f2fs: fix double lock in handle_failed_inode

In handle_failed_inode, there is a potential deadlock which can happen
in below call path:

- f2fs_create
 - f2fs_lock_op   down_read(cp_rwsem)
 - f2fs_add_link
  - __f2fs_add_link
   - init_inode_metadata
    - f2fs_init_security    failed
    - truncate_blocks    failed
 - handle_failed_inode
  - f2fs_truncate
   - truncate_blocks(..,true)
					- write_checkpoint
					 - block_operations
					  - f2fs_lock_all  down_write(cp_rwsem)
    - f2fs_lock_op   down_read(cp_rwsem)

So in this path, we pass parameter to f2fs_truncate to make sure
cp_rwsem in truncate_blocks will not be locked again.

Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This commit is contained in:
Chao Yu
2015-07-16 18:19:02 +08:00
committed by Jaegeuk Kim
parent ecbaa4068f
commit 55f57d2c42
4 changed files with 7 additions and 7 deletions

View File

@@ -343,7 +343,7 @@ void f2fs_evict_inode(struct inode *inode)
i_size_write(inode, 0);
if (F2FS_HAS_BLOCKS(inode))
f2fs_truncate(inode);
f2fs_truncate(inode, true);
f2fs_lock_op(sbi);
remove_inode_page(inode);
@@ -385,7 +385,7 @@ void handle_failed_inode(struct inode *inode)
i_size_write(inode, 0);
if (F2FS_HAS_BLOCKS(inode))
f2fs_truncate(inode);
f2fs_truncate(inode, false);
remove_inode_page(inode);