ZFS deadlock (?)

[Ivan Voras]

Maybe it's just me, but I really still have a hard time trusting ZFS.
Here's what happened today in a machine with FreeNAS 8.3:

41526 root          1  45    0 17856K  1860K tx->tx  0   0:00  0.00%
/sbin/zfs create -o compression=lzjb -V 100g store/test
41641 root          1  44    0 17856K  1860K tx->tx  0   0:00  0.00%
/sbin/zfs create -o compression=lzjb -V 200g store/bla
41492 root          1  45    0 17856K  1860K tx->tx  2   0:00  0.00%
/sbin/zfs create -o compression=lzjb -V 100g store/test
41392 root          1  44    0 17856K  1860K tx->tx  0   0:00  0.00%
/sbin/zfs create -o compression=lzjb -V 100G store/test

These processes were started from the web GUI and are apparently
deadlocked or not proceeding because of some other reason. I tried
creating the "test" volume first (the earliest PID), and when it didn't
finish in a few minutes I cancelled it from the web GUI and started
again. Apparently, all of these processes are in some kind of deadlock
state.

No IO is happening on the machine. An attempt to do a "zpool scrub
store" resulted in zpool also being (dead)locked, in a different place:
[ctrl-t]
load: 0.00  cmd: zpool 41724 [scl->scl_cv)] 37.22r 0.00u 0.00s 0% 1696k

Now, all IO on the file system is (dead) locked and cannot proceed.

There is nothing significant in /var/log/messages.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 260 bytes
Desc: OpenPGP digital signature
URL: <http://lists.freebsd.org/pipermail/freebsd-fs/attachments/20130306/53eecc41/attachment.sig>