May 2003 - Darwin-kernel - lists.apple.com

Re: Startup disk check
by Steven Bytnar 02 May '03

02 May '03

By the way, if your test disk has the developer tools installed, there are +-26807 files that you could nuke in sudo rm -rf /Developer/Documentation Every file node gets counted and checked, so the more files you have, the longer the disk check takes. I've got the dev tools installed on multiple disks with multiple OS versions, so getting rid of these directories sped up my disk checks. :) It would be nice if HeaderDoc generated .tar.gz'd documentation wads. A viewer application to view the wads would be great. Heck, a web-archive like thing that's viewable in Apple Help Viewer or Safari would be the next best thing! Even better, make a read-only disk image of /Developer/Documentation and make it overmount on /Developer/Documentation at each boot. There are drawbacks to taking this approach. Namely, upgrading the dev tools or installing CHUD or USB DDK's, which assume that they can write to /Developer/Documentation, wont be happy. I wonder... can you could journal a HFS+ disk image? Then, you wouldn't need to make the /Developer/Documentation image read-only! Next, journal the disk containing your journaled disk image and see what hilarity ensues. ;) --Steve On Wed, Apr 30, 2003 at 07:13:22PM -0700, Eric Long wrote: > > You can reduce the reboot delay if you turn on HFS+ journaling for > > all file systems. If you boot into MacOS 8 or 9, you should shutdown > > MacOS X cleanly before doing so. > > I'll take a look at that thought. Thanks. > > > Have you tried booting into single user mode and manually loading > > the necessary kernel extensions? > > Loading and unloading isn't the problem. The problem is that the test case > produces a panic. I have to reboot after that. I tried a lot of things to > try and isolate the trigger. What I found is that if I set the > kextcb->e_fcb value to anything other than null in my test kext's socreate() > hook, the third party extension will call panic(). I'm thinking, what > business is it of theirs what that value is... besides the issue of why a > kernel extension would ever call panic(). Seems to me if their kext is > confused it should just bow out gracefully in any case. > > Eric > _______________________________________________ > darwin-kernel mailing list | darwin-kernel(a)lists.apple.com > Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel > Do not post admin requests to the list. They will be ignored. _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: mach_msg_rpc_from_kernel
by Jim Magee 01 May '03

01 May '03

On Thursday, May 1, 2003, at 11:51 AM, Steven Bytnar wrote: On Wed, Apr 30, 2003 at 10:32:38PM -0400, Dave Koziol wrote: The other example I know of which is similar is KUNCUserNotificationDisplayAlert, although they are not actually trying to interact with the user, just do some things with aren't easy to do in the kernel. Which versions of OS X support the KUNC functions? Is this a 10.2 thing, or are they available all the way back to 10.0.0? They are available all the way back to 10.0.0 (although some bugs existed in the earlier versions). But their use (especially these GUI-centric routines) is HIGHLY discouraged. You should either user a helper application that contacts the driver directly, or use the IOKit registry to indicate conditions that are not part of the standard protocol for your family. Check the archive for references on how to do those things. Why? There are times during boot when KUNC will not work because the catcher for them isn't available yet. And worse, that timing is different for each release, so what works in 10.0 and 10.2 may not work in 10.1 just because of timing. Paths have to be hard-coded into the kext. Unattended systems, or systems where nobody has logged in yet, will just drop the notifications. Besides, we don't want people to have to write "screen-scrapers" just to handle these events programmatically. Notifications requiring responses may not always route the response back to your KEXT correctly (and you cannot safely unload the kext until the response DOES come back). --Jim _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: mach_msg_rpc_from_kernel
by Steven Bytnar 01 May '03

01 May '03

On Wed, Apr 30, 2003 at 10:32:38PM -0400, Dave Koziol wrote: > The other example I know of which is similar is > KUNCUserNotificationDisplayAlert, although they are not actually trying > to interact with the user, just do some things with aren't easy to do > in the kernel. Which versions of OS X support the KUNC functions? Is this a 10.2 thing, or are they available all the way back to 10.0.0? Thanks, --Steve _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: Startup disk check
by Peter Lovell 01 May '03

01 May '03

Hi Eric (again :), one thought is to use your (faster) desktop machine with a small external disk for this testing. Don't have your large (internal) disk mounted. Just have the minimal stuff needed to test. Loading and unloading isn't the problem. The problem is that the test case produces a panic. I have to reboot after that. I tried a lot of things to try and isolate the trigger. What I found is that if I set the kextcb->e_fcb value to anything other than null in my test kext's socreate() hook, the third party extension will call panic(). I'm thinking, what business is it of theirs what that value is... besides the issue of why a kernel extension would ever call panic(). Seems to me if their kext is confused it should just bow out gracefully in any case. That's rather tacky. Why on earth would a kext be walking the kextcb chain to see what other kexts have put in that field of their own kextcb? The other kext can't just be accidentally stumbling across yours, as it is given a kextcb of its very own. So it has to be starting back at the head of the list and walking down. What this is really saying is that the other kext will panic() if there's any kext other than itself (perhaps, before itself). Regards.....Peter _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: Startup disk check
by Brian Tabone 01 May '03

01 May '03

Eric, You can turn on Journaling on your startup disk, this will cause the system to bypass the fsck and allow you to have faster re-boot. I believe the command is (as root) diskutil enableJournal /dev/xxxx where xxxx is your device for the root drive. You may wish to check on google to be sure about that command, it's been a while since I turned it on. Your reboots will be much faster (After a crash) with journalling turned on, and your filesystem will be less likely to suffer catastrophic damage. -Hope that helps, Brian Tabone On Wednesday, April 30, 2003, at 06:47 PM, Eric Long wrote: I have been testing to isolate a kernel panic caused by third party kext, which only occurs when my test kext loads before it. The test kext does almost nothing, but I have been playing with it to see if there's anything I can do. If the third party kext isn't loaded, no panic occurs. The panic is happening because the third party kext is actually calling panic(). While screwing with this I've had to do lots of rebooting. I'm using a laptop that's a couple of years old. I don't want to install the third party software on my desktop Mac. Rebooting takes forever and a day on Jaguar because of the disk verification check. Is there anyway to skip this check? Is it too risky to bypass? Thanks, Eric _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored. _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: mach_msg_rpc_from_kernel
by Dave Koziol 01 May '03

01 May '03

On Wednesday, April 30, 2003, at 01:44 PM, Godfrey van der Linden wrote: Excuse me, what thread are you trying to block in the kernel? What sort of driver are you writing? In general we do not allow what you are trying to do. Some folks I'm trying to help are writing a CoreAudio Driver. (Note, I don't know anything about CoreAudio...) They would like to handle some processing for certain requests in user space. So, what they want to do is when they get one of these requests, make a request to a user process, block waiting for the response and when it comes back, return the results to the caller. I went looking around to see if there was anything that did anything like this, and I found mach_msg_rpc_from_kernel which looked very interesting... The other example I know of which is similar is KUNCUserNotificationDisplayAlert, although they are not actually trying to interact with the user, just do some things with aren't easy to do in the kernel. So, if there a preferred way to do something like this? At 13:02 -0400 03-4-30, Dave Koziol wrote: I've found this routine in ipc_mig.c, and it looks like it does exactly what I want to do for a particular project, but I haven't been able to find any code which actually uses it. Can anyone point out anywhere in Darwin where this is used? Is this a "safe" facility to use? I need to perform some RPC calls from a KEXT into a user process, and block waiting for the results. This routine appears to do exactly that... Thanks for any pointers in the right direction, Dave Koziol VP Engineering IOXperts, Inc _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: Startup disk check
by Eric Long 01 May '03

01 May '03

> You can reduce the reboot delay if you turn on HFS+ journaling for > all file systems. If you boot into MacOS 8 or 9, you should shutdown > MacOS X cleanly before doing so. I'll take a look at that thought. Thanks. > Have you tried booting into single user mode and manually loading > the necessary kernel extensions? Loading and unloading isn't the problem. The problem is that the test case produces a panic. I have to reboot after that. I tried a lot of things to try and isolate the trigger. What I found is that if I set the kextcb->e_fcb value to anything other than null in my test kext's socreate() hook, the third party extension will call panic(). I'm thinking, what business is it of theirs what that value is... besides the issue of why a kernel extension would ever call panic(). Seems to me if their kext is confused it should just bow out gracefully in any case. Eric _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: vfs sample source
by Brian Bergstrand 01 May '03

01 May '03

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Wednesday, April 30, 2003, at 04:43 PM, raldrich(a)mac.com wrote: > On Thursday, April 24, 2003, at 09:03 AM, Chris Rudolph wrote: > >> Hi all, >> >> I've heard that there is some sample code for writing a vfs for >> macosx. I know I could just look at some of the vfs bits in the >> darwin sources, but trying to weed through everything is quite >> daunting. Does anyone have any suggestions? >> >> Thanks, >> Rudy. >> > > Rudy, > > I've found that the cddafs source code (darwin/src/notlive/cddafs) > makes for good sample code for getting started. Its far better > structured and more well commented than any of the other file systems > (It feels like it was written by a 'classic' Mac engineer, rather than > a Unix guru). It doesn't support writing, but by the time you reach > that point, you'll have a pretty good understanding of what's needed > for that anyway. > > Ron Aldrich > Software Architects, Inc. > If you want to look at a third party VFS plug-in, check out Ext2 FS <http://sourceforge.net/projects/ext2fsx/>. Brian Bergstrand <http://www.classicalguitar.net/brian/> PGP Key ID: 0xB6C7B6A2 VI VI VI : Editor of the Beast -----BEGIN PGP SIGNATURE----- Version: PGP 8.0.2 iQA/AwUBPrBjcXnR2Fu2x7aiEQKB9ACg07xk7yqxPf8MXdCWtudugLkwCZMAn0RZ XySxeuHJmrphfA0bOSkHo5ee =5/Ew -----END PGP SIGNATURE----- _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: Startup disk check
by Steven Bytnar 01 May '03

01 May '03

Hi Eric, You can reduce the reboot delay if you turn on HFS+ journaling for all file systems. If you boot into MacOS 8 or 9, you should shutdown MacOS X cleanly before doing so. sudo diskutil enableJournal /dev/disk#s# You could also setup NetBoot for the test machine. The root filesystem will always be in a consistent state then. ;) Setting up NetBoot is non-trivial and requires OS X Server components. Have you tried booting into single user mode and manually loading the necessary kernel extensions? (Or do they load even on a single user boot?) You could unmount non-required volumes. Have you tried shift-booting too? If the kext's don't load, you could load them manually to possibly reproduce the problem. In these last two cases, you could remount all filesystems as read-only to avoid disk checks for those file systems. mount -o remount,ro / --Steve On Wed, Apr 30, 2003 at 04:47:39PM -0700, Eric Long wrote: > I have been testing to isolate a kernel panic caused by third party kext, > which only occurs when my test kext loads before it. The test kext does > almost nothing, but I have been playing with it to see if there's anything I > can do. If the third party kext isn't loaded, no panic occurs. The panic > is happening because the third party kext is actually calling panic(). > > While screwing with this I've had to do lots of rebooting. I'm using a > laptop that's a couple of years old. I don't want to install the third party > software on my desktop Mac. Rebooting takes forever and a day on Jaguar > because of the disk verification check. > > Is there anyway to skip this check? Is it too risky to bypass? > > > Thanks, > > Eric > _______________________________________________ > darwin-kernel mailing list | darwin-kernel(a)lists.apple.com > Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel > Do not post admin requests to the list. They will be ignored. _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0