dm zoned: Fix reclaim zone selection

When dm zoned has multiple devices, random zones are never selected for
reclaim if all reserved sequential write zones are in use and no
sequential write required zones can be selected for reclaim. This can
lead to deadlocks as selecting a cache zone allows reclaiming a
sequential zone, ensuring forward progress.

Fix this by always defaulting to selecting a random zone when no
sequential write required zone can be selected.

[Damien: fix commit message]

Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>
Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>

diff --git a/drivers/md/dm-zoned-metadata.c b/drivers/md/dm-zoned-metadata.c
index b40e643..5cf6f5f 100644 (file)
--- a/drivers/md/dm-zoned-metadata.c
+++ b/drivers/md/dm-zoned-metadata.c
@@ -2026,7 +2026,7 @@ static struct dm_zone *dmz_get_seq_zone_for_reclaim(struct dmz_metadata *zmd,
 struct dm_zone *dmz_get_zone_for_reclaim(struct dmz_metadata *zmd,
                                         unsigned int dev_idx, bool idle)
 {
-       struct dm_zone *zone;
+       struct dm_zone *zone = NULL;
 
        /*
         * Search for a zone candidate to reclaim: 2 cases are possible.
@@ -2039,7 +2039,7 @@ struct dm_zone *dmz_get_zone_for_reclaim(struct dmz_metadata *zmd,
        dmz_lock_map(zmd);
        if (list_empty(&zmd->reserved_seq_zones_list))
                zone = dmz_get_seq_zone_for_reclaim(zmd, dev_idx);
-       else
+       if (!zone)
                zone = dmz_get_rnd_zone_for_reclaim(zmd, dev_idx, idle);
        dmz_unlock_map(zmd);