All of lore.kernel.org
 help / color / mirror / Atom feed
From: Vivek Goyal <vgoyal@redhat.com>
To: Jens Axboe <axboe@kernel.dk>
Cc: linux kernel mailing list <linux-kernel@vger.kernel.org>,
	Tejun Heo <tj@kernel.org>, Chris Mason <chris.mason@oracle.com>
Subject: Re: Kernel crash in icq_free_icq_rcu
Date: Tue, 17 Jan 2012 16:01:14 -0500	[thread overview]
Message-ID: <20120117210114.GA23527@redhat.com> (raw)
In-Reply-To: <20120117205816.GF19223@redhat.com>

On Tue, Jan 17, 2012 at 03:58:16PM -0500, Vivek Goyal wrote:

[..]

> > >> Can you try this?
> > > 
> > > Nope, this does not help either. Can reproduce the issue with below
> > > patch applied.
> > > 
> > >>
> > >>
> > >> diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
> > >> index 163263d..ee55019 100644
> > >> --- a/block/cfq-iosched.c
> > >> +++ b/block/cfq-iosched.c
> > >> @@ -3117,18 +3117,17 @@ cfq_should_preempt(struct cfq_data *cfqd, struct cfq_queue *new_cfqq,
> > >>   */
> > >>  static void cfq_preempt_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
> > >>  {
> > >> -	struct cfq_queue *old_cfqq = cfqd->active_queue;
> > >> -
> > >>  	cfq_log_cfqq(cfqd, cfqq, "preempt");
> > >> -	cfq_slice_expired(cfqd, 1);
> > >>  
> > >>  	/*
> > >>  	 * workload type is changed, don't save slice, otherwise preempt
> > >>  	 * doesn't happen
> > >>  	 */
> > >> -	if (cfqq_type(old_cfqq) != cfqq_type(cfqq))
> > >> +	if (cfqq_type(cfqd->active_queue) != cfqq_type(cfqq))
> > >>  		cfqq->cfqg->saved_workload_slice = 0;
> > >>  
> > >> +	cfq_slice_expired(cfqd, 1);
> > >> +
> > > 
> > > cfq_slice_expired() will overwrite the value of
> > > cfqq->cfqg->saved_workload_slice. So we need to set it to zero after
> > > cfq_slice_expired.
> > 
> > Good point, lets just fix that up afterwards, the use-after-free needs
> > to go asap.
> > 
> > > I was thinking of just saving the workload type of cfqq before
> > > cfq_slice_expired() so that we don't access old cfqq after
> > > cfq_slice_expired(). But then I noticed that we don't drop a cfqq
> > > reference in cfq_slice_expired(). So not sure how cfq_slice_expired()
> > > can lead to freeing up of queue. It should happen only when process
> > > has exited and last request on the queue if finished. 
> > 
> > It does, it drops a ref to the cic which in turn drops the active async
> > and sync queues.
> 
> Ok, I see it now. Thanks.
> 
> I modified your patch a bit. It does not seem to solve my problem but
> might help with Chris Mason's boot issue.
> 
> Chris, can you please give it a try.

Oops, old mail id of chris. Fixing it now.

Thanks
Vivek

> 
> Thanks
> Vivek
> 
> ---
>  block/cfq-iosched.c |    4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> Index: linux-2.6/block/cfq-iosched.c
> ===================================================================
> --- linux-2.6.orig/block/cfq-iosched.c	2012-01-18 02:49:33.000000000 -0500
> +++ linux-2.6/block/cfq-iosched.c	2012-01-18 02:50:31.000000000 -0500
> @@ -3117,7 +3117,7 @@ cfq_should_preempt(struct cfq_data *cfqd
>   */
>  static void cfq_preempt_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
>  {
> -	struct cfq_queue *old_cfqq = cfqd->active_queue;
> +	enum wl_type_t old_cfqq_wl_type = cfqq_type(cfqd->active_queue);
>  
>  	cfq_log_cfqq(cfqd, cfqq, "preempt");
>  	cfq_slice_expired(cfqd, 1);
> @@ -3126,7 +3126,7 @@ static void cfq_preempt_queue(struct cfq
>  	 * workload type is changed, don't save slice, otherwise preempt
>  	 * doesn't happen
>  	 */
> -	if (cfqq_type(old_cfqq) != cfqq_type(cfqq))
> +	if (old_cfqq_wl_type != cfqq_type(cfqq))
>  		cfqq->cfqg->saved_workload_slice = 0;
>  
>  	/*

  reply	other threads:[~2012-01-17 21:01 UTC|newest]

Thread overview: 30+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2012-01-17 20:18 Kernel crash in icq_free_icq_rcu Vivek Goyal
2012-01-17 20:19 ` Jens Axboe
2012-01-17 20:40   ` Vivek Goyal
2012-01-17 20:42     ` Jens Axboe
2012-01-17 20:58       ` Vivek Goyal
2012-01-17 21:01         ` Vivek Goyal [this message]
2012-01-17 21:48 ` Tejun Heo
2012-01-17 22:07   ` Vivek Goyal
2012-01-18  1:01     ` Shaohua Li
2012-01-18  1:03       ` Tejun Heo
2012-01-18  1:05         ` Shaohua Li
2012-01-18  1:11           ` Tejun Heo
2012-01-18  1:30             ` Shaohua Li
2012-01-18  2:26               ` Shaohua Li
2012-01-18  4:23                 ` Shaohua Li
2012-01-18  6:03               ` Shaohua Li
2012-01-18 13:51                 ` Vivek Goyal
2012-01-18 14:20                   ` Vivek Goyal
2012-01-18 16:09                     ` Tejun Heo
2012-01-18 16:24                       ` Jens Axboe
2012-01-18 16:31                         ` Jens Axboe
2012-01-18 16:36                           ` Vivek Goyal
2012-01-18 17:10                             ` Tejun Heo
2012-01-18 19:07                               ` Jens Axboe
2012-01-18 19:05                             ` Jens Axboe
2012-01-18 16:55                           ` Tejun Heo
2012-01-18 16:07                 ` Tejun Heo
2012-01-19  1:41                   ` [patch]block: fix NULL icq_cache reference Shaohua Li
2012-01-19  1:43                     ` Tejun Heo
2012-01-19  8:20                     ` Jens Axboe

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20120117210114.GA23527@redhat.com \
    --to=vgoyal@redhat.com \
    --cc=axboe@kernel.dk \
    --cc=chris.mason@oracle.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=tj@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.