Btrfs: make backref walking code handle skinny metadata
authorJosef Bacik <jbacik@fusionio.com>
Fri, 28 Jun 2013 17:11:22 +0000 (13:11 -0400)
committerJosef Bacik <jbacik@fusionio.com>
Tue, 2 Jul 2013 15:51:02 +0000 (11:51 -0400)
I missed fixing the backref stuff when I introduced the skinny metadata.  If you
try and do things like snapshot aware defrag with skinny metadata you are going
to see tons of warnings related to the backref count being less than 0.  This is
because the delayed refs will be found for stuff just fine, but it won't find
the skinny metadata extent refs.  With this patch I'm not seeing warnings
anymore.  Thanks,

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
fs/btrfs/backref.c

index 431ea92..eaf1333 100644 (file)
@@ -597,6 +597,7 @@ static int __add_inline_refs(struct btrfs_fs_info *fs_info,
        int slot;
        struct extent_buffer *leaf;
        struct btrfs_key key;
+       struct btrfs_key found_key;
        unsigned long ptr;
        unsigned long end;
        struct btrfs_extent_item *ei;
@@ -614,17 +615,21 @@ static int __add_inline_refs(struct btrfs_fs_info *fs_info,
 
        ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
        flags = btrfs_extent_flags(leaf, ei);
+       btrfs_item_key_to_cpu(leaf, &found_key, slot);
 
        ptr = (unsigned long)(ei + 1);
        end = (unsigned long)ei + item_size;
 
-       if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
+       if (found_key.type == BTRFS_EXTENT_ITEM_KEY &&
+           flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
                struct btrfs_tree_block_info *info;
 
                info = (struct btrfs_tree_block_info *)ptr;
                *info_level = btrfs_tree_block_level(leaf, info);
                ptr += sizeof(struct btrfs_tree_block_info);
                BUG_ON(ptr > end);
+       } else if (found_key.type == BTRFS_METADATA_ITEM_KEY) {
+               *info_level = found_key.offset;
        } else {
                BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
        }
@@ -796,8 +801,11 @@ static int find_parent_nodes(struct btrfs_trans_handle *trans,
        INIT_LIST_HEAD(&prefs_delayed);
 
        key.objectid = bytenr;
-       key.type = BTRFS_EXTENT_ITEM_KEY;
        key.offset = (u64)-1;
+       if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
+               key.type = BTRFS_METADATA_ITEM_KEY;
+       else
+               key.type = BTRFS_EXTENT_ITEM_KEY;
 
        path = btrfs_alloc_path();
        if (!path)
@@ -862,7 +870,8 @@ again:
                slot = path->slots[0];
                btrfs_item_key_to_cpu(leaf, &key, slot);
                if (key.objectid == bytenr &&
-                   key.type == BTRFS_EXTENT_ITEM_KEY) {
+                   (key.type == BTRFS_EXTENT_ITEM_KEY ||
+                    key.type == BTRFS_METADATA_ITEM_KEY)) {
                        ret = __add_inline_refs(fs_info, path, bytenr,
                                                &info_level, &prefs);
                        if (ret)
@@ -1276,12 +1285,16 @@ int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
 {
        int ret;
        u64 flags;
+       u64 size = 0;
        u32 item_size;
        struct extent_buffer *eb;
        struct btrfs_extent_item *ei;
        struct btrfs_key key;
 
-       key.type = BTRFS_EXTENT_ITEM_KEY;
+       if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
+               key.type = BTRFS_METADATA_ITEM_KEY;
+       else
+               key.type = BTRFS_EXTENT_ITEM_KEY;
        key.objectid = logical;
        key.offset = (u64)-1;
 
@@ -1294,9 +1307,15 @@ int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
                return ret;
 
        btrfs_item_key_to_cpu(path->nodes[0], found_key, path->slots[0]);
-       if (found_key->type != BTRFS_EXTENT_ITEM_KEY ||
+       if (found_key->type == BTRFS_METADATA_ITEM_KEY)
+               size = fs_info->extent_root->leafsize;
+       else if (found_key->type == BTRFS_EXTENT_ITEM_KEY)
+               size = found_key->offset;
+
+       if ((found_key->type != BTRFS_EXTENT_ITEM_KEY &&
+            found_key->type != BTRFS_METADATA_ITEM_KEY) ||
            found_key->objectid > logical ||
-           found_key->objectid + found_key->offset <= logical) {
+           found_key->objectid + size <= logical) {
                pr_debug("logical %llu is not within any extent\n",
                         (unsigned long long)logical);
                return -ENOENT;