Fatal trap 12: page fault panic with recent kernel with ZFS

[Thomas Backman]

On May 21, 2009, at 08:31 PM, Kip Macy wrote:

>> Hey, I (not the OP) am still having trouble with this. The panics  
>> appear to
>> be gone since I set arc_min to 30M and arc_max to 100M, though (2GB  
>> RAM).
>> I get/got them during a full zfs send -R | zfs recv -Fvd of a  
>> ~3.3GB pool.
>> The only modification I've done to the source tree is the  
>> libzfs_sendrecv
>> patch here:
>> http://lists.freebsd.org/pipermail/freebsd-current/2009-May/006814.html
>
>
> I'll apply the patch this weekend.
>
> If I get time I'll also try to reproduce your panic.
>
>
> Thanks,
> Kip
Two weekends later and still the same deal with send | recv. ;)
Just worried that this'll make it into 8.0-RELEASE, that's all. That  
would be bad. The patch linked above seems to solve the problem for  
me, anyway. I just tried a fresh build (sources from June 5th at 10AM  
CEST), without the patch, and I got the same core dump on zfs recv,  
with a SIGSEGV in strcmp(). (See the linked thread.) Applied the  
patch, rebuild libzfs, rebooted, and it works again.

It seems something along the lines of
zfs snapshot -r tank at initial-snapshot
zfs send -R tank at initial-snapshot | zfs recv -vFd slave
and then
zfs snapshot -r tank at testsnap # some other, unrelated snapshot in  
between may or may not be needed to crash, I'm not sure
zfs snapshot -r tank at second-snapshot
zfs send -R -I initial-snapshot tank at second-snapshot | zfs recv -Fvd  
slave
causes the core dump.

Regards,
Thomas