btrfs: fix possible deadlock by clearing __GFP_FS flag

Using the GFP_HIGHUSER_MOVABLE flag to allocate the metadata's page may cause
deadlock.
  Task1
  open()
    ...
    btrfs_search_slot()
      ...
      btrfs_cow_block()
	...
	alloc_page()
	  wait for reclaiming
					shrink_slab()
					  ...
					  shrink_icache_memory()
					    ...
					    btrfs_evict_inode()
					      ...
					      btrfs_search_slot()

If the path is locked by task1, the deadlock happens.

So the btree's page cache is different with the file's page cache, it can not
allocate pages by GFP_HIGHUSER_MOVABLE flag, we must clear __GFP_FS flag in
GFP_HIGHUSER_MOVABLE flag.

Reported-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

fs/btrfs/disk-io.c

@@ -1724,6 +1724,8 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 		goto fail_bdi;
 	}
 
+	fs_info->btree_inode->i_mapping->flags &= ~__GFP_FS;
+
 	INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
 	INIT_LIST_HEAD(&fs_info->trans_list);
 	INIT_LIST_HEAD(&fs_info->dead_roots);

fs/btrfs/inode.c

@@ -2536,6 +2536,8 @@ static void btrfs_read_locked_inode(struct inode *inode)
 	BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
 
 	alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
+	if (location.objectid == BTRFS_FREE_SPACE_OBJECTID)
+		inode->i_mapping->flags &= ~__GFP_FS;
 
 	/*
 	 * try to precache a NULL acl entry for files that don't have
@@ -4084,7 +4086,6 @@ struct inode *btrfs_iget(struct super_block *s, struct btrfs_key *location,
 		BTRFS_I(inode)->root = root;
 		memcpy(&BTRFS_I(inode)->location, location, sizeof(*location));
 		btrfs_read_locked_inode(inode);
-
 		inode_tree_add(inode);
 		unlock_new_inode(inode);
 		if (new)