virtio_blk: fix the discard_granularity and discard_alignment queue limits
authorChristoph Hellwig <hch@lst.de>
Mon, 18 Apr 2022 04:53:07 +0000 (06:53 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 9 Jun 2022 08:22:53 +0000 (10:22 +0200)
[ Upstream commit 62952cc5bccd89b76d710de1d0b43244af0f2903 ]

The discard_alignment queue limit is named a bit misleading means the
offset into the block device at which the discard granularity starts.

On the other hand the discard_sector_alignment from the virtio 1.1 looks
similar to what Linux uses as discard granularity (even if not very well
described):

  "discard_sector_alignment can be used by OS when splitting a request
   based on alignment. "

And at least qemu does set it to the discard granularity.

So stop setting the discard_alignment and use the virtio
discard_sector_alignment to set the discard granularity.

Fixes: 1f23816b8eb8 ("virtio_blk: add discard and write zeroes support")
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20220418045314.360785-5-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>
drivers/block/virtio_blk.c

index c0b8a26..c05138a 100644 (file)
@@ -855,11 +855,12 @@ static int virtblk_probe(struct virtio_device *vdev)
                blk_queue_io_opt(q, blk_size * opt_io_size);
 
        if (virtio_has_feature(vdev, VIRTIO_BLK_F_DISCARD)) {
-               q->limits.discard_granularity = blk_size;
-
                virtio_cread(vdev, struct virtio_blk_config,
                             discard_sector_alignment, &v);
-               q->limits.discard_alignment = v ? v << SECTOR_SHIFT : 0;
+               if (v)
+                       q->limits.discard_granularity = v << SECTOR_SHIFT;
+               else
+                       q->limits.discard_granularity = blk_size;
 
                virtio_cread(vdev, struct virtio_blk_config,
                             max_discard_sectors, &v);