November 2013 - Darwin-dev - lists.apple.com

Re: fork and pthreads
by Stephen J. Butler 18 Nov '13

18 Nov '13

_______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.ap… This email sent to site_archiver(a)lists.apple.com

1 0

Re: fork and pthreads
by Ariel Burton 18 Nov '13

18 Nov '13

On Sun 17 Nov 19:28, Stephen J. Butler wrote: > That method makes 100% sense to me. Where did you hear lock/unlock is the > way to share pthread structures? The objective of the standard idiom is not to share the pthread entities between processes, but to prevent the child from deadlocking. See, e.g., the dicsussion in Butenhof's book, or here: http://h30097.www3.hp.com/docs/base_doc/DOCUMENTATION/V51B_HTML/MAN/MAN3/22… Note also that the scheme outlined in the link you provided is just as susceptible to the deadlock problem if the forking process is multi-threaded. Ariel Burton _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.ap… This email sent to site_archiver(a)lists.apple.com

1 0

Re: fork and pthreads
by Stephen J. Butler 18 Nov '13

18 Nov '13

_______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.ap… This email sent to site_archiver(a)lists.apple.com

1 0

Re: fork and pthreads
by Stephen J. Butler 18 Nov '13

18 Nov '13

_______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.ap… This email sent to site_archiver(a)lists.apple.com

1 0

fork and pthreads
by ariel.burton＠roguewave.com 18 Nov '13

18 Nov '13

Can anyone help me with this following? The standard idiom for protecting mutexes across fork is to register atfork handlers which lock the mutex in the `prepare' handler before the fork, and unlock them in the post-fork handlers. The toy program given below follows this convention. To aid debugging I've used error checking mutexes. The problem is trying to unlock the mutex in the child, where the attempt fails with EPERM. #include <assert.h> #include <pthread.h> #include <unistd.h> #include <stdlib.h> #include <stdio.h> static pthread_mutex_t mutex = PTHREAD_ERRORCHECK_MUTEX_INITIALIZER; static void lock_mutex ( pthread_mutex_t *m ) { int err = pthread_mutex_lock ( m ); if ( 0 != err ) { printf ( "error locking mutex: %s\n", sys_errlist [ err ] ); exit ( 1 ); } } /* lock_mutex */ static void unlock_mutex ( pthread_mutex_t *m, char *id ) { int err = pthread_mutex_unlock ( m ); if ( 0 != err ) { printf ( "%s: error locking mutex: %s\n", id, sys_errlist [ err ] ); exit ( 1 ); } } /* unlock_mutex */ static void atfork_prepare (void) { lock_mutex ( &mutex ); } /* atfork_prepare */ static void atfork_post_common ( char *id ) { unlock_mutex ( &mutex, id ); } /* atfork_post_common */ static void atfork_parent (void) { atfork_post_common ( "parent" ); } /* atfork_parent */ static void atfork_child (void) { atfork_post_common ( "child" ); } /* atfork_child */ int main ( int argc, char *argv [] ) { int err; pid_t pid; err = pthread_atfork ( atfork_prepare, atfork_parent, atfork_child ); assert ( 0 == err ); pid = fork (); if ( 0 == pid ) { printf ( "child\n" ); } else if ( pid > 0 ) { printf ( "parent\n" ); } else { perror ( "error forking\n" ); } return 0; } I build like this: cc -g t_atfork.c -lpthread -o t_atfork and I see the following output: i386-darwin% ./t_atfork parent child: error unlocking mutex: Operation not permitted What am I not getting right? Thanks, Ariel Burton _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.ap… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Code Signing Kernel Extensions on 10.8 and Earlier
by Chris Suter 13 Nov '13

13 Nov '13

Hi Nick, On Wed, Nov 13, 2013 at 1:04 PM, Nick Blievers <nblievers(a)threatmetrix.com> wrote: > And it works? doh! I took it at face value and have different kext’s signed on different machines! It seems to, but I’m reluctant to ship it without knowing the reason Apple said it was an issue. It would be nice to avoid having to do separate builds for different versions of OS X. It’s possible that Apple meant that you cannot use the 10.9 SDK and ship on earlier versions of OS X (which has always been an issue) and that’s fine for me as I’m not doing anything that requires it. Kind regards, Chris _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.ap… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Code Signing Kernel Extensions on 10.8 and Earlier
by Nick Blievers 13 Nov '13

13 Nov '13

And it works? doh! I took it at face value and have different kext’s signed on different machines! On 13 Nov 2013, at 12:56 pm, Chris Suter <chris(a)sutes.me> wrote: > Nope. Like this: > > codesign --keychain <keychain> -s <identity> <kext> > > -- Chris > > On Wed, Nov 13, 2013 at 12:37 PM, Derrick Brashear <shadow(a)gmail.com> wrote: >> are you signing with --no-macho? >> >> Derrick >> >> >> On Nov 12, 2013, at 20:10, Chris Suter <chris(a)sutes.me> wrote: >> >>> Hi all, >>> >>> In the WWDC 2013 session 707, it says that that OS X 10.9 signed kexts >>> *will not load* on releases prior to OS X 10.8. >>> >>> Our signed kext (built with an earlier SDK) appears to load fine on >>> earlier releases as well as 10.9. I have tested 10.5, 10.6 and 10.8 >>> and 10.9. >>> >>> I’m hoping that someone knowledgable could comment on this discrepancy. >>> >>> Kind regards, >>> >>> Chris >>> >>> _______________________________________________ >>> Do not post admin requests to the list. They will be ignored. >>> Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) >>> Help/Unsubscribe/Update your Subscription: >>> https://lists.apple.com/mailman/options/darwin-dev/shadow%40gmail.com >>> >>> This email sent to shadow(a)gmail.com > > _______________________________________________ > Do not post admin requests to the list. They will be ignored. > Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) > Help/Unsubscribe/Update your Subscription: > https://lists.apple.com/mailman/options/darwin-dev/nblievers%40threatmetrix… > > This email sent to nblievers(a)threatmetrix.com _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.ap… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Code Signing Kernel Extensions on 10.8 and Earlier
by Chris Suter 13 Nov '13

13 Nov '13

Nope. Like this: codesign --keychain <keychain> -s <identity> <kext> -- Chris On Wed, Nov 13, 2013 at 12:37 PM, Derrick Brashear <shadow(a)gmail.com> wrote: > are you signing with --no-macho? > > Derrick > > > On Nov 12, 2013, at 20:10, Chris Suter <chris(a)sutes.me> wrote: > >> Hi all, >> >> In the WWDC 2013 session 707, it says that that OS X 10.9 signed kexts >> *will not load* on releases prior to OS X 10.8. >> >> Our signed kext (built with an earlier SDK) appears to load fine on >> earlier releases as well as 10.9. I have tested 10.5, 10.6 and 10.8 >> and 10.9. >> >> I’m hoping that someone knowledgable could comment on this discrepancy. >> >> Kind regards, >> >> Chris >> >> _______________________________________________ >> Do not post admin requests to the list. They will be ignored. >> Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) >> Help/Unsubscribe/Update your Subscription: >> https://lists.apple.com/mailman/options/darwin-dev/shadow%40gmail.com >> >> This email sent to shadow(a)gmail.com _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.ap… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Code Signing Kernel Extensions on 10.8 and Earlier
by Derrick Brashear 13 Nov '13

13 Nov '13

are you signing with --no-macho? Derrick On Nov 12, 2013, at 20:10, Chris Suter <chris(a)sutes.me> wrote: > Hi all, > > In the WWDC 2013 session 707, it says that that OS X 10.9 signed kexts > *will not load* on releases prior to OS X 10.8. > > Our signed kext (built with an earlier SDK) appears to load fine on > earlier releases as well as 10.9. I have tested 10.5, 10.6 and 10.8 > and 10.9. > > I’m hoping that someone knowledgable could comment on this discrepancy. > > Kind regards, > > Chris > > _______________________________________________ > Do not post admin requests to the list. They will be ignored. > Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) > Help/Unsubscribe/Update your Subscription: > https://lists.apple.com/mailman/options/darwin-dev/shadow%40gmail.com > > This email sent to shadow(a)gmail.com _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.ap… This email sent to site_archiver(a)lists.apple.com

1 0

Code Signing Kernel Extensions on 10.8 and Earlier
by Chris Suter 13 Nov '13

13 Nov '13

Hi all, In the WWDC 2013 session 707, it says that that OS X 10.9 signed kexts *will not load* on releases prior to OS X 10.8. Our signed kext (built with an earlier SDK) appears to load fine on earlier releases as well as 10.9. I have tested 10.5, 10.6 and 10.8 and 10.9. I’m hoping that someone knowledgable could comment on this discrepancy. Kind regards, Chris _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.ap… This email sent to site_archiver(a)lists.apple.com

1 0