Re: WARNING: at block/genhd.c:1556 disk_clear_events+0xdc/0xf0()

[Tejun Heo <tj@kernel.org>]

Hello, Sitsofe.

Sorry about late replay.  I've been and still am travelling.  :-(

On Sun, Apr 10, 2011 at 03:59:20PM +0100, Sitsofe Wheeler wrote:
> Shortly after saying this the problem reappeared by itself. I've found a
> method to reproduce it too: by running
> for i in `seq 1 100`; do udevadm trigger --subsystem-match=block ; done
> and then waiting at the desktop for a five or so minutes (so the
> requests that are backed up drain away) then running it again etc. the
> problem will usually trigger. The warning is still here in the most
> recent 2.6.39-rc2-00120-g94c8a98 kernel too. Applying the patch to the
> first kernel that showed th problem doesn't fix the issue (although it
> appears to make the issue a lot more difficult to trigger).

Thanks a lot for finding out the test case.  I'll see whether I can
reproduce the problem in qemu.

> I will note that by EeePC's onboard "gen-0" SSD mysteriously developed
> bad blocks after all this so perhaps the above is especially stressful
> for disks...

Yeah, it probably hammers the same sector over and over again.
Earlier/cheap SSDs often lack proper wear leveling and writing
repeatedly to the same sector can easily create bad blocks.  Sorry
about that.

Thanks.

-- 
tejun