKM^vQMspGMӱ +ɫ%4u6JkvVtNulj{#(ll˖_~w>j.D:ٽ^)z||A 16D0yJ5ՀNyɈ1sgY|MCN"X~RQpxp<됆l2&  Òw[ϸa[Eql㺭o7Vv_GAD\fz[88>/ƇG} Y@׎<<"My.nVCroZ+N-(g_V}+H̍26;p1tz͂p/.+EpY44;ml9Vj3XC? ~Sdy8F#; ^ʾ2_OGߴ-/~V~Rg6,δG`f {=îҶF3 25k4mk4s5AfFqfcɿ'#FqDչ[dƊ܋?+endstream +endobj +3256 0 obj +1658 +endobj +3254 0 obj << +/Type /Page +/Contents 3255 0 R +/Resources 3253 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 3213 0 R +>> endobj +3252 0 obj << +/D [3254 0 R /XYZ 71.731 705.352 null] +>> endobj +3202 0 obj << +/D [3254 0 R /XYZ 119.551 647.569 null] +>> endobj +3203 0 obj << +/D [3254 0 R /XYZ 230.603 647.569 null] +>> endobj +3204 0 obj << +/D [3254 0 R /XYZ 274.707 647.569 null] +>> endobj +3205 0 obj << +/D [3254 0 R /XYZ 428.827 647.569 null] +>> endobj +3206 0 obj << +/D [3254 0 R /XYZ 151.919 634.618 null] +>> endobj +3207 0 obj << +/D [3254 0 R /XYZ 181.807 634.618 null] +>> endobj +3214 0 obj << +/D [3254 0 R /XYZ 71.731 622.498 null] +>> endobj +3215 0 obj << +/D [3254 0 R /XYZ 71.731 589.973 null] +>> endobj +3216 0 obj << +/D [3254 0 R /XYZ 71.731 580.01 null] +>> endobj +3217 0 obj << +/D [3254 0 R /XYZ 279.502 569.761 null] +>> endobj +948 0 obj << +/D [3254 0 R /XYZ 71.731 520.247 null] +>> endobj +385 0 obj << +/D [3254 0 R /XYZ 254.511 494.743 null] +>> endobj +3218 0 obj << +/D [3254 0 R /XYZ 71.731 488.616 null] +>> endobj +3219 0 obj << +/D [3254 0 R /XYZ 71.731 488.616 null] +>> endobj +3220 0 obj << +/D [3254 0 R /XYZ 71.731 455.724 null] +>> endobj +3221 0 obj << +/D [3254 0 R /XYZ 71.731 445.762 null] +>> endobj +3222 0 obj << +/D [3254 0 R /XYZ 71.731 445.762 null] +>> endobj +3223 0 obj << +/D [3254 0 R /XYZ 71.731 432.945 null] +>> endobj +3224 0 obj << +/D [3254 0 R /XYZ 139.476 417.035 null] +>> endobj +3225 0 obj << +/D [3254 0 R /XYZ 481.41 391.132 null] +>> endobj +3226 0 obj << +/D [3254 0 R /XYZ 204.671 378.18 null] +>> endobj +3227 0 obj << +/D [3254 0 R /XYZ 345.769 378.18 null] +>> endobj +3228 0 obj << +/D [3254 0 R /XYZ 502.948 378.18 null] +>> endobj +3229 0 obj << +/D [3254 0 R /XYZ 240.486 365.229 null] +>> endobj +3230 0 obj << +/D [3254 0 R /XYZ 374.969 365.229 null] +>> endobj +3231 0 obj << +/D [3254 0 R /XYZ 71.731 340.158 null] +>> endobj +3232 0 obj << +/D [3254 0 R /XYZ 71.731 340.158 null] +>> endobj +3233 0 obj << +/D [3254 0 R /XYZ 71.731 329.244 null] +>> endobj +3234 0 obj << +/D [3254 0 R /XYZ 139.476 311.431 null] +>> endobj +3235 0 obj << +/D [3254 0 R /XYZ 71.731 286.734 null] +>> endobj +3236 0 obj << +/D [3254 0 R /XYZ 71.731 286.734 null] +>> endobj +3237 0 obj << +/D [3254 0 R /XYZ 71.731 275.446 null] +>> endobj +3238 0 obj << +/D [3254 0 R /XYZ 139.476 257.633 null] +>> endobj +3239 0 obj << +/D [3254 0 R /XYZ 71.731 232.936 null] +>> endobj +3240 0 obj << +/D [3254 0 R /XYZ 71.731 232.936 null] +>> endobj +3241 0 obj << +/D [3254 0 R /XYZ 71.731 221.648 null] +>> endobj +3242 0 obj << +/D [3254 0 R /XYZ 139.476 203.835 null] +>> endobj +3243 0 obj << +/D [3254 0 R /XYZ 71.731 179.138 null] +>> endobj +3244 0 obj << +/D [3254 0 R /XYZ 71.731 179.138 null] +>> endobj +3245 0 obj << +/D [3254 0 R /XYZ 71.731 167.85 null] +>> endobj +3246 0 obj << +/D [3254 0 R /XYZ 139.476 150.037 null] +>> endobj +3253 0 obj << +/Font << /F31 805 0 R /F30 731 0 R /F25 681 0 R /F37 1210 0 R /F62 669 0 R /F58 2248 0 R /F42 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3313 0 obj << +/Length 3314 0 R +/Filter /FlateDecode +>> +stream +xYK62bve Q)E3T]x3f#7_clf2UTՕ<o<Ûȇ"?$|7ƌ~9$"l|Ƒ\r_%]-. ="|k5JU%]am^ 5;?~pퟻ__ ""wnc7PR9+ܿ:^_ܚ#_7 +992zI}s>mqpeGPH)/ + VY^md:{h+|Z]u`QLt~䴐5=1q6 MP(5{ғwNQZިS\t-SeՉVW5מff2X36 {NUT{&!ofs^v4 #Tsy3V".*AzuEkBAFPUQ"oPF0:içBgpj_)`!]ǥ] ǃw'0M!F8B0EqddZ { >ORù1VfK6h\ ՜L=e{⋬Ss\5kŴhνd@{ovbzU]^gv՚VB2&>ƿg~4jא=A5)g.4W&8 2{Hs5 {ID{;ҝh]@7VFyeQuy44ƛoY;6=T1{C65Rnj[(A!Zs^ OA +HYl + %X\IChkyqac 3<*E1a<0]c~BXXpҪ?kA@jyw5+%qs(.R\wKvX)$O 5$+k4mcXiGS`࣋M8Iw? +vG +mL|w.Ì$ZuMׯNBV?g+l/PWUK++ӈ sй=pу ڒ4F~aY TII^&NHZt|+ܯ~zՐfomtsNW> endobj +3310 0 obj << +/D [3312 0 R /XYZ 71.731 705.352 null] +>> endobj +3247 0 obj << +/D [3312 0 R /XYZ 71.731 657.532 null] +>> endobj +3248 0 obj << +/D [3312 0 R /XYZ 71.731 657.532 null] +>> endobj +3249 0 obj << +/D [3312 0 R /XYZ 71.731 645.547 null] +>> endobj +3250 0 obj << +/D [3312 0 R /XYZ 139.476 629.637 null] +>> endobj +3251 0 obj << +/D [3312 0 R /XYZ 139.476 590.782 null] +>> endobj +3257 0 obj << +/D [3312 0 R /XYZ 71.731 567.769 null] +>> endobj +3258 0 obj << +/D [3312 0 R /XYZ 71.731 567.769 null] +>> endobj +3259 0 obj << +/D [3312 0 R /XYZ 71.731 552.895 null] +>> endobj +3260 0 obj << +/D [3312 0 R /XYZ 139.476 536.984 null] +>> endobj +3261 0 obj << +/D [3312 0 R /XYZ 201.732 536.984 null] +>> endobj +3262 0 obj << +/D [3312 0 R /XYZ 234.1 511.082 null] +>> endobj +3263 0 obj << +/D [3312 0 R /XYZ 220.83 498.13 null] +>> endobj +3264 0 obj << +/D [3312 0 R /XYZ 71.731 486.011 null] +>> endobj +3265 0 obj << +/D [3312 0 R /XYZ 71.731 486.011 null] +>> endobj +3266 0 obj << +/D [3312 0 R /XYZ 71.731 475.097 null] +>> endobj +3267 0 obj << +/D [3312 0 R /XYZ 139.476 457.284 null] +>> endobj +3268 0 obj << +/D [3312 0 R /XYZ 361.849 457.284 null] +>> endobj +3269 0 obj << +/D [3312 0 R /XYZ 429.913 457.284 null] +>> endobj +3270 0 obj << +/D [3312 0 R /XYZ 139.476 444.332 null] +>> endobj +3271 0 obj << +/D [3312 0 R /XYZ 139.476 431.381 null] +>> endobj +3272 0 obj << +/D [3312 0 R /XYZ 413.356 431.381 null] +>> endobj +3273 0 obj << +/D [3312 0 R /XYZ 317.098 418.429 null] +>> endobj +3274 0 obj << +/D [3312 0 R /XYZ 297.691 405.478 null] +>> endobj +3275 0 obj << +/D [3312 0 R /XYZ 334.493 405.478 null] +>> endobj +3276 0 obj << +/D [3312 0 R /XYZ 472.962 405.478 null] +>> endobj +3277 0 obj << +/D [3312 0 R /XYZ 196.651 392.527 null] +>> endobj +3278 0 obj << +/D [3312 0 R /XYZ 263.061 392.527 null] +>> endobj +3279 0 obj << +/D [3312 0 R /XYZ 154.141 379.575 null] +>> endobj +3280 0 obj << +/D [3312 0 R /XYZ 71.731 369.493 null] +>> endobj +3281 0 obj << +/D [3312 0 R /XYZ 71.731 369.493 null] +>> endobj +3282 0 obj << +/D [3312 0 R /XYZ 71.731 356.542 null] +>> endobj +3283 0 obj << +/D [3312 0 R /XYZ 139.476 338.729 null] +>> endobj +3284 0 obj << +/D [3312 0 R /XYZ 172.432 325.777 null] +>> endobj +3285 0 obj << +/D [3312 0 R /XYZ 232.158 325.777 null] +>> endobj +3286 0 obj << +/D [3312 0 R /XYZ 71.731 300.706 null] +>> endobj +3287 0 obj << +/D [3312 0 R /XYZ 71.731 300.706 null] +>> endobj +3288 0 obj << +/D [3312 0 R /XYZ 71.731 289.792 null] +>> endobj +3289 0 obj << +/D [3312 0 R /XYZ 139.476 271.979 null] +>> endobj +3290 0 obj << +/D [3312 0 R /XYZ 405.156 259.028 null] +>> endobj +3291 0 obj << +/D [3312 0 R /XYZ 71.731 246.908 null] +>> endobj +3292 0 obj << +/D [3312 0 R /XYZ 71.731 246.908 null] +>> endobj +3293 0 obj << +/D [3312 0 R /XYZ 71.731 234.091 null] +>> endobj +3294 0 obj << +/D [3312 0 R /XYZ 139.476 218.181 null] +>> endobj +3295 0 obj << +/D [3312 0 R /XYZ 300.829 205.23 null] +>> endobj +3296 0 obj << +/D [3312 0 R /XYZ 355.175 205.23 null] +>> endobj +3297 0 obj << +/D [3312 0 R /XYZ 139.476 179.327 null] +>> endobj +3298 0 obj << +/D [3312 0 R /XYZ 422.601 179.327 null] +>> endobj +3299 0 obj << +/D [3312 0 R /XYZ 139.476 166.375 null] +>> endobj +3311 0 obj << +/Font << /F31 805 0 R /F30 731 0 R /F25 681 0 R /F37 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3359 0 obj << +/Length 3360 0 R +/Filter /FlateDecode +>> +stream +xYˎ8X%6TKyn D,F(UquEQEs:qn`1E)Ǿ/s+UHIl+ҕy?b0VQ('<}Ifۤ_c+ : k5UU> 6J×<\r'.1qYr x4"_9~Hˆ&s:)TҬh=7]94$18XċǾCO#>x DT@JjW$W$Mk)^>e~YuemWo=5ڼ,NZ5b +\942CIJEULt wzA؈kuI^ڹFO1| 3FGR/K'\m>ͯ$ϧ&ZXXz_Փf:_:QgheN^)%+"iҊ-mY$#k-O}xD״0n*џ C;1℥eJ_a)0OvyՈE[-dgݢjX!k@(+p1kuya#DnsT :Z܍r+fvt +ꅎEsHu AjZ(BC=т. SeNzl/Tf* ),ME1*vm&d]z+];LI1Bh]ޥ&R|8]Sd䔵qt8rH|FIդk/E!pa t lA)0 OrAOSnw(9SMH!o %R]8+}< +_w؎ Z1?kLoKI< <'e{H䅿[j%>p%CD`H/J0M) \\.=BEr+iS՛Tv 0$Qb`A$Yc&O$* *<u0HC}D q4K(]JqLV_H8'ń=gRcE!M (c.1"WWøGx`kz!/#Vϑ>i( g !ƌ86e*m#{݀/kGh|ƌ> ]Y%5;'!:.oаQҰ`ƔڨCbkoVb:(xC(0ZSI +ym?Gx.+v7e 'Bd +˚3S+>`ݤUk893H`Kv L4Sy:ƼDSF##Z*Y\\.cTa-ѷ`R4:N9<\d%)IM$g[Őt~=Rtu#gDGqB#OrbJ?7F~E%`.`܅Li:7XW76BoVsIW(xYSc)}@N{ia{,M Ɉ eMp[zjYWjqeㅣJ8"ȱ^Åb1$/[p{c.A$v4 _v%(q1' C6[Y +6jٌpendstream +endobj +3360 0 obj +1906 +endobj +3358 0 obj << +/Type /Page +/Contents 3359 0 R +/Resources 3357 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 3213 0 R +/Annots [ 3361 0 R ] +>> endobj +3361 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [397.903 264.243 442.735 273.264] +/Subtype /Link +/A << /S /GoTo /D (configuration_file_elements) >> +>> endobj +3356 0 obj << +/D [3358 0 R /XYZ 71.731 705.352 null] +>> endobj +3300 0 obj << +/D [3358 0 R /XYZ 71.731 657.532 null] +>> endobj +3301 0 obj << +/D [3358 0 R /XYZ 71.731 657.532 null] +>> endobj +3302 0 obj << +/D [3358 0 R /XYZ 71.731 647.45 null] +>> endobj +3303 0 obj << +/D [3358 0 R /XYZ 139.476 629.637 null] +>> endobj +3304 0 obj << +/D [3358 0 R /XYZ 139.476 629.637 null] +>> endobj +3305 0 obj << +/D [3358 0 R /XYZ 385.73 577.831 null] +>> endobj +3306 0 obj << +/D [3358 0 R /XYZ 380.2 564.88 null] +>> endobj +3307 0 obj << +/D [3358 0 R /XYZ 383.379 551.928 null] +>> endobj +3308 0 obj << +/D [3358 0 R /XYZ 139.476 538.977 null] +>> endobj +3309 0 obj << +/D [3358 0 R /XYZ 172.124 538.977 null] +>> endobj +3315 0 obj << +/D [3358 0 R /XYZ 71.731 526.857 null] +>> endobj +3316 0 obj << +/D [3358 0 R /XYZ 71.731 526.857 null] +>> endobj +3317 0 obj << +/D [3358 0 R /XYZ 71.731 515.943 null] +>> endobj +3318 0 obj << +/D [3358 0 R /XYZ 139.476 498.13 null] +>> endobj +3319 0 obj << +/D [3358 0 R /XYZ 191.212 498.13 null] +>> endobj +3320 0 obj << +/D [3358 0 R /XYZ 71.731 486.011 null] +>> endobj +3321 0 obj << +/D [3358 0 R /XYZ 71.731 486.011 null] +>> endobj +3322 0 obj << +/D [3358 0 R /XYZ 71.731 473.194 null] +>> endobj +3323 0 obj << +/D [3358 0 R /XYZ 139.476 457.284 null] +>> endobj +3324 0 obj << +/D [3358 0 R /XYZ 139.476 457.284 null] +>> endobj +3325 0 obj << +/D [3358 0 R /XYZ 405.027 444.332 null] +>> endobj +3326 0 obj << +/D [3358 0 R /XYZ 139.476 431.381 null] +>> endobj +3327 0 obj << +/D [3358 0 R /XYZ 315.643 431.381 null] +>> endobj +3328 0 obj << +/D [3358 0 R /XYZ 301.238 418.429 null] +>> endobj +3329 0 obj << +/D [3358 0 R /XYZ 442.735 418.429 null] +>> endobj +3330 0 obj << +/D [3358 0 R /XYZ 475.382 418.429 null] +>> endobj +3331 0 obj << +/D [3358 0 R /XYZ 71.731 393.359 null] +>> endobj +3332 0 obj << +/D [3358 0 R /XYZ 71.731 393.359 null] +>> endobj +3333 0 obj << +/D [3358 0 R /XYZ 71.731 380.542 null] +>> endobj +3334 0 obj << +/D [3358 0 R /XYZ 139.476 364.631 null] +>> endobj +3335 0 obj << +/D [3358 0 R /XYZ 160.129 364.631 null] +>> endobj +949 0 obj << +/D [3358 0 R /XYZ 71.731 347.531 null] +>> endobj +389 0 obj << +/D [3358 0 R /XYZ 367.808 324.183 null] +>> endobj +3336 0 obj << +/D [3358 0 R /XYZ 71.731 315.546 null] +>> endobj +3337 0 obj << +/D [3358 0 R /XYZ 238.234 292.303 null] +>> endobj +3338 0 obj << +/D [3358 0 R /XYZ 343.927 292.303 null] +>> endobj +3339 0 obj << +/D [3358 0 R /XYZ 119.551 279.351 null] +>> endobj +3340 0 obj << +/D [3358 0 R /XYZ 398.321 279.351 null] +>> endobj +3341 0 obj << +/D [3358 0 R /XYZ 346.437 266.4 null] +>> endobj +3342 0 obj << +/D [3358 0 R /XYZ 71.731 259.262 null] +>> endobj +3343 0 obj << +/D [3358 0 R /XYZ 71.731 215.426 null] +>> endobj +3344 0 obj << +/D [3358 0 R /XYZ 71.731 205.464 null] +>> endobj +3345 0 obj << +/D [3358 0 R /XYZ 71.731 205.464 null] +>> endobj +3346 0 obj << +/D [3358 0 R /XYZ 71.731 194.55 null] +>> endobj +3347 0 obj << +/D [3358 0 R /XYZ 139.476 176.737 null] +>> endobj +3348 0 obj << +/D [3358 0 R /XYZ 340.53 176.737 null] +>> endobj +3357 0 obj << +/Font << /F31 805 0 R /F30 731 0 R /F25 681 0 R /F42 669 0 R /F37 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3397 0 obj << +/Length 3398 0 R +/Filter /FlateDecode +>> +stream +xX[o6H梾n+m[ݑPxL_clLj;糱^^ygXӻGAx!"i6ݿG>CYzӰo}󮠍dv#=ezx'ov \z;aauζz~T'1ۍ:<.)BoGᤲVH%+ AQWtZ +GA逬L;_xWZz.NzT,6D{֫~߱qƬfG1Q֑;4pk⥠a4 2YӰ}67$LP<`m +0@40B⪯K^qkt}Ū}G%N7AdU#:Rղ9l}87 +0uȗu$FAִAI27v.#e'^zKSĘ.}kf%hnP2膵'њL{%ϭ?!UZ؝cdh&ZK[VNwK}$uSi0K^/Tn&Z%D7cN ;ƖW[wAA9ի4R5YkMuFNWQ.~ͱpc^;Tځ؏ prlgE8FqlQy8FiБ Q%g֧CM{!(sQUzDkvAmѰuˆB^O_`Ŧ`tp{QB.(SQuZPA@s%{r:2F[Z]n.+~*B6޹w 0,.~4ui=dG4ZU"}&t04oy"ex#PƎ7,Y/f}' ݯcY>kS\pTgEL©׉fϰD͸t}mn4T]:9j՝2 osLukrkKYUCsW)gx(غx/W oͦeyQM\˄NEBcȎojk 7uꑳD~ mW7$"Z3`Mϊ)8AIŒEȿ},/'x)FoZd/endstream +endobj +3398 0 obj +1517 +endobj +3396 0 obj << +/Type /Page +/Contents 3397 0 R +/Resources 3395 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 3213 0 R +>> endobj +3394 0 obj << +/D [3396 0 R /XYZ 71.731 705.352 null] +>> endobj +951 0 obj << +/D [3396 0 R /XYZ 71.731 717.307 null] +>> endobj +3349 0 obj << +/D [3396 0 R /XYZ 71.731 657.532 null] +>> endobj +3350 0 obj << +/D [3396 0 R /XYZ 71.731 657.532 null] +>> endobj +3351 0 obj << +/D [3396 0 R /XYZ 71.731 647.45 null] +>> endobj +3352 0 obj << +/D [3396 0 R /XYZ 139.476 629.637 null] +>> endobj +3353 0 obj << +/D [3396 0 R /XYZ 434.735 629.637 null] +>> endobj +3354 0 obj << +/D [3396 0 R /XYZ 340.082 564.88 null] +>> endobj +3355 0 obj << +/D [3396 0 R /XYZ 423.627 564.88 null] +>> endobj +3362 0 obj << +/D [3396 0 R /XYZ 71.731 541.866 null] +>> endobj +3363 0 obj << +/D [3396 0 R /XYZ 71.731 541.866 null] +>> endobj +3364 0 obj << +/D [3396 0 R /XYZ 71.731 528.895 null] +>> endobj +3365 0 obj << +/D [3396 0 R /XYZ 139.476 511.082 null] +>> endobj +3366 0 obj << +/D [3396 0 R /XYZ 478.83 511.082 null] +>> endobj +3367 0 obj << +/D [3396 0 R /XYZ 71.731 498.962 null] +>> endobj +3368 0 obj << +/D [3396 0 R /XYZ 71.731 498.962 null] +>> endobj +3369 0 obj << +/D [3396 0 R /XYZ 71.731 488.048 null] +>> endobj +3370 0 obj << +/D [3396 0 R /XYZ 139.476 470.235 null] +>> endobj +3371 0 obj << +/D [3396 0 R /XYZ 438.611 470.235 null] +>> endobj +950 0 obj << +/D [3396 0 R /XYZ 71.731 453.134 null] +>> endobj +393 0 obj << +/D [3396 0 R /XYZ 302.832 429.787 null] +>> endobj +3372 0 obj << +/D [3396 0 R /XYZ 71.731 423.66 null] +>> endobj +3373 0 obj << +/D [3396 0 R /XYZ 71.731 390.768 null] +>> endobj +3374 0 obj << +/D [3396 0 R /XYZ 71.731 380.806 null] +>> endobj +3375 0 obj << +/D [3396 0 R /XYZ 71.731 380.806 null] +>> endobj +3376 0 obj << +/D [3396 0 R /XYZ 71.731 367.989 null] +>> endobj +3377 0 obj << +/D [3396 0 R /XYZ 139.476 352.078 null] +>> endobj +3378 0 obj << +/D [3396 0 R /XYZ 71.731 339.959 null] +>> endobj +3379 0 obj << +/D [3396 0 R /XYZ 71.731 339.959 null] +>> endobj +3380 0 obj << +/D [3396 0 R /XYZ 71.731 329.045 null] +>> endobj +3381 0 obj << +/D [3396 0 R /XYZ 139.476 311.232 null] +>> endobj +3382 0 obj << +/D [3396 0 R /XYZ 227.644 298.28 null] +>> endobj +3383 0 obj << +/D [3396 0 R /XYZ 257.383 285.329 null] +>> endobj +3384 0 obj << +/D [3396 0 R /XYZ 71.731 260.258 null] +>> endobj +3385 0 obj << +/D [3396 0 R /XYZ 71.731 260.258 null] +>> endobj +3386 0 obj << +/D [3396 0 R /XYZ 71.731 247.441 null] +>> endobj +3387 0 obj << +/D [3396 0 R /XYZ 139.476 231.531 null] +>> endobj +3388 0 obj << +/D [3396 0 R /XYZ 178.669 218.58 null] +>> endobj +3389 0 obj << +/D [3396 0 R /XYZ 71.731 208.429 null] +>> endobj +3390 0 obj << +/D [3396 0 R /XYZ 71.731 208.429 null] +>> endobj +3391 0 obj << +/D [3396 0 R /XYZ 71.731 195.546 null] +>> endobj +3392 0 obj << +/D [3396 0 R /XYZ 139.476 177.733 null] +>> endobj +3395 0 obj << +/Font << /F31 805 0 R /F30 731 0 R /F25 681 0 R /F37 1210 0 R /F42 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3431 0 obj << +/Length 3432 0 R +/Filter /FlateDecode +>> +stream +xXK6r_ XmAPEh[XYtDy5rj%ofh!$/P"ߑ,Q<68C+nNsiY{I_/c+4 CjY.Q^qcWMo.j~)C-R%#ѱ$2fo'۹BdJE@q#Ǧe-,{ݯyBz{I u=o5^^ȸ:@S(yu> endobj +3427 0 obj << +/D [3430 0 R /XYZ 71.731 705.352 null] +>> endobj +3428 0 obj << +/D [3430 0 R /XYZ 71.731 717.307 null] +>> endobj +3393 0 obj << +/D [3430 0 R /XYZ 196.651 647.569 null] +>> endobj +397 0 obj << +/D [3430 0 R /XYZ 298.206 594.17 null] +>> endobj +3399 0 obj << +/D [3430 0 R /XYZ 71.731 588.043 null] +>> endobj +3400 0 obj << +/D [3430 0 R /XYZ 71.731 578.08 null] +>> endobj +3401 0 obj << +/D [3430 0 R /XYZ 71.731 578.08 null] +>> endobj +3402 0 obj << +/D [3430 0 R /XYZ 71.731 563.256 null] +>> endobj +3403 0 obj << +/D [3430 0 R /XYZ 139.476 547.345 null] +>> endobj +3404 0 obj << +/D [3430 0 R /XYZ 420.37 547.345 null] +>> endobj +3405 0 obj << +/D [3430 0 R /XYZ 71.731 522.275 null] +>> endobj +3406 0 obj << +/D [3430 0 R /XYZ 71.731 522.275 null] +>> endobj +3407 0 obj << +/D [3430 0 R /XYZ 71.731 511.361 null] +>> endobj +3408 0 obj << +/D [3430 0 R /XYZ 139.476 493.547 null] +>> endobj +3409 0 obj << +/D [3430 0 R /XYZ 428.06 493.547 null] +>> endobj +3410 0 obj << +/D [3430 0 R /XYZ 71.731 456.242 null] +>> endobj +3411 0 obj << +/D [3430 0 R /XYZ 71.731 456.242 null] +>> endobj +3412 0 obj << +/D [3430 0 R /XYZ 71.731 444.611 null] +>> endobj +3413 0 obj << +/D [3430 0 R /XYZ 139.476 426.798 null] +>> endobj +3414 0 obj << +/D [3430 0 R /XYZ 348.161 426.798 null] +>> endobj +3415 0 obj << +/D [3430 0 R /XYZ 496.722 400.895 null] +>> endobj +3416 0 obj << +/D [3430 0 R /XYZ 71.731 375.824 null] +>> endobj +3417 0 obj << +/D [3430 0 R /XYZ 71.731 375.824 null] +>> endobj +3418 0 obj << +/D [3430 0 R /XYZ 71.731 364.91 null] +>> endobj +3419 0 obj << +/D [3430 0 R /XYZ 139.476 347.097 null] +>> endobj +3420 0 obj << +/D [3430 0 R /XYZ 310.473 347.097 null] +>> endobj +3421 0 obj << +/D [3430 0 R /XYZ 71.731 322.026 null] +>> endobj +3422 0 obj << +/D [3430 0 R /XYZ 71.731 317.045 null] +>> endobj +3423 0 obj << +/D [3430 0 R /XYZ 278.805 307.546 null] +>> endobj +952 0 obj << +/D [3430 0 R /XYZ 71.731 251.755 null] +>> endobj +401 0 obj << +/D [3430 0 R /XYZ 228.892 226.251 null] +>> endobj +3424 0 obj << +/D [3430 0 R /XYZ 71.731 217.613 null] +>> endobj +3425 0 obj << +/D [3430 0 R /XYZ 353.649 194.37 null] +>> endobj +3426 0 obj << +/D [3430 0 R /XYZ 119.551 181.419 null] +>> endobj +3429 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F37 1210 0 R /F42 669 0 R /F30 731 0 R /F62 669 0 R /F58 2248 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3456 0 obj << +/Length 3457 0 R +/Filter /FlateDecode +>> +stream +xڵXn6>uhKlm"-}h@#qfK1{/E^j邁MIz! ?ox$ q#Dy+]ӏM3 +H&jgWo`~F8Xq7o뗇(8~co߽On^b/^F;$kt^X$&+C;0X9QB4"\dyz:S_J\ѽWìRƋ rowQŒcnpi釼,m;VUz=I+v(x>"SlwJ*7fľG>TF +Њִ8E.j2zz0Oy}B+YX={?auן;;nǃaQ ܫI+$qgu~֋+/~_N,즻 /FGх$1*6[m"ϝDL[tˎWl0N[nˆ:o;~#[20t&sp;&/d?|;ݢSsfΉĝJfd <2v ;*&I꼪,#o%%))Se&81xW2v۽{6c_ہTK)"Cl-n!`BajqtnjCpi>dWxЅ`FZO +R-"$Z1gjC! 8 |z/Ƌ U#[i)tȯ޽OT밪si ([Ex}vi׭f~yXI̪Tu [UUxHWR@&$ 68KB/5CHSR`%8ɘM &uyϊd!g7EUG#@ڲ0M +9 +|H!u!eՃȴ4@Q[GbEoC3 4`|y-.R`fnM +C'+M[d&3j-Vƭ5XSc8~<11a dPm@:R-X;DFI93xj ؖ͢Ng_nyxb>!aN5Htn:BtNJLE8o ? +Ts萁1hBGdmThSx@z}ਣ3 K&}*|9XfVį)> endobj +3458 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [268.749 336.074 338.487 345.095] +/Subtype /Link +/A << /S /GoTo /D (rrset_ordering) >> +>> endobj +3459 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [257.631 253.384 327.369 262.405] +/Subtype /Link +/A << /S /GoTo /D (topology) >> +>> endobj +3453 0 obj << +/D [3455 0 R /XYZ 71.731 705.352 null] +>> endobj +3433 0 obj << +/D [3455 0 R /XYZ 71.731 555.362 null] +>> endobj +3434 0 obj << +/D [3455 0 R /XYZ 71.731 512.476 null] +>> endobj +3435 0 obj << +/D [3455 0 R /XYZ 71.731 487.505 null] +>> endobj +3436 0 obj << +/D [3455 0 R /XYZ 71.731 466.635 null] +>> endobj +3437 0 obj << +/D [3455 0 R /XYZ 71.731 456.673 null] +>> endobj +3438 0 obj << +/D [3455 0 R /XYZ 182.819 446.424 null] +>> endobj +3439 0 obj << +/D [3455 0 R /XYZ 337.289 446.424 null] +>> endobj +953 0 obj << +/D [3455 0 R /XYZ 71.731 408.566 null] +>> endobj +405 0 obj << +/D [3455 0 R /XYZ 301.84 383.062 null] +>> endobj +3440 0 obj << +/D [3455 0 R /XYZ 71.731 376.935 null] +>> endobj +3441 0 obj << +/D [3455 0 R /XYZ 167.77 338.23 null] +>> endobj +3442 0 obj << +/D [3455 0 R /XYZ 71.731 280.004 null] +>> endobj +3443 0 obj << +/D [3455 0 R /XYZ 137.534 268.492 null] +>> endobj +3444 0 obj << +/D [3455 0 R /XYZ 293.227 268.492 null] +>> endobj +3445 0 obj << +/D [3455 0 R /XYZ 153.862 255.541 null] +>> endobj +3446 0 obj << +/D [3455 0 R /XYZ 460.339 255.541 null] +>> endobj +3447 0 obj << +/D [3455 0 R /XYZ 175.57 242.589 null] +>> endobj +3448 0 obj << +/D [3455 0 R /XYZ 309.507 229.638 null] +>> endobj +3449 0 obj << +/D [3455 0 R /XYZ 71.731 209.549 null] +>> endobj +3450 0 obj << +/D [3455 0 R /XYZ 285.556 159.9 null] +>> endobj +3451 0 obj << +/D [3455 0 R /XYZ 386.934 159.9 null] +>> endobj +3454 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F37 1210 0 R /F62 669 0 R /F58 2248 0 R /F42 669 0 R /F30 731 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3472 0 obj << +/Length 3473 0 R +/Filter /FlateDecode +>> +stream +xڭXKo6z̯QVHz MwѴ@-|+zPl:VRIlCp"~3͐3 +BQ/$yճ`YƘljH%q{ug$=h?uL~Lge + +g_V;sէmoMb z "{HHd&&QPea)G*驔p&_ϲ)d/Bm6ɽmy;냣<ٶl~FQ]+KR%eH& G}jMP) ۀ[Zֻ,ꮮ*묷^665_[ao cݙ7rd?!#4N f,#fU 펹1H,2U +S8Wrԯjdoz=bUpRV( F(yI{aK < bRs|؜bHҢSN;.9Ȯ>a?] .&b4ɘEX + 9 1)}>vWO-x)\1^$MW8 XPP݀*Կƕϛ!šf"E`0Q0x˧{y(@qa +]XtmX!2aތ?fiTAIȳ5\kb昵ʼmqx_'UA`Fz6;{`Ή$ojDFh^jn^=X|qhå~%$KIĜ4?'U"?Dendstream +endobj +3473 0 obj +1121 +endobj +3471 0 obj << +/Type /Page +/Contents 3472 0 R +/Resources 3470 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 3460 0 R +>> endobj +3468 0 obj << +/D [3471 0 R /XYZ 71.731 705.352 null] +>> endobj +3469 0 obj << +/D [3471 0 R /XYZ 71.731 717.307 null] +>> endobj +3452 0 obj << +/D [3471 0 R /XYZ 71.731 657.532 null] +>> endobj +3461 0 obj << +/D [3471 0 R /XYZ 71.731 557.742 null] +>> endobj +3462 0 obj << +/D [3471 0 R /XYZ 71.731 339.074 null] +>> endobj +3463 0 obj << +/D [3471 0 R /XYZ 356.389 314.221 null] +>> endobj +3464 0 obj << +/D [3471 0 R /XYZ 71.731 263.247 null] +>> endobj +954 0 obj << +/D [3471 0 R /XYZ 71.731 197.805 null] +>> endobj +3470 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F37 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3507 0 obj << +/Length 3508 0 R +/Filter /FlateDecode +>> +stream +xڭX_o6OaIbV$%Qlvaِ(Tʒ qac_wG)KlXx$OgaCC6K/ ?'XCj&}qByxK"#>{X{&ZU/,$oqW/h%mVzְ OZ|xTDgl5IgPN(%2 2Gʴ"z|oxfa_%#4l~(vY+UW HjHXmwuVٝI76{bZ',]ֻmcCڍ%rmwlNF F*OңƦ*FY+?lB#DT$E% +3B3QKٴIWfm3y9mKI`ygc+KM1,' +lYyִcL8%1;46XdT?pr{J1b LeOeNC舤HͧbDQh3~l* $ Y` 3+uTbe z=O+HYFm,gKFID1/~'M;"#.lh>VYPl2mz&9ǒ!GKG:_$+ -Wv`L _Fat4$M6-5aAP,\2c±~vVP[!4 A9\> vTv*1ڷ +|Wo{@q` Sô1% f1. }7D.qmSS(ʃ;2) D tt&Nu[e)Y/' TW.28`G෺\C8TfLRX#I,t7&HuO8W^|5.w Muo[p 8G( V2 "WAa=f|fnAx3X7YHNn`uR@f55hȺ_zBbBNUHB8 +5}_4=y~CMmkSbO +`H:'#Y.^:|(ǻTgqm/*PSA4 -xXݑWW}l)(vש;ukV;_(Rz&ҫ,ׂ4KBw8ƽk[{ CW&./uzUZÂ"ed+ӓ>L J,vX.]+5w֑Z/ݍk"ۀG +fx. #rh1볗Scl̮ $Bi/5^8J%"^pZ9lݵl I)I tL5vה h)"*Fl]\C y#(ľӠYi#ZD^<4{$Qi4ZiPh@`&pE'XHwu ;lvUoe>&CiKL#s h]%[NL\M}lɌ3)J tS=8Mx񆥪r88z/-~"+`EXhL#u<x_fUu'8ȬpE¤6Bi}Ϻ%J7L"sbRzZ$ͯON˟yyP`fM=4SWSx`Caj*8.'\y ~ ( | 0XK$g:.T|endstream +endobj +3508 0 obj +1756 +endobj +3506 0 obj << +/Type /Page +/Contents 3507 0 R +/Resources 3505 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 3460 0 R +/Annots [ 3509 0 R ] +>> endobj +3509 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [365.256 600.581 434.994 609.602] +/Subtype /Link +/A << /S /GoTo /D (the_sortlist_statement) >> +>> endobj +3504 0 obj << +/D [3506 0 R /XYZ 71.731 705.352 null] +>> endobj +409 0 obj << +/D [3506 0 R /XYZ 264.158 647.569 null] +>> endobj +3465 0 obj << +/D [3506 0 R /XYZ 71.731 638.932 null] +>> endobj +3466 0 obj << +/D [3506 0 R /XYZ 238.524 615.689 null] +>> endobj +3467 0 obj << +/D [3506 0 R /XYZ 290.268 602.738 null] +>> endobj +3474 0 obj << +/D [3506 0 R /XYZ 71.731 572.82 null] +>> endobj +3475 0 obj << +/D [3506 0 R /XYZ 71.731 572.82 null] +>> endobj +3476 0 obj << +/D [3506 0 R /XYZ 162.59 563.186 null] +>> endobj +3477 0 obj << +/D [3506 0 R /XYZ 227.147 563.186 null] +>> endobj +3478 0 obj << +/D [3506 0 R /XYZ 264.806 563.186 null] +>> endobj +3479 0 obj << +/D [3506 0 R /XYZ 323.984 563.186 null] +>> endobj +3480 0 obj << +/D [3506 0 R /XYZ 361.642 563.186 null] +>> endobj +3481 0 obj << +/D [3506 0 R /XYZ 184.109 551.53 null] +>> endobj +3482 0 obj << +/D [3506 0 R /XYZ 71.731 540.159 null] +>> endobj +3483 0 obj << +/D [3506 0 R /XYZ 266.378 528.616 null] +>> endobj +3484 0 obj << +/D [3506 0 R /XYZ 437.554 528.616 null] +>> endobj +3485 0 obj << +/D [3506 0 R /XYZ 217.871 515.664 null] +>> endobj +3486 0 obj << +/D [3506 0 R /XYZ 71.731 508.526 null] +>> endobj +3487 0 obj << +/D [3506 0 R /XYZ 200.995 497.732 null] +>> endobj +3488 0 obj << +/D [3506 0 R /XYZ 71.731 485.612 null] +>> endobj +3489 0 obj << +/D [3506 0 R /XYZ 71.731 480.631 null] +>> endobj +3490 0 obj << +/D [3506 0 R /XYZ 71.731 480.631 null] +>> endobj +3491 0 obj << +/D [3506 0 R /XYZ 71.731 399.137 null] +>> endobj +3492 0 obj << +/D [3506 0 R /XYZ 71.731 374.066 null] +>> endobj +3493 0 obj << +/D [3506 0 R /XYZ 71.731 318.586 null] +>> endobj +3494 0 obj << +/D [3506 0 R /XYZ 382.91 306.684 null] +>> endobj +3495 0 obj << +/D [3506 0 R /XYZ 71.731 286.594 null] +>> endobj +3496 0 obj << +/D [3506 0 R /XYZ 164.382 275.8 null] +>> endobj +3497 0 obj << +/D [3506 0 R /XYZ 71.731 263.68 null] +>> endobj +3498 0 obj << +/D [3506 0 R /XYZ 71.731 258.699 null] +>> endobj +3499 0 obj << +/D [3506 0 R /XYZ 182.819 249.2 null] +>> endobj +3500 0 obj << +/D [3506 0 R /XYZ 374.176 249.2 null] +>> endobj +955 0 obj << +/D [3506 0 R /XYZ 71.731 188.029 null] +>> endobj +413 0 obj << +/D [3506 0 R /XYZ 320.897 162.525 null] +>> endobj +3501 0 obj << +/D [3506 0 R /XYZ 71.731 154.073 null] +>> endobj +3505 0 obj << +/Font << /F31 805 0 R /F42 669 0 R /F25 681 0 R /F30 731 0 R /F37 1210 0 R /F60 1503 0 R /F62 669 0 R /F58 2248 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3555 0 obj << +/Length 3556 0 R +/Filter /FlateDecode +>> +stream +xYM&Ɯl͈ivvH6=rH-mbd+J),to.ٝmYd{U~*$ %IUqzKW +,cl8K#IaoW}ϝh7[넘pyMo74YN6 2Lע.査?a\iB·9hFpSfssu\64X4ܟk{]-J(_W|/*|#A>D(]e`.hJ#kT"{FlN7unbӆk*1rx߰`w櫲1uc/|njXLRFf>~`YmU\P#^X{cEӖpegW% I o|!Z1Q7JB88ZJ0nÆ rQHu+ESwmj+U_ T!{A]n.+.Ɍ9L(O|gW5g]1|0$Q +{I| FIgN (ldp`E$GԴs[%=EϽhEHiaDM3, eِiEysvd[w_XBh[wURu9 y[\pg<Ɋ kIz^rUl^#WUz$EVjK>o\;U!z/+VlE QiϝVX@20%Ig۽CZ-' 3/ ChBrBw`F4$A$A07 v~sP{LNPf*4)+C?ْuY 80 z9'e[^%#~m21ms ]~Ɔ)a]+۲ikzRՁcdunTjuܞUDϛ&uLAvm"g&J]W ;t>\5v|؎(iuߟg2`Gkdf ע2rb2Á^7Pnp+ a]@EO sj#şդG~Xמm%+ӶXU!"ffD&PrF#M=Œ eIan?͘ ޡҬj`pțT'Wqsdu$vѣ 3jg|-n%Lzn(M~%e:{ݥ>m:َ^'Kq+~eZ ¯ +|ȒOYjJ^ A^rN`ٜ1 GMRF$P/7҅?ˤ)) kendstream +endobj +3556 0 obj +2010 +endobj +3554 0 obj << +/Type /Page +/Contents 3555 0 R +/Resources 3553 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 3460 0 R +>> endobj +3551 0 obj << +/D [3554 0 R /XYZ 71.731 705.352 null] +>> endobj +3552 0 obj << +/D [3554 0 R /XYZ 71.731 717.307 null] +>> endobj +3502 0 obj << +/D [3554 0 R /XYZ 304.665 647.569 null] +>> endobj +3503 0 obj << +/D [3554 0 R /XYZ 386.307 634.618 null] +>> endobj +3510 0 obj << +/D [3554 0 R /XYZ 71.731 627.48 null] +>> endobj +3511 0 obj << +/D [3554 0 R /XYZ 71.731 583.644 null] +>> endobj +3512 0 obj << +/D [3554 0 R /XYZ 119.551 559.898 null] +>> endobj +3513 0 obj << +/D [3554 0 R /XYZ 205.846 559.898 null] +>> endobj +3514 0 obj << +/D [3554 0 R /XYZ 377.959 559.898 null] +>> endobj +3515 0 obj << +/D [3554 0 R /XYZ 422.243 559.898 null] +>> endobj +3516 0 obj << +/D [3554 0 R /XYZ 261.845 521.044 null] +>> endobj +3517 0 obj << +/D [3554 0 R /XYZ 391.926 521.044 null] +>> endobj +3518 0 obj << +/D [3554 0 R /XYZ 240.376 508.093 null] +>> endobj +3519 0 obj << +/D [3554 0 R /XYZ 494.948 508.093 null] +>> endobj +3520 0 obj << +/D [3554 0 R /XYZ 71.731 500.955 null] +>> endobj +3521 0 obj << +/D [3554 0 R /XYZ 71.731 465.089 null] +>> endobj +3522 0 obj << +/D [3554 0 R /XYZ 71.731 460.108 null] +>> endobj +3523 0 obj << +/D [3554 0 R /XYZ 164.878 450.609 null] +>> endobj +956 0 obj << +/D [3554 0 R /XYZ 71.731 412.751 null] +>> endobj +417 0 obj << +/D [3554 0 R /XYZ 214.82 387.246 null] +>> endobj +3524 0 obj << +/D [3554 0 R /XYZ 71.731 378.609 null] +>> endobj +3525 0 obj << +/D [3554 0 R /XYZ 71.731 368.646 null] +>> endobj +3526 0 obj << +/D [3554 0 R /XYZ 71.731 368.646 null] +>> endobj +3527 0 obj << +/D [3554 0 R /XYZ 71.731 358.235 null] +>> endobj +3528 0 obj << +/D [3554 0 R /XYZ 139.476 340.422 null] +>> endobj +3529 0 obj << +/D [3554 0 R /XYZ 497.359 340.422 null] +>> endobj +3530 0 obj << +/D [3554 0 R /XYZ 245.617 327.471 null] +>> endobj +3531 0 obj << +/D [3554 0 R /XYZ 396.828 327.471 null] +>> endobj +3532 0 obj << +/D [3554 0 R /XYZ 71.731 315.725 null] +>> endobj +3533 0 obj << +/D [3554 0 R /XYZ 71.731 315.725 null] +>> endobj +3534 0 obj << +/D [3554 0 R /XYZ 71.731 304.437 null] +>> endobj +3535 0 obj << +/D [3554 0 R /XYZ 139.476 286.624 null] +>> endobj +3536 0 obj << +/D [3554 0 R /XYZ 139.476 273.673 null] +>> endobj +3537 0 obj << +/D [3554 0 R /XYZ 187.516 260.721 null] +>> endobj +3538 0 obj << +/D [3554 0 R /XYZ 262.772 260.721 null] +>> endobj +3539 0 obj << +/D [3554 0 R /XYZ 367.419 260.721 null] +>> endobj +3540 0 obj << +/D [3554 0 R /XYZ 71.731 235.651 null] +>> endobj +3541 0 obj << +/D [3554 0 R /XYZ 71.731 235.651 null] +>> endobj +3542 0 obj << +/D [3554 0 R /XYZ 71.731 224.736 null] +>> endobj +3543 0 obj << +/D [3554 0 R /XYZ 139.476 206.923 null] +>> endobj +3544 0 obj << +/D [3554 0 R /XYZ 139.476 206.923 null] +>> endobj +3545 0 obj << +/D [3554 0 R /XYZ 71.731 181.852 null] +>> endobj +3546 0 obj << +/D [3554 0 R /XYZ 71.731 181.852 null] +>> endobj +3547 0 obj << +/D [3554 0 R /XYZ 71.731 170.938 null] +>> endobj +3548 0 obj << +/D [3554 0 R /XYZ 139.476 153.125 null] +>> endobj +3549 0 obj << +/D [3554 0 R /XYZ 220.711 140.174 null] +>> endobj +3553 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F37 1210 0 R /F30 731 0 R /F50 1590 0 R /F62 669 0 R /F58 2248 0 R /F42 669 0 R /F39 1242 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3586 0 obj << +/Length 3587 0 R +/Filter /FlateDecode +>> +stream +xڽXɒ6_[&o28)#0"ABQ~ĿpƧHBF׍&(!"Dhٗ/F}ط`MKY0׷?CYL6C/=S˛~{ǻzq-=֊ҫv՞o}|a\I0·k$IMOc97)}T!EQp2I6 Dy*xɫfOu i$4'AIk?nqB=R:O2Rxy6w(E]y@>!]:/UW[lYEĮg]578y;ѫ2X%`jϊyS$+e7vb&ᯀ )dV+D)j+q"ڣJ2/*W ( +ƷzO'pS;zXi q>hqLw;c_%(AaH^K+J< ueLƕ>!Pue=ԝ=?ֲ۾kfY5`ކfY7J0Ӭjk,5~UZuL5 +CЀ^C)*"A?;vɞ3ms(&,H?Equ(J㺇A`mbɛGղxӖROZ =0- qDؒٸ#C>f4*!EamX%R"'9mjm?*T`mūvW쬖L61tJ;,tZM.$Dim Zӈfȹ ;ī)()+Z< Vɜj?*W6ѶK`q,Xq wpGw vg8`i] )HM"%ASRO.hҦ҂vcJ;0k;@j P ˎjYF RO=)Gˣ_rt`W%9$|͔iZlb5Nd_l!KqWW?O!\h`)s}sc|4(?@aWr㦇w=npj0Em +Q$7梮V3E8DSh dp ȒMD)\udgIw"$|+VcCendstream +endobj +3587 0 obj +1620 +endobj +3585 0 obj << +/Type /Page +/Contents 3586 0 R +/Resources 3584 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 3460 0 R +/Annots [ 3588 0 R ] +>> endobj +3588 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [247.938 550.768 292.769 559.679] +/Subtype /Link +/A << /S /GoTo /D (dynamic_update) >> +>> endobj +3583 0 obj << +/D [3585 0 R /XYZ 71.731 704.356 null] +>> endobj +3550 0 obj << +/D [3585 0 R /XYZ 71.731 657.532 null] +>> endobj +3557 0 obj << +/D [3585 0 R /XYZ 71.731 596.76 null] +>> endobj +3558 0 obj << +/D [3585 0 R /XYZ 71.731 596.76 null] +>> endobj +3559 0 obj << +/D [3585 0 R /XYZ 71.731 581.786 null] +>> endobj +3560 0 obj << +/D [3585 0 R /XYZ 139.476 565.876 null] +>> endobj +3561 0 obj << +/D [3585 0 R /XYZ 403.473 552.925 null] +>> endobj +3562 0 obj << +/D [3585 0 R /XYZ 71.731 516.96 null] +>> endobj +3563 0 obj << +/D [3585 0 R /XYZ 71.731 516.96 null] +>> endobj +3564 0 obj << +/D [3585 0 R /XYZ 71.731 503.988 null] +>> endobj +3565 0 obj << +/D [3585 0 R /XYZ 71.731 503.988 null] +>> endobj +3566 0 obj << +/D [3585 0 R /XYZ 71.731 491.037 null] +>> endobj +3567 0 obj << +/D [3585 0 R /XYZ 71.731 491.037 null] +>> endobj +3568 0 obj << +/D [3585 0 R /XYZ 71.731 476.183 null] +>> endobj +3569 0 obj << +/D [3585 0 R /XYZ 71.731 476.183 null] +>> endobj +3570 0 obj << +/D [3585 0 R /XYZ 71.731 463.231 null] +>> endobj +3571 0 obj << +/D [3585 0 R /XYZ 139.476 447.321 null] +>> endobj +3572 0 obj << +/D [3585 0 R /XYZ 71.731 414.28 null] +>> endobj +957 0 obj << +/D [3585 0 R /XYZ 71.731 344.706 null] +>> endobj +421 0 obj << +/D [3585 0 R /XYZ 277.703 319.202 null] +>> endobj +3573 0 obj << +/D [3585 0 R /XYZ 71.731 313.075 null] +>> endobj +3574 0 obj << +/D [3585 0 R /XYZ 243.235 300.273 null] +>> endobj +3575 0 obj << +/D [3585 0 R /XYZ 469.822 300.273 null] +>> endobj +3576 0 obj << +/D [3585 0 R /XYZ 71.731 293.135 null] +>> endobj +3577 0 obj << +/D [3585 0 R /XYZ 280.186 282.34 null] +>> endobj +3578 0 obj << +/D [3585 0 R /XYZ 330.058 217.583 null] +>> endobj +3579 0 obj << +/D [3585 0 R /XYZ 71.731 197.494 null] +>> endobj +3580 0 obj << +/D [3585 0 R /XYZ 71.731 174.58 null] +>> endobj +3581 0 obj << +/D [3585 0 R /XYZ 71.731 169.598 null] +>> endobj +3582 0 obj << +/D [3585 0 R /XYZ 71.731 169.598 null] +>> endobj +3584 0 obj << +/Font << /F31 805 0 R /F62 669 0 R /F58 2248 0 R /F30 731 0 R /F25 681 0 R /F37 1210 0 R /F42 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3623 0 obj << +/Length 3624 0 R +/Filter /FlateDecode +>> +stream +xXmo6Be`bER$%S_nX +d0+ +FrҴQ$e)!@H{yxw4RÑHPR(#Q~#! +aQ"E" KWp@鲧;/SznVyoXvu۸TĪ)O^*`QJ(tS8p2%L ӈK΢";@pabOD&YQ9;}Ͷ,1VǯeTlgvFb]wVnqkY](},iث` +{F!Yn6nRG\7E]^VV7DJo^,4~sTlxCZ0IdEJ-Ws( )9x3 !f`XU繧 9 2;p<J/O4Uus}dp#{@&ˑvde|+2GqXBxuwTNUZA/9+"@Sʭ ̧pDFejE'8"dc;FR֫'Ww桪ڰ/Zh8~M0d$C3rsEم}AJ-r _V¯Z5QZj(*e,ݠ5)StGI{YUsho ~"i[/ +4V}70F ƀc^֕J+BK4?W+w}٪j^J1ʘiTЋb,yaO((WxД 12 5iWv}l8,uM|frZ6~l}e1nh%p ]~(M=wB W˜zl[BX_~yxLRQGV J.+^o;}ɃoF.!i6gRa>4@bmf,Ěi>%Z|MWfZB91-k9]+9Tm\+NxTVS( -M +w6uVVjA/iQz=G^:<[qZ)|Y%6$&A eIoD FvЖ422[ziD4%!\]4z[F뻫NVabXdbx@^n^B}ހ՞fď]3Ɗq &Xm )FD}{B*Nٜ6BRͅ6x}~g@x$D~S#IOIPendstream +endobj +3624 0 obj +1364 +endobj +3622 0 obj << +/Type /Page +/Contents 3623 0 R +/Resources 3621 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 3460 0 R +>> endobj +3619 0 obj << +/D [3622 0 R /XYZ 71.731 705.352 null] +>> endobj +3620 0 obj << +/D [3622 0 R /XYZ 71.731 717.307 null] +>> endobj +958 0 obj << +/D [3622 0 R /XYZ 71.731 430.982 null] +>> endobj +425 0 obj << +/D [3622 0 R /XYZ 303.905 401.573 null] +>> endobj +3589 0 obj << +/D [3622 0 R /XYZ 71.731 391.431 null] +>> endobj +3590 0 obj << +/D [3622 0 R /XYZ 157.21 379.954 null] +>> endobj +3591 0 obj << +/D [3622 0 R /XYZ 141.07 368.298 null] +>> endobj +3592 0 obj << +/D [3622 0 R /XYZ 184.109 368.298 null] +>> endobj +3593 0 obj << +/D [3622 0 R /XYZ 141.07 356.641 null] +>> endobj +3594 0 obj << +/D [3622 0 R /XYZ 221.767 356.641 null] +>> endobj +3595 0 obj << +/D [3622 0 R /XYZ 141.07 344.985 null] +>> endobj +3596 0 obj << +/D [3622 0 R /XYZ 221.767 344.985 null] +>> endobj +3597 0 obj << +/D [3622 0 R /XYZ 141.07 333.329 null] +>> endobj +3598 0 obj << +/D [3622 0 R /XYZ 178.729 333.329 null] +>> endobj +3599 0 obj << +/D [3622 0 R /XYZ 141.07 321.673 null] +>> endobj +3600 0 obj << +/D [3622 0 R /XYZ 205.628 321.673 null] +>> endobj +3601 0 obj << +/D [3622 0 R /XYZ 141.07 310.016 null] +>> endobj +3602 0 obj << +/D [3622 0 R /XYZ 237.907 310.016 null] +>> endobj +3603 0 obj << +/D [3622 0 R /XYZ 141.07 298.36 null] +>> endobj +3604 0 obj << +/D [3622 0 R /XYZ 178.729 298.36 null] +>> endobj +3605 0 obj << +/D [3622 0 R /XYZ 237.907 298.36 null] +>> endobj +3606 0 obj << +/D [3622 0 R /XYZ 297.085 298.36 null] +>> endobj +959 0 obj << +/D [3622 0 R /XYZ 71.731 265.73 null] +>> endobj +429 0 obj << +/D [3622 0 R /XYZ 382.808 237.369 null] +>> endobj +3607 0 obj << +/D [3622 0 R /XYZ 71.731 227.004 null] +>> endobj +3608 0 obj << +/D [3622 0 R /XYZ 137.534 217.245 null] +>> endobj +3609 0 obj << +/D [3622 0 R /XYZ 71.731 212.144 null] +>> endobj +3610 0 obj << +/D [3622 0 R /XYZ 137.534 199.312 null] +>> endobj +3611 0 obj << +/D [3622 0 R /XYZ 449.31 199.312 null] +>> endobj +3612 0 obj << +/D [3622 0 R /XYZ 119.551 186.36 null] +>> endobj +3613 0 obj << +/D [3622 0 R /XYZ 268.411 186.36 null] +>> endobj +3614 0 obj << +/D [3622 0 R /XYZ 276.172 173.409 null] +>> endobj +3615 0 obj << +/D [3622 0 R /XYZ 406.292 173.409 null] +>> endobj +3621 0 obj << +/Font << /F31 805 0 R /F30 731 0 R /F25 681 0 R /F40 669 0 R /F37 1210 0 R /F60 1503 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3670 0 obj << +/Length 3671 0 R +/Filter /FlateDecode +>> +stream +xڽYێq6`3"uGLnI-ɶВ2=N~>KQ$;A@[ET)n<؇/$~6Yn,cl0?0>%ioOo.Bo_~WA4OhR4 Fhb?~ٚ@x, >4!91ߚ¥ɟOzE^vpECg mQ vE<\v\S6g] f#b{}Y>Ybf8*-jƌ8 mC@@IC!YŇ"9^ bz) +qD+}9ɏۚx0׼띇LUuWzײ*: I!1ЌַNh2`N̚%]o1όXHf9jOOor9ߏC Z]'eV@44(1 +qj239Êsef$E E%n&h_r:G$8KWf%Iٹө/ +8W6o*HRR#|%G-kfGЙ +3[[Y +tZt%n'B+]-O;˳E#Gip>fVg-.yъLpEn.țN5[!O?L# Z_J ,ukϸ!Z4=oֺ4z$a{86kQ@<_h)X~>FB!nPOmW7K[9`ZpF(]>:~6&^M/~`qBN mfɟ5=R1~%OR"®{F~m"S8R-~qaNr,??[LelF̞LRm:-̃mckju-I7=,aTPZ$UYM)M7++{4-2_ Cxj;\zBp[+BX$Po&l@Y8M+_-J?[X97=IJoqەBMт9 +~oL4ss1mZ49-ʊ-edzB>8rc^r#[3)Q] @! ytny][άiBb/0:QC WpוnJYp^\m7_dR2ٖqӈYx? bh;/x69 Gjv39A8OyWhR;/+B08q:_F?ݛ?8 +endstream +endobj +3671 0 obj +2214 +endobj +3669 0 obj << +/Type /Page +/Contents 3670 0 R +/Resources 3668 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 3672 0 R +>> endobj +3667 0 obj << +/D [3669 0 R /XYZ 71.731 705.352 null] +>> endobj +3616 0 obj << +/D [3669 0 R /XYZ 71.731 657.532 null] +>> endobj +3617 0 obj << +/D [3669 0 R /XYZ 254.224 634.618 null] +>> endobj +3618 0 obj << +/D [3669 0 R /XYZ 290.767 634.618 null] +>> endobj +3625 0 obj << +/D [3669 0 R /XYZ 455.786 603.734 null] +>> endobj +3626 0 obj << +/D [3669 0 R /XYZ 331.494 590.782 null] +>> endobj +3627 0 obj << +/D [3669 0 R /XYZ 290.846 577.831 null] +>> endobj +3628 0 obj << +/D [3669 0 R /XYZ 71.731 559.799 null] +>> endobj +3629 0 obj << +/D [3669 0 R /XYZ 137.534 546.947 null] +>> endobj +3630 0 obj << +/D [3669 0 R /XYZ 439.567 533.996 null] +>> endobj +3631 0 obj << +/D [3669 0 R /XYZ 71.731 513.906 null] +>> endobj +3632 0 obj << +/D [3669 0 R /XYZ 119.551 477.209 null] +>> endobj +3633 0 obj << +/D [3669 0 R /XYZ 297.064 477.209 null] +>> endobj +3634 0 obj << +/D [3669 0 R /XYZ 367.618 477.209 null] +>> endobj +3635 0 obj << +/D [3669 0 R /XYZ 71.731 444.168 null] +>> endobj +3636 0 obj << +/D [3669 0 R /XYZ 137.534 433.373 null] +>> endobj +3637 0 obj << +/D [3669 0 R /XYZ 270.374 420.422 null] +>> endobj +3638 0 obj << +/D [3669 0 R /XYZ 71.731 413.418 null] +>> endobj +3639 0 obj << +/D [3669 0 R /XYZ 349.227 402.489 null] +>> endobj +3640 0 obj << +/D [3669 0 R /XYZ 231.849 389.538 null] +>> endobj +3641 0 obj << +/D [3669 0 R /XYZ 119.551 376.586 null] +>> endobj +3642 0 obj << +/D [3669 0 R /XYZ 408.026 376.586 null] +>> endobj +3643 0 obj << +/D [3669 0 R /XYZ 444.828 376.586 null] +>> endobj +3644 0 obj << +/D [3669 0 R /XYZ 165.887 363.635 null] +>> endobj +3645 0 obj << +/D [3669 0 R /XYZ 451.562 363.635 null] +>> endobj +3646 0 obj << +/D [3669 0 R /XYZ 159.122 350.684 null] +>> endobj +3647 0 obj << +/D [3669 0 R /XYZ 307.354 350.684 null] +>> endobj +3648 0 obj << +/D [3669 0 R /XYZ 441.191 350.684 null] +>> endobj +3649 0 obj << +/D [3669 0 R /XYZ 71.731 332.651 null] +>> endobj +3650 0 obj << +/D [3669 0 R /XYZ 71.731 332.651 null] +>> endobj +3651 0 obj << +/D [3669 0 R /XYZ 132.004 306.848 null] +>> endobj +3652 0 obj << +/D [3669 0 R /XYZ 374.949 306.848 null] +>> endobj +3653 0 obj << +/D [3669 0 R /XYZ 71.731 299.71 null] +>> endobj +3654 0 obj << +/D [3669 0 R /XYZ 137.534 288.916 null] +>> endobj +3655 0 obj << +/D [3669 0 R /XYZ 266.657 288.916 null] +>> endobj +3656 0 obj << +/D [3669 0 R /XYZ 356.221 288.916 null] +>> endobj +3657 0 obj << +/D [3669 0 R /XYZ 309.486 275.964 null] +>> endobj +3658 0 obj << +/D [3669 0 R /XYZ 456.035 275.964 null] +>> endobj +3659 0 obj << +/D [3669 0 R /XYZ 71.731 229.972 null] +>> endobj +3660 0 obj << +/D [3669 0 R /XYZ 238.255 219.177 null] +>> endobj +960 0 obj << +/D [3669 0 R /XYZ 71.731 189.125 null] +>> endobj +433 0 obj << +/D [3669 0 R /XYZ 345.781 161.872 null] +>> endobj +3661 0 obj << +/D [3669 0 R /XYZ 71.731 148.94 null] +>> endobj +3668 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F30 731 0 R /F40 669 0 R /F37 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3717 0 obj << +/Length 3718 0 R +/Filter /FlateDecode +>> +stream +xڽX[o6 +bN$%QBˊn@74SQEBed߱CmaEs?(.%%b{xBaV=Ҁ0D0]?'x?%i׻nun|/Wk˘7/pyU^ܶKUwߪR[z*bd6|5OcRXD{`fyyZǑXac̏y|f-~0b zyD) ԧfbn)%qwC[h@Dý=8!ߣmRldǧ((0u(H”pubMLElV7*[P+-4޽jdl uQ6S.~rE%8 jgQ_ZV7 ezoFI #YA3Qح\{K3Ti\xsu9tնUh2\;)z[崤9 +(!& Go_a wU{}DL!IqU +.5“D/#iD1\}°Yqtp}W%Zn|71A]]UB⥬ʵl} +aVX3>*)PV+-KԶY`6p+˲}Fjȼtlu}/qr +٨low•VU޶e+ӹzm[0fW3:0CY^Z1l)mq6;+60G3nJ6Sqny/!w&ȳۥ|++<81gtUK6ʲ?6'}޷ʢdvIhNɪ]FP$L:uSM~,hմG}UQyÛ^{ SY1 WLӕMHf@c`QܾNs6</5y_{tHYok g4^c0~6>7Gގvٜ9^Ԟ'}1v7upFJ(F#_~O~8#_8yAzendstream +endobj +3718 0 obj +1858 +endobj +3716 0 obj << +/Type /Page +/Contents 3717 0 R +/Resources 3715 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 3672 0 R +/Annots [ 3719 0 R ] +>> endobj +3719 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [464.244 552.641 509.075 561.662] +/Subtype /Link +/A << /S /GoTo /D (DNSSEC) >> +>> endobj +3714 0 obj << +/D [3716 0 R /XYZ 71.731 705.352 null] +>> endobj +3662 0 obj << +/D [3716 0 R /XYZ 141.07 647.569 null] +>> endobj +3663 0 obj << +/D [3716 0 R /XYZ 179.24 647.569 null] +>> endobj +3664 0 obj << +/D [3716 0 R /XYZ 217.221 647.569 null] +>> endobj +3665 0 obj << +/D [3716 0 R /XYZ 255.203 647.569 null] +>> endobj +3666 0 obj << +/D [3716 0 R /XYZ 293.184 647.569 null] +>> endobj +3673 0 obj << +/D [3716 0 R /XYZ 141.07 635.913 null] +>> endobj +3674 0 obj << +/D [3716 0 R /XYZ 151.83 635.913 null] +>> endobj +3675 0 obj << +/D [3716 0 R /XYZ 190 635.913 null] +>> endobj +3676 0 obj << +/D [3716 0 R /XYZ 227.981 635.913 null] +>> endobj +3677 0 obj << +/D [3716 0 R /XYZ 265.962 635.913 null] +>> endobj +3678 0 obj << +/D [3716 0 R /XYZ 303.944 635.913 null] +>> endobj +3679 0 obj << +/D [3716 0 R /XYZ 352.873 635.913 null] +>> endobj +961 0 obj << +/D [3716 0 R /XYZ 71.731 603.283 null] +>> endobj +437 0 obj << +/D [3716 0 R /XYZ 424.685 574.922 null] +>> endobj +3680 0 obj << +/D [3716 0 R /XYZ 71.731 564.557 null] +>> endobj +3681 0 obj << +/D [3716 0 R /XYZ 137.534 554.797 null] +>> endobj +3682 0 obj << +/D [3716 0 R /XYZ 71.731 495.854 null] +>> endobj +3683 0 obj << +/D [3716 0 R /XYZ 137.534 485.059 null] +>> endobj +962 0 obj << +/D [3716 0 R /XYZ 71.731 455.007 null] +>> endobj +441 0 obj << +/D [3716 0 R /XYZ 291.582 427.754 null] +>> endobj +3684 0 obj << +/D [3716 0 R /XYZ 71.731 417.613 null] +>> endobj +3685 0 obj << +/D [3716 0 R /XYZ 146.45 406.136 null] +>> endobj +3686 0 obj << +/D [3716 0 R /XYZ 200.248 406.136 null] +>> endobj +3687 0 obj << +/D [3716 0 R /XYZ 205.628 406.136 null] +>> endobj +3688 0 obj << +/D [3716 0 R /XYZ 237.907 394.479 null] +>> endobj +3689 0 obj << +/D [3716 0 R /XYZ 264.806 382.823 null] +>> endobj +3690 0 obj << +/D [3716 0 R /XYZ 275.565 371.167 null] +>> endobj +3691 0 obj << +/D [3716 0 R /XYZ 151.83 359.511 null] +>> endobj +3692 0 obj << +/D [3716 0 R /XYZ 162.59 359.511 null] +>> endobj +3693 0 obj << +/D [3716 0 R /XYZ 151.83 347.854 null] +>> endobj +3694 0 obj << +/D [3716 0 R /XYZ 248.666 347.854 null] +>> endobj +3695 0 obj << +/D [3716 0 R /XYZ 151.83 336.198 null] +>> endobj +3696 0 obj << +/D [3716 0 R /XYZ 162.59 336.198 null] +>> endobj +963 0 obj << +/D [3716 0 R /XYZ 71.731 303.568 null] +>> endobj +445 0 obj << +/D [3716 0 R /XYZ 370.485 275.207 null] +>> endobj +3697 0 obj << +/D [3716 0 R /XYZ 71.731 264.842 null] +>> endobj +3698 0 obj << +/D [3716 0 R /XYZ 137.534 255.083 null] +>> endobj +3699 0 obj << +/D [3716 0 R /XYZ 314.099 255.083 null] +>> endobj +3700 0 obj << +/D [3716 0 R /XYZ 71.731 222.042 null] +>> endobj +3701 0 obj << +/D [3716 0 R /XYZ 141.957 211.247 null] +>> endobj +3702 0 obj << +/D [3716 0 R /XYZ 350.332 198.296 null] +>> endobj +3703 0 obj << +/D [3716 0 R /XYZ 119.551 185.344 null] +>> endobj +3704 0 obj << +/D [3716 0 R /XYZ 374.939 185.344 null] +>> endobj +3705 0 obj << +/D [3716 0 R /XYZ 119.551 172.393 null] +>> endobj +3706 0 obj << +/D [3716 0 R /XYZ 317.675 172.393 null] +>> endobj +3707 0 obj << +/D [3716 0 R /XYZ 394.038 172.393 null] +>> endobj +3708 0 obj << +/D [3716 0 R /XYZ 341.476 159.442 null] +>> endobj +3709 0 obj << +/D [3716 0 R /XYZ 435.89 146.49 null] +>> endobj +3710 0 obj << +/D [3716 0 R /XYZ 404.817 133.539 null] +>> endobj +3715 0 obj << +/Font << /F31 805 0 R /F60 1503 0 R /F37 1210 0 R /F40 669 0 R /F25 681 0 R /F30 731 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3744 0 obj << +/Length 3745 0 R +/Filter /FlateDecode +>> +stream +xڥXK6_aV\zPžEEEh=h%:&"K(y-;|ɢ$; D73MRBV*~|uOQǭ:X~˚D:F^񦮮>uw߳- U9ӫ'08Aa"h I4G]1b^?n Lp-ٳ>2V@0J"{DtYԫ'uUvt9?DWg^rc ẅg]l1< (9b"ρ)E)+FX%iLᨅfUXi}ˎ7V!Yܘ\xSdѝg\^;lprݸ xFKt[R3)S5UV^?TBL2#{$o~5{4;0pu2zT|RŊ%P"v c 9~/HI$ʰ8QS'Y;fκe.YU0LrwG6~U{ Unv^VS\,fL (\8Q^ ` nfaLЁ-lC|cHM6k-ΚDtrP^ȧwtN`PB};VӲioH٭UQ ́O/Ud(/:&ͦ +L8MKfv=?{VuǜtmvҪ<{"sk^W]6)PF )]!Lt0٤}KVxV͌Rp+[Ϙ%(\,b24NzTVڢ4TuZh] 7xA5cG:uu2^MUjtu%;Ҿ<ѳs=FrxєƂ@plmt  `byPH|iJ)F}Dx qpQh"ǩdhyZIGP( 58 QN8d=?tuj{id*l)E"酜t\l"HNƪɒ Ӄ+1N2wj_#(*p%Օ[}QYĮB \2VޮX\mBxgĉA]Q#MwJ#:0i냤@ +~0_k-d̪;ؿg^|/-:C3acU_DF`H\* MG3|3\5t8 ;Ȝd$1JRRdDIGJےzE;[ň$r:aj iendstream +endobj +3745 0 obj +1520 +endobj +3743 0 obj << +/Type /Page +/Contents 3744 0 R +/Resources 3742 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 3672 0 R +>> endobj +3741 0 obj << +/D [3743 0 R /XYZ 71.731 705.352 null] +>> endobj +3711 0 obj << +/D [3743 0 R /XYZ 71.731 657.532 null] +>> endobj +3712 0 obj << +/D [3743 0 R /XYZ 213.079 647.569 null] +>> endobj +3713 0 obj << +/D [3743 0 R /XYZ 478.161 647.569 null] +>> endobj +3720 0 obj << +/D [3743 0 R /XYZ 71.731 614.528 null] +>> endobj +3721 0 obj << +/D [3743 0 R /XYZ 251.555 603.734 null] +>> endobj +3722 0 obj << +/D [3743 0 R /XYZ 428.1 603.734 null] +>> endobj +3723 0 obj << +/D [3743 0 R /XYZ 119.551 577.831 null] +>> endobj +3724 0 obj << +/D [3743 0 R /XYZ 508.627 577.831 null] +>> endobj +3725 0 obj << +/D [3743 0 R /XYZ 409.858 564.88 null] +>> endobj +3726 0 obj << +/D [3743 0 R /XYZ 71.731 557.742 null] +>> endobj +3727 0 obj << +/D [3743 0 R /XYZ 71.731 526.857 null] +>> endobj +3728 0 obj << +/D [3743 0 R /XYZ 178.191 516.063 null] +>> endobj +3729 0 obj << +/D [3743 0 R /XYZ 231.998 503.111 null] +>> endobj +3730 0 obj << +/D [3743 0 R /XYZ 344.584 490.16 null] +>> endobj +3731 0 obj << +/D [3743 0 R /XYZ 471.855 490.16 null] +>> endobj +3732 0 obj << +/D [3743 0 R /XYZ 192.327 477.209 null] +>> endobj +3733 0 obj << +/D [3743 0 R /XYZ 71.731 472.108 null] +>> endobj +3734 0 obj << +/D [3743 0 R /XYZ 390.023 459.276 null] +>> endobj +3735 0 obj << +/D [3743 0 R /XYZ 71.731 447.157 null] +>> endobj +964 0 obj << +/D [3743 0 R /XYZ 71.731 160.246 null] +>> endobj +3742 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F30 731 0 R /F37 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3859 0 obj << +/Length 3860 0 R +/Filter /FlateDecode +>> +stream +xY[oV_Sshv{Q[m"6NPwwk8qon xR #&L.gwfR1u-⇟~dպZgM>x&=홱Y6$rQګkY..~ %; \JD sr2"J*J>gih*2L´Ƀ$ +x< jBz>h +k_Dژf%H$ )P@vE1гt}Kg/PO.佒(w +gIPY2싔ۛޅu +Uسw5C1_q}Ӭ;{0`âX$A[QQ#Bpbo>\ÐTl?0:KCl)'rj0|8lqsO[._29z>FUN\c$uRZd|nNçfQcE'~w܍4i\1;gh~?ʣ1"~q"ϏQEtCX$ Έ> endobj +3856 0 obj << +/D [3858 0 R /XYZ 71.731 704.356 null] +>> endobj +449 0 obj << +/D [3858 0 R /XYZ 293.376 647.066 null] +>> endobj +3736 0 obj << +/D [3858 0 R /XYZ 71.731 636.925 null] +>> endobj +3737 0 obj << +/D [3858 0 R /XYZ 146.45 625.448 null] +>> endobj +3738 0 obj << +/D [3858 0 R /XYZ 200.248 625.448 null] +>> endobj +3739 0 obj << +/D [3858 0 R /XYZ 205.628 625.448 null] +>> endobj +3740 0 obj << +/D [3858 0 R /XYZ 243.287 625.448 null] +>> endobj +3746 0 obj << +/D [3858 0 R /XYZ 141.07 602.135 null] +>> endobj +3747 0 obj << +/D [3858 0 R /XYZ 232.527 602.135 null] +>> endobj +3748 0 obj << +/D [3858 0 R /XYZ 141.07 590.479 null] +>> endobj +3749 0 obj << +/D [3858 0 R /XYZ 227.147 590.479 null] +>> endobj +3750 0 obj << +/D [3858 0 R /XYZ 141.07 578.823 null] +>> endobj +3751 0 obj << +/D [3858 0 R /XYZ 243.287 578.823 null] +>> endobj +3752 0 obj << +/D [3858 0 R /XYZ 141.07 567.166 null] +>> endobj +3753 0 obj << +/D [3858 0 R /XYZ 232.527 567.166 null] +>> endobj +3754 0 obj << +/D [3858 0 R /XYZ 141.07 555.51 null] +>> endobj +3755 0 obj << +/D [3858 0 R /XYZ 237.907 555.51 null] +>> endobj +3756 0 obj << +/D [3858 0 R /XYZ 340.123 555.51 null] +>> endobj +3757 0 obj << +/D [3858 0 R /XYZ 141.07 543.854 null] +>> endobj +3758 0 obj << +/D [3858 0 R /XYZ 291.705 543.854 null] +>> endobj +3759 0 obj << +/D [3858 0 R /XYZ 141.07 532.198 null] +>> endobj +3760 0 obj << +/D [3858 0 R /XYZ 167.969 520.541 null] +>> endobj +3761 0 obj << +/D [3858 0 R /XYZ 211.331 520.541 null] +>> endobj +3762 0 obj << +/D [3858 0 R /XYZ 243.609 520.541 null] +>> endobj +3763 0 obj << +/D [3858 0 R /XYZ 302.787 520.541 null] +>> endobj +3764 0 obj << +/D [3858 0 R /XYZ 313.547 520.541 null] +>> endobj +3765 0 obj << +/D [3858 0 R /XYZ 356.908 520.541 null] +>> endobj +3766 0 obj << +/D [3858 0 R /XYZ 389.187 520.541 null] +>> endobj +3767 0 obj << +/D [3858 0 R /XYZ 141.07 508.885 null] +>> endobj +3768 0 obj << +/D [3858 0 R /XYZ 221.767 508.885 null] +>> endobj +3769 0 obj << +/D [3858 0 R /XYZ 246.515 508.885 null] +>> endobj +3770 0 obj << +/D [3858 0 R /XYZ 271.262 508.885 null] +>> endobj +3771 0 obj << +/D [3858 0 R /XYZ 141.07 497.229 null] +>> endobj +3772 0 obj << +/D [3858 0 R /XYZ 189.489 497.229 null] +>> endobj +3773 0 obj << +/D [3858 0 R /XYZ 141.07 485.573 null] +>> endobj +3774 0 obj << +/D [3858 0 R /XYZ 178.729 485.573 null] +>> endobj +3775 0 obj << +/D [3858 0 R /XYZ 141.07 473.916 null] +>> endobj +3776 0 obj << +/D [3858 0 R /XYZ 200.248 473.916 null] +>> endobj +3777 0 obj << +/D [3858 0 R /XYZ 224.995 473.916 null] +>> endobj +3778 0 obj << +/D [3858 0 R /XYZ 141.07 462.26 null] +>> endobj +3779 0 obj << +/D [3858 0 R /XYZ 173.349 450.604 null] +>> endobj +3780 0 obj << +/D [3858 0 R /XYZ 216.71 450.604 null] +>> endobj +3781 0 obj << +/D [3858 0 R /XYZ 248.989 450.604 null] +>> endobj +3782 0 obj << +/D [3858 0 R /XYZ 308.167 450.604 null] +>> endobj +3783 0 obj << +/D [3858 0 R /XYZ 318.927 450.604 null] +>> endobj +3784 0 obj << +/D [3858 0 R /XYZ 362.288 450.604 null] +>> endobj +3785 0 obj << +/D [3858 0 R /XYZ 394.567 450.604 null] +>> endobj +3786 0 obj << +/D [3858 0 R /XYZ 141.07 438.948 null] +>> endobj +3787 0 obj << +/D [3858 0 R /XYZ 205.628 438.948 null] +>> endobj +3788 0 obj << +/D [3858 0 R /XYZ 141.07 427.291 null] +>> endobj +3789 0 obj << +/D [3858 0 R /XYZ 227.147 427.291 null] +>> endobj +3790 0 obj << +/D [3858 0 R /XYZ 141.07 415.635 null] +>> endobj +3791 0 obj << +/D [3858 0 R /XYZ 254.046 415.635 null] +>> endobj +3792 0 obj << +/D [3858 0 R /XYZ 141.07 403.979 null] +>> endobj +3793 0 obj << +/D [3858 0 R /XYZ 194.868 403.979 null] +>> endobj +3794 0 obj << +/D [3858 0 R /XYZ 227.147 403.979 null] +>> endobj +3795 0 obj << +/D [3858 0 R /XYZ 286.325 403.979 null] +>> endobj +3796 0 obj << +/D [3858 0 R /XYZ 329.686 403.979 null] +>> endobj +3797 0 obj << +/D [3858 0 R /XYZ 361.965 403.979 null] +>> endobj +3798 0 obj << +/D [3858 0 R /XYZ 410.383 403.979 null] +>> endobj +3799 0 obj << +/D [3858 0 R /XYZ 437.282 403.979 null] +>> endobj +3800 0 obj << +/D [3858 0 R /XYZ 470.305 403.979 null] +>> endobj +3801 0 obj << +/D [3858 0 R /XYZ 141.07 392.323 null] +>> endobj +3802 0 obj << +/D [3858 0 R /XYZ 248.666 392.323 null] +>> endobj +3803 0 obj << +/D [3858 0 R /XYZ 141.07 380.666 null] +>> endobj +3804 0 obj << +/D [3858 0 R /XYZ 264.806 380.666 null] +>> endobj +3805 0 obj << +/D [3858 0 R /XYZ 141.07 369.01 null] +>> endobj +3806 0 obj << +/D [3858 0 R /XYZ 270.186 369.01 null] +>> endobj +3807 0 obj << +/D [3858 0 R /XYZ 141.07 357.354 null] +>> endobj +3808 0 obj << +/D [3858 0 R /XYZ 264.806 357.354 null] +>> endobj +3809 0 obj << +/D [3858 0 R /XYZ 141.07 345.698 null] +>> endobj +3810 0 obj << +/D [3858 0 R /XYZ 270.186 345.698 null] +>> endobj +3811 0 obj << +/D [3858 0 R /XYZ 141.07 334.041 null] +>> endobj +3812 0 obj << +/D [3858 0 R /XYZ 189.489 334.041 null] +>> endobj +3813 0 obj << +/D [3858 0 R /XYZ 254.046 334.041 null] +>> endobj +3814 0 obj << +/D [3858 0 R /XYZ 141.07 322.385 null] +>> endobj +3815 0 obj << +/D [3858 0 R /XYZ 189.489 322.385 null] +>> endobj +3816 0 obj << +/D [3858 0 R /XYZ 227.47 322.385 null] +>> endobj +3817 0 obj << +/D [3858 0 R /XYZ 265.451 322.385 null] +>> endobj +3818 0 obj << +/D [3858 0 R /XYZ 303.433 322.385 null] +>> endobj +3819 0 obj << +/D [3858 0 R /XYZ 141.07 310.729 null] +>> endobj +3820 0 obj << +/D [3858 0 R /XYZ 243.287 310.729 null] +>> endobj +3821 0 obj << +/D [3858 0 R /XYZ 302.787 310.729 null] +>> endobj +3822 0 obj << +/D [3858 0 R /XYZ 318.389 310.729 null] +>> endobj +3823 0 obj << +/D [3858 0 R /XYZ 350.667 310.729 null] +>> endobj +3824 0 obj << +/D [3858 0 R /XYZ 141.07 299.073 null] +>> endobj +3825 0 obj << +/D [3858 0 R /XYZ 259.426 299.073 null] +>> endobj +3826 0 obj << +/D [3858 0 R /XYZ 318.927 299.073 null] +>> endobj +3827 0 obj << +/D [3858 0 R /XYZ 334.528 299.073 null] +>> endobj +3828 0 obj << +/D [3858 0 R /XYZ 366.807 299.073 null] +>> endobj +3829 0 obj << +/D [3858 0 R /XYZ 141.07 287.416 null] +>> endobj +3830 0 obj << +/D [3858 0 R /XYZ 232.527 287.416 null] +>> endobj +3831 0 obj << +/D [3858 0 R /XYZ 292.028 287.416 null] +>> endobj +3832 0 obj << +/D [3858 0 R /XYZ 307.629 287.416 null] +>> endobj +3833 0 obj << +/D [3858 0 R /XYZ 339.908 287.416 null] +>> endobj +3834 0 obj << +/D [3858 0 R /XYZ 141.07 275.76 null] +>> endobj +3835 0 obj << +/D [3858 0 R /XYZ 248.666 275.76 null] +>> endobj +3836 0 obj << +/D [3858 0 R /XYZ 308.167 275.76 null] +>> endobj +3837 0 obj << +/D [3858 0 R /XYZ 323.768 275.76 null] +>> endobj +3838 0 obj << +/D [3858 0 R /XYZ 356.047 275.76 null] +>> endobj +3839 0 obj << +/D [3858 0 R /XYZ 141.07 264.104 null] +>> endobj +3840 0 obj << +/D [3858 0 R /XYZ 237.907 264.104 null] +>> endobj +3841 0 obj << +/D [3858 0 R /XYZ 141.07 252.448 null] +>> endobj +3842 0 obj << +/D [3858 0 R /XYZ 270.186 252.448 null] +>> endobj +3843 0 obj << +/D [3858 0 R /XYZ 141.07 240.791 null] +>> endobj +3844 0 obj << +/D [3858 0 R /XYZ 200.248 240.791 null] +>> endobj +3845 0 obj << +/D [3858 0 R /XYZ 141.07 229.135 null] +>> endobj +3846 0 obj << +/D [3858 0 R /XYZ 243.287 229.135 null] +>> endobj +3847 0 obj << +/D [3858 0 R /XYZ 141.07 217.479 null] +>> endobj +3848 0 obj << +/D [3858 0 R /XYZ 243.287 217.479 null] +>> endobj +3849 0 obj << +/D [3858 0 R /XYZ 141.07 205.823 null] +>> endobj +3850 0 obj << +/D [3858 0 R /XYZ 232.527 205.823 null] +>> endobj +3851 0 obj << +/D [3858 0 R /XYZ 141.07 194.167 null] +>> endobj +3852 0 obj << +/D [3858 0 R /XYZ 232.527 194.167 null] +>> endobj +965 0 obj << +/D [3858 0 R /XYZ 71.731 149.88 null] +>> endobj +3857 0 obj << +/Font << /F31 805 0 R /F40 669 0 R /F37 1210 0 R /F60 1503 0 R /F52 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3864 0 obj << +/Length 3865 0 R +/Filter /FlateDecode +>> +stream +xڍWˎ6fRƌHUM1 cAKF^z3S(=sIzRp "ƉW^Q_=ϰ7Beݫ`|N$v+yİٲ8׷c+˻; M|ŧ +F/Bl XmG FIBXm㔰|D;H8J搡#$k>+I +oߋXᖷ%6Hlߠ0̣qD)HQ1nn!aYIF"z WKce$w|Ж0$pe Z_'< ; +TqCc9%68^l]tRw]W*nXy.иM{olKuE[@+jC>֥tVn %/bDV,KsP&;Cc,$4YI@/=>+lxA7VoXHi^4 >r$MÅlEQ \C2jƆB/z4hfҀKifGM &% ŕki a#VB1Tt<`VDX#(|kF#jU#o|h[ nơu;(F>[D_kqA44SиA3yu-XС=8 na$s2xC6eհ]iL..\Re0ƇhOɝB4@M;"*[ld\ʱ/:>Sxա|5my?nWt`ڳ*}VyfZސ~DpP,smnlE:n頊[ed `䕗hB*>(K)] \LZt'8^^Ӓ{D#(HyӠڞV*T xJcgQ.u6Ok[~p``A6-|< ӊ۝%@0]҆SټlNp meVؗŶ0Z~UQ+XHdmј0NHrC?_@67gDb3?|W J +]WIZ}8*l /܆c漛(&[z uS]j\n\G؉tYz bk1=K vS>CgPx7Y8Z5=\;vW(Đ{qԧ>Q Wۥ.m6;H? 4>N( T + Wbh}esڭRJ >>Q,\@bG1endstream +endobj +3865 0 obj +1381 +endobj +3863 0 obj << +/Type /Page +/Contents 3864 0 R +/Resources 3862 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 3672 0 R +>> endobj +3861 0 obj << +/D [3863 0 R /XYZ 71.731 704.356 null] +>> endobj +453 0 obj << +/D [3863 0 R /XYZ 372.279 647.066 null] +>> endobj +966 0 obj << +/D [3863 0 R /XYZ 71.731 633.912 null] +>> endobj +457 0 obj << +/D [3863 0 R /XYZ 235.12 612.596 null] +>> endobj +3853 0 obj << +/D [3863 0 R /XYZ 71.731 600.158 null] +>> endobj +3854 0 obj << +/D [3863 0 R /XYZ 71.731 594.181 null] +>> endobj +3855 0 obj << +/D [3863 0 R /XYZ 71.731 594.181 null] +>> endobj +3862 0 obj << +/Font << /F31 805 0 R /F40 669 0 R /F42 669 0 R /F37 1210 0 R /F25 681 0 R /F30 731 0 R /F56 1242 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3869 0 obj << +/Length 3870 0 R +/Filter /FlateDecode +>> +stream +xڝXK6r̯r[e).!큖hD %J(Xəo^ BX>'{/ކw̎Ei'8}zq{wX},H#hFIOft45su_jn~2>(oo=YZvE.yS/'^.nw;nwA_$:xk?Y%CraG0彉BԣjN͈uU$sw\Vun0YaRGF #ϖ6%ˈ`j4{F1 _ns\y1Op"eB^3.ߒ],w`( +70Gp"V <'-lH+/+ uDkhd#b[j]KjJZP}=cb͟Dܕ-Xs 8R+ײ. >Fui!AL5KĔS-$WI g PܝNJsR!W5Iѓ[e'mC\N Ee@[N ̡CeY4@"Z񱵈UKd.5 jңlUPHëtExfCUC-m+28"63%#f;4t}k8:ΤcT +P❟!5t7:JV2 A;TO ֭`t-#h^ + S޶il+Ӗaũ]>7qLpS(62q׳idh7@3_H5Z5<*Y'kLl9ml`{Vg "wTol +6TW% LL]~ ٺѕRcM]!^]U9$z;xA; ҙD1p7_Lyhb6 `CjoZߍ0oXȸ} 5j;1Ց @#\(`t =gUyH :ڇS?2\| ܘ^Ta;yr mu5%/_vR^g̡axendstream +endobj +3870 0 obj +1650 +endobj +3868 0 obj << +/Type /Page +/Contents 3869 0 R +/Resources 3867 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 3672 0 R +>> endobj +3866 0 obj << +/D [3868 0 R /XYZ 71.731 704.356 null] +>> endobj +3867 0 obj << +/Font << /F31 805 0 R /F37 1210 0 R /F25 681 0 R /F30 731 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3914 0 obj << +/Length 3915 0 R +/Filter /FlateDecode +>> +stream +xXn6]bDRqMzy%JRRw(q@EQ9s̙b/?.C//b} ̙$ !ģ%^b+x?EiDͶ[)~[]C',XE\_\߼4\dUkF̾[x+^|nzp$۰d1a4a)m 0HPE( KhNT+#0%vfǍȭ*)LFھl /][ 3syY fe+m#28͡lhL5pa i\? +m}*&.;Rzy9Ѕ{LOy~O(ŒFüJe}V[IgE,\/*qk }haJH%ͼE&lkI7KN KY" B&"D N<Aخ%}>8  +1 !81x\CZ@iXfت"یh:h]A.^ [=ܾ;NcB[bYMe9.e`}10:g44*/dHQ@aS|1lmK^X7;qe.}c\38γϬnq7y逌4Zs|ڙyB]൐78r-1Q8{sJW7"wdiF?^o ps 0T fz8(Naq700b6pjNe%dJԱnxiC5ٝR[Αۢ{%ۃ9(#nWZ,>.,0p-R=vΜQL>V><VF ԙA5~FZPj8~RsRbb7)\"W1T8`~-W𿧼aQYU3y>;: <;<iJhvScV(i;^pb՘j{ԽD.o/fCϭ({4T TyvgkVȭ⮗ǹ n.OȢ/WIZ>V$dN1K7< endstream +endobj +3915 0 obj +1504 +endobj +3913 0 obj << +/Type /Page +/Contents 3914 0 R +/Resources 3912 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 3920 0 R +/Annots [ 3916 0 R 3917 0 R 3918 0 R 3919 0 R ] +>> endobj +3916 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [292.122 337.966 356.878 346.987] +/Subtype /Link +/A << /S /GoTo /D (access_control) >> +>> endobj +3917 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [292.122 297.12 356.878 306.141] +/Subtype /Link +/A << /S /GoTo /D (access_control) >> +>> endobj +3918 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [301.516 256.273 366.273 265.294] +/Subtype /Link +/A << /S /GoTo /D (access_control) >> +>> endobj +3919 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [330.537 161.628 395.294 170.54] +/Subtype /Link +/A << /S /GoTo /D (dynamic_update_policies) >> +>> endobj +3911 0 obj << +/D [3913 0 R /XYZ 71.731 705.352 null] +>> endobj +967 0 obj << +/D [3913 0 R /XYZ 71.731 559.5 null] +>> endobj +461 0 obj << +/D [3913 0 R /XYZ 201.299 533.996 null] +>> endobj +3871 0 obj << +/D [3913 0 R /XYZ 71.731 527.869 null] +>> endobj +3872 0 obj << +/D [3913 0 R /XYZ 469.872 515.067 null] +>> endobj +3873 0 obj << +/D [3913 0 R /XYZ 119.551 502.115 null] +>> endobj +3874 0 obj << +/D [3913 0 R /XYZ 71.731 494.977 null] +>> endobj +3875 0 obj << +/D [3913 0 R /XYZ 137.534 484.183 null] +>> endobj +3876 0 obj << +/D [3913 0 R /XYZ 522.654 471.231 null] +>> endobj +3877 0 obj << +/D [3913 0 R /XYZ 71.731 451.142 null] +>> endobj +3878 0 obj << +/D [3913 0 R /XYZ 204.222 427.396 null] +>> endobj +968 0 obj << +/D [3913 0 R /XYZ 71.731 410.295 null] +>> endobj +465 0 obj << +/D [3913 0 R /XYZ 246.453 386.948 null] +>> endobj +3879 0 obj << +/D [3913 0 R /XYZ 71.731 378.495 null] +>> endobj +3880 0 obj << +/D [3913 0 R /XYZ 71.731 368.533 null] +>> endobj +3881 0 obj << +/D [3913 0 R /XYZ 71.731 368.533 null] +>> endobj +3882 0 obj << +/D [3913 0 R /XYZ 71.731 356.034 null] +>> endobj +3883 0 obj << +/D [3913 0 R /XYZ 139.476 340.123 null] +>> endobj +3884 0 obj << +/D [3913 0 R /XYZ 228.571 340.123 null] +>> endobj +3885 0 obj << +/D [3913 0 R /XYZ 71.731 328.004 null] +>> endobj +3886 0 obj << +/D [3913 0 R /XYZ 71.731 328.004 null] +>> endobj +3887 0 obj << +/D [3913 0 R /XYZ 71.731 315.187 null] +>> endobj +3888 0 obj << +/D [3913 0 R /XYZ 139.476 299.277 null] +>> endobj +3889 0 obj << +/D [3913 0 R /XYZ 228.571 299.277 null] +>> endobj +3890 0 obj << +/D [3913 0 R /XYZ 71.731 287.157 null] +>> endobj +3891 0 obj << +/D [3913 0 R /XYZ 71.731 287.157 null] +>> endobj +3892 0 obj << +/D [3913 0 R /XYZ 71.731 276.243 null] +>> endobj +3893 0 obj << +/D [3913 0 R /XYZ 139.476 258.43 null] +>> endobj +3894 0 obj << +/D [3913 0 R /XYZ 228.571 258.43 null] +>> endobj +3895 0 obj << +/D [3913 0 R /XYZ 71.731 246.311 null] +>> endobj +3896 0 obj << +/D [3913 0 R /XYZ 71.731 246.311 null] +>> endobj +3897 0 obj << +/D [3913 0 R /XYZ 71.731 233.494 null] +>> endobj +3898 0 obj << +/D [3913 0 R /XYZ 139.476 217.583 null] +>> endobj +3899 0 obj << +/D [3913 0 R /XYZ 71.731 192.512 null] +>> endobj +3900 0 obj << +/D [3913 0 R /XYZ 71.731 192.512 null] +>> endobj +3901 0 obj << +/D [3913 0 R /XYZ 71.731 179.696 null] +>> endobj +3902 0 obj << +/D [3913 0 R /XYZ 139.476 163.785 null] +>> endobj +3912 0 obj << +/Font << /F31 805 0 R /F37 1210 0 R /F25 681 0 R /F42 669 0 R /F30 731 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +3958 0 obj << +/Length 3959 0 R +/Filter /FlateDecode +>> +stream +xYˎ6f0 DU4EZ`RdfS]=&F]J-(gE²,^ǹrƇ?I1JCc_ߘz:}&Dqʇ޾x(OaxeHN;/mۏU>;lIx~X[t/.4A~m<%g( +7^4 Gˤ?a[.(Ys:o<<|:i(=I){C,AD vE[Ї#+N% xlhnt\}f~nH +F]Lβ[x3<i,!Cfe(ԤȊ㭆խR_iY+ B1EOcw^{.QKYC^jӚƤH>͐ +}ê&DF$AY$'*`qMK. !:omVp^IK"^`/s3ҜM,s&Tha(fQل!TA f1\F_T6V.Vl@$> +1 vo{HY})xnO}1W +-lg>+C9mOiٞU;R/O +tkGWZg;M-B+vonjd-,2ueE5T(zbfzGIhOʻx-b3r2ŨFX'WY +[=i 2 :RܷoL$J%adJr9qQ0)h1N(B89SdQZsa#R8Oe/ 0}Ef{9M`S7qȿƙ1T J24C%1 +8cr$vݐ`:8.8-uDǓn2p5\b~3vP!'p㇟ц@ =C.Gg5[MJK+ OŀtSnǾ;&&u'j":)q(3xu#8U)m#878b*WNX} 8Y[͓>?c|0aR#֝rS؊0yMI˲SAJɮ6qOziO.+dO ]} +X4](pB(adDlP@A{H{oQcX y9 8Z2QZq6*a6kf{[n6TXH0(ؔ&ƶJ`ܚJZ~(_+9 27ݵi iV9;^;⮡ݐ.ї|FFNRslx}-Oi3Q bqjA#vרp H,ND,o +La/-]w>QX&--bb)Ǚ' { #+9YY$=Ff:L+13Nv/>l2(Ub~A~%eCÑu}^I+V3ԼAtfDYưD և<*iY +:P'>-W]cHO:L#i =Ol{˞s8%GL~68q9Ry(VHˎ僗b߽\92h˜3,C[d)YJZ Zz +^'m3!qY р_s-_F=ٌ8mM"I%ժb #F9nG]t KDN XST}irR2Y{z?* e> endobj +3960 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [404.249 544.79 449.081 553.701] +/Subtype /Link +/A << /S /GoTo /D (dynamic_update_security) >> +>> endobj +3961 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [268.441 167.207 333.197 176.228] +/Subtype /Link +/A << /S /GoTo /D (boolean_options) >> +>> endobj +3955 0 obj << +/D [3957 0 R /XYZ 71.731 705.352 null] +>> endobj +3903 0 obj << +/D [3957 0 R /XYZ 71.731 657.532 null] +>> endobj +3904 0 obj << +/D [3957 0 R /XYZ 71.731 657.532 null] +>> endobj +3905 0 obj << +/D [3957 0 R /XYZ 71.731 645.547 null] +>> endobj +3906 0 obj << +/D [3957 0 R /XYZ 139.476 629.637 null] +>> endobj +3907 0 obj << +/D [3957 0 R /XYZ 252.551 616.685 null] +>> endobj +3908 0 obj << +/D [3957 0 R /XYZ 336.665 603.734 null] +>> endobj +3909 0 obj << +/D [3957 0 R /XYZ 255.44 590.782 null] +>> endobj +3910 0 obj << +/D [3957 0 R /XYZ 322.518 590.782 null] +>> endobj +3921 0 obj << +/D [3957 0 R /XYZ 71.731 570.693 null] +>> endobj +3922 0 obj << +/D [3957 0 R /XYZ 71.731 519.052 null] +>> endobj +3923 0 obj << +/D [3957 0 R /XYZ 71.731 519.052 null] +>> endobj +3924 0 obj << +/D [3957 0 R /XYZ 71.731 504.078 null] +>> endobj +3925 0 obj << +/D [3957 0 R /XYZ 139.476 488.168 null] +>> endobj +3926 0 obj << +/D [3957 0 R /XYZ 218.34 488.168 null] +>> endobj +3927 0 obj << +/D [3957 0 R /XYZ 491.182 488.168 null] +>> endobj +3928 0 obj << +/D [3957 0 R /XYZ 367.867 462.265 null] +>> endobj +3929 0 obj << +/D [3957 0 R /XYZ 160.218 449.313 null] +>> endobj +3930 0 obj << +/D [3957 0 R /XYZ 139.476 436.362 null] +>> endobj +3931 0 obj << +/D [3957 0 R /XYZ 71.731 424.243 null] +>> endobj +3932 0 obj << +/D [3957 0 R /XYZ 71.731 424.243 null] +>> endobj +3933 0 obj << +/D [3957 0 R /XYZ 71.731 413.329 null] +>> endobj +3934 0 obj << +/D [3957 0 R /XYZ 139.476 395.515 null] +>> endobj +3935 0 obj << +/D [3957 0 R /XYZ 280.047 369.613 null] +>> endobj +3936 0 obj << +/D [3957 0 R /XYZ 71.731 357.493 null] +>> endobj +3937 0 obj << +/D [3957 0 R /XYZ 71.731 357.493 null] +>> endobj +3938 0 obj << +/D [3957 0 R /XYZ 71.731 346.579 null] +>> endobj +3939 0 obj << +/D [3957 0 R /XYZ 139.476 328.766 null] +>> endobj +3940 0 obj << +/D [3957 0 R /XYZ 497.818 328.766 null] +>> endobj +3941 0 obj << +/D [3957 0 R /XYZ 71.731 282.774 null] +>> endobj +3942 0 obj << +/D [3957 0 R /XYZ 196.651 271.979 null] +>> endobj +3943 0 obj << +/D [3957 0 R /XYZ 71.731 251.89 null] +>> endobj +3944 0 obj << +/D [3957 0 R /XYZ 71.731 200.248 null] +>> endobj +3945 0 obj << +/D [3957 0 R /XYZ 71.731 200.248 null] +>> endobj +3946 0 obj << +/D [3957 0 R /XYZ 71.731 185.275 null] +>> endobj +3947 0 obj << +/D [3957 0 R /XYZ 139.476 169.364 null] +>> endobj +3948 0 obj << +/D [3957 0 R /XYZ 228.571 169.364 null] +>> endobj +3956 0 obj << +/Font << /F31 805 0 R /F30 731 0 R /F25 681 0 R /F39 1242 0 R /F37 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4021 0 obj << +/Length 4022 0 R +/Filter /FlateDecode +>> +stream +xXmo61B'8QDִx؀mȢޯQ$n7C #ysG(F@o, Ɣ亮ER)f}{ 4r.:TXٮ,7wުSͭۺX`U=LB<7֭ +qCnuyIJ}hH:ʛ/Eqdc٘"F>{1zޮ?<;)<|3;ݨq[.*P Hx?eiYiϭw#-fL!L_IA4p_(6qT TPnRmUĂ,%!ln.b@_IUѪsEy:[^Uj + <6|4!þpQPZCԙ<2!a9ܑbe#D@Jk'h"ah7&aqdDW_JXahxQ  u(l3+i5 fLĴ4(mX3!|O<`-OicVgTƆfRdfKs ڄNG@$АGӦcT, B@TXՊ^#6ٜ77ў1-S@b}0D&t񝏢.(.V_oyFZCUZB> g@ߟ;-IVTihRu\012Df9}Eϓ6EM蛣]{nzz}S,e\ L20s pWLc pbk;u" xyxϮ>{>\fEdh'7Z +yZ?<'Yᅞ- (_Pޅ|ܻz^}F兙x /b1\A$1/0(b@Pȋ4;hѯwj,])" |7O8}'YF }y =WWybK._G@eM9(v]4ZF?֤8SZPfFU- iyjvrw]w+/AUӳ +|`z;wy4.Bp|caǙN"AԽendstream +endobj +4022 0 obj +1317 +endobj +4020 0 obj << +/Type /Page +/Contents 4021 0 R +/Resources 4019 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 3920 0 R +/Annots [ 4023 0 R 4024 0 R 4025 0 R 4026 0 R 4027 0 R ] +>> endobj +4023 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [330.567 412.287 395.324 421.309] +/Subtype /Link +/A << /S /GoTo /D (zone_transfers) >> +>> endobj +4024 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [327.259 371.441 392.016 380.462] +/Subtype /Link +/A << /S /GoTo /D (zone_transfers) >> +>> endobj +4025 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [336.096 330.594 400.853 339.615] +/Subtype /Link +/A << /S /GoTo /D (zone_transfers) >> +>> endobj +4026 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [332.789 289.747 397.545 298.768] +/Subtype /Link +/A << /S /GoTo /D (zone_transfers) >> +>> endobj +4027 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [266.209 248.901 330.966 257.922] +/Subtype /Link +/A << /S /GoTo /D (boolean_options) >> +>> endobj +4017 0 obj << +/D [4020 0 R /XYZ 71.731 705.352 null] +>> endobj +4018 0 obj << +/D [4020 0 R /XYZ 71.731 717.307 null] +>> endobj +3949 0 obj << +/D [4020 0 R /XYZ 71.731 657.532 null] +>> endobj +3950 0 obj << +/D [4020 0 R /XYZ 71.731 657.532 null] +>> endobj +3951 0 obj << +/D [4020 0 R /XYZ 71.731 647.45 null] +>> endobj +3952 0 obj << +/D [4020 0 R /XYZ 139.476 629.637 null] +>> endobj +3953 0 obj << +/D [4020 0 R /XYZ 357.675 629.637 null] +>> endobj +3954 0 obj << +/D [4020 0 R /XYZ 342.891 616.685 null] +>> endobj +3962 0 obj << +/D [4020 0 R /XYZ 71.731 604.566 null] +>> endobj +3963 0 obj << +/D [4020 0 R /XYZ 71.731 604.566 null] +>> endobj +3964 0 obj << +/D [4020 0 R /XYZ 71.731 593.652 null] +>> endobj +3965 0 obj << +/D [4020 0 R /XYZ 139.476 575.838 null] +>> endobj +3966 0 obj << +/D [4020 0 R /XYZ 471.935 575.838 null] +>> endobj +3967 0 obj << +/D [4020 0 R /XYZ 71.731 550.768 null] +>> endobj +3968 0 obj << +/D [4020 0 R /XYZ 71.731 550.768 null] +>> endobj +3969 0 obj << +/D [4020 0 R /XYZ 71.731 539.854 null] +>> endobj +3970 0 obj << +/D [4020 0 R /XYZ 139.476 522.04 null] +>> endobj +3971 0 obj << +/D [4020 0 R /XYZ 189.828 522.04 null] +>> endobj +3972 0 obj << +/D [4020 0 R /XYZ 167.152 509.089 null] +>> endobj +3973 0 obj << +/D [4020 0 R /XYZ 506.306 509.089 null] +>> endobj +3974 0 obj << +/D [4020 0 R /XYZ 71.731 486.076 null] +>> endobj +3975 0 obj << +/D [4020 0 R /XYZ 71.731 486.076 null] +>> endobj +3976 0 obj << +/D [4020 0 R /XYZ 71.731 471.201 null] +>> endobj +3977 0 obj << +/D [4020 0 R /XYZ 139.476 455.291 null] +>> endobj +3978 0 obj << +/D [4020 0 R /XYZ 270.115 455.291 null] +>> endobj +3979 0 obj << +/D [4020 0 R /XYZ 350.632 455.291 null] +>> endobj +3980 0 obj << +/D [4020 0 R /XYZ 71.731 443.172 null] +>> endobj +3981 0 obj << +/D [4020 0 R /XYZ 71.731 443.172 null] +>> endobj +3982 0 obj << +/D [4020 0 R /XYZ 71.731 432.258 null] +>> endobj +3983 0 obj << +/D [4020 0 R /XYZ 139.476 414.444 null] +>> endobj +3984 0 obj << +/D [4020 0 R /XYZ 228.571 414.444 null] +>> endobj +3985 0 obj << +/D [4020 0 R /XYZ 71.731 402.325 null] +>> endobj +3986 0 obj << +/D [4020 0 R /XYZ 71.731 402.325 null] +>> endobj +3987 0 obj << +/D [4020 0 R /XYZ 71.731 391.411 null] +>> endobj +3988 0 obj << +/D [4020 0 R /XYZ 139.476 373.598 null] +>> endobj +3989 0 obj << +/D [4020 0 R /XYZ 228.571 373.598 null] +>> endobj +3990 0 obj << +/D [4020 0 R /XYZ 71.731 361.478 null] +>> endobj +3991 0 obj << +/D [4020 0 R /XYZ 71.731 361.478 null] +>> endobj +3992 0 obj << +/D [4020 0 R /XYZ 71.731 350.564 null] +>> endobj +3993 0 obj << +/D [4020 0 R /XYZ 139.476 332.751 null] +>> endobj +3994 0 obj << +/D [4020 0 R /XYZ 228.571 332.751 null] +>> endobj +3995 0 obj << +/D [4020 0 R /XYZ 71.731 320.632 null] +>> endobj +3996 0 obj << +/D [4020 0 R /XYZ 71.731 320.632 null] +>> endobj +3997 0 obj << +/D [4020 0 R /XYZ 71.731 309.717 null] +>> endobj +3998 0 obj << +/D [4020 0 R /XYZ 139.476 291.904 null] +>> endobj +3999 0 obj << +/D [4020 0 R /XYZ 228.571 291.904 null] +>> endobj +4000 0 obj << +/D [4020 0 R /XYZ 71.731 279.785 null] +>> endobj +4001 0 obj << +/D [4020 0 R /XYZ 71.731 279.785 null] +>> endobj +4002 0 obj << +/D [4020 0 R /XYZ 71.731 266.968 null] +>> endobj +4003 0 obj << +/D [4020 0 R /XYZ 139.476 251.058 null] +>> endobj +4004 0 obj << +/D [4020 0 R /XYZ 228.571 251.058 null] +>> endobj +4005 0 obj << +/D [4020 0 R /XYZ 71.731 238.938 null] +>> endobj +4006 0 obj << +/D [4020 0 R /XYZ 71.731 238.938 null] +>> endobj +4007 0 obj << +/D [4020 0 R /XYZ 71.731 226.121 null] +>> endobj +4008 0 obj << +/D [4020 0 R /XYZ 139.476 210.211 null] +>> endobj +4009 0 obj << +/D [4020 0 R /XYZ 150.266 210.211 null] +>> endobj +4010 0 obj << +/D [4020 0 R /XYZ 374.801 197.26 null] +>> endobj +4019 0 obj << +/Font << /F31 805 0 R /F30 731 0 R /F25 681 0 R /F37 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4079 0 obj << +/Length 4080 0 R +/Filter /FlateDecode +>> +stream +xYK6_ DHtl6(Ȕ+:>Cd6A]$rH,^1Fq/ϛ,5ICDYD%!~~ݏ>E)mCX>gZTpɑ}>}Kynvj2e(^ +?=aݲcB,k 'F(NpR@FQ=$DS~CǼ 56vkn.Ep77D}E +, +fåGGi.ږաgN{Kmߌ݁{MVvbIL-r5* +HehӀE(N/n(zScgr;UO0enn=i`=l¾ܽ&ͦ#0 'J i:HRbkzи@.=dM1/13O7%Y6<熈cl D>sJ/+,lW#{WY$8";ic2D؁bgJA!l?2mh͸[ێ4ZMdnk;Ԁʑz*QvCcQQ KR֫=Nލ 쭰)+׭(+\8}Ni}hH%Q/ HȽ,PVn%нl(XuUQi0l(zp`Vy@$|opCS#Ku*|7"jϹ Ϲ|,sclS”]J'tHq2/pYk1Zf3 +N} 8e 9nǷ4bz}[l'\OPZnm(U{cP"0'|{N]p]]1ژOH/jpy { M]y!AGHTa@d +.]n[f_BY_> endobj +4081 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [324.579 573.682 394.317 582.703] +/Subtype /Link +/A << /S /GoTo /D (tuning) >> +>> endobj +4082 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [306.039 532.835 370.796 541.856] +/Subtype /Link +/A << /S /GoTo /D (zone_transfers) >> +>> endobj +4083 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [319.319 491.988 384.076 501.009] +/Subtype /Link +/A << /S /GoTo /D (zone_transfers) >> +>> endobj +4084 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [297.013 451.142 361.77 460.163] +/Subtype /Link +/A << /S /GoTo /D (zone_transfers) >> +>> endobj +4085 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [310.294 410.295 375.05 419.316] +/Subtype /Link +/A << /S /GoTo /D (zone_transfers) >> +>> endobj +4086 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [230.514 330.594 300.252 339.506] +/Subtype /Link +/A << /S /GoTo /D (tuning) >> +>> endobj +4076 0 obj << +/D [4078 0 R /XYZ 71.731 705.352 null] +>> endobj +4011 0 obj << +/D [4078 0 R /XYZ 71.731 657.532 null] +>> endobj +4012 0 obj << +/D [4078 0 R /XYZ 71.731 657.532 null] +>> endobj +4013 0 obj << +/D [4078 0 R /XYZ 71.731 647.45 null] +>> endobj +4014 0 obj << +/D [4078 0 R /XYZ 139.476 629.637 null] +>> endobj +4015 0 obj << +/D [4078 0 R /XYZ 148.602 629.637 null] +>> endobj +4016 0 obj << +/D [4078 0 R /XYZ 139.476 616.685 null] +>> endobj +4028 0 obj << +/D [4078 0 R /XYZ 71.731 591.749 null] +>> endobj +4029 0 obj << +/D [4078 0 R /XYZ 139.476 575.838 null] +>> endobj +4030 0 obj << +/D [4078 0 R /XYZ 228.571 575.838 null] +>> endobj +4031 0 obj << +/D [4078 0 R /XYZ 71.731 563.719 null] +>> endobj +4032 0 obj << +/D [4078 0 R /XYZ 71.731 563.719 null] +>> endobj +4033 0 obj << +/D [4078 0 R /XYZ 71.731 552.805 null] +>> endobj +4034 0 obj << +/D [4078 0 R /XYZ 139.476 534.992 null] +>> endobj +4035 0 obj << +/D [4078 0 R /XYZ 228.571 534.992 null] +>> endobj +4036 0 obj << +/D [4078 0 R /XYZ 71.731 522.872 null] +>> endobj +4037 0 obj << +/D [4078 0 R /XYZ 71.731 522.872 null] +>> endobj +4038 0 obj << +/D [4078 0 R /XYZ 71.731 511.958 null] +>> endobj +4039 0 obj << +/D [4078 0 R /XYZ 139.476 494.145 null] +>> endobj +4040 0 obj << +/D [4078 0 R /XYZ 228.571 494.145 null] +>> endobj +4041 0 obj << +/D [4078 0 R /XYZ 71.731 482.026 null] +>> endobj +4042 0 obj << +/D [4078 0 R /XYZ 71.731 482.026 null] +>> endobj +4043 0 obj << +/D [4078 0 R /XYZ 71.731 469.209 null] +>> endobj +4044 0 obj << +/D [4078 0 R /XYZ 139.476 453.298 null] +>> endobj +4045 0 obj << +/D [4078 0 R /XYZ 228.571 453.298 null] +>> endobj +4046 0 obj << +/D [4078 0 R /XYZ 71.731 441.179 null] +>> endobj +4047 0 obj << +/D [4078 0 R /XYZ 71.731 441.179 null] +>> endobj +4048 0 obj << +/D [4078 0 R /XYZ 71.731 428.362 null] +>> endobj +4049 0 obj << +/D [4078 0 R /XYZ 139.476 412.452 null] +>> endobj +4050 0 obj << +/D [4078 0 R /XYZ 228.571 412.452 null] +>> endobj +4051 0 obj << +/D [4078 0 R /XYZ 71.731 400.332 null] +>> endobj +4052 0 obj << +/D [4078 0 R /XYZ 71.731 400.332 null] +>> endobj +4053 0 obj << +/D [4078 0 R /XYZ 71.731 389.418 null] +>> endobj +4054 0 obj << +/D [4078 0 R /XYZ 71.731 389.418 null] +>> endobj +4055 0 obj << +/D [4078 0 R /XYZ 71.731 376.467 null] +>> endobj +4056 0 obj << +/D [4078 0 R /XYZ 71.731 376.467 null] +>> endobj +4057 0 obj << +/D [4078 0 R /XYZ 71.731 361.613 null] +>> endobj +4058 0 obj << +/D [4078 0 R /XYZ 71.731 361.613 null] +>> endobj +4059 0 obj << +/D [4078 0 R /XYZ 71.731 348.661 null] +>> endobj +4060 0 obj << +/D [4078 0 R /XYZ 139.476 332.751 null] +>> endobj +969 0 obj << +/D [4078 0 R /XYZ 71.731 315.65 null] +>> endobj +469 0 obj << +/D [4078 0 R /XYZ 311.106 292.303 null] +>> endobj +4061 0 obj << +/D [4078 0 R /XYZ 71.731 283.851 null] +>> endobj +4062 0 obj << +/D [4078 0 R /XYZ 71.731 283.851 null] +>> endobj +4063 0 obj << +/D [4078 0 R /XYZ 215.56 260.422 null] +>> endobj +4064 0 obj << +/D [4078 0 R /XYZ 290.18 260.422 null] +>> endobj +4065 0 obj << +/D [4078 0 R /XYZ 71.731 253.284 null] +>> endobj +4066 0 obj << +/D [4078 0 R /XYZ 137.439 242.49 null] +>> endobj +4067 0 obj << +/D [4078 0 R /XYZ 408.802 242.49 null] +>> endobj +4068 0 obj << +/D [4078 0 R /XYZ 71.731 222.4 null] +>> endobj +4069 0 obj << +/D [4078 0 R /XYZ 137.534 211.606 null] +>> endobj +4070 0 obj << +/D [4078 0 R /XYZ 263.748 211.606 null] +>> endobj +4071 0 obj << +/D [4078 0 R /XYZ 71.731 165.613 null] +>> endobj +4072 0 obj << +/D [4078 0 R /XYZ 222.205 154.819 null] +>> endobj +4073 0 obj << +/D [4078 0 R /XYZ 134.216 141.867 null] +>> endobj +4074 0 obj << +/D [4078 0 R /XYZ 408.712 141.867 null] +>> endobj +4077 0 obj << +/Font << /F31 805 0 R /F30 731 0 R /F25 681 0 R /F39 1242 0 R /F42 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4111 0 obj << +/Length 4112 0 R +/Filter /FlateDecode +>> +stream +xXY61Bȓ YE}YA7ih^aHr w(l'HS hj8K { ##('=z {"x! S7O{ SơtbhVkB?FKJE]]}ɺһb'C*80Q@od$: BoMbID&t):#f_ )U֊ Nu!GM;f(ж]3Q +S/^Jz$JVWւN?~ma|m}l2y^X6swlflnE +"`;#~]"֬Es{{Zfm4dǚ+P|O牳JLnh'@3/dgQv{SP+"['(N(m` ;-S[%u(41&:y]r0_ 'E(My̰ 8G/F (v*,iB!Ji5}[P˜% dQ_ b`yu&u ?"tu6%1ҺW|#ZYC IH i\gm e + g-޾d X6&}i.Lߙ|WxK[dSt +ժrؿ@\`RUϱW`a.~_6y9=gjJ>5Ǻ﫺_DТSI H\" ef7t:ml;q=pjҿXҨLPwo~4n*ysfؖt&[1L E8xzduyCڂ__]/sd MpC +39۾nv)'L%JM5 {5c*^V-8DD0,A-LQȈ;*v}ʁPC~&NV!՞ů:BaKkAؔ' t>mEc‚7+cGm4zrk|xB/pFvhFn-vVWfS=3`~jd" 7YP;3'-d6mrfҿ5%xϕ/5šR7*S4blQr^%^7 #a Q \5xpk{ )5B Q(bBNVQy?l=3ralF;]ݔbv\`&PX!6_l~ CL:ρI:F-endstream +endobj +4112 0 obj +1574 +endobj +4110 0 obj << +/Type /Page +/Contents 4111 0 R +/Resources 4109 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 3920 0 R +>> endobj +4108 0 obj << +/D [4110 0 R /XYZ 71.731 705.352 null] +>> endobj +4075 0 obj << +/D [4110 0 R /XYZ 171.287 647.569 null] +>> endobj +4087 0 obj << +/D [4110 0 R /XYZ 71.731 629.537 null] +>> endobj +4088 0 obj << +/D [4110 0 R /XYZ 71.731 606.623 null] +>> endobj +4089 0 obj << +/D [4110 0 R /XYZ 130.311 595.066 null] +>> endobj +4090 0 obj << +/D [4110 0 R /XYZ 173.349 595.066 null] +>> endobj +4091 0 obj << +/D [4110 0 R /XYZ 211.008 595.066 null] +>> endobj +4092 0 obj << +/D [4110 0 R /XYZ 260.17 595.066 null] +>> endobj +4093 0 obj << +/D [4110 0 R /XYZ 308.588 595.066 null] +>> endobj +4094 0 obj << +/D [4110 0 R /XYZ 335.487 595.066 null] +>> endobj +4095 0 obj << +/D [4110 0 R /XYZ 346.247 595.066 null] +>> endobj +4096 0 obj << +/D [4110 0 R /XYZ 71.731 582.862 null] +>> endobj +4097 0 obj << +/D [4110 0 R /XYZ 71.731 539.111 null] +>> endobj +4098 0 obj << +/D [4110 0 R /XYZ 473.19 528.317 null] +>> endobj +4099 0 obj << +/D [4110 0 R /XYZ 119.551 515.365 null] +>> endobj +4100 0 obj << +/D [4110 0 R /XYZ 172.951 515.365 null] +>> endobj +4101 0 obj << +/D [4110 0 R /XYZ 237.847 515.365 null] +>> endobj +4102 0 obj << +/D [4110 0 R /XYZ 71.731 503.963 null] +>> endobj +4103 0 obj << +/D [4110 0 R /XYZ 71.731 498.982 null] +>> endobj +4104 0 obj << +/D [4110 0 R /XYZ 71.731 498.982 null] +>> endobj +4105 0 obj << +/D [4110 0 R /XYZ 71.731 372.656 null] +>> endobj +970 0 obj << +/D [4110 0 R /XYZ 71.731 309.728 null] +>> endobj +473 0 obj << +/D [4110 0 R /XYZ 180.806 276.593 null] +>> endobj +1057 0 obj << +/D [4110 0 R /XYZ 71.731 266.415 null] +>> endobj +477 0 obj << +/D [4110 0 R /XYZ 462.53 237.221 null] +>> endobj +4106 0 obj << +/D [4110 0 R /XYZ 71.731 227.078 null] +>> endobj +1058 0 obj << +/D [4110 0 R /XYZ 71.731 179.074 null] +>> endobj +481 0 obj << +/D [4110 0 R /XYZ 267.912 156.723 null] +>> endobj +4107 0 obj << +/D [4110 0 R /XYZ 71.731 150.596 null] +>> endobj +4109 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F30 731 0 R /F37 1210 0 R /F56 1242 0 R /F60 1503 0 R /F36 669 0 R /F40 669 0 R /F42 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4125 0 obj << +/Length 4126 0 R +/Filter /FlateDecode +>> +stream +xYYoF_G \ˣoMj +h"W"W );C&c?xcBhm_Ϸ7-ՏsQ GSpQ;A>2: 9`rr/+ַE,4d/j,[jC>B$`-Ms%e,P(\:Dx(P$Ab !F hPQ;YH'e&r:iEܐABAʲ\E3Vub0KJ9.f7̕ UtY7͚k(`vHbDblfCWdV OYbIPpq/b-lg>g&ZQ~mվC`2]*/0ڤ)D6l/O#h9^a(`EYU|\hV`M¦$uX> u]K[93 +UCb]9[F5]Ywײ +>6D?J{Coyos}%DDJՖׅҁSiKnhTb7= yx<>݅0q^;2Iew +yv B +Cm: 9/۠p4>?eSspp:jb(4uE+kJ~(NªD%4iy>('l@6C\6U0-`zm%+ְ׭EPsˆzT~iHԳ)FT^UY7ڊses#j.9s=^@qlAOxzCI>/il؟s ݖ6gdx B{q<@!NiZfP&_v(B9EP_~y? |]g3{)qofp]2G&MK&~墣Mu*rV]o-Cf= +rN@g0:H+;Bmh+{[B|`3b]j 7GwW#26IGVݧ[`q~+\bil'j.Me/w;i||=RIu_\݂>ғoF ɿM:勒=tt+B!.UcD@-.m$ a?_ QhH"'qrt[Rox}>Eendstream +endobj +4126 0 obj +1815 +endobj +4124 0 obj << +/Type /Page +/Contents 4125 0 R +/Resources 4123 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 3920 0 R +/Annots [ 4127 0 R 4128 0 R ] +>> endobj +4127 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [310.851 606.558 380.589 615.47] +/Subtype /Link +/A << /S /GoTo /D (the_sortlist_statement) >> +>> endobj +4128 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [399.956 606.558 469.694 615.47] +/Subtype /Link +/A << /S /GoTo /D (rrset_ordering) >> +>> endobj +4122 0 obj << +/D [4124 0 R /XYZ 71.731 704.356 null] +>> endobj +4113 0 obj << +/D [4124 0 R /XYZ 71.731 601.577 null] +>> endobj +4114 0 obj << +/D [4124 0 R /XYZ 71.731 578.663 null] +>> endobj +4115 0 obj << +/D [4124 0 R /XYZ 71.731 573.682 null] +>> endobj +4116 0 obj << +/D [4124 0 R /XYZ 71.731 573.682 null] +>> endobj +4117 0 obj << +/D [4124 0 R /XYZ 71.731 392.562 null] +>> endobj +4118 0 obj << +/D [4124 0 R /XYZ 193.154 379.61 null] +>> endobj +4119 0 obj << +/D [4124 0 R /XYZ 71.731 354.539 null] +>> endobj +4120 0 obj << +/D [4124 0 R /XYZ 71.731 349.558 null] +>> endobj +4121 0 obj << +/D [4124 0 R /XYZ 71.731 349.558 null] +>> endobj +4123 0 obj << +/Font << /F31 805 0 R /F25 681 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4142 0 obj << +/Length 4143 0 R +/Filter /FlateDecode +>> +stream +xڵX]oH_Re>֦6lk7>Ǩ_wl0`Rx {gl.\KlWC}kr5 D1\ⷫ|CzZV_&*^Y6a!y|{uNpY/=3j"W| w[~{~Ava Mqk>A> a``BjTXw c(`naAҲ8s, iU/opBC3ׅ`ɹ +å =N`C&A!-7/C:pDZpxӹ>,!ז77e"׹̓P?ȧ Ue=ײ 皃O"=!Y餈0 ;<tx7oZ͡4Za$?AQ)֣ 1Ls]F1ήg?7˪{S1SZFu"y>!jOnuu-IVj^/Ž17Gxq#WfzvӶf]_—.'(Xϛӓ< 'υ9JP!UmlxQ۷ySM=T=DqzJ@%(I+7Z] + t{?n2KoWuzG W;"z+h7r<Y]9ͱ;z֎[~ip]̍UѪYBoҟ!=a-1CN35ܿAfmFV,W&:OKQs VX vf#A5.>?;R7ΥꖀdQR?rc|Mǎ}BjE,jx HW.n0R΍d,>1y- ,P7Y$b:OLAŶyWbm +HQtjhz#cuv{p#v Q\7{0 PK='LJzƃ:Z-,t녖{{д^ y83u ɰbX=xe;A+<:C? 0b}D|w!/[ + =❴0&[0f!r=t7bm'yu$~=+JnÔԇ2 ~{LR$^.Pnendstream +endobj +4143 0 obj +1587 +endobj +4141 0 obj << +/Type /Page +/Contents 4142 0 R +/Resources 4140 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 4145 0 R +/Annots [ 4144 0 R ] +>> endobj +4144 0 obj << +/Type /Annot +/Border [0 0 0] /H /I /C [1 0 0] +/Rect [346.577 177.369 401.092 193.309] +/Subtype /Link +/A << /S /GoTo /D (classes_of_resource_records) >> +>> endobj +4139 0 obj << +/D [4141 0 R /XYZ 71.731 704.356 null] +>> endobj +4129 0 obj << +/D [4141 0 R /XYZ 71.731 239.7 null] +>> endobj +4130 0 obj << +/D [4141 0 R /XYZ 178.519 226.749 null] +>> endobj +4131 0 obj << +/D [4141 0 R /XYZ 71.731 214.629 null] +>> endobj +4132 0 obj << +/D [4141 0 R /XYZ 71.731 209.648 null] +>> endobj +4133 0 obj << +/D [4141 0 R /XYZ 71.731 209.648 null] +>> endobj +4134 0 obj << +/D [4141 0 R /XYZ 71.731 167.008 null] +>> endobj +4135 0 obj << +/D [4141 0 R /XYZ 71.731 167.008 null] +>> endobj +4136 0 obj << +/D [4141 0 R /XYZ 71.731 141.937 null] +>> endobj +4137 0 obj << +/D [4141 0 R /XYZ 71.731 136.956 null] +>> endobj +4138 0 obj << +/D [4141 0 R /XYZ 71.731 136.956 null] +>> endobj +4140 0 obj << +/Font << /F31 805 0 R /F25 681 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4155 0 obj << +/Length 4156 0 R +/Filter /FlateDecode +>> +stream +xY[6c~cUwا46g @ h,jE*'ίߡ!ue^曡,|Mm"%" +iw?3=:^Kw`ۦb/ߞI-iZL=}[^=fK"=hUNW|B6bm1^Abd^(TÓEl ^ ֱy~]Ӎ5a2u͛:I6ȍ{DGKE4cWR C?EAea +HlO!x(6ijmáQb5&k|BjaK/Uڈ\A9򨿽0y,W7|$LX(Ջ,C@z[ంŪ54MIIn×$vƍN4rͼ{|?MOʋ|h0<[0Z =qp d+o{7=.*} 9@X%{Z>Oj-L9X>}0ouf6A:7L]Yy?~\={X3D`8p̅%$/, \2?`mqS&_r቎ȆQMO?*MC4ؖiMJd¸ ~dui # +Y6hY&me92L>E[ GI٥.YΤip|*٥ɺ&Q:v;OoNMH Ϯ3190Toԫcf3Fͅl\ 3RnjNUn?*&LSq*[v·^X5* j2ހ8CR +,5=~ߗ3C2L!:&uh9t&d +ߴ9\53wۄWB69bc vgqmm 3B dq +dH]Xb{yh~ېORp#I~X{3ñ65& L&1-m wK9]b,D9l!㊰i1tL'נ5V0e rEcR%.3H)̡L9!-&*hsa0 㘏5deq6#8k+zL;΋ `tp -4we}Z? i쫅^eٖ9Ҹ3耐A%}A6 `eظތ=pVdаK{D]#^C]vDžO;|߽EDw8Ц^^dSldۥ|ŰG] 68r0V\UeSIIr4޺rTt~T +<;rU:%$K+zM+޽5'~?!s".xd^놱=.DwӋX\i=%}AU; +?0m'6u  ǂ5e*nz3 `MV*{p(jOgqKV=QCň5:r +]5꧸-qAwnCQ +*TZQWAYe6oJR}zVtb& SM3yfžv"0`& +U^}%=UA *M?uN](NR/6pm8Fo-?endstream +endobj +4156 0 obj +2060 +endobj +4154 0 obj << +/Type /Page +/Contents 4155 0 R +/Resources 4153 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 4145 0 R +>> endobj +4152 0 obj << +/D [4154 0 R /XYZ 71.731 705.352 null] +>> endobj +4146 0 obj << +/D [4154 0 R /XYZ 71.731 444.531 null] +>> endobj +4147 0 obj << +/D [4154 0 R /XYZ 71.731 385.588 null] +>> endobj +4148 0 obj << +/D [4154 0 R /XYZ 71.731 289.947 null] +>> endobj +1059 0 obj << +/D [4154 0 R /XYZ 71.731 249.1 null] +>> endobj +485 0 obj << +/D [4154 0 R /XYZ 311.823 225.753 null] +>> endobj +4149 0 obj << +/D [4154 0 R /XYZ 71.731 217.3 null] +>> endobj +4150 0 obj << +/D [4154 0 R /XYZ 71.731 160.831 null] +>> endobj +4153 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F42 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4176 0 obj << +/Length 4177 0 R +/Filter /FlateDecode +>> +stream +xXY61B2fE[A&Amh0D: ERֵ P0`oof>؉1}B7џ/sz5I=Dq|&Zn?Ei;]nQf$FH[}ףT,pRJϮv˪-aAq<85'(e8;$B~[E īF#L|ϽsOYQ^o6"y>23UYA0FF_wEFu$hcYx&@%&( q+]n*fΠZU}Mn=%ޚuvu3`den9}A<@_5-ߟO-8W呴81z'Uyɥl PvM]Vo<vUs'aB56QeRoQvdVTRU]`"ab,PLlDsfItQ[|Ϟ930R{{ÎUR +;ۯDӄfu|z(Ikc6 AleZ2!-o*L@8v|(3t<䧉0gCf( I,q]2B$@Q^WWi]?>=ߢn9%q0Xy-5wfq8 +JQrDZYA`-_m$0vCIC&ÀҟЩok 5>NH~\g/GT7O +jtNYxk+~ŒNR#y*:|xX>"<;+'ǰL uVrkK]Wټ=vuI-}Vdh;iCF#9)$gژVQR9~@HIGL;M39s>ښn<[')4mR$d҆yKEDq~>2O`Q|9>pTu0G]}Dpz{m@$j>>w1Ombz8P}ȡqJ ^Ew$/aB&&^"<dN/1SL֓jb[j PM8 a8kTHGzs`y4^H-%)ՇApeH ̽"; wźpRkx@bcphr8Pcr jsZI\pqd0Ledž׍~9ӷ [iIқ I^{kGXendstream +endobj +4177 0 obj +1636 +endobj +4175 0 obj << +/Type /Page +/Contents 4176 0 R +/Resources 4174 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 4145 0 R +>> endobj +4173 0 obj << +/D [4175 0 R /XYZ 71.731 705.352 null] +>> endobj +4151 0 obj << +/D [4175 0 R /XYZ 71.731 657.532 null] +>> endobj +4157 0 obj << +/D [4175 0 R /XYZ 71.731 601.577 null] +>> endobj +4158 0 obj << +/D [4175 0 R /XYZ 71.731 572.75 null] +>> endobj +4159 0 obj << +/D [4175 0 R /XYZ 71.731 547.779 null] +>> endobj +4160 0 obj << +/D [4175 0 R /XYZ 71.731 542.798 null] +>> endobj +4161 0 obj << +/D [4175 0 R /XYZ 71.731 542.798 null] +>> endobj +4162 0 obj << +/D [4175 0 R /XYZ 71.731 426.434 null] +>> endobj +4163 0 obj << +/D [4175 0 R /XYZ 71.731 395.451 null] +>> endobj +4164 0 obj << +/D [4175 0 R /XYZ 71.731 375.461 null] +>> endobj +4165 0 obj << +/D [4175 0 R /XYZ 71.731 352.547 null] +>> endobj +4166 0 obj << +/D [4175 0 R /XYZ 71.731 347.566 null] +>> endobj +4167 0 obj << +/D [4175 0 R /XYZ 71.731 347.566 null] +>> endobj +4168 0 obj << +/D [4175 0 R /XYZ 71.731 294.963 null] +>> endobj +4169 0 obj << +/D [4175 0 R /XYZ 275.106 282.012 null] +>> endobj +1060 0 obj << +/D [4175 0 R /XYZ 71.731 254.948 null] +>> endobj +489 0 obj << +/D [4175 0 R /XYZ 292.286 227.695 null] +>> endobj +4170 0 obj << +/D [4175 0 R /XYZ 71.731 220.343 null] +>> endobj +4171 0 obj << +/D [4175 0 R /XYZ 71.731 150.685 null] +>> endobj +4174 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F37 1210 0 R /F40 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4194 0 obj << +/Length 4195 0 R +/Filter /FlateDecode +>> +stream +xڝXɎ68_lňHI +t z| Ai[Y4#E=3kWka/(pBD+WϞ +hQෛW?}pmݸջcznD Vۻ,чw9UwޤJXՇ#{p GQ9Ȅ!) 2= zs\ʛ}ɲdQU;HV;kgG}*`BD [l%*3k^c4nOycp+R$FQ6Ζ=m-n-w{3֑]0,U#*|;;ZJ5gNaghÀrⅡѤ3D}./Kz}Y)/̘b)W4jޕiK > 5c̜[&~Ip%5 %R]m-抴ſT{ +Ɇr!Td!c.$kpžv0ʒfl6܂zvtw. R<;`S.V,nWCzGHi(&ڌQ'fȢO->e1Xu-<ؠ{ZEׇ5e7(]I7YGJ^BdvyKjטVZV NLXt_U_bD4W!b? QdTՆopk[rf42/ۗ=VuN@6vCУ'$*)U7w 8!CrƔ"ܗÜ,ӹ('J7eDۯNpxwN@WEE3(pi o11I A s7U ,qĀe`)AqZ"('Adמ (9ia(9Oa;٧#]tT#s\FlE w#q*P!\TK[;_arL%]tc:8n\0ᐂA|^Qui*>?lT}]-mtܞ(EQ4S{׿i×iTQQk-UУ>Op\#"|<͋T14Qo>>ՠ)m& +jBmatQvBif +}2ː6j]W6ͅn]q' +y򭯕!tiڧ$~CKnendstream +endobj +4195 0 obj +1691 +endobj +4193 0 obj << +/Type /Page +/Contents 4194 0 R +/Resources 4192 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 4145 0 R +>> endobj +4191 0 obj << +/D [4193 0 R /XYZ 71.731 704.356 null] +>> endobj +4172 0 obj << +/D [4193 0 R /XYZ 456.882 595.764 null] +>> endobj +4178 0 obj << +/D [4193 0 R /XYZ 71.731 577.731 null] +>> endobj +4179 0 obj << +/D [4193 0 R /XYZ 71.731 528.915 null] +>> endobj +4180 0 obj << +/D [4193 0 R /XYZ 71.731 523.933 null] +>> endobj +4181 0 obj << +/D [4193 0 R /XYZ 71.731 523.933 null] +>> endobj +4182 0 obj << +/D [4193 0 R /XYZ 71.731 377.682 null] +>> endobj +4183 0 obj << +/D [4193 0 R /XYZ 71.731 357.593 null] +>> endobj +4184 0 obj << +/D [4193 0 R /XYZ 257.512 346.798 null] +>> endobj +4185 0 obj << +/D [4193 0 R /XYZ 362.956 346.798 null] +>> endobj +4186 0 obj << +/D [4193 0 R /XYZ 274.09 333.847 null] +>> endobj +1061 0 obj << +/D [4193 0 R /XYZ 71.731 316.746 null] +>> endobj +493 0 obj << +/D [4193 0 R /XYZ 198.492 289.493 null] +>> endobj +4187 0 obj << +/D [4193 0 R /XYZ 71.731 279.128 null] +>> endobj +4188 0 obj << +/D [4193 0 R /XYZ 71.731 231.347 null] +>> endobj +4189 0 obj << +/D [4193 0 R /XYZ 71.731 226.365 null] +>> endobj +4190 0 obj << +/D [4193 0 R /XYZ 71.731 226.365 null] +>> endobj +4192 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F37 1210 0 R /F40 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4228 0 obj << +/Length 4229 0 R +/Filter /FlateDecode +>> +stream +xڕXKs61%@MNg)E|(>dow} {1Fq/;s&i!^Xk7?)JzmuuUc+OX`N\bkx+Uen 㕨2v$9> 'b(L$(IW o7i J=?PB+–^m6vYtuͳ¶3n(5f+apTn7Vrs=j&\c!GYZ:r QF J(Ԛգ-U7'k6f [1#[E S@a1܋ D3zay±|Ȑ<' +n0y1w?@#ԧ\ӕRm3U`\:u\6[[BGl'(2y!Kck,?Kqc { xiE}NSdSE!1E1Sh/Q]qX ,}'~<ʞ9𣧧@h T蓉Ѿa6,?S-Uvvw୉])GmcW9~砬%ؒ>^KYXm~l-c32~&#SVF|F\+uƅ%SuVz %.Ecc22cְXmE<.S +޴~jھnWmE).ϝݬj/Zȶp)fGV LR9H>5t6.H]Uc-VT}a +j[tk +XR9-\F1_SaDP8:OנmF>}&`T>`p싡b1 Y@eSLfrEO[_гdN\,?l1Qw>s7%4Rcu^V6냪!蒯*qʄꍎ^ +n (So"~&~9&(pfNi N'8E1O;dۻO XM 7ɚiXFOA ϣw_%PEl` +S#/0J8R" +,HFDnܩV3 C4pdX9h'LSp7X1;;8JAâSzkrUO~[)!M\`^-}%K&l-8Ƴ9w g@zrXs/;`S$tYe }:v)]v$Ά/5á{Y]ŔWǬǟ2 N7! qѼfۺ V 7ŭSr '` u U`8aurTCQ:oo>' uCuoUR 5&+SDǕGݠ? oRKE)I{e?U-n /d' ~2-!ZɯKE % ̿&2C2Zo ]Ā*?(I4-^QvHSendstream +endobj +4229 0 obj +1677 +endobj +4227 0 obj << +/Type /Page +/Contents 4228 0 R +/Resources 4226 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 4145 0 R +>> endobj +4224 0 obj << +/D [4227 0 R /XYZ 71.731 705.352 null] +>> endobj +4225 0 obj << +/D [4227 0 R /XYZ 71.731 717.307 null] +>> endobj +4196 0 obj << +/D [4227 0 R /XYZ 71.731 608.317 null] +>> endobj +4197 0 obj << +/D [4227 0 R /XYZ 119.551 582.414 null] +>> endobj +1062 0 obj << +/D [4227 0 R /XYZ 71.731 567.282 null] +>> endobj +497 0 obj << +/D [4227 0 R /XYZ 274.267 538.06 null] +>> endobj +4198 0 obj << +/D [4227 0 R /XYZ 71.731 527.695 null] +>> endobj +4199 0 obj << +/D [4227 0 R /XYZ 119.551 504.984 null] +>> endobj +4200 0 obj << +/D [4227 0 R /XYZ 134.216 440.228 null] +>> endobj +4201 0 obj << +/D [4227 0 R /XYZ 71.731 428.108 null] +>> endobj +4202 0 obj << +/D [4227 0 R /XYZ 71.731 423.127 null] +>> endobj +4203 0 obj << +/D [4227 0 R /XYZ 71.731 423.127 null] +>> endobj +4204 0 obj << +/D [4227 0 R /XYZ 71.731 380.487 null] +>> endobj +4205 0 obj << +/D [4227 0 R /XYZ 71.731 370.524 null] +>> endobj +4206 0 obj << +/D [4227 0 R /XYZ 182.819 358.868 null] +>> endobj +1063 0 obj << +/D [4227 0 R /XYZ 71.731 297.698 null] +>> endobj +501 0 obj << +/D [4227 0 R /XYZ 289.401 268.288 null] +>> endobj +4207 0 obj << +/D [4227 0 R /XYZ 71.731 260.936 null] +>> endobj +4208 0 obj << +/D [4227 0 R /XYZ 71.731 228.074 null] +>> endobj +4209 0 obj << +/D [4227 0 R /XYZ 242.558 217.28 null] +>> endobj +4210 0 obj << +/D [4227 0 R /XYZ 290.159 217.28 null] +>> endobj +4211 0 obj << +/D [4227 0 R /XYZ 362.935 217.28 null] +>> endobj +1064 0 obj << +/D [4227 0 R /XYZ 71.731 205.877 null] +>> endobj +505 0 obj << +/D [4227 0 R /XYZ 290.316 182.809 null] +>> endobj +4212 0 obj << +/D [4227 0 R /XYZ 71.731 175.397 null] +>> endobj +4213 0 obj << +/D [4227 0 R /XYZ 152.487 163.88 null] +>> endobj +4214 0 obj << +/D [4227 0 R /XYZ 197.597 163.88 null] +>> endobj +4215 0 obj << +/D [4227 0 R /XYZ 265.841 163.88 null] +>> endobj +4216 0 obj << +/D [4227 0 R /XYZ 271.649 163.88 null] +>> endobj +4226 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F37 1210 0 R /F40 669 0 R /F62 669 0 R /F58 2248 0 R /F30 731 0 R /F42 669 0 R /F48 1503 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4272 0 obj << +/Length 4273 0 R +/Filter /FlateDecode +>> +stream +xڵXMo8r X~HT,HSn.\o@ŦcaeGߑ8%KM"@DRlBM#JD|ڽ`Ķ~wf4)OU~| +$b4Om/uܧSIiqgU098KۅL:]ie )I(&n9≉+qL__\ apP ,4.zJzZ^8ǾhZqdoSoG~$>ruj\02T5bwe +2NB5N:B78#8k ®xOU\tǨ$ N /hEMuO$ vmb8Zt͐$ ,ݹU: @k6\iE0n׫æwvgQr0$Z0}vf >ާ,J62HȽ>p%Su4Yef:pU3qۙڷslq:8 +\bӯ(i=fF.>Ʈ'NhB JSĉ:YUt͟ogrQ?AB6>|Kb!e5nģ6561h4|`#8ӈNgy`hܷ!`ElÈz$Q#Bx&gDZ̀|D{ѪYbcg+] Y('4? mb<*j}HF @{` hK׭%>כ }B{U;ڿJEp)Dh6Z ao +aomiak;4/~aFu\Q l+i}] zT관L{}//ŅB~|) @8%{P߮׫ !O؃Ng?n!0$3#~pu$0!j.,u6>%o#Yʹ1xendstream +endobj +4273 0 obj +1525 +endobj +4271 0 obj << +/Type /Page +/Contents 4272 0 R +/Resources 4270 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 4145 0 R +>> endobj +4269 0 obj << +/D [4271 0 R /XYZ 71.731 705.352 null] +>> endobj +4217 0 obj << +/D [4271 0 R /XYZ 71.731 657.532 null] +>> endobj +4218 0 obj << +/D [4271 0 R /XYZ 71.731 657.532 null] +>> endobj +4219 0 obj << +/D [4271 0 R /XYZ 226.359 634.618 null] +>> endobj +4220 0 obj << +/D [4271 0 R /XYZ 277.088 634.618 null] +>> endobj +4221 0 obj << +/D [4271 0 R /XYZ 331.125 634.618 null] +>> endobj +4222 0 obj << +/D [4271 0 R /XYZ 384.793 634.618 null] +>> endobj +4223 0 obj << +/D [4271 0 R /XYZ 182.365 621.666 null] +>> endobj +4230 0 obj << +/D [4271 0 R /XYZ 71.731 578.259 null] +>> endobj +4231 0 obj << +/D [4271 0 R /XYZ 71.731 553.358 null] +>> endobj +4232 0 obj << +/D [4271 0 R /XYZ 71.731 553.358 null] +>> endobj +1065 0 obj << +/D [4271 0 R /XYZ 71.731 523.764 null] +>> endobj +509 0 obj << +/D [4271 0 R /XYZ 300.932 498.429 null] +>> endobj +4233 0 obj << +/D [4271 0 R /XYZ 71.731 491.017 null] +>> endobj +4234 0 obj << +/D [4271 0 R /XYZ 152.487 479.5 null] +>> endobj +4235 0 obj << +/D [4271 0 R /XYZ 205.896 479.5 null] +>> endobj +4236 0 obj << +/D [4271 0 R /XYZ 256.208 479.5 null] +>> endobj +4237 0 obj << +/D [4271 0 R /XYZ 262.016 479.5 null] +>> endobj +4238 0 obj << +/D [4271 0 R /XYZ 306.18 479.5 null] +>> endobj +4239 0 obj << +/D [4271 0 R /XYZ 311.988 479.5 null] +>> endobj +4240 0 obj << +/D [4271 0 R /XYZ 71.731 472.362 null] +>> endobj +4241 0 obj << +/D [4271 0 R /XYZ 221.647 461.567 null] +>> endobj +4242 0 obj << +/D [4271 0 R /XYZ 462.261 461.567 null] +>> endobj +4243 0 obj << +/D [4271 0 R /XYZ 220.551 448.616 null] +>> endobj +4244 0 obj << +/D [4271 0 R /XYZ 419.801 448.616 null] +>> endobj +4245 0 obj << +/D [4271 0 R /XYZ 71.731 441.478 null] +>> endobj +4246 0 obj << +/D [4271 0 R /XYZ 443.582 430.683 null] +>> endobj +4247 0 obj << +/D [4271 0 R /XYZ 71.731 407.67 null] +>> endobj +4248 0 obj << +/D [4271 0 R /XYZ 71.731 402.689 null] +>> endobj +4249 0 obj << +/D [4271 0 R /XYZ 442.597 391.132 null] +>> endobj +1066 0 obj << +/D [4271 0 R /XYZ 71.731 329.961 null] +>> endobj +513 0 obj << +/D [4271 0 R /XYZ 269.717 304.457 null] +>> endobj +4250 0 obj << +/D [4271 0 R /XYZ 71.731 297.045 null] +>> endobj +4251 0 obj << +/D [4271 0 R /XYZ 152.487 285.528 null] +>> endobj +4252 0 obj << +/D [4271 0 R /XYZ 179.894 285.528 null] +>> endobj +4253 0 obj << +/D [4271 0 R /XYZ 248.138 285.528 null] +>> endobj +4254 0 obj << +/D [4271 0 R /XYZ 253.946 285.528 null] +>> endobj +4255 0 obj << +/D [4271 0 R /XYZ 71.731 278.39 null] +>> endobj +4256 0 obj << +/D [4271 0 R /XYZ 71.731 247.506 null] +>> endobj +4257 0 obj << +/D [4271 0 R /XYZ 71.731 247.506 null] +>> endobj +1067 0 obj << +/D [4271 0 R /XYZ 71.731 210.854 null] +>> endobj +517 0 obj << +/D [4271 0 R /XYZ 483.402 182.395 null] +>> endobj +4258 0 obj << +/D [4271 0 R /XYZ 71.731 173.501 null] +>> endobj +4259 0 obj << +/D [4271 0 R /XYZ 152.487 162.271 null] +>> endobj +4260 0 obj << +/D [4271 0 R /XYZ 214.922 162.271 null] +>> endobj +4261 0 obj << +/D [4271 0 R /XYZ 247.301 162.271 null] +>> endobj +4262 0 obj << +/D [4271 0 R /XYZ 267.724 162.271 null] +>> endobj +4263 0 obj << +/D [4271 0 R /XYZ 294.125 162.271 null] +>> endobj +4264 0 obj << +/D [4271 0 R /XYZ 314.548 162.271 null] +>> endobj +4265 0 obj << +/D [4271 0 R /XYZ 320.357 162.271 null] +>> endobj +4270 0 obj << +/Font << /F31 805 0 R /F30 731 0 R /F25 681 0 R /F37 1210 0 R /F52 1210 0 R /F42 669 0 R /F48 1503 0 R /F62 669 0 R /F58 2248 0 R /F40 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4288 0 obj << +/Length 4289 0 R +/Filter /FlateDecode +>> +stream +xڝXKo8rz&"[omDBeIKMݠ}I Ùo3C#KQ`!ã߀G䅁"~ys^P;ͮ$_&7{R +ʦsYAnu "?OHNt c:rS`>p82\x!r P:PrFdT^^CcӴ)Q&1DPh8e) lh(/,քK4$M7Dw8@xLoO ɕ}ncy6h CD7AmA^LL)(QJ +J d X"F;ϗ˲` й.yw5QꏩM(zfT}|{dchnVzr(m$9;ʆG2ڜOU C (jٗz9|}Fq3CeK8/ tVrPvM"鑉h2Q0r|`d!' +Gv/u\䇑CF&T |[Brpg$gCWе"p$h\ Ĝ tHn_sAK ; S*iہLlDֆJVt/xe]*!goT yg}81 BeS6vG^ߓ ێP3׆\d;&C9Vصr Xz^YA=y/ʎɜ99-m(eN:qk;}]`b9.CaZN<3p?"RO[ZG&M"lO=j~(Hvi6*&{NE6(ʣOG֏Kx̆mI-s}U=䆓ҕfХ^9% ̦Bv0w '$x:2Sl5աH+YKal(hQݮ>d0f&fD{'vKۇF -7ݓ_ұ3]nvm7#g0?_f_fJo~k+ =?X1sfoz#)grdDjXL{FaᘉZ6V ɺC NXUTiDŽr)86ܑITD2r61!g_K#5usњY/:Do80F"; +cK6"ȓ~dU9RA;txš$l,&(,5W[X`^ RMʺӺ~x:wQvSS'i,o)M誌]RDp3@!syZ\^ĩ|[s)o!u`oi\SMq>s:Dz|/fNWW^SO9 }M1XWq +pz80%+by_6Ӄ`̀n> endobj +4285 0 obj << +/D [4287 0 R /XYZ 71.731 705.352 null] +>> endobj +4266 0 obj << +/D [4287 0 R /XYZ 71.731 657.532 null] +>> endobj +4267 0 obj << +/D [4287 0 R /XYZ 71.731 657.532 null] +>> endobj +4268 0 obj << +/D [4287 0 R /XYZ 152.756 634.618 null] +>> endobj +4274 0 obj << +/D [4287 0 R /XYZ 71.731 609.547 null] +>> endobj +4275 0 obj << +/D [4287 0 R /XYZ 71.731 609.547 null] +>> endobj +4276 0 obj << +/D [4287 0 R /XYZ 71.731 565.828 null] +>> endobj +4277 0 obj << +/D [4287 0 R /XYZ 71.731 541.702 null] +>> endobj +4278 0 obj << +/D [4287 0 R /XYZ 71.731 541.702 null] +>> endobj +4279 0 obj << +/D [4287 0 R /XYZ 71.731 452.133 null] +>> endobj +4280 0 obj << +/D [4287 0 R /XYZ 71.731 442.17 null] +>> endobj +4281 0 obj << +/D [4287 0 R /XYZ 71.731 442.17 null] +>> endobj +4282 0 obj << +/D [4287 0 R /XYZ 71.731 173.379 null] +>> endobj +4283 0 obj << +/D [4287 0 R /XYZ 137.534 160.428 null] +>> endobj +4284 0 obj << +/D [4287 0 R /XYZ 252.969 160.428 null] +>> endobj +4286 0 obj << +/Font << /F31 805 0 R /F30 731 0 R /F25 681 0 R /F52 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4294 0 obj << +/Length 4295 0 R +/Filter /FlateDecode +>> +stream +xڅ? 0t\̥YPACXo4 0 +BHTDAeED"&pk͓k_wiFF +ƨ|52K?^x]Mpj+궪ӭ_$S,w_3I!3G&!V|埭""Kendstream +endobj +4295 0 obj +186 +endobj +4293 0 obj << +/Type /Page +/Contents 4294 0 R +/Resources 4292 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 4290 0 R +>> endobj +4291 0 obj << +/D [4293 0 R /XYZ 71.731 705.352 null] +>> endobj +4292 0 obj << +/Font << /F31 805 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4320 0 obj << +/Length 4321 0 R +/Filter /FlateDecode +>> +stream +xڭWKo6r̯0re{f7AQl(hH*Qu9,ŎCa$G$̃EDK,Vt癥~Jo-"I(kOfG Zr笑/(v"?D*ቧ}[Eo,}X3|E""ޥ)DssDǢ tOH.*"?[ACDo^W MoJr\gY idy9HDDj$:nPs*Veo;Cw<.$ >*+K1_Béy41μ=[$fg QA8B,% M\Lx@ 樛lYm!O Zv)QnqLh ΈH\j]QCdžE:{3x1)p̙ +os8Ejs DPE}n00`6q'bԻŅ}!sKR}8: :_ӀxɫPHO@lI [)VMң i'Df8;$PuB|^t1XLfNeFSm;p@o9`APr5uPr^B@PDژk( GJH;+NKp7xhՇYՔFĶ8M+ޖ4{g$TMh!Eyе$JQ)ƻG$TdHlĮj.Š}QHmJ>#&4IװFUp>/TQSQUV*vU]` +ѤUG_YjS'XB +E#qN's/Ł?_BK+@OhY!0Ffh`Bփ]%@K8=0Ŗ3cFUtpՆ/LNMnX;?k+zdF?SJ Ip3!9U}~VK_y}7dޥpMP׳rñȸ֕|T9maտ1fZ} +h֎SlvQ)pғԥ I/rpHQ93 3:0we7+:D6{=(eC]wѯ~YF"0mx$ONmnUɏ9@;>>sc_d,VtObd,y🰾Ky+зݦ߸ZDqA$!"Y ֺ$0|!2%ߖۧeE^} dݪendstream +endobj +4321 0 obj +1366 +endobj +4319 0 obj << +/Type /Page +/Contents 4320 0 R +/Resources 4318 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 4290 0 R +>> endobj +4317 0 obj << +/D [4319 0 R /XYZ 71.731 705.352 null] +>> endobj +1068 0 obj << +/D [4319 0 R /XYZ 71.731 657.532 null] +>> endobj +521 0 obj << +/D [4319 0 R /XYZ 489.584 642.462 null] +>> endobj +1069 0 obj << +/D [4319 0 R /XYZ 71.731 621.4 null] +>> endobj +525 0 obj << +/D [4319 0 R /XYZ 276.126 590.385 null] +>> endobj +4296 0 obj << +/D [4319 0 R /XYZ 71.731 581.562 null] +>> endobj +4297 0 obj << +/D [4319 0 R /XYZ 119.551 555.875 null] +>> endobj +4298 0 obj << +/D [4319 0 R /XYZ 175.351 555.875 null] +>> endobj +4299 0 obj << +/D [4319 0 R /XYZ 231.151 555.875 null] +>> endobj +4300 0 obj << +/D [4319 0 R /XYZ 302.264 555.875 null] +>> endobj +4301 0 obj << +/D [4319 0 R /XYZ 348.211 555.875 null] +>> endobj +4302 0 obj << +/D [4319 0 R /XYZ 71.731 548.871 null] +>> endobj +4303 0 obj << +/D [4319 0 R /XYZ 71.731 517.852 null] +>> endobj +4304 0 obj << +/D [4319 0 R /XYZ 144.178 507.058 null] +>> endobj +4305 0 obj << +/D [4319 0 R /XYZ 71.731 486.968 null] +>> endobj +4306 0 obj << +/D [4319 0 R /XYZ 71.731 464.054 null] +>> endobj +4307 0 obj << +/D [4319 0 R /XYZ 71.731 222.075 null] +>> endobj +4308 0 obj << +/D [4319 0 R /XYZ 71.731 192.14 null] +>> endobj +4309 0 obj << +/D [4319 0 R /XYZ 411.782 179.288 null] +>> endobj +4310 0 obj << +/D [4319 0 R /XYZ 119.551 166.337 null] +>> endobj +1070 0 obj << +/D [4319 0 R /XYZ 71.731 149.236 null] +>> endobj +4318 0 obj << +/Font << /F29 669 0 R /F36 669 0 R /F25 681 0 R /F30 731 0 R /F31 805 0 R /F37 1210 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4363 0 obj << +/Length 4364 0 R +/Filter /FlateDecode +>> +stream +xڥY[۸_a60EշfMHt,ql6RLCQd[̃h";MtQE%ᦺe_`άIHgjil<מ>LmFߙQac y_qQ#l˞ïݿ{w>bBdw$>NS4 QV i J,V;lr[!z3Wt3HL?` +y4t*eAg8#yه$Ikk cw`33> endobj +4360 0 obj << +/D [4362 0 R /XYZ 71.731 705.352 null] +>> endobj +529 0 obj << +/D [4362 0 R /XYZ 399.869 644.973 null] +>> endobj +4311 0 obj << +/D [4362 0 R /XYZ 71.731 632.802 null] +>> endobj +4312 0 obj << +/D [4362 0 R /XYZ 272.974 623.414 null] +>> endobj +4313 0 obj << +/D [4362 0 R /XYZ 316.968 623.414 null] +>> endobj +4314 0 obj << +/D [4362 0 R /XYZ 410.317 623.414 null] +>> endobj +4315 0 obj << +/D [4362 0 R /XYZ 123.616 610.463 null] +>> endobj +4316 0 obj << +/D [4362 0 R /XYZ 371.414 610.463 null] +>> endobj +4322 0 obj << +/D [4362 0 R /XYZ 71.731 590.373 null] +>> endobj +4323 0 obj << +/D [4362 0 R /XYZ 306.428 579.579 null] +>> endobj +4324 0 obj << +/D [4362 0 R /XYZ 144.448 566.627 null] +>> endobj +4325 0 obj << +/D [4362 0 R /XYZ 157.698 566.627 null] +>> endobj +4326 0 obj << +/D [4362 0 R /XYZ 431.179 566.627 null] +>> endobj +4327 0 obj << +/D [4362 0 R /XYZ 71.731 559.489 null] +>> endobj +4328 0 obj << +/D [4362 0 R /XYZ 287.629 548.695 null] +>> endobj +4329 0 obj << +/D [4362 0 R /XYZ 331.623 548.695 null] +>> endobj +4330 0 obj << +/D [4362 0 R /XYZ 406.422 548.695 null] +>> endobj +4331 0 obj << +/D [4362 0 R /XYZ 502.899 548.695 null] +>> endobj +4332 0 obj << +/D [4362 0 R /XYZ 119.551 535.743 null] +>> endobj +4333 0 obj << +/D [4362 0 R /XYZ 71.731 530.642 null] +>> endobj +4334 0 obj << +/D [4362 0 R /XYZ 71.731 530.642 null] +>> endobj +1071 0 obj << +/D [4362 0 R /XYZ 71.731 507.116 null] +>> endobj +533 0 obj << +/D [4362 0 R /XYZ 276.118 478.438 null] +>> endobj +4335 0 obj << +/D [4362 0 R /XYZ 71.731 471.086 null] +>> endobj +4336 0 obj << +/D [4362 0 R /XYZ 174.873 458.314 null] +>> endobj +4337 0 obj << +/D [4362 0 R /XYZ 119.551 445.363 null] +>> endobj +4338 0 obj << +/D [4362 0 R /XYZ 436.108 445.363 null] +>> endobj +4339 0 obj << +/D [4362 0 R /XYZ 119.551 432.411 null] +>> endobj +4340 0 obj << +/D [4362 0 R /XYZ 210.25 432.411 null] +>> endobj +4341 0 obj << +/D [4362 0 R /XYZ 186.051 419.46 null] +>> endobj +4342 0 obj << +/D [4362 0 R /XYZ 244.521 419.46 null] +>> endobj +4343 0 obj << +/D [4362 0 R /XYZ 71.731 412.322 null] +>> endobj +4344 0 obj << +/D [4362 0 R /XYZ 345.65 401.527 null] +>> endobj +4345 0 obj << +/D [4362 0 R /XYZ 427.183 401.527 null] +>> endobj +4346 0 obj << +/D [4362 0 R /XYZ 176.188 375.624 null] +>> endobj +4347 0 obj << +/D [4362 0 R /XYZ 229.588 375.624 null] +>> endobj +4348 0 obj << +/D [4362 0 R /XYZ 293.747 375.624 null] +>> endobj +4349 0 obj << +/D [4362 0 R /XYZ 369.711 375.624 null] +>> endobj +1072 0 obj << +/D [4362 0 R /XYZ 71.731 358.524 null] +>> endobj +537 0 obj << +/D [4362 0 R /XYZ 287.767 331.271 null] +>> endobj +4350 0 obj << +/D [4362 0 R /XYZ 71.731 320.906 null] +>> endobj +4351 0 obj << +/D [4362 0 R /XYZ 200.357 311.146 null] +>> endobj +4352 0 obj << +/D [4362 0 R /XYZ 298.588 311.146 null] +>> endobj +4353 0 obj << +/D [4362 0 R /XYZ 145.005 298.195 null] +>> endobj +4354 0 obj << +/D [4362 0 R /XYZ 447.547 298.195 null] +>> endobj +4355 0 obj << +/D [4362 0 R /XYZ 160.228 285.244 null] +>> endobj +1073 0 obj << +/D [4362 0 R /XYZ 71.731 245.229 null] +>> endobj +541 0 obj << +/D [4362 0 R /XYZ 312.829 212.094 null] +>> endobj +4356 0 obj << +/D [4362 0 R /XYZ 71.731 199.923 null] +>> endobj +4357 0 obj << +/D [4362 0 R /XYZ 459.995 190.535 null] +>> endobj +4358 0 obj << +/D [4362 0 R /XYZ 205.229 164.632 null] +>> endobj +4359 0 obj << +/D [4362 0 R /XYZ 459.631 151.681 null] +>> endobj +4361 0 obj << +/Font << /F31 805 0 R /F36 669 0 R /F25 681 0 R /F30 731 0 R /F37 1210 0 R /F48 1503 0 R /F39 1242 0 R /F40 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4372 0 obj << +/Length 4373 0 R +/Filter /FlateDecode +>> +stream +x}Un0zWh+aY&i@P =cE_ߕv)+@;3;$yǃ<&XA~/c:.beYloVWaEuv+Z'n]F(T*[)pL-gVb9XY.6夢0$n^ q$m;W +{\l׭BYSĆ=P{M+2DN1 Ӝ-cx [D 5?{JSV)fendstream +endobj +4373 0 obj +822 +endobj +4371 0 obj << +/Type /Page +/Contents 4372 0 R +/Resources 4370 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 4290 0 R +>> endobj +4369 0 obj << +/D [4371 0 R /XYZ 71.731 705.352 null] +>> endobj +4365 0 obj << +/D [4371 0 R /XYZ 71.731 627.48 null] +>> endobj +4366 0 obj << +/D [4371 0 R /XYZ 289.441 603.734 null] +>> endobj +4367 0 obj << +/D [4371 0 R /XYZ 332.43 590.782 null] +>> endobj +4368 0 obj << +/D [4371 0 R /XYZ 71.731 583.644 null] +>> endobj +4370 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F30 731 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4377 0 obj << +/Length 4378 0 R +/Filter /FlateDecode +>> +stream +xڅ;0 P'qBV^ 1B> endobj +4374 0 obj << +/D [4376 0 R /XYZ 71.731 705.352 null] +>> endobj +4375 0 obj << +/Font << /F31 805 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4398 0 obj << +/Length 4399 0 R +/Filter /FlateDecode +>> +stream +xڥXێ6[l Vt>nӸA k%VW"UZ!-M>ȹ93)̲"/JY^?f~fGa ~e7.[7ކY{immHEbf3ʟEy[ȹ*aWotd^$F2 =?:+/@)793OF +)^NQ/?fƸ2S'R܃Uߙ#_ wyvh5CZG2}2fkc i3oΖaQKJeFW*;a[}-x=/Teai+{J)[*x(U{ 'Ѐsa:P@"LpjvԔHEzz$x8͍BPXdZߛcɔ!S2PV[`@!hY&~kPUhޓ1A@&_<Bř@':,_0f7T\Yh`;R.m) +cIY83~?3 +GBF8)&zXI.5c1S4.s5 +=v9ꙷMށ354SwxCR(&(&#mͨbVot ,dq2 |~ÙBmm_$Wqr~oؔXSg~e4 tM>9n%AOx⎢n>kB_Պ^*Ll׶@i|N=$2 eEvqP'+2–׌8 +K H-> endobj +4395 0 obj << +/D [4397 0 R /XYZ 71.731 705.352 null] +>> endobj +1074 0 obj << +/D [4397 0 R /XYZ 71.731 657.532 null] +>> endobj +545 0 obj << +/D [4397 0 R /XYZ 339.193 642.462 null] +>> endobj +1075 0 obj << +/D [4397 0 R /XYZ 71.731 621.08 null] +>> endobj +549 0 obj << +/D [4397 0 R /XYZ 262.544 590.385 null] +>> endobj +1076 0 obj << +/D [4397 0 R /XYZ 71.731 580.207 null] +>> endobj +553 0 obj << +/D [4397 0 R /XYZ 463.82 551.013 null] +>> endobj +4379 0 obj << +/D [4397 0 R /XYZ 71.731 540.648 null] +>> endobj +1077 0 obj << +/D [4397 0 R /XYZ 71.731 477.922 null] +>> endobj +557 0 obj << +/D [4397 0 R /XYZ 484.998 444.788 null] +>> endobj +4380 0 obj << +/D [4397 0 R /XYZ 71.731 432.35 null] +>> endobj +4381 0 obj << +/D [4397 0 R /XYZ 71.731 351.333 null] +>> endobj +4382 0 obj << +/D [4397 0 R /XYZ 71.731 320.449 null] +>> endobj +1078 0 obj << +/D [4397 0 R /XYZ 71.731 266.651 null] +>> endobj +561 0 obj << +/D [4397 0 R /XYZ 290.811 233.517 null] +>> endobj +4383 0 obj << +/D [4397 0 R /XYZ 71.731 221.345 null] +>> endobj +4384 0 obj << +/D [4397 0 R /XYZ 261.829 211.957 null] +>> endobj +4385 0 obj << +/D [4397 0 R /XYZ 513.632 211.957 null] +>> endobj +4386 0 obj << +/D [4397 0 R /XYZ 136.428 199.006 null] +>> endobj +4387 0 obj << +/D [4397 0 R /XYZ 119.551 186.055 null] +>> endobj +4388 0 obj << +/D [4397 0 R /XYZ 284.172 173.103 null] +>> endobj +4389 0 obj << +/D [4397 0 R /XYZ 119.551 147.201 null] +>> endobj +4390 0 obj << +/D [4397 0 R /XYZ 163.267 147.201 null] +>> endobj +4396 0 obj << +/Font << /F29 669 0 R /F36 669 0 R /F40 669 0 R /F25 681 0 R /F31 805 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4404 0 obj << +/Length 4405 0 R +/Filter /FlateDecode +>> +stream +x}Rn0F*m0Fv,[!l'~FmRExwaHT(0ʳ$ ++A jsrQ8EK(=]mOjV2qq9G&fzl&^չ +>ϟ0"(]eU t0l6<1rhT73[}4Jd㨍 t'kݰ> x- 渗AGsl7gs +mx ɠ(Y(ֹq49GQj9tFrzm\n"ȁ6 loq5fWgP2+ vQendstream +endobj +4405 0 obj +342 +endobj +4403 0 obj << +/Type /Page +/Contents 4404 0 R +/Resources 4402 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 4400 0 R +>> endobj +4401 0 obj << +/D [4403 0 R /XYZ 71.731 705.352 null] +>> endobj +4391 0 obj << +/D [4403 0 R /XYZ 71.731 657.532 null] +>> endobj +4392 0 obj << +/D [4403 0 R /XYZ 299.982 647.569 null] +>> endobj +4393 0 obj << +/D [4403 0 R /XYZ 488.711 647.569 null] +>> endobj +4394 0 obj << +/D [4403 0 R /XYZ 150.535 634.618 null] +>> endobj +4402 0 obj << +/Font << /F31 805 0 R /F25 681 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4429 0 obj << +/Length 4430 0 R +/Filter /FlateDecode +>> +stream +xڕXَ80elXe\vNMPt<msB +I|\b% Js&/,h1?4 ;>17{6GI +7$Zo Dt9x>T3 EAEξb9\Gͨݛ|/$NqDmŸ(%g!Gq:|4?dHT ̛E3]EE ;?wу{i#UōHƜ{>_uB0Vj>2!1 WGss҉n$7nhOoG&]2[9;˶LpϏGwm=,+sr|akKQ9ˈaφc38Vi\9tqoW䶳C& Kۯ~uЗěh#:Slc=V%e\JiιWrZ$ǿ`h ׂ7I:(X[ؓpW#sr(M޼a-`qBƈ_E Nc*|+bܘ8PZEG,\|Л9 g乢Z:;68CM_E +K890`ke&sSeW~&ïLIa}YzlV%:+/1Ttr^aW6lIƘƳ+ vȧT J-eV:Jia @x!KѹP$~OtȮ +9( O3nFhmʽ5d3èq{ӾR9\pcpWЌr ɬ[3CM#C9* hʽGeg2Ob9jYh޶q Vi3Ji,Ј!AYe`}6K-7)p>V .d(7DqPm@nz4<5t/H>\BCM͟SKvʟ!yM0vv&^.cTd4ՆК?6݃'>칪֏]jLv؞{6o$Vs(e/]-,]k)u*@Ho!DCmV';*YKbBe'V?5bN)rUA}z_ʶ@FQ@ (ՔlE*_!s負*f?GqW~;j8^LOhMT{mh/&,8 vR_~sk~ABa,TAՉgVvzBT h" p170CrnRD̀AowKejuȇ +Km^H{M!cDޥ8xW# S*ږxj΢UgG■WdZzQ׍gNXrc|Ok=jXV_`o6TMT j[E q)QvR--"pԦk~<1rC$Th}I9Q{pvۈ53ݿSj)kƝ[|QOo`%[irJڗloi!I)嫏AO/!q~P1䐐߉ѿ 4C1"&p3G( +*T)JDuk-遍; +-܃ª@UC8+z{Èx4TSw*݋'"V +Zf@4/_`_StJa%~ uY#4/|#,Z4Yŧg!ֺla{'uJ~9|eXt:Ơ7i\O$C +gAXwjGhendstream +endobj +4430 0 obj +2101 +endobj +4428 0 obj << +/Type /Page +/Contents 4429 0 R +/Resources 4427 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 4400 0 R +>> endobj +4425 0 obj << +/D [4428 0 R /XYZ 71.731 705.352 null] +>> endobj +1079 0 obj << +/D [4428 0 R /XYZ 71.731 657.532 null] +>> endobj +565 0 obj << +/D [4428 0 R /XYZ 315.085 642.462 null] +>> endobj +1080 0 obj << +/D [4428 0 R /XYZ 71.731 621.4 null] +>> endobj +569 0 obj << +/D [4428 0 R /XYZ 269.327 590.385 null] +>> endobj +1081 0 obj << +/D [4428 0 R /XYZ 71.731 576.592 null] +>> endobj +573 0 obj << +/D [4428 0 R /XYZ 360.429 551.013 null] +>> endobj +4406 0 obj << +/D [4428 0 R /XYZ 71.731 540.87 null] +>> endobj +4407 0 obj << +/D [4428 0 R /XYZ 158.843 440.229 null] +>> endobj +4408 0 obj << +/D [4428 0 R /XYZ 71.731 433.091 null] +>> endobj +4409 0 obj << +/D [4428 0 R /XYZ 504.612 396.393 null] +>> endobj +4410 0 obj << +/D [4428 0 R /XYZ 375.897 383.442 null] +>> endobj +4411 0 obj << +/D [4428 0 R /XYZ 425.38 357.539 null] +>> endobj +4412 0 obj << +/D [4428 0 R /XYZ 365.366 331.636 null] +>> endobj +4413 0 obj << +/D [4428 0 R /XYZ 343.947 305.733 null] +>> endobj +4414 0 obj << +/D [4428 0 R /XYZ 248.018 292.782 null] +>> endobj +4415 0 obj << +/D [4428 0 R /XYZ 391.009 279.831 null] +>> endobj +4416 0 obj << +/D [4428 0 R /XYZ 71.731 259.741 null] +>> endobj +4417 0 obj << +/D [4428 0 R /XYZ 71.731 259.741 null] +>> endobj +4418 0 obj << +/D [4428 0 R /XYZ 348.081 235.995 null] +>> endobj +4419 0 obj << +/D [4428 0 R /XYZ 71.731 190.003 null] +>> endobj +4420 0 obj << +/D [4428 0 R /XYZ 71.731 190.003 null] +>> endobj +4421 0 obj << +/D [4428 0 R /XYZ 425.789 179.208 null] +>> endobj +4422 0 obj << +/D [4428 0 R /XYZ 71.731 159.119 null] +>> endobj +4423 0 obj << +/D [4428 0 R /XYZ 71.731 159.119 null] +>> endobj +4426 0 obj << +/D [4428 0 R /XYZ 71.731 83.686 null] +>> endobj +4427 0 obj << +/Font << /F29 669 0 R /F36 669 0 R /F40 669 0 R /F25 681 0 R /F31 805 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4448 0 obj << +/Length 4449 0 R +/Filter /FlateDecode +>> +stream +xڥWo6ؿBoVM%D\e _Hʒ+!Ew}wG(?ee!>ǑٯNS^ Lӈ<\>5@rݟ[gN2Iȏnfi󟖃"CQk]re5ROK6q v8طq{s|^ld hf%ʓl"mN58OѽP2;r# +]$ YxuRO2tԦ"F ӈ ^qs/ã2^[`̂G֜!DNK7{{pox; ĴR{ ,y:\y eQp1 &s#~m_UǴU~k7!4c*^3ʻ;ݷ;l]Bt ҩ? {Ml~pi`%$!F p"Z}7M¯nE|AfA O2{ߧaNgjbY2ߙd7 PZC8`|ᑔ0IT,`%S:ӗ)u_T endstream +endobj +4449 0 obj +1338 +endobj +4447 0 obj << +/Type /Page +/Contents 4448 0 R +/Resources 4446 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 4400 0 R +>> endobj +4445 0 obj << +/D [4447 0 R /XYZ 71.731 704.356 null] +>> endobj +4424 0 obj << +/D [4447 0 R /XYZ 361.859 647.569 null] +>> endobj +1082 0 obj << +/D [4447 0 R /XYZ 71.731 591.679 null] +>> endobj +577 0 obj << +/D [4447 0 R /XYZ 326.79 556.487 null] +>> endobj +1083 0 obj << +/D [4447 0 R /XYZ 71.731 546.309 null] +>> endobj +581 0 obj << +/D [4447 0 R /XYZ 315.928 517.115 null] +>> endobj +1084 0 obj << +/D [4447 0 R /XYZ 71.731 506.973 null] +>> endobj +585 0 obj << +/D [4447 0 R /XYZ 233.506 482.644 null] +>> endobj +4431 0 obj << +/D [4447 0 R /XYZ 71.731 476.517 null] +>> endobj +4432 0 obj << +/D [4447 0 R /XYZ 137.534 463.715 null] +>> endobj +4433 0 obj << +/D [4447 0 R /XYZ 522.654 450.764 null] +>> endobj +1085 0 obj << +/D [4447 0 R /XYZ 71.731 420.712 null] +>> endobj +589 0 obj << +/D [4447 0 R /XYZ 230.063 397.364 null] +>> endobj +4434 0 obj << +/D [4447 0 R /XYZ 71.731 391.237 null] +>> endobj +4435 0 obj << +/D [4447 0 R /XYZ 137.534 378.435 null] +>> endobj +1086 0 obj << +/D [4447 0 R /XYZ 71.731 330.396 null] +>> endobj +593 0 obj << +/D [4447 0 R /XYZ 400.299 295.323 null] +>> endobj +1087 0 obj << +/D [4447 0 R /XYZ 71.731 285.146 null] +>> endobj +597 0 obj << +/D [4447 0 R /XYZ 250.194 255.951 null] +>> endobj +4436 0 obj << +/D [4447 0 R /XYZ 71.731 245.809 null] +>> endobj +4437 0 obj << +/D [4447 0 R /XYZ 119.551 222.875 null] +>> endobj +4438 0 obj << +/D [4447 0 R /XYZ 431.805 222.875 null] +>> endobj +4439 0 obj << +/D [4447 0 R /XYZ 185.851 209.924 null] +>> endobj +4440 0 obj << +/D [4447 0 R /XYZ 379.044 209.924 null] +>> endobj +4441 0 obj << +/D [4447 0 R /XYZ 71.731 189.834 null] +>> endobj +4442 0 obj << +/D [4447 0 R /XYZ 71.731 166.92 null] +>> endobj +4443 0 obj << +/D [4447 0 R /XYZ 71.731 161.939 null] +>> endobj +4444 0 obj << +/D [4447 0 R /XYZ 71.731 161.939 null] +>> endobj +4446 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F36 669 0 R /F40 669 0 R /F42 669 0 R /F30 731 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4471 0 obj << +/Length 4472 0 R +/Filter /FlateDecode +>> +stream +xXnF]K +''ɠ-&660 [Em4H@Rv%y!єE k H1+6!&Wo MLfٶ+ elpnuv!R7pua;}ec7͗-~4EQkwOT P7Ľ*sQ~=2J\Ʌuct<K+LVD.LQq8Q( Fx SUPjV9-ځS[Mou+-:a(*:au:׼ )+@;'{\ekA!,bt88ZuHkösY&,FĜńhߧ(S/ +AnRig:ZAfD9EKY6@=MUSZ02QKntrY!W(T@ICV^tmZk[8l mA9҇p[ylG:e#9Z{w?UĀKicA/=&MSQ?ԝx(I?N?+;1ChM@E[ITS,95^Yۭ6]ϳEZQ(*jK&NLJc)UCw>ҫ<+l;hے!J}Nء9BIΟ +t̩q3CDb|LšN9A󑎔rD6Mw-̆|=BLzH]vFG={E&UJߨϽĬ}Ggg]_q5[ 1-=ZeK^Tz.@;Xp T˪WtNzJd@72_ 9Elknb1Q79r“JOiy:N1Lb ފ[cJnqTڇ +5ltb`szZl.΋f FKkE %(4a>x:?ڋ27tf&Eendstream +endobj +4472 0 obj +1637 +endobj +4470 0 obj << +/Type /Page +/Contents 4471 0 R +/Resources 4469 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 4400 0 R +>> endobj +4467 0 obj << +/D [4470 0 R /XYZ 71.731 705.352 null] +>> endobj +4468 0 obj << +/D [4470 0 R /XYZ 71.731 717.307 null] +>> endobj +4450 0 obj << +/D [4470 0 R /XYZ 71.731 563.485 null] +>> endobj +4451 0 obj << +/D [4470 0 R /XYZ 71.731 540.471 null] +>> endobj +4452 0 obj << +/D [4470 0 R /XYZ 71.731 530.509 null] +>> endobj +4453 0 obj << +/D [4470 0 R /XYZ 71.731 530.509 null] +>> endobj +4454 0 obj << +/D [4470 0 R /XYZ 71.731 346.4 null] +>> endobj +4455 0 obj << +/D [4470 0 R /XYZ 137.534 333.448 null] +>> endobj +4456 0 obj << +/D [4470 0 R /XYZ 119.551 320.497 null] +>> endobj +4457 0 obj << +/D [4470 0 R /XYZ 287.937 320.497 null] +>> endobj +4458 0 obj << +/D [4470 0 R /XYZ 366.841 307.546 null] +>> endobj +4459 0 obj << +/D [4470 0 R /XYZ 71.731 287.456 null] +>> endobj +4460 0 obj << +/D [4470 0 R /XYZ 71.731 256.572 null] +>> endobj +4461 0 obj << +/D [4470 0 R /XYZ 71.731 233.658 null] +>> endobj +4462 0 obj << +/D [4470 0 R /XYZ 71.731 228.677 null] +>> endobj +4463 0 obj << +/D [4470 0 R /XYZ 71.731 228.677 null] +>> endobj +4464 0 obj << +/D [4470 0 R /XYZ 71.731 175.676 null] +>> endobj +4465 0 obj << +/D [4470 0 R /XYZ 71.731 142.635 null] +>> endobj +4469 0 obj << +/Font << /F31 805 0 R /F25 681 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4491 0 obj << +/Length 4492 0 R +/Filter /FlateDecode +>> +stream +xڍX[8汿"&i;z:Cpk4{lv \3YϙD<,$;p&81L„8^0HGrOHzV[_uM˜=no/Vr &s+gf}'&7bcCb >s)  +7w"xjw ZdvڳKEpL{Ң#ރ`'% +V\82x}Z(Es1"-SֿluIhoyfTd5+-^ )ǽb +YZڷum[Tg($mu׫`NJ=T{;GywpX_F.@B=Ԛ^=st ӑ\ F=Gܱ.eֳXV$TMIݭW],y&j#Դke? cpآe73v+x0VU(n@5MܱfC~@===/bw$:RBI$Ț +E?jRQ#Sg=.)P"endstream +endobj +4492 0 obj +1698 +endobj +4490 0 obj << +/Type /Page +/Contents 4491 0 R +/Resources 4489 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 4400 0 R +>> endobj +4488 0 obj << +/D [4490 0 R /XYZ 71.731 705.352 null] +>> endobj +4466 0 obj << +/D [4490 0 R /XYZ 71.731 657.532 null] +>> endobj +4473 0 obj << +/D [4490 0 R /XYZ 71.731 596.596 null] +>> endobj +4474 0 obj << +/D [4490 0 R /XYZ 71.731 565.712 null] +>> endobj +4475 0 obj << +/D [4490 0 R /XYZ 71.731 495.973 null] +>> endobj +4476 0 obj << +/D [4490 0 R /XYZ 71.731 495.973 null] +>> endobj +4477 0 obj << +/D [4490 0 R /XYZ 71.731 480.078 null] +>> endobj +1088 0 obj << +/D [4490 0 R /XYZ 71.731 414.28 null] +>> endobj +601 0 obj << +/D [4490 0 R /XYZ 425.142 381.145 null] +>> endobj +1089 0 obj << +/D [4490 0 R /XYZ 71.731 367.352 null] +>> endobj +605 0 obj << +/D [4490 0 R /XYZ 322.527 341.773 null] +>> endobj +4478 0 obj << +/D [4490 0 R /XYZ 71.731 331.63 null] +>> endobj +4479 0 obj << +/D [4490 0 R /XYZ 396.26 321.649 null] +>> endobj +4480 0 obj << +/D [4490 0 R /XYZ 239.649 282.794 null] +>> endobj +4481 0 obj << +/D [4490 0 R /XYZ 365.206 282.794 null] +>> endobj +4482 0 obj << +/D [4490 0 R /XYZ 149.698 269.843 null] +>> endobj +4483 0 obj << +/D [4490 0 R /XYZ 435.891 269.843 null] +>> endobj +1090 0 obj << +/D [4490 0 R /XYZ 71.731 267.686 null] +>> endobj +609 0 obj << +/D [4490 0 R /XYZ 158.983 240.433 null] +>> endobj +4484 0 obj << +/D [4490 0 R /XYZ 71.731 227.279 null] +>> endobj +613 0 obj << +/D [4490 0 R /XYZ 155.569 195.179 null] +>> endobj +4485 0 obj << +/D [4490 0 R /XYZ 71.731 186.357 null] +>> endobj +4486 0 obj << +/D [4490 0 R /XYZ 71.731 157.874 null] +>> endobj +4487 0 obj << +/D [4490 0 R /XYZ 71.731 134.96 null] +>> endobj +4489 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F30 731 0 R /F36 669 0 R /F40 669 0 R /F48 1503 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4511 0 obj << +/Length 4512 0 R +/Filter /FlateDecode +>> +stream +xڕWr6c䌅䤶ǗZJ3m-y^;?(Du.2țInQ/7_b?fl5qCDsN(ȾJ02[G-4p@D0{+DǔCHc4Q8CCDf``7E5oy\9Էeܦÿ\!:U*>!u +t H|R|d>u3@ +psrN^aEQ0f!Qԙz(TAnx. ֣;&.Eܟ&?-DoEV4ϱ݋}U RtShIaDF~ͣ,']hap\-`L'ɠQԃ~k݈@8dt #9NaVlwPX\7QݤSnvS.߫՟֥)~m5#BD*Tl\9DޡjH(xpl -!o!DZǷv\|'6.̓)ƒCGlle:Nť(@B'٬c:9vP*A"09tZ~+Z5)#W~1LO(6Z..+ąPBPs:b`To^ū>SŤُn֢LW;jNt1F)r9 S]w۸)yIUxrY ÆRoAWQmwWJ9/g(t_$/.@>OcӮ} mc@_OR2m|q1^ADw7AL@5A> endobj +4508 0 obj << +/D [4510 0 R /XYZ 71.731 705.352 null] +>> endobj +1740 0 obj << +/D [4510 0 R /XYZ 71.731 630.469 null] +>> endobj +617 0 obj << +/D [4510 0 R /XYZ 239.407 597.334 null] +>> endobj +4493 0 obj << +/D [4510 0 R /XYZ 71.731 585.163 null] +>> endobj +4494 0 obj << +/D [4510 0 R /XYZ 71.731 560.029 null] +>> endobj +4495 0 obj << +/D [4510 0 R /XYZ 71.731 537.234 null] +>> endobj +4496 0 obj << +/D [4510 0 R /XYZ 71.731 514.201 null] +>> endobj +4497 0 obj << +/D [4510 0 R /XYZ 71.731 491.287 null] +>> endobj +4498 0 obj << +/D [4510 0 R /XYZ 71.731 455.422 null] +>> endobj +4499 0 obj << +/D [4510 0 R /XYZ 71.731 409.594 null] +>> endobj +621 0 obj << +/D [4510 0 R /XYZ 442.908 376.459 null] +>> endobj +4500 0 obj << +/D [4510 0 R /XYZ 71.731 364.288 null] +>> endobj +4501 0 obj << +/D [4510 0 R /XYZ 71.731 339.273 null] +>> endobj +4502 0 obj << +/D [4510 0 R /XYZ 71.731 316.24 null] +>> endobj +4503 0 obj << +/D [4510 0 R /XYZ 71.731 283.363 null] +>> endobj +625 0 obj << +/D [4510 0 R /XYZ 479.524 250.229 null] +>> endobj +4504 0 obj << +/D [4510 0 R /XYZ 71.731 238.058 null] +>> endobj +4505 0 obj << +/D [4510 0 R /XYZ 71.731 200.092 null] +>> endobj +4506 0 obj << +/D [4510 0 R /XYZ 71.731 164.227 null] +>> endobj +4509 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F36 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4534 0 obj << +/Length 4535 0 R +/Filter /FlateDecode +>> +stream +xXMo8z̯Qj?Ni8b"Ff!;IY{]`yo޼L0|D$|N _}5q' R4O'.~gSo6_8!mWSO'd%!dڝ|,i9c'D +"Yؖ!Ab +!a%)څ{rnlP.N$Fo3B${E(qIY{(Tp%$8(*cv|Ͳs󕽸 !E7.SO>,1M,?}/܂yuZ#Bc?O E +tţ)D)^ʐ"% +[IcM`X;po ^8ΊdJ b~v现Ig&]i Ą1yxB!&>J|O\2@>*z +ζMœL>z_L^k۪~vZFGpfbnkU!}1CL}>msQ'uBwkGkdj5e( &/i^ >s1:wdmZjGߗ".%UdBr;-QMY|,K._;#B# +1}wmXg(+1i<0Iҧׁ=F{"7另aCB(&ިZZC=XK,+9-1*]_G +*MzL]X?l-W; +yl´$F{"֛M},FJ$S1REXެ~ t}tʐAE4 YC`XޜLf{pV.@Z;KM*P mjhlo2FgPPqwx`+.K<Z?i2#Ik?ʯ1jqҚpD1q:۔iQL@RCglS5SJ },*c Yڷv)ʑQ!p=+O.wr<"u]nڶ~35ÒM^2yz  `D8dr;[\.ˢ]q(Pе ,U5g3 4[܅ +w值Aܼ@{'nL z +abN)!ףS0Y^{Uw?׵Kaif;hW e*8"? FCE2uYĨcb`@$ yaA A8;zepc&7|"{̇endstream +endobj +4535 0 obj +1381 +endobj +4533 0 obj << +/Type /Page +/Contents 4534 0 R +/Resources 4532 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 4513 0 R +>> endobj +4530 0 obj << +/D [4533 0 R /XYZ 71.731 705.352 null] +>> endobj +4531 0 obj << +/D [4533 0 R /XYZ 71.731 717.307 null] +>> endobj +4507 0 obj << +/D [4533 0 R /XYZ 71.731 657.532 null] +>> endobj +629 0 obj << +/D [4533 0 R /XYZ 267.089 645.343 null] +>> endobj +4514 0 obj << +/D [4533 0 R /XYZ 71.731 633.172 null] +>> endobj +4515 0 obj << +/D [4533 0 R /XYZ 71.731 597.025 null] +>> endobj +4516 0 obj << +/D [4533 0 R /XYZ 71.731 572.173 null] +>> endobj +4517 0 obj << +/D [4533 0 R /XYZ 71.731 536.427 null] +>> endobj +4518 0 obj << +/D [4533 0 R /XYZ 71.731 500.442 null] +>> endobj +4519 0 obj << +/D [4533 0 R /XYZ 71.731 466.515 null] +>> endobj +4520 0 obj << +/D [4533 0 R /XYZ 71.731 428.712 null] +>> endobj +4521 0 obj << +/D [4533 0 R /XYZ 71.731 395.955 null] +>> endobj +633 0 obj << +/D [4533 0 R /XYZ 241.989 362.7 null] +>> endobj +4522 0 obj << +/D [4533 0 R /XYZ 71.731 353.878 null] +>> endobj +4523 0 obj << +/D [4533 0 R /XYZ 71.731 325.396 null] +>> endobj +4524 0 obj << +/D [4533 0 R /XYZ 71.731 302.601 null] +>> endobj +4525 0 obj << +/D [4533 0 R /XYZ 71.731 279.687 null] +>> endobj +4526 0 obj << +/D [4533 0 R /XYZ 71.731 246.81 null] +>> endobj +637 0 obj << +/D [4533 0 R /XYZ 203.737 213.556 null] +>> endobj +4527 0 obj << +/D [4533 0 R /XYZ 71.731 201.385 null] +>> endobj +4528 0 obj << +/D [4533 0 R /XYZ 71.731 176.371 null] +>> endobj +4529 0 obj << +/D [4533 0 R /XYZ 71.731 153.337 null] +>> endobj +4532 0 obj << +/Font << /F31 805 0 R /F36 669 0 R /F25 681 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4552 0 obj << +/Length 4553 0 R +/Filter /FlateDecode +>> +stream +xڥWn8+d fŗKIIE Zm"䴙KT%Y :dZ$9s)< '1F1Kd~'٧i"L"4ynY0#wߗϮi0YkaNQ6O$!.\$DQbkbmo>O1&{i꧒&>LfgASQ+]N,U-K%쯍.BX܉c|2+A^]$8NQGns;>J5qFD(dFJ1EثEkZA8w|׫s\ +2'|{F c :vi,!3'"jmRTyqb2\: (!Xpj,BPX[ 'bN˞yom Dv'E+ 8v!*u&/㞱\g+E43Y +HJNR\Pp):M͹wm.X=jRn5h'z󐕅 P!Fs8uUI|U&ݪa/KH#IMFFxt hLj DpϹTuBkuFG8C.2n{;7-S:H3/2M?G9(NxkLC=Z}G{f h XVqzW^=3C.TJTuN{/A:ޖz s:skK6>ѕ5y7|?v[zǓ-ʥEZ+f#@!tO> endobj +4549 0 obj << +/D [4551 0 R /XYZ 71.731 705.352 null] +>> endobj +4536 0 obj << +/D [4551 0 R /XYZ 71.731 640.431 null] +>> endobj +4537 0 obj << +/D [4551 0 R /XYZ 71.731 607.555 null] +>> endobj +641 0 obj << +/D [4551 0 R /XYZ 270.687 574.42 null] +>> endobj +4538 0 obj << +/D [4551 0 R /XYZ 71.731 565.597 null] +>> endobj +4539 0 obj << +/D [4551 0 R /XYZ 71.731 526.101 null] +>> endobj +4540 0 obj << +/D [4551 0 R /XYZ 71.731 501.369 null] +>> endobj +4541 0 obj << +/D [4551 0 R /XYZ 71.731 478.336 null] +>> endobj +4542 0 obj << +/D [4551 0 R /XYZ 71.731 455.422 null] +>> endobj +4543 0 obj << +/D [4551 0 R /XYZ 71.731 421.494 null] +>> endobj +4544 0 obj << +/D [4551 0 R /XYZ 71.731 386.68 null] +>> endobj +645 0 obj << +/D [4551 0 R /XYZ 466.819 353.545 null] +>> endobj +4545 0 obj << +/D [4551 0 R /XYZ 71.731 341.374 null] +>> endobj +1091 0 obj << +/D [4551 0 R /XYZ 71.731 284.081 null] +>> endobj +649 0 obj << +/D [4551 0 R /XYZ 212.034 256.111 null] +>> endobj +4546 0 obj << +/D [4551 0 R /XYZ 71.731 248.758 null] +>> endobj +1092 0 obj << +/D [4551 0 R /XYZ 71.731 167.08 null] +>> endobj +653 0 obj << +/D [4551 0 R /XYZ 318.826 139.827 null] +>> endobj +4547 0 obj << +/D [4551 0 R /XYZ 71.731 132.475 null] +>> endobj +1093 0 obj << +/D [4551 0 R /XYZ 71.731 132.475 null] +>> endobj +4550 0 obj << +/Font << /F31 805 0 R /F25 681 0 R /F36 669 0 R /F40 669 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4557 0 obj << +/Length 4558 0 R +/Filter /FlateDecode +>> +stream +xڅMO@{l:HG-b-[ ALfy,F'*KFʷI|/\Nx-XI)mg2I 9t#yo|p%IsP +WK̨DG\c-q_6E0u30p i`Vqkzfߡ.U(6\u64no(L珝l/b@FA8=wPt6!8U>stB~uendstream +endobj +4558 0 obj +285 +endobj +4556 0 obj << +/Type /Page +/Contents 4557 0 R +/Resources 4555 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 4513 0 R +>> endobj +4554 0 obj << +/D [4556 0 R /XYZ 71.731 705.352 null] +>> endobj +657 0 obj << +/D [4556 0 R /XYZ 158.983 647.066 null] +>> endobj +4548 0 obj << +/D [4556 0 R /XYZ 71.731 636.701 null] +>> endobj +4555 0 obj << +/Font << /F31 805 0 R /F40 669 0 R /F25 681 0 R >> +/ProcSet [ /PDF /Text ] +>> endobj +4562 0 obj << +/Length 4563 0 R +/Filter /FlateDecode +>> +stream +x3T0BC 67730VH2Tl˛B&PRD҄bendstream +endobj +4563 0 obj +62 +endobj +4561 0 obj << +/Type /Page +/Contents 4562 0 R +/Resources 4560 0 R +/MediaBox [0 0 609.711 789.038] +/Parent 4513 0 R +>> endobj +4559 0 obj << +/D [4561 0 R /XYZ 0 777.083 null] +>> endobj +4560 0 obj << +/ProcSet [ /PDF ] +>> endobj +4564 0 obj << +/Type /Encoding +/Differences [ 0/.notdef 1/dotaccent/fi/fl/fraction/hungarumlaut/Lslash/lslash/ogonek/ring 10/.notdef 11/breve/minus 13/.notdef 14/Zcaron/zcaron/caron/dotlessi/dotlessj/ff/ffi/ffl 22/.notdef 30/grave/quotesingle/space/exclam/quotedbl/numbersign/dollar/percent/ampersand/quoteright/parenleft/parenright/asterisk/plus/comma/hyphen/period/slash/zero/one/two/three/four/five/six/seven/eight/nine/colon/semicolon/less/equal/greater/question/at/A/B/C/D/E/F/G/H/I/J/K/L/M/N/O/P/Q/R/S/T/U/V/W/X/Y/Z/bracketleft/backslash/bracketright/asciicircum/underscore/quoteleft/a/b/c/d/e/f/g/h/i/j/k/l/m/n/o/p/q/r/s/t/u/v/w/x/y/z/braceleft/bar/braceright/asciitilde 127/.notdef 130/quotesinglbase/florin/quotedblbase/ellipsis/dagger/daggerdbl/circumflex/perthousand/Scaron/guilsinglleft/OE 141/.notdef 147/quotedblleft/quotedblright/bullet/endash/emdash/tilde/trademark/scaron/guilsinglright/oe 157/.notdef 159/Ydieresis 160/.notdef 161/exclamdown/cent/sterling/currency/yen/brokenbar/section/dieresis/copyright/ordfeminine/guillemotleft/logicalnot/hyphen/registered/macron/degree/plusminus/twosuperior/threesuperior/acute/mu/paragraph/periodcentered/cedilla/onesuperior/ordmasculine/guillemotright/onequarter/onehalf/threequarters/questiondown/Agrave/Aacute/Acircumflex/Atilde/Adieresis/Aring/AE/Ccedilla/Egrave/Eacute/Ecircumflex/Edieresis/Igrave/Iacute/Icircumflex/Idieresis/Eth/Ntilde/Ograve/Oacute/Ocircumflex/Otilde/Odieresis/multiply/Oslash/Ugrave/Uacute/Ucircumflex/Udieresis/Yacute/Thorn/germandbls/agrave/aacute/acircumflex/atilde/adieresis/aring/ae/ccedilla/egrave/eacute/ecircumflex/edieresis/igrave/iacute/icircumflex/idieresis/eth/ntilde/ograve/oacute/ocircumflex/otilde/odieresis/divide/oslash/ugrave/uacute/ucircumflex/udieresis/yacute/thorn/ydieresis] +>> endobj +2248 0 obj << +/Type /Font +/Subtype /Type1 +/Encoding 4564 0 R +/BaseFont /Helvetica +>> endobj +1590 0 obj << +/Type /Font +/Subtype /Type1 +/Encoding 4564 0 R +/BaseFont /Courier-BoldOblique +>> endobj +1503 0 obj << +/Type /Font +/Subtype /Type1 +/Encoding 4564 0 R +/BaseFont /Courier-Oblique +>> endobj +1242 0 obj << +/Type /Font +/Subtype /Type1 +/Encoding 4564 0 R +/BaseFont /Courier-Bold +>> endobj +1210 0 obj << +/Type /Font +/Subtype /Type1 +/Encoding 4564 0 R +/BaseFont /Courier +>> endobj +805 0 obj << +/Type /Font +/Subtype /Type1 +/Encoding 4564 0 R +/BaseFont /Times-Italic +>> endobj +731 0 obj << +/Type /Font +/Subtype /Type1 +/Encoding 4564 0 R +/BaseFont /Times-Bold +>> endobj +681 0 obj << +/Type /Font +/Subtype /Type1 +/Encoding 4564 0 R +/BaseFont /Times-Roman +>> endobj +669 0 obj << +/Type /Font +/Subtype /Type1 +/Encoding 4564 0 R +/BaseFont /Helvetica-Bold +>> endobj +670 0 obj << +/Type /Pages +/Count 6 +/Parent 4565 0 R +/Kids [661 0 R 673 0 R 678 0 R 684 0 R 727 0 R 848 0 R] +>> endobj +1056 0 obj << +/Type /Pages +/Count 6 +/Parent 4565 0 R +/Kids [973 0 R 1096 0 R 1207 0 R 1239 0 R 1276 0 R 1305 0 R] +>> endobj +1329 0 obj << +/Type /Pages +/Count 6 +/Parent 4565 0 R +/Kids [1326 0 R 1332 0 R 1349 0 R 1375 0 R 1387 0 R 1398 0 R] +>> endobj +1447 0 obj << +/Type /Pages +/Count 6 +/Parent 4565 0 R +/Kids [1441 0 R 1500 0 R 1587 0 R 1649 0 R 1695 0 R 1716 0 R] +>> endobj +1746 0 obj << +/Type /Pages +/Count 6 +/Parent 4565 0 R +/Kids [1743 0 R 1775 0 R 1829 0 R 1836 0 R 1846 0 R 1873 0 R] +>> endobj +1903 0 obj << +/Type /Pages +/Count 6 +/Parent 4565 0 R +/Kids [1900 0 R 1931 0 R 1955 0 R 1985 0 R 2019 0 R 2043 0 R] +>> endobj +2066 0 obj << +/Type /Pages +/Count 6 +/Parent 4566 0 R +/Kids [2063 0 R 2087 0 R 2101 0 R 2106 0 R 2133 0 R 2138 0 R] +>> endobj +2161 0 obj << +/Type /Pages +/Count 6 +/Parent 4566 0 R +/Kids [2157 0 R 2166 0 R 2199 0 R 2224 0 R 2245 0 R 2264 0 R] +>> endobj +2322 0 obj << +/Type /Pages +/Count 6 +/Parent 4566 0 R +/Kids [2316 0 R 2357 0 R 2416 0 R 2458 0 R 2509 0 R 2528 0 R] +>> endobj +2551 0 obj << +/Type /Pages +/Count 6 +/Parent 4566 0 R +/Kids [2548 0 R 2567 0 R 2618 0 R 2753 0 R 2833 0 R 2888 0 R] +>> endobj +2936 0 obj << +/Type /Pages +/Count 6 +/Parent 4566 0 R +/Kids [2932 0 R 2995 0 R 3044 0 R 3094 0 R 3128 0 R 3173 0 R] +>> endobj +3213 0 obj << +/Type /Pages +/Count 6 +/Parent 4566 0 R +/Kids [3210 0 R 3254 0 R 3312 0 R 3358 0 R 3396 0 R 3430 0 R] +>> endobj +3460 0 obj << +/Type /Pages +/Count 6 +/Parent 4567 0 R +/Kids [3455 0 R 3471 0 R 3506 0 R 3554 0 R 3585 0 R 3622 0 R] +>> endobj +3672 0 obj << +/Type /Pages +/Count 6 +/Parent 4567 0 R +/Kids [3669 0 R 3716 0 R 3743 0 R 3858 0 R 3863 0 R 3868 0 R] +>> endobj +3920 0 obj << +/Type /Pages +/Count 6 +/Parent 4567 0 R +/Kids [3913 0 R 3957 0 R 4020 0 R 4078 0 R 4110 0 R 4124 0 R] +>> endobj +4145 0 obj << +/Type /Pages +/Count 6 +/Parent 4567 0 R +/Kids [4141 0 R 4154 0 R 4175 0 R 4193 0 R 4227 0 R 4271 0 R] +>> endobj +4290 0 obj << +/Type /Pages +/Count 6 +/Parent 4567 0 R +/Kids [4287 0 R 4293 0 R 4319 0 R 4362 0 R 4371 0 R 4376 0 R] +>> endobj +4400 0 obj << +/Type /Pages +/Count 6 +/Parent 4567 0 R +/Kids [4397 0 R 4403 0 R 4428 0 R 4447 0 R 4470 0 R 4490 0 R] +>> endobj +4513 0 obj << +/Type /Pages +/Count 5 +/Parent 4568 0 R +/Kids [4510 0 R 4533 0 R 4551 0 R 4556 0 R 4561 0 R] +>> endobj +4565 0 obj << +/Type /Pages +/Count 36 +/Parent 4569 0 R +/Kids [670 0 R 1056 0 R 1329 0 R 1447 0 R 1746 0 R 1903 0 R] +>> endobj +4566 0 obj << +/Type /Pages +/Count 36 +/Parent 4569 0 R +/Kids [2066 0 R 2161 0 R 2322 0 R 2551 0 R 2936 0 R 3213 0 R] +>> endobj +4567 0 obj << +/Type /Pages +/Count 36 +/Parent 4569 0 R +/Kids [3460 0 R 3672 0 R 3920 0 R 4145 0 R 4290 0 R 4400 0 R] +>> endobj +4568 0 obj << +/Type /Pages +/Count 5 +/Parent 4569 0 R +/Kids [4513 0 R] +>> endobj +4569 0 obj << +/Type /Pages +/Count 113 +/Kids [4565 0 R 4566 0 R 4567 0 R 4568 0 R] +>> endobj +4570 0 obj << +/Type /Outlines +/First 3 0 R +/Last 567 0 R +/Count 10 +>> endobj +659 0 obj << +/Title 660 0 R +/A 658 0 R +/Parent 647 0 R +/Prev 655 0 R +>> endobj +655 0 obj << +/Title 656 0 R +/A 654 0 R +/Parent 647 0 R +/Prev 651 0 R +/Next 659 0 R +>> endobj +651 0 obj << +/Title 652 0 R +/A 650 0 R +/Parent 647 0 R +/Next 655 0 R +>> endobj +647 0 obj << +/Title 648 0 R +/A 646 0 R +/Parent 567 0 R +/Prev 643 0 R +/First 651 0 R +/Last 659 0 R +/Count -3 +>> endobj +643 0 obj << +/Title 644 0 R +/A 642 0 R +/Parent 567 0 R +/Prev 639 0 R +/Next 647 0 R +>> endobj +639 0 obj << +/Title 640 0 R +/A 638 0 R +/Parent 567 0 R +/Prev 635 0 R +/Next 643 0 R +>> endobj +635 0 obj << +/Title 636 0 R +/A 634 0 R +/Parent 567 0 R +/Prev 631 0 R +/Next 639 0 R +>> endobj +631 0 obj << +/Title 632 0 R +/A 630 0 R +/Parent 567 0 R +/Prev 627 0 R +/Next 635 0 R +>> endobj +627 0 obj << +/Title 628 0 R +/A 626 0 R +/Parent 567 0 R +/Prev 623 0 R +/Next 631 0 R +>> endobj +623 0 obj << +/Title 624 0 R +/A 622 0 R +/Parent 567 0 R +/Prev 619 0 R +/Next 627 0 R +>> endobj +619 0 obj << +/Title 620 0 R +/A 618 0 R +/Parent 567 0 R +/Prev 615 0 R +/Next 623 0 R +>> endobj +615 0 obj << +/Title 616 0 R +/A 614 0 R +/Parent 567 0 R +/Prev 603 0 R +/Next 619 0 R +>> endobj +611 0 obj << +/Title 612 0 R +/A 610 0 R +/Parent 603 0 R +/Prev 607 0 R +>> endobj +607 0 obj << +/Title 608 0 R +/A 606 0 R +/Parent 603 0 R +/Next 611 0 R +>> endobj +603 0 obj << +/Title 604 0 R +/A 602 0 R +/Parent 567 0 R +/Prev 595 0 R +/Next 615 0 R +/First 607 0 R +/Last 611 0 R +/Count -2 +>> endobj +599 0 obj << +/Title 600 0 R +/A 598 0 R +/Parent 595 0 R +>> endobj +595 0 obj << +/Title 596 0 R +/A 594 0 R +/Parent 567 0 R +/Prev 579 0 R +/Next 603 0 R +/First 599 0 R +/Last 599 0 R +/Count -1 +>> endobj +591 0 obj << +/Title 592 0 R +/A 590 0 R +/Parent 583 0 R +/Prev 587 0 R +>> endobj +587 0 obj << +/Title 588 0 R +/A 586 0 R +/Parent 583 0 R +/Next 591 0 R +>> endobj +583 0 obj << +/Title 584 0 R +/A 582 0 R +/Parent 579 0 R +/First 587 0 R +/Last 591 0 R +/Count -2 +>> endobj +579 0 obj << +/Title 580 0 R +/A 578 0 R +/Parent 567 0 R +/Prev 571 0 R +/Next 595 0 R +/First 583 0 R +/Last 583 0 R +/Count -1 +>> endobj +575 0 obj << +/Title 576 0 R +/A 574 0 R +/Parent 571 0 R +>> endobj +571 0 obj << +/Title 572 0 R +/A 570 0 R +/Parent 567 0 R +/Next 579 0 R +/First 575 0 R +/Last 575 0 R +/Count -1 +>> endobj +567 0 obj << +/Title 568 0 R +/A 566 0 R +/Parent 4570 0 R +/Prev 547 0 R +/First 571 0 R +/Last 647 0 R +/Count -13 +>> endobj +563 0 obj << +/Title 564 0 R +/A 562 0 R +/Parent 547 0 R +/Prev 559 0 R +>> endobj +559 0 obj << +/Title 560 0 R +/A 558 0 R +/Parent 547 0 R +/Prev 551 0 R +/Next 563 0 R +>> endobj +555 0 obj << +/Title 556 0 R +/A 554 0 R +/Parent 551 0 R +>> endobj +551 0 obj << +/Title 552 0 R +/A 550 0 R +/Parent 547 0 R +/Next 559 0 R +/First 555 0 R +/Last 555 0 R +/Count -1 +>> endobj +547 0 obj << +/Title 548 0 R +/A 546 0 R +/Parent 4570 0 R +/Prev 523 0 R +/Next 567 0 R +/First 551 0 R +/Last 563 0 R +/Count -3 +>> endobj +543 0 obj << +/Title 544 0 R +/A 542 0 R +/Parent 523 0 R +/Prev 531 0 R +>> endobj +539 0 obj << +/Title 540 0 R +/A 538 0 R +/Parent 531 0 R +/Prev 535 0 R +>> endobj +535 0 obj << +/Title 536 0 R +/A 534 0 R +/Parent 531 0 R +/Next 539 0 R +>> endobj +531 0 obj << +/Title 532 0 R +/A 530 0 R +/Parent 523 0 R +/Prev 527 0 R +/Next 543 0 R +/First 535 0 R +/Last 539 0 R +/Count -2 +>> endobj +527 0 obj << +/Title 528 0 R +/A 526 0 R +/Parent 523 0 R +/Next 531 0 R +>> endobj +523 0 obj << +/Title 524 0 R +/A 522 0 R +/Parent 4570 0 R +/Prev 267 0 R +/Next 547 0 R +/First 527 0 R +/Last 543 0 R +/Count -3 +>> endobj +519 0 obj << +/Title 520 0 R +/A 518 0 R +/Parent 475 0 R +/Prev 503 0 R +>> endobj +515 0 obj << +/Title 516 0 R +/A 514 0 R +/Parent 503 0 R +/Prev 511 0 R +>> endobj +511 0 obj << +/Title 512 0 R +/A 510 0 R +/Parent 503 0 R +/Prev 507 0 R +/Next 515 0 R +>> endobj +507 0 obj << +/Title 508 0 R +/A 506 0 R +/Parent 503 0 R +/Next 511 0 R +>> endobj +503 0 obj << +/Title 504 0 R +/A 502 0 R +/Parent 475 0 R +/Prev 499 0 R +/Next 519 0 R +/First 507 0 R +/Last 515 0 R +/Count -3 +>> endobj +499 0 obj << +/Title 500 0 R +/A 498 0 R +/Parent 475 0 R +/Prev 495 0 R +/Next 503 0 R +>> endobj +495 0 obj << +/Title 496 0 R +/A 494 0 R +/Parent 475 0 R +/Prev 491 0 R +/Next 499 0 R +>> endobj +491 0 obj << +/Title 492 0 R +/A 490 0 R +/Parent 475 0 R +/Prev 479 0 R +/Next 495 0 R +>> endobj +487 0 obj << +/Title 488 0 R +/A 486 0 R +/Parent 479 0 R +/Prev 483 0 R +>> endobj +483 0 obj << +/Title 484 0 R +/A 482 0 R +/Parent 479 0 R +/Next 487 0 R +>> endobj +479 0 obj << +/Title 480 0 R +/A 478 0 R +/Parent 475 0 R +/Next 491 0 R +/First 483 0 R +/Last 487 0 R +/Count -2 +>> endobj +475 0 obj << +/Title 476 0 R +/A 474 0 R +/Parent 267 0 R +/Prev 299 0 R +/First 479 0 R +/Last 519 0 R +/Count -6 +>> endobj +471 0 obj << +/Title 472 0 R +/A 470 0 R +/Parent 455 0 R +/Prev 467 0 R +>> endobj +467 0 obj << +/Title 468 0 R +/A 466 0 R +/Parent 455 0 R +/Prev 463 0 R +/Next 471 0 R +>> endobj +463 0 obj << +/Title 464 0 R +/A 462 0 R +/Parent 455 0 R +/Prev 459 0 R +/Next 467 0 R +>> endobj +459 0 obj << +/Title 460 0 R +/A 458 0 R +/Parent 455 0 R +/Next 463 0 R +>> endobj +455 0 obj << +/Title 456 0 R +/A 454 0 R +/Parent 299 0 R +/Prev 451 0 R +/First 459 0 R +/Last 471 0 R +/Count -4 +>> endobj +451 0 obj << +/Title 452 0 R +/A 450 0 R +/Parent 299 0 R +/Prev 447 0 R +/Next 455 0 R +>> endobj +447 0 obj << +/Title 448 0 R +/A 446 0 R +/Parent 299 0 R +/Prev 443 0 R +/Next 451 0 R +>> endobj +443 0 obj << +/Title 444 0 R +/A 442 0 R +/Parent 299 0 R +/Prev 439 0 R +/Next 447 0 R +>> endobj +439 0 obj << +/Title 440 0 R +/A 438 0 R +/Parent 299 0 R +/Prev 435 0 R +/Next 443 0 R +>> endobj +435 0 obj << +/Title 436 0 R +/A 434 0 R +/Parent 299 0 R +/Prev 431 0 R +/Next 439 0 R +>> endobj +431 0 obj << +/Title 432 0 R +/A 430 0 R +/Parent 299 0 R +/Prev 427 0 R +/Next 435 0 R +>> endobj +427 0 obj << +/Title 428 0 R +/A 426 0 R +/Parent 299 0 R +/Prev 363 0 R +/Next 431 0 R +>> endobj +423 0 obj << +/Title 424 0 R +/A 422 0 R +/Parent 363 0 R +/Prev 419 0 R +>> endobj +419 0 obj << +/Title 420 0 R +/A 418 0 R +/Parent 363 0 R +/Prev 415 0 R +/Next 423 0 R +>> endobj +415 0 obj << +/Title 416 0 R +/A 414 0 R +/Parent 363 0 R +/Prev 411 0 R +/Next 419 0 R +>> endobj +411 0 obj << +/Title 412 0 R +/A 410 0 R +/Parent 363 0 R +/Prev 407 0 R +/Next 415 0 R +>> endobj +407 0 obj << +/Title 408 0 R +/A 406 0 R +/Parent 363 0 R +/Prev 403 0 R +/Next 411 0 R +>> endobj +403 0 obj << +/Title 404 0 R +/A 402 0 R +/Parent 363 0 R +/Prev 399 0 R +/Next 407 0 R +>> endobj +399 0 obj << +/Title 400 0 R +/A 398 0 R +/Parent 363 0 R +/Prev 395 0 R +/Next 403 0 R +>> endobj +395 0 obj << +/Title 396 0 R +/A 394 0 R +/Parent 363 0 R +/Prev 391 0 R +/Next 399 0 R +>> endobj +391 0 obj << +/Title 392 0 R +/A 390 0 R +/Parent 363 0 R +/Prev 387 0 R +/Next 395 0 R +>> endobj +387 0 obj << +/Title 388 0 R +/A 386 0 R +/Parent 363 0 R +/Prev 383 0 R +/Next 391 0 R +>> endobj +383 0 obj << +/Title 384 0 R +/A 382 0 R +/Parent 363 0 R +/Prev 379 0 R +/Next 387 0 R +>> endobj +379 0 obj << +/Title 380 0 R +/A 378 0 R +/Parent 363 0 R +/Prev 375 0 R +/Next 383 0 R +>> endobj +375 0 obj << +/Title 376 0 R +/A 374 0 R +/Parent 363 0 R +/Prev 371 0 R +/Next 379 0 R +>> endobj +371 0 obj << +/Title 372 0 R +/A 370 0 R +/Parent 363 0 R +/Prev 367 0 R +/Next 375 0 R +>> endobj +367 0 obj << +/Title 368 0 R +/A 366 0 R +/Parent 363 0 R +/Next 371 0 R +>> endobj +363 0 obj << +/Title 364 0 R +/A 362 0 R +/Parent 299 0 R +/Prev 359 0 R +/Next 427 0 R +/First 367 0 R +/Last 423 0 R +/Count -15 +>> endobj +359 0 obj << +/Title 360 0 R +/A 358 0 R +/Parent 299 0 R +/Prev 355 0 R +/Next 363 0 R +>> endobj +355 0 obj << +/Title 356 0 R +/A 354 0 R +/Parent 299 0 R +/Prev 351 0 R +/Next 359 0 R +>> endobj +351 0 obj << +/Title 352 0 R +/A 350 0 R +/Parent 299 0 R +/Prev 339 0 R +/Next 355 0 R +>> endobj +347 0 obj << +/Title 348 0 R +/A 346 0 R +/Parent 339 0 R +/Prev 343 0 R +>> endobj +343 0 obj << +/Title 344 0 R +/A 342 0 R +/Parent 339 0 R +/Next 347 0 R +>> endobj +339 0 obj << +/Title 340 0 R +/A 338 0 R +/Parent 299 0 R +/Prev 335 0 R +/Next 351 0 R +/First 343 0 R +/Last 347 0 R +/Count -2 +>> endobj +335 0 obj << +/Title 336 0 R +/A 334 0 R +/Parent 299 0 R +/Prev 331 0 R +/Next 339 0 R +>> endobj +331 0 obj << +/Title 332 0 R +/A 330 0 R +/Parent 299 0 R +/Prev 327 0 R +/Next 335 0 R +>> endobj +327 0 obj << +/Title 328 0 R +/A 326 0 R +/Parent 299 0 R +/Prev 323 0 R +/Next 331 0 R +>> endobj +323 0 obj << +/Title 324 0 R +/A 322 0 R +/Parent 299 0 R +/Prev 319 0 R +/Next 327 0 R +>> endobj +319 0 obj << +/Title 320 0 R +/A 318 0 R +/Parent 299 0 R +/Prev 315 0 R +/Next 323 0 R +>> endobj +315 0 obj << +/Title 316 0 R +/A 314 0 R +/Parent 299 0 R +/Prev 311 0 R +/Next 319 0 R +>> endobj +311 0 obj << +/Title 312 0 R +/A 310 0 R +/Parent 299 0 R +/Prev 307 0 R +/Next 315 0 R +>> endobj +307 0 obj << +/Title 308 0 R +/A 306 0 R +/Parent 299 0 R +/Prev 303 0 R +/Next 311 0 R +>> endobj +303 0 obj << +/Title 304 0 R +/A 302 0 R +/Parent 299 0 R +/Next 307 0 R +>> endobj +299 0 obj << +/Title 300 0 R +/A 298 0 R +/Parent 267 0 R +/Prev 271 0 R +/Next 475 0 R +/First 303 0 R +/Last 455 0 R +/Count -22 +>> endobj +295 0 obj << +/Title 296 0 R +/A 294 0 R +/Parent 287 0 R +/Prev 291 0 R +>> endobj +291 0 obj << +/Title 292 0 R +/A 290 0 R +/Parent 287 0 R +/Next 295 0 R +>> endobj +287 0 obj << +/Title 288 0 R +/A 286 0 R +/Parent 271 0 R +/Prev 275 0 R +/First 291 0 R +/Last 295 0 R +/Count -2 +>> endobj +283 0 obj << +/Title 284 0 R +/A 282 0 R +/Parent 275 0 R +/Prev 279 0 R +>> endobj +279 0 obj << +/Title 280 0 R +/A 278 0 R +/Parent 275 0 R +/Next 283 0 R +>> endobj +275 0 obj << +/Title 276 0 R +/A 274 0 R +/Parent 271 0 R +/Next 287 0 R +/First 279 0 R +/Last 283 0 R +/Count -2 +>> endobj +271 0 obj << +/Title 272 0 R +/A 270 0 R +/Parent 267 0 R +/Next 299 0 R +/First 275 0 R +/Last 287 0 R +/Count -2 +>> endobj +267 0 obj << +/Title 268 0 R +/A 266 0 R +/Parent 4570 0 R +/Prev 255 0 R +/Next 523 0 R +/First 271 0 R +/Last 475 0 R +/Count -3 +>> endobj +263 0 obj << +/Title 264 0 R +/A 262 0 R +/Parent 255 0 R +/Prev 259 0 R +>> endobj +259 0 obj << +/Title 260 0 R +/A 258 0 R +/Parent 255 0 R +/Next 263 0 R +>> endobj +255 0 obj << +/Title 256 0 R +/A 254 0 R +/Parent 4570 0 R +/Prev 135 0 R +/Next 267 0 R +/First 259 0 R +/Last 263 0 R +/Count -2 +>> endobj +251 0 obj << +/Title 252 0 R +/A 250 0 R +/Parent 223 0 R +/Prev 247 0 R +>> endobj +247 0 obj << +/Title 248 0 R +/A 246 0 R +/Parent 223 0 R +/Prev 243 0 R +/Next 251 0 R +>> endobj +243 0 obj << +/Title 244 0 R +/A 242 0 R +/Parent 223 0 R +/Prev 231 0 R +/Next 247 0 R +>> endobj +239 0 obj << +/Title 240 0 R +/A 238 0 R +/Parent 231 0 R +/Prev 235 0 R +>> endobj +235 0 obj << +/Title 236 0 R +/A 234 0 R +/Parent 231 0 R +/Next 239 0 R +>> endobj +231 0 obj << +/Title 232 0 R +/A 230 0 R +/Parent 223 0 R +/Prev 227 0 R +/Next 243 0 R +/First 235 0 R +/Last 239 0 R +/Count -2 +>> endobj +227 0 obj << +/Title 228 0 R +/A 226 0 R +/Parent 223 0 R +/Next 231 0 R +>> endobj +223 0 obj << +/Title 224 0 R +/A 222 0 R +/Parent 135 0 R +/Prev 199 0 R +/First 227 0 R +/Last 251 0 R +/Count -5 +>> endobj +219 0 obj << +/Title 220 0 R +/A 218 0 R +/Parent 199 0 R +/Prev 215 0 R +>> endobj +215 0 obj << +/Title 216 0 R +/A 214 0 R +/Parent 199 0 R +/Prev 211 0 R +/Next 219 0 R +>> endobj +211 0 obj << +/Title 212 0 R +/A 210 0 R +/Parent 199 0 R +/Prev 207 0 R +/Next 215 0 R +>> endobj +207 0 obj << +/Title 208 0 R +/A 206 0 R +/Parent 199 0 R +/Prev 203 0 R +/Next 211 0 R +>> endobj +203 0 obj << +/Title 204 0 R +/A 202 0 R +/Parent 199 0 R +/Next 207 0 R +>> endobj +199 0 obj << +/Title 200 0 R +/A 198 0 R +/Parent 135 0 R +/Prev 195 0 R +/Next 223 0 R +/First 203 0 R +/Last 219 0 R +/Count -5 +>> endobj +195 0 obj << +/Title 196 0 R +/A 194 0 R +/Parent 135 0 R +/Prev 191 0 R +/Next 199 0 R +>> endobj +191 0 obj << +/Title 192 0 R +/A 190 0 R +/Parent 135 0 R +/Prev 155 0 R +/Next 195 0 R +>> endobj +187 0 obj << +/Title 188 0 R +/A 186 0 R +/Parent 155 0 R +/Prev 183 0 R +>> endobj +183 0 obj << +/Title 184 0 R +/A 182 0 R +/Parent 155 0 R +/Prev 179 0 R +/Next 187 0 R +>> endobj +179 0 obj << +/Title 180 0 R +/A 178 0 R +/Parent 155 0 R +/Prev 175 0 R +/Next 183 0 R +>> endobj +175 0 obj << +/Title 176 0 R +/A 174 0 R +/Parent 155 0 R +/Prev 171 0 R +/Next 179 0 R +>> endobj +171 0 obj << +/Title 172 0 R +/A 170 0 R +/Parent 155 0 R +/Prev 159 0 R +/Next 175 0 R +>> endobj +167 0 obj << +/Title 168 0 R +/A 166 0 R +/Parent 159 0 R +/Prev 163 0 R +>> endobj +163 0 obj << +/Title 164 0 R +/A 162 0 R +/Parent 159 0 R +/Next 167 0 R +>> endobj +159 0 obj << +/Title 160 0 R +/A 158 0 R +/Parent 155 0 R +/Next 171 0 R +/First 163 0 R +/Last 167 0 R +/Count -2 +>> endobj +155 0 obj << +/Title 156 0 R +/A 154 0 R +/Parent 135 0 R +/Prev 151 0 R +/Next 191 0 R +/First 159 0 R +/Last 187 0 R +/Count -6 +>> endobj +151 0 obj << +/Title 152 0 R +/A 150 0 R +/Parent 135 0 R +/Prev 147 0 R +/Next 155 0 R +>> endobj +147 0 obj << +/Title 148 0 R +/A 146 0 R +/Parent 135 0 R +/Prev 139 0 R +/Next 151 0 R +>> endobj +143 0 obj << +/Title 144 0 R +/A 142 0 R +/Parent 139 0 R +>> endobj +139 0 obj << +/Title 140 0 R +/A 138 0 R +/Parent 135 0 R +/Next 147 0 R +/First 143 0 R +/Last 143 0 R +/Count -1 +>> endobj +135 0 obj << +/Title 136 0 R +/A 134 0 R +/Parent 4570 0 R +/Prev 91 0 R +/Next 255 0 R +/First 139 0 R +/Last 223 0 R +/Count -8 +>> endobj +131 0 obj << +/Title 132 0 R +/A 130 0 R +/Parent 115 0 R +/Prev 119 0 R +>> endobj +127 0 obj << +/Title 128 0 R +/A 126 0 R +/Parent 119 0 R +/Prev 123 0 R +>> endobj +123 0 obj << +/Title 124 0 R +/A 122 0 R +/Parent 119 0 R +/Next 127 0 R +>> endobj +119 0 obj << +/Title 120 0 R +/A 118 0 R +/Parent 115 0 R +/Next 131 0 R +/First 123 0 R +/Last 127 0 R +/Count -2 +>> endobj +115 0 obj << +/Title 116 0 R +/A 114 0 R +/Parent 91 0 R +/Prev 111 0 R +/First 119 0 R +/Last 131 0 R +/Count -2 +>> endobj +111 0 obj << +/Title 112 0 R +/A 110 0 R +/Parent 91 0 R +/Prev 107 0 R +/Next 115 0 R +>> endobj +107 0 obj << +/Title 108 0 R +/A 106 0 R +/Parent 91 0 R +/Prev 95 0 R +/Next 111 0 R +>> endobj +103 0 obj << +/Title 104 0 R +/A 102 0 R +/Parent 95 0 R +/Prev 99 0 R +>> endobj +99 0 obj << +/Title 100 0 R +/A 98 0 R +/Parent 95 0 R +/Next 103 0 R +>> endobj +95 0 obj << +/Title 96 0 R +/A 94 0 R +/Parent 91 0 R +/Next 107 0 R +/First 99 0 R +/Last 103 0 R +/Count -2 +>> endobj +91 0 obj << +/Title 92 0 R +/A 90 0 R +/Parent 4570 0 R +/Prev 67 0 R +/Next 135 0 R +/First 95 0 R +/Last 115 0 R +/Count -4 +>> endobj +87 0 obj << +/Title 88 0 R +/A 86 0 R +/Parent 67 0 R +/Prev 83 0 R +>> endobj +83 0 obj << +/Title 84 0 R +/A 82 0 R +/Parent 67 0 R +/Prev 79 0 R +/Next 87 0 R +>> endobj +79 0 obj << +/Title 80 0 R +/A 78 0 R +/Parent 67 0 R +/Prev 75 0 R +/Next 83 0 R +>> endobj +75 0 obj << +/Title 76 0 R +/A 74 0 R +/Parent 67 0 R +/Prev 71 0 R +/Next 79 0 R +>> endobj +71 0 obj << +/Title 72 0 R +/A 70 0 R +/Parent 67 0 R +/Next 75 0 R +>> endobj +67 0 obj << +/Title 68 0 R +/A 66 0 R +/Parent 4570 0 R +/Prev 7 0 R +/Next 91 0 R +/First 71 0 R +/Last 87 0 R +/Count -5 +>> endobj +63 0 obj << +/Title 64 0 R +/A 62 0 R +/Parent 23 0 R +/Prev 55 0 R +>> endobj +59 0 obj << +/Title 60 0 R +/A 58 0 R +/Parent 55 0 R +>> endobj +55 0 obj << +/Title 56 0 R +/A 54 0 R +/Parent 23 0 R +/Prev 39 0 R +/Next 63 0 R +/First 59 0 R +/Last 59 0 R +/Count -1 +>> endobj +51 0 obj << +/Title 52 0 R +/A 50 0 R +/Parent 39 0 R +/Prev 47 0 R +>> endobj +47 0 obj << +/Title 48 0 R +/A 46 0 R +/Parent 39 0 R +/Prev 43 0 R +/Next 51 0 R +>> endobj +43 0 obj << +/Title 44 0 R +/A 42 0 R +/Parent 39 0 R +/Next 47 0 R +>> endobj +39 0 obj << +/Title 40 0 R +/A 38 0 R +/Parent 23 0 R +/Prev 35 0 R +/Next 55 0 R +/First 43 0 R +/Last 51 0 R +/Count -3 +>> endobj +35 0 obj << +/Title 36 0 R +/A 34 0 R +/Parent 23 0 R +/Prev 31 0 R +/Next 39 0 R +>> endobj +31 0 obj << +/Title 32 0 R +/A 30 0 R +/Parent 23 0 R +/Prev 27 0 R +/Next 35 0 R +>> endobj +27 0 obj << +/Title 28 0 R +/A 26 0 R +/Parent 23 0 R +/Next 31 0 R +>> endobj +23 0 obj << +/Title 24 0 R +/A 22 0 R +/Parent 7 0 R +/Prev 19 0 R +/First 27 0 R +/Last 63 0 R +/Count -6 +>> endobj +19 0 obj << +/Title 20 0 R +/A 18 0 R +/Parent 7 0 R +/Prev 15 0 R +/Next 23 0 R +>> endobj +15 0 obj << +/Title 16 0 R +/A 14 0 R +/Parent 7 0 R +/Prev 11 0 R +/Next 19 0 R +>> endobj +11 0 obj << +/Title 12 0 R +/A 10 0 R +/Parent 7 0 R +/Next 15 0 R +>> endobj +7 0 obj << +/Title 8 0 R +/A 6 0 R +/Parent 4570 0 R +/Prev 3 0 R +/Next 67 0 R +/First 11 0 R +/Last 23 0 R +/Count -4 +>> endobj +3 0 obj << +/Title 4 0 R +/A 2 0 R +/Parent 4570 0 R +/Next 7 0 R +>> endobj +4571 0 obj << +/Names [(0) 664 0 R (1.0) 1 0 R (10) 1176 0 R (10.0) 565 0 R (10.33.1) 569 0 R (10.33.61.2) 573 0 R (10.34.1) 577 0 R (10.34.62.2) 581 0 R (10.34.62.41.3) 585 0 R (10.34.62.42.3) 589 0 R (10.35.1) 593 0 R (10.35.63.2) 597 0 R (10.36.1) 601 0 R (10.36.64.2) 605 0 R (10.36.65.2) 609 0 R (10.37.1) 613 0 R (10.38.1) 617 0 R (10.39.1) 621 0 R (10.40.1) 625 0 R (10.41.1) 629 0 R (10.42.1) 633 0 R (10.43.1) 637 0 R (10.44.1) 641 0 R (10.45.1) 645 0 R (10.45.66.2) 649 0 R (10.45.67.2) 653 0 R (10.45.68.2) 657 0 R (1000) 823 0 R (1002) 2038 0 R (1003) 2039 0 R (1004) 2040 0 R (1005) 824 0 R (1007) 2048 0 R (1008) 2049 0 R (1009) 825 0 R (1011) 2050 0 R (1012) 2051 0 R (1013) 2052 0 R (1014) 2053 0 R (1015) 2054 0 R (1016) 2055 0 R (1017) 2056 0 R (1018) 2057 0 R (1019) 2058 0 R (1020) 826 0 R (1022) 2059 0 R (1023) 2060 0 R (1024) 2067 0 R (1025) 2068 0 R (1026) 827 0 R (1028) 2069 0 R (1029) 2070 0 R (1030) 2071 0 R (1031) 2072 0 R (1032) 2073 0 R (1033) 828 0 R (1035) 2074 0 R (1036) 2075 0 R (1037) 2076 0 R (1038) 2077 0 R (1039) 2078 0 R (104) 691 0 R (1040) 829 0 R (1042) 2079 0 R (1043) 2080 0 R (1044) 2081 0 R (1045) 2082 0 R (1046) 2083 0 R (1047) 2084 0 R (1048) 2090 0 R (1049) 2091 0 R (1050) 2092 0 R (1051) 2093 0 R (1052) 2094 0 R (1053) 2095 0 R (1054) 2096 0 R (1055) 2097 0 R (1056) 2098 0 R (1060) 831 0 R (1062) 2109 0 R (1063) 2110 0 R (1064) 2111 0 R (1065) 2112 0 R (1068) 2113 0 R (1069) 2114 0 R (107) 1216 0 R (1070) 2115 0 R (1071) 2116 0 R (1072) 2117 0 R (1073) 2118 0 R (1074) 2119 0 R (1075) 2120 0 R (1076) 2121 0 R (1077) 2122 0 R (1078) 2123 0 R (1079) 2124 0 R (108) 1217 0 R (1080) 2125 0 R (1081) 2126 0 R (1082) 2127 0 R (1083) 2128 0 R (1084) 2129 0 R (1085) 2130 0 R (1089) 2141 0 R (109) 1218 0 R (1090) 2142 0 R (1091) 2143 0 R (1092) 2144 0 R (1093) 2145 0 R (1094) 2146 0 R (1095) 2147 0 R (1096) 2148 0 R (1097) 2149 0 R (11) 1177 0 R (110) 1219 0 R (1100) 2150 0 R (1101) 2151 0 R (1102) 2152 0 R (1103) 2153 0 R (111) 692 0 R (113) 1220 0 R (114) 1221 0 R (115) 1222 0 R (116) 1223 0 R (117) 1224 0 R (118) 693 0 R (12) 688 0 R (120) 1225 0 R (121) 1226 0 R (122) 1227 0 R (123) 1228 0 R (124) 1229 0 R (125) 1230 0 R (126) 1231 0 R (1264) 836 0 R (1266) 2162 0 R (1267) 2163 0 R (1268) 2169 0 R (1269) 2170 0 R (127) 1232 0 R (1270) 2171 0 R (1271) 2172 0 R (1272) 837 0 R (1274) 2173 0 R (1275) 2174 0 R (1276) 2175 0 R (1277) 2176 0 R (1278) 2177 0 R (1279) 2178 0 R (128) 1233 0 R (1280) 2179 0 R (1281) 2180 0 R (1282) 2181 0 R (1283) 2182 0 R (1284) 2183 0 R (1285) 2184 0 R (1286) 2185 0 R (1287) 2186 0 R (1288) 2187 0 R (1289) 2188 0 R (129) 1234 0 R (1290) 2189 0 R (1291) 2190 0 R (1292) 2191 0 R (1293) 2192 0 R (1294) 2193 0 R (1295) 2194 0 R (1296) 2195 0 R (1297) 2196 0 R (1298) 2202 0 R (1299) 2203 0 R (130) 1235 0 R (1300) 2204 0 R (1301) 838 0 R (1303) 2205 0 R (1304) 2206 0 R (1305) 2207 0 R (1306) 839 0 R (1308) 2208 0 R (1309) 2209 0 R (131) 1236 0 R (1310) 2210 0 R (1311) 2211 0 R (1312) 2212 0 R (1313) 2213 0 R (1314) 2214 0 R (1315) 840 0 R (1317) 2215 0 R (1318) 2216 0 R (1319) 2217 0 R (132) 1243 0 R (1320) 2218 0 R (1321) 2219 0 R (1322) 2220 0 R (1323) 2222 0 R (1324) 2227 0 R (1325) 2228 0 R (1326) 2229 0 R (1327) 2230 0 R (1328) 2231 0 R (1329) 2232 0 R (133) 1244 0 R (1330) 2233 0 R (1331) 2234 0 R (1332) 2235 0 R (1333) 2236 0 R (1336) 2237 0 R (1337) 2238 0 R (1338) 2239 0 R (1339) 2240 0 R (134) 1245 0 R (1340) 2241 0 R (135) 1246 0 R (136) 1247 0 R (137) 1248 0 R (138) 1249 0 R (14) 1178 0 R (140) 1250 0 R (1405) 2249 0 R (1406) 2250 0 R (1407) 2251 0 R (1408) 842 0 R (1411) 2252 0 R (1412) 2253 0 R (1416) 2254 0 R (1417) 2255 0 R (1418) 2256 0 R (1419) 2257 0 R (142) 694 0 R (1420) 2258 0 R (1421) 2259 0 R (1422) 2260 0 R (144) 1251 0 R (145) 1252 0 R (1450) 2267 0 R (1451) 2268 0 R (1452) 2269 0 R (1453) 2270 0 R (1454) 2271 0 R (1455) 844 0 R (1458) 2272 0 R (1459) 2273 0 R (146) 1253 0 R (1460) 2274 0 R (1461) 2275 0 R (1462) 2276 0 R (1463) 2277 0 R (1467) 2278 0 R (1468) 2279 0 R (1469) 2280 0 R (147) 1254 0 R (1470) 2281 0 R (1471) 2282 0 R (1472) 2283 0 R (1473) 2284 0 R (1474) 2285 0 R (1475) 2286 0 R (1476) 2287 0 R (1477) 2288 0 R (1478) 2289 0 R (1479) 2290 0 R (148) 1255 0 R (1480) 2291 0 R (1481) 2292 0 R (1482) 2293 0 R (1483) 2294 0 R (1484) 2295 0 R (1485) 2296 0 R (1486) 2297 0 R (1487) 2298 0 R (1488) 2299 0 R (1489) 2300 0 R (149) 1256 0 R (1492) 2301 0 R (1493) 2302 0 R (1494) 2303 0 R (1495) 2304 0 R (1496) 2305 0 R (1497) 2306 0 R (1498) 2307 0 R (1499) 2308 0 R (15) 1179 0 R (150) 1257 0 R (1500) 2309 0 R (1501) 2310 0 R (1502) 2311 0 R (1503) 2312 0 R (1504) 2314 0 R (1506) 2323 0 R (1507) 2324 0 R (1508) 2325 0 R (1509) 2326 0 R (151) 1258 0 R (1510) 2327 0 R (1511) 2328 0 R (1512) 2329 0 R (1513) 2330 0 R (1514) 2331 0 R (1515) 2332 0 R (1516) 2333 0 R (1517) 2334 0 R (1518) 2335 0 R (1519) 2336 0 R (152) 1259 0 R (1520) 2337 0 R (1521) 2338 0 R (1522) 2339 0 R (1523) 2340 0 R (1524) 2341 0 R (1525) 2342 0 R (1526) 2343 0 R (1527) 2344 0 R (1528) 931 0 R (153) 1260 0 R (1531) 2345 0 R (1532) 2346 0 R (1533) 932 0 R (1536) 2347 0 R (1537) 2348 0 R (1538) 2349 0 R (1539) 2350 0 R (154) 1261 0 R (1540) 933 0 R (1543) 2351 0 R (1544) 2352 0 R (1545) 2353 0 R (1546) 2354 0 R (1547) 934 0 R (155) 1262 0 R (1550) 2360 0 R (1551) 2361 0 R (1553) 2362 0 R (1554) 2363 0 R (1555) 2364 0 R (1556) 2365 0 R (1557) 2366 0 R (1559) 2367 0 R (156) 1263 0 R (1560) 2368 0 R (1561) 2369 0 R (1562) 2370 0 R (1563) 2371 0 R (1564) 2372 0 R (1565) 935 0 R (1568) 2373 0 R (1569) 2374 0 R (157) 1264 0 R (1570) 2375 0 R (1571) 2376 0 R (1572) 2377 0 R (1573) 2378 0 R (1574) 2379 0 R (1575) 2380 0 R (1576) 2381 0 R (1577) 2382 0 R (1578) 2383 0 R (1579) 2384 0 R (158) 1265 0 R (1580) 2385 0 R (1581) 2386 0 R (1582) 2387 0 R (1583) 2388 0 R (1584) 2389 0 R (1585) 2390 0 R (1586) 2391 0 R (1587) 2392 0 R (1588) 2393 0 R (1589) 2394 0 R (159) 1266 0 R (1590) 2395 0 R (1591) 2396 0 R (1592) 2397 0 R (1593) 2398 0 R (1594) 2399 0 R (1595) 2400 0 R (1596) 2401 0 R (1597) 2402 0 R (1598) 2403 0 R (1599) 2404 0 R (16) 1180 0 R (160) 1267 0 R (1600) 2405 0 R (1601) 2406 0 R (1602) 2407 0 R (1603) 2408 0 R (1604) 2409 0 R (1605) 936 0 R (1608) 2410 0 R (1609) 2411 0 R (161) 1268 0 R (1610) 2412 0 R (1611) 2413 0 R (1612) 2421 0 R (1613) 2422 0 R (1614) 2423 0 R (1615) 2424 0 R (1616) 2425 0 R (1617) 2426 0 R (1618) 2427 0 R (1619) 2428 0 R (162) 1269 0 R (1620) 2429 0 R (1621) 937 0 R (1624) 2430 0 R (1625) 2431 0 R (1626) 2432 0 R (1627) 2433 0 R (1628) 2434 0 R (1629) 2435 0 R (163) 1270 0 R (1630) 2436 0 R (1631) 2437 0 R (1632) 2438 0 R (1633) 2439 0 R (1634) 2440 0 R (1635) 2441 0 R (1636) 2442 0 R (1637) 2443 0 R (1638) 2444 0 R (1639) 2445 0 R (164) 1271 0 R (1640) 2446 0 R (1641) 2447 0 R (1642) 2448 0 R (1643) 2449 0 R (1644) 2450 0 R (1645) 2451 0 R (1646) 2452 0 R (1647) 2453 0 R (1648) 2454 0 R (1649) 2455 0 R (165) 695 0 R (1650) 2461 0 R (1651) 2462 0 R (1652) 2463 0 R (1653) 2464 0 R (1654) 2465 0 R (1655) 2466 0 R (1656) 2467 0 R (1657) 2468 0 R (1658) 2469 0 R (1659) 2470 0 R (1660) 2471 0 R (1661) 2472 0 R (1662) 2473 0 R (1663) 2474 0 R (1664) 2475 0 R (1665) 2476 0 R (1666) 2477 0 R (1667) 2478 0 R (1668) 2479 0 R (1669) 2480 0 R (167) 1272 0 R (1670) 2481 0 R (1671) 2482 0 R (1672) 2483 0 R (1673) 2484 0 R (1674) 2485 0 R (1675) 2486 0 R (1676) 2487 0 R (1677) 2488 0 R (1678) 2489 0 R (1679) 2490 0 R (168) 1273 0 R (1680) 2491 0 R (1681) 2492 0 R (1682) 2493 0 R (1683) 2494 0 R (1684) 2495 0 R (1685) 2496 0 R (1686) 2497 0 R (1687) 2498 0 R (1688) 2499 0 R (1689) 2500 0 R (169) 1281 0 R (1690) 2501 0 R (1691) 2502 0 R (1692) 2503 0 R (1693) 2504 0 R (1694) 2505 0 R (1695) 2506 0 R (1696) 2512 0 R (1697) 2513 0 R (1698) 2514 0 R (1699) 2515 0 R (17) 1181 0 R (170) 1282 0 R (1701) 2516 0 R (1702) 2517 0 R (1703) 2518 0 R (1704) 2519 0 R (1705) 2520 0 R (1706) 2521 0 R (1707) 2522 0 R (1708) 2523 0 R (1709) 2524 0 R (1710) 2525 0 R (1711) 2532 0 R (1715) 2533 0 R (1716) 2534 0 R (1717) 2535 0 R (1718) 2536 0 R (1719) 2537 0 R (172) 696 0 R (1720) 2538 0 R (1721) 2539 0 R (1722) 2540 0 R (1723) 2541 0 R (1724) 2542 0 R (1725) 2543 0 R (1726) 2544 0 R (174) 1283 0 R (175) 1284 0 R (176) 1285 0 R (177) 1286 0 R (178) 1287 0 R (179) 697 0 R (18) 1182 0 R (181) 1288 0 R (182) 1289 0 R (183) 1290 0 R (1836) 939 0 R (1839) 2552 0 R (184) 1291 0 R (1840) 2553 0 R (1841) 2554 0 R (1842) 2555 0 R (1843) 2556 0 R (1844) 2557 0 R (1845) 2558 0 R (1846) 2559 0 R (1847) 2560 0 R (1848) 2561 0 R (1849) 2562 0 R (185) 698 0 R (1850) 2563 0 R (1851) 2564 0 R (1852) 2570 0 R (1853) 2571 0 R (1854) 2572 0 R (1855) 2573 0 R (1856) 2574 0 R (1857) 2575 0 R (1858) 2576 0 R (1859) 2577 0 R (1860) 940 0 R (1863) 2578 0 R (1864) 2579 0 R (1866) 2580 0 R (1867) 2581 0 R (1868) 2582 0 R (1869) 2583 0 R (187) 1292 0 R (1870) 2584 0 R (1871) 2585 0 R (1872) 2586 0 R (1873) 2587 0 R (1874) 2588 0 R (1875) 2589 0 R (1876) 2590 0 R (1877) 2591 0 R (1878) 2592 0 R (1879) 941 0 R (188) 1293 0 R (1882) 2593 0 R (1883) 2594 0 R (1884) 2595 0 R (1885) 2596 0 R (1886) 2597 0 R (1887) 2598 0 R (1888) 2599 0 R (1889) 2600 0 R (189) 1294 0 R (1890) 2601 0 R (1891) 2602 0 R (1892) 2603 0 R (1893) 2604 0 R (1894) 2605 0 R (1895) 2606 0 R (1896) 2607 0 R (1897) 2608 0 R (1898) 2609 0 R (1899) 2610 0 R (19) 689 0 R (190) 1295 0 R (1900) 2611 0 R (1901) 2612 0 R (1902) 2613 0 R (1903) 2614 0 R (1904) 2615 0 R (1905) 2623 0 R (1906) 2624 0 R (1907) 2625 0 R (1908) 2626 0 R (1909) 2627 0 R (191) 1296 0 R (1910) 2628 0 R (1911) 2629 0 R (1912) 2630 0 R (1913) 2631 0 R (1914) 2632 0 R (1915) 2633 0 R (1916) 2634 0 R (1917) 2635 0 R (1918) 2636 0 R (1919) 2637 0 R (192) 1297 0 R (1920) 2638 0 R (1921) 2639 0 R (1922) 2640 0 R (1923) 2641 0 R (1924) 2642 0 R (1925) 2643 0 R (1926) 2644 0 R (1927) 2645 0 R (1928) 2646 0 R (1929) 2647 0 R (193) 1298 0 R (1930) 2648 0 R (1931) 2649 0 R (1932) 2650 0 R (1933) 2651 0 R (1934) 2652 0 R (1935) 2653 0 R (1936) 2654 0 R (1937) 2655 0 R (1938) 2656 0 R (1939) 2657 0 R (194) 699 0 R (1940) 2658 0 R (1941) 2659 0 R (1942) 2660 0 R (1943) 2661 0 R (1944) 2662 0 R (1945) 2663 0 R (1946) 2664 0 R (1947) 2665 0 R (1948) 2666 0 R (1949) 2667 0 R (1950) 2668 0 R (1951) 2669 0 R (1952) 2670 0 R (1953) 2671 0 R (1954) 2672 0 R (1955) 2673 0 R (1956) 2674 0 R (1957) 2675 0 R (1958) 2676 0 R (1959) 2677 0 R (196) 1299 0 R (1960) 2678 0 R (1961) 2679 0 R (1962) 2680 0 R (1963) 2681 0 R (1964) 2682 0 R (1965) 2683 0 R (1966) 2684 0 R (1967) 2685 0 R (1968) 2686 0 R (1969) 2687 0 R (197) 1300 0 R (1970) 2688 0 R (1971) 2689 0 R (1972) 2690 0 R (1973) 2691 0 R (1974) 2692 0 R (1975) 2693 0 R (1976) 2694 0 R (1977) 2695 0 R (1978) 2696 0 R (1979) 2697 0 R (198) 1301 0 R (1980) 2698 0 R (1981) 2699 0 R (1982) 2700 0 R (1983) 2701 0 R (1984) 2702 0 R (1985) 2703 0 R (1986) 2704 0 R (1987) 2705 0 R (1988) 2706 0 R (1989) 2707 0 R (199) 1302 0 R (1990) 2708 0 R (1991) 2709 0 R (1992) 2710 0 R (1993) 2711 0 R (1994) 2712 0 R (1995) 2713 0 R (1996) 2714 0 R (1997) 2715 0 R (1998) 2716 0 R (1999) 2717 0 R (2.0) 5 0 R (2.1.1) 9 0 R (2.2.1) 13 0 R (2.3.1) 17 0 R (2.4.1) 21 0 R (2.4.1.2) 25 0 R (2.4.2.2) 29 0 R (2.4.3.2) 33 0 R (2.4.4.1.3) 41 0 R (2.4.4.2) 37 0 R (2.4.4.2.3) 45 0 R (2.4.4.3.3) 49 0 R (2.4.5.2) 53 0 R (2.4.5.4.3) 57 0 R (2.4.6.2) 61 0 R (200) 1309 0 R (2000) 2718 0 R (2001) 2719 0 R (2002) 2720 0 R (2003) 2721 0 R (2004) 2722 0 R (2005) 2723 0 R (2006) 2724 0 R (2007) 2725 0 R (2008) 2726 0 R (2009) 2727 0 R (201) 1310 0 R (2010) 2728 0 R (2011) 2729 0 R (2012) 2730 0 R (2013) 2731 0 R (2014) 2732 0 R (2015) 2733 0 R (2016) 2734 0 R (2017) 2735 0 R (2018) 2736 0 R (2019) 2737 0 R (202) 1311 0 R (2020) 2738 0 R (2021) 2739 0 R (2022) 2740 0 R (2023) 2741 0 R (2024) 2742 0 R (2025) 2743 0 R (2026) 2744 0 R (2027) 2745 0 R (2028) 2746 0 R (2029) 2747 0 R (203) 1312 0 R (2030) 2748 0 R (2031) 2749 0 R (2032) 2750 0 R (2033) 2756 0 R (2034) 2757 0 R (2035) 2758 0 R (2036) 2759 0 R (2037) 2760 0 R (2038) 2761 0 R (2039) 2762 0 R (204) 700 0 R (2040) 2763 0 R (2041) 2764 0 R (2042) 2765 0 R (2043) 2766 0 R (2044) 2767 0 R (2045) 2768 0 R (2046) 2769 0 R (2047) 2770 0 R (2048) 2771 0 R (2049) 2772 0 R (2050) 2773 0 R (2051) 2774 0 R (2052) 2775 0 R (2053) 2776 0 R (2054) 2777 0 R (2055) 2778 0 R (2056) 2779 0 R (2057) 2780 0 R (2058) 2781 0 R (2059) 2782 0 R (206) 1313 0 R (2060) 2783 0 R (2061) 2784 0 R (2062) 2785 0 R (2063) 2786 0 R (2064) 2787 0 R (2065) 2788 0 R (2066) 2789 0 R (2067) 2790 0 R (2068) 2791 0 R (2069) 2792 0 R (207) 1314 0 R (2070) 2793 0 R (2071) 2794 0 R (2072) 2795 0 R (2073) 2796 0 R (2074) 2797 0 R (2075) 2798 0 R (2076) 2799 0 R (2077) 2800 0 R (2078) 2801 0 R (2079) 2802 0 R (208) 1315 0 R (2080) 2803 0 R (2081) 2804 0 R (2082) 2805 0 R (2083) 2806 0 R (2084) 2807 0 R (2085) 2808 0 R (2086) 2809 0 R (2087) 2810 0 R (2088) 2811 0 R (2089) 2812 0 R (209) 1316 0 R (2090) 2813 0 R (2091) 2814 0 R (2092) 2815 0 R (2093) 2816 0 R (2094) 2817 0 R (2095) 2818 0 R (2096) 2819 0 R (2097) 942 0 R (21) 1183 0 R (210) 1317 0 R (2100) 2820 0 R (2101) 2821 0 R (2102) 2822 0 R (2103) 2823 0 R (2104) 2824 0 R (2106) 2825 0 R (2107) 2826 0 R (2108) 2827 0 R (2109) 2828 0 R (211) 1318 0 R (2110) 2829 0 R (2111) 2830 0 R (2113) 2836 0 R (2114) 2837 0 R (2115) 2838 0 R (2116) 2839 0 R (2117) 2840 0 R (2118) 2841 0 R (212) 701 0 R (2120) 2842 0 R (2121) 2843 0 R (2122) 2844 0 R (2123) 2845 0 R (2124) 2846 0 R (2125) 2847 0 R (2126) 2848 0 R (2127) 2849 0 R (2128) 2850 0 R (2130) 2851 0 R (2131) 2852 0 R (2132) 2853 0 R (2133) 2854 0 R (2134) 2855 0 R (2135) 2856 0 R (2136) 2857 0 R (2137) 2858 0 R (2138) 2859 0 R (2139) 2860 0 R (214) 1319 0 R (2140) 2861 0 R (2142) 2862 0 R (2143) 2863 0 R (2144) 2864 0 R (2145) 2865 0 R (2146) 2866 0 R (2148) 2867 0 R (2149) 2868 0 R (215) 1320 0 R (2150) 2869 0 R (2151) 2870 0 R (2152) 2871 0 R (2153) 2872 0 R (2155) 2873 0 R (2156) 2874 0 R (2157) 2875 0 R (2158) 2876 0 R (2159) 2877 0 R (216) 1321 0 R (2160) 2878 0 R (2161) 2879 0 R (2162) 2880 0 R (2164) 2881 0 R (2165) 2882 0 R (2166) 2883 0 R (2167) 2884 0 R (2168) 2885 0 R (217) 1322 0 R (2170) 2891 0 R (2171) 2892 0 R (2172) 2893 0 R (2173) 2894 0 R (2174) 2895 0 R (2175) 2896 0 R (2178) 2897 0 R (2179) 2898 0 R (218) 1323 0 R (2180) 2899 0 R (2181) 2900 0 R (2183) 2901 0 R (2184) 2902 0 R (2185) 2903 0 R (2186) 2904 0 R (2187) 2905 0 R (2188) 2906 0 R (2191) 2907 0 R (2193) 2908 0 R (2194) 2909 0 R (2195) 2910 0 R (2196) 2911 0 R (2197) 2912 0 R (2198) 2913 0 R (2199) 2914 0 R (22) 1184 0 R (2200) 2915 0 R (2201) 2916 0 R (2203) 2917 0 R (2204) 2918 0 R (2205) 2919 0 R (2206) 2920 0 R (2207) 2921 0 R (2208) 2922 0 R (2210) 2923 0 R (2211) 2924 0 R (2212) 2925 0 R (2213) 2926 0 R (2214) 2927 0 R (2215) 2928 0 R (2216) 2929 0 R (2217) 2937 0 R (2218) 2938 0 R (2219) 2939 0 R (222) 703 0 R (2220) 2940 0 R (2221) 2941 0 R (2222) 2942 0 R (2223) 2943 0 R (2224) 2944 0 R (2225) 2945 0 R (2226) 2946 0 R (2227) 2947 0 R (2228) 2948 0 R (2229) 2949 0 R (2230) 2950 0 R (2232) 2951 0 R (2233) 2952 0 R (2234) 2953 0 R (2235) 2954 0 R (2236) 2955 0 R (2237) 2956 0 R (2239) 2957 0 R (224) 1335 0 R (2240) 2958 0 R (2241) 2959 0 R (2242) 2960 0 R (2243) 2961 0 R (2245) 2962 0 R (2246) 2963 0 R (2247) 2964 0 R (2248) 2965 0 R (2249) 2966 0 R (225) 1336 0 R (2250) 2967 0 R (2251) 2968 0 R (2252) 2969 0 R (2253) 2970 0 R (2254) 2971 0 R (2255) 2972 0 R (2256) 2973 0 R (2258) 2974 0 R (2259) 2975 0 R (226) 1337 0 R (2260) 2976 0 R (2261) 2977 0 R (2263) 2978 0 R (2264) 2979 0 R (2265) 2980 0 R (2266) 2981 0 R (2267) 2982 0 R (2268) 2983 0 R (2269) 2984 0 R (227) 1338 0 R (2270) 2985 0 R (2271) 2986 0 R (2273) 2987 0 R (2274) 2988 0 R (2275) 2989 0 R (2276) 2990 0 R (2277) 2991 0 R (2278) 2992 0 R (228) 1339 0 R (2280) 2998 0 R (2281) 2999 0 R (2282) 3000 0 R (2283) 3001 0 R (2284) 3002 0 R (2285) 3003 0 R (2287) 3004 0 R (2288) 3005 0 R (2289) 3006 0 R (229) 1340 0 R (2290) 3007 0 R (2291) 3008 0 R (2293) 3009 0 R (2294) 3010 0 R (2295) 3011 0 R (2296) 3012 0 R (2297) 3013 0 R (2298) 3014 0 R (2299) 3015 0 R (23) 1185 0 R (230) 704 0 R (2300) 3016 0 R (2301) 3017 0 R (2303) 3018 0 R (2304) 3019 0 R (2305) 3020 0 R (2306) 3021 0 R (2307) 3022 0 R (2308) 3023 0 R (2309) 3024 0 R (2310) 3025 0 R (2312) 3026 0 R (2313) 3027 0 R (2314) 3028 0 R (2315) 3029 0 R (2316) 3030 0 R (2317) 3031 0 R (2318) 3032 0 R (2319) 3033 0 R (232) 1341 0 R (2320) 3034 0 R (2322) 3035 0 R (2323) 3036 0 R (2324) 3037 0 R (2325) 3038 0 R (2326) 3039 0 R (2327) 3040 0 R (2329) 3042 0 R (233) 1342 0 R (2331) 3048 0 R (2332) 3049 0 R (2333) 3050 0 R (2334) 3051 0 R (2335) 3052 0 R (2338) 3053 0 R (2339) 3054 0 R (234) 705 0 R (2340) 3055 0 R (2341) 3056 0 R (2342) 3057 0 R (2343) 3058 0 R (2347) 3059 0 R (2348) 3060 0 R (2349) 3061 0 R (2350) 3062 0 R (2351) 3063 0 R (2354) 3064 0 R (2355) 3065 0 R (2356) 3066 0 R (2357) 3067 0 R (2358) 3068 0 R (236) 1343 0 R (2361) 3069 0 R (2362) 3070 0 R (2363) 3071 0 R (2364) 3072 0 R (2365) 3073 0 R (2366) 3074 0 R (2367) 3075 0 R (2368) 3076 0 R (2369) 3077 0 R (237) 1344 0 R (2371) 3078 0 R (2372) 3079 0 R (2373) 3080 0 R (2374) 3081 0 R (2375) 3082 0 R (2376) 3083 0 R (2377) 3084 0 R (2378) 3085 0 R (2379) 3086 0 R (238) 1345 0 R (2380) 3087 0 R (2381) 3088 0 R (2382) 3089 0 R (2383) 3090 0 R (2384) 3092 0 R (2386) 3102 0 R (2387) 3103 0 R (2388) 3104 0 R (2389) 3105 0 R (239) 706 0 R (2390) 3106 0 R (2392) 3107 0 R (2393) 3108 0 R (2394) 3109 0 R (2395) 3110 0 R (2396) 3111 0 R (2397) 944 0 R (2399) 3112 0 R (24) 1186 0 R (2400) 3113 0 R (2402) 3114 0 R (2403) 3115 0 R (2404) 3116 0 R (2405) 3117 0 R (2406) 3118 0 R (2407) 3119 0 R (2409) 3120 0 R (241) 1346 0 R (2410) 3121 0 R (2411) 3122 0 R (2412) 3123 0 R (2413) 3124 0 R (2414) 3125 0 R (2418) 3131 0 R (242) 707 0 R (2420) 3132 0 R (2422) 3133 0 R (2423) 3134 0 R (2424) 3135 0 R (2425) 3136 0 R (2426) 3137 0 R (2427) 3138 0 R (2428) 3139 0 R (2430) 3140 0 R (2431) 3141 0 R (2432) 3142 0 R (2433) 3143 0 R (2434) 3144 0 R (2435) 3145 0 R (2436) 3146 0 R (2438) 3147 0 R (2439) 3148 0 R (244) 1352 0 R (2440) 3149 0 R (2441) 3150 0 R (2443) 3151 0 R (2444) 3152 0 R (2445) 3153 0 R (2446) 3154 0 R (2449) 3155 0 R (245) 1353 0 R (2450) 3156 0 R (2451) 3157 0 R (2452) 3158 0 R (2453) 3159 0 R (2454) 3160 0 R (2455) 3161 0 R (2457) 3162 0 R (2458) 3163 0 R (2459) 3164 0 R (246) 1354 0 R (2460) 3165 0 R (2461) 3166 0 R (2462) 946 0 R (2464) 3167 0 R (2465) 3168 0 R (2466) 3169 0 R (2467) 3170 0 R (2468) 3180 0 R (2469) 3181 0 R (247) 1355 0 R (2470) 3182 0 R (2471) 3183 0 R (2472) 3184 0 R (2473) 3185 0 R (2474) 3186 0 R (2475) 3187 0 R (2476) 3188 0 R (2477) 3189 0 R (2478) 3190 0 R (2479) 3191 0 R (248) 1356 0 R (2480) 3192 0 R (2481) 3193 0 R (2482) 3194 0 R (2483) 3195 0 R (2484) 3196 0 R (2485) 3197 0 R (2486) 3198 0 R (2487) 3199 0 R (2488) 947 0 R (249) 1357 0 R (2490) 3200 0 R (2491) 3201 0 R (2492) 3202 0 R (2493) 3203 0 R (2494) 3204 0 R (2495) 3205 0 R (2496) 3206 0 R (2497) 3207 0 R (2498) 3214 0 R (2499) 3215 0 R (25) 1187 0 R (250) 1358 0 R (2500) 3216 0 R (2501) 3217 0 R (2504) 3218 0 R (2505) 3219 0 R (2506) 3220 0 R (2508) 3221 0 R (2509) 3222 0 R (251) 1359 0 R (2510) 3223 0 R (2511) 3224 0 R (2512) 3225 0 R (2513) 3226 0 R (2514) 3227 0 R (2515) 3228 0 R (2516) 3229 0 R (2517) 3230 0 R (2519) 3231 0 R (252) 1360 0 R (2520) 3232 0 R (2521) 3233 0 R (2522) 3234 0 R (2524) 3235 0 R (2525) 3236 0 R (2526) 3237 0 R (2527) 3238 0 R (2529) 3239 0 R (253) 1361 0 R (2530) 3240 0 R (2531) 3241 0 R (2532) 3242 0 R (2534) 3243 0 R (2535) 3244 0 R (2536) 3245 0 R (2537) 3246 0 R (2539) 3247 0 R (254) 1362 0 R (2540) 3248 0 R (2541) 3249 0 R (2542) 3250 0 R (2543) 3251 0 R (2545) 3257 0 R (2546) 3258 0 R (2547) 3259 0 R (2548) 3260 0 R (2549) 3261 0 R (255) 1363 0 R (2550) 3262 0 R (2551) 3263 0 R (2553) 3264 0 R (2554) 3265 0 R (2555) 3266 0 R (2556) 3267 0 R (2557) 3268 0 R (2558) 3269 0 R (2559) 3270 0 R (256) 1364 0 R (2560) 3271 0 R (2561) 3272 0 R (2562) 3273 0 R (2563) 3274 0 R (2564) 3275 0 R (2565) 3276 0 R (2566) 3277 0 R (2567) 3278 0 R (2568) 3279 0 R (257) 1365 0 R (2570) 3280 0 R (2571) 3281 0 R (2572) 3282 0 R (2573) 3283 0 R (2574) 3284 0 R (2575) 3285 0 R (2577) 3286 0 R (2578) 3287 0 R (2579) 3288 0 R (258) 1366 0 R (2580) 3289 0 R (2581) 3290 0 R (2583) 3291 0 R (2584) 3292 0 R (2585) 3293 0 R (2586) 3294 0 R (2587) 3295 0 R (2588) 3296 0 R (2589) 3297 0 R (259) 1367 0 R (2590) 3298 0 R (2591) 3299 0 R (2593) 3300 0 R (2594) 3301 0 R (2595) 3302 0 R (2596) 3303 0 R (2597) 3304 0 R (2598) 3305 0 R (2599) 3306 0 R (26) 1188 0 R (260) 1368 0 R (2600) 3307 0 R (2601) 3308 0 R (2602) 3309 0 R (2604) 3315 0 R (2605) 3316 0 R (2606) 3317 0 R (2607) 3318 0 R (2608) 3319 0 R (261) 1369 0 R (2610) 3320 0 R (2611) 3321 0 R (2612) 3322 0 R (2613) 3323 0 R (2614) 3324 0 R (2615) 3325 0 R (2616) 3326 0 R (2617) 3327 0 R (2618) 3328 0 R (2619) 3329 0 R (262) 1370 0 R (2620) 3330 0 R (2622) 3331 0 R (2623) 3332 0 R (2624) 3333 0 R (2625) 3334 0 R (2626) 3335 0 R (2627) 949 0 R (2629) 3336 0 R (263) 1371 0 R (2630) 3337 0 R (2631) 3338 0 R (2632) 3339 0 R (2633) 3340 0 R (2634) 3341 0 R (2636) 3342 0 R (2637) 3343 0 R (2639) 3344 0 R (264) 1372 0 R (2640) 3345 0 R (2641) 3346 0 R (2642) 3347 0 R (2643) 3348 0 R (2645) 3349 0 R (2646) 3350 0 R (2647) 3351 0 R (2648) 3352 0 R (2649) 3353 0 R (2650) 3354 0 R (2651) 3355 0 R (2653) 3362 0 R (2654) 3363 0 R (2655) 3364 0 R (2656) 3365 0 R (2657) 3366 0 R (2659) 3367 0 R (2660) 3368 0 R (2661) 3369 0 R (2662) 3370 0 R (2663) 3371 0 R (2664) 950 0 R (2666) 3372 0 R (2667) 3373 0 R (2669) 3374 0 R (267) 1378 0 R (2670) 3375 0 R (2671) 3376 0 R (2672) 3377 0 R (2674) 3378 0 R (2675) 3379 0 R (2676) 3380 0 R (2677) 3381 0 R (2678) 3382 0 R (2679) 3383 0 R (2681) 3384 0 R (2682) 3385 0 R (2683) 3386 0 R (2684) 3387 0 R (2685) 3388 0 R (2687) 3389 0 R (2688) 3390 0 R (2689) 3391 0 R (2690) 3392 0 R (2691) 3393 0 R (2692) 951 0 R (2694) 3399 0 R (2696) 3400 0 R (2697) 3401 0 R (2698) 3402 0 R (2699) 3403 0 R (27) 1189 0 R (270) 710 0 R (2700) 3404 0 R (2702) 3405 0 R (2703) 3406 0 R (2704) 3407 0 R (2705) 3408 0 R (2706) 3409 0 R (2708) 3410 0 R (2709) 3411 0 R (2710) 3412 0 R (2711) 3413 0 R (2712) 3414 0 R (2713) 3415 0 R (2715) 3416 0 R (2716) 3417 0 R (2717) 3418 0 R (2718) 3419 0 R (2719) 3420 0 R (272) 1379 0 R (2720) 3421 0 R (2721) 3422 0 R (2722) 3423 0 R (2725) 3424 0 R (2726) 3425 0 R (2727) 3426 0 R (2728) 3428 0 R (2729) 3433 0 R (273) 1380 0 R (2730) 3434 0 R (2731) 3435 0 R (2732) 3436 0 R (2733) 3437 0 R (2734) 3438 0 R (2735) 3439 0 R (2739) 3440 0 R (274) 711 0 R (2740) 3441 0 R (2742) 3442 0 R (2743) 3443 0 R (2744) 3444 0 R (2745) 3445 0 R (2747) 3446 0 R (2748) 3447 0 R (2749) 3448 0 R (2750) 3449 0 R (2751) 3450 0 R (2752) 3451 0 R (2753) 3452 0 R (2754) 3461 0 R (2755) 3462 0 R (2756) 3463 0 R (2757) 3464 0 R (276) 1381 0 R (2760) 3465 0 R (2761) 3466 0 R (2762) 3467 0 R (2764) 3469 0 R (2766) 3474 0 R (2767) 3475 0 R (2768) 3476 0 R (2769) 3477 0 R (277) 1382 0 R (2770) 3478 0 R (2771) 3479 0 R (2772) 3480 0 R (2773) 3481 0 R (2774) 3482 0 R (2775) 3483 0 R (2776) 3484 0 R (2777) 3485 0 R (2778) 3486 0 R (2779) 3487 0 R (278) 1383 0 R (2780) 3488 0 R (2781) 3489 0 R (279) 1384 0 R (28) 1190 0 R (280) 712 0 R (2803) 3491 0 R (2804) 3492 0 R (2805) 3493 0 R (2806) 3494 0 R (2807) 3495 0 R (2808) 3496 0 R (2809) 3497 0 R (2810) 3498 0 R (2811) 3499 0 R (2812) 3500 0 R (2815) 3501 0 R (2816) 3502 0 R (2817) 3503 0 R (2818) 3510 0 R (2819) 3511 0 R (282) 1390 0 R (2820) 3512 0 R (2821) 3513 0 R (2822) 3514 0 R (2823) 3515 0 R (2824) 3516 0 R (2825) 3517 0 R (2826) 3518 0 R (2827) 3519 0 R (2828) 3520 0 R (2829) 3521 0 R (283) 1391 0 R (2830) 3522 0 R (2831) 3523 0 R (2834) 3524 0 R (2836) 3525 0 R (2837) 3526 0 R (2838) 3527 0 R (2839) 3528 0 R (284) 1392 0 R (2840) 3529 0 R (2841) 3530 0 R (2842) 3531 0 R (2844) 3532 0 R (2845) 3533 0 R (2846) 3534 0 R (2847) 3535 0 R (2848) 3536 0 R (2849) 3537 0 R (285) 1393 0 R (2850) 3538 0 R (2851) 3539 0 R (2853) 3540 0 R (2854) 3541 0 R (2855) 3542 0 R (2856) 3543 0 R (2857) 3544 0 R (2859) 3545 0 R (286) 1394 0 R (2860) 3546 0 R (2861) 3547 0 R (2862) 3548 0 R (2863) 3549 0 R (2864) 3550 0 R (2865) 3552 0 R (2868) 3557 0 R (2869) 3558 0 R (2870) 3559 0 R (2871) 3560 0 R (2873) 3561 0 R (2875) 3562 0 R (2876) 3563 0 R (2877) 3564 0 R (2878) 3565 0 R (2879) 3566 0 R (2880) 3567 0 R (2881) 3568 0 R (2882) 3569 0 R (2883) 3570 0 R (2884) 3571 0 R (2885) 3572 0 R (2888) 3573 0 R (2889) 3574 0 R (2890) 3575 0 R (2891) 3576 0 R (2892) 3577 0 R (2893) 3578 0 R (2894) 3579 0 R (2895) 3580 0 R (2896) 3581 0 R (29) 1191 0 R (2939) 3589 0 R (2940) 3590 0 R (2941) 3591 0 R (2942) 3592 0 R (2943) 3593 0 R (2944) 3594 0 R (2945) 3595 0 R (2946) 3596 0 R (2947) 3597 0 R (2948) 3598 0 R (2949) 3599 0 R (2950) 3600 0 R (2951) 3601 0 R (2952) 3602 0 R (2953) 3603 0 R (2954) 3604 0 R (2955) 3605 0 R (2956) 3606 0 R (2960) 3607 0 R (2961) 3608 0 R (2962) 3609 0 R (2963) 3610 0 R (2964) 3611 0 R (2965) 3612 0 R (2966) 3613 0 R (2967) 3614 0 R (2968) 3615 0 R (2969) 3616 0 R (2970) 3617 0 R (2971) 3618 0 R (2972) 3620 0 R (2974) 3625 0 R (2975) 3626 0 R (2976) 3627 0 R (2977) 3628 0 R (2978) 3629 0 R (2979) 3630 0 R (2980) 3631 0 R (2981) 3632 0 R (2982) 3633 0 R (2983) 3634 0 R (2984) 3635 0 R (2985) 3636 0 R (2986) 3637 0 R (2987) 3638 0 R (2988) 3639 0 R (2989) 3640 0 R (2990) 3641 0 R (2991) 3642 0 R (2992) 3643 0 R (2993) 3644 0 R (2994) 3645 0 R (2995) 3646 0 R (2996) 3647 0 R (2997) 3648 0 R (2998) 3649 0 R (2999) 3650 0 R (3.0) 65 0 R (3.5.1) 69 0 R (3.6.1) 73 0 R (3.7.1) 77 0 R (3.8.1) 81 0 R (3.9.1) 85 0 R (30) 1192 0 R (3000) 3651 0 R (3001) 3652 0 R (3002) 3653 0 R (3003) 3654 0 R (3004) 3655 0 R (3005) 3656 0 R (3006) 3657 0 R (3007) 3658 0 R (3008) 3659 0 R (3009) 3660 0 R (3010) 960 0 R (3013) 3661 0 R (3014) 3662 0 R (3015) 3663 0 R (3016) 3664 0 R (3017) 3665 0 R (3018) 3666 0 R (3019) 3673 0 R (3020) 3674 0 R (3021) 3675 0 R (3022) 3676 0 R (3023) 3677 0 R (3024) 3678 0 R (3025) 3679 0 R (3026) 961 0 R (3029) 3680 0 R (3030) 3681 0 R (3032) 3682 0 R (3033) 3683 0 R (3034) 962 0 R (3037) 3684 0 R (3038) 3685 0 R (3039) 3686 0 R (3040) 3687 0 R (3041) 3688 0 R (3042) 3689 0 R (3043) 3690 0 R (3044) 3691 0 R (3045) 3692 0 R (3046) 3693 0 R (3047) 3694 0 R (3048) 3695 0 R (3049) 3696 0 R (3050) 963 0 R (3053) 3697 0 R (3054) 3698 0 R (3055) 3699 0 R (3056) 3700 0 R (3057) 3701 0 R (3058) 3702 0 R (3059) 3703 0 R (3060) 3704 0 R (3061) 3705 0 R (3062) 3706 0 R (3063) 3707 0 R (3064) 3708 0 R (3065) 3709 0 R (3066) 3710 0 R (3067) 3711 0 R (3068) 3712 0 R (3069) 3713 0 R (3070) 3720 0 R (3071) 3721 0 R (3072) 3722 0 R (3073) 3723 0 R (3074) 3724 0 R (3075) 3725 0 R (3076) 3726 0 R (3077) 3727 0 R (3078) 3728 0 R (3079) 3729 0 R (3080) 3730 0 R (3081) 3731 0 R (3082) 3732 0 R (3083) 3733 0 R (3084) 3734 0 R (3085) 3735 0 R (3089) 3736 0 R (3090) 3737 0 R (3091) 3738 0 R (3092) 3739 0 R (3093) 3740 0 R (3094) 3746 0 R (3095) 3747 0 R (3096) 3748 0 R (3097) 3749 0 R (3098) 3750 0 R (3099) 3751 0 R (31) 1193 0 R (3100) 3752 0 R (3101) 3753 0 R (3102) 3754 0 R (3103) 3755 0 R (3104) 3756 0 R (3105) 3757 0 R (3106) 3758 0 R (3107) 3759 0 R (3108) 3760 0 R (3109) 3761 0 R (3110) 3762 0 R (3111) 3763 0 R (3112) 3764 0 R (3113) 3765 0 R (3114) 3766 0 R (3115) 3767 0 R (3116) 3768 0 R (3117) 3769 0 R (3118) 3770 0 R (3119) 3771 0 R (3120) 3772 0 R (3121) 3773 0 R (3122) 3774 0 R (3123) 3775 0 R (3124) 3776 0 R (3125) 3777 0 R (3126) 3778 0 R (3127) 3779 0 R (3128) 3780 0 R (3129) 3781 0 R (3130) 3782 0 R (3131) 3783 0 R (3132) 3784 0 R (3133) 3785 0 R (3134) 3786 0 R (3135) 3787 0 R (3136) 3788 0 R (3137) 3789 0 R (3138) 3790 0 R (3139) 3791 0 R (3140) 3792 0 R (3141) 3793 0 R (3142) 3794 0 R (3143) 3795 0 R (3144) 3796 0 R (3145) 3797 0 R (3146) 3798 0 R (3147) 3799 0 R (3148) 3800 0 R (3149) 3801 0 R (3150) 3802 0 R (3151) 3803 0 R (3152) 3804 0 R (3153) 3805 0 R (3154) 3806 0 R (3155) 3807 0 R (3156) 3808 0 R (3157) 3809 0 R (3158) 3810 0 R (3159) 3811 0 R (3160) 3812 0 R (3161) 3813 0 R (3162) 3814 0 R (3163) 3815 0 R (3164) 3816 0 R (3165) 3817 0 R (3166) 3818 0 R (3167) 3819 0 R (3168) 3820 0 R (3169) 3821 0 R (3170) 3822 0 R (3171) 3823 0 R (3172) 3824 0 R (3173) 3825 0 R (3174) 3826 0 R (3175) 3827 0 R (3176) 3828 0 R (3177) 3829 0 R (3178) 3830 0 R (3179) 3831 0 R (3180) 3832 0 R (3181) 3833 0 R (3182) 3834 0 R (3183) 3835 0 R (3184) 3836 0 R (3185) 3837 0 R (3186) 3838 0 R (3187) 3839 0 R (3188) 3840 0 R (3189) 3841 0 R (3190) 3842 0 R (3191) 3843 0 R (3192) 3844 0 R (3193) 3845 0 R (3194) 3846 0 R (3195) 3847 0 R (3196) 3848 0 R (3197) 3849 0 R (3198) 3850 0 R (3199) 3851 0 R (32) 1194 0 R (3200) 3852 0 R (3201) 965 0 R (3204) 966 0 R (3206) 3853 0 R (3207) 3854 0 R (3261) 967 0 R (3263) 3871 0 R (3264) 3872 0 R (3265) 3873 0 R (3266) 3874 0 R (3267) 3875 0 R (3268) 3876 0 R (3269) 3877 0 R (3270) 3878 0 R (3271) 968 0 R (3273) 3879 0 R (3275) 3880 0 R (3276) 3881 0 R (3277) 3882 0 R (3278) 3883 0 R (3279) 3884 0 R (3282) 3885 0 R (3283) 3886 0 R (3284) 3887 0 R (3285) 3888 0 R (3286) 3889 0 R (3289) 3890 0 R (3290) 3891 0 R (3291) 3892 0 R (3292) 3893 0 R (3293) 3894 0 R (3296) 3895 0 R (3297) 3896 0 R (3298) 3897 0 R (3299) 3898 0 R (33) 1195 0 R (3301) 3899 0 R (3302) 3900 0 R (3303) 3901 0 R (3304) 3902 0 R (3307) 3903 0 R (3308) 3904 0 R (3309) 3905 0 R (3310) 3906 0 R (3311) 3907 0 R (3312) 3908 0 R (3313) 3909 0 R (3314) 3910 0 R (3315) 3921 0 R (3318) 3922 0 R (3319) 3923 0 R (3320) 3924 0 R (3321) 3925 0 R (3322) 3926 0 R (3323) 3927 0 R (3324) 3928 0 R (3325) 3929 0 R (3326) 3930 0 R (3328) 3931 0 R (3329) 3932 0 R (3330) 3933 0 R (3331) 3934 0 R (3332) 3935 0 R (3334) 3936 0 R (3335) 3937 0 R (3336) 3938 0 R (3337) 3939 0 R (3338) 3940 0 R (3339) 3941 0 R (3340) 3942 0 R (3341) 3943 0 R (3343) 3944 0 R (3344) 3945 0 R (3345) 3946 0 R (3346) 3947 0 R (3347) 3948 0 R (3350) 3949 0 R (3351) 3950 0 R (3352) 3951 0 R (3353) 3952 0 R (3354) 3953 0 R (3355) 3954 0 R (3357) 3962 0 R (3358) 3963 0 R (3359) 3964 0 R (3360) 3965 0 R (3361) 3966 0 R (3363) 3967 0 R (3364) 3968 0 R (3365) 3969 0 R (3366) 3970 0 R (3367) 3971 0 R (3368) 3972 0 R (3369) 3973 0 R (3371) 3974 0 R (3372) 3975 0 R (3373) 3976 0 R (3374) 3977 0 R (3375) 3978 0 R (3376) 3979 0 R (3378) 3980 0 R (3379) 3981 0 R (3380) 3982 0 R (3381) 3983 0 R (3382) 3984 0 R (3385) 3985 0 R (3386) 3986 0 R (3387) 3987 0 R (3388) 3988 0 R (3389) 3989 0 R (3392) 3990 0 R (3393) 3991 0 R (3394) 3992 0 R (3395) 3993 0 R (3396) 3994 0 R (3399) 3995 0 R (34) 1196 0 R (3400) 3996 0 R (3401) 3997 0 R (3402) 3998 0 R (3403) 3999 0 R (3406) 4000 0 R (3407) 4001 0 R (3408) 4002 0 R (3409) 4003 0 R (3410) 4004 0 R (3413) 4005 0 R (3414) 4006 0 R (3415) 4007 0 R (3416) 4008 0 R (3417) 4009 0 R (3418) 4010 0 R (3420) 4011 0 R (3421) 4012 0 R (3422) 4013 0 R (3423) 4014 0 R (3424) 4015 0 R (3425) 4016 0 R (3427) 4018 0 R (3429) 4028 0 R (3430) 4029 0 R (3431) 4030 0 R (3434) 4031 0 R (3435) 4032 0 R (3436) 4033 0 R (3437) 4034 0 R (3438) 4035 0 R (3441) 4036 0 R (3442) 4037 0 R (3443) 4038 0 R (3444) 4039 0 R (3445) 4040 0 R (3448) 4041 0 R (3449) 4042 0 R (3450) 4043 0 R (3451) 4044 0 R (3452) 4045 0 R (3455) 4046 0 R (3456) 4047 0 R (3457) 4048 0 R (3458) 4049 0 R (3459) 4050 0 R (3462) 4051 0 R (3463) 4052 0 R (3464) 4053 0 R (3465) 4054 0 R (3466) 4055 0 R (3467) 4056 0 R (3468) 4057 0 R (3469) 4058 0 R (3470) 4059 0 R (3471) 4060 0 R (3475) 4061 0 R (3476) 4062 0 R (3477) 4063 0 R (3478) 4064 0 R (3479) 4065 0 R (3480) 4066 0 R (3481) 4067 0 R (3482) 4068 0 R (3483) 4069 0 R (3484) 4070 0 R (3485) 4071 0 R (3486) 4072 0 R (3487) 4073 0 R (3488) 4074 0 R (3489) 4075 0 R (3490) 4087 0 R (3491) 4088 0 R (3492) 4089 0 R (3493) 4090 0 R (3494) 4091 0 R (3495) 4092 0 R (3496) 4093 0 R (3497) 4094 0 R (3498) 4095 0 R (3499) 4096 0 R (35) 1197 0 R (350) 1401 0 R (3500) 4097 0 R (3501) 4098 0 R (3502) 4099 0 R (3503) 4100 0 R (3504) 4101 0 R (3505) 4102 0 R (3506) 4103 0 R (351) 1402 0 R (352) 1403 0 R (353) 1404 0 R (3534) 4105 0 R (3535) 970 0 R (3539) 4106 0 R (354) 1405 0 R (3540) 1058 0 R (3542) 4107 0 R (3545) 4113 0 R (3546) 4114 0 R (3547) 4115 0 R (356) 1407 0 R (3576) 4117 0 R (3577) 4118 0 R (3578) 4119 0 R (3579) 4120 0 R (359) 1408 0 R (36) 1198 0 R (360) 1409 0 R (361) 1410 0 R (362) 1411 0 R (363) 1412 0 R (364) 1413 0 R (366) 1414 0 R (368) 714 0 R (37) 1199 0 R (370) 715 0 R (372) 1415 0 R (3723) 4129 0 R (3724) 4130 0 R (3725) 4131 0 R (3726) 4132 0 R (3739) 4134 0 R (3740) 4135 0 R (3741) 4136 0 R (3742) 4137 0 R (375) 1416 0 R (378) 1418 0 R (3786) 4146 0 R (3787) 4147 0 R (3788) 4148 0 R (3789) 1059 0 R (379) 1419 0 R (3791) 4149 0 R (3792) 4150 0 R (3793) 4151 0 R (3794) 4157 0 R (3795) 4158 0 R (3796) 4159 0 R (3797) 4160 0 R (38) 1200 0 R (380) 1420 0 R (381) 1421 0 R (382) 1422 0 R (383) 1423 0 R (384) 1424 0 R (385) 1425 0 R (3859) 4162 0 R (386) 1426 0 R (3860) 4163 0 R (3861) 4164 0 R (3862) 4165 0 R (3863) 4166 0 R (387) 1427 0 R (388) 1428 0 R (3888) 4168 0 R (3889) 4169 0 R (389) 1429 0 R (3890) 1060 0 R (3892) 4170 0 R (3893) 4171 0 R (3894) 4172 0 R (3895) 4178 0 R (3896) 4179 0 R (3897) 4180 0 R (39) 690 0 R (390) 1430 0 R (391) 1431 0 R (392) 1432 0 R (393) 1433 0 R (394) 1434 0 R (395) 1435 0 R (396) 1436 0 R (397) 1437 0 R (398) 1439 0 R (3980) 4182 0 R (3981) 4183 0 R (3982) 4184 0 R (3983) 4185 0 R (3984) 4186 0 R (3987) 4187 0 R (3988) 4188 0 R (3989) 4189 0 R (399) 1448 0 R (4.0) 89 0 R (4.10.1) 93 0 R (4.10.7.2) 97 0 R (4.10.8.2) 101 0 R (4.11.1) 105 0 R (4.12.1) 109 0 R (4.13.1) 113 0 R (4.13.10.2) 129 0 R (4.13.9.2) 117 0 R (4.13.9.5.3) 121 0 R (4.13.9.6.3) 125 0 R (400) 1449 0 R (4009) 4196 0 R (401) 1450 0 R (4010) 4197 0 R (4011) 1062 0 R (4013) 4198 0 R (4014) 4199 0 R (4015) 4200 0 R (4016) 4201 0 R (4017) 4202 0 R (402) 1451 0 R (4035) 4204 0 R (4036) 4205 0 R (4037) 4206 0 R (4038) 1063 0 R (404) 1452 0 R (4040) 4207 0 R (4041) 4208 0 R (4042) 4209 0 R (4043) 4210 0 R (4044) 4211 0 R (4045) 1064 0 R (4048) 4212 0 R (4049) 4213 0 R (405) 1453 0 R (4050) 4214 0 R (4051) 4215 0 R (4052) 4216 0 R (4053) 4217 0 R (4054) 4218 0 R (4055) 4219 0 R (4056) 4220 0 R (4057) 4221 0 R (4058) 4222 0 R (4059) 4223 0 R (406) 1454 0 R (4060) 4225 0 R (4062) 4230 0 R (4063) 4231 0 R (4064) 4232 0 R (4065) 1065 0 R (4068) 4233 0 R (4069) 4234 0 R (407) 1455 0 R (4070) 4235 0 R (4071) 4236 0 R (4072) 4237 0 R (4073) 4238 0 R (4074) 4239 0 R (4075) 4240 0 R (4076) 4241 0 R (4077) 4242 0 R (4078) 4243 0 R (4079) 4244 0 R (408) 1456 0 R (4080) 4245 0 R (4081) 4246 0 R (4082) 4247 0 R (4083) 4248 0 R (4084) 4249 0 R (4085) 1066 0 R (4088) 4250 0 R (4089) 4251 0 R (409) 1457 0 R (4090) 4252 0 R (4091) 4253 0 R (4092) 4254 0 R (4093) 4255 0 R (4094) 4256 0 R (4095) 4257 0 R (4096) 1067 0 R (41) 1201 0 R (410) 1458 0 R (4100) 4258 0 R (4101) 4259 0 R (4102) 4260 0 R (4103) 4261 0 R (4104) 4262 0 R (4105) 4263 0 R (4106) 4264 0 R (4107) 4265 0 R (4108) 4266 0 R (4109) 4267 0 R (411) 1459 0 R (4110) 4268 0 R (4111) 4274 0 R (4112) 4275 0 R (4113) 4276 0 R (4114) 4277 0 R (4115) 4278 0 R (4116) 4279 0 R (4117) 4280 0 R (412) 1460 0 R (413) 1461 0 R (414) 1462 0 R (415) 1463 0 R (416) 1464 0 R (4165) 4282 0 R (4166) 4283 0 R (4167) 4284 0 R (417) 1465 0 R (4173) 4296 0 R (4174) 4297 0 R (4175) 4298 0 R (4176) 4299 0 R (4177) 4300 0 R (4178) 4301 0 R (4179) 4302 0 R (418) 1466 0 R (4180) 4303 0 R (4181) 4304 0 R (4182) 4305 0 R (4183) 4306 0 R (4184) 4307 0 R (4185) 4308 0 R (4186) 4309 0 R (4187) 4310 0 R (4188) 1070 0 R (419) 1467 0 R (4192) 4311 0 R (4193) 4312 0 R (4194) 4313 0 R (4195) 4314 0 R (4196) 4315 0 R (4197) 4316 0 R (4198) 4322 0 R (4199) 4323 0 R (42) 1202 0 R (420) 1468 0 R (4200) 4324 0 R (4201) 4325 0 R (4202) 4326 0 R (4203) 4327 0 R (4204) 4328 0 R (4205) 4329 0 R (4206) 4330 0 R (4207) 4331 0 R (4208) 4332 0 R (4209) 4333 0 R (421) 1469 0 R (4210) 4334 0 R (4211) 1071 0 R (4214) 4335 0 R (4215) 4336 0 R (4216) 4337 0 R (4217) 4338 0 R (4218) 4339 0 R (4219) 4340 0 R (422) 1470 0 R (4220) 4341 0 R (4221) 4342 0 R (4222) 4343 0 R (4223) 4344 0 R (4224) 4345 0 R (4225) 4346 0 R (4226) 4347 0 R (4227) 4348 0 R (4228) 4349 0 R (4229) 1072 0 R (423) 1471 0 R (4232) 4350 0 R (4233) 4351 0 R (4234) 4352 0 R (4235) 4353 0 R (4236) 4354 0 R (4237) 4355 0 R (424) 1472 0 R (4240) 4356 0 R (4241) 4357 0 R (4242) 4358 0 R (4243) 4359 0 R (4244) 4365 0 R (4245) 4366 0 R (4246) 4367 0 R (4247) 4368 0 R (425) 1473 0 R (4250) 1075 0 R (4252) 1076 0 R (4254) 4379 0 R (4255) 1077 0 R (4257) 4380 0 R (4258) 4381 0 R (4259) 4382 0 R (426) 1474 0 R (4260) 1078 0 R (4262) 4383 0 R (4263) 4384 0 R (4264) 4385 0 R (4265) 4386 0 R (4266) 4387 0 R (4267) 4388 0 R (4268) 4389 0 R (4269) 4390 0 R (427) 1475 0 R (4270) 4391 0 R (4271) 4392 0 R (4272) 4393 0 R (4273) 4394 0 R (4276) 1080 0 R (4278) 1081 0 R (428) 1476 0 R (4282) 4406 0 R (4283) 4407 0 R (4284) 4408 0 R (4285) 4409 0 R (4286) 4410 0 R (4287) 4411 0 R (4288) 4412 0 R (4289) 4413 0 R (4290) 4414 0 R (4291) 4415 0 R (4292) 4416 0 R (4293) 4417 0 R (4294) 4418 0 R (4295) 4419 0 R (4296) 4420 0 R (4297) 4421 0 R (4298) 4422 0 R (4299) 4423 0 R (43) 1203 0 R (430) 1477 0 R (4300) 4424 0 R (4301) 4426 0 R (4308) 1084 0 R (431) 1478 0 R (4310) 4431 0 R (4311) 4432 0 R (4312) 4433 0 R (4313) 1085 0 R (4315) 4434 0 R (4316) 4435 0 R (4317) 1086 0 R (432) 1479 0 R (4322) 4436 0 R (4323) 4437 0 R (4324) 4438 0 R (4325) 4439 0 R (4326) 4440 0 R (4327) 4441 0 R (4328) 4442 0 R (4329) 4443 0 R (433) 1480 0 R (434) 1481 0 R (435) 1482 0 R (436) 1483 0 R (437) 1484 0 R (438) 1485 0 R (439) 1486 0 R (4396) 4450 0 R (4397) 4451 0 R (4398) 4452 0 R (440) 1487 0 R (441) 1488 0 R (442) 1489 0 R (443) 1490 0 R (444) 1491 0 R (4445) 4454 0 R (4446) 4455 0 R (4447) 4456 0 R (4448) 4457 0 R (4449) 4458 0 R (445) 1492 0 R (4450) 4459 0 R (4451) 4460 0 R (4452) 4461 0 R (4453) 4462 0 R (446) 1493 0 R (447) 1494 0 R (4477) 4464 0 R (4478) 4465 0 R (4479) 4466 0 R (448) 1495 0 R (4480) 4468 0 R (4481) 4473 0 R (4482) 4474 0 R (4483) 4475 0 R (4484) 4476 0 R (4485) 4477 0 R (4490) 4478 0 R (4491) 4479 0 R (4492) 4480 0 R (4493) 4481 0 R (4494) 4482 0 R (4495) 4483 0 R (4496) 1090 0 R (4497) 4484 0 R (4499) 4485 0 R (4506) 4486 0 R (451) 1496 0 R (4513) 4487 0 R (452) 1498 0 R (4522) 4493 0 R (4530) 4494 0 R (4538) 4495 0 R (4546) 4496 0 R (455) 1505 0 R (4553) 4497 0 R (456) 1506 0 R (457) 1507 0 R (4570) 4498 0 R (458) 1508 0 R (4589) 4499 0 R (459) 1509 0 R (4594) 4500 0 R (460) 1510 0 R (4606) 4501 0 R (461) 1511 0 R (4618) 4502 0 R (462) 1512 0 R (4626) 4503 0 R (4629) 4504 0 R (463) 1513 0 R (4637) 4505 0 R (464) 1514 0 R (465) 1515 0 R (4658) 4506 0 R (4669) 4507 0 R (4671) 4514 0 R (468) 1517 0 R (4689) 4515 0 R (469) 1518 0 R (470) 1519 0 R (4701) 4516 0 R (471) 1520 0 R (4712) 4517 0 R (472) 1521 0 R (4729) 4518 0 R (473) 1522 0 R (474) 1523 0 R (4741) 4519 0 R (4749) 4520 0 R (475) 1524 0 R (4757) 4521 0 R (476) 1525 0 R (4760) 4522 0 R (4768) 4523 0 R (477) 1526 0 R (4775) 4524 0 R (478) 1527 0 R (4782) 4525 0 R (479) 1528 0 R (4796) 4526 0 R (4799) 4527 0 R (480) 1529 0 R (4807) 4528 0 R (4815) 4529 0 R (4823) 4531 0 R (483) 1531 0 R (4834) 4536 0 R (484) 1532 0 R (4846) 4537 0 R (485) 1533 0 R (4852) 4538 0 R (4859) 4539 0 R (486) 1534 0 R (4867) 4540 0 R (487) 1535 0 R (4875) 4541 0 R (488) 1536 0 R (4882) 4542 0 R (489) 1537 0 R (4896) 4543 0 R (490) 1538 0 R (4903) 4544 0 R (4905) 4545 0 R (491) 1539 0 R (492) 1540 0 R (4925) 4546 0 R (4926) 1092 0 R (4929) 4547 0 R (493) 1541 0 R (4930) 1093 0 R (4931) 4548 0 R (494) 1542 0 R (495) 1543 0 R (496) 1544 0 R (497) 1545 0 R (498) 1546 0 R (499) 1547 0 R (5.0) 133 0 R (5.14.1) 137 0 R (5.14.11.2) 141 0 R (5.15.1) 145 0 R (5.16.1) 149 0 R (5.17.1) 153 0 R (5.17.12.2) 157 0 R (5.17.12.7.3) 161 0 R (5.17.12.8.3) 165 0 R (5.17.13.2) 169 0 R (5.17.14.2) 173 0 R (5.17.15.2) 177 0 R (5.17.16.2) 181 0 R (5.17.17.2) 185 0 R (5.18.1) 189 0 R (5.19.1) 193 0 R (5.20.1) 197 0 R (5.20.18.2) 201 0 R (5.20.19.2) 205 0 R (5.20.20.2) 209 0 R (5.20.21.2) 213 0 R (5.20.22.2) 217 0 R (5.21.1) 221 0 R (5.21.23.2) 225 0 R (5.21.24.10.3) 237 0 R (5.21.24.2) 229 0 R (5.21.24.9.3) 233 0 R (5.21.25.2) 241 0 R (5.21.26.2) 245 0 R (5.21.27.2) 249 0 R (500) 1548 0 R (501) 1549 0 R (502) 1550 0 R (503) 1551 0 R (504) 1552 0 R (506) 1553 0 R (507) 1554 0 R (508) 1555 0 R (509) 1556 0 R (511) 1557 0 R (512) 1558 0 R (513) 1559 0 R (514) 1560 0 R (515) 1561 0 R (516) 1562 0 R (517) 1563 0 R (518) 1564 0 R (519) 1565 0 R (521) 1566 0 R (522) 1567 0 R (523) 1568 0 R (524) 1569 0 R (525) 1570 0 R (526) 1571 0 R (527) 1572 0 R (528) 1573 0 R (529) 1574 0 R (531) 1575 0 R (532) 1576 0 R (533) 1577 0 R (534) 1578 0 R (535) 1579 0 R (537) 1580 0 R (538) 1581 0 R (539) 1582 0 R (540) 1583 0 R (542) 1585 0 R (544) 1591 0 R (545) 1592 0 R (546) 1593 0 R (547) 1594 0 R (548) 1595 0 R (549) 1596 0 R (550) 1597 0 R (552) 1598 0 R (553) 1599 0 R (554) 1600 0 R (555) 1601 0 R (557) 1602 0 R (558) 1603 0 R (559) 1604 0 R (560) 1605 0 R (562) 1606 0 R (563) 1607 0 R (564) 1608 0 R (565) 1609 0 R (567) 1610 0 R (568) 1611 0 R (569) 1612 0 R (570) 1613 0 R (572) 1614 0 R (573) 1615 0 R (574) 1616 0 R (575) 1617 0 R (576) 1618 0 R (578) 1619 0 R (579) 1620 0 R (580) 1621 0 R (581) 1622 0 R (583) 1623 0 R (584) 1624 0 R (585) 1625 0 R (586) 1626 0 R (588) 1627 0 R (589) 1628 0 R (590) 1629 0 R (591) 1630 0 R (592) 1631 0 R (593) 1632 0 R (594) 1633 0 R (595) 1634 0 R (596) 1635 0 R (597) 1636 0 R (598) 1637 0 R (599) 1638 0 R (6.0) 253 0 R (6.22.1) 257 0 R (6.23.1) 261 0 R (600) 1639 0 R (601) 1640 0 R (602) 1641 0 R (603) 1642 0 R (604) 1643 0 R (605) 1644 0 R (606) 1645 0 R (607) 1646 0 R (609) 1652 0 R (610) 1653 0 R (611) 1654 0 R (612) 1655 0 R (613) 1656 0 R (614) 1657 0 R (615) 1658 0 R (616) 1659 0 R (617) 1660 0 R (618) 1661 0 R (619) 1662 0 R (620) 1663 0 R (621) 1664 0 R (622) 1665 0 R (623) 1666 0 R (624) 1667 0 R (625) 1668 0 R (626) 1669 0 R (627) 1670 0 R (628) 1671 0 R (629) 1672 0 R (630) 1673 0 R (631) 1674 0 R (632) 1675 0 R (633) 1676 0 R (634) 1677 0 R (635) 1678 0 R (636) 1679 0 R (637) 1680 0 R (638) 1681 0 R (639) 1682 0 R (640) 1683 0 R (641) 1684 0 R (642) 1685 0 R (643) 1686 0 R (644) 1687 0 R (645) 1688 0 R (646) 1689 0 R (647) 1690 0 R (648) 1691 0 R (649) 1692 0 R (650) 1699 0 R (651) 1700 0 R (652) 1701 0 R (653) 1702 0 R (654) 1703 0 R (655) 1704 0 R (656) 1705 0 R (657) 1706 0 R (658) 1707 0 R (659) 1708 0 R (660) 718 0 R (662) 1709 0 R (663) 1710 0 R (664) 1711 0 R (665) 1712 0 R (692) 1719 0 R (693) 1720 0 R (694) 1721 0 R (695) 1722 0 R (696) 1723 0 R (697) 1724 0 R (7.0) 265 0 R (7.24.1) 269 0 R (7.24.28.11.3) 277 0 R (7.24.28.12.3) 281 0 R (7.24.28.2) 273 0 R (7.24.29.13.3) 289 0 R (7.24.29.14.3) 293 0 R (7.24.29.2) 285 0 R (7.25.1) 297 0 R (7.25.30.2) 301 0 R (7.25.31.2) 305 0 R (7.25.32.2) 309 0 R (7.25.33.2) 313 0 R (7.25.34.2) 317 0 R (7.25.35.2) 321 0 R (7.25.36.2) 325 0 R (7.25.37.2) 329 0 R (7.25.38.2) 333 0 R (7.25.39.15.3) 341 0 R (7.25.39.16.3) 345 0 R (7.25.39.2) 337 0 R (7.25.40.2) 349 0 R (7.25.41.2) 353 0 R (7.25.42.2) 357 0 R (7.25.43.17.3) 365 0 R (7.25.43.18.3) 369 0 R (7.25.43.19.3) 373 0 R (7.25.43.2) 361 0 R (7.25.43.20.3) 377 0 R (7.25.43.21.3) 381 0 R (7.25.43.22.3) 385 0 R (7.25.43.23.3) 389 0 R (7.25.43.24.3) 393 0 R (7.25.43.25.3) 397 0 R (7.25.43.26.3) 401 0 R (7.25.43.27.3) 405 0 R (7.25.43.28.3) 409 0 R (7.25.43.29.3) 413 0 R (7.25.43.30.3) 417 0 R (7.25.43.31.3) 421 0 R (7.25.44.2) 425 0 R (7.25.45.2) 429 0 R (7.25.46.2) 433 0 R (7.25.47.2) 437 0 R (7.25.48.2) 441 0 R (7.25.49.2) 445 0 R (7.25.50.2) 449 0 R (7.25.51.2) 453 0 R (7.25.51.32.3) 457 0 R (7.25.51.33.3) 461 0 R (7.25.51.34.3) 465 0 R (7.25.51.35.3) 469 0 R (7.26.1) 473 0 R (7.26.52.2) 477 0 R (7.26.52.36.3) 481 0 R (7.26.52.37.3) 485 0 R (7.26.53.2) 489 0 R (7.26.54.2) 493 0 R (7.26.55.2) 497 0 R (7.26.56.2) 501 0 R (7.26.56.38.3) 505 0 R (7.26.56.39.3) 509 0 R (7.26.56.40.3) 513 0 R (7.26.57.2) 517 0 R (700) 1725 0 R (701) 1726 0 R (702) 1727 0 R (703) 1728 0 R (704) 1729 0 R (705) 1730 0 R (706) 1731 0 R (707) 1732 0 R (708) 1733 0 R (709) 1734 0 R (710) 1735 0 R (711) 1736 0 R (712) 1737 0 R (713) 1738 0 R (716) 1739 0 R (718) 1747 0 R (719) 1748 0 R (72) 1211 0 R (720) 1749 0 R (721) 1750 0 R (722) 1751 0 R (723) 1752 0 R (724) 723 0 R (726) 1753 0 R (727) 1754 0 R (728) 1755 0 R (729) 1756 0 R (73) 1212 0 R (730) 1757 0 R (731) 1758 0 R (732) 1759 0 R (733) 1760 0 R (734) 1761 0 R (735) 1762 0 R (736) 1763 0 R (737) 1764 0 R (738) 1765 0 R (739) 1766 0 R (74) 1213 0 R (740) 1767 0 R (741) 1768 0 R (742) 1769 0 R (743) 1770 0 R (744) 1771 0 R (745) 1772 0 R (746) 1779 0 R (747) 1780 0 R (748) 1781 0 R (749) 1782 0 R (75) 1214 0 R (750) 1783 0 R (751) 1784 0 R (752) 1785 0 R (753) 1786 0 R (754) 1787 0 R (755) 1788 0 R (756) 1789 0 R (757) 1790 0 R (758) 1791 0 R (759) 1792 0 R (760) 1793 0 R (761) 1794 0 R (762) 1795 0 R (763) 1796 0 R (764) 1797 0 R (765) 1798 0 R (766) 1799 0 R (767) 1800 0 R (768) 1801 0 R (769) 1802 0 R (770) 1803 0 R (771) 1804 0 R (772) 1805 0 R (773) 1806 0 R (774) 1807 0 R (775) 1808 0 R (776) 1809 0 R (777) 1810 0 R (778) 1811 0 R (779) 1812 0 R (780) 1813 0 R (781) 1814 0 R (782) 1815 0 R (783) 1816 0 R (784) 1817 0 R (785) 1818 0 R (786) 1819 0 R (787) 1820 0 R (788) 1821 0 R (789) 1822 0 R (790) 1823 0 R (791) 1824 0 R (793) 1825 0 R (794) 1827 0 R (796) 1833 0 R (797) 1839 0 R (798) 1840 0 R (799) 1841 0 R (8.0) 521 0 R (8.27.1) 525 0 R (8.28.1) 529 0 R (8.28.58.2) 533 0 R (8.28.59.2) 537 0 R (8.29.1) 541 0 R (800) 1842 0 R (801) 1844 0 R (804) 1849 0 R (805) 1850 0 R (806) 1851 0 R (807) 1852 0 R (808) 1853 0 R (809) 1854 0 R (810) 1855 0 R (812) 1856 0 R (813) 1857 0 R (814) 1858 0 R (815) 806 0 R (817) 1859 0 R (818) 1860 0 R (819) 1861 0 R (820) 807 0 R (822) 1862 0 R (823) 1863 0 R (824) 1864 0 R (825) 1865 0 R (826) 1866 0 R (827) 1867 0 R (828) 1868 0 R (829) 1869 0 R (830) 1870 0 R (831) 808 0 R (833) 1877 0 R (834) 1878 0 R (835) 1879 0 R (836) 809 0 R (838) 1880 0 R (839) 810 0 R (841) 1881 0 R (842) 1882 0 R (843) 1883 0 R (844) 1884 0 R (845) 1885 0 R (846) 1886 0 R (847) 1887 0 R (848) 1888 0 R (849) 1889 0 R (850) 1890 0 R (851) 811 0 R (853) 1891 0 R (854) 1892 0 R (855) 1893 0 R (856) 1894 0 R (857) 1895 0 R (858) 1896 0 R (859) 1898 0 R (861) 1904 0 R (862) 1905 0 R (863) 1906 0 R (864) 1907 0 R (865) 1908 0 R (866) 1909 0 R (867) 812 0 R (869) 1910 0 R (870) 1911 0 R (871) 1912 0 R (872) 1913 0 R (873) 1914 0 R (874) 1915 0 R (875) 1916 0 R (876) 1917 0 R (877) 1918 0 R (878) 1919 0 R (880) 813 0 R (882) 1920 0 R (883) 1921 0 R (884) 814 0 R (886) 1922 0 R (887) 1923 0 R (888) 1924 0 R (889) 1925 0 R (890) 1926 0 R (891) 1927 0 R (892) 1928 0 R (893) 1935 0 R (894) 1936 0 R (895) 1937 0 R (896) 1938 0 R (897) 1939 0 R (898) 1940 0 R (899) 815 0 R (9) 1175 0 R (9.0) 545 0 R (9.30.1) 549 0 R (9.30.60.2) 553 0 R (9.31.1) 557 0 R (9.32.1) 561 0 R (901) 1941 0 R (902) 1942 0 R (903) 1943 0 R (904) 1944 0 R (905) 1945 0 R (906) 1946 0 R (909) 1947 0 R (910) 1948 0 R (911) 1949 0 R (912) 1950 0 R (913) 1951 0 R (914) 1952 0 R (915) 1958 0 R (916) 817 0 R (918) 1959 0 R (919) 1960 0 R (920) 1961 0 R (921) 1962 0 R (922) 1963 0 R (923) 1964 0 R (924) 1965 0 R (925) 1966 0 R (926) 1967 0 R (927) 1968 0 R (928) 1969 0 R (929) 1970 0 R (930) 1971 0 R (931) 1972 0 R (932) 1973 0 R (933) 1974 0 R (934) 1975 0 R (935) 1976 0 R (936) 818 0 R (938) 1977 0 R (939) 1978 0 R (940) 1979 0 R (941) 1980 0 R (942) 1981 0 R (943) 1983 0 R (944) 1988 0 R (945) 1989 0 R (946) 1990 0 R (947) 1991 0 R (948) 819 0 R (950) 1992 0 R (951) 1993 0 R (952) 1994 0 R (953) 1995 0 R (954) 1996 0 R (955) 1997 0 R (956) 1998 0 R (957) 1999 0 R (958) 2000 0 R (959) 2001 0 R (960) 2002 0 R (961) 820 0 R (963) 2003 0 R (964) 2004 0 R (965) 2005 0 R (966) 2006 0 R (967) 2007 0 R (968) 2008 0 R (969) 2009 0 R (970) 2010 0 R (971) 2011 0 R (972) 2012 0 R (973) 2013 0 R (974) 2014 0 R (975) 2015 0 R (976) 2016 0 R (977) 821 0 R (979) 2022 0 R (980) 2023 0 R (981) 2024 0 R (982) 2025 0 R (983) 2026 0 R (984) 822 0 R (987) 2027 0 R (988) 2028 0 R (989) 2029 0 R (990) 2030 0 R (991) 2031 0 R (992) 2032 0 R (993) 2033 0 R (994) 2034 0 R (995) 2035 0 R (996) 2036 0 R (998) 2037 0 R (Access_Control_Lists) 1069 0 R (Configuration_File_Grammar) 841 0 R (DNSSEC) 816 0 R (Doc-Start) 663 0 R (Setting_TTLs) 1061 0 R (access_control) 945 0 R (acl) 843 0 R (address_match_lists) 835 0 R (admin_tools) 717 0 R (bibliography) 1088 0 R (boolean_options) 943 0 R (ch01) 687 0 R (ch02) 702 0 R (ch03) 708 0 R (ch04) 719 0 R (ch05) 830 0 R (ch06) 833 0 R (ch07) 1068 0 R (ch08) 1074 0 R (ch09) 1079 0 R (classes_of_resource_records) 1083 0 R (configuration_file_elements) 834 0 R (controls_statement_definition_and_usage) 845 0 R (diagnostic_tools) 716 0 R (dig) 1417 0 R (dynamic_update) 720 0 R (dynamic_update_policies) 969 0 R (dynamic_update_security) 1073 0 R (historical_dns_information) 1082 0 R (incremental_zone_transfers) 722 0 R (internet_drafts) 1091 0 R (ipv6addresses) 1087 0 R (journal) 721 0 R (lwresd) 832 0 R (named-checkconf) 1504 0 R (named-checkzone) 1516 0 R (notify) 713 0 R (page.1) 665 0 R (page.10) 1237 0 R (page.100) 4360 0 R (page.101) 4369 0 R (page.102) 4374 0 R (page.103) 4395 0 R (page.104) 4401 0 R (page.105) 4425 0 R (page.106) 4445 0 R (page.107) 4467 0 R (page.108) 4488 0 R (page.109) 4508 0 R (page.11) 1274 0 R (page.110) 4530 0 R (page.111) 4549 0 R (page.112) 4554 0 R (page.113) 4559 0 R (page.12) 1303 0 R (page.13) 1324 0 R (page.14) 1330 0 R (page.15) 1347 0 R (page.16) 1373 0 R (page.17) 1385 0 R (page.18) 1396 0 R (page.19) 1438 0 R (page.2) 671 0 R (page.20) 1497 0 R (page.21) 1584 0 R (page.22) 1647 0 R (page.23) 1693 0 R (page.24) 1714 0 R (page.25) 1741 0 R (page.26) 1773 0 R (page.27) 1826 0 R (page.28) 1834 0 R (page.29) 1843 0 R (page.3) 676 0 R (page.30) 1871 0 R (page.31) 1897 0 R (page.32) 1929 0 R (page.33) 1953 0 R (page.34) 1982 0 R (page.35) 2017 0 R (page.36) 2041 0 R (page.37) 2061 0 R (page.38) 2085 0 R (page.39) 2099 0 R (page.4) 682 0 R (page.40) 2104 0 R (page.41) 2131 0 R (page.42) 2136 0 R (page.43) 2155 0 R (page.44) 2164 0 R (page.45) 2197 0 R (page.46) 2221 0 R (page.47) 2243 0 R (page.48) 2262 0 R (page.49) 2313 0 R (page.5) 725 0 R (page.50) 2355 0 R (page.51) 2414 0 R (page.52) 2456 0 R (page.53) 2507 0 R (page.54) 2526 0 R (page.55) 2546 0 R (page.56) 2565 0 R (page.57) 2616 0 R (page.58) 2751 0 R (page.59) 2831 0 R (page.6) 846 0 R (page.60) 2886 0 R (page.61) 2930 0 R (page.62) 2993 0 R (page.63) 3041 0 R (page.64) 3091 0 R (page.65) 3126 0 R (page.66) 3171 0 R (page.67) 3208 0 R (page.68) 3252 0 R (page.69) 3310 0 R (page.7) 971 0 R (page.70) 3356 0 R (page.71) 3394 0 R (page.72) 3427 0 R (page.73) 3453 0 R (page.74) 3468 0 R (page.75) 3504 0 R (page.76) 3551 0 R (page.77) 3583 0 R (page.78) 3619 0 R (page.79) 3667 0 R (page.8) 1094 0 R (page.80) 3714 0 R (page.81) 3741 0 R (page.82) 3856 0 R (page.83) 3861 0 R (page.84) 3866 0 R (page.85) 3911 0 R (page.86) 3955 0 R (page.87) 4017 0 R (page.88) 4076 0 R (page.89) 4108 0 R (page.9) 1205 0 R (page.90) 4122 0 R (page.91) 4139 0 R (page.92) 4152 0 R (page.93) 4173 0 R (page.94) 4191 0 R (page.95) 4224 0 R (page.96) 4269 0 R (page.97) 4285 0 R (page.98) 4291 0 R (page.99) 4317 0 R (proposed_standards) 1740 0 R (rfcs) 1089 0 R (rndc) 1530 0 R (rrset_ordering) 954 0 R (rrset_ordering_title) 1406 0 R (sample_configuration) 709 0 R (server_statement_definition_and_usage) 959 0 R (server_statement_grammar) 958 0 R (statsfile) 957 0 R (synthesis) 955 0 R (table.1) 1204 0 R (table.10) 3582 0 R (table.11) 3855 0 R (table.12) 4104 0 R (table.13) 4116 0 R (table.14) 4121 0 R (table.15) 4133 0 R (table.16) 4138 0 R (table.17) 4161 0 R (table.18) 4167 0 R (table.19) 4181 0 R (table.2) 1215 0 R (table.20) 4190 0 R (table.21) 4203 0 R (table.22) 4281 0 R (table.23) 4444 0 R (table.24) 4453 0 R (table.25) 4463 0 R (table.3) 1395 0 R (table.4) 1713 0 R (table.5) 2154 0 R (table.6) 2242 0 R (table.7) 2261 0 R (table.8) 2545 0 R (table.9) 3490 0 R (the_category_phrase) 938 0 R (the_sortlist_statement) 953 0 R (topology) 952 0 R (tsig) 724 0 R (tuning) 956 0 R (types_of_resource_records_and_when_to_use_them) 1057 0 R (zone_statement_grammar) 964 0 R (zone_transfers) 948 0 R] +/Limits [(0) (zone_transfers)] +>> endobj +4572 0 obj << +/Kids [4571 0 R] +>> endobj +4573 0 obj << +/Dests 4572 0 R +>> endobj +4574 0 obj << +/Type /Catalog +/Pages 4569 0 R +/Outlines 4570 0 R +/Names 4573 0 R + /PageMode /UseOutlines /URI << /Base () >> /ViewerPreferences << >> +/OpenAction 662 0 R +>> endobj +4575 0 obj << + /Author () /Title () /Subject () /Creator (LaTeX with hyperref package) /Producer (pdfTeX13.d) /Keywords () +/Creator (TeX) +/Producer (pdfTeX-0.13d) +/CreationDate (D:20020701165700) +>> endobj +xref +0 4576 +0000000000 65535 f +0000034031 00000 n +0000000009 00000 n +0000515791 00000 n +0000000048 00000 n +0000087652 00000 n +0000000083 00000 n +0000515669 00000 n +0000000122 00000 n +0000087964 00000 n +0000000164 00000 n +0000515596 00000 n +0000000206 00000 n +0000088405 00000 n +0000000247 00000 n +0000515510 00000 n +0000000289 00000 n +0000089675 00000 n +0000000342 00000 n +0000515424 00000 n +0000000384 00000 n +0000092411 00000 n +0000000441 00000 n +0000515314 00000 n +0000000483 00000 n +0000092790 00000 n +0000000537 00000 n +0000515240 00000 n +0000000581 00000 n +0000093231 00000 n +0000000623 00000 n +0000515153 00000 n +0000000667 00000 n +0000097749 00000 n +0000000717 00000 n +0000515066 00000 n +0000000761 00000 n +0000099209 00000 n +0000000792 00000 n +0000514942 00000 n +0000000836 00000 n +0000101938 00000 n +0000000888 00000 n +0000514868 00000 n +0000000934 00000 n +0000102379 00000 n +0000000980 00000 n +0000514781 00000 n +0000001026 00000 n +0000102755 00000 n +0000001067 00000 n +0000514707 00000 n +0000001113 00000 n +0000105763 00000 n +0000001156 00000 n +0000514583 00000 n +0000001200 00000 n +0000106395 00000 n +0000001246 00000 n +0000514522 00000 n +0000001292 00000 n +0000106897 00000 n +0000001330 00000 n +0000514448 00000 n +0000001374 00000 n +0000109769 00000 n +0000001430 00000 n +0000514323 00000 n +0000001470 00000 n +0000109891 00000 n +0000001526 00000 n +0000514249 00000 n +0000001568 00000 n +0000110396 00000 n +0000001613 00000 n +0000514162 00000 n +0000001655 00000 n +0000110646 00000 n +0000001695 00000 n +0000514075 00000 n +0000001737 00000 n +0000110961 00000 n +0000001780 00000 n +0000513988 00000 n +0000001822 00000 n +0000112378 00000 n +0000001885 00000 n +0000513914 00000 n +0000001927 00000 n +0000115341 00000 n +0000001978 00000 n +0000513786 00000 n +0000002018 00000 n +0000115527 00000 n +0000002072 00000 n +0000513673 00000 n +0000002115 00000 n +0000115650 00000 n +0000002160 00000 n +0000513597 00000 n +0000002205 00000 n +0000115900 00000 n +0000002257 00000 n +0000513520 00000 n +0000002303 00000 n +0000117781 00000 n +0000002362 00000 n +0000513429 00000 n +0000002406 00000 n +0000121354 00000 n +0000002445 00000 n +0000513337 00000 n +0000002489 00000 n +0000121921 00000 n +0000002520 00000 n +0000513220 00000 n +0000002564 00000 n +0000122046 00000 n +0000002610 00000 n +0000513102 00000 n +0000002656 00000 n +0000122234 00000 n +0000002723 00000 n +0000513023 00000 n +0000002771 00000 n +0000130695 00000 n +0000002816 00000 n +0000512944 00000 n +0000002864 00000 n +0000148787 00000 n +0000002913 00000 n +0000512865 00000 n +0000002960 00000 n +0000151500 00000 n +0000002994 00000 n +0000512733 00000 n +0000003035 00000 n +0000151623 00000 n +0000003083 00000 n +0000512615 00000 n +0000003127 00000 n +0000152127 00000 n +0000003166 00000 n +0000512550 00000 n +0000003213 00000 n +0000155865 00000 n +0000003256 00000 n +0000512457 00000 n +0000003300 00000 n +0000156433 00000 n +0000003360 00000 n +0000512364 00000 n +0000003404 00000 n +0000168202 00000 n +0000003438 00000 n +0000512232 00000 n +0000003482 00000 n +0000168962 00000 n +0000003511 00000 n +0000512114 00000 n +0000003558 00000 n +0000169278 00000 n +0000003628 00000 n +0000512035 00000 n +0000003677 00000 n +0000171995 00000 n +0000003726 00000 n +0000511956 00000 n +0000003775 00000 n +0000172310 00000 n +0000003821 00000 n +0000511863 00000 n +0000003868 00000 n +0000172498 00000 n +0000003937 00000 n +0000511770 00000 n +0000003984 00000 n +0000173258 00000 n +0000004055 00000 n +0000511677 00000 n +0000004102 00000 n +0000176462 00000 n +0000004166 00000 n +0000511584 00000 n +0000004213 00000 n +0000177220 00000 n +0000004269 00000 n +0000511505 00000 n +0000004316 00000 n +0000179800 00000 n +0000004349 00000 n +0000511412 00000 n +0000004393 00000 n +0000180752 00000 n +0000004422 00000 n +0000511319 00000 n +0000004466 00000 n +0000181252 00000 n +0000004499 00000 n +0000511187 00000 n +0000004543 00000 n +0000184156 00000 n +0000004574 00000 n +0000511108 00000 n +0000004621 00000 n +0000185419 00000 n +0000004663 00000 n +0000511015 00000 n +0000004710 00000 n +0000188230 00000 n +0000004754 00000 n +0000510922 00000 n +0000004801 00000 n +0000189051 00000 n +0000004854 00000 n +0000510829 00000 n +0000004901 00000 n +0000192531 00000 n +0000004944 00000 n +0000510750 00000 n +0000004991 00000 n +0000192971 00000 n +0000005037 00000 n +0000510632 00000 n +0000005081 00000 n +0000193791 00000 n +0000005128 00000 n +0000510553 00000 n +0000005175 00000 n +0000195790 00000 n +0000005236 00000 n +0000510421 00000 n +0000005283 00000 n +0000196041 00000 n +0000005342 00000 n +0000510342 00000 n +0000005391 00000 n +0000196733 00000 n +0000005429 00000 n +0000510263 00000 n +0000005479 00000 n +0000199067 00000 n +0000005534 00000 n +0000510170 00000 n +0000005581 00000 n +0000199508 00000 n +0000005651 00000 n +0000510077 00000 n +0000005698 00000 n +0000199948 00000 n +0000005771 00000 n +0000509998 00000 n +0000005818 00000 n +0000204716 00000 n +0000005897 00000 n +0000509865 00000 n +0000005938 00000 n +0000204840 00000 n +0000006000 00000 n +0000509786 00000 n +0000006044 00000 n +0000205217 00000 n +0000006101 00000 n +0000509707 00000 n +0000006145 00000 n +0000209404 00000 n +0000006195 00000 n +0000509574 00000 n +0000006236 00000 n +0000210098 00000 n +0000006297 00000 n +0000509456 00000 n +0000006341 00000 n +0000214794 00000 n +0000006393 00000 n +0000509338 00000 n +0000006440 00000 n +0000214919 00000 n +0000006486 00000 n +0000509259 00000 n +0000006536 00000 n +0000215425 00000 n +0000006571 00000 n +0000509180 00000 n +0000006621 00000 n +0000219136 00000 n +0000006670 00000 n +0000509062 00000 n +0000006717 00000 n +0000219451 00000 n +0000006758 00000 n +0000508983 00000 n +0000006808 00000 n +0000220018 00000 n +0000006843 00000 n +0000508904 00000 n +0000006893 00000 n +0000222907 00000 n +0000006942 00000 n +0000508771 00000 n +0000006986 00000 n +0000225395 00000 n +0000007037 00000 n +0000508692 00000 n +0000007084 00000 n +0000225646 00000 n +0000007132 00000 n +0000508599 00000 n +0000007179 00000 n +0000229875 00000 n +0000007240 00000 n +0000508506 00000 n +0000007287 00000 n +0000230382 00000 n +0000007340 00000 n +0000508413 00000 n +0000007387 00000 n +0000236589 00000 n +0000007453 00000 n +0000508320 00000 n +0000007500 00000 n +0000236840 00000 n +0000007552 00000 n +0000508227 00000 n +0000007599 00000 n +0000237220 00000 n +0000007664 00000 n +0000508134 00000 n +0000007711 00000 n +0000240067 00000 n +0000007759 00000 n +0000508041 00000 n +0000007806 00000 n +0000241020 00000 n +0000007867 00000 n +0000507948 00000 n +0000007914 00000 n +0000246271 00000 n +0000007966 00000 n +0000507816 00000 n +0000008013 00000 n +0000247221 00000 n +0000008079 00000 n +0000507737 00000 n +0000008129 00000 n +0000259583 00000 n +0000008177 00000 n +0000507658 00000 n +0000008227 00000 n +0000262454 00000 n +0000008276 00000 n +0000507565 00000 n +0000008323 00000 n +0000266339 00000 n +0000008374 00000 n +0000507472 00000 n +0000008421 00000 n +0000267419 00000 n +0000008485 00000 n +0000507379 00000 n +0000008532 00000 n +0000284793 00000 n +0000008585 00000 n +0000507246 00000 n +0000008632 00000 n +0000294418 00000 n +0000008698 00000 n +0000507167 00000 n +0000008748 00000 n +0000316196 00000 n +0000008793 00000 n +0000507074 00000 n +0000008843 00000 n +0000319973 00000 n +0000008883 00000 n +0000506981 00000 n +0000008933 00000 n +0000324107 00000 n +0000008977 00000 n +0000506888 00000 n +0000009027 00000 n +0000325751 00000 n +0000009067 00000 n +0000506795 00000 n +0000009117 00000 n +0000328712 00000 n +0000009160 00000 n +0000506702 00000 n +0000009210 00000 n +0000340374 00000 n +0000009254 00000 n +0000506609 00000 n +0000009304 00000 n +0000344390 00000 n +0000009366 00000 n +0000506516 00000 n +0000009416 00000 n +0000347864 00000 n +0000009468 00000 n +0000506423 00000 n +0000009518 00000 n +0000349568 00000 n +0000009571 00000 n +0000506330 00000 n +0000009621 00000 n +0000353013 00000 n +0000009660 00000 n +0000506237 00000 n +0000009710 00000 n +0000358174 00000 n +0000009763 00000 n +0000506144 00000 n +0000009813 00000 n +0000360193 00000 n +0000009858 00000 n +0000506051 00000 n +0000009908 00000 n +0000363928 00000 n +0000009963 00000 n +0000505958 00000 n +0000010013 00000 n +0000369046 00000 n +0000010050 00000 n +0000505879 00000 n +0000010100 00000 n +0000371672 00000 n +0000010150 00000 n +0000505786 00000 n +0000010197 00000 n +0000372935 00000 n +0000010249 00000 n +0000505693 00000 n +0000010296 00000 n +0000378760 00000 n +0000010361 00000 n +0000505600 00000 n +0000010408 00000 n +0000382156 00000 n +0000010466 00000 n +0000505507 00000 n +0000010513 00000 n +0000382535 00000 n +0000010584 00000 n +0000505414 00000 n +0000010631 00000 n +0000383485 00000 n +0000010681 00000 n +0000505321 00000 n +0000010728 00000 n +0000389395 00000 n +0000010791 00000 n +0000505228 00000 n +0000010838 00000 n +0000398436 00000 n +0000010888 00000 n +0000505110 00000 n +0000010935 00000 n +0000398561 00000 n +0000010998 00000 n +0000505031 00000 n +0000011048 00000 n +0000403577 00000 n +0000011088 00000 n +0000504938 00000 n +0000011138 00000 n +0000404211 00000 n +0000011173 00000 n +0000504845 00000 n +0000011223 00000 n +0000422393 00000 n +0000011265 00000 n +0000504766 00000 n +0000011315 00000 n +0000426662 00000 n +0000011368 00000 n +0000504648 00000 n +0000011412 00000 n +0000426788 00000 n +0000011446 00000 n +0000504530 00000 n +0000011493 00000 n +0000426976 00000 n +0000011566 00000 n +0000504451 00000 n +0000011616 00000 n +0000435788 00000 n +0000011661 00000 n +0000504372 00000 n +0000011711 00000 n +0000438944 00000 n +0000011765 00000 n +0000504279 00000 n +0000011812 00000 n +0000441922 00000 n +0000011863 00000 n +0000504186 00000 n +0000011910 00000 n +0000444569 00000 n +0000011949 00000 n +0000504093 00000 n +0000011996 00000 n +0000445264 00000 n +0000012046 00000 n +0000503961 00000 n +0000012093 00000 n +0000445705 00000 n +0000012146 00000 n +0000503882 00000 n +0000012196 00000 n +0000448775 00000 n +0000012238 00000 n +0000503789 00000 n +0000012288 00000 n +0000449970 00000 n +0000012330 00000 n +0000503710 00000 n +0000012380 00000 n +0000450603 00000 n +0000012422 00000 n +0000503631 00000 n +0000012469 00000 n +0000456653 00000 n +0000012537 00000 n +0000503498 00000 n +0000012578 00000 n +0000456777 00000 n +0000012639 00000 n +0000503419 00000 n +0000012683 00000 n +0000460480 00000 n +0000012728 00000 n +0000503287 00000 n +0000012772 00000 n +0000461817 00000 n +0000012835 00000 n +0000503208 00000 n +0000012882 00000 n +0000462897 00000 n +0000012931 00000 n +0000503129 00000 n +0000012978 00000 n +0000463406 00000 n +0000013030 00000 n +0000503050 00000 n +0000013074 00000 n +0000467760 00000 n +0000013122 00000 n +0000502917 00000 n +0000013163 00000 n +0000467885 00000 n +0000013209 00000 n +0000502799 00000 n +0000013253 00000 n +0000468011 00000 n +0000013293 00000 n +0000502734 00000 n +0000013340 00000 n +0000468199 00000 n +0000013419 00000 n +0000502641 00000 n +0000013463 00000 n +0000468513 00000 n +0000013531 00000 n +0000502562 00000 n +0000013575 00000 n +0000472636 00000 n +0000013621 00000 n +0000502442 00000 n +0000013663 00000 n +0000472760 00000 n +0000013705 00000 n +0000502324 00000 n +0000013750 00000 n +0000472886 00000 n +0000013791 00000 n +0000502259 00000 n +0000013839 00000 n +0000476032 00000 n +0000013901 00000 n +0000502127 00000 n +0000013946 00000 n +0000476157 00000 n +0000013997 00000 n +0000502023 00000 n +0000014045 00000 n +0000476283 00000 n +0000014099 00000 n +0000501944 00000 n +0000014150 00000 n +0000476600 00000 n +0000014190 00000 n +0000501865 00000 n +0000014241 00000 n +0000476853 00000 n +0000014280 00000 n +0000501733 00000 n +0000014325 00000 n +0000476979 00000 n +0000014383 00000 n +0000501668 00000 n +0000014431 00000 n +0000483262 00000 n +0000014479 00000 n +0000501536 00000 n +0000014524 00000 n +0000483388 00000 n +0000014587 00000 n +0000501457 00000 n +0000014635 00000 n +0000483895 00000 n +0000014691 00000 n +0000501378 00000 n +0000014739 00000 n +0000484021 00000 n +0000014771 00000 n +0000501285 00000 n +0000014816 00000 n +0000486165 00000 n +0000014845 00000 n +0000501192 00000 n +0000014890 00000 n +0000486669 00000 n +0000014928 00000 n +0000501099 00000 n +0000014973 00000 n +0000486983 00000 n +0000015035 00000 n +0000501006 00000 n +0000015080 00000 n +0000489130 00000 n +0000015145 00000 n +0000500913 00000 n +0000015190 00000 n +0000489697 00000 n +0000015231 00000 n +0000500820 00000 n +0000015276 00000 n +0000490072 00000 n +0000015316 00000 n +0000500727 00000 n +0000015361 00000 n +0000492275 00000 n +0000015395 00000 n +0000500634 00000 n +0000015440 00000 n +0000492777 00000 n +0000015482 00000 n +0000500516 00000 n +0000015527 00000 n +0000492966 00000 n +0000015590 00000 n +0000500437 00000 n +0000015638 00000 n +0000493154 00000 n +0000015680 00000 n +0000500344 00000 n +0000015728 00000 n +0000494028 00000 n +0000015781 00000 n +0000500265 00000 n +0000015829 00000 n +0000016187 00000 n +0000015861 00000 n +0000016368 00000 n +0000016429 00000 n +0000016307 00000 n +0000016490 00000 n +0000015913 00000 n +0000016166 00000 n +0000497185 00000 n +0000497281 00000 n +0000016845 00000 n +0000016906 00000 n +0000016725 00000 n +0000016562 00000 n +0000016705 00000 n +0000017398 00000 n +0000017459 00000 n +0000017278 00000 n +0000016947 00000 n +0000017257 00000 n +0000497092 00000 n +0000017827 00000 n +0000017888 00000 n +0000017707 00000 n +0000017544 00000 n +0000017687 00000 n +0000087590 00000 n +0000087902 00000 n +0000088343 00000 n +0000089613 00000 n +0000092349 00000 n +0000092728 00000 n +0000093169 00000 n +0000097687 00000 n +0000099147 00000 n +0000101876 00000 n +0000102317 00000 n +0000102693 00000 n +0000103258 00000 n +0000106333 00000 n +0000106835 00000 n +0000109707 00000 n +0000109831 00000 n +0000110334 00000 n +0000110585 00000 n +0000110899 00000 n +0000112316 00000 n +0000115279 00000 n +0000115465 00000 n +0000115588 00000 n +0000115838 00000 n +0000117719 00000 n +0000121292 00000 n +0000121859 00000 n +0000121984 00000 n +0000122172 00000 n +0000128793 00000 n +0000148725 00000 n +0000151438 00000 n +0000151563 00000 n +0000152065 00000 n +0000153073 00000 n +0000156372 00000 n +0000168140 00000 n +0000033970 00000 n +0000034091 00000 n +0000021932 00000 n +0000017929 00000 n +0000021910 00000 n +0000022656 00000 n +0000497000 00000 n +0000022807 00000 n +0000022959 00000 n +0000023109 00000 n +0000023259 00000 n +0000023408 00000 n +0000023557 00000 n +0000023706 00000 n +0000023855 00000 n +0000024005 00000 n +0000024156 00000 n +0000024307 00000 n +0000024458 00000 n +0000024609 00000 n +0000024760 00000 n +0000024911 00000 n +0000025062 00000 n +0000025212 00000 n +0000025363 00000 n +0000025514 00000 n +0000025665 00000 n +0000025816 00000 n +0000025967 00000 n +0000026118 00000 n +0000026269 00000 n +0000026420 00000 n +0000026571 00000 n +0000026722 00000 n +0000026873 00000 n +0000027024 00000 n +0000027175 00000 n +0000027326 00000 n +0000027478 00000 n +0000027628 00000 n +0000027778 00000 n +0000027929 00000 n +0000028080 00000 n +0000028231 00000 n +0000028382 00000 n +0000028533 00000 n +0000028684 00000 n +0000028835 00000 n +0000028986 00000 n +0000029138 00000 n +0000029290 00000 n +0000029457 00000 n +0000029624 00000 n +0000029775 00000 n +0000029926 00000 n +0000030077 00000 n +0000030228 00000 n +0000030379 00000 n +0000030530 00000 n +0000030684 00000 n +0000030838 00000 n +0000030989 00000 n +0000031140 00000 n +0000031291 00000 n +0000031442 00000 n +0000031605 00000 n +0000031768 00000 n +0000031927 00000 n +0000032086 00000 n +0000032237 00000 n +0000032388 00000 n +0000032540 00000 n +0000032692 00000 n +0000032854 00000 n +0000033016 00000 n +0000033170 00000 n +0000033324 00000 n +0000033496 00000 n +0000033668 00000 n +0000033819 00000 n +0000496906 00000 n +0000168900 00000 n +0000169216 00000 n +0000171933 00000 n +0000172248 00000 n +0000172436 00000 n +0000173196 00000 n +0000176400 00000 n +0000177159 00000 n +0000177409 00000 n +0000180690 00000 n +0000181190 00000 n +0000184094 00000 n +0000185357 00000 n +0000188168 00000 n +0000188989 00000 n +0000187726 00000 n +0000192909 00000 n +0000193729 00000 n +0000195728 00000 n +0000195979 00000 n +0000196671 00000 n +0000199005 00000 n +0000199446 00000 n +0000199886 00000 n +0000204654 00000 n +0000204779 00000 n +0000205155 00000 n +0000209342 00000 n +0000210036 00000 n +0000212375 00000 n +0000214857 00000 n +0000215363 00000 n +0000219074 00000 n +0000219389 00000 n +0000219956 00000 n +0000222845 00000 n +0000225333 00000 n +0000225584 00000 n +0000229813 00000 n +0000230320 00000 n +0000051164 00000 n +0000051225 00000 n +0000038159 00000 n +0000034202 00000 n +0000038137 00000 n +0000038931 00000 n +0000039082 00000 n +0000039233 00000 n +0000039384 00000 n +0000039535 00000 n +0000039686 00000 n +0000039837 00000 n +0000039986 00000 n +0000040136 00000 n +0000040287 00000 n +0000040438 00000 n +0000040589 00000 n +0000040740 00000 n +0000040891 00000 n +0000041042 00000 n +0000041193 00000 n +0000041344 00000 n +0000041495 00000 n +0000041646 00000 n +0000041797 00000 n +0000041948 00000 n +0000042098 00000 n +0000042248 00000 n +0000042402 00000 n +0000042556 00000 n +0000042707 00000 n +0000042858 00000 n +0000043009 00000 n +0000043160 00000 n +0000043311 00000 n +0000043462 00000 n +0000043612 00000 n +0000043763 00000 n +0000043913 00000 n +0000044063 00000 n +0000044212 00000 n +0000044362 00000 n +0000044514 00000 n +0000044666 00000 n +0000044818 00000 n +0000044970 00000 n +0000045122 00000 n +0000045274 00000 n +0000045426 00000 n +0000045578 00000 n +0000045729 00000 n +0000045881 00000 n +0000046033 00000 n +0000046185 00000 n +0000046337 00000 n +0000046489 00000 n +0000046641 00000 n +0000046793 00000 n +0000046945 00000 n +0000047097 00000 n +0000047251 00000 n +0000047405 00000 n +0000047557 00000 n +0000047709 00000 n +0000047883 00000 n +0000048057 00000 n +0000048224 00000 n +0000048391 00000 n +0000048543 00000 n +0000048695 00000 n +0000048847 00000 n +0000048999 00000 n +0000049151 00000 n +0000049303 00000 n +0000049455 00000 n +0000049607 00000 n +0000049758 00000 n +0000049909 00000 n +0000050082 00000 n +0000050255 00000 n +0000050406 00000 n +0000050558 00000 n +0000050709 00000 n +0000050860 00000 n +0000051012 00000 n +0000236528 00000 n +0000236778 00000 n +0000237158 00000 n +0000234681 00000 n +0000240958 00000 n +0000243440 00000 n +0000247160 00000 n +0000259522 00000 n +0000262392 00000 n +0000266277 00000 n +0000267357 00000 n +0000284731 00000 n +0000294356 00000 n +0000316134 00000 n +0000319913 00000 n +0000322311 00000 n +0000325689 00000 n +0000328650 00000 n +0000340312 00000 n +0000344328 00000 n +0000343191 00000 n +0000349506 00000 n +0000352951 00000 n +0000355775 00000 n +0000360131 00000 n +0000363866 00000 n +0000368984 00000 n +0000371610 00000 n +0000372874 00000 n +0000378698 00000 n +0000382094 00000 n +0000382473 00000 n +0000383423 00000 n +0000387594 00000 n +0000396577 00000 n +0000398499 00000 n +0000403517 00000 n +0000404149 00000 n +0000422332 00000 n +0000426600 00000 n +0000068362 00000 n +0000068423 00000 n +0000054905 00000 n +0000051323 00000 n +0000054883 00000 n +0000055734 00000 n +0000055921 00000 n +0000056108 00000 n +0000056260 00000 n +0000056412 00000 n +0000056563 00000 n +0000056714 00000 n +0000056865 00000 n +0000057017 00000 n +0000057169 00000 n +0000057321 00000 n +0000057472 00000 n +0000057623 00000 n +0000057775 00000 n +0000057927 00000 n +0000058077 00000 n +0000058228 00000 n +0000058395 00000 n +0000058562 00000 n +0000058714 00000 n +0000058866 00000 n +0000059017 00000 n +0000059168 00000 n +0000059319 00000 n +0000059471 00000 n +0000059624 00000 n +0000059777 00000 n +0000059941 00000 n +0000060105 00000 n +0000060258 00000 n +0000060411 00000 n +0000060574 00000 n +0000060737 00000 n +0000060890 00000 n +0000061043 00000 n +0000061195 00000 n +0000061347 00000 n +0000061510 00000 n +0000061673 00000 n +0000061826 00000 n +0000061979 00000 n +0000062132 00000 n +0000062285 00000 n +0000062438 00000 n +0000062591 00000 n +0000062748 00000 n +0000062905 00000 n +0000063076 00000 n +0000063247 00000 n +0000063410 00000 n +0000063573 00000 n +0000063731 00000 n +0000063889 00000 n +0000064044 00000 n +0000064199 00000 n +0000064357 00000 n +0000064515 00000 n +0000064688 00000 n +0000064861 00000 n +0000065047 00000 n +0000065233 00000 n +0000065386 00000 n +0000065539 00000 n +0000065692 00000 n +0000065845 00000 n +0000065998 00000 n +0000066151 00000 n +0000066304 00000 n +0000066457 00000 n +0000066628 00000 n +0000066799 00000 n +0000066952 00000 n +0000067105 00000 n +0000067258 00000 n +0000067411 00000 n +0000067563 00000 n +0000067716 00000 n +0000067868 00000 n +0000068020 00000 n +0000068191 00000 n +0000497399 00000 n +0000426725 00000 n +0000426913 00000 n +0000435727 00000 n +0000438881 00000 n +0000441859 00000 n +0000444506 00000 n +0000445201 00000 n +0000445642 00000 n +0000448712 00000 n +0000449907 00000 n +0000450540 00000 n +0000456590 00000 n +0000456716 00000 n +0000457792 00000 n +0000461754 00000 n +0000462834 00000 n +0000463343 00000 n +0000467697 00000 n +0000467823 00000 n +0000467948 00000 n +0000468136 00000 n +0000468450 00000 n +0000472573 00000 n +0000472699 00000 n +0000472823 00000 n +0000475969 00000 n +0000476094 00000 n +0000476220 00000 n +0000476537 00000 n +0000476790 00000 n +0000476916 00000 n +0000483200 00000 n +0000483325 00000 n +0000483832 00000 n +0000492903 00000 n +0000493092 00000 n +0000493280 00000 n +0000085407 00000 n +0000085470 00000 n +0000072663 00000 n +0000068521 00000 n +0000072640 00000 n +0000073483 00000 n +0000073635 00000 n +0000073787 00000 n +0000073982 00000 n +0000074177 00000 n +0000074330 00000 n +0000074483 00000 n +0000074635 00000 n +0000074787 00000 n +0000074940 00000 n +0000075093 00000 n +0000075254 00000 n +0000075415 00000 n +0000075568 00000 n +0000075721 00000 n +0000075873 00000 n +0000076025 00000 n +0000076177 00000 n +0000076329 00000 n +0000076482 00000 n +0000076635 00000 n +0000076788 00000 n +0000076941 00000 n +0000077094 00000 n +0000077247 00000 n +0000077400 00000 n +0000077553 00000 n +0000077720 00000 n +0000077887 00000 n +0000078040 00000 n +0000078193 00000 n +0000078346 00000 n +0000078499 00000 n +0000078652 00000 n +0000078805 00000 n +0000078977 00000 n +0000079149 00000 n +0000079302 00000 n +0000079455 00000 n +0000079608 00000 n +0000079761 00000 n +0000079913 00000 n +0000080066 00000 n +0000080219 00000 n +0000080372 00000 n +0000080525 00000 n +0000080678 00000 n +0000080830 00000 n +0000080982 00000 n +0000081135 00000 n +0000081288 00000 n +0000081440 00000 n +0000081593 00000 n +0000081768 00000 n +0000081943 00000 n +0000082118 00000 n +0000082293 00000 n +0000082446 00000 n +0000082599 00000 n +0000082752 00000 n +0000082905 00000 n +0000083058 00000 n +0000083211 00000 n +0000083373 00000 n +0000083535 00000 n +0000083696 00000 n +0000083857 00000 n +0000084010 00000 n +0000084163 00000 n +0000084316 00000 n +0000084469 00000 n +0000084632 00000 n +0000084796 00000 n +0000084948 00000 n +0000085101 00000 n +0000085254 00000 n +0000087713 00000 n +0000087774 00000 n +0000087838 00000 n +0000088025 00000 n +0000088088 00000 n +0000088152 00000 n +0000088216 00000 n +0000088280 00000 n +0000088467 00000 n +0000088530 00000 n +0000088592 00000 n +0000088656 00000 n +0000088720 00000 n +0000088784 00000 n +0000088848 00000 n +0000088911 00000 n +0000088975 00000 n +0000089039 00000 n +0000089103 00000 n +0000089167 00000 n +0000089231 00000 n +0000089295 00000 n +0000089358 00000 n +0000089421 00000 n +0000089485 00000 n +0000089549 00000 n +0000089736 00000 n +0000089799 00000 n +0000089862 00000 n +0000089924 00000 n +0000087527 00000 n +0000089986 00000 n +0000087403 00000 n +0000085569 00000 n +0000087380 00000 n +0000496816 00000 n +0000092033 00000 n +0000092096 00000 n +0000092160 00000 n +0000092223 00000 n +0000092286 00000 n +0000092473 00000 n +0000092536 00000 n +0000092600 00000 n +0000092664 00000 n +0000092852 00000 n +0000092915 00000 n +0000092978 00000 n +0000093042 00000 n +0000093106 00000 n +0000093293 00000 n +0000093356 00000 n +0000093417 00000 n +0000093480 00000 n +0000096668 00000 n +0000096731 00000 n +0000096795 00000 n +0000096859 00000 n +0000096922 00000 n +0000096986 00000 n +0000097050 00000 n +0000097114 00000 n +0000091970 00000 n +0000093544 00000 n +0000091846 00000 n +0000090112 00000 n +0000091823 00000 n +0000496721 00000 n +0000097178 00000 n +0000097241 00000 n +0000097305 00000 n +0000097369 00000 n +0000097433 00000 n +0000097496 00000 n +0000097560 00000 n +0000097624 00000 n +0000097811 00000 n +0000097874 00000 n +0000097938 00000 n +0000098002 00000 n +0000098065 00000 n +0000098129 00000 n +0000098193 00000 n +0000098256 00000 n +0000098320 00000 n +0000098384 00000 n +0000098448 00000 n +0000098512 00000 n +0000098575 00000 n +0000098638 00000 n +0000098702 00000 n +0000098766 00000 n +0000098829 00000 n +0000098893 00000 n +0000098956 00000 n +0000099019 00000 n +0000099083 00000 n +0000099271 00000 n +0000099333 00000 n +0000096605 00000 n +0000099397 00000 n +0000096104 00000 n +0000093684 00000 n +0000096081 00000 n +0000096258 00000 n +0000096452 00000 n +0000101749 00000 n +0000101812 00000 n +0000101999 00000 n +0000102061 00000 n +0000102125 00000 n +0000102189 00000 n +0000102253 00000 n +0000102441 00000 n +0000102504 00000 n +0000102567 00000 n +0000102631 00000 n +0000102816 00000 n +0000102879 00000 n +0000102942 00000 n +0000103006 00000 n +0000103068 00000 n +0000103131 00000 n +0000103195 00000 n +0000105825 00000 n +0000105888 00000 n +0000105952 00000 n +0000106016 00000 n +0000101686 00000 n +0000103319 00000 n +0000101376 00000 n +0000099510 00000 n +0000101353 00000 n +0000101521 00000 n +0000106080 00000 n +0000106142 00000 n +0000106206 00000 n +0000106270 00000 n +0000106457 00000 n +0000106519 00000 n +0000106582 00000 n +0000106645 00000 n +0000106708 00000 n +0000106771 00000 n +0000106959 00000 n +0000107022 00000 n +0000107086 00000 n +0000107149 00000 n +0000107213 00000 n +0000105700 00000 n +0000107277 00000 n +0000105576 00000 n +0000103431 00000 n +0000105553 00000 n +0000497523 00000 n +0000107775 00000 n +0000107838 00000 n +0000107651 00000 n +0000107389 00000 n +0000107629 00000 n +0000109953 00000 n +0000110016 00000 n +0000110079 00000 n +0000110143 00000 n +0000110206 00000 n +0000110270 00000 n +0000110458 00000 n +0000110521 00000 n +0000110708 00000 n +0000110771 00000 n +0000110835 00000 n +0000111023 00000 n +0000109644 00000 n +0000111086 00000 n +0000109520 00000 n +0000107911 00000 n +0000109497 00000 n +0000112440 00000 n +0000112503 00000 n +0000112567 00000 n +0000112630 00000 n +0000112693 00000 n +0000112757 00000 n +0000112820 00000 n +0000112884 00000 n +0000112947 00000 n +0000113011 00000 n +0000113074 00000 n +0000113138 00000 n +0000113201 00000 n +0000113265 00000 n +0000113328 00000 n +0000113392 00000 n +0000113455 00000 n +0000113519 00000 n +0000113582 00000 n +0000113646 00000 n +0000113709 00000 n +0000112253 00000 n +0000113773 00000 n +0000112129 00000 n +0000111211 00000 n +0000112107 00000 n +0000115403 00000 n +0000115712 00000 n +0000115775 00000 n +0000115963 00000 n +0000116026 00000 n +0000116090 00000 n +0000116154 00000 n +0000115216 00000 n +0000116217 00000 n +0000115092 00000 n +0000113885 00000 n +0000115069 00000 n +0000117844 00000 n +0000117907 00000 n +0000117971 00000 n +0000118034 00000 n +0000118097 00000 n +0000118160 00000 n +0000117656 00000 n +0000118223 00000 n +0000117532 00000 n +0000116356 00000 n +0000117509 00000 n +0000120911 00000 n +0000120974 00000 n +0000121038 00000 n +0000121100 00000 n +0000121164 00000 n +0000000000 00000 n +0000121228 00000 n +0000121416 00000 n +0000121479 00000 n +0000121542 00000 n +0000121605 00000 n +0000121668 00000 n +0000121731 00000 n +0000121795 00000 n +0000122109 00000 n +0000122296 00000 n +0000122359 00000 n +0000122421 00000 n +0000122483 00000 n +0000122546 00000 n +0000122610 00000 n +0000122674 00000 n +0000124788 00000 n +0000124851 00000 n +0000124915 00000 n +0000124979 00000 n +0000125043 00000 n +0000125107 00000 n +0000125171 00000 n +0000125235 00000 n +0000125299 00000 n +0000125363 00000 n +0000125427 00000 n +0000125491 00000 n +0000125555 00000 n +0000125619 00000 n +0000125683 00000 n +0000120785 00000 n +0000120848 00000 n +0000122737 00000 n +0000120128 00000 n +0000118336 00000 n +0000120105 00000 n +0000120291 00000 n +0000120459 00000 n +0000120621 00000 n +0000497648 00000 n +0000125747 00000 n +0000125810 00000 n +0000125873 00000 n +0000125936 00000 n +0000126000 00000 n +0000126063 00000 n +0000126126 00000 n +0000126189 00000 n +0000126253 00000 n +0000126317 00000 n +0000126380 00000 n +0000126443 00000 n +0000126506 00000 n +0000126569 00000 n +0000126633 00000 n +0000126697 00000 n +0000126760 00000 n +0000126824 00000 n +0000126888 00000 n +0000126951 00000 n +0000127015 00000 n +0000127079 00000 n +0000127143 00000 n +0000127206 00000 n +0000127270 00000 n +0000127334 00000 n +0000127398 00000 n +0000127462 00000 n +0000127525 00000 n +0000127589 00000 n +0000127652 00000 n +0000127715 00000 n +0000127778 00000 n +0000127842 00000 n +0000127906 00000 n +0000127969 00000 n +0000128032 00000 n +0000128095 00000 n +0000128159 00000 n +0000128222 00000 n +0000128286 00000 n +0000128350 00000 n +0000128414 00000 n +0000128477 00000 n +0000128539 00000 n +0000128602 00000 n +0000128665 00000 n +0000128729 00000 n +0000130758 00000 n +0000124662 00000 n +0000124725 00000 n +0000128855 00000 n +0000124538 00000 n +0000122889 00000 n +0000124515 00000 n +0000496623 00000 n +0000130821 00000 n +0000130883 00000 n +0000130945 00000 n +0000131008 00000 n +0000131072 00000 n +0000131136 00000 n +0000131200 00000 n +0000131263 00000 n +0000131326 00000 n +0000131390 00000 n +0000131454 00000 n +0000131518 00000 n +0000131582 00000 n +0000131645 00000 n +0000131708 00000 n +0000131771 00000 n +0000131835 00000 n +0000131899 00000 n +0000131962 00000 n +0000132025 00000 n +0000132089 00000 n +0000132153 00000 n +0000132217 00000 n +0000132281 00000 n +0000132345 00000 n +0000132409 00000 n +0000132473 00000 n +0000132536 00000 n +0000132599 00000 n +0000132662 00000 n +0000132726 00000 n +0000132790 00000 n +0000132854 00000 n +0000132917 00000 n +0000132980 00000 n +0000133044 00000 n +0000133108 00000 n +0000133172 00000 n +0000133236 00000 n +0000133300 00000 n +0000133364 00000 n +0000133428 00000 n +0000133492 00000 n +0000133556 00000 n +0000133620 00000 n +0000133684 00000 n +0000133748 00000 n +0000133811 00000 n +0000133874 00000 n +0000133937 00000 n +0000134000 00000 n +0000134063 00000 n +0000134126 00000 n +0000134190 00000 n +0000134253 00000 n +0000134316 00000 n +0000134379 00000 n +0000134442 00000 n +0000134505 00000 n +0000134567 00000 n +0000134630 00000 n +0000134693 00000 n +0000134757 00000 n +0000134820 00000 n +0000134883 00000 n +0000134947 00000 n +0000135011 00000 n +0000135075 00000 n +0000135139 00000 n +0000135203 00000 n +0000135266 00000 n +0000135330 00000 n +0000135393 00000 n +0000135456 00000 n +0000135519 00000 n +0000135583 00000 n +0000137431 00000 n +0000137494 00000 n +0000137557 00000 n +0000137619 00000 n +0000130569 00000 n +0000130632 00000 n +0000135647 00000 n +0000130445 00000 n +0000128968 00000 n +0000130422 00000 n +0000496521 00000 n +0000137683 00000 n +0000137746 00000 n +0000137809 00000 n +0000137872 00000 n +0000137936 00000 n +0000137999 00000 n +0000138063 00000 n +0000138126 00000 n +0000138189 00000 n +0000138252 00000 n +0000138315 00000 n +0000138379 00000 n +0000138442 00000 n +0000138505 00000 n +0000138568 00000 n +0000138632 00000 n +0000138695 00000 n +0000138758 00000 n +0000138821 00000 n +0000138885 00000 n +0000138948 00000 n +0000139011 00000 n +0000139074 00000 n +0000139138 00000 n +0000139201 00000 n +0000139264 00000 n +0000139328 00000 n +0000139391 00000 n +0000139455 00000 n +0000139518 00000 n +0000139581 00000 n +0000139644 00000 n +0000139708 00000 n +0000139771 00000 n +0000139834 00000 n +0000139897 00000 n +0000139961 00000 n +0000140024 00000 n +0000140087 00000 n +0000140150 00000 n +0000140214 00000 n +0000140277 00000 n +0000140341 00000 n +0000140404 00000 n +0000140468 00000 n +0000140532 00000 n +0000140596 00000 n +0000140659 00000 n +0000140723 00000 n +0000143617 00000 n +0000143681 00000 n +0000143745 00000 n +0000143809 00000 n +0000143873 00000 n +0000143936 00000 n +0000143999 00000 n +0000137368 00000 n +0000140787 00000 n +0000137244 00000 n +0000135815 00000 n +0000137221 00000 n +0000144062 00000 n +0000144125 00000 n +0000144189 00000 n +0000144253 00000 n +0000144317 00000 n +0000144381 00000 n +0000144445 00000 n +0000144507 00000 n +0000144571 00000 n +0000144635 00000 n +0000144699 00000 n +0000144763 00000 n +0000144827 00000 n +0000144891 00000 n +0000144955 00000 n +0000145019 00000 n +0000145083 00000 n +0000145146 00000 n +0000145209 00000 n +0000145272 00000 n +0000145336 00000 n +0000145400 00000 n +0000145464 00000 n +0000145528 00000 n +0000145592 00000 n +0000145656 00000 n +0000145719 00000 n +0000145783 00000 n +0000145847 00000 n +0000145911 00000 n +0000145975 00000 n +0000146039 00000 n +0000146103 00000 n +0000146166 00000 n +0000146229 00000 n +0000146292 00000 n +0000146356 00000 n +0000146419 00000 n +0000146482 00000 n +0000146546 00000 n +0000146609 00000 n +0000143554 00000 n +0000146672 00000 n +0000143223 00000 n +0000140928 00000 n +0000143200 00000 n +0000143368 00000 n +0000148094 00000 n +0000148157 00000 n +0000148221 00000 n +0000148284 00000 n +0000148348 00000 n +0000148412 00000 n +0000148474 00000 n +0000148538 00000 n +0000148598 00000 n +0000148662 00000 n +0000148850 00000 n +0000148913 00000 n +0000148977 00000 n +0000149040 00000 n +0000149103 00000 n +0000148031 00000 n +0000149166 00000 n +0000147907 00000 n +0000146799 00000 n +0000147884 00000 n +0000151686 00000 n +0000151749 00000 n +0000151812 00000 n +0000151876 00000 n +0000151940 00000 n +0000152003 00000 n +0000152190 00000 n +0000152252 00000 n +0000152316 00000 n +0000152379 00000 n +0000152442 00000 n +0000152505 00000 n +0000152568 00000 n +0000152631 00000 n +0000152694 00000 n +0000152758 00000 n +0000152822 00000 n +0000152885 00000 n +0000152948 00000 n +0000153011 00000 n +0000155928 00000 n +0000486102 00000 n +0000151375 00000 n +0000153135 00000 n +0000151251 00000 n +0000149292 00000 n +0000151228 00000 n +0000497773 00000 n +0000155991 00000 n +0000156054 00000 n +0000156118 00000 n +0000156181 00000 n +0000156244 00000 n +0000156308 00000 n +0000156496 00000 n +0000156559 00000 n +0000156623 00000 n +0000156686 00000 n +0000156748 00000 n +0000156811 00000 n +0000156874 00000 n +0000156938 00000 n +0000157001 00000 n +0000157064 00000 n +0000157127 00000 n +0000157190 00000 n +0000157254 00000 n +0000157318 00000 n +0000157381 00000 n +0000159903 00000 n +0000159967 00000 n +0000160031 00000 n +0000160095 00000 n +0000160159 00000 n +0000155802 00000 n +0000157445 00000 n +0000155490 00000 n +0000153287 00000 n +0000155467 00000 n +0000155635 00000 n +0000160223 00000 n +0000160286 00000 n +0000160350 00000 n +0000160414 00000 n +0000160477 00000 n +0000160540 00000 n +0000160603 00000 n +0000160667 00000 n +0000160731 00000 n +0000160795 00000 n +0000160859 00000 n +0000160922 00000 n +0000160986 00000 n +0000161050 00000 n +0000161113 00000 n +0000161176 00000 n +0000161239 00000 n +0000161302 00000 n +0000161365 00000 n +0000161428 00000 n +0000161492 00000 n +0000161555 00000 n +0000161619 00000 n +0000161682 00000 n +0000161745 00000 n +0000161809 00000 n +0000161873 00000 n +0000161937 00000 n +0000162000 00000 n +0000162064 00000 n +0000162128 00000 n +0000162192 00000 n +0000162255 00000 n +0000162318 00000 n +0000162381 00000 n +0000162445 00000 n +0000162507 00000 n +0000162570 00000 n +0000162633 00000 n +0000162696 00000 n +0000162759 00000 n +0000162822 00000 n +0000162885 00000 n +0000162949 00000 n +0000163013 00000 n +0000163076 00000 n +0000164303 00000 n +0000159777 00000 n +0000159840 00000 n +0000163139 00000 n +0000159464 00000 n +0000157571 00000 n +0000159441 00000 n +0000159609 00000 n +0000164366 00000 n +0000164240 00000 n +0000164429 00000 n +0000164116 00000 n +0000163266 00000 n +0000164094 00000 n +0000165558 00000 n +0000165621 00000 n +0000168013 00000 n +0000168076 00000 n +0000165432 00000 n +0000165495 00000 n +0000165684 00000 n +0000165308 00000 n +0000164543 00000 n +0000165286 00000 n +0000168265 00000 n +0000168328 00000 n +0000168392 00000 n +0000168456 00000 n +0000168519 00000 n +0000168582 00000 n +0000168646 00000 n +0000168709 00000 n +0000168773 00000 n +0000168836 00000 n +0000169025 00000 n +0000169088 00000 n +0000169152 00000 n +0000169341 00000 n +0000169404 00000 n +0000169467 00000 n +0000169530 00000 n +0000169593 00000 n +0000169657 00000 n +0000169721 00000 n +0000169784 00000 n +0000169847 00000 n +0000167950 00000 n +0000169911 00000 n +0000167638 00000 n +0000165784 00000 n +0000167615 00000 n +0000167783 00000 n +0000172058 00000 n +0000172121 00000 n +0000172184 00000 n +0000172373 00000 n +0000172561 00000 n +0000172624 00000 n +0000172688 00000 n +0000172752 00000 n +0000172816 00000 n +0000172879 00000 n +0000172942 00000 n +0000173006 00000 n +0000173070 00000 n +0000173133 00000 n +0000173320 00000 n +0000173383 00000 n +0000173446 00000 n +0000173510 00000 n +0000173574 00000 n +0000175956 00000 n +0000171807 00000 n +0000171870 00000 n +0000173637 00000 n +0000171683 00000 n +0000170077 00000 n +0000171660 00000 n +0000497898 00000 n +0000176019 00000 n +0000176083 00000 n +0000176146 00000 n +0000176210 00000 n +0000176274 00000 n +0000176337 00000 n +0000176525 00000 n +0000176588 00000 n +0000176651 00000 n +0000176715 00000 n +0000176779 00000 n +0000176842 00000 n +0000176905 00000 n +0000176968 00000 n +0000177032 00000 n +0000177095 00000 n +0000177283 00000 n +0000177346 00000 n +0000179863 00000 n +0000179926 00000 n +0000179989 00000 n +0000180053 00000 n +0000180117 00000 n +0000180180 00000 n +0000180244 00000 n +0000175893 00000 n +0000177471 00000 n +0000175576 00000 n +0000173776 00000 n +0000175553 00000 n +0000175721 00000 n +0000180307 00000 n +0000180370 00000 n +0000180434 00000 n +0000180498 00000 n +0000180562 00000 n +0000180626 00000 n +0000180814 00000 n +0000180876 00000 n +0000180938 00000 n +0000181001 00000 n +0000181064 00000 n +0000181127 00000 n +0000181315 00000 n +0000181378 00000 n +0000181442 00000 n +0000181505 00000 n +0000181569 00000 n +0000183968 00000 n +0000179737 00000 n +0000181633 00000 n +0000179613 00000 n +0000177597 00000 n +0000179590 00000 n +0000184031 00000 n +0000184217 00000 n +0000184280 00000 n +0000184344 00000 n +0000184407 00000 n +0000184469 00000 n +0000184532 00000 n +0000184596 00000 n +0000184659 00000 n +0000184722 00000 n +0000184785 00000 n +0000184849 00000 n +0000184913 00000 n +0000184977 00000 n +0000185040 00000 n +0000185104 00000 n +0000185167 00000 n +0000185230 00000 n +0000185294 00000 n +0000185482 00000 n +0000185545 00000 n +0000185609 00000 n +0000187788 00000 n +0000187851 00000 n +0000183842 00000 n +0000183905 00000 n +0000185671 00000 n +0000183718 00000 n +0000181759 00000 n +0000183695 00000 n +0000187915 00000 n +0000187978 00000 n +0000188041 00000 n +0000188104 00000 n +0000188293 00000 n +0000188356 00000 n +0000188420 00000 n +0000188483 00000 n +0000188547 00000 n +0000188611 00000 n +0000188675 00000 n +0000188738 00000 n +0000188800 00000 n +0000188862 00000 n +0000188925 00000 n +0000189114 00000 n +0000189176 00000 n +0000189240 00000 n +0000189303 00000 n +0000189367 00000 n +0000189431 00000 n +0000189495 00000 n +0000189559 00000 n +0000189622 00000 n +0000189686 00000 n +0000189749 00000 n +0000192340 00000 n +0000192403 00000 n +0000192467 00000 n +0000187663 00000 n +0000189812 00000 n +0000187539 00000 n +0000185811 00000 n +0000187516 00000 n +0000192594 00000 n +0000192657 00000 n +0000192720 00000 n +0000192783 00000 n +0000192845 00000 n +0000193034 00000 n +0000193097 00000 n +0000193160 00000 n +0000193223 00000 n +0000193287 00000 n +0000193349 00000 n +0000193413 00000 n +0000193477 00000 n +0000193540 00000 n +0000193603 00000 n +0000193666 00000 n +0000193854 00000 n +0000193916 00000 n +0000193979 00000 n +0000192277 00000 n +0000194042 00000 n +0000191808 00000 n +0000189952 00000 n +0000191785 00000 n +0000191962 00000 n +0000192115 00000 n +0000195853 00000 n +0000195916 00000 n +0000196104 00000 n +0000196167 00000 n +0000196230 00000 n +0000196293 00000 n +0000196356 00000 n +0000196419 00000 n +0000196482 00000 n +0000196545 00000 n +0000196607 00000 n +0000196796 00000 n +0000196858 00000 n +0000195665 00000 n +0000196921 00000 n +0000195541 00000 n +0000194181 00000 n +0000195518 00000 n +0000498023 00000 n +0000198878 00000 n +0000198941 00000 n +0000199130 00000 n +0000199193 00000 n +0000199256 00000 n +0000199320 00000 n +0000199383 00000 n +0000199570 00000 n +0000199633 00000 n +0000199697 00000 n +0000199760 00000 n +0000199823 00000 n +0000200011 00000 n +0000200074 00000 n +0000200138 00000 n +0000200201 00000 n +0000200265 00000 n +0000200329 00000 n +0000198815 00000 n +0000200391 00000 n +0000198691 00000 n +0000197047 00000 n +0000198668 00000 n +0000201514 00000 n +0000201577 00000 n +0000201641 00000 n +0000201704 00000 n +0000201767 00000 n +0000201831 00000 n +0000201894 00000 n +0000201957 00000 n +0000202020 00000 n +0000201451 00000 n +0000202083 00000 n +0000201327 00000 n +0000200517 00000 n +0000201305 00000 n +0000202574 00000 n +0000202637 00000 n +0000202450 00000 n +0000202183 00000 n +0000202428 00000 n +0000204902 00000 n +0000204965 00000 n +0000205028 00000 n +0000205091 00000 n +0000205280 00000 n +0000205343 00000 n +0000205407 00000 n +0000205469 00000 n +0000205533 00000 n +0000205597 00000 n +0000205661 00000 n +0000205724 00000 n +0000205788 00000 n +0000205851 00000 n +0000205915 00000 n +0000205975 00000 n +0000206035 00000 n +0000206097 00000 n +0000206161 00000 n +0000206225 00000 n +0000206289 00000 n +0000206353 00000 n +0000204591 00000 n +0000206417 00000 n +0000204467 00000 n +0000202710 00000 n +0000204444 00000 n +0000206963 00000 n +0000207026 00000 n +0000206839 00000 n +0000206556 00000 n +0000206817 00000 n +0000209467 00000 n +0000209529 00000 n +0000209591 00000 n +0000209655 00000 n +0000209719 00000 n +0000209782 00000 n +0000209846 00000 n +0000209909 00000 n +0000209972 00000 n +0000210161 00000 n +0000210224 00000 n +0000210288 00000 n +0000210351 00000 n +0000210414 00000 n +0000209279 00000 n +0000210477 00000 n +0000208968 00000 n +0000207099 00000 n +0000208945 00000 n +0000209113 00000 n +0000498148 00000 n +0000214982 00000 n +0000215045 00000 n +0000212312 00000 n +0000212437 00000 n +0000212188 00000 n +0000210630 00000 n +0000212165 00000 n +0000215108 00000 n +0000215172 00000 n +0000215235 00000 n +0000215299 00000 n +0000215488 00000 n +0000215551 00000 n +0000215615 00000 n +0000215678 00000 n +0000215741 00000 n +0000215805 00000 n +0000215868 00000 n +0000215932 00000 n +0000215995 00000 n +0000216059 00000 n +0000216122 00000 n +0000216186 00000 n +0000216250 00000 n +0000216313 00000 n +0000216377 00000 n +0000216439 00000 n +0000216502 00000 n +0000216565 00000 n +0000216628 00000 n +0000216691 00000 n +0000216754 00000 n +0000216817 00000 n +0000216881 00000 n +0000216945 00000 n +0000214731 00000 n +0000217007 00000 n +0000214607 00000 n +0000212565 00000 n +0000214584 00000 n +0000218883 00000 n +0000218946 00000 n +0000219010 00000 n +0000219198 00000 n +0000219261 00000 n +0000219325 00000 n +0000219514 00000 n +0000219577 00000 n +0000219640 00000 n +0000219704 00000 n +0000219767 00000 n +0000219830 00000 n +0000219892 00000 n +0000220081 00000 n +0000220144 00000 n +0000220207 00000 n +0000220270 00000 n +0000220333 00000 n +0000220395 00000 n +0000218757 00000 n +0000218820 00000 n +0000220458 00000 n +0000218633 00000 n +0000217173 00000 n +0000218610 00000 n +0000222216 00000 n +0000222279 00000 n +0000222342 00000 n +0000222403 00000 n +0000222466 00000 n +0000222530 00000 n +0000222593 00000 n +0000222655 00000 n +0000222718 00000 n +0000222781 00000 n +0000222970 00000 n +0000223033 00000 n +0000223097 00000 n +0000223159 00000 n +0000223222 00000 n +0000223285 00000 n +0000222153 00000 n +0000223348 00000 n +0000222029 00000 n +0000220597 00000 n +0000222006 00000 n +0000496429 00000 n +0000225142 00000 n +0000225205 00000 n +0000225269 00000 n +0000225458 00000 n +0000225521 00000 n +0000225709 00000 n +0000225772 00000 n +0000225836 00000 n +0000225899 00000 n +0000225963 00000 n +0000226026 00000 n +0000226089 00000 n +0000226152 00000 n +0000225079 00000 n +0000226215 00000 n +0000224955 00000 n +0000223501 00000 n +0000224932 00000 n +0000229496 00000 n +0000229559 00000 n +0000229623 00000 n +0000229687 00000 n +0000229749 00000 n +0000229938 00000 n +0000230001 00000 n +0000230064 00000 n +0000230128 00000 n +0000230192 00000 n +0000230256 00000 n +0000230443 00000 n +0000230506 00000 n +0000230570 00000 n +0000230634 00000 n +0000230697 00000 n +0000230760 00000 n +0000230823 00000 n +0000230886 00000 n +0000230950 00000 n +0000231014 00000 n +0000231078 00000 n +0000231142 00000 n +0000231206 00000 n +0000231270 00000 n +0000231334 00000 n +0000231397 00000 n +0000231461 00000 n +0000231525 00000 n +0000231589 00000 n +0000231653 00000 n +0000231717 00000 n +0000231781 00000 n +0000231845 00000 n +0000231909 00000 n +0000231972 00000 n +0000232036 00000 n +0000232100 00000 n +0000232163 00000 n +0000232227 00000 n +0000234743 00000 n +0000234807 00000 n +0000234871 00000 n +0000234935 00000 n +0000234999 00000 n +0000235063 00000 n +0000229370 00000 n +0000229433 00000 n +0000232291 00000 n +0000228737 00000 n +0000226355 00000 n +0000228714 00000 n +0000228900 00000 n +0000229052 00000 n +0000229211 00000 n +0000498273 00000 n +0000235127 00000 n +0000235191 00000 n +0000235255 00000 n +0000235319 00000 n +0000235383 00000 n +0000235447 00000 n +0000235511 00000 n +0000235575 00000 n +0000235638 00000 n +0000235702 00000 n +0000235766 00000 n +0000235830 00000 n +0000235894 00000 n +0000235958 00000 n +0000236021 00000 n +0000236084 00000 n +0000236147 00000 n +0000236211 00000 n +0000236275 00000 n +0000236339 00000 n +0000236402 00000 n +0000236465 00000 n +0000236652 00000 n +0000236715 00000 n +0000236903 00000 n +0000236966 00000 n +0000237030 00000 n +0000237094 00000 n +0000237283 00000 n +0000237346 00000 n +0000237409 00000 n +0000237473 00000 n +0000234618 00000 n +0000237535 00000 n +0000234494 00000 n +0000232445 00000 n +0000234471 00000 n +0000240130 00000 n +0000240193 00000 n +0000240257 00000 n +0000240320 00000 n +0000240384 00000 n +0000240448 00000 n +0000240512 00000 n +0000240576 00000 n +0000240639 00000 n +0000240703 00000 n +0000240766 00000 n +0000240830 00000 n +0000240894 00000 n +0000241083 00000 n +0000241146 00000 n +0000241209 00000 n +0000241272 00000 n +0000241336 00000 n +0000241399 00000 n +0000241463 00000 n +0000241527 00000 n +0000241591 00000 n +0000241655 00000 n +0000241719 00000 n +0000241783 00000 n +0000241847 00000 n +0000241911 00000 n +0000241975 00000 n +0000242039 00000 n +0000242102 00000 n +0000242166 00000 n +0000242230 00000 n +0000242293 00000 n +0000242357 00000 n +0000242421 00000 n +0000242485 00000 n +0000242549 00000 n +0000242613 00000 n +0000242676 00000 n +0000242740 00000 n +0000242804 00000 n +0000242867 00000 n +0000242931 00000 n +0000242995 00000 n +0000243058 00000 n +0000243122 00000 n +0000243186 00000 n +0000243249 00000 n +0000243313 00000 n +0000243376 00000 n +0000246334 00000 n +0000246397 00000 n +0000246461 00000 n +0000246524 00000 n +0000240004 00000 n +0000243502 00000 n +0000239511 00000 n +0000237689 00000 n +0000239488 00000 n +0000239665 00000 n +0000239817 00000 n +0000246587 00000 n +0000246650 00000 n +0000246714 00000 n +0000246778 00000 n +0000246841 00000 n +0000246904 00000 n +0000246968 00000 n +0000247032 00000 n +0000247096 00000 n +0000247284 00000 n +0000247347 00000 n +0000247409 00000 n +0000247472 00000 n +0000247536 00000 n +0000247600 00000 n +0000247663 00000 n +0000247727 00000 n +0000247790 00000 n +0000247854 00000 n +0000247917 00000 n +0000247981 00000 n +0000248045 00000 n +0000248109 00000 n +0000248173 00000 n +0000248237 00000 n +0000248301 00000 n +0000248365 00000 n +0000248429 00000 n +0000248493 00000 n +0000248557 00000 n +0000248621 00000 n +0000248684 00000 n +0000248747 00000 n +0000248810 00000 n +0000251532 00000 n +0000246208 00000 n +0000248874 00000 n +0000246084 00000 n +0000243684 00000 n +0000246061 00000 n +0000251596 00000 n +0000251659 00000 n +0000251723 00000 n +0000251787 00000 n +0000251850 00000 n +0000251913 00000 n +0000251977 00000 n +0000252041 00000 n +0000252105 00000 n +0000252169 00000 n +0000252233 00000 n +0000252297 00000 n +0000252360 00000 n +0000252424 00000 n +0000252488 00000 n +0000252552 00000 n +0000252615 00000 n +0000252679 00000 n +0000252743 00000 n +0000252806 00000 n +0000252869 00000 n +0000252933 00000 n +0000252997 00000 n +0000253061 00000 n +0000253125 00000 n +0000253186 00000 n +0000253248 00000 n +0000253310 00000 n +0000253372 00000 n +0000253436 00000 n +0000253499 00000 n +0000253563 00000 n +0000253627 00000 n +0000253691 00000 n +0000253755 00000 n +0000253818 00000 n +0000253881 00000 n +0000256045 00000 n +0000256108 00000 n +0000256172 00000 n +0000256236 00000 n +0000256300 00000 n +0000256364 00000 n +0000256428 00000 n +0000256492 00000 n +0000256556 00000 n +0000251469 00000 n +0000253945 00000 n +0000251345 00000 n +0000249013 00000 n +0000251322 00000 n +0000256620 00000 n +0000256683 00000 n +0000256746 00000 n +0000256809 00000 n +0000256873 00000 n +0000256936 00000 n +0000256999 00000 n +0000257063 00000 n +0000259078 00000 n +0000259141 00000 n +0000259204 00000 n +0000259267 00000 n +0000259331 00000 n +0000259395 00000 n +0000255982 00000 n +0000257127 00000 n +0000255670 00000 n +0000254058 00000 n +0000255647 00000 n +0000255815 00000 n +0000259459 00000 n +0000259646 00000 n +0000259709 00000 n +0000259773 00000 n +0000259835 00000 n +0000259898 00000 n +0000259961 00000 n +0000260024 00000 n +0000260088 00000 n +0000260151 00000 n +0000260214 00000 n +0000260278 00000 n +0000260341 00000 n +0000260404 00000 n +0000259015 00000 n +0000260467 00000 n +0000258891 00000 n +0000257240 00000 n +0000258868 00000 n +0000498398 00000 n +0000262517 00000 n +0000262580 00000 n +0000262644 00000 n +0000262708 00000 n +0000262771 00000 n +0000265259 00000 n +0000265322 00000 n +0000265385 00000 n +0000265449 00000 n +0000265512 00000 n +0000265576 00000 n +0000265640 00000 n +0000265704 00000 n +0000262329 00000 n +0000262834 00000 n +0000262205 00000 n +0000260593 00000 n +0000262182 00000 n +0000265768 00000 n +0000265831 00000 n +0000265895 00000 n +0000265958 00000 n +0000266022 00000 n +0000266086 00000 n +0000266150 00000 n +0000266213 00000 n +0000266402 00000 n +0000266465 00000 n +0000266529 00000 n +0000266593 00000 n +0000266656 00000 n +0000266720 00000 n +0000266783 00000 n +0000266847 00000 n +0000266910 00000 n +0000266974 00000 n +0000267038 00000 n +0000267102 00000 n +0000267165 00000 n +0000267229 00000 n +0000267293 00000 n +0000267482 00000 n +0000267545 00000 n +0000267609 00000 n +0000267673 00000 n +0000267736 00000 n +0000267799 00000 n +0000267863 00000 n +0000267926 00000 n +0000267990 00000 n +0000268053 00000 n +0000268117 00000 n +0000268180 00000 n +0000268244 00000 n +0000268307 00000 n +0000268371 00000 n +0000268435 00000 n +0000268497 00000 n +0000268560 00000 n +0000268623 00000 n +0000268687 00000 n +0000268750 00000 n +0000270634 00000 n +0000270697 00000 n +0000265196 00000 n +0000268814 00000 n +0000264733 00000 n +0000262974 00000 n +0000264710 00000 n +0000264887 00000 n +0000265042 00000 n +0000270761 00000 n +0000270824 00000 n +0000270888 00000 n +0000270951 00000 n +0000271015 00000 n +0000271078 00000 n +0000271142 00000 n +0000271205 00000 n +0000271269 00000 n +0000271332 00000 n +0000271396 00000 n +0000271459 00000 n +0000271523 00000 n +0000271586 00000 n +0000271650 00000 n +0000271713 00000 n +0000271777 00000 n +0000271840 00000 n +0000271904 00000 n +0000271967 00000 n +0000272031 00000 n +0000272094 00000 n +0000272158 00000 n +0000272222 00000 n +0000272285 00000 n +0000272349 00000 n +0000272412 00000 n +0000272476 00000 n +0000272539 00000 n +0000272603 00000 n +0000272666 00000 n +0000272730 00000 n +0000272793 00000 n +0000272857 00000 n +0000272921 00000 n +0000272984 00000 n +0000273048 00000 n +0000273111 00000 n +0000273175 00000 n +0000273239 00000 n +0000273303 00000 n +0000273367 00000 n +0000273431 00000 n +0000273494 00000 n +0000273558 00000 n +0000273622 00000 n +0000273686 00000 n +0000273750 00000 n +0000273814 00000 n +0000273878 00000 n +0000273941 00000 n +0000274005 00000 n +0000274068 00000 n +0000274132 00000 n +0000274195 00000 n +0000274259 00000 n +0000274322 00000 n +0000274386 00000 n +0000274449 00000 n +0000274513 00000 n +0000274576 00000 n +0000274640 00000 n +0000274703 00000 n +0000274767 00000 n +0000274831 00000 n +0000274895 00000 n +0000274958 00000 n +0000275022 00000 n +0000275086 00000 n +0000275150 00000 n +0000275213 00000 n +0000275277 00000 n +0000275341 00000 n +0000275405 00000 n +0000275469 00000 n +0000275533 00000 n +0000275597 00000 n +0000275660 00000 n +0000275724 00000 n +0000275787 00000 n +0000275851 00000 n +0000275914 00000 n +0000275978 00000 n +0000276041 00000 n +0000276105 00000 n +0000276168 00000 n +0000276232 00000 n +0000276295 00000 n +0000276359 00000 n +0000276422 00000 n +0000276486 00000 n +0000276549 00000 n +0000276613 00000 n +0000276676 00000 n +0000276740 00000 n +0000276803 00000 n +0000276867 00000 n +0000276930 00000 n +0000276994 00000 n +0000277056 00000 n +0000277119 00000 n +0000277182 00000 n +0000277246 00000 n +0000277310 00000 n +0000277374 00000 n +0000277438 00000 n +0000277501 00000 n +0000277565 00000 n +0000277629 00000 n +0000277693 00000 n +0000277757 00000 n +0000277820 00000 n +0000277884 00000 n +0000277948 00000 n +0000278012 00000 n +0000278076 00000 n +0000278139 00000 n +0000278203 00000 n +0000278267 00000 n +0000278331 00000 n +0000280155 00000 n +0000280218 00000 n +0000280282 00000 n +0000280346 00000 n +0000280410 00000 n +0000280474 00000 n +0000280538 00000 n +0000280602 00000 n +0000270571 00000 n +0000278395 00000 n +0000270447 00000 n +0000268954 00000 n +0000270424 00000 n +0000280666 00000 n +0000280729 00000 n +0000280793 00000 n +0000280856 00000 n +0000280920 00000 n +0000280983 00000 n +0000281047 00000 n +0000281110 00000 n +0000281174 00000 n +0000281237 00000 n +0000281301 00000 n +0000281364 00000 n +0000281428 00000 n +0000281491 00000 n +0000281555 00000 n +0000281618 00000 n +0000281682 00000 n +0000281745 00000 n +0000281809 00000 n +0000281872 00000 n +0000281936 00000 n +0000281999 00000 n +0000282063 00000 n +0000282126 00000 n +0000282190 00000 n +0000282254 00000 n +0000282318 00000 n +0000282381 00000 n +0000282445 00000 n +0000282508 00000 n +0000282572 00000 n +0000282635 00000 n +0000282699 00000 n +0000282762 00000 n +0000282826 00000 n +0000282889 00000 n +0000282953 00000 n +0000283016 00000 n +0000283080 00000 n +0000283143 00000 n +0000283207 00000 n +0000283270 00000 n +0000283334 00000 n +0000283397 00000 n +0000283461 00000 n +0000283524 00000 n +0000283588 00000 n +0000283651 00000 n +0000283715 00000 n +0000283778 00000 n +0000283842 00000 n +0000283905 00000 n +0000283969 00000 n +0000284032 00000 n +0000284096 00000 n +0000284159 00000 n +0000284223 00000 n +0000284286 00000 n +0000284350 00000 n +0000284413 00000 n +0000284477 00000 n +0000284540 00000 n +0000284604 00000 n +0000284667 00000 n +0000284856 00000 n +0000284919 00000 n +0000284983 00000 n +0000285047 00000 n +0000285111 00000 n +0000287286 00000 n +0000287349 00000 n +0000287412 00000 n +0000287474 00000 n +0000287538 00000 n +0000287602 00000 n +0000280092 00000 n +0000285174 00000 n +0000279968 00000 n +0000278510 00000 n +0000279945 00000 n +0000287666 00000 n +0000287729 00000 n +0000287792 00000 n +0000287855 00000 n +0000287919 00000 n +0000287983 00000 n +0000288047 00000 n +0000288110 00000 n +0000288173 00000 n +0000288236 00000 n +0000288300 00000 n +0000288364 00000 n +0000288428 00000 n +0000288492 00000 n +0000288556 00000 n +0000288620 00000 n +0000288683 00000 n +0000288746 00000 n +0000288809 00000 n +0000288873 00000 n +0000288937 00000 n +0000289001 00000 n +0000289065 00000 n +0000289129 00000 n +0000289193 00000 n +0000289257 00000 n +0000289321 00000 n +0000289384 00000 n +0000289447 00000 n +0000289510 00000 n +0000289574 00000 n +0000289638 00000 n +0000289701 00000 n +0000289764 00000 n +0000289827 00000 n +0000289891 00000 n +0000289955 00000 n +0000290019 00000 n +0000290082 00000 n +0000290145 00000 n +0000290208 00000 n +0000290272 00000 n +0000290336 00000 n +0000290399 00000 n +0000290462 00000 n +0000293029 00000 n +0000293092 00000 n +0000293155 00000 n +0000293218 00000 n +0000293282 00000 n +0000287223 00000 n +0000290526 00000 n +0000287099 00000 n +0000285314 00000 n +0000287076 00000 n +0000293346 00000 n +0000293409 00000 n +0000293472 00000 n +0000293533 00000 n +0000293597 00000 n +0000293661 00000 n +0000293724 00000 n +0000293787 00000 n +0000293850 00000 n +0000293913 00000 n +0000293977 00000 n +0000294040 00000 n +0000294103 00000 n +0000294166 00000 n +0000294230 00000 n +0000294292 00000 n +0000294481 00000 n +0000294544 00000 n +0000294607 00000 n +0000294670 00000 n +0000294732 00000 n +0000294796 00000 n +0000294860 00000 n +0000294924 00000 n +0000294988 00000 n +0000295052 00000 n +0000295116 00000 n +0000295179 00000 n +0000295242 00000 n +0000295303 00000 n +0000295367 00000 n +0000295431 00000 n +0000295495 00000 n +0000295558 00000 n +0000295621 00000 n +0000295684 00000 n +0000295748 00000 n +0000298132 00000 n +0000298196 00000 n +0000292966 00000 n +0000295812 00000 n +0000292664 00000 n +0000290666 00000 n +0000292641 00000 n +0000292809 00000 n +0000498523 00000 n +0000298260 00000 n +0000298323 00000 n +0000298387 00000 n +0000298451 00000 n +0000298515 00000 n +0000298579 00000 n +0000298642 00000 n +0000298704 00000 n +0000298768 00000 n +0000298831 00000 n +0000298894 00000 n +0000298958 00000 n +0000299022 00000 n +0000299086 00000 n +0000299150 00000 n +0000299213 00000 n +0000299276 00000 n +0000299339 00000 n +0000299403 00000 n +0000299467 00000 n +0000299531 00000 n +0000299594 00000 n +0000299657 00000 n +0000299720 00000 n +0000299784 00000 n +0000299848 00000 n +0000299911 00000 n +0000299974 00000 n +0000300037 00000 n +0000300101 00000 n +0000300165 00000 n +0000300229 00000 n +0000300293 00000 n +0000300357 00000 n +0000300421 00000 n +0000300485 00000 n +0000300549 00000 n +0000300613 00000 n +0000300676 00000 n +0000300739 00000 n +0000300802 00000 n +0000300864 00000 n +0000300927 00000 n +0000300990 00000 n +0000301053 00000 n +0000301117 00000 n +0000301181 00000 n +0000301245 00000 n +0000301308 00000 n +0000301372 00000 n +0000303871 00000 n +0000303934 00000 n +0000303997 00000 n +0000304060 00000 n +0000304124 00000 n +0000304188 00000 n +0000298069 00000 n +0000301436 00000 n +0000297945 00000 n +0000295952 00000 n +0000297922 00000 n +0000304252 00000 n +0000304315 00000 n +0000304378 00000 n +0000304441 00000 n +0000304505 00000 n +0000304569 00000 n +0000304633 00000 n +0000304696 00000 n +0000304759 00000 n +0000304822 00000 n +0000304886 00000 n +0000304950 00000 n +0000305014 00000 n +0000305077 00000 n +0000305141 00000 n +0000305205 00000 n +0000305269 00000 n +0000305332 00000 n +0000305396 00000 n +0000305460 00000 n +0000305524 00000 n +0000305586 00000 n +0000305648 00000 n +0000305711 00000 n +0000305775 00000 n +0000305839 00000 n +0000305903 00000 n +0000305967 00000 n +0000306031 00000 n +0000306093 00000 n +0000306155 00000 n +0000306218 00000 n +0000306282 00000 n +0000306346 00000 n +0000306410 00000 n +0000306473 00000 n +0000306536 00000 n +0000309980 00000 n +0000310043 00000 n +0000310106 00000 n +0000310169 00000 n +0000310233 00000 n +0000310297 00000 n +0000303745 00000 n +0000303808 00000 n +0000306600 00000 n +0000303446 00000 n +0000301549 00000 n +0000303423 00000 n +0000303591 00000 n +0000310361 00000 n +0000310424 00000 n +0000310487 00000 n +0000310550 00000 n +0000310614 00000 n +0000310678 00000 n +0000310741 00000 n +0000310804 00000 n +0000310867 00000 n +0000310930 00000 n +0000310993 00000 n +0000311057 00000 n +0000311119 00000 n +0000311181 00000 n +0000311244 00000 n +0000311308 00000 n +0000311372 00000 n +0000311435 00000 n +0000311498 00000 n +0000311561 00000 n +0000311625 00000 n +0000311689 00000 n +0000311752 00000 n +0000311815 00000 n +0000311878 00000 n +0000311942 00000 n +0000312006 00000 n +0000312070 00000 n +0000312134 00000 n +0000312197 00000 n +0000312261 00000 n +0000312324 00000 n +0000312387 00000 n +0000312450 00000 n +0000312513 00000 n +0000312575 00000 n +0000312639 00000 n +0000312702 00000 n +0000312762 00000 n +0000312825 00000 n +0000312889 00000 n +0000315373 00000 n +0000315437 00000 n +0000309854 00000 n +0000309917 00000 n +0000312952 00000 n +0000308783 00000 n +0000306740 00000 n +0000308760 00000 n +0000308964 00000 n +0000309122 00000 n +0000309308 00000 n +0000309483 00000 n +0000309669 00000 n +0000315501 00000 n +0000315565 00000 n +0000315628 00000 n +0000315691 00000 n +0000315754 00000 n +0000315817 00000 n +0000315880 00000 n +0000315943 00000 n +0000316006 00000 n +0000316070 00000 n +0000316259 00000 n +0000316322 00000 n +0000316385 00000 n +0000316447 00000 n +0000316509 00000 n +0000316572 00000 n +0000316636 00000 n +0000316700 00000 n +0000316763 00000 n +0000316826 00000 n +0000316889 00000 n +0000316952 00000 n +0000319786 00000 n +0000319849 00000 n +0000315310 00000 n +0000317016 00000 n +0000315186 00000 n +0000313079 00000 n +0000315163 00000 n +0000320036 00000 n +0000320099 00000 n +0000320160 00000 n +0000320223 00000 n +0000320286 00000 n +0000320348 00000 n +0000320411 00000 n +0000320475 00000 n +0000320539 00000 n +0000320603 00000 n +0000320666 00000 n +0000320729 00000 n +0000320792 00000 n +0000320856 00000 n +0000320920 00000 n +0000320984 00000 n +0000321047 00000 n +0000321110 00000 n +0000321173 00000 n +0000321236 00000 n +0000321299 00000 n +0000321362 00000 n +0000321425 00000 n +0000321488 00000 n +0000321551 00000 n +0000321614 00000 n +0000321677 00000 n +0000321740 00000 n +0000321804 00000 n +0000321868 00000 n +0000321932 00000 n +0000321996 00000 n +0000322058 00000 n +0000322120 00000 n +0000322183 00000 n +0000322247 00000 n +0000324170 00000 n +0000324233 00000 n +0000324296 00000 n +0000324360 00000 n +0000319723 00000 n +0000322373 00000 n +0000318898 00000 n +0000317156 00000 n +0000318875 00000 n +0000319070 00000 n +0000319241 00000 n +0000319408 00000 n +0000319566 00000 n +0000324424 00000 n +0000324487 00000 n +0000324551 00000 n +0000324614 00000 n +0000324677 00000 n +0000324740 00000 n +0000324803 00000 n +0000324866 00000 n +0000324930 00000 n +0000324993 00000 n +0000325057 00000 n +0000325121 00000 n +0000325184 00000 n +0000325247 00000 n +0000325310 00000 n +0000325374 00000 n +0000325437 00000 n +0000325500 00000 n +0000325563 00000 n +0000325626 00000 n +0000325814 00000 n +0000325877 00000 n +0000328014 00000 n +0000328078 00000 n +0000328142 00000 n +0000328206 00000 n +0000328270 00000 n +0000328334 00000 n +0000324044 00000 n +0000325941 00000 n +0000323920 00000 n +0000322499 00000 n +0000323897 00000 n +0000498648 00000 n +0000328398 00000 n +0000328461 00000 n +0000328524 00000 n +0000328586 00000 n +0000328775 00000 n +0000328838 00000 n +0000328901 00000 n +0000328964 00000 n +0000329027 00000 n +0000329090 00000 n +0000329153 00000 n +0000329217 00000 n +0000329280 00000 n +0000329343 00000 n +0000329406 00000 n +0000329469 00000 n +0000329533 00000 n +0000329597 00000 n +0000329660 00000 n +0000329723 00000 n +0000329786 00000 n +0000329850 00000 n +0000329913 00000 n +0000329976 00000 n +0000330039 00000 n +0000330103 00000 n +0000330166 00000 n +0000330229 00000 n +0000330292 00000 n +0000330356 00000 n +0000330419 00000 n +0000330482 00000 n +0000330544 00000 n +0000332794 00000 n +0000332857 00000 n +0000332920 00000 n +0000332983 00000 n +0000333047 00000 n +0000327951 00000 n +0000330608 00000 n +0000327827 00000 n +0000326067 00000 n +0000327804 00000 n +0000333111 00000 n +0000333174 00000 n +0000333237 00000 n +0000333300 00000 n +0000333364 00000 n +0000333428 00000 n +0000333490 00000 n +0000333552 00000 n +0000333615 00000 n +0000333678 00000 n +0000333741 00000 n +0000333805 00000 n +0000333869 00000 n +0000333933 00000 n +0000333997 00000 n +0000334061 00000 n +0000334125 00000 n +0000334189 00000 n +0000334253 00000 n +0000334317 00000 n +0000334381 00000 n +0000334445 00000 n +0000334509 00000 n +0000334573 00000 n +0000334636 00000 n +0000334699 00000 n +0000334762 00000 n +0000334826 00000 n +0000334890 00000 n +0000334954 00000 n +0000335017 00000 n +0000335080 00000 n +0000335143 00000 n +0000335207 00000 n +0000335271 00000 n +0000335334 00000 n +0000335397 00000 n +0000335460 00000 n +0000335524 00000 n +0000335587 00000 n +0000335650 00000 n +0000335714 00000 n +0000335778 00000 n +0000338347 00000 n +0000338410 00000 n +0000338473 00000 n +0000338535 00000 n +0000338599 00000 n +0000338663 00000 n +0000338726 00000 n +0000338787 00000 n +0000338851 00000 n +0000338915 00000 n +0000332731 00000 n +0000335842 00000 n +0000332607 00000 n +0000330761 00000 n +0000332584 00000 n +0000338979 00000 n +0000339042 00000 n +0000339105 00000 n +0000339168 00000 n +0000339231 00000 n +0000339294 00000 n +0000339357 00000 n +0000339420 00000 n +0000339483 00000 n +0000339547 00000 n +0000339611 00000 n +0000339675 00000 n +0000339739 00000 n +0000339803 00000 n +0000339867 00000 n +0000339931 00000 n +0000339995 00000 n +0000340058 00000 n +0000340121 00000 n +0000340184 00000 n +0000340248 00000 n +0000340437 00000 n +0000340500 00000 n +0000340564 00000 n +0000340628 00000 n +0000340692 00000 n +0000340756 00000 n +0000340818 00000 n +0000340881 00000 n +0000340944 00000 n +0000341007 00000 n +0000341070 00000 n +0000341132 00000 n +0000341196 00000 n +0000343253 00000 n +0000343316 00000 n +0000343379 00000 n +0000343441 00000 n +0000343505 00000 n +0000343569 00000 n +0000343632 00000 n +0000338284 00000 n +0000341259 00000 n +0000337963 00000 n +0000335955 00000 n +0000337940 00000 n +0000338108 00000 n +0000343695 00000 n +0000343758 00000 n +0000343821 00000 n +0000343884 00000 n +0000343948 00000 n +0000344011 00000 n +0000344074 00000 n +0000344137 00000 n +0000344200 00000 n +0000344264 00000 n +0000344453 00000 n +0000344515 00000 n +0000344578 00000 n +0000344641 00000 n +0000344704 00000 n +0000344767 00000 n +0000344831 00000 n +0000344894 00000 n +0000344957 00000 n +0000345020 00000 n +0000345084 00000 n +0000345147 00000 n +0000345211 00000 n +0000345274 00000 n +0000345337 00000 n +0000345400 00000 n +0000345464 00000 n +0000345527 00000 n +0000345590 00000 n +0000345653 00000 n +0000345716 00000 n +0000347800 00000 n +0000343128 00000 n +0000345780 00000 n +0000343004 00000 n +0000341385 00000 n +0000342981 00000 n +0000347926 00000 n +0000347989 00000 n +0000348051 00000 n +0000348113 00000 n +0000348176 00000 n +0000348240 00000 n +0000348303 00000 n +0000348366 00000 n +0000348429 00000 n +0000348492 00000 n +0000348556 00000 n +0000348619 00000 n +0000348682 00000 n +0000348745 00000 n +0000348808 00000 n +0000348872 00000 n +0000348936 00000 n +0000349000 00000 n +0000349063 00000 n +0000349126 00000 n +0000349188 00000 n +0000349252 00000 n +0000349316 00000 n +0000349379 00000 n +0000349442 00000 n +0000349631 00000 n +0000349694 00000 n +0000349757 00000 n +0000347674 00000 n +0000347737 00000 n +0000349821 00000 n +0000347550 00000 n +0000345906 00000 n +0000347527 00000 n +0000352508 00000 n +0000352571 00000 n +0000352634 00000 n +0000352697 00000 n +0000352760 00000 n +0000352823 00000 n +0000352887 00000 n +0000353075 00000 n +0000353138 00000 n +0000353200 00000 n +0000353263 00000 n +0000353327 00000 n +0000353391 00000 n +0000353455 00000 n +0000353519 00000 n +0000353582 00000 n +0000353646 00000 n +0000353709 00000 n +0000353771 00000 n +0000355459 00000 n +0000352445 00000 n +0000353833 00000 n +0000351971 00000 n +0000349974 00000 n +0000351948 00000 n +0000352125 00000 n +0000352288 00000 n +0000498773 00000 n +0000355522 00000 n +0000355585 00000 n +0000355648 00000 n +0000355712 00000 n +0000358237 00000 n +0000358300 00000 n +0000358364 00000 n +0000355333 00000 n +0000355396 00000 n +0000355837 00000 n +0000355209 00000 n +0000353986 00000 n +0000355186 00000 n +0000358428 00000 n +0000358490 00000 n +0000358552 00000 n +0000358615 00000 n +0000358679 00000 n +0000358743 00000 n +0000358807 00000 n +0000358871 00000 n +0000358934 00000 n +0000358997 00000 n +0000359061 00000 n +0000359125 00000 n +0000359189 00000 n +0000359252 00000 n +0000359316 00000 n +0000359379 00000 n +0000359442 00000 n +0000359505 00000 n +0000359568 00000 n +0000359631 00000 n +0000359694 00000 n +0000359757 00000 n +0000359820 00000 n +0000359882 00000 n +0000359944 00000 n +0000360007 00000 n +0000360069 00000 n +0000360256 00000 n +0000362848 00000 n +0000362912 00000 n +0000358111 00000 n +0000360319 00000 n +0000357795 00000 n +0000355937 00000 n +0000357772 00000 n +0000357940 00000 n +0000362976 00000 n +0000363038 00000 n +0000363101 00000 n +0000363165 00000 n +0000363229 00000 n +0000363293 00000 n +0000363357 00000 n +0000363421 00000 n +0000363485 00000 n +0000363549 00000 n +0000363613 00000 n +0000363676 00000 n +0000363739 00000 n +0000363802 00000 n +0000363990 00000 n +0000364053 00000 n +0000364116 00000 n +0000364179 00000 n +0000364242 00000 n +0000364306 00000 n +0000364370 00000 n +0000364434 00000 n +0000364498 00000 n +0000364561 00000 n +0000364624 00000 n +0000364687 00000 n +0000364751 00000 n +0000364815 00000 n +0000364879 00000 n +0000364943 00000 n +0000365007 00000 n +0000365070 00000 n +0000365133 00000 n +0000365196 00000 n +0000365260 00000 n +0000365324 00000 n +0000365387 00000 n +0000365450 00000 n +0000365513 00000 n +0000365577 00000 n +0000367915 00000 n +0000362722 00000 n +0000362785 00000 n +0000365641 00000 n +0000362598 00000 n +0000360486 00000 n +0000362575 00000 n +0000367978 00000 n +0000368040 00000 n +0000368102 00000 n +0000368165 00000 n +0000368229 00000 n +0000368293 00000 n +0000368355 00000 n +0000368417 00000 n +0000368480 00000 n +0000368543 00000 n +0000368606 00000 n +0000368669 00000 n +0000368732 00000 n +0000368795 00000 n +0000368858 00000 n +0000368922 00000 n +0000369109 00000 n +0000369172 00000 n +0000369236 00000 n +0000369300 00000 n +0000369363 00000 n +0000369426 00000 n +0000369490 00000 n +0000369553 00000 n +0000369615 00000 n +0000369678 00000 n +0000367852 00000 n +0000369741 00000 n +0000367544 00000 n +0000365822 00000 n +0000367521 00000 n +0000367689 00000 n +0000371735 00000 n +0000371798 00000 n +0000371861 00000 n +0000371924 00000 n +0000371988 00000 n +0000372051 00000 n +0000372115 00000 n +0000372178 00000 n +0000372242 00000 n +0000372305 00000 n +0000372369 00000 n +0000372432 00000 n +0000372496 00000 n +0000372559 00000 n +0000372623 00000 n +0000372685 00000 n +0000372748 00000 n +0000372811 00000 n +0000372998 00000 n +0000373061 00000 n +0000373125 00000 n +0000373188 00000 n +0000373252 00000 n +0000373315 00000 n +0000373378 00000 n +0000373441 00000 n +0000373505 00000 n +0000376212 00000 n +0000376275 00000 n +0000376339 00000 n +0000371484 00000 n +0000371547 00000 n +0000373569 00000 n +0000371360 00000 n +0000369894 00000 n +0000371337 00000 n +0000376403 00000 n +0000376467 00000 n +0000376531 00000 n +0000376595 00000 n +0000376658 00000 n +0000376722 00000 n +0000376786 00000 n +0000376849 00000 n +0000376913 00000 n +0000376977 00000 n +0000377041 00000 n +0000377104 00000 n +0000377168 00000 n +0000377232 00000 n +0000377295 00000 n +0000377359 00000 n +0000377423 00000 n +0000377487 00000 n +0000377551 00000 n +0000377615 00000 n +0000377679 00000 n +0000377743 00000 n +0000377807 00000 n +0000377871 00000 n +0000377935 00000 n +0000377998 00000 n +0000378061 00000 n +0000378125 00000 n +0000378189 00000 n +0000378251 00000 n +0000378315 00000 n +0000378379 00000 n +0000378443 00000 n +0000378507 00000 n +0000378571 00000 n +0000378634 00000 n +0000378823 00000 n +0000381334 00000 n +0000381397 00000 n +0000381460 00000 n +0000381524 00000 n +0000381588 00000 n +0000376149 00000 n +0000378885 00000 n +0000376025 00000 n +0000373709 00000 n +0000376002 00000 n +0000498898 00000 n +0000381652 00000 n +0000381715 00000 n +0000381778 00000 n +0000381838 00000 n +0000381902 00000 n +0000381966 00000 n +0000382030 00000 n +0000382219 00000 n +0000382282 00000 n +0000382346 00000 n +0000382409 00000 n +0000382598 00000 n +0000382661 00000 n +0000382724 00000 n +0000382788 00000 n +0000382852 00000 n +0000382916 00000 n +0000382980 00000 n +0000383044 00000 n +0000383107 00000 n +0000383170 00000 n +0000383233 00000 n +0000383297 00000 n +0000383360 00000 n +0000383548 00000 n +0000383611 00000 n +0000383675 00000 n +0000383739 00000 n +0000383802 00000 n +0000383866 00000 n +0000383930 00000 n +0000383994 00000 n +0000384058 00000 n +0000384122 00000 n +0000384186 00000 n +0000384250 00000 n +0000384314 00000 n +0000384376 00000 n +0000386389 00000 n +0000386452 00000 n +0000386516 00000 n +0000381271 00000 n +0000384440 00000 n +0000380971 00000 n +0000379011 00000 n +0000380948 00000 n +0000381116 00000 n +0000386580 00000 n +0000386643 00000 n +0000386707 00000 n +0000386769 00000 n +0000386833 00000 n +0000386897 00000 n +0000386960 00000 n +0000387023 00000 n +0000387086 00000 n +0000387150 00000 n +0000387214 00000 n +0000387277 00000 n +0000387340 00000 n +0000387404 00000 n +0000387467 00000 n +0000387531 00000 n +0000389458 00000 n +0000389521 00000 n +0000389584 00000 n +0000389648 00000 n +0000389712 00000 n +0000386326 00000 n +0000387656 00000 n +0000386202 00000 n +0000384580 00000 n +0000386179 00000 n +0000389776 00000 n +0000389839 00000 n +0000389903 00000 n +0000389966 00000 n +0000390030 00000 n +0000390093 00000 n +0000390157 00000 n +0000390220 00000 n +0000390284 00000 n +0000390346 00000 n +0000390409 00000 n +0000390472 00000 n +0000390535 00000 n +0000390599 00000 n +0000390662 00000 n +0000390726 00000 n +0000390790 00000 n +0000390854 00000 n +0000390918 00000 n +0000390982 00000 n +0000391046 00000 n +0000391110 00000 n +0000391173 00000 n +0000391237 00000 n +0000391301 00000 n +0000391365 00000 n +0000391428 00000 n +0000391492 00000 n +0000391555 00000 n +0000391619 00000 n +0000391682 00000 n +0000391746 00000 n +0000391810 00000 n +0000391872 00000 n +0000391936 00000 n +0000391999 00000 n +0000392063 00000 n +0000392127 00000 n +0000392191 00000 n +0000392255 00000 n +0000392319 00000 n +0000392382 00000 n +0000392446 00000 n +0000392509 00000 n +0000392573 00000 n +0000392636 00000 n +0000392700 00000 n +0000392763 00000 n +0000392827 00000 n +0000392891 00000 n +0000392955 00000 n +0000393019 00000 n +0000393083 00000 n +0000393147 00000 n +0000393211 00000 n +0000393275 00000 n +0000393338 00000 n +0000393402 00000 n +0000393465 00000 n +0000393529 00000 n +0000393591 00000 n +0000393654 00000 n +0000393717 00000 n +0000393781 00000 n +0000393844 00000 n +0000393908 00000 n +0000393971 00000 n +0000394035 00000 n +0000394099 00000 n +0000394162 00000 n +0000394226 00000 n +0000394289 00000 n +0000394353 00000 n +0000394417 00000 n +0000394480 00000 n +0000394544 00000 n +0000394608 00000 n +0000394672 00000 n +0000394736 00000 n +0000394799 00000 n +0000394863 00000 n +0000394927 00000 n +0000394991 00000 n +0000395055 00000 n +0000395118 00000 n +0000395182 00000 n +0000395246 00000 n +0000395310 00000 n +0000395374 00000 n +0000395436 00000 n +0000395499 00000 n +0000395562 00000 n +0000395625 00000 n +0000395688 00000 n +0000395751 00000 n +0000395815 00000 n +0000395878 00000 n +0000395942 00000 n +0000396005 00000 n +0000396069 00000 n +0000396132 00000 n +0000396196 00000 n +0000396259 00000 n +0000396323 00000 n +0000396386 00000 n +0000396450 00000 n +0000396513 00000 n +0000398623 00000 n +0000398686 00000 n +0000398749 00000 n +0000389332 00000 n +0000396638 00000 n +0000389208 00000 n +0000387769 00000 n +0000389185 00000 n +0000398373 00000 n +0000398812 00000 n +0000398249 00000 n +0000396766 00000 n +0000398226 00000 n +0000400841 00000 n +0000400904 00000 n +0000400717 00000 n +0000398965 00000 n +0000400694 00000 n +0000403640 00000 n +0000403703 00000 n +0000403767 00000 n +0000403831 00000 n +0000403894 00000 n +0000403958 00000 n +0000404022 00000 n +0000404085 00000 n +0000404274 00000 n +0000404337 00000 n +0000404400 00000 n +0000404463 00000 n +0000404526 00000 n +0000404590 00000 n +0000404654 00000 n +0000404717 00000 n +0000404780 00000 n +0000404843 00000 n +0000404907 00000 n +0000404971 00000 n +0000405034 00000 n +0000405097 00000 n +0000405160 00000 n +0000405223 00000 n +0000405286 00000 n +0000405349 00000 n +0000405412 00000 n +0000405475 00000 n +0000405539 00000 n +0000405602 00000 n +0000405665 00000 n +0000405728 00000 n +0000408466 00000 n +0000408529 00000 n +0000408592 00000 n +0000408655 00000 n +0000408719 00000 n +0000408783 00000 n +0000408847 00000 n +0000408910 00000 n +0000403454 00000 n +0000405792 00000 n +0000402623 00000 n +0000401017 00000 n +0000402600 00000 n +0000402795 00000 n +0000402958 00000 n +0000403120 00000 n +0000403283 00000 n +0000499023 00000 n +0000408974 00000 n +0000409037 00000 n +0000409100 00000 n +0000409163 00000 n +0000409226 00000 n +0000409290 00000 n +0000409353 00000 n +0000409417 00000 n +0000409481 00000 n +0000409545 00000 n +0000409609 00000 n +0000409672 00000 n +0000409735 00000 n +0000409798 00000 n +0000409862 00000 n +0000409926 00000 n +0000409989 00000 n +0000410052 00000 n +0000410115 00000 n +0000410179 00000 n +0000410243 00000 n +0000410306 00000 n +0000410370 00000 n +0000410432 00000 n +0000410495 00000 n +0000410558 00000 n +0000410621 00000 n +0000410685 00000 n +0000413418 00000 n +0000413481 00000 n +0000413544 00000 n +0000413606 00000 n +0000413670 00000 n +0000413734 00000 n +0000408403 00000 n +0000410749 00000 n +0000407914 00000 n +0000405918 00000 n +0000407891 00000 n +0000408068 00000 n +0000408239 00000 n +0000413798 00000 n +0000413861 00000 n +0000413924 00000 n +0000413987 00000 n +0000414051 00000 n +0000414115 00000 n +0000414178 00000 n +0000414241 00000 n +0000414304 00000 n +0000414367 00000 n +0000414430 00000 n +0000414494 00000 n +0000414558 00000 n +0000414621 00000 n +0000414684 00000 n +0000414747 00000 n +0000414811 00000 n +0000414875 00000 n +0000414939 00000 n +0000415002 00000 n +0000415065 00000 n +0000415128 00000 n +0000415192 00000 n +0000415256 00000 n +0000415319 00000 n +0000415382 00000 n +0000415445 00000 n +0000415509 00000 n +0000415573 00000 n +0000415636 00000 n +0000415699 00000 n +0000415762 00000 n +0000415826 00000 n +0000415890 00000 n +0000415953 00000 n +0000416016 00000 n +0000416079 00000 n +0000416143 00000 n +0000416207 00000 n +0000416270 00000 n +0000416333 00000 n +0000416396 00000 n +0000416460 00000 n +0000416524 00000 n +0000416587 00000 n +0000416650 00000 n +0000416713 00000 n +0000416777 00000 n +0000416841 00000 n +0000419862 00000 n +0000419925 00000 n +0000419988 00000 n +0000420050 00000 n +0000420114 00000 n +0000420178 00000 n +0000413292 00000 n +0000413355 00000 n +0000416904 00000 n +0000412295 00000 n +0000410876 00000 n +0000412272 00000 n +0000412476 00000 n +0000412639 00000 n +0000412802 00000 n +0000412965 00000 n +0000413128 00000 n +0000420242 00000 n +0000420305 00000 n +0000420369 00000 n +0000420433 00000 n +0000420496 00000 n +0000420559 00000 n +0000420622 00000 n +0000420686 00000 n +0000420750 00000 n +0000420813 00000 n +0000420876 00000 n +0000420939 00000 n +0000421003 00000 n +0000421067 00000 n +0000421130 00000 n +0000421193 00000 n +0000421256 00000 n +0000421320 00000 n +0000421384 00000 n +0000421447 00000 n +0000421510 00000 n +0000421573 00000 n +0000421637 00000 n +0000421701 00000 n +0000421764 00000 n +0000421827 00000 n +0000421890 00000 n +0000421953 00000 n +0000422016 00000 n +0000422079 00000 n +0000422142 00000 n +0000422205 00000 n +0000422268 00000 n +0000422456 00000 n +0000422519 00000 n +0000422582 00000 n +0000422645 00000 n +0000422708 00000 n +0000422771 00000 n +0000422834 00000 n +0000422897 00000 n +0000422958 00000 n +0000423022 00000 n +0000423086 00000 n +0000423149 00000 n +0000423213 00000 n +0000423277 00000 n +0000425330 00000 n +0000419799 00000 n +0000423341 00000 n +0000418649 00000 n +0000417017 00000 n +0000418626 00000 n +0000418839 00000 n +0000418994 00000 n +0000419157 00000 n +0000419320 00000 n +0000419482 00000 n +0000419644 00000 n +0000425394 00000 n +0000425457 00000 n +0000425520 00000 n +0000425584 00000 n +0000425648 00000 n +0000425712 00000 n +0000425775 00000 n +0000425839 00000 n +0000425903 00000 n +0000425967 00000 n +0000426030 00000 n +0000426093 00000 n +0000426156 00000 n +0000426220 00000 n +0000426284 00000 n +0000426348 00000 n +0000426411 00000 n +0000426474 00000 n +0000426537 00000 n +0000426850 00000 n +0000427039 00000 n +0000425267 00000 n +0000427102 00000 n +0000425143 00000 n +0000423467 00000 n +0000425120 00000 n +0000429748 00000 n +0000429811 00000 n +0000429874 00000 n +0000429937 00000 n +0000430000 00000 n +0000430063 00000 n +0000430126 00000 n +0000430189 00000 n +0000430252 00000 n +0000429685 00000 n +0000430315 00000 n +0000429199 00000 n +0000427282 00000 n +0000429176 00000 n +0000429353 00000 n +0000429523 00000 n +0000432474 00000 n +0000432535 00000 n +0000432599 00000 n +0000432662 00000 n +0000432725 00000 n +0000432788 00000 n +0000432851 00000 n +0000432914 00000 n +0000432977 00000 n +0000433040 00000 n +0000432411 00000 n +0000433103 00000 n +0000432090 00000 n +0000430401 00000 n +0000432067 00000 n +0000432235 00000 n +0000499148 00000 n +0000435538 00000 n +0000435601 00000 n +0000435664 00000 n +0000435851 00000 n +0000435912 00000 n +0000437999 00000 n +0000435475 00000 n +0000435975 00000 n +0000435351 00000 n +0000433189 00000 n +0000435328 00000 n +0000438062 00000 n +0000438125 00000 n +0000438187 00000 n +0000438250 00000 n +0000438313 00000 n +0000438376 00000 n +0000438439 00000 n +0000438502 00000 n +0000438565 00000 n +0000438628 00000 n +0000438691 00000 n +0000438754 00000 n +0000438817 00000 n +0000439007 00000 n +0000439070 00000 n +0000441226 00000 n +0000437936 00000 n +0000439133 00000 n +0000437812 00000 n +0000436074 00000 n +0000437789 00000 n +0000441290 00000 n +0000441353 00000 n +0000441416 00000 n +0000441479 00000 n +0000441542 00000 n +0000441605 00000 n +0000441668 00000 n +0000441732 00000 n +0000441796 00000 n +0000441985 00000 n +0000442048 00000 n +0000442111 00000 n +0000442174 00000 n +0000441163 00000 n +0000442237 00000 n +0000441039 00000 n +0000439246 00000 n +0000441016 00000 n +0000444379 00000 n +0000444442 00000 n +0000444631 00000 n +0000444694 00000 n +0000444758 00000 n +0000444822 00000 n +0000444885 00000 n +0000444948 00000 n +0000445011 00000 n +0000445074 00000 n +0000445137 00000 n +0000445327 00000 n +0000445390 00000 n +0000445453 00000 n +0000445516 00000 n +0000445579 00000 n +0000445768 00000 n +0000445831 00000 n +0000445894 00000 n +0000445957 00000 n +0000446020 00000 n +0000448077 00000 n +0000448140 00000 n +0000448203 00000 n +0000448267 00000 n +0000448331 00000 n +0000448395 00000 n +0000448459 00000 n +0000444253 00000 n +0000444316 00000 n +0000446083 00000 n +0000444129 00000 n +0000442350 00000 n +0000444106 00000 n +0000448523 00000 n +0000448586 00000 n +0000448649 00000 n +0000448838 00000 n +0000448901 00000 n +0000448963 00000 n +0000449025 00000 n +0000449087 00000 n +0000449149 00000 n +0000449210 00000 n +0000449272 00000 n +0000449335 00000 n +0000449399 00000 n +0000449463 00000 n +0000449527 00000 n +0000449591 00000 n +0000449654 00000 n +0000449718 00000 n +0000449780 00000 n +0000449843 00000 n +0000450033 00000 n +0000450096 00000 n +0000450160 00000 n +0000450224 00000 n +0000450288 00000 n +0000450352 00000 n +0000450414 00000 n +0000450477 00000 n +0000450666 00000 n +0000450729 00000 n +0000450793 00000 n +0000450857 00000 n +0000450921 00000 n +0000450985 00000 n +0000451049 00000 n +0000451113 00000 n +0000453392 00000 n +0000453455 00000 n +0000453518 00000 n +0000448014 00000 n +0000451177 00000 n +0000447890 00000 n +0000446263 00000 n +0000447867 00000 n +0000453582 00000 n +0000453645 00000 n +0000453708 00000 n +0000453771 00000 n +0000453834 00000 n +0000453897 00000 n +0000453960 00000 n +0000454022 00000 n +0000454084 00000 n +0000454147 00000 n +0000454211 00000 n +0000453329 00000 n +0000454275 00000 n +0000453205 00000 n +0000451371 00000 n +0000453182 00000 n +0000499273 00000 n +0000454799 00000 n +0000454862 00000 n +0000454675 00000 n +0000454388 00000 n +0000454653 00000 n +0000456840 00000 n +0000456903 00000 n +0000456967 00000 n +0000457031 00000 n +0000457095 00000 n +0000457159 00000 n +0000457223 00000 n +0000457286 00000 n +0000457349 00000 n +0000457413 00000 n +0000457476 00000 n +0000457539 00000 n +0000457602 00000 n +0000457664 00000 n +0000457728 00000 n +0000460543 00000 n +0000460606 00000 n +0000460670 00000 n +0000460734 00000 n +0000460798 00000 n +0000460862 00000 n +0000456527 00000 n +0000457855 00000 n +0000456403 00000 n +0000454935 00000 n +0000456380 00000 n +0000460926 00000 n +0000460989 00000 n +0000461053 00000 n +0000461117 00000 n +0000461181 00000 n +0000461245 00000 n +0000461308 00000 n +0000461372 00000 n +0000461436 00000 n +0000461500 00000 n +0000461564 00000 n +0000461628 00000 n +0000461691 00000 n +0000461880 00000 n +0000461943 00000 n +0000462007 00000 n +0000462071 00000 n +0000462135 00000 n +0000462199 00000 n +0000462262 00000 n +0000462325 00000 n +0000462388 00000 n +0000462451 00000 n +0000462514 00000 n +0000462578 00000 n +0000462642 00000 n +0000462706 00000 n +0000462770 00000 n +0000462960 00000 n +0000463023 00000 n +0000463087 00000 n +0000463151 00000 n +0000463215 00000 n +0000463279 00000 n +0000463469 00000 n +0000463532 00000 n +0000463596 00000 n +0000463660 00000 n +0000460417 00000 n +0000463724 00000 n +0000460293 00000 n +0000457994 00000 n +0000460270 00000 n +0000465001 00000 n +0000465063 00000 n +0000465127 00000 n +0000465190 00000 n +0000464938 00000 n +0000465253 00000 n +0000464814 00000 n +0000463891 00000 n +0000464792 00000 n +0000465764 00000 n +0000465827 00000 n +0000465640 00000 n +0000465352 00000 n +0000465618 00000 n +0000468073 00000 n +0000468262 00000 n +0000468324 00000 n +0000468387 00000 n +0000468576 00000 n +0000468639 00000 n +0000468703 00000 n +0000468767 00000 n +0000468831 00000 n +0000468895 00000 n +0000468959 00000 n +0000469023 00000 n +0000469842 00000 n +0000469905 00000 n +0000469969 00000 n +0000470033 00000 n +0000467634 00000 n +0000469087 00000 n +0000467510 00000 n +0000465900 00000 n +0000467487 00000 n +0000499398 00000 n +0000469779 00000 n +0000470097 00000 n +0000469655 00000 n +0000469212 00000 n +0000469633 00000 n +0000472949 00000 n +0000473011 00000 n +0000473075 00000 n +0000473138 00000 n +0000473202 00000 n +0000473266 00000 n +0000473329 00000 n +0000473393 00000 n +0000473457 00000 n +0000473521 00000 n +0000473585 00000 n +0000473648 00000 n +0000473711 00000 n +0000473775 00000 n +0000473838 00000 n +0000473901 00000 n +0000473965 00000 n +0000474028 00000 n +0000475905 00000 n +0000472510 00000 n +0000474091 00000 n +0000474153 00000 n +0000472386 00000 n +0000470183 00000 n +0000472363 00000 n +0000476346 00000 n +0000476409 00000 n +0000476473 00000 n +0000476663 00000 n +0000476726 00000 n +0000477042 00000 n +0000477105 00000 n +0000477169 00000 n +0000477233 00000 n +0000477297 00000 n +0000477361 00000 n +0000477424 00000 n +0000477486 00000 n +0000477549 00000 n +0000475842 00000 n +0000477612 00000 n +0000475718 00000 n +0000474278 00000 n +0000475695 00000 n +0000479739 00000 n +0000479802 00000 n +0000479865 00000 n +0000479928 00000 n +0000479991 00000 n +0000480052 00000 n +0000480116 00000 n +0000480180 00000 n +0000480244 00000 n +0000480308 00000 n +0000480371 00000 n +0000480434 00000 n +0000480497 00000 n +0000480560 00000 n +0000480623 00000 n +0000480686 00000 n +0000482822 00000 n +0000479613 00000 n +0000479676 00000 n +0000480749 00000 n +0000479489 00000 n +0000477750 00000 n +0000479466 00000 n +0000482885 00000 n +0000482948 00000 n +0000483011 00000 n +0000483074 00000 n +0000483137 00000 n +0000483451 00000 n +0000483513 00000 n +0000483576 00000 n +0000483640 00000 n +0000483704 00000 n +0000483768 00000 n +0000483958 00000 n +0000484084 00000 n +0000484147 00000 n +0000484210 00000 n +0000482759 00000 n +0000484272 00000 n +0000482635 00000 n +0000480835 00000 n +0000482612 00000 n +0000486228 00000 n +0000486291 00000 n +0000486354 00000 n +0000486417 00000 n +0000486480 00000 n +0000486543 00000 n +0000486606 00000 n +0000486732 00000 n +0000486795 00000 n +0000486858 00000 n +0000486920 00000 n +0000487046 00000 n +0000487109 00000 n +0000487172 00000 n +0000489067 00000 n +0000486039 00000 n +0000487235 00000 n +0000485915 00000 n +0000484411 00000 n +0000485892 00000 n +0000499523 00000 n +0000489193 00000 n +0000489256 00000 n +0000489319 00000 n +0000489382 00000 n +0000489445 00000 n +0000489508 00000 n +0000489571 00000 n +0000489634 00000 n +0000489758 00000 n +0000489821 00000 n +0000489884 00000 n +0000489947 00000 n +0000490010 00000 n +0000490135 00000 n +0000490198 00000 n +0000490261 00000 n +0000488941 00000 n +0000489004 00000 n +0000490324 00000 n +0000488817 00000 n +0000487334 00000 n +0000488794 00000 n +0000492149 00000 n +0000492212 00000 n +0000492337 00000 n +0000492400 00000 n +0000492463 00000 n +0000492526 00000 n +0000492589 00000 n +0000492652 00000 n +0000492715 00000 n +0000492840 00000 n +0000493029 00000 n +0000493217 00000 n +0000494091 00000 n +0000492086 00000 n +0000493343 00000 n +0000491962 00000 n +0000490423 00000 n +0000491939 00000 n +0000493965 00000 n +0000494154 00000 n +0000493841 00000 n +0000493455 00000 n +0000493819 00000 n +0000494539 00000 n +0000494597 00000 n +0000494415 00000 n +0000494253 00000 n +0000494394 00000 n +0000494639 00000 n +0000499639 00000 n +0000499764 00000 n +0000499890 00000 n +0000500016 00000 n +0000500096 00000 n +0000500188 00000 n +0000515863 00000 n +0000568050 00000 n +0000568091 00000 n +0000568131 00000 n +0000568312 00000 n +trailer +<< +/Size 4576 +/Root 4574 0 R +/Info 4575 0 R +>> +startxref +568519 +%%EOF diff --git a/Operating Systems/Linux/DNS/DNS and BIND, 5th ed.chm b/Operating Systems/Linux/DNS/DNS and BIND, 5th ed.chm new file mode 100644 index 0000000000000000000000000000000000000000..84b7111f18a7ac2329fb42b4aca3acf4f8e55cd3 --- /dev/null +++ b/Operating Systems/Linux/DNS/DNS and BIND, 5th ed.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:33bf83e1d97cb159d2fd52f1de0aea830057f0fc2d4b4872023bc8da9f6137f3 +size 2470225 diff --git a/Operating Systems/Linux/DNS/Dns in Action A Detailed and Practical Guide to Dns.pdf b/Operating Systems/Linux/DNS/Dns in Action A Detailed and Practical Guide to Dns.pdf new file mode 100644 index 0000000000000000000000000000000000000000..771d936f1b3b156ddc7cc6bb03eae12a08944dc1 --- /dev/null +++ b/Operating Systems/Linux/DNS/Dns in Action A Detailed and Practical Guide to Dns.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:896c12de2298687cf12816c7374b6cbc9ee2d0df5f9e08b4f2f0c313016b1dcf +size 3578252 diff --git a/Operating Systems/Linux/DNS/Pro DNS and BIND.pdf b/Operating Systems/Linux/DNS/Pro DNS and BIND.pdf new file mode 100644 index 0000000000000000000000000000000000000000..76026a8e620aba890024840ad0e117b0ed557be4 --- /dev/null +++ b/Operating Systems/Linux/DNS/Pro DNS and BIND.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:05c4f7cef8bd406e0014a9103e6054fa526ce31a3dbcde129d128e4be8ff6743 +size 3291293 diff --git a/Operating Systems/Linux/Essentials Books/Backup And Recovery.pdf b/Operating Systems/Linux/Essentials Books/Backup And Recovery.pdf new file mode 100644 index 0000000000000000000000000000000000000000..9f121149239587aefae708dfc0ba910e50155854 --- /dev/null +++ b/Operating Systems/Linux/Essentials Books/Backup And Recovery.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fd445e1dbc551822352e53d28fbc749514b1cb7abfb71354843500e66fb06167 +size 2830542 diff --git a/Operating Systems/Linux/Essentials Books/Ebook Teach Yourself Linux In 24 Hours Sharereactor.pdf b/Operating Systems/Linux/Essentials Books/Ebook Teach Yourself Linux In 24 Hours Sharereactor.pdf new file mode 100644 index 0000000000000000000000000000000000000000..eb09bb9929815aa37ccbe9e82dc241c31289b7e8 --- /dev/null +++ b/Operating Systems/Linux/Essentials Books/Ebook Teach Yourself Linux In 24 Hours Sharereactor.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d71341bb6af629f01390eb54f3ff8a7b67d1046ea8e9b1bf1c1d4254013b1410 +size 4017954 diff --git a/Operating Systems/Linux/Essentials Books/Hacking Linux Exposed.pdf b/Operating Systems/Linux/Essentials Books/Hacking Linux Exposed.pdf new file mode 100644 index 0000000000000000000000000000000000000000..6ee2c004ca10d7ebb2298c69ebbea76201210a23 --- /dev/null +++ b/Operating Systems/Linux/Essentials Books/Hacking Linux Exposed.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fc91115cfd380e9932f1e941a12cd44c6d5c4873b70a08b401b81e27c8745a9f +size 101733939 diff --git a/Operating Systems/Linux/Essentials Books/How Linux Works - What Every Super-User Should Know.chm b/Operating Systems/Linux/Essentials Books/How Linux Works - What Every Super-User Should Know.chm new file mode 100644 index 0000000000000000000000000000000000000000..91d55f58500dfbca4b192e80e9ce187924c72c71 --- /dev/null +++ b/Operating Systems/Linux/Essentials Books/How Linux Works - What Every Super-User Should Know.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bfa7ae673c80068af4cfb01f00f0d9d8a4a9b738ef42ba423ea69619167ad3fa +size 1048833 diff --git a/Operating Systems/Linux/Essentials Books/Linux Annoyances For Geeks.chm b/Operating Systems/Linux/Essentials Books/Linux Annoyances For Geeks.chm new file mode 100644 index 0000000000000000000000000000000000000000..2d12b9e7acccdad36f3f05b52e456b9d0de7a044 --- /dev/null +++ b/Operating Systems/Linux/Essentials Books/Linux Annoyances For Geeks.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:09ca66c9c89b42f8f0aa7b210ddcf162dd051b474e1823d321a63602376badaf +size 2381604 diff --git a/Operating Systems/Linux/Essentials Books/Linux Bible 2008 Edition.pdf b/Operating Systems/Linux/Essentials Books/Linux Bible 2008 Edition.pdf new file mode 100644 index 0000000000000000000000000000000000000000..86da6998017e802f608bc4e33f754434d04db5f8 --- /dev/null +++ b/Operating Systems/Linux/Essentials Books/Linux Bible 2008 Edition.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c573bbe3b07ed24a362201444d2799a894d341d4340327fd6077bf4a4d7119aa +size 27049039 diff --git a/Operating Systems/Linux/Essentials Books/Linux Newbie Administrator Guide.pdf b/Operating Systems/Linux/Essentials Books/Linux Newbie Administrator Guide.pdf new file mode 100644 index 0000000000000000000000000000000000000000..f17fe4995a0e61623749abdcfc85ba5f591b8160 Binary files /dev/null and b/Operating Systems/Linux/Essentials Books/Linux Newbie Administrator Guide.pdf differ diff --git a/Operating Systems/Linux/Essentials Books/Linux The Complete Reference.pdf b/Operating Systems/Linux/Essentials Books/Linux The Complete Reference.pdf new file mode 100644 index 0000000000000000000000000000000000000000..4edd78c56c17b5ac6873380221f1c97d34392f48 --- /dev/null +++ b/Operating Systems/Linux/Essentials Books/Linux The Complete Reference.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:25a8511bf02d9c5ce0fb2ec86a327a2ecacce091149cc2fea09f24621262bd1c +size 9346436 diff --git a/Operating Systems/Linux/Essentials Books/Linux Timesaving Techniques for Dummies.pdf b/Operating Systems/Linux/Essentials Books/Linux Timesaving Techniques for Dummies.pdf new file mode 100644 index 0000000000000000000000000000000000000000..88ab34a4963e7d171cb8878a6e5b105554b404fb --- /dev/null +++ b/Operating Systems/Linux/Essentials Books/Linux Timesaving Techniques for Dummies.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c366476a26b181a2fd50966e3efca3409cf4e4d3e004571929d9fe3d70cf1c60 +size 15979902 diff --git a/Operating Systems/Linux/Essentials Books/Linux Troubleshooting For System Administrators And Power Users .chm b/Operating Systems/Linux/Essentials Books/Linux Troubleshooting For System Administrators And Power Users .chm new file mode 100644 index 0000000000000000000000000000000000000000..18f9b84eb2f3e7ac7c3fad625175ac26b8f4beec --- /dev/null +++ b/Operating Systems/Linux/Essentials Books/Linux Troubleshooting For System Administrators And Power Users .chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4b4607161dae9cdd0be891df116622a1a34cc855cf7a793ae58402ce381e069c +size 4259091 diff --git a/Operating Systems/Linux/Essentials Books/Linux+ Certification_Bible.pdf b/Operating Systems/Linux/Essentials Books/Linux+ Certification_Bible.pdf new file mode 100644 index 0000000000000000000000000000000000000000..71ed0ab53fb1f76b88cf977aecbc254e80a4da6b --- /dev/null +++ b/Operating Systems/Linux/Essentials Books/Linux+ Certification_Bible.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d71d52084c723822710ddc071e85723c0016bf970f0bcbca20188911426618d9 +size 3777001 diff --git a/Operating Systems/Linux/Essentials Books/O'Reilly - Essential System Administration 3rd Edition.chm b/Operating Systems/Linux/Essentials Books/O'Reilly - Essential System Administration 3rd Edition.chm new file mode 100644 index 0000000000000000000000000000000000000000..bfdeee55d4b9ac48fa789fe7d8d784895d07060a --- /dev/null +++ b/Operating Systems/Linux/Essentials Books/O'Reilly - Essential System Administration 3rd Edition.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8adec4ae488802b37b8d676022e729e6c31a34870999d5d7b81368e7fe1b5235 +size 4495794 diff --git a/Operating Systems/Linux/Essentials Books/O'Reilly - Learning Red Hat Enterprise Linux & Fedora 4th Edition.chm b/Operating Systems/Linux/Essentials Books/O'Reilly - Learning Red Hat Enterprise Linux & Fedora 4th Edition.chm new file mode 100644 index 0000000000000000000000000000000000000000..7f47639124a9a745ae1597701171b0a9df20b714 --- /dev/null +++ b/Operating Systems/Linux/Essentials Books/O'Reilly - Learning Red Hat Enterprise Linux & Fedora 4th Edition.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9d963e7f7ada6b1c386c0647d0a591ec0570e74e250ad94d244a9efc4a9f6483 +size 4594902 diff --git a/Operating Systems/Linux/Essentials Books/O'Reilly - Learning the Vi Editor 6th Edition.pdf b/Operating Systems/Linux/Essentials Books/O'Reilly - Learning the Vi Editor 6th Edition.pdf new file mode 100644 index 0000000000000000000000000000000000000000..d34faae25db791597a3d7c8633331d4961e0c427 --- /dev/null +++ b/Operating Systems/Linux/Essentials Books/O'Reilly - Learning the Vi Editor 6th Edition.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b8603097d64a2d7b1328b6aea022eb0e8ec9a8b98add26b1195896cc1ec87fe2 +size 1992369 diff --git a/Operating Systems/Linux/Essentials Books/O'Reilly - Learning the bash Shell 2nd Edition.chm b/Operating Systems/Linux/Essentials Books/O'Reilly - Learning the bash Shell 2nd Edition.chm new file mode 100644 index 0000000000000000000000000000000000000000..cb350f9033463068bc8dcefbe272287ac2251278 Binary files /dev/null and b/Operating Systems/Linux/Essentials Books/O'Reilly - Learning the bash Shell 2nd Edition.chm differ diff --git a/Operating Systems/Linux/Essentials Books/O'Reilly - UNIX Power Tools 3rd Edition.chm b/Operating Systems/Linux/Essentials Books/O'Reilly - UNIX Power Tools 3rd Edition.chm new file mode 100644 index 0000000000000000000000000000000000000000..a9007448d91787e9fb39e0756af70da042ee86a8 --- /dev/null +++ b/Operating Systems/Linux/Essentials Books/O'Reilly - UNIX Power Tools 3rd Edition.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8f51794271a61b46467fd4f6a6a58e322035085c480ed39f519417d808a9a686 +size 2349896 diff --git a/Operating Systems/Linux/Essentials Books/O'Reilly - UNIX in a Nutshell 3rd Edition.chm b/Operating Systems/Linux/Essentials Books/O'Reilly - UNIX in a Nutshell 3rd Edition.chm new file mode 100644 index 0000000000000000000000000000000000000000..dbd2028b8197766d5d00038289933c2887e9d272 --- /dev/null +++ b/Operating Systems/Linux/Essentials Books/O'Reilly - UNIX in a Nutshell 3rd Edition.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:325a8803bb2f1b798d27d9ba71052796109544ab2e9e1dd832a30176b3c559ff +size 1228300 diff --git a/Operating Systems/Linux/Essentials Books/OReilly Linux System Administration.pdf b/Operating Systems/Linux/Essentials Books/OReilly Linux System Administration.pdf new file mode 100644 index 0000000000000000000000000000000000000000..80da8aa0976a91b2e75f13ef7ae326fdf5cf0333 --- /dev/null +++ b/Operating Systems/Linux/Essentials Books/OReilly Linux System Administration.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:acd4f70d7c26d7a49834f6e7ab5eef8c9e6a81594b8b4a52a4a3a22e7225f984 +size 3581533 diff --git a/Operating Systems/Linux/Essentials Books/Red Hat Linux Security And Optimization.pdf b/Operating Systems/Linux/Essentials Books/Red Hat Linux Security And Optimization.pdf new file mode 100644 index 0000000000000000000000000000000000000000..cca4987c7e39e53c4d30348b7796971404f51009 --- /dev/null +++ b/Operating Systems/Linux/Essentials Books/Red Hat Linux Security And Optimization.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:882d67c2e85bf2e152287ad5850c1b559a44d1aa586abb12ed02a66d90a52f5d +size 5378196 diff --git a/Operating Systems/Linux/Essentials Books/User Mode Linux.chm b/Operating Systems/Linux/Essentials Books/User Mode Linux.chm new file mode 100644 index 0000000000000000000000000000000000000000..3cb5dc0fbd89b7a53d8c11e193ce13365c621e2c --- /dev/null +++ b/Operating Systems/Linux/Essentials Books/User Mode Linux.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3ccd33a4ac06e8b69b36a6d1b83bedcd1f2ef03ffa5d99cacd13b5641a98a5b4 +size 1195931 diff --git a/Operating Systems/Linux/Essentials Networking (VPN,SNMP,Load Balancing)/O'Reilly - Server Load Balancing.pdf b/Operating Systems/Linux/Essentials Networking (VPN,SNMP,Load Balancing)/O'Reilly - Server Load Balancing.pdf new file mode 100644 index 0000000000000000000000000000000000000000..b9d3df4bcb09adc46d855a2032d73043d71e2881 --- /dev/null +++ b/Operating Systems/Linux/Essentials Networking (VPN,SNMP,Load Balancing)/O'Reilly - Server Load Balancing.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6aaf88f119b7d93879e1f08bb052e86844c39d1dc10a89b5fb8a04f47e7f4830 +size 3272622 diff --git a/Operating Systems/Linux/Essentials Networking (VPN,SNMP,Load Balancing)/OpenVPN - Building And Integrating Virtual Private Networks.pdf b/Operating Systems/Linux/Essentials Networking (VPN,SNMP,Load Balancing)/OpenVPN - Building And Integrating Virtual Private Networks.pdf new file mode 100644 index 0000000000000000000000000000000000000000..b4dafdaed98251267d8363c628618fa05b8d2d3f --- /dev/null +++ b/Operating Systems/Linux/Essentials Networking (VPN,SNMP,Load Balancing)/OpenVPN - Building And Integrating Virtual Private Networks.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d6d4bc784c54aac1a6f8ac5d70523e0cd7c7fd6ea5b7c00b4fa9e6bf34acdf0f +size 9774672 diff --git a/Operating Systems/Linux/For Dummies - Linux All in One Desk Reference.pdf b/Operating Systems/Linux/For Dummies - Linux All in One Desk Reference.pdf new file mode 100644 index 0000000000000000000000000000000000000000..2c3273d7d8ba46fcf5b2ccfa6336bbdaf0343815 --- /dev/null +++ b/Operating Systems/Linux/For Dummies - Linux All in One Desk Reference.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ecb42c3994a1eba2d4b4cdb8de98022f55828702ba20147a5a357c6bc715a055 +size 20699083 diff --git a/Operating Systems/Linux/For Dummies - Linux Timesaving Techniques.pdf b/Operating Systems/Linux/For Dummies - Linux Timesaving Techniques.pdf new file mode 100644 index 0000000000000000000000000000000000000000..a845087e950464f120a13c65a6d33da32f6bd625 --- /dev/null +++ b/Operating Systems/Linux/For Dummies - Linux Timesaving Techniques.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8734ccdd6643a2f8db8b92bef9070ebee77b2c53fe8b34a3033603efa02d51c3 +size 16130050 diff --git a/Operating Systems/Linux/IDS Firewall & Security/Configuring IPCop Firewalls - Closing Borders With Open Source.pdf b/Operating Systems/Linux/IDS Firewall & Security/Configuring IPCop Firewalls - Closing Borders With Open Source.pdf new file mode 100644 index 0000000000000000000000000000000000000000..7989a39618701b1c1bfd5bc1ba302b8a128c3007 --- /dev/null +++ b/Operating Systems/Linux/IDS Firewall & Security/Configuring IPCop Firewalls - Closing Borders With Open Source.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8f57c567d5b0696d5a2706a9ec266e7ca5269afa7acf48b8dc172eddf083633a +size 11311241 diff --git a/Operating Systems/Linux/IDS Firewall & Security/O'Reilly Building Internet Firewalls 2nd Edition.pdf b/Operating Systems/Linux/IDS Firewall & Security/O'Reilly Building Internet Firewalls 2nd Edition.pdf new file mode 100644 index 0000000000000000000000000000000000000000..8124f822e4b8b1fdc26f59505d115cef6e81b3de --- /dev/null +++ b/Operating Systems/Linux/IDS Firewall & Security/O'Reilly Building Internet Firewalls 2nd Edition.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5da0b17d634e0082506d7241f7cd5b22da7e20afa84009af7c3ad1c9a1275f6a +size 5134597 diff --git a/Operating Systems/Linux/IDS Firewall & Security/O'Reilly - Building Secure Servers with Linux.chm b/Operating Systems/Linux/IDS Firewall & Security/O'Reilly - Building Secure Servers with Linux.chm new file mode 100644 index 0000000000000000000000000000000000000000..f1cfea03c7d36d202da4cef881723640d73edb3a --- /dev/null +++ b/Operating Systems/Linux/IDS Firewall & Security/O'Reilly - Building Secure Servers with Linux.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0406e531763398cfa95ae9d27f64faa5afb7e9c7c146864bb7ff555ea1353f76 +size 1597318 diff --git a/Operating Systems/Linux/IDS Firewall & Security/O'Reilly - Building Secure Servers with Linux.pdf b/Operating Systems/Linux/IDS Firewall & Security/O'Reilly - Building Secure Servers with Linux.pdf new file mode 100644 index 0000000000000000000000000000000000000000..3300c8f352480e375196a82e799c8b1808f75782 --- /dev/null +++ b/Operating Systems/Linux/IDS Firewall & Security/O'Reilly - Building Secure Servers with Linux.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4c7fb8700e4fe32ec64b47ca63bf4117441b81985b5a1ea3179b8cc15b5e092a +size 2972965 diff --git a/Operating Systems/Linux/IDS Firewall & Security/O'Reilly - Network Security with OpenSSL.pdf b/Operating Systems/Linux/IDS Firewall & Security/O'Reilly - Network Security with OpenSSL.pdf new file mode 100644 index 0000000000000000000000000000000000000000..43b6d1edd705aad08741c4d5979e6fb210fcdfcd --- /dev/null +++ b/Operating Systems/Linux/IDS Firewall & Security/O'Reilly - Network Security with OpenSSL.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cbae72ebbc361313e75298b0894d3ad7ca398517c064941e281d50587550b0b9 +size 2526496 diff --git a/Operating Systems/Linux/IDS Firewall & Security/O'Reilly - RADIUS.chm b/Operating Systems/Linux/IDS Firewall & Security/O'Reilly - RADIUS.chm new file mode 100644 index 0000000000000000000000000000000000000000..70142b492a5c1a577f0a36700be0d53ad91c9f9e Binary files /dev/null and b/Operating Systems/Linux/IDS Firewall & Security/O'Reilly - RADIUS.chm differ diff --git a/Operating Systems/Linux/IDS Firewall & Security/Security Warrior.chm b/Operating Systems/Linux/IDS Firewall & Security/Security Warrior.chm new file mode 100644 index 0000000000000000000000000000000000000000..450e34f4ebc0cca9e39bb3ea59518a0f85c0b3bc --- /dev/null +++ b/Operating Systems/Linux/IDS Firewall & Security/Security Warrior.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:440515ee9f16dbb761af1843aafcc0544524617b96ead5cb0da873b5a841f956 +size 4725317 diff --git a/Operating Systems/Linux/IDS Firewall & Security/Snort For Dummies.pdf b/Operating Systems/Linux/IDS Firewall & Security/Snort For Dummies.pdf new file mode 100644 index 0000000000000000000000000000000000000000..dfb71712f500347c2d78f48542250e5690956255 --- /dev/null +++ b/Operating Systems/Linux/IDS Firewall & Security/Snort For Dummies.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b6ce86007f3db8b0e0d83e9d06a718cc59ac0c45d3a7f217d135f00946366637 +size 7389326 diff --git a/Operating Systems/Linux/IDS Firewall & Security/The Best Damn Firewall Book Period .pdf b/Operating Systems/Linux/IDS Firewall & Security/The Best Damn Firewall Book Period .pdf new file mode 100644 index 0000000000000000000000000000000000000000..3e2b69416c26a4e4a6cce07de3f8f7a25789a8fb --- /dev/null +++ b/Operating Systems/Linux/IDS Firewall & Security/The Best Damn Firewall Book Period .pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:85b66d92a79cfad72dad0493aaaef6b3afc7cf51001b92ad080735abf485ee7c +size 22623758 diff --git a/Operating Systems/Linux/IDS Firewall & Security/The Tao Of Network Security Monitoring - Beyond Intrusion Detection.chm b/Operating Systems/Linux/IDS Firewall & Security/The Tao Of Network Security Monitoring - Beyond Intrusion Detection.chm new file mode 100644 index 0000000000000000000000000000000000000000..34f7ae4d07243dc0bd85fbaa59bbdfc6991db3dc --- /dev/null +++ b/Operating Systems/Linux/IDS Firewall & Security/The Tao Of Network Security Monitoring - Beyond Intrusion Detection.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ddee147c52bc9f0df298fb6ba4213185d0b2de1eef44ee7a9614033b411c68c2 +size 16217260 diff --git a/Operating Systems/Linux/LDAP/O'Reilly - LDAP System Administration.chm b/Operating Systems/Linux/LDAP/O'Reilly - LDAP System Administration.chm new file mode 100644 index 0000000000000000000000000000000000000000..eca7c9ec6a583c84f52a3dcac6b874fc8acf1087 --- /dev/null +++ b/Operating Systems/Linux/LDAP/O'Reilly - LDAP System Administration.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:78045b6765767f39d666acd9b0aa16207ace25d8bea72ec1b9f06b7adaa85255 +size 1160305 diff --git a/Operating Systems/Linux/Linux All-In-One Desk Reference For Dummies (Feb 2005).pdf b/Operating Systems/Linux/Linux All-In-One Desk Reference For Dummies (Feb 2005).pdf new file mode 100644 index 0000000000000000000000000000000000000000..ee60715c13935f0c48f81ef14d2ea5b533433733 --- /dev/null +++ b/Operating Systems/Linux/Linux All-In-One Desk Reference For Dummies (Feb 2005).pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5ca75dc027a1c372fb848003746eb24876aa862f802e81a037ce746467629836 +size 19921677 diff --git a/Operating Systems/Linux/Linux For Dummies, 6th Edition (Feb 2005).pdf b/Operating Systems/Linux/Linux For Dummies, 6th Edition (Feb 2005).pdf new file mode 100644 index 0000000000000000000000000000000000000000..858fd843b00b8b114e4a2a9076c7e569f7280087 --- /dev/null +++ b/Operating Systems/Linux/Linux For Dummies, 6th Edition (Feb 2005).pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b78d17e1676b88614362bb9bea659578b47c093bbbb6fd3cc71f24a31207e051 +size 11633856 diff --git a/Operating Systems/Linux/Mail Servers/No Starch Press - The Book of Postfix.pdf b/Operating Systems/Linux/Mail Servers/No Starch Press - The Book of Postfix.pdf new file mode 100644 index 0000000000000000000000000000000000000000..b37e3d8b2dcc2337dd8f1d6ebef648af3d02969d --- /dev/null +++ b/Operating Systems/Linux/Mail Servers/No Starch Press - The Book of Postfix.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0650a22c82051aa2f0e09dde7dff2cbc7b7db237ac1c29a8b85867e891fb7106 +size 56279597 diff --git a/Operating Systems/Linux/Mail Servers/O'Reilly - Postfix The Definitive Guide.pdf b/Operating Systems/Linux/Mail Servers/O'Reilly - Postfix The Definitive Guide.pdf new file mode 100644 index 0000000000000000000000000000000000000000..e19b2da32c2d1e47ba8f3d2078160bacf4fbe5b1 --- /dev/null +++ b/Operating Systems/Linux/Mail Servers/O'Reilly - Postfix The Definitive Guide.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2766105855797c12b06fa272e87ca1328a2e8d5abb00126c74d035156b91df8c +size 2721408 diff --git a/Operating Systems/Linux/Mail Servers/O'Reilly SendMail Cookbook.chm b/Operating Systems/Linux/Mail Servers/O'Reilly SendMail Cookbook.chm new file mode 100644 index 0000000000000000000000000000000000000000..b9a475e2bb71af2b535ca4d3f40dcc46d1a6eb9c Binary files /dev/null and b/Operating Systems/Linux/Mail Servers/O'Reilly SendMail Cookbook.chm differ diff --git a/Operating Systems/Linux/Mail Servers/OReilly sendmail 4th Edition.pdf b/Operating Systems/Linux/Mail Servers/OReilly sendmail 4th Edition.pdf new file mode 100644 index 0000000000000000000000000000000000000000..740b9bcf3d5bbf7ce9c2a23ad26646faa0fe7cbb --- /dev/null +++ b/Operating Systems/Linux/Mail Servers/OReilly sendmail 4th Edition.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:053ecd4610e8aa99ad7728f28386cd6f297011ab0c6868732dc39c25e5dc3b69 +size 7502611 diff --git a/Operating Systems/Linux/Mail Servers/Packtpub.Linux.Email.Nov.2009.pdf b/Operating Systems/Linux/Mail Servers/Packtpub.Linux.Email.Nov.2009.pdf new file mode 100644 index 0000000000000000000000000000000000000000..6cef1dfbb0eb212a877c62c0b82ef2c03a7f6b5e --- /dev/null +++ b/Operating Systems/Linux/Mail Servers/Packtpub.Linux.Email.Nov.2009.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:69f1e975f52be99e5dc1e9f81f0024a1b35bf75723c95b88f31a4eb766278f34 +size 4072962 diff --git a/Operating Systems/Linux/NFS & NIS/O'Reilly - Managing NFS and NIS 2nd Edt.pdf b/Operating Systems/Linux/NFS & NIS/O'Reilly - Managing NFS and NIS 2nd Edt.pdf new file mode 100644 index 0000000000000000000000000000000000000000..1c46f96d81bdee2375fb8f8b31b9994a2804f158 --- /dev/null +++ b/Operating Systems/Linux/NFS & NIS/O'Reilly - Managing NFS and NIS 2nd Edt.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8bfba826a71479d2d9602e0e68497e29fd181d74d22db0287c9e09f39797a4a2 +size 2035071 diff --git a/Operating Systems/Linux/No Starch Press - Linux Appliance Design.pdf b/Operating Systems/Linux/No Starch Press - Linux Appliance Design.pdf new file mode 100644 index 0000000000000000000000000000000000000000..c446bfdc2d95dc4aff02405a0539ac306a26c685 --- /dev/null +++ b/Operating Systems/Linux/No Starch Press - Linux Appliance Design.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1ae4d431043d8c295ba3011278d85de581401350a698dc2ff61d762bf6a6003e +size 8881651 diff --git a/Operating Systems/Linux/No Starch Press - Linux Firewalls.pdf b/Operating Systems/Linux/No Starch Press - Linux Firewalls.pdf new file mode 100644 index 0000000000000000000000000000000000000000..d6d9e0fc61f5e2e1ff932f43195d0995f416942d --- /dev/null +++ b/Operating Systems/Linux/No Starch Press - Linux Firewalls.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8126f2851d97b2e783ac9a99f9fea379eea16f2e633c11f0334b06d26c61d794 +size 6823226 diff --git a/Operating Systems/Linux/No Starch Press - Linux For Non-Geeks (2004).chm b/Operating Systems/Linux/No Starch Press - Linux For Non-Geeks (2004).chm new file mode 100644 index 0000000000000000000000000000000000000000..abcc7d27893bc56f6b3a275b273be95f698224b4 --- /dev/null +++ b/Operating Systems/Linux/No Starch Press - Linux For Non-Geeks (2004).chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:04c3ba137ba59e4b2d83a3e4709f2ce5c05f96cc988b52c44dab88377bd38eff +size 15491153 diff --git a/Operating Systems/Linux/No Starch Press - Programming Linux Games.pdf b/Operating Systems/Linux/No Starch Press - Programming Linux Games.pdf new file mode 100644 index 0000000000000000000000000000000000000000..b44ab046b2ebeaba0e6b57ea94c6c6239c9947fa --- /dev/null +++ b/Operating Systems/Linux/No Starch Press - Programming Linux Games.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d98a2a59334ce74610ba0ca955cd2a890020327cee61c7f56de94ed02e79730a +size 1859038 diff --git a/Operating Systems/Linux/O'Reilly - Linux Cookbook.chm b/Operating Systems/Linux/O'Reilly - Linux Cookbook.chm new file mode 100644 index 0000000000000000000000000000000000000000..2b93e4af874b4a077614606351866a49f2a28df3 --- /dev/null +++ b/Operating Systems/Linux/O'Reilly - Linux Cookbook.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6436c29dd22526c3ecc769e2f71acc173f9940721ed6f8b0bcf00e9a71b6ba3d +size 1698083 diff --git a/Operating Systems/Linux/O'Reilly - Linux Networking Cookbook.pdf b/Operating Systems/Linux/O'Reilly - Linux Networking Cookbook.pdf new file mode 100644 index 0000000000000000000000000000000000000000..102314ceeff419bebaa53f0b54d2417332c30b58 --- /dev/null +++ b/Operating Systems/Linux/O'Reilly - Linux Networking Cookbook.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e93af0158b868196f9579f399facf192f515aeb2bea912b234bffcb5d03014ab +size 4869213 diff --git a/Operating Systems/Linux/O'Reilly - Linux Security Cookbook.chm b/Operating Systems/Linux/O'Reilly - Linux Security Cookbook.chm new file mode 100644 index 0000000000000000000000000000000000000000..7177f1972cd6aba4a0f5e3ae86d143e3894728eb Binary files /dev/null and b/Operating Systems/Linux/O'Reilly - Linux Security Cookbook.chm differ diff --git a/Operating Systems/Linux/O'Reilly Fedora Linux.chm b/Operating Systems/Linux/O'Reilly Fedora Linux.chm new file mode 100644 index 0000000000000000000000000000000000000000..9f429df944e2d10d0a9c8659fc886e9f1eead86d --- /dev/null +++ b/Operating Systems/Linux/O'Reilly Fedora Linux.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2b99101d8e299291c98e5c901211e956e7f1cb4c9e2606a05e50eebadab07b74 +size 8351634 diff --git a/Operating Systems/Linux/O'Reilly LPI Linux Certification In A Nutshell (2nd Edition).chm b/Operating Systems/Linux/O'Reilly LPI Linux Certification In A Nutshell (2nd Edition).chm new file mode 100644 index 0000000000000000000000000000000000000000..89d6a28221d941b2d25f9c98a744e766b669904a --- /dev/null +++ b/Operating Systems/Linux/O'Reilly LPI Linux Certification In A Nutshell (2nd Edition).chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6a2110275ea794f8f873fed5bfc55d15f6a9666cf05e6caaf9ad04c7749eb082 +size 2403736 diff --git a/Operating Systems/Linux/O'Reilly Linux Desktop Hacks.chm b/Operating Systems/Linux/O'Reilly Linux Desktop Hacks.chm new file mode 100644 index 0000000000000000000000000000000000000000..23d7f877eeba423e5e4a4ab303bd0b72ea4e7b9b --- /dev/null +++ b/Operating Systems/Linux/O'Reilly Linux Desktop Hacks.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:413b043dfc5d113438e9015a3f8677f8c116d880e08e75bd52c4603a1b445114 +size 2592644 diff --git a/Operating Systems/Linux/O'Reilly Linux Device Drivers (3rd Edition).chm b/Operating Systems/Linux/O'Reilly Linux Device Drivers (3rd Edition).chm new file mode 100644 index 0000000000000000000000000000000000000000..2d396ba8508a3612d9327a5d43464b1dca091ec1 --- /dev/null +++ b/Operating Systems/Linux/O'Reilly Linux Device Drivers (3rd Edition).chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1536f166c4ab7b3a6473a3e0f49614227b5905eeaedc39e664162750e75c337e +size 1379968 diff --git a/Operating Systems/Linux/O'Reilly Linux Kernel in a Nutshell.pdf b/Operating Systems/Linux/O'Reilly Linux Kernel in a Nutshell.pdf new file mode 100644 index 0000000000000000000000000000000000000000..6794a0b50b46558e99f4128c7d563787d44f3443 --- /dev/null +++ b/Operating Systems/Linux/O'Reilly Linux Kernel in a Nutshell.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5adae9578d23d62f10d8970dc62d455877f7862dfe76c47583001cbc7e21e58f +size 3347565 diff --git a/Operating Systems/Linux/O'Reilly Linux Multimedia Hacks.chm b/Operating Systems/Linux/O'Reilly Linux Multimedia Hacks.chm new file mode 100644 index 0000000000000000000000000000000000000000..1eb0e3458593700a3e024f2d50801f50efc92821 --- /dev/null +++ b/Operating Systems/Linux/O'Reilly Linux Multimedia Hacks.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:258fa3f636b77bd11b7126104d15c450098151ffe323e3a3516c39d7d961f2f0 +size 1749844 diff --git a/Operating Systems/Linux/O'Reilly Linux Network Administrator's Guide (3rd Edition).chm b/Operating Systems/Linux/O'Reilly Linux Network Administrator's Guide (3rd Edition).chm new file mode 100644 index 0000000000000000000000000000000000000000..59923ca2a5e538dc38f87ad6bbefdac50ce63c4e --- /dev/null +++ b/Operating Systems/Linux/O'Reilly Linux Network Administrator's Guide (3rd Edition).chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2843b014a034d9609faed7a17ffe05513e77662de78ada13550595452a94de85 +size 1132401 diff --git a/Operating Systems/Linux/O'Reilly Linux Server Hacks.chm b/Operating Systems/Linux/O'Reilly Linux Server Hacks.chm new file mode 100644 index 0000000000000000000000000000000000000000..d8678f8385c39d9175cd7b5be9bb8dc20f3b409b Binary files /dev/null and b/Operating Systems/Linux/O'Reilly Linux Server Hacks.chm differ diff --git a/Operating Systems/Linux/O'Reilly Linux System Administration.pdf b/Operating Systems/Linux/O'Reilly Linux System Administration.pdf new file mode 100644 index 0000000000000000000000000000000000000000..686be853798b03ef6e626ecfab3e2357117a98f5 --- /dev/null +++ b/Operating Systems/Linux/O'Reilly Linux System Administration.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ec6d11336c92c16e884ca25efdefa2d903cfba8483dc30aca1f068b5c11bc020 +size 2884412 diff --git a/Operating Systems/Linux/O'Reilly Linux System Programming.pdf b/Operating Systems/Linux/O'Reilly Linux System Programming.pdf new file mode 100644 index 0000000000000000000000000000000000000000..b152b8f28a204266eca16477fbbd48c407e68616 --- /dev/null +++ b/Operating Systems/Linux/O'Reilly Linux System Programming.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a0011e9b5dcd7897953be8aafc43b119d0e76539ccd019f80cb7eccb59131a22 +size 1276445 diff --git a/Operating Systems/Linux/O'Reilly Linux in a Nutshell (5th Edition).chm b/Operating Systems/Linux/O'Reilly Linux in a Nutshell (5th Edition).chm new file mode 100644 index 0000000000000000000000000000000000000000..5d28d979e37f873795d9b24a1a707fed64febcaf --- /dev/null +++ b/Operating Systems/Linux/O'Reilly Linux in a Nutshell (5th Edition).chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:70485cedae7bb503b024851eab2b5813336b1ac3368423cc7ccec344687cd6d0 +size 1428677 diff --git a/Operating Systems/Linux/O'Reilly Running Linux (5th Edition).chm b/Operating Systems/Linux/O'Reilly Running Linux (5th Edition).chm new file mode 100644 index 0000000000000000000000000000000000000000..5cb6aa9e0521eb359bd9206718967e4582d0fe27 --- /dev/null +++ b/Operating Systems/Linux/O'Reilly Running Linux (5th Edition).chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:215bcafcb564545cd079b9dac5f8db06bd8ec49b7ff0d8391d77263c0ac11a3b +size 5673818 diff --git a/Operating Systems/Linux/O'Reilly SUSE Linux.chm b/Operating Systems/Linux/O'Reilly SUSE Linux.chm new file mode 100644 index 0000000000000000000000000000000000000000..78eff38e63c6b2fc189aad8e0fbc468c5412580f --- /dev/null +++ b/Operating Systems/Linux/O'Reilly SUSE Linux.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:af1e1b96ff8193d578f1973c627722c9e884de9ea65a68b82020961c6bf7e02f +size 5744234 diff --git a/Operating Systems/Linux/O'Reilly Understanding The Linux Kernel (3rd Edition).pdf b/Operating Systems/Linux/O'Reilly Understanding The Linux Kernel (3rd Edition).pdf new file mode 100644 index 0000000000000000000000000000000000000000..7b15409d0096dd04db313028a9b86548257fbb2a --- /dev/null +++ b/Operating Systems/Linux/O'Reilly Understanding The Linux Kernel (3rd Edition).pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:24dbcd934604f04d190d82d26fb127ae6f50e88cf12856d5687d4bbdb0fb6bba +size 11048654 diff --git a/Operating Systems/Linux/O'Reilly_-_Linux_Command_Directory.pdf b/Operating Systems/Linux/O'Reilly_-_Linux_Command_Directory.pdf new file mode 100644 index 0000000000000000000000000000000000000000..564fa17fc53c16b71e9802bd18bf03a7b03f901c --- /dev/null +++ b/Operating Systems/Linux/O'Reilly_-_Linux_Command_Directory.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4f89e7a1ba0c492544c63c90bf231de74b9e2c9476ac20211323d32815d364e2 +size 10896088 diff --git a/Operating Systems/Linux/OReilly - Linux Networking Cookbook.pdf b/Operating Systems/Linux/OReilly - Linux Networking Cookbook.pdf new file mode 100644 index 0000000000000000000000000000000000000000..28770daa655949c0eb4f9fda15d5b734a5ffaac0 --- /dev/null +++ b/Operating Systems/Linux/OReilly - Linux Networking Cookbook.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1d9dd78ab710faa84945f064e17ea2e4b4c2947d6cba13c2fb470434d280e517 +size 4723710 diff --git a/Operating Systems/Linux/OReilly.Building.Embedded.Linux.Systems.Aug.2008.eBook-DDU.pdf b/Operating Systems/Linux/OReilly.Building.Embedded.Linux.Systems.Aug.2008.eBook-DDU.pdf new file mode 100644 index 0000000000000000000000000000000000000000..05a845dcac056f4f49e45f53399fb3ed25360b45 --- /dev/null +++ b/Operating Systems/Linux/OReilly.Building.Embedded.Linux.Systems.Aug.2008.eBook-DDU.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9a930926d2a08126ade14a5597cbeb119df34b3dc90187170fa64cc05282c10f +size 4429254 diff --git a/Operating Systems/Linux/Red Hat Enterprise Linux 4 for Dummies (Apr 2005).pdf b/Operating Systems/Linux/Red Hat Enterprise Linux 4 for Dummies (Apr 2005).pdf new file mode 100644 index 0000000000000000000000000000000000000000..d79a04b7b8ec33e50514750af1a7a24e88a1b807 --- /dev/null +++ b/Operating Systems/Linux/Red Hat Enterprise Linux 4 for Dummies (Apr 2005).pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cc0b96e61e00b9a53196d09fdb87ffa7ee24525c615dcd987eb9879d416b6698 +size 13862619 diff --git a/Operating Systems/Linux/Red Hat Linux Fedora for Dummies.chm b/Operating Systems/Linux/Red Hat Linux Fedora for Dummies.chm new file mode 100644 index 0000000000000000000000000000000000000000..3fefe6882831dd87a9129ccf1d3aa7cad6e23126 --- /dev/null +++ b/Operating Systems/Linux/Red Hat Linux Fedora for Dummies.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d9b0c8f9366206b62a7bb311a7f12b2a64285423bf58bf11a9c08e817c967fb +size 5255188 diff --git a/Operating Systems/Linux/Red Hat Software - Linux Complete Command Reference.pdf b/Operating Systems/Linux/Red Hat Software - Linux Complete Command Reference.pdf new file mode 100644 index 0000000000000000000000000000000000000000..9e66c1d422d305bc9fdf0564aeb250cbb5957be8 --- /dev/null +++ b/Operating Systems/Linux/Red Hat Software - Linux Complete Command Reference.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c0588722793f8d1318c13ce2a0b4ae4ad70c3f8f387fa36e942ebcf83e668ec +size 19262299 diff --git a/Operating Systems/Linux/SELinux/SELinux By Example - Using Security Enhanced Linux .chm b/Operating Systems/Linux/SELinux/SELinux By Example - Using Security Enhanced Linux .chm new file mode 100644 index 0000000000000000000000000000000000000000..d8080997738a53fb7c2efcc1628d2c56d933b96f --- /dev/null +++ b/Operating Systems/Linux/SELinux/SELinux By Example - Using Security Enhanced Linux .chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:849123096edb4d464ea5ca2d067c7e67ed97e94cddadd6056d1ce3f35fbb88dc +size 4121770 diff --git a/Operating Systems/Linux/SSH & VPN/IPsec Virtual Private Network Fundamentals.chm b/Operating Systems/Linux/SSH & VPN/IPsec Virtual Private Network Fundamentals.chm new file mode 100644 index 0000000000000000000000000000000000000000..b9671d626e508944ac1e0cd754777552709d1964 --- /dev/null +++ b/Operating Systems/Linux/SSH & VPN/IPsec Virtual Private Network Fundamentals.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cf3b4bf19e7923be9f03de77302373c971816637feb140fce86d12f96884373e +size 19213006 diff --git a/Operating Systems/Linux/SSH & VPN/O'Reilly - SSH The Secure Shell The Definitive Guide.pdf b/Operating Systems/Linux/SSH & VPN/O'Reilly - SSH The Secure Shell The Definitive Guide.pdf new file mode 100644 index 0000000000000000000000000000000000000000..5be8805b7fd7bfe9fb9309c9063806332fe281bd --- /dev/null +++ b/Operating Systems/Linux/SSH & VPN/O'Reilly - SSH The Secure Shell The Definitive Guide.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b0024215f2e97cbc70614ba5d645efdfaee9994a524f00b6136511633a172272 +size 3297413 diff --git a/Operating Systems/Linux/SSH & VPN/O'Reilly - The Secure Shell The Definitive Guide.pdf b/Operating Systems/Linux/SSH & VPN/O'Reilly - The Secure Shell The Definitive Guide.pdf new file mode 100644 index 0000000000000000000000000000000000000000..f5fffc2366095785b1e187682c902d42fd4b0c5e --- /dev/null +++ b/Operating Systems/Linux/SSH & VPN/O'Reilly - The Secure Shell The Definitive Guide.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e7d97ee90a3ad9a7fcefeef100c305927bc964f58c9b1830e2567b9cd733fb7f +size 3301272 diff --git a/Operating Systems/Linux/SUSE Linux 10 for Dummies.pdf b/Operating Systems/Linux/SUSE Linux 10 for Dummies.pdf new file mode 100644 index 0000000000000000000000000000000000000000..5abd907c4b5cd931f18f8ed435151c9958c4a535 --- /dev/null +++ b/Operating Systems/Linux/SUSE Linux 10 for Dummies.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f9e8828d63bf71b1e12efbb56ff59123ce2420b5eeee952c25b608dedffa11dd +size 13670327 diff --git a/Operating Systems/Linux/SUSE Linux 9.3 For Dummies (May 2005).pdf b/Operating Systems/Linux/SUSE Linux 9.3 For Dummies (May 2005).pdf new file mode 100644 index 0000000000000000000000000000000000000000..3cd628e65b245ee4abbce40e7052f6a2fb884a80 --- /dev/null +++ b/Operating Systems/Linux/SUSE Linux 9.3 For Dummies (May 2005).pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f294d2a80b1121d198cbb66855ea48024e16b9493846cc1bd32c54335ebd9a92 +size 11143507 diff --git a/Operating Systems/Linux/Samba/O'Reilly Samba 2nd Edition.chm b/Operating Systems/Linux/Samba/O'Reilly Samba 2nd Edition.chm new file mode 100644 index 0000000000000000000000000000000000000000..2cd3963d992a0d7fadd37cdd8051fce9f29f5636 --- /dev/null +++ b/Operating Systems/Linux/Samba/O'Reilly Samba 2nd Edition.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fcedc369f58eff5f09eebe485c4363375d9f6ba3026135910acd95b84ffcdd3c +size 3310308 diff --git a/Operating Systems/Linux/Sams - Teach Yourself Linux in 24 Hours.pdf b/Operating Systems/Linux/Sams - Teach Yourself Linux in 24 Hours.pdf new file mode 100644 index 0000000000000000000000000000000000000000..ad37cd8834822984db55373c8acf1e2beca7d3c9 --- /dev/null +++ b/Operating Systems/Linux/Sams - Teach Yourself Linux in 24 Hours.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d2546f0ad7ab45c6564679c555887116b7ec1eae8f6101a528e3982a444c85f +size 12620189 diff --git a/Operating Systems/Linux/Squid/O'Reilly - Squid The Definitive Guide.chm b/Operating Systems/Linux/Squid/O'Reilly - Squid The Definitive Guide.chm new file mode 100644 index 0000000000000000000000000000000000000000..ea3bc5ba957843bebed65a4acef6261278009bcc --- /dev/null +++ b/Operating Systems/Linux/Squid/O'Reilly - Squid The Definitive Guide.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e4336218c4cd5b64b9a1b24aa8d2bcdcafa112650fef4e41f64d42693d91adc +size 1083938 diff --git a/Operating Systems/Linux/Squid/O'Reilly - Squid The Definitive Guide.pdf b/Operating Systems/Linux/Squid/O'Reilly - Squid The Definitive Guide.pdf new file mode 100644 index 0000000000000000000000000000000000000000..33e25f8c33704ac58c7d995dc07beebea023a9c2 --- /dev/null +++ b/Operating Systems/Linux/Squid/O'Reilly - Squid The Definitive Guide.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a150502f66898a140aff6b467d58aed546e9c06993ebaf1a63f7c8459f57a2c4 +size 4355546 diff --git a/Operating Systems/Linux/Web Server - Apache/O'Reilly Apache The Definitive Guide 3rd Edition.chm b/Operating Systems/Linux/Web Server - Apache/O'Reilly Apache The Definitive Guide 3rd Edition.chm new file mode 100644 index 0000000000000000000000000000000000000000..8d68108c51b2a84683783418334559328ad7de97 Binary files /dev/null and b/Operating Systems/Linux/Web Server - Apache/O'Reilly Apache The Definitive Guide 3rd Edition.chm differ diff --git a/Operating Systems/Linux/Wiley.Linux.Bible.2009.Edition.Jan.2009.pdf b/Operating Systems/Linux/Wiley.Linux.Bible.2009.Edition.Jan.2009.pdf new file mode 100644 index 0000000000000000000000000000000000000000..23af20360631eb983e28daa5111cdc26baae0453 --- /dev/null +++ b/Operating Systems/Linux/Wiley.Linux.Bible.2009.Edition.Jan.2009.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fe63e9c00aa446046f4893b91a3d35d78f5a99fef888db36e5003605600dc937 +size 37924523 diff --git a/Operating Systems/OSX/Mac OS X GUI Customization for Dummies v2.1.pdf b/Operating Systems/OSX/Mac OS X GUI Customization for Dummies v2.1.pdf new file mode 100644 index 0000000000000000000000000000000000000000..219168c46e32c23b2eb8568cd1c2596dcd470abf --- /dev/null +++ b/Operating Systems/OSX/Mac OS X GUI Customization for Dummies v2.1.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a80b989d2fd9996ec8cca28dd4954a6235b102ff7c3118eae3dc573a83b099d1 +size 9024324 diff --git a/Operating Systems/OSX/Mac OS X Panther Timesaving Techniques for Dummies (2004).pdf b/Operating Systems/OSX/Mac OS X Panther Timesaving Techniques for Dummies (2004).pdf new file mode 100644 index 0000000000000000000000000000000000000000..4c5dc4ed5b4e4e13d5addbbd7f90b6e656039ef8 --- /dev/null +++ b/Operating Systems/OSX/Mac OS X Panther Timesaving Techniques for Dummies (2004).pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3fcbaf268b3be4be983e83d919b7f395286225658ae13589f245c8b4535a92d0 +size 25608339 diff --git a/Operating Systems/OSX/Mac OS X Tiger Timesaving Techniques for Dummies.pdf b/Operating Systems/OSX/Mac OS X Tiger Timesaving Techniques for Dummies.pdf new file mode 100644 index 0000000000000000000000000000000000000000..4f0de2e8cc1396ce60f8b017c43fdb1b22c876e3 --- /dev/null +++ b/Operating Systems/OSX/Mac OS X Tiger Timesaving Techniques for Dummies.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fa4da082cc9289ac4168b4d369e8b94943255bc0568e9b96150aaa2ff8b8ab27 +size 40309222 diff --git a/Operating Systems/OSX/Mac mini Hacks & Mods for Dummies.pdf b/Operating Systems/OSX/Mac mini Hacks & Mods for Dummies.pdf new file mode 100644 index 0000000000000000000000000000000000000000..becf481d3b899b6059a215030449d4a7c2209589 --- /dev/null +++ b/Operating Systems/OSX/Mac mini Hacks & Mods for Dummies.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:91a258200965060225c2ed54a1f57883ab685dec22bfabb9df7b1504dae89d00 +size 10709496 diff --git a/Operating Systems/OSX/Macs for Dummies, 8th Edition (2004).pdf b/Operating Systems/OSX/Macs for Dummies, 8th Edition (2004).pdf new file mode 100644 index 0000000000000000000000000000000000000000..e1ddb2be4818becd595a717ab7e7171385def655 --- /dev/null +++ b/Operating Systems/OSX/Macs for Dummies, 8th Edition (2004).pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:175aae0f5f260139abe715dbdf53edcd619276cc0134716ee4d7c390df028dba +size 10316296 diff --git a/Operating Systems/OSX/O'Reilly Essential Mac OS X Panther Server Administration.chm b/Operating Systems/OSX/O'Reilly Essential Mac OS X Panther Server Administration.chm new file mode 100644 index 0000000000000000000000000000000000000000..15f2dc4ebdede53e8306f7f2573ca0500dc9ec9b --- /dev/null +++ b/Operating Systems/OSX/O'Reilly Essential Mac OS X Panther Server Administration.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6c4e5a41ee4e4b7ccc2c1dcee116ef547c324c0490a0d5108f16ae7687298b55 +size 21598715 diff --git a/Operating Systems/OSX/O'Reilly Learning Unix for Mac OS X.pdf b/Operating Systems/OSX/O'Reilly Learning Unix for Mac OS X.pdf new file mode 100644 index 0000000000000000000000000000000000000000..78dfe5898033ca08dd7e61f1fc51a1243d8b5992 --- /dev/null +++ b/Operating Systems/OSX/O'Reilly Learning Unix for Mac OS X.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0e741015e6e4559b05713bd5aad8b2e96c858935abeac73d41496866377f16df +size 17114191 diff --git a/Operating Systems/OSX/O'Reilly Mac OS X Panther In a Nutshell (2nd Edition).chm b/Operating Systems/OSX/O'Reilly Mac OS X Panther In a Nutshell (2nd Edition).chm new file mode 100644 index 0000000000000000000000000000000000000000..2ef252abd4b99d07fc2957d51f33362f949c1f46 --- /dev/null +++ b/Operating Systems/OSX/O'Reilly Mac OS X Panther In a Nutshell (2nd Edition).chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6c9165c7287a1df9c9c6d4d9aaa1c201c99297eae85e498f77cf3eebb31d0c20 +size 9935018 diff --git a/Operating Systems/OSX/O'Reilly Mac OS X The Missing Manual (Panther Edition).chm b/Operating Systems/OSX/O'Reilly Mac OS X The Missing Manual (Panther Edition).chm new file mode 100644 index 0000000000000000000000000000000000000000..ab0991fe37a1df212d0406d6b0ea1934db881ece --- /dev/null +++ b/Operating Systems/OSX/O'Reilly Mac OS X The Missing Manual (Panther Edition).chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b21d1415f7f94d5bbfba5bd50ba943343272f4f0c602ef71ed0fe49eea60705a +size 13617251 diff --git a/Operating Systems/OSX/O'Reilly Mac OS X Unwired.chm b/Operating Systems/OSX/O'Reilly Mac OS X Unwired.chm new file mode 100644 index 0000000000000000000000000000000000000000..ca7ed8c24628b47e0a6c19b93fbd8c448d0d2def --- /dev/null +++ b/Operating Systems/OSX/O'Reilly Mac OS X Unwired.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:39a81ac1b1445b0f6a8276cc996359dadc6d8a015aa2329d85a5bbf23573b4c8 +size 5030107 diff --git a/Operating Systems/OSX/O'Reilly Switching to the Mac The Missing Manual (Tiger Edition).chm b/Operating Systems/OSX/O'Reilly Switching to the Mac The Missing Manual (Tiger Edition).chm new file mode 100644 index 0000000000000000000000000000000000000000..11437e2ead6bb577a411a832db9c1ec041119113 --- /dev/null +++ b/Operating Systems/OSX/O'Reilly Switching to the Mac The Missing Manual (Tiger Edition).chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8968511cf40e91babcf9d7f1494fb8e7374eaf82d11fab4fe187754242708358 +size 5266824 diff --git a/Operating Systems/Operating Systems - Internals and Design Principles - 7th E.pdf b/Operating Systems/Operating Systems - Internals and Design Principles - 7th E.pdf new file mode 100644 index 0000000000000000000000000000000000000000..f45e9f5a0cb4f1f6b64889e7cc08e7fcab8ccdc2 --- /dev/null +++ b/Operating Systems/Operating Systems - Internals and Design Principles - 7th E.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a99804cd6a4c169f25f29290088e044b9c83f5c236cc53a30788164bd9b038d4 +size 8902179 diff --git a/Operating Systems/Operating Systems and Middleware.pdf b/Operating Systems/Operating Systems and Middleware.pdf new file mode 100644 index 0000000000000000000000000000000000000000..834488adebafda1a24be2db10ae80b43e27ea8fe --- /dev/null +++ b/Operating Systems/Operating Systems and Middleware.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:089356df3e22a9afb376deee894198db4a2f482dca4fada4dcf7c12cdc5e4887 +size 2827100 diff --git a/Operating Systems/Operating.Systems.-.A.concept-based.approach.pdf b/Operating Systems/Operating.Systems.-.A.concept-based.approach.pdf new file mode 100644 index 0000000000000000000000000000000000000000..51979fbcd1440f49c38fa66c8d46d548eb78f8b3 --- /dev/null +++ b/Operating Systems/Operating.Systems.-.A.concept-based.approach.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b28785f2f00288425f8fc1ab9913c1803e2f0a98a5075801f719daa4d3d15028 +size 14256530 diff --git a/Operating Systems/Plan9/Introduction_to_Operating_Systems_Abstractions_Using_Plan_9_from_Bell_Labs_-_Ballesteros.pdf b/Operating Systems/Plan9/Introduction_to_Operating_Systems_Abstractions_Using_Plan_9_from_Bell_Labs_-_Ballesteros.pdf new file mode 100644 index 0000000000000000000000000000000000000000..204a5008a98016c5da1a610e58d2b57ad1e7f086 --- /dev/null +++ b/Operating Systems/Plan9/Introduction_to_Operating_Systems_Abstractions_Using_Plan_9_from_Bell_Labs_-_Ballesteros.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6dee726ee0409f31f4a4026c389c3a93b5e0638a70363671ee070b2a1527cc65 +size 1320610 diff --git a/Operating Systems/Plan9/Notes_on_The_Plan_9_3rd_Edition_Kernel_Source_-_Ballesteros.pdf b/Operating Systems/Plan9/Notes_on_The_Plan_9_3rd_Edition_Kernel_Source_-_Ballesteros.pdf new file mode 100644 index 0000000000000000000000000000000000000000..fb88e8bce8b5a0be9205de4ddd2f0b60dcce80ea --- /dev/null +++ b/Operating Systems/Plan9/Notes_on_The_Plan_9_3rd_Edition_Kernel_Source_-_Ballesteros.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fedb7146abbfdbc428ca5c50ca9edd56ce8c3ce9bc00c62cf570b78638036153 +size 1947318 diff --git a/Operating Systems/Plan9/Notes_on_The_Plan_9_3rd_Edition_Kernel_Source_-_Ballesteros.tar.gz b/Operating Systems/Plan9/Notes_on_The_Plan_9_3rd_Edition_Kernel_Source_-_Ballesteros.tar.gz new file mode 100644 index 0000000000000000000000000000000000000000..e676a070ae338559bc79dcafa7c462ec02ec038a --- /dev/null +++ b/Operating Systems/Plan9/Notes_on_The_Plan_9_3rd_Edition_Kernel_Source_-_Ballesteros.tar.gz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:551d279c0a5c97153dca3bd72d330c25ee4aaec859cae1ece0a7593d0cfda1ed +size 665004 diff --git a/Operating Systems/Plan9/Notes_on_The_Plan_9_3rd_Edition_Kernel_Source_-_Ballesteros.txt b/Operating Systems/Plan9/Notes_on_The_Plan_9_3rd_Edition_Kernel_Source_-_Ballesteros.txt new file mode 100644 index 0000000000000000000000000000000000000000..abe1427c314aeaa9ecd39bfc3824772378049048 --- /dev/null +++ b/Operating Systems/Plan9/Notes_on_The_Plan_9_3rd_Edition_Kernel_Source_-_Ballesteros.txt @@ -0,0 +1,16204 @@ + + Notes on the Plan 9(tm) 3rd edition Kernel Source + + Francisco J Ballesteros + + September 16, 2004 + +Contents + + * [1]Contents + * [2]Trademarks + * [3]License + * [4]Preface + * [5]Fixed since the first version + * [6]Introduction + + [7]How to read this document + o [8]Coming up next + + [9]Other documentation + o [10]Manual pages + o [11]Papers + + [12]Introduction to Plan 9 + + [13]Source code + o [14]Notes on C + o [15]mk + + [16]PC hardware facilities + o [17]Registers + o [18]Instructions and addressing modes + o [19]Memory + o [20]Interrupts and exceptions + * [21]System source + + [22]Quick tour to the source + o [23]Interesting include files + o [24]Interesting source files + + [25]System structures + * [26]Starting up + + [27]Introduction + + [28]Running the loader + o [29]Preparing for loading + o [30]Loading the kernel + + [31]Booting the kernel + + [32]Processors and system configuration + + [33]I/O ports + o [34]Port allocation + o [35]Back to I/O initialization + + [36]Memory allocation + + [37]Architecture initialization + o [38]Traps and interrupts + o [39]Virtual Memory + o [40]Traps and interrupts (continued) + + [41]Setting up I/O + + [42]Preparing to have processes + + [43]Devices + + [44]Files and Channels + o [45]Using local files + o [46]Starting to serve files + o [47]Setting up the environment + + [48]Memory pages + + [49]The first process + o [50]Hand-crafting the first process: The data structures + o [51]Hand-crafting the first process: The state + o [52]Starting the process + * [53]Processes + + [54]Trap handling continued + + [55]System calls + + [56]Error handling + o [57]Exceptions in C + o [58]Error messages + + [59]Clock, alarms, and time handling + o [60]Clock handling + o [61]Time handling + o [62]Alarm handling + + [63]Scheduling + o [64]Context switching + o [65]Context switching + o [66]FPU context switch + o [67]The scheduler + + [68]Locking + o [69]Disabling interrupts + o [70]Test and set locks + o [71]Queuing locks + o [72]Read/write locks + + [73]Synchronization + o [74]Rendezvous + o [75]Sleep and wakeup + + [76]Notes + o [77]Posting notes + o [78]Notifying notes + o [79]Terminating the handler + + [80]Rfork + + [81]Exec + o [82]Locating the program + o [83]Executing the program + + [84]Dead processes + o [85]Exiting and aborting + o [86]Waiting for children + + [87]The proc device + o [88]Overview + o [89]Reading under /proc + o [90]Writing under /proc + o [91]A system call? A file operation? Or what? + * [92]Files + + [93]Files for users + + [94]Name spaces + o [95]Path resolution + o [96]Adjusting the name space + + [97]File I/O + o [98]Read + o [99]Write + o [100]Seeking + o [101]Metadata I/O + + [102]Other system calls + o [103]Current directory + o [104]Pipes + + [105]Device operations + o [106]The pipe device + o [107]Remote files + + [108]Caching + o [109]Caching a new file + o [110]Using the cached file + + [111]I/O + o [112]Creating a queue + o [113]Read + o [114]Other read procedures + o [115]Write + o [116]Other write procedures + o [117]Terminating queues + o [118]Other queue procedures + o [119]Block handling + o [120]Block allocation + + [121]Protection + o [122]Your local kernel + o [123]Remote files + * [124]Memory Management + + [125]Processes and segments + o [126]New segments + o [127]New text segments + + [128]Page faults or giving pages to segments + o [129]Anonymous memory pages + o [130]Text and data memory pages + o [131]Physical segments + o [132]Hand made pages + + [133]Page allocation and paging + o [134]Allocation and caching + o [135]Paging out + o [136]Configuring a swap file + o [137]Paging in + o [138]Weird paging code? + + [139]Duplicating segments + + [140]Terminating segments + + [141]Segment system calls + o [142]Attaching segments + o [143]Detaching segments + o [144]Resizing segments + o [145]Flushing segments + o [146]Segment profiling + + [147]Intel MMU handling + o [148]Flushing entries + o [149]Adding entries + o [150]Adding and looking up entries + * [151]Epilogue + * [152]Bibliography + + Trademarks + + Plan 9 is a trademark of Lucent Technologies Inc. + + The contents herein describe software initially developed by Lucent + Technologies Inc. and others, and is subject to the terms of the + Lucent Technologies Inc. Plan 9 Open Source License Agreement. A copy + of the Plan 9 open Source License Agreement is available at: + http://plan9.bell-labs.com/plan9dist/download.html or by contacting + Lucent Technologies at http: //www.lucent.com. All software + distributed under such Agreement is distributed on an "AS IS" basis, + WITHOUT WARRANTY OF ANY KIND, either express or implied. See the + Lucent Technologies Inc. Plan 9 Open Source License Agreement for the + specific language governing all rights, obligations and limitations + under such Agreement. Portions of the software developed by Lucent + Technologies Inc. and others are Copyright (c) 2000. All rights + reserved. + + License + + Copyright $\copyright$ 2001 by Francisco J. Ballesteros. This material + may be distributed only subject to the terms and conditions set forth + in the Open Publication License, v1.0 or later (the latest version is + presently available at http://www.opencontent.org/openpub/). + + Preface + + The very first time I understood how an operating system works was + while reading the source code of Minix. Years after, I had the + pleasure of reading John Lions ``Commentary on UNIX'' along with the + source code of UNIX v6. + + Although time has passed, I still feel that the best way to learn how + an operating system works is by reading its code. However, + contemporary UNIX (read Linux, Solaris, etc.) source code is a mess: + hard to follow, complex, full of special cases, plenty of compiler + tricks and plenty of bugs. When Plan 9 source code was released to the + public on its 3rd edition, I knew it was just the material I needed + for my Operating Systems Design course. This commentary is an attempt + to provide a guide to the source code of Plan 9 3rd edition. + + The concepts included are those covered on the ``Operating Systems + Design'' course of 4th year at Universidad Rey Juan Carlos de Madrid + [[153]2]. + + Any reader, specially when following the course, is encouraged to read + the source along with the commentary as well as to modify and enhance + the system. + + Fixed since the first version + + Since the last version of this document, these things have been fixed. + + * Introduction chapter: + + It now says which kernel tree the notes refer to (3rd + edition; june release). + + Figures added for Intel stuff + + Brief description of assembler. + * Starting up chapter. + + Note about why using 9load. + + 9load jumps to tokzero as I was saying, but that does not set + the PC to KZERO. + + fixes in 9load probe routine. + + fixes about the mapping for the Mach structure. + + fixes in ram scanning. + + Small section about PC hardware added. + + Horrendous bug about how do QIDs work fixed (I said that vers + was used to detect removed-and-recreated files, which is not + the case; In fact I think I understood that was not the case + long ago before starting this document. Too sleepy that day? + Too few caffeine? Who knows?). + + pageinit has nothing to do with software for paging, it is + just initializing the page allocator. The starting up chapter + was saying it had to do with software MMU, although on the + next paragraph it was described and the reader could know + what it really does. + + psstate only holds the process state for ps. + + process groups have name spaces, not processes. The + description was confusing. + + Some figures added. + + Note added about the check for pidalloc wrap. + + Note about why to reserve room in the bottom of the kernel + stack. + + Fix about eflags in touser. + + sectioning for functions added. + * Process chapter. + + Notes added about why to use a scheduler stack. + + Note added about why to have two scheduling classes. + + Note added about other synch. means. + + Notes added about asking other processors for things. + + Fix in clockintr. Last lines were for user-level code, not + for non-kernel processes. + + postnote fix. + + noted fix. + + Note for wakeup in pexit. + + procstopwait fix. + + More comments added. + + Fixes in fork. + + note about the lack of incref on slash. + + Some figures added. + + sectioning for functions. + * Files chapter. + + Note added about why to have a limit on file descriptors. + + Non-empty directories can be removed. + + Hint about saveregisters. + + The example for ``..'' now uses walk and not cd, which was + unfortunate because cd does not crosses the mount point. + + Note about why a read from a union must clone the mounted + channels anyway. The comment could be suggesting that the + clone could be avoided, that is not the case. + + fixes in walk. + + Fixes in the pipe device. + + Note about how to avoid the CHDIR trick for exportfs. + + Nested calls to the mount driver are not a problem; msg sizes + are. + + Note about coherence of distributed caches. + + Note about using just pages to do caching. + + Some figures added. + * Memory chapter. + + Note about why to use just physical memory for caching remote + files. + + Note about why the pager avoids waiting for locks. + + Note about the lock+unlock when paging-in. + + Note fix about the use of share while cloning segments. + + Note fix about the check for ESEG while attaching segments. + + Other notes added. + + Figures added. + + Please, if you feel that a figure should be added to clarify + something, that a piece of text should be moved somewhere else to + clarify the exposition, that anything is missing, that anything should + be removed, fixed, etc. just send me a mail to + nemo@gsyc.escet.urjc.es. I will pay attention to it. + + I am grateful to Jean Mehat for the fixes he sent me. Yes, as you can + see, yet another thing to fix is the missing acknowledgements section. + + Introduction + + An OS does mainly two things: it multiplexes the hardware and provides + abstractions built on it. Plan 9 does it for a network of machines. + The nice thing of Plan 9 is that it is centered around a single + abstraction: the file. Almost everything in the system is presented as + files. Therefore, most of the complexity lies on the ``multiplexes the + hardware'' part, and not on the ``provides abstractions'' part. By not + optimizing the system where it is not necessary, even the + ``multiplexes the hardware'' part is kept simple (You should compare + the source code with that of Linux if you don't believe this). + + Before proceeding with the source code, I give you a piece of advice + regarding how to read this document, which shouldn't be read as a + regular book. + + How to read this document + + This commentary is that, a commentary to the Plan 9 kernel source. I + have used the source for the June release of the third edition. It + should be read like any commentary of a program, by keeping both the + commentary and the source side by side. In fact, you should try to + read the system code without reading the commentary. If while you are + reading the code and the commentary, you feel curiosity about what + else is done at a particular file, you should go read it all: remember + that nobody can teach you what you don't want to learn. + + The final goal is to understand the system, how it is built, what + services it provides and how are them provided. As with any program, + it is better to focus on the tasks the system has to carry out. Most + of the commentary will be centered on them. Before understanding the + code of the system, it is wise to take a view to the system as a user. + You are strongly advised to read the manual pages relevant for each + chapter, as well as to use the system either at the Plan 9 laboratory, + at home, or at both. Ask for help if feel you can't install Plan 9 at + home. Once you know the set of services provided, you also know that + has to be implemented, and you will understand the code better. + + While you read the commentary, you will see that I refer to the + authors of the source code as ``the author''. Each time I mention the + author, I am referring to the author(s) of the particular piece of + code discussed. Plan 9 is the joint result of many people. The main + authors of the code are Rob Pike, Dave Presotto, Sean Dorward, Bob + Flandrena, Ken Thompson, Howard Trickey, and Phil Winterbottom. As far + as I know, Rob Pike and Dave Presotto were the system architects; and + Ken Thompson was the architect for the file server. + + Also, whenever I say ``he'' or ``his'', you should understand that I + am saying ``he or she'' and ``his or her''. I do not like typing so + much and for me it is hard to find impersonal sentences where ``he'' + and ``his'' can be avoided. So excuse me, no offense intended. + +Coming up next + + In the next section, I give you some pointers you should follow. They + are mostly research papers about Plan 9. You should read them (well, + at least you are expected to read the first one) to learn more about + the system before looking at its internals. It is good to learn to + follow the documentation pointers ``on demand'', as you feel you need + to know more about a particular topic to understand the code. + + What remains of this chapter is whatever I think is the bare minimum + to understand the source code. Next section gives a quick introduction + to reading code written in C, the language used for the Plan 9 kernel. + You can skip this whole section but take a loot at it when you find + something that is not ``ANSI C'' in the code. The following one is a + quick introduction to PC hardware facilities. + + Remaining chapters describe different topics of the system and can be + read randomly, although it would be good to read chapter [154]2, about + system source code organization, and chapter [155]3, on system + startup, before proceeding with the following ones. Besides describing + how the system boots, chapter [156]3 describes several important + concepts to understand the design of Plan 9. + + To save trees, the source code is not printed on paper. All chapters + refer to code using pointers like /dir/file.c:30,35. They focus on a + given line (or lines). These pointers can be used as ``addresses'' on + the Plan 9 editors you will be using during the course. It is very + convenient to print this commentary, open the acme editor[157]5.1 + full-screen, and then follow the commentary giving the pointers to + acme as they appear. It is even better to use a text version of the + manuscript and open it on acme. Then you can jump to the source by + clicking button 3 on the pointer. What? you don't know how to use + acme? Don't worry, forget this and the next couple of paragraphs and + reread them when you get started with acme. To get started you can + read the paper on acme from volume 2 of the Plan 9 manual [[158]7]. + + If you open the text version on acme, I suggest you execute these + commands by using button 2 on them: +Local bind -a . /sys/src/9/port +Local bind -a . /sys/src/9/pc + + If you used button 2 to execute them, your namespace in acme will have + been arranged so that the files in this directory appear to be also in + the directories with the Plan 9 source code. This way, by using button + 3 on file pointers acme will jump to the given location in a different + window. So, now that your namespace is ready, close this file, go to + /sys/src/9/pc and open this file there. This document will be jumping + to code in other directories (e.g. port); in that case, I suggest you + simply edit the tag of the Acme window for this file to update its + directory (e.g. so that the tag is /sys/src/9/pc/9.txt while reading + files in the pc directory, and it is /sys/src/9/port/9.txt while + reading files in the port directory). Do not Put this file. + + More on acme advice, to get line numbers on a file, select it all by + typing :, and using button 3; then type | awk + '{printf("%-5d\t%s\n",NR,$0)}' (or |cat -n if you are on UNIX), select + it and use button 2. Don't Put the file. To locate identifiers through + the source you can create a script to grep the parameter in *.[ch]. + For your convenience, a copy of the kernel source with line numbers is + installed both at the Linux laboratory and at the Plan 9 laboratory. + + Whenever we refer to a file, a relative path has as the working + directory the directory where we are looking source files on Plan 9. + Absolute path names start always at the Plan 9 root. If you are + browsing on Linux, and Plan 9 is installed at /plan9, that means + /plan9/absolute path name instead. Remember that if you use Linux you + still have wily, an acme look-alike. You have also Inferno, where you + have acme (See the web page for the ``Advanced Operating Systems'' + course [[159]1]). + + I suggest you install Plan 9 on your PC and then use it to read the + source code as I said before. By using the system you will ``feel'' + how it works better, and you will use something that is neither UNIX + nor Windows. There are excellent pieces of advice regarding how to + install Plan 9 in volume 2 of the manual [[160]10]. + + If you feel emotionally attached to Linux, you can at least install + wily, an acme-look-alike for UNIX; but you will be missing something + great. + + When discussing a particular data structure or function, it is good to + see where is it used through the system. To find that, you can use the + grep program. By using it within acme, you can simply click with your + (three button) mouse jump through the occurrences found by grep -n. + + When a particular section of a classical OS textbook would further + discuss a topic being addressed, a pointer of the form [n]/section + will appear, where the ``[n]'' part is a reference to the + bibliography. You do not need to follow this kind pointer immediately, + although that might help you if you feel lost. + + Several times I will be discussing code implementing a system call or + used by a popular command. References such like man(1) instruct you to + read the manual page on ``man'' on section ``1'' of the manual as a + definitive reference on the program or system call discussed. You + should at least browse the manual pages as they are cited; and you can + skip parts that you don't understand there. + + One of the abilities you are expected to learn is that of browsing + through a reasonably sized piece of code or documentation. While doing + that, remember that it is important to ignore at first things we don't + understand and try focus on what you can understand. Of course, unless + you know the not-understood part is not relevant for you, you should + try to understand that part too, and ask for help if you can't. + + NOTE: A chapter describing the disk device driver? at the end? + + NOTE: Convince a colleague to write a companion for the ip + directory? + + Other documentation + + The third edition of Plan 9 comes with a two volume programmer's + manual [[161]9,[162]10]. The first volume, ``the manual'', is the set + of manual pages for the system. Manual pages are packaged into + sections. There are several sections, including a section on commands + and another on system calls and library functions. Manual pages are + similar to that of the ``man'' command on UNIX, although the set of + sections vary. + + The second Plan 9 programmer's manual volume, ``the documents'' is a + set of papers relevant for Plan 9. They discuss one aspect or another + of the system. I expect you to read at the very least several ones, + and I highly recommend you read all of them. You will find that papers + on volume two are not like typical research papers these days, on the + contrary, they are simple, show a new idea or a new way of doing + something, and can be understood by themselves; moreover, they are + implemented. Reading Plan 9 papers is a fine way of get a kind + introduction to the system. + +Manual pages + + The manual [[163]9] is divided in sections. When you refer to a manual + page like man(1), you are referring to the manual page for ``man'' on + section 1 of the manual. Manual pages can be found at several places: + + * Using the man command on Plan 9, like in man 1 man. + * Writing the name of the page (e.g. ``man(1)'') on the acme editor, + and clicking on it with mouse button-2. + * running nroff on Linux for the manual page found at + /sys/man/manX/xxxx. For example, if your Plan 9 tree is at /plan9, + you can: + nroff -man /plan9/sys/man/1/man + On the Linux laboratory, you have also the 9man command that + refers to the manual of Plan 9 installed on the Linux file system; + and ignores Linux manual pages. + * Using your favorite web browser and looking at + http://plan9.bell-labs.com/sys/man + + If, as I recommended, you are using acme to read the source, method 2 + is the most convenient one. + + Now go, read intro(1), and drink some coffee. Give yourself enough + time to assimilate what you read there. + + Done? Ok, if you are really done, you should now know that + + * Section 1 of the manual is for general user commands. You type + them on a shell, or click on their names with button 2 in acme. + * Section 2 is for library functions and system calls. This is the + programatic interface to the system. You are studying how the + system calls described here are implemented. + * Section 3 shows kernel devices, which supply ``kernel files'' you + need to access to use the system. These files show up typically + under /dev. You will be interested mostly on manual pages for + devices we discuss. + * Section 4 has manual pages on file systems that you can mount. + They are supplied usually by user programs that implement some + service. For instance, access to FAT file systems is provided + through a program that services a FAT file system--using the FAT + partition as the storage medium. + * Section 5 shows how you talk to files on Plan 9. Plan 9 is a + distributed system that permits remote access to files. This + section shows the 9P protocol used for that purpose. It is at the + very heart of the system. During the chapter on file systems, you + should be reading this section. + * Section 6 discusses several file formats. For example, the format + of manual pages is shown at man(6). + * Section 7 addresses databases and programs that access them. + * Section 8 is about system administration. Commands needed to + install and maintain the system are found here. Some of them will + appear while reading the code, and you should read their manual + pages. + + Too many things to read? I recommend you read manual pages on demand, + as they are mentioned on the commentary, or as you use the tools + described on them. The very first time you use a new Plan 9 program or + tool, it is good to take a look to its manual page. In that way, as + you use the system, you will be learning what it has to offer. + +Papers + + Documents from the manual [[164]10] can be found at several places + too. You can use page on Plan 9 (or gv on Linux) with postscript files + in the Plan 9 directory /sys/doc. You can also use your favorite web + browser and look at http://plan9.bell-labs.com/sys/doc. These are the + papers: + + Plan 9 From Bell Labs + is an introductory paper. It gives you an overview of the + system. Reading it you will find that Plan 9 is not UNIX and + also that networks are central to the design of Plan 9. + + You are expected to read this one soon. + + The Use of Name Spaces in Plan 9 + gives you more insight into one key feature of Plan 9: every + process has its own name space. You can think that every + process has a ``UNIX mount table'' for itself; although that is + not the whole truth. + + You are expected to read this one. + + Getting Started with Plan 9 + is an introductory document with information you need to know + to start running Plan 9. + + The Organization of Networks in Plan 9 + shows how networking works on Plan 9. The section on Streams is + no longer relevant (Streams are gone on 3rd edition), although + it is worth reading it because the spirit remains the same. + + How to Use the Plan 9 C Compiler + will be helpful for you to do your assignments. Once you know + how to use C, this paper tells you how to do it on Plan 9. More + on this on the next section. + + Maintaining Files on Plan 9 with Mk + describes a tool similar to make. It is used to build programs + (and documents) on Plan 9. This paper will be also of help for + doing your assignments; as you are expected to use both C and + mk. More on this on the next section. + + The conventions for using mk in Plan 9 + is also good to read. It shows how mk is used to build the + system. This paper can save you some time. + + Acid: A Debugger Built From A Language + is an introduction to the debugger. You will find that it is + not similar to the kind of debuggers you have been using, and + it is highly instructive to debug using Acid. + + Acid Manual + is the reference manual for the debugger. + + Rc The Plan 9 Shell + shows you the shell you will be using. If you have used a UNIX + shell that is probably all you need. You can learn more of rc + as you use it. Remember that rc is installed also on Linux. + + The Text Editor sam + describes the editor used on Plan 9. It is a fine editor + although you can go with acme instead. Indeed, I heavily + suggest you start by using Acme. Of course, it is healthy to + try sam too. The sam editor is installed on Linux too. + + Acme: A User Interface for Programmers + describes the Acme editor. Well, as the title says, it is more + like an environment. You have a clone for Linux called wily. I + used Emacs for years until I found acme, and the same may + happen to you. You should read this document and play with acme + or wily, to navigate the source code. By the way, it is named + ``acme'' because it does everything. + + Installing the Plan 9 Distribution + is something to print and keep side by side with the keyboard + if you intend to run Plan 9. The title says it all. + + Lexical File Names in Plan 9 or Getting Dot-Dot Right + describes how file name resolution works despite the existence + of the bind(2) system call. Read this before you read the + chapter on Plan 9 files. + + There are several other papers, good to read too, that I have omitted + here for the sake of brevity. + + I recommend you fork now another process in your brain and read all of + them in background. Whenever I feel its better for you to read first + any of them, I will let you know. + + Introduction to Plan 9 + + This section intentionally left blank[165]5.2 + + Source code + + Plan 9 is written in assembly (only a few parts) and C. You must read + C code to understand how the system works. Moreover, you are expected + to write your own C code to modify the kernel in your assignments. + + This section will introduce you to C to let you read it. Nevertheless, + you should read the following two books if you have not done so: + + The C programming Language, 2nd ed. + is a good, kind, introduction to the language [[166]5]. It is + easy to read and it pays to do so. The compiler used on the + book is the UNIX C compiler. You can find how to use the Plan 9 + C compiler on the paper ``How to use the Plan 9 C compiler'' + [[167]8]--these guys use descriptive titles, don't do them? + + The practice of programming + is a ``must read'' [[168]4]. It will teach you those things you + should have been taught during the programming courses. + + Now that you have pointers, I will first comment a bit of Plan 9 C, + then a bit of how to use mk to avoid the need to manually call the C + compiler. + +Notes on C + + I include this section on C mostly to document a few differences with + respect to ANSI C, and for the sake of completeness. But I really + recommend you to read The C programming Language [[169]5] if you don't + know C yet. + + Where is the kernel C code? + + The system source code is structured as a set of directories contained + on /sys/src/9. Although there are valuable include files at + /386/include (and similar directories for other architectures) and + /sys/include. You will be using mostly these directories: + + /sys/src/9/pc + contains machine dependent code for PC computers. This code + assumes that you are running on a PC. + + /sys/src/9/port + contains portable code. This code is shared among different + architectures. + + /sys/src/9/boot + contains code used to bring up the system. + + Other directories under /sys/src/9 contain source for other + architectures, but for the ip directory--which contains a TCP/IP + protocol stack. For Plan 9 file systems, the kernel source code is + found at /sys/src/fs instead. Subdirectories of fs/ follow the same + conventions that subdirectories of 9/. I do not comment the file + system kernel, it is a specialized kernel (borrowing a lot from the + generic kernel) designed to serve files fast. + + C and its preprocessor + + If you take a look to any of those directories, you will find files + named ``xxx.c'', ``xxx.h'', and ``sss.s''. Files terminated on ``.s'' + are assembly language files. They contain low-level glue code and are + used where either C is not low-level enough to let the programmer do + the job, or where it is more natural to use assembler than it is to + use C. Files named ``xxx.c'' and ``xxx.h'' are the subject of this + section: they contain C source code. + + The C language has a compiler proper, and a preprocessor. Files are + first fed to the preprocessor, which does some work, and the result is + finally sent to the compiler. The compiler generates assembly code + that will be translated to binary and linked into an executable file. + On Plan 9, the compiler is usually in charge of preprocessing the + source too, so a single program is run on the source; nevertheless, + you better think that source is first fed into the preprocessor and + the result goes automatically to the compiler proper. + + C source files can be thought as ``implementation modules'' or package + bodies. H source files can be thought as ``definition modules'' or + package specs. When someone writes a C module to be used on a program, + the module has a header file (a ``.h'' file) with declarations needed + to interface to the module and a C file (a ``.C'' file) with the + implementation. + + Consider these three files: +/* this is main.c */ /* this is msg.h */ /* this is msg.c */ +#include "msg.h" typedef char *Msg; #include "msg.h" +main() void set(Msg *m,char*s); +{ char *get(Msg m); void set(Msg*m,char*s){ + Msg m; *m=strdup(s); + set(&m,"Hi world!"); } + char*get(Msg m){ + print(get(m)); return m; +} } + + The point to get here, is that msg.h has the interface to msg.c. It + contains a type definition for Msg and the header of a couple of + functions. The main program (always called main in C) can include + these definitions and then use them. Files ``main.c'' and ``msg.c'' + can be compiled separately into object files and then linked together. + + When compiling main.c, the preprocessor will notice the #include + directive and replace it by the set of lines found in the named file + (msg.h in the example). It is textual substitution. The preprocessor + knows nothing about C. The resulting (preprocessed) file would be sent + to the C compiler proper. In the example, some includes must be + missing since there is no prototype for the print function (and the + same happens with strdup). + + Another useful preprocessor directive is #define, which lets you + define symbols. Note that again, this is textual substitution--the + preprocessor knows nothing about C. + +#define SPANISH 0 +#define ENGLISH 1 +extern int lang; +void hi() { + if (lang == SPANISH) + print("hola"); + else print("hello"); +} + + After the #define lines, the preprocessor will replace any ``SPANISH'' + text with ``0'' and ``ENGLISH'' with ``1''. The compiler will see none + of these symbols. + + Functions + + C has no procedures: every subroutine is a function. The result of a + function may be ignored though. Look at this function: + +int add(int c, int l) +{ + return c+l; +} + + It receives two integer parameters named ``c'' and ``l''--parameters + are always passed by value on C. It returns an integer value (the + first ``int'' before the function name). The ``return'' statement + builds the return value for the function and transfers control back to + the caller routine. + + A function can return ``void'' (which means ``nothing'') is provided + to let a function return nothing. There is another use of void, a + pointer to void is actually a pointer to anything. + + When a parameter passed to a function is not used, you can declare it + without a name, as in +int add3(int c, int) +{ + return c+3; +} + + This is not allowed by ANSI C. In this example, of course, it is silly + to declare a parameter and not to use it. However, when a function + should present a generic interface, and a concrete implementation of + the function does not need a particular parameter, it is wise to leave + the interface untouched and not to use the parameter. + + For instance, imagine that to open a file you should use a function + with this prototype: + + int open(char *name, int just_for_read); + + Now imagine a particular file on a CDROM is opened. In the + implementation, it is not necessary to specify the open mode because + it has to be ``read-only''. Now look at this function: + + int open_cdrom_file(char *name, int) + { ... + } + + It implements the above interface, and has an unused parameter. + + To use a function it suffices to know its header. We can know it + because we #included a file where the header is kept, or because we + are calling the function after its implementation. + + Of course, functions can be (mutually) recursive. + + Data types + + There are several primitive data types: char for characters, int for + integers, long for long integers, long long for longer integers, + double for real numbers. These are signed, and you have types defined + with a leading ``u'' for the unsigned versions (e.g.: ulong, which is + actually unsigned long). + + Arithmetic operators are as usual, with the addition of ++ and - which + increment and decrement the operand. They may be used either prefix or + postfix. When prefix, the argument is incremented (or decremented) + prior to its use in the expression; when postfix, the argument is + modified after used for the expression. For instance, i++=j++ means: +i=j; +i++; +j++; + + whereas ++i=++j means +i++; +j++; +i=j; + + The modulus operator is ``%''. Assignment is done with =. Assignment + is sometimes ``folded'' with another operator. For instance, i%=3 + means i=i%3, x|=0x4 means x=x|4, which does a bitwise OR with for. `&' + is the bitwise AND operator. The operator ``x'' negates each bit in x, + so x=x inverts every bit in x. + + Booleans and conditions + + There is no boolean in C, any non-zero integer value (or convertible + to integer) is understood as ``TRUE''. Zero, means false. + + Relational operators are ==, !=, <=, >=, <, >, where != means ``not + equal''. You can use ! to negate a boolean expression. More complex + boolean expressions can be built using && (and), || (or) and ! (not). + Once the compiler knows that a boolean expression will be true (or + false) it will not evaluate the rest of the expression. Some would say + that C has shortcut evaluation of boolean expressions. For those of + you how know Ada, in C you are always using ``and then'' and ``or + then''. For instance: on 1 || f(), function f would never be called. + The same would happen to 0 && f(). This is very useful because you can + check at a pointer is not nil and dereference it within the same + condition. + + As Plan 9 is meant to run anywhere in the world, it has to cope with + every language. A Rune data type is defined (it is actually an + unsigned short) to support strings of ``characters'' in any language. + Rune is used to represent a character or a symbol, hence the name + (some languages use symbols for words, or lexems). Remember that char + has only 256 values which are not enough to accommodate symbols on all + languages. The character encoding system is called Unicode, encoded + using UTF-8. UTF-8 is compatible with the first 128 ASCII characters, + but beware that it will use several bytes when needed. And beware too + that it is not compatible with ISO.8859.1 that you use on Linux. + + Given primitive types, you can build more complex types as follows. + + pointers + A pointer to a given type is declared using *; e.g. char *p is + a pointer to character. You refer to the pointed-to value by + using also *, like in *p--which is the character pointed by p. + + The operator & gives the address of a variable; hence, given + int i, the declaration int *p=&i would declare a pointer p and + initialize it to point to i. A function name can be used as a + pointer to the function, like in + +int (*f)(int a, int b) = add3; /* the previous add3 function */ +... +g=(*f)(1,2); + + arrays + An array in C is simply a pointer with some storage associated, + do not forget this. For instance, char s[3] declares an array + named s of three char slots. The slots are s[0], s[1], and + s[2]. Array indexes go from zero to the array length minus one. + + Arrays can be initialized as in + +int arry[3] = { 10, 20, 30 }; + +int tokens[256] = { + ['$'] DOLLAR, + ['/'] SLASH +}; + + The last example is an array of 256 integers. We plan to index + it using a character (which is a small integer in C). And we + only initialize slots corresponding to characters '$' and '/'. + This initialization style is an addition to ANSI C in Plan 9, + and it is very useful: instead of using conditionals to check + for dollars and slashes and generate a number, we can spend a + few extra bytes and allocate one array holding an entry for + every character; for those of interest, we place there the + values desired; for others, we don't care. + + Because arrays are actually pointers with some storage, C has + pointer arithmetic. Assume this + +int a[3]; +int *p=a; + + Here, p points to a[0]. Well, p+1 is a pointer pointing to + a[1], p-1 would point to the integer right before a[0]. + + Also, if p points to a[1] and q points to a[0], then p-q would + be 1: the number of elements between the two pointers. + + As you can imagine, these to expressions are equivalent: a[i] + and *(a+i). + + Beware, p=a will copy pointers, not array contents. Use memmove + to do that. + + structures + An struct is the equivalent of a record in Pascal. It is + declared by giving a set of field declarations. E.g.: + +typedef struct Point Point; +struct Point { + int x; + int y; +}; + +Point p = (Point){3,2}; +Point q = (Point){ .x 3, .y 2 }; + + declares a new struct tag, Point; declares a point p of type + Point; initializes it with a copy of the Point {3,2}. + ``Literals'' of structures are called ``structure displays'' in + Plan 9's C. They are an extension to ANSI C. + + In the example above, struct Point {...} declares the structure + with an structure tag Point, so that you can say struct Point + p. But it is customary to give a synonymous for the new type + ``struct Point'' by using typedef. Typedef defines a new name + (the Point on the right in the example) for an existing type + (struct Point in the example). + + Once p has been declared, p.x and p.y are names for the members + (i.e. fields) of the p structure. + + Structures can be nested like in: + +struct Line { + Point origin; + Point end; +}; + + And we would say l.origin.x, provided that l is a Line. + + If a struct, member of another struct, has no name its members + are ``promoted'' to the outer struct. That is to save some + typing. For example: + +typedef struct Circle Circle; +struct Circle { + Point; /* has no name! */ + int radius; +}; +Circle c; + + And we could say things like c.x, c.y, and c.radius. Both x and + y come from Point! You can even say c.Point, although it would + be tasteless on this case. Member promotion and unnamed fields + are an extension to ANSI C. + + When you have a pointer to an structure, you can refer to + members of the structure in several ways: + +/* The way that you should know by now: */ +Point *p; +(*p).x =3; + +/* A more convenient way */ +p->x = 3; + + Everybody uses the -> form, and nobody uses the (*x).y form. + + Unions + A union is a struct where only one of its fields will be used + at a time. It is used to build variant records, although it is + a bit more flexible. For instance: + +typedef struct Vehicle Vehicle; +struct Vehicle { +#define CAR 0 +#define SHIP 1 + int kind; + union { + int number_of_wheels; + char can_go_underwater; + }; +}; + + Note how we used an anonymous union (one with no name) within + Vehicle. We can use either one field or another of the union: + they will be sharing storage. We cannot use both at the same + time. + + More examples: +/* array of 4 points; to use like in: arry[3].x, arry[2].y, etc. */ +Point arry[4]; +/* A polygon that contains the number of points, and an array with them. */ +struct Polygon { + int npoints; + Point points[MAXPOINTS]; +}; + + Control structures + + Control structures are very easy to learn. All of them use sentences + as their building blocks. Sentences are always terminated by + semicolons. Several sentences may be grouped to form a block using { + and }. At the beginning of a block, you may declare some variables. + + while + To repeat while the condition holds: + + while(*p){ + *q++ = *p++; + } + + A variant tests the condition at the end and iterates at least + once: + + do { + x[i]++; + x[j]--; + } while(i != j); + + for + is a generic iteration tool. + + for (i=0; ipdb) are updated for the upper memory. The range for video + memory is set write-through (you don't want the cache to retain + just written pixel values). The range used by ROMs and devices + you want to be uncached, to interact with the devices directly. + The routine mmuwalk returns a page table entry (pte) given the + virtual address. Remember that below 4M physical addresses were + mapped one-to-one at KZERO? KADDR remembers: The author knows + it cannot use physical addresses, because the kernel is also + running with paging enabled (i.e. using virtual addresses). A + physical address is converted to a kernel virtual address by + using KADDR (i.e. by adding KZERO). That is because for kernel + usage, physical memory is mapped at virtual addresses starting + at KZERO. Although right now not all physical memory is mapped + at KZERO, that is going to change soon. + The page table being updated it the one for the boot processor, + it will be used later as a template to setup new page tables. + memory.c:429 + Every protection change on the page table requires a TLB flush. + Otherwise, until the next context switch, entries within the + TLB would retain the old protection flags. You will see + mmuflushtlb in the virtual memory chapter. + memory.c:431,432 + These two routines scan available memory and fill up Rmaps + accordingly. The author probably wrote two routines because + unlike regular RAM, upper memory must take into account video + memory and the like. These routines update the page table for + the kernel to make KZERO be the starting of a map for all + physical memory. Bby now, KADDR can be used only for physical + addresses within the map at KZERO, which are just 4M. Starting + to fix that now. + memory.c:440,452 + conf is updated to keep the address (base) and number of pages + (npage) of the first RAM map; it is also updated to keep the + address and number of pages of the biggest map. Hopefully, on + the PC, the first map would be conventional memory, and the + second one will contain all extended memory. + + But, how do umbscan and ramscan work? + + Filling up allocators + + main... + meminit + umbscan() Scans for UMB blocks. + + memory.c:208,253 + umbscan starts looking past the end of video memory up to the + ROM signature at 0xc0000. It does not go up to 0xf0000 because + a two-byte check at 0xc0000 can tell us if there is a ROM + mapped there by the hardware or not. + memory.c:228,229 + At each pass, it writes the first and last byte of every 2K + chunk with 0xcc. + memory.c:230,233 + If reading back those bytes does not yield the just written + value, it is not real RAM. It must be a ROM then. So rewrite + p[0] and p[1] just to be sure that if they were on registers + (as dictated by the compiler), their values are in sync with + the real value in memory. Not sure why they write p[2], but + could be for a similar reason. p[2] seems to contain the number + of 512 byte blocks at the ROM scanned. + memory.c:234 + If the two bytes starting the 2K block match the values just + written (the signature of a ROM), skip the number of 512 byte + blocks recorded by the ROM in the third byte. This portion is + not kept in the allocator. + memory.c:238,239 + If the two bytes are 0xff, make the memory available for + read-only allocation by calling mapfree (Note the rmapumb map + and not the rmapumbrw map). Remember that it is not regular RAM + because the read value was not what the routine wrote. But + blocks marked with 0xff seem to hold device RO memory for us to + read. + memory.c:241,242 + It was regular ram, so make it available for allocation. + Adjacent maps will be coalesced. + memory.c:246,252 + Finally, looks like if the first two bytes at 0xe0000 are 0xff, + not signed by a ROM, and not regular RAM, indicate that there + is device memory (64K) for us to read. Place the memory in + rmapumb. + + main... + meminit + ramscan() Scans for regular RAM blocks and updates the page table. + + memory.c:256 + umbscan was easy, tricky because of PC messy memory management + for IO devices, but easy. Now, ramscan has the important task + of updating the the kernel page table to reflect the installed + ram. Besides, it fills the ram map as memory gets scanned. + memory.c:274,276 + The routine leaves untouched the range from 0 up to 0x5fff. If + you look at mem.h:33,39 you will see some stuff, going from the + interrupt descriptor table, and information from 9load up to + the Mach structure for this processor at 0x5000 (see + figure [180]3.2). Therefore, start by putting into the rmapram + allocator memory going after the Mach structure up to the end + of conventional memory (640K). To determine the end, it reads + BIOS information at 0x400--again, a very low address better + left untouched. + memory.c:278,280 + From 640 up to 1M we had the UMBs scanned before, and from 1M + on we had the kernel loaded. (Looking that the mkfile you see + how the image is linked to start at kernel virtual address + 0x80100020 which leads to 0x00100020 physical address). So, + only memory from the end of the kernel upwards may be used now. + The author gives to the allocator the memory starting at the + end of the kernel. The linker places the symbol end at the very + end of the kernel image, past the data and bss segments (note + that our stack ``segment'' is actually a bunch of bytes after + our Mach). How much memory do you have? By now, the author + places in the allocator at least MemMinMB MBytes. That has to + be a reasonable low value. By allocating at least that, the + author can use the allocator in the following code that scans + for more available memory. + memory.c:290,301 + If no hint was given about how much memory is installed, it + makes a guess by reading from CMOS the configured value. In any + case, pretend to have at least 24 MBytes. The PC is not so good + at letting the system know how much memory is installed, there + are many variants out there. Most of the complexity of umbscan + and ramscan has to do with the idiosyncratic nature of the PC. + memory.c:309,314 + Starting at the page after end+MemMaxMB, it is going to check + one MByte at a time if the memory is there or not. The trick is + to write a silly value (line :312) at the first word of the + Mbyte being tested, and see if we can read it back. If the + write did work, there is memory. The author is saving the value + actually stored at address KZERO, can you guess why? + memory.c:320,321 + While it scans for memory, a page table reflecting the actual + memory installed is built. So, the physical address scanned is + converted to a kernel virtual address, and used to index into + the first level page table. table points to the entry in the + first level page table (page directory, or PD) corresponding to + the virtual address scanned. Now going to update the ``image'' + of physical memory mapped at address KZERO to reflect the + memory actually installed. + memory.c:322,328 + If the entry is null, there is no secondary page table, and it + must be allocated. In line :326 the entry in the PD is updated + to be valid and point to the secondary page table (PT) just + allocated. By zeroing it, the routine invalidates all its + entries--valid bit is zero. Line :327 is reseting a counter + that we will discuss below. Saw how it can allocate memory + while filling up the allocator? It was convenient to place a + few Mbytes there at line :280. + memory.c:329,330 + Now getting a pointer to the secondary page table entry for the + virtual address scanned. The macro PPN gives the physical page + number (a physical address, actually). + When the author gets a pointer he pretends to use, it must be a + kernel virtual address (i.e. bigger than KZERO). However, page + tables keep physical addresses. Do you get the picture? + memory.c:332,332 + Establish the mapping by setting the physical page address, the + valid bit, and write permission. The flush must be done too, + remember why? + Since the PTEUSER bit is not set in the entry, only the kernel + (ring 0) can access this virtual memory page. + memory.c:344,372 + Here is the actual scan for memory. The commentary is a ``must + read''. mapfree is used to place memory under the allocator; + its parameter is one of rmapumb (UMB memory), rmapram (Real + memory for use), and rmapupa (Memory that seems not to be + there). For regular RAM, enable write permission for the MMU; + for UMBs (i.e. device memory), set it uncached to get straight + to the device memory; and for ``phantom'' RAM, clear the map. + The *pte++=...is used to update the mapping and advance the + pointer to the next page table entry. Each of the three if arms + map a whole Mbyte at a time once the check for the first word + of the Mbyte tells us the kind of memory there. To know why one + MByte at a time, ask yourself whether you can install on your + PC less than a MByte of extra memory. Another thing to note is + that the author is counting in nvalid how many pages are + available. The counters are reset at the beginning of a 4MByte + block, i.e. at the beginning of a page mapped with the first + entry of a secondary page table. + memory.c:383,393 + And here is why. On the PC, a first level page table entry can + be used to map a whole bunch of 4MBytes (called a + ``super-page''), without using any second level page table. If + the page address starts at a 4Mbyte boundary, and the pages on + the 4Mbyte block just scanned are of one kind, the entry in the + first level page table (*table) can be used to map the 4Mbyte + block. nvalid knows how many pages there are of each kind. + memory.c:392 + When the ``super-page'' map is not used, the pointer to the + secondary page table at map is cleared. Next time it is checked + at line :323, a new (secondary) page table is allocated for the + next 4Mbyte chunk. If a super-page map was used, the pointer is + not released and the memory used by the old secondary page + table (now unused) will be reused for the next non-existing but + needed second level page table. + memory.c:398,399 + Perhaps the routine used a ``super-page'' and saved an + allocated second-level page table that now is unnecessary. + memory.c:340,401 + And perhaps maxmem is not page-aligned. Place the remaining + part of maxmem within the not-existing memory allocator. + memory.c:402,403 + In any case, ensure that from maxmem to the end of the memory + range the ``memory'' is registered as not backed up. Imagine + that later someone plugs in a PCMCIA memory card, memory will + be moved (allocated) from rmapupa to (deallocated) xrmapupa. + Now the kernel can ask for memory at xrmapupa and use it. + memory.c:406 + Finally, restore the word at KZERO messed up previously for + memory checking purposes. Kernel allocators are filled up + reflecting the actual memory installed in the system, and the + mapping starting at KZERO has entries appropriate for the whole + physical memory. + + Dynamic kernel memory + + As we saw, the RAM map allocator is not enough to provide dynamic + kernel memory. It was good at registering (un)allocated contiguous + portions of memory areas in the system, but it is good just for big + chunks of memory. Also, it is not to allocate and free repeated times + a given portion of memory because it would not tolerate + fragmentation--remember that it may even drop the last fragment? + + The map is used during boot to allocate UMB areas for devices, and to + collect available RAM for dynamic memory allocation. It is xalloc that + collects that RAM for later use by the dynamic memory allocator. + main.c:142 + the call to xinit initializes the allocator. + + main + xinit() Initializes xalloc with blocks from the Rmaps and tells + palloc. + + /sys/src/9/port/xalloc.c:45,84 + It initializes a list of ``holes'' (i.e. memory to allocate) + given the information in conf. It uses the first and the + biggest chunk of RAM, as recorded in npage0/base0 and + npage1/base1. Perhaps it would have been better to turn Rmaps + into the interface between the machine dependent and the + portable part; the PC part could fill it up, and xalloc could + collect those RAM maps found in the appropriate rmap. Probably + the author's reason not to do so is that some other + architecture may not need resource maps at all and its + implementation (conf memory banks) is admittedly more simple. + Although the collected memory should be enough, xalloc could + run out of memory, yet have more memory available in + rmaps--despite the recombination of fragments done by rmapfree + will make its best to end up with a big map holding most of the + installed memory. In any case, you better avoid dynamic memory + if you can allocate an array of structures and keep on using + it. Memory fragmentation is not to be underestimated. + xalloc.c:57,66 + Pages (page frames) are removed from the bank 1 in conf and + added with xhole to the allocator. palloc.p1 is initialized + with the starting address for the pages removed; this is not + relevant for us now, but palloc is the source for page (frame) + allocation in the system, and fields p0, np0, p1 and np1 are + used for that. As it happens in conf, the author maintains + information about just two memory banks. + At most conf.upages are placed in xalloc. Those pages are to be + used for user stuff. + xalloc.c:68,75 + Pages are removed from the bank 0 in conf and added with xhole + to the allocator. palloc.p0 is initialized too. + xalloc.c:77,78 + The number of pages placed under allocation is recorded in + palloc's np0 and np1. + xalloc.c:79,82 + Until now, conf recorded physical addresses for banks 0 and 1 + (xhole receives physical addresses, as rmaps do). But from now + on, conf records kernel virtual addresses for both banks. One + point here is that from now on npage0/npage1 do not keep the + number of pages in each bank, but the uppermost bound for each + back; perhaps this is a bit confusing. + + Try to understand yourself how other xalloc routines work. Take a look + first to the data structures near the top of the file. While reading + them, note how ilock is used to prevent race conditions--you will + learn why in the process chapter. I suggest you start by reading + xalloc (xallocz, actually), thenxfree, finally the other ones. + + There are 128 (i.e. Nhole) memory pools. Are they enough to use xalloc + as a generic allocator? No. xalloc should be used just for big, + long-lived, kernel data structures. + ../../libc/port/pool.c + For actual dynamic memory the kernel uses pools. pool.c + implements generic memory pools where memory can be allocated + and deallocated. By using different pools for different + purposes, fragmentation can be fought. Pools are like + ``arenas'' in some programming languages. In fact, pool.c is in + libc/ and is used by user programs as the C library dynamic + memory provider. Pools are appropriate for allocation of even + small and short-lived data structures. Fragmentation would be + contained within the pool. + /sys/src/9/port/alloc.c + As the pool interface is generic, hence more complex than it + ought to be, a regular malloc interface is built on top of the + pool interface, in alloc.c. malloc is very much like poolalloc, + but allocates memory from a 4MBytes pool. There are a few + differences regarding the C library malloc interface. + alloc.c:21,36 + A pool is declared for malloc. Note the generic programming + again. Routines are placed into the pool to allocate more + memory for the pool, merge, lock, unlock, print, and panic on + the pool. xalloc is used as the allocation routine. You already + saw a bit of generic programming before, when the routines to + process plan9.ini worked independently of how to read and seek + on particular devices. The trick was to use pointers to + functions and stick to a well defined interface. The code + called read and seek using the interface (the function + prototype) without knowing what was the actual function used. + As pools are generic, they use routines noted in lock and + unlock fields to allow concurrent usage of the pool. For + malloc, the routines end up using ilock on a Private.lk field. + This contortion is needed because it is malloc who knows how to + lock things in the kernel, yet pools must be able to acquire + locks. + + smalloc() Allocates zeroed dynamic memory (must succeed). + + alloc.c:172,177 + A smalloc routine is included to request dynamic kernel memory + when the allocation must succeed. When there is not enough + memory, it makes the caller sleep for a while and tries again; + and so on until it gets the memory. There are several places in + the kernel where the author prefers to wait for memory instead + of reporting a ``not enough memory'' error; smalloc is used + there instead of malloc. smalloc zeroes the memory allocated. + + malloc() Allocates zeroed dynamic memory. + + alloc.c:186,199 + malloc is like malloc, but it zeroes the memory allocated. That + is both for security issues and to be sure that anything + allocated there gets a reasonable initial value: nil pointers, + zero counters, etc. Shouldn't malloc just call mallocz? + In all these routines, setmalloctag is used to record the PC + that called the allocation routine. That is to find out who is + guilty for bad usage of memory allocation routines; and also to + know who is using a given portion of memory. All in all, for + debug purposes. + Apart from these details, the code should be easy to + understand. While reading the file, ignore the pimagmem pool, + used for program Images but not related to the malloc + interface. It seems to be declared in alloc.c to reuse the + locking and debugging routines that operate on pools near the + beginning of the file. By making them receive a Pool and not + use directly the mainmem one, they can be applied to pimagmem + too. Shouldn't these generic routines be moved to pool.c? + + So, regular dynamic kernel memory (i.e. malloced one) comes from + malloc (or smalloc, or ...) in alloc.c, which in turn uses a pool + (initially 4MByte) supplied by pool.c, confining fragmentation to be + inside the pool. Several pools are used for different things (malloc + and Images, by now). Pools are fed from RAM maps corresponding to + installed memory. The lowest level is necessary to discriminate RAM of + different kinds, the next one to reduce fragmentation, and the upper + one as a convenient interface. + + Architecture initialization + + main.c:141 + archinit sets up a generic interface to operate on this + particular architecture model. Yes, it is a PC, but there are + very different kinds of PC. + dat.h:216,229 + This is the interface to highly architecture dependent + routines, they vary from one PC model to another or they must + be performed only at particular PC models--this is very useful + to make the code work for both multiprocessor PCs and regular + PCs. + + main + archinit() Initializes architecture specific procedures. + + devarch.c:535,542 + The way to select concrete implementations is to scan a table + of known architecture models and set the global arch pointing + to the right entry. A generic entry is used if the exact model + is unknown. + devarch.c:544,556 + Conventions are used to make more simple the table of + knownarchs. If any routine is not specified by the selected + entry in knownarchs, it is defined as the generic routine. + Looks like, in Object-Oriented words, knownarchs is redefining + routines for concrete archs and inheriting everything else from + the generic one. Simple yet effective. + devarch.c:560,563 + Pentiums and above have a Time Stamp Counter (tsc) builtin that + counts the number of cycles gone in the processor. It is better + to perform time measures than the external programmable clock. + If you know the Hertz the machine is running at, you know the + exact time since the last time you reset the tsc counter. + Forget everything else in archinit by now. + +Traps and interrupts + + Traps are important because system calls, page faults, and other traps + (exceptions) are used to request some service from the kernel (perform + a system call, repair a page fault, etc.). The intel has several + protection rings. Plan 9 uses 0 for kernel and 3 for users. Each ring + has its own stack. The stack used for ring 0 is the kernel stack. When + the hardware detects a trap, it saves the processor context in the + kernel stack. After that, what happens depends on the entry for the + given trap number in a table called IDT (interrupt descriptor table). + That table contains ``pointers'' to routines that handle the trap. As + the Intel has segmentation, each entry contains a small number to + describe on which segment the routine resides. The small number is + used as an index into a Global Descriptor Table (or GDT) that contains + descriptors for segments in the system (base address, length, + protection). The whole picture was seen at figure [181]1.2. You + already knew this, right? + + Remember that code segments determine the protection ring you are + running at. There are different text and data segments (as well as + other extra segments courtesy of Intel) for users (ring 3) and kernel + (ring 0). Other than protection, hardware segments are used almost for + nothing else in Plan 9. The paging hardware is all the kernel needs. + + By the way, you know which one is your current kernel stack, but + beware that it would be a different one as soon as you get real + processes running and such processes issue system calls. Each process + has its own kernel stack used by the kernel to service its traps. + + I won't say more about how the Intel hardware deals with traps and + interrupt as I feel this is enough to understand the code. + + main.c:143 + trapinit is called to initialize trap handling. + + main + trapinit() Initializes interrupt and trap vectors. + + trap.c:142,163 + trapinit fills up the IDT with entries for the 256 trap + numbers. Each entry in the IDT has several fields. Fields d0 + and d1 hold the address for the handling routine. Plan 9 keeps + in vectortable the routines handling traps and interrupts. + Lines :145, :160, and :161 store the routine address (vaddr) + using the two fields d0 and d1. Ask Intel why you must use two + fields to store one address. The KESEL at line :161 is + specifying that the routine address refers to the kernel + executable code segment; i.e. the processor will jump to ring 0 + and execute the routine in kernel mode. See figure [182]1.2. + trap.c:145 + For all traps, set the ``present'' bit in the entry (SEGP). + That tells the hardware that the entry is valid. + trap.c:148,154 + Why not fold these two branches? For breakpoints and system + calls the entry is set up as an ``interrupt gate'' at privilege + level 3, that is, user code is allowed to issue an int + instruction to perform a system call or to notify a breakpoint. + trap.c:156,159 + For any other kind of trap, the privilege level is set to 0 + (kernel). That means that those traps should not be ``called''. + Well, the page fault trap and others are among them. And they + should not be called. It is just that the hardware can generate + them, but users cannot use an int instruction to request such + traps. If users try to do so, they will get a protection + fault--which is yet another trap generated by the hardware. + trap.c:162 + Use the next entry in the vector table for the next trap. The + pointer is incremented in 6 characters each time, not in 4. + Why? + l.s:549,805 + In l.s you see that the vector table does not contain pointers + to handlers, but instead, it contains binary code to call the + handler (the byte after the call is a parameter specifying the + trap number). The code calls strayintr (or strayintrx) in l.s. + These are the interrupt handling procedures pointed to by IDT + entries. By using the table, trapinit can forget about which + traps get an error code pushed on the stack, and which ones do + not get it. Depending on that fact, the author calls strayintrx + or strayintr to ensure the kernel stack has always the same + layout after a trap. The latter pushes the ``error code'' (the + interrupt number, actually) by software, as the hardware did + not do so. + Should the intel hardware push always the error code, + vectortable could go away and entries in the IDT point just to + strayintr or to syscallintr. + l.s:614 + For system calls, the vectortable does not call strayintr + (common trap handling), but syscallintr instead. Having a + common trap handling piece of code simplifies things (it avoids + duplicated code), but it can make you run slower. For system + calls, the call path continues at syscallintr, in plan9l.s--it + does only the strictly necessary to prepare for calling syscall + and proceed with the system call. More on this later, let's get + back to regular traps. + + strayintr() Interrupt handler (no error code pushed by the hardware). + + l.s:514 + strayintr simply provides common code to jump into intrcommon. + It pushes the trap number so that all traps have an error code + pushed in the stack together with the saved processor state. + + intrcommon() Entry point for interrupt/trap handling. + + l.s:521 + Once the stack looks the same for all traps, intrcommon saves + the data segment (which may be the user data segment) and loads + the kernel data segment (Remember that the text segment is + already ok, because the hardware did set it up as described in + the IDT.) Afterwards, data memory references refer to the + kernel data segment. This can be done because after the trap, + the processor is running at ring 0; the user cannot load + segment registers because that are privileged instructions. + l.s:528,534 + Now, after other segment descriptors are saved (user's) and + loaded (kernel's), and after the whole set of registers is + pushed in the stack,intrcommon prepares for calling trap to do + the trap processing. The stack at line :535 has the saved + processor status (made by the hardware) together with the + registers just saved. If you look at /386/include/ureg.h you + will see how: + l.s:534 + General registers (top of stack) were last pushed. + l.s:532,533 + Some extra user segment registers (fs and gs) were pushed + before. + l.s:524 and :528 + Another user extra segment (es) and the user data segment + (ds) were pushed before. And before that, as briefly + pointed out before, the trap number and error code where + pushed, either by the hardware or by strayintr. Finally, + even before that, the hardware pushed the processor + context that was going to be clobbered when the hardware + calls the trap service procedure. + The kernel has now in the kernel stack anUreg for the saved + context. + l.s:536,537 + By pushing the stack pointer, the author sets up the Ureg* + argument of trap, and finally calls trap. + The just saved user register set will be used when returning + from the trap to restore the user processor context. By + restoring it, you jump back to user code. If any register is + modified in the Ureg, it will be modified the next time the + user process runs, after returning from the trap. + + forkret() Returns from interrupt/trap handler. + + l.s:539,547 + Although a routine on its own, forkret is the code executed if + trap returns. If you look at it, it restores the processor + context from the Ureg pushed by the hardware and intrcommon. + The IRETL instruction is a very interesting one because it + reloads the processor with the context saved by the hardware on + the trap. This means that it also restores the code segment and + the stack segment and places the processor back in protection + ring 3 (i.e. userland). + + By now we skip trap handling. We will get back to it when discussing + processes. But we had a pending discussion of syscallintr. + + syscallintr() Entry point for system call trap handling + + plan9l.s:31,52 + This is the call and return path from user to kernel and + vice-versa during system calls. After the vectortable + dispatches to syscallintr there is lean (i.e. fast) code to get + the Ureg on the stack; after the call to syscall there is again + lean code until the IRETL. Compare this call path with the one + the processor would follow by going from strayintr down to + trap, and then switching on the trap number and checking some + stuff, calling syscall and back. System calls are discussed + later. The code is in trap.c:471,539 though; and it dispatches + using a system call table found at ../port/systab.c. + + The point is that after main.c:143, the kernel is prepared to service + system calls as well as other exceptions. After reading all this code, + you now know what that means. + +Virtual Memory + + We skip printinit by now, to continue with low-level glue + initialization. + main.c:148 + Interesting things begin to happen. mmuinit will initialize the + MMU data structures. Yes, the kernel already has a page table, + but it would not even be able to do a context switch. Let's see + how this thing works. + + main + mmuinit() Initializes the MMU. + + mmu.c:48 + The intel is quite bizarre at supporting processes for systems + software. It tries to do it all, and most operating systems + have to do some contortions to use what it provides whenever + they prefer to implement a different thing. This line is + allocating a ``Task state segment''. That is the data structure + used by Intel to describe a Task. It is needed because the + processor uses that ``segment'' to switch to a different + protection ring on a trap. Remember that I said that the + hardware saves the processor context on the kernel stack when a + trap happens? How does the hardware know what is the kernel + stack? It might not be your current stack when you have a + process running. + At any time, the ``Task Register'' is loaded with a descriptor + into a task state segment (TSS). It is a memory segment as any + other, but it describes the current task for the hardware. The + selector loaded into the task register selects a descriptor + from the GDT (Global descriptor table) that points to a TSS. + dat.h:109,136 + The TSS contains among other interesting things, esp0, the + stack pointer to be used at ring 0. When a trap places the + processor into kernel mode, the hardware obtains the esp0 in + the TSS loaded at the task register, and uses that stack to + save the faulting context. If the processor was already in + kernel mode, the context is saved in the current stack. + Remaining fields of the TSS contain the supposedly last saved + context for the task. On Intels, you can use a call or a jump + instruction into a TSS to switch from one task to another, and + the hardware will save the previous task context, and load the + next one. This is so slow that almost nobody uses the TSS to + implement the context switch for OS processes. In fact, in + mmu.c:20 you see how there is just one TSS descriptor for all + processes, which means that the TSS is used just to make the + hardware work. + mmu.c:51 + The GDT used until now, just for booting, is copied into the + machine structure for the current processor. The kernel is + getting out of the initial (and weird) data structures used + just to boot. The current flow of control is on its way to + become a regular process on a regular processor. + mmu.c:53,34 + The descriptor in the GDT for the TSS is updated to point to + the just allocated TSS. It will run at ring 0. + mmu.c:56,60 + Concoct a descriptor for the GDT and load it into the GDT + register. Now using the GDT for this processor. The kernel is + almost prepared to officially switch to the new TSS. + mmu.c:62,66 + Remember that the IDT was initialized at address IDTADDR? Now + the kernel loads the IDT register. Until now the kernel were + not really prepared to service traps nor interrupts--I lied. + But now the hardware knows that the IDT has some new stuff in, + because the register was reloaded. Did you notice that + interrupts are still disabled? + mmu.c:68,74 + Protections for the kernel text pages are changed not to be + writable. This could be done before while the page table for + the kernel was initialized during memory scan, but that's not a + big deal. + mmu.c:76 + taskswitch loads in the TSS of the current processor the given + stack pointer for all protection rings (and also sets the stack + segment as the kernel data segment there). It also loads in the + TSS cr3 register the given page directory pointer. cr3 is the + pointer to the page table. Besides noting it within TSS, + taskswitch loads the pdb into the processor cr3 register. Note + the ``kernel stack'' passed to taskswitch. It is the address of + the Mach processor plus the page size. The kernel stack resides + after the machine data structure. Not a big kernel stack. + Remember that such stack is given to taskswitch so that it + could initialize esp0 to point to the kernel stack for the + current task. + mmu.c:77 + Finally, the task register is loaded with the selector for the + TSS just created, and the kernel becomes an official task for + Intel. This just means that when we get up to user level, the + processor will use the right stack to service traps and + interrupts: the kernel stack specified in the TSS. + +Traps and interrupts (continued) + + main.c:149,150 + If the arch structure filled up by archinit has a routine to + initialize interrupt handling code, it is called now. If no + such routine exists, it is assumed that interrupts are already + initialized. + + main + i8259init() Initializes the PICs. + + devarch.c:377 + For the ``generic'' PC, i8259init is used as intrinit. + i8259.c:33,102 + This file contains code for the i8259 programmable interrupt + controllers (PICs). In the PC, there are two ones routing 8 + interrupts each. The two i8259s are cascaded to dispatch up to + 15 interrupts to the processor (one of the 16 ones is used to + cascade the chips). Even though the programmable timer was + initialized time ago, no timer interrupt ever reached the + processor because the intermediate i8259s were not initialized. + See figure [183]3.4. + + CAPTION: Figure 3.4: Interrupts arrive from the device though the + PICs. The PIC may mask each of the interrupt lines. The processor must + have interrupt enabled in flags to notice interrupts. + + \resizebox{14cm}{!}{\includegraphics{intr.eps}} + + i8259.c:37,38 + Allocate control ports for both chips. + i8259.c:39,102 + Comments make the code self explanatory. In any case, by the + end of the routine both chips route interrupts in their way to + the processor; The PIC is supposed to dispatch its 16 + interrupts starting at VectorPIC. The hardware uses the + interrupt vector offset to index into the IDT. Therefore, + interrupts numbered VectorPIC through VectorPIC+15 correspond + to PIC dispatched interrupts. + i8259.c:29 + The mask (i8259mask) programmed on the i8259s and the cleared + interrupt enable flag in the processor status word are avoiding + interrupts from happening. But since the kernel already has a + working TSS, IDT, and handlers for the IDT entries, it is + mostly ready to service interrupts. + main.c:151 + The PICs just initialized are used by ns16552install, which + shouldn't be called here, but by chandevreset instead. My guess + is that for kbdinit and to allow a serial line based console, + the serial line (i.e. UART) initialization is being done here, + or maybe this is a fossil if any time back in the past devices + were initialized right from main. Although the code is still + clean and easy to follow, it could be an alternative to place + most regular initialization routines under the control of + chandevreset, and let it resolve dependencies. But that could + also be a recipe for disaster if the (generated) chandevreset + code would not honor dependencies. Forget this brief guess if + you didn't understand, it is not relevant. + + main + ns16552install() Sets up the uarts. + + ns16552.h:82,111 + ns16552install allocates ports for the serial lines (two ones, + eia0 and eia1). Then ns16552setup initializes the two UARTS. + ns16552.h:99 and :103 + intrenable is called to request that IrqUART0 and IrqUART1 + interrupts be enabled (not masked by the PICs) and handled by + the ns16552intrx routine. We'll get to intrenable a bit later. + ns16552.h:106,110 + If plan9.ini said to use a serial console, setup the specified + serial port so that its input and output queues are used for + kbdq (the queue for keyboard I/O) and printq (the queue for + console output I/O, as we saw before). ns16552special sets the + pointers passed (kbdq and printq) to point to the Uart + input/output queues. So, anyone reading from the ``keyboard'' + will be reading from the Uart input queue, that will in turn be + written by the serial line driver as characters come in. The + rest of ns16552intall simply initializes the several kinds of + serial boards you may have installed. We will ignore that. + + main + ... + intrenable() Enables an interrupt and installs its handler. + + trap.c:19,57 + Going down to intrenable, it enables the interrupt irq after + setting up an Vctl structure for the interrupt. Although we did + not look into trap, note that it is also called by interrupts, + which are handled like traps. The trap function uses a Vctl + array to index with the trap number and obtain a ``vector + control'' structure to learn how to handle the trap--more + generic programming. + io.h:44,56 + Among other things, the Vctl structure holds the interrupt + number, a tbdf field which identifies the place of the device + in the bus hierarchy, and a pointer for the interrupt handler + f. The handler admits an argument and there is a place for the + argument (a) in the Vctl. When an interrupt (or trap) happens, + trap takes the Vctl and calls f(a). That is an easy way to + reuse a given f by supplying different arguments to it. + trap.c:24,31 + A newly allocated Vctl is initialized. The name supplied to + intrenable is stored in v->name. That way, the kernel (and its + users) can know who allocated the interrupt. The interrupt + number, handler, and its argument are stored too. In our case, + name would be eia0 or eia1, and the argument for the handler + would be 0 or 1--telling ns16552intrx which one of the two + UARTS is interrupting. + trap.c:33,34 + Locking the Vctl to avoid someone changing the handler under + our feet, call the architecture specific intrenable supplying + the Vctl. + + main... + intrenable + i8259enable() Enables an interrupt in the PIC. + + i8259.c:131,147 + For the ``generic'' architecture, intrenable is i8259enable. It + updates the i8259mask and installs it on the PICs. The + interrupt number is checked to be valid. + i8259.c:148,152 + Also, if the interrupt is not level-triggered (it is + edge-triggered) it is not allowed to be shared. Shared? Yes, + look at the Vctl and see how there is a pointer for a next + handler. On the PC, interrupt numbers are scarce, and devices + may share them. The kernel will call the drivers sharing the + interrupts, and they should cooperate to determine which one + was actually responsible for the interrupt. + i8259.c:153,157 + Here is where the new interrupt mask is programmed, and the + interrupt is enabled. It is now passing through the PICs from + the device to the processor. If the interrupt enable flag is + set, the processor may be interrupted now by this interrupt and + trap will call its handler(s) using the Vctl(s). + + main + + main.c:152 + mathinit enables some traps and interrupts, used by the + coprocessor to notify FPU (Floating point unit) errors. + mathinit()Enables FPU traps/interrupts + main.c:552,559 + Calls to trapenable and intrenable would be setting up Vctl + structures to let trap know that the kernel is prepared to + service FPU related events. + main.c:153 + kbdinit (in kbd.c:397) allocates ports for the keyboard, as + well as the keyboard interrupt. After consuming any character + from the keyboard (from the impatient user) it enables the + keyboard interrupt. The keyboard interrupt handler translates + keyboard generated keycodes into Runes, that are the + ``characters'' of the Unicode standard (Plan 9 uses unicode + instead of ascii, have you Japanese friends?) + i8253enable()Enables the clock interrupt + main.c:154,155 + If a clockenable exists for the current architecture, it is + called. For us, it is i8253enable that enables the clock + interrupt using clockintr as a handler. Clock ticks from the + programmable timer are now arriving to the processor. clockintr + will be discussed later; it is the heartbeat of the system. + + Setting up I/O + + main + printinit() Initializes console output. + + main.c:144 + the call to printinit initialized the queue used for print in + the console. What? the queue? Let's see that. + /sys/src/9/port/qio.c:25,49 + Time ago, Plan 9 used Streams [[184]15] to do I/O. The data + structure found here, is the distilled replacement for Plan 9 + 3rd edition. It defines a Queue. A queue is the structure used + to read/write bytes from/to a device or any other kernel beast. + The author uses that because it would not be good to block a + process writing on the console just because it takes a long + time to put bytes in the serial line. It is better to place the + characters in a queue and, when the line is ready, process them + and put the bytes through. This is an example, but there are + many similar situations and I hope you get the picture. + qio.c:29,30 + Queues hold Blocks. A Block is a buffer waiting for I/O. Some + part of the kernel puts a buffer in a queue, and another part + is expected to consume the buffer some time in the future. + portdat.h:123,135 + Block is defined here, and provides pointers (next and list to + link up blocks sitting in a queue). base points to the start of + the buffer, and lim determines the end of the buffer. The other + two pointer rp and wp are the read pointer and the write + pointer. Routines writing to the Block advance the write + pointer, and routines reading advance the read pointer. The + portion of the buffer still to be read lies between rp and wp. + Note the pointer to a free routine. The allocator of a Block + can supply the buffer from whatever memory allocator it chooses + (maybe just static memory), and set free appropriately so that + when the buffer is no longer used memory is released. + qio.c:32,35 + These values summarize the memory held by the queue (i.e. by + the blocks in the queue). + qio.c:43,46 + These locks are used to queue processes waiting for data to + read in the queue, as well as processes waiting for buffer + space in the queue so they could write. + Now let's see a bit of the implementation. + qio.c:74,364 + Several routines provide the programmatic interface for Blocks. + They are all you need to manipulate and access the buffering + provided by the Blocks. They know that blocks are linked + together. + qio.c:370,403 + qget is a routine called by readers of a queue. + qio.c:375,383 + After locking the queue, it sets its state to Qstarve if there + is no data to read, and returns a null block. + qio.c:384,387 + If there is a block to read, it is removed from the queue. + qio.c:390,396 + When the state is Qflow (the writer was stopped because it was + writing too fast), and the queue has less than half its + capacity, the writer is awakened (it will continue and write) + and the Qflow flag is removed so that any other write can + proceed. + + We saw this to get a flavor of I/O using queues, but we'll see queues + in chapter [185]5 that discusses files and I/O. + + Preparing to have processes + + The kernel has almost booted, and we have gone a long way already. You + now know a bit of the data structures used and how the system glues to + the hardware. The things to come are more interesting and a bit + higher-level than what is past. + + ../pc/main.c:156 + procinit0 should be called actually procinit, but there is + another routine with the same name. + procinit()Initializes the process table. + ../port/proc.c:386,400 + it creates a process table containing the number of processes + initialized previously in conf. All process table entries are + linked into a free process list. Each entry contains what the + system knows about a particular program in execution. + initseg()Initializes allocation for segment (images). + ../pc/main.c:157 + Processes need (program) images to run. initseg initializes the + Image allocator in ../port/segment.c by doing the same other + allocators do: Images are allocated and linked into a free + list. You will learn later what is an image. + + Devices + + main.c:158 + links is called to initialize devices. However, there is no C + source file with a links function definition. What happens + here? + ../port/portmkfile:45,46 + the script mkdevc is called to generate 9pcdisk.c from 9pcdisk + or any other configuration file. + mkdevc + generates a source file from the configuration file (i.e. the + value of the $CONF variable as given to mk). + What does it generate? Let's look at both pcdisk and pcdisk.c. + ../pc/pcdisk.c:9,31 + First, external Dev structures are declared for entries under + (i.e. indented below) ``dev'' in pcdisk. You see, rootdevtab + for root, consdevtab for cons, etc. What is a Dev? By now, + think of it as a bunch of procedures (i.e. pointers to + functions) describing how to operate on a device. The + interesting thing for you now is that they have a reset + procedure. To pick up one, etherdevtab is declared in + devether.c:431,450, and it contains a reference to etherreset. + pcdisk.c:32,57 + Now, a devtab array with pointers to Dev structures is built by + mkdevc, it is null terminated. + pcdisk.c:59,66 + For entries in the configuration file named *.root, *code, + array declarations are generated together with a length + variable (The actual arrays are not being declared now). Looks + like ``.roots'' need this to get initialized; but forget this + now. + pcdisk.c:67,76 + This is the interesting thing for us now. For each thing in the + conf file under link, an external thinglink function is + declared. That is our initial entry point for ethernet and + communication devices. The convention is that a thing.c file + would contain the thinglink function. By generating this code + automatically, to add a new ``linked'' device, the author only + needs to write a new C source file, add the entry for the + device in the configuration file, and recompile the kernel. Of + course, the scripts won't work unless the author follows name + conventions. + + main + links() Links device drivers into other drivers. + + pcdisk.c:77,92 + The generated links function, calls the link procedures for the + ``link'' devices configured into the kernel. What are links? + Well, you see how most entries under links are for ethernet + cards; concrete ethernet drivers can be linked to a generic + driver supplying the common functionality. You get the picture. + (Note also how for *.root entries, a call to addrootfile is + generated. That is initializing some kernel-supplied ``files'' + used to get the system working; More on this later). + ether8003link()Links the WD8003 ethernet driver + ether8003.c:267,270 + For ether8003, you only have to go to file ether8003.c and look + at function ether8003link (saw the name convention?). If calls + addethercard supplying a card name and a pointer to a reset + procedure. + addethercard()Links and ethernet ethernet driver + devether.c:302,311 + addethercard is simply registering (linking!) the card into a + table with configured cards. Later, the reset procedure of the + card will be called to prepare it for use. We don't discuss it + here, but devether is a generic ethernet device that uses + services of concrete ethernet card drivers. For instance, the + kernel uses devether to start and stop ethernet cards, and + devether uses Ether structures to locate procedures for + starting and stopping the concrete ethernet card involved. + How is the Ether structure being filled up? When later, + devether calls the reset procedure registered by addethercard, + it is supplied an Ether structure that it must fill up. + Starting to see how things fit together? + pcdisk.c:95,98 + Just for curiosity, see how the knownarchs array mentioned + before is also generated. For our local configuration, the only + specific architecture is that for Intel based multiprocessors; + everything else is a regular PC. + main.c:160 + now that generic devices have their concrete devices linked + into, call chandevreset. + + main + chandevreset() Resets device drivers. + + ../port/chan.c:50,56 + Forgetting about the ``chan'' thing, chandevreset iterates + through the devtab generated by mkdevc and calls all reset + procedures. That prepares each device for operation. As one + device may depend on another, the configuration file has (at + the right of the line configuring a device) the name of + ``it-depends-on'' devices. mkdevc must honor dependencies when + generating devtab. + etherreset()Resets the ethernet ethernet driver + ../pc/devether.c:337 + Using again our ethernet card as an example, etherreset may + look rather complex for us now, but it just tries to detect + (linked) ethernet cards and prepare them for operation. + devether.c:361,362 + This is where reset for the ethernet card linked before is + called. The purpose of this routine is to try to detect cards, + and initialize any interface (i.e. shared memory between the + host and the card, or I/O ports, or whatever) used to talk to + them. If a card is detected, reset should say so with its + return value so that devether knows whether there is yet + another ethernet card or not. + ether8003.c:71,142 + Most of the code of reset has to do with playing the dance from + the card manual to determine if the card is there and what kind + of card is it. + + Files and Channels + + . + + First a quick remark, I am still discussing initialization carried out + by main. But since files and channels are so central in Plan 9, let's + say a bit about them before continuing with the source. You are + advised to read intro(2) until the point where processes are + discussed. Manual pages for system calls mentioned below are also + relevant. + + In Plan 9, a file is an entity serviced by a server process over the + network. That is a generic definition. Of course, the ``server + process'' can be the kernel, or a user process, and the ``network'' + may be some kernel code to glue a local, in-kernel, file server with a + local user of the file. Note that ``local'' here means ``within the + same node''. + + Files are used with the traditional unix interface: create, open, + close, dup, read, seek, and write. And files are still (as they were + in UNIX) a (named) sequence of bytes. Files are deleted with remove + (kind of UNIX's unlink, but a bit different). stat and wstat are used + to read and write file attributes. Directories are read like files, + but they are written either by using dirwstat to change attributes of + a directory entry, or by using create or remove (they add and delete + directory entries for the file affected). + + This (procedural) definition of what is a file, refers to procedures + that can be applied to files, either to obtain new files or to + manipulate and destroy them. However, when there is a network between + the file and the file user (the caller of the procedures), something + has to be done instead of calling the procedure with a procedure call. + + What Plan 9 does is to translate calls to file procedures in the + client machine to RPCs to the server. In case you didn't know, an RPC + is a remote procedure call. It works by sending a message from the + client to the server when a procedure is called, and receiving later + another message with the procedure result. The steps are mostly as + follows: + 1. + The client (the caller of let's say, write) calls the procedure + (write). + 2. + A piece of code (stub) in the client implements the local + procedure actually called by the client, but that is not the + real procedure being called. The client stub builds a message + with the identifier of the procedure being called, and a copy + of the parameters passed to the procedure. + 3. + The client stub sends the message to the server process (the + one implementing the procedure being called). + 4. + The server process is a process listening for messages + requesting procedure calls. It receives the message sent by the + client. + 5. + The server process unpacks the message and determines the + procedure to be called. Depending on the procedure being + called, the server knows what parameters are in the message, + and unpacks them. + 6. + The server process calls the actual procedure (write) with the + parameters just unpacked. + 7. + The procedure returns, yielding some results. + 8. + The server process builds a reply message with the procedure + result. + 9. + The server sends the reply message back to the client + 10. + The client stub receives the reply + 11. + The client stub unpacks any output parameter and result + received, and returns pretending that it was the stub the + procedure that computed the result. + + Now, the protocol defining the request (called transaction in Plan 9) + and reply messages to perform operations on Plan 9 files is called 9P. + It is defined in section 5 of the manual. It is a protocol, and not a + bunch of unrelated RPCs, that means that both the client and the + server using 9P are expected to follow 9P rules. You can take a look + at intro(5) to see what's going on. + + How does the client get in touch with the server to issue 9P requests? + 9P does not specify that--read: 9P permits you to use any way you can + imagine to get in touch with the server. You are expected to get a + network connection between the client and the server by any other + means. Once you have the connection, the client and the server can + talk 9P on it. + + For remote files, the connection is likely to be either a TCP or an IL + stream (IL is the Plan 9 transport of choice) over an IP network. For + local files, you still have a ``connection'' between the client and + the server. I'll now describe this one. + +Using local files + + The client is your local machine. Consider a process calling a file + procedure like write, it specifies a small integer (a file descriptor) + representing the file where to write. File descriptors are obtained + with open as in UNIX. + + /sys/src/9/port/portdat.h:555 + The kernel knows which one is the current process, and locates + the fgrp field of its Proc structure. The fgrp points to a File + Descriptor Group structure. See figure [186]3.5. + + CAPTION: Figure 3.5: The user uses file descriptors (indexes into the + Fgrp descriptor array) to specify files; but the kernel uses channels + to point to routines knowing how to perform file operations. + + \resizebox{14cm}{!}{\includegraphics{files.eps}} + + portdat.h:431,437 + The Fgrp contains an array of pointers to Chan. Every Chan + represents a file being used, and every file descriptor is just + an index into the array in the Fgrp. open allocates a new entry + in the Fgrp and places a Chan on it. The index for the entry is + given to the user as the descriptor for the file just opened. + + We know what is a file from the point of view of the client (a + descriptor), but where is the server? and where is the file on the + server? The answers reside in the Chan structure, with a bit of help + from other structures elsewhere. To answer the questions, let's follow + a bit of the path that a write system call walks. + + syswrite() write system call. + + sysfile.c:444 + syswrite is called, with arg holding a file descriptor, a + pointer to a buffer, and a number of bytes. + sysfile.c:452 + fd2chan translates the (integer) file descriptor into a Chan + structure, by looking at the Fgrp for the current process. + sysfile.c:453,468 + Ignore this by now. It is for handling directories and checking + errors. + sysfile.c:469 + Here it is! The type field of the Chan structure is used to + determine the kind of device implementing the file (the device + can be just ``software'', of course). Now, by indexing with + type into devtab, the author gets a reference to the Dev + structure for the device implementing the file. + Say that ``file'' is a printer, the pointer to a Dev structure + found at devtab[type] would be a pointer to lptdevtab, the Dev + declared at ../pc/devlpt.c:209,228. This is because the type in + channels pointing to local printers is simply the index in + devtab for the lptdevtab entry. + Now, still in ../port/sysfile.c:469, the procedure pointed to + by write in devtab[type] is called. If you look at lptdevtab, + it is lptwrite the procedure actually called. lptwrite is given + a pointer to the Chan for the file being written. Besides, note + how the offset where to write in the file is taken from the + offset field of the Chan. You can see how a Chan in Plan 9 is a + reference to a server file. You also start to see some + implications, like that to share a file offset, the Chan must + be shared, which means that processes sharing the Chan must + reside on the same machine. + + syswrite + lptwrite() Writes on a file in a lpt device. + + ../pc/devlpt.c:148 + In our example, lptwrite gets called, with the Chan for the + file. + devlpt.c:155 + At this point, something interesting happens. The lpt driver + (the file server in this case) takes the qid field from the + Chan. The QID identifies the file in the server. It only has + sense within the server. Here, the server is just the local + printer driver, and the driver is checking the QID to see which + files should be written (it services several files). + The QID is made of two numbers, path and vers. Actually, it is + path that identifies the file. In this case, path holds the + value Qdata for the printer data file. But it could be any of + Qdlr, Qpsr, and Qpcr for files with names dlr, psr, and pcr + also serviced by the printer driver. Two files (within the same + server) are the same file if they have the same path value in + their QIDs. This means that a client of the file server can + check if two Chans refer to the same file by checking their + paths (assuming the files are within the same server. If a file + is removed and created, it should get a new path. + The vers field of the QID is used to distinguish different + versions of the file. It is useful because someone might be + caching a file, or might like to know if the file has changed + since it last got its QID. + Two files are exactly the same file, and thus have the same + contents if they reside on the same server and have the same + QID. This is also important for caching. If a cache has a copy + of a file, and the server still has the same QID for that file, + there is no need to refresh the file copy in the cache: it is + the same one! + devlpt.c:165 + Another interesting thing happens. There can be several + printers. Which one is the one used for write? The Chan + structure has a dev field that specifies which particular + device is being used. So, type and dev fields together identify + a device in the kernel. The type field of the Chan is used to + select the appropriate implementation for file operations, and + the dev field is then used to select the appropriate device of + that kind. + + So, what is a file for the client process? A file descriptor that + leads to a channel. Where is the server? The type and dev fields of + the channel know. Which one is the file? The qid field of the cannel + knows. + + An what about remote files? For remote files, (not discussed now), the + type field would select mntdevtab among devtab entries. mntdevtab + (devmnt.c:920,939) is the Dev for the mount driver, a driver that + implements device operations by issuing RPCs to the server process. + The mount driver uses the connection to the server supplied by the the + caller of mount(2) to talk 9P with the server implementing the remote + file. + +Starting to serve files + + Going back to main, the kernel is already servicing some files (see + root(3)). + + ../pc/main.c:158 + main calls links + ../pc/pcdisk.c:78,81 + which calls addrootfile for cfs, kfs, and ppp. + main.c:160 + main also calls chandevreset, which calls reset for configured + devices, including a call to rootreset: the reset procedure for + rootdevtab. + ../port/devroot.c:78,90 + rootreset is simply calling to addrootdir and addrootfile + several times. + + So, main makes multiple (indirect) calls to addrootfile and + addrootdir. Let's discuss them now. + + ../port/devroot.c + This is the ``device driver'' for the root of the file system. + It services a flat directory implementing the well known ``/'' + directory. + + main + ... + addrootfile() Adds a new file to the root device file free. + + devroot.c:62,66 + addrootfile ``creates'' a new file in the directory serviced by + the root device. addroot()Adds an entry to the tree. + devroot.c:46,47 + At most Nfiles can be placed in the directory serviced. + devroot.c:48 + This is a admittedly simple file system. When other parts of + the kernel create a file into devroot, they supply file + contents as well. Remember buffers named cfscode, kfscode, etc. + declared by mkdevc? + rootdata (devroot.c:19) is an array of pointers to the data of + the (at most Nfiles) files serviced. So, addroot uses the first + free entry to plug the file contents in. There are nroot files, + from 0 to nroot-1. + Where do file contents come from? Consider for example the + kfscode array. The mkfile is calling ../port/mkroot using kfs + as an argument, which is calling data2s. data2s takes a binary + from the running system where you are compiling the kernel + (kfs, which is a program), and generates an assembly file + (kfs.root.s) with the definition of an array (kfscode). The + array contents are the contents of the file. That file is later + assembled and linked into the system. That is how regular + Plan 9 binaries are linked into the kernel to be used as ``root + files''. You can imagine that root files are files needed to + boot the system (i.e. to connect to a true file system, etc.) + and cannot be loaded from the disk/network file system (chicken + and the egg problem). + devroot.c:49 + That was the contents, and the name, permissions, etc? rootdir + is an array of Dirtab structures, containing attributes for the + files serviced. In Plan 9, attributes reside within directories + (well, they can reside at any place the file server wants to + put them at). So, d is the pointer to the directory entry for + the file being ``created''. + devroot.c:50,52 + File name, length, and permissions are set in the directory + entry for the file. + devroot.c:53 + Crucial! a QID for the file is invented. For devroot, files are + numbered 1 to Nfiles, so that file i resides at rootdata[i-1] + and rootdir[i-1]. Clients using the directory serviced will + obtain QIDs for their files and pass them back to devroot when + requesting a file operation. The server must be able to locate + the file quickly given its QID. An index is a just fine way. + devroot.c:54,55 + In Plan 9, directories are also created with create, as files + are. When the CHDIR bit is set in the permissions given to + create, the file server understands that it must create a + directory; not a file. It is the convention in 9P (yes, 9P) + that the path component of QIDs have the CHDIR bit set for + directories. So, these two lines of devroot.c are adjusting the + QID of the file to look like a directory in case the + permissions said to create a directory. + devroot.c:56 + The number of created files is incremented. The next time a + file is created, it would be placed in the next entry. + devroot.c:62,75 + Pay now attention to the arguments given to addroot in both + addrootfile and addrootdir. Understood? Well, it's true, I + didn't say that directories have by convention 0-length. + devroot.c:80,90 + The calls being made by main to the root driver are simply + populating the root device with empty directories holding + nothing. But now that there are directories, they can be + populated! + However, although devroot is prepared to service its files, no + one has a ``connection'' to devroot yet. Well, as devroot is + local, I mean that no one has a Chan with the type set so that + devroot will service the file at the other end. + +Setting up the environment + + Most of the environment for the first user process, and the ones to + come is provided by files. For instance, environment variables, names + for the host, the user, etc. are provided by files serviced from the + kernel. Now that you understand a bit of what does this mean, let's + enumerate some files implementing part of the user's environment. If + you want a user's description of what is provided, refer to manual + section 3. It describes devices that service file trees from the + kernel, as root does. You should read intro(3) at least. + + Take a quick look at any of them and don't be worried too much if you + don't understand what is going on. Some of them (eg. env) are fairly + easy to understand though. + ../port/devcons.c + implements the console device. It supplies files for console + I/O and also for other diverse tasks like user authentication, + time of day, rebooting, etc. + ../pc/devarch.c + We saw a bit of it. It supplies files to identify the + processor, see allocated irqs and I/O ports, and files to do + I/O from user processes. + ../port/devenv.c + Provides environment variables. They are serviced from the + kernel as files in Plan 9. + ../port/devpipe.c + Pipes + ../port/dev....c and ../pc/dev...c + and many others. + + I will comment some kernel drivers through this commentary, as I did + with the root driver. + + Memory pages + + After our incursion into the files serviced by the kernel, we are back + to main. + + You are assumed to have at the very least some concepts on virtual + memory from a basic operating systems or architecture course. You + don't? I think you will survive, but study that topic a bit... + + ../pc/main.c:161 + Just initialized the device files, but have more work to do. + Now pageinit initializes page (frame) allocation, to prepare + for paging. + + main + pageinit() Initializes the page allocator. + + ../port/page.c:13 + pageinit has the important task of initializing the palloc page + allocator. Right now the kernel cannot create a process because + it can not allocate pages for its code, data, and stack; not + yet. + portdat.h:444,459 + The page allocator holds a Page structure for each available + page frame known by the system. You can image that it is a big + data structure--in fact, remember that when calculating + available kernel memory the author had to take into account the + size of this array. By now, see how it looks: looks like the + allocator implements LRU for free pages. + portdat.h:279,293 + The Page structure keeps for each one the physical address in + memory, virtual address for user, and disk address on a file. + It also has a virtualized copy of the modified/referenced bits + maintained by the MMU. Page contents come from either a file or + a swap file, image points to that. If you look at Image you see + how it contains Pte structures that keep pointers to Pages. + Don't be worried. It's simple: pages are either in use by + in-memory images of files or they are free in the palloc array. + page.c:19,20 + A big Page array is allocated at once for the number of pages + defined in np0 and np1 in palloc. Both np0 and np1 were + initialized at xalloc.c:77,78 with the number of pages + ``eaten'' by xalloc from the two memory banks defined in the + conf structure. + See how the author fights fragmentation? It has no sense to + allocate and deallocate Page structures as they are needed. + There will be at most as many free pages as free page frames at + boot time. The author allocates a big array, and then links + Pages on the appropriate list--Images or palloc as they are + allocated/released. + + Lesson: To avoid fragmentation, avoid using dynamic memory whenever + possible. Allocate many resources at once and then keep on using + them. + page.c:21,22 + More defensive programming. You must have enough memory for the + page array, but maybe you don't. + page.c:24,36 + One page at a time, bank 0 pages are linked into the free list. + p0 is kept pointing to the physical address where a page is + being ``moved'' to the free list; np0 has the number of pages + in bank0 not in the free list; and freecount has the number of + pages in the free list. + color has to do with caching. Two different pages can have a + cache conflict because both ones collide on a cache sitting + between the page and the page user. For instance, think that + even pages use entry 0 in the processor cache, and odd pages + use entry 1. It is better to use pages 3 and 4 than it is to + use pages 3 and 5. Got the picture? + The algorithm uses NCOLOR ``colors'' to ``paint'' the pages. If + a page is ever allocated for a given user, and that user has + color ``1'', it is better to give him a page of color ``1''. + That is because pages of the same color are far apart in the + memory and the author assumes that the bigger the distance two + pages are at, the less the chance they will collide. + page.c:37,47 + The same is done for pages on bank 1. + page.c:48,50 + The list is set. + page.c:52 + The number of pages for the user is the distance between the + head and one past the tail. + page.c:52 + The size of the physical memory in Kbytes is computed, just for + letting the user know. + page.c:53 + Remember the nswap ``limit'' set when the kernel first knew how + many pages would be available? That was actually the maximum + number of pages not found in-memory. By adding the physical + memory size, you get the virtual memory size. This is just to + let the user know. + page.c:57,58 + These two values are used for paging. Eg. when less than a 5% + of pages are free, the kernel should be moving pages to disk, + to make more room. It is not healthy to let all the pages be + occupied. It could be that (due to some bug) the kernel needs + free memory for the algorithm that gets more free memory... + + main + swapinit() Initializes the page allocator. + + ../pc/main.c:162 + Back to main, swapinit initializes swapping. + ../port/swap.c:21,33 + Set swapimage.notext; which means that there is no swap file + yet. Perhaps an explicit initialization would make things more + clear: swapimage has been initialized to all-zeroes by the + loader. Its c member that points to the Chan for the swap file + is still zero, which means that there is no channel to the swap + file. In any case, a ``swap map'' (swmap) is allocated. It + reflects the state of portions in the swap file. A page array + for the max number of pages being paged out to swap + (conf.nswppo) is also allocated. + + The first process + + You should know already what a process is, at least from a theoretical + perspective and as a user. To remind you, Plan 9 can use a machine + with just one CPU and pretend that there are multiple programs + executing at the same time. The way to do it is to let each program + run a small fraction of time. As the processor is fast enough, if + programs are given small fractions of processor time repeatedly, they + would appear to be executing all at the same time. A program in + execution is called a process. + + A process is more than a program, it has open files, is run on the + name of a user, has memory for data allocated while the program is + running, a stack to keep track of nested procedure calls and to + maintain local variables, etc. + + To multiplex the processor among processes, a timer interrupt is + typically used. The system arranges a timer to interrupt the + processor, say, 200ms in the future, and loads the processor context + with register values appropriate for executing the user process code. + When the interrupt arrives, the hardware (and the interrupt handling + software) saves the context of the processor as it was right before + the interrupt. That set of registers, if reloaded, would permit the + process to continue where it was. But usually, the kernel loads the + context of a different process instead, letting it run for another + quantum of 200ms in this example. At some time in the future, the + saved context for the process in this example would be reloaded and it + will not even know that another process was using the processor + before. + + In a previous section, you were advised to read intro(2) until the + point where processes are discussed, now go and read all of intro(2) + and fork(2). Most of what the code you are going to read is doing is + also described at boot(8). You are also advised to refresh your + ``theory'' of what a process is in case you are lost by this point. + + From now on, I will be describing some important fields of the process + data structure. I suggest you try to guess how they are used. One fine + way of doing it is (after reading the manual pages I suggested) using + grep on the source to see where are those fields used. Try to guess + what is the field being used for. + +Hand-crafting the first process: The data structures + + Right now, Plan 9 is just a flow of control running a program (the + kernel) loaded into memory. That is not enough to provide system + services, there should be a real process that could later use + rfork/exec to create new ones and execute other programs. + + You can see figure [187]3.6 to get a glance of how our current flow of + control will evolve into a set of processes. In the figure you see how + initially, the kernel started executing and then it creates the memory + image for a first process. It will later jump to protection ring 3 to + execute the user code for it. After that, the kernel would only + execute to service interrupts and system calls--(in the mean time the + process would be ready to run, but would not be running). This first + process would make rfork system calls to create new processes, and the + clock interrupt will be used to multiplex the processor among existing + processes. + + CAPTION: Figure 3.6: The kernel is just a program, and there is just a + flow of control. However, users believe that their processes are + executing sequential programs. + + \resizebox{14cm}{!}{\includegraphics{procs.eps}} + + ../pc/main.c:163 + About to call userinit in main. It is the procedure that + creates the first user process. After that is done, what + remains is to jump to that process and let the system run when + interrupts, exceptions, or system calls request any system + service. + + main + userinit() Initializes the user environment and creates the first + process(es). + + main.c:236 + Previously, the kernel allocated an array with Proc entries. + That was the process table. Each entry is a Proc struct with + the information needed to implement processes. For example, the + Proc structure is used to get to the saved processor context + needed to put that process back in the processor. Now the + author allocates a free entry in the process table. That entry + will be filled up to initialize (create) a new process. + + main + userinit + newproc() Allocates a process table entry. + + ../port/proc.c:307,314 + newproc waits until the free list has at least a free entry. + The resrcwait call would make the current process wait due to + the specified reason (``no procs''); it would be preempted or + moved out of the processor. Think that the current process, + willing to create a new one, has to wait until a free entry be + available. By the way, what is the current process? By now, + none. However, there are free entries and we break the loop at + line :309. The kernel is creating the first process and by now + there is no current one. Note the use of the loop around the + wait, because there is no guarantee that by the time the + current process gains the lock in line :313 any free entry be + still there. Another process could run instead, gain the lock, + and allocate the just released entry. + proc.c:315 + The process entry is removed from the free list. p is the + process begin created. + proc.c:318 + This will be clear when we discuss processes in chapter [188]4. + Processes have a ``state''. The process state tells the system + on what condition the process is in. In this case, the process + is being handled by the scheduler, the part of the system that + decides who runs next and switches from one process to another. + portdat.h:493,504 + You should recognize at least the typical states Ready and + Running. To get an idea of how state is used, consider that + dead processes (those that terminate or abort) have their state + set to Dead; functions doing things to processes while they are + alive can check that the state is Dead and do nothing to dead + processes. + proc.c:319 + This is the state name as printed by ps(1). It is a new + process. psstate holds a representative string for the state + the process is in. That can effectively turn ps(1) into a + debugging tool. It also permits you to inspect a process from a + different machine, psstate would make sense even on a different + architecture: It is just a string that is understood + everywhere. + proc.c:320 + mach points to the Mach structure of the processor where the + process is running. Did not run yet. It is necessary to know on + which processor the process is running at, because on that + processor is where the actual process context is (while a + process is running, the set of registers last saved for him by + the hardware is irrelevant). The convention is that mach is + zero only when the process context is saved elsewhere and its + user code is not running. + proc.c:321 + Not on the free list anymore. + proc.c:324 + This is discussed in the chapter for processes. It is used to + make the process wait due to some reason. Not waiting anything + now. + proc.c:325,328 + Some resources used by the process cleared. Let's say something + about it now before continuing. + + Take a coffee, go read rfork(2) (reread intro(2) if you forgot), and + come back later. + + proc.c:325 + Processes have a name space. More precisely, processes are + within a process group and process groups have a name space. + pgrp is a pointer to a Pgrp structure representing the + ``process group''. For me, the initial ``p'' is confusing + because it sounds like ``process'' and not ``name'', but the + author would have a good reason for the name. The name space + determines what names lead to what files. Actually, the name + space is very similar to that of a UNIX machine. In UNIX, there + is a mount table (for the whole system) that specifies paths + that lead to file systems. For instance, ``/'' leads to the + root file system, ``/cdrom'' may lead to a file system on a + cdrom, etc. Now, back to Plan 9, every process has a name + space. The name space is made of a mount table that associates + paths to file systems. For example, one thing the first process + will do is to associate ``/'' to the root file system + implemented by the root device. + The interesting thing in Plan 9 is that every process (group) + has its own name space and can add and remove entries without + affecting other processes. Adding a new entry is called + mounting a file system or binding a file (see mount(2)). + Deleting entries is called unmounting a file system. + This is very important because it provides an engineering + solution to several important problems on distributed systems. + For instance, by binding /$objtype/bin onto /bin, each process + will find appropriate binaries for the current architecture; + binding files under /dev is a fine way to make a particular + process see a different device-perhaps at a different node! + Processes can share their name space (i.e. their process group) + too. That means that several processes may have the same value + in their pgrp field. + proc.c:326 + Processes have environment variables. Both variables and values + are strings. An example of an environment variable is PATH + (which, by the way, is mostly unused in Plan 9). Again, each + process has its own environment (made of a set of variables). + egrp points to an Environment Group or Egrp structure. + portdat.h:413,429 + It maintains variable names and value strings. Several + processes may share their environment; egrp may be the same for + more than one process. + proc.c:327 + Processes have open files. As we saw before, each process has a + file descriptor group. fgrp points to an Fgrp structure. + portdat.h:431,437 + It has the set of file descriptor entries for the process. You + already know that each file descriptor leads to a Chan + structure that is all you need to perform an operation to the + file referenced by the Chan. Again, processes may share their + file descriptor group. Several processes may have the same + value for fgrp. + proc.c:328 + Processes can synchronize by using rendezvous(2) . To + rendezvous, processes call rendezvous supplying a tag that must + be the same for the two processes doing the rendezvous. Tags + are actually local to a ``process rendezvous group''. rgrp + points to a Rgrp structure. + portdat.h:407,411 + This implements the rendezvous group. It is obvious that + processes can share their Rgrp. + proc.c:329 + Sometimes, a process gets broken and has to be debugged. The + debugger is another process that controls the execution of the + debugged process by placing breakpoints, altering the debugged + process state, and processing events of interest for debugging. + pdbg in a process being debugged points to the debugger process + Proc structure. + proc.c:330 + This has to do with how math coprocessors work. It is really + expensive to save and reload the coprocessor context. + Therefore, the system tries to avoid unnecessary coprocessor + context switches by not loading the coprocessor when a process + does not use it. fpstate records the state of the coprocessor. + The author uses it to know if the coprocessor is being used. By + now, it was not even initialized to a reasonable state. FPinit + says so. + proc.c:331 + You probably think that all processes execute user programs, + running with the processor in non-privileged mode (ring 3). And + that these processes enter the kernel only to service an + interrupt, a trap, or a system call. That is not true. There + are processes that spend their lives within the kernel. These + are called kernel processes. For example, in Plan 9, there is a + process called ``pager'' that has the important task of doing + page outs when low on free page frames, by stealing page frames + from other processes. This is better implemented as a process + with its own (sequential) flow of control that mostly sleeps + until a point when it starts moving pages to disk. You can + imagine that to move pages to disk, the process must do I/O and + must have special privileges, since it must have access to the + pages for the processes involved. For both reasons, it is + implemented as a process that never leaves the kernel, it + starts executing a at a given function within the kernel. After + that, it is handled as a regular process--with special + privileges, admittedly. kp is true when the process is a kernel + process. + proc.c:335 + Processes move. They move from one processor to another. This + is important because if a process has just run at a given + processor, the processor's cache will still have cached memory + for the process. It is much cheaper to run this process again + on that processor than it is to run it at a new one: The new + processor will remove from its cache other memory and it will + have to move in new entries for this process; besides, the old + processor has now unuseful cached memory. movetime keeps the + earliest allowed next time for a move, to help in taking into + account ``processor affinity'': a process has affinity for the + processor where it did run last time. + proc.c:336 + Processes can be wired to a processor, so they always run on + it. wired points to the Mach structure for the processor where + the process is wired to. + proc.c:337 + ureg points to a saved Ureg for the process used for notes. + Forget this now if you don't understand. + proc.c:338 + error maintains the error string for the process (see + errstr(2)). Errors are represented by strings (portable, + readable) in Plan 9. They are set either by errstr or by a + failed system call. + proc.c:339 + Processes have a virtual address space. That means that + different processes have different page tables and their + virtual to physical address translations differ. In Plan 9, the + user part of the address space (what the user process sees) is + organized as a set of segments. Every process has a TEXT + segment with the code, a DATA segment with its data, and a + STACK segment with the stack. There are other segment kinds + like BSS, for uninitialized global data, etc. Some of them are + initialized later for this handcrafted process. + Processes may share segments, although some segments, like + stacks, are not shareable--that would make no sense. The + chapter on virtual memory discusses this. Each process can have + up to NSEG segments, and at this line, newproc is clearing all + the pointers to the NSEG segments. + proc.c:340 + Processes have a pid that identifies them. The pid is unique + within a node. incref takes a Ref structure, that contains a + counter, and increments it. The author uses incref and not ++ + because some other processor might be creating a process--well, + not now. incref uses locks to avoid races. So, in this line, a + new pid has been allocated. It is ``1'' because pidalloc was + initialized to zero early during kernel bootstrap. + proc.c:341 + noteid identifies a ``process note group''. Processes may be + posted notes, which are kind of UNIX signals. Notes are + strings, though--and may be posted through the network using + proc(3)! The system itself posts notes to processes when they + get an exception, and they can prepare to handle the note as + said in notify(2). You can post a note to a group of processes + (a set of processes with the same noteid), and all processes + with that noteid will get the note. A new note group has been + allocated. + proc.c:342,343 + If so many processes have been created that pidalloc or + noteidalloc get wrapped down to zero (by a set of increments), + panic. + proc.c:344,345 + If there is no kernel stack for the process (which is the case) + a new stack is allocated. Remember that smalloc keeps on trying + until the stack can be allocated. Right now the kernel is + running into a rather borrowed kernel stack. The current stack + belongs to the current processor, not to the current process, + and each process should have its own kernel stack where the + hardware will push the processor context on traps and + interrupts, and where kernel procedures will be called during + system calls made by the process. + The author tries not to waste time by keeping the kernel stack + for a process that existed before; its kstack would be non-nil + and the one used in the previous life of its Proc would be + reused for the new process. + proc.c:347 + All set. If you take a look to it all, the author has + initialized everything to just nothing--except for the kernel + stack, which comes along with the allocated Proc. The process + must now be given some resources to start working. + + main + userinit + + ../pc/main.c:237 + Back to userinit in main, it starts to give resources to the + process. At this line, create a new namespace pgrp. + newpgrp()Allocate a new process group + ../port/pgrp.c:43,51 + newpgrp simply allocates a Pgrp structure with everything set + to null. It also allocates a new process group id for the just + created group and sets ref to 1. + portdat.h:49,53 + Perhaps it's time to say a bit about Refs. a Ref structure + contains a lock and a counter. It is used to do reference + counting--hence the name--although it seems to be reused to + allocate pids, etc. too. A reference counter (Ref) is simply a + counter with an integer value that corresponds to the number of + users (i.e. references) of the data structure the counter is + associated with. For example, the Pgrp just allocated is used + only by the process being initialized. Therefore, the counter + must be one. Whenever a new user of the referenced structure + appears (e.g. a new process starts sharing our Pgrp), the Ref + is incremented. When a user disappears (e.g. a process using + the Pgrp ceases to exist) the counter is decremented. If the + counter ever reaches zero, the data structure is no longer used + and can be deallocated. The Lock is necessary because multiple + processes (and processors) could be updating the counter at the + same time. You might say that it would have been better to + implement reference counters by atomic increments and + decrements on word-sized values as the Linux kernel and some + others do. However, the Plan 9 approach has two advantages: it + is portable and simple to understand. By not optimizing what + does not need to be optimized, the machine dependent part can + be kept small, and the code can be kept more + understandable--ever looked inside of the Linux kernel? give it + a try. + + Lesson: Do not optimize; ever. Do optimize only when you really + have a performance problem. And be very reticent regarding what is + a performance problem: processors are increasing speed + dramatically. + Another reason I did not say is that the ``atomic increment'' + done by Linux is not so atomic--not a single instruction; at + least not in all the cases and for all the architectures. + ../pc/main.c:238,239 + The Egrp (environment variables) is also allocated--initialized + to zero, and its reference counter set to one. It puzzles me + why there is no newegrp routine. Admittedly newpgrp assigns an + id, but it is definitely not more complex than it would be + ``newegrp''. + main.c:240 + dupfgrp (which duplicates an Fgrp) is called with a null value + to create a new file descriptor group for the process. Again, + maybe a newfgrp routine would avoid surprises for the reader, + although the code is pretty clean and the author knows the good + reason that newfgrp does not exist. + dupfgrp()Create/duplicate a file descriptor group + ../port/pgrp.c:170,183 + First, a new Fgrp is allocated and if no Fgrp to duplicate was + given, the Fgrp is initialized by allocating a chunk of + entries. The DELTAFD name suggests that the fd array is resized + on demand to contain DELTAFD extra entries. That is in fact the + case. It is a usual technique employed to avoid fragmentation + and save time to extend arrays dynamically with ``delta'' new + entries every time they run out of entries--instead of + allocating a rather big array or refusing to admit more + entries, or allocating one extra entry at a time. + It returns with a brand new file descriptor group with no open + file (no channel linked into) and DELTAFD ready entries. The + reference counter is again 1. + ../pc/main.c:241 + newrgrp creates a new rendezvous group for the process. + newrgrp()Create a redezvous group + ../port/pgrp.c:53,61 + Again, this just allocates the Rgrp and sets the reference + counter to 1. + ../pc/main.c:242 + procmode is set to 640 (octal). Processes are handled by + files--like everything else. devproc implements the driver + supplying /proc files, that represent processes. This field is + simply the permissions for the process, in the file system + view. Using a file for everything is a nice way of fixing how + permissions are checked: in the same way file systems check + permissions on real files. + Although processes are files, there is no way to create new + processes by using file system operations. The author + considered that it wouldn't pay the effort, which is reasonable + since processes are created locally, within the same node. The + same happens to several other system calls that cannot be + performed using the file system. + main.c:244 + text contains the name of the file where the process text + (code) comes from. In this case the author uses the string + *init* because there is no file. + main.c:245 + Positive discrimination at work. eve is the name of the user + who boots the machine. Well, eve is the name of the array in + ../port/auth.c:37 that holds the name of the user who boots the + machine. Conventionally, in Plan 9, that user is referred to as + ``eve''. You know, ``Adam and Eve''. In Plan 9 there is no + ``root'' (superuser), but ``eve'' is given more privileges than + other users. The first process certainly runs on the name of + the user who booted the machine, hence this line. + main.c:246 + Ok, the kernel did that before. But the code is cool, isn't it? + fpoff()Places the FPU into an ``inactive'' state + main.c:247 + This executes some instructions to place the coproprocessor in + an ``inactive'' state. FPOFF is defined at l.s:373,378. By the + way, this and the previous line should be replaced by a call to + procsetup (main.c:564,569). + +Hand-crafting the first process: The state + + main + userinit + + main.c:250,256 + Preparing to switch to the first process. sched is a member of + Proc of type Label. A Label is like an oversimplification of a + jumpbuf in C. It is buffer where a copy of a program counter + and stack pointer is kept. Labels are used to implement + coroutines. More soon. + main.c:255 + The program counter registered in sched is the address of the + init0 routine. That function will be the very first thing + executed by the process being created and its purpose is to do + some final arrangements before jumping into the user program + for the process. + main.c:256 + The stack pointer saved in sched is the address of the + allocated kernel stack (kstack) plus the size of the allocated + stack minus something. On the Intel, stacks grow downwards; to + lower addresses. The address of the allocated stack is the + lowest address in the stack, but the stack pointer for the + empty stack would point to the biggest address of the allocated + stack on an Intel. The ``minus something'' thing is to reserve + some space at the bottom of the stack. What's the reservation + for? + ../port/portdat.h:91 + MAXSYSARG is the maximum number of arguments for a system call. + Here, the assumption is that each argument occupies at most a + word (if it occupies more a pointer would be used). + ../pc/main.c:256 + The number of words for arguments is multiplied by the number + of bytes in a word, because adding to kstack makes it move + counting characters, not words. By the way, the ``-12'' in the + comment seems to be a relic: 5 words times 4 bytes per word is + 20. I guess at some point there was a kstack-(12+something) and + the 12 was removed and the comment kept obsolete. But who + knows. + Taking into account that the reservation is for system call + arguments, for me it looks like the author wanted to ensure + that the (theoretically) pushed arguments for a system call + made within the kernel always are valid memory. There is a + check in trap.c:504 that precisely ensures that. For a kernel + process, the arguments would not stand in the ``user stack'', + but in the kernel stack instead. + What is needed is a kernel stack that has a fake return PC on + it so that gotolabel could overwrite it to ``return'' to the PC + in the label. But I am not sure more extra room be needed. + + NOTE: Is this ok? + main.c:261 + The first segment for the new process! (Not to be confused with + a hardware segment) It is going to be an STACK segment. Must + start at USTKTOP (initial top of user stack) minus USTKSIZE + (the size of the stack: it grows downwards!). The number of + pages in the segment must be the number of pages to get + USTKSIZE bytes. Forget a bit about how is the segment created + by now. But think that the Segment structure is created and it + contains in the end a bunch of (indirect) references to Page + structures. By now all the pointers to Pages are null. Note + that addresses mentioned here are virtual addresses. + The virtual memory for this process will look like the one + depicted in figure [189]3.7 (in that figure you can see the + virtual memory for a second process too, so you could get the + feeling of how this works). Right now the kernel is running on + the stack near the Mach structure; although it will later run + at the kernel stack for the process being created. + + CAPTION: Figure 3.7: Virtual memory for user processes. Permissions in + the hardware page tables will be set so that user code (ring 3) is + unable to access the last two gigabytes of the virtual address space. + + \resizebox{14cm}{!}{\includegraphics{vmem.eps}} + + main.c:262 + The stack segment is placed into the seg array for the process. + The stack segment always resides in the SSEG entry. + main.c:263 + Got the segment, but it has no memory. The call to newpage + obtains a new fresh page; later segpage plugs the page into the + segment. The page is requested to be cleared (the ``1'' for + newpage). UTKSTOP-BY2PG is the virtual address for the page. + The ``0'' for newpage is a pointer to the Segment reference and + we are not (yet) interested what that is for. Ignore how + newpage works by now, but take into account that a page is + allocated from the page (frame) allocator, its reference count + gets incremented, its referenced/modified field (modref) is + cleared, and its virtual address (va) is set to the one given + to newpage. + main.c:264 + segpage attaches the Page into the segment at the virtual + address specified by the va field in the page. One of the + pointers of the Segment points now to the page. The segment has + memory! And the user has an initial stack to issue procedure + calls on it. + One thing to note is that the MMU knows nothing about this page + yet. Only the kernel data structures know that the segment has + a page there. The MMU page table will be updated when the page + is first used. + main.c:265 + The page is ``mapped'' for the segment, but we are running in + kernel now, and not even within the context of the process + holding the segment (i.e. not using its address space). kmap + maps the page for the kernel and returns the address the kernel + should use to access the page. + dat.h:202 + Nice. kmap only has to add KZERO to the physical page address. + Remember that the kernel did set up a one-to-one mapping for + physical memory at address KZERO? That's why. No need to add + temporary entries to the current page table just to access + physical memory. But, why is kmap named ``map''? It is + conceptually mapping the page for kernel usage. It doesn't + matter if all the feasible maps were done before. The use of + kmap instead of just ORing KZERO also allows the author to + change his mind and use another implementation in the future + without changing the code that depends on kmap. Interfaces are + important in that they isolate pieces of code. + + Lesson: Keep clean interfaces between different components of your + programs. Interfaces allow you to modify one component without + affecting the others. + main.c:266 + VA gives the (user) virtual address for a given kernel address, + and bootargs places arguments for the initial process in the + page whose virtual address was given. By convention, Plan 9 + processes receive two arguments for their main program, the + argument count, and the array of arguments. Each argument is + just a string. And by convention, the argument zero is the + program name. See exec(2). + By using arguments and environment variables, users can control + what the executed program will do. + dat.h:201 + Because on PCs, the kernel shares the virtual address space of + the process running, VA simply returns its argument. + + NOTE: Not sure at all about my assumption about what VA is and why + it does return its argument + . + + main + userinit + bootargs() Sets up arguments for the first user process. + + main.c:305 + bootargs is building arguments for the user process entry point + in the user stack. sp is set to the top of the (empty) user + stack. Again, space is reserved for MAXSYSARG words. + main.c:307 + ac is the argument count. zero by now. + main.c:308 + first argument. av is the argument vector (i.e. like argv) and + its first entry points to the string ``/386/9dos'' just pushed + on the user stack by pusharg. The first argument is the program + name. In this case, it is customary that the program be called + 9dos, the Plan 9 version for PCs. + pusharg()Pushes an argument in the user stack. + main.c:286,294 + pusharg only advances the stack pointer to make room for the + string given, and copies it there. It returns the new stack + pointer (which points to the first character in the pushed + string). + main.c:309 + cp was set to point to the address where the kernel loader + placed the bootline. Ensure that it is null-terminated. + main.c:310 + In buf, the author is going to adjust bootline. The adjusted + version will be pushed as an argument. The 64 in the + declaration should probably be replaced by BOOTLINELEN. + main.c:311,318 + If bootline started with fd, place the canonical full name for + that: ``local!#f/fd0disk''. That saves typing for the boot user + and keeps the first process happy. The ``#f'' is the path to + the kernel floppy device. The name is pushed on the user stack + and placed into the argument vector. + main.c:314,316 + The same for hard disks. + main.c:317,318 + If option ``n'' is given to the first process, it understands + we are booting from the network. Surprisingly, option ``n'' is + missing from the boot(8) manual page (boot is the first + process). + main.c:319,324 + If the argument was a floppy or a hard disk, put the name of + the disk into the conf structure as if the user said + bootdisk=... in plan9.ini. + main.c:328 + The stack pointer points to words, but pusharg puts characters + in the stack. This is ensuring that dummy bytes are pushed on + the stack in case that is needed to complete a word. But all + the arguments are now pushed. + main.c:331 + Make room in the stack to put the argument vector--with + pointers to the arguments just pushed. The +1 is because it has + to be null-terminated (did you read exec(2)?). + main.c:332,335 + lsp is the argument vector passed to the user program. It is + built by pushing in the stack each of the recorded arguments in + av and null-terminating it. av[i] was the kernel virtual + address for the pushed argument. The added expression is to + ``move'' that address into the one seen by the process. It is a + translation. The length of the translation is the difference + between what the user code thinks is the start of the page + allocated for the stack, and what the kernel thinks is the + start of that page. This addition is necessary: remember that + the kernel is using a virtual address that depends on the KZERO + mapping for the stack (e.g. bigger than KZERO), but the user + will use its own user virtual address for the stack (e.g. lower + than KZERO). The KZERO map has no permissions to be used by + code running at ring 3. Otherwise, a user program would be able + to access all physical memory installed. + Perhaps the usage of a ``kernel-virtual-to-user-virtual'' or + ``kv2uv'' macro would make this more evident. + main.c:336 + The same adjustment has to be done to the stack pointer itself. + Beware, I did not tell, but sp is a global variable pointing to + the user stack for the boot process. That is why it is adjusted + and the routine just returns. It is probably a global because + init0 is the one using sp to jump into the user code. Because + we are going to loose our stack when jumping to init0 later, + the author cannot easily pass the argument using the kernel + stack of the new process. The sizeof ulong adjustment is for + the argument count. By the way, what happened to the argument + count? Does boot not use it and nobody noticed? Or are we + missing something? + + main + userinit + + main.c:267 + The user stack is all set. So release the kernel map for the + stack page; i.e. do nothing. + main.c:272 + A segment of type TEXT is created for the process. It starts at + UTZERO (the zero address within the text segment). Just one + page suffices (that's 4K). + main.c:273 + flushme has to do with caching. Forget it. + main.c:274,277 + The segment is placed in the list of segments for the process + (slot TSEG for the text segment). A page allocated at address + UTZERO and linked into the segment as before. Ignore the + memset; it is honoring flushme. + main.c:278,280 + The text segment for the process is initialized with the + contents of initcode. That is the program run by this process. + If you look into ../pc/initcode you'll see that it simply execs + boot. + Hint: You run new programs in new processes by doing a fork and + then an exec. By hand-crafting the first user process, the + system is doing the fork (It cannot use fork because there is + no process to fork). The best way to avoid hand-crafting an + exec is to use a silly user program that calls the regular exec + system call. You're done, and you can start as the first + process any program you want! + main.c:282 + ready changes the process state to Ready (ready to execute) and + links the Proc into the scheduler ready queue--the queue of + processes ready to run. More on that later. + +Starting the process + + main.c:164 + userinit is done, and schedinit starts up the scheduler now + that there is a process to schedule. schedinit never returns, + so main is really done. + + main + schedinit() Initialize and start scheduling. + + ../port/proc.c:57 + m points to the Mach structure for the current processor. In + ../pc/dat.h:160 you can see how it contains a Label (you know, + PC and SP buffer). setlabel()Record PC and SP. + ../pc/l.s:497,498 + setlabel receives a pointer to a Label. It first loads ax with + the pointer to the label passed. FP is the ``frame pointer'' + that points to the activation record for setlabel in the stack. + l.s:499 + the current stack pointer is copied into label[0] (considering + label as an array of integers now). + l.s:500,501 + The stack has the typical layout for a function call. The + function right now is setlabel. The top of the stack contains + the ``return pc'', i.e. the address where to continue executing + upon return from setlabel. That address is stored in BX, and + then BX is placed into frame[1] (considering label as an array + of integers now, assembly uses offset 4 because an integer has + 4 bytes). The return PC points to the instruction in schedinit + right after the call to setlabel. + So, m->sched has the PC and SP corresponding to the point of + execution right after setlabel was called. + ../port/proc.c:58 + up is a pointer to the current user process. Right now, it has + a null value. So, why does schedinit check for up? You will + know in the next chapter; You are not expected to, but can you + guess why? + It is obvious that up cannot be a global variable, because, how + could then different processors have different user processes? + up is defined at ../pc/dat.h:263 to be the externup field of + the Mach structure for the current processor. For each + processor, that field points to its current process. On + uniprocessors, a global up would suffice. + proc.c:83 + The scheduler is called. sched picks up a process and switches + to it. Right now there is only a first (boot) process. + + main + schedinit + sched() Schedule a process. + + proc.c:93 + If up is not null, there is a current process and the routine + must save its state; otherwise the kernel won't be able to come + back to the process when it be allowed to run again. But there + is no current process yet; forget this now. + proc.c:107 + runproc chooses among the set of ready processes one to be run. + up is set to point to it. Until the next time sched runs it + will be the current process (for this processor). + + main... + sched + runproc() Elect a process to run. + + proc.c:201 + Interrupts allowed. From now on, the timer, and any device can + request our attention. Interrupts will be serviced using the + kernel stack of the boot processor, by now. + proc.c:202,203 + idlehands does whatever the kernel should be doing while there + is nothing to run on the processor. It is defined as ``do + nothing'' at ../pc/fns.h:43. So, whenever no process is ready + to execute the kernel would be looping here doing nothing until + an interrupt (or another processor) changes things. + proc.c:204,243 + The loop keeps on looping until a process is found for running. + Whenever it gets that process, it jumps to the found label, + where the fortunate process will be put on the processor. The + only thing to notice right now is that there is an initial + process and the jump to found is done. Although it is also + interesting that the loops leaves rq pointing to the run queue + where the selected process resides. There are several run + queues as we will see in the next chapter. + proc.c:246 + Interrupts cleared: The routine is messing up with the queue of + ready processes, if an interrupt arrived and could change + things, the kernel could crash. + proc.c:247,248 + What if another processor is also picking up a process to run? + The routine should wait until that processor be done with the + scheduler queue. So, if it cannot gain the lock of the run + queue, it goes back to the loop. Interrupts were disabled + before locking the queue, but that affects just to the + processor running this code, other processors are free-running. + The goto will again enable interrupts and reenter the idle loop + that searches for ready processes. The routine will again + detect that a process can be run, and try to lock again the + scheduler queue (the process that it found before the goto loop + might still be there if not picked up by another processor). + This is busy waiting (i.e. no semaphores), but since the + routine cannot even know who should run, what else can it do? + You might say that the author could let the old current process + run a bit more time, but it could be that 1) there is no such + process or 2) such process is now blocked waiting for something + to happen. + This is not the case now. There is one process (boot) for us to + run. + proc.c:250,255 + Got the lock, now search for a process p that runs at this + processor (affinity!). head is the head of the processor ready + queue and the rnext field of Proc is used to link processes + into this queue. The loop selects any process p that either + 1. + did run in the processor running runproc, for affinity. + The routine can know that because the mp field of Proc has + the number of the field machno in the Mach structure where + it did run last. + 2. + did not move recently (to avoid trashing, i.e. moving a + process repeatedly between several processors). The code + knows that p did not move recently because in movetime p + has the earliest time when it is allowed to move; more + below. + l points to the Proc right before p in the run queue. + For us, p points to the ``*init*'' process just created by + main. + proc.c:260,263 + If the run queue is empty, go back to the loop. It could be + empty because all processes can be blocked (not the case). It + also goes back to the loop when the process is running on a + different processor (the process state is at the processor, and + not saved within the kernel; see the comment to learn how to + know that). + proc.c:264,271 + Note how easily the process is removed from the run queue. It + is no longer ready, it is going to be running. The counters for + the number of processes in the run queue and the number of + ready processes are updated. + proc.c:272,273 + This should not happen. The kernel (as any program) tries not + to flood the user with messages. Whenever something important + has to be said, it does so; otherwise it remains silent, + because, who cares? Guess why this shouldn't happen? + proc.c:274 + Done with the run queue, let other processors use it. + proc.c:276 + The process is being handled by the scheduler. Perhaps this + could be moved right before line :274, to make sure that if by + any bug, the process gets linked back to a run queue, everybody + will see that it is scheding and report the bug in line :273. + proc.c:277,279 + Must honor conventions. mp must have the number of the Mach + (processor) where the process last run on. mach, the pointer to + the structure is a different thing and may be null. Also, + movetime is set to the current time at the boot processor plus + 1/10 of second. HZ has the number of ticks per second and ticks + gets incremented every tick. The author uses the time at + processor 0, to make all processors agree on the current time + on an MP machine (every processor has its own vision of time + depending on the exact point when it was initialized). + So, p is removed from the ready queue and attached to the + processor where it is going to run (it is not going to be + ready, it is running). + + main + schedinit + sched() Schedule a process. + + proc.c:107 + Back in sched, the pointer to the current process is setup as I + said before. + proc.c:108,109 + Update the state and set the mach pointer. Its state is going + to be in the processor. Also, the proc field of m is updated. + You can go from Mach to the current Proc and vice-versa. + proc.c:111 + mmuswitch changes the address space to that of the new process. + mmuswitch()Switches the MMU to another page table. + ../pc/mmu.c:126 + This is the only line executed now, it is not really switching + the ``Intel task''. It will load the page table given: the page + table for the processor. If the process has its own page table + (lines above) that one is used instead; not the case! The + kernel stack for this process--used by interrupts and traps + that occur while running at user level--will be setup to the + top of the currently empty kernel stack of the process. kstack + points to the memory allocated for the stack but that is not + the top of the empty stack. + taskswitch()Switches the Intel idea of the current task. + mmu.c:27,40 + It is updating the TSS used by this processor to ensure that + the kernel stack pointer will be set at stack, within the + address space determined by the page table pointed to by pdb. + Besides, it loads the new page table pointer into the MMU. The + kernel has just switched to the address space of the just + created process, but it is still running using our boot + processor stack. Don't worry, the kernel is mapped the same way + at all page tables used for Plan 9 processes, so our (kernel) + virtual addresses keep on pointing to the same place in memory. + The chapter on virtual memory will make this more clear. + ../port/proc.c:112 + Our mind is going. Remember that sched was setup to have the PC + for init0 and the SP pointing to the end of init0's parameters + pushed manually by main. + gotolabel()Jumps to a saved context + ../pc/l.s:489,490 + gotolabel resets the stack and program counter with those of + label. It now fetches the pointer to the label into a AX. We + consider label as an array of int, although it is not. + l.s:491 + The stack pointer is set to that in label[0], the old boot + processor stack is gone. At this point the processor starts + using the kernel stack for the new process. + l.s:492,493 + The trick!, when gotolabel returns, the machine jumps to the + return PC (theoretically) pushed last on the stack by a call + instruction. By pushing the PC saved in label[1] (i.e. the + start of init0 code), the ret at line :495 ``returns'' to the + PC in the label. + l.s:494 + The convention is that functions returning an int use register + ax to pass the value back to the caller. Usually, the 1 just + returned will appear to be the return value of setlabel, but + that is explained in the next chapter. + + init0() (Kernel) entry point for the first process. + + main.c:190 + We are running at the address space of *init*, using its kernel + stack, and starting at the first instruction of init0. Now we + are a regular process executing within the kernel. Not just the + flow of control taken after the machine was reset. + main.c:197 + Interrupts were disabled during the context switch, but now the + kernel can handle them again. + main.c:199,206 + The comment states that chandevinit will not call any + rootinit--there is none. Remember chandevreset? The same thing + again. The comment regarding early kernel processes means this: + you do not have root and current directories for this process, + and although you want to setup such things for the user code + about to run, it is good to setup them now so that kernel + processes started before jumping to user code could at least + have silly root and dot directories, provided by the root + device. I don't know why there is no rootinit function with + this code in, the author knows. + main.c:203 + slash is a pointer to a Chan structure for the root directory. + It is set to point to a channel given by namec (``name + channel''), which corresponds to a file with the given name in + the current name space. What? I said namespace and there is no + root yet? Yes. Plan 9 have absolute paths, relative paths, and + paths for kernel device files. ``#/'' refers to the root of the + file tree serviced by the root device. Devices service file + systems named by ``#character''. ``/'' is the character for the + root device. The reason to have device paths is that no matter + how many adjustments a process makes in its name space, it + still has access to kernel devices and can access system + provided files--this is not the whole truth. + So now slash points to the directory serviced by root--which is + the root of root's file tree. You will see how this happens in + the chapter devoted to files. But you already got a glance + about it when we discussed channels and root. Try to read the + code anyway. + main.c:204,205 + cnameclose simply removes the given name--not the whole truth. + Right now, it is ``#/'', and that's a funny name for our /. So + the author creates a new name for the channel. What? channels + have names? Yes they do. Think that they are caching the file + name. + main.c:206 + cclone clones (dups) a channel. Now dot ``points'' to the file + where slash points. + main.c:208 + chandevinit calls the init routine for every Dev configured in + devtab. We are now a regular process, have a regular + (reasonably sized) kernel stack, are handling interrupts once + in a while, and devices can be initialized. I are not going to + describe how the various init functions work. Only when they be + relevant for the topic of one of the given chapters I'll do so. + main.c:210,223 + Various environment variables are set up for the current + process. The lines for setting the terminal environment + variable are generating a suitable value from the current + architecture name. The kind of cpu is set to 386, you know + that, right?. This is very important, because $cputype is used + among other things to choose what binaries are appropriate for + this architecture (saw /386 on your Plan 9 box?). The + environment variable service is also important, because boot + uses it to choose the rc script used to start system services. + On terminals you want rio, on cpu servers you probably don't. + For configuration (ini) parameters with names not starting with + *, the author sets environment variables to reflect their + settings. This is very important, because if the boot initial + program is ever updated to take into account some peculiarity + of the Plan 9 installation, an environment variable can be used + to indicate that; just by adding a line to the plan9.ini file. + Ignore the waserror and the poperror. They are described in the + next chapter. + main.c:224 + A kernel process is created named ``alarm''. It will start + executing at the kernel function alarmkproc. That function + iterates through the list of alarms searching for expired + alarms. Then it posts ``alarm'' notes to the processes with + expired alarms and sleeps until the next alarm expires. This is + really good because it turns alarm handling into a sequential + activity: not all processes must be setting up timers for + pending alarms while in the kernel. But we are not interested + on that now. + main.c:225 + touser transfers control to the user program within the given + process context. The global variable sp is used to tell touser + what should be the current user stack pointer. This was + initialized before by bootargs. + + init0() + touser() Perform an upcall from the kernel to the user code. + + plan9l.s:12,25 + The trick is to pretend that we return from an interrupt + suffered while running the user program. The processor is silly + and obeys, reloading the processor context with that of the + ``previously'' running user program. + plan9l.s:13,19 + All these pushes are building the fake stack frame after the + (not-existent) interrupt: User PC, user code segment, user + flags, user stack pointer, user stack segment. The code and + stack segments are the UDSEL and UESEL, which are the + appropriate selectors for UDSEG and UESEG. + plan9l.s:16,17 + A fake flags word with just the interrupt enable bit set is + pushed on the stack (as if it was a real flags pushed during + the interrupt that saved the frame being built). After the + iret, we are sure that interrupts will be enabled. It would be + a disaster if that was not the case, because no timer could + ever preempt the process. + mem.h:75,76 + Users run at ring 3, with non-privileged mode. Interrupts and + traps lead to a switch back to kernel mode as dictated by the + IDT entries. + plan9l.s:20,24 + The user data segment descriptor is copied into the descriptors + for all other user's extra segments. These are not loaded by + the iret, so they are updated by hand. + plan9l.s:25 + Up to userland. After this instruction, the program in initcode + is running at user-level, with the kernel fully initialized + below to handle traps and interrupts. + initcode:14 + When the program reaches this point, a trap is raised with + number 0x64, and the exec system call to execute the program in + /boot is issued, with arguments for /boot taken verbatim as + given to us. Remaining initialization is up to boot! + + Now, go read again boot(8) and take a look at what it does. That is so + clearly stated there that I'd rather interfere your learning by + repeating it here. + + One final consideration. Most of the work done by userinit in main + would be portable and could be said to belong to the port directory. + However, some machine dependent assumptions (e.g. the growing + direction of stacks) are being made. The code is more simple having a + single userinit than it would be having a machine independent userinit + and then a machine dependent userinit with the code that cannot be + made portable. Portability refers to the difficulty of porting the + code to a new system, not to the number of lines of the pc directory. + If the pc directory holds more files, but is easier to understand, and + easier to rewrite/adapt for a new architecture; that's fine! + + So you now know how the kernel boots, you have an operational system + and everything else is done by issuing system calls to the kernel. In + the following chapters we will read the code related to servicing + system calls for the major components in the system. + + Processes + + On this chapter, we will read the code related to processes. As I will + do with following chapters too, I try to follow the execution path for + system calls related to the chapter topic; but I make exceptions to + this discussion order whenever I feel its necessary. + + Although I am not discussing source code files, one file at a time, I + suggest you still try to read files as they appear. The worst thing + that can happen is that you don't understand the code and come back to + the commentary; but another thing that can happen is that you + understand most of it or it all on your own! That would be a big + progress! During this chapter, you will be reading these files: + * Files at /sys/src/9/port: + + portdat.h + Portable data structures. + + portfns.h + Portable functions. + + proc.c + Processes. + + pgrp.c + Process groups. + + devcons.c + Console device. + + devproc.c + Process device. + + alarm.c + Alarm handling. + + tod.c + Time of day. + + taslock.c + Test and set locks. + + qlock.c + Queuing locks. + + sysproc.c + Process system calls. + + * Files at /sys/src/9/pc: + + trap.c + Trap handling procedures. + + dat.h + Machine dependent data structures. + + fns.h + Machine dependent functions. + + clock.c + Clock handling. + + main.c + Machine dependent process handling. + + l.s + coroutines, locking and other low-level routines. + + ...and several other ones used as examples. + + Trap handling continued + + Before looking at how processes work, let's revisit trap handling once + more. I will be using the clock interrupt as an example of the trap + source. The clock interrupt is important because it is the mechanism + used to multiplex the processor among processes. In what follows, + assume that a user process was running while a clock interrupt happen. + + trap() C entry point for traps. + + /sys/src/9/pc/trap.c:218 + l.s dispatches the interrupt to the trap procedure, supplying a + pointer to the Ureg structure. + trap.c:225 + intrts is set to the result of fastticks. This Mach field + records the time stamp for the interrupt. That is necessary for + devintrts that handles interrupt time stamps. + devarch.c:597,601 + The time stamp is the result of the architecture specific + fastclock routine, which is defined to be cycletimer + devarch.c:586,595 + that returns the time stamp counter of the processor. (In case + you don't know, Intel processors from the Pentium up have a tsc + register which is incremented by the hardware at every clock + tick). + trap.c:226,230 + user is set to true if the interrupt (or trap) happened while + running user code. It was so, if the saved code segment within + the Ureg is the user code segment. + trap.c:232 + The trap number is kept in vno (vector number). It is going to + be used heavily and caching it in a variable will both enhance + the readability of the code and make it run faster. + What can be the value of vno? + trap.c:175,208 + excname contains names for the first 32 trap numbers, which are + generated by the processor. If vno is less than 32, it must be + an exception generated by the processor. See figure [190]4.1 if + you got lost. + + CAPTION: Figure 4.1: External interrupts (from the PICs) are + dispatched by IDT entries at VectorPIC; processor exceptions (caused + by a faulting instruction) are dispatched by the first 32 IDT entries; + other exceptions can be ``called'' by int instructions. In the end, + either trap or syscall service them. + + \resizebox{14cm}{!}{\includegraphics{traps.eps}} + + Otherwise, vno is either within VectorPIC (32!) and + VectorPIC+16, or it is above VectorPIC+16. In the first case, + the event corresponds to an external interrupt routed through + the PIC; in the second case, it must be a ``software generated + interrupt'' (i.e. int n). + trap.c:233 + Back to the trap routine, ctl is the vector control structure + for this trap. If there was no Vctl, we are in trouble: only + traps and interrupts that the kernel is prepared to service had + their Vctl set up by either intrenable or trapenable. Ah, and + beware of the assignment! + trap.c:234,266 + The kernel is handling a trap or interrupt that someone enabled + before. Things go well... + trap.c:234,238 + Only for interrupts, increment the number of interrupts + recorded in the Mach structure. And if the trap number is a + ``real interrupt'' (not a processor exception, and not a system + call), take the interrupt number from the Vctl. lastintr holds + the value of the last interrupt. That is used for debugging. + Perhaps a local variable would suffice at the expense of + reporting unwanted interrupts only when they happen--see below. + + NOTE: is this ok? + The check against VectorSYSCALL seems to be unnecessary, since + the trap for system calls is routed through plan9l.s to enter + syscall (below in trap.c) directly. + trap.c:240,241 + isr is the interrupt service routine. If the Vctl has one, call + it with the trap number. But, what is an ISR? + i8259.c:159,162 + Our PIC is the i8259. When we did enable it, the i8259elcr + (edge/level control register) had one bit set per edge + triggered interrupt. The i8259isr interrupt service procedure + was set as the service procedure for level triggered + interrupts, and it was set as the end of interrupt procedure + for edge triggered interrupts. Note how fields eoi and isr are + used to turn Vctl into a programmable interrupt handler. + i8259isr()Interrupt service routine for the 8259. + i8259.c:105,128 + All i8259isr does, is to tell the i8259 that the interrupt has + been serviced by acknowledging the interrupt. The chip now + knows that, and forgets about the interrupt until it is raised + again. If the interrupt is not acknowledged, the i8259 would + ignore that interrupt when raised again, because it would + assume that the processor is still handling the previous one + and is not prepared for servicing it again. The real interrupt + number is not the one coming in vno. vno was set to + VectorPIC+number because PIC serviced interrupts start at + VectorPIC in the IDT. By adjusting the number, irq contains the + interrupt number that the i8259 knows about: from 0 to 15. + trap.c:242,245 + Here is where the functions registered as interrupt handlers + (or trap handlers) are called. For the clock interrupt, the + handler will be clockintr, as said in i8253.c:130 by + i8253enable. Let's defer a bit what clockintr does, but note + how this is the point where the trap (or interrupt) is actually + serviced. Vctls for the same trap are linked through the next + field--there can be several handlers for the same interrupt. + The check in line :243 shows that a vector number can be + enabled (the kernel knows it is ok to get that trap) even when + there is nothing to do to handle it. + trap.c:246,247 + You now know what this does. When necessary, it calls the ``end + of interrupt'' routine. + trap.c:250,265 + The author preempts processes here. This is discussed in the + next section. + trap.c:267,271 + There is no vctl for the trap number: no part of the kernel + requested to handle it. Besides, the trap number has a name in + excname, and the trap comes from the user program. The user did + something weird and got a trap generated by the processor. + After indexing with the number to get a name for the trap, a + note for the trap is posted to the process. The process is + likely to die. Interrupts (disabled since the trap) are enabled + before posting the note. + You can get into these lines when any of the bad things named + in the excname array happen to the process. + trap.c:272,294 + Not a ``handled'' interrupt, and not a processor exception + while running user code. The condition checks that vno + corresponds to an interrupt and not to a processor exception. + Again, is the check for VectorSYSCALL necessary? + In this case, the number of spurious interrupts for the + processor is incremented and trap returns doing nothing more. A + message is printed because this shouldn't happen. Now that we + bothered the user, the for loop also reports other unwanted + interrupts at different processors. If you read the comment, + you see that the author is suspicious of IRQ 7. Most PCs keep + on delivering that interrupt under certain circumstances even + if you are not allowing interrupts. + trap.c:295,308 + Really into trouble. It must be a processor exception while + running kernel code. So the kernel has a serious bug. The + dumpregs call prints processor registers to aid in debugging + the kernel, and then the kernel calls panic. Only the boot + processor (machno zero) panics, other processors sit in a loop + until the panic at the boot processor causes the system to go + down. + trap.c:310,313 + If something was posted for the process, honor it. More on that + when we read note-handling code. When trap returns, l.s will + reload the processor context from the Ureg and it will resume + the activity previous to the trap. + panic()Issue a message and halt. + ../port/devcons.c:183,204 + Just to satisfy your curiousity, panic prints the given message + (dumps the stack to aid in debugging the kernel) and halts the + system by calling exit. + exit()Stop system operation. + ../pc/main.c:606,630 + exit prints the ``exiting'' message once for each processor + (note the use of the active.machs bit field. Then it waits for + a while to let the console print the panic and exiting message + (important if going through the serial line). Finally, if + running at the boot processor on a terminal, it loops forever. + For CPU servers, seems like a reboot (reset) is preferred to + restart CPU server operation--after giving some time to let a + human read the message. The exiting field of active is set to + true. + As each processor exits, its bit in active.matchs is cleared, + and other processors will exit too because they notice the + active.exiting bit (see ../pc/clock.c:53,54). This is shown + later. + + System calls + + Some system services (eg. clock handling) are requested by interrupts; + some others are requested by explicit system calls. Let's complete now + the discussion of system call handling. + /sys/src/libc/9syscall/mkfile:51,61 + This script generates for all system calls listed in sys.h, a + procedure that puts the system call number into AX, issues an + int instruction, and returns. These procedures are called to + issue system calls from user code, and are linked along with + every user program. Although there are many system calls, all + of them use the same trap number, and it is the parameter in ax + that determines which one is being requested. + + syscall() C entry point for the system call trap. + + ../pc/trap.c:471 + System calls get routed to syscall in trap.c by plan9l.s. + trap.c:477,478 + If a system call was issued from the kernel something is wrong. + The saved Ureg CS selector is used to check if the system was + running at ring 3 while the system call was issued. + trap.c:480,481 + Accounting for number of system calls services, and noting that + the process is servicing a system call. + trap.c:483 + registers for the user program are those saved in the Ureg. The + name is dbgreg because this is very useful for debuggers, to + fix up a faulting program and let it continue. The last known + PC for the process is that saved in the Ureg; remember that in + the Proc. + trap.c:485 + This line recovers the system call number from the ax register + from the saved user context. + trap.c:487,490 + Coprocessor stuff. If doing a fork (to create a new process) + and the coprocessor was used, save its state in the fpsave for + the current process. The new FPU state is inactive. By doing + this, the author ensures that both the parent and the child + process start with an PFinactive coprocessor state. The child + may not use the floating point unit, ever; in that case, its + FPU would be kept FPinactive. + trap.c:491 + System call servicing may take some time; enable interrupts. + All the information needed now is kept in up (and ureg). If an + interrupt arrives, the current (kernel) stack will be used to + service it and the current processing will continue after + returning from that interrupt. + trap.c:497,502 + (Ignore the error handling; just assume that line :497 is + entered). If the number is not that of an existing system call + (between 0 and nsyscall-1), post a note for the process and + report the error. As you will see, error would cause the + routine to continue at line :513 in this case. + trap.c:504,505 + When the stack pointer for the user code (sp) is not in the + first page mapped for the stack, or is so near the top of an + empty stack that it seems to be no space for the system call + arguments, check that the addresses going from sp to the end of + the system call arguments are indeed okay. The first stack page + is known to be okay because the kernel mapped it when the + process was brought to life; but we cannot be sure otherwise + that the user stack pointer looks fine and points into existing + stack space. The kernel checks before accessing the user stack. + + Lesson: Don't trust your users! If you implement any kind of + service, assume that users would be as malicious as you can image. + Usually, they will not be malicious, but they will have bugs that + could make them behave as if they were really malicious. + trap.c:507 + s in the Proc structure is set with the arguments for the + current system call. s is of type Sargs, which holds as many + words as MAXSYSARG says--i.e. as many arguments as a system + call may take. The word in the top of the user stack is + ignored; that would be the return PC for the ``system call'' + assembler routine called by the user code. + trap.c:508 + The string with the ``ps'' state of the process is updated to + contain the system call name. + trap.c:510 + And the system call is called. Note how the Sargs is supplied. + Arguments reside within kernel memory and can be used at will, + without checking that the stack addresses are still valid. + While the system call is executing, the kernel could switch to + a different process. + trap.c:530,538 + The result of the system call is placed into AX (will be + returned by the assembler user-level stub); and any note posted + (will be discussed later) is handled. + + Error handling + + Errors are handled by several routines within the kernel. Error + handling also includes routines used to report errors, be they fatal + or not. See pages perror(2) and errstr(2). + +Exceptions in C + + Errors are handled using error, waserror, nexterror, and poperror. + Handled with care, these routines provide a clean and fast error + handling mechanism similar to exceptions in other languages. + + ../port/portdat.h:593,595 + Each process has an array of up to NERR Labels for error + handling together with the number of entries in the array + (nerrlab). There is also a place to put an error message of up + to ERRLEN characters. + To see how this is used, let's see what syscall does for error + handling. In figure [191]4.2 you can see the whole picture. + Initially, the user calls a library function to perform a + system call, and it traps into the kernel (figure [192]4.2(a)). + ../pc/trap.c:495,496 + The return value for the system call is set to -1, which means + failure. waserror is called inside a conditional. If it returns + false, the system call will be called and return value set + accordingly; otherwise, syscall would return the error + condition. + waserror()Prepares for errors setting up an error label. + fns.h:123 + waserror increments the number of error labels for the process + (initially zero) and fills up another label in errlab + (initially the first error label). + trap.c:496 + Go back to this line. When syscall was called and it called + waserror, setlabel in fns.h:123 returns zero. The expression + (a,b) returns the value of b; therefore waserror returns zero, + which means that the then-arm of the if is taken. + During the system call the first error label is set and keeps + the SP and PC for the kernel as they were in trap.c:496 during + the early system call steps (see figure [193]4.2(b)). + poperror()Removes an error label. + trap.c:511 + The system call completed, and poperror is called. + ../port/portfns.h:199 + poperror simply decrements the number of error labels: it + removes the label added by waserror in trap.c:496. The state is + like that of figure [194]4.2(a) (of course, the PC and SP would + be that for trap.c:511), and not the ones as of line :496. + Now, imagine the system call number is wrong. + ../pc/trap.c:501 + error is called with Ebadarg to report the error. error()Raises + an error. + ../port/proc.c:1132,1138 + error copies the given string into the error string for the + process. (yes!, Ebadarg is a string with a descriptive text for + the error: portable, human readable, simple). Once the error + reason is noted, nexterror is called. nexterror()Re-raise an + error. + proc.c:1140,1144 + nexterror is where things start to move. A gotolabel jumps to + the kernel state as recorded in the last saved error label; and + the number of error labels is decremented to `pop' it off the + array. In our example, the error label was set by syscall and + both the PC and SP would be set as they were when waserror was + called. + ../pc/trap.c:496 + this time, waserror returns true, because after gotolabel, the + corresponding setlabel appears to return true. Therefore, the + conditional is not taken. In effect, nexterror is ``raising an + exception'', so that the flow of control resumes where it was + at the last waserror. Stack (local, or automatic) variables are + deallocated and function calls made since the waserror are gone + without returning. Can you see how the combination of waserror + and error forgets about an ongoing computation and resumes in + the top-level routine where an appropriate action can be taken? + In the figure [195]4.2, you can see how this works. To make it + more clear, the figure corresponds to a situation where a + system call is called ([196]4.2(a)), waserror in syscall sets + the error label ([197]4.2(b)), another kernel procedure + (syssleep) is called ([198]4.2(c)), and this procedure calls + error to raise an error ([199]4.2(d)). Noticed how it works? + + CAPTION: Figure 4.2: Error handling: labels are set in errlab and used + to raise ``exceptions''. waserror remembers the context; error + notifies an error and restores the context. + + [Initial context: The kernel starts executing a system call. PC and SP + correspond to the current kernel context.] + \resizebox{6cm}{!}{\includegraphics{err0.eps}} [Waserror is called: A + label is set to remember the context where to continue after any + error.] \resizebox{6cm}{!}{\includegraphics{err1.eps}} [More calls: + the kernel continues executing, more records pushed on the stack, + etc.] \resizebox{6cm}{!}{\includegraphics{err2.eps}} [Error called: An + error string is set, and the context is restored as it was when + waserror was called. This time, waserror returns a different value.] + \resizebox{6cm}{!}{\includegraphics{err3.eps}} + + Things are more interesting because waserror/(next)error pairs can be + nested. For example, the system call called in the previous example + could call waserror again, and a routine called by the system call + could call error. In this case, there would be two (nested!) error + labels in the error stack. The first call to error would jump to the + last label pushed by waserror; Then, a nexterror could be used to jump + back again (re-raise the exception), or alternatively execution could + proceed. + + To clarify things, the scheme looks like this +top_routine() { + // one typical idiom... + if (!waserror()){ // (1) + do_the_job(); + do other things; + poperror(); + } + + // another typical idiom... + alloc(some_memory) + lock(a_lock_held); + if (waserror()){ // (2) + free(some_memory); + unlock(a_lock_held); + nexterror(); + } + poperror(); +} + +do_the_job(){ + if (something fails) + error(msg); +} + + Things to note: poperror would remove the label pushed by waserror; + error would jump right to the line of waserror again, but that would + make top_routine follow the else arm in 1, and the if arm in 2; + nexterror would jump to whatever outer routine called waserror before + top_routine was called, telling the caller that we suffered an error. + In the case nexterror is called, the reason for the error would still + be msg. do_the_job does not need to return the error condition to the + caller function, and how top_Routine does not need to check for any + error condition returned by do_the_job. + + One final note, special care has to be taken when calling error (or + nexterror) because some resources could have been allocated (or locks + acquired) since the last call to waserror. Take into account that + nexterror does not know anything about either resources or locks; + therefore, the routine that did allocate/acquire those resources/locks + must call waserror to release them on errors (like 2 in the example). + To pick up an example, add the line +char *p=malloc(BIGSIZE); + + right at the beginning of do_the_job, and consider that ``something + fails''. Got the picture? + +Error messages + + You already saw panic and a couple other routines. They are easy to + follow, therefore I will not comment on them. Nevertheless, the pprint + routine is curious. + + The pprint routine reports errors not on the console as panic does, + but on the standard error stream for the process. The kernel is + assuming that stderr is descriptor number two and is opened for + writing; not assuming too much. This is a detail that shows how Plan 9 + was built with the network in mind from the ground up. + + UNIX would print in the console any message about problems related to + the current process but not causing a panic (e.g. an ``NFS server not + responding'' when a network file cannot be used due to server + problems). The message is of interest to the process but not to the + whole system. By printing the message in the console, the user sitting + there can see the message, but the process could be started from a + terminal miles away! So it's better to print the message to wherever + the process prints diagnostics (stderr) and let the process (owner) + know. Besides, the user sitting at the console usually does not care + of any problem for processes he has not started. + + Clock, alarms, and time handling + + The clock is used to maintain the system time (when the TSC is not + available) and to implement alarms. An alarm causes a function to be + called after an specified amount of time. There is a system call + alarm(2) that can be used to request an ``alarm'' note to be posted to + the process after the given period of time. By handing the note, the + user process can achieve the effect of the alarm: calling a function + after a period of time. + + time(2), cons(3), alarm(2) are manual pages that can be of interest + for you now. + +Clock handling + + Let's see how the clock works starting at the clock interrupt. + + trap + + /sys/src/9/pc/trap.c:218 + The clock interrupt happens and l.s dispatches it to the trap + procedure, supplying a pointer to the Ureg structure. + Interrupts stay disabled. + trap.c:242,245 + For the clock interrupt, the handler was clockintr. + + trap + clockintr() Services the clock interrupt. + + clock.c:43,44 + The increment notes in the Mach structure that another tick + passed by. The call to fastticks updates the fastclock field of + Mach; that is used by non-boot processors to update their own + TSCs!. Looks like although the boot processor is in charge of + time, other processors try to be in sync. + clock.c:45,46 + The PC image in the Proc for the running process is updated to + be real one. This is also done when entering a system call. + clock.c:48 + Do some time accounting, as we will see in the next section. + clock.c:49,50 + Record execution statistics for kernel profiling, if needed. + kproftimer is a pointer to _kproftimer only when the kprof + device has been init'ed. If not doing kernel profiling, the + pointer will be nil and ignored. + clock.c:51,52 + This processor is not really active. It may be exiting (or + halted!) but it got yet another clock interrupt because + interrupts are enabled. Ignore it. + clock.c:53,54 + Some processor paniced (or started shutdown) and the kernel is + exiting. If this is running at a different processor, it + notices now and calls exit to terminate operation at this + processor. exit resets the bit for the processor in machs so + that other clock interrupts are ignored at lines :51,52. + Noticed how one processor does not perform immediate actions on + another processor? The best the author can do is to ``kindly + request'' the other processor to do something: processors are a + living thing. + + Lesson: When using multiple processes (processors) to do something, + do not directly interfere with the execution of others; ask them + for anything you want instead. This prevents dangerous race + conditions because only you can do things to yourself. + clock.c:56,62 + Alarms and clock0links serviced here! (see next section). + clock.c:64,68 + Will see in the virtual memory chapter. Some other processor + asked we to flush our MMU by reloading our page table. We do + so. The clock interrupt is a good place to see if anyone else + is asking this processor to do anything: the requester does not + need to wait too much (although it needs to wait!). + clock.c:70,75 + More scheduling affairs. Forget this now. + clock.c:76,79 + If the code interrupted was running at ring 3, account for + another tick in a counter maintained in the bottom of the + user's stack, and call segclock to do profiling on the user + code. Interrupts will be reenabled after clockintr returns to + trap, and trap returns to recover the user state. + While kernel is servicing an interrupt or a system call, + ureg->cs would not be UESEL, and no time will be charged to the + (interrupted) user. + clock.c:11,18 + To avoid synchronization problems due to multiple clock + interrupts on machines with several processors, the boot + processor does clock handling. That is the meaning of ``0'' in + clock0link. To service ``kernel alarms'', i.e. stuff that needs + to be done every tick for the kernel, links a established into + clock0link. Lines :57,62 are servicing these links. A link is + just a pointer to a clock procedure to notify of the clock + tick. addclock0linkEstablishes a procedure called at clock 0 + ticks. + clock.c:21,34 + Which ones are the links? addclock0link inserts a new link into + the list. So, grep for addclock0link! + ns16552.h:104 + The serial line UART wants uartclock be called every tick. + ../port/devcons.c:1015 + The console driver wants randomclock be called every tick. + That is to maintain the random number generator (see + cons(3)). + ../port/devmouse.c:85 + mouseclock should be called to redraw the cursor. The + author knows the user can be kept happy if the cursor + appears to be responsive, even if the machine is heavily + loaded. + It is important both to be fast, and to appear to be fast! + ../port/tod.c:47 + todfix should be called to maintain (fix!) the time of + day. + +Time handling + + The clock ticks, and every tick the tod (time of day) module updates + the system idea of the time of day. + + main... + consinit + todinit() Initializes time of day handling. + + ../port/devcons.c:471,475 + The console driver init function calls todinit (and + randominit). Remember that this driver was initialized during + boot as every other driver configured into the system. + tod.c:43,48 + todinit calls fastticks to update the hz field of the tod (time + of day) structure. Perhaps a local variable instead of tod.hz + would make it clearer that tod gets its hz when todsetfreq is + called, and not now. todsetfreq()Initialize TOD frequency. + tod.c:54,60 + todsetfreq is calculating the multiplier mentioned in the + comment at lines :8,24; read that comment now. + todfix()Updates the time of day. + tod.c:147,156 + Once per tick, todfix gets called. The last variable retains + its value from call to call, and is used to know if the last + call was issued at least one second ago. The author does not + want to do todget too often (to avoid wasting processor time), + but he wants it to run often enough to avoid overflows in the + counters used (note that there are many ticks per second). One + second appears to be a reasonable compromise. + todget()Updates the time of day and returns it. + tod.c:95,141 + todget is doing the actual time of day updating. It is used + both to update the time of day and to get the time of day. It + is usual that a routine to get something can be reused to + update such thing before accessing it. + tod.c:101,104 + If not yet initialized, the routine initializes the module by + calling fastticks and setting up the hz field. In any case, + ticks has the value for our TSC based fast clock. + tod.c:105 + tod.last has the time when todget was last called (1 second + ago). Now diff has the number of ticks passed since then. + Initially, tod.last is zero until either todset is called to + set the time of day or todget runs and notices that tod.last is + too far away in the past. + tod.c:108,109 + x has the number of nanoseconds since the epoch time. Note how + tod.off is added. It will be updated later. + tod.c:111,129 + Only the boot processor does time of day handling. Other + processors may be calling todget through devcons (cons(3)) or + devaudio (audio(3)) to get the time of day. + tod.c:114,121 + If the time of day must be adjusted, change it a bit at a time. + Values sstart and send are set by todset. The ilock is used to + prevent others from accessing tod in the mean time and also to + prevent interrupts while it is being updated. ilock is + discussed later. + tod.c:124,127 + Not too often, last is updated to record the interval since the + last call and off to record the time since epoch. + Where are overflows? last is used to get in diff the number of + ticks since the last adjustment. If diff gets too big, x could + perhaps overflow. In fact it doesn't matter where would the + actual overflow happen, what matters is that the author assumes + that diff would never be too big and the algorithm is coded + assuming that. The author is ensuring that the assumption + holds. + tod.c:132,135 + Time could go backward because time can be changed. In no case + a time change will report an earlier time next time the user + asks--that could really hurt programs that depend on the time + behaving properly. Because of the same reason, the author + adjusted time a bit at a time in the above lines, just to + permit the user to adjust the time without big jumps into the + future. + lasttime holds the time reported by gettod. One thing is what + the routine reports, and another what it believes. + tod.c:137,140 + Time of day finally reported. + tod.c:66,89 + Time is set by this routine. To see some call, look at + devcons.c:1211,1241, where the console driver accepts writes + from the user to set or adjust the time of day. A -1 t is the + convention for adjusting time a bit at a time: todset sets + sstart and ssend (and delta) so that todget adjusts time + slowly. If the time is not negative, the time is reset to the + given value. todset can adjust the time in multiple ways. Go to + devcons.c and try to see when is todset called. Correlate that + with the cons(3) driver manual page. seconds()Returns the time + of day in seconds. + tod.c:158,168 + Just to complete tod, seconds returns the time of day in + seconds, and is used mostly by drivers for time outs and time + stamps recorded in seconds. + +Alarm handling + + You now know how time goes by. User processes can know by reading from + the console driver's time file. However, users also may want to be + notified after a given amount of time. They also may want to sleep + (i.e. be kept blocked and not ready to run) during a given amount of + time. + + portdat.h:79,89 + Both Alarms and Talarm are headers for lists of processes. + portdat.h:75,576 and :585 + The Alarms and Talarm lists are linked using the palarm and the + tlink fields of the Proc structure--perhaps the names should be + more uniform here. The author keeps the Alarms list holding all + processes with alarms in the current machine. Each node (Proc) + in the list keeps the alarm time in alarm and the list is kept + sorted by call time. The list Talarm is analogous but maintains + sleeping processes. + + syssleep() sleep system call. + + sysproc.c:478,493 + syssleep calls tsleep to put the calling process to sleep. + + sysalarm() alarm system call. + + sysproc.c:495:499 + sysalarm calls procalarm for the same purpose. In both + routines, arg[0] is the period of time for sleeping or for the + alarm. (see alarm(3)). Forget a bit about tsleep and look into + procalarm. + + sysalarm + procalarm() Sets up an alarm for the process. + + alarm.c:84,87 + old set to the previously set value for the process alarm. (did + you read alarm(3)?). The previous value is recovered by looking + at time in the boot processor. + alarm.c:88,91 + Canceling an alarm. alarm is set to zero, and it will be + ignored later by checkalarms. + alarm.c:92 + The absolute time for the alarm computed by looking at time in + processor 0. Using absolute times for alarms lets the author + compare the processor time with the alarm time without much + arithmetic nor race conditions. It also avoids adjusting the + alarm field as time goes by. + alarm.c:94,102 + First the Alarms list locked, to avoid races with other + processes using alarm or the alarm list--e.g. alarmkproc uses + the list to notify expired alarms, and should not use the list + while it is being modified. Then search any existing alarm + entry for the current process. (Saw the + pointer-to-pointer-to-node thing again?) Line :98 removes the + entry if it exists. + alarm.c:104 + By this line, the process has no previous alarm registered in + the system: not in the list, no pointer to any ``next'' alarm + entry. + alarm.c:105,116 + Alarms list not empty, the node f with the first alarm after + the one being installed is located. That node is linked after + the current process at line:109, and the current process linked + in place of that node at line :110. If this alarm is going to + be the longer one, l in line :115 points to the ``next'' + pointer in the last node, and the current process is linked + there. + alarm.c:117,118 + Alarms list empty, easy. + alarm.c:119,123 + One way or another, the current process is now linked into the + alarm list, the alarm time is recorded, and the list unlocked. + The goto is used to share the code at done. It is breaking the + loop and the conditional in a clear way. Remove the goto, and + the code would become less clear. + + trap... + clockintr + checkalarms() Checks for expired alarms. + + alarm.c:44 + checkalarms will be called later by clockintr. + alarm.c:49,53 + It looks at the head of the alarms list (no lock!) and calls + wakeup if the first alarm expired. The list is not locked + because the kernel still runs with interrupts disabled and + because it is checkalarms the one removing alarms from the + list. So, it is safe to look into the first node because it + would not disappear under checkalarms feet. Can you tell know + why to cancel an alarm the alarm field of Proc is set to zero? + By the way, what happens if while the alarm is pending the + process dies? Hints: the alarm list points to processes; a zero + alarm is ignored; alarm removes any pending alarm before + installing the new one. + Ignore the rest of checkalarms by now. + + alarmkproc() Entry point for the alarm kernel process. + + alarm.c:12,38 + Remember from the previous chapter that a kernel process is + running alarmkproc? We have an endless process scanning the + alarm list. The first time it entered the loop, it locked + alarms, saw that no alarm was pending, unlocked it, and called + sleep on alarmmr. So, by the time a process is setting up an + alarm in the code just discussed, alarmkproc was probably + ``sleeping on alarmmr''. The call to wakeup at line :53 awakes + alarmkaproc and it continues running right after line :36. (The + return0 is a function that returns zero, but ignore that now). + The author uses a kernel process to post alarm notes to + processes with expired alarms. This process is subject to + regular process scheduling and may sleep when locks cannot be + gained. + alarm.c:18,19 + The current time recorded in now, and the list locked. If the + list cannot be locked because other processor is calling alarm, + the kernel process will wait here until the lock is gained. + alarm.c:20 + Pick up the head of the alarm list, and keep on scanning while + the alarm time is past. The alarm does not happen at the exact + time it was scheduled at; it may happen later. + alarm.c:21 + If the alarm was canceled, alarm is zero and we must ignore the + entry. + alarm.c:22 + The lock on debug is needed to post the alarm note. If we + cannot get it, better break the loop and go to sleep until next + time checkalarms awakes alarmkproc again--yes, starvation is + theoretically feasible, but so improbable that who cares? + I hope you will appreciate that there are algorithms that are + theoretically not good, but the author still prefers to use + them than to modify them to keep theoreticians happy and incur + into more overhead. This does not means that theory is not + important; this only means that you have to balance theory with + what you know from practice. + alarm.c:23,28 + An alarm note is posted for the process, debug unlocked and the + alarm reset (to zero). Any error during postnote is ignored + (poperror!). The alarm kernel process better keeps on trying to + post alarm notes, than die if an error happens while posting + one. This is one place where waserror is called not to + deallocate resources on errors, but to ignore errors. + + trap... + clockintr + checkalarms + + alarm.c:55,56 + Back to checkalarms, it does by itself the processing of + ``sleep alarms'' using the Talarm list. If the list is empty, + nothing else has to be done. + Can you guess why the author uses two different alarm lists? + alarm.c:58,73 + For any talarm expired, a wakeup on trend is issued (and the + process removed from the talarm list). When twhen is zero, the + timer is ignored. + This is less serious than posting a note, and I guess that is + the reason why talarms are handled directly by checkalarms. + Talking about reasons, there is one for having two lists + (Talarm and Alarms), a process may setup an alarm and go to + sleep waiting for the alarm to happen. So, two different lists + have to be maintained. The Alarms list always has a postnote as + the associated action and does not need to accept a + user-supplied handler. + Talarm keeps kernel timers used whenever the author wants to be + notified after a while. Alarms keeps alarms set for user + processes. More clear now? + + syssleep + tsleep() Sets up a timer and puts the process to sleep. + + proc.c:525,570 + tsleep is the one placing processes into the talarm list. A + function and an argument is given. The ``t'' is Talarm is for + ``timer''. The kernel uses Talarms to setup timers that call a + given function when expired. After the discussion of Alarms, I + think the code should be mostly clear. tsleep can be called + several times for the (current) process (cf. lines :538,548) + and the previous timer is canceled in that case. The sleep call + in line :567 is the one that actually makes the process sleep. + I defer the discussion on sleep/wakeup until later. + delay()Waits a bit. + clock.c:83,100 + To complete the discussion about timing, delay routines should + delay for so few time that they loop to implement the delay. It + would not pay to setup a timer because of the small delay time. + (Well, admittedly, delay is called to make the kernel wait for + long periods of time too; guess why?) + + Scheduling + + Plan 9 has preemptive scheduling, which means that from time to time + processes are preempted and moved out of the processor. To now when a + process should be preempted, the system clock is used. Most scheduling + is done by sched and runproc in ../port/proc.c, as we saw while + learning about system startup. Interestingly, there is a resched + function defined in ../port/portfns.h, but does not seem to be + defined; just a relic. + +Context switching + + sched() Switches to another process. + + proc.c:91,113 + We saw sched before. If you grep for it, you will find that it + is called mostly whenever the current process should yield the + processor to let another process run. In particular there are + two points of interest where sched is called: + ../pc/trap.c:250,265 + The author preempts processes here when a higher-priority + process is waiting for a processor. + ../pc/clock.c:70,75 + The author calls sched from the clockintr routine. Let's + start here. Assume that a process is running as shown in + figure [200]4.3 and look at that figure while reading + below. + + [proc.c:91,113] We saw <#9481#>sched before. If you grep for it, you + will find that it is called mostly whenever the current process should + yield the processor to let another process run. In particular there + are two points of interest where sched is called: + ../pc/trap.c:250,265 + The author preempts processes here when a higher-priority + process is waiting for a processor. + ../pc/clock.c:70,75 + The author calls sched from the clockintr routine. Let's start + here. Assume that a process is running as shown in + figure [201]4.3 and look at that figure while reading below. + +Context switching + + trap + clockintr + sched() Switches to another process. + + [clock.c:70,71] Another clock interrupt caused a call to clockintr. + After checking for Alarms and Talarm, if there is no current process + (yet) or the state of the current process is not Running we return + from the interrupt. If the state is not Running, it is likely that the + process is being moved from one scheduling state to another; + therefore, we better do nothing. For example, if the process is going + into sleep, sched will be called by sleep. Routines like clockintr try + not to interfere when the job will be done by someone else. + [clock.c:73,74] If any process is ready to run, it calls sched. + anyready checks the nrdy global in proc.c:32. Things look as shown in + figure [202]4.3(b). + + trap + clockintr + sched() Switches to another process. + + ../port/proc.c:91 + sched is called. Another process should run on this processor. + This involves both choosing the next process and switching to + it. The first task (implementing the scheduling policy) is done + by runproc, the second task is done by sched. + + CAPTION: Figure 4.3: The kernel uses a scheduler stack to switch + processes. There is one label per processor to switch to the scheduler + and one label per process to remember where the process was within the + kernel. Thick arrows show where the processor is running. + + [The user code is running] + \resizebox{7cm}{!}{\includegraphics{intr0.eps}} [An interrupt leads to + sched. The user state is in the Ureg, and the kernel state is saved in + Proc.sched.] \resizebox{7cm}{!}{\includegraphics{intr1.eps}} + [Switching to the scheduler stack, using Mach.sched.] + \resizebox{7cm}{!}{\includegraphics{intr2.eps}} [The scheduler uses + the sched label of another proc to switch to it.] + \resizebox{7cm}{!}{\includegraphics{intr3.eps}} [The kernel resumes + within the new process context; it starts returning] + \resizebox{7cm}{!}{\includegraphics{intr4.eps}} [The new process user + code runs] \resizebox{7cm}{!}{\includegraphics{intr5.eps}} + + proc.c:93,94 + Assume there is a current process (interrupted by the clock + interrupt). We disable interrupts from now on. + proc.c:97 + Account the number of context switches. + proc.c:99 + procsave saves the machine dependent part of the process. + procsave()Saves mach. dep. context. + ../pc/main.c:575 + procsave starts running using the pointer to the current + process as a parameter. + main.c:577 + If the process used the coprocessor + main.c:578,579 + If the process is dying, there is nothing to do but to reset + the FPU. + main.c:581,589 + But if the process is not dying, the FPU state is saved into + the fpsave entry in the Proc structure. Next time we switch + into this process, the state will be reloaded into the FPU. Pay + attention to the comment. + main.c:591 + Once saved, the process fpstate is inactive, meaning that there + is no FPU state for this process in the real FPU. We will see + more about FPU handling soon. + ../port/proc.c:100 + setabel called with sched for the current process. It saves the + current stack pointer and program counter into label (see + fig. [203]4.3(b)). And returns 0! (note the line + ../pc/l.s:502). The kernel stack for the current process has + activation frames for sched, clockintr, trap, and the saved + Ureg for the process moving out of the processor. + proc.c:105 + the gotolabel reloads the program counter and stack pointer + with the ones recorded in the sched label of the Mach structure + for the current processor. If you remember from the starting up + chapter, the label was set in schedinit. + + schedinit() Calls the scheduler. + + proc.c:58 + And here we are!, the stack was the initial kernel stack used + during boot (above the Mach structure), and the program counter + was set to point right before line :58. The old process is + mostly gone, although up still points to it--it is not null. + Things are like in fig. [204]4.3(c); the scheduler stack is + used to switch from one process to another. + proc.c:59 + Starting to switch. We set the proc pointer in Mach to null. + proc.c:60,63 + If the process is still runnable, but is being preempted, ready + makes arrangements so it gets Ready to run in the future. + proc.c:64,79 + If the process is dying, the state is adjusted, MMU machine + dependent structures (page tables) are released (the prototype + page table set up during boot will be used thereafter), and the + process will be linked into the free process list. More about + process death later. + proc.c:80,81 + The process state is saved, so set mach to zero, and forget + about the current process. + proc.c:83 + sched called again, with no current process. We are still + running in the scheduler stack, like shown in fig. [205]4.3(d). + It is a good thing to have a scheduler stack. It allows + procedure calls in occasions when the previous process kernel + stack should not be used; i.e. to switch to a another process, + the author does not need to keep on using the current stack. + Besides, it is convenient when there is no current process. + + schedinit + sched() Switches to another process + + proc.c:107 + Back in sched, runproc selects another process to run. You know + from the previous chapter that it loops when no ready process + exists. + proc.c:108,110 + The process is linked to the processor and set running. + proc.c:111 + Switched to the page table for the next process. We are in the + address space of the coming process, but kernel addresses are + shared, so don't worry. + proc.c:112 + the gotolabel reloads the saved kernel context (stack and PC) + for the coming process. That context was probably saved at line + :100 when the new current process was last preempted. gotolabel + reloads the the stack and PC, pretending to return 1 from the + setlabel that filled up the label. the kernel switches to the + kernel stack for the coming process. We end up as shown in + figure [206]4.3(e). + proc.c:101 + setlabel returned true, so procrestore is called to reload + machine dependent processor context, interrupts are allowed + again, and sched returns. For the PC, procrestore is defined to + do nothing (../pc/fns.h:98). Probably, the setlabel for the + current process was made while running sched, called from + clockintr, called from trap, so the return starts the unwinding + of the kernel stack. When trap finally returns, the IRET in l.s + will reload the saved Ureg for the current process--see + fig. [207]4.3(f). + + One more note: the state of the new current process could be other + than sown in this example execution. In general, that process could + have a kernel stack corresponding to any path of execution leading to + a call to sched. You will see more examples of that during this + chapter. + +FPU context switch + + You saw how procsave and procrestore were used to save and restore the + FPU state. But how is the FPU context really handled? + mathinit()Initializes FPU traps/interrupts. + ../pc/main.c:552,559 + FPU traps were initially enabled by main. procsetup()Initialize + FPU for a new process. + main.c:565,569 + Initially, the fpstate is set to FPinit, as you saw in the + previous chapter, and the FPU set to an offline state. + mathemu()Services the FPU emulation fault. + main.c:517,520 + If the process uses the FPU, it will get an emulation fault, + because the FPU was set off. mathemu calls fpinit to initialize + the FPU (enable it) and sets the fpstate to FPactive (because + the process is known to use the FPU). + ../port/proc.c:99 + If a new process is getting switched in, the FPU state for the + previous process (which used the FPU in our example) is + saved... + ../pc/main.c:575,592 + because it was FPactive. Its FPU state is now FPinactive. + If the next process does not use the FPU, its state will be + FPinactive. + When the first process (that used the FPU) is switched back + again, procrestore does nothing!. + main.c:521,535 + When the process starts using the FPU again, another FPU fault + leads to mathemu. As it is FPinactive, its FPU context was + saved and must be reloaded. Lines :533,534 do that. The process + is now FPactive, as it was before the first time it was + preempted in our example. I hope you will see how the author + avoids switching the FPU context when the process does not use + the FPU. + Some other OSes try to avoid even saving the FPU state, by + keeping track of who did use the FPU last and saving that state + only when it is absolutely necessary--i.e. before another + process uses it. But, take into account that Plan 9 runs on MP, + and the FPU state might be loaded into a different processor + the next time. Things are more simple in the way they are in + the code, and fast enough. + +The scheduler + + How does runproc select the next process to run? It applies a + scheduling policy to select a process. Let's look at it. But we should + start by the routine allowing processes to run. + + Getting ready + + ../port/proc.c:24,33 + Processes willing to run are either Running, or they are Ready + to run. Ready processes are linked into Nrq scheduler queues. + Each queue has processes of a given priority, and high-priority + processes get more CPU than low-priority ones. Unlike + UNIX[208]8.1, Plan 9 uses higher priority values for higher + priorities! + + ready() Puts a process in the ready queue and recalculates priority. + + ../port/proc.c:142 + Before running, the process must be set Ready and linked into a + ready queue--so that later runproc can pick it up. + proc.c:147 + With interrupts disabled (because nobody should touch the + scheduler queues), + proc.c:150,153 + if the process was running (is being preempted), rt is + incremented. rt counts the ``running time'' for the process + measured in quanta. Every time the process goes from Running to + Ready, rt is incremented. So rt measures how many full quanta + the process just consumed. pri is set to the formula: + + \begin{displaymath}\frac{\frac{\mathtt{art}+2\mathtt{rt}}{4}}{\mathtt{ + Squantum}} \end{displaymath} + proc.c:153,157 + if the process was not running (it is not being preempted), the + average running time, art, is set to + \(\frac{\mathtt{art}+2\mathtt{rt}}{4}\) , and pri is set to + + \begin{displaymath}\frac{\frac{\mathtt{art}+2\mathtt{rt}}{4}}{\mathtt{ + Squantum}} \end{displaymath} + --rt is reset to zero. Wait a bit to understand what is going + on and keep on reading. + Squantum is set to \(\frac{\mathtt{HZ}+\mathtt{Nrq}-1}{ + \mathtt{Nrq}}\) in proc.c:138. If the number of run queues is + small with respect to the machine HZ (ticks per second), + Squantum is almost \(\frac{\mathtt{HZ}}{\mathtt{Nrq}}\) , + dividing the HZ evenly among the run queues. If Nrq is big + regarding HZ, Squantum is almost + \(\frac{\mathtt{Nrq}}{\mathtt{Nrq}}\) . But this formula is + empirical and only the author knows how it was adjusted to + yield the current expression. For our PC, HZ is 82 and Nrq is + 20, yielding a Squantum of 5. To augment pri in one, rt has to + be incremented by 5/2, or art has to be incremented by 5. + So, every time the process gets Ready, (i.e. is preempted), pri + increases slowly as rt and art increase. rt influences more pri + than art does. Things can change because if the process leaves + the processor voluntarily (i.e. gets blocked, and after a while + it goes from a non-ready state to Ready) its rt is set to zero, + and its art gets decreased. The order of lines :154,157 is + ensuring that changed values are used next time and not now. + proc.c:158,160 + Here it is, pri is set to basepri minus the pri value just + computed. If the computed pri was small, the process priority + would be close to basepri; otherwise it can go all way down to + lower values, or even zero. So, for the process, it is bad when + the computed pri gets big. Should the process not block, + neither rt nor art would be decreased, so the computed pri gets + bigger and basepri-pri would be smaller; should the process + block, rt will be reset, and art will be decreased so that the + computed pri gets smaller and basepri-pri would get closer to + basepri. If the same process keeps on blocking (e.g. gets + Running, computes a bit, reads a file and blocks, gets + unblocked, and so on), its final pri will end up being basepri. + If the process keeps on running, its final pri will be very + low. Interactive processes tend to block, and they are not + penalized; CPU intensive processes are penalized. By the way, + do not pay much attention to the exact details of the formulas, + other similar ones are likely to work too; these things come + out of the author's experience with the system: they are + empirical. + Did I already said that these things are empirical? Don't + forget. + proc.c:162,170 + If the process basepri was above PriNormal, avoid the priority + decreasing too much. If the process is waiting for a lock, its + priority would be just PriLock. Otherwise, its priority is the + pri just computed. + portdat.h:522,525 + PriLock is 0, a very low priority value. If the process is + waiting to gain a lock, it is not doing anything useful (yet), + so the system penalizes it. It is better to let the process + holding the lock run, and penalizing ourselves is a fine way of + favoring that. Besides, note that PriKproc (basepri for kernel + processes) and PriRoot (basepri for processes running /* files) + are above PriNormal. That means that both kernel processes and + ``root processes'' will be kept above priNormal, no matter what + they do. The system gives them priority over normal process, + that are below PriNormal. Processes in both priority classes + (above and below PriNormal) get their priorities adjusted + during time, but they stay within the same class. + Both root processes and kernel processes are working on behalf + of the whole system. It makes sense to give them priority + because otherwise the whole system would suffer. Try to find + out which ones are the kernel and root processes in your Plan 9 + box (hints: how are you using your local disk files? how are + you talking to other nodes in the network? how are you using + your swap file? did you look at the mkfile?) + proc.c:171 + There is a run queue for each priority level. Set rq to be the + queue for the process priority this time. + proc.c:173,184 + The process is set Ready and linked into its priority queue. + readytime is set to the time the process was set Ready. + proc.c:185 + If interrupts were enabled at line :147, they are enabled now; + otherwise they stay disabled. The reason for using splx is that + ready must both lock the run queue and disable interrupts; as + ready can be called either with interrupts enabled or disabled, + the author restores things as they were. + + In few words, you now know that Plan 9 uses dynamic priorities within + two priority classes. + + Picking up a process + + Finally, sched calls runproc to pick up a process to run. You already + read runproc in the previous chapter, but let's look at some details + now. + + sched + runproc() Elects a process for running. + + proc.c:204 + Once out of four times, runproc forgets about processor + affinity and priority, and picks up the process waiting + longer--trying to avoid starvation and to balance the load of + processors here. + proc.c:211,219 + Low priority queues are scanned first! xrq points to the run + queue with the minimum readytime found for the head process, + and rt is set to the minimum readytime for that process. By + line :219 the lowest priority process sitting at the head of a + run queue that was ready before the other ones is located by + xrq. Just the head is used, to avoid locking the ready queues + while scanning for processes. + proc.c:220,226 + If there is such a process, rq is set to the queue where it + stands, and p to the process selected. The goto goes to the + place where runproc tries to run it. If there is no such + process, runproc loops again; but next time it will honor + priorities and affinity. If the process is not wired to the + processor, movetime is set to zero, that is a really small + value and will allow the process to move. + proc.c:232,241 + Three out of four times, run queues are scanned from high to + low priority; as they should. If there are no processes, + runproc keeps on looping. The process chosen is the first one + that either + + is the first in the queue that reached its movetime (did not + moved recently; same reasons as above), or + + is the first in the queue that did run on this processor (to + keep processes running within their favorite processors). + By the way, the author didn't really choose a process, but a + run queue instead! + The queues were not locked. But the worst thing that may happen + is that the process gets removed from the ready queue (by + another processor) and runproc will find its rnext pointer to + be nil. Since Procs are allocated statically in a big process + table, there are no worries about crossing a dangling pointer. + proc.c:245 + got a process. + proc.c:246,248 + Someone is using the queues (more than one processor), try + again. + proc.c:250,255 + Now a process is chosen, but using the rq selected before. If + things have changed, the fortunate one may be different from + the process than caused this ready queue to be selected--this + is the price for avoiding locks before. l is set to the process + before the one selected, and p to the first one with affinity + for this processor: the selected one. + proc.c:260,263 + No process selected, try again. If the reason was that no + process had affinity for the processor, in a couple of loops + runproc will ignore this fact. + proc.c:264,271 + The process removed from the ready queue. It is no longer ready + but running. l is used to remove it from the list. Counters + adjusted accordingly. + The process rnext is not cleared!, if any other processor is + scanning through the ready queues, it could still jump over to + the previously next process in the ready queue, even though the + process is not ready. The author is careful to avoid + unnecessary locks in places where locking would mean severe + performance degradation. + proc.c:272,273 + A scheduling bug? + proc.c:276,280 + And there it goes. sched will switch now to it, as we saw + before. + + Processes tend to be chosen from the head of the list, and are + inserted at the tail in ready. The effect would be a round-robin for + processes with the same priority--if you forget about affinity or if + there is only one processor. + + There is one thing to consider. What if a high-priority process was + blocked, and it suddenly gets Ready? That can happen because we are + running a low priority process and an interrupt notified the kernel + that the event the process was waiting for, just happened. The answer + to the question is in trap. + + trap + + ../pc/trap.c:255,265 + Conditions say that: it is an interrupt but not the clock or a + timer; there is a currently running process, and there are + higher priority processes waiting. Interrupts happen often, but + clock and timer interrupts happen really often. Checking for a + higher priority process when a frequent (but not ubiquitous) + interrupt happens is a way of checking often (but not always!) + for a high priority process. The preempted flag is set when the + author commits to preempting the current process in favor of + the higher priority one, the check at line :258 ensures that + this code would be ignored if the process is already being + preempted. A simple call to sched ensures that the + high-priority process will be able to run. + sched will not return before the other process runs. When it + returns (the current process has been switched back to the + processor), preempted will be reset, and another interrupt may + yield to a new preemption. + The slphi at line :262 is problematic, because the new + interrupt might cause a preemption before doing a + return-from-interrupt for the current one. Think that each + interrupt pushes more frames into the kernel stack, and they + are popped only when returning from the interrupt. If enough + interrupts arrive, the kernel stack would overflow and the + system would probably crash. But the preemted check seems to + suffice to avoid that. + + You know affinity, but processes can be also wired to a processor. + + procwired() Wires a process to a processor. + + proc.c:354,383 + procwired wires p to any processor (if bm is less than zero), + or to the processor specified by bm otherwise. Most of the + routine is picking up a processor to wire the process to. The + one with less wired processes is used. Perhaps a new m->nwired + field would save most of this code. A big movetime is given to + the process so it never(?) moves. + + Finally, I don't discuss it, but accounttime in + ../port/proc.c:1210,1233 maintains values for processor load averages + and process run times. + + Locking + + During the description of what the code is doing, you saw lots of + locks and locking routines. I skipped all of them. Now it's time to + discuss locking one you know about process scheduling and process + priorities. + + Because Plan 9 runs on MP machines, there are several locking + primitives employed. Let's see them from the most simple to the most + complex. + +Disabling interrupts + + Remember that within the kernel setlabel and gotolabel are used to + provide coroutines. Therefore, within the kernel, the kernel decides + when to leave the processor by using gotolabel in favor of another + kernel routine. So, you could say that ongoing system calls for other + processes are not an issue regarding mutual exclusion for critical + regions within the kernel. + + What? You don't understand the meaning of ``mutual exclusion'' or + ``critical region'' (go, reread the material for the OS course you + attended several years ago and come back later). + + However, while the kernel is executing a routine in favor of a user + process, an interrupt may arise. The interrupt will start a different + routine of the kernel while the previous one is stopped. If a system + call is being executed and an interrupt arrives, the interrupt routine + can try to access resources used by the previously ongoing system + call. Therefore, the kernel must prevent interrupts from happening + while touching data structures that can be manipulated by the code + executing after the interrupt. If you take into account that an + interrupt can lead to the suspension of an ongoing system call and a + context switch to another process, you can imagine that code executing + after the interrupt can touch almost any data structure in the kernel. + + According to what I just said, if there is a single processor, + disabling interrupts would ensure mutual exclusion among processes + while executing within the kernel. While touching an important data + structure, the kernel can disable interrupts and it knows nothing will + ``preempt'' it in the meanwhile. There can be more than one processor, + but even so, there are structures that are handled (read: written) + only by one processor (e.g. Mach for each processor) and can be + protected by disabling interrupts. Other critical regions, like the + code doing a context switch for a process, are also protected this + way. How are interrupts enabled and disabled? + + The abstraction used to enable/disable interrupts is the ``processor + priority level''. Imagine that the processor is running at a given + priority (0 or 1). While running at a low priority, interrupts + (high-priority events) can interrupt the processor. While running at + high priority, interrupts cannot interrupt the processor. This comes + from the days UNIX was implemented because the processor used actually + worked this way. + + spllo() Sets processor priority low. + + ../pc/l.s:433,437 + spllo (set priority level low) is used to enable interrupts. It + pushes the flags word in the stack and pops it back into ax + (the return value of spllo). The interesting part is sti, which + sets the ``interrupt enable'' bit in the flags word. The author + uses the stack to move flags into ax because the only way flags + can be accessed on the Intel is by pushing/popping it on/from + the stack. The value returned by spllo can be used to restore + the previous ``processor priority level'' (i.e. the interrupt + enable bit) as it was before the spllo. + + splhi() Sets processor priority high. + + ../pc/l.s:423,431 + splhi (set priority level high) is used to disable interrupts. + Lines :424,426 set in m->splpc the return PC as saved in the + stack by the call to splhi. That is, m->splpc holds the PC of + the instruction that called splhi; That is for profiling, but + can be used for debugging too. If somehow interrupts are + disabled and they shouldn't be, you could know who is guilty + for that. The real work is done at line :430: interrupt enable + cleared. The old value of flags is returned as in spllo. + + splx() Sets processor priority as given. + + ../pc/l.s:439,448 + splx (beware that it continues until the ret), exchanges the + flags and the value passed as a parameter. If the kernel calls + spllo and, later, passes the value it returns to splx, flags + would be restored as they were; i.e. the priority level would + be restored. Lines :440,442 are saving the caller's PC in + m->splpc for the same reason as above. Line :445 is taking the + first parameter (FP is the frame pointer). + ../pc/l.s:450,451 + spldone is not used as a function. If spllo or splx is + executing, the PC is between spllo and spldone. If you look at + ../port/devkprof.c:38,43 you will see how are m->splpc and + spldone used. The routine _kproftimer maintains statistics + about what parts of the kernel execute during what times, the + author seems to want to charge the spl times to the caller and + not to the routines themselves. + +Test and set locks + + By using an atomic tas instruction, which tests for the value of a + word and sets it to true value, mutual exclusion can be achieved even + with interrupts enabled. That is important because it is overlay + expensive to disable interrupts on all processors--and it is also + complex. So, kernel routines can agree that when a lock word has a + true value (non-zero), the critical region cannot be entered. By using + an atomic tas, the previous value of the lock can be tested and set + without race conditions. The first one to set the lock, acquires it. + Another useful variant of tas is xchg, which exchanges a register with + a memory position atomically. + + tas() Tests and sets a word. + + ../pc/l.s:462,466 + The Intel only has xchg, so tas uses the intel instruction to + emulate a tas. The parameter passed is the lock being tested + and set. The value set (0xdeaddead) is irrelevant. Line :465 is + atomic! + + xchgw() Exchanges a two words. + + l.s:472,476 + xchgw is a wrapper for the Intel instruction xchg. Two + parameters passed are the ``register'' and the memory position + being exchanged atomically. By the way, xchgw seems to be used + only the astar device, while tas is the routine actually used + for mutual exclusion by the kernel. So, if astar had another + way of doing its business, xchgw could go away--as seemed to + happen with xchgl. + + The kernel does not use tas directly, but uses routines provided by + taslock.c instead. With the help of several machine dependent + routines, most of test-and-set code is kept portable. + + lock() Acquires a test-and-set lock. + + ../port/taslock.c:31 + lock acquires a lock using tas. The Lock structure is defined + in ../pc/dat.h:24,31; you will see how it works now. + getcallerpc()Gets the PC after the instruction that called it. + ../port/taslock.c:36 + getcallerpc uses the address of the lock parameter to locate + the PC of the caller. If you now the address of the first + parameter, you can know where in the stack is the return PC and + obtain that value. The PC of the caller is stored in Lock.pc + for debugging purposes. If a lock has problems, it is useful to + know who did set the lock. + taslock.c:39 + Here it is. tas tries to set the lock, which is actually + Lock.key. If the lock was cleared before tas executed, the + return value would be zero; otherwise, the return value shows + that the lock is already held by someone else. + taslock.c:40,43 + The Lock structure is updated to record the process setting the + lock and the PC that called lock; all done. isilock is a + boolean stating that the lock was set by ilock and not by lock. + More about this soon. Most of the time, the lock would be not + set and the kernel would execute these lines. If the lock is + found to be set most of the times, that would show contention + for a given lock within the kernel. The affected data structure + would better split in several ones, or some of its parts locked + separately, and perhaps a different kind of lock used (Noticed + that Proc has several locks?). + taslock.c:46 + glare counts how many times the lock couldn't be acquired at + the first attempt. + taslock.c:47 + Trying to get the lock. If there is a current process and it is + running (i.e. the lock is not requested once the process was + committed to block) the kernel can call the scheduler to wait + for a while until the lock be released. + taslock.c:48,53 + If the scheduler can be called, save the process priority so it + can be restored later, and set the priority to PriLock (i.e. to + zero, so that almost every other process would be preferred by + the scheduler). If the process priority is kept as it is, and + the holder of the lock has a lower priority, the scheduler + could prefer to switch back to the current process instead; So, + the author is making sure that such thing would not happen. + lockwait is set in Proc to state that the process is waiting + for a lock; the scheduler would maintain the priority set to + PriLock no matter what the process does. + taslock.c:55,56 + Now waiting while trying to get the lock repeatedly. The number + of ``attempts'' is recorded in inglare. When contention for + locks appear, the author can at least now that in mean, + inglare/glare loops are needed. If the relation gets too big, + it can be a symptom that locking has to be adjusted. + taslock.c:58 + Important loop! If there are several processors, the current + one has l->key on its cache. If the processor does not write + the lock, it will not interfere other processors because the + lock value will be read from the cache. If the processor would + tas the lock instead, it could lead to a high bus + contention--because the value would be written and that usually + means that the value has to be put into main memory, or that + the processors must talk to update their caches. So, do not + even try to set the lock until we know it is no longer set. + taslock.c:59,65 + Just one processor, and can call the scheduler. Do so (see + figure [209]4.4(a)). The i counter is used to report we are + ``looping many times'' once per thousand iterations. + + CAPTION: Figure 4.4: Behavior of test-and-set locks. + + [test-and-set lock on uniprocessors. It is better to context switch if + the lock cannot be acquired.] + \resizebox{8cm}{!}{\includegraphics{tas.eps}} [test-and-set lock on + MP. It is better to wait for a while.] + \resizebox{8cm}{!}{\includegraphics{tasmp.eps}} [test-and-set lock on + MP in case a context switch was made. The time used for context + switching at processor one is wasted time, since no user process is + running in the mean time.] + \resizebox{8cm}{!}{\includegraphics{tasmpsw.eps}} + + taslock.c:65,70 + More than one processor. It is better to wait for a while until + the lock is released by another processor (see + figure [210]4.4(b)). tas locks should be held only during small + amounts of time, so it would not pay to do a context switch + (see figure [211]4.4(c)). In the case of a monoprocessor, it + would be a waste to keep on looping because unless we release + the processor the lock holder would not run. Note the high + number of iterations until reporting that we have problems. + taslock.c:72 + Out of the while, so the lock was released when we last checked + it in line :58. Now try to set it. + taslock.c:73,80 + If got the lock, update the Lock structure and restore the + previous priority in case it was set to PriLock. If did not get + the lock try it again--note that no tas will be tried until the + ``lock-read-only'' while finishes. + + canlock() Tries to acquire a test-and-set lock. + + taslock.c:136,145 + canlock tries to set the lock but just once. It lets the caller + know whether the lock was acquired or not. Routines willing to + give up or to do something else when a lock cannot be acquired + can use canlock. You already saw how canlock was used by the + scheduler to give up and try again. The implementation of + canlock uses tas as lock does, and initializes the Lock + structure appropriately. + + unlock() Releases a test-and-set lock. + + taslock.c:148,158 + Releasing a tas lock is easy, just set the lock (l->key) to + zero and the next tas will return zero. Note how the lock is + checked to be set. The call to coherence ensures that the lock + value is seen by other processors--a ``no-op'' on Intels. + + ilock() Interrupt-safe version of lock. + + taslock.c:85,86 + ilock is a variant of lock. The lock routine works fine when it + comes to mutual exclusion between different processors. + However, the kernel may also want mutual exclusion between a + process running inside the kernel (e.g. a system call) and an + interrupt handler. Imagine that the kernel gets a lock on the + memory allocator, and an interrupt arrives. If the interrupt + handler wants to allocate memory, it would try to acquire the + memory allocator lock. You get a deadlock! This is shown in + figure [212]4.5. + + CAPTION: Figure 4.5: Acquiring locks without ilock. The process 1 + acquires a lock, and gets interrupted. The interrupt handler tries to + acquire the lock and has to block!. The lock will never be released. + + \resizebox{9cm}{!}{\includegraphics{ilock.eps}} + + The way to avoid the deadlock is to disable interrupts besides + acquiring the lock. By disabling interrupts, no interrupt + handler can request the lock because no interrupt handler will + run (in the current processor) while the lock is held. Other + processors are not an issue because they can try to acquire the + lock without a deadlock. + The author could disable interrupts and then call lock, but if + the lock was not acquired, interrupts might be cleared for just + too long. The ilock routine tries to acquire the lock and + restores the previous interrupt state when the lock cannot be + acquired. + taslock.c:95 + Interrupts disabled, the previous priority level kept at x. + taslock.c:96,102 + If tas got the lock, return with interrupts disabled and the + lock set. (isilock records that the lock disabled interrupts + too). The previous priority level is saved in l->sr. The unlock + routine needs that to restore interrupts to their previous + state. + taslock.c:112,113 + A single processor available, how could the lock be released if + interrupts are disabled? This message would mean that usage of + locking primitives has to be fixed. + taslock.c:117,120 + While the lock is held by another process, restore the + interrupt status. Once the lock is released, interrupts are + cleared before trying again to acquire the lock. + + iunlock() Interrupt-safe version of unlock. + + taslock.c:161,177 + Locks acquired with ilock are released with iunlock. Again, the + author ensures that it is an ilocked lock. The last two lines + are manually setting splpc to be the PC of the caller of + iunlock, and then doing the rest of splx (remember that splx + did fall down to splxpc?). + +Queuing locks + + A Lock can be used to protect small critical regions. When the lock is + being held for a long time, or when there is much contention on a + lock, another kind of lock is needed. If processes would have to wait + for a long time to acquire a lock, they better sleep while waiting. + When there is much contention for a lock, it is also better to respect + the arrival order of the lock requesters; otherwise you can get + starvation. + + ../port/portdat.h:60,66 + A Qlock is a lock that maintains a queue of processes waiting + for the lock. The head and tail fields maintain the list, the + locked flag shows whether the lock is held or not. And finally, + the Qlock structure has to be protected for race conditions: + use is a Lock that must be held to operate on the QLock. use + will be held during a small amount of time--as soon as the + process either gets the Qlock or gets queued, the use lock is + released. The author is building locks appropriate for big + critical regions and high contention using the simple tas locks + as the building block. + + qlock() Acquires a queuing lock. + + qlock.c:17 + qlock is the routine called to acquire a QLock. + qlock.c:21 + Important! to protect q, a tas lock is acquired. + qlock.c:23,27 + If the q lock is not set, set it and return. In this case the + tas lock is released; it was held only while the QLock was + manipulated. + qlock.c:28 + The number of processes queued on a QLock increased. The author + can know whether the queues of QLocks are really used or not. A + kernel is a living thing, the only way to issue a good + diagnostic is by asking it about its symptoms. + qlock.c:29,38 + The process is queued in the QLock. There must be a current + process, otherwise there is nothing to queue. See how the + process is queued in the tail of the queue? + qlock.c:39 + The process was not Ready, now it is Queueing and will not run + again until extracted from the lock queue and placed back in a + ready queue. + qlock.c:40 + The PC of the caller recorded to find out guilties for locking + bugs. + qlock.c:41,42 + The QLock was protected by the use lock during all this time; + now release the tas lock and switch to a different process. The + current process is queued waiting for the lock. + + canqlock() Tries to acquire a queuing lock. + + qlock.c:45,57 + Like canlock, canqlock tries to acquire the lock and returns + reporting whether the lock was acquired or not. canlock is used + on the use lock (because canqlock should not block). + + qunlock() Releases a queuing lock. + + qlock.c:59,76 + qunlock releases a QLock. If there are processes in the queue, + the one at the head is extracted and set Ready. That process + has the lock (Hence locked is kept set to true). If there is no + process waiting for the lock locked is set to false. QLocks are + acquired in FIFO order regarding the request time. + + There are many places where QLocks are used. They are used wherever + the lock is likely to be held for a long time--and processes are + likely to wait for the lock a long time. For instance, I/O devices + like cons, pipe, etc. use QLocks because I/O operations are likely to + be slow and take a significant amount of time. + +Read/write locks + + The locks discussed above could suffice. However, many times there are + several processes accessing a data structure just for reading it, and + there are several other processes writing the data structure. It makes + no sense to serialize the readers of the data structure. When there is + a single processor, readers would necessarily read the data structure + once at a time; but on multiprocessors, several processors could be + reading the same data structure at the same time without any problem. + With the locks seen before, multiple processors willing to read the + same data structure could be stalled, waiting for another reader to + finish. That is why there are read/write locks. While you read below, + see figure [213]4.6 as an example of how R/W locks would block + readers/writers. The figure should be clear by the end of this + section. + + CAPTION: Figure 4.6: Read/Write locks. Typical scenario showing how + readers/writers acquire the lock. + + \resizebox{10cm}{!}{\includegraphics{rwlocks.eps}} + + rlock() Acquires a R/W lock for reading. + + qlock.c:79 + rlock tries to acquire a RWLock for a reader process. If will + be able to acquire the lock if there is no one writing the + locked data structure--or waiting to write. + qlock.c:83 + Again, using a tas lock to protect the lock data structure. + qlock.c:84 + Now it is important to know an overall number of readers and + writers for RWLocks; that can reveal symptoms that RWLocks are + used where a more simple lock could be used instead. + qlock.c:85,90 + writer is a counter for the number of writers. Perhaps it + should have been named writers, but there is only one writer + allowed at a time, hence the name. So, if there is no writer, + and there is no process waiting for the lock, the lock is + acquired. When there is no writer, but there are processes + waiting for the lock, that processes would be writers waiting + to acquire the lock. In this case, it is important to queue and + give priority to the writers who arrived before. Otherwise, a + continuous flow of readers might starve a writer. readers is + incremented to reflect that there is one more reader holding + this lock. There is no ``locked'' field in the RWlock; if there + are readers or writers, the lock may be locked or not depending + on who you are and who is holding the lock. + qlock.c:92 + One more reader had to wait, let the author know. + qlock.c:93,102 + The process (and there must be one) is queued in the list of + processes queueing at the lock. The queue is maintained using + the qnext field of the Proc structure. + qlock.c:103,105 + The process state shows that the queued process is queueing for + reading. As with QLocks, the process will not run until + dequeued from the lock queue and placed back in a ready queue. + + runlock() Releases a R/W lock for reading. + + qlock.c:109 + Readers use runlock to release a RWlock for a reader. + qlock.c:115 + If there are more readers holding the lock besides the one + releasing the lock, nothing else has to be done: the lock is + still held by remaining readers. Besides, if there is no + process waiting to acquire the lock, there is nothing more to + do because by updating readers, the lock is released when + readers gets down to zero. + qlock.c:120,122 + The last reader went away and there are processes waiting. The + first process waiting in the queue must be a writer--note that + should it have been a reader, it would have entered acquiring + the lock, because there was no writer waiting by that time. + There can be readers in the queue too, but they are queued + because they saw there was another process queued (and a writer + was among them) and decided not to pass it. The process state + is checked to see whether the queued process is waiting for + read or for write--see figure [214]4.7. + + CAPTION: Figure 4.7: Queueing locks. The process is just removed from + the ready queue while waiting to acquire the lock. In this case, the + lock is a read/write lock and the process state is used to see what + the process is waiting for. + + \resizebox{12cm}{!}{\includegraphics{qlock.eps}} + + qlock.c:123,128 + As the last reader is gone, awake the writer by removing it + from the queue and setting it ready. The writer counter is set + to one to reflect that one writer holds the lock. That would + prevent further readers/writers to acquire the lock. Now that p + is ready, the scheduler can elect it for running. + + wlock() Acquires a R/W lock for writing. + + qlock.c:132 + Writers use wlock to acquire the RWlock for writing. + qlock.c:138,145 + If there are no readers and there are no writers (no writer, + actually) the lock can be acquired. The PC that called wlock + and the process that acquired the lock are noted, for + debugging. For readers, the author thought it was not worth to + record that information, probably because there are multiple + readers. + qlock.c:148,158 + The process must wait until either the last reader releases the + lock, or the current writer releases the lock. The process + state is updated to reflect ``waiting to acquire a RWlock'' for + writing, and the scheduler is called. The process won't run + again until dequeued by either the last reader or by the + writer. Even if the lock is held by readers, further rlocks + will have to wait. + + wunlock() Releases a R/W lock for writing. + + qlock.c:165 + Writers releasing the lock call wunlock. + qlock.c:169,175 + If there is no process waiting, the lock is released by setting + writer to zero. + qlock.c:176,184 + If the first process waiting is a writer, it is given the lock + (note that writer is kept set to 1). The writer is allowed to + run by removing it from the lock queue and setting it ready. + qlock.c:186,187 + The process must be one waiting to acquire the lock for + reading. Otherwise, some bug caused other process to enter the + queue, or some bug caused the waiting process to change its + state. + qlock.c:189,197 + The first process waiting is a reader, but as the lock is going + to be acquired by a reader, any other reader waiting can + acquire the lock at the same time too. The author scans the + queue seeking for processes queueing for read. All of them are + removed from the queue and set ready. readers is updated to + reflect that the lock is held by that many readers. The + scanning stops as soon as a writer process is found, because + remaining readers should yield to the writer who arrived + before. + qlock.c:198,199 + It did not harm that writer was set until now, because the tas + lock was held. But better update it now. + + canrlock() Tries to acquire a R/W lock for reading. + + qlock.c:202,216 + canrlock is used as canlock, but for a RWlock locked for + reading. Perhaps a canwlock could be added for completeness, + although no one is using such a thing now. + + One comment before proceeding. When a process is being awakened to + acquire the lock, you saw how the author does the bookkeeping for the + awakened process (updating counters et al.) in the process holding the + tas lock. You will also see that when author implements note posting + and notifying, whatever can be done easily within the notifying + process is not done by the notified process. This is a general rule + that when you hold a lock and some bookkeeping has to be done for the + process being awakened, you better do it in the awakening process + before releasing the lock. Otherwise, the awakened one would have + necessarily to acquire the lock for the resource and that could lead + to more race conditions. Keep the code simple. Of course there is a + counterpart rule that if you can do something more easily in the + process awaken than it can be done in the awakening process, you + should do it where it is more simple. Keep the code simple, did I say + that? + + Synchronization + + There are several forms of synchronization in Plan 9. User processes + can use rendezvous(2) to synchronize. + + Besides, the OCHEXCL bit for permissions given to the create(2) system + call, allow processes to synchronize on their access to files: only a + process may have the file open at a time. This simple mechanism avoids + the need for complex file locking primitives found on other systems + like UNIX. This is very important since there is distributed access to + files (noticed the nfslockd process on your UNIX box?). + + Moreover, the CHAPPEND permission bit, together with the guarantee + that small writes are likely to be serviced atomically by the file + server, can be used together to add new data to a file without race + conditions. + + While in the kernel, processes sleep waiting for an event and wakeup + other processes, as you saw while discussing timers; and of course, + you have the various lock primitives discussed before. + + In this section you are going to read the code for redezvous, sleep, + and wakeup. + +Rendezvous + + See rendezvous(2) before continuing. + + sysrendezvous() rendezvous system call. + + ../port/sysproc.c:696 + sysrendezvous is called with two arguments + sysproc.c:702 + tag is arg[0] and val is arg[1]. + portdat.h:407,411 + A Rgrp (rendezvous group) is a hash table with RENDHASH + entries. + sysproc.c:703 + REND is defined in portat.h:393 to use the hash function to + locate the entry in the table. The author just applies the + modulus to the tag, that seems to be a good enough hash + function for the case. l is our entry in the hash table. + sysproc.c:706 + The list in the hash bucket is searched for a process with the + same tag. You see how Procs doing rendezvous are linked into + the Rgrp hash table using the rendhash field as the ``next'' + pointer. The code is clear, but perhaps names like ``rqnext'', + ``rendnext'', etc. would make it more clear. + sysproc.c:707 + One process called rendezvous with tag! + sysproc.c:708,710 + It is removed from the list. The value the first process gave + to rendezvous will be the value returned to the 2nd process + calling rendezvous for the same tag. The val supplied by this + 2nd process, is passed to rendval for the first process. + sysproc.c:712,713 + Waiting until the first caller of rendezvous stops running (may + be at a different processor). This is busy waiting, because the + author thinks it would be a waste to sleep until the first + process stops, and wakeup later. On uniprocessors, the process + will never be running. + sysproc.c:714,716 + Now that the first caller of rendezvous is not running, we can + mess it up. Remember, the first caller did call rendezvous. If + it was running, it was about to stop, waiting for a second + process calling rendezvous--Agree now that it would be a waste + to sleep? As the first caller is now sleeping waiting for us; + make it ready again. It will pickup rendval as the return + value--It was sleeping also in sysrendezvous. + sysproc.c:722,725 + This is the starting point for the first caller of + sysrendezvous with a given tag. Record the tag and the value so + the 2nd caller notices; and link the process in the Rgrp hash. + sysproc.c:726 + The process calling rendezvous was running; hence it was not in + the scheduler ready queue. Set the state to Rendezvous to + reflect that it is no longer Running; later it will be moved to + Ready as you saw before. + sysproc.c:729,731 + sched yields the processor. Other processes will run. The + current one will not because it is not linked in a ready queue. + When the 2nd caller calls ready for us, this process will + eventually be running, and continue by returning from + sysrendezvous with the 2nd caller's value. + One thing to note: rgrp is unlocked before returning or + scheding. After sched, no lock is necessary to complete the + rendezvous. + + This is a common scheme that you already saw several times: a process + is moved out of the ready queue, it runs and blocks due to some + reason; so it gets linked into the structure representing the reason. + Later, the structure will be scanned and the process moved back to a + ready queue. + +Sleep and wakeup + + sleep() Waits for something. + + proc.c:403,452 + Sleep and wakeup are complicated, as the plenty of comments + suggest. sleep(r,f,arg) puts a process to sleep on r due to a + reason represented by (*f)(arg). wakeup(r) wakes up a process + on r. If wakeup is called, the reason for sleeping no longer + holds. There are several problems though. + + A process may call sleep, and in the mean time, right after + starting to call sleep, another process can call wakeup for + him. + + Right after calling sleep, the condition may change and we + may change our mind and don't sleep. Therefore, wakeup can be + called for a process that is no longer sleeping. + + While a process is sleeping, it can be posted a note with + postnote. That may even make the process die. So once more, + wakeup can be called for a non-sleeping process. + Read this comment at lines :403,452, and think about it. + Perhaps the only way to make things more simple would be to + change the semantics of sleep and wakeup. + By the way, the Rendez parameter of sleep and wakeup may be + called so because it is used to rendezvous the processes + calling sleep and wakeup; but it has nothing to do with + rendezvous(2). + proc.c:458,460 + Without interrupts and locking rlock. + proc.c:466,473 + Important action, see the comment. + proc.c:475,481 + The condition changed while calling sleep; no need to sleep any + more. Note how r->p is set to nil, so wakeup knows there is + nobody sleeping there. + Can you understand now why tfn (proc.c:518,522) is given as an + argument to sleep at line :567? Beware that tfn is not up->tfn! + proc.c:487,488 + No longer Running, Wakeme is the state for sleeping processes. + Besides, the process is sleeping on r. Any postnote will notice + the state and update p->r to be nil; so that a later wakeup + notices that there is no process to wake up. + proc.c:490,506 + Doing the work of sched, instead of calling it. Why? + proc.c:509,513 + A note was posted, report the abortion of the sleep. + + wakeup() Wakes up a sleeping process. + + proc.c:576 + wakeup will wake up the process sleeping on r, if still there. + proc.c:582,585 + Important action, see the comment. + proc.c:589,590 + The sleeper was gone. No locks by now! + proc.c:592,593 + Now getting the lock + proc.c:595 + The process could go away between line :588 and this line. So + check that r->p is still there and is the p we know. Both + checks are necessary since a different process could sleep on r + between both lines. + proc.c:596,601 + The process is awaken and placed back in the ready queue. The + return value is true if a process was awaken. + proc.c:604,607 + One way or another, we are done. + + postnote is discussed together with notes in the next section. + + There is a routine used to put the process to sleep for a while, + awaiting for a resource. + + resrcwait() Waits for some time due to a reason. + + pgrp.c:261,277 + resrcwait uses tsleep to wait for a resource. It sets the + ``ps'' state to let the user know what is the process waiting + for, and sleeps for 300 ms. return0 (a procedure returning + zero) is used so that when sleep checks the condition function, + it returns zero and sleep puts the process to sleep. rescrcwait + seems to be used just to await for free slots in the process + table. + + Notes + + Users make system calls to notify and noted to handle notes. See + notify(2). Notes are posted by writing to a note file or by the + system; see proc(3). + + The desing of Plan 9 notes is nice is that it services several needs: + the kernel can use it to notify of exceptional conditions to the user + process, and the proc(3) files can be used by user code to notify + anything even through the network. Other systems (e.g. UNIX) do not + have a means to asynchronously notifying to processes over the network + (e.g. you must use either sockets or signals for that, and signals do + not work across the network!). + +Posting notes + + sysnotify() Sets up the process handler for notes. + + sysproc.c:572,578 + sysnotify registers the handler for notes in the field notify + for the current process. The address is checked to be valid + because the user could lie--if the address is zero, the handler + is being canceled. + sysproc.c:580,586 + sysnoted simply does nothing. Why? the noted system call has + nothing to do, the work is done in trap.c. + Let's start by posting a note to a process. + + syswrite + procwrite() Handles writes for proc(3). + + devproc.c:720 + A write to /proc/n/note leads to a call to procwrite with Qid + Qnote. Remember the section on files in the previous chapter? + devproc.c:721,724 + It is an error to post a note for a kernel process. It is an + error to post a note message longer than ERRLEN characters. + devproc.c:727 + Here is where the note is posted. postnote does the work. If + you grep for postnote, the kernel calls it in several other + places, where it feels that the system must post a note to + notify something. + devproc.c:677,679 + There is another way for the user to post a note, send it to a + group of processes. If the file written is notepg, pgrpnote + posts the note... notepg()Sends a note to a group of processes. + pgrp.c:12,40 + ...by scanning the whole process table searching for not-dead + processes with the same noteid. In the end, postnote is called + to post notes; kernel processes are not notified. The author + scans the table without locking the processes, and when he + thinks he got a process, a lock is acquired and the check + repeated--now without races. This pattern is used in several + other places as you will see. Any error in postnote is ignored. + + syswrite + ... + postnote() Posts a note to a process. + + proc.c:611 + The note is n, the process notified is p, not the current + process. postnote must lock debug in the Proc affected, but + will do so only if dolock--i.e. some caller of postnote does + not hold the lock. + proc.c:623,624 + flag is NUser if postnote is called by devproc--the user wrote + the note file. nnote is the number of notes posted (but not yet + notified) for the process. So, if the kernel posts the note and + there is no handler or p->notified, the number of notes is set + to zero. notified is true while the process is being notified. + Rationale: if there is no handler, there is no point in keeping + previous notes so set the number to zero; if the process is + being notified, and the note comes from the kernel, forget + about pending notes after then one being notified, because the + kernel one is likely to be important. Only when there is a + handler and a pending note, nnote is preserved so that the + previously posted note is kept for the user before the one + posted by the kernel. + proc.c:626,631 + The note array holds the at most NNOTE notes posted to the + process. If there is space, nnote is adjusted and the note + copied (posted) to the slot in the array. Both msg (the note + text) and flag (the note flag) are kept in the array. postnote + returns true if the note was posted. + proc.c:632 + There is a note for the process, let it know. + proc.c:636,645 + Race against sleep/wakeup. Get the lock and look at p->r. If + non-nil, the process is sleeping. Note the ``paranoid'' check + to ensure that the race did not mess things up. Locking for + these three routines is so complex that security must go first. + The process is pulled out of the sleep and made Ready again. A + later wakeup would notice that r->p is zero and do nothing. + proc.c:649,650 + Unless the process is doing a rendezvous, the post is done. The + process will notice the post and handle things itself. + proc.c:653,664 + Besides sleeping, the other reason a process may be waiting is + on a rendezvous. Both sleep and rendezvous may be interrupted + due to a note. If the state is Rendezvous (note the double + check once the lock is gained!) the value to be returned is set + to the representation of -1 in two's complement (see + rendezvous(2)). The process is then removed from the Rgrp hash + queue and set Ready. Compare this code with the code in + sysproc.c:708,715. In postnote, there is no need to wait until + the process stops running (if it was so). + + Now that the note is posted, and the notified process is ready, it + will run sooner or later. + +Notifying notes + + Imagine the just notified process starts running. + proc.c:475 + If the process was in sleep, or enters sleep, notepending is + seen, and the ongoing system call is aborted at line :512. The + notepending flag is reset (now the process knows it has notes), + but nnote is still non-zero as there are notes in the process' + note array. Something similar happens to processes with notes + posted while doing a rendezvous. + ../pc/trap.c:532,538 + In any case, when the notified process runs again, it will be + inside the kernel, probably in sched or aborting a sleeping + system call. One way or another, as procedures return, the + process will reach trap (or the last lines of syscall). Ignore + lines :532,533 by now. Right now, the process is as depicted in + figure [215]4.8(a). + + CAPTION: Figure 4.8: Notes are handled by the notified process. notify + sets up the context for the handler. The previous state is recorded in + the user stack. + + [The process right before being notified.] + \resizebox{6cm}{!}{\includegraphics{note.eps}} [The process after + notify has setup the user stack for the handler. The next iret will + make the handler run.] \resizebox{6cm}{!}{\includegraphics{note1.eps}} + + If not doing a fork, and the process has notes posted (nnote + not zero), notify is called. The notified process was returning + from an interrupt or a system call when notify gets called. + + trap + notify() Notifies of a note for this process. + + trap.c:546 + notify is the routine actually receiving the posted note. It + will take appropriate actions depending on the note. You should + note how the notified process handles the note itself. That is + more simple than doing it in the notifying process because we + are now running in the notified process context (e.g. user + stack addresses can be used safely). + trap.c:552,555 + Remember the check for procctl in trap? notify is taking care + of ``procctl'' here--I defer the discussion of this until later + in this chapter. If the process was posted a note, we pass + these lines. + trap.c:557,558 + You know that debug is the lock to acquire when posting notes. + trap.c:559 + The note is being handled, no longer pending. + trap.c:560 + First, pick up the first note. If there more ones, they will be + handled after the process has been notified of the current note + (i.e. when the process enters the kernel again and starts to + leave it in trap). + trap.c:561,566 + If the note starts with ``sys:'', the kernel posted the note. + Ensure that there is room to add to the note the user program + counter and add it to the note. If the kernel posted the note, + it is likely that the instruction pointed to by the user PC, + caused a trap that caused the postnote. Therefore, the value + for the PC is valuable to fix the bug. + trap.c:568,574 + If the note was not posted by the user, and there is no handler + (notify is null) or the process is currently handling a note, + the process is killed. pexit does the job. This is reasonable + since the handler is probably faulting and it makes no sense to + give it a second chance. + trap.c:575,579 + The process is handling a note right now, do nothing; The check + could be done at line :537, but notify would then have no + chance to kill the process in case something caused a kernel + posted note. Think of a process using an illegal instruction + while running the handler for the note. + trap.c:581,584 + Default action for user posted notes when there is no handler: + die. + trap.c:585,586 + Starting to setup the user stack to run the note handler (see + figure [216]4.8(b)). We know there is a handler. Here the + author makes room for a copy of the saved Ureg in the user + stack. When the process is being notified, the note handler is + supplied with a copy of the saved Ureg; the handler can make + changes in the copied Ureg and the kernel will reflect those + changes in the real saved Ureg. That way, a user can repair the + cause for a note by changing the noted process context. Can you + see that the kernel is using ``user virtual addresses'' + directly? That is feasible because the kernel now runs within + the context (i.e. address space) of the notified process. User + virtual addresses can still be used from the kernel, they must + be checked though to ensure that they really exist and have + appropriate permissions. + trap.c:588,593 + The user could lie regarding the address of the handler or its + stack. Ensure that both the handler entry point and the place + in the stack where handler arguments are copied are valid + addresses. If they are not, the process dies. The space for the + arguments must hold a copy of Ureg, plus room for an error + message of at most ERRLEN characters, plus 4 machine words. See + later. + trap.c:595,597 + up->ureg is set pointing to the copy of the saved Ureg in the + user stack. This is the copy for the user, and not the real + Ureg (which is the parameter ureg). The real Ureg is copied + into the user stack and a pointer to the just copied Ureg + pushed on the user stack. + trap.c:598 + What? did this before. Perhaps line :595 should be deleted? + Looks like the old up->ureg should be saved in the user stack + before being updated to point to the current User's copy of + Ureg. Note the comment. + trap.c:599,600 + Make room in the stack for the just pushed Ureg* and the note + message; copy the note message. + trap.c:601,604 + Add room for three words: the return program counter, the + pointer the the copied Ureg, and the pointer to the note + message in the user stack; and initialize them accordingly. Why + is the return PC being set to zero? + trap.c:605,606 + The real (note: ureg and not up->ureg) saved user stack pointer + set to the new value for the handler. The real saved user PC + set to the address of the handler (kept in notify). After trap + returns leading to a return from interrupt, the reloaded + process context would make it run the handler as if it had been + called. The arguments are as they should be, but the return PC + for the handler is zero! A return will cause a page fault + (because address zero is not valid) and the process will surely + die--because the system would post a note during the execution + of the handler. But you know that a note handler must not + return, you read noted(2), right? + trap.c:607,610 + One the note is notified, shift remaining notes (if any) to + remove the first one. lastnote keeps a copy of the note just + notified--used for debugging and while returning from the + handler. Besides notified is set to record that the handler is + running. + +Terminating the handler + + Assume now that the process note handler behaves correctly, and it + calls noted(2) before returning from the handler. + + syscall + sysnoted() Does nothing. The system call is just to enter the kernel. + + ../port/sysproc.c:581,586 + When syscall calls sysnoted (the system call for noted(2)) it + does nothing (Although I think that the code in noted discussed + below could be moved to sysnoted). + ../pc/trap.c:532,533 + The system call number is NOTED, and noted is called. The Ureg + given corresponds to the context for the user within the user + library noted function, right before the handler terminates. + The second argument is a pointer to the arguments of noted(2), + which are just an integer (the return PC for noted is in the + top of the user stack). Figure [217]4.9(a) shows how the stacks + look like. + + CAPTION: Figure 4.9: Returning from a note handler and restoring the + user context. + + [The process right after calling noted(2).] + \resizebox{5cm}{!}{\includegraphics{noted0.eps}} [The process after + noted(NCONT) did its job. It will resume where it was before the + note.] \resizebox{5cm}{!}{\includegraphics{noted1.eps}} [The process + after noted(NSAVE) did its job. It will start another handler.] + \resizebox{5cm}{!}{\includegraphics{noted2.eps}} + + syscall + noted() Handles a posted note. + + trap.c:621 + noted must restore the user context as it was before the note + was notified. + trap.c:627,631 + If notified not set, and the argument to noted is not NRSTR, + abort. Read the noted(2) manual page if you have not done so. + notified should be true, because noted is being used to return + from a handler to the previous context, but looks like + noted(NRSTR) can be used even when there seems to be no + handler. + trap.c:632 + The handler is no longer running. See how notified is used to + report that the user context corresponds to a note handler? + trap.c:634,637 + nureg and oureg set to the handler copy of the Ureg saved + before the process was notified. + trap.c:638,642 + The copied Ureg* and Ureg must be valid addresses. Kill the + process otherwise. + trap.c:651,660 + Can we trust that the copied Ureg is reasonable? The user could + try to mess up with segment selectors in the copied Ureg and + the kernel could crash or compromise security if the user was + trusted. The same can happen to bits in the flag word that + enable/disable interrupts and affect system issues; however, + other bits in the flag word can be changed. + trap.c:662 + This is it!, now that we trust the handler Ureg, copy it back + to the currently saved Ureg. After noted returns, the return + from interrupt will reload the (fixed) process context. + Usually, the user PC and SP in the fixed Ureg would be those + corresponding to the user context before the process was + notified (as shown in figure [218]4.9(b)). + trap.c:664 + The argument to noted specifies what to do next. + trap.c:665,674 + Be it NCONT or NRSTR, the pointer to current copy of Ureg for + note handlers is set to its old value. (But remember line :595! + A bug there?). + trap.c:676,690 + NSAVE arranges for the user stack to be almost preserved as it + stands (see figure [219]4.9(c)). The user stack is set above + the old Ureg, leaving place for three parameters and a fake + return PC; a pointer to the oureg is set as the first + parameter. But, parameters for who? If you ask this, did you + read noted(2)?. The note handler calling noted has adjusted the + PC in its copy of the Ureg so that a different routine is + called when it returns; the NSAVE is set for noted to let it + know that it should keep a handler stack frame for the routine. + Using this ``trick'', the user can chain handlers for notes. + trap.c:691,695 + None of the known flags, let the user know and continue as if + NDFLT was said. + trap.c:696,702 + The process did not say NCONT to let the process continue, and + did not say NSAVE to chain another handler. The reason for the + note is not likely to be repaired and the process must die. + When noted returns, the saved Ureg is reloaded (including any + fix from the note handler), and the process resumes operation. + + By the way, most of the blocks with qunlock, pprint, and pexit used to + handle errors could be folded into a common error handling block by + using a goto like other kernel routines do; and perhaps the arithmetic + done with the user stack pointer could be simplified a bit. + + Rfork + + Now it's time to see how are processes created and destroyed. The + system call used to create a process is rfork. Read the rfork(2) + manual page. + + Plan 9 follows UNIX (as with many other things) in that processes are + arranged into a hierarchy. The system creates an initial process, and + remaining ones are created using rfork as descendant of the first one. + + Using a hierarchy of processes is good in that provides a natural + means to share resources, by setting them up in the parent before + spawning any child. Unlike UNIX, Plan 9 is able to adjust the + resources a process has, including its name space, so that any process + can get a brand new instance of the resource, or a clone of the + resource. But you already read this in the manual page, right? + + Although the proc(3) device permits handling of processes using files + (over the network), typical operations of process creation and program + execution are performed by regular system calls on the local node. + Moreover, processes can only share resources (namespace, descriptors, + etc.) within a node. This is not a severe problem, because name spaces + can be constructed to use the same resources (files) over the network. + The approach chosen by the author is simple, yet effective. + + sysrfork() Entry point for rfork. Creates new processes or adjusts the + current process resources. + + ../port/sysproc.c:20 + sysrfork is the entry point for the rfork system call. It is + called from syscall in ../pc/trap.c. + sysproc.c:31 + The flag supplied to rfork is very important, because it + controls what rfork will do. It is made of an OR of bits, + stating that particular resources for the process should be + (re)created, duplicated, or shared. + sysproc.c:32,38 + The user cannot request that file descriptor group be both + copied (RFFDG) and and cleaned (RFCFDG). The same for the name + space and the environment. Flag names are not so hard to + remember: they all start with RF (for RFork). Now, take the + file descriptor group (FDG) as an example: to share it, say + nothing; to duplicate it, say RF and FDG, i.e. DFFDG; to clean + it, put a C before the flag name, i.e. RF and C and FDG, that + is RFCFDG. Calls to error will jump to the label set by the + last call to waserror--at ../pc/trap.c:496. + sysproc.c:40 + Important!, if RFPROC is said, the system must create a new + process and use remaining flags to set up its resources. If + RFPROC is not said, changes affect the current process. The set + of flags for rfork is a kind of micro-language, used both to + adjust resources in the current process and to control the + initialization of resources for the new process. Lines :41,78 + are executed when rfork is adjusting resources for the current + process. + sysproc.c:41,42 + RFMEM requests data and bss segments to be shared between the + parent and the child, but there is no child. RFNOWAIT request + the child to be ``independent'' of the parent--more about this + later. As there is no new process, these flags have no sense. + sysproc.c:43 + the fgrp has to be either copied or cleared. It makes sense to + copy the fgrp even when there is no new process. The fgrp may + be shared among the current and other processes, the current + process is probably going to adjust its fgrp and does not want + to disturb the other processes. By calling rfork with RFFDG + set, the process can ``clone'' the fgrp and get its own copy. + sysproc.c:44,48 + up->fgrp set to either a duplicate of up->fgrp, or to a + duplicate of nil--i.e. to a fresh new one. You already saw in + the last chapter how dupfgrp works when creating a new group. + dupfgrp()Duplicates an Fgrp. + pgrp.c:185,208 + When the fgrp is not being created, but being cloned from an + existing one, these lines execute. Lines :187,190 get the + number of used entries in the cloned group--and round that + number to a multiple of DELTAFD entries. Later, memory for the + array is allocated, the reference count set to one and the + array initialized from the cloned one. incref(c) adds an extra + reference to each channel in the cloned group. The author + ensures that the fgrp appears to grow in chunks of DELTAFD + entries, no matter if that was really the case or not: he + sticks to his design. + closefgrp()Releases a reference to a Fgrp. + sysproc.c:49 + The process got a new fgrp, so the old one is no longer + referenced by the process. closefgrp releases the reference to + the previous up->fgrp; if the reference count gets down to zero + (no other process using this fgrp), all file descriptors in the + fgrp are closed and the fgrp deallocated. + sysproc.c:51,59 + The name space adjusted. pgrp is actually the name space group. + First, a new pgrp is created (an empty mount table for the + process). If the pgrp is to be copied, pgrpcpy duplicates in + up->pgrp the old name space. There is no duppgrp routine + (although a simple wrapper for pgrpcpy could be created to make + the code look like the one for fgrp). The noattach flag + prevents mount and attach from being used on the name space. + The old pgrp value is set for the new name space. Otherwise, a + process could bypass the noattach flag by duplicating the name + space. I think that a duppgrp routine could take care of this + detail too. Finally, the reference to the old name space group + is released. + sysproc.c:60,61 + If RFNOMNT was set, forbid mounts and attachments on the name + space by setting the flag. + sysproc.c:62,66 + Start a new rendezvous group for the process. That is used to + avoid clashes in the rendezvous tag namespace, and to prevent + some processes to rendezvous with others. A new rgrp is created + and the reference to the one dropped. + sysproc.c:67,74 + Environment group adjusted. Again, I miss the existence of a + newegrp and/or dupegrp routine--But that's just a naming issue + mostly. The creation of a new egrp is done by allocating it and + setting the reference counter to one. If the environment is to + be copied, envcpy will recreate in up->egrp the variables found + in the old environment. + sysproc.c:75,76 + Create a new note group for the process. Similar to what was + done for rendezvous, but more simple. Remember that the whole + process list is scanned to determine who belongs to a process + group when posting notes? To create a new note group it + suffices to get a new noteid value. + sysproc.c:80 + The previous lines were executed only when resources for the + current process should be adjusted. Did not return at line :77, + so the caller wants a new process. Allocate it at this line. If + you remember from the previous chapter, newproc allocates a + free Proc entry and initializes it with everything set to null + but for the kernel stack, the process pid and the process + noteid. The process state, the ``ps'' state, and the FPU state + are set to initial values too. + sysproc.c:82,84 + The newly created process has the same FPU saved state, and + appears to be executing the same system call (number and + arguments). + sysproc.c:85 + No errors for the new process. + sysproc.c:86,88 + The new process uses the same root directory and the same + current directory. rfork usually duplicates the calling process + unless told otherwise. Now there is another reference for dot. + An incref on slash seems to be missing. I think that the author + considered it unnecessary because all processes have the same + value for slash (boot gets one pointing to the root device and + rfork makes the child have the parent's slash), the slash + channel will never be released. Nevertheless, I think it would + be better to incref/decref it. + sysproc.c:90,96 + The set of notes for the current process is duplicated for the + new one. dbgreg points to the saved Ureg after traps, none by + now. Also, there is no note handler running for the new + process, set notified to 0. + sysproc.c:98,104 + Going to work with segments, so gain the lock seglock and + prepare to release it if there is an error. waserror is used to + jump back to it on an error, release the lock, and re-raise the + error to the waserror in syscall (trap.c). dupseg()Clones or + shares a segment. + sysproc.c:105,107 + For all segments, call dupseg to duplicate the ith segment in + the seg array. The n lets dupseg know that the segment should + be duplicated and not shared or cleared. I'll get back to + dupseg on when talking about virtual memory on chapter [220]6. + The only things you should know by now is that: + + dupseg returns the segment given incrementing its reference + counter for TEXT, PHYSICAL, and SHARED segments. You see how + text segments (read only) and physical memory segments are + shared between the parent and the child no matter what rfork + is told. + + dupseg creates a fresh new stack segment and returns it, when + the segment given is a stack. The base address and size for + the new stack are the same as in the passed segment. + + for DATA and BSS segments, dupsep will add a new reference + and return the segment given (i.e. share it) or it will + create a copy of the given segment, depending on the share + flag (n). + So, after the calls to dupseg, the new process has its seg + array setup either sharing the parent's segments or with a copy + of parent's segments. Of course, text segments are always + shared with the parent and the child always gets a fresh new + stack. Apart from that, everything else in virtual memory looks + like the parent's memory; see figure [221]4.10. + + CAPTION: Figure 4.10: Virtual memory layout for a forked process. The + layout of physical memory is not really like the one shown; besides, + Plan 9 uses paging, and does not map whole segments. + + \resizebox{14cm}{!}{\includegraphics{forkvm.eps}} + + sysproc.c:108,109 + Lock released and the last error label removed. A call to error + to notify an error will jump now to the waserror at syscall: + there is no cleanup to be done here now and the direct jump to + syscall can be permitted upon errors. + sysproc.c:111,155 + File descriptor group, name space, rendezvous group, and + environment group are either duplicated from the parent for the + new process, or created, or shared. It all is the same that was + done by rfork to adjust resources for the current process when + no RFPROC flag was given (The main difference is that resources + are shared when they are neither cleared nor cloned). Perhaps + some code could be shared and rfork made shorter; nevertheless + the code is simple and easy to follow. + sysproc.c:156 + hang is a flag stating that the process should stop when doing + an exec to give the user a chance to debug it. The child gets + the same flag than the parent. As creating new processes is + usually an rfork plus an exec, it makes sense to propagate the + flag. + sysproc.c:157 + ``permissions'' for the file representing the new process are + the same they were in the parent. + sysproc.c:159,162 + Read the comment! When you do an rfork requesting the creation + of a new process, the parent is given the pid of the child, and + the child appears to return from rfork just like the parent, + but returns zero instead. forkchild sets things up in the child + so that it would appear to be returning from rfork with a + return value of zero. Note that trap (../pc/trap.c:227,230) did + set dbgreg to point to the Ureg saved by the hardware when + rfork was called. + + sysrfork + forkchild() Handcrafts the child kernel stack. + + ../pc/trap.c:772 + forkchild has to be machine dependent because it assumes the + stack layout for the current architecture. + trap.c:777,782 + When the scheduler jumps to the new process, it will jump to + the sched label in p. The author initializes the label so that + the kernel code executing is not sched (which usually sets the + label when the process is leaving the processor), but the first + instruction of forkret. forkret will then return from the rfork + system call in the child as if it had called rfork. The kernel + stack pointer is set to the end of the kernel stack for the new + process, but leaving room for a copy of an Ureg structure and + two extra words. The two extra words are the return PC and the + argument (the Ureg*) of the syscall routine. Yes, for the + syscall routine and not for the forkret routine. More later. + trap.c:784 + cureg points two words after the top of the kernel stack for + the new process, that is where an Ureg is going to be copied. + trap.c:785 + Important!, the ureg passed to forkchild is copied into the + kernel stack for the new process. That Ureg was the one saved + by the hardware when the user called rfork. Therefore, forkret + has its own copy of that processor context. + trap.c:787 + This is where rfork is forced to return zero at the child. The + return value will be taken from the return-value register, + which is ax. ax is set to zero in the Ureg copied for the child + process. The whole picture can be seen in figure [222]4.11. + + CAPTION: Figure 4.11: Kernel stacks and Procs after forkchild. + + \resizebox{14cm}{!}{\includegraphics{fork.eps}} + + trap.c:791 + insyscall is set and reset by syscall upon starting/terminating + a system call. Reset it for the child since it is completing + its ``call to syscall''. + + sysrfork + + sysproc.c:164,165 + The parent of the child is the current process. + sysproc.c:166,172 + If RFNOWAIT was set, the parent will not call wait(2) for the + child, so make the parent pid be the pid for the initial + process. Every process likes to have a parent who cares for it! + That process will wait for the child. More about wait in a + following section. If it was not set, increment the number of + children for the parent. + sysproc.c:173,174 + No request to start a new note group, so keep the parent's. + Remember that newproc did set noteid to be a new group. + sysproc.c:176,181 + Initialize the state of the FPU, zero the time counters, and + record at time[TReal] the starting time for the new process. By + subtracting that value from ticks at processor 0, the system + can know how much (real) time passed since the process was + born. Names for the text file (binary file) and the user named + duplicated. + sysproc.c:183,187 + The comment says it all. The reason is that when a segment gets + shared, permissions on the page table for the memory affected + can change too. Therefore the MMU has to be flushed to drop the + old permissions from cached page table entries. This will + become clear in a following chapter. + sysproc.c:188,189 + Priority (both base and actual) inherited from the parent. + sysproc.c:190 + The child appears to have run at the same processor the parent + was running at. + sysproc.c:191,193 + If the parent is wired to a processor, the child gets wired to + that processor too. procwired wires the process as you saw + before. + sysproc.c:194,195 + All set. The child gets linked into the ready queue and set + Ready. When the scheduler is called, it could elect the child. + sysproc.c:196 + When the current process gets back to the processor after sched + runs other processes, the pid of the child is returned as the + result of the system call. + That was okay for the parent, but what does the child now? + + forkret Appears to return from syscall. + + ../pc/l.s:539 + When the scheduler picks up the child for running, it jumps to + the sched label for the process and forkret starts running. + forkret does exactly what is done after syscall returns from + the call at plan9l.s:43. The only difference is that syscall + was never called by the current process. The stack was set by + forkchild as if syscall was called, so that forkret could + believe in that. + One thing to see here is that forkret is actually assuming that + the process returns from trap and not from syscall; but the + code in forkret and plan9l.s:45,52 is exactly the same, which + means that it would work anyway. Perhaps it would be better to + move the forkret declaration from l.s to plan9l.s:44, since it + is returning via syscall and not via trap. + l.s:540 + throw away the fake Ureg* argument in the stack. + l.s:541,545 + Reload the processor registers and segments from the Ureg saved + in the child kernel stack by forkchild. + l.s:546 + ignore the couple of words in the Ureg above the hardware saved + processor context in the stack. + l.s:547 + Here we go! The iret reloads the processor PC, SP and their + segments so that the process continues back in user-level + returning from the rfork system call. The ax register restored + at line :541 was set to zero by forkchild, therefore, rfork + returns zero to the child. + ../pc/trap.c:535,538 + To complete the discussion of rfork, here is my guess about the + reason for the scallnr!=RFORK in trap. + Suppose that the child was setup by forkchild to start running + in trap and not in forkret--probably by copying the kernel + stack for the new process in this hypothetic previous version + of the kernel source. If the system call was rfork (and it was + called by trap), a new process would be created and both + processes would return from the rfork system call back to trap. + If that would be the case, and the RFORK check was removed, + both the parent and the child would check for pending procctls + and pendingnotes. Perhaps the code used variables in the stack + that could cause the child to be posted a note that was really + for the parent. + Regarding the actual code, the only utility I can see for this + check is to avoid posting a note to the child before giving it + a chance to either install its own note handler or issue an + exec system call and be forbidden for parent's faults. In any + case, the child has its notify field as the parent has it. + + NOTE: Is this correct? Or note handler setup code did not work + properly on forks? Or this has something to do with procctl and I'm + missing that? + + Exec + + Now that you know how a new process is created, let's see how it can + execute a new program. It needs to both locate the program to be + executed and execute it. The separation of concerns between rfork + (creating a process) and exec (executing a program) allows a parent to + perform adjustments on the child process before executing its program. + This comes back from the days of UNIX. + + An executable in Plan 9 is any file with the execute permission set. + Unlike other systems, the file name has nothing to do with the fact + that it could be executed. The file must be either a text file or an + a.out file. A text file to be executed usually starts with ``#!'' and + the path of the program to interpret the file; for example, rc scripts + start with ``#!/bin/rc''. a.out files are generated by the Plan 9 + assembler (see a.out(6)), and contain, among other things, the + following items: + * An Exec header, with information about the image of the program + (sizes for segments, etc.). + * The executable code for the text segment. + * The image of the data segment with initialized variables. + +Locating the program + + sysexec() exec system call. Executes a new program. + + ../port/sysproc.c:209 + A process willing to destroy its memory in favor of executing a + brand new program calls exec; this is the entry point for the + system call. + sysproc.c:226,227 + The first argument is a file name, where the executable for the + new program is to be found. So, check that the address is + valid, and get a pointer for it. Remember that the user virtual + addresses are valid, therefore, the pointer supplied by the + user is ok for kernel usage. + sysproc.c:228 + Ignore this by now. To satisfy your curiosity, indir seems to + mean ``indirection''. + sysproc.c:230,234 + tc is the text channel, or the channel pointing to the text + file for the new program. namec resolves the name in the + current name space and returns an open channel checking for + execute permission. The waserror prepares for closing the + channel and re-raising the error if the following code raises + an error. + sysproc.c:235,236 + namec did set up->elem with the file name--without any previous + path component. So now elem contains the name for the text + file. Again, ignore the indir thing; just notice that it is + zero now. + sysproc.c:238,240 + You know this, right? Using the channel type to call the + appropriate read routine to get the Exec header for the text + file. At least two characters wanted. Yes, the Exec header is + more than two characters, but keep on reading. If the error is + raised, you get back to line :231, the channel is closed and + the error re-raised. + sysproc.c:241,243 + Extracting the magic number from the header, as well as the + size of the text segment and the entry point. The Exec header + is defined at /sys/include/a.out.h:2,12. read could get just + two characters and all these fields could be trash. The numbers + just extracted are stored in big-endian order in the Exec + header, l2be is ``little to big endian'', however, that + transformation on a big endian yields a little endian value; + never mind, the fact is that l2be convert those values to a + little endian representation--shouldn't be this a machine + dependent operation? + sysproc.c:224,250 + If the whole exec header was read and the magic number states + that it is indeed an a.out file, it can be executed. The break + would break the loop used to search for the text file, and + execution would continue at line :275 were a.out binaries are + loaded. The error is raised in case the entry point is set + before the start of the text segment for the user plus the size + of the Exec header, or in case it is beyond the text segment + (plus the size of the Exec header). The image in memory will + contain the Exec header and then the text segment, hence the + range--text images look very much like the file. Besides, the + entry point should be within the user portion of the virtual + address space (not with the KZERO bit set). + sysproc.c:252,254 + Not an a.out. It is a file interpreted by another program. + sysproc.c:255 + The exec header is copied into a character array. + sysproc.c:256,257 + If the line does not start with ``#!'', it is not an script, so + don't know what kind of binary it is. Ignore indir once more. + shargs()Builds arguments for shell scripts. + sysproc.c:258 + shargs takes the line array, the number of characters kept at + line, and builds in progarg an argument vector for the program. + If you read lines :441,469 you will see how that is done. Can + you guess why the loop at :447,449? The number of arguments + filled up is kept in n upon shargs return. + sysproc.c:261 + indir is set when the file is to be interpreted by another + program! + sysproc.c:265,266 + shargs filled up progarg according to what follows #!. Now, + consider an rc script named ``/tmp/f'' starting with + ``#!/bin/rc -e -s'', when the file gets execed, /bin/rc should + run with the command line /bin/rc -e -s /tmp/f. shargs would + have filled up progarg for the command line /bin/rc -e -s, but + it knows nothing about the final missing argument. These two + lines at exec are supplying as the final argument, the name of + the file to be interpreted--note that the argument vector must + be null-terminated. + sysproc.c:267,268 + The first parameter in the argument array supplied as the + second argument to exec is no longer valid, so remove it from + the argument array. + sysproc.c:269,270 + The file being executed is not the script, but its interpreter. + The name of the interpreter is at progarg[0]. Besides, the + interpreter should believe that it is named after the script + name, not after the file containing the interpreter code. The + first parameter for the program executed is set as the script + file name, which was elem. + sysproc.c:271,272 + Now let's get back to indir. You see how the channel for the + script file is closed (and the error label popped because we + already closed the channel). The loop will iterate once more + with file set to the file being execed (the interpreter) and + indir set to one (at line :261). + Should the interpreter file on this new iteration be another + ``#!'' file, the test at line :256 would raise an error. The + author does not want an interpreter to be an interpreted file! + That can appear to be a restriction but it is not--interpreters + are usually binary files, and if they are to be scripts, the + can be easily wrapped with a silly binary file that calls the + script. Should the author allow nested interpreters, a loop + could arise because a malicious (or dumb) user could setup two + files to interpret themselves recursively; e.g. file /a starts + with #!/b and file /b starts with #!/a. It's more simple to + forbid nested interpreters than it is to check for looks in the + nested interpreter call. Besides, allowing nested interpreters + would require more complex code in exec. Despite that, the + author wrote sysexec in a way that makes it easier to allow it + to handle nested interpreters. + If the interpreter is a binary and not an interpreted file, the + check for indir at line :235 preserves elem as the script file + name even though it is the interpreter the one being executed, + and the break at line :249 would lead to the code execing an + a.out file. + +Executing the program + + sysexec + + sysproc.c:275,276 + Starting to execute an a.out, be it an interpreter or not. Now + extract the lengths for the data and bss segments. The lengths + for the text segment and the entry point had to be extracted + before to check for illegal entry points. + sysproc.c:277 + t is set to the end of the text segment. That is the first + address of the segment (UTZERO), plus the sizes for the Exec + header, and text segment proper. The +BY2PG-1 and &~(BY2PG-1) + is rounding the computed value to a page boundary. A page can + be either text or data, but not both. + sysproc.c:278 + d set to the end of the data segment, computed by adding the + size of the data segment to the just computed (and rounded) end + of the text segment. + sysproc.c:279,280 + The end of the BSS (b) computed the same way. + sysproc.c:281,282 + Don't trust the exec header. The end of text, data and bss + segments should not invade the high part of the address space, + used for the kernel. The error would jump to line :231, where + the channel is closed, and the error re-raised. + sysproc.c:287 + nbytes counts how many bytes are to be pushed in the user + stack. You already know that the bottom of the stack is used + for a profiling clock. This ``first pass'' counts the number of + bytes to be pushed on the stack. + sysproc.c:288 + No arguments pushed yet. + sysproc.c:289,296 + If exec-ing with an indirection (i.e. an interpreter), the + argument array is the progarg computed by shargs. Count the + bytes for the null-terminated strings in progarg. + sysproc.c:297 + evenaddr seems to fix the passed parameter to start at an even + address. Some busses would raise an alignment error exception + otherwise; but on the PC, evenaddr does nothing. + sysproc.c:298,307 + Besides any argument counted in the case of an interpreter, the + arguments given as the second parameter to exec have to be + accounted for too--note that for interpreters, the first + parameter (the script name) was removed from the argument + array; that is to avoid counting it twice. The calls to + validaddr are ensuring that both argp and the strings kept + there reside at valid user virtual addresses. The call at line + :299 checks the first word (the first page actually) for argp; + when the page offset for the argp address is less than then + size of the word, argp is jumping into the next page, so call + validaddr once more to verify that the next page is still in + place. Calling validaddr every pass in the loop would be a + waste. The number of bytes to be pushed is incremented with the + length for each argument, as reported by vmemchr (plus one for + the final zero). vmemchr is like memchr, which returns the + pointer to the first occurrence of a character (0) in a string + (a); unlike memchr, vmemchr checks that the memory where the + string resides is valid user virtual memory. By subtracting the + start of the array (a), its length is computed. + sysproc.c:308 + the size of the user stack is now known: One pointer per + argument plus a null terminator for argv; plus the actual size + of the arguments, rounded to a multiple of the word size. + sysproc.c:310,315 + The comment says it all. On Intels it can waste a bit of memory + but who cares. The author is still computing the size of the + stack. + sysproc.c:316 + Count the number of pages needed for the initial stack. + sysproc.c:321,322 + Ensure the the user stack does not get too big. TSTKSIZ is the + maximum allowed size for the ``temporary'' user stack being + setup now. + sysproc.c:324,328 + Going to operate on process's segments, qlock it. Note the use + of a QLock (long waiting, maybe), and the use of waserror to + release the lock in case of errors. + sysproc.c:329 + A new stack created. ESEG is an extra segment slot used for + exec. Right know exec could still fail, and you don't want to + loose your user-level stack yet. This stack segment goes from + TSTKTOP-USTKSIZE to TSTKTOP; noticed it is not USTKTOP? The + author does not want to mess up the current user stack because + exec can still fail, therefore, a temporary stack segment is + created right below the user stack. TSTKTOP (../pc/mem.h:54) is + precisely USTKTOP-USTKSIZE. Even though the stack is not at + TSTKTOP, pointers pushed on it have values assuming that it + starts at USTKTOP. This stack is going to move to its proper + location, but later. By the way, I think that the comment that + says ``putting it in kernel virtual'' is a bit confusing, since + the stack is being built at the user portion of the virtual + address space. + sysproc.c:331,350 + setup the stack arguments for the new process. Arguments are + copied appropriately including progarg when indir is one. You + should understand the code. Remember that the pointers pushed + (i.e. :346) assume that the stack is mapped at USTKTOP, and not + at TSTKTOP. + sysproc.c:352 + elem was kept with either the binary file name (indir not set) + or the script name (indir set); copy it as the name for the + process text. + sysproc.c:354,363 + Old segments ``released''. This is a point of no return. Only + segments between SSEG and BSEG are released. That includes the + current user stack (which is not shareable), text (which is + being replaced by exec), data (which is being replaced by exec) + and BSS (also replaced by exec). putseg decrements the + reference counter for the segment and releases resources + (memory, mostly) held by the segment when the counter gets down + to zero. + sysproc.c:364,370 + From the BSS on, only segments marked as ``close on exec'' are + released. Remaining segments are kept. Shared segments created + by the process would lie between BSS and NSEG; so they would be + kept shared between the parent and the child. fdclose()Closes + file descriptors with a matching flag. + sysproc.c:375,377 + File descriptors marked as ``close con exec'' on the fgrp for + the new process are released. fdclose closes all open file + descriptors which happen to have set the flag passed it. In + this case, all open file descriptors marked as CCEXEC would be + closed. + sysproc.c:379,383 + tc is the channel to the text file, attachimage returns an + Image corresponding to that channel. The thing going on is + caching. The Image structure, discussed later in the virtual + memory chapter, is responsible for caching images of text + files. If someone else is executing the program found at the + file pointed to by tc, the memory used to keep the text loaded + will be shared because the Image used would be the same. Don't + worry too much about this, just think that the Image contains a + segment (img->s) used as a cache for the program text. ts is + kept pointing to the text segment and seg[TSEG] is set + accordingly. + The comment states that the image is ``locked'' when returned. + That is because the author is going to update the segment held + by the Image. The text segment may be shared by different + processes and it would make no sense to acquire the seglock on + one of them to operate on the segment. Instead, the Image must + be locked to work on the text segment. + sysproc.c:384,387 + You will know when virtual memory be discussed. Just to record + that all the text should be ``flushed'' because it is now + shared, and also to know where is the text in the image. + sysproc.c:389,398 + A new data segment created and set in place. The Image for the + data segment (the place where memory comes from) corresponds to + the binary file where the text was found, but starting after + the text. You know that a.out files keep both text and data. + sysproc.c:399,400 + The BSS segment created. The ``zero-fill on demand'' means that + pages will be brought in for the segment as needed, they will + be filled to all zeros when brought. + sysproc.c:402,412 + exec passed the point of no return, so there is no problem to + relocate ESEG into its place, SSEG, which is the proper + location for the user stack. Now that the seglock is released + there is no need to jump back to line :325 on errors. The base + address and top of the stack is set, and relocateseg adjusts + the segment so it starts at USTKTOP, where it belongs. The + movement is done by changing the virtual memory address + translations. + sysproc.c:414,419 + Read the comment, you already know about priorities. The + ``device character'' for the root device is ``/'', so the + kernel is checking that the file comes from the root device. If + that is the case, the priority is adjusted accordingly; + otherwise the process keeps the priority it had (probably + inherited from the parent at rfork). + sysproc.c:420,421 + Remove the error label first, so that if the close for the + channel fails, exec would not close it again. + sysproc.c:428,433 + No notes yet, and FPU state initialized. + sysproc.c:434,435 + If hang was set, honor it by by setting procctl to Proc_stopme, + which means that the process will be stopped for debugging + before returning to user level. + sysproc.c:437 + Finally, execregs initializes the user stack pointer and + program counter. + + sysexec + execregs() Initializes user registers so the program starts in its + entry point. + + ../pc/trap.c:706,719 + execregs starts by setting up sp to the actual top of the user + stack, with ssize bytes on it. Then it pushes the number of + arguments to complete the main entry point arguments. As + syscall did set dbgreg to point to the ureg saved by the + hardware, the only thing to be done is to update on it the user + stack pointer and the program counter. The return value for + sysexec is the address of the profiling clock, which might be + used by the user-level library code, but seems to be not + relevant for the kernel. Remember that exec does not return + when successful, so the return value can be only of interest to + the assembler entry point for Plan 9 processes. + + By the way, in case you didn't guess, the loop at lines + sysproc.c:447,449 is to ensure that the first line of the script fits + within the size of an Exec header. Remember that exec read the header + and then copied it to line? If the line is longer than the size of the + header, exec would miss the trailing part of the line, so better fail. + This is a tradeoff for simplicity, as exec could perfectly keep on + reading until a whole line is read. + + Dead processes + + Processes can terminate existence in several ways. First a process can + call exits to terminate itself (see exits(2)). A message can be passed + to exists, which will be passed by the kernel to the parent process + calling wait(2). Thus, the concept of a process hierarchy also helps + in controlling how processes went in their lives. The parent calls + wait and receives reports about its dead children; every child tells + the parent. + + The message passed is more meaningful for humans than the UNIX error + code, and what is actually more important, is portable to different + architectures! The convention is that a null string means ``ok''. + + Another way to (almost) terminate is by faulting, either voluntarily + (see abort(2)) or involuntarily. Faulted processes are kept hanging + around for debugging in a Broken state. That is better than saving a + core file for several reasons: first, no more core files hanging + around in the file system; second, a broken process is still ``alive'' + and can be inspected for more than just data values, the broke(1) rc + script can be used to locate and terminate broken processes. + + Yet another way is by using the proc(3) device ctl file for the + process. A write of kill to that file, terminates the process. This + way works fine over the network, since the file can be used remotely. + +Exiting and aborting + + /sys/src/libc/386/main9.s:1,7 + The entry point for the user process is usually _main. _main is + an assembly stub that calls the C entry point, main, after + doing some work for the profiling clock and the main arguments. + Remember that the return value of exec was the profile clock? + That value was ``returned'' to the new program. + main9.s:9,13 + If main ever returns without calling exits, exits would still + be called with the ``main'' string as the argument. + + sysexits() Terminates the process reporting a reason. + + /sys/src/9/port/sysproc.c:502 + One way or another, sysexits is the entry point called by the + process terminating. + sysproc.c:505 + In case the user error string (the first argument) does not + look fine, this is the string reported. + sysproc.c:509,522 + If no status string was supplied, that is ok. If it was + supplied, copy it to the kernel buffer buf. validaddr and + vmemchr are used to be sure that addresses are valid. If + addresses are not valid, an error is raised and status is set + to inval. + sysproc.c:523,424 + pexit kills the process; the return is to make the compiler + happy--all system calls should return a value. + proc.c:732 + The 1 as a second argument asked pexit to release the process + memory; it is really being killed. + + sysexits + pexit() Terminates the process. + + proc.c:745 + By setting alarm to zero, any alarm is canceled. The Proc may + still be linked into the alarm list. This is not a problem + because if a new process reuses the Proc entry, and it does not + use alarms before expiration of a previous alarm for this Proc, + alarmkproc will find its alarm set to zero and ignore it. If + the new process ever sets an alarm, it will be first removed + from the alarm list. + proc.c:747,759 + All resources cleared while the lock was held. Releasing them + may take some time, so do not hold the lock for more than + needed. + proc.c:761,770 + Now resources are released by calling routines that decrement + their reference counters; if a reference counter gets down to + zero, the resource is released--perhaps causing other + decrements in reference counters for structures used by the + resource; e.g. the fgrp uses channels that are cclosed when the + fgrp goes away. + proc.c:776 + Kernel processes are always there, and the author does not do + housekeeping for them; but user processes have parents and + there is a relationship to be maintained. + proc.c:777,782 + All processes have a parent. You will see what happens to a + process when its parent is not there. Hint: read the panic + message! + proc.c:784,788 + A waserror but in a while loop. Remember that waserror returns + false when first called to set the error label, and then it + returns true when an error jumps back to the waserror? The + effect of the while is to call waserror again when an error + happens--i.e. to restore the error label popped by error. That + means that the process keeps on trying to smalloc a Waitq + structure, no matter what errors happen. But, smalloc provides + guaranteed allocation. What error could be raised by smalloc? + smalloc calls tsleep to wait for free memory if the pool is + exhausted, tsleep calls sleep, and sleep raises an Eintr if the + process is interrupted. So, no matter how many interrupts the + process gets, exits will not be aborted returning to the + process with an Eintr error; exits does not return, ever! + proc.c:790 + readnum prints the number given into the buffer passed + (shouldn't it be ``printnum''?) It returns the number of + characters used to print the number, or zero if it did not fit. + The buffer is wq->w.pid, and the number id up->pid. So, the pid + field in the message for the parent kept (in the Waitq + allocated) is being filled up with the ascii representation of + the pid. + proc.c:791,798 + These calls to readnum are filling up the the wait message for + the parent with times as said in wait(2). TK2MS converts ticks + to milliseconds and the time entry at Treal is used to know for + how long the process had lived. Saw how the message can be + understood at any architecture? Guess why? + proc.c:799,805 + If a non-null (and not empty!) error string was supplied by the + process, it is copied into the error message in the wait + message--note how the message is prefixed with the name for the + text file and the process pid. That is very important when the + parent process is a shell, like rc, to let any human user know + who did die. It is also important if the parent cares about who + died. + proc.c:807,830 + If the parent's pid (p->pid) does not match parentpid, the + parent is not ``associated to the child'' (see rfork) and does + not care about the wait message from this child; if the parent + is broken it does not care either; and if more than 128 wait + messages are queued for the parent, the author thinks that the + parent does not care either. Daemon processes that fork a child + per request, but ``forget'' to call wait would be able to leave + an indeterminate number of wait records behind them but the + 128-check enforces a 128 limit. This is yet another detail + where you can see how the author tries to protect the system + against buggy processes. + To pass the wait message for the caring parent, just queue it + in the parent's waitq (queue of wait records). If the parent + cares, the number of child processes and wait records is + adjusted. The wakeup, awakes the parent in case he is sleeping + waiting for a wait record. By the way, remember that nchild was + incremented in rfork only if RFNOWAIT was not set? + proc.c:833,834 + User processes bookkeeping is complete. This code is executed + for both kernel and user processes. If the memory should not be + released, the caller wants the process to hang around in a + broken state. + + ... + pexit + addbroken() Keep the process in a Broken state. + + proc.c:670,696 + addbroken moves the process to an array of broken processes, + and changes the state to Broken. By calling the scheduler, + addbroken will not return unless the process is set again + ready; that happens when the process is really terminated (e.g. + by a write of kill to its ctl file). Should this happen, + addbroken returns and the remaining code at pexit would + terminate the process. It is nice how the code to terminate the + process is shared in this way for both processes exiting and + processes aborting. When NBROKEN broken processes exist, the + first who broke is terminated to make room for the new broken + process. One thing to note is that too many broken processes + are a waste because they would probably never be debugged. + Perhaps for CPU server kernels it would be better to keep a + broken structure per user using the CPU server, but the author + thinks this suffices. Another thing to note is that the broken + process is terminated just by placing it in the ready + queue--when it runs, pexit will terminate the process. Just + simple. + + ... + pexit + + proc.c:836,844 + Segments released. The process' mind is going. Only when the + last reference to each segment is gone, it is released. Do you + think that these lines would destroy the stack segment? And the + text segment? + Although you are not expected to answer this before the chapter + on virtual memory, what happens if there is an ongoing page + fault on one of the segments released? How can the page fault + handler ensure that the segment will not go away under its + feet? + proc.c:846,849 + By setting pid to zero, no child will leave a wait record + because of the test at line :815. The wakeup is not for this + parent process (which is dying and not in pwait), it would + awake any process waiting in devproc.c:561 for a child to die. + proc.c:851,854 + Now that nobody is linking more wait records, release all wait + records queued--the parent could terminate without calling + wait. + proc.c:856,862 + Awake any debugger waiting for us (e.g. waiting for a note to + be posted for us) and dissociate from the debugger. + proc.c:864,866 + After a couple of locks are taken, the state is set to Moribund + and sched is called. sched will never return because the + process is really destroyed and will not get back to the ready + queue. If a bug makes sched return, panic. Why does the author + acquire these locks here? + proc.c:64,79 + sched calls gotolabel for m->sched, which leads to code in + schedinit. This time, this branch is taken and the process + state is set to Dead. After releasing MMU data structures for + the current process (using the prototype page table for the + current processor afterwards), the process is linked into the + free process list. Releasing MMU data structures and linking + the process in the free queue, requires both palloc and + procalloc locks to be held. However, right now in schedinit, + which one is the current process? There is no process. What if + lock couldn't acquire the lock at the first attempt? What if it + even called sched? That is why the locks are acquired while + there the dying process is still alive enough for requesting a + tas lock. + The kernel stack is kept bound to the Proc (and reused by the + next process using that Proc). After the locks are released, + any other processor could pickup the Proc and its kernel stack + could be reused. This is no problem since the current stack is + the ``scheduler stack'' kept near Mach. Using a scheduler stack + allows the author to step back out of the dying process while + killing it. + proc.c:80,83 + The current process is gone, sched called and it will call + runproc to run another (existing) process. + + In case you didn't notice, processes aborting, generate a fault that + (as you will see) end up calling pexit with an indication not to + release the process memory. + + By the way, it would be nice not to release the process resources + (fgrp et al.) for broken processes (at least when explicitly + requesting so), so that the process could be debugged even looking at + the set of open file descriptors; and perhaps it would be feasible to + fix things up a bit and let the process get ready again. One simple + way to do so would be to move the process into the Broken state before + line :747, and to add control operations to fix up the process state + and set it back to ready. + + I think it's time now for you to look at figure [223]4.12 and see how + a process changes its state. Probably you did draw a scheduling + diagram while you learned how processes are born, get ready, etc. + Compare yours with the one in the figure. For the sake of simplicity, + I have not shown the Scheding state, which is used while the processes + is changing its scheduling state in several places (e.g. from Ready to + Running and from Dead to Ready). + + CAPTION: Figure 4.12: (Simplified) process state transition diagram. + Do not take it verbatim: the Scheding state is missing, Broken + processes appear to go right to Moribund, without passing through + Ready. + + \resizebox{10cm}{!}{\includegraphics{sched.eps}} + +Waiting for children + + syswait() Waits for dead children. + + sysproc.c:528 + syswait is the entry point for wait(2). After checking that the + Waitmsg pointed to by the first argument resides at valid user + addresses, pwait does the work. + + syswait + pwait() Wait for dead children. + + proc.c:883 + pwait receives the Waitmsg to be filled up. + proc.c:888,889 + If the wait queue is being manipulated (a child dying right now + at a different processor?) just give up. Why? seems to be for + avoiding deadlocks between devproc and proc. In any case, the + parent is likely keep on calling wait until he gets the desired + wait record. + + NOTE: is this true? + proc.c:891,894 + Now the lock held. + proc.c:896,901 + If there is no child (dead or alive), abort. + proc.c:903 + sleep until haswaitq; haswaitq returns true if the wait queue + is not nil. Therefore, if there are children in the wait queue, + the process will not even sleep. If all children are alive, the + process sleeps until it gets awaken by a dying process, or by a + note. + proc.c:905,909 + Got a dead child, remove a wait queue entry. + proc.c:911,912 + Nothing else to do, release the lock. + proc.c:914,918 + Return extracted information by copying it to the parameter + passed (if it was not nil), and returning the dead child pid. + + The proc device + + Although process creation is done only with system calls, processes + are represented as files for most other purposes. Read the proc(3) + manual page to learn what files are serviced by the proc driver. Proc + can be mounted over the network to operate on remote processes as if + they were local; not the UNIX's /proc, definitely. Files under + /proc/n/ correspond to views of running processes; e.g. two successive + cats for /proc/n/status would return different file contents, because + the file contents are the status of the process, and the status + changes over time. Reads under /proc/n are used to inspect processes, + and writes under /proc/n are used to change various things on running + processes. + + In the next chapter, you will learn more about file systems in Plan 9, + and will be able to understand better devproc.c, that is where the + proc device implements its file system. However, I think you can + understand most of devproc now. I am going to discuss the code related + to inventing a file hierarchy from the set of processes; but skipping + code related just to the file system, which will be clear after the + chapter on file systems. + +Overview + + devproc.c:10,28 + Several Qid types defined. + devproc.c:54,67 + Here you see how actual Qids are built from the Qid types + defined above and the process numbers. Why does the author do + this? + devproc.c:31,49 + This data structure is a template for the directory serviced + for each process. You can see the file names, the Qid for the + file, the file length and the file mode. + devproc.c:757,776 + This is the Dev structure linked into devtab. The `p' is the + letter name for the driver, and most fields are pointers to + routines used when a channel type corresponds to devproc. + Routines with names starting with dev are default + implementations for channel operations in dev.c. Many of the + routines supplied by proc, relay on generic implementations + that use the procgen routine to iterate through a proc + directory. Let's see some of the routines now. + + open... + procopen() Opens a proc file. + + devproc.c:161 + procopen is the routine used when a file under /proc is being + opened. The file to be open is represented by the c channel. In + Plan 9, opening a file means to check permissions and prepare + the file for I/O. For example, after you walk to + /proc/3/notepg, you have to open it before writing it. + devproc.c:168,169 + The CHDIR bit is set in Qid, therefore, a directory is being + opened; rely on the generic routine for that. + devproc.c:171,176 + The process slot is kept in the Qid; give the slot to proctab + and obtain the Proc for the c channel. Remember that the file + being opened is not a real file, but some aspect of a process. + procopen is locating the process and locking it so that the + process could be inspected without race conditions. + By keeping both the slot and the kind of file in the Qid, the + author knows quickly what kind of processing (and on which + process) should be done given the Qid for the file. + devproc.c:177,179 + The process could have died since the channel was obtained (by + a walk) and the open was requested. If the process died (even + if its Proc was reused by a different process), the PID in Proc + will not match the PID in the Qid. + devproc.c:181 + openmode checks omode for invalid bits. + devproc.c:183 + Each file under /proc/n has a type encoded in the Qid (see + lines :10,28). + devproc.c:184,191 + The text file for the process is the one being opened + (/proc/n/text). Only opening for reading is allowed, since the + text is being used for executing the process. proctext is the + routine doing the job. proctext()Gets the channel for the + process text. + devproc.c:785,797 + Check that the process text is still there and get a reference + to the Image for the process text segment. + devproc.c:805,807 + The image contains a channel to be used for accessing its file. + devproc.c:809,812 + Increment the reference counter for the channel (someone opened + it), and check that channel is still opened for reading. When + the process is still there, the code is not assuming that the + image is there; and when the image is there, the code does not + assume that the channel is set up for reading. Why? Hint: + processes are living things. + devproc.c:820 + Now got a channel for reading the file for the process text + segment image. It has been incref'ed, and will be the channel + resulting from procopen. + devproc.c:193,200 + For these files (/proc/n/proc,etc.) nothing has to be done but + to check that the open is for reading--processing continues + after the switch. + devproc.c:202,210 + Nothing done; can be opened for read or for write. + devproc.c:212,216 + For /proc/n/ns, mode has to be read and temporary storage for + walking the mount table is allocated. + devproc.c:218,226 + For /proc/n/notepg, only writes are allowed, and not for the + first process group (boot). The id of the process group and the + noteid for the process are kept as a Qid in the channel. + devproc.c:228,231 + Defense against bugs; no other Qid types known. + devproc.c:233,246 + After checking that the process is still there, the generic + devopen routine is called. devopen uses procgen to iterate + through the proc/n/ directory searching for a file matching the + channel supplied by the user (i.e. the file being opened, like, + /proc/n/ns). Once found, devopen checks permissions and either + raises an error or returns the channel. + + wstat... + procwstat() Updates attributes of a proc file. + + devproc.c:250,256 + procwstat is used to modify file attributes, including + permissions. No wstat is permitted on directories. + devproc.c:268,269 + Only the user who started the process, and eve can change + attributes in proc files. + devproc.c:271 + convM2D converts the machine independent representation of file + attributes (given by the caller) into a Dir structure, more + amenable for processing. The file could be wstat'ed from a + different machine with a different architecture. + devproc.c:272,279 + If the user in the Dir structure (to be written) is not the + owner of the process, a chown is being done. Only eve is + allowed to do such thing. + devproc.c:280 + Honor a chmod in the file. + + close... + procclose() No more I/O on a proc file. + + devproc.c:337,342 + The temporary storage allocated in procopen is released when + the file is closed. + +Reading under /proc + + read... + procread() Reads from a proc file. + + devproc.c:363,364 + procread services reads under /proc files. It corresponds to a + call read(f,va,n) with the file offset set to off. + devproc.c:378,379 + Use the generic routine for reading directory entries. + devproc.c:381,383 + The process could have died since the open was done. + devproc.c:386,414 + /proc/n/mem represents the process memory. Reading is achieved + by doing a memmove to copy the memory being read into the + buffer transferred to the caller of read. The mem file + represents virtual memory: offset 0 is virtual address 0. Not + all addresses are valid. + devproc.c:387,389 + Addresses before KZERO or within the user stack are read + with procctlmemio, which checks that addresses are valid + and lie within process' segments. Perhaps some time ago + the user stack was within the kernel portion of the + address space; right now, the first part of the ``or'' at + :387,388 is true whenever the second part is true. + devproc.c:391,401 + Addresses between KZERO and end correspond to kernel + addresses and are read by a direct memmove without further + checks. + devproc.c:402,413 + Remaining addresses correspond to memory found at the two + memory banks in conf--also read with memmove by the grace + of the direct map between physical and virtual memory. + Remember that addresses in conf were updated to be kernel + virtual addresses--although early when booting they were + physical addresses instead. + If you trace the kernel execution after the open of + /proc/n/mem, you will see how permissions are checked using the + mode kept in the Proc structure for the process. The author is + permissive in allowing any user with permissions to read kernel + memory (even though he protects the memory used to keep user + keys, all memory allocated from xalloc can be read). However, + this permissiveness is good to make it easier to debug and + inspect the kernel state. + devproc.c:415,427 + Profiling is not discussed now, but see proc(3). + devproc.c:428,451 + Copy to the user buffer the text for the first note posted for + the process, and decrement the number of notes. You can see how + posted notes can be read/canceled by reading this file. + devproc.c:452,458 + The Proc structure is read. Useful to debug the kernel: No need + to put more prints nor to attach a debugger just to see a value + in Proc, just read it. + devproc.c:460,483 + Useful!, dbgreg pointed to the saved Ureg while the process was + switched out. A read here returns the user context. The code + below regread: simply copies the memory read from the Ureg + pointed by rptr. The kregs file corresponds to the kernel + context. When the process in on its way to be switched out, + there is an Ureg saved when last entering the kernel which + holds the user register set; but then the process is really + being switched out, there is a label set by the scheduler + before jumping to other process' kernel label: setkernur sets + in the ``kernel Ureg'' the PC and SP saved in the process + scheduling label. There is no kernel Ureg, although users are + told so. Remaining registers in the ``kernel Ureg'' are + reported as zero. + devproc.c:485,518 + The status file is invented to contain the name for the text + file (:495), the owner of the process (:496) and the process + state (as kept in psstate). If pssate is not set, the process + state name is obtained by translating the process scheduling + state state to a printable representation. Besides, various + times and the size for the process are reported as said in the + proc(3) manual page. To compute the size, the lengths (top + minus base) for the various segments are accounted for. NAMELEN + and NUMSIZE are used to pad the various pieces of status at + fixed positions in the ``file''. That can simplify a lot the + code to read a particular field of the status file. + devproc.c:520,539 + segment contains a printable representation of the segments for + the process. The segment array is iterated to obtain segment + names, types, and boundaries. + devproc.c:541,575 + The contents of the wait file are the next wait message from a + died process. The code uses the waitq as you saw before for + wait(2). The read on /proc/n/wait would block until a child + dies. If you see line :554, the read would fail if there is no + children and the read of /proc/n/wait is being done by the + process /proc/n/. That is, the parent can use read to block + waiting for a dead child; other (unrelated) process can read + this file to cause a ``wait'' for the child. Perhaps the wait + system call could be removed in favor of reading /proc/n/wait. + devproc.c:577,611 + Not to be discussed now, but the code synthesizes the text + corresponding to commands to reproduce the name space for the + process. That text can be fed to a shell to recreate the name + space even at a different machine. + devproc.c:613,616 + noteid is simple. procfds is generating a text representation + of open file descriptors. + + read... + procread + procfds() Reads a proc file descriptors file. + + devproc.c:286,335 + The Fgrp for the process is iterated and for each open file + descriptor, the channel is inspected to obtain the open mode, + device type, Qid, file offset, etc. + +Writing under /proc + + write... + procwrite() Writes a proc file. + + devproc.c:661 + procwrite is analogous to procread, but does a write instead. + The write is for file referenced by c, and corresponds to a + write(f,va,n) when file offset is off. + devproc.c:669,670 + No writes on directories. + devproc.c:677,680 + A write to notepg would post a note to the process group. You + saw how pgrpnote did that. Remember that procopen set pgrpid in + c to contain the noteid? When the user opened the file, the + notepg file represented a concrete note group. Should the + process change its note group in the mean time, the file still + points to the old note group. Besides, the process is not + checked for death before pgrpnote is called. So, imagine you + want to kill all processes in a process group by posting a note + to the group. Imagine that you open the notepg file, and then + the process starts dying voluntarily; by using the saved + noteid, the file would still cause a note post to remaining + processes in the note group. + devproc.c:691,697 + A write to mem is used to modify process memory. A debugger can + use this file to update variables in the debugged process. The + process should be stopped though--because it would be + unpredictable what could happen if memory could be updated + while the process is running. procctlmemio is used to operate + on process memory, as happened in procread. If you look at the + last parameter it was 1 in procread and it is 0 in procwrite; + it is deciding what to do: read or write. + devproc.c:698,706 + A write to regs updates the saved registers for the process. A + debugger can use this to update the process PC, SP, etc. + (dbgreg points to the saved Ureg for the process). + setregisters()Updates the process Ureg. + ../pc/trap.c:734,747 + setregisters copies the supplied registers into the Ureg for + the process. Both flags and code and stack segments are ensured + to be valid ones. Otherwise the user could cause a system crash + or break system security. + ../port/devproc.c:708,714 + The same for FPU registers. In this case, the user can update + all of the FPU context and no machine dependent routine is + needed to ensure that a valid state remains. If the user is + writing a wrong state, he would just harm himself. + devproc.c:716,718 + A write to ctl can be used to perform control operations on the + process. + + write... + procwrite + procctlreq() Writes a procctl request. + + devproc.c:873,881 + procctlreq does the job after copying the request string to a + kernel buffer. + devproc.c:883,884 + A write of ``stop'' to ctl leads to a call to procstopwait with + the last parameter set to Proc_stopme. procstopwait()Wais for a + process to stop. + devproc.c:828,835 + procstopwait attaches the current process (up) as the debugger + for the process whose ctl file is being written. To setup a + debugger for process p, the pdbg field of p is set pointing to + the debugger process, and procctl in p is set to be the process + control operation. The author wrote things so that the control + operation is known to p, and it will honor it if needed. If the + pdbg field of the process' Proc was set, there was already a + debugger and the call fails. If the process was already + stopped, the write of ctl results in a non-operation. + The write of ctl can be done through the network, and in that + case, the debugger would be running at a different machine. So, + who is the debugger process? The write request would have been + sent through the network, but there is a (file system server) + process in the node of p doing the actual write to the ctl + file. That process would be setup as the debugger in the Proc + structure. For the kernel, it does not matter if that is the + real debugger or a remote delegate for the debugger. + devproc.c:838 + The (debugger) process psstate is set to Stopwait. sleep will + make the process wait. The p process state can still be Running + or Ready, so the scheduler can elect it for running. When the + to-be-debugged process runs again because the scheduler elects + it, it will reach soon either the end of trap or the end of + syscall in trap.c. + + trap + notify + procctl() Checks for process control operations. + + ../pc/trap.c:310,313 + If it is trap the first one to notice, it would call notify + when noticing that procctl is set. + trap.c:535,538 + syscall would do the same. + trap.c:552,553 + notify checks for notes, but it calls proctl too-when it sees + that a control request is pending. + ../port/proc.c:1096,1098 + If the control operation is to terminate the process because it + consumed too much physical memory, do so. The process is + killing itself upon request. + proc.c:1100,1102 + If the process is being killed, do so. + proc.c:1104,1107 + This is for tracing processes, you can ignore it now; although + you can see how it does the same of Proc_stopme when the + process has pending notes. + proc.c:1109,1126 + The process stops itself voluntarily. The ``ps'' state is + updated to reflect that the process already stopped. The + scheduling state (p->state) is set to Stopped. The call to the + scheduler switches to a different process and the debugged one + will not run again until it is set Ready (by the debugger). The + local state is used to resume the process in the state it was + before being stopped, because it could be a different thing + each time the process is stopped. Before discussing the wakeup + call, note that interrupts were disabled since notify was + called. That makes sense since it messes up the user stack and + besides it can stop the process via procctl. + proc.c:1116,1119 + If there was a debugger waiting for the process to stop, wake + it up. The debugger expects the write to /proc/n/ctl not to + return before the process is stopped. + devproc.c:844 + So the debugger sleeps until the process is stopped. If the + wakeup runs before the sleep, the procstopped function will + notice and the debugger will not even sleep. Otherwise the + debugger sleeps until the process stops itself and notifies the + debugger. + proc.c:1118 + One more thing, pdbg is set to nil when the operation is done. + pdbg is used to let p know who is its debugger (so it could + awake it, etc.), but as soon as p does not need to know who is + its debugger, the ``connection'' is reset. pdbg acts as a lock + in that if a debugger is already (waiting for) stopping the + process, no other debugger would be allowed to do so. Once the + process is stopped, a different debugger process can operate on + the process. + devproc.c:847,848 + The debugged process could die due to a note post or a control + operation. + + write... + procwrite + procctlreq + + devproc.c:885,900 + Back to procctlreq, a write to ctl with a ``kill'' string would + kill the process. Should the process be broken (did fault), + unbreak sets it ready again. unbreak()Terminates a broken + process. + proc.c:699,713 + It does so by scanning the broken process array and setting as + Ready the one passed as a parameter--the array is updated to + reflect the deallocated entry. + Although it may look silly to scan the array instead of using + p, remember that at most NBROKEN processes are kept broken. In + general, processes must be either running, linked into a ready + queue, or linked into the data structure that prevents the + process from being ready. In this case, broken does the job. + devproc.c:891,895 + Should the process be stopped, it is killed by the system. The + Proc_exitme control operation will be handled later by procctl. + The process is set back into Ready state so it could run and + kill itself. + devproc.c:896,899 + In any other state, the process will either be Ready, or get + back to Ready if it was sleeping or doing a rendezvous. So just + post the note and setup the control operation. + devproc.c:901,906 + hang requests that the process stops when doing an exec. Just + update the flag accordingly. It is honored by exec, which uses + the Proc_stopme control operation to stop the process doing the + exec. + devproc.c:908,909 + A write of ``waitstop'' uses procstopwait again to stop the + process. Unlike the previous usage, ctl is now zero, which + means that no control operation is posted (:833,834). So, the + writer of ctl would sleep until the process is stopped, but it + does not stop the process: it waits until the process stops. + For example, a debugger process may write ``hang'' to ctl and + then waitstop, to wait until the debugged process does an exec. + devproc.c:911,917 + A write of ``startstop'' resumes an stopped process (sets it + ready) and then waits until the process stops. Although + procstopwait would set procctl to Proc_traceme, it is set by + hand before allowing the process to run; otherwise the process + could be free running before honoring the Proc_traceme + operation. + proc.c:1104,1107 + The Proc_traceme operation is handled by the started process + like a stop one, but only when the started process has a note + posted. For instance, a debugger may set a breakpoint and let + the process run until it reaches the breakpoint or faults. + Whenever that happens, procctl would not return at line :1106 + because there are posted notes, and the process will stop after + awaking the debugger process. + devproc.c:919,923 + Just set ready an stopped process. + devproc.c:925,936 + procctlfgrp closes all open file descriptors. I don't know what + this control operation is really for, but it could be useful if + other control operations allowed file redirection to be done by + means of proc(3). + devproc.c:938,949 + A write of ``pri N'' to ctl would set a new base priority for + the process. Only eve is allowed to raise priority this way--it + is her machine, isn't it? Can you see the difference between + the ``root'' user on UNIX and ``eve'' in Plan 9? + devproc.c:951,956 + To wire a process to a processor number. You already saw how + procwired does it. + devproc.c:958,968 + A write of ``profile'' to ctl is clearing the profile + information, which hangs from the text segment profile field. + +A system call? A file operation? Or what? + + Let's look briefly at how the user C library implements some services + using the system calls and system services that you now know. I hope + you will be reading more of that library as you learn how the kernel + works. + + abort() Aborts execution. + + /sys/src/libc/9sys/abort.c + Just crosses a null pointer. The process gets a page fault and + enters the Broken state. + + fork() Creates a new process. + + fork.c + Just calls rfork asking for a new process, with a copy of the + file descriptor and rendezvous groups. + + postnote() Posts a note. + + postnote.c + Writes to /proc/n/note or to /proc/n/notepg. + + getenv() Get the value of an environment variable. + + getenv.c + Just read the file ``/env/name''. + + getpid() Gets the process it. + + getpid.c + Just read the file ``#c/pid''--see cons(3); more on the next + chapter. + + Can you see how even most of the user utility functions are actually + using file operations? + + Files + + File systems are central to Plan 9. Remember that the key point is + that everything is a file and files can be accessed over the network. + In section [224]3.11, ``Files and Channels'', you already learned a + bit about files and channels. You should reread that section if you + forgot it and then continue with this chapter. + + In this chapter you are going to read the implementation of the + various system calls related to files, including bind, chdir, close, + seek, dup, open, read, create, fd2path, remove, and wstat. Besides, as + an example of a file system, you are going to revisit kernel devices + (e.g. pipe), looking at the generic routines provided in dev.c. + Finally, the device translating calls to file procedures into RPCs, + mnt, is also discussed in this chapter. + + During this chapter, you will be reading these files: + * Files at /sys/src/9/port: + + sysfile.c + File system calls. + + chan.c + Channels. + + cleanname.c + Name cleanup. + + portdat.h + Portable data structures. + + pgrp.c + Name spaces. + + dev.c + Generic device routines. + + devpipe.c + Pipe device driver. + + devmnt.c + Mount driver (remote files). + + cache.c + Caching remote files. + + qio.c + Queue based I/O. + + ...and several other ones used as examples. + + Files for users + + Users operate on files using the system calls provided at sysfile.c. + As you already know, such system calls are serviced by using channel + operations. Let's describe now how such system calls work, without + looking too much into the channels. I hope you get a better image of + what is going on after seeing how your system calls are translated + into channel operations. See open(2), dup(2), and close(2) manual + pages. + + sysopen() Entry point for the open system call. Prepares a file for + I/O. + + /sys/src/9/port/sysfile.c:221 + Users operate on files by obtaining file descriptors using the + open (also create) system call. sysopen is the entry point for + open. + openmode()Checks the open mode for a file. + sysfile.c:226 + openmode is called with the second argument. It does a cleanup + of the omode parameter for open and returns a clean omode. In + this case, the cleaned mode (the return value) is not being + used. The author does this call to let openmode raise an error + in case the open mode is not valid. + sysfile.c:227,231 + Cleanup the channel for the file being opened on errors. + sysfile.c:232,233 + namec opens the file and returns a channel for the file name + given. It resolves the name in the current name space; namec is + discussed below. + sysfile.c:234,236 + A new file descriptor allocated to the new channel. + sysfile.c:237,238 + The descriptor points to the channel used. open is done. + + sysopen + newfd() Installs a new file descriptor for the given channel. + + sysfile.c:46,65 + The Fgrp contains an array of pointers to channels. File + descriptors are simply indexes into this array. For example, if + file descriptor 3 is open, fd[3] would point to the channel for + the file. After locking fgrp, the array of pointers to channels + is searched for a nil entry. The first entry unused (:54) is + selected. Lines :56,59 allocate new entries in case all entries + in the array are being used. maxfd records the end of the used + part of the array, and the new allocated file descriptor is set + to point to c at line :62. + + sysopen + newfd + growfd() Resizes the file descriptor set. + + sysfile.c:13,43 + When the descriptor array is exhausted, growfd resizes it, + DELTAFD new entries at a time. The routine does nothing if the + array is already big enough to hold the descriptor desired + (fd); it can be used just to check that f is big enough (nfd is + the number of entries, used or not, in the array). The array + will never contain more than 100 entries going from 0 to 99 . A + good reason to limit the number of file descriptors to 100 is + that nobody could allocate a big amount of file descriptors to + exhaust kernel memory. Another good reason is that this limit + can convince users to close unused file descriptors, which + would also release resources on the file servers involved. + Lines :27,31 are double checking that the number of allocated + descriptors is kept under a reasonable value, for the same + reason. Note also the use of malloc, and not smalloc. If no + more memory is available, the open will fail, and the process + will not sleep waiting for memory. + + sysdup() dup(2) entry point. Duplicates a file descriptor. + + sysfile.c:180 + sysdup is used to duplicate a file descriptor. After dup, two + file descriptors point to the same channel. + sysfile.c:189,190 + fdtochan returns a reference to the channel given the file + descriptor. c holds the channel for the file descriptor being + duplicated, and fd holds the file descriptor number where the + user wants to place the duplicate. + sysfile.c:191,205 + The user specified where to place the duplicate. growfd ensures + that the entry for the descriptor exists in the array. maxfd + has to be updated in case the new fd is the biggest one. + Finally, the channel is linked into the descriptor entry. The + dance with oc is to close the previous channel in case the + ``duplicate descriptor'' was already opened. The channel c has + now another reference (at fd). + Can you find out where is the incref for the channel? + + ... + fdtochan() Gets the Chan for a file descriptor. + + sysfile.c:68 + fdtochan is used wherever the kernel wants the channel given a + file descriptor specified by the user. + sysfile.c:77,80 + Has the file descriptor a channel? nfd is checked before fd[fd] + to be sure that the entry exists. + sysfile.c:81,82 + If the caller of fdtochan said iref, a new reference is added. + This is where the reference was added for the duplicated + channel in sysdup. + The reference is added while f is locked, so that the channel + does not disappear even if someone else is releasing the Fgrp. + sysfile.c:85,89 + Ignore this by now. + sysfile.c:91,104 + If the caller gave a valid mode to specify that the channel is + going to be used according to mode, check permissions. If mode + is -1, or if the channel mode allows everything, no check is + done (the kernel can pass -1 to fdtochan to request the channel + no matter what is going to be used for). Otherwise, the checks + ensure that a read only channel is not used to truncate a file + and that whatever be mode (read or write), the channel has the + bits on. + + syscreate() create(2) entry point. Creates a file and opens it. + + sysfile.c:735,753 + create creates a new file and returns an open file descriptor + for it. It is similar to open, but note how Acreate (and not + Aopen) is given to namec; and how the last argument of namec + specifies the permissions for the new file. Besides resolving + the path given by the user, namec would create the file and + return a new channel for it. Probably it would be good to use a + single system call for both open and create--even though users + could still have to different routines to prevent accidents. + + sysremove() remove(2) entry point. Removes a file. + + sysfile.c:755,776 + remove removes a file (be it a file or a directory). It uses + namec to get a channel for the file, and then it calls the + device specific remove function, which removes the file. + Removing a directory does not require that the directory be + empty, it depends on what the file server implementing the + directory wants to do (e.g. see upas/fs in mail(1)). + + sysclose() close(2) entry point. Closes a file descriptor. + + sysfile.c:271,277 + Users close their file descriptors using close(2). sysclose is + the entry point for that. fdtochan is called, but the channel + returned is not used. If the file descriptor is not valid, + fdtochan will raise an error and sysclose would be done. + Otherwise, fdclose is called to close the open file descriptor. + + sysclose + fdclose() Closes a file descriptor with the matching flag. + + sysfile.c:242 + fdclose closes fd. + sysfile.c:248,254 + First, the channel (c) for the file descriptor is obtained. + After fdtochan and before line :248 another process could have + closed the file descriptor, so c has to be checked to be still + there. The call to fdtochan not only checked that the file + descriptor was open, it also checked that the entry in the + array was allocated, so the author can index on it safely--the + array can grow but it does not shrink. + sysfile.c:255,260 + The flag given to fdclose is checked against the channel flags. + If the flag is not set in the channel, the routine + returns--without closing the descriptor! + That is useful to close only those descriptors that have a flag + set. Saw how the author writes routines that can be used in + more than one way? Did you read ``The Practice of + Programming''? + sysfile.c:261,267 + The first line closes the descriptor, the last line drops the + reference to the channel (which is actually ``closed'' if this + was the only reference for it). Besides, if the descriptor + closed is the biggest one, maxfd is updated to mark the end of + the used part in the file descriptor array. + + But for namec and cclose you already know how file descriptors are + added and removed from the process Fgrp. You also know how are new + Fgrps allocated, duplicated, and deleted when processes are created + and deleted. So, what remains for you to understand what the user sees + of the file system (names and descriptors) is to discuss name spaces. + + Name spaces + + In Plan 9, every process has a name space. Well, every process group + has a name space. Usually, processes sharing a ``session'' share their + name space. For instance, every rio window starts usually with a new + name space, which is a copy of the name space for the process that + started the window. + + The name space is simply a mapping from names to files (actually, from + names to channels to files) that can be adjusted to alter the set of + existing mappings. Name spaces are implemented in + /sys/src/9/port/pgrp.c (because each process group has its own name + space) with tight cooperation from the implementation of channels in + chan.c. To understand name spaces, it is crucial to remember that + every Proc has a Pgrp (name space), as well as a slash channel (root + directory) and a dot channel (current directory). To understand the + implementation of name spaces, I think it is better to see how are + names resolved; then you will understand better the code used to + customize the name spaces. + + As you will see through this section, name spaces can be customized by + using mount(2) and bind(2) to add new entries. Figure [225]5.1 shows + an example of this. You should read the paper ``The Use of Name Spaces + in Plan 9'' [[226]12] from volume 2 of the manual if you are confused. + + CAPTION: Figure 5.1: Name spaces for a couple of processes. Note how + they differ. Each name space has in /bin appropriate binaries for the + architecture used. Besides, the file used for the ethernet device + seems to come from a different machine. + +Path resolution + + namec() Get the channel for a given file name. + + chan.c:630 + namec translates a name into a channel, using the current name + space. It receives an access mode that shows what will be done + with the file (e.g. create it, open it, etc.) as well as an + open mode that specifies whether the caller is going to open + the file for reading, writing, etc. Perm is used to initialize + permissions for files being created. + chan.c:641,642 + namec is called with paths that come from user code; don't + trust them and check that they are at least a non-empty string. + If the caller of namec ``consumed'' all the path, there is no + such file. + chan.c:644,651 + The (virtual) address is not bigger than KZERO, which means + that it is an user-supplied path (not a kernel supplied one). + So, verify that the memory from name to the end of the string + is valid; note how BY2PG is used to do only one check per page + (perhaps this could be embedded into vmemchr?). + chan.c:653,658 + In cname, the author keeps the name for the file pointed to by + channel; release it on errors. You will understand soon the + reason for keeping names on channels. Can you guess it now? + Hint: consider how symbolic links mix with cd on UNIX. + chan.c:660,663 + Names are resolved differently depending on the first character + of the path supplied. The switch is just selecting the starting + point for resolving the name. + Names starting with / are resolved within the pgrp, but + starting at slash; names starting with # are resolved within + the kernel driver name space (they are kind of absolute names + for kernel devices); everything else is resolved starting at + dot. Once the starting point is selected, a name can be + resolved by iteration, resolving one path component at a time. + Lines :660,726 are setting up things so that lines :728,732 + could iterate to resolve the entire path. But how are things + set up? + Callers of namec usually want to know what is the name (without + any previous path; just the file name) for the file just + resolved; e.g. when executing a new program, up->text is set + with the file name (ls) for the file (/bin/ls) being executed. + The author sets elem pointing to the elem field of the current + Proc. As each path component is resolved, elem is updated to + contain the path component. So, the caller of namec can later + use up->elem to recover the file name. + The author sets mntok to reflect whether a mount operation is + allowed in the file name being resolved or not; by default, it + is allowed. More on that later. + isdot is set whenever the path being resolved corresponds to + the current directory. By default, assume it does not. + chan.c:664,665 + Resolving an absolute path (starts with ``/''). newcname + creates a channel name from a string; so cname is set to keep + the given path. + newcname()Creates a channel name. + chan.c:108,122 + A Cname is used to share names among channels with the same + name. It is reference counted. The actual memory allocated is + CNAMESLOP (20) bytes more than needed to hold the path. That + seems to be to permit changes in the name that increase the + path slightly. The ncname counter is used to keep track of how + many channel names there are; If the author sees that the + number of names is close to the number of channels, there is no + point on sharing channel names. The path length is kept within + Cname. Although it could be recomputed by calling strlen, the + author prefers to call strlen just once, and reuse the computed + length whenever it is needed. + chan.c:666 + So, what follows the ``/''? + skipslash()Advances a name past the prefix slash (if any). + chan.c:862,872 + skipslash returns a pointer past the /. The path could be + //xxx, and skipslash would skip all the adjacent slashes. Users + seldom write //, but programs do; just define a variable v1 to + be /a/b/c/, and then add a relative path v2 of the form x/y: + you get /a/b/c//x/y. The system should cope with that. Lines + :867,870 remove any ``.'' component, so that file servers do + not see unnecessary ``.'' names. The system is replacing paths + like ``/./a'',``./a'', and ``/a/.'', with ``/a'',``a'', and + ``/a'' respectively. By understanding ``.'' here, this code + does not need to be duplicated at every file server in the + system, and what is better, the meaning of ``.'' can be kept + consistent across file servers. + I lied a bit, file servers can still see ``.'' as a path. Keep + on reading. + chan.c:667 + Iteration to resolve an absolute path should start at slash, so + get a clone of the slash channel for the current process. A + clone is needed because the iteration used to resolve the path + will ``move'' (actually, walk) the channel to point to each + file/directory along the path name. If the author used + up->slash to walk, the root directory for the process would + change. + After this line, cname is the entire absolute path, name points + to the first component name (after the slash, once any dot has + been removed), and c is a channel pointing to the root + directory for the process. Besides, mntok and isdot are set + appropriately. Although elem has not been set, nextelem will + take care of that later. + chan.c:669 + The name is referring to a kernel device path (e.g. ``#SsdC0'' + to specify the disk sdC0 from the sd device). Kernel device + paths allow you to use devices even though you may have an + empty name space (Remember the implementation of the getpid + function in the C library?) + chan.c:670,679 + As with absolute paths, the channel name is the supplied name. + Mounts were allowed for absolute paths, but not for kernel + device paths. The author wants kernel paths to remain the same, + so mntok is set to zero. Besides, elem is set to contain both + the initial ``#'' and whatever follows until the next slash. + The n<2 check would copy the slash in``#/'', which is the name + for the root device's root directory. So, elem is set to + contain the file name. + chan.c:680,697 + When you use a kernel device path, you are actually + ``attaching'' to (mounting) the file tree serviced by the + kernel device to your name space. Once it is attached, you can + resolve path components within the device's file tree. + Lines :692,695 are extracting the first ``rune'' (e.g. + character for Spanish, but something different for a Japanese) + and looking if it is an ``M'' (utfrune is an strchr for runes). + If it is an ``M'', the path is ``#M'', which corresponds to a + mount driver path (see mnt(3)). The mount driver is the one + issuing RPCs for remote files when you perform a file operation + on them. If users could attach to mount driver paths, they + could bypass file permissions. Imagine that a server checks + credentials from a client when the server file system is + attached to the client name space. Once the client has been + allowed access, another process could try to ``borrow'' some + files serviced by the mount driver on behalf of the previous + process. By denying attaches to mount driver path names, the + only way to get files from the mount driver is by attaching to + (and authenticating with) a remote file server. + Lines :696,697 just check that the Pgrp does not have the + noattach attribute set, which would prevent attachments. This + flag can be used to prevent a process from acquiring more files + than found on its name space. If you don't trust a program too + much, you can build a name space where the program cannot hurt + anybody else, and set the noattach flag. These lines forbid + attachments when noattach is set and the device is any of + ``#|'', ``#d'', ``#e'', ``#c'', or ``#p''; see the comment to + learn which devices they are. The comment says that it is okay + to allow attachments on these devices (i.e. if r is contained + in ``|decp'', noattach should be ignored). However, that would + need a not ('!') before utfrune. I think that the comment is + right, and the if condition is missing a ``not''. + chan.c:698,700 + r holds the first character after the '#'. That character + identifies the kernel device. devno returns an index into + devtab for the given device character. The 1, is to tell devno + that it is the user the one specifying a device name; it is + okay if the user is mistaken: the device may not exist. If a 0 + was given, devno would panic if the device is missing--that + would be a kernel bug. + chan.c:702 + The initial channel to resolve a kernel device path is the + channel obtained by ``attaching'' to the device file tree. That + channel points to the root file for the device. c is setup + properly now. The channel does not need to be cloned because it + is a brand new channel. + chan.c:703 + Once the initial name is processed (the device name), advance + to the next one; as the author did after processing the ``/'' + for absolute paths. + chan.c:705,707 + Must be a relative path. So, the name for the channel is the + name for the current directory followed by the relative path + supplied by the user. up->dot->name->s is the C string in the + Cname for the dot directory of the current process. Once cname + has a Cname built from the current directory name, addelem + adds, as a suffix, the relative path. + addelem()Adds a suffix to a name. + chan.c:137,148 + If the name is shared between several channels, make a new copy + so that adding a suffix to the name will not change the name of + other channels using the shared copy. For instance, when a + channel is cloned, the name is shared among the clones; if a + clone changes its name (e.g. because of a walk), other clones + should be kept untouched. + chan.c:150,158 + More space is allocated to hold the suffix (s) and the small + extra space. + chan.c:159,160 + The new name is prefix+/+suffix, unless prefix was really + ``prefix/'' or suffix was ``/suffix''. + chan.c:708,711 + The starting point to resolve the path is dot, get a clone of + the dot channel in c--the clone is needed for the same reasons + it was needed for absolute paths. The call to skipslash would + simplify things like ./x and .//x down to x. If name is the + empty string after simplifying the path, name referred to the + current directory, so the author sets isdot. + When isdot is set, there are no further elems in the name; this + case is handled as a special case by the code following. + chan.c:714,717 + Starting to walk on the channel, close it on failures. + chan.c:719 + nextelem obtains the next ``component'' name in the remaining + path name, placing it in elem and advancing name past the + element; the advanced name is returned, hence the assignment. + nextelem()Get the next element from the name. + chan.c:903,926 + nextelem is called after skipslash, so there is a bug if the + first character is a slash. If there are no more slashes, the + whole name is the next component name. The loop copies the + component name to elem (up->elem), one rune at a time. isfrog + contains characters that cannot appear as a component name. + Looks like Rob Pike decided to try allowing blanks in component + names. After the element name has been obtained, skipslash does + more ``dot'' and ``slash'' cleanup and advances to the next + component name. + For paths like ``/'' and ``.'', nextelem would set elem to be + the empty string and it would advance name up to the end. + chan.c:721,726 + mount is discussed later. + chan.c:728,732 + The heart of path name lookup. For each component, walk updates + the channel to refer to the next directory/file in the path. + nextelem advances the name for the next component (and cleans + it up), and walk updates the channel (receives a pointer to + it). + You now see that because the channel moves down the path, slash + and dot had to be cloned. + When the loop finishes, the name has been resolved, but for the + last component! In /x/y/z, c would then correspond to /x/y, and + elem would be z; name would be an empty string after nextelem + extracted z to elem. + It is important to stop before the last component because it + could be that it is a file to be created, and it would make no + sense to walk to it. When the path is ``.'' or ``/'', the whole + path was resolved when looking up the initial directory for the + iteration. In this special case, the loop does not execute + because name was already exhausted. + chan.c:734 + How to resolve the last component depends on what is the + channel for. That is why amode is used. + chan.c:735,742 + Aaccess means that the file is checked (for existence, + attributes, etc.) The author walks to the final component. For + the ``.'' special case, no walk is done, since the whole path + was already traversed (ignore domount). For the ``/'' special + case, isdot is not set and walk is called; however, elem is the + empty string and walk would notice and return without doing + anything. + chan.c:744,753 + The name refers to a directory the user is trying to get into. + The author walks to it, and checks that it is actually a + directory. For the special case, nothing is done but to check + the directory flag. + chan.c:755,761 + The file is being opened. Directories can be opened too. + Forgetting about domount, in the normal case, just walk to the + file being opened; in the special case, no walk is really done + (not called, or it does nothing). + The author used an else instead of the break at line :738 + because more things have to be done by continuing after the + Open label. + chan.c:762,778 + If you remember sysopen, a channel was obtained for the given + path, but where was the file opened? Here is where the file is + opened. c points to the file to be opened. So, c->type is used + to obtain from devtab a pointer to the open routine--how to + prepare a file for I/O depends on who is implementing the file. + The ``close on exec'' bit is removed from the open mode + supplied to namec; the file server does not care about it. open + may return a different channel than the one for the file being + opened. That is useful, and allows the driver to ``generate + files'' when other files are opened, like when you open a clone + file. In that case, c already contains its own Cname and the + cname created before is unnecessary; the author sets newname to + reflect that. + Bits to flag the channel as close on exec, and remove on close, + are kept in the channel flags (remember that channels with + CCEXEC are closed on exec?). + Guess what is saveregisters doing? Hints: the compiler saves + registers when doing a procedure call (so that the called + procedure could use registers at will); if an error is raised + and you get back to line :714, which c are you closing? + chan.c:780,788 + Mount is not discussed yet, but note how in the special case + nothing is done and, in any case, c would be pointing to the + file where the Amount is to be performed. + chan.c:790,792 + The file is being created; remember that in the normal case,c + points to the directory where the file is being created at. If + the file name was ``.'', forbid creation. (See below for the + ``/'' special case). + chan.c:794,800 + Get a clone for c; read the comment. Clwalk does both a clone + and a walk for the channel; it is not used from the kernel (see + clwalk(5)). By getting a clone before walking, there are no + race conditions regarding clwalk--otherwise one of the + walk/clwalk could fail because both of them arrived to the file + server. + nameok()Name contains valid characters. + chan.c:805 + nameok checks that only valid characters are in the file name + (using the isfrog array). The 0 means that ``/'' is not ok. + nameok is also used to verify that a whole path is valid, in + which case ``/'' would be considered to be a valid character. + chan.c:806,812 + If the walk succeeds, the (cloned) channel points to the file + to be created; and the file already exists! c points to the + directory containing the file and is no longer needed. By + setting OTRUNC and going to Open, the file is opened with + truncation to zero bytes--achieving the same effect of create. + For the ``/'' path, walk would succeed and ``/'' would be + opened with truncation, which would typically lead to an + (Eperm) error. + chan.c:813,814 + The walk failed (did not raise an error, but failed because of + its return code). That means that the file does not exist. The + cloned channel is closed, and c will be used to create the + file. + chan.c:819,820 + Forget this by now. It is just ensuring c is the appropriate + directory for creating a file; assume it is and createdir is + not executed. + chan.c:822,832 + When syscreate is called, the caller expects that the file + would be opened and empty after syscreate returns. The + algorithm would be simply ``if the file does not exist, create + it; else truncate it''. However, things can change during the + algorithm. Just in case file creation fails, assume that it + could be because another process created the file (after the + failure of the previous walk and before the create operation + was executed). That can happen more easily here than on a + centralized system because the latency of file system + operations while going through the network is not to be + underestimated--see figure [227]5.2. + + CAPTION: Figure 5.2: Open/Create races. Not so probable... + + Should the create fail, try once more to walk to the file. If + the file was indeed created, the walk in the error handling + code will succeed and namec behaves as if the previous walk + succeeded. If this second walk fails, it makes no sense to try + over and over to ``create it if walk fails'' and ``walk to it + if create fails''. A real fix could be done by folding the open + and create operations into a single one, so that the file + server could hold a lock for the file while deciding whether to + create the file or truncate it. + chan.c:833 + Here is where the actual create is done. c is the directory + where the file is being created and elem holds the file + name--special cases were dealt with before. The final argument + perm is used now to establish initial permissions for the file. + chan.c:834,839 + Close on exec and remove on close noted in the channel. + chan.c:847,852 + newname was set to true before starting to resolve the path. It + is only set to false after Open, when the channel returned by + open was not the one supplied. Only when open returns an + already constructed channel, the name in the channel is + ``old''. Otherwise, the name for the channel has been built + from the user supplied path, and perhaps from the current + directory name. Thus, most of the times, the channel name is a + new one built by namec. Should the channel name be new, + cleancname does some cleanup on it, and it is set as the new + name for the channel--after releasing the previous name in the + channel, if any. Should the channel name be an ``old'' one, the + new cname just built is not used and has to be released. + By the way, how does cleancname work? + cleancname()Cleans a channel name. + chan.c:604,623 + cleancname calls cleanname for both device paths and other + paths. However, for kernel device paths, only the portion of + the part after the slash (e.g. ``/a/b'' in ``#S/a/b'') is + handed over to cleanname. As cleanname may leave a final slash + for directories, it has to be removed (:619) for all kernel + device names but for the root driver, whose name is #/. + cleanname()Cleans a file name. + /sys/src/libc/port/cleanname.c:9,52 + The code looks more complex than it is. Although you should try + to implement the algorithm yourself if you think the code could + be easily written in a more compact way. It does several + things: removes duplicated slashes, removes any ``.'' (but for + the case when the path is just ``.''), and simplifies ``b/..'' + by removing both. The ``..'' in ``/..'' is removed because the + parent of the root directory is the root directory by + convention. The path may contain ``..'' when no simplification + can be done (as in ``../x''). + Try to understand the code yourself after reading carefully the + comment. If you get lost, exercise the algorithm with several + paths. + By the way, the author put much effort into name cleanup at + several places, as you saw. Although that can be worth just to + keep cleaner names, there are good reasons for this effort, as + you will see later. + +Adjusting the name space + + Name spaces are adjusted (on a per-process basis) by using bind(2) and + mount(2) system calls. Read their manual pages now. + + Both system calls are essentially the same thing: They modify the name + space so that a path will lead to a different file tree. The + difference between them come from where is the ``different tree'' + located. For bind, it is a portion of the file tree the process sees, + whereas for mount it is a file tree serviced by a different process. + In what follows, unless I explicitly tell otherwise, I use the term + ``mount'' to refer both to ``mount'' and to ``bind''. Besides, I use + the term ``mounted file'' to refer to either a file or a directory + which is either bound or mounted; I use the term ``mount point'' to + refer to either a file or a directory where either a file or a + directory has been bound or mounted. Got confused? read bind(2) and + reread this paragraph. + + The name space structure is mostly kept under the Pgrp structure for + the process. + + When a directory is being mounted (or bound) onto an existing + directory, it is feasible to keep both the previous and new contents + in place. That is, by mounting a new directory onto an existing one, + you can ``add'' the contents of the new one to the existing one. That + is done often with /bin, where new ``binaries'' are added by mounting + other directories like /386/bin, and /usr/$user/bin onto /bin. When a + file is looked up later in /bin, it can be found at any of the mounted + directories. A directory where several other directories are mounted + is called a union in Plan 9. + + When a new directory is mounted onto a union, the user requesting the + operation can specify where to place the new directory within the + union. That is important because to lookup a file on a union, each + directory mounted in the union is searched in order until the file is + found. Flags like MAFTER/MBEFORE request that the new directory be + added after/before the previous ones in the union. For example, + depending on the flag, /bin/cat may be either /386/bin/cat or + /usr/$user/bin/cat. Another useful flag is MREPL, which dictates that + the previous directories be omitted from the union so that the new one + replaces previous contents. + + Now, consider a file being created on an union. On which one of the + mounted directories is the file created? In the example, is it created + in /bin, in /386/bin, or in /usr/$user/bin? The author added an + MCREATE flag that can be given to any mounted directory. Any new file + is created in the first directory mounted at the union that has the + MCREATE flag set. + + You now have an overall picture of how mounts work. But before reading + how mount and bind work, it is better to see what is the final effect + of a mount/bind. To do so, let's look at what does namec to resolve a + path taking into account mount points. + + Walking mount points + + namec() Get the channel for a given file name. + + chan.c:630 + namec resolved a name into a channel by walking a file + hierarchy. Now you know that, at some point, a new file tree + may be bound to the file tree, and namec should walk through + the mounted tree. + chan.c:725,726 + Before this point, the initial channel for the iteration has + been set to be c and elem contains the name for the first + component to resolve. Now, unless the path is for a kernel + device (!mntok), the path corresponds to the current directory, + or the path is not being resolved to mount the root directory + (elem is empty), domount is called. The channel to start the + iteration is the one returned by domount, which may be c or may + be not. + + namec + domount() Process mount points for a channel. + + chan.c:392 + domount is just ``doing'' the effect of a mount on a file. It + receives a channel for a file, c, and takes care of what + channel should be used instead when something has been mounted + on c. The processing for a mount point is done at a channel, + and not at a file. That is reasonable if you think that mount + is a client operation, and not a file server operation. The + client kernel sees the files through channels, therefore the + processing to honor a mount point can be done by looking into + the channel of interest. + chan.c:398,399 + First the name space (the process group Pgrp) for the current + process is locked for read. This is important since domount is + called often for dealing with possible mounts at a channel. + Since processes walking through the file hierarchy are not + modifying it, they can ``read'' the information for the + channels traversed at the same time; a read lock prevents any + change while there are readers, but allows multiple readers at + the same time. + chan.c:400,403 + A channel has an mh field, that points to a ``mount head'' or + Mhead. An Mhead (portdat.h:232,239) contains information about + what is mounted upon the channel: from is a reference to the + channel used as a mount point, mount is a reference to whatever + is mounted there. The information about what is mounted in the + channel is actually found in the list starting at mount; from + is used just to recover the mount point given the Mhead. + At these lines (:400,403), any previous reference to an Mhead + for the channel is released (putmhead only drops the reference + to the Mhead). More later. + chan.c:405 + domount iterates over a set of MHeads. + portdat.h:394,405 + Apart from the noattach flag, and a read/write lock, you see an + array of MNTHASH Mhead hash entries. This is the real mount + table; see figure [228]5.3. Each ``process group'' has a + namespace with mount entries dictating how is the namespace + modified for the Pgrp. The MOUNTH macro applies a simple hash + function to a channel and returns the hash bucket for it on the + mount hash table. The hash function uses the qid.path to hash + the channel. + + CAPTION: Figure 5.3: Mount structures. Here you can see how two + directories have been mounted over /bin, with the MBEFORE flag. + + The loop iterates on the hash slot for the channel, which has a + list of Mheads for channels with the same hash value. The list + is built using the hash field of Mhead. + The mount information is kept at Pgrp, in the hash table, and + not at the mh field of Chan; mh is just a ``cache'' for the + mount information--so that the table does not need to be + scanned all the times. + chan.c:406,407 + Different channels may have Mheads (because they were used as + mount points), and different channels may have the same + mount-hash (MOUNTH) value. Therefore, not all channels in the + hash list point to c's mount point--if any. After acquiring a + read lock on the Mhead, c is compared with the from field of + the Mhead. Now, the check is done using eqchan, and not ==. + Why? + eqchan()Are both channels referring to the same file? + chan.c:212,223 + eqchan is needed because two different Chans can refer to the + same file. Regarding the kernel, if two channels point to the + same file, they can be considered to be the same--for mount + purposes and other things. + Remember that two channels are a reference to the same file if + their Qids, device type, and device number are the same. In our + case, pathonly is true, which means that vers in the Qid is + ignored. Remember, the author is locating Mheads for the c + channel, he doesn't care if the mount point is modified or not + since the time when m->from was set as a mount point (Mhead) in + the mount table. No matter if c is physically the same channel + kept in the table, if its file is used as a mount point, it is + recognized in the table. + chan.c:407 + If the channel in the Mhead does not point to the file c points + to, it is a different mount point; ignore it. If the channel + does not correspond to a mount point, it will not be in the + hash list, and the routine returns the original c channel. In + this case, namec would continue using the initial c channel. + chan.c:408,413 + The channel is the same, so c is a mount point. cclone gets a + clone of m->mount->to, where m was the Mhead for c. Let's see + what this means. + portdat.h:220,230 + Mheads have a mount field pointing to Mount structures. The + Mount (list) represents file(s) mounted at the channel whose + Mhead has the Mount linked at. Several fields are + self-describing: next points to following Mount'ed trees at the + same mount mount; head points to the Mhead, so that both the + mount point (head->from) and the list of mounted trees + (head->mount) are accessible; and to points to the channel for + the mounted tree. See figure [229]5.3 if you got confused. + So, if you mount /usr/nemo/bin onto /bin, there would be + channel for /bin, with an Mhead entry in your Pgrp. That Mhead + would have from pointing to the channel for /bin, and mount + pointing to a list of Mount structures. That list would have a + Mount structure with to being a channel to /usr/nemo/bin. + chan.c:413 + nc is now a clone of the channel for the file mounted at the + file pointed by c. It points to the same file to points to. A + clone channel is needed because the returned channel is going + to be walked upon return from domount; the author does not want + m->mount->to to walk, it should be kept pointing to the mounted + tree. If you read nc as ``new c'', it is clear what the code is + doing. + chan.c:414,415 + As it was done with c, if nc has an Mhead cached at mh, release + it. + chan.c:416,418 + So, what was mh in Chan? It is set to the mount header for a + channel that is used as a mounted file tree. The usual picture + is to get a channel c, then do a domount for it, and later use + its mh field to operate on the mounted tree. + It is important to see that since nc->mh points to the MHead in + the list, successive mounted files can be quickly found by + following next pointers in the Mount list. You should remember + that nc represents not just the first mounted file; it also + provides access to any other mounted file in the same mount + point (by means of mh). + The xmh field is set as mh when domount crosses a mount point, + therefore it is the ``last mount point crossed''; But it seems + to be unused. Either a previous version of the kernel used xmh, + or it is there for debugging purposes. The last line adds a + reference--because of mh. + chan.c:419,421 + Channels keep the file name used to create them, as you know. + The name for nc is not the name of the mounted tree, but the + name of the mount-point--which was the name given by the user + to namec. The Cname is now shared. + chan.c:422,426 + c is released, its name will stay because of the added + reference. From now on, the caller of domount should use the + channel to the mounted tree instead of the channel to the mount + point. + + namec + + chan.c:725,732 + Back to namec, c is now the channel to the mounted file--or the + original channel if nothing was mounted on its name. At lines + :728,732, walk would call domount for every path component + resolved if mntok is set to true. That means that during path + traversal, what you read about doing a domount for the initial + path component, is also done for each remaining path components + (because each one could also be a mount point). + In the first two lines, if the path is ``.'' , isdot is set and + domount does not execute. That means that for the current + directory, c is the original channel and not the one for any + file mounted on it. Besides, if the path is ``/'' (elem is + empty) and it is being accessed for mounting, domount does not + execute either and c is also the original channel. Why? keep on + reading... + chan.c:736,739, :756,757 + The domount for ``.'' is done if the access is for Aopen or for + Aaccess (e.g., if you open ``.'', you would be opening the + mounted file, in case there is one) However, should the access + be for Atodir or Amount, the original channel is used (i.e. the + mount point is not traversed). By default, the author does not + execute domount at line :726 and then executes it later for the + two cases where it should be done. Why shouldn't it be done for + Atodir and Amount? + chan.c:745,753 + All Atodir (cd) accesses got domount executed at line + :726, but for an Atodir into ``.''. Thus, for Atodir + accesses, ``.'' resolves always to the mount point for + ``.'', and not to the mounted file--if any. + Suppose you ``cd'' into a directory. ``.'' is resolved and + up->dot is set to the mount point for that directory. + Suppose you later use (not for ``cd'') a relative path, + either ``./something'' or ``.'', domount will be called + (for a clone of up->dot) to resolve ``.''--it is called + either by walk or by explicit calls at lines :736,739 and + :756,757. This means that if you mount something at ``.'', + any future relative path will traverse the mount entry + just added for ``.'', even though up->dot was set + (resolved) before the mount was done. That is the meaning + of the comment at lines :745,748; if there were a domount + here for isdot, you would resolve the mount when cd'ing + into ``.'', any ulterior mount would not be processed by + your relative path resolution, because up->dot already + crossed the mount point. + chan.c:780,788 + Regarding the other case when domount is not called for + ``.'', it also affects ``/'' paths. When namec resolves + ``/'' or ``.'' to mount something on it, domount is not + called. When the path has more elements, domount is not + called for the final path element (walk called domount for + all but for the last element). So, in few words, domount + is never called for the last path element when it is being + resolved to mount something on it. Why? + That is because the new Mount is going to be added to an + Mhead for the original channel. When you resolve a path to + mount something on it, the channel from namec is the + original channel for the path, and not the channel for + anything previously mounted on it. For example, if you do + three mounts on /bin, the three Mounts would be linked at + an Mhead setup for the first mount. That Mhead would be + the header of a list of three Mount entries. To say it + other way, you mount to the mount point, not to something + mounted on it. + + What remains to learn how are mount points traversed is to see what + does ``walk'' when walking through a mount point. Of course you + already know how domount works, but that is only part of the story + (given a file, go to the thing mounted on it); the other part of the + story is what to do when you have a directory that is a mount point, + and you have to lookup names on it, or to create names on it. + + Directory walking + + walk() Walks to a name on a channel. + + chan.c:493 + walk is called to resolve a file name (no paths!) on a channel. + The channel points to a directory (mounted or not). After walk, + the channel points to the file named name within the directory + pointed to by the channel given to walk. When the directory is + mounted, there many be more than one directory mounted. domnt + tells walk whether the given directory is considered to be + valid as a mount point; if it is not, no mount related job has + to be done. + chan.c:499 + ac is the ancestor channel (the parent you are walking + through). cp is a pointer to a Chan pointer; that is to update + the passed Chan. After a walk, the caller might be using a + different channel than the one given to walk. + chan.c:501,502 + I told you. + chan.c:504,508 + redundant ``..'' names were removed, but what about the ``..'' + path? In this case, undomount returns the channel for the + ``..'' file; more later. + chan.c:510 + Walk is going to move the channel to a subdirectory or to a + file contained in the directory represented by the channel. If + the passed channel had a CCREATE flag stating that it was + mounted allowing file creation, clean it up so it does not + pollute the inner file. + chan.c:511,517 + First, the device specific walk routine is called. That is to + tell the file server that the file being used by this client + (the kernel), and identified by the channel, should now point + to name instead. If the device walk succeeds, the ancestor file + contained a component named name; + Domount is called on the channel after the walk. The channel + points to the file named name, but if something was mounted on + that file, domount will find a mount hash entry for the channel + Qid, and traverse it. Next time walk be called on the channel + it will walk within the mounted tree, not within the mount + point. + An interesting implication of calling the device walk before + looking at mount points, is that to use bind on a file, the + file must exist, not just its name. However, that is reasonable + given that all you have to bind something on a file, is the + channel to the file. The ancestor of the file has nothing to do + with it. What I mean, is that bind is not like link on UNIX. + updatecname()Updates the name of a channel after a walk. + chan.c:477,490 + updatecname adds name to the cname in the channel, so that + cname still corresponds to the name of the channel file. Now ac + is a channel for the file named name. For ``..'', updatecname + uses undomount as walk did before; more about that later. + chan.c:519,520 + The device walk failed; i.e. there was no such file below the + file pointed to by ac. Now what? + The device walk just tried the name in the served file tree; + But it could be that the ancestor channel is a directory with + other directories mounted on it (i.e. you are walking through a + union). In this case, the channel has a non-nil mh pointing to + the Mhead for the mount point. Besides, it is clear that in + this case the walk could fail yet the file may be there because + of a mount. + Remember that the channel mh was set by domount, in case the + channel had something mounted on it. Well, not exactly, domount + received a channel with something mounted on its file name, and + domount returned a substitute channel for the mounted file, + with mh set; but you get the picture. What follows is done only + for channels that correspond to mount points. + chan.c:522,530 + Going to exercise mounted files in turn. Hold a read lock on + the Mhead. + chan.c:531 + Remember, mh->mount is the first Mount entry for the channel, + and they are linked through next. The author is iterating over + the mounted files. + chan.c:532,537 + After getting a clone for the channel pointing to the mounted + tree, try the walk on it. Should it work, we are done: got the + file!. Otherwise, try with the next mounted thing. Why is a + clone needed? Hint: consider what would happen to the mount + entry if its to channel was used for the walk. + chan.c:542,543 + No luck, no such file at any unioned directory. + chan.c:545,548 + Because c is already resolved, drop its mh reference, if any. + cclone asked the device for a clone channel, it may come with + an mh reference, but it should have none; because it is already + resolved and it does not represent the mount point--it + represents just the file mounted on it. Should mh remain set, + further calls to walk could try to iterate over the mount + entries found through it. + chan.c:551,556 + The name for the channel updated. The name coming out of the + mounted tree is mostly ignored, and the channel name is updated + to contain the name for the channel given to walk plus the + element name: the channel name is the name used by the user to + walk from the root to the channel. Two channels may point to + the same file, yet have different names because of bind. Start + to understand why channels have cnames? + chan.c:557,561 + The channel given is updated with the new channel, and it is + checked as a possible mount point by domount. Although the + channel points to a file in a mounted file tree, its name can + be a mount point too. domount would replace the channel with + that of the mounted file, if any. + + To completely understand walk, you still must read how undomount + works. Undomount is the one who understands the meaning of ``..'' when + it must cross a mount point (hence its name). + + Consider the case when /usr/nemo/bin/rc is bound to /bin/rc. Now, + imagine there is a directory /usr/nemo/bin/386. If you walk to /bin/rc + you end up in the same directory as if you walk to /usr/nemo/bin/rc. + Consider the case when you do the last walk. Your file is + /usr/nemo/bin/rc. Imagine that you walk to ../386, your file is + /usr/nemo/bin/386. Now consider that you walk to /bin/rc, if you later + walk to ../386, you should end up in /bin/386, not in + /usr/nemo/bin/386; even tough ``../386'' was applied to a channel + really at /usr/nemo/bin/rc! + + When doing a ``walk("..")'', the new file should be the parent of the + current file, but the parent regarding the path used to get to the + file. To pick up yet another example, if you use the path /bin/rc, you + do not care that it was really /usr/nemo/bin/rc, for you, ``..'' means + /bin, not /usr/nemo/bin/rc. + + By recording in channels the name used to create the channel, ``..'' + can be implemented that way. Both walk and updatecname call undomount. + + To make it more clear, consider in our example a walk to ``..'' on a + channel pointing to the file /usr/nemo/bin/rc whose cname is + ``/bin/rc'': + 1. + walk calls undomount (chan.c:506), which returns a channel + pointing to /bin/rc, given the channel pointing to + /usr/nemo/bin/rc. The channel name is still /bin/rc, and the + channel still points to ``rc''; but now you are at the /bin/rc + tree, not any more in the /usr/nemo/bin/rc file tree. + 2. + Later (chan.c:511), walk calls the device specific walk + routine, which would do a walk("..") in the channel for + /bin/rc, making the channel point to the file /bin. You are + mostly done, but for the channel name, which is still /bin/rc. + 3. + At line chan.c:513, updatecname would build a name /bin/rc/.. + and simplify it, leading to /bin. You are done! + updatecname calls undomount too. When /bin is also a union, + updatecname would return the mount point for the union, and + later chan.c:515, domount would return the channel for the + first mounted directory at that mount point. That channel has + the mh field set so that lookups in unions could work for it + too. + If walk fails when calling the device specific walk for ``..'', + it would continue below line :517; if the channel is a union, + the same device_walk+updatecname+domount sequence is played at + each unioned directory. + + ... + undomount() Goes from the mounted file to the mount point. + + chan.c:436 + Back to undomount, it does part of the ``walk'' for ``..'' on c + . In our example, the channel for /bin/rc after the bind of + /usr/nemo/bin/rc, was really a channel to /usr/nemo/bin/rc; + only that its Cname was /bin/rc. Undomount steps back from the + mounted file, to the mount point. + chan.c:443,448 + The meaning of ``..'' depends on the mount table. + chan.c:450,453 + he is the end of the mount table. The loops are iterating over + the entire mount table: all hash buckets, all mount points, all + mounted files. A very expensive operation! (hence the effort to + remove unnecessary ``..'' elements in path names). Hopefully, + there will not be so many mount entries and this would not have + a noticeable impact on system performance. + chan.c:454 + The channel c corresponds to a mounted file (it would be an + entry for /usr/nemo/bin/rc mounted somewhere). + chan.c:455,462 + But, in our example, you don't know whether it is the entry for + the mount at /bin/rc, or it is an entry for a mount at a + different mount point. t->head is the Mhead for the mount + entry, t->head->from is the channel for the mount point; its + name->s is the C string for the mount point channel name. + Should the string be the same that the one in c's Cname, the + mount point is the one we are looking for (e.g. it would be + /bin/rc). Otherwise the loop continues searching. + chan.c:463,467 + Strings did match, so to step back the union, get a clone of + the mount point and stop the search. + One final note about this. The author breaks just the inner + loop and I don't see the point on searching remaining mount + points once one has been found--I mean that the two outer loops + would continue. I think this is a bug which affects just + efficiency, although the author may have a good reason for + doing so. + + Creating files on mounted places + + You may remember that + + namec + + chan.c:819,820 + while resolving a name for creation, namec tries a walk to the + file, and reaches these lines if the walk failed. The file + named elem is going to be created under the file pointed to by + c. However, if mh is not nil in c, it is a mounted file. + Besides, it is one of (maybe) many files mounted at the + directory where the elem file has to be created. On which one + of these mounted directories must the file be created: on the + first in the union that has the MCREATE bit set. Channels to + mounted files have the CCREATE bit set if their mount had the + MCREATE flag. So, if this channel (the first in the union) has + the CCREATE flag, it is the channel to the directory where the + file should be created. + + namec + createdir() Chooses where to create a file in a (union) directory. + + chan.c:567,593 + Otherwise, createdir locks the Mhead and iterates over the set + of mount entries. The first one with the CCREATE bit set is + cloned, and the clone returned. namec would then use the + channel for the first directory mounted with MCREATE to create + the file on it. + The mh field of the cloned channel is cleaned up, and reset to + be the mh for the channel c (i.e. for the first mount entry). + If the creation fails and another walk is tried once more, the + walk would use the mh field to lookup names in the union. + When no mount entry has the create bit set, an error is raised + and file creation is aborted. + + Mounting and binding + + sysbind() Entry point for the bind system call. Binds a name to + another name. + + sysmount() Entry point for the mount system call. Mounts a file tree. + + sysfile.c:687,697 + Both bind and mount are implemented by calling bindmount. The + last parameter is an indication that a mount is being done. + + sysmount + bindmount() Mounts or binds a file to another. + + sysfile.c:624,629 + The third parameter flag controls how the operation behaves. It + is a bitmap of bits defined in lib.h:85,91. The two low bits + are used to specify the order for the new directory. The user + cannot request both that the directory be mounted before and + after the previous ones. Besides, only bits in MMASK are valid + flags. MCACHE requests that file data should be cached, and is + noted in bogus. Bogus is being initialized to contain the + information about the mount. + sysfile.c:631,662 + For mount, a file descriptor is supplied as a first argument to + bindmount. Must convert it into a channel. For bind, it is a + path, which must be converted to a channel too. + sysfile.c:632,633 + mounts are forbidden if the noattach bit is set. It is okay to + bind because no new files are brought into the name space, but + no new file trees can be attached. + sysfile.c:635,640 + fdtochan takes a file descriptor and returns a channel for it. + The last parameter specifies to add a new reference to the + channel. Should an error occur, cclose would drop the + reference. The channel corresponds to the file descriptor being + mounted, and is noted in bogus. On the other end of the + channel, there should be someone speaking 9P, to service file + requests. + sysfile.c:642,650 + For mount, a request is going to be issued to the file server + to attach its file tree to our name space. As a server can + service several trees, the fourth argument of mount specifies + (as a string) which file tree to mount. The author is ensuring + that the string is in valid virtual memory, and noting the + string into bogus.spec. nameok is used to check that the spec + has a valid path (i.e. valid characters); it is okay if spec + contains slashes, hence the 1. + sysfile.c:652,656 + The file tree being mounted is serviced by a remote process + which speaks 9P. The client is going to be the kernel mount + driver, which translates procedure calls into 9P RPCs as the + mounted tree is used. The attach procedure of the mnt driver is + used to get a channel to the remote process: First, knowing + that the name for mnt is #M, devtab is searched by devno. devno + returns a valid channel type (an index into devtab) for the + mount driver; Second, the attach procedure for the driver is + called supplying the bogus structure just filled up. If you + look at it, bogus contains a channel to the server, the spec + for the file tree requested, and an indication of whether files + are being cached or not; everything attach needs to get in + touch with the file server and attach to it. This is discussed + later. + If attach completes without error, c0 is a channel to the (root + of the) file tree in the server, which (after completion of + attach) recognizes us as a valid client and is willing to talk + 9P with us. + sysfile.c:658,662 + The first argument for bindmount was the path given to bind; + therefore, to obtain a channel it suffices to use namec to + resolve the path into a channel. c0 is now a channel to the + file tree being mounted. + sysfile.c:669,674 + The second argument was the path for the directory where the + file tree is being mounted at--or for the file where the new + file is being bound at. namec is used to get a channel c1 for + the mount (or bind) point. Note how Amount access mode is used + in namec; c1 would be the very first mount point for arg[1]. + sysfile.c:676,685 + The first line is where the mount is being done; what remains + is to close both channels (mount point/mounted file) because + the mount table already holds what it needs, and to close the + descriptor supplied to mount in case it was a sysmount. The + descriptor is closed because the kernel is going to exchange 9P + messages on it with the file server; the user would only + interfere and cause problems. But how does cmount work? + + sysmount + bindmount + cmount() Adds a mount entry. + + chan.c:226 + cmount is called with channels for the mount point (old) and + the mounted file (new), the flags and any spec are passed too. + chan.c:233,234 + It is okay to bind a file to a file, and a directory to a + directory; but not for any other case. Don't trust users! + chan.c:236,239 + If mounting with MBEFORE or MAFTER, ensure channels are for + directories; otherwise, it has no sense--for files, you only + can replace entire file contents. + chan.c:241,242 + Going to write the Pgrp, stop any further lookup/change to the + name space. + chan.c:244,249 + Search the name space for any Mhead for this mount point (note + eqchan again!). If an Mhead is found, m points to it; + otherwise, m would be nil. + chan.c:251,269 + The comment says it all. The from field of the Mhead is set to + the mount point--and the reference noted by incref. As you now + know, the point is not that this particular channel is set in + the from; the point is that a channel with its cname, its type, + its dev, and its qid is sitting at the from. + Lines :267,268 may add a mount entry for the mount point itself + to the Mhead. Guess why? + Exactly, if the mount is not an MREPL, previous contents at the + mount point are still visible; therefore, the channel to the + mount point is added as if it was one of the directories + mounted on it. When later the mount entry list be searched, + names originally at the mount point will be searched too. So, + the Mhead has now either the original mount point channel, or + it is empty. + newmount()Adds a new mount entry. + pgrp.c:231,245 + newmount simply allocates a Mount entry, and initializes it. + The author sets to i to the channel for the mounted file; and + sets the pointer to its Mhead. + mountfree()Releases mount entries. + pgrp.c:247,259 + The counterpart of newmount is mountfree, which releases + references to the channels for the mounted files as well as the + mount entries. + chan.c:270,275 + After the mount entry is locked, there is no need to keep + locked the whole name space. + chan.c:277 + An entry for the new mounted file is created. + chan.c:278,292 + As the mounted file could itself be a union, any mount entry + for the mounted file must be copied to the list of entries for + the mount point. The author links a copy of each such entry + starting at the next pointer for the new mount point being set + (nm). + If the mounted file is to replace the mount point, its mount + entries are set with the MAFTER flag. I don't think this flag + is used for anything. The point is that the Mhead for the mount + point has either its previous mount entries, or an entry for + the old contents if appropriate; besides, the new Mount has all + entries for the mounted file linked on it through the next + pointers. + chan.c:294,297 + If this mount was an MREPL, cleanup all previous mount entries + for the mount point. mountfree releases all the mount entries. + chan.c:299,300 + Cache the MCREATE flag on the channel, so createdir only needs + to look at the channel. + chan.c:302,306 + If mounting after, skip any previous mount entry at the Mhead, + and link the new mount point (and all trailing mount entries + for directories mounted at the mounted file!) at the end. + chan.c:307,312 + Link the list of new entries before the previous ones. If mount + was MREPL and not MBEFORE, the list was already cleaned up, so + a ``link before'' works too. In the case of MREPL it could be + more efficient not to iterate the list being added to the Mhead + (because that is just to set the last next pointer to nil), but + nobody cared to optimize that. Would the user notice the + optimization? + + Mounted files can be unmounted by calling unmount. + + sysunmount() Removes a mount entry. + + sysfile.c:700 + This is the entry point for unmount + sysfile.c:706,707 + The name for the mount point is resolved (note the Amount + access) and cmount is now a channel to it. + sysfile.c:709,717 + If the first argument is not nil, it specifies the mounted + file--in this case the user wants to unmount that specific + mounted file, and not any other one. The address is checked and + namec used to get a channel for the mounted file (note the + Aopen access mode). Should the first argument be nil, cmounted + remains set to nil. + sysfile.c:726 + cunmount undoes the mount. Note, not undomount! + + sysunmount + cunmount() Undoes a mount. + + chan.c:329,334 + The Mhead is located for the mount point--yes, perhaps a + getmount routine could be created to do the lookup and share + these lines among routines looking up mount entries. + chan.c:336,339 + The user could call unmount on a file with nothing mounted on + it. + chan.c:341,352 + If shouldn't care about unmounting a particular mounted file, + mountfree is called on the whole list of Mount entries. All + entries are released and the head released too. All mounts are + undone. The unlock is done in any case after locking the + particular MHead of interest. + chan.c:354,376 + The user cares of unmounting a particular mounted file. So, + iterate the set of mount entries for the mount point. Iteration + stops when an entry is found such that its to channel is eqchan + to mounted. (Line :358 checks if the channel used to talk 9P to + the server of the mounted directory is eqchan with the mounted + directory; this can happen with exportfs, as you will see later + when you read about the mount driver). If the entry is found, + it is removed from the list and released. If the list gets + empty, the Mhead is released too. If no entry is found, an + error is raised--that ``mounted file'' was not mounted at the + union. + + Creating and destroying name spaces + + Other routines that operate on name spaces are the ones creating an + empty name space, copying an existing name space and deleting a + namespace. That happens as a consequence of rfork and process death. + + sysrfork + newpgrp() Creates a new name space. + + pgrp.c:42,51 + An empty Pgrp is created with everything set to nil. As a + result, the mount table hash is empty and noattach cleared. You + saw in a previous chapter how pgrpid was assigned later. + + sysrfork + pgrpcpy() Copies a name space. + + pgrp.c:124,130 + pgrpcpy is used to copy a name space into another. Only the + source is locked because the target is still being built. + pgrp.c:132,159 + Hash entries are replicated so that to holds a copy of the + mount entries in from. Note the increfs when structures are + shared (Channels are!, because the namespace is copied, but + channels to mounted file systems and mount points are not!). + The loop is simple to understand if you notice that for each + pass (:135) an Mhead f is being copied into a new one, mh, (l + is used to build the list); and at each inner pass (:144), + Mount m is being copied into a new one, n. + The field copy of m is set to n (the place it is being copied + to). pgrpcpy relies on pgrpinsert to link mount entries being + added through the Mount.order field. Both copy and order seem + to be in Mount just for this occasion--to save the author the + burden of allocating a whole bunch of data structures during a + brief amount of time just to copy the set of mount entries. + pgrpinsert()Inserts a mount into a list ordered by mountids. + pgrp.c:100,118 + What does pgrpinsert do? It receives the pointer to the Mount + being copied and a pointer for the start of the ``order'' list. + If the list is empty, it is set to the node just copied. If the + list is not empty, f advances until its mountid is bigger than + the one for the node copied. So, pgrpinsert is building a + sorted list of Mount entries. The list is sorted in ascending + order of mountids. + pgrp.c:163,167 + That was the list for, mountids for the copied entries are + assigned in the same order they were assigned for the original + Pgrp. Why? + The mountids can tell the order in which the user issued mount + requests that resulted in the set of Mount entries. That order + does not correspond to the order of entries in the Mhead, + because of the MBEFORE/MAFTER flags. The proc driver services + an ns file that returns (when read) commands to replicate the + namespace. It relies on the order of mountids to reproduce the + commands in the appropriate order. + Execute in your Plan 9 box a couple of mounts and then execute + cat /proc/$pid/ns. More clear now? + + closepgrp() Releases a name space. + + pgrp.c:71,97 + closepgrp is called to release a reference to the pgrp. When it + comes down to zero, all entries are released. Two locks are + needed: devproc uses the debug lock, although routines using + the namespace use the ns lock. + + File I/O + + Once a file descriptor is open, the user can use read(2) and write(2) + on it. Although the actual implementation is done by the server + servicing the file, functionality such like the file offset is kept at + the client side and is provided by channels (You already know what a + file offset is, if you don't, read the ``File I/O'' section of + intro(2)). + + That is a fine way the author has to side-step the problem of sharing + file offsets on distributed file systems: by not doing it! Take this + as an example of the principle that, before adding a new feature to + your system (e.g. sharing file offsets among several nodes) you should + ask yourself what is the benefit, and what is the cost; then decide. + + The most useful feature of shared file offsets, i.e. allowing several + related processes to consume/produce the same file, is still here: + * Several processes (within the same node) can share a file + descriptor and that means they would share the file offset too. + * The ``append only'' bit of file permissions can be used to allow + several processes to produce contents for a file no matter the + node they are running at. This is in effect ``sharing'' the file + offset, which is always kept at the end of file for write + purposes. + +Read + + sysread() Entry point for the read system call. Reads from a file. + + sysfile.c:375 + sysread receives a descriptor and a buffer together with its + length. It is expected to fill up the buffer with bytes coming + from that descriptor. The bytes can come from an actual file, + or from a file synthesized by a file server, nobody knows. + sysfile.c:381,383 + Buffer addresses are verified and c is set to the channel for + the descriptor. The channel should have the OREAD bit set. + Remember that when you open a file, its mode is cached in the + channel. Although permissions are checked by the file server, + when the device specific open routine is called, once the file + server granted access, the mode is kept in the channel. Future + access checks can be done by the client without disturbing the + server (although the server is likely to check that the file id + has the OREAD bit set too). + The last 1 to fdtochan requests an incref for the channel. + Should the descriptor be closed by a different process, the + channel would stay alive because of the added reference. Forget + by now the 1 asking checkmnt to fdtochan. + sysfile.c:390,396 + If the channel Qid has the CHDIR bit (it is a directory), a + read should return a integral number of directory entries (See + the read(2) manual page). The buffer length (n) is adjusted to + be a multiple of the directory entry size DIRLEN. Should the + channel offset be not a multiple of DIRLEN (shouldn't happen) + or the buffer too small to keep even a single entry, Etoosmall + is raised. + sysfile.c:398,405 + If c is a union, read should get entries from all the mounted + directories. + It is easy to know if this is a union; remember that the + channel installed for the file descriptor was obtained with + namec and Aopen access mode. That means that any mount point + was traversed and mh initialized in the channel to point to the + Mhead. + Thus, if there is an mh in the channel, unionread reads entries + from each directory mounted. Otherwise, the device specific + read is called. The device should honor the convention that a + read from a directory should return an integral number of + directory entries. Should the device fail to honor that + convention, the channel offset would be set to a non-multiple + of DIRLEN at line :404, and the next read will fail. The offset + for the file is kept by the channel; you already knew this. For + directories, the file offset counts bytes, and not directory + entries! as it should be. + The channel is locked just while changing its offset. Apart + from that, it can be used safely without locking because it is + not going to be deleted, nor its device type is going to + change. The real read is done by the device, which could stand + miles away, and there is no guarantee that while this read is + in progress no other read could be made. Nevertheless, the + server is likely to serialize read/write requests for the file. + + sysread + unionread() Reads from a union. + + sysfile.c:280 + unionread does the work for reading the union while holding a + lock on it. + sysfile.c:291,292 + Where to start reading? If you read a union from the beginning + to the end, you should get all directory entries in all + directories mounted, as you find them in the list of Mounts. + The problem is that users tend to declare a buffer and read + repeatedly from a file; each read should start past the + previous read (remember offset, right?). + The author could use the channel offset and skip as many Mount + entries and directory entries as needed to fill up offset + bytes. However, that would be a waste because reading a union + would be O(n2) regarding the number of entries. You don't know + how many entries there are at each mount entry. Therefore, you + cannot compute how many entries to skip unless you read them. + Can you think of more alternatives besides re-reading them and + doing what is said in the next paragraph? What are the + benefits? What are the penalties? + Instead, an uri (union read index) field in the channel is used + to record how many mount entries were exhausted by previous + reads. That saves lots of reads that are likely to be serviced + from the network and not from a local cache. The loop at lines + :291,292 is skipping over already exhausted mount entries. + sysfile.c:294 + Keep on reading while there are mount entries to read from. + sysfile.c:299,300 + As far as I know, to is only released by mountfree, and that is + done with the lock on the Mhead held. Therefore I would say + that in no case to should be nil, but the check does not hurt + anyway. + + NOTE: Is this ok? + sysfile.c:301 + Going to read from to, so clone it. Otherwise, the channel + would be modified (e.g. its offset would change). Perhaps in + this case it would be more simple to save the previous offset, + explicitly set it up, use the channel and restore the offset. + But that would require that all channels for mounted + directories be kept open all the time; that would mean ``use + more resources'' for the file servers. Besides, it is a good + thing to keep channels for Mount entries untouched; the author + can rely on that to make the code more simple. + Don't you see any other problem here? You cannot walk on an + open file, so the clone is necessary anyway. + sysfile.c:305,308 + This is not the usual error handling block. Should an error + occur, the channel to the mounted directory is closed and the + routine continues at the next entry. That means that if a file + server goes down, only people really using it would be + affected. If a directory serviced by the crashed file server is + mounted, its entries will be ignored due to errors, but + remaining entries in the union would still work. This is the + kind of thing that makes a distributed system more reliable: be + prepared to tolerate remote failures bothering the user as few + as possible. + sysfile.c:310 + The device specific open for the mounted directory is called. + The clone channel is used. + sysfile.c:311,314 + The author is indeed adjusting offset by hand between + successive unioreads. The offset is saved in the channel for + the mounted file; But note how the offset is used only within + the same mounted directory. + sysfile.c:318,321 + If could read something, return that. If more entries remain to + be read, the user would call read(2) again. offset is adjusted + before returning so that next time the read would start where + it was left at. + sysfile.c:323,329 + If there are no more in the current mounted directory, nr would + be zero and this code execute. When an error happens, line :307 + would jump here too. Just increment uri to remember that the + current mount entry should be skipped next time. If there are + no more entries, break the loop and return 0 (eof). Otherwise, + reset the offset and iterate again, so that the next mounted + directory starts to be read at offset zero. + +Write + + syswrite() Entry point for the write system call. Writes in a file. + + sysfile.c:444 + syswrite has the same interface of sysread, but it writes + rather than reads. + sysfile.c:453,459 + Should an error happen, restore the offset in the channel. The + code following advances offset, yet it could fail. If an error + occurs, the write failed but the offset should be kept + untouched. Perhaps this would be more clear if the offset were + simply saved to a variable and restored from there, but the + author does not do so. Guess why? + sysfile.c:461,462 + The only way to write to a directory is by creating or deleting + files (also by changing file attributes). + sysfile.c:464,467 + Advance the offset by the number of bytes to be written. oo + keeps the old offset, and could perhaps be used to restore the + initial channel offset on errors, perhaps not. + sysfile.c:469 + Here is the actual write. The old offset is supplied and the + device specific write would write at that position. + sysfile.c:471,475 + The device could write less than n bytes. This is usually due + to an error (e.g. ``disk full''). If the device wrote m bytes, + the offset has to account for those m bytes. However, it was n + that was added to it, by subtracting n-m, it gets m units more + than it had before syswrite. Should the device write all n + bytes, offset is kept n bytes beyond its value before syswrite. + + Could you guess the reason for the dance around offset? sysread did it + in a more straightforward way, by simply adding n bytes to the offset + after the read was done. + + I think that the reason is that the author does not want the channel + to be locked during the whole syswrite. Suppose the channel offset is + o. For sysread, the worst thing that can happen if two different + processes are reading the same file (through the same channel) is as + follows: + * The first sysread calls read in the device and reads at offset o. + * The second sysread calls read, in the device, which also reads at + offset o! + * Both sysread increment the channel offset. + + Although this could be considered to be a bug (If I am not missing + anything here), the net effect is not harmful, as the file is kept in + a consistent state. Actually, I would say that the bug is at the + application, which didn't synchronize on its access to the file. + + Now consider syswrite. writes are different in that if they are mixed + the file could be left in an inconsistent state. Suppose again that + offset is o, and two syswrites are being done through the channel. If + syswrite worked like sysread (by adding n to offset after calling the + device write), the two writes could overwrite the same portion of the + file; e.g. the two calls to the device write are performed, then the + offset incremented twice (holding the lock for the offset). + + By incrementing the offset in the channel before calling write, any + following write through the same channel would ask the device to write + past the n bytes theoretically being written. Of course that means + that the offset must be restored (in case of errors) by doing + arithmetic and not by restoring the initial value. The reason is that + if the first syswrite fails, but in the mean time the second syswrite + could really write, offset should account for the written bytes by + both the first and the second syswrite. If you didn't understand, try + to exercise concurrent syswrites and see how are the file and the + channel left. I think that the approach used in syswrite should be + used in sysread too, but the author may disagree. + + Finally, note that when two different channels are used for the same + file (applications could run at different nodes), it is the + application responsibility to synchronize. The CHEXCL bit could help + here. As it is said in create(2), if a file is created with that + permission bit set, only one client may have the file open at a time. + That functionality is implemented by the file server, which serializes + file usage when notices the CHEXCL mode bit. + +Seeking + + The seek(2) system call can be used to move the offset in an open file + to a desired position. + + sysseek() Entry point for the seek system call. Moves the file offset. + + sysfile.c:535,541 + After checking the argument, sseek does the work. + + sysseek + sseek() Seeks on a channel. + + sysfile.c:495 + The first argument is the file descriptor to seek on. Get the + channel. But note the arg[1] (not arg[0]!) for the first + argument. + sysfile.c:500,501 + Should seek be allowed on directories, sysread could found + offsets not aligned to DIRLEN. So the author forbids that. + sysfile.c:503,504 + No seeks on pipes. Other file servers could perfectly ignore + seeks as well as they could ignore file offsets (i.e. they + could read always from the very first byte). But that's the + file server choice. + sysfile.c:506,509 + An vlong (second argument to seek) uses two longs. Now, the + arguments for system calls were assumed to be machine words + (longs). These lines use the u member of the o (offset) union + to fill up the two words of the vlong. Arguments 2 and 3 are + actually the second argument (two words) of seek. The first + argument was kept at arg[1] and not at arg[0]. That is because + arg[0] is kept unused to make the vlong stand aligned to an + vlong size boundary regarding the argument array. That is a + good thing if the code is pretended to be portable, as some + machines (not the PC) are very picky regarding alignment + issues. The user-level library stub to issue the system call + for seek must honor the same convention of wasting arg[0] and + pushing the vlong in the order in which it is being extracted + here. If the stub (machine dependent) honors this convention, + this code can be kept portable. Now you know also that arg[4] + is actually the third argument for seek, right? + By the way, look /sys/src/libc/9syscall/mkfile:50,61. + Understand the if now? + sysfile.c:509,528 + Depending on type, the n kept at o should be either used as the + new offset, added to the current offset, or used as the new + offset but counting backward from the end of the file. When + arithmetic is being done with the channel offset, a lock is + held. Does it makes sense to hold a lock just to assign the + offset? + If the offset can be written atomically, it doesn't matter; but + on Intels, in this case, two words are written as the new + channel offset. There is a very thin critical region here, and + the offset could be mangled in the very improbable case that a + sseek writes the offset, while another sseek does too. Just too + improbable, but perhaps that should be fixed. + Stat is used to get DIRLEN bytes with status information from + the file, and convM2D is used to convert that into a Dir + structure, from where recover the length of the file. stat is + discussed later, but convM2D is not. The DIRLEN bytes are in a + ``standard format'' and have to be converted to a native format + before used (byte ordering et al.). The reason for this all is + that Plan 9 is a distributed system. + sysfile.c:529,531 + The final offset value given to the user (line :506 is not + needed but it is good to keep it there for safety). uri is + reset, because unionread could use that to read entries? That + couldn't happen. If this is a directory, seek is forbidden. + However, it is a good practice to set uri to a reasonable + value, in case a bug (no check for directory?) is introduced in + the code by future changes. + sysfile.c:543,560 + You may have noticed that there is a sysoseek routine below + sysseek. If you look at system call numbers for seek and oseek + (/sys/src/libc/9syscall/sys.h), you would notice that seek is + placed the last one, and oseek is placed near read. That means + that the author had once just the seek system call, and it was + number 16 (OSEEK now). But seek was changed in a so + incompatible way, that the author preferred to keep both the + previous and the new version for seek in place. Old Plan 9 + binaries would have the library stub named seek that does a + system call with number 16, and that would call sysoseek. New + Plan 9 binaries would be built with the stub which calls seek + instead. This is a common technique to reduce the impact of + changes. What sysoseek does is to rearrange the argument array + so that a[0] contains the address of the vlong, and the vlong + is filled up with arg[1], which means that the change was + probably that seek received a long, and now it receives an + vlong. In other words, sysoseek is adapting the old interface + to the new interface, but the implementation of seek stands the + same. Try to learn from this all how to minimize the impact of + changes. + +Metadata I/O + + Files have attributes, and you already know some. They have + permissions, a name, owner, etc. The system calls reading/updating + attributes are stat(2), fstat(2), wstat(2), and fwstat(2). The former + two ones read attributes, and the later two ones update attribute + values. Services like chmod(2) are implemented by doing a wstat(2) on + the affected file. File attributes are read and written in a machine + independent format with DIRLEN bytes per file attribute set. Read + stat(5) if you are curious about file attributes. + + Let's see the system calls that read attributes before, and then the + ones that write them. + + sysfstat() Entry point for the fstat system call. Reads file + attributes (given a file descriptor). + + sysstat() Entry point for the stat system call. Reads file attributes + (given a file name). + + sysfile.c:563,578 + sysfstat takes an open file descriptor, and tries to read new + values for attributes. All the routine does is to get a channel + and call device specific stat to read the attributes. It is the + file server the one filling up the buffer with information + (probably) coming for Dir structures. + sysfile.c:581,597 + sysstat is the same, but takes a file name instead of a file + descriptor. namec does the job of getting a channel, and then + the device specific stat routine reads any attribute. Perhaps + both system calls could be folded into one, but it's not a big + deal. + + syswstat() Entry point for the wstat system call. Writes file + attributes + + sysfile.c:778,813 + The ``write attributes'' version of the routines simply call + the device wstat instead of the device stat. The only thing the + author verifies regarding the new attributes, is that the new + name for the file (first bytes in the buffer with attributes + supplied) looks fine and has valid characters. That is the only + attribute that would hurt 9P. Should the name be ok, the call + to wstat can be done, and remaining attributes should be + checked by the file server. + + This is another place where you can see how the design of Plan 9 works + well for a distributed system. File attributes are kept (together with + the file) within the file server providing the files. The system does + not impose any particular way of implementing file attributes. All + Plan 9 cares about is that the device should either be in-kernel, or + service 9P requests. Besides, by forwarding all calls related to file + metadata to the file server process, Plan 9 does not introduce new + problems related to metadata sharing over a distributed system. Yet + another point is that the file server is free to trust you and accept + wstat requests; it would do so if you authenticated the connection to + the file server. Saw how all pieces fit together? + + Other system calls + +Current directory + + The getwd(2) function is not a system call. It is a library function + that opens ``.'' and calls fd2path(2) on it. + + sysfd2path() Entry point for the fd2path system call. Returns the name + for a file descriptor. + + sysfile.c:120,135 + sysfd2path receives a file descriptor and a buffer together + with the buffer length. It verifies that the buffer has valid + addresses (arg[1] is the pointer to the buffer and arg[2] is + its length). Then it uses fdtochan to get the channel for the + open descriptor. The work was really done when the channel + (Cname) was built. The only thing fd2path has to do is to + extract the name (if there is any) and print it in the user + buffer. The channel name is the name used by the user to get to + the file. If the user used a relative path to open the + descriptor given to fd2path, the name of up->dot was used to + build c's name, in namec. + + Another useful system call is chdir(2) + + syschdir() Entry point for the chdir system call. Changes the current + directory. + + sysfile.c:599,610 + It simply verifies the name given, and resolves it to a channel + using namec (Note the Atodir access, that was explained + before). Then dot for the current process is set to that + channel, after releasing the reference to the previous dot. + +Pipes + + There is a system call to build a pipe. A pipe is a buffered channel + used to let two processes communicate. Surprisingly, the pipe system + call is not a real system call; I mean that although it is a system + call, it uses files serviced by a pipe device to provide its service. + The system call is provided as a convenience, although the user is + perfectly capable of using the pipe device himself without relying on + the system call. Using the device has the good thing that the user is + conscious that pipes are provided using files serviced by pipe, and + they could be even mounted through the network. + + syspipe() Entry point for the pipe system call. Creates a full-duplex + pipe. + + sysfile.c:138,146 + syspipe is called to create a pipe (see figure [230]5.4). The + pipe interconnects two file descriptors so that at one you read + what was written at the other. arg[0] is an array with space to + put the descriptors numbers in. evenaddr is used because + integers are going to be written into arg; some machines issue + alignment exceptions when you write an integer into a location + not aligned to an even address. + + CAPTION: Figure 5.4: Pipes are bidirectional channels accessed through + files provided by the pipe device. + + sysfile.c:147 + d holds the Dev entry for the pipe device, which is named #|. + sysfile.c:148 + c[0] is a channel to the root of the pipe device file tree. The + name works despite what the user did with his mount table. Here + is where the pipe was actually setup. If you read the pipe(3) + manual page, it says that an attach of the pipe device causes a + new pipe to be created. The pipe endpoints are files named data + and data1 below the root supplied by the pipe device. Different + attachments to the pipe device cause different pipes to be + created; files data and data1 would be different for each + attach. + As a side note, Plan 9 does not have the UNIX mkfifo system + call, which creates a pipe with a name in the file system. In + Plan 9, all pipes have names, even those created with pipe(2). + Execute ``bind '#|' /tmp/dir'', and then try to read/write + files in /tmp/dir. + sysfile.c:150,151 + By default, descriptors given to the user are invalid ones. On + error, the user will know. + sysfile.c:152,161 + Preparing to cleanup on errors. Descriptors are closed only if + they were open. fd holds the the descriptor numbers, which are + indexes for the Fgrp descriptor array f->fd. + sysfile.c:162,166 + Both c[0] and c[1] are channels to the root of the pipe device. + So, walk them to data and data1 to get channels to both ends of + the pipe. (why did not the author choose names ``data0'' and + ``data1'' for both endpoints?) + sysfile.c:167,168 + Important. In Plan 9, the walk merely changed the channel to + point to a different file, but you have to open the file before + doing I/O. The mode is ORDWR because Plan 9 pipes are + bidirectional. + sysfile.c:169,170 + The caller of pipe(2) is unaware of pipe files, the author + plumbs the channels to a couple of file descriptors and pipe is + done. + + Device operations + + In the code you read before, you noticed that the actual file system + work was done by device specific operations. In fact, you already knew + this since chapter [231]3, ``Starting Up''. Let's read now the code + still unread regarding devices and device operations. + + portdat.h:175,195 + The Dev structure (which you already saw), contains the + information for a known device type. Instances of Dev are + configured into devtab when compiling the kernel. The first two + fields contain the ``device character'' and the device name. + You know that kernel devices have names like ``#C'', the ``C'' + is at dc; it identifies the device type. To locate the entry in + devtab for a particular device, you only have to iterate + through it searching for an entry with the wanted dc. The name + is mostly for debugging. + Regarding device operations, they correspond to 9P messages + (read intro(5)). When a user process (or the kernel) performs a + file operation, that operation translates into 9P requests + (transactions). For example, you know that to open a file you + need to clone a channel, walk on it, and open it. These + procedures (clone,walk,open,etc.) correspond actually to + Tclone, Twalk, and Topen 9P requests. When the file server is + within the kernel, the kernel looks up the device in devtab and + calls its implementation for the request (the Dev.open, + Dev.walk, etc.); when the file server is remote, the kernel + driver for the file is the mount driver, which issues 9P + messages (Topen, Twalk, etc.). Do you get the picture? + 9P uses the term ``transaction'' for requests. So, what is each + 9P transaction for? One fine way of learning it is to look at + the implementation of each transaction for a particular device. + I'm going to use the pipe device. While you read the devpipe.c + source, you will learn what is each transaction for; and you + will see how the file dev.c provides default implementations + for most 9P operations. Such default routines are handy when a + device has nothing to do (but for replying to the request) to + service a particular 9P transaction. All devices are file + servers, and all file servers are likely to share much code; + that shared code is located into generic utility routines in + dev.c. + +The pipe device + + Initialization + + devpipe.c:370,389 + This is the ``entry point'' for the pipe device. The Dev + structure is linked into devtab so that the kernel can locate + it. The kernel device name is ``#|'', among routines linked + here, there are routines corresponding to 9P transactions. + devreset()Generic procedure to reset a device. + dev.c:62,65 + The ``reset'' procedure is not a 9P operation, but a routine + provided to ``reset'' the device to an initial state so that + its init procedure could be called. You saw how it worked for + ethernet devices while learning how the system boots. For + pipes, nothing has to be done to reset the device; and that is + very common. The dev.c file contains a devreset routine to use + as a generic reset procedure. It does nothing. Instead of + declaring an empty routine for each device without resetting + needs, this one is used. + Remember that chandevreset calls all reset procedures for + configured devices at boot time. + + pipeinit() Initializes the pipe device. + + devpipe.c:41,50 + The ``init'' procedure is not a 9P operation. It is used to + initialize the device and prepare it for operation. After init + is called, other procedures can be called. For pipes, the + configured pipeqsize parameter determines the size of the queue + used by each pipe. Should it be unspecified in the + configuration file, 256K are set for multiprocessor machines, + and 32K for monoprocessors. For multiprocessors, processes at + both ends of the pipe can execute at different processors. It + is not a problem if a process is allowed to run until it puts + 256K in the pipe, even if the reader has not read a single + byte. By giving more room to the pipe, the writer can write + more without blocking, and the reader can read more without + blocking--even if the other process is not attending the pipe. + On monoprocessors, the author thinks that it is better not to + let the process run for so long before being blocked; after + all, the process at the other end of the pipe has to use the + only processor in the system. + devinit()Generic procedure to initialize a device. + dev.c:62,70 + The default implementation (used if the device does not care + about init) is one that does nothing. + You know that chandevinit calls all the init procedures for + configured devices (after chandevreset) at boot time. + + Attaching to the server + + pipeattach() Attaches to the pipe device. + + devpipe.c:55,56 + The first 9P request issued to a file server is an attach. It + attaches a client to the file server. Any authentication is + done here. (see attach(5)). Usually, attach would just attach + to the server's file tree. However, for devpipe, attach also + creates a new pipe and attaches the client to its corresponding + file tree. To create multiple pipes, you attach multiple times + to #|. An string spec is supplied to attach. That is useful in + case a server services multiple file trees, to select one of + them. + devpipe.c:61 + devattach is a generic attach procedure that contains common + stuff for attach procedures. Both the name of the device and + spec are given to it. + + pipeattach + devattach() Generic code to attach to a device. + + dev.c:72,78 + devattach creates a new channel, c, which would point to the + root of the device file tree. + + pipeattach + devattach + newchan() Setup a new channel. + + chan.c:67,103 + newchan tries to get a channel structure from a free list found + in chanalloc (a channel allocator). Should the free list be + empty, smalloc is used to allocate a Chan, and it is linked + into the chanalloc list. Channels in use are linked into + chanalloc.list through the link field of Chan; channels not in + use (deallocated) are linked into chanalloc.free through the + next field of Chan. The fid field is given an unique value, + different from other channels in the system. chanalloc.fid + contains a number which is incremented every time a channel is + created. So, every Chan in the kernel has its own fid. The FID, + represents a file for the client in 9P. When requests are + issued from a client to a 9P file server, every file in use by + the client has its own fid. You will learn more about FIDS when + discussing remote files. By now, note how this kernel ensures + that all its channels have different fids; So, file servers + attending this kernel (including the servers within the kernel) + see different fids for different files used by this kernel. + Remaining fields of Chan are reset, but for the reference held + by whoever is allocating the channel. + + pipeattach + devattach + + dev.c:79 + The channel allocated is to be given to the client (the caller + of attach for devpipe). In the same way the client identifies + files by FIDs, the server identifies files by QIDs. A QID is an + unique number within the server, identifying a file on it. The + qid field of Chan holds the QID for the file pointed to by the + channel. In this case, it is a directory and the convention is + that directory QIDs have the CHDIR bit set. The path of the Qid + is just CHDIR and its vers is zero. + dev.c:80 + The type in the channel is used to index back to devtab and + obtain 9P procedures for this channel. The index is the one for + tc (|) as found by devno. + dev.c:81,82 + The name for the channel is an absolute path for the file + serviced. By now, it is the name for the device root (#| in + this case). + dev.c:83 + The channel returned. All this code in devattach is the typical + work that all attach routines must do. dev is giving the author + a means to share that code. + + pipeattach + + devpipe.c:62,64 + The channel is setup, but devpipe has to do its job. First, + allocate a Pipe structure. (exhausted raises an error with the + appropriate message, see proc.c:1147,1154). + devpipe.c:67,77 + Pipes have two queues because they are bidirectional, unlike + UNIX pipes. qopen creates a new queue, using the configured + size. Queues are discussed later. + devpipe.c:79,83 + path relates to the path field in the QID. It identifies a file + within the server. In this case, a pipe has two files (data and + data1). To give each file an unique path value, the author + increments pipealloc.path every time a pipe is created. + NETQID()Builds a Qid.path from two values. The macro NETQID + takes two values and builds a Qid.path field. The reason for + that is that it is handy to use Qid.path as two fields, one + specifying the file and another specifying the file type. In + this case, Qdir is used as the type for the pipe root + directory, Qdata0 and Qdata1 are used for the data files for + the pipe. By looking at this tiny field within Qid.path, + devpipe knows which kind of file is being used. Regarding the + other little field in Qid.path, specifying which file is the + QID for, the value is set to 2 x path; even values refer to one + data file for the pipe, odd values to the other. Nevertheless, + in this case the CHDIR bit is kept set (this is the root + directory for this pipe). + devpipe.c:84,85 + The aux field of Chan is provided to let the drivers place + there their state for the channel. In this case, the pipe + structure is linked there. As there are no multiple ``pipe + devices'', the device number is set to zero (another + alternative would have been to use dev to multiplex the device + among multiple pipes). + devpipe.c:86 + So, after calling attach for the device, the client (the + kernel) has a channel to the root of the specified file tree in + the server. + + Navigating + + Once the client has a channel to the root directory, the client can + move it through the file hierarchy by issuing walk requests. More + precisely, the client would clone the channel (to keep the channel to + the root intact) and then walk on the clone. + + By using walk to move a file descriptor (a channel, with a fid) to + point to a different file in the server (to change its qid), + navigation of paths can be done in a more simple way that it is done + by protocols such like NFS [[232]16] or RFS [[233]14]. For instance, + there is no need to read directory entries, nor to check that an entry + is there in the client, nor to open/close intermediate directories. + Just one walk per path component suffices. + + pipeclone() Clones a channel for the pipe device. + + devpipe.c:89,90 + pipeclone is the clone procedure for the pipe device. The + kernel calls clone when it wants a copy of a channel it has. + The procedure is supplied the original channel (c), and the new + (clone wannabe) channel (nc). It is expected to return the + cloned channel unless there are errors. + devpipe.c:94,95 + Recover the state for this channel (the Pipe structure), and + call the generic devclone procedure provided by dev.c, which + contains common stuff done almost by every clone procedure. + + pipeclone + devclone() Generic clone procedure for devices. + + dev.c:86,110 + To get a clone, first ensure that the channel is not open. + Should it be open, somebody could be doing I/O to the file and + it is not polite to mess up with the file in the mean time. As + the client in this case is the kernel, if the channel was open, + it would be a kernel bug; hence the panic. + Later, if the caller of clone did not bother to allocate a + Chan, devclone does the work. All fields are copied; but see + how a new reference is added for any Mhead structure pointed to + by mh. Besides, Chan links are kept untouched and the clone has + no name. The reason for not cloning the name is that it is + likely to be computed differently by the caller. + + pipeclone + + devpipe.c:96,109 + The pipe is locked, and a new reference accounted. (Another + client is using this pipe). By the way, the if would never be + executed because of the panic in devclone. + + pipewalk() Walks on a pipe device file. + + devpipe.c:147,151 + After cloning the channel to the root, the kernel would + probably walk on it to make it point to a different file. As + this is almost the same processing for all devices, devwalk + does all the job. The only thing devwalk has to know is how to + obtain directory entries for the file being walked. You will + understand this right now, while reading devwalk. + + pipewalk + devwalk() Generic walk procedure for devices. + + dev.c:113 + devwalk is a generic walk routine. It performs a walk to name + on the c channel. However, it does not know what is the file + hierarchy serviced by the device. Therefore, it needs some help + from the device to learn what names it is servicing. The help + comes in the form of an array of Dirtab entries, and a Devgen + procedure. + portdat.h:197,203 + A Dirtab entry contains the name and the Qid for a file. That + is mostly what is needed to scan directory entries. Besides, + file length and permissions are found here too. The directory + table supplied to devwalk is a fake one, file attributes are + not kept there; they are kept wherever the file server wants. + Of course, it is convenient to use an array of Dirtab entries + for directories, but it is not a must. + portdat.h:43 + A Devgen procedure is an iterator for directory entries. It + receives an array of Dirtab entries, and an index for a file + (third parameter). It is expected to both update the channel to + point to the i-th file in the directory table, and to fill up a + Dir structure with attributes for the file. + dev.c:118 + Back to devwalk, it first ensures that c represents a + directory. isdir raises an error when the QID has not the CHDIR + bit set. + dev.c.:119,120 + Should the name be ``.'', the walk is already done: c already + points to ``.''. The procedure returns true to indicate + success. + dev.c:121,125 + Should the name be ``..'', it is gen the one who knows how to + walk to it. The channel and the directory table are given to + gen. That is why devwalk received the table, to pass it back to + gen, to let it iterate through the table if needed. In this + case, the ``file index'' supplied to gen is DEVDOTDOT, which is + -1. The convention is that gen should walk to ``..'' when + DEVDOTDOT is given as an index. How to do that, only the device + (pipe in this case) knows. + A Dir structure is supplied to gen. Devgen routines should fill + up the Dir with attributes for the file determined by the file + index. Once gen did its job, the QID is extracted from the just + filled Dir, and updated in the channel. From now on, c points + to the file found by gen. + + pipewalk + devwalk + pipegen() Directory entry iterator for the pipe device. + + devpipe.c:112,122 + In the case of pipe, the Devgen routine is pipegen (see line + :150). Should the index be DEVDOTDOT, the file is the root of + the pipe file tree. Why? Because if file was one of the data + files, the parent is the root; but if file was the root, its + parent is also the root by convention. To fill up a Dir + structure with file attributes (which is a common job for + devices), there is a generic devdir routine. + devdir()Fills up a Dir structure. + dev.c:25,43 + devdir simply receives as parameters the qid, name, length, + owner, and permissions for the file, and puts all that + information into the Dir structure. If the channel has the + CHDIR bit set, the mode field of the Dir structure is kept with + CHDIR set. That is the convention for directories. Access time + and modification time are set accordingly with the current + time. Every time devdir fills up a Dir structure, times are + updated. The group for the file is set to eve by default. Of + course, should the device supplying the file disagree regarding + modification time or group id for the file, it can update the + Dir entry before using it. + dev.c:126,127 + Back to devwalk, the name is neither ``.'' nor ``..''. The + procedure iterates through the given tab to find the file. + Starting with directory index zero, it calls gen with + successive indexes until gen returns -1. + dev.c:128,130 + The convention is that gen should return -1 when the index + given is not valid. In this case, that means that the index is + past the entries in the directory: there are no more entries. + dev.c:131,132 + gen returned zero. That means that the entry does not exit, + however, the index is valid and no major error occurred. So, + continue iteration. + dev.c:133,138 + A valid entry was found and gen filled up dir with attributes + for the file. If the name in dir is the name to walk to, the + procedure uses the QID for the file (dir.qid) as said by gen to + make the channel point to that file. A return of true from + devwalk means that the walk could be done. Iteration is + continued (linear search) until the entry is found. + dev.c:141 + Just safety. Be sure that if you reach this line, a false is + returned to say that devwalk couldn't walk to the file. + devpipe.c:124 + In the case devwalk was called for a file not being ``.'', nor + ``..'', you saw how it called gen (pipegen in this case) to + iterate. In this case, this line is reached, and tab + corresponds to pipedir (see line :150). + devpipe.c:34,39 + pipedir has Dirtab entries for a typical pipe root directory. + It contains two entries (for two files) named ``data'' and + ``data1''. Their Qids are Qdata0 and Qdata1, but note that + these are actually ``file types'' to build the real Qids. Their + mode is 0600, which makes sense for pipe data files, and their + length is set to zero--just to give them a length. + NETID()Extracts the id from a Qid. + devpipe.c:124,126 + NETID takes the Qid's path, and extracts the file id from the + Qid path. Remember that the Qid for a pipe file keeps both the + ``type'' and the ``id'' for the file. Pipe services multiple + file trees (one per pipe), although for its clients, it is + servicing just one. The way pipe has to distinguish among + different pipes is to use the Qid path. Qid.path must be unique + for the three files used for each pipe. Let's revisit how are + Qid paths built: + + The root directory for the n-th pipe has (2 x pipealloc.path; + Qdir) as Qid.path. (cf. line :83). + + The 0th file in the pipe directory has (2 x pipealloc.path; + Qdata0) as Qid.path. (cf. line :124). + + The 1st file in the pipe directory has (1+2 x pipealloc.path; + Qdata1) as Qid.path. (cf. lines :124,126). + Just Qdir and Qdata would suffice to distinguish among + different files; because the ``id'' is different for data0 and + data1. Nevertheless, the author thought it was convenient to + have two types for the data files. + devpipe.c:127,128 + No table given, or index out of range. Return failure. + devpipe.c:129 + tab points now to tab[i], the entry for the i-th file. + devpipe.c:130 + The Pipe state for the file is recovered from the channel. + devpipe.c:131,141 + The index given by devwalk, selected an entry in pipetab. For + each entry (data0/data1) set the file length reported as the + length of the queue associated with the file. qlen returns the + queue length. The default should never execute because the + index was within range and pipetab has two entries. But just in + case something changes, the routine does its best by looking at + the length field for the file in the table. + devpipe.c:142,143 + Fill up the Dir for the file. The length was computed, the name + came from the pipetab entry for the file, as well as + permissions came from there too. Regarding the Qid, vers is set + to zero (the author does not care), and path is set by placing + in it both the Qid.path from the table (actually the file + type!) and the file id. + + Opening pipes + + The client using devpipe, after attaching to it would clone the + channel to the root and walk on it. Once the channel is positioned + into the file of interest, the device open routine is called. This + pattern of usage is common in 9P. open(2) would issue multiple + requests on its own (attach?, clone, walk, open). + + pipeopen() Open procedure for the pipe device. + + devpipe.c:181 + pipeopen is the open routine for devpipe. It receives the + channel being opened and the open mode. It should check that + permissions allow the requested open mode, and prepare the file + for I/O. + devpipe.c:185,192 + A directory is opened (the root), only OREAD mode is allowed. + If the mode is OREAD, it is noted in the channel for further + use (by I/O routines) and the channel is flagged as open. + devpipe.c:194 + An open for a data file. A process is about to read/write one + end of the pipe. p is the Pipe structure for the channel. + devpipe.c:196,203 + The process is going to use one queue. The qref array keeps + reference counts for both queues--because several processes + could open the pipe files, and these files should stay as long + as at least one process is using them. + devpipe.c:206,209 + openmode checks that omode has valid bits on it. The offset is + set to zero so that any read/write would be done at the + beginning of the file. + In fact, the pipe device ignores offsets while servicing reads + and writes because pipes are streams of bytes; as you learned + before, a file server is free to ignore offsets in read/write + requests. The only thing that matters is that the server should + offer a consistent view of its files. + + There is also a generic open routine provided by dev.c. Let's look at + it. + + devopen() Generic open procedure for devices. + + dev.c:212,251 + The gen routine is used to iterate through the directory, + searching for the file being opened. When the entry is found + (same path in Qid), the information found in the Dir structure + filled up by gen is used to check permissions (you know: + ``rwx'' bits for owner, group, others). The routine is doing + nothing but for checking that the open can be done and to + initialize channel flags and mode accordingly. + + devcreate() Generic create procedure for devices. + + devremove() Generic remove procedure for devices. + + devwstat() Generic wstat procedure for devices. + + dev.c:253,257 + pipe uses devcreate as the routine for file creation. It simply + denies permission to create files. + dev.c:287,297 + The same happens for remove and wstat (9P transactions for + removing files and updating attributes). + + Read/Write + + pipewrite() write procedure for the pipe device. + + devpipe.c:306 + pipewrite is the write procedure for devpipe. It mimics the + write(2) system call. + devpipe.c:310,311 + Interrupts should not be disabled. Looks like the author was + bitten by a bug supposedly caused by calling pipewrite with + interrupts disabled, and the author preferred to ensure that + wouldn't happen. + devpipe.c:312,317 + The code could be like it is, without line :314. However, if + CMSG is set in the channel flag, the channel is being used to + talk to a file server servicing a mounted file tree. In this + case, the system would allow the write without posting a note. + Why is this necessary? I think that the write could be done by + a server to reply to a request issued by the client. Now, the + client could have gone and its side of the pipe closed. It is + not fair to kill the server with a note in this case. + devpipe.c:319,336 + qwrite does all the job. Since only OREAD is allowed when + opening a directory, this must be a pipe data file--there is a + panic just in case. By the way, qwrite would block the process + when the queue is full because the reader is slow. + + piperead() Read procedure for the pipe device. + + devpipe.c:265,282 + piperead is the routine called for reading pipe files. Its + interface is like read(2). A generic routine devdirread is used + when the read refers to a directory, and qread is the one doing + the job of reading from a pipe (data file). qread would block + the process is there is nothing to be read in the + queue--because the writer is slow. + A write to data0 writes to q[1], and a read from data1 reads + from q[1]. Thus, what is written to data0 is read from + data1--and the other way around. + + piperead + devdirread() Generic read procedure for the device directories. + + dev.c:185,210 + devdirread is called by piperead to read from the (root) + directory. It reads an integral number of directory entries + each time. Line :191 determines the number of entries (each + DIRLEN bytes) read so far. The loop controls that at most n + bytes are placed in d, and also increments the file index for + gen. Gen fills up dir, and convD2M places dir information into + d in a machine-independent format. + + There are two read/write routines that do not correspond with 9P + requests. They are bread and bwrite. Their purpose is to read and + write blocks of data. By using specialized versions of read and write + that operate on blocks, block I/O can be made more efficient. This is + mainly used by the code in /sys/src/9/ip, which reads/writes blocks of + data received/sent through network devices. + + pipebwrite() Block write procedure for the pipe device. + + devpipe.c:338,368 + The routine should write the block bp. It is like pipewrite, + but qbwrite is used instead of qwrite. The ignored long + argument is an offset--ignored because pipes are FIFO. + + devbwrite() Generic block write procedure for devices. + + dev.c:276,285 + There is a generic devbwrite, which simply adapts the request + to the device write procedure. It also releases the block + (because the block is usually allocated by the source of the + data and deallocated by the drain of the data). Pipe does not + use this procedure. + + pipebread() Block read procedure for the pipe device. + + devpipe.c:284,299 + The routine is like piperead, but uses qbread (not qread) to + read pipe data files. To read the directory, a generic devbread + routine is used. + + devbread() Generic block read procedure for devices. + + dev.c:259,274 + The generic devbread allocates a Block, (releasing it on + errors) and calls the device specific read routine to fill up + the Block. In the case of data files, the block comes from the + queue. This is simply an adaptor to the device read routine. + + Attributes + + pipestat() Stat procedure for the pipe device. Gets file attributes. + + devpipe.c:153,175 + stat is the 9P request to obtain file attributes. wstat allows + to update attribute values, and you now know that this is not + allowed for pipe files. Depending on the Qid, devdir is called + to fill up DIRLEN bytes into db. The structure is the same that + is obtained by reading the directory. Qlen is used to provide + lengths for pipe data files. + + devstat() Generic attribute read procedure for devices. + + dev.c:145,152 + There is a generic devstat routine (not used by pipe). It uses + again a Dirtab and gen to locate the file of interest (same + Qid), and fill up the stat buffer (db) with file attributes as + filled up by gen in dir. + dev.c:153,171 + The index was out of range--which is the case when the channel + points to the directory and not to any file on it. If the Qid + is for a directory, a name is given to it when the channel has + no name, the name is set to ``/'' when the channel name is + ``/'', and if the channel name is not ``/'', the name is set to + the last name after the the last ``/''--i.e. elem is the file + name for the path in the channel name. If it was not a + directory, it is a true error. + dev.c:174,181 + The file was found--Qids match. + +Remote files + + What you have read about file systems works without depending on where + is the file server (where are the files). The kernel (e.g. for Fgrps + and the Pgrps) use channels to represent files being used, and + channels use device operations to implement the functionality needed + to operate on files. Until now, you have seen how all device + operations are called by procedure calls dispatched by the device + type. For remote files, the system works mostly in the same way; the + mnt(3) device implements 9P operations for remote files. + + Initialization + + mntreset() Reset procedure for the mount driver. + + devmnt.c:64,71 + mntreset is the reset procedure for mnt, (:920,939). It is + initializing the mntalloc global, which is an allocator for Mnt + structures (:portdat.h:241,253) used to service remote mounts + (Everything else in mntalloc was set to zero by the loader). + Besides, it calls cinit to initialize a cache (caching is + discussed later). + + There is no init procedure (the generic null one is used). + + Attach + + sysmount + bindmount + mntattach() Attach procedure for the mount driver. + + sysfile.c:652,656 + When mount is used instead of bind, the file tree bound is + serviced behind a file descriptor. That means that 9P + transactions must be issued through its channel to operate on + the file tree. The file server can be a local user process, a + remote user process, or a remote kernel used as a file server; + for the local kernel, it doesn't matter. + Remember that bogus contains a channel to the server, the spec + for the file tree requested, and an indication of whether files + are being cached or not; everything attach needs to get in + touch with the file server and attach to it. The casting is + needed because attach should receive a character string, but + mntattach knows it receives an structure and not a string. + By the way, mount is the only way to attach the mount driver, + as attach is forbidden from ``#M'' paths. Every time you attach + to it, you are attaching to one particular file tree, which is + in fact serviced through the network. + devmnt.c:73,86 + mntattach knows it receives a bogus structure and recovers it + from the pretended string spec. Perhaps a common header (or a + comment saying who else is using the structure) would help to + prevent errors. The c channel corresponds to the file + descriptor being mounted. A 9P server sits at the other end of + the channel. + devmnt.c:88,109 + The mntalloc list is scanned, looking for a entry (m) with the + same channel and a non-zero id. That list contains entries used + to handle each mounted file tree. If such an entry is found, it + is locked and checked to be still valid (ref). The checks for + id and c are repeated because the lock was not held before--to + avoid blocking other processes using the entry. + This loop is an attempt to share the entry (when the same file + is being mounted several times); In this case a reference is + added, and another channel obtained using mntchan. That channel + is the one to be used for the root directory of the file tree. + The procedure mattach issues a 9P attach transaction to the + remote file server, and the CACHE flag is cached in the + channel. You will learn soon how mntchan and mattach work. + Noticed how eqchan is not used to compare channels in mntalloc? + The mount driver is interested on finding an entry with exactly + the same channel structure. + devmnt.c:111,124 + No entry with the same channel, so allocate a new entry from + either the free list of mntalloc or from fresh new memory (the + list used now is mntfree and not list). A new id is allocated + and placed into the Mnt structure created. Can you see how id + is non-zero, and mclose resets it to zero? The id check at :90 + is for safety. + devmnt.c:127,131 + Starting to initialize the Mnt for this mount. The channel used + to talk to the server is recorded in m->c (So the loop used + above to locate an Mnt entry was searching for an entry using + exactly the same 9P connection to the server). The flag CMSG in + the channel states that it is being used to talk to a remote + file server. Although the descriptor supplied by the user was + closed by sysmount, it could be duped, so the flag in the + channel is needed to prevent interferences in the 9P + conversation. + Remember the check at sysfile.c:85,89? When fdtochan is used to + get a channel for a file descriptor, and it is being used by + the mount driver, an error is raised. The mount driver is very + jealous about this channel. That happens only when a true + chkmnt is given to fdto2chan, which happens when sysread9p, + sysread, syswrite9p, syswrite, sseek, and sysfwstat call + fdtochan. Routines reading, writing or updating the state are + forbidden for channels used by the mount driver. + devmnt.c:132,140 + blocksize in m is set to the size for data blocks exchanged + between the mount driver and the file server. Unless the spec + string given to mount was mntblk=n, it is set to MAXFDATA. In + this case spec is set to null, not to disturb the remote + server. Perhaps a writable file for the mount driver would be + better than using the spec string; although the method used by + the author allows different block sizes for different mounts. + devmnt.c:141 + All flags are kept in Mnt flags, but for MCACHE, which is kept + in c. Surprisingly, only the MCACHE flag was set in bogus.flags + by bindmount. Therefore, m->flags is now zero. + devmnt.c:143 + Right now, the channel is still there, but when bindmount + closes it it could go away. This new reference keeps the + channel alive. + devmnt.c:149 + mntchan returns the root channel to be given to the caller. + + sysmount... + mntattach + mntchan() Returns a channel to the mount driver impersonating the + channel to the remote file tree. + + devmnt.c:185,196 + It uses devattach to build a channel for a directory in the #M + device. This is very important. The user thinks it got a + channel to the remote root directory, but he got a channel to + the mount driver instead. The real channel to the server (not + to the root, but to the server) is kept by the mount driver, as + shown in figure [234]5.5. An important thing here is that dev + in the channel built is set with an unique id. This id is used + by the mount driver to recognize that mount entries remain + valid. + + CAPTION: Figure 5.5: The mount driver sits between the server and the + client. The kernel uses channels serviced by the mount driver. + Requests for these channels are implemented speaking 9P with the file + server. + + devmnt.c:150,157 + Note how to cleanup on errors, the ``cclose'' is done. The + recursive call would be to mntclose, which would try to use the + channel to send a Tclunk request. That would make no sense at + this point. But perhaps a channel flag could be added to the + channel so that mntclose would do nothing to a channel not yet + set. + devmnt.c:159 + Starting to speak 9P. mattach performs a 9P Tattach transaction + (client side). If it finishes without raising an error, the + attach RPC has been done, m->c is a channel to the server, and + c is a channel to the mount driver. + devmnt.c:170,177 + If the if is entered, you knows that: + 1. + The channel to the server (m->c) is a channel serviced by + the mount driver. You know this is the case because the + channel to the server has as type the index for the mount + driver in devtab. + 2. + The file server at the other end of the channel is + exportfs. Tricky!!. You know this because in + /sys/src/cmd/exportfs/exportsrv.c:574, exportfs is + clearing the CHDIR bit from the QID for the served + directory (c->qid). That should never happen because the + server is servicing its root directory. However, exportfs + clears that bit, and the mount driver notices that, does + this hack, and repairs the missing bit at line :172. + Perhaps it would be more clean if the Rattach reply could + carry a string of information back to the client--in the + same way the Tattach carries an spec string. Nevertheless, + the protocol and the code are cool, aren't they? + What is the hack doing? Well, the Mnt entry for c is closed, + i.e. not going to behave as if the remote c was mounted through + m->c, which it was. The mntptr (which points to the Mnt + structure) of the client channel (c) is set to point to the one + linked at the server channel. That means that mc->mntptr is + going to be shared among all file trees serviced through mc. + Should this be removed from the kernel, requests for the tree + being mounted would be translated to 9P transactions by this + mount driver, and such transactions would be written to + c->mchan (which is m->c here). Now, to write the 9P request r + to a file serviced by the mount driver, a 9P request s has to + be issued to m->c->mchan, with the r request being the data. + That makes no sense, when you could speak 9P right to the final + server[235]9.1. Hence the hack. This is one of the few places + in the kernel source when a trick which is not clearly exposed + in the system interfaces is being used to prevent the system + from doing a silly thing. + Now, what if mc is remote (serviced by the mount driver), but + the server is not known to be exportfs? The author cannot + assume that the remote server would multiplex its connection + among trees mounted from it, so the best the mount driver can + do is to use 9P requests (to the server of the channel for the + file server) to write 9P requests for the channel (to the file + server). + + sysmount... + mntattach + mattach() Uses a Tattach to attach to the server file tree and + initializes the channel to point to its root directory. + + devmnt.c:198,210 + We still have pending the discussion of mattach, which + allocates an Mntrpc structure to manage the RPC done by the + mount driver. The structure is deallocated on errors. + devmnt.c:8,22 + An Mntrpc contains among other things two Fcall fields (request + and reply) which are the request message and the reply message + for the RPC (see /sys/include/fcall.h). Those are 9P + transactions and replies. + devmnt.c:212,215 + The request is a Tattach. The fid for the Tattach is the fid in + c. Each kernel channel has its own fid (chan.c:81), so there is + no ambiguity regarding which client ``file descriptor'' would + point to the root of the remote tree. The user name is that of + the current process (can user contain null characters?). + devmnt.c:216,218 + Fields ticket and auth are set by authrequest to authenticate + the client to the server. m->c->session is the structure used + to maintain authentication/encryption information for the + session maintained through the channel. m->c is the channel to + the server, which is not c, that is the channel for the client. + The value returned by authrequest is used later by authreply to + check that the reply received for the request is valid and not + a fake one. Should it not pass the security check, authreply + raises an error. mountrpc does the actual work of sending the + request and receiving the reply. + devmnt.c:220,226 + The channel QID is set from the qid in the reply (so that it + really points now to the server root file). Besides c.mchan is + set to the channel for the server, now that it is attached. + Future requests for c would use its mchan field to issue RPCs. + mquid also holds the QID for server's root directory. + + RPCs + + But how is the RPC done? + + mountrpc() Performs an RPC for a mounted file tree. + + devmnt.c:607,631 + mountrpc issues the request in r and receives any reply for it. + It is actually a wrapper for mountio, which does the job. First + it sets the reply tag and type to poisoned values, in case + anyone checks them before the reply arrives. Then mountio does + the job. Finally, the reply type is analyzed: an Rerror causes + a raise of the error reported (note that strings are + portable!); an Rflush interrupts the request; an Rattach + (Tattach+1) returns without errors; and everything else should + not happen. + + mountrpc + mountio() Performs I/O to issue a 9P request and receive the reply. + + devmnt.c:634 + m is used to manage the remote mount point and r is the RPC to + be done. + devmnt.c:638,646 + The while restores the error label after an error is raised. + The loop body executes each error, (unless the error is + re-raised by nexerror). Ignore this by now. + devmnt.c:648,652 + All RPCs r keep in r->m a link to the mount entry they are for. + Besides, the mount entry keeps in queue a lists of RPCs being + done. The list is done through the list field of Mntrpc, as + shown in figure [236]5.6. + + CAPTION: Figure 5.6: Mntrpcs are used to maintain the state for + ongoing RPCs for a mounted file tree, represented by Mnt. + + devmnt.c:655,657 + The server and client machines could have different + architectures. convS2M packs the request into the buffer at + rpc, in a machine independent format. The buffer at rpc is + allocated by mntralloc with MAXRPC bytes, which is the limit + for a message. The panic shouldn't happen, but just in case 9P + is changed and convS2M is not updated, let the author know. + devmnt.c:658,664 + Some times, the channel used to talk to the server is local + (e.g. the file /srv/kfs is a channel to the local KFS server). + In this case, the channel m->c is not serviced by the mount + driver and the request would be serviced by the device specific + write (:662). But some other times, the channel used to talk to + the server may correspond to a file which is mounted using the + mount driver (e.g. /n/remote/tmp/pipe). If the channel to the + server corresponds to a mount driver channel, the rpc buffer is + written using mnt9prdwr. For mount driver channels, the device + specific write is not called. Why does the author do so? The + answer relates to message sizes as you will learn later. + devmnt.c:665,666 + stime keeps the time when the request was sent. The Tattach is + now traveling to the file server and stime has the current + time. n bytes were written to the server. + devmnt.c:669,681 + The routine did set rip (reader in progress) to nil before. So, + unless other process changed the state in m, rip is nil and the + loop is broken at line :672. + If no process is reading a reply for this mount point, the + current process is the one reading. From now on, other callers + of mountio would notice that rip is not nil, and enter the + loop. They will stay in the loop until the process in rip stops + reading. Instead of spinning all the time, sleep is used to + block them until rpcattn says that r->done or there is no + r->rip process. So, processes awaiting for RPC replies sleep + and read one at a time. r->r is the Rendez used to sleep. In + figure [237]5.6 you can see a process servicing reads while + another is awaiting its reply. + devmnt.c:683,686 + done is set to false by mntralloc, it is set to true when the + RPC should be considered to be done. So, unless the RPC + finished, call mntrpcread and mountmux. + The first process doing an RPC on m is calling mntrpcread and + mountmux, and remaining processes are looping/sleeping waiting + for their replies. + + mountrpc + mountio + mntrpcread() Reads from the 9P channel until gets a reply message. + + devmnt.c:692,711 + mntrpcread loops until a valid 9P reply message is read from + the channel to the server. As it happen with write, mnt9prdwr + or the device specific read is called depending on whether the + channel to the server is a mount driver channel or not. Any + reply message is read into the rpc buffer of the Mntrpc--now + that the request was sent, the buffer can be reused for the + reply. Lines :708,709 call convM2S to convert the machine + independent reply found at rpc into a reply structure. If the + conversion fails (the reply message is not a 9P message), the + loop continues and another reply is read. The mount driver is + ignoring invalid messages. Once a valid reply is read, the + return executes and mountmux is called. + + mountrpc + mountio + mountmux() Multiplexes the 9P channel among multiple RPCs. + + devmnt.c:728,763 + mountmux multiplexes the channel to the server among multiple + processes doing RPCs. m->queue was a list of ongoing RPCs + through m. The list is searched for an RPC whose request had + the tag found in the reply just read. If you read intro(5), you + know that 9P replies carry the tag of their request message. If + no RPC in the list has a request with the same tag of the + reply, no one is waiting for this reply. In this case the + routine completes without doing anything, and mntrpcread would + run again and read another reply message into the rpc buffer. + The reply (without a request) has been ignored. + devmnt.c:740,749 + If the reply message is for the RPC being done by the process + that called mountmux (the first reader), r would be equal to q + (the node with matching tag). If r and q differ, the reply is + for some other process. In this case, the rpc buffer for the + current process is exchanged with the rpc buffer of the RPC + replied. Besides, the reply structure for the process replied + is set to point to the reply structure containing the unpacked + reply. In this case the current Mntrpc has an empty buffer in + rpc to receive another reply. + This buffer exchange is very important to avoid fragmentation + and also to improve performance. The author tries not to + allocate/deallocate buffers unless it is really necessary. + devmnt.c:750,759 + done is set in the Mntrpc whose reply arrived. If it was + another process, line :757 would awake it; then it would notice + that done is true and pick up its reply. If it was the current + process, our done is set. + + mountrpc + mountio + + devmnt.c:683,686 + If the reply was for us, the loop breaks. In any other case, + the process keeps on reading from the channel and servicing + replies to waiting processes (including himself). + Can you see how when one process has to do some work for + himself it tries to do useful work for others too? It would be + silly if all processes were spinning trying to read from the + channel. + devmnt.c:675,679 + Another processes awaken would check its RPC done field. Should + it be true, the process has a reply in r, and mountio returns. + Should it be false, the process re-enters the loop and sleeps + again if there is another process reading from the channel + (rip). If there is no other process reading (the reader got its + reply and its mountio returned), the process breaks the loop + and becomes the channel reader, servicing replies for others. + devmnt.c:687 + Unless the reply to the RPC was serviced by another process (a + channel reader on our behalf), mntgate is called. + + mountrpc + mountio + mntgate() Elects a new reader for the channel. + + devmnt.c:713,726 + What happens is that we were the channel reader (servicing + requests to others), but our reply finally arrived and we are + leaving mountio. Another process must read the channel, if + other RPCs are pending. mntgate awakes the first process found + with a pending RPCs in the list. It becomes the reader (see + line :719). + + mountrpc + mountio + + devmnt.c:638,646 + Should an error occur during all this time, the process calls + mntgate (if it's the reader) to let other process take its + place, because due to the error the current process can abort + the RPC and return. But how are errors handled? + Consider that an Emountrpc is raised at line :663, after + getting back to :638, the loop body is entered. The error is + not Eintr, therefore an mntflushalloc is done on r. + mntflushalloc()Allocates a Flush request for a failed RPC. + devmnt.c:769,784 + mntflushalloc allocates an Mntrpc for a Tflush message, and + links r (the RPC suffering the error) in the flushed field of + the Tflush Mntrpc. A list of flushed requests is being built + through the flushed field of Mntrpc. The field oldtag of the + Tflush is set to the tag of the RPC failing. If the failing + request was a Tflush, the RPC failing was the one that caused + the Tflush. + devmnt.c:646 + So on a error, mountio runs again with the request being a + Tflush. If an RPC fails, a Tflush is sent to the server. If the + Tflush fails, another Tflush is sent. All those Tflush are for + the (oldtag) RPC that failed. If the Tflush can proceed (the + server is running and serviced the Tflush), either :677 or :689 + call mntflushfree. + mntflushfree()Releases flushed RPCs + devmnt.c:793,808 + mntflushfree removes from the RPC queue (mntqrm) any undone RPC + flushed by r (a Tflush), and releases the structure (mntfree). + The reply is set to Rflush, because mntqrm sets the done field + of the RPC to true--the affected process might check done in + the mean time, and it should notice that no actual reply + arrived. + devmnt.c:641,644 + Should the first Tflush abort with an error raised, another is + sent. New Tflush requests are sent until an Rflush is received + or an Eintr error is raised. In the RPC is interrupted, any + flush request is released and the interrupt error re-raised. + + mountrpc + + devmnt.c:616,631 + The RPC finished (note the Rflush case). Only if the + transaction got a non-error reply, mountrpc finishes without + raising an error. + + Using remote files + + After an attach is successfully done, the client has a channel to the + mount driver, which from the client point of view is pointing to the + root directory of the remote file server. Typically, the next things + done by the client are clones and walks to navigate the mounted tree. + + cclone + mntclone() Clones a channel for a remote file. + + devmnt.c:228,229 + The clone done by the client is done by cclone, which calls to + the device specific clone: mntclone. + mntchk()Checks that the file is still mounted. + devmnt.c:235 + mntchk (devmnt.c:883,897) checks that c->mntptr points to an + Mnt whose id is not 0 (still allocated) and is less than the + dev in the channel. Line :192 sets the ``device number'' for + the channel to the Mnt id. If the check fails, either the Mnt + has been released (id set to zero at :402) or it has been both + released and reallocated for a different mount (whose id would + be bigger than the copy kept in the channel dev. + All mount driver device operations for a remote channel use + mntck to check that the connection is still alive. + devmnt.c:236 + The routine allocates the RPC structure, as it was done for + mntattach. + devmnt.c:237,240 + clone can be called with or without the ``cloned channel'', so + better be sure that mntclone has an nc. + devmnt.c:241,246 + Release the RPC on errors, and nc if it was allocated by + mntclone. + devmnt.c:248,251 + A Tclone is sent and the reply is received (or an error + raised). newfid is the FID for the clone, which was set when + the clone channel was created. devclone()Generic procedure for + cloning channels. + devmnt.c:253 + devclone does the job of copying the state in c to nc. The RPC + was just to let the file server know that newfid should be + understood as another ``file descriptor'' pointing to the file + where fid was pointing to. + devmnt.c:254,255 + What? That was done by devclone, but it does not hurt. As + another channel (nc) is using the Mnt for c, count one more + reference. Even if unmount is used, Mnt will not go away until + its reference count gets down to zero--because all channels + going through it have been closed. + devmnt.c:257 + To keep the compiler happy--alloc is used even if no error is + raised. Otherwise, the compiler might issue a warning. + + walk + mntwalk() Generic walk procedure for devices. + + devmnt.c:263,285 + Regarding walks, you now how the global namec and walk routines + work. When the device specific walk routine is called, mntwalk + executes. It issues a Twalk RPC, so that c would refer to the + file named name, within the directory pointed to by c-i.e. + within the directory for c->fid. After the Twalk is sent, and + an Rwalk is received, the QID from the Rwalk is set in the + channel. This kernel now knows that c points to the file + represented by the new QID, and the server knows that c->fid is + now pointing to that file. + + File attributes + + mntstat() Stat procedure for remote files. + + devmnt.c:287,307 + mntstat issues a Tstat transaction when an stat operation is + done on the remote file. It works like clone or walk. The + different thing is the call to mntdirfix with both the + attributes read for the file, and the channel for the file. + mntdirfix()Fix attributes for remote files. + devmnt.c:900,909 + mntdirfix changes some attributes read, to reflect that the + file is serviced by the mount driver. In particular, it writes + in the last two `shorts' of dirbuf, the letter for the device + (M) and the mount id. Although the stat(5) manual page states + that those two shorts are for kernel use, they are used + (/sys/src/libc/9sys/convM2D.c) to report the device type and + device number for the file (Can you guess where does the ``M'' + listed by ls come from?) + + mntwstat() Wstat procedure for remote files. + + devmnt.c:439,457 + mntwstat is also similar, but it issues a Twstat instead. + + Open and close + + mntopen() Open procedure for remote files. + + devmnt.c:309,337 + The QID is set from that in the reply (because the server could + have created a new file), the offset is reset to zero, and the + channel is flagged to be open. Lines :333,334 report to the + cache that the file is in use for I/O--if the file is to be + cached. That is to give the cache an opportunity to invalidate + old versions for the file and do other things. + + cclose + mntclose() Close procedure for remote files. + + devmnt.c:427,431 + This is the device operation called by cclose when the last + reference to the channel goes away. It issues a Tclunk request + to let the server know that the fid is no longer in use. + + cclose + mntclose + mntclunk() Issues a clunk RPC. + + devmnt.c:368,388 + There is no close in 9P. mntclunk issues a Tclunk, or a Tremove + is the file is being removed (devmnt.c:433,437). Seems that + mntclunk is being reused to issue both kinds of transactions. + + Read and write + + The actual device specific routines are mntread and mntwrite, but if + you look at read9p(2), you will notice that to encapsulate 9P on 9P + without problems because of the maximum message limit, read9p and + write9p have to be used to write 9P requests to a file serviced + through 9P. + + sysread... + mntread() Read procedure for remote files. + + devmnt.c:466,500 + A read to a file serviced by mount driver leads to mntread as + the device specific read procedure. cache is set if the channel + has the CCACHE bit set (i.e. if it comes from a tree mounted + with MCACHE) and it is not a directory. Caching file contents + is one thing, but caching directory entries is one of the + things that makes distributed file systems complicated (race + conditions, too much locking for clients using entries etc). + Plan 9 sidesteps that problem by not caching directories. After + all, the design of 9P (i.e. walk) allows the client to walk + paths without needing to cache directory entries. This is also + good in that if the file server changes its mind regarding + which files exist, its clients would know without any problem. + If the file is cached, the read is serviced from the cache by + cread. If the bytes cached (nc) do not suffice to satisfy the + read request (n bytes), a Tread is issued to read the bytes not + kept in the cache. cupdate updates later the cache with the + bytes read by the Tread. Besides, the device type and number + for any directory entries read are set by calls to mntdirfix. + By making directory reads return an integral number of + directory entries, processing of entries is greatly simplified. + Compare the routine with the one needed in case Tread could + return any number of bytes. + + syswrite... + mntwrite() Write procedure for remote files. + + devmnt.c:508,512 + A write is serviced by issuing a Twrite request. Both Treads + and Twrites are serviced by mntrdwr. The author reuses code as + much as he can: reads and writes have much in common. + + syswrite... + mntwrite + mntrdwr() Issues Tread/Twrite RPCs. + + devmnt.c:552,565 + either a Tread or a Twrite (type) on the channel. Note the + checks for the mount point and caching. + devmnt.c:566,602 + The routine loops sending Treads/Twrites, with the buffer for + the request being the buf given as a parameter. cnt is set with + the number of bytes read/writen. + devmnt.c:576,582 + One fine reason for looping. The caller could want to + read/write more than blocksize bytes (MAXFDATA), in which case + multiple Tread/Twrite must be issued for at most blocksize + bytes each. + devmnt.c:584,587 + Will not read (write?) more bytes than requested. + devmnt.c:589,592 + The only difference between read and write; not enough to + justify two different routines. For reads, copy the data read + into the user buffer. For writes, let the cache know of the + bytes writen to the file. + devmnt.c:596,601 + Next time, read/write past the bytes read/writen. The procedure + adjusts the file offset and number of bytes processed (cnt). + The loop continues until n is zero, which means that cnt is the + initial value of n; or until read/write could not service as + many bytes as requested (no more bytes to read/disk full); or + until a note has been posted for the process. + + syswrite9p + mntwrite9p + mnt9prdwr() Reads/Writes 9P requests (encapsulated in 9P). + + devmnt.c:515 + mnt9prdwr implements both mntread9p (devmnt.c:459,463) and + mntwrite9p (:502,506); it is also used by mntrpcread and + mountio to read and write 9P requests. It is not a mnt device + specific procedure, but a generic 9P tool. + The sysread9p and syswrite9p system calls call mntread9p and + mntwrite9p to do the work when the channel to the file server + is serviced by the mount driver. Otherwise, sysread9p + (sysfile.c:335,372 and syswrite9p (sysfile.c:414,441) call the + device read/write procedure or unionread as they should. + devmnt.c:521,525 + At most MAXRPC-32 bytes read/written (and for write this limit + should not be ever reached). The Tread (or Twrite) is sent as + usually and the reply processed as usually too. So, what's the + difference with respect to mntrdwr? First, no cache is ever + used (would you cache a connection to a server?); Second, the + routine does not loop, and it reads/writes a single chunk of at + most MAXRPC bytes. The routines transmit a 9P message verbatim. + If you compare sysread9p and sysread, you will notice how in no + case the the mount driver device specific procedure is called + to read the 9P request, and the same happens to syswrite9p and + syswrite. Besides, note that mntrdwr uses messages of blocksize + length (which can be much lower than MAXRPC) while mnt9prdwr + uses messages of at most MAXRPC bytes, independently of the + configured blocksize. Perhaps both routines could be unified + into a single one, but the author preferred to keep them + separate. + + Caching + + In the third edition of Plan 9, authors considered that it was + important (due to performance reasons) to cache files mounted from + remote file servers. That can save many 9P transactions by satisfying + reads and writes from a local cache in the client kernel. The + implementation stands at cache.c. + + In this section, you will be reading the code related to caching in + the kernel. Besides a kernel cache, Plan 9 has a user-level program + called cfs (see cfs(4)) that caches remote files. cfs is started at + boot time and services reads from a local cache (kept on a local + disk). This is interesting when it is better to read files from the + local disk than to read them from the network. Surprisingly, this is + not always the case, because when you have a fast network and a fast + file server node (plenty of memory) it can be much faster to read a + file from the network than reading it from the slow local disk. + Nevertheless, for slow network connections the performance improvement + can be dramatic. + + Regarding the kernel, cfs is just a ``remote'' file server. Therefore, + in what follows, I focus just on the caching done by the kernel. + Figure [238]5.7 shows how all the pieces fit together. + + CAPTION: Figure 5.7: Caching remote files. The best thing (1) is to + keep them cached in kernel memory. The next best thing (for slow + connections) is to keep them cached in a local disk (2). Finally, you + always have the network (3). + + chandevreset + mntreset + cinit() Initializes the kernel cache for remote files. + + cache.c:105,127 + You already know that cinit is called by mntreset at boot time. + It initializes the cache global (:39,47) by allocating NFILE + Mntcache entries, double linking them through next and prev + fields using head and tail as the list header. xalloc is used, + and not malloc. When the machine is plenty of memory (more than + 200MB), the maximum number of bytes to cache in a file + (maxcache) is not set to its usual value (MAXCACHE) but to + 10 times more. + +Caching a new file + + sysopen... + mntopen + copen() Prepares a cached remote file for I/O. + + cache.c:210 + The cache starts to work when copen is called. A copen is meant + to prepare the cache for I/O on a file. It is the mount driver + that calls copen when a remote file is being opened or created + (i.e. before doing any I/O on it). By ``remote'' you should + understand ``not in kernel'' now. Even if the file is serviced + locally by a user program, caching it can avoid unnecessary + data copies and context switches. + The cache does not keep copies of intermediate directories used + to walk to the files of interest. Therefore, cache memory is + used just for files being really used. The user controls which + file systems should be cached (i.e. by means of the MCACHE + flag). + cache.c:216,217 + The mount driver checked the CHDIR bit, but the author ensures + that it is innocuous to call copen on a directory. + Plan 9 does not cache file attributes (walk works well enough). + The alternative would be to cache attributes (including + directory contents) and perform walks locally. However, this + would require that all contents of all intermediate directories + walked be sent to the client. Moreover, this would introduce + severe coherency problems (all file server clients should see + the same set of files, with the same attributes). + cache.c:219,223 + Entries in the cache are linked at NHASH hash buckets (:40,47). + The hash function is a modulus on the channel qid.path. + Multiple entries are linked at the bucket through the hash link + of Mntcache. The routine searches the hash bucket for any entry + for the same file. The check is done using qid.path, dev, and + type fields of Mntcache, which keep the qid.path, dev and type + fields for the cached file. If multiple channels point to the + same file, these fields would be same in all of them. vers is + not compared. The cached file could be a previous version of + the file, but it would still be its cache entry. + Figure [239]5.8 shows the structures involved. + + CAPTION: Figure 5.8: Caching for remote files. Mntcache structures are + caching one file each. They are kept in an LRU list and rely on + Extents, which use kernel pages, to cache file contents. + + cache.c:224 + An entry found. The mcp (Mntcache pointer) of the channel is + set to point to its cache entry; read and write procedures can + avoid scanning the whole cache to lookup the Mntcache for the + channel. The assignment is needed because multiple channels + could be opened for the same file. All of their mcp fields + would have the same value after copen. + ctail()Sets an Mntcache at the tail of the LRU. + cache.c:225 + Remember that Mntcaches are double linked on a queue starting + at Mntcache.head and .tail. ctail (:182,207) unlinks the node + given from the list (hence the double links) and links the node + at the tail. By doing so, the file last opened is found last on + the list of entries. Probably, the author would reclaim cache + entries starting from the head of the list. It makes sense not + to reclaim this entry because it has been just opened, and is + likely to be needed soon. + cache.c:226,235 + Once that the entry is placed on the list, the lock for the + cache can be released. If vers (also copied in the Mntcache) is + older than it is in the channel, the file has changed and cache + contents are useless. The vers field in Mntcache is updated + whenever new file contents are written through the cache. The + routine cnodata does the job of disposing any (useless) cached + bytes for the file--more soon. The cache is prepared to service + the file. + One quick note about vers. vers in the channel is updated + whenever 9P requests carry back a QID for the file. If other + nodes are using the (cached) file too, it could be that the + cache contents are actually out of date, and the kernel + wouldn't notice. The actual responsible for this lack of + coherence is the user, who mounted the file with caching, or + the server owner, who did not set the OCHEXCL bit in the file + being cached. To maintain a set of distributed caches in + coherent state is just too expensive and complex. The tools the + author gives you allow you both to cache files and to use them + coherently, you only have to use the tools in the right way. + Other distributed file systems tried to do distributed caching, + but they either supplied ``session coherence'' (i.e. only after + a close can others see our changes) or were so complex that a + node failure could bring the whole system down. + cache.c:239,248 + No Mntcache entry found for the file. Should use one of the + existing entries to service the file. The hash bucket for the + head of the Mntcache list is searched. If the head entry is + there, it is removed from the hash list. You should note here + that it is the head of the list the one reused. A file could be + loosing its entry in favor of another one. You should note also + that used entries are linked into the hash bucket (hashing with + the QID). + cache.c:250,252 + The old file (if the entry was used) is forgotten. Now this + entry is for the file represented by the channel. + cache.c:254,257 + The entry is linked into the hash bucket for the c channel, + where it belongs now. ctail is used to move the entry to the + tail of the list, as it has just been used. By allocating from + the head, and moving the entry to the tail whenever it is used, + the author is doing a ``Least recently used'' policy to + maintain cache entries. + cache.c:259,269 + Finally, mcp in the channel is set to the entry allocated for + it and any extent linked into the Mntcache (which would be for + the previously cached file) is released. The cache is unlocked + as soon as no pointer is being moved. The entry has to be kept + locked because extents are being released. + + Extents + + copen + cnodata() Invalidates a cache entry. + + cache.c:166,180 + We had pending the discussion of cnodata. The cache does not + cache bytes, but Extents. (cache.c:16,24). An Extent is a len + bytes portion of the file starting at start offset. All cached + extents for a file are linked through their next pointers, at + the list field of the file's Mntcache. The routine is simply + calling extentfree for the whole list. + extentfree()Deallocates an extent. + cache.c:63,71 + extentfree is releasing the extent. It links the extent into + the head of the extent cache (:50,56). The next time the cache + allocates Extents, the ones from the ecache.head will be + reused. extentalloc()Allocates an extent. + cache.c:73,102 + Initially, the ecache has all its fields set to zero by the + kernel loader; the first time extentalloc is called, it would + notice that the free list (head) is empty, and allocate NEXTENT + extents. + The cache could do the same for the Mntcache entries; or + alternatively, cinit could also contain lines :81,92. One + reason the author could had to initialize extents this way, is + that the user could never use the MCACHE flag for mounts, and + the cache would never be used. But in that case, there is no + need to keep Mntcache entries allocated. + The actual allocation of the extent is done at lines :95,98. + The routine clears the contents of e (security first!). + +Using the cached file + + The next time the cache works, is when mntread calls cread + (devmnt.c:480) and cupdate (devmnt.c:489), and when mntrdwr calls + cwrite for Twrites (devmnt.c:592). + + Reading from the cache + + sysread + mntread + cread() Services a read from the cache. + + cache.c:288 + You know cread is called to read from the cache instead of + using Treads, when feasible--i.e. when the cache has the bytes. + cache.c:297,298 + off is the file offset where to read and len is the number of + bytes. As only the initial maxcache bytes for the file are + cached, cread ignores any request which passes that limit. + Perhaps the condition is too strong, as off could be below + maxcache and part of the len bytes could be cached. + cache.c:300,302 + Either the mcp has the Mntcache to use, or it is nil--stating + that the channel is not being cached. + cdev()Checks that the cache is still for a channel. + cache.c:304,308 + cdev tells whether the entry is valid as a cache for the given + channel (:273,285). It checks path, dev, type, and vers. + Mntcache entries are stolen from the head of the LRU list. The + entry for the channel could have been reused and the only way + to know is to check these fields. (The alternative would be to + iterate through channels, or to link channels using an Mntcache + entry, which is more complex and inefficient). + When an Mntcache is stolen from a channel, that channel remains + uncached--cread (also cupdate and cwrite) ignores it and does + not allocate another Mntcache for it. This is a fine way of + preventing Mntcache entries from being stolen repeatedly due to + reads and writes; that only happens during open. + cache.c:310,321 + The list of Entents for the entry is searched for an entry + containing offset (the first byte read). If there is no such + entry, there is nothing of interest cached for the file. + Extents are sorted accordingly with their addresses + ([start/start+len]). + cache.c:323,361 + Starting to use the cache. cread copies bytes from the Extent + located previously (and following ones if necessary) to the + read buffer. total keeps the total number of bytes serviced, + and len maintains the number of bytes yet to be read. + cache.c:324,331 + Each extent uses at most on page to cache file contents. By + using extents instead of pages in Mntcache, the author can keep + track of byte ranges cached for the file. cpage returns the + Page structure for the extent; should it be nil, the extent + does not have memory caching anything and it is removed from + the list (:327) and released. If an extent did loose its page, + the author assumes that any following one (which could contain + cached bytes) has lost its page too. cpage (cache.c:156,164) + just calls lookpage, which is discussed in the memory + management chapter, to lookup the page used as a cache. + The author uses just pages to do caching. Other systems use + several kinds of caches, which in the end, are using pages too. + By using always pages to do caching, actual caching has to be + implemented just once: by caching pages. In the next chapter + you will see how Images (used to keep a memory image of used + files) are using pages too, as segments do. Simple, isn't it? + If you don't agree, try to think how caching could work if the + author did cache ``files'' or ``blocks'' instead of pages + coming from files--you should exercise open/close/read/write + requests on this imaginary cache hierarchy. + cache.c:333,336 + The extent has a page caching some bytes. o is set as the + offset within the extent corresponding to the offset in the + file. l determines how many bytes can be read from the Extent. + cache.c:338,344 + kmap ensures that the cached page can be read from the kernel + (you know it's a nop here), and the author ensures that the + page is released and kunmapped (nop) on errors. cpage returns a + page, and it should not be unmapped while the routine is using + it; putpage lets the memory system know that the page is no + longer in use by the caller of cpage. See lines :348,351 too. + cache.c:346 + Here is the read from the cache! memory copied from the extent + to the read buffer. + cache.c:353,361 + After memory has been copied from the extent, if the read + request needs more bytes, go to the next extent which has the + following bytes. Note the check: there could be no next extent, + or it could be that the next extent is not caching the bytes + right after the current one (it does not start at offset, but + starts later). In this case there is nothing more cached for + this read. + When some of the following bytes are missing from the cache, + the author refuses to check if any posterior extent would have + bytes within the range read. The benefit may not be worth the + effort. Besides, the caller of cread assumes that it reads a + contiguous initial portion of the region being read; the + trailing portion is read by the mount driver. Any change here + to bypass holes in the cache would require changes in the mount + driver too. The author assumption is used to keep the code + simple, yet provides effective caching (You should take into + account that most applications read entire file contents). + + Updating the cache + + sysread + mntread + cupdate() Updates cache contents. + + cache.c:460,478 + cupdate is called by the mount driver after reading the + trailing portion of the file region being read--that portion + was missing from the cache and cupdate adds it to the cache. + Any further cread would now find that portion too. Initial + checks are like those in cread, for the same reasons. + cache.c:483,489 + The Extent list for the Mntcache entry is to be kept sorted by + file offsets cached. + cache.c:491,499 + f is the extent starting past the offset added to the cache (if + any), and eblock the end of the new portion cached. So, if the + portion read overlaps with the next extent, only bytes missing + up to the start of that extent must be inserted (remember that + the author stops at the first hole while reading). If all the + portion being updated in the cache overlaps with the next + extent, just forget about the update. This could happen because + several processes could read from the channel, find the hole in + the cache, issue Treads for the file, and try to update the + cache. Only one copy should be added. + cache.c:501,509 + These lines are the special case for insertion when it has to + be done at the head of the extent list. cchain does the actual + work of allocating extents and memory to cache the bytes + updated. It returns a pair of pointers to the first (returned + value) and last (in tail) nodes linked as a (sub)list to be + inserted. When inserting at the head, the next of the last new + node is the first node of the list (f, although using m->list + would be more clear); and the new head node is the first of the + list allocated. cchain is discussed later. + cache.c:511,522 + Not the first node, so cupdate ensures that this does not + overlap with any previous node (as it was done before with any + posterior node). + cache.c:524,540 + ee was set as the end of the previous extent (:512). If offset + is precisely that, the updated portion is contiguous to the + previous extent. If the previous extent length is less than the + page size, it could accommodate up to BY2PG - p->len bytes for + the updated portion. Do so. It could be the case that it could + accommodate all the updated portion, in which case there is no + need to allocate more extents (cchain would be never called). + cpgmove is just a memmove that gets and kmaps the extent page + while copying memory (:435,457). + cache.c:524,547 + As it was done with the head, cchain allocates more extents to + accommodate bytes updated. This means that the previous node + was either not contiguous or not with enough space on its page + to cache all the updated bytes. How does cchain work? + + sysread... + cupdate + cchain() Creates a chain of extents to update the cache. + + cache.c:368,381 + You know its interface. It loops creating extents to keep more + bytes from buf until the number of bytes yet to be updated is + zero. It is not considered a serious problem if no new extent + can be allocated; in this case, nothing more is cached. When + all NEXTENT extents are in use, nothing more is cached. + cache.c:383,387 + An extent caches at most one page. auxpage returns a page that + can be used by the caller. Should no page be available, there + is no need to keep the the extent. + cache.c:388,409 + After noting in the extent what it would be caching, the bid + field of the extent is set to the pgno counter in cache. This + value is also noted in the daddr (disk address) field of the + Page used, and can be used both for consistency checks + (:160,161) and to lookup a page with a given bid. pgno is + incremented not by one, but by BY2PG instead; it is being used + as the address in a fictitious device where all pages used by + the cache are kept in allocation order. + cache.c:410,417 + The memory updated in the cache is copied into the extent page. + cache.c:419,420 + Before releasing the page, cachepage is used to add the page to + a page cache. That is discussed in the virtual memory chapter, + but note how extent pages are kept in the page cache. Extents + are just adapting the page cache to serve as an extent cache + for remote files. + + Writing to the cache + + syswrite + mntwrite + cwrite() Updates the contents of a cached file (new version). + + cache.c:551,574 + cwrite is called to write to cached files. It increments vers + both in the Mntcache and in the channel QID, as the file is + being updated. + The mount driver calls cwrite just for Twrites on cached files, + and that is the case when it increments the QID vers for the + file; if somebody else is using the file and the server is + incrementing its vers field, this client wouldn't notice. Other + file systems tried to enforce coherence to the limit, and as a + result, forced clients to be blocked while transactions for + other clients were ongoing. In Plan 9, if this relaxed + coherence model is a problem, the application should use OEXCL + to ensure that only one process at a time has the file open--or + rely on any other synchronization means. + cache.c:576,599 + The first extent for the portion written is located. If such + extent is not the first one or does not exist (:583), the + routine tries to use any final hole in the page for the + previous extent--if contiguous. This is the same of cupdate. + cache.c:601,621 + The portion written could overlap an extent which would be past + the one being written. This could happen if the file has been + read into the cache (updated) and another process writes the + file or the reader does a seek to rewrite the file. The new + bytes written are the valid ones, and any previous copy has to + be released. extentfree is used to release the extent, rather + than reusing it. In that way, cchain can be used to add more + extents to the cache as it was done in cupdate. + I think that cupdate and cwrite could be serviced by a single + routine, like happens typically with read and write. But the + author may disagree. + + I/O + + You now know how files work in Plan 9, but you still have to look at + how actual I/O is done. Prior to the 3rd edition, Streams [[240]13] + were used as the framework to do I/O (i.e. to read/write from + devices). In the 3rd edition, streams were replaced by a more simple + (and less flexible) queue based I/O module. In this chapter you saw + how pipes used qio facilities to do I/O, and in previous chapters you + saw how the console and serial lines used qio too. The kernel uses + queues for I/O wherever there is an I/O flow of bytes from a source to + a drain. This happens when using devices and also when using artifacts + like pipes. I suggest you revisit the pipe device after reading qio in + this section, so you could fit the pieces together. + +Creating a queue + + qio.c:25,50 + a queue is a flow of bytes from the source to the drain (e.g. + from the keyboard device to the reader of the console + keyboard). The queue maintains Blocks, each with a block of + bytes. Queues perform flow control activities; they block the + reader when the queue has nothing for the drain, as well as + they block the writer when there is no room in the queue. Let's + see all this while you learn how routines to use Queues work. + If you look at the file, the initial part is implementing + routines that operate on Blocks and the final part is + implementing queue routines. + + qopen() Opens (prepares) a queue for I/O. + + qio.c:740 + A queue starts its service when a call to qopen creates it + (devpipe.c:67 and :72). qopen receives the maximum number of + bytes to be kept buffered by the queue (limit). If you look the + comment, you will notice that qopen uses malloc and should not + be called from an interrupt handler, because malloc could try + to acquire locks to allocate memory and cause a deadlock with + the process being interrupted. + msg is true if the queue is queueing messages and not bytes. + This is an important parameter. For a pipe, it does not matter + whether the bytes fed to the pipe were fed in a single write or + in a couple of writes; if the reader reads N bytes, it should + get those N bytes if they are present in the + queue--independently of how were they written. However, for a + network transport and other devices, it is important to read + what was written, no more, no less. If a network device places + two network packets in a queue, and a network transport wants + to get the next packet received, it should be able to read just + the last ``message'' written (i.e. the last packet queued). + Other way to say this is that queue can either delimit data + from different writes or not. msg is a way to make Queue a + generic queueing tool. + The kick and arg parameters are used to let the queue user do + something before flow controlled processes are awaken. This is + a convenient thing to have to make more simple the code of the + queue users. devpipe has nothing special to do and passes nil + as kick. However, the ns16552 serial device passes ns16552kick + as its kick procedure (to restart serial output). + qio.c:744,759 + inilim is the initial value for limit; more soon. state holds + the state for the queue, which is initially Qstarve (no bytes + in) and Qmsg if the queue is message oriented. eof and noblock + are cleared and you will see what this means. + +Read + + qread() Reads bytes from a queue. + + qio.c:860,873 + qread is the procedure to read bytes from a queue. Read the + comment. It calls qbread, who does the actual job of reading + len bytes and returning a Block with the bytes. qbread returns + zero if there is nothing more to be read (the queue was closed + and nobody would write more bytes on it, and the queue is also + empty). BLEN()Returns the length of data in a block. + portdat.h:134 + BLEN is defined to take a pointer to a Block, and return the + difference between its wp and its rp. That is the number of + bytes yet to be read in the Block. + portdat.h:123,133 + It is clear what is happening here, a Block contains a series + of bytes at base (the first one is base[0], and the last one is + base[lim-1]). Bytes written into the block are written starting + at wp (which would be initially base). Bytes read from the + block are read from rp (which would be initially base, and + should never go beyond wp). + qio.c:873,877 + qread copies the bytes in the block (len bytes) to the user + buffer (vp), and then it releases the block. freeb is discussed + later. + + qbread() Reads a block from a queue. + + qio.c:773 + qbread does the actual work of reading from the queue. Besides + helping qread, it is a queue read routine on its own--it is + very useful to implement devbread procedures. + qio.c:778 + A queuing lock is gained on rlock, where queue readers + synchronize. qlock maintains a list of blocked processes so + that the first who called qbread would be the first getting + bytes from the queue, if it blocks while acquiring the lock. + qio.c:785,807 + On this loop, an ilock is done on the queue, this prevents any + interrupt handler from using the queue, because no interrupts + can arrive while holding the lock. Once the lock on q is + gained, no one is messing up with the queue block pointers and + they can be used safely. If the list of blocks in the queue + (bfirst) is not nil, b is the block to read from, so break the + loop. If there are no blocks (!b), and the queue state is + Qclosed there will never be a block, so the author releases the + locks, sets the eof flag in the queue and the routine returns + zero as the number of bytes read. A read count of zero is the + convention in Plan 9 for signalling EOF. When EOF is signalled + more than three times, the reader could be ignoring EOF and the + routine raises an error instead. q->err contains the error + string to raise; e.g. qclose sets it to Ehungup. + If the state is not Qclosed, and there is no block in the + queue, this process must wait until the source (the writer) + puts more bytes in the queue. So, the author sets the Qstarve + flag to let the writer know that a reader is starving (waiting + for bytes), the q lock is released (so that the writer could + add more bytes to the queue) and the process sleeps on rr until + notempty. notempty lets sleep know that it shouldn't sleep if + there are bytes to be read in the queue. The parameter for + notempty is the queue being read. When the process wakes up + later, it would repeat the loop, check for Qclosed again (the + writer could close the queue instead of writing to it) and that + process repeats until there is a block in q->bfirst. + BALLOC()Returns the number of bytes allocated to a block. + qio.c:809,814 + Got a block. The routine removes it from the head of the list + (Blocks are linked through their next field). dlen counts the + number of bytes in the queue, as a block of n bytes has been + removed. len counts the number of allocated bytes in the queue + (i.e. number of bytes from base to lim-1 in all blocks + linked)--BALLOC (portdat.h:135) is a macro counting the number + of allocated bytes in a block). + qio.c:818,837 + If the block has more bytes and the queue is not message + oriented, remaining bytes (unread) should be kept in the queue. + If the queue is message oriented, the reader should read the + first message; no matter how many bytes it wants. Any unread + portion of the first message is discarded. Line :819 checks + that there are more bytes in the block (n) than wanted (len); + the next line checks that the queue is not message oriented. + Note the iunlock (lock held since :787). The lock is released + while allocating a block for n-len bytes, and copying those + bytes into the block allocated (wp is advanced to point past + the bytes written in the block). The ilock is done to mess up + with block pointers again while inserted the new block in the + queue--holding the unread portion of the previous block. By + releasing the lock, the queue can be used by others in the mean + time. + Line :836 sets wp in the block being read to point after the + bytes to be read from the block. Should the application write + to the block, it would not overwrite the data already placed in + the block. + qio.c:839,846 + The writer could be sleeping because there was no free room in + the queue to write more bytes. That is signaled by Qflow in + state. Should that be the case, if the used space in the queue + is below half its maximum number of bytes, the writer can be + awaken. Even though the queue may have (less than limit/2) free + room, the writer is kept sleeping. That is to avoid sequences + of sleep/wakeup/sleep/wakeup/...because the writer is awakened + too soon, fills up the empty room, and has to sleep again. The + queue state is changed before releasing the lock. After the + lock is released, other processes can lock the queue and + interrupts can arrive (if they were enabled before the ilock). + qio.c:848,853 + With the lock released, any writer sleeping awakened. If a kick + procedure is supplied to qopen, it is called. Usually, a + process waiting to write the queue corresponds to a process + doing output to the queue (possibly drained by a device). On + the other hand, when it is a device the one doing output to the + queue (an input device), the kick procedure can be used to + resume the device and allow it to put more of its data into the + queue--device input would happen at interrupt handlers and it + makes no sense to block (put to sleep) an innocent process just + because it happened that it was running while a device received + an interrupt stating that there are more bytes to add to an + input queue. + qio.c:856,857 + Until now, any other reader would be blocked on the rlock--one + reader at a time!. Now other readers are allowed to enter the + critical region and the block with the bytes being read is + returned to the caller. If the queue is message oriented, all + the bytes in the first block of the queue are returned; + otherwise, just the bytes wanted. + +Other read procedures + + qconsume() Reads from a queue even within interrupt handlers. + + qio.c:451,517 + Within an interrupt handler, qread cannot be used. qconsume is + like a qread which can be used by interrupt handlers (e.g. + devns16552.c:475). First, qconsume does not call qlock, which + may call sleep and perform context switches. (Is there any + context to switch while you are running within an interrupt + handler?) Second, qconsume does not call sleep to block when + there are no bytes to be read--qconsume would return -1 + instead. Both routines, qread and qconsume, can be used on + system-call (and trap) handling kernel code, within the context + of a process. + As you see, routines for use on interrupt handlers (and those + that must synchronize with them) use ilock. If the kernel is + servicing a regular system call and a queue routine uses ilock, + no interrupts are allowed, and no interrupt handler can even + try to use a queue routine: no deadlock. If the kernel is + servicing an interrupt and the queue calls ilock, no interrupts + would arrive and there can be no context switch, so there can + be no deadlock. That is why the author avoids carefully any + call to sleep and context switches here. + Of course, other processors can still try to use the queue, but + would notice the lock, and would not interfere (they would + either block the process or spin to wait a bit). + qconsume takes care of being as lightweight as possible. Unlike + qbread, which would split the initial block into two ones when + only part of the block is read, qconsume uses the rp block + pointer to read only part of the block when len dictates that. + In that way, qconsume avoids block allocation. Along with this + line, any initial empty block is skipped and linked into the + tofree list passed later to freeblist. + + qget() Gets a block from a queue. + + qio.c:369,403 + There is yet another read procedure for queues, qget. It is the + most simple read procedure, and it is specialized just to get + the first block in the queue, if any. It never blocks, and is + appropriate for use on interrupt handlers too. qconsume can + read any number of bytes, but qget can only get a block. qget + is used mainly by the code in ../ip. The tcp/ip protocol stack + uses blocks to store protocol data units (e.g. network + packets). qget and other queue routines operating on queue + blocks, allow the tcp/ip code to use queues to do packet (i.e. + block) i/o. + + qdiscard() Discards bytes from a queue. + + qio.c:408,445 + qdiscard is not a ``read'' routine, but removes bytes from the + queue. It iterates through the queue blocks until len bytes are + discarded. If a whole block is discarded it is freeb'ed, + otherwise the rp pointer is used to ``read'' the bytes + discarded. There is no synchronization regarding qread, and the + routine does not block. It is also appropriate for interrupt + handlers. This routine is useful for ../ip code, which discards + data when it is known to be received by the peer node (e.g. + when data is acknowledged). + + qcanread() Is there anything to be read from the queue? + + qio.c:1160,1164 + qcanread is a small procedure, albeit an important one. Some + queue users would not like to read the queue when that would + block them (e.g. devcons). qcanread returns non-zero when there + is something to read. The caller can later call qread. There is + no guarantee (specially on multiprocessors) that the queue + would be still non-empty when qread is later called. The caller + of qcanread must ensure that by any other means if that is + important. + +Write + + qwrite() Writes bytes on a queue. + + qio.c:961 + qwrite is called to write bytes in the queue (e.g. + devpipe.c:327). If it is message oriented, the bytes written + are considered to be the message. + qio.c:970,989 + The routine writes at most Maxatomic bytes (32K) at a time in + the queue. It is reasonable to limit how many bytes can be + placed in the queue at a single qbwrite to avoid a writer + flooding the queue with a request so big that locks are going + to be held while acquiring resources to queue an unreasonable + amount of bytes (e.g. lots of pages in memory, etc.) It is also + good to keep this limit to put a reasonable limit on message + lengths, so that readers of message oriented queues do not have + to cope with unreasonable long messages. Important lines are + :976, which allocates a block for the n bytes being added at + this pass; :982,984, which copy the bytes from the user buffer + into the block using the wp pointer; and :986 which calls + qbwrite to do the job of queuing another chunk of (at most + Maxatomic) bytes. The while condition checks for Qmsg; if the + queue is message oriented, and a message is at most Maxatomic + bytes, the routine would not qbwrite any byte that does not fit + into a message. + + qbwrite() writes a block in a queue. + + qio.c:891,901 + qbwrite adds the block to the queue (e.g. devpipe.c:354). The + lock used is wlock. The queue can be read and written at the + same time, but writers serialize their access to the queue (in + the same way readers do). Like qbread, this routine is useful + to implement devbwrite procedures. + qio.c:904,927 + While adding the block, a lock on q is held. Again, an ilock is + used (know why?). If the queue is closed, there is no point on + writing on it, so the block is released and the queue error + returned. The loop keeps the writer there until the block can + be added; but if the length of the queue is beyond its limit, + no more bytes should be added. + If noblock is set (it was set initially to false by qopen), a + write on a ``full'' queue is discarded and qbwrite pretends + that n bytes in the block were written. If noblock is not set, + the writer of a full queue sleeps until the queue is below its + limit--and Qflow is flagged so that a reader would wake up the + sleeping writer. + qio.c:929,936 + The block is added to the queue and the queue len and dlen + fields are updated. Blocks are added at blast--they are read + from bfirst. + qio.c:939,942 + If a reader is sleeping waiting for bytes in the queue, the + routine wakes it up. If there are multiple readers, the first + one holds rlock while sleeping, so other readers would not even + enter to read the queue until the first one is awakened and + gets the bytes. That is the reason for having just one bit to + signal ``readers waiting'' ``writers waiting''. + +Other write procedures + + qiwrite() Writes to a queue (for console). + + qio.c:999,1045 + qiwrite is a version of qwrite folded with qbwrite. It exists + because console routines may want to write on queues during + boot time, even before there are real processes (see + devcons.c). ilock is used (to prevent further interrupts too), + but there is no qlock (read the comment). That means that only + the lock on the queue is gained, but in no case the ``current + process'' (which could be simply the flow of control existing + at boot time) would call the scheduler within qlock. Besides, + flow control is not obeyed, the caller will never block because + the queue has gone beyond its limit. + In any case, it works like qbwrite, and would wakeup any reader + waiting. + + qproduce() Writes bytes on a queue even within interrupt handlers. + + qio.c:634,683 + There is yet another routine that writes to the queue, it is + qproduce. Like qiwrite, it does not use qlock. It does not + enforce Maxatomic, and never sleeps. qproduce is to qwrite what + qconsume is to qread. qproduce is intended to be called by + interrupt handlers (E.g. devns16552.c:725). When the queue goes + out of limits the block is not added and an error signalled by + returning -1. Besides, iallocb is used to allocate a block, and + not allocb. The iallocb version of allocb knows it runs within + interrupt handlers. + + qpass() Writes a (list of) block to a queue. + + qio.c:519,564 + qpass is the counterpart of qget. It writes a block in a queue. + There is no lock on q->wlock, and there is no call to sleep for + a full queue. The routine is also appropriate for interrupt + handlers. One interesting thing is that the routine adds not + just one block, but a list of blocks (:534,537) and accounts + for that (:541,546). Another interesting thing is that it + enables Qflow not when len goes above limit, but when it goes + above limit/2 (:551). + Queues get full when the limit is passed. Usually, it is a + write which makes the queue go above limit the one that sets + Qflow. In this case, the routine is used to write whole blocks + (maybe more than one), so the author takes care not go too far + above limit, and half the limit is used as a limit. + This routine is very useful for the code in ../ip , to place + protocol packets (queue blocks) into queues to be serviced + later. + + qpassnolim() Writes a (list of) block to a queue without obeying + limits. + + qio.c:566,606 + qpassnolim is exactly as qpass, but does not check for limit. + In this case the author wants the list of blocks to be written, + no matter the queue fill state. I don't know why the author did + not add a flag to qpass, to avoid checking the limit, and wrote + instead qpassnolim. Perhaps nobody cared to do code cleanup in + qio.c, or perhaps careful measuring suggested the multiplicity + of routines. By the way, qpassnolim seems to be used for the + ../ip code when the author knows it is okay to overflow the + queue to pass data to another part of the protocol stack. + + qwindow() Is there room in the queue for writing? + + qio.c:1146,1155 + qwindow is to be used like qcanread, a caller of qwrite can use + qwindow to see if the qwrite would block or not. + +Terminating queues + + qhangup() Hangs up on a queue. + + qio.c:1095,1109 + qhangup is used to state that no one else will write anything + more to the queue. However, the queue is kept with any block + not yet read. The err field is used to report that the queue is + hunged up (or the message supplied by the caller), and state is + set to Qclosed. Any reader would notice the Qclosed and it will + not block waiting for more bytes. Any writer will just discard + the bytes being written. notempty (:766) returns true when the + queue is closed, so that sleep will consider that there is no + need to sleep on a closed queue. The two wakeups would wake up + any reader or writer sleeping, and they will behave as I just + said. Did you noticed the ilock? + + qclose() Closes a queue. + + qio.c:1062,1088 + qclose closes the queue--like qhangup. Unlike qhangup, it + releases any block in the queue (:1083). qhangup is intended to + be used when one end of the queue hangs up, and how qclose is + more like a ``free'' routine (e.g. devpipe.c:228,229). Another + way to see it is that qhangup can be used by writers to signal + that there is nothing more to come; while qclose can be used to + shutdown the queue. Qflow and Qstarve are cleared. Perhaps they + should be cleared by qhangup too. + + qreopen() Reuses a closed queue. + + qio.c:1123,1132 + qreopen can be used to undo the effect of a close. It clears + the Qclose and sets the Qstarve and eof fields as in qopen. The + purpose is to reuse closed queues instead of allocating new + ones (e.g. devpipe.c:247,248). + + qfree() Deallocates a queue. + + qio.c:1051,1056 + When the queue is no longer needed, qfree does the close and + then calls free. It can be called for an already qfree'd queue, + as free checks for nil pointers and qclose does so too. The + comment suggests that perhaps qclose should add reference + counting and free the queue when the reference goes down to + zero. + +Other queue procedures + + qcopy() Copies bytes from a queue. + + qio.c:688,734 + qcopy is used to copy bytes from the queue into a new block. + Bytes are not read from the start of the queue. Instead, qcopy + copies bytes from the given offset in the queue. Lines :701,715 + locate the block and ``rp'' (p) where to start reading, and + later lines perform the copy. The queue blocks and their rp are + kept untouched. This routine is used by the ../ip code, which + usually likes to copy data out of network messages. + +Block handling + + Routines early in qio.c perform operations on blocks. They are mostly + of interest to protocol stacks using queues as their I/O mechanism. + For instance, code adding headers, extracting data, etc. from network + messages use these routines. I think you should be able to understand + these routines: + + qio.c:91,133 + padblock takes a block and returns a new block (or the same bp + if it has enough space) with size extra bytes of padding at the + front or at the back. That can be used to add headers or + trailers. + qio.c:138,150 + blocklen uses BLEN to return the length of a list of blocks. + Some times, specially when a message is traveling through a + protocol stack, a message may end up being a sequence of + blocks. + qio.c:155,174 + concatblock takes care of merging all the linked blocks into a + single one. Some routines assume that a message is contained + within a single block, concatblock can be used to ensure that. + qio.c:179,232 + pullupblock checks that there are n bytes after the bytes in + the block (after rp). It allocates a new block if needed. This + is useful to add n bytes to a message without turning the + message into a block list. + qio.c:237,273 + trimblock trims a block to a subset of the bytes on it. Useful + to remove unwanted headers and trailers. This can be used to + trim bytes at the front, at the end, or at both sides. + qio.c:278,302 + copyblock copies bytes to a new block. + qio.c:304,330 + adjustblock truncates the block to len bytes. Perhaps, + trimblock could be used instead of providing a new routine, + although this routine would run faster. + qio.c:334,364 + pullblock removes bytes from the front of a block list. + + Finally, note how most queue routines update statistics declared at + qio.c:109,14. Those counters tell the author how intensively are used + the routines involved. For instance, if qcopycnt goes too far, it may + be a symptom that queue copies should be avoided if there is a + performance problem involved. Statistics are important in that they + let the author know the real usage of the code; most of the author + assumptions would not correspond to the real system usage as seen by + the statistics. + +Block allocation + + allocb.c:24,56 + allocb is the routine allocating blocks always but for + interrupt handlers. The memory allocated is for the Block + itself, and also for the data to be kept in the block. Hdrspc + empty bytes are kept allocated besides the n bytes requested + (the total allocated space is size+Hdrspc+sizeof(Block)). That + is to allow protocol stacks to place their headers before the + data in the block. Should the author not do so, almost every + step in a protocol stack would require allocating new blocks, + concatenating them, and releasing previous blocks. The system + I/O for networks would go unbearablely slow. Another + interesting thing is that the routine raises an error (unlike + iallocb). + base is set pointing past the Block in the allocated memory, + and rp and wp are set pointing after the Hdrspc (which is + computed by subtracting size from the limit of the allocated + memory). + The dance around BLOCKALIGN is ensuring that pointers are + aligned to BLOCKALIGN (8) bytes. That can prevent alignment + errors on machines that are picky regarding where can integer + values be placed in memory. Not the case, but this does not + hurt. + The memory held by the block would be released when free is + called in the block--the free block routine is appropriately + set to nil. + allocb.c:61,108 + iallocb is a version of allocb for interrupt handlers. The + difference with respect to allocb is that ialloc does not raise + any error (returns nil instead) and sets the BINTR flag in the + block (some ialloc accounting too, admittedly). The flag is + only used by freeb to do accounting, but is not used for other + things. + allocb.c:110,140 + freeb calls the free procedure, does accounting for iallocated + blocks, and releases the block. If a free procedure is + provided, free is not called on the block; providers of block + storage are responsible to reclaim unused storage. All the + pointers are set to Bdead, which is a meaningless value that + can be recognized quickly when the debugger prints pointer + values. + + Protection + + In Plan 9, there are several system calls (see auth(2) and fauth(2)) + that have to do with protection. However, before looking at their + code, it is better to understand the overall architecture of Plan 9 + regarding security. Read also auth(6) and cons(3). What I comment here + is just what I think you need to know to understand the code. + +Your local kernel + + Each Plan 9 machine is either a terminal, a CPU server, or a file + server. Each machine run its own Plan 9 kernel, customized to perform + well for the given task. Terminals are machines used to interface the + Plan 9 network to its users. For example, each user runs rio(1) (the + window system) at its terminal. Terminal machines use services from + other nodes in the network. In particular, a terminal uses a CPU + server to execute commands on it, and a file server to get files from + it. Besides, machines where you run your programs in the network (e.g. + cpu servers) use files serviced from your terminal (e.g. your mouse). + + Everything is a file in Plan 9, and file permissions are what the + system uses to provide protection. Each file has rwx bits for its + owner, its group, and others (you already know how that works, since + you did learn that for UNIX). Thus, one barrier of protection is + placed at the file server that services the files accessed. + + In Plan 9 there is no `superuser' as in UNIX. In UNIX, a user with + uid 0 is granted special privileges by the system, which has + conditionals in the kernel to allow such uid to do almost anything. In + Plan 9, no user is granted permission to do everything. + + Each Plan 9 kernel is booted by a a user, and that kernel only trusts + that user. The user who boots a node is referred to as ``eve'', as you + know. Each kernel services some files, and eve is granted special + permissions on those files--noticed the checks for ``eve'' while + reading the code? + + By trusting only the user who did boot the node, Plan 9 does not allow + other users (nor other kernels) in the network to do things to your + local files. Why does Plan 9 give special permissions to eve? + + If you have physical access to the system and can boot it, nobody can + prevent you from using another system (e.g. msdos) and access the disk + files without Plan 9 even knowing. Therefore, there is no security + breach in allowing you to bypass permissions for local files. + + To check permissions for a process accessing a file, each process has + a user identifier (Proc.user). The initial process belongs to the user + eve, who booted the machine. That user, types its user name and its + password and the boot process uses that information to authenticate + the user. Before this point, the boot process belongs to the user + ``none''. Say that the typed user name was nemo; once authenticated, + the boot process continues and processes forked would run on nemo's + name too. To authenticate, the user process uses auth(2) services to + get in touch with the authentication server (another machine) and gain + tickets for the user. While authenticating, authwrite (auth.c:422) is + called to respond to a challenge, and if the reply is ok, the user + field of the process is changed according to the user who is + authenticating. So, each machine trusts the authentication server and + itself; it usually trust nobody else. + + For terminals, this is mostly what happens. For CPU servers, the user + who boots the CPU server has some processes on its name, and must be + able to create processes for other users willing to compute on the CPU + server considered. What happens, is that the user owning the CPU + server is granted permission (by the authentication server) to speak + on behalf of the user that wants to compute on the CPU server. + +Remote files + + According to what I said, other nodes will not trust you. How could + they trust you? Actually, the Plan 9 kernel does not care--mostly. As + far as the kernel is concerned, your files come from a server (a set + of servers actually) speaking 9P through a set of mounted file + descriptors. It is you who get those file descriptors by setting up + network connections to file servers. If you can authenticate to a file + server in the network, and convince it to speak 9P for you, you can + later give the descriptor to mount(2) and bring the server files to + your name space. In principle, your local kernel does nothing to let + you authenticate to the remote server and get your 9P session up. What + your local kernel does is to check protections for your local files. + + As an example, you must first authenticate to a Plan 9 file server to + use its files. (e.g. you authenticate with a file server kernel to + access your files; you authenticate with your local kernel to get + access to files serviced by the local kernel; etc). This can be + considered to be a first barrier of protection: convincing the file + server to speak 9P with you. Later, the file server will be checking + permissions, given the attributes of its files and your + (authenticated) identity; you can consider this as a second barrier of + protection. + + By placing authentication mechanisms outside the system (which only + has to handle 9P), and letting you obtain the authenticated + connections to file servers, Plan 9 can be as secure (and as insecure) + as you want it to be. + + One thing the kernel does for you is to keep your tickets--after you + gave your user name and password while booting--to authenticate + connections for which you already have tickets. Of course, you can + still use any other means to protect connections with your file + servers, and then mount the connection descriptors. + + The code keeping your user id and your ticket (generated from your + password) is found at /sys/src/9/port/auth.c, with console files + serviced by /sys/src/9/port/devcons.c. I think you should be able to + read that and understand it, provided you understood auth(6) and + auth(2). + + Memory Management + + Plan 9 uses paged virtual memory. Although on Intels there is + segmentation hardware, hardware segments are used just to implement + protection ring 0 for the kernel and ring 3 for the user--go back to + the introduction chapter if you forgot. Hardware segments are not to + be confused with process segments, which is an abstraction implemented + in software by Plan 9. + + Before discussing memory management system calls like segbrk, + segattach, segdettach, segfree and segflush, I start by discussing how + memory management works. You already know a bit about this, from + chapter [241]3. I hope that way you will learn what data structures + are involved, and you will understand better the code related to + memory management system calls and memory management trap handlers. + + During this chapter, you will be reading these files: + * Files at /sys/src/9/port + + fault.c + Page fault handling. + + page.c + Paging code. + + segment.c + Process segments. + + swap.c + Swapping code. + + sysproc.c + Process system calls. + + devproc.c + Process device. + + devcons.c + Console device. + + * Files at /sys/src/9/pc + + mem.h + Memory management definitions. + + memory.c + Actually discussed at chapter ch:start, but you may want + to reread it here. + + mmu.c + Memory Management Unit handling code. + + trap.c + Entry points for MMU faults. + + dat.h + Machine dependent data structures. + + Processes and segments + + ../pc/mem.h:29,54 + To remind you, the kernel uses the paging hardware (two-level + page tables) to implement virtual memory. Each process has its + own virtual address space, split into two regions, one for the + kernel and another for the user. The user portion of the + virtual address space is using addresses from 0 to 2G + (0x00000000 to 0x7fffffff). The kernel portion goes from 2G up + to 4G (0x80000000 to 0xffffffff). The last two Gbytes, for + kernel usage, are shared among all Plan 9 processes, which + means that their entries in the hardware page tables are the + same[242]10.1. You should remember among other things the + identity mapping for physical memory. + From the point of view of the process, things are different + (see figure [243]6.1). Its 2G of the virtual address space + (what it can see) are structured into segments. A process knows + it has a set of segments attached at concrete virtual addresses + with concrete lengths. For instance, all processes have a text + segment (with instructions) at address UTZERO, past the first + page--which is kept unmapped to catch dereferences for nil + pointers. Besides, processes have stack, data, and BSS + segments. + + CAPTION: Figure 6.1: The user view of a virtual address space: A text + segment with the program code, a data segment with initialized data, a + BSS segment with uninitialized data, and a stack segment. + + mem.h:60,77 + Do not confuse the process (software) segments with the + hardware segments used by Plan 9. + +New segments + + Let's start by looking at how are new segments created, considering + first a stack segment. + ../port/sysproc.c:324,329 + The boot process was given segments by hand by the kernel + bootstrap code and the first thing it did was an exec system + call to execute the code for the boot process. sysexec then + calls newseg to create a stack segment for the new program. To + remind you, segments for a Proc are kept linked to its seg + array, which has NSEG entries. If you remember from the chapter + on processes, ESEG is an slot for an extra segment (SSEG is the + slot for the stack segment). + + newseg() Creates a segment. + + segment.c:47,54 + This procedure creates a segment of a given type, base and + length. It aborts if the size is beyond the maximum size + allowed for a segment--size is in pages, as segments must + contain an integral number of virtual memory pages because the + paging hardware is used to implement them. + portdat.h:365,383 + If you look at Segment, you can see how there is a map + array with pointers to Pte structure (see + figure [244]6.2). + + CAPTION: Figure 6.2: A Segment maintains a virtual MMU data structure + holding Page structures for pages in the segment. + + portdat.h:323,329 + A Pte contains at most PTEPERTAB pointers to Page + structures, each one responsible of a (virtual) memory + page. + What is happening is that a segment is using a virtual MMU + as its data structure. So, the map in Segment is like a + two-level page table that lets the segment hold pointers + to all Page structures for the pages it has. The reason + for using this two-level structure is the same reason the + hardware has for using two-level page tables: to save + memory yet to be efficient when looking up entries. + ../pc/mem.h:104 + As map will have at most SEGMAPSIZE entries, and each + entry has (../port/portdat.h:325) at most PTEPERTAB + pointers to pages, the maximum number of pages is the + limit checked at segment.c:53. + swapfull()Running out of swap space? + segment.c:56,57 + To implement virtual memory, all pages that do not fit into + main memory are kept in a swap file (which could be a swap + partition, since partitions are files). swapfull + (swap.c:406,409) returns true when the swap file has less than + a one tenth of free space. The kernel refuses to create new + segments when it thinks that there will be no space in swap for + backing up the segment. + segment.c:59,64 + The new Segment is created. It is reference counted because + processes can share segments. The type, base address, and top + address (the first address past the last address in the + segment) are kept in the Segment structure. + segment.c:66 + mapsize is set to the number of map entries (PTEPERTAB entries + each) needed to hold size pages. Part of a the last Pte in map + could be unused. + segment.c:67,73 + nelem is a macro returning the number of entries in an array + (portfns.h:173). If more entries are needed than the number of + entries in the (small) ssegmap array kept in Segment, map is + allocated to contain twice the entries needed--unless that + value goes over the maximum number of entries, in which case, + just the maximum is allocated. + The author is allocating twice the space required because he + thinks that in the future the segment could grow. In that case, + the author wants to be sure that allocated space would suffice + most of the times. That makes unnecessary to reallocate + existing entries. Right now, all pointers in the map are nil, + because smalloc is used. mapsize holds the number in entries in + the map array. + segment.c:74,77 + What happens when there are less entries in map than entries in + ssegmap? map is set pointing to ssegmap, instead of allocating + a fresh new map. The author made a provision for small + segments, so that they do not incur in the overhead of + allocating/deallocating maps. Small segments are serviced just + with the Segment structure. This is also a help to fight + fragmentation, not just execution time, because less structures + have to be allocated. + segment.c:79 + Finally, a new Segment structure with enough entries in map + (all nil) is returned. + +New text segments + + sysexec + + sysproc.c:381 + During sysexec, a new text segment for the code found in the tc + channel is created. + + sysexec + attachimage() Creates a segment attached to a file image. + + segment.c:246 + attachimage tries to create a new segment of the given type. + Unlike newseg, attachimage attaches a file image to the + segment, so that the segment would appear to contain whatever + is contained in the file referenced by the channel c (see + figure [245]6.3). + + CAPTION: Figure 6.3: A Segment can be attached to a file image. + + segment.c:251,252 + imagechanreclaim is discussed later--it is just closing unused + channels which were used to fill up other images. + segment.c:254 + Locking the imagealloc, which is an allocator of Image + structures. + segment.c:19,31 + The imagealloc contains a free list of Images, and a hash table + for Images. You will be seeing how it is used. + portdat.h:309,321 + An Image represents the image in memory for a portion of a + given file. As attachimage takes a channel to a file and builds + a segment upon its contents, Images are very important here. + Just note how an Image contains a channel to the file used as + the source of data for the image (c), and how there is a link + to the Segment which is using the image of the text file. + segment.c:260,261 + All Images under imagealloc are kept hashed on the QID of they + file they come from. ihash selects the appropriate hash entry + in imagealloc. Each hash bucket has images linked through the + hash field of Image. The author is searching for an Image which + comes from the same file; therefore, the qid.path of the + channel is compared to the qid.path of the Image (Images keep + the qid for the file they are maintaining). + segment.c:262,270 + Should an Image for c's file be found, the image is locked and + the QIDs compared (with eqqid this time). The check at :261 was + a quick guess to avoid locking all images in the cache just to + find out that they are not the ones of interest. In the real + check, the QID, the QID for the channel to the mounted file + (which could be zero if not mounted), the channel to the + mounted server and the device type are compared. If you + remember, two files are the very same file if their QIDs match, + they are serviced by the same device type, and the actual + server is the same. This is the check being done here. Lines + :264,265 are needed to distinguish between different channels + going through the mount driver, but pointing to different files + on different servers. + If such an image is found, a reference is added and the routine + continues at :298, with the image locked. The author knows that + only a copy of the file text has to be kept in memory. If you + run different processes for the /bin/rc program, the text has + to be in memory just once (because it can be shared due to its + read-only nature). Therefore, all such text segments would be + sharing their Images, which would be just a single Image for a + channel going to /bin/rc. + segment.c:274,285 + No Image was found for the text, so the author allocates a new + one. The loop, which calls imagereclaim, tries to deallocate + Images back to the free list. The process doing an exec would + be looping calling imagereclaim, and letting other processes + run until it can get an Image from the free list. This process + could do not much else, because exec did commit to execute a + new process and it cannot be even aborted. The choice is either + wait for an image or die. The imagealloc lock is released while + allowing others to release images. + segment.c:287,299 + The new image is locked, a reference added to it, and its + fields initialized. The Image is linked into the hash bucket + corresponding to the file's qid. imagealloc is unlocked but the + image remains locked. + segment.c:301,312 + i could be a newly allocated one, or one reused (shared) from + the hash. If the Image comes from the hash, its s field points + to a text segment, which would be shared, so a new reference is + added to it. If the image is new, it has no segment yet, so + newseg allocates a new segment of the type desired, and its + image field is set pointing to the image. You can go from the + segment to the image using image, and back to the segment using + s. The waserror handling code does not call nexterror, because + the caller is gone during sysexec. Instead, the process is + killed on errors. + segment.c:314,315 + All set. Either a fresh new image created or a previous one + shared. + + sysexec + + sysproc.c:382,387 + attachimage returns the image locked. The caller adjusts the + fields fstart and flen in the Segment using the image to record + the portion of the Image which should be used to fill up + segment memory. In the case of the text segment, its bytes come + right from the beginning of the file--even the header is + ``mapped'' within the text segment (looks like the real text + file, doesn't it?) + sysproc.c:393,397 + In the case of the data segment, its bytes come from the text + file, past the code. + To summarize, an Image is an image of a program file kept on + memory, it is attached to one or more segments, and each + segment attaches to a portion (fstart, flen) of the file image. + We will get back to images later. + + Page faults or giving pages to segments + +Anonymous memory pages + + I use the term ``anonymous memory'' (as others do) to refer to memory + which does not come from a file. For example, text and data segments + have their contents coming from the file with the program being + executed. Stacks and BBS segments, on the other hand, are created with + cleared memory, which does not come from any file. Let's pick up the + stack segment as an example to see how it gets some pages. + + sysexec + + sysproc.c:329,333 + Once the process has a segment, it can reference addresses + within the segment. When are pages given to segments? When does + a segment get actual memory? If you read the comment, you'll + get a hint. + Segments are made of virtual memory pages. Those pages can be + either at physical memory, or at secondary storage. For the + stack segment just created at :329, there are no pages yet. But + the code below will write to stack addresses! + + trap + + ../pc/trap.c:218 + The first time the kernel writes to an address within the first + page of the stack (the last page in the segment due to stack + growing direction on intels), a page fault trap is generated. + That makes sense since the hardware MMU has the translation for + the stack page marked as absent. + trap.c:242,245 + The handler in the Vctl for the page fault trap is called--the + handler was set at :170 to be fault386. + + trap + fault386() Services a 386 page fault. + + trap.c:435,442 + fault386 receives the Ureg for the faulted processor context + (which would be code running within the kernel in the example). + Due to Intel nature, the faulting address is not saved by the + hardware in the Ureg, but is located in the cr2 register + instead. The routine saves the address in addr--another page + fault in the mean time would overwrite the cr2 and loose the + previous faulting address. + trap.c:443,445 + user is true if the saved context had the UESEL as the code + segment selector (i.e. it was code running at user-level). If + the page fault happened while running inside the kernel and + mmukmapsync can handle addr, nothing else is done--the page + fault should be fixed. mmukmapsync()Synchronizes kernel maps + for the MMU. + mmu.c:273,288 + We will see later, but mmukmapsync tries to get the hardware + page table entry (pte) for the faulting address. It uses the + pdb pointer for the boot processor, which points to the + ``prototype'' page table kept for processor 0. mmuwalk walks + through the page table to get the entry. If pte is nil, there + is no second-level page table and mmukmapsync does nothing. If + the entry in the second level page table is is nil, the same + happens. In our stack page fault example, there is no entry + added for the new stack page. Therefore, in our case, + mmukmapsync does nothing. + What is mmukmapsync doing then? Try to guess, later I'll tell + you. + trap.c:446 + If bit 2 is set in the trap error code pushed by the processor, + the fault was due to a write operation. So read means that it + was a read the operation faulting at addr. + trap.c:447,449 + insyscall records whether the faulting process was executing + within the kernel (e.g. sysexec) or was running user code. In + any case, you are now running within the kernel. fault is + called to do the actual processing... + + trap + fault386 + fault() Services a page fault. + + ../port/fault.c:9 + ...and it is given the faulting address and an indication of + whether it was a read the operation causing the fault or not. + fault.c:14,15 + The routine saves the previous process `ps' state (which would + be restored later) and lets `ps' know that the process is + faulting. As you can see, the author uses psstate to be more + descriptive regarding the process state; the process scheduling + state is a different thing. + fault.c:16 + Until now, interrupts were disabled--which also prevented + context switches to a different process. Now that the page + fault is being handled (and has saved cr2), interrupts can be + allowed. Servicing a page fault may take a long time. + fault.c:18 + Accounting for the local processor. + fault.c:19,38 + seg locates the segment where addr stands, if there is no such + segment, the address was outside segments used by the process, + and the fault cannot be repaired (hence the return -1). If the + fault was for write and the segment was a read only segment, + the fault cannot be repaired either. Otherwise, fixfault would + do its best to repair the fault--e.g. by allocating a new page + frame, filling it with the contents of the faulting page, and + repairing the address translation. As fixfault may fail due to + allocation failures, etc., fault loops until the fault is + either repaired, or is known not to be repairable. Finally, the + saved `ps' state is restored and fault returns 0 to indicate + that it repaired the fault (because fixfault returned zero and + the loop was broken). + ../pc/trap.c:450,459 + Before looking fixfault and seg, note that when fault returns, + if it returns zero, the insyscall state is restored, and + fault386 returns to trap, which would return (or context switch + to another process) causing a return from interrupt. The iret + restores the processor context and the faulting instruction is + retried. However, when fault returns -1, fault386 would either + cause a panic (if the faulting instruction was within the + kernel) or post a ``sys:trap:fault'' note to the faulting + process. That note can kill the faulting process. In our + current example, the fault will be repaired. + seg()Locates a segment given the address. + ../port/fault.c:359,380 + First, fault calls seg with the pointer to the current Proc, + the faulting address, and dolock set to true. seg iterates + through the seg array of up, looking for a segment in use (they + have a Seg hanging from seg[]) whose addresses contain addr + (note n->base and n->top). If such segment is found, a pointer + to the Segment is returned. + If dolock was true, the Segment is locked and the check is + repeated; to ensure that the segment was still there and its + addresses still contain the faulting address. + + trap... + fault + fixfault() Tries to fix a repairable page fault. + + fault.c:50,51 + Should a segment contain the faulting address (seg[ESEG] in our + case), fixfault is called for it. In this case, doputmmu is + true--because fault wants the address translation to be ok for + the hardware too. + fault.c:61,64 + va is the faulting address; addr is set to the page address (by + clearing the offset bits). soff is set to be the offset in s + for the faulting address. p is a pointer to the map entry for + the segment offset. + Segment maps contain entries relative to the base address of + the segment. The segment offset is used as an address to be + translated by the map--very much like the hardware does with + its page tables. PTEMAPMEM is the number of bytes addressed by + each map entry (../pc/mem.h:102,103 defines it as 1M, and + defines PTEPERTAB as the number of pages needed to cover that + Mbyte). + fault.c:65,66 + One thing which can happen (that's the case for us), is that + the segment does not even have a map entry allocated for the + faulting offset. ptealloc()Allocates a Pte. + page.c:466,475 + In this case, ptealloc allocates a Pte structure with PTEPERTAB + entries to link Pages on it. This is like allocating the second + level page table for the ``virtual MMU'' used to implement the + segment. All entries in the Pte are still nil. + fault.c:68 + etp is now a pointer to the Pte which should contain the Page + structure for the faulting address. + fault.c:69 + pg is set to point to the entry in pages where the Page for the + faulting address should be. The index is the offset within a + map for the segment offset, divided by the number of bytes in a + page. In our case, *pg is nil as nobody allocated a page for + the stack. + fault.c:70 + type contains the kind of segment handled. More later. + fault.c:72,75 + A Pte, contains first and last pointers that point to the first + and last used entries. They are used to iterate through all + used pages without having to iterate through all entries in the + Pte--which could contain just a few contiguous pages. These + lines update first and last accordingly. + fault.c:77,80 + How to repair the fault, depends on the kind of segment; note + the defensive programming once more. + fault.c:82,88 + For page faults within text segments, the page should be paged + in from the text file. pio does the job, as discussed later. + fault.c:90,105 + For BSS segments, shared segments, stack segments, and segments + mapped (from devices?), which is our case, the pg is checked. + If it is nil, there was no page for the segment and a new one + should be added. This is called ``demand loading'' or + ``zero-fill on demand'' depending on whether the new page + should be loaded from a file or should be just initialized to + all-zero; the ``on demand'' part means that the system does it + only when a page fault shows that the process demands the page + involved. + + trap... + fixfault + newpage() Allocates a new page (frame) for a segment. + + page.c:119,131 + newpage is called to add a page to the segment at the given + page address. More precisely, the segment had its page + ``officially'', but that page had no page frame; newpage + allocates a Page that represents a page frame and gives it to + the segment virtual memory page. + Clear was set to true to request the new page be cleared. By + now, consider that there are more free pages than the value of + swapalloc.highwater and the loop trying to allocate a page + breaks at :131. Forget most of newpage now, which is discussed + later, but note that by allocating a Page, the author allocates + a page frame too (the one at p->pa). + page.c:185,197 + The other thing of interest for us now is that a reference is + added to the Page (it is being added to a segment), its va is + set to the virtual address for the page, modref set to zero + (the cache of the hardware bits in the page table) and the + actual page frame for the page (at VA(kmap(p))) set to al + zeroes. The page frame is allocated for the page, and the page + is represented by the Page structure. More clear now? + + trap... + fixfault + newpage() Allocates a new page (frame) for a segment. + + fault.c:100,101 + Back to fixfault, after new is our new Page, it could be that + newpage did set s to zero because it had problems to allocate a + new page for the segment. In this case, the fault cannot be + repaired and the routine returns -1. Despite fixfault saying + that the fault is not yet repaired, fault retries + later-hopefully, a Page for the segment could be allocated in + the future. + fault.c:103 + The entry in the segment Pte for the page (pg) is set to point + to the new page. The segment has a new page which has a page + frame for it. The goto does not need to be there because the + case would fall through the next case. Probably in a previous + version there was a goto common at some other point in the + routine. + fault.c:107,108 + For our stack segment, a page is allocated on demand if no page + was there (the stack is growing), and the same happens for BSS + segments (which are all zero, so zero filled pages can be + allocated on demand). For a page fault on a data segment, + processing would start here. + pagedout()Is the page paged out? + fault.c:110,111 + pagedout (portdat.h:350) returns true if the segment actually + has the page, but the page is not really in memory because + either it was paged out (its page frame reclaimed for other + uses) or it was never paged in (never read from whatever file + it comes from). onswap (portdat.h:349) checks whether the + PG_ONSWAP bit is set in the pointer to the Page; which is the + convention for pages paged out. Thus, if the page was never + paged in, pagedout notices that the pointer is nil and says + that it was paged out (it lies). If the page was actually paged + out, its page exists, but the pointer has the PG_ONSWAP bit + set. In any case, the page has to be brought into a page frame, + before it could be used. pio does the job of paging in the + page. + In our current example, pagedout would return false. + fault.c:113,117 + If the access was for read, mmuphys is set to be the contents + of the page table entry (for the hardware MMU) for the faulting + page. PPN returns the page frame (physical page) number for the + entry, and bits for ``read-only'' and ``valid'' are set on it. + Besides, the software copy of the ``referenced'' bit is set. I + defer the discussion of copymode until later. Just note that if + the page fault was because the page was missing, it is now + in-memory, and the ``valid'' bit is set. The switch is broken + because that is all that has to be done to repair the + fault--but for updating the MMU page table entry. + fault.c:119,148 + The number of references for the page is computed. For us image + in the page is nil, so the number of references is just the ref + field in the Page. In our case, there is just one reference to + the page and code in :127,140 does not execute. As there is no + image for this page, only the unlock is done--no duppage + called. All this will become more clear for you later. But + let's concentrate on how are pages added to our stack segment. + fault.c:149,151 + The fault was for a write access, so fill up the entry + (mmuphys) for the hardware page table with the page frame + number and the write and valid bits. The ``modified'' and + ``referenced'' software bits are set in modref. putmmu()Updates + an MMU entry. + fault.c:173,176 + If the caller requested that the hardware MMU page table should + be updated, putmmu updates the entry for addr with the the + prototype in mmuphys. The Page structure is passed because on + some architectures putmmu might need to use/update Page + information, but that's not the case for the Intel. After + putmmu returns, the hardware page table has a valid address + translation from the faulting page to the just allocated page + frame. fixfault returns zero to state that the fault was + repaired, and fault would return to 386fault which would return + to trap. + At last, the faulting processor context would be reloaded and + resumed by the iret in l.s. In this case, the faulting + instruction was one in sysexec.c, filling up the stack for the + process; execution would continue from that point on. + The processing just described would also be the one when + addresses within the BSS segment are first referenced. The BSS + is a data segment initialized to all zeroes. By attaching pages + to it as they are used, and initializing their page frames to + all zeroes, the process can believe that the whole BSS segment + was there right from the beginning. The same happens for stack + segments, as you now know. + +Text and data memory pages + + trap... + fixfault + + fault.c:83,88 + If the process first references a page within the text segment, + these lines are reached. Processing is mostly like servicing a + page fault for the stack segment, but there are important + differences. Assuming that the page faulting was never brought + from the text file to the text segment, pagedout would find *pg + to be nil, and return true. pio is called to do page I/O on the + text segment. After it loads the program text that should go in + the page from the text file, mmuphys is updated with a + translation to the page frame for reading (that means + ``execute'' permission too). Later putmmu will install that + translation in the MMU page table. + + trap... + fixfault + pio() Performs I/O to do a ``page-in'' for a page. + + fault.c:180,191 + pio tries to get into *p a Page (with the associated page + frame) with its corresponding memory filled up according to + what is said in s. + fault.c:192,199 + If there is no Page pointer (which means that the Pte had a nil + pointer for this page), the page contents must be brought in + from the image attached to the segment. daddr is the address in + disk for page contents. The address is fstart (the address in + the image corresponding to the start of the segment) plus the + offset within the segment for the faulting page. (Remember that + there are different fstart values for text and data segments?). + lookpage takes the Image attached and the address on it, and + returns a cached Page for that image portion. Hopefully, the + Image would be caching that page most of the times, and no + access to disk would be required: new would be non-nil and pio + is done. Should new be nil, you have to go to disk to read page + contents. This is the if arm taken for the first reference to a + text (or data) page. + fault.c:200,208 + Should there be a pointer to a Page, that means that the page + was paged out. The pointer (as you will see) is not really a + pointer to a page, but a daddr with the PG_ONSWAP bit set. When + low on memory, Plan 9 reclaims page frames from user pages. If + a stack or a BSS page is reclaimed its page frame, page + contents must be stored somewhere else[246]10.2; i.e. on the + swap area. In this case, swapaddr returns the address in swap + where the page copy stands, and lookpage uses the swapimage + Image instead of the segment image. putswap marks the space + allocated for the page within the swap area as no longer used. + Swap space is allocated just to keep the pages moved out from + system memory. Unlike other systems (e.g. some UNIXes), Plan 9 + does not keep the swap space allocated when the page is kept in + memory. If the page ever needs to be paged out again, another + piece of swap space would be allocated for it at that point in + time. + fault.c:211,215 + The page was not found in the Image. It is definitely not in + memory. A new Page (with a fresh new page frame) is allocated. + The page frame is mapped at kaddr. + fault.c:217,252 + The page has been first referenced (see above). + fault.c:218,225 + About to read from the channel to the file where s->image + memory comes from. In case of error, release the page just + allocated and call faulterror--which would either pexit or post + a debug note. If page contents cannot be retrieved, there is no + much else to do. The routine cannot return by calling nexterror + because, in the end, the faulting context would be reloaded, + another page fault happen, another I/O error for the channel + happen, etc. + fault.c:227,231 + The read procedure for the channel is called to read into kaddr + (the page frame), ask bytes, starting at offset daddr (the + address for the page in the file). Lines :227,229 set ask to + the page size or the number of bytes from the faulting address + to the end of the segment--whatever is the minimum. The end of + a segment does not need to be aligned at page boundaries. For + example, a compiled file can have initialized variables (data + segment contents) which could occupy just 1K bytes, much less + than a page size; it would not make sense to read more than + that Kbyte from the file. + fault.c:234,235 + Remaining bytes in the page (3K in the example) would be set to + zero. After these lines, the page is loaded in memory. + fault.c:239,251 + While the page was being read into memory, the lock on s->lk + was released--reads take a long time. That means that some + other process could fault on the page too, and start to read it + too. The first process reaching line :239, would notice that + the Pte entry (*p) is still nil. So that process takes the + responsibility of attaching the page to the segment: its daddr + is set to the daddr computed, cachepage is called to let the + Image keep the page cached, and the Pte entry is set to point + to the page. The second process arriving here, would notice + that *p is not nil, which means that the work is done, so it + does nothing but to return (cachectl is not discussed here). + The call to putpage releases the reference to new, which could + cause the page to be deallocated when the number of references + becomes zero. + The author prefers to let one process do some unuseful work + some times (when faulting on a page being faulted by other + too), than to keep the whole segment locked (which would block + processes using that segment) just to avoid this (not so + probable) race condition. + One more note, pio does not fill up any MMU page table entry. + It just handles the virtual page table used by the segment, and + does Page I/O. The caller should call putmmu to let the + hardware know. + + trap... + fixfault + + fault.c:110,111 + For data pages first referenced, pio is called too, and + processing is like above. + fault.c:144,145 + However, for data segment pages first referenced (unlike stack + pages), duppage is called when there is enough space in the + swap area. What is happening is that data pages can be written. + If the data page is written, its contents would differ from the + disk file contents. + Now that the page is still fresh (it is just read), the author + prefers to employ a bit of time and memory to make a copy of + the page. The copy is to be kept by the Image, so that when + another process faults on this page, the initial contents do + not need to be read from disk, but from the Image page cache + instead. + +Physical segments + + fault.c:154,169 + If the segment is a bunch of physical memory, servicing the + page fault is done by allocating a Page structure for the + physical page already assigned to the segment. The way to + allocate the Page depends on whether the segment has a + pseg->pgalloc function or not. If it has one, it is the + provider of Pages, otherwise a Page is allocated and its pa is + set to point to the pseg->pa address of the segment plus the + offset for the faulting page in the segment. Physical segments + are discussed together with segattach. + +Hand made pages + + main + userinit + segpage() Adds a page to a segment eagerly; not on demand. + + segment.c:222,243 + segpage is used only during boot to add a page to a segment. + Page is supposed to be initialized by the caller, and segpage + only plugs the page in the appropriate Pte for the segment. + + Page allocation and paging + + You now know that segments are filled up with pages on demand. Let's + see now in more detail how are page frames allocated when segments + reclaim more memory. + +Allocation and caching + + auxpage() Allocates a page frame. + + page.c:240,246 + auxpage is called to allocate a Page structure, along with an + associated page frame. It is used by the code in cache.c to + allocate page frames for extents. Pages come from a free list + of pages in palloc. They are taken from head. + page.c:247,250 + freecount was initialized by pageinit to the number of free + page frames. For each page frame, a Page structure was + initialized (with its pa set to the page frame address) and + linked into palloc list (palloc.head/palloc.tail). + swapalloc.highwater was also initialized by pageinit to be + 5/100 of the number of page frames available for users (not for + kernel). So, if the number of free pages goes down a 5% of + available memory for users, auxpage refuses to allocate one of + the (now precious) free page (frames). + When the author uses auxpage, allocation could fail. An example + is cchain (cache.c:383), which does caching only if free pages + are available. So, what is happening is that the kernel cache + for remote files consumes only free pages, but refuses to grow + when memory is scarce. + Could the author use virtual pages to do caching? Yes, but in + that case they could go to disk, and reading from a disk can + (sometimes) be slower than reading from the network. Besides, + in any case, a local file server can be used to cache remote + files (e.g. cfs). + pageunchain()Removes a page from the palloc list. + page.c:251 + pageunchain (page.c:65,80) removes p from the palloc list and + adjusts freecount (one less page). The list is double linked + using the next and prev fields of Page--to remove any entry. + Saw how the routine checks that the palloc lock is held? + page.c:253,261 + The page should not be used (hence the ref check). A reference + is added to the page, uncachepage called, and the page returned + to the caller. The reason to call uncachepage is that the + caller is going to use this page for new stuff. Let's see what + this means. + + Images are used to represent an image in memory (read: cache) of file + contents. You should remember that lookpage is called with an Image + and a disk address to recover a cached Page for that offset within the + image. How can that be done? + + cachepage() Adds a page as a cache for part of an Image. + + page.c:365,385 + cachepage is called (as you know) when an image page is read + from its file. It locks the palloc.hash table, and adds the + page to the hash bucket for the page. pghash uses the daddr + (ignoring page offset) to hash the page. Since the page is + being kept for caching part of i, its image field is set to + point to the image and a new reference added to the image. The + whole picture is shown in figure [247]6.4. + + CAPTION: Figure 6.4: Palloc cooperates with Images, so that a cache of + pages is kept for Images. Pages are hashed on their daddrs and + maintained on a LRU list. + + lookpage() Lookup a page in the cache. + + page.c:411,439 + When later a routine wants a page from the image, lookpage + scans the hash bucket (given the disk address) for a page with + the same Image and the same daddr. If the page is found (note + the double check to avoid locking all the pages) a reference is + added to it (:428). + When an image is released by the last segment using it, its + pages are still kept in palloc.hash. If no one is using such + pages, their reference count would be zero. However, such pages + still contain file data which would save disk (or network) + reads. Besides being kept in the hash, free pages are linked + through the list starting at palloc.head. The call to + pageunchain removes the page from its list (the palloc.head + list). If this is the first new reference, the page must be + removed from the free list, but second and posterior new + references would find the page out of the free list (the page + might be reused multiple times if found by lookpage for + different segments). + + uncachepage() Removes a page from the cache. + + page.c:342,363 + Finally, pages leave the cache (are removed from the hash) when + uncachepage is called (e.g. after auxpage allocates an unused + page for a different purpose). At this point, the page is no + longer caching part of the file image in memory. + + To summarize, pages are initialized to represent fresh page frames + during boot. They are allocated later from the free list (initialized + at boot) and added to the hash for use as a cache. When their last + reference goes away, they are added again to the free list. Only when + they are used for a different purpose, they are removed from the + cache--in the hope that the same file will be used again (e.g. ls + would be executed once more). As long as there are free page frames, + nothing else happens. But what happens when physical memory is not + enough for segment pages and to cache file contents? + + newpage() Allocates a page. + + page.c:118,119 + newpage is called to allocate a new page for the given segment. + page.c:130,133 + For user processes, less than a 5% of free user pages is + considered as ``no more free pages''; kernel processes would + still get their pages until really out of memory. During this + loop, the caller process would be waiting to get a free page + for its usage. + page.c:135 + While waiting, release the palloc lock. This allows other + processes to use it (e.g. to release pages). + page.c:136,141 + If the caller supplied a pointer to its Segment*, which means + that it was allocating a page for that segment, the author + releases the segment lock, clears the Segment* used by the + caller and sets dontalloc. Later, after trying to get more free + pages, the routine would return (:154,161) without trying to + allocate a page. Can you guess why the author does this? + page.c:142,152 + In this critical region, the process calls kickpager and sleeps + for a while waiting for free pages. Hopefully the pager process + notified by kickpager would make more pages available, by + stealing pages from someone else. If there are several + processes allocating free pages, they will enter this region + too, until the point when one of the checks at :130,133 + succeed. + page.c:154,161 + When memory is considered to be full, the process would loop, + waking up the pager and sleeping for a while repeated times, + until the pager gets free pages. This usually requires disk or + network I/O and can be a really slow process. Now, if the + caller is servicing a page fault on a segment, it would make no + sense to keep the whole segment locked just to await for a free + page. The segment could be the text of ls and many other ls + processes could perhaps keep on running without the faulting + page. What the author does is to let freepage release the lock + of the segment (:138), and tell the caller (:139) that the + segment is no longer locked. For example, in fault.c:99,101, + fixfault would notice that it had to wait, and did loose the + segment lock, for slow I/O. It would fail and let fault retry + later. + page.c:166,169 + Got some free pages (according to freecount). Get one from the + free list. Due to processor caching issues, not all pages are + the same. The algorithm assigns ``colors'' to pages, so that + it's best to allocate a ``red'' page frame for a ``red'' page + (:127). Only a few colors are needed (like colors in a map). + To remind you, this is because the processor (hardware) cache + sometimes use the same cache entry depending on the (physical, + or virtual, depending on the architecture) address of a page. + For example, a processor with two cache entries (ridiculous, + but just an example) could use entry 0 for even page frames, + and entry 1 for odd page frames. If a process is using a data + structure between two pages, it is better to allocate even page + frames to even pages, and odd page frames to odd pages. This + way, the two pages can be kept in the processor cache at the + same time. I hope you get the picture with this silly example, + but note that for the Intel, getpgcolor always gives 0 as the + color (../pc/mem.h:120). So the author does no page coloring + for the PC. + page.c:171,176 + No free page for our color, just take the first one. ct is used + to control the cache. By now, note that PG_NOFLUSH or PG_NEWCOL + is set depending on whether a page of the right color was found + or not. + page.c:178 + The page removed from the free list. + page.c:180,187 + The author checks that the page was indeed free (no + references), and removes it from the cache. This page could + belong to a different image and be placed on the free list by + the pager. In any case, the page is going to be used for a + different purpose now. + page.c:188,189 + The page has a cachectl array with an entry per processor. For + all processors, entries are set to either PG_NOFLUSH or + PG_NEWCOL depending on the page color. On Intels, cachectl is + not used, but other architectures might use it to determine + whether the cache entry for the page should be flushed or not. + This is because there are architectures that fill up the cache + using virtual addresses, if a page has changed its virtual + address (the frame is reused for a different page), its entry + on the cache might need to be flushed if the hardware wouldn't + notice that and flush the entry automatically. + page.c:193,197 + Zero the memory if requested. + +Paging out + + kickpager() Starts the pager process or wakes it up. + + swap.c:90,101 + kickpager is used to ask the pager to do its job. Its purpose + is to get some free pages. Because of the static started, the + first time the system is running out of (physical) memory, a + kernel process is started to run the pager function[248]10.3. + Next times, only a wakeup for swapalloc.r is issued. Let's see + pager. + + pager() Main routine for the pager process. + + swap.c:103,118 + pager keeps running under the loop to get more free pages. To + prevent it from running even when there is free memory, the + author makes it sleep at line :118. It will stay there until + awakened by a process calling newpage. When the first call to + kickpager starts the pager, needpages can prevent the pager + from sleeping. + swap.c:120 + The pager will not sleep until it has managed to free some + pages. + swap.c:122 + swapimage.c is the channel to the swap file (or partition). If + there is such channel, some pages can be paged out to swap + space (i.e. copied to the swap file and their frames reused as + free memory). + swap.c:123,125 + p is going in a round-robin fashion among existing processes. + All configured process entries are scanned (:133,114). + swap.c:127,128 + Dead processes are not using memory and are skipped (note that + dead is not broken), and kernel processes are given kept + untouched (they run within the kernel, don't they?) + swap.c:130,131 + If canqlock acquires the look, it is the turn for this process + to donate some of its pages to the Plan 9 cause. Otherwise, the + process is touching its segments. Instead of blocking the pager + process, the author chooses another victim. Although the + algorithm is not fair, it is better to use an unfair algorithm + than it is to let the pager block while free memory is needed. + swap.c:133,137 + Starting to iterate over the process segments, seeking for + pages to steal. As soon as needpages says so, no more pages are + stolen. The segment lock is kept for the whole search. + swap.c:139,160 + Got a segment for the process (an used entry). Only segment + types mentioned in the switch get pages stolen. pageout is the + page thief. When it is a text page the one stolen, the `ps' + state is not changed (because the page comes from the text file + (read-only) and there is no need to maintain the process locked + in pageout for too long. For data pages (including stack and + others), the `ps' state is set to Pageout during the call to + pageout and maybe later to I/O, during the call to executeio. + Let's defer a bit pageout and executeio. As you just saw, all + (user) processes are candidates to get their pages stolen in a + round-robin manner. + swap.c:164,174 + If there is no swap file defined yet, freebroken is called on + terminals to kill broken processes and reuse their memory. On + CPU servers, killbig is used instead. A message is printed to + let the user know that either more memory should be added to + the system, or a swap file configured. The call to tsleep + prevents the message from appearing too frequently. + + pager + killbig() Kills a big process and reclaims its memory. + + proc.c:1156,1190 + killbig locates the process with the biggest memory image (sum + of the lengths for its segments), and kills it. The author + knows this is not fair, and prints a diagnostic to let the user + know. However, the author hopes that by killing this big + process, the CPU server could get out of the out of memory + condition. Nevertheless, the injured user could ask the CPU + server administrator to configure a swap file if that's the + problem. The action really killing the process is a procctl + order of exitbig, so the process will kill itself later when it + checks its procctl. But in any case, as the memory is needed + now, mfreeseg is called to release the memory of all user + segments in that process, so that it be available now; mfreeseg + is discussed later. + + pager + pageout() Steals pages from a process segment. + + swap.c:179,180 + pageout tries to steal pages from the given process and + segment. + swap.c:186,187 + When the author services page faults, lk is acquired and + newpage can be called. Now, newpage can awake the pager which + can try to get the lock to do some page outs--skipping locked + segments, (note that this actually means Pte map locking, and + not seg locking). + By using canqlock, the worst thing that may happen is that + other segment is used to steal pages from, not a big deal when + compared with a deadlock. You should note that the pager does + not want to block as it should get free memory soon. Besides, + by avoiding races against page operations on the locked + segments, the author can forget about race conditions in that + respect. + swap.c:189,192 + steal is non-zero while procctlmemio (devproc.c:981,1049) is + doing I/O on segment memory. That is precisely to prevent the + pager from paging out segment pages under devproc feet. + swap.c:194,198 + Only if canflush, are pages stolen from this segment. As + canflush adds an extra reference to the segment, putseg must be + called to release the reference. The reference avoids segment + deletion while it is being used to page out some of its pages. + + pager + pageout + canflush() Anyone running on pages from this segment? + + swap.c:235,265 + canflush returns true if canpage returns true for all (alive) + processes using the given segment. When there is more than one + reference, all processes must be searched to find other users + of the segment. + proc.c:283,299 + canpage returns true only if the process is not running, and it + sets newtlb to true in such case. + What is going on? If the process is running (pager is just + another process, and there can be multiple processors), the + author refuses to check if the page is really being used or + not. It is more simple to remove pages from processes not + running (note the mach check!). When the process runs again, it + will notice the newtlb flag and flush its MMU (because page + translations are going to change in pageout). So, only when + none of the processes using the page is running, can pageout + steal the page. + What if a process which said it canpage runs after the call to + canpage but before pageout completes? That is no problem, + mmuswitch would notice newtlb and will call mmuptefree to set + as invalid the entries in the MMU page table for every user + page in that process. So, pageout really hurts to the process + affected. Even if it gets just a few pages paged out, it will + suffer many page faults. + + pager + pageout + + swap.c:207,228 + For all (user) Pages in the segment (first and last are used to + avoid scanning all the map entries), if the Page is pagedout + (never paged in, or paged out) it is ignored. If the page has + the referenced bit set, it is forgiven and it has a new chance + to be referenced again before the next pass of the pager for + this page. If the page is not referenced (or was forgiven and + not referenced again before being reached once more), pagepte + steals the page. This is a second-chance paging out policy. If + the author did not forgave pages referenced, pages really in + use by the process could be paged-in soon, and then paged out + again, and the system could end up trashing (it would do + nothing else but to service page-ins and page-outs). + swap.c:225,226 + ioptr points to the current page I/O transaction. If nswppo + page I/O requests are in place, the system refuses to do more + page I/O. I think that the author tries to avoid trashing here + too. If after finishing the current transactions, memory is + still scarce, more page outs would happen. Another good reason + not to do too much I/O to get free memory is that the user is + waiting for his application to run, and the application is + waiting because the processor is being used for the pager too. + + pager + pageout + pagepte() Updates a Pte for a page out, adding the page to the I/O + list. + + swap.c:267,274 + Paging out a page. A pointer to the pointer used by the caller + is passed, so that pagepte can update it for the caller. + swap.c:275,278 + If the page is a text page, it can be paged-in later from the + text file. The caller Page* is cleared and the reference to the + page released. That's all to do here. This is the reason why + psstate was not changed for the process while doing a page out. + Because, in fact, there are no ``page outs'' for text pages, + they are simply discarded. + swap.c:280,290 + For these segments, a copy of the page memory must be made + before reusing its frame. newswap allocates a new disk address + (within the swap file) where to copy the page. + newswap()Allocates space for a page in swap. + swap.c:35,56 + newswap scans a bitmap swapalloc for a zero byte--which + represents room for a page in the corresponding offset for the + swap file. last and top are used to do kind of a next-fit + policy for swap space allocation; last is kept set as the last + slot found at look. Initially, last is initialized to point to + the start of the ``byte-map'' in swapinit. Line :51 is marking + the byte as allocated. The author trades space for time, by + using bytes and not bits in the map. It is more simple to use a + whole byte than it is to use a bit (it should be masked and + checked). Memory is cheap these days. + By the way, now you can understand why fault.c:122,123 called + swapcount to account extra references for Pages that had + swapimage as their image. swapcount (swap.c:84,88) returns the + ``1'' or the ``0'' in the swap ``bytemap'' entry for the page. + So, code in fault.c:122,123 accounts one extra reference for a + swap file Page if the swap bitmap states that such page is + being used. As you will see soon, when (non-text) pages are + paged out, their ref could reach zero. + swap.c:291,292 + newswap returns -1 (in two's complement, note the unsigned + return value from newswap) when there are no more free swap + slots for pages. The routine refuses to page out if there is no + place to copy page contents. + swap.c:293 + cachedel removes any cached page for the swap image at disk + address daddr. Remember that lookpage can try to get a cached + page for a disk address? In the past, this disk address could + haven been used to keep other page, and it could be that the + palloc hash (cache) still has a cached pages pretending to be + the contents for this file slot. No longer the case. The pages + caching daddr would still be in the free list, but it would not + be in the hash list any more. + swap.c:295,298 + In the same way, the page being paged out can be linked into + the palloc hash, as a cache for part of its image. uncachepage + removes the page from the hash, and sets its image and daddr to + zero. Now the page is no longer for its old image--uncachepage + expects the page to be locked. + swap.c:300,317 + The comments say it all. But note that PG_ONSWAP is set (see + pagedout). Although the page is not yet sent to the swap file, + the kernel considers it as swapped out. The space for the + pointer to the Page in the segment is used to keep the daddr + for the page in the swap file. + Now pagepte completes and pageout would perhaps loop adding + more I/O requests to iolist by calling pagepte more times. When + the configured I/O limit is reached, or when the segment is + entirely scanned, pageout completes too. Later, lines :154,156 + would call executeio. + + pager + executeio() Performs I/O for page outs. + + swap.c:331,366 + All I/O requests placed in iolist are serviced at a go. For + each page set in iolist, kmap is called to set a temporary + mapping so that write can be called for the swap file channel + to write the page contents. The segment reference to the page + is not released (putpage called) until write returns; i.e. + after page contents are safe in swap. Besides the segment (Pte + entry) reference, there was another extra reference which is + removed at :361. As the process had the segment unlocked before + executeio executes, the segment could call putpage on its own. + If you look at putpage, (page.c:209,238), you will see that it + checks the PG_ONSWAP bit and calls putswap to release pages + with the bit set. However, pagepte did set the bit in the + Segment pointer to the page, but not in the iolist pointer to + the page. Therefore, the putpage call from executeio really + does a putpage. + By deferring page I/O requests until after the segment is + unlocked, the time the segment lock is held is reduced a lot + (I/O takes a long time). In the mean time, processes could + service other page faults for the segment. This is very + important since mmuptefree clears page table entries for the + process affected and it will have to service many page faults + for the segment. + As a final comment, note that the PG_MOD bit is not checked to + avoid calling write for a page which has not been modified + since it was last read (e.g. from the data section of the text + file). Although that could save some I/O, the author probably + thinks that it is not worth to do so. Take into account that he + duplicate made for data pages before they could be written + helps here. + Should the author change his mind in this respect, pages backed + up by swap space would need to keep swap space allocated even + while in memory, so that pages not changed since their last + page-in could be discarded without I/O. + +Configuring a swap file + + syswrite + conswrite + + devcons.c:847,864 + A swap file is configured by a write to #c/swap. Usually, the + string written is the number of an open file descriptor for the + file to be used as a swap area. A write of the string start + would kickpager instead of configuring the swap file. For CPU + servers, only the boot user (Eve) can configure a swap + file--otherwise, any user could read memory from other user's + processes by configuring a swap file and forcing the server + into a low-memory condition. Since terminals run processes on + the name of eve, the author does not check anything for them. + The work is done by setswapchan. + + syswrite + conswrite + setswapchan() Configures a swap channel. + + swap.c:374,403 + The important work is done at line :402. Lines up to :386 take + care of unconfiguring a previous swap file if it was not used + (All nswap pages are free); the new file will be used instead. + Lines :392,400 limit the number of pages in the swap file to be + those that fit in the partition. Surprisingly, if a previous + swap file existed, nswap would only decrease, and not increase. + It will always be at or below the value configured at boot + time. By the way, the check for `M' is because all `files' come + from the mount driver--this is a CPU/terminal kernel; so, if + the device is not #M, it is likely to be a kernel device who + provided file, and that's usually a disk partition. Perhaps a + more explicit check against the storage device could be + made--but what about using a kfs file? Hint: what if kfs data + was swapped out? + +Paging in + + You already saw most of the code needed to page-in some pages for user + segments, while learning how are pages added to segments: Plan 9 uses + demand load for pages. Let's see now the part of the code we didn't + read. + + To remind you, fixfault called pio to do a page-in for the faulting + page. You saw how the cache for the image (or the swap file image) was + tried by pio, and how pio worked when there was not a Page for the + page. + + trap... + fixfault + pio + + fault.c:253,269 + Now, when it appears to be a pointer to a Page hanging from + loadrec (from the segment map), the page faulting was paged + out. Moreover, the page faulting is not a text page because + text pages have their Pages deallocated on page-outs. So, the + page must be read from the swap file pointed to by + (swapimage.c). Remember that at this point in pio, new has a + fresh page (frame) to be used for the faulting page. + The calls to qlock/qunlock within the error recovery block at + lines :258,259 seem to be to wait until sure that no other + process is holding the lk lock. This routine acquires that + lock. faulterror can call pexit to kill the process. Therefore, + if the I/O fails to do a page-in, the process is killed after + nobody is running within the critical region. + fault.c:271,286 + During the call to read (which can block), the segment was + unlocked. Another process could initiate a page in. This case + is easy to check because it that did not happen, loadrec would + still be *p, the entry in the Segment. If loadrec differs, that + must be because at line :290 the other process did set the + entry to the page it allocated. The first one getting past line + :269 wins. If another process did the page-in, the routine + releases the page allocated (and read!!) by this process: all + done. It could also be that the pager stole this page, which is + known because pagedout returns true (and *p changed too!). In + this case the page has to be paged in again, hence the goto. + Perhaps the author could save some reads by allowing at most + one process to do the read for a page-in. Nevertheless, it is + not clear that would be worth because several processes must be + faulting on the same page for it to be worth. Besides, the + saved time would be minimized because of caching. Remember? + Measure and then optimize, not vice-versa. + +Weird paging code? + + If you are curious, you already noticed how there are still portions + of paging code that remain to be read. In particular, + + trap... + fixfault + + fault.c:113,117 + This code restores the hardware MMU permissions for the + faulting page when it is a read-fault and copymode is zero and + avoids doing any other thing. + fault.c:127,140 + This code does something which is not calling duppage when + there is more than one reference to the Page (including as + references pages swapped out). If you look at the code, it + allocates a new page and calls copypage to install a + translation to a copy of the faulting page. + + To understand what is going on, you must read dupseg first. That + procedure clones a segment during a rfork. + + Duplicating segments + + sysrfork() + dupseg() Duplicates or shares a segment. + + segment.c:143,144 + During rfork (sysproc.c:107), dupseg is called to duplicate a + segment or request that it be shared with the parent process. + seg is the process segment array. The routine is expected to + return a pointer to the duplicated/shared segment. + segment.c:153 + Segments are duplicated one way or another depending on the + kind of segment. + segment.c:154,159 + For segments that are read-only (text), shared (SG_SHARED and + SG_SHDATA), or physical memory (which is shared), the reference + count is incremented and the parent's segment is used as-is by + the child. It does not matter what share says. + segment.c:161,169 + Stack segments are never shared. newseg creates a new stack + segment with the same base and size as the original segment. + This new segment is still empty (although it should be a copy + of the parent's stack). + segment.c:212,219 + Later, dupseg calls ptecpy to copy each Pte in the original + segment to the new (duplicate) segment. Let's see what happens + to other segments before looking at ptecpy. + segment.c:171,186 + For BSS segments (and MAPs), one of two things can happen. + If the segment is to be shared, and nobody is sharing the + segment (its reference is one), the segment type is changed to + be SG_SHARED. From now on, calls to dupseg for this segment + would just add another reference, as seen before. You now know + that a SG_SHARED segment is a BSS or MAP segment that is being + shared. The routine adds the extra reference and returns the + segment as the duplicate one. + If the segment is not to be shared (!share) a new segment is + created, and later, ptecpy would copy Pte entries for the new + segment. Let's see now ptecpy. + + sysrfork() + dupseg + ptecopy() Copies PTE entries. + + page.c:442,464 + ptecpy is an innocent looking function, which allocates new + Ptes for the duplicated segment and iterates through all (used) + Pages in the old Pte. For each entry used, if the page was + swapped out, dupswap is called. But otherwise, an extra + reference is just added to the Page, and the the Page* in the + destination entry is copied from the source entry! Noticed that + the segment is being duplicated, although Pages are shared? + Although the duplicate segment was expected to get a copy of + the pages, it gets the same pages. What's going on? + The duplicated segment has no real MMU page table entry + updated, so it is going to page fault on the pages ``copied'', + although its entries are set in its Ptes. + + trap... + fixfault + + fault.c:126,141 + When the process later has a page fault due to a write on the + ``copied'' BSS segment, the page has more than one reference + (for read accesses, the MMU entry is given read permission and + nothing else is done). + In this case, fixfault knows that the extra references are + there because although the page is being shared, it should not + be shared officially (pages really shared on shared segments + have a reference count of 1, it is the Segment that has the + reference count bigger than 1). This is called ``copy on + write'', or COW. When the segment is duplicated, the segment + and its Ptes are duplicated, but the pages are shared (with a + reference count bigger than 1). On a (write) page fault, the + page is copied (copypage) to a new page frame (new), and the + new page is given to the faulting process. there is a call to + putpage because this process is no longer using the shared copy + of the page. In figure [249]6.5 you can see a COW segment and a + shared segment. + + CAPTION: Figure 6.5: A copy on write (or copy on reference) segment is + like a copy of another segment: but both segments share unmodified + (unreferenced) pages. This is not to be confused with a shared + segment. The figure shows how each Page has an associated page frame. + + If other processes copied the segment (on write), the number of + references in the page after putpage is still greater than one, + and more copies will be made, if the page is referenced. If no + other process is sharing (copying on write) the page, its + reference is one, and lines :143,150 would execute instead: the + page is used as is, after saving a copy for the image cache and + restoring permissions in the MMU entry. + The check for copymode at line :113 is deciding when to really + copy the page. I have been saying that the page is copied on + write. I lied. Only when copymode is not-zero is the kernel + restoring MMU permissions (without any copying) for read + faults. When it is zero, even a page fault for reading causes + the page to be copied. So, if conf.copymode is zero, Plan 9 + does copy on reference, otherwise, copy on write is used. + copymode is zero unless archinit or mpinit set it to one, which + happens only for multiprocessor machines. So, Plan 9 uses copy + on reference for monoprocessors and copy on write for (Intel) + multiprocessors. + + NOTE: Why? + It is clear that the copied segment has all its MMU + translations invalid; but what about the original segment + copied on write? Any write to its pages should also cause a + page fault, but since the segment was read/write, translations + would still have write access in the MMU. + sysproc.c:182,187 + Is it clear the comment now? All translations are set invalid + by flushmmu. Page faults for read would just repair the MMU + translations. Page faults for write would copy the COW pages. + Although it would be faster to downgrade the translations to + read-only for COWed segments, the author prefers to keep the + code simple. Should this be a performance problem, perhaps + flushmmu would be replaced by a more clever routine which could + downgrade entries too. + + sysrfork() + dupseg + + segment.c:188,211 + Ignore the data2txt thing. If the segment is to be shared, add + a reference to it and change its type to SG_SHDATA; you now + know what's a SG_SHDATA segment. If the segment is not being + shared, but copied, COW is used and a new segment is created as + before. The difference with respect to COW for BSS segments is + that the image is added an extra reference and attached to the + segment too. BSS segments have their storage initialized to + zero, and they are never paged in from any image (well, they + are paged in from swap, but you know how that is done). On the + other hand, data segments page their pages in from the image + corresponding to the text file with initial contents for the + data segment. Even though the segment is COW, the copied + segment should also page in from the image those pages which + have never been referenced. So, the segment must be attached to + the image too. Besides this reason, it is good design to keep + the copied segment attached to the same image, as it would be + if it was the original segment: it is a copy, isn't it? + segment.c:189,190 + Back to the first two lines, you must read a bit of devproc to + understand what is going on. + procctlmemio()Does I/O to memory of a process. + devproc.c:998,999 + If procctlmemio is servicing a write for the text segment, it + calls txt2data. + txt2data()Replaces a text segment with a data segment. + devproc.c:1051,1078 + txt2data takes a text segment and replaces it by a data + segment. The data segment is a regular SG_DATA segment but, if + the text segment was the one in seg[TSEG], the entry in TSEG + now contains an SG_DATA, which is not usual. As a data segment, + it accepts writes, and that seems to be the reason why the + author is replacing an SG_TEXT with an SG_DATA. Although the + author could have set a mode field in Segment, and provide some + means to change it, it is more clear to have the type of the + segment determine what can be done to the segment. As this kind + of write seems to be most useful for debugging, it is not + likely to be a frequent operation and its efficiency is not an + issue. + It is useful to be able to write to the text segment to change + instructions, and to set breakpoints on it. + segment.c:189,190 + However, if the process forks, the child should get its regular + SG_TEXT segment in TSEG. dupseg knows and calls data2txt to + recover the text segment corresponding to the weird data + segment. + data2txt()Restores a text segment from a replacement data + segment. + devproc.c:1080,1093 + The only thing data2txt does is to recreate the text segment + from the image, as happened before during exec. + + By the way, the checks for ref==1 besides checking share during dupseg + seem to be more of defensive programming; since the segment type would + be changed just the very first time. But I may be missing something + here. + + Terminating segments + + putseg() Drops a reference to a segment. + + segment.c:83 + When a segment is being released, putseg drops the reference to + it. + segment.c:91,100 + Segments and images are linked circularly, the convention is to + lock the image before the segment. If the last reference to the + segment is being released, it will go away and the image should + no longer point to the segment. In this case, i->s is cleared. + segment.c:108,123 + The last reference is going. The routine clears this thing up, + including a call to putimage, if there is an image attached. + putimage does not call to putseg to release its reference to + the segment. You know why, don't you? + + putseg + putimage() Drops a reference to an image. + + segment.c:385,391 + putimage releases the image. It does nothing for images with + notext. You already saw in the starting up chapter that notext + was set for the first process text image; it is also set for + the swap image by swapinit and for the fscache image used to + cache remote files in cache.c. + segment.c:394,426 + For images with ``text'', after their last reference is gone, + their QID is cleared, and their channel to the text is closed. + The Image is also removed from the Image hash table and linked + to the free list (attachimage would no longer find this image + when searching for other users of the same text). One thing to + note here is how the channel is not really closed here, but + placed into freechan (which is resized on demand) to be closed + later. Closing a channel may block and also may take a long + time. The author wants that to happen after all locks have been + released. + + attachimage + imagechanreclaim() Releases channels used for images (as well as + images). + + segment.c:358,382 + imagechanreclaim is the routine actually calling close for + these channels. It is called from attachimage. Read the comment + regarding locks, it is very explicative. Although keeping these + channels open require file servers to keep resources that are + not really useful ( channels are about to be closed), the + author prefers to close the channels on calls to attachimage + rather than after calling unlock in putimage. One fine reason + can be that putimage is also called during page faults (which + already may take a long time). Perhaps the pager could be also + in charge of closing image channels. + By the way, you now know why imagereclaim puts pages instead of + images to get some Image structures released, because pages + keep the cached contents of the image and the image will not go + away until all its pages have been released. + + putseg + + segment.c:112,115 + Back to putseg, freepte releases the Pages used by the segment. + + putseg + freepte() Releases pages in a Pte. + + page.c:478,516 + For physical segments, a pgfree function is called to + deallocate pages (in the same way that the pgalloc routine was + used by fixfault). Besides calling pgfree, the Page reference + count is decremented and it is freed if no more references. + putpage is not called, because the physical segment allocates + and deallocates pages in a rather specific way. + page.c:508,514 + The usual thing. putpage is called for all Pages found in the + segment. + + putseg + freepte + putpage() Releases a page. + + page.c:208,238 + if the page was swapped out, p is not a pointer to the page, + but a swap address. + putswap()Releases a swapped out page. putswap (swap.c:58,73) + marks the swap page as no longer allocated in the swap + ``bytemap''. + Otherwise, p can point to a real Page, and reference counting + is used. If the image for the page is not swapimage, the page + is linked to the tail of the free list, otherwise it is linked + to the head. Since pages are allocated from the head, the + author is trying to keep the page in the list for a longer time + if the page still caches part of a image. Since the swap image + is just used as backing storage, their pages are to be reused + soon. It should be clear now, but note how the page free list + is actually used as a cache. + Finally, the author wakes up a process that was sleeping + waiting for pages--if any. The reason to wake up a sleeping + process is that there is now a free page available for + allocation (Remaining requesters could still be blocked in + palloc.pwait, only one of them did pass and sleep in palloc.r). + The reason not to issue the wakeup when no process is sleeping + is that in that case, nobody is waiting for memory. + Nevertheless, wakeup would no nothing in that case and calling + it wouldn't hurt, or did the author measure that it would hurt + performance? Or am I missing something here? + + Segments are usually released by putseg, however, there is another + routine (the one called by killbig, which you already saw) that is + called to release memory held by a segment. It is also used by a + couple other routines besides killbig. + + mfreeseg() Releases the memory used by a segment. + + segment.c:503,536 + First, mfreeseg scans all entries in the segment that are for + the pages starting at start. Unused map entries are ignored + (they have no memory to free); all entries in the segment Ptes + are set to nil and pages kept linked at list. + segment.c:537,551 + Second, mfreeseg calls putpage for all pages linked (all + segment pages). Before doing so, procflushseg is called if the + segment is shared. The reason is that when the segment is + shared, page reference counts are one, but pages should not go + away before letting all processes using the segment know that + the segment memory is being released. Other processes using the + segment could be even running right now at a different + processor. + + mfreeseg + procflushseg() Flushes MMU for processes using the given segment. + + proc.c:973,998 + procflushseg iterates through all processes, searching for seg + entries with the s segment. newtlb is set for all processes + with such a segment, so that its entries are invalidated before + it runs again. Besides, (:991) if the process is running at any + processor, flushmmu is set for that processor and nwait + incremented. + proc.c:1007,1001 + If nwait was not zero, the current process waits until flushmmu + is reset to zero for all processors. The current processor can + only `kindly request' to other processors that their MMU + entries be flushed, they are running and they will flush such + entries when they notice the flushmmu flag. Hopefully, that + will happen soon. + + Segment system calls + + You now know how typical text, data, bss and stack segments are + created, copied during rfork and how are page faults serviced. That is + most of what you should know about memory management. However, there + are several system calls related to memory management that remain yet + to be seen. + +Attaching segments + + syssegattach() Entry point for the segattach system call. Attaches a + new segment. + + sysproc.c:614,618 + syssegattach is used to create a new memory segment. System + call arguments are passed verbatim to segattach. + + syssegattach + segattach() Attaches a new segment. + + segment.c:600,601 + It is segattach who does the job. + segment.c:607,608 + va is given as zero when segattach should choose the address + where to map the segment in the process address space. If it is + not zero, the author checks that the address is not a kernel + address. I don't know the exact reason for the ``BUG'' comment, + but it seems to me that the author plans to be able to attach + segments within the address space of the kernel. + segment.c:610,611 + Checking that name is a null terminated string at existing + virtual memory. + segment.c:613,615 + sno is the slot for the new segment. The first empty slot is + used. The check for ESEG at line :614 is ensuring that the + ``extra segment'' used during exec is kept available. + Otherwise, the process could not use sysexec. + segment.c:617,618 + No more segments for this process. + segment.c:620,622 + len is now an integral number of pages. + segment.c:624,642 + The comment is very descriptive. It is very likely that a big + hole exist in virtual memory right below the user stack. Most + checking is done by isoverlap. By the way, wouldn't it be + better to ensure that virtual addresses used by ESEG are kept + unused? It doesn't matter too much because that portion of the + user address space is used while the maps are set for the + temporary stack mapping. + isoverlap()Would the segment overlap with an existing one? + segment.c:554,571 + isoverlap must iterate through the whole segment array for the + process, checking that neither va nor newtop are contained + within any segment. During all this time, seg is not locked. + However, only the current process could deallocate its seg + entries or attach new entries, and the current process is + currently doing the segattach, so the lock is not really + needed. Nevertheless, the lock could prevent future BUGs, in + case the author changes his mind and uses segattach for a + non-current (not up) process. + segment.c:644,646 + A hole found, isoverlap is called once more after rounding va + to a page boundary. Perhaps va should be truncated before line + :634, and these lines avoided. Besides, it is not clear for me + why Esoverlap (and not Enovmem) is raised, although the author + knows why, I think that Enovmem could be perfectly raised here + too. + segment.c:648,650 + name is the segment ``class'' specified by the user. On each + machine, an array of physical segments is kept at physseg. If + the segment class name matches the name of one of these + physical segments, the routine continues at :653. Otherwise + Ebadarg is raised. You see how a segattach can only be done for + segments declared in physseg. By the way, since the found label + is used only to avoid returning Ebadarg, perhaps an if could + have been better. The code is clear anyway, isn't it? + ../pc/segment.h:1,8 + For Intel PCs, the only known segments are ``shared'' and + ``memory'' (there are more ones, as you will see). But for + other architectures, physseg may contain exotic physical + segments (read the manual page). I think that the name is phys + because usually, processes attach to segments representing + physical memory like device-provided memory, memory locks, etc. + ../port/segment.c:654,664 + Unless the segment length be bigger than SEGMAXSIZE, the + segment is attached. newseg is creating the new segment, it + will be either a SHARED or a BSS segment, and page faults with + zero fill it on demand. + + Physical segments + + addphysseg() Adds a new segment class. + + segment.c:573,598 + I told you that ../pc/segment.h had just a couple of ``shared'' + and ``memory'' segments (classes) defined. There is a routine + addphysseg which is called by device drivers to declare the + existence of physical memory segments important for them. For + example, in ../pc/vgas3.c:103, the S3 video card driver calls + addphysseg to add an entry to physseg, with attribute + SG_PHYSICAL (physical memory used for the card) and name + s3screen (the video memory). This call is done by s3linear, + which is the S3 VGAdev procedure used to enable a linear mode + in the card (see ../pc/devvga.c and ../pc/screen.c). After this + segment is added by calling addphysseg, a segattach can be done + for s3screen to get to the video memory. + The routine addphysseg only checks that there is free room + after the initialized entries (those with a name) and before + the last entry (which must be a null entry) to add the new + segment. Should there be space, the new segment given is linked + into the array. + +Detaching segments + + syssegdetach() segdetach(2) system call. Detaches a segment. + + sysproc.c:621,631 + syssegdetach is the counterpart of syssegattach. It detaches a + segment from the address space. The seglock must be acquired + now. Routines that do not want seg entries to disappear under + their feet acquire this lock too. + sysproc.c:633,644 + The first argument is an address contained in the segment to be + detached. The seg array is searched until the entry number (i) + is found and s is set to the segment being detached. + sysproc.c:646,651 + arg is a pointer to the user arguments for the system call, + therefore it resides within the user stack. if this address is + contained in the segment to be detached, it is the stack + segment the one detached. In this case, the system refuses to + detach the segment. A process always needs a stack. Although + the check is nice, perhaps a check against SG_STACK would be + more clear. + The system does not seem to refuse detaches for the text + segment, although the manual page suggests so. + sysproc.c:652,660 + The segment is released by the call to putseg. + By the way, perhaps a segdettach routine could contain most of + syssegdetach code, as it happens with segattach. + + syssegfree() segfree(2) system call. Releases (part of) a segment. + + sysproc.c:664,686 + syssegfree releases (note the call to mfreeseg) part of the + memory held by the segment. This routine calls seg instead of + locating the segment itself to find the segment containing from + and lock it. Perhaps the routine could be generalized to return + the index for the segment in the seg array so that others (e.g. + syssegdetach) could benefit from it too. What happens to the + portion of the address space released depends on the kind of + segment. For instance, should it be a BSS segment, pages would + be later zero-filled on demand. + +Resizing segments + + syssegbrk() segbrk(2) system call. Resizes a segment. + + sysproc.c:588,608 + syssegbrk resizes the segment. Only SHARED, BSS, and SHDATA + segments can be resized. The reason is that text and data + segments correspond (and are paged from) an image of a text + file. Only for ``anonymous memory'' does brk make sense. ibrk + does all the work. + + syssegbrk + ibrk() Resizes a segment. + + segment.c:431,454 + addr is the new end address for the segment. If should be at + least the segment base. The comment says that BSS might be + overlapping the data segment, in which case :449 is checking + that addr is smaller than base for a BSEG and addr is bigger + than base for the DSEG (otherwise an error is raised). However, + segbrk finds the first segment where addr is contained, and + :448 would never be true. + Nevertheless, in a previous implementation of segbrk (note + :sysproc.c:688,693), the segment resized was always the BSS + segment, in which case :448 could be true for old Plan 9 + binaries and lines :448,454 would leave addr being the start of + the BSS segment. Perhaps it would have been better to let + sysbrk_ do the check, and either keep ibrk assuming that addr + always resides within segment bounds, or keep ibrk checking + that addr is within the segment. The reason for doing so is + that should sysbrk_ disappear, the weird BSS check in ibrk may + be forgotten and kept there. + segment.c:456,463 + The new top and size for the segment is computed. If the + segment is to shrink, mfreeseg releases the (now unused) memory + of the segment. Since segment base and top are not updated, any + page fault on the released part of the segment would make the + segment grow again. Perhaps lines :494,495 should be copied + before line :462. Although the current behavior is perfectly + reasonable (and compatible with what is said in segbrk(2).). + segment.c:465,468 + The segment is growing. Since the only segments resized + (sysproc.c:600,607) are SHARED, BSS, and SHDATA segments, which + have the swap file as their backing store, no resize is allowed + if there is no free swap space. + segment.c:470,478 + The whole list of segments is scanned searching for a segment + containing newtop. If newtop is not contained within other + segment, the space from the current top up to newtop is + available. However, there could be a case when there is a + segment starting after top, but ending before newtop, in which + the check would miss that there is another segment overlapping + the portion of the virtual address space being allocated. + Perhaps the check could be changed to see if base for any + segment is between top and newtop. + segment.c:480,492 + mapsize recomputed and map is reallocated to have enough space + for the new Ptes--map could be using the small ssegmap array. + segment.c:494,497 + Segment bounds updated. Segments grow, but they really never + shrink--only memory is reused if they are pretending to shrink. + +Flushing segments + + syssegflush() segflush(2) system call. Flushes a segment cache. + + segment.c:681,729 + syssegflush does a flush of the processor cache for memory + within the segment. This seems to be used only by instruction + simulator commands. The routine flushes a range of the user + address space, which may spawn several segments. For each + segment, pteflush is called to flush Ptes affected. + + syssegflush + pteflush() Flushes the cache for a Pte. + + segment.c:667,678 + pteflush sets the cachectl entry for the page (if not paged + out) to PG_TXTFLUSH. As you may remember, this means that for + several architectures (not the Intel), the processor cache + entries for the given pages may be flushed later by the + flushmmu call at :727. + +Segment profiling + + segclock() Update segment profiling counters. + + segment.c:731,745 + Not really a system call, but each time the clock ticks, if the + processor was running user code, ../pc/clock.c:76,79 would call + segclock giving the current saved program counter for the user + process. The call is made after adding to the word in the + bottom of the user stack the number of ticks in a millisecond. + The kernel is maintaining a ``clock'' for the running process + in the bottom of its user stack. If profile is set in the TSEG + for the process, segclock adds the time passed to the value in + s->profile[0]. Moreover, if the program counter for the user + was within this segment, it is converted to a relative offset + within TSEG and another time counter incremented. The + >>LRESPROF is used to have one time counter per LRESPROF + instructions in the text segment. + If you think of it, the author is filling up profile with an + array of clocks for the process. The first clock counts the + time the process has been executing (one ms added every time a + ms happens and the process was running), following clocks have + the time spent by the process within the first, second,... + group of LRESPROF instructions. The user can later inspect + these clocks and learn where is his program spending time. This + is crucial to let the user know what should be optimized, and + what not. + + Intel MMU handling + + During the chapter about system startup you already read some code for + MMU handling in the Intel PC. In this section you will be reading the + code which has been used by the machine independent code for memory + management, and was not discussed previously. + +Flushing entries + + flushmmu() Flushes MMU translations. + + ../pc/mmu.c:80,89 + flushmmu was called whenever the page table was changed, and + translations should be updated. Remember that the TLB may keep + cached entries. All it does is to set the newtlb flag and + switch to the current page table. On the Intel, a load of the + page directory base register flushes the TLB too. + + flushmmu + mmuswitch() Switches the address space in the MMU. + + mmu.c:110,127 + mmuswitch (in this case), notices the newtlb and calls + mmuptefree to flush the current set of entries. Later, the + switch is done. As you should know by now, if the process has a + mmupdb installed, that page table is used, otherwise, the + prototype page table for the processor is used. The virtual + address of the Page used to keep the first level page table is + used to access its entries, and its physical address is used to + set the pdbr (aka cr3). + One important thing here is that the routine maps the Mach + structure (together with the small scheduler stack) for the + current processor at MACHADDR. PDX is getting the index in the + first level page table for the virtual address MACHADDR. Each + processor has at MACHADDR a map of its Mach, as you saw in the + starting up chapter. + + flushmmu + mmuswitch + mmuptefree() Releases translations. + + mmu.c:91,108 + mmuptefree starts with the Page noted in mmuused, and iterates + through the whole set of pages linked through the next field. + For each such page, its entry in the pdb is set to nil (i.e. + invalid). pdb is set to be the virtual address of the mmupdb, + to update later its contents. The Page keeps in daddr the index + in the PDB for it. mmuused is a list of Pages used to keep + secondary page tables for the process. So, mmuptefree is just + clearing (invalidating) entries in the first level page table. + Al those pages are linked later into mmufree. All second level + page tables linked through mmuused are free now. The process + will have to suffer page faults to get new second level tables + again. This time, the machine independent MMU code will + instruct mmu.c to fill up the entries according to changes in + the Segment entries. + Can you see how the portable virtual memory data structures + dictate what the machine dependent code should do? Can you see + how they can differ from what the MMU has installed? + dat.h:90,95 + mmupdb, mmufree, and mmuused are just Pages. The fields va and + pa of such pages are used to update entries using the virtual + address space and to get the page frame address for giving it + to either the MMU or to an entry in PD. (By the why, see + ../port/portdat.h:628 if you don't understand why I told you to + look into PMMU and not into Proc). + +Adding entries + + putmmu() Updates a translation. + + mmu.c:195,196 + putmmu is called by the machine independent code to add an + entry to the MMU page table for the current process. This + happens when fixfault is repairing a page fault. How does this + work? + You know that each processor has a prototype page table, which + includes mappings for kernel space as well as an identity + mapping for physical memory. The kernel uses this prototype + page table if the process has not its own one. So, when a new + process is created, it starts using this page table until it + suffers a page fault and fixfault calls putmmu. + mmu.c:203,204 + If the process did not have its own page table, one is created. + mmupdballoc does the job. Following calls to putmmu due to page + faults will find (and use) the existing PDB allocated here. + mmupdballoc()Allocates a PDB . + mmu.c:173,193 + mmupdballoc takes a new Page, either by allocating it (with + newpage) or by using one from a free list at pdbpool. va is set + in the Page as the kernel virtual address for the page frame + and the page is kept mapped for kernel usage (no-op on Intels). + It is initialized by copying the prototype page table from the + processor, which you know was initialized. Initialized PDBs are + placed back to the pdbpool when they are no longer used by the + process, so that their allocation could be faster. Since + newpage is given a nil s pointer, it will keep on trying to + allocate a page instead of returning failure (as it does when + called from fixfault). + mmu.c:205,206 + Back to putmmu, pdb is now a kernel pointer to the PDB and pdbx + is the entry in the first level page table for the virtual + address to be mapped to pa using the Page supplied (note it's + not used on Intels). + mmu.c:208,217 + If there is no second-level page table (entry invalid in the + PDB), must allocate one. They are allocated on demand. Again, + Pages for second level page tables are reused. If there is one + in the free list (mmufree) that is used, otherwise newpage is + called to allocate a new one. clear is set for newpage, so it + is zeroed, setting all entries as invalid. If the page is new, + a kernel map is made, otherwise, the kernel map was already + done at page allocation time. + mmu.c:218 + The entry in the PDB is set as valid, usable for protection + ring 3, and allowing writes. The Intel checks permissions on + the entry in the PDB, and then it checks permissions on the + entry in the second label page table, every time an address is + used (of course, TLB caches such things). + mmu.c:219,221 + mmuused is updated to contain the new page, so that mmuptefree + could know which entries are used. As only a few entries of the + 1024 existing entries are really used, that saves a lot of + time. daddr (not used here as a disk address) is used to keep + the index in the PDB. PDB used entries can be found pretty + quickly. + mmu.c:224,225 + At this point, the second level page table exists. pte is set + pointing to the second level page table. page->va should be + equal to KADDR(PPN(pdb[pdbx])). However, the author prefers to + obtain the pointer through the entry in the PDB, probably to be + sure that things are working properly. The entry in the second + level page table for va is set with pa together with the USER + bit, which says that ring 3 can access the page. When fixfault + calls putmmu, pa is not just the page frame address for va, + fixfault sets some of its offset bits to specify which + permissions should be enabled (e.g. PTEWRITE, etc.), so the + author ORs pa to keep those bits set. + + NOTE: perhaps page's va cannot be used because the second level + page table could be shared? Check that. + mmu.c:227,231 + Again ensuring that the Mach page is mapped, although it should + be mapped already if the PDB did exist, and it will be mapped + by mmuswitch in any case. Just for safety. + +Adding and looking up entries + + mmuwalk() Walks to an MMU entry (and perhaps creates it). + + mmu.c:233,234 + You saw how mmuwalk was called to get a pointer to the entry in + the MMU page table for a given va. If level is 1, the entry + wanted is the one in the first level page table, should it + be 2, the entry wanted is the one in the second level page + table. Other architectures may accept more than two levels + (note the clean generic interface for the routine), but intels + have just two. create can be set to request that the second + level table be created. mmu.c itself uses mmuwalk, and meminit + uses it too to create a prototype table. Unlike putmmu, mmuwalk + does not require up to have a valid current process. + mmu.c:245,247 + table is set to point to the entry in PDB for the given va. If + the entry is nil, and it shouldn't be created, there is nothing + to do. + mmu.c:249,268 + Just two levels on Intels. For level 1, table is the pointer to + the entry. For level 2, if the entry in *table was not valid, a + second level page table is allocated and the entry in the PDB + updated. This only happens when create was set. Finally, table + is set to point to the virtual address for the second level + page table, and a pointer to its entry for va returned. + + mmukmap() Adds a kernel map to the MMU. + + mmu.c:307,308 + mmukmap adds a translation to the MMU page table for a page + within the kernel portion of the address space. You saw how + mmukmap was called during boot time, while initializing memory. + mmu.c:314,318 + mach0 is set to point to the Mach structure for the boot + processor. If it has bit 0x10 set in its cr4 and the processor + is not so old that does not support it, super-pages can be used + (Remember, using entries in the first-level page table to map + 4MB). pse is set if super-pages can be used. + mmu.c:321,326 + If va is not given, mmukmap understands that it is doing the + identity map at KZERO. Otherwise, it is truncated to be a page + address. + mmu.c:328,399 + This loop maps the range of physical memory going from pa to + pa+size-1. The routine can be used both to map a single page + and to map a whole range of (4K) pages. + mmu.c:331 + table is pointing to the first level entry in the prototype PDB + at processor zero. + mmu.c:335,379 + The entry is valid. If the entry had the PTESIZE bit, it was a + map for 4M without using a second level page table (:337,357). + In this case, x is the start of a 4M super- page frame, which + should be the same of pa--at each pass, pa is the physical + address being mapped. pa is set either to the end of the mapped + memory or to the end of the super-page, preparing things up for + the next iteration or for terminating the loop. The loop + continues after the portion mapped by the super-page. If pa is + already pae, all remaining pages to be mapped were already + mapped by the super-page and the loop finishes. + If the entry (:336) did not have the PTESIZE bit, it is + pointing to a second level page table (:360,376). In this case, + mmuwalk is used to get the entry in the second level page table + (note again: processor zero PDB). If the entry is valid, pa is + advanced past the already mapped page. + sync is not-zero when a second level page table entry has been + scanned. + mmu.c:382,298 + The entry was not mapped. If pse says so, a whole super-page is + used to map the desired address. This assumes that all the 4MB + of physical memory (4MB aligned) where pa stands are to be + mapped; hence the checks at line :387. It is very important to + use super-pages since the Intel has different TLB entries for + super-pages and regular pages. By making the kernel use + super-pages, the user TLB entries are not disturbed while + servicing interrupts and system calls. If no super-pages can be + used (or the mapping does not contain a whole 4M super-page), + mmuwalk is used to create the second level entry. pgsz is set + so that :396,397 can prepare the next pass. sync is not-zero + when any entry has been added to the processor zero PDB. + mmu.c:402,407 + mmukmapsync is called if mmukmap was called and either the + mapping was at a second-level page table, or an entry was + updated--you also saw a call to mmukmapsync before. + + mmukmap + mmukmapsync() Synchronizes kernel maps in page tables. + + mmu.c:272,273 + mmukmapsync updates prototype page tables for all processors by + looking at the boot processor page table. It fixes any missing + entry. So, apart of initial memory mappings created during boot + time, any kernel map added later only has to be set at + processor 0. That is precisely what mmukmap does! + Other processors will fault when using memory mapped at a + different processor, but mmukmapsync will notice that there is + a map in mach0->pdb, and will copy such map to the faulting + processor page table. Note also how it flushes the tlb, and how + the page table for the current process is updated to repair any + page fault there. + Faults at these kernel mapped pages will happen only for memory + not mapped initially during boot time. + + Epilogue + + And now what? + + Although we have gone a long way already, you still have devices in + the Plan 9 kernel that remain to be read. You should try to understand + them. + + Besides, Plan 9 is much more than its kernel. For example, the code + for the local file system provider (kfs) is really interesting to + learn how to structure a partition into a set of files. + + It is also illustrative to read the code for user commands, including + the plumber, rio, sam, and acme. You can learn a lot by doing so; not + just a lot about operating systems, but also a lot about good + programming practices. + + Finally, you could be so kind to let me know whatever suggestion you + may have about this document. + + Have fun with Plan 9! + +Bibliography + + 1 + Francisco J. Ballesteros. + Advanced Operating Systems Course Web site. + http://gsyc.escet.urjc.es/docencia/asignaturas/ampliacion_ssoo, + 2000. + 2 + Francisco J. Ballesteros. + Operating Systems Design Course Web site. + http://gsyc.escet.urjc.es/docencia/asignaturas/osd, 2000. + 3 + Intel. + Pentium iii processors - manuals. + http://developer.intel.com/design/pentiumiii/manuals, 2000. + 4 + Brian W Kernighan and Rob Pike. + The Practice of Programming. + Addison-Wesley Publishing Company, 1999. + 5 + Kerninghan and Ritchie. + The C Programming Language. + Prentice-Hall, 1988. + 6 + Rob Pike. + A manual for the Plan 9 assembler. + Plan 9 Programmer's manual, 3rd ed., vol. 2, 2000. + 7 + Rob Pike. + Acme: A User Interface for Programmers. + Plan 9 Programmer's manual, 3rd ed., vol. 2, 2000. + 8 + Rob Pike. + How to Use the Plan 9 C Compiler. + Plan 9 Programmer's manual, 3rd ed., vol. 2, 2000. + 9 + Rob Pike et al. + Plan 9 Programmer's Manual: Volume 1: The Manuals. 3rd ed. + Harcourt Brace and Co., New York, NY, USA, 2000. + 10 + Rob Pike et al. + Plan 9 Programmer's Manual: Volume 2: The Documents. 3rd ed. + Harcourt Brace and Co., New York, NY, USA, 2000. + 11 + Rob Pike, Dave Presotto, Sean Dorward, Bob Flandrena, Ken + Thompson, Howard Trickey, and Phil Winterbottom. + Plan 9 from Bell Labs. + Computing Systems, 8(3):221-254, Summer 1995. + 12 + Rob Pike, Dave Presotto, Ken Thompson, Howard Trickey, and Phil + Winterbottom. + The use of name spaces in Plan 9. + Operating Systems Review, 27(2):72-76, April 1993. + 13 + D. L. Presotto. + Multiprocessor streams for Plan 9. + In UKUUG. UNIX - The Legend Evolves. Proceedings of the Summer + 1990 UKUUG Conference, pages 11-19 (of xi + 260), Buntingford, + Herts, UK, ???? 1990. UK Unix Users Group. + 14 + et al. Rifkin, A.P. + Rfs architectural overview. + In Summer Usenix Conference. USENIX, 1986. + 15 + Dennis M. Ritchie. + A Stream Input-Output System. + AT&T Bell Laboratories Technical Journal, 63(8):1897-1910, + October 1984. + Also available from + http://cm.bell-labs.com/cm/cs/who/dmr/st.html. + 16 + Sun Microsystems, Inc. + NFS: Network file system protocol specification. + RFC 1094, Network Information Center, SRI International, March + 1989. + _________________________________________________________________ + + Footnotes + + ... editor[250]5.1 + As you will see, acme is much more than a editor; it is a full + environment to do your daily work on Plan 9. + + ... blank[251]5.2 + Read ``Plan 9 From Bell Labs'' [[252]11] instead. The paper is + so clear that nothing else has to be said. + + ... ones[253]5.3 + Of course, if you plan to do so, you should reuse the existing + source. The machine dependent part of the compiler is the only + thing that needs to be done + + ... set[254]5.4 + I used the typical convention found in Linux and popular + assemblers for the PC. + + ... UNIX[255]8.1 + In UNIX, priority 0 is better than priority 10. + + ... server[256]9.1 + The thing is even worse because exportfs is used to export part + of the local name space to other processes, and it is likely + that exportfs would lead to more 9P requests to service its + file tree. + + ... same[257]10.1 + It does not work exactly this way, but you will know. + + ... else[258]10.2 + This also happens for other pages, as you will see + + ... function[259]10.3 + The console device may start also the pager. + _________________________________________________________________ + + + Francisco J Ballesteros + 2001-01-08 + +References + + 1. file://localhost/usr/nemo/doc/os/9/9/9.html + 2. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00200000000000000000 + 3. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00300000000000000000 + 4. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00400000000000000000 + 5. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00500000000000000000 + 6. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00600000000000000000 + 7. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00610000000000000000 + 8. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00611000000000000000 + 9. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00620000000000000000 + 10. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00621000000000000000 + 11. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00622000000000000000 + 12. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00630000000000000000 + 13. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00640000000000000000 + 14. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00641000000000000000 + 15. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00642000000000000000 + 16. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00650000000000000000 + 17. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00651000000000000000 + 18. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00652000000000000000 + 19. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00653000000000000000 + 20. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00654000000000000000 + 21. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00700000000000000000 + 22. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00710000000000000000 + 23. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00711000000000000000 + 24. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00712000000000000000 + 25. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00720000000000000000 + 26. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00800000000000000000 + 27. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00810000000000000000 + 28. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00820000000000000000 + 29. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00821000000000000000 + 30. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00822000000000000000 + 31. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00830000000000000000 + 32. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00840000000000000000 + 33. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00850000000000000000 + 34. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00851000000000000000 + 35. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00852000000000000000 + 36. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00860000000000000000 + 37. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00870000000000000000 + 38. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00871000000000000000 + 39. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00872000000000000000 + 40. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00873000000000000000 + 41. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00880000000000000000 + 42. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00890000000000000000 + 43. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION008100000000000000000 + 44. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION008110000000000000000 + 45. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION008111000000000000000 + 46. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION008112000000000000000 + 47. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION008113000000000000000 + 48. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION008120000000000000000 + 49. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION008130000000000000000 + 50. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION008131000000000000000 + 51. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION008132000000000000000 + 52. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION008133000000000000000 + 53. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00900000000000000000 + 54. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00910000000000000000 + 55. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00920000000000000000 + 56. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00930000000000000000 + 57. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00931000000000000000 + 58. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00932000000000000000 + 59. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00940000000000000000 + 60. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00941000000000000000 + 61. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00942000000000000000 + 62. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00943000000000000000 + 63. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00950000000000000000 + 64. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00951000000000000000 + 65. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00952000000000000000 + 66. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00953000000000000000 + 67. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00954000000000000000 + 68. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00960000000000000000 + 69. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00961000000000000000 + 70. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00962000000000000000 + 71. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00963000000000000000 + 72. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00964000000000000000 + 73. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00970000000000000000 + 74. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00971000000000000000 + 75. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00972000000000000000 + 76. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00980000000000000000 + 77. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00981000000000000000 + 78. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00982000000000000000 + 79. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00983000000000000000 + 80. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION00990000000000000000 + 81. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION009100000000000000000 + 82. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION009101000000000000000 + 83. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION009102000000000000000 + 84. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION009110000000000000000 + 85. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION009111000000000000000 + 86. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION009112000000000000000 + 87. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION009120000000000000000 + 88. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION009121000000000000000 + 89. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION009122000000000000000 + 90. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION009123000000000000000 + 91. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION009124000000000000000 + 92. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001000000000000000000 + 93. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001010000000000000000 + 94. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001020000000000000000 + 95. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001021000000000000000 + 96. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001022000000000000000 + 97. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001030000000000000000 + 98. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001031000000000000000 + 99. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001032000000000000000 + 100. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001033000000000000000 + 101. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001034000000000000000 + 102. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001040000000000000000 + 103. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001041000000000000000 + 104. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001042000000000000000 + 105. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001050000000000000000 + 106. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001051000000000000000 + 107. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001052000000000000000 + 108. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001060000000000000000 + 109. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001061000000000000000 + 110. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001062000000000000000 + 111. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001070000000000000000 + 112. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001071000000000000000 + 113. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001072000000000000000 + 114. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001073000000000000000 + 115. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001074000000000000000 + 116. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001075000000000000000 + 117. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001076000000000000000 + 118. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001077000000000000000 + 119. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001078000000000000000 + 120. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001079000000000000000 + 121. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001080000000000000000 + 122. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001081000000000000000 + 123. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001082000000000000000 + 124. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001100000000000000000 + 125. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001110000000000000000 + 126. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001111000000000000000 + 127. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001112000000000000000 + 128. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001120000000000000000 + 129. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001121000000000000000 + 130. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001122000000000000000 + 131. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001123000000000000000 + 132. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001124000000000000000 + 133. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001130000000000000000 + 134. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001131000000000000000 + 135. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001132000000000000000 + 136. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001133000000000000000 + 137. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001134000000000000000 + 138. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001135000000000000000 + 139. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001140000000000000000 + 140. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001150000000000000000 + 141. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001160000000000000000 + 142. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001161000000000000000 + 143. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001162000000000000000 + 144. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001163000000000000000 + 145. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001164000000000000000 + 146. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001165000000000000000 + 147. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001170000000000000000 + 148. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001171000000000000000 + 149. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001172000000000000000 + 150. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001173000000000000000 + 151. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001200000000000000000 + 152. file://localhost/usr/nemo/doc/os/9/9/9.html#SECTION001300000000000000000 + 153. file://localhost/usr/nemo/doc/os/9/9/9.html#osdw3 + 154. file://localhost/usr/nemo/doc/os/9/9/9.html#ch:src + 155. file://localhost/usr/nemo/doc/os/9/9/9.html#ch:start + 156. file://localhost/usr/nemo/doc/os/9/9/9.html#ch:start + 157. file://localhost/usr/nemo/doc/os/9/9/9.html#foot590 + 158. file://localhost/usr/nemo/doc/os/9/9/9.html#acme + 159. file://localhost/usr/nemo/doc/os/9/9/9.html#aosw3 + 160. file://localhost/usr/nemo/doc/os/9/9/9.html#vol2 + 161. file://localhost/usr/nemo/doc/os/9/9/9.html#vol1 + 162. file://localhost/usr/nemo/doc/os/9/9/9.html#vol2 + 163. file://localhost/usr/nemo/doc/os/9/9/9.html#vol1 + 164. file://localhost/usr/nemo/doc/os/9/9/9.html#vol2 + 165. file://localhost/usr/nemo/doc/os/9/9/9.html#foot593 + 166. file://localhost/usr/nemo/doc/os/9/9/9.html#C + 167. file://localhost/usr/nemo/doc/os/9/9/9.html#plan9c + 168. file://localhost/usr/nemo/doc/os/9/9/9.html#tpop + 169. file://localhost/usr/nemo/doc/os/9/9/9.html#C + 170. file://localhost/usr/nemo/doc/os/9/9/9.html#foot451 + 171. file://localhost/usr/nemo/doc/os/9/9/9.html#intelw3 + 172. file://localhost/usr/nemo/doc/os/9/9/9.html#pnasm + 173. file://localhost/usr/nemo/doc/os/9/9/9.html#foot523 + 174. file://localhost/usr/nemo/doc/os/9/9/9.html#intelw3 + 175. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:vmtbl + 176. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:idt + 177. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:lmem + 178. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:bmem + 179. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:mach + 180. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:bmem + 181. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:idt + 182. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:idt + 183. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:intr + 184. file://localhost/usr/nemo/doc/os/9/9/9.html#ritchie84:_streams + 185. file://localhost/usr/nemo/doc/os/9/9/9.html#ch:fs + 186. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:files + 187. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:procs + 188. file://localhost/usr/nemo/doc/os/9/9/9.html#ch:proc + 189. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:vmem + 190. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:traps + 191. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:err + 192. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:err0 + 193. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:err1 + 194. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:err0 + 195. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:err + 196. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:err0 + 197. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:err1 + 198. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:err2 + 199. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:err3 + 200. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:sw + 201. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:sw + 202. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:sw1 + 203. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:sw1 + 204. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:sw2 + 205. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:sw3 + 206. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:sw4 + 207. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:sw5 + 208. file://localhost/usr/nemo/doc/os/9/9/9.html#foot6975 + 209. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:tasup + 210. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:tasmp + 211. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:tasmpsw + 212. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:ilock + 213. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:rwlocks + 214. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:qlock + 215. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:note0 + 216. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:note1 + 217. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:noted0 + 218. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:noted1 + 219. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:noted2 + 220. file://localhost/usr/nemo/doc/os/9/9/9.html#ch:mm + 221. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:forkvm + 222. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:fork + 223. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:sched + 224. file://localhost/usr/nemo/doc/os/9/9/9.html#sc:introfs + 225. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:ns + 226. file://localhost/usr/nemo/doc/os/9/9/9.html#Pike:1993:UNS + 227. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:orace + 228. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:mount + 229. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:mount + 230. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:pipe + 231. file://localhost/usr/nemo/doc/os/9/9/9.html#ch:start + 232. file://localhost/usr/nemo/doc/os/9/9/9.html#nfs + 233. file://localhost/usr/nemo/doc/os/9/9/9.html#ux:rfs + 234. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:mnt + 235. file://localhost/usr/nemo/doc/os/9/9/9.html#foot15176 + 236. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:mntrpc + 237. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:mntrpc + 238. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:cache + 239. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:mcache + 240. file://localhost/usr/nemo/doc/os/9/9/9.html#Presotto:1990:MSP + 241. file://localhost/usr/nemo/doc/os/9/9/9.html#ch:start + 242. file://localhost/usr/nemo/doc/os/9/9/9.html#foot19196 + 243. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:umem + 244. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:seg + 245. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:segimg + 246. file://localhost/usr/nemo/doc/os/9/9/9.html#foot19610 + 247. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:palloc + 248. file://localhost/usr/nemo/doc/os/9/9/9.html#foot19789 + 249. file://localhost/usr/nemo/doc/os/9/9/9.html#fig:cow + 250. file://localhost/usr/nemo/doc/os/9/9/9.html#tex2html1 + 251. file://localhost/usr/nemo/doc/os/9/9/9.html#tex2html2 + 252. file://localhost/usr/nemo/doc/os/9/9/9.html#Pike:1995:PBL + 253. file://localhost/usr/nemo/doc/os/9/9/9.html#tex2html3 + 254. file://localhost/usr/nemo/doc/os/9/9/9.html#tex2html4 + 255. file://localhost/usr/nemo/doc/os/9/9/9.html#tex2html17 + 256. file://localhost/usr/nemo/doc/os/9/9/9.html#tex2html32 + 257. file://localhost/usr/nemo/doc/os/9/9/9.html#tex2html36 + 258. file://localhost/usr/nemo/doc/os/9/9/9.html#tex2html40 + 259. file://localhost/usr/nemo/doc/os/9/9/9.html#tex2html42 diff --git a/Operating Systems/Unix/BSD & Unix/O'Reilly - Learning the UNIX Operating System.pdf b/Operating Systems/Unix/BSD & Unix/O'Reilly - Learning the UNIX Operating System.pdf new file mode 100644 index 0000000000000000000000000000000000000000..f41bfcb89179ae64d4985f530c9d3b37146f1cce --- /dev/null +++ b/Operating Systems/Unix/BSD & Unix/O'Reilly - Learning the UNIX Operating System.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0a3af25641cc20ec393f333d3ee1257867aaf18e72afc21989cc2e55418ead16 +size 1145128 diff --git a/Operating Systems/Unix/BSD & Unix/O'Reilly - The Complete FreeBSD.pdf b/Operating Systems/Unix/BSD & Unix/O'Reilly - The Complete FreeBSD.pdf new file mode 100644 index 0000000000000000000000000000000000000000..ea9c823dab2a14e06084f28b0c3acc7c53a26058 --- /dev/null +++ b/Operating Systems/Unix/BSD & Unix/O'Reilly - The Complete FreeBSD.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:efba373724f6b070d6dbaca925a3cad30990532995a1b1a4239c7b125bb4f503 +size 10573178 diff --git a/Operating Systems/Unix/BSD & Unix/UNIX Filesystems - Evolution, Design, & Implementation.pdf b/Operating Systems/Unix/BSD & Unix/UNIX Filesystems - Evolution, Design, & Implementation.pdf new file mode 100644 index 0000000000000000000000000000000000000000..0bf18a169d0a49e730452fa362939500b1ef4c00 --- /dev/null +++ b/Operating Systems/Unix/BSD & Unix/UNIX Filesystems - Evolution, Design, & Implementation.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:01fe10b7028f150669b05cd2a7c833733d1e93aeaf99f5864dcf2bcde0670cc2 +size 4252465 diff --git a/Operating Systems/Unix/Classic Shell Scripting.chm b/Operating Systems/Unix/Classic Shell Scripting.chm new file mode 100644 index 0000000000000000000000000000000000000000..8419112e5942cb708b8239417596198b60fd0758 --- /dev/null +++ b/Operating Systems/Unix/Classic Shell Scripting.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:458be8400526bbbe1d57c47934ffe1440a5d6469406fc09959ac735ec842ef14 +size 1089345 diff --git a/Operating Systems/Unix/Learning the vi and Vim Editors, 7e.pdf b/Operating Systems/Unix/Learning the vi and Vim Editors, 7e.pdf new file mode 100644 index 0000000000000000000000000000000000000000..a509e4edb2fe790d0317e7030735b57604abcb2c --- /dev/null +++ b/Operating Systems/Unix/Learning the vi and Vim Editors, 7e.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c1300ee002a78aee41f94707c0676773f416c83bac69520435648bd23b430726 +size 7261405 diff --git a/Operating Systems/Unix/Linux Complete Command Reference.pdf b/Operating Systems/Unix/Linux Complete Command Reference.pdf new file mode 100644 index 0000000000000000000000000000000000000000..5bd240dc72a173de7a8f0d471068e386b00c8497 --- /dev/null +++ b/Operating Systems/Unix/Linux Complete Command Reference.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b53010f433edc21814730563eec1052519b851ef95829cbce713ddf2f9a33f16 +size 10631201 diff --git a/Operating Systems/Unix/Sams Publishing - Unix Shell Programming, 3rd Edition.chm b/Operating Systems/Unix/Sams Publishing - Unix Shell Programming, 3rd Edition.chm new file mode 100644 index 0000000000000000000000000000000000000000..3f67dc57932d528f36f57e913f0253af42efe985 --- /dev/null +++ b/Operating Systems/Unix/Sams Publishing - Unix Shell Programming, 3rd Edition.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e5c90d1e8ed6fce609b87c469a50e9dbf33c87062f6648c9b7d0118c6b68981a +size 1127049 diff --git a/Operating Systems/Unix/Unix For Dummies.chm b/Operating Systems/Unix/Unix For Dummies.chm new file mode 100644 index 0000000000000000000000000000000000000000..b990b8aac266da03a27feb3ef797b94904c956c7 --- /dev/null +++ b/Operating Systems/Unix/Unix For Dummies.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:25c61998e69e855a9a76691b7b567b1286f700cc04198ea331fc5986bc36ecc2 +size 3095776 diff --git a/Operating Systems/Unix/linux-Addison.Wesley.The.Art.Of.Unix.Programming.chm b/Operating Systems/Unix/linux-Addison.Wesley.The.Art.Of.Unix.Programming.chm new file mode 100644 index 0000000000000000000000000000000000000000..5373e643a2efb518fff9f041b9d890f56a57fd24 --- /dev/null +++ b/Operating Systems/Unix/linux-Addison.Wesley.The.Art.Of.Unix.Programming.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1638a5721bf0d29a30456983446cc3a26324423813089dc19c41f60a356b253c +size 1565802 diff --git a/Operating Systems/Windows Server/Active Directory/O'Reilly Active Directory (2nd Edition).chm b/Operating Systems/Windows Server/Active Directory/O'Reilly Active Directory (2nd Edition).chm new file mode 100644 index 0000000000000000000000000000000000000000..7c22e7649903d37c8b04f2c73291a64f91af228e --- /dev/null +++ b/Operating Systems/Windows Server/Active Directory/O'Reilly Active Directory (2nd Edition).chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:827b09b1251ea37fe468f6029de1980411ef97e0bd1d2abe06b60f86a66894a0 +size 3758526 diff --git a/Operating Systems/Windows Server/Active Directory/O'Reilly Active Directory Cookbook for Windows Server 2003 & Windows 2000.chm b/Operating Systems/Windows Server/Active Directory/O'Reilly Active Directory Cookbook for Windows Server 2003 & Windows 2000.chm new file mode 100644 index 0000000000000000000000000000000000000000..06e5e6dd2ac4f7679a125373dae501ee1fa3d801 --- /dev/null +++ b/Operating Systems/Windows Server/Active Directory/O'Reilly Active Directory Cookbook for Windows Server 2003 & Windows 2000.chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b413f67c0fe1e4960b4a525d08ecd9dd6b95c159b3f889b270fea9613766c007 +size 1052501 diff --git a/Operating Systems/Windows Server/Sybex.Mastering.System.Center.Configuration.Manager.2007.R2.Apr.2009.pdf b/Operating Systems/Windows Server/Sybex.Mastering.System.Center.Configuration.Manager.2007.R2.Apr.2009.pdf new file mode 100644 index 0000000000000000000000000000000000000000..5eae02df452dfe64fbe7e968839b1813276ff936 --- /dev/null +++ b/Operating Systems/Windows Server/Sybex.Mastering.System.Center.Configuration.Manager.2007.R2.Apr.2009.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e668e2a29f60a5aa3848349f663b23e9ceb6bd684fab3288d4d1aeb70effa47b +size 19219558 diff --git a/Operating Systems/Windows Server/Sybex.Windows.Server.2008.Hyper-V.Insiders.Guide.to.Microsofts.Hypervisor.Apr.2009.pdf b/Operating Systems/Windows Server/Sybex.Windows.Server.2008.Hyper-V.Insiders.Guide.to.Microsofts.Hypervisor.Apr.2009.pdf new file mode 100644 index 0000000000000000000000000000000000000000..c0aeb002cc7647bfff4a0df28f550a667176f046 --- /dev/null +++ b/Operating Systems/Windows Server/Sybex.Windows.Server.2008.Hyper-V.Insiders.Guide.to.Microsofts.Hypervisor.Apr.2009.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:84cf4732cfb54bfc6654d7e22e7bbc4df9b3f235a5c5a99bb45cc6ccdffb9dff +size 14931251 diff --git a/Operating Systems/Windows Server/Wiley.Windows.Server.2008.Bible.Jul.2008.pdf b/Operating Systems/Windows Server/Wiley.Windows.Server.2008.Bible.Jul.2008.pdf new file mode 100644 index 0000000000000000000000000000000000000000..0a115db94b4794893f16cc49a0fefcbde3158d17 --- /dev/null +++ b/Operating Systems/Windows Server/Wiley.Windows.Server.2008.Bible.Jul.2008.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ddf7298309b0a8d5a05f73cb35e2118f6c7ba7905f59da189727fba08c6b4839 +size 16967226 diff --git a/Operating Systems/Windows Server/Windows Server 2003 For Dummies.pdf b/Operating Systems/Windows Server/Windows Server 2003 For Dummies.pdf new file mode 100644 index 0000000000000000000000000000000000000000..fee77ff78a19a0c4b4767ca02565c8206c1988aa --- /dev/null +++ b/Operating Systems/Windows Server/Windows Server 2003 For Dummies.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e016d54c898a3eb086c6b8ed397402bc27a113a70dfd694244bb30661854aa34 +size 47646428 diff --git a/Operating Systems/Windows Server/Windows Server 2003 for Dummies (2003).chm b/Operating Systems/Windows Server/Windows Server 2003 for Dummies (2003).chm new file mode 100644 index 0000000000000000000000000000000000000000..42bc31564b62f6aa5328c2faea4d6a123610f524 --- /dev/null +++ b/Operating Systems/Windows Server/Windows Server 2003 for Dummies (2003).chm @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c7bbfcf21546da3cbccfdba050410aa5e1aabb5d8dd60fa704deb14c686cc09f +size 10473457 diff --git a/Operating Systems/Windows XP/Windows XP Digital Music For Dummies (Nov 2004).pdf b/Operating Systems/Windows XP/Windows XP Digital Music For Dummies (Nov 2004).pdf new file mode 100644 index 0000000000000000000000000000000000000000..e16f870d1e00246f24899a24af9096cfa76ee090 --- /dev/null +++ b/Operating Systems/Windows XP/Windows XP Digital Music For Dummies (Nov 2004).pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9e0e3f8dd57aa0082d0eda6b99a791ecc94568285324f7881fba67b5f4d87431 +size 7694554 diff --git a/Operating Systems/Windows XP/Windows XP Gigabook For Dummies (2004).pdf b/Operating Systems/Windows XP/Windows XP Gigabook For Dummies (2004).pdf new file mode 100644 index 0000000000000000000000000000000000000000..d2a5b0f4aa0d2ed1571684869229f020211e9265 --- /dev/null +++ b/Operating Systems/Windows XP/Windows XP Gigabook For Dummies (2004).pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:17cbaa7e1d41a80e42df91caa7e684e3294e9a0bcbbcd639548f57842336cd69 +size 32901333 diff --git a/Operating Systems/Windows XP/Windows XP Hacks & Mods For Dummies (Sep 2005).pdf b/Operating Systems/Windows XP/Windows XP Hacks & Mods For Dummies (Sep 2005).pdf new file mode 100644 index 0000000000000000000000000000000000000000..4a724d6e279e28dff74962a0eb6e3b22fea22f40 --- /dev/null +++ b/Operating Systems/Windows XP/Windows XP Hacks & Mods For Dummies (Sep 2005).pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7dd09abd83717b67fb823abf54afa0299520e4c677a0dbb3ead1b77dd4f8fe74 +size 14914902 diff --git a/Operating Systems/Windows XP/Windows XP Just the Steps for Dummies (2005).pdf b/Operating Systems/Windows XP/Windows XP Just the Steps for Dummies (2005).pdf new file mode 100644 index 0000000000000000000000000000000000000000..d80d9a4b5e2c1fe0ea3e89971d8f1a104f58fc89 --- /dev/null +++ b/Operating Systems/Windows XP/Windows XP Just the Steps for Dummies (2005).pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6e81222fbe6db5b6fb295decec52c205c4d7636e46e9346ac3802d2548d0be7d +size 37388466 diff --git a/Operating Systems/Windows XP/Windows XP Timesaving Techniques for Dummies, 2nd Edition (Jan 2005).pdf b/Operating Systems/Windows XP/Windows XP Timesaving Techniques for Dummies, 2nd Edition (Jan 2005).pdf new file mode 100644 index 0000000000000000000000000000000000000000..8511f1efdaadb63619899e0c203a0c8bac37d986 --- /dev/null +++ b/Operating Systems/Windows XP/Windows XP Timesaving Techniques for Dummies, 2nd Edition (Jan 2005).pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cc0e27cf5a9325d47117ad221c121b33098fa825f0225bab54fa1a84b856d147 +size 25469610 diff --git a/Operating Systems/Windows XP/Windows XP for Dummies, 2nd Edition.pdf b/Operating Systems/Windows XP/Windows XP for Dummies, 2nd Edition.pdf new file mode 100644 index 0000000000000000000000000000000000000000..a9ab00cccc48a304676cdb8bf25a622850a70273 --- /dev/null +++ b/Operating Systems/Windows XP/Windows XP for Dummies, 2nd Edition.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d390b8c9188a4e0362066a49528bb1c50127d0752b732c84a06358bdebc7d351 +size 12037813 diff --git a/Operating Systems/Windows XP/Windows Xp For Dummies, 1st Edition (2001).pdf b/Operating Systems/Windows XP/Windows Xp For Dummies, 1st Edition (2001).pdf new file mode 100644 index 0000000000000000000000000000000000000000..9b867ddb55919c7ad6d923430c430d17975a1b39 --- /dev/null +++ b/Operating Systems/Windows XP/Windows Xp For Dummies, 1st Edition (2001).pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d00b16ab6d2f7c1ebb2f082fc5e9abd0664f47e47d081aac267dc0c6420851a9 +size 6473963