[PATCH] blk_queue_start_tag() shared map race fix

If we share the tag map between two or more queues, then we cannot
use __set_bit() to set the bit. In fact we need to make sure we
atomically acquire this tag, so loop using test_and_set_bit() to
protect from that.

Noticed by Mike Christie <michaelc@cs.wisc.edu>

Signed-off-by: Jens Axboe <axboe@kernel.dk>

diff --git a/block/ll_rw_blk.c b/block/ll_rw_blk.c
index f757ed4132147a27beb7c18e6bdd4e916b68cf27..83425fb3c8dba6e2b62122aaa6a02a25d2436ab4 100644 (file)
--- a/block/ll_rw_blk.c
+++ b/block/ll_rw_blk.c
@@ -1171,11 +1171,16 @@ int blk_queue_start_tag(request_queue_t *q, struct request *rq)
                BUG();
        }
 
-       tag = find_first_zero_bit(bqt->tag_map, bqt->max_depth);
-       if (tag >= bqt->max_depth)
-               return 1;
+       /*
+        * Protect against shared tag maps, as we may not have exclusive
+        * access to the tag map.
+        */
+       do {
+               tag = find_first_zero_bit(bqt->tag_map, bqt->max_depth);
+               if (tag >= bqt->max_depth)
+                       return 1;
 
-       __set_bit(tag, bqt->tag_map);
+       } while (test_and_set_bit(tag, bqt->tag_map));
 
        rq->cmd_flags |= REQ_QUEUED;
        rq->tag = tag;