December 2007 - Darwin-kernel

Re: missing driver (AppleMacRiscPCI) at startup on G5
by Al Ciplickas 19 Dec '07

19 Dec '07

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=AM479gxTatb8i8rlrV123DIq4s3FNbaFjd2F6bRM9nSlqoknH9QtcPQ63yb1Wj8g; h=Received:In-Reply-To:References:Mime-Version:Content-Type:Message-Id:Cc:Content-Transfer-Encoding:From:Subject:Date:To:X-Mailer:X-ELNK-Trace:X-Originating-IP; -Al- On Dec 18, 2007, at 2:38 PM, Terry Lambert wrote: You need to rebuild the KEXT cache, not the KEXT. man kextcache or http://developer.apple.com - Terry -Al- On Dec 18, 2007, at 1:25 PM, Terry Lambert wrote: Instead of removing the KEXT cache, rebuild it. _______________________________________________ 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.… Thanks again, Terry. I tried the rebuild and the complaint was that there was not a usr/../bootcache.plist (or something to that effect). Can't imagine how that could have happened, since I was not aware of that file. So I thought doing an 'upgrade' from an Install DVD would solve the problem. No way. The upgrade finished normally, but there was neither the bootcache.plist nor the kextentions.cache (sp?) files. This is confusing, as I expect a system upgrade to make all the necessary files to do a normal startup. I will locate an active bootcache.plist from an existing system partition and copy it to the usr/../ directory to see if it will help me do the kextcache rebuild. Do you think this is an anomaly for me or has this happened many times before and I should file a radar report? On Dec 18, 2007, at 2:09 PM, Al Ciplickas wrote: Thanks, Terry. I'll rebuild the KEXT. But forgot how at the moment. Any hints? On Dec 18, 2007, at 12:00 PM, Al Ciplickas wrote: This is a new one for me: G5 complains about 'com.apple.driver.AppleMacRiscPCI' missing in the startup from 'kld_load_from_memory()'. I got that using <cmd-V> at startup. I used another system partition to look it the file was there (in System/Library/Extensions/) and it was! So I took out the kextensions.cache and its partner and tried again: still got the same complaint. (By the way, it says there are 1310 buffer headers and 1310 cluster IO buffer header, if that matters.) Do anyone know of something else to try? This is under OSX 10.4.10 as well as 10.4.2 (which I reloaded in hopes of clearing up "lost" files). The error is indicating that it's unable to load it from the cache that was loaded into memory by the boot loader. It can't read it off the disk itself because it can't even talk to the PCI bus on which the disk controller lives without this KEXT loaded. -- Terry This email sent to site_archiver(a)lists.apple.com

1 0

Re: Darwin-kernel Digest, Vol 4, Issue 283
by Terry Lambert 18 Dec '07

18 Dec '07

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Given that subdirectories in /dev are not supported, and may be removed, if we end up needing to do that for some technical reason, it's probably better to just go that route on both 10.4 and 10.5 instead. This is probably the approach we will end up taking... just have to actually do it :) Oi! That's my code you'd be ripping out! ;o) Subdirectories do not remove the node name length limit. -- Terry _______________________________________________ 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.… On Dec 13, 2007, at 8:41 PM, Sam Vaughan wrote: On 14/12/2007, at 7:05 AM, Terry Lambert <tlambert(a)apple.com> wrote: The reason for the subdirectories in the first place was to allow LUNs to be identified by the port and node WWNs on the host and RAID controller that uniquely described their path on the fibre channel network. The WWNs are too long to include in the one devfs node name, so the only option is to use a hierarchy. This email sent to site_archiver(a)lists.apple.com

1 0

Re: missing driver (AppleMacRiscPCI) at startup on G5
by Terry Lambert 18 Dec '07

18 Dec '07

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com You need to rebuild the KEXT cache, not the KEXT. man kextcache or http://developer.apple.com - Terry -Al- On Dec 18, 2007, at 1:25 PM, Terry Lambert wrote: Instead of removing the KEXT cache, rebuild it. _______________________________________________ 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.… On Dec 18, 2007, at 2:09 PM, Al Ciplickas wrote: Thanks, Terry. I'll rebuild the KEXT. But forgot how at the moment. Any hints? On Dec 18, 2007, at 12:00 PM, Al Ciplickas wrote: This is a new one for me: G5 complains about 'com.apple.driver.AppleMacRiscPCI' missing in the startup from 'kld_load_from_memory()'. I got that using <cmd-V> at startup. I used another system partition to look it the file was there (in System/Library/Extensions/) and it was! So I took out the kextensions.cache and its partner and tried again: still got the same complaint. (By the way, it says there are 1310 buffer headers and 1310 cluster IO buffer header, if that matters.) Do anyone know of something else to try? This is under OSX 10.4.10 as well as 10.4.2 (which I reloaded in hopes of clearing up "lost" files). The error is indicating that it's unable to load it from the cache that was loaded into memory by the boot loader. It can't read it off the disk itself because it can't even talk to the PCI bus on which the disk controller lives without this KEXT loaded. -- Terry This email sent to site_archiver(a)lists.apple.com

1 0

Re: missing driver (AppleMacRiscPCI) at startup on G5
by Al Ciplickas 18 Dec '07

18 Dec '07

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=QmqmBCjZ6grFarFBsyn9QXE67hkJzwXdVZSrytRPgmckryS4mZGeNGxvYfQCt6BU; h=Received:In-Reply-To:References:Mime-Version:Content-Type:Message-Id:Cc:Content-Transfer-Encoding:From:Subject:Date:To:X-Mailer:X-ELNK-Trace:X-Originating-IP; -Al- On Dec 18, 2007, at 1:25 PM, Terry Lambert wrote: Instead of removing the KEXT cache, rebuild it. _______________________________________________ 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.… Thanks, Terry. I'll rebuild the KEXT. But forgot how at the moment. Any hints? On Dec 18, 2007, at 12:00 PM, Al Ciplickas wrote: This is a new one for me: G5 complains about 'com.apple.driver.AppleMacRiscPCI' missing in the startup from 'kld_load_from_memory()'. I got that using <cmd-V> at startup. I used another system partition to look it the file was there (in System/Library/Extensions/) and it was! So I took out the kextensions.cache and its partner and tried again: still got the same complaint. (By the way, it says there are 1310 buffer headers and 1310 cluster IO buffer header, if that matters.) Do anyone know of something else to try? This is under OSX 10.4.10 as well as 10.4.2 (which I reloaded in hopes of clearing up "lost" files). The error is indicating that it's unable to load it from the cache that was loaded into memory by the boot loader. It can't read it off the disk itself because it can't even talk to the PCI bus on which the disk controller lives without this KEXT loaded. -- Terry This email sent to site_archiver(a)lists.apple.com

1 0

Re: missing driver (AppleMacRiscPCI) at startup on G5
by Terry Lambert 18 Dec '07

18 Dec '07

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Instead of removing the KEXT cache, rebuild it. -- Terry _______________________________________________ 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.… On Dec 18, 2007, at 12:00 PM, Al Ciplickas wrote: This is a new one for me: G5 complains about 'com.apple.driver.AppleMacRiscPCI' missing in the startup from 'kld_load_from_memory()'. I got that using <cmd-V> at startup. I used another system partition to look it the file was there (in System/Library/Extensions/) and it was! So I took out the kextensions.cache and its partner and tried again: still got the same complaint. (By the way, it says there are 1310 buffer headers and 1310 cluster IO buffer header, if that matters.) Do anyone know of something else to try? This is under OSX 10.4.10 as well as 10.4.2 (which I reloaded in hopes of clearing up "lost" files). The error is indicating that it's unable to load it from the cache that was loaded into memory by the boot loader. It can't read it off the disk itself because it can't even talk to the PCI bus on which the disk controller lives without this KEXT loaded. This email sent to site_archiver(a)lists.apple.com

1 0

missing driver (AppleMacRiscPCI) at startup on G5
by Al Ciplickas 18 Dec '07

18 Dec '07

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=jbbYdYTnNwKRI6IwCUqeI6+bj8PomRP1HCuut2VSWLAtCo7BPNmVIqVSjtq0CcQu; h=Received:Mime-Version:Content-Transfer-Encoding:Message-Id:Content-Type:To:From:Subject:Date:X-Mailer:X-ELNK-Trace:X-Originating-IP; Thanks. -Al- _______________________________________________ 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 is a new one for me: G5 complains about 'com.apple.driver.AppleMacRiscPCI' missing in the startup from 'kld_load_from_memory()'. I got that using <cmd-V> at startup. I used another system partition to look it the file was there (in System/Library/Extensions/) and it was! So I took out the kextensions.cache and its partner and tried again: still got the same complaint. (By the way, it says there are 1310 buffer headers and 1310 cluster IO buffer header, if that matters.) Do anyone know of something else to try? This is under OSX 10.4.10 as well as 10.4.2 (which I reloaded in hopes of clearing up "lost" files). This email sent to site_archiver(a)lists.apple.com

1 0

Re: Darwin-kernel Digest, Vol 4, Issue 283
by Sam Vaughan 14 Dec '07

14 Dec '07

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On 14/12/2007, at 7:05 AM, Terry Lambert <tlambert(a)apple.com> wrote: Or you can choose to not try creating subdirectores in /dev, and put the information into node names themselves instead. For example, a SCSI disc device on controller 0 with unit 3 and lun 8 with three partitions on it, where you wanted the second one could be named something like "sdc0u3l8s3". UNIX systems have used this type of convention forever. Think of the letters other than the initial "sd" as path component separators, if it helps you sleep. Given that subdirectories in /dev are not supported, and may be removed, if we end up needing to do that for some technical reason, it's probably better to just go that route on both 10.4 and 10.5 instead. This is probably the approach we will end up taking... just have to actually do it :) Oi! That's my code you'd be ripping out! ;o) Sam _______________________________________________ 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 The reason for the subdirectories in the first place was to allow LUNs to be identified by the port and node WWNs on the host and RAID controller that uniquely described their path on the fibre channel network. The WWNs are too long to include in the one devfs node name, so the only option is to use a hierarchy. If you ask me, even if the required information could be squashed into one name, it's a hell of a lot neater and less cryptic to break it up into a hierarchy. Why punish yourself with things like "sdc0u3l8s3"? Not to mention the sorting problems once any of those numbers goes double-digits... smime.p7s

1 0

Re: IOCommandGate commandSleep/commandWakeup
by Michael Smith 14 Dec '07

14 Dec '07

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On Dec 13, 2007, at 11:23 AM, Eric Long wrote: You probably want THREAD_UNINT. = 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.… kr = driver->myCommandGate->commandSleep(&(driver- >isSleeping), THREAD_ABORTSAFE); THREAD_ABORTSAFE means that it is safe to abort this thread while it is asleep, in which case it will never return. If you are certain that you can tear down your operation while it is in progress, you might consider THREAD_INTERRUPTIBLE, but you will have to handle the THREAD_INTERRUPTED return case as well, and the fact that you may be racing with whatever is responsible for waking you up "normally". This email sent to site_archiver(a)lists.apple.com

1 0

IOCommandGate commandSleep/commandWakeup
by Eric Long 13 Dec '07

13 Dec '07

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thread-index: Acg9vZ4P3JwFbKmwEdyV5gAX8si2KA== Thread-topic: IOCommandGate commandSleep/commandWakeup User-agent: Microsoft-Entourage/11.3.6.070618 Hi, I'm encountering a problem where calling commandWakeup() on an IOCommandGate is failing to wake up the sleeping thread. This describes how I though things were supposed to work: Some code calls: myCommandGate->runCommand(NULL, NULL, NULL, NULL); The command gate runs its action: IOReturn com_myCompany_driver::waitForMonkey( OSObject *owner, void*, void*, void*, void* ) { com_myCompany_driver *driver; IOReturn kr = kIOReturnSuccess; driver = OSDynamicCast( com_myCompany_driver, owner); // Make sure that this driver type is the sender of the command. SIGNAL_IF(driver == NULL); if (driver) { driver->tellMonkeyToFling(); // calls messageClients driver->isSleeping = true; kr = driver->myCommandGate->commandSleep(&(driver->isSleeping), THREAD_ABORTSAFE); driver->isSleeping = false; if (kr != THREAD_AWAKENED) { DEBUG_LOG("commandSleep() returned error %08x!", kr); } else { DEBUG_LOG("Driver %p resuming thread after sleep", driver); } } } Then when the monkey responds, wake the thread up: void com_myCompany_driver::handleMonkeyResponse( bool inResponse ) { monkeyResponse = inResponse; myCommandGate->commandWakeup(&isSleeping, false); DEBUG_LOG("Called commandWakeup for driver %p", this); } In the system log, I see the log messages I expect. It looks like everything happened that should happen, except in some instances, the sleeping thread never wakes up (and thus I never see the "Driver x resuming thread after sleep" message, but I do see the "Called commandWakeup for driver x" message. It's like commandWakeup did nothing. Eric _______________________________________________ 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: [darwin-kernel] devfs_make_node mutilates pathname on Leopard?
by Nick Blievers 13 Dec '07

13 Dec '07

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com User-agent: Thunderbird 2.0.0.0 (Macintosh/20070326) Nick _______________________________________________ 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.… Terry Lambert wrote: On Dec 12, 2007, at 5:08 PM, Nick Blievers <nickb(a)sgi.com> wrote: Specifically, creation of subdirectories in /dev is not officially supported, don't do that. The change of function precludes kernel buffer overflow based crashes for some device drivers using name information supplied from user space. Even with the "off by one" introduced in the changeover, the buffet overrun is avoided, which is a good thing for kernel stability. In general, device node length is constrained overall to a value "reasonable" for a single path component, which means the more /'s you used, the shorter the name allowed for the eventual device anyway. I think the overall limit is something like 32 characters; if that's not exactly correct, whatever it is is a heck of a lot smaller than PATH_MAX at 1024 bytes. -- Terry >From what I remember the limit discussed was 32, however looking at the code it appears the limit is DEVMAXPATHSIZE (128) for the full path and DEVMAXNAMESIZE (32) for each element. Also I believe this limit existed pre-Leopard, just wasn't enforced, and hence buffer-overrun's where possible. So currently our solution is to test the kernel version and add a bogus character to the path element in out devfs_make_node() call... this is the only way to do it without using functions that are not exported. Or you can choose to not try creating subdirectores in /dev, and put the information into node names themselves instead. For example, a SCSI disc device on controller 0 with unit 3 and lun 8 with three partitions on it, where you wanted the second one could be named something like "sdc0u3l8s3". UNIX systems have used this type of convention forever. Think of the letters other than the initial "sd" as path component separators, if it helps you sleep. Given that subdirectories in /dev are not supported, and may be removed, if we end up needing to do that for some technical reason, it's probably better to just go that route on both 10.4 and 10.5 instead. This is probably the approach we will end up taking... just have to actually do it :) This email sent to site_archiver(a)lists.apple.com

1 0