ZFS System crashes on "zpool status tank" on 11.0-RELEASE-p2

I have a zfs pool encrypted with geli. After a clean system boot, I mount the geli devices, then do zfs status tank so the pool is discovered. The system panics. I found and old thread documenting a similar issue (involving SSDs and GELI), but it still seems to be happening. Here's my stack trace:

Code:
Fatal double fault
...
panic: double fault
cpuid = 2
KDB: stack backtrace:
#0 0xffffffffB8b24077 at kdb_backtrace+0x67
#1 0xffffffff8Oad93e2 at vpanic+0x182
#2 0xffffffff88ad9253 at panic+0x43
#3 0xffffffff8efa18d2 at dblfault_handler+0xa2
#4 0xffffffff88f8423c at Xdblfault+0xac
#5 0xffffffff8223c6e3 at vdev_queue_io_done+0x83
#6 0xffffffff822594a9 at zio_vdev_io_done+0xd9
#7 0xffffffffB2255b26 at zio_execute+0x236
#8 0xffffffff822592ed at zio_vdev_io_start+0x34d
#9 0xffffffff82255b26 at zio_execute+0x236
#10 0xffffffff8223c72d at vdev_queue_io_done+0xcd
#11 0xffffffff822594b9 at zio_vdev_io_done+0xd9
#12 0xffffffff82255b26 at zio_execute+0x236
#13 0xffffffff822592ed at zio_vdev_io_start+0x34d
#14 0xffffffffB2255b26 at zio_execute+0x236
#15 0xffffffff8223c72d at vdev_queue_io_done+0xcd
#16 0xffffffff82259489 at zio_vdev_io_done+0xd9
#17 0xffffffff82255b26 at zio_execute+0x236
 
I think you should do zpool import for discovery. When you do zpool status you implicitly ask the system to actually import the pool, which will fail because of Xdblfault; most probably because of stack overrun.
By the way, this kernel panic comes from the function vdev_queue_io_done - a simple function that waits for io to complete. From the view of your stack it seems that these vdev_io functions are recursively calling each other, some block pointer loop.

Another command that might give more info, try this on all the devices in your pool (XXX):
zdb -ulAAA /dev/XXX
 
Back
Top