December 2009 - Darwin-kernel

Debugging a hard system wedge
by Godfrey van der Linden 11 Dec '09

11 Dec '09

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com G'day, All. Really there is no other place to ask this. I mean if you have a hammer (some knowledge of kernel programming) everything looks like a nail. On my brand new MacBookPro5,3 (2.66Ghz) 10.6.2 system I have been getting unexplained hard wedges. The symptoms:- - The menu clock seconds counter keeps ticking. - The mouse still moves - The fans do not spin up. - top in terminal stops reporting - ssh into the system doesn't work. - After a wakeup the login window works to type the password in, but it hangs around with a spinning cog and never logs me in. - I suspect that mDNS isn't going either as the machines <name>.local doesn't seem to resolve, though I'm not sure of that. - May be unrelated, the system never, ever goes to idle sleep. I must close the lid or it will happily sit there for hours with the screen off but otherwise running. I haven't attempted to connect to the system with GDB yet as the symptoms feels wrong to me as a kernel hang. I'd like the groups opinion on the utility of showallstacks and other magic kgmacros to track this down. I don't really have any interesting kernel level apps running, for instance I do not run any of the virtualisers. I do run teleport that I think may have a kext but it seems pretty harmless generally -- I have deactivated it in the hope of localising the problem. _______________________________________________ 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.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: audit warning?
by Joe van Tunen 10 Dec '09

10 Dec '09

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thread-index: Acp59BoUWHJWVOXnEd65QwAX8uBReA== Thread-topic: audit warning? User-agent: Microsoft-Entourage/11.4.0.080122 > From: Godfrey van der Linden <gvdl(a)mac.com> > Date: Fri, 11 Dec 2009 10:27:42 +1100 > > Probably the wrong list, but darwin-dev seems to have died and the cocoa lists > aren't really the right place either. > > Saw these few lines go past in my secure.log yesterday. I don't recognise > them and I'm a bit concerned. Can anybody out there explain them? > > Dec 10 19:47:13 [snipped] [ loginwindow[38]: in pam_sm_authenticate(): Failed > to determine Kerberos principal name. > Dec 10 19:47:14 [snipped] [9430]: audit warning: allsoft > Dec 10 19:47:14 [snipped] [[9431]: audit warning: soft /var/audit > Dec 10 19:47:14 [snipped] [[9433]: audit warning: closefile > /var/audit/20091208225521.20091210084714 > Dec 10 19:47:52 [snipped] [[9458]: audit warning: soft /var/audit > > > Godfrey I googled and got http://www.apple.com/support/security/commoncriteria/CC_AdminGuide.pdf What version of Mac OS X are you using? _______________________________________________ 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.… This email sent to site_archiver(a)lists.apple.com

1 0

audit warning?
by Godfrey van der Linden 10 Dec '09

10 Dec '09

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Probably the wrong list, but darwin-dev seems to have died and the cocoa lists aren't really the right place either. Saw these few lines go past in my secure.log yesterday. I don't recognise them and I'm a bit concerned. Can anybody out there explain them? Dec 10 19:47:13 [snipped] [ loginwindow[38]: in pam_sm_authenticate(): Failed to determine Kerberos principal name. Dec 10 19:47:14 [snipped] [9430]: audit warning: allsoft Dec 10 19:47:14 [snipped] [[9431]: audit warning: soft /var/audit Dec 10 19:47:14 [snipped] [[9433]: audit warning: closefile /var/audit/20091208225521.20091210084714 Dec 10 19:47:52 [snipped] [[9458]: audit warning: soft /var/audit Godfrey _______________________________________________ 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.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: About sudo and su on 10.6
by Chris Suter 09 Dec '09

09 Dec '09

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi Yuji Oishi, 2009/12/9 Bigstone <bigstone.gm(a)hscjpn.co.jp> > The gloval variable tgetpass_flags dose not initialzed before using macro SET(x,y). I don't know what your problem is, but all global variables are implicitly initialised to zero, so that’s not your problem. I would guess that the problem lies with your code. In any case, unless it’s for debugging/testing purposes, I can’t think of a reason why you’d ever want to run sudo that way. There are almost always better ways of doing things so perhaps you could explain what you’re trying to do, although this list perhaps isn’t the best place for that (sudo isn’t really the darwin Kernel). Kind regards, Chris _______________________________________________ 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.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Thread join?
by Michael Smith 09 Dec '09

09 Dec '09

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On Dec 9, 2009, at 12:42 PM, Duane Murphy wrote: > How would one do the equivalent of join() for a thread in the kernel? Typically, you don't. > I think I have a situation where the underlying thread is exiting, > possibly aborting, so that my built in thread exit indicators do not > get called. This causes some threads to hang waiting for the thread to > exit. Threads don't normally abort; if your toplevel thread routine is responsible for tracking your threads' existence then control will pass back through it when a thread is terminating. > Or maybe the right question is what is the appropriate way to wait on > a thread that may abort or already be aborted? You don't; you should not terminate threads that you know (may) have waiters. In general, you should not be creating and destroying kernel threads in your workflow. = Mike _______________________________________________ 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.… This email sent to site_archiver(a)lists.apple.com

1 0

About sudo and su on 10.6
by Bigstone 09 Dec '09

09 Dec '09

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.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.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Thread join?
by Duane Murphy 09 Dec '09

09 Dec '09

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256 Well, I'm still trying to work out what is happening. I am working on a disk driver (override). When the disk is removed (without being ejected) things sometimes go awry. It seems that if the timing is write and there is a thread that is making a request to the disk at the time the disk is removed then after a timeout, it seems that thread may be aborted. I'm not entirely sure this is true, but I can end up with another thread waiting on a mutex that the thread that is making the request to the drive is supposed to release. But the thread is gone and the mutex is still locked. In summary, I'm using a mutex to indicate that a thread has completed. One thread waits on that mutex until the IO thread exits. Under these circumstances, the IO thread has gone, but the mutex is still locked. Still doing debugging, but thread synchronization seems to be something useful here. I guess a related question would be how to tell that a thread is still valid and active. That is a thread's state (given a thread reference of course). ...Duane On Dec 8, 2009, at 10:03 PM, Godfrey van der Linden wrote: > IIRC there is no automatic way of doing this. In any case I had to > roll my own with complicated locking and other weird hand off code. > How is your thread terminating, are you returning from main or is it > being aborted by another thread (if that is even possible)? > > I've just had a look at the IOWorkLoop termination code, it is > pretty complicated to get right it seems to be necessary to call > thread_deallocate() followed by thread_terminate(), this is all new > to me so something has changed since I last worked in this area. > > Godfrey > > On 2009-12-09, at 12:42 PM, Duane Murphy wrote: > >> How would one do the equivalent of join() for a thread in the kernel? >> >> I think I have a situation where the underlying thread is exiting, >> possibly aborting, so that my built in thread exit indicators do not >> get called. This causes some threads to hang waiting for the thread >> to >> exit. >> >> Or maybe the right question is what is the appropriate way to wait on >> a thread that may abort or already be aborted? >> >> ...Duane > ...Duane -----BEGIN PGP SIGNATURE----- Version: PGP Desktop 9.10.0 (Build 500) Charset: US-ASCII wsBVAwUBSx9GJkrg9acQ4r2CAQjmRAgAzkArJ+FT4MRIsTspfYbkS+3Drr8T9fG7 wWyDppgKXD1vqTFbsbUIQkaArcL6g+eVwmzOnKO8bIta4pi6UMzQGclz6fNlNqiK ZiDbHpbSqGwW9LE/BwjwdwqCifyYUGFmITFp6YZ9SsbFzRqHaMHtKbstJSb3U1so 3yxPlRDQpvKetppzKFXgLRCquC//jy06omp1cxOMh0oA/5Cs7YVRzYsjxfe/yv4u 4pD5qade7XIu0Y7j0CR+j+hXncq9ymuwkTF29t+yv0E0pWsQ/9gGIJKYoMOpNeaK bOKkKFxTHtcF8PKv4FVWeosgaBgT4xvBP6HoNOSHc5wPHwOGvea5Tg== =cUsb -----END PGP SIGNATURE----- _______________________________________________ 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.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Thread join?
by Godfrey van der Linden 09 Dec '09

09 Dec '09

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com IIRC there is no automatic way of doing this. In any case I had to roll my own with complicated locking and other weird hand off code. How is your thread terminating, are you returning from main or is it being aborted by another thread (if that is even possible)? I've just had a look at the IOWorkLoop termination code, it is pretty complicated to get right it seems to be necessary to call thread_deallocate() followed by thread_terminate(), this is all new to me so something has changed since I last worked in this area. Godfrey On 2009-12-09, at 12:42 PM, Duane Murphy wrote: > How would one do the equivalent of join() for a thread in the kernel? > > I think I have a situation where the underlying thread is exiting, > possibly aborting, so that my built in thread exit indicators do not > get called. This causes some threads to hang waiting for the thread to > exit. > > Or maybe the right question is what is the appropriate way to wait on > a thread that may abort or already be aborted? > > ...Duane _______________________________________________ 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.… This email sent to site_archiver(a)lists.apple.com

1 0

Thread join?
by Duane Murphy 09 Dec '09

09 Dec '09

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256 How would one do the equivalent of join() for a thread in the kernel? I think I have a situation where the underlying thread is exiting, possibly aborting, so that my built in thread exit indicators do not get called. This causes some threads to hang waiting for the thread to exit. Or maybe the right question is what is the appropriate way to wait on a thread that may abort or already be aborted? ...Duane -----BEGIN PGP SIGNATURE----- Version: PGP Desktop 9.10.0 (Build 500) Charset: US-ASCII wsBVAwUBSx8AlUrg9acQ4r2CAQitdQf+L55ZFiZJXY2eb7pCuuMBVnxiB8/Sw5v/ JiRNvGuaSwUJAuZx6DwQHp3I3TK4UmXOuOzRnyuYUncLcA31FzPzNoLvTFScAn/h mPtnVsEKxK5oyxCRg29ww10BddxNnRCfxtx7pKzqBbdfJxpwko3I/Rn/CizK0NDJ 6yWwykhhOXshLpaaCzizeP7m9NWXil8WkWSfJYCtkWd41gRfasymoaXnXnIxzRwO lr24mNx2grdqesHhuFhgEcZ5oUFDtjladlgRPZV1TmmeMG5v7sok29qyeJa/f8Yr ZBCRdO1Q3EpAVPjIfKZw16j3ExwjuYYF2oqv7R0taOmp9G1B5Cda+w== =swQS -----END PGP SIGNATURE----- _______________________________________________ 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.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: undefined symbols: DiskArbInit/DiskArbDiskAppearedWithMountpointPing_auto
by Quinn 04 Dec '09

04 Dec '09

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com The best list for DiskArb questions is filesystem-dev. <http://lists.apple.com/mailman/listinfo/filesystem-dev> And I will answer here anyway (-: 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 15:49 -0800 3/12/09, mogambo wrote: Not sure where this question belongs - here or on xcode-userslist or drivers. I will ask here anyway. DiskArbDiskAppearedWithMountpointPing_auto is not necessary since 10.5. Thus, you should ifdef out the calls in your 64-bit build. This means, if you still support 10.4.x, you must ensure that your 32-bit code runs on 10.4.x. Depending on how your code is structured, this may require a re-exec dance. This email sent to site_archiver(a)lists.apple.com

1 0