On Fri, Jul 11, 2008 at 01:50:10PM +1000, Mark Goodwin wrote:
>
>
> Dave Chinner wrote:
>> On Fri, Jul 11, 2008 at 12:48:11PM +1000, Mark Goodwin wrote:
>>> Thanks for the report Eric. This looks very similar to a
>>> deadlock Lachlan recently hit in the patch for
>>> "Use atomics for iclog reference counting"
>>> http://oss.sgi.com/archives/xfs/2008-02/msg00130.html
>>>
>>> It seems this patch can cause deadlocks under heavy log traffic.
>>> I don't think anyone has a fix yet ... Lachlan is out this week,
>>> but Tim can follow-up here ...
>>
>> Nice to know - why didn't anyone email me or report this to the
>> list when the bug was first found? I mean, I wrote that code, I know
>> what it is supposed to be doing and as a result should be able
>
> Only recently found and didn't think it was this easy to hit.
> But no excuses ...
>
>> help find and fix the bug. Can you please post what details you have
>> about the problem (test case, stack traces, debugging info, etc)
>> so I can try to find the problem.
>
> See Tim's follow-up.
>
>> This is a regression that is in the mainline kernel that is due to
>> be released probably in the next couple of days. Having a little
>> bit of time to try and find the bug would have been nice...
>
> At this stage, I think it would be safest to back-out the commit,
> all the way to mainline. Tim, can you please work thru that today
> with priority.
No, do not back it out. I just posted the fix.
Cheers,
Dave.
--
Dave Chinner
david@xxxxxxxxxxxxx
|