blkcg: make sure blkg_lookup() returns %NULL if @q is bypassing

Currently, blkg_lookup() doesn't check @q bypass state.  This patch
updates blk_queue_bypass_start() to do synchronize_rcu() before
returning and updates blkg_lookup() to check blk_queue_bypass() and
return %NULL if bypassing.  This ensures blkg_lookup() returns %NULL
if @q is bypassing.

This is to guarantee that nobody is accessing policy data while @q is
bypassing, which is necessary to allow replacing blkio_cgroup->pd[] in
place on policy [de]activation.

v2: Added more comments explaining bypass guarantees as suggested by
    Vivek.

v3: Added more comments explaining why there's no synchronize_rcu() in
    blk_cleanup_queue() as suggested by Vivek.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

1 files changed, 13 insertions, 2 deletions

diff --git a/block/blk-core.c b/block/blk-core.c
index 991c1d6ef245..f2db628aa509 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -416,7 +416,8 @@ void blk_drain_queue(struct request_queue *q, bool drain_all)
  * In bypass mode, only the dispatch FIFO queue of @q is used.  This
  * function makes @q enter bypass mode and drains all requests which were
  * throttled or issued before.  On return, it's guaranteed that no request
- * is being throttled or has ELVPRIV set.
+ * is being throttled or has ELVPRIV set and blk_queue_bypass() %true
+ * inside queue or RCU read lock.
  */
 void blk_queue_bypass_start(struct request_queue *q)
 {
@@ -426,6 +427,8 @@ void blk_queue_bypass_start(struct request_queue *q)
 	spin_unlock_irq(q->queue_lock);
 
 	blk_drain_queue(q, false);
+	/* ensure blk_queue_bypass() is %true inside RCU read lock */
+	synchronize_rcu();
 }
 EXPORT_SYMBOL_GPL(blk_queue_bypass_start);
 
@@ -462,7 +465,15 @@ void blk_cleanup_queue(struct request_queue *q)
 
 	spin_lock_irq(lock);
 
-	/* dead queue is permanently in bypass mode till released */
+	/*
+	 * Dead queue is permanently in bypass mode till released.  Note
+	 * that, unlike blk_queue_bypass_start(), we aren't performing
+	 * synchronize_rcu() after entering bypass mode to avoid the delay
+	 * as some drivers create and destroy a lot of queues while
+	 * probing.  This is still safe because blk_release_queue() will be
+	 * called only after the queue refcnt drops to zero and nothing,
+	 * RCU or not, would be traversing the queue by then.
+	 */
 	q->bypass_depth++;
 	queue_flag_set(QUEUE_FLAG_BYPASS, q);