Btrfs: bail out if block group has different mixed flag
authorLiu Bo <bo.li.liu@oracle.com>
Fri, 26 Aug 2016 01:08:27 +0000 (18:08 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Fri, 6 Jan 2017 10:16:10 +0000 (11:16 +0100)
commit 49303381f19ab16a371a061b67e783d3f570d56e upstream.

Currently we allow inconsistence about mixed flag
 (BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA).

We'd get ENOSPC if block group has mixed flag and btrfs doesn't.
If that happens, we have one space_info with mixed flag and another
space_info only with BTRFS_BLOCK_GROUP_METADATA, and
global_block_rsv.space_info points to the latter one, but all bytes
from block_group contributes to the mixed space_info, thus all the
allocation will fail with ENOSPC.

This adds a check for the above case.

Reported-by: Vegard Nossum <vegard.nossum@oracle.com>
Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
[ updated message ]
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/btrfs/extent-tree.c

index 47cdc6f3390b2f72f0f861c3e6de157bf49c0046..431364ab9c0eaa3b8379834930ae3908c238be20 100644 (file)
@@ -9686,6 +9686,11 @@ int btrfs_read_block_groups(struct btrfs_root *root)
        struct extent_buffer *leaf;
        int need_clear = 0;
        u64 cache_gen;
+       u64 feature;
+       int mixed;
+
+       feature = btrfs_super_incompat_flags(info->super_copy);
+       mixed = !!(feature & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS);
 
        root = info->extent_root;
        key.objectid = 0;
@@ -9739,6 +9744,15 @@ int btrfs_read_block_groups(struct btrfs_root *root)
                                   btrfs_item_ptr_offset(leaf, path->slots[0]),
                                   sizeof(cache->item));
                cache->flags = btrfs_block_group_flags(&cache->item);
+               if (!mixed &&
+                   ((cache->flags & BTRFS_BLOCK_GROUP_METADATA) &&
+                   (cache->flags & BTRFS_BLOCK_GROUP_DATA))) {
+                       btrfs_err(info,
+"bg %llu is a mixed block group but filesystem hasn't enabled mixed block groups",
+                                 cache->key.objectid);
+                       ret = -EINVAL;
+                       goto error;
+               }
 
                key.objectid = found_key.objectid + found_key.offset;
                btrfs_release_path(path);