Re: netbsd-5 crash with enabled quota

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

On Sun, Jan 25, 2009 at 08:10:16PM +0100, 
6bone%6bone.informatik.uni-leipzig.de@localhost wrote:
> Hello list,
> 
> I am using an actual version of NetBSD 5.0_BETA (amd). With enabled quota 
> the kernel sometimes crashes with:
> 
> vnode @ 0xffff800096eade58, flags (1030<MPSAFE,LOCKSWORK,XLOCK>)
>         tag UNKNOWN(0), type VREG(1), usecount 2, writecount 0, holdcount 
> 0
>         freelisthd 0x0, mount 0xffff800071b70000, data 0x0 lock 
> 0xffff800096ead0
> panic: dead but not clean
> 
> fatal breakpoint trap in supervisor mode
> trap type 1 code 0 rip ffffffff8051e215 cs 8 rflags 246 cr2 
> ffff8000717ae000 c0
> Stopped in pid 0.59 (system) at netbsd:breakpoint+0x5:  leave
> db{1}> trace
> breakpoint() at netbsd:breakpoint+0x5
> panic() at netbsd:panic+0x260
> vpanic() at netbsd:vpanic+0x27
> vrelel() at netbsd:vrelel+0xad
> vget() at netbsd:vget+0x160
> qsync() at netbsd:qsync+0xea
> ffs_sync() at netbsd:ffs_sync+0x29c
> VFS_SYNC() at netbsd:VFS_SYNC+0x33
> sync_fsync() at netbsd:sync_fsync+0x76
> VOP_FSYNC() at netbsd:VOP_FSYNC+0x71
> sched_sync() at netbsd:sched_sync+0x14f
> 
> 
> Can anyone have a look at the code?
> 
> 
> Thank you for your efforts.

It's a race condition affecting only the assertion. That is to say, the
assertion is broken but the system is in fact OK. I have a fix for this and
am going to get it pulled into the 5.0 branch.

Andrew