[PATCH 4/6 v7] ext4: Lock i_mutex for punch hole

[Allison Henderson <achender@linux.vnet.ibm.com>]

This patch modifies the fallocate routine to lock i_mutex during
the punch hole operation.

Yongqiang noticed that the vfs layer locks i_mutex for truncate,
but not fallocate, so the fallocate routine will need to take
care of locking i_mutex.  Otherwise a page may be mapped after
punch hole has released the pages, but before i_data_sem is
locked to release the blocks in the extent tree.

Signed-off-by: Allison Henderson <achender@linux.vnet.ibm.com>
---
:100644 100644 9124cd2... 007fb08... M	fs/ext4/extents.c
 fs/ext4/extents.c |   10 +++++++---
 1 files changed, 7 insertions(+), 3 deletions(-)

diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index 9124cd2..007fb08 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -3774,8 +3774,13 @@ long ext4_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
 		return -EOPNOTSUPP;
 
-	if (mode & FALLOC_FL_PUNCH_HOLE)
-		return ext4_punch_hole(file, offset, len);
+	mutex_lock(&inode->i_mutex);
+
+	if (mode & FALLOC_FL_PUNCH_HOLE) {
+		ret = ext4_punch_hole(file, offset, len);
+		mutex_unlock(&inode->i_mutex);
+		return ret;	
+	}
 
 	trace_ext4_fallocate_enter(inode, offset, len, mode);
 	map.m_lblk = offset >> blkbits;
@@ -3789,7 +3794,6 @@ long ext4_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
 	 * credits to insert 1 extent into extent tree
 	 */
 	credits = ext4_chunk_trans_blocks(inode, max_blocks);
-	mutex_lock(&inode->i_mutex);
 	ret = inode_newsize_ok(inode, (len + offset));
 	if (ret) {
 		mutex_unlock(&inode->i_mutex);
-- 
1.7.1