block: Fix a race between the cgroup code and request queue initialization

Initialize the request queue lock earlier such that the following
race can no longer occur:

blk_init_queue_node()             blkcg_print_blkgs()
  blk_alloc_queue_node (1)
    q->queue_lock = &q->__queue_lock (2)
    blkcg_init_queue(q) (3)
                                    spin_lock_irq(blkg->q->queue_lock) (4)
  q->queue_lock = lock (5)
                                    spin_unlock_irq(blkg->q->queue_lock) (6)

(1) allocate an uninitialized queue;
(2) initialize queue_lock to its default internal lock;
(3) initialize blkcg part of request queue, which will create blkg and
    then insert it to blkg_list;
(4) traverse blkg_list and find the created blkg, and then take its
    queue lock, here it is the default *internal lock*;
(5) *race window*, now queue_lock is overridden with *driver specified
    lock*;
(6) now unlock *driver specified lock*, not the locked *internal lock*,
    unlock balance breaks.

The changes in this patch are as follows:
- Move the .queue_lock initialization from blk_init_queue_node() into
  blk_alloc_queue_node().
- Only override the .queue_lock pointer for legacy queues because it
  is not useful for blk-mq queues to override this pointer.
- For all all block drivers that initialize .queue_lock explicitly,
  change the blk_alloc_queue() call in the driver into a
  blk_alloc_queue_node() call and remove the explicit .queue_lock
  initialization. Additionally, initialize the spin lock that will
  be used as queue lock earlier if necessary.

Reported-by: Joseph Qi <joseph.qi@linux.alibaba.com>
Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
Reviewed-by: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Philipp Reisner <philipp.reisner@linbit.com>
Cc: Ulf Hansson <ulf.hansson@linaro.org>
Cc: Kees Cook <keescook@chromium.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

1 files changed, 16 insertions, 8 deletions

diff --git a/block/blk-core.c b/block/blk-core.c
index e873a24bf82d1..41c74b37be85e 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -888,6 +888,19 @@ static void blk_rq_timed_out_timer(struct timer_list *t)
 	kblockd_schedule_work(&q->timeout_work);
 }
 
+/**
+ * blk_alloc_queue_node - allocate a request queue
+ * @gfp_mask: memory allocation flags
+ * @node_id: NUMA node to allocate memory from
+ * @lock: For legacy queues, pointer to a spinlock that will be used to e.g.
+ *        serialize calls to the legacy .request_fn() callback. Ignored for
+ *	  blk-mq request queues.
+ *
+ * Note: pass the queue lock as the third argument to this function instead of
+ * setting the queue lock pointer explicitly to avoid triggering a sporadic
+ * crash in the blkcg code. This function namely calls blkcg_init_queue() and
+ * the queue lock pointer must be set before blkcg_init_queue() is called.
+ */
 struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id,
 					   spinlock_t *lock)
 {
@@ -940,11 +953,8 @@ struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id,
 	mutex_init(&q->sysfs_lock);
 	spin_lock_init(&q->__queue_lock);
 
-	/*
-	 * By default initialize queue_lock to internal lock and driver can
-	 * override it later if need be.
-	 */
-	q->queue_lock = &q->__queue_lock;
+	if (!q->mq_ops)
+		q->queue_lock = lock ? : &q->__queue_lock;
 
 	/*
 	 * A queue starts its life with bypass turned on to avoid
@@ -1031,13 +1041,11 @@ blk_init_queue_node(request_fn_proc *rfn, spinlock_t *lock, int node_id)
 {
 	struct request_queue *q;
 
-	q = blk_alloc_queue_node(GFP_KERNEL, node_id, NULL);
+	q = blk_alloc_queue_node(GFP_KERNEL, node_id, lock);
 	if (!q)
 		return NULL;
 
 	q->request_fn = rfn;
-	if (lock)
-		q->queue_lock = lock;
 	if (blk_init_allocated_queue(q) < 0) {
 		blk_cleanup_queue(q);
 		return NULL;