[OpenAFS] Reproducible kernel panic on MacOS x 10.3.9 with 1.3.81
Samuel L. Bayer
Tue, 23 Aug 2005 12:32:51 -0400
Hi all -
I've submitted this to the bug tracker (#21046), but I thought I'd run
it past you all to see if anyone has any ideas.
We have OpenAFS installed on a number of machines, all of them dual
processor G5s. My machine seem to be the heaviest user of our AFS space
among those machines, and it reliable generates a kernel panic on
shutdown. This would merely be an annoyance, except the AFS performance
tends to get flakier and flakier the longer the machine is up: strange
file contents, etc. My standard solution is to boot the machine, but if
I happen not to be on site, and I get a kernel panic on shutdown,
somebody has to intervene physically in order for the machine to reboot
properly, to the best of my knowledge.
The kernel panic is clearly in the afs extension, according to the PC
counter in the panic log and the kextstat command. The kernel panic I get is
panic(cpu 0): unmount: dangling vnode
Latest stack backtrace for cpu 0:
0x00083498 0x0008397C 0x0001EDA4 0x000C60B4 0x000C39DC
0x00218030 0x00220BA4 0x00246D84
Proceeding back via exception chain:
Exception state (sv=0x2D566280)
PC=0x9005F5CC; MSR=0x0000D030; DAR=0x0030E300; DSISR=0x40000000;
LR=0x000027A4; R1=0xBFFFFEB0; XCP=0x00000030 (0xC00 - System call)
Darwin Kernel Version 7.9.0:
Wed Mar 30 20:11:17 PST 2005; root:xnu/xnu-517.12.7.obj~1/RELEASE_PPC
Sometimes it's the other CPU, but it's always the same first several
items of the backtrace, PC, etc.
If I haven't done anything in AFS space, I don't get the kernel panic;
but it doesn't seem like I have to do much to trigger it.
None of the other machines seem to be having this problem, but I can't
exactly tell whether it's because of their relatively light AFS use or
because of possibly conflicting kernel extensions.
Any ideas? I've looked at ticket 18358, and it looks like it might be
related, but I can't really tell.
Thanks for any advice you can provide.