[PATCH] nvme: respect user max_discard_sectors setting

[PATCH] nvme: respect user max_discard_sectors setting

[Keith Busch]

From: Keith Busch <kbusch@kernel.org>

A user can change that attribute to 0 through sysfs, but they can't
change the hw limit. Check the hw limit instead to know if the driver
previously set up discard capabilities.

Fixes: 70200574cc229 ("block: remove QUEUE_FLAG_DISCARD")
Signed-off-by: Keith Busch <kbusch@kernel.org>
---
 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 50818dbcfa1ae..8d56c102974ee 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -1748,7 +1748,7 @@ static void nvme_config_discard(struct nvme_ctrl *ctrl, struct gendisk *disk,
 	 * updating the hardware limits when overridden through sysfs.  This is
 	 * harmless because discard limits in NVMe are purely advisory.
 	 */
-	if (queue->limits.max_discard_sectors)
+	if (queue->limits.max_hw_discard_sectors)
 		return;
 
 	blk_queue_max_discard_sectors(queue, max_discard_sectors);
-- 
2.34.1

Re: [PATCH] nvme: respect user max_discard_sectors setting

[Christoph Hellwig]

On Thu, Jan 04, 2024 at 11:12:44AM -0800, Keith Busch wrote:
> From: Keith Busch <kbusch@kernel.org>
> 
> A user can change that attribute to 0 through sysfs, but they can't
> change the hw limit. Check the hw limit instead to know if the driver
> previously set up discard capabilities.

blk_queue_max_discard_sectors updates the hardware and user limit.
It's a bit of a mess, but Jens didn't quite like my half-hearted fix
last time.  If we want to do this properly we probably need a separate
user limit and recalculate the effective limit when either the hw or
user limit changes.  For early in the 6.9 merge window I plan to redo
how all the limits are set, and that should give us a better opportunity
for these kinds of things

Re: [PATCH] nvme: respect user max_discard_sectors setting

[Keith Busch]

On Fri, Jan 05, 2024 at 05:34:01AM +0100, Christoph Hellwig wrote:
> On Thu, Jan 04, 2024 at 11:12:44AM -0800, Keith Busch wrote:
> > From: Keith Busch <kbusch@kernel.org>
> > 
> > A user can change that attribute to 0 through sysfs, but they can't
> > change the hw limit. Check the hw limit instead to know if the driver
> > previously set up discard capabilities.
> 
> blk_queue_max_discard_sectors updates the hardware and user limit.
> It's a bit of a mess, but Jens didn't quite like my half-hearted fix
> last time.  If we want to do this properly we probably need a separate
> user limit and recalculate the effective limit when either the hw or
> user limit changes.  For early in the 6.9 merge window I plan to redo
> how all the limits are set, and that should give us a better opportunity
> for these kinds of things

Cool, I look forward to seeing what improvements you have in mind for
this error prone area. In the meantime, is this patch okay with you or
is that a 'Nak'?

Re: [PATCH] nvme: respect user max_discard_sectors setting

[Christoph Hellwig]

On Fri, Jan 05, 2024 at 09:07:40AM -0700, Keith Busch wrote:
> > blk_queue_max_discard_sectors updates the hardware and user limit.
> > It's a bit of a mess, but Jens didn't quite like my half-hearted fix
> > last time.  If we want to do this properly we probably need a separate
> > user limit and recalculate the effective limit when either the hw or
> > user limit changes.  For early in the 6.9 merge window I plan to redo
> > how all the limits are set, and that should give us a better opportunity
> > for these kinds of things
> 
> Cool, I look forward to seeing what improvements you have in mind for
> this error prone area. In the meantime, is this patch okay with you or
> is that a 'Nak'?

It's not a hard NAK, but I think this actually leaves us in a slightly
worse position than before.  So in doubt I'd not merge it.