Re: BUG at scsi_lib.c:1108 [Was: mmotm 2009-11-24-16-47 uploaded]

[Jiri Slaby <jirislaby@gmail.com>]

On 11/25/2009 04:19 PM, James Bottomley wrote:
> On Wed, 2009-11-25 at 16:12 +0100, Jiri Slaby wrote:
>> On 11/25/2009 01:47 AM, akpm@linux-foundation.org wrote:
>>> The mm-of-the-moment snapshot 2009-11-24-16-47 has been uploaded to
>>
>> Hi, I'm hitting the BUG below in the past two -mmotm's
>> (2009-11-24-16-47, 2009-11-17-14-03):
> 
> So this looks like some type of bug in the barrier code. What the BUG_ON
> is saying is that something sent us a REQ_TYPE_FS (which should be a
> filesystem read or write) with no attached data, so we can't process it.
> 
> I've cc'd Jens to see what he thinks.
> 
> Could you bisect this to find the offending commit?

Hmm. I bisected it twice to
commit 1bebedd653e1bb0440ebf40724c55791c21ad7cc
Merge: a8d5ddf 62fa36a
Author: Stephen Rothwell <sfr@canb.auug.org.au>
Date:   Wed Nov 25 17:42:47 2009 +1100

    Merge remote branch 'md/for-next'


It doesn't make sense at all. How can empty merge cause a regression?
And md/for-next doesn't produce the BUG.

After the 1st bisection I did
 git checkout 1bebedd653e1bb0440ebf40724c55791c21ad7cc
and built it. It crashed. Then I did
 git bisect start 1bebedd origin/stable
where origin is next. And got back to the 1bebedd by bisection.

-- 
js
Faculty of Informatics, Masaryk University
Suse Labs, Novell