June 2005 - Darwin-dev - lists.apple.com

Re: cmpxchgl on x86?
by Jonas Maebe 24 Jun '05

24 Jun '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On 24 jun 2005, at 16:40, David Leimbach wrote: Also, there are compare-and-swap routines already in libSystem: <libkern/OSAtomic.h> /* Compare and swap. They return true if the swap occured. */ bool OSAtomicCompareAndSwap32( int32_t oldValue, int32_t newValue, int32_t *theValue ); bool OSAtomicCompareAndSwap32Barrier( int32_t oldValue, int32_t newValue, int32_t *theValue ); Seems a shame to have to include an entire other framework just for a single assembly instruction :). It'd be really cool if apple had a lighter weight wrapper around such lock code :). libSystem = libc Jonas _______________________________________________ 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: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: cmpxchgl on x86?
by David Leimbach 24 Jun '05

24 Jun '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:reply-to:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=bHhkEV9he6h9xHkNVM/VY/9SubklswJu4oYGgpoX5HgMVo8j3FdgUlNPozQrYWJrpUXcTbX6TPJtn2rxcLugNGU3MZuwN6j6aHeb9rCyjKLpW8KfVzNKCBmWc4AMwPieLi4zPDNhnkt8hFR0k6m+ZlMNciwqB2FDRNO54EFSWfM= > Also, there are compare-and-swap routines already in libSystem: > > <libkern/OSAtomic.h> > > /* Compare and swap. They return true if the swap occured. > */ > bool OSAtomicCompareAndSwap32( int32_t oldValue, int32_t newValue, > int32_t *theValue ); > bool OSAtomicCompareAndSwap32Barrier( int32_t oldValue, int32_t > newValue, int32_t *theValue ); > Seems a shame to have to include an entire other framework just for a single assembly instruction :). It'd be really cool if apple had a lighter weight wrapper around such lock code :). Looking at atomic.s from xnu is what I usually do to find out what I need to hand-code so I can avoid that extra dependency and write more maintainable, and portable code. Dave _______________________________________________ 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: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: How to get the name of an opened semaphore?
by Peter Lovell 24 Jun '05

24 Jun '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Jun 23, 2005, at 9:48 PM, Tron Thomas wrote: Cheers.....Peter _______________________________________________ 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: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… It seems that to solve my problem with semaphores and device space, that I should try unlinking the named semaphores when I close them. Opening a semaphore provides a sem_t pointer that can be used to close the semaphore. Unlinking a semaphore, however, requires the name used when opening the semaphore. How can I obtain the name of the semaphore without having to maintain the relationship between the semaphore name and its corresponding sem_t pointer myself? You can of course unlink the semaphore before then, depending upon the needs of your application. Once it is unlinked, no-one can open it but the existing semaphore remains available until the last reference is closed. In one instance I have, for example, the semaphore is used between threads so the semaphore is unlinked immediately after being opened. Your case may be different (I don't have your original post here right now). You might also look in darwin for the source to ipcs and ipcrm. Obviously they can find the names, but it's possible that you'd need root access. However, it might be that you could find the ones you created. This email sent to site_archiver(a)lists.apple.com

1 0

How to get the name of an opened semaphore?
by Tron Thomas 24 Jun '05

24 Jun '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Mozilla Thunderbird 1.0.2 (Windows/20050317) _______________________________________________ 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: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… It seems that to solve my problem with semaphores and device space, that I should try unlinking the named semaphores when I close them. Opening a semaphore provides a sem_t pointer that can be used to close the semaphore. Unlinking a semaphore, however, requires the name used when opening the semaphore. How can I obtain the name of the semaphore without having to maintain the relationship between the semaphore name and its corresponding sem_t pointer myself? This email sent to site_archiver(a)lists.apple.com

1 0

Does any public interface between Mac system preference network and third-party drivers?
by CPLiang0＠Winbond.com.tw 24 Jun '05

24 Jun '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Thread-index: AcV4WuLLvXA3R+FVRESYKSBoYeAT5Q== Thread-topic: Does any public interface between Mac system preference network and third-party drivers? _______________________________________________ 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: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

What's going on with IOATABlockStorage?
by Kevin Elliott 24 Jun '05

24 Jun '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Is this getting worked on? -- ______________________________________________________ Arguing with an engineer is like wrestling with a pig in mud. After a while, you realize the pig is enjoying it. ______________________________________________________ Kevin Elliott <mailto:kelliott@mac.com> AIM/iChatAV: kelliott(a)mac.com (video chat available) ______________________________________________________ _______________________________________________ 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: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… I went to download the current source to IOATABlockStorage today, and imagine my surprise when I discovered that it hasn't been available since 10.3.5! What's going on? It's listed in both OpenDarwin and Apple's source site, but neither site actually has it available. This email sent to site_archiver(a)lists.apple.com

1 0

Re: cmpxchgl on x86?
by Stan Shebs 23 Jun '05

23 Jun '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Mozilla/5.0 (Macintosh; U; PPC Mac OS X Mach-O; en-US; rv:1.7.3) Gecko/20040910 Andrew Gallatin wrote: I'm trying to compile a unix command-line program on Darwin 8.0 x86, and I'm hitting this problem: /var/tmp/cczbJSUe.s:1993:Can't relocate expression. Absolute 0 assumed. /var/tmp/cczbJSUe.s:1993:operands given don't match any known 386 instruction The code in it is complaining about is an inline cmpxchgl asm routine: static inline uint32_t my__cmpswap_u32(uint32_t *ptr, uint32_t old, uint32_t new) { asm volatile("lock cmpxchgl %1,%2" : "=a" (old) : "q" (new), "m" (*ptr),"0" (old) : "memory"); return old; } Our assembler doesn't accept "lock" as a valid prefix byte. This came up at a WWDC lab too; adding newline is a workaround, but it's reasonable to expect that the assembler does what everybody else's assembler can do, so please file a bug also. Stan _______________________________________________ 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: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This seems to happen with both gcc-4.0 and gcc-3.3 Can somebody tell me what I'm doing wrong, please? This email sent to site_archiver(a)lists.apple.com

1 0

Re: cmpxchgl on x86?
by Andrew Gallatin 23 Jun '05

23 Jun '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Matt Watson writes: > > Can somebody tell me what I'm doing wrong, please? > > Try putting a / after the lock. Thanks.. lock; cmpxchgl seemed to do the trick. > > Also, there are compare-and-swap routines already in libSystem: > Yes, but this code already compiles on Linux, FreeBSD, and Solaris/x86, and that's just the x86 OSes. The last thing I want is to add another ifdef. Thanks for the asm tip! Drew _______________________________________________ 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: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: cmpxchgl on x86?
by Matt Watson 23 Jun '05

23 Jun '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Jun 23, 2005, at 10:11 AM, Andrew Gallatin wrote: I'm trying to compile a unix command-line program on Darwin 8.0 x86, and I'm hitting this problem: The code in it is complaining about is an inline cmpxchgl asm routine: This seems to happen with both gcc-4.0 and gcc-3.3 Can somebody tell me what I'm doing wrong, please? Try putting a / after the lock. Assembler quirk. Also, there are compare-and-swap routines already in libSystem: <libkern/OSAtomic.h> matt. _______________________________________________ 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: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… /var/tmp/cczbJSUe.s:1993:Can't relocate expression. Absolute 0 assumed. /var/tmp/cczbJSUe.s:1993:operands given don't match any known 386 instruction static inline uint32_t my__cmpswap_u32(uint32_t *ptr, uint32_t old, uint32_t new) { asm volatile("lock cmpxchgl %1,%2" : "=a" (old) : "q" (new), "m" (*ptr),"0" (old) : "memory"); return old; } /* Compare and swap. They return true if the swap occured. */ bool OSAtomicCompareAndSwap32( int32_t oldValue, int32_t newValue, int32_t *theValue ); bool OSAtomicCompareAndSwap32Barrier( int32_t oldValue, int32_t newValue, int32_t *theValue ); This email sent to site_archiver(a)lists.apple.com

1 0

cmpxchgl on x86?
by Andrew Gallatin 23 Jun '05

23 Jun '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com I'm trying to compile a unix command-line program on Darwin 8.0 x86, and I'm hitting this problem: /var/tmp/cczbJSUe.s:1993:Can't relocate expression. Absolute 0 assumed. /var/tmp/cczbJSUe.s:1993:operands given don't match any known 386 instruction The code in it is complaining about is an inline cmpxchgl asm routine: static inline uint32_t my__cmpswap_u32(uint32_t *ptr, uint32_t old, uint32_t new) { asm volatile("lock cmpxchgl %1,%2" : "=a" (old) : "q" (new), "m" (*ptr),"0" (old) : "memory"); return old; } This seems to happen with both gcc-4.0 and gcc-3.3 Can somebody tell me what I'm doing wrong, please? I realize that cmpxchgl is not in the 386abi, but I'm using a P4. I've passed -march=pentium to gcc, but to no avail. I've even tried saving the assembler to a temporary file and running: /usr/libexec/gcc/darwin/i386/as -arch pentium foo.s but it still complains. Thanks, Drew _______________________________________________ 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: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0