From: Akinobu Mita <akinobu.mita@gmail.com>
To: linux-kernel@vger.kernel.org, Jens Axboe <axboe@kernel.dk>
Cc: Akinobu Mita <akinobu.mita@gmail.com>
Subject: [PATCH 1/4] blk-mq: fix sysfs registration/unregistration race
Date: Sun, 21 Jun 2015 22:52:28 +0900 [thread overview]
Message-ID: <1434894751-6877-2-git-send-email-akinobu.mita@gmail.com> (raw)
In-Reply-To: <1434894751-6877-1-git-send-email-akinobu.mita@gmail.com>
There is a race between cpu hotplug handling and adding/deleting
gendisk for blk-mq, where both are trying to register and unregister
the same sysfs entries.
cpu hotplug handling for blk-mq (blk_mq_queue_reinit) does unregister
and register sysfs entries of hctx for all request queues in
all_q_list. On the other hand, these entries for a request queue may
have already been unregistered when cpu hotplug event occurs. Because
those sysfs entries are unregistered by blk_mq_unregister_disk()
firstly, and then a request queue is deleted from all_q_list by
blk_mq_free_queue(). So there is a race window that cpu hotplug must
not occur.
Fix it by serializing sysfs registration/unregistration with using
per hardware queue mutex and BLK_MQ_F_SYSFS_UP flag.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: Jens Axboe <axboe@kernel.dk>
---
block/blk-mq-sysfs.c | 19 +++++++++++++++++--
block/blk-mq.c | 1 +
include/linux/blk-mq.h | 1 +
3 files changed, 19 insertions(+), 2 deletions(-)
diff --git a/block/blk-mq-sysfs.c b/block/blk-mq-sysfs.c
index b79685e..d8ef3a3 100644
--- a/block/blk-mq-sysfs.c
+++ b/block/blk-mq-sysfs.c
@@ -371,7 +371,10 @@ void blk_mq_unregister_disk(struct gendisk *disk)
int i, j;
queue_for_each_hw_ctx(q, hctx, i) {
+ mutex_lock(&hctx->sysfs_up_lock);
blk_mq_unregister_hctx(hctx);
+ hctx->flags &= ~BLK_MQ_F_SYSFS_UP;
+ mutex_unlock(&hctx->sysfs_up_lock);
hctx_for_each_ctx(hctx, ctx, j)
kobject_put(&ctx->kobj);
@@ -423,10 +426,17 @@ int blk_mq_register_disk(struct gendisk *disk)
kobject_uevent(&q->mq_kobj, KOBJ_ADD);
queue_for_each_hw_ctx(q, hctx, i) {
+ mutex_lock(&hctx->sysfs_up_lock);
+
hctx->flags |= BLK_MQ_F_SYSFS_UP;
ret = blk_mq_register_hctx(hctx);
- if (ret)
+
+ if (ret) {
+ hctx->flags &= ~BLK_MQ_F_SYSFS_UP;
+ mutex_unlock(&hctx->sysfs_up_lock);
break;
+ }
+ mutex_unlock(&hctx->sysfs_up_lock);
}
if (ret) {
@@ -443,8 +453,11 @@ void blk_mq_sysfs_unregister(struct request_queue *q)
struct blk_mq_hw_ctx *hctx;
int i;
- queue_for_each_hw_ctx(q, hctx, i)
+ queue_for_each_hw_ctx(q, hctx, i) {
+ mutex_lock(&hctx->sysfs_up_lock);
blk_mq_unregister_hctx(hctx);
+ mutex_unlock(&hctx->sysfs_up_lock);
+ }
}
int blk_mq_sysfs_register(struct request_queue *q)
@@ -453,7 +466,9 @@ int blk_mq_sysfs_register(struct request_queue *q)
int i, ret = 0;
queue_for_each_hw_ctx(q, hctx, i) {
+ mutex_lock(&hctx->sysfs_up_lock);
ret = blk_mq_register_hctx(hctx);
+ mutex_unlock(&hctx->sysfs_up_lock);
if (ret)
break;
}
diff --git a/block/blk-mq.c b/block/blk-mq.c
index 594eea0..ec94258 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -1660,6 +1660,7 @@ static int blk_mq_init_hctx(struct request_queue *q,
INIT_DELAYED_WORK(&hctx->delay_work, blk_mq_delay_work_fn);
spin_lock_init(&hctx->lock);
INIT_LIST_HEAD(&hctx->dispatch);
+ mutex_init(&hctx->sysfs_up_lock);
hctx->queue = q;
hctx->queue_num = hctx_idx;
hctx->flags = set->flags;
diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h
index 2056a99..78cd4f3 100644
--- a/include/linux/blk-mq.h
+++ b/include/linux/blk-mq.h
@@ -59,6 +59,7 @@ struct blk_mq_hw_ctx {
struct blk_mq_cpu_notifier cpu_notifier;
struct kobject kobj;
+ struct mutex sysfs_up_lock;
};
struct blk_mq_tag_set {
--
1.9.1
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
Please read the FAQ at http://www.tux.org/lkml/
next prev parent reply other threads:[~2015-06-21 13:53 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-06-21 13:52 [PATCH 0/4] blk-mq: fix race conditions on cpu hotplug handling Akinobu Mita
2015-06-21 13:52 ` Akinobu Mita [this message]
2015-06-24 8:59 ` [PATCH 1/4] blk-mq: fix sysfs registration/unregistration race Ming Lei
2015-06-21 13:52 ` [PATCH 2/4] blk-mq: fix q->mq_map access race Akinobu Mita
2015-06-24 9:23 ` Ming Lei
2015-06-21 13:52 ` [PATCH 3/4] blk-mq: establish new mapping before cpu starts handling requests Akinobu Mita
2015-06-24 9:46 ` Ming Lei
2015-06-24 14:34 ` Akinobu Mita
2015-06-24 16:24 ` Ming Lei
2015-06-25 2:56 ` Akinobu Mita
2015-06-25 8:07 ` Ming Lei
2015-06-25 12:49 ` Akinobu Mita
2015-06-25 15:40 ` Ming Lei
2015-06-25 23:35 ` Akinobu Mita
2015-06-26 17:14 ` Akinobu Mita
2015-06-27 16:08 ` Ming Lei
2015-06-29 14:25 ` Akinobu Mita
2015-06-30 9:06 ` Ming Lei
2015-06-21 13:52 ` [PATCH 4/4] blk-mq: fix mq_usage_counter race when switching to percpu mode Akinobu Mita
2015-06-24 12:35 ` Ming Lei
2015-06-29 14:31 ` Akinobu Mita
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=1434894751-6877-2-git-send-email-akinobu.mita@gmail.com \
--to=akinobu.mita@gmail.com \
--cc=axboe@kernel.dk \
--cc=linux-kernel@vger.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
Powered by JetHome