[PATCHv2] nvme: skip noiob for zoned devices

Zoned block devices reuse the chunk_sectors queue limit to define zone
boundaries. If a such a device happens to also report an optimal
boundary, do not use that to define the chunk_sectors as that may
intermittently interfere with io splitting and zone size queries.

Signed-off-by: Keith Busch <kbusch@kernel.org>
---
v1->v2: Fixed the if condition to check for *not* zoned.

 drivers/nvme/host/core.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index 88cff309d8e4..4e0225d85e70 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c

@@ -2075,7 +2075,7 @@ static int __nvme_revalidate_disk(struct gendisk *disk, struct nvme_id_ns *id)
 		}
 	}
 
-	if (iob)
+	if (iob && !blk_queue_is_zoned(ns->queue))
 		blk_queue_chunk_sectors(ns->queue, rounddown_pow_of_two(iob));
 	nvme_update_disk_info(disk, ns, id);
 #ifdef CONFIG_NVME_MULTIPATH

-- 
2.24.1


_______________________________________________
Linux-nvme mailing list
Linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

Queued for nvme-5.9-rc

_______________________________________________
Linux-nvme mailing list
Linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

On Fri, Aug 07, 2020 at 09:32:35AM -0700, Keith Busch wrote:

Zoned block devices reuse the chunk_sectors queue limit to define zone
boundaries. If a such a device happens to also report an optimal
boundary, do not use that to define the chunk_sectors as that may
intermittently interfere with io splitting and zone size queries.

Instead of skipping shouldn't we use the max of noiob and the zone
size, even if that is mostly theoretical?

---end quoted text---

_______________________________________________
Linux-nvme mailing list
Linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

On Mon, Aug 10, 2020 at 02:33:13PM +0200, Christoph Hellwig wrote:

On Fri, Aug 07, 2020 at 09:32:35AM -0700, Keith Busch wrote:

quoted

Zoned block devices reuse the chunk_sectors queue limit to define zone
boundaries. If a such a device happens to also report an optimal
boundary, do not use that to define the chunk_sectors as that may
intermittently interfere with io splitting and zone size queries.

Instead of skipping shouldn't we use the max of noiob and the zone
size, even if that is mostly theoretical?

No, I don't think so. If noiob is larger, we'll return the wrong value
for blk_queue_zone_sectors() and blkdev_nr_zones(), which will have
undesirable consequences.

_______________________________________________
Linux-nvme mailing list
Linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

On Mon, Aug 10, 2020 at 07:56:22AM -0700, Keith Busch wrote:

On Mon, Aug 10, 2020 at 02:33:13PM +0200, Christoph Hellwig wrote:

quoted

On Fri, Aug 07, 2020 at 09:32:35AM -0700, Keith Busch wrote:

quoted

Zoned block devices reuse the chunk_sectors queue limit to define zone
boundaries. If a such a device happens to also report an optimal
boundary, do not use that to define the chunk_sectors as that may
intermittently interfere with io splitting and zone size queries.

Instead of skipping shouldn't we use the max of noiob and the zone
size, even if that is mostly theoretical?

No, I don't think so. If noiob is larger, we'll return the wrong value
for blk_queue_zone_sectors() and blkdev_nr_zones(), which will have
undesirable consequences.

Then we at least need to warn about a larger noiob.  And add a
comment that we are ignoring it deliberately.

_______________________________________________
Linux-nvme mailing list
Linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

quoted

quoted

On Fri, Aug 07, 2020 at 09:32:35AM -0700, Keith Busch wrote:

quoted

Zoned block devices reuse the chunk_sectors queue limit to define zone
boundaries. If a such a device happens to also report an optimal
boundary, do not use that to define the chunk_sectors as that may
intermittently interfere with io splitting and zone size queries.

Instead of skipping shouldn't we use the max of noiob and the zone
size, even if that is mostly theoretical?

No, I don't think so. If noiob is larger, we'll return the wrong value
for blk_queue_zone_sectors() and blkdev_nr_zones(), which will have
undesirable consequences.

Then we at least need to warn about a larger noiob.  And add a
comment that we are ignoring it deliberately.

I'm fine with that, Keith if you agree you can send a patch
and I'll fold it into the original one.

_______________________________________________
Linux-nvme mailing list
Linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme