sbitmap: Advance the queue index before waking up a queue
authorGabriel Krisman Bertazi <krisman@suse.de>
Tue, 15 Nov 2022 22:45:51 +0000 (17:45 -0500)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Fri, 10 Mar 2023 08:34:34 +0000 (09:34 +0100)
commit 976570b4ecd30d3ec6e1b0910da8e5edc591f2b6 upstream.

When a queue is awaken, the wake_index written by sbq_wake_ptr currently
keeps pointing to the same queue.  On the next wake up, it will thus
retry the same queue, which is unfair to other queues, and can lead to
starvation.  This patch, moves the index update to happen before the
queue is returned, such that it will now try a different queue first on
the next wake up, improving fairness.

Fixes: 4f8126bb2308 ("sbitmap: Use single per-bitmap counting to wake up queued tags")
Reported-by: Jan Kara <jack@suse.cz>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Gabriel Krisman Bertazi <krisman@suse.de>
Link: https://lore.kernel.org/r/20221115224553.23594-2-krisman@suse.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
lib/sbitmap.c

index e918cd8..443accf 100644 (file)
@@ -566,13 +566,19 @@ static struct sbq_wait_state *sbq_wake_ptr(struct sbitmap_queue *sbq)
        for (i = 0; i < SBQ_WAIT_QUEUES; i++) {
                struct sbq_wait_state *ws = &sbq->ws[wake_index];
 
+               /*
+                * Advance the index before checking the current queue.
+                * It improves fairness, by ensuring the queue doesn't
+                * need to be fully emptied before trying to wake up
+                * from the next one.
+                */
+               wake_index = sbq_index_inc(wake_index);
+
                if (waitqueue_active(&ws->wait)) {
                        if (wake_index != atomic_read(&sbq->wake_index))
                                atomic_set(&sbq->wake_index, wake_index);
                        return ws;
                }
-
-               wake_index = sbq_index_inc(wake_index);
        }
 
        return NULL;