ext4: fix missing unlock on error in collapse range

Add i_write_mutex unlock on the error handling path in collapse_range().

Cc Dan Carpenter <[email protected]>

Signed-off-by Namjae Jeon <[email protected]>

Signed-off-by Ashish Sangwan <[email protected]>

---
 fs/ext4/extents.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

Quoted 8 lines

Add i_write_mutex unlock on the error handling path in collapse_range().

Cc Dan Carpenter <[email protected]>

Signed-off-by Namjae Jeon <[email protected]>

Signed-off-by Ashish Sangwan <[email protected]>

---
 fs/ext4/extents.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

Thanks for the patch!

This is the increment diff which I'm going to apply instead. It results in a net decrease of one line of code, instead of a net increase of two, and it is more consistent in terms of the error handling style.

Cheers,

- Ted

 fs/ext4/extents.c | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

fs/ext4/extents.c +4 -5

index 8ffab40..cb23a34 100644
@@ -5423,10 +5423,8 @@ int ext4_collapse_range(struct inode *inode, loff_t offset, loff_t len)
 	/* Call ext4_force_commit to flush all data in case of data=journal. */
 	if (ext4_should_journal_data(inode)) {
 		ret = ext4_force_commit(inode->i_sb);
-		if (ret) {
-			mutex_unlock(&EXT4_I(inode)->i_write_mutex);
-			return ret;
-		}
+		if (ret)
+			goto out_i_write_mutex;
 	}
 
 	/*

@@ -5439,7 +5437,7 @@ int ext4_collapse_range(struct inode *inode, loff_t offset, loff_t len)
 	ret = filemap_write_and_wait_range(inode->i_mapping, ioffset,
 					   LLONG_MAX);
 	if (ret)
-		return ret;
+		goto out_i_write_mutex;
 
 	/* Take mutex lock */
 	mutex_lock(&inode->i_mutex);

@@ -5512,6 +5510,7 @@ out_dio:
 	ext4_inode_resume_unlocked_dio(inode);
 out_mutex:
 	mutex_unlock(&inode->i_mutex);
+out_i_write_mutex:
 	mutex_unlock(&EXT4_I(inode)->i_write_mutex);
 	return ret;
 }