Re: [GIT PULL] Core block IO bits for 2.6.39 - early Oops

[Dave Chinner <david@fromorbit.com>]

On Thu, Mar 24, 2011 at 08:34:41PM +0100, Markus Trippelsdorf wrote:
> On 2011.03.24 at 19:58 +0100, Jens Axboe wrote:
> > On 2011-03-24 19:54, Markus Trippelsdorf wrote:
> > > On 2011.03.24 at 19:51 +0100, Jens Axboe wrote:
> > >> On 2011-03-24 19:36, Jens Axboe wrote:
> > >>> On 2011-03-24 19:30, Markus Trippelsdorf wrote:
> > >>>> On 2011.03.24 at 14:43 +0100, Jens Axboe wrote:
> > >>>>>
> > >>>>> This is the main pull request for the block IO layer and friends for
> > >>>>> 2.6.39.
> > >>>>
> > >>>> This merge results in an early oops on my system (amd64, xfs).
> > >>>> See the attached photo.
> > >>>>
> > >>>
> > >>> Auch. Can you ensure that you have CONFIG_DEBUGINFO=y in your .config
> > >>> and then do:
> > >>>
> > >>> $ gdb vmlinux
> > >>> ...
> > >>> l *cfq_insert_request+0x32
> > >>>
> > >>> and send that output?
> > >>
> > >> I took a closer look at the oops, and it most likely looks like q ==
> > >> NULL (offset 0x18 == q->elevator). You left out the Code part, so I
> > >> can't verify that for certain. Which makes very little sense. I take it
> > >> this is 100% reproducible? When you send the gdb output, please also
> > >> attach your .config.
> > > 
> > > Yes, it's 100% reproducible here. My .config follows:
> > 
> > Can you try this patch and see if it makes a difference?
> 
> There's no patch ;-)
> 
> > If you boot without the patch and add elevator=noop, does it then work?
> 
> It works insofar as the Oops is gone. But my xfs partitions apparently
> still get corrupted (I had to run xfs_repair on several of them, because
> they would not mount otherwise).

So the patchset is causing repeatable filesystem corruption? Sounds
to me like this series is not yet ready for mainline merging. Last
thing I want to spend the .39 cycle helping people recover busted
filesystems as a result of undercooked block layer changes...

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com