btrfs: set cache_block_group_error if we find an error

We set cache_block_group_error if btrfs_cache_block_group() returns an
error, this is because we could end up not finding space to allocate and
mistakenly return -ENOSPC, and which could then abort the transaction
with the incorrect errno, and in the case of ENOSPC result in a
WARN_ON() that will trip up tests like generic/475.

However there's the case where multiple threads can be racing, one
thread gets the proper error, and the other thread doesn't actually call
btrfs_cache_block_group(), it instead sees ->cached ==
BTRFS_CACHE_ERROR.  Again the result is the same, we fail to allocate
our space and return -ENOSPC.  Instead we need to set
cache_block_group_error to -EIO in this case to make sure that if we do
not make our allocation we get the appropriate error returned back to
the caller.

CC: stable@vger.kernel.org # 4.14+
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>

fs/btrfs/extent-tree.c

@@ -4310,8 +4310,11 @@ static noinline int find_free_extent(struct btrfs_root *root,
 			ret = 0;
 		}
 
-		if (unlikely(block_group->cached == BTRFS_CACHE_ERROR))
+		if (unlikely(block_group->cached == BTRFS_CACHE_ERROR)) {
+			if (!cache_block_group_error)
+				cache_block_group_error = -EIO;
 			goto loop;
+		}
 
 		if (!find_free_extent_check_size_class(ffe_ctl, block_group))
 			goto loop;