Re: zfs solaris assert panic in 11.0-ALPHA5 r302256

On 18/07/2016 20:40, Don Lewis wrote:
> On 18 Jul, Andriy Gapon wrote:
>> On 08/07/2016 07:13, Don Lewis wrote:
>>> My package buiding machine just crashed with this panic during a
>>> poudriere run:
>>>
>>> panic: solaris assert: (dn->dn_phys->dn_nlevels == 0 && db->db_level == 0) || dn->dn_phys->dn_nlevels > db->db_level || dn->dn_next_nlevels[txgoff] > db->db_level || dn->dn_next_nlevels[(tx->tx_txg-1) & TXG_MASK] > db->db_level || dn->dn_next_nlevels[(tx->tx_txg
>>
>> Don,
>>
>> do you have a crash dump?
>> It would be interesting to see a pretty-print of dn, dn->dn_phys, db and
>> tx in the frame where the assert is hit.
> 
> I do.  Unfortunately kgdb reports that the values of dn and db were
> optimized out.
> 

Well... You can try to use kgdb7111 from ports, perhaps it would work
better. Also, it's often possible to find values of wanted variables by
finding a relevant value that's not optimized out and then following
through pointers, etc to get to the right values.  in other cases it's
possible to get the values by examining the disassembly and values of
registers.

Finally, if you can get anything useful out of the dump it would make
sense to follow up to
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=203419


-- 
Andriy Gapon