Re: stability problems

[haad <haaaad%gmail.com@localhost>]

On Wed, Mar 16, 2011 at 2:34 PM, Thomas Klausner <wiz%netbsd.org@localhost> 
wrote:
> On Sun, Mar 13, 2011 at 06:13:32PM +0100, Thomas Klausner wrote:
>> Since updating from 5.99.47/amd64 (2011-02-28) to 5.99.48 (2011-03-11)
>> my usually very stable desktop machine has rebooted twice without any
>> apparent reason or syslog entry.
>
> I've now got a reproducible case where it does that: When I'm running
> the tests for m4-1.4.16 in my bulk build tmpfs sandbox.
>
> Since after two times I was expecting it, I got a (handcopied)
> backtrace now:
>
> vmem_alloc() at netbsd:vmem_alloc+0x38
> tmpfs_alloc_node() at netbsd:tmpfs_alloc_node+0x232
> tmpfs_alloc_file() at netbsd:tmpfs_alloc_file+0xb0
> VOP_SYMLINK() at netbsd:VOP_SYMLINK+0x71
> do_sys_symlink() at netbsd:do_sys_symlink+0x16a
> syscall() at netbsd:syscall+0xaa

Do you use tmpfs kmod, is it build with -g ? If not can you build it with -g
and -O0 and try to reproduce your problem. Then after panic can you run

gdb tmpfs.kmod
info line *(tmpfs_alloc_node+0x232)

To get exact line where we got this panic. I saw simillar panic in zfs lately
caused by

kmem_alloc(0, 1);

-- 


Regards.

Adam