Re: [PATCH v6 6/6] nbd: fix uaf in nbd_handle_reply()

On 2021/09/15 16:20, Christoph Hellwig wrote:

On Wed, Sep 15, 2021 at 04:15:37PM +0800, Yu Kuai wrote:

quoted

+++ b/block/blk-core.c

@@ -489,6 +489,7 @@ void blk_queue_exit(struct request_queue *q)
  {
  	percpu_ref_put(&q->q_usage_counter);
  }
+EXPORT_SYMBOL(blk_queue_exit);

These needs to be an EXPORT_SYMBOL_GPL.  But more importantly it
needs to be a separate properly documented patch, and this function
needs to grow a kerneldoc comment as well.

quoted

+		/*
+		 * Get q_usage_counter can prevent accessing freed request
+		 * through blk_mq_tag_to_rq() in nbd_handle_reply(). If
+		 * q_usage_counter is zero, then no request is inflight, which
+		 * means something is wrong since we expect to find a request to
+		 * complete here.
+		 */
+		if (!percpu_ref_tryget(&q->q_usage_counter)) {
+			dev_err(disk_to_dev(nbd->disk), "%s: no io inflight\n",
+				__func__);
+			break;
+		}

And this needs a properly documented wrapper as well.

quoted

+
  		cmd = nbd_handle_reply(nbd, args->index, &reply);
-		if (IS_ERR(cmd))
+		if (IS_ERR(cmd)) {
+			blk_queue_exit(q);
  			break;
+		}
  
  		rq = blk_mq_rq_from_pdu(cmd);
  		if (likely(!blk_should_fake_timeout(rq->q)))
  			blk_mq_complete_request(rq);
+		blk_queue_exit(q);

That being said I can't say I like how this exposed block layer
internals.  We don't really need a reference to the queue here
anywhere, you just use it as a dumb debug check.  If we really want to
reuse (abuse?) q_usage_counter a helper to just grab a reference and
immediately drop it might be a better fit.
.

Hi,

The uaf is because blk_mq_sched_free_requests() can concurrent with
nbd_read_stat(), and hold the ref during nbd_read_stat() can break the
concurrency, thus the ref can't be dropped immediately.

I'll use percpu_ref_put() directly as Ming suggestted.

Thanks,
Kuai