lkml.org 
[lkml]   [2002]   [Sep]   [16]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: end_request error procedure in 2.5?
On Mon, Sep 16 2002, Peter T. Breuer wrote:
> "A month of sundays ago Jens Axboe wrote:"
> > > end_request( req, (req->errors == 0) ? 1 : 0 );
> > > ..
> > >
> > > static void end_request(struct request *req, int uptodate) {
> > > struct bio *bio;
> > > while ((bio = req->bio) != NULL) {
> > > blk_finished_io(bio_sectors(bio));
> > > req->bio = bio->bi_next;
> > > bio->bi_next = NULL;
> > > bio_endio(bio, uptodate);
> > > }
> > > blk_put_request(req);
> > > }
> > >
> > >
> > > It works fine except on error. Kernel 2.5.31. I understand that
> > > put_request adds the request back to a free list (if gotten from there
> > > via get_request). The request is ordinary, except out of range ...
> > > it's produced by an e2fsck of the device when the device itself is
> > > unformatted, and the out of range request gets passed to the driver and
> > > is errored there, and "kapow" ..
> >
> > The error is most likely in the driver calling end_that_request_first(),
>
> Hmmm ... it's not called. The above is exactly all that is called
> and LOCAL_END_REQUEST is set. OK. I see what you are saying. Yes, I
> will direct my attention to that function instead ...
>
> ... and yes, I see a possible path in which the queue spinlock may be
> taken twice. OK!
>
> Thanks!
>
> > not the function itself. Maybe you can try to do at least some
> > debugging, I hope you are not expecting anyone to be able to help you
> > from the above report.
>
> !! :-)
>
> Thanks, yes I know! However, it's taken me about 4 days to get it this
> far. As you know, complete lockups are hard to debug! There's a race
> condition between the printk appearing on the console and the machine
> stopping :-(.

NMI watchdog is invaluable for this sort of thing, beats printk by a
wide margin :-). Of course that only works of you have the hardware it
works on, but that's what test machines are for.

--
Jens Axboe

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:28    [W:0.038 / U:0.940 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site