block: fix revalidate performance regression

commit 47fe1c3064 upstream.

The scsi driver function sd_read_block_characteristics() always calls
disk_set_zoned() to a disk zoned model correctly, in case the device
model changed. This is done even for regular disks to set the zoned
model to BLK_ZONED_NONE and free any zone related resources if the drive
previously was zoned.

This behavior significantly impact the time it takes to revalidate disks
on a large system as the call to disk_clear_zone_settings() done from
disk_set_zoned() for the BLK_ZONED_NONE case results in the device
request queued to be frozen, even if there are no zone resources to
free.

Avoid this overhead for non-zoned devices by not calling
disk_clear_zone_settings() in disk_set_zoned() if the device model
was already set to BLK_ZONED_NONE, which is always the case for regular
devices.

Reported by: Brian Bunker <brian@purestorage.com>

Fixes: 508aebb805 ("block: introduce blk_queue_clear_zone_settings()")
Cc: stable@vger.kernel.org
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20230529073237.1339862-1-dlemoal@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This commit is contained in:
Damien Le Moal 2023-05-29 16:32:37 +09:00 committed by Justin Hammond
parent a0b56b1f13
commit 7a3e707916

View file

@ -875,6 +875,7 @@ static bool disk_has_partitions(struct gendisk *disk)
void blk_queue_set_zoned(struct gendisk *disk, enum blk_zoned_model model)
{
struct request_queue *q = disk->queue;
unsigned int old_model = q->limits.zoned;
switch (model) {
case BLK_ZONED_HM:
@ -912,7 +913,7 @@ void blk_queue_set_zoned(struct gendisk *disk, enum blk_zoned_model model)
*/
blk_queue_zone_write_granularity(q,
queue_logical_block_size(q));
} else {
} else if (old_model != BLK_ZONED_NONE) {
blk_queue_clear_zone_settings(q);
}
}