From: John Garry <john.garry@huawei.com>
To: <hare@suse.de>, <bvanassche@acm.org>, <ming.lei@redhat.com>,
<axboe@kernel.dk>, <hch@lst.de>
Cc: <linux-block@vger.kernel.org>, <linux-kernel@vger.kernel.org>,
<pragalla@codeaurora.org>, <kashyap.desai@broadcom.com>,
<yuyufen@huawei.com>, John Garry <john.garry@huawei.com>
Subject: [RFC PATCH v3 1/3] blk-mq: Clean up references to old requests when freeing rqs
Date: Fri, 5 Mar 2021 23:14:52 +0800 [thread overview]
Message-ID: <1614957294-188540-2-git-send-email-john.garry@huawei.com> (raw)
In-Reply-To: <1614957294-188540-1-git-send-email-john.garry@huawei.com>
It has been reported many times that a use-after-free can be intermittently
found when iterating busy requests:
- https://lore.kernel.org/linux-block/8376443a-ec1b-0cef-8244-ed584b96fa96@huawei.com/
- https://lore.kernel.org/linux-block/5c3ac5af-ed81-11e4-fee3-f92175f14daf@acm.org/T/#m6c1ac11540522716f645d004e2a5a13c9f218908
- https://lore.kernel.org/linux-block/04e2f9e8-79fa-f1cb-ab23-4a15bf3f64cc@kernel.dk/
The issue is that when we switch scheduler or change queue depth, there may
be references in the driver tagset to the stale requests.
As a solution, clean up any references to those requests in the driver
tagset. This is done with a cmpxchg to make safe any race with setting the
driver tagset request from another queue.
Signed-off-by: John Garry <john.garry@huawei.com>
---
block/blk-mq-sched.c | 2 +-
block/blk-mq-tag.c | 2 +-
block/blk-mq.c | 20 ++++++++++++++++++--
block/blk-mq.h | 2 ++
4 files changed, 22 insertions(+), 4 deletions(-)
diff --git a/block/blk-mq-sched.c b/block/blk-mq-sched.c
index ddb65e9e6fd9..bc19bd8f8c7b 100644
--- a/block/blk-mq-sched.c
+++ b/block/blk-mq-sched.c
@@ -615,7 +615,7 @@ void blk_mq_sched_free_requests(struct request_queue *q)
queue_for_each_hw_ctx(q, hctx, i) {
if (hctx->sched_tags)
- blk_mq_free_rqs(q->tag_set, hctx->sched_tags, i);
+ blk_mq_free_rqs_ext(q->tag_set, hctx->sched_tags, i, hctx->tags);
}
}
diff --git a/block/blk-mq-tag.c b/block/blk-mq-tag.c
index ce813b909339..7ff1b20d58e7 100644
--- a/block/blk-mq-tag.c
+++ b/block/blk-mq-tag.c
@@ -580,7 +580,7 @@ int blk_mq_tag_update_depth(struct blk_mq_hw_ctx *hctx,
return -ENOMEM;
}
- blk_mq_free_rqs(set, *tagsptr, hctx->queue_num);
+ blk_mq_free_rqs_ext(set, *tagsptr, hctx->queue_num, hctx->tags);
blk_mq_free_rq_map(*tagsptr, flags);
*tagsptr = new;
} else {
diff --git a/block/blk-mq.c b/block/blk-mq.c
index d4d7c1caa439..9cb60bf7ac24 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -2286,8 +2286,8 @@ blk_qc_t blk_mq_submit_bio(struct bio *bio)
return BLK_QC_T_NONE;
}
-void blk_mq_free_rqs(struct blk_mq_tag_set *set, struct blk_mq_tags *tags,
- unsigned int hctx_idx)
+void __blk_mq_free_rqs_ext(struct blk_mq_tag_set *set, struct blk_mq_tags *tags,
+ unsigned int hctx_idx, struct blk_mq_tags *ref_tags)
{
struct page *page;
@@ -2296,10 +2296,14 @@ void blk_mq_free_rqs(struct blk_mq_tag_set *set, struct blk_mq_tags *tags,
for (i = 0; i < tags->nr_tags; i++) {
struct request *rq = tags->static_rqs[i];
+ int j;
if (!rq)
continue;
set->ops->exit_request(set, rq, hctx_idx);
+ /* clean up any references which occur in @ref_tags */
+ for (j = 0; ref_tags && j < ref_tags->nr_tags; j++)
+ cmpxchg(&ref_tags->rqs[j], rq, 0);
tags->static_rqs[i] = NULL;
}
}
@@ -2316,6 +2320,18 @@ void blk_mq_free_rqs(struct blk_mq_tag_set *set, struct blk_mq_tags *tags,
}
}
+void blk_mq_free_rqs_ext(struct blk_mq_tag_set *set, struct blk_mq_tags *tags,
+ unsigned int hctx_idx, struct blk_mq_tags *ref_tags)
+{
+ __blk_mq_free_rqs_ext(set, tags, hctx_idx, ref_tags);
+}
+
+void blk_mq_free_rqs(struct blk_mq_tag_set *set, struct blk_mq_tags *tags,
+ unsigned int hctx_idx)
+{
+ __blk_mq_free_rqs_ext(set, tags, hctx_idx, NULL);
+}
+
void blk_mq_free_rq_map(struct blk_mq_tags *tags, unsigned int flags)
{
kfree(tags->rqs);
diff --git a/block/blk-mq.h b/block/blk-mq.h
index 3616453ca28c..031e29f74926 100644
--- a/block/blk-mq.h
+++ b/block/blk-mq.h
@@ -53,6 +53,8 @@ struct request *blk_mq_dequeue_from_ctx(struct blk_mq_hw_ctx *hctx,
*/
void blk_mq_free_rqs(struct blk_mq_tag_set *set, struct blk_mq_tags *tags,
unsigned int hctx_idx);
+void blk_mq_free_rqs_ext(struct blk_mq_tag_set *set, struct blk_mq_tags *tags,
+ unsigned int hctx_idx, struct blk_mq_tags *references);
void blk_mq_free_rq_map(struct blk_mq_tags *tags, unsigned int flags);
struct blk_mq_tags *blk_mq_alloc_rq_map(struct blk_mq_tag_set *set,
unsigned int hctx_idx,
--
2.26.2
next prev parent reply other threads:[~2021-03-05 15:20 UTC|newest]
Thread overview: 30+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-03-05 15:14 [RFC PATCH v3 0/3] blk-mq: Avoid use-after-free for accessing old requests John Garry
2021-03-05 15:14 ` John Garry [this message]
2021-03-06 2:52 ` [RFC PATCH v3 1/3] blk-mq: Clean up references to old requests when freeing rqs Khazhy Kumykov
2021-03-08 11:18 ` John Garry
2021-03-06 18:13 ` Bart Van Assche
2021-03-08 10:37 ` John Garry
2021-03-05 15:14 ` [RFC PATCH v3 2/3] blk-mq: Freeze and quiesce all queues for tagset in elevator_exit() John Garry
2021-03-06 4:32 ` Bart Van Assche
2021-03-08 10:50 ` John Garry
2021-03-08 19:35 ` Bart Van Assche
2021-03-10 15:57 ` Bart Van Assche
2021-03-11 0:58 ` Ming Lei
2021-03-11 8:21 ` John Garry
2021-03-12 23:05 ` Bart Van Assche
2021-03-16 16:15 ` John Garry
2021-03-16 17:00 ` Bart Van Assche
2021-03-16 17:43 ` John Garry
2021-03-16 19:59 ` Bart Van Assche
2021-03-19 18:19 ` John Garry
2021-03-19 18:32 ` Bart Van Assche
2021-03-05 15:14 ` [RFC PATCH v3 3/3] blk-mq: Lockout tagset iterator when exiting elevator John Garry
2021-03-06 4:43 ` Bart Van Assche
2021-03-08 11:17 ` John Garry
2021-03-08 19:59 ` Bart Van Assche
2021-03-09 17:47 ` John Garry
2021-03-09 19:21 ` Bart Van Assche
2021-03-10 8:52 ` John Garry
2021-03-10 16:00 ` Bart Van Assche
2021-03-10 17:26 ` John Garry
2021-03-18 10:26 ` [RFC PATCH v3 0/3] blk-mq: Avoid use-after-free for accessing old requests Shinichiro Kawasaki
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=1614957294-188540-2-git-send-email-john.garry@huawei.com \
--to=john.garry@huawei.com \
--cc=axboe@kernel.dk \
--cc=bvanassche@acm.org \
--cc=hare@suse.de \
--cc=hch@lst.de \
--cc=kashyap.desai@broadcom.com \
--cc=linux-block@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=ming.lei@redhat.com \
--cc=pragalla@codeaurora.org \
--cc=yuyufen@huawei.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