blk-mq: fix race condition in active queue accounting

If multiple CPUs are sharing the same hardware queue, it can
cause leak in the active queue counter tracking when __blk_mq_tag_busy()
is executed simultaneously.

Fixes: ee78ec1077d3 ("blk-mq: blk_mq_tag_busy is no need to return a value")
Signed-off-by: Tian Lan <tian.lan@twosigma.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: John Garry <john.g.garry@oracle.com>
Link: https://lore.kernel.org/r/20230522210555.794134-1-tilan7663@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

diff --git a/block/blk-mq-tag.c b/block/blk-mq-tag.c
index d6af9d4..dfd81ca 100644 (file)
--- a/block/blk-mq-tag.c
+++ b/block/blk-mq-tag.c
@@ -39,16 +39,20 @@ void __blk_mq_tag_busy(struct blk_mq_hw_ctx *hctx)
 {
        unsigned int users;
 
+       /*
+        * calling test_bit() prior to test_and_set_bit() is intentional,
+        * it avoids dirtying the cacheline if the queue is already active.
+        */
        if (blk_mq_is_shared_tags(hctx->flags)) {
                struct request_queue *q = hctx->queue;
 
-               if (test_bit(QUEUE_FLAG_HCTX_ACTIVE, &q->queue_flags))
+               if (test_bit(QUEUE_FLAG_HCTX_ACTIVE, &q->queue_flags) ||
+                   test_and_set_bit(QUEUE_FLAG_HCTX_ACTIVE, &q->queue_flags))
                        return;
-               set_bit(QUEUE_FLAG_HCTX_ACTIVE, &q->queue_flags);
        } else {
-               if (test_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
+               if (test_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state) ||
+                   test_and_set_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
                        return;
-               set_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state);
        }
 
        users = atomic_inc_return(&hctx->tags->active_queues);