blk-mq: use the right hctx when getting a driver tag fails

While dispatching requests, if we fail to get a driver tag, we mark the
hardware queue as waiting for a tag and put the requests on a
hctx->dispatch list to be run later when a driver tag is freed. However,
blk_mq_dispatch_rq_list() may dispatch requests from multiple hardware
queues if using a single-queue scheduler with a multiqueue device. If
blk_mq_get_driver_tag() fails, it doesn't update the hardware queue we
are processing. This means we end up using the hardware queue of the
previous request, which may or may not be the same as that of the
current request. If it isn't, the wrong hardware queue will end up
waiting for a tag, and the requests will be on the wrong dispatch list,
leading to a hang.

The fix is twofold:

1. Make sure we save which hardware queue we were trying to get a
   request for in blk_mq_get_driver_tag() regardless of whether it
   succeeds or not.
2. Make blk_mq_dispatch_rq_list() take a request_queue instead of a
   blk_mq_hw_queue to make it clear that it must handle multiple
   hardware queues, since I've already messed this up on a couple of
   occasions.

This didn't appear in testing with nvme and mq-deadline because nvme has
more driver tags than the default number of scheduler tags. However,
with the blk_mq_update_nr_hw_queues() fix, it showed up with nbd.

Signed-off-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>

1 files changed, 13 insertions, 12 deletions

diff --git a/block/blk-mq.c b/block/blk-mq.c
index 935f2cc7c8c3..828f4bd193f2 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -845,12 +845,8 @@ bool blk_mq_get_driver_tag(struct request *rq, struct blk_mq_hw_ctx **hctx,
                 .flags = wait ? 0 : BLK_MQ_REQ_NOWAIT,
         };
 
-        if (rq->tag != -1) {
-done:
-                if (hctx)
-                        *hctx = data.hctx;
-                return true;
-        }
+        if (rq->tag != -1)
+                goto done;
 
         if (blk_mq_tag_is_reserved(data.hctx->sched_tags, rq->internal_tag))
                 data.flags |= BLK_MQ_REQ_RESERVED;
@@ -862,10 +858,12 @@ done:
                         atomic_inc(&data.hctx->nr_active);
                 }
                 data.hctx->tags->rqs[rq->tag] = rq;
-                goto done;
         }
 
-        return false;
+done:
+        if (hctx)
+                *hctx = data.hctx;
+        return rq->tag != -1;
 }
 
 static void __blk_mq_put_driver_tag(struct blk_mq_hw_ctx *hctx,
@@ -962,14 +960,17 @@ static bool blk_mq_dispatch_wait_add(struct blk_mq_hw_ctx *hctx)
         return true;
 }
 
-bool blk_mq_dispatch_rq_list(struct blk_mq_hw_ctx *hctx, struct list_head *list)
+bool blk_mq_dispatch_rq_list(struct request_queue *q, struct list_head *list)
 {
-        struct request_queue *q = hctx->queue;
+        struct blk_mq_hw_ctx *hctx;
         struct request *rq;
         LIST_HEAD(driver_list);
         struct list_head *dptr;
         int errors, queued, ret = BLK_MQ_RQ_QUEUE_OK;
 
+        if (list_empty(list))
+                return false;
+
         /*
          * Start off with dptr being NULL, so we start the first request
          * immediately, even if we have more pending.
@@ -980,7 +981,7 @@ bool blk_mq_dispatch_rq_list(struct blk_mq_hw_ctx *hctx, struct list_head *list)
          * Now process all the entries, sending them to the driver.
          */
         errors = queued = 0;
-        while (!list_empty(list)) {
+        do {
                 struct blk_mq_queue_data bd;
 
                 rq = list_first_entry(list, struct request, queuelist);
@@ -1051,7 +1052,7 @@ bool blk_mq_dispatch_rq_list(struct blk_mq_hw_ctx *hctx, struct list_head *list)
                  */
                 if (!dptr && list->next != list->prev)
                         dptr = &driver_list;
-        }
+        } while (!list_empty(list));
 
         hctx->dispatched[queued_to_index(queued)]++;