[PATCH AUTOSEL for 4.15 057/124] blk-mq: fix race between updating nr_hw_queues and switching io sched

From: Sasha Levin
Date: Mon Mar 19 2018 - 18:28:59 EST


From: Ming Lei <ming.lei@xxxxxxxxxx>

[ Upstream commit fb350e0ad99359768e1e80b4784692031ec340e4 ]

In both elevator_switch_mq() and blk_mq_update_nr_hw_queues(), sched tags
can be allocated, and q->nr_hw_queue is used, and race is inevitable, for
example: blk_mq_init_sched() may trigger use-after-free on hctx, which is
freed in blk_mq_realloc_hw_ctxs() when nr_hw_queues is decreased.

This patch fixes the race be holding q->sysfs_lock.

Reviewed-by: Christoph Hellwig <hch@xxxxxx>
Reported-by: Yi Zhang <yi.zhang@xxxxxxxxxx>
Tested-by: Yi Zhang <yi.zhang@xxxxxxxxxx>
Signed-off-by: Ming Lei <ming.lei@xxxxxxxxxx>
Signed-off-by: Jens Axboe <axboe@xxxxxxxxx>
Signed-off-by: Sasha Levin <alexander.levin@xxxxxxxxxxxxx>
---
block/blk-mq.c | 4 ++++
1 file changed, 4 insertions(+)

diff --git a/block/blk-mq.c b/block/blk-mq.c
index 4385c5cbf57b..1372ad3a21d3 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -2388,6 +2388,9 @@ static void blk_mq_realloc_hw_ctxs(struct blk_mq_tag_set *set,
struct blk_mq_hw_ctx **hctxs = q->queue_hw_ctx;

blk_mq_sysfs_unregister(q);
+
+ /* protect against switching io scheduler */
+ mutex_lock(&q->sysfs_lock);
for (i = 0; i < set->nr_hw_queues; i++) {
int node;

@@ -2432,6 +2435,7 @@ static void blk_mq_realloc_hw_ctxs(struct blk_mq_tag_set *set,
}
}
q->nr_hw_queues = i;
+ mutex_unlock(&q->sysfs_lock);
blk_mq_sysfs_register(q);
}

--
2.14.1