ZFS-related panic, 8.0-RC1

[Kamigishi Rei]

On 07.10.2009 21:36, Gleb Kurtsou wrote:
> panic: _sx_xlock_hard: recursed on non-recursive sx
>> zfsvfs->z_hold_mtx[i] @
>> /usr/src/sys/modules/zfs/../../cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_znode.c:1023
>>      
> Do you run nullfs on top of zfs by any chance?
>    
Indeed. I'm using ezjail suite, and it uses nullfs to link jail 
subdirectories to the base structure.

> I've seen something similar. It was caused by
> zfs ->  getnewvnode ->  zfs_inactive
> But information provided is not enough for debugging. Can you get full
> backtrace?
>    
I've disabled gmirror so next time I'll probably get the backtrace.

Note that with 8.0-BETA3, the previous revision I ran, nothing happened 
through 39+ days uptime with the same ezjail-based structure.

--
Kamigishi Rei
KREI-RIPE