Bad hardware + zfs = panic

[Pawel Jakub Dawidek]

On Thu, May 06, 2010 at 03:22:17AM +0200, Ståle Kristoffersen wrote:
> I've been debugging a hardware error for the past few days, and I think it
> was the CPU and that it is now fixed. But reading a file that was written to a
> zfs-pool when stuff got corrupted still triggered a panic in ZFS code:
> 
> Fatal trap 12: page fault while in kernel mode
> cpuid = 0; apic id = 00
> fault virtual address   = 0x28
> fault code              = supervisor read data, page not present
> instruction pointer     = 0x20:0xffffffff8106f2d3
> stack pointer           = 0x28:0xffffff80774914e0
> frame pointer           = 0x28:0xffffff8077491510
> code segment            = base 0x0, limit 0xfffff, type 0x1b
>                         = DPL 0, pres 1, long 1, def32 0, gran 1
> processor eflags        = interrupt enabled, resume, IOPL = 0
> current process         = 1350 (smbd)
> trap number             = 12
> panic: page fault
> cpuid = 0
> Uptime: 2m53s
> 
> The lines in the backtrace that got my attention was:
> #6  0xffffffff80847c73 in calltrap () at
> /usr/src/sys/amd64/amd64/exception.S:224
> #7  0xffffffff8106f2d3 in vdev_is_dead (vd=0x0) at
> /usr/src/sys/modules/zfs/../../cddl/contrib/opensolaris/uts/common/fs/zfs/vdev.c:1847
> #8  0xffffffff8106f2ed in vdev_readable (vd=0x0) at
> /usr/src/sys/modules/zfs/../../cddl/contrib/opensolaris/uts/common/fs/zfs/vdev.c:1854
> 
> The complete bt is available here:
> http://heim.ifi.uio.no/staalebk/zfs-panic.txt
> 
> As you can see vd=0x0, and I think that caused the panic, since it
> tried to follow that pointer:
>  return (vd->vdev_state < VDEV_STATE_DEGRADED);
> 
> I then tried to remove the file and I got this:
> Solaris: WARNING: metaslab_free_dva(): bad DVA
> 199476166:1296607792756162560
> Solaris: WARNING: metaslab_free_dva(): bad DVA 4236221:7256850009726709760
> Solaris: WARNING: metaslab_free_dva(): bad DVA
> 935912721:16480078061480073216
> 
> Maybe there should be a test to check if vd was zero, and
> throw an io-error or something, instead of panicing?

Well, I don't think it should be possible for vdev to be NULL.
But if you still have this panic, can you try this patch:

	http://people.freebsd.org/~pjd/patches/vdev_mirror.c.patch

-- 
Pawel Jakub Dawidek                       http://www.wheelsystems.com
pjd at FreeBSD.org                           http://www.FreeBSD.org
FreeBSD committer                         Am I Evil? Yes, I Am!
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 196 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-fs/attachments/20100512/b9e8fe71/attachment.pgp