f2fs: fix race condition on setting FI_NO_EXTENT flag

[ Upstream commit 07725ad ]

The following scenarios exist.
process A:               process B:
->f2fs_drop_extent_tree  ->f2fs_update_extent_cache_range
                          ->f2fs_update_extent_tree_range
                           ->write_lock
 ->set_inode_flag
                           ->is_inode_flag_set
                           ->__free_extent_tree // Shouldn't
                                                // have been
                                                // cleaned up
                                                // here
  ->write_lock

In this case, the "FI_NO_EXTENT" flag is set between
f2fs_update_extent_tree_range and is_inode_flag_set
by other process. it leads to clearing the whole exten
tree which should not have happened. And we fix it by
move the setting it to the range of write_lock.

Fixes:5f281fab9b9a3 ("f2fs: disable extent_cache for fcollapse/finsert inodes")
Signed-off-by: Zhang Qilong <zhangqilong3@huawei.com>
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

fs/f2fs/extent_cache.c

@@ -804,9 +804,8 @@ void f2fs_drop_extent_tree(struct inode *inode)
 	if (!f2fs_may_extent_tree(inode))
 		return;
 
-	set_inode_flag(inode, FI_NO_EXTENT);
-
 	write_lock(&et->lock);
+	set_inode_flag(inode, FI_NO_EXTENT);
 	__free_extent_tree(sbi, et);
 	if (et->largest.len) {
 		et->largest.len = 0;