block, bfq: do not insert oom queue into position tree
authorPaolo Valente <paolo.valente@linaro.org>
Mon, 3 Feb 2020 10:40:55 +0000 (11:40 +0100)
committerJens Axboe <axboe@kernel.dk>
Mon, 3 Feb 2020 13:58:15 +0000 (06:58 -0700)
BFQ maintains an ordered list, implemented with an RB tree, of
head-request positions of non-empty bfq_queues. This position tree,
inherited from CFQ, is used to find bfq_queues that contain I/O close
to each other. BFQ merges these bfq_queues into a single shared queue,
if this boosts throughput on the device at hand.

There is however a special-purpose bfq_queue that does not participate
in queue merging, the oom bfq_queue. Yet, also this bfq_queue could be
wrongly added to the position tree. So bfqq_find_close() could return
the oom bfq_queue, which is a source of further troubles in an
out-of-memory situation. This commit prevents the oom bfq_queue from
being inserted into the position tree.

Tested-by: Patrick Dung <patdung100@gmail.com>
Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
block/bfq-iosched.c

index 55d4328..15dfb08 100644 (file)
@@ -613,6 +613,10 @@ bfq_pos_tree_add_move(struct bfq_data *bfqd, struct bfq_queue *bfqq)
                bfqq->pos_root = NULL;
        }
 
+       /* oom_bfqq does not participate in queue merging */
+       if (bfqq == &bfqd->oom_bfqq)
+               return;
+
        /*
         * bfqq cannot be merged any longer (see comments in
         * bfq_setup_cooperator): no point in adding bfqq into the