On Wed, 2020-05-27 at 08:22 +0200, Hannes Reinecke wrote:
> Prefer full zones when selecting the next zone for reclaim.
>
> Signed-off-by: Hannes Reinecke <hare@xxxxxxx>
> ---
> drivers/md/dm-zoned-metadata.c | 9 ++++++++-
> 1 file changed, 8 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/md/dm-zoned-metadata.c b/drivers/md/dm-zoned-metadata.c
> index b89b3d3b9ec9..f161ef4e3d3d 100644
> --- a/drivers/md/dm-zoned-metadata.c
> +++ b/drivers/md/dm-zoned-metadata.c
> @@ -1936,7 +1936,7 @@ static struct dm_zone *dmz_get_rnd_zone_for_reclaim(struct dmz_metadata *zmd,
> unsigned int idx, bool idle)
> {
> struct dm_zone *dzone = NULL;
> - struct dm_zone *zone;
> + struct dm_zone *zone, *last = NULL;
> struct list_head *zone_list;
>
> /* If we have cache zones select from the cache zone list */
> @@ -1953,6 +1953,13 @@ static struct dm_zone *dmz_get_rnd_zone_for_reclaim(struct dmz_metadata *zmd,
> dzone = zone->bzone;
> if (dzone->dev->dev_idx != idx)
> continue;
> + if (!last) {
> + last = dzone;
> + continue;
> + }
> + if (last->weight < dzone->weight)
> + continue;
> + dzone = last;
> } else
> dzone = zone;
> if (dmz_lock_zone_reclaim(dzone))
If all random/cache zones are used but none of them satisfy the
condition last->weight < dzone->weight, we may end up starving reclaim
and having user IOs accessing a new chunk wait a loooong time, if not
forever, No ? I agree that aiming at reclaim of full zones first is
more efficient, but we need a fallback to ensure forward progress.
--
Damien Le Moal
Western Digital Research
--
dm-devel mailing list
dm-devel@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/dm-devel
