July 2008 - Darwin-kernel - lists.apple.com

Re: Documentation bug: host vs target vs development
by Terry Lambert 31 Jul '08

31 Jul '08

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Let's take your example: The target machine is a x86. The development machine is a PPC. -- Terry _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… On Jul 31, 2008, at 10:20 AM, Stéphane Sudre wrote: On Jul 31, 2008, at 6:59 PM, Terry Lambert wrote: Symbol files are created via the kextload command and must specify the load address of the KEXT, which is only known after it is loaded into a kernel. Consider the case where you are developing on a PPC machine, targetting an x86 machine, because you intend to do two machine debugging of the x86 machine. Maybe you are doing this because these are the only two machines you have. The issue is not with the command line. It's that it would be better to keep on using the same name for the machines. Now, I introduce the "host machine". Which one is it? The x86 or PPC? Whichever one you run the kextload command on; in this case, it had better be the x86 machine, AFTER your KEXT is loaded, if you want your symbols to be useful for two machine debugging. In my case, I had 2 x86 Macs so I used the development machine for this section of the instructions and was able to debug remotely (and it works really well). The PPC <-> x86 debug will end up being a particular problem for this exercise, since you must successfully load the KEXT into the x86 machine before you can run the kextstat to get the address you need to supply to kextload to relocate the kext symbols to the correct address so the symbol file can be loaded into gdb to use it for two machine debugging. That basically would mean printf debugging until the load works, and intentionally delaying probe or anything else that might break until you are ready to debug; then kextstat; then kextload to load the symbols on the x86 machine; then copying the symbols over to the PPC machine, then debugging. I think the intent was to not tie the kextload command to the machine doing the debugging vs. the machine being debugged. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Documentation bug (bis): another source of confusion
by Garth Cummings 31 Jul '08

31 Jul '08

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi Stéphane, On Jul 31, 2008, at 6:02 AM, Stéphane Sudre wrote: http://developer.apple.com/DOCUMENTATION/Darwin/Conceptual/KEXTConcept/KEXT… And point 5 is: "Optionally, you can rename the symbol file." Best regards, --gc ____________________________________________________________________ Garth Cummings gcummings(a)apple.com Sr. Software Engineer Apple Developer Technical Support <http://developer.apple.com/technicalsupport> _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… In the "Creating a Symbol File on the Target Machine" chapter, we can read this on point 4: "[...]Do not rename the symbol file if you are debugging with a DWARF symbol file (.dSYM). If you are debugging with a STABS symbol file (.sym), you may rename the file so long as the name still ends with .sym.[...]" So optionally, you can rename something you should not rename. Documentation bugs can be filed the same way as OS or hardware bugs: <http://developer.apple.com/bugreporter >. That's the official bug reporting mechanism. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Documentation bug: host vs target vs development
by Stéphane Sudre 31 Jul '08

31 Jul '08

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On Jul 31, 2008, at 6:59 PM, Terry Lambert wrote: There's a naming bug IMHO in the documentation at http://developer.apple.com/DOCUMENTATION/Darwin/Conceptual/KEXTConcept/KEXT… This is a source of confusion. Let's take your example: The target machine is a x86. The development machine is a PPC. Now, I introduce the "host machine". Which one is it? The x86 or PPC? My $0.02 _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… On Jul 31, 2008, at 5:51 AM, Stéphane Sudre <ssudre(a)intego.com> wrote: The documentation is using "Target machine" and "Development machine" in almost all the document. But in the "Create a Symbol File" chapter, it's using "Host machine". Which one is supposed to be the host: the target or the development machine? It's even more confusing to Frenchies as a host in French can be the person who receives or is received. Symbol files are created via the kextload command and must specify the load address of the KEXT, which is only known after it is loaded into a kernel. Consider the case where you are developing on a PPC machine, targetting an x86 machine, because you intend to do two machine debugging of the x86 machine. Maybe you are doing this because these are the only two machines you have. The issue is not with the command line. It's that it would be better to keep on using the same name for the machines. In my case, I had 2 x86 Macs so I used the development machine for this section of the instructions and was able to debug remotely (and it works really well). This email sent to site_archiver(a)lists.apple.com

1 0

Re: msleep : question and typo
by Stéphane Sudre 31 Jul '08

31 Jul '08

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On Jul 31, 2008, at 6:40 PM, Terry Lambert wrote: Typo: "you should with either use wait_queue functions or use" -> with--; Question: Does it: - wake up all the sleeping "threads" at once? or - wake up only one sleeping thread? Thanks. _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… On Jul 31, 2008, at 8:39 AM, Rick Macklem <rmacklem(a)uoguelph.ca> wrote: On Thu, 31 Jul 2008, Stéphane Sudre wrote: http://developer.apple.com/documentation/Darwin/Conceptual/KernelProgrammin… /apple_ref/doc/uid/TP30000905-CH219-CHDFIGGH Let's say you are in a re-entrant callback and you make a call to msleep(void * channel, ...) on multiple calls with the same channel and you then issue a wakeup(void * channel) call. Normally wakes up all threads sleeping on that event at the time of the wake up call. (A common practice is to put the msleep() in a loop checking for the condition it is sleeping for. ie. Using "while(..) msleep();" instead of "if(...) msleep();". This does result in threads waking up and then going back to sleep, so if you can use separate events easily, that could be more efficient.) Both wakeup ("wake all sleepers") and wakeup_one ("wakeup exactly one sleeper") are exported KPI. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Documentation bug: host vs target vs development
by Terry Lambert 31 Jul '08

31 Jul '08

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com There's a naming bug IMHO in the documentation at http://developer.apple.com/DOCUMENTATION/Darwin/Conceptual/KEXTConcept/KEXT… This is a source of confusion. -- Terry _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… On Jul 31, 2008, at 5:51 AM, Stéphane Sudre <ssudre(a)intego.com> wrote: The documentation is using "Target machine" and "Development machine" in almost all the document. But in the "Create a Symbol File" chapter, it's using "Host machine". Which one is supposed to be the host: the target or the development machine? It's even more confusing to Frenchies as a host in French can be the person who receives or is received. Symbol files are created via the kextload command and must specify the load address of the KEXT, which is only known after it is loaded into a kernel. Consider the case where you are developing on a PPC machine, targetting an x86 machine, because you intend to do two machine debugging of the x86 machine. Maybe you are doing this because these are the only two machines you have. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Documentation bug (ter): missing instruction
by Terry Lambert 31 Jul '08

31 Jul '08

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com http://developer.apple.com/DOCUMENTATION/Darwin/Conceptual/KEXTConcept/KEXT… [...] " -- Terry _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… On Jul 31, 2008, at 6:35 AM, Stéphane Sudre <ssudre(a)intego.com> wrote: In the "Creating a Symbol File on the Host Machine" chapter, the first instructions are: "To create a symbol file on the host machine, you must have a copy of the kernel extension on the host machine. The host machine must have the same architecture as the target machine (for example, two Intel machines). You must also have the kernel debug kit that correspondes with the version of Mac OS X installed on the target machine. • Create the symbol file. sudo kextload -n -k /Volumes/KernelDebugKit/mach_kernel -s /tmp/ syms/ HelloIOKit.kext Adjust the path to the kernel debug kit and your kernel extension as appropriate. This assumes that the /tmp/syms folder exists by default on Mac OS X. That's not the case: /tmp/syms/: No such file or directory When you specify the output file to any command via a command line argument, the directory components prior to the terminal path component are assummed to exist in order to permit the creation of the output file. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Documentation bug (ter): missing instruction
by Terry Lambert 31 Jul '08

31 Jul '08

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com http://developer.apple.com/DOCUMENTATION/Darwin/Conceptual/KEXTConcept/KEXT… [...] " /tmp/syms/: No such file or directory _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/tlambert%40apple.com _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… On Jul 31, 2008, at 6:35 AM, Stéphane Sudre <ssudre(a)intego.com> wrote: In the "Creating a Symbol File on the Host Machine" chapter, the first instructions are: "To create a symbol file on the host machine, you must have a copy of the kernel extension on the host machine. The host machine must have the same architecture as the target machine (for example, two Intel machines). You must also have the kernel debug kit that correspondes with the version of Mac OS X installed on the target machine. • Create the symbol file. sudo kextload -n -k /Volumes/KernelDebugKit/mach_kernel -s /tmp/ syms/ HelloIOKit.kext Adjust the path to the kernel debug kit and your kernel extension as appropriate. This assumes that the /tmp/syms folder exists by default on Mac OS X. That's not the case: This email sent to tlambert(a)apple.com This email sent to site_archiver(a)lists.apple.com

1 0

Re: msleep : question and typo
by Terry Lambert 31 Jul '08

31 Jul '08

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Typo: http://developer.apple.com/documentation/Darwin/Conceptual/KernelProgrammin… "you should with either use wait_queue functions or use" -> with--; Question: Does it: - wake up all the sleeping "threads" at once? or - wake up only one sleeping thread? -- Terry _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… On Jul 31, 2008, at 8:39 AM, Rick Macklem <rmacklem(a)uoguelph.ca> wrote: On Thu, 31 Jul 2008, Stéphane Sudre wrote: Let's say you are in a re-entrant callback and you make a call to msleep(void * channel, ...) on multiple calls with the same channel and you then issue a wakeup(void * channel) call. Normally wakes up all threads sleeping on that event at the time of the wake up call. (A common practice is to put the msleep() in a loop checking for the condition it is sleeping for. ie. Using "while(..) msleep();" instead of "if(...) msleep();". This does result in threads waking up and then going back to sleep, so if you can use separate events easily, that could be more efficient.) Both wakeup ("wake all sleepers") and wakeup_one ("wakeup exactly one sleeper") are exported KPI. This email sent to site_archiver(a)lists.apple.com

1 0

Re: inode for a given vnode or filepath.
by Quinn 31 Jul '08

31 Jul '08

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com I would like to find out, how can I get an inode number for a given vnode or filename. For a vnode I can use vnode_getattr(), but it requires vfs context as one of it's parameters. I can get the context by using vfs_context_current(), but it seems to be in com.apple.kpi.unsupported. Are there any other *supported* ways to get And at last, does inode number and/or device id change over time and, if they do, under what circumstances? Share and Enjoy -- Quinn "The Eskimo!" <http://www.apple.com/developer/> Apple Developer Relations, Developer Technical Support, Core OS/Hardware _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… At 15:23 +0400 31/7/08, Maxim Zhuravlev wrote: You can create a VFS context for the current thread by calling vfs_context_create, passing NULL to the ctx parameter. That's a complex question. We discussed it recently over <filesystem-dev(a)lists.apple.com>. If you search the list archive for "inode number", you'll find lots of useful info. Specifically, the thread "Can i rely on inode numbers?". This email sent to site_archiver(a)lists.apple.com

1 0

Re: msleep : question and typo
by Quinn 31 Jul '08

31 Jul '08

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com <http://developer.apple.com/bugreporter/> S+E -- Quinn "The Eskimo!" <http://www.apple.com/developer/> Apple Developer Relations, Developer Technical Support, Core OS/Hardware _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… At 11:39 -0400 31/7/08, Rick Macklem wrote: ps: I remember spotting the same typo, so that doc is getting read;-) You can file bugs against the documentation using the standard bug reporting mechanism. I get the feeling that our Developer Publications guy is too busy writing documentation to read this mailing list (-: This email sent to site_archiver(a)lists.apple.com

1 0