May 2009 - Darwin-dev - lists.apple.com

Re: xinetd and launchd in Mac OS X 10.5
by Terry Lambert 27 May '09

27 May '09

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hello, My backup/restore application is supposed to run on Mac OS X 10.4 and 10.5. I need the xinetd startup script. But with 10.4 launchd is been used. However, 10.4 also supports xinetd. Does 10.5 also provide xinetd It's launchd. -- Terry _______________________________________________ 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… On May 27, 2009, at 3:14 AM, rohan a <info1686(a)gmail.com> wrote: Which is the default startup script : xinetd or launchd ? This email sent to site_archiver(a)lists.apple.com

1 0

Retrieve user, host, port and volume for a mounted filesystem
by Mikael Hakman 27 May '09

27 May '09

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.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: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… I need to retrieve user, host, port and volume for a mounted filesystem. When mounted fs is smb then f_mntfromname field in struct statfs returned by statfs() call gives this information as // user@host:port/volume. When mounted fs is afp then this field contains a character string such as afp_31Mvug000bmX0000oM0000VU-1.2f000006. I haven't been able to determine the meaning of this string, neither how to get required information from using it. Any help would be greatly appreciated. Thanks. This email sent to site_archiver(a)lists.apple.com

1 0

xinetd and launchd in Mac OS X 10.5
by rohan a 27 May '09

27 May '09

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:received:date:message-id:subject :from:to:content-type:content-transfer-encoding; bh=XU+SGkBZ5pOfP/bSLl7tUUpO4BCITPBTniIuSBUY8a8=; b=IkJnw7nVkTDzXoGuzIhnqStALyKvnCgDb3nVJ9nEZMZTerAtjNJBYaBmbJ+FsYv7WH bNWfO17wVs6USgzO77IL2n0289nNUuCe91tHlLVILWgIjXLjj5AqnjDvvXBA/TEU+Fex fMT2axslkpE8C2WlwavGgrb5AKQssqMdi2ZQk= Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:date:message-id:subject:from:to:content-type :content-transfer-encoding; b=ARQwlINljF2WY9ugFzUfpuXP3ur74nRKUIIzydYwHbbEPVTgtP9bwIGVKFSHY0yflr 1Ai4wTVRB7lVcWZQnDeQjRATwFqo2UjFKoG7TyAFDaxqA/nRs6Tt30bT4hEq1s4UB9Qu WUx4t900DVBz23LGXnYTOMImt6vjiFEqJTU1A= Hello, My backup/restore application is supposed to run on Mac OS X 10.4 and 10.5. I need the xinetd startup script. But with 10.4 launchd is been used. However, 10.4 also supports xinetd. Does 10.5 also provide xinetd Which is the default startup script : xinetd or launchd ? Thanks _______________________________________________ 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: pread and EACCES error?
by Brian Bergstrand 21 May '09

21 May '09

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hi Ed, On May 21, 2009, at 10:42 PM, Ed Wynne wrote: Hi all, I'm seeing a weird error from pread in the following setup: The copy is done in DKIOCGETMAXBLOCKCOUNTREAD chunks (128KB for ATA). Mac OS X 10.5.7 (Darwin 9.7) on Intel. Hey Brian, Thanks. _______________________________________________ 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… On May 20, 2009, at 2:55 AM, Brian Bergstrand wrote: Copying from a 300GB PATA drive connected via USB to another 300GB ATA drive connected via Firewire using the raw char devices for each. After copying about 37GB of data, pread fails on the source USB drive with EACCES (13). According to the man page, this is not a valid error return (although the man page may be out of date). So where is the EACCES coming from? The IOKit block drivers? And what does it mean in this situation? Thanks for any help?_______________________________________________ Potentially strange question... have you tried retrying the failed read? Or retrying after deblocking it to something smaller? Even better would be to watch an ATA or USB bus trace of the failed read. I suspect your issue is originating in the USB bridge, or the USB stack talking to the USB bridge. I'd bet something odd is happening and its bubbling up as the EACCES return. Also, you didn't say the drive was dead after this happens, so I suspect it's something transient. I've seen this type of thing crop up from time to time with various USB access paths to optical devices. Scary as all get out. -Ed Yes, I was thinking the same thing and we told the customer about as much. I was just wondering if there was a code path in the kernel that could return this error. I will add a retry and see where that gets us. This email sent to site_archiver(a)lists.apple.com

1 0

Re: pread and EACCES error?
by Ed Wynne 21 May '09

21 May '09

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hi all, I'm seeing a weird error from pread in the following setup: The copy is done in DKIOCGETMAXBLOCKCOUNTREAD chunks (128KB for ATA). Mac OS X 10.5.7 (Darwin 9.7) on Intel. Hey Brian, -Ed _______________________________________________ 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 On May 20, 2009, at 2:55 AM, Brian Bergstrand wrote: Copying from a 300GB PATA drive connected via USB to another 300GB ATA drive connected via Firewire using the raw char devices for each. After copying about 37GB of data, pread fails on the source USB drive with EACCES (13). According to the man page, this is not a valid error return (although the man page may be out of date). So where is the EACCES coming from? The IOKit block drivers? And what does it mean in this situation? Thanks for any help?_______________________________________________ Potentially strange question... have you tried retrying the failed read? Or retrying after deblocking it to something smaller? Even better would be to watch an ATA or USB bus trace of the failed read. I suspect your issue is originating in the USB bridge, or the USB stack talking to the USB bridge. I'd bet something odd is happening and its bubbling up as the EACCES return. Also, you didn't say the drive was dead after this happens, so I suspect it's something transient. I've seen this type of thing crop up from time to time with various USB access paths to optical devices. Scary as all get out. smime.p7s

1 0

Re: pread and EACCES error?
by Brian Bergstrand 21 May '09

21 May '09

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On May 21, 2009, at 9:19 PM, Terry Lambert wrote: Hi Terry, Thanks for your help. -- Terry Thanks. On May 21, 2009, at 10:32 AM, Terry Lambert wrote: -- Terry Hi all, I'm seeing a weird error from pread in the following setup: Mac OS X 10.5.7 (Darwin 9.7) on Intel. Thanks for any help?_______________________________________________ Brian Bergstrand <http://www.bergstrand.org/brian/> PGP Key ID: 0xB6C7B6A2 Brian Bergstrand <http://www.bergstrand.org/brian/> PGP Key ID: 0xB6C7B6A2 _______________________________________________ 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… On May 21, 2009, at 1:26 AM, Brian Bergstrand wrote: The code is not multi-threaded. It's a fairly simple command line tool. I don't have a custom KEXT installed either. I do a bunch of preflight to get disk sizes, make sure the devices are not mounted, etc. I then open a read handle on the source device and a read-write handle on the target device. The tool then enters a loop and just preads from the source and pwrites to the target. All on one thread. The tool is linked against CoreFoundation and libSystem - nothing else. If I have a read handle open on the device and something causes diskarb to attempt to remount any volumes on the device could that cause the problem? I don't see how, because I can open a read handle on a disk containing mounted volumes or on the mounted volumes themselves (just not a write handle). A forced revocation of the descriptor could result in an EBADF, but not an EACCES. Also, in general, you should be prepared for any error back from any system call, not just the ones on the man pages. People write all sorts of plugins and there are all sorts of third party drivers for Firewire and thumb drives and other devices, that the man page error numbers merely document what the code paths would return without third party software being involved. Yes, I understand that. And the code does handle any error, but it is an error and therefore the copy is aborted. If I knew why the error occurred I might be able to handle it a different way. 3rd party software is likely not the cause. This error is seen running as root on the 10.5.6 DDK environment DVD and it's been seen running as root on a stock 10.5.7 normal install. It's not reproducible at will, so it's hard to pin anything down. All I have are logs. Also the offset where the error occurs is a native disk block boundary (although if it wasn't I should get back EIO). There is nothing directly in the normal code path that does kauth checks after the open is successful. So it's your driver or your code. Is you code multithreaded? If so, does it serialize the close? On May 20, 2009, at 2:55 AM, Brian Bergstrand wrote: Copying from a 300GB PATA drive connected via USB to another 300GB ATA drive connected via Firewire using the raw char devices for each. After copying about 37GB of data, pread fails on the source USB drive with EACCES (13). According to the man page, this is not a valid error return (although the man page may be out of date). The copy is done in DKIOCGETMAXBLOCKCOUNTREAD chunks (128KB for ATA). So where is the EACCES coming from? The IOKit block drivers? And what does it mean in this situation? This email sent to site_archiver(a)lists.apple.com

1 0

Re: pread and EACCES error?
by Terry Lambert 21 May '09

21 May '09

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hi Terry, -- Terry Thanks. On May 21, 2009, at 10:32 AM, Terry Lambert wrote: -- Terry Hi all, I'm seeing a weird error from pread in the following setup: Mac OS X 10.5.7 (Darwin 9.7) on Intel. Thanks for any help?_______________________________________________ Brian Bergstrand <http://www.bergstrand.org/brian/> PGP Key ID: 0xB6C7B6A2 _______________________________________________ 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… On May 21, 2009, at 1:26 AM, Brian Bergstrand wrote: The code is not multi-threaded. It's a fairly simple command line tool. I don't have a custom KEXT installed either. I do a bunch of preflight to get disk sizes, make sure the devices are not mounted, etc. I then open a read handle on the source device and a read-write handle on the target device. The tool then enters a loop and just preads from the source and pwrites to the target. All on one thread. The tool is linked against CoreFoundation and libSystem - nothing else. If I have a read handle open on the device and something causes diskarb to attempt to remount any volumes on the device could that cause the problem? I don't see how, because I can open a read handle on a disk containing mounted volumes or on the mounted volumes themselves (just not a write handle). A forced revocation of the descriptor could result in an EBADF, but not an EACCES. Also, in general, you should be prepared for any error back from any system call, not just the ones on the man pages. People write all sorts of plugins and there are all sorts of third party drivers for Firewire and thumb drives and other devices, that the man page error numbers merely document what the code paths would return without third party software being involved. There is nothing directly in the normal code path that does kauth checks after the open is successful. So it's your driver or your code. Is you code multithreaded? If so, does it serialize the close? On May 20, 2009, at 2:55 AM, Brian Bergstrand wrote: Copying from a 300GB PATA drive connected via USB to another 300GB ATA drive connected via Firewire using the raw char devices for each. After copying about 37GB of data, pread fails on the source USB drive with EACCES (13). According to the man page, this is not a valid error return (although the man page may be out of date). The copy is done in DKIOCGETMAXBLOCKCOUNTREAD chunks (128KB for ATA). So where is the EACCES coming from? The IOKit block drivers? And what does it mean in this situation? This email sent to site_archiver(a)lists.apple.com

1 0

Re: pread and EACCES error?
by Brian Bergstrand 21 May '09

21 May '09

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hi Terry, Thanks. On May 21, 2009, at 10:32 AM, Terry Lambert wrote: -- Terry Hi all, I'm seeing a weird error from pread in the following setup: The copy is done in DKIOCGETMAXBLOCKCOUNTREAD chunks (128KB for ATA). Mac OS X 10.5.7 (Darwin 9.7) on Intel. Thanks for any help?_______________________________________________ Brian Bergstrand <http://www.bergstrand.org/brian/> PGP Key ID: 0xB6C7B6A2 _______________________________________________ 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… The code is not multi-threaded. It's a fairly simple command line tool. I don't have a custom KEXT installed either. I do a bunch of preflight to get disk sizes, make sure the devices are not mounted, etc. I then open a read handle on the source device and a read-write handle on the target device. The tool then enters a loop and just preads from the source and pwrites to the target. All on one thread. The tool is linked against CoreFoundation and libSystem - nothing else. If I have a read handle open on the device and something causes diskarb to attempt to remount any volumes on the device could that cause the problem? I don't see how, because I can open a read handle on a disk containing mounted volumes or on the mounted volumes themselves (just not a write handle). There is nothing directly in the normal code path that does kauth checks after the open is successful. So it's your driver or your code. Is you code multithreaded? If so, does it serialize the close? On May 20, 2009, at 2:55 AM, Brian Bergstrand wrote: Copying from a 300GB PATA drive connected via USB to another 300GB ATA drive connected via Firewire using the raw char devices for each. After copying about 37GB of data, pread fails on the source USB drive with EACCES (13). According to the man page, this is not a valid error return (although the man page may be out of date). So where is the EACCES coming from? The IOKit block drivers? And what does it mean in this situation? This email sent to site_archiver(a)lists.apple.com

1 0

Re: pread and EACCES error?
by Terry Lambert 21 May '09

21 May '09

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com -- Terry Hi all, I'm seeing a weird error from pread in the following setup: The copy is done in DKIOCGETMAXBLOCKCOUNTREAD chunks (128KB for ATA). Mac OS X 10.5.7 (Darwin 9.7) on Intel. Thanks for any help?_______________________________________________ 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/tlambert%40apple.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: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… There is nothing directly in the normal code path that does kauth checks after the open is successful. So it's your driver or your code. Is you code multithreaded? If so, does it serialize the close? On May 20, 2009, at 2:55 AM, Brian Bergstrand wrote: Copying from a 300GB PATA drive connected via USB to another 300GB ATA drive connected via Firewire using the raw char devices for each. After copying about 37GB of data, pread fails on the source USB drive with EACCES (13). According to the man page, this is not a valid error return (although the man page may be out of date). So where is the EACCES coming from? The IOKit block drivers? And what does it mean in this situation? This email sent to tlambert(a)apple.com This email sent to site_archiver(a)lists.apple.com

1 0

Re: Powerdown notification
by Eric Gouriou 21 May '09

21 May '09

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On May 20, 2009, at 10:38 AM, Matthew Mashyna wrote: Eric Matt On Mar 18, 2009, at 6:03 PM, Ethan Bold wrote: -Ethan On Mar 18, 2009, at 1:14 PM, Brian Bergstrand wrote: On Mar 18, 2009, at 8:07 PM, Ethan Bold wrote: Hi Matt- What about the following constants: #define kIOMessageCanSystemPowerOff iokit_common_msg(0x240) #define kIOMessageSystemWillPowerOff iokit_common_msg(0x250) #define kIOMessageSystemWillNotPowerOff iokit_common_msg(0x260) They exist in the headers so one assumes they will be sent. Brian Bergstrand <http://www.bergstrand.org/brian/> PGP Key ID: 0xB6C7B6A2 _______________________________________________ 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/eric.gouriou%40pobox.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: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… My client asked me to take this issue up again and I found an acceptable solution. To recap... I have a Cocoa based daemon that monitors, uh, "stuff" on client computers. This daemon regularly checks in with a server to report the client's status. The problem I was having was that when the system shuts down, the daemon wasn't able to get the notification that a shutdown was imminent and report the shutdown to the server. I got a lot of suggestions about using IOKit and distributed notifications. None of that stuff worked when my daemon was launched from launchd. I started looking at catching the SIGTERM. I wasn't able to catch it on shutdown until... I changed the launch daemon plist to include this option: HopefullyExitsFirst When HopefullyExitsFirst is turned on, the shutdown waits for my daemon to exit and voila! I get the signal and have time to report the shutdown. At least I think it waits. I'm not completely sure but I get the result I am looking for. Is it acceptable to slowdown every shutdown ? How long does it take for the daemon to handle the SIGTERM and exit ? ---------------- HopefullyExitsFirst <boolean> This optional key causes programs to exit earlier during system shutdown. This key exists because some jobs do more than flush buffers and exit like they're supposed to. The use of this key should be considered a temporary solution until the software can be changed to only flush dirty buffers and then exit. ---------------- The use of that key is discouraged and considered a kludge. I know that Damien would love to get rid of it. The semantics (carefully left unspecified) may also change. "earlier" is pretty vague, a 10ms headstart on the SIGTERM could be what this will mean in a future release. The not HopefullyExitsFirst case may not work for this daemon due to dependencies on other daemons which are being shutdown concurrently. A solution would be to stop relying on those services, or at the very least to stop relying on those services to handle the shutdown case. Potential solutions: - there could be a simplistic BSD-only daemon which just invokes sigwait() and notifies the server upon reception of SIGTERM. - the server could understand that a severed connection implies a shutdown. - the daemon register a (WAN) Bonjour service and rely on mDNSResponder to handle the unregistration at shutdown for you. It's difficult to provide advice given the dearth of information about this daemon. In general the mention of a "Cocoa based [launch] daemon" makes me cringe. Did you mean a Foundation-based daemon ? At the very least I'd encourage you to file an enhancement request stating your needs and constraints. Though your assumption is reasonable, those constants don't work. IOKit doesn't send out messages for those. IOKit does not provide any notifications of system shutdown. The IORegisterForSystemPower() API call that you make below only refers to system sleep & wake power events, not shutdown events - the argument messageType may be kIOPMMessageCanSystemSleep, kIOPMMessageSystemWillSleep, kIOPMMessageSystemWillNotSleep, kIOPMMessageSystemHasPoweredOn, kIOPMMessageSystemWillPowerOn. That's all - no messages sent on shutdown. This email sent to eric.gouriou(a)pobox.com This email sent to site_archiver(a)lists.apple.com

1 0