btrfs: zoned: fix comparison of alloc_offset vs meta_write_pointer
authorNaohiro Aota <naohiro.aota@wdc.com>
Wed, 4 May 2022 23:12:48 +0000 (16:12 -0700)
committerDavid Sterba <dsterba@suse.com>
Tue, 17 May 2022 18:15:25 +0000 (20:15 +0200)
The block_group->alloc_offset is an offset from the start of the block
group. OTOH, the ->meta_write_pointer is an address in the logical
space. So, we should compare the alloc_offset shifted with the
block_group->start.

Fixes: 99777e1aeef4 ("btrfs: zoned: implement active zone tracking")
CC: stable@vger.kernel.org # 5.16+
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>
fs/btrfs/zoned.c

index 488577efd8267ee1d2bc38957b6c0f8e66c30ea4..b60767492b3caa2526542b4f0e862c2ee0d71bda 100644 (file)
@@ -1890,7 +1890,7 @@ static int do_zone_finish(struct btrfs_block_group *block_group, bool fully_writ
        /* Check if we have unwritten allocated space */
        if ((block_group->flags &
             (BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_SYSTEM)) &&
-           block_group->alloc_offset > block_group->meta_write_pointer) {
+           block_group->start + block_group->alloc_offset > block_group->meta_write_pointer) {
                spin_unlock(&block_group->lock);
                return -EAGAIN;
        }