ZFS import panic with r219703

[Freddie Cash]

On Thu, Mar 17, 2011 at 10:00 AM, Olivier Smedts <olivier at gid0.org> wrote:
> 2011/3/17 Freddie Cash <fjwcash at gmail.com>:
>>> Hrm, it looks like the "pool roll-back on import" feature is working.
>>>
>>> # zpool import -F -d /dev/hast storage
>>>
>>> The above command imported the pool successfully.  No dmu_free_range()
>>> errors.  No solaris assert.  No kernel panic.  Will try hammering on
>>> the system a bit to see if that sticks or whether the space_map errors
>>> show up again.
>>
>> Damn, of course that would be too easy.  :(  Adding or removing any
>> data from the pool still causes it to panic with the dmu_free_range()
>> assertion.
>
> Does resilvering help after the forced import ?

I think this pool is hooped.  :(  It won't import in any way now, no
matter what combination of options I use: readonly, force, roll-back,
without the corrupted hast device so it's in a degraded state, etc.

The latest panic is:

solaris assert: zio-io_type != ZIO_TYPE_WRITE || spa_writable(spa),
file: /usr/src/sys/modules/zfs/../../cddl/contrib/opensolaris/uts/common/fs/zfs/zio.c,
line 2321

In case anyone is interested in the results, I've put core.txt.12 up
at http://www.sd73.bc.ca/downloads/crash/ which is the core file
relating to the above panic.

I think after lunch I'm going to destroy the pool and start over.
This box went through a lot of crashes and hangs while finding the
right loader.conf tunables for hast/zfs and issues with CompactFlash
for the OS.  Now that I've got those set and figured out, I'm going to
start over and see how things go.

-- 
Freddie Cash
fjwcash at gmail.com