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

Re: Kauth
by Terry Lambert 16 May '06

16 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On May 16, 2006, at 1:12 AM, Jérome Foucher wrote: Hi folks ! I've started to use kauth for a future application. This is on purpose for this sample code. -- 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.… I've looked at KauthORama sample code, and have a few questions for which I didn't manage to find any answers in the archives. First of all, the KauthORama's kext does not load automatically at startup. Is it done on purpose, or is it a bug ? There's a lot of documentation on KEXT loading in this lists' archives, actually; it has to do with when and why KEXTs are loaded, what keeps them unloaded, whether an application should be used to load and unload them, etc.. The answer depends on your particular use and implementation: the problem you are trying to solve. In general, there's not a lot of documentation on this particular topic that has been sorted and sifted into one place, though. I suspect it's because if you force a KEXT to load each time the system boots, no matter what, then you can end up with a KEXT with a bug that can turn your system into a doorstop until you delete the KEXT or reinstall the system. If you have particular needs, you should look over this lists' archives carefully. More annoying, if I do not call "kextunload KauthORama" and shutdown my Mac, I get a kernel panic (each time, and this happened on all the machines I've tested). I checked the code a few times and could not find a bug in it that coulw cause a kernel panic. Please file a bug via Bugreporter <http://developer.apple.com/bugreporter/ >. Include the stack traceback and other information from having set the debug flags, so that it's possible to tell where things are crashing. Als indicate if the code is the stock Apple sample code, or if you have modified it. If all of this is "normal", does this mean my user-land app has to load the kext itself and make sure it's unloaded before the computer shuts down ? In general, this should not be necessary, so long as you maintain appropriate reference/use counts so that a KEXT does not get unloaded while it's in use. This is covered pretty well in the KEXT programming information on <http://developer.apple.com/>. This email sent to site_archiver(a)lists.apple.com

1 0

Kauth
by Jérome Foucher 16 May '06

16 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi folks ! I've started to use kauth for a future application. Thanks Jérome _______________________________________________ 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 looked at KauthORama sample code, and have a few questions for which I didn't manage to find any answers in the archives. First of all, the KauthORama's kext does not load automatically at startup. Is it done on purpose, or is it a bug ? More annoying, if I do not call "kextunload KauthORama" and shutdown my Mac, I get a kernel panic (each time, and this happened on all the machines I've tested). I checked the code a few times and could not find a bug in it that coulw cause a kernel panic. If all of this is "normal", does this mean my user-land app has to load the kext itself and make sure it's unloaded before the computer shuts down ? This email sent to site_archiver(a)lists.apple.com

1 0

Re: pgrp kpi ?
by Terry Lambert 15 May '06

15 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com -- Terry -- Chris 15 maj 2006 kl. 21.55 skrev Terry Lambert: On May 15, 2006, at 12:41 PM, Christer Bernérus wrote: Hi. In what context? _______________________________________________ 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 somewhat problematic, because you mentioned the cred structure. I assume that the reason it wants this is that it's squirreling away data in the cr_groups list, like AFS does? In general, this information is technically dereferenceable out of the proc structure itself, which has a p->p_pgrp pointer. In practice, it would probably be better, at least until a later release, to pass the PID itself back, and then use the KERN_PROC sysctl from user space to obtain the information on the basis of the PIS of the requesting process. On May 15, 2006, at 2:19 PM, Christer Bernérus wrote: This is within the coda file system kernel extension when the kernel code expects to build a callback message to the userland cache manager. That message is expected to contain an opcode, an unique number, a process ID, a process group id, a session ID and a creds struct. I don't do any userland coding, so I'd like to stick to the already defined protocol. The functions that wants the pgid are typical file system open and close routines. In a kext, it seems like the proc structure is opaque these days. How do I get hold of the pgrp given a proc pointer ? If you are writing a tty driver (for example), it's passed in as part of your context due to the tty line discipline. If you need it for some other reason, can you tell us the reason and the context in which it is needed / will be used? -- Terry This email sent to site_archiver(a)lists.apple.com

1 0

Re: pgrp kpi ?
by Christer Bernérus 15 May '06

15 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com -- Chris 15 maj 2006 kl. 21.55 skrev Terry Lambert: On May 15, 2006, at 12:41 PM, Christer Bernérus wrote: Hi. In what context? _______________________________________________ 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 This is within the coda file system kernel extension when the kernel code expects to build a callback message to the userland cache manager. That message is expected to contain an opcode, an unique number, a process ID, a process group id, a session ID and a creds struct. I don't do any userland coding, so I'd like to stick to the already defined protocol. The functions that wants the pgid are typical file system open and close routines. In a kext, it seems like the proc structure is opaque these days. How do I get hold of the pgrp given a proc pointer ? If you are writing a tty driver (for example), it's passed in as part of your context due to the tty line discipline. If you need it for some other reason, can you tell us the reason and the context in which it is needed / will be used? -- Terry smime.p7s

1 0

Re: pgrp kpi ?
by Terry Lambert 15 May '06

15 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On May 15, 2006, at 12:41 PM, Christer Bernérus wrote: Hi. In what context? -- 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.… In a kext, it seems like the proc structure is opaque these days. How do I get hold of the pgrp given a proc pointer ? If you are writing a tty driver (for example), it's passed in as part of your context due to the tty line discipline. If you need it for some other reason, can you tell us the reason and the context in which it is needed / will be used? This email sent to site_archiver(a)lists.apple.com

1 0

pgrp kpi ?
by Christer Bernérus 15 May '06

15 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi. -- Chris _______________________________________________ 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 In a kext, it seems like the proc structure is opaque these days. How do I get hold of the pgrp given a proc pointer ? smime.p7s

1 0

Re: kauth_getuid returns 0?
by Terry Lambert 13 May '06

13 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com To elaborate on what Brian said. -- Terry On May 11, 2006, at 8:30 PM, Brian Bergstrand wrote: -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On May 11, 2006, at 9:28 PM, Russell Seehafer (RIT Student) wrote: Hi, Thanks! Russ HTH. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.3 (Darwin) iD8DBQFEZAFTedHYW7bHtqIRAv2XAJ9Gny13NuUFYFNx0BSHsLXXC8NshQCg1mrF /VozuSFlTTBia/GT0p26xis= =4446 -----END PGP SIGNATURE----- _______________________________________________ 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.… kauth_getuid() returns the UID of the currently running thread, if there is one. Otherwise it returns the uid of the currently running process. In practice, there are a number of cases where operations are enqueued into a work queue for later operation. In general, if the credential is important, then a reference is enqueued as part of the metadata on the operation being performed. This is then used as the credential by the proxy operation. The credential you should use in your code depends on whether the operation is occurring in a proxy context, or it's occurring in the real context - in other words, if you are being called into from a worker thread, it's likely you are being passed a credential as one of your arguments, or as part of the metadata for the work item in scope at the time of the call, and there will either be an accessor function for this, or (worst case) the structure internals will be visible and you can dereference to get it yourself. In other words, blindly calling kauth_getuid() is not necessarily the right thing to do in all cases. For an NKE, in general, work will come in on a worker thread when being pushed up the TCP stack, and the credential you will want to use for the operation is based on the process associated with the socket to which the data is destined, not the currently running thread. Most likely you are either not using this information, or you don't have the information in scope because you are trying to perform your operation at the wrong layer. If you need more help, some of the networking guys read this list, but you will likely need to tell them more about what you are doing and when (what's getting hooked, what you are trying to accomplish, etc.) for them to help you. If this information is sensitive, and not suitable for a public list, you should probably work through bugreporter or DTS. I have a nke that keeps track of its associated user space clients based on the value returned from kauth_getuid() I have run into some instances where this call returns 0. I cannot find any documentation on this method does anyone have any idea what could be causing this call to return 0? My understanding is that if a packet is being transmitted by safari, for example, that kauth_getuid() would return the user's id of the user that is running safari. Is there more I should consider when using this method? 0 is a valid UID. It is the UID of the root user. 0 could also be returned because there is no current task at the point you call kauth_getuid() so you are getting the root user by default (ie the kernel task). This email sent to site_archiver(a)lists.apple.com

1 0

Re: kauth_getuid returns 0?
by Brian Bergstrand 12 May '06

12 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On May 11, 2006, at 9:28 PM, Russell Seehafer (RIT Student) wrote: Hi, Thanks! Russ HTH. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.3 (Darwin) iD8DBQFEZAFTedHYW7bHtqIRAv2XAJ9Gny13NuUFYFNx0BSHsLXXC8NshQCg1mrF /VozuSFlTTBia/GT0p26xis= =4446 -----END PGP SIGNATURE----- _______________________________________________ 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 have a nke that keeps track of its associated user space clients based on the value returned from kauth_getuid() I have run into some instances where this call returns 0. I cannot find any documentation on this method does anyone have any idea what could be causing this call to return 0? My understanding is that if a packet is being transmitted by safari, for example, that kauth_getuid() would return the user's id of the user that is running safari. Is there more I should consider when using this method? 0 is a valid UID. It is the UID of the root user. 0 could also be returned because there is no current task at the point you call kauth_getuid() so you are getting the root user by default (ie the kernel task). This email sent to site_archiver(a)lists.apple.com

1 0

kauth_getuid returns 0?
by Russell Seehafer (RIT Student) 12 May '06

12 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thread-index: AcZ1a9NAyvPBO1bNT+mhUy8nZYySKg== Thread-topic: kauth_getuid returns 0? _______________________________________________ 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 Brian Bergstrand 09 May '06

09 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On May 9, 2006, at 11:21 AM, Jeremy Pereira wrote: On May 8, 2006, at 23:01, Terry Lambert wrote: diskutil unmount force disk? or : umount -f /dev/disk? -- Terry -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.3 (Darwin) iD8DBQFEYNvQedHYW7bHtqIRAoO+AJ4nOdACh8RXTcLTccxTPFYBK5gjCQCdEvPg //jZaSj7k+XwU0SUywpdovo= =SryW -----END PGP SIGNATURE----- _______________________________________________ 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. I don't think Finder/CarbonDiskManager ever do a force unmount. You have use diskutil to do so 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