May 2003 - Darwin-kernel - lists.apple.com

Having problems with .gdbinit kernel macros...
by Shawn Erickson 31 May '03

31 May '03

Most of the kernel debugging macros fail but some do work. I am using "xnu/osfmk/.gdbinit". Do I have a mismatch between the .gdbinit file and the kernel I am using? Or do I need to have a debug/symboled build of the kernel for them to work? For example... (gdb) showallstacks Attempt to extract a component of a value that is not a structure. (gdb) paniclog Argument to arithmetic operation not a number or boolean. (gdb) showalltasks task vm_map ipc_space #acts pid proc command Attempt to extract a component of a value that is not a structure. (gdb) showallacts Attempt to extract a component of a value that is not a structure. (gdb) showallvm Attempt to extract a component of a value that is not a structure. (gdb) showallkmods kmod address size id refs version name 0x10037a9c 0x10024000 0x00014000 96 0 1.0 com.ftsw.driver.qlc22xx 0x0ff77398 0x0ff69000 0x0000f000 95 1 1.0 com.ftsw.driver.qlcfchba 0x100a36c8 0x10093000 0x00011000 81 0 1.2.4 com.apple.iokit.IOSCSIBlockCommandsDevice 0x0ff501bc 0x0ff48000 0x00009000 78 1 1.2.4 com.apple.iokit.IOSCSIParallelFamily 0x0ff32980 0x0ff1f000 0x00014000 72 2 1.2.4 com.apple.iokit.IOSCSIArchitectureModelFamily 0x0fd6515c 0x0fd60000 0x00006000 71 0 1.2.1 com.apple.nke.IPFirewall 0x0fc735c0 0x0fc6e000 0x00006000 70 0 1.3.1 com.apple.nke.SharedIP 0x0f829bdc 0x0f827000 0x00003000 69 0 1.0.8 com.apple.driver.AppleLED 0x0f8110ac 0x0f80e000 0x00004000 66 0 1.0.7 com.apple.driver.AppleCPUThermo 0x0f789708 0x0f779000 0x00011000 65 0 1.0.5 com.apple.iokit.IOFireWireIP 0x0f735348 0x0f733000 0x00003000 64 0 1.0.2 com.apple.driver.AppleFPButton 0x0f7305d8 0x0f72e000 0x00003000 63 0 1.0.3 com.apple.driver.AppleBlower [snip] _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: Scanning devices from inside the kernel
by Sam Vaughan 30 May '03

30 May '03

Thanks Dan, getMatchingServices() does exactly what I need. Having the IOService object for the devices I want to talk to is going to be a huge help. The key was finding the OS classes in libkern that let you do CoreFoundation type things in the kernel. Here's the code for anyone who's interested: OSDictionary* services = IOService::serviceMatching(kIOMediaClass); if (!services) return; OSIterator* iterator = IOService::getMatchingServices(services); if (!iterator) return; while (true) { IOService* service = OSDynamicCast(IOService, iterator->getNextObject()); if (!service) break; printf("name: %s\n", service->getName()); service->release(); } iterator->release(); services->release(); Sam On Thursday, May 29, 2003, at 15:00 Australia/Melbourne, darwin-kernel-request(a)lists.apple.com wrote: If the volume management could be done in mostly in user space, all the better, easier, safer, etc, but if not, what you want is doable of course. You can look up the media objects by BSD name or all media objects by IOMedia class, just as in user-space. See getMatchingServices(). _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: Execve
by Quinn 29 May '03

29 May '03

At 15:29 -0700 28/5/03, Justin C. Walker wrote: The Dock is either using Finder or doing what Finder does (I don't know the code). GUI applications are launched by means of high-level APIs (Workspace for Cocoa, Launch Services for Carbon and Cocoa, Process Manager for Carbon and Cocoa). These are layered on top of Core Processes (a private, internal API), which is implemented within the window server. When the Finder launches a program, Launch Services communicates with the window server, which does the fork/exec. This is why the window server is the parent process of all GUI applications. Ultimately, however, all processes are launched via fork/exec. S+E -- Quinn "The Eskimo!" <http://www.apple.com/developer/> Apple Developer Technical Support * Networking, Communications, Hardware _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: Locking and synchronization primitives
by Wally Crooze 29 May '03

29 May '03

A couple of more things on locking/synchronization primitives... How effective are the implementations of funnels and other locking mechanisms (e.g. simple_locks) for SMP systems? I noticed that simple_lock is defined as either 1 (as funnels are in effect), disable_preemption (on Uni-processors) or usimple_lock (on SMP systems). I was under the impression that usimple_locks were for uni-processor systems... this is confusing. Also, can the locking and synchronization primitives be nested safely? i.e. simple_lock(lock1) simple_lock(lock2) simple_unlock(lock2) simple_unlock(lock1) Does simple_locks disable interrupts, etc? Also what interface is equivalent to synch-variables... such as sv_init, sv_rmproc? I'm also looking for a mechanism that will allow me to attempt to upgrade (i.e. read to write) an mrlock, but if it can't get it immediately it will return still holding the read (similar to the try versions of the other locks). Cheers. -- Wally Crooze +-------------------------------------------------------+ | ,-_|\ Wally Crooze <wallycrooze(a)pobox.com> | | / \ | | \_,-._/ | | v | +-------------------------------------------------------+ _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: Scanning devices from inside the kernel
by Dan Markarian 29 May '03

29 May '03

Hi Sam, If the volume management could be done in mostly in user space, all the better, easier, safer, etc, but if not, what you want is doable of course. You can look up the media objects by BSD name or all media objects by IOMedia class, just as in user-space. See getMatchingServices(). http://developer.apple.com/techpubs/macosx/Darwin/General/ WritingDeviceDriver/IOService/chapter_4_section_3.html http://developer.apple.com/techpubs/macosx/Darwin/Reference/Kernel/ IOService/Classes/IOService/Functions/Functions.html http://developer.apple.com/techpubs/macosx/Darwin/Reference/Kernel/ IORegistryEntry/Classes/IORegistryEntry/Functions/Functions.html Dan On Wednesday, May 28, 2003, at 09:13 PM, Sam Vaughan wrote: I'm porting a volume manager that is part of a file system to Mac OS X. One of the first things I need to do is to scan the device tree and find the disks I'm interested in. The only related documentation I've found describes how to access hardware from applications using IOKit. What I ended up doing in my kext is including a Darwin header, xnu/bsd/miscfs/devfs/devfsdefs.h, and walking the device tree to find the disks and their dev_t (and thus major and minor) numbers. I know this is frowned upon, but I couldn't find another way of doing it from inside the kernel. The problem I have now is distinguishing the disk entries I'm interested in from the ones I'm not. How can I talk to the driver associated with each device? I'd like to get target and lun numbers for the disks that are connected via a fibre channel HBA. Whilst I can see these numbers in IORegistryExplorer, I can't find a way of accessing them from inside the kernel. What is the recommended approach for getting information on devices in the device tree from inside the kernel? Also, am I wrong in my assumption that the IOKit is of little use to me inside a kext? Sam _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored. _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Scanning devices from inside the kernel
by Sam Vaughan 29 May '03

29 May '03

I'm porting a volume manager that is part of a file system to Mac OS X. One of the first things I need to do is to scan the device tree and find the disks I'm interested in. The only related documentation I've found describes how to access hardware from applications using IOKit. What I ended up doing in my kext is including a Darwin header, xnu/bsd/miscfs/devfs/devfsdefs.h, and walking the device tree to find the disks and their dev_t (and thus major and minor) numbers. I know this is frowned upon, but I couldn't find another way of doing it from inside the kernel. The problem I have now is distinguishing the disk entries I'm interested in from the ones I'm not. How can I talk to the driver associated with each device? I'd like to get target and lun numbers for the disks that are connected via a fibre channel HBA. Whilst I can see these numbers in IORegistryExplorer, I can't find a way of accessing them from inside the kernel. What is the recommended approach for getting information on devices in the device tree from inside the kernel? Also, am I wrong in my assumption that the IOKit is of little use to me inside a kext? Sam _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: Execve
by Justin C. Walker 28 May '03

28 May '03

On Wednesday, May 28, 2003, at 03:03 PM, Robert Sandilands wrote: I'm trying to determine how program execution is started in the kernel. I know that there is an entry points trough syscall[SYS_execve] and syscall[SYS_execv]. This seems to only catch programs executed from the Terminal. But if you launch programs through the dock, it does not seem to pass through those interfaces. Any ideas of how executing something like iCal goes through the kernel? All roads lead to execve. See xnu/bsd/kern/kern_exec.c. Terminal plays no part in this. The shell is what you are talking to when running Terminal, and it does just what Finder does, albeit in a more limited way. In either case, your shell or Finder finally arrives at a path to an executable, and it runs that executable with a variant of the exec() system call. For "apps" that actually aren't executables (in the Unix sense), another program is launched (or contacted some how) to read in the content of the file to be "executed" and set it up to execute in some way (e.g., the Carbon/CFM apps are handled LaunchCFM; shell scripts by a named shell). The Dock is either using Finder or doing what Finder does (I don't know the code). Regards, Justin -- /~\ The ASCII Justin C. Walker, Curmudgeon-at-Large \ / Ribbon Campaign X Help cure HTML Email / \ _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Execve
by Robert Sandilands 28 May '03

28 May '03

I'm trying to determine how program execution is started in the kernel. I know that there is an entry points trough syscall[SYS_execve] and syscall[SYS_execv]. This seems to only catch programs executed from the Terminal. But if you launch programs through the dock, it does not seem to pass through those interfaces. Any ideas of how executing something like iCal goes through the kernel? Robert Sandilands --------------------------------------------------------------------- #include http://rsandila.webhop.org/disclaimer.html _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: How much VM is available to a process?
by Pelle Johansson 28 May '03

28 May '03

onsdagen den 28 maj 2003 kl 07.53 skrev Rob Clevenger: I have a user app that is a memory hog. When I view it in "Process Viewer" I see that it consumes close to 3 gb of virtual memory before the program complains about not enough memory. Is there a way to bump up the amount of VM available or is this a hard limitation (based on number of address bits maybe)? The whole 32 bit address space (4GB) is available for usage. However, your program and those of all loaded shared libraries and the stack(s) take up some space. The biggest amount of memory you can allocate with malloc() is about 2.5 GB, because malloc() needs a continuous hole and after that it runs into allocated space. You can still allocate the rest of the address space using vm_allocate() directly though. If you need more than 4 GB, you'll have to use some trick. The easiest is probably to use several processes to do the work. They can even write into each other's address space using vm_write() (or temporarily map it up). Another way is creating several temporary files and mmap() one of them. Then when it runs out switch to a new file. The problem is of course that before every data access in this part of the memory you'll have to make sure the correct file is the one mapped. -- Pelle Johansson <morth(a)morth.org> _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

How much VM is available to a process?
by Rob Clevenger 28 May '03

28 May '03

I have a user app that is a memory hog. When I view it in "Process Viewer" I see that it consumes close to 3 gb of virtual memory before the program complains about not enough memory. Is there a way to bump up the amount of VM available or is this a hard limitation (based on number of address bits maybe)? thanks, - rob _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0