From: John Garry <john.garry@huawei.com>
To: Ming Lei <ming.lei@redhat.com>, Jens Axboe <axboe@kernel.dk>,
"Christoph Hellwig" <hch@lst.de>
Cc: <linux-block@vger.kernel.org>,
<syzbot+77ba3d171a25c56756ea@syzkaller.appspotmail.com>
Subject: Re: [PATCH] blk-mq: fix use-after-free in blk_mq_exit_sched
Date: Wed, 9 Jun 2021 09:59:43 +0100 [thread overview]
Message-ID: <f5fbc650-5bd3-32ee-1d31-8b1dd1d7fa19@huawei.com> (raw)
In-Reply-To: <20210609063046.122843-1-ming.lei@redhat.com>
On 09/06/2021 07:30, Ming Lei wrote:
Thanks for the fix
> tagset can't be used after blk_cleanup_queue() is returned because
> freeing tagset usually follows blk_clenup_queue(). Commit d97e594c5166
> ("blk-mq: Use request queue-wide tags for tagset-wide sbitmap") adds
> check on q->tag_set->flags in blk_mq_exit_sched(), and causes
> use-after-free.
>
> Fixes it by using hctx->flags.
>
The tagset is a member of the Scsi_Host structure. So it is true that
this memory may be freed before the request_queue is exited?
> Reported-by: syzbot+77ba3d171a25c56756ea@syzkaller.appspotmail.com
> Fixes: d97e594c5166 ("blk-mq: Use request queue-wide tags for tagset-wide sbitmap")
> Cc: John Garry <john.garry@huawei.com>
> Signed-off-by: Ming Lei <ming.lei@redhat.com>
> ---
> block/blk-mq-sched.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/block/blk-mq-sched.c b/block/blk-mq-sched.c
> index a9182d2f8ad3..80273245d11a 100644
> --- a/block/blk-mq-sched.c
> +++ b/block/blk-mq-sched.c
> @@ -680,6 +680,7 @@ void blk_mq_exit_sched(struct request_queue *q, struct elevator_queue *e)
> {
> struct blk_mq_hw_ctx *hctx;
> unsigned int i;
> + unsigned int flags = 0;
>
> queue_for_each_hw_ctx(q, hctx, i) {
> blk_mq_debugfs_unregister_sched_hctx(hctx);
> @@ -687,12 +688,13 @@ void blk_mq_exit_sched(struct request_queue *q, struct elevator_queue *e)
> e->type->ops.exit_hctx(hctx, i);
> hctx->sched_data = NULL;
> }
> + flags = hctx->flags;
I know the choice is limited, but it is unfortunate that we must set
flags in a loop
> }
> blk_mq_debugfs_unregister_sched(q);
> if (e->type->ops.exit_sched)
> e->type->ops.exit_sched(e);
> blk_mq_sched_tags_teardown(q);
> - if (blk_mq_is_sbitmap_shared(q->tag_set->flags))
> + if (blk_mq_is_sbitmap_shared(flags))
> blk_mq_exit_sched_shared_sbitmap(q);
this is
blk_mq_exit_sched_shared_sbitmap(struct request_queue *queue)
{
sbitmap_queue_free(&queue->sched_bitmap_tags);
..
}
And isn't it safe to call sbitmap_queue_free() when
sbitmap_queue_init_node() has not been called?
I'm just wondering if we can always call
blk_mq_exit_sched_shared_sbitmap()? I know it's not an ideal choice either.
Thanks,
John
> q->elevator = NULL;
> }
>
next prev parent reply other threads:[~2021-06-09 9:05 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-06-09 6:30 [PATCH] blk-mq: fix use-after-free in blk_mq_exit_sched Ming Lei
2021-06-09 8:59 ` John Garry [this message]
2021-06-09 10:16 ` Ming Lei
2021-06-09 11:42 ` John Garry
2021-06-14 22:07 ` Ming Lei
2021-06-15 10:02 ` John Garry
2021-06-18 6:03 ` Ming Lei
2021-06-18 14:50 ` 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=f5fbc650-5bd3-32ee-1d31-8b1dd1d7fa19@huawei.com \
--to=john.garry@huawei.com \
--cc=axboe@kernel.dk \
--cc=hch@lst.de \
--cc=linux-block@vger.kernel.org \
--cc=ming.lei@redhat.com \
--cc=syzbot+77ba3d171a25c56756ea@syzkaller.appspotmail.com \
/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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).