btrfs: zoned: do not account freed region of read-only block group as zone_unusable

We migrate zone unusable bytes to read-only bytes when a block group is
set to read-only, and account all the free region as bytes_readonly.
Thus, we should not increase block_group->zone_unusable when the block
group is read-only.

Fixes: 169e0da ("btrfs: zoned: track unusable bytes for zones")
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>

Author: naota

fs/btrfs/free-space-cache.c

@@ -2555,7 +2555,12 @@ static int __btrfs_add_free_space_zoned(struct btrfs_block_group *block_group,
 	to_unusable = size - to_free;
 
 	ctl->free_space += to_free;
-	block_group->zone_unusable += to_unusable;
+	/*
+	 * If the block group is read-only, we should account freed space into
+	 * bytes_readonly.
+	 */
+	if (!block_group->ro)
+		block_group->zone_unusable += to_unusable;
 	spin_unlock(&ctl->tree_lock);
 	if (!used) {
 		spin_lock(&block_group->lock);