dm writecache: fix performance degradation in ssd mode

commit cb72848 upstream.

Fix a thinko in ssd_commit_superblock. region.count is in sectors, not
bytes. This bug doesn't corrupt data, but it causes performance
degradation.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Fixes: dc8a01a ("dm writecache: optimize superblock write")
Cc: stable@vger.kernel.org # v5.7+
Reported-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drivers/md/dm-writecache.c

@@ -523,7 +523,7 @@ static void ssd_commit_superblock(struct dm_writecache *wc)
 
 	region.bdev = wc->ssd_dev->bdev;
 	region.sector = 0;
-	region.count = PAGE_SIZE;
+	region.count = PAGE_SIZE >> SECTOR_SHIFT;
 
 	if (unlikely(region.sector + region.count > wc->metadata_sectors))
 		region.count = wc->metadata_sectors - region.sector;