Re: kern/38107: recent crash in -current

[Andrew Doran <ad%netbsd.org@localhost>]

The following reply was made to PR kern/38107; it has been noted by GNATS.

From: Andrew Doran <ad%netbsd.org@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: 
Subject: Re: kern/38107: recent crash in -current
Date: Tue, 26 Feb 2008 12:41:31 +0000

 On Tue, Feb 26, 2008 at 08:15:01AM +0000, kefren%ngnetworks.ro@localhost wrote:
 
 >      Under heavy diskload (extracting firefox sources, rm -rf /usr/src) I 
 > get the following:
 > 
 > #0  0xc046cff2 in cpu_reboot (howto=260, bootstr=0x0) at 
 > /usr/work/src/sys/arch/i386/i386/machdep.c:952
 > 952                     dumpsys();
 > (gdb) bt
 > #0  0xc046cff2 in cpu_reboot (howto=260, bootstr=0x0) at 
 > /usr/work/src/sys/arch/i386/i386/machdep.c:952
 > #1  0xc03d80fa in panic (fmt=0xc067ef24 "trap") at 
 > /usr/work/src/sys/kern/subr_prf.c:260
 > #2  0xc0471123 in trap (frame=0xcb708af8) at 
 > /usr/work/src/sys/arch/i386/i386/trap.c:372
 > #3  0xc01030c8 in calltrap ()
 > #4  0xc040d0ba in VFS_SYNC (mp=0xca4fb600, a=0, b=0x0) at 
 > /usr/work/src/sys/kern/vfs_subr2.c:1102
 > #5  0xc040c16d in vfs_shutdown () at /usr/work/src/sys/kern/vfs_subr.c:1810
 > #6  0xc046d0c9 in cpu_reboot (howto=256, bootstr=0x0) at 
 > /usr/work/src/sys/arch/i386/i386/machdep.c:938
 > #7  0xc03d80fa in panic (fmt=0xc067ef24 "trap") at 
 > /usr/work/src/sys/kern/subr_prf.c:260
 > #8  0xc0471123 in trap (frame=0xcb708c70) at 
 > /usr/work/src/sys/arch/i386/i386/trap.c:372
 > #9  0xc01030c8 in calltrap ()
 > #10 0xc040d0ba in VFS_SYNC (mp=0xcb708cdc, a=65554, b=0xc05e89a0) at 
 > /usr/work/src/sys/kern/vfs_subr2.c:1102
 > #11 0xc0417a73 in VOP_FSYNC (vp=0xcba39008, cred=0xca4e4f00, flags=8, 
 > offlo=0, offhi=0)
 >     at /usr/work/src/sys/kern/vnode_if.c:804
 > #12 0xc041eeba in sched_sync (v=0xca4fb600) at 
 > /usr/work/src/sys/miscfs/syncfs/sync_subr.c:223
 > #13 0xc01002cf in lwp_trampoline ()
 
 If it happens again, please go to the frame with the original VOP_SYNC()
 - not the one done during vfs_shutdown() - and do "print *mp" and "print
 *(mp->mnp_op)".
 
 Did you happen to unmount a file system just before the crash? It's likely
 something using the kmem allocator is corrupting kernel memory. I've seen
 evidence that indicates there is such a bug currently, so it may not be
 directly caused by VFS.
 
 Andrew