May 2006 - Darwin-kernel - lists.apple.com

Re: Eject problem with optical disk
by Jeremy Pereira 09 May '06

09 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On May 8, 2006, at 23:01, Terry Lambert wrote: -- 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.… The best thing for this is two machine debugging and the command "showallstacks". This will tell you exactly who has things open, and why. I've tried that, but it's more or less impossible to capture the machine state in the window during which the unmount syscall is occurring. But it's most likely that you just need to support the forced unmount option in your VFS's unmount vfsop. That would be a nice idea, but I know for a fact that the unmount operation does not get as far as the vfsop. I inserted a call to panic() as the first line of the vfsop and there was no panic. At least not until I rebooted the machine where the panic did occur at the expected line either because the force flag is set when the machine is being shut down or because none of the other readdir activities are happening. Basically, this will cause the vnodes to be ripped out from under whoever is trying to use them, and cause further operations to be invalid. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Deciphering error
by Eric Long 08 May '06

08 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com User-agent: Microsoft-Entourage/11.1.0.040913 > void > myCommandHandler( OSObject *owner, void *arg0, void *arg1, void *arg2, void > *arg3 ); > Oops. There's the problem. It's supposed to return IOReturn. That's the cause. So... Never mind. 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

Deciphering error
by Eric Long 08 May '06

08 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com User-agent: Microsoft-Entourage/11.1.0.040913 Hi, Does this error I'm getting from an IOCommandGate call makes sense to anyone: IOReturn result = gate->runCommand(msgPtr, &msgSize, (void *)0, (void *)0)) result: 30593024: system: 0x0, sub: 0x74b, code: 0x1000 What's really odd is that nothing seemed to go wrong. The command went through and every else appeared to operation cleanly. I don't have any clues why result isn't coming back 0. My hunch is that the error code is bogus, like maybe the stack is off. My command handler looks something like this: void myCommandHandler( OSObject *owner, void *arg0, void *arg1, void *arg2, void *arg3 ); I use the same handler in other places without seeing this error, but in the other instances I pass 0 for all args. The function itself is pretty simple. It calls enqueue() to add data to a queue. If the args are 0 it pulls data from another queue, else it queues the data from the args. The IODataQueue enqueue() call in my command handler returns true. So I don't see an error happening there. 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: Eject problem with optical disk
by Terry Lambert 08 May '06

08 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On May 7, 2006, at 10:02 PM, Jeremy Pereira wrote: -- 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.… Apologies if this is the wrong list, I'm not sure exactly where the problem lies, so I thought I would start here. I have written a VFS driver for a proprietary optical disk format and it is mostly working fine. However, when I attempt to eject the disk from the Finder, if there are more than about 500 files in the trash, the disk fails to eject. I've turned on debugging in diskarbitrationd and it appears that several programs attempt to enumerate the contents of the .Trashes/ 501 folder (501 is the uid of my user) by calling the readdir vnode operation. diskarbitrationd is making the unmount system call, but I think EBUSY is being returned before it gets to the VFS unmount call in my code (I put a call to panic() in as the very first line of the function and no panic happened). Having checked the kernel code, I think one of the readdirs is overrunning some time-out and is still holding a reference to the .Trashes/501 vnode when the unmount checks if any are still in use. Am I on the right track with the above analysis? Am I doing something wrong and can anybody suggest some areas I might look in to resolve this? OS is OS X 10.4.6 The best thing for this is two machine debugging and the command "showallstacks". This will tell you exactly who has things open, and why. But it's most likely that you just need to support the forced unmount option in your VFS's unmount vfsop. Basically, this will cause the vnodes to be ripped out from under whoever is trying to use them, and cause further operations to be invalid. This email sent to site_archiver(a)lists.apple.com

1 0

Re: dumb xnu-792 build question
by Terry Lambert 08 May '06

08 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On May 8, 2006, at 4:27 AM, Peter Lovell wrote: Ok, I feel dumb asking this, but I can't seem to find it, so... That's the correct file, if you want to let things build with warnings. bsd/vfs/vfs_subr.c and bsd/sys/vnode.h Which is what I'd recommend. -- 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 May 7, 2006, at 4:09 PM, rick(a)snowhite.cis.uoguelph.ca wrote: I've added a bunch of dirs to xnu-792 and the make finds them and builds them ok. (added to bsd/conf/files) However, as soon as I get a warning on any of them, the build stops. I can see that other parts of the system build with warnings and keep going. So, what's the trick to get the build environment to not stop on warnings? (I'm building on Darwin8.0.1 for i386, if that makes any difference.) I have succeeded in getting rid of the warnings, but had to add crud like the following to avoid the warnings, which isn't exactly pretty:-) /* XXX really dumb, but the compiler complains since the last arg * of vnode_waitforwrites() isn't "const char *". */ bcopy("nfsflush", stupid, 9); while ((error = vnode_waitforwrites(vp, 0, slpflag, slptimeo, stupid))) { Thanks in advance for any help, rick Hi Rick, have a look at xnu/bsd/conf/Makefile.i386 for the list of files for OBJS_NO_WERROR If you add your files to the list then the build will warn but not stop. I know it looks weird but this file with "i386" suffix is used for ppc builds. He's actually doing an i386 build... in 792, this is only the case for the i386 code anyway (which makes warnings errors for everything except the listed files). Alternately, you can change the last argument to vnode_waitforwrites() to a const char * by modifying In general, if you are getting warnings, then it's best, particularly for new files which haven't been vetted, for them to be treated as errors. This email sent to site_archiver(a)lists.apple.com

1 0

Re: dumb xnu-792 build question
by Peter Lovell 08 May '06

08 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On May 7, 2006, at 4:09 PM, rick(a)snowhite.cis.uoguelph.ca wrote: Ok, I feel dumb asking this, but I can't seem to find it, so... If you add your files to the list then the build will warn but not stop. Regards.....Peter _______________________________________________ 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.… I've added a bunch of dirs to xnu-792 and the make finds them and builds them ok. (added to bsd/conf/files) However, as soon as I get a warning on any of them, the build stops. I can see that other parts of the system build with warnings and keep going. So, what's the trick to get the build environment to not stop on warnings? (I'm building on Darwin8.0.1 for i386, if that makes any difference.) I have succeeded in getting rid of the warnings, but had to add crud like the following to avoid the warnings, which isn't exactly pretty:-) /* XXX really dumb, but the compiler complains since the last arg * of vnode_waitforwrites() isn't "const char *". */ bcopy("nfsflush", stupid, 9); while ((error = vnode_waitforwrites(vp, 0, slpflag, slptimeo, stupid))) { Thanks in advance for any help, rick Hi Rick, have a look at xnu/bsd/conf/Makefile.i386 for the list of files for OBJS_NO_WERROR I know it looks weird but this file with "i386" suffix is used for ppc builds. This email sent to site_archiver(a)lists.apple.com

1 0

Re: loading kext from a daemon program
by Quinn 08 May '06

08 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com 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 2:51 -0700 6/5/06, Michael Smith wrote: ... don't get too cocky about "absolute paths". / is relative (think chroot(2)). ... don't forget to quote your paths. That, in itself, is enough to make me run away screaming from "system" and "popen". This email sent to site_archiver(a)lists.apple.com

1 0

Eject problem with optical disk
by Jeremy Pereira 08 May '06

08 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com OS is OS X 10.4.6 _______________________________________________ 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.… Apologies if this is the wrong list, I'm not sure exactly where the problem lies, so I thought I would start here. I have written a VFS driver for a proprietary optical disk format and it is mostly working fine. However, when I attempt to eject the disk from the Finder, if there are more than about 500 files in the trash, the disk fails to eject. I've turned on debugging in diskarbitrationd and it appears that several programs attempt to enumerate the contents of the .Trashes/ 501 folder (501 is the uid of my user) by calling the readdir vnode operation. diskarbitrationd is making the unmount system call, but I think EBUSY is being returned before it gets to the VFS unmount call in my code (I put a call to panic() in as the very first line of the function and no panic happened). Having checked the kernel code, I think one of the readdirs is overrunning some time-out and is still holding a reference to the .Trashes/501 vnode when the unmount checks if any are still in use. Am I on the right track with the above analysis? Am I doing something wrong and can anybody suggest some areas I might look in to resolve this? This email sent to site_archiver(a)lists.apple.com

1 0

Re: dumb xnu-792 build question
by Herb Petschauer 07 May '06

07 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=CE/AUoFTohixj7NqeieVrdwfbwf5BR94A7uKc+TvTVGSW0Z94xZPDTho3vjWfrfTTPqGeeLAwOZw+IezDDuMTUUl15ViXKm1C9MOuXIlVAx7/VSD/rYKvEx9ASQVO+1G2uo/BmwmBSk3oVs4qGeBeDLAh17PFNscCEszvh83v3k= This is going to be vague because it's been a few months since I've built a kernel. Somewhere in the config files of the build system there is a file that indicates which files are to be built with the "treat warnings as errors" directive. I can't recall what that directive is exactly (no dev tools on this laptop) but simple grep on "error" may be enlightening. Anyway, it's Sunday so I hope some of that makes sense :-) -H. On 07/05/06, rick(a)snowhite.cis.uoguelph.ca <rick(a)snowhite.cis.uoguelph.ca> wrote: Ok, I feel dumb asking this, but I can't seem to find it, so... I've added a bunch of dirs to xnu-792 and the make finds them and builds them ok. (added to bsd/conf/files) However, as soon as I get a warning on any of them, the build stops. I can see that other parts of the system build with warnings and keep going. So, what's the trick to get the build environment to not stop on warnings? (I'm building on Darwin8.0.1 for i386, if that makes any difference.) I have succeeded in getting rid of the warnings, but had to add crud like the following to avoid the warnings, which isn't exactly pretty:-) /* XXX really dumb, but the compiler complains since the last arg * of vnode_waitforwrites() isn't "const char *". */ bcopy("nfsflush", stupid, 9); while ((error = vnode_waitforwrites(vp, 0, slpflag, slptimeo, stupid))) { Thanks in advance for any help, rick _______________________________________________ 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/bogvardi%40gmail.com This email sent to bogvardi(a)gmail.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.… Perhaps in your case the default config is "treat warnings as errors"? Check to see what options your added files are being compiled with. Look for the word "error" in the compile directive itself. Then you may have to do some reverse engineering to figure out the build system... This email sent to site_archiver(a)lists.apple.com

1 0

dumb xnu-792 build question
by rick＠snowhite.cis.uoguelph.ca 07 May '06

07 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Ok, I feel dumb asking this, but I can't seem to find it, so... I've added a bunch of dirs to xnu-792 and the make finds them and builds them ok. (added to bsd/conf/files) However, as soon as I get a warning on any of them, the build stops. I can see that other parts of the system build with warnings and keep going. So, what's the trick to get the build environment to not stop on warnings? (I'm building on Darwin8.0.1 for i386, if that makes any difference.) I have succeeded in getting rid of the warnings, but had to add crud like the following to avoid the warnings, which isn't exactly pretty:-) /* XXX really dumb, but the compiler complains since the last arg * of vnode_waitforwrites() isn't "const char *". */ bcopy("nfsflush", stupid, 9); while ((error = vnode_waitforwrites(vp, 0, slpflag, slptimeo, stupid))) { Thanks in advance for any help, rick _______________________________________________ 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