[08/28] ocfs2: add ip_alloc_sem in direct IO to protect allocation changes
diff mbox

Message ID 55de399c.PCB+X4ihy+Ps4440%akpm@linux-foundation.org
State New
Headers show

Commit Message

Andrew Morton Aug. 26, 2015, 10:11 p.m. UTC
From: WeiWei Wang <wangww631@huawei.com>
Subject: ocfs2: add ip_alloc_sem in direct IO to protect allocation changes

In ocfs2, ip_alloc_sem is used to protect allocation changes on the node. 
In direct IO, we add ip_alloc_sem to protect date consistent between
direct-io and ocfs2_truncate_file race (buffer io use ip_alloc_sem
already).  Although inode->i_mutex lock is used to avoid concurrency of
above situation, i think ip_alloc_sem is still needed because protect
allocation changes is significant.

Other filesystem like ext4 also uses rw_semaphore to protect data
consistent between get_block-vs-truncate race by other means, So
ip_alloc_sem in ocfs2 direct io is needed.

Signed-off-by: Weiwei Wang <wangww631@huawei.com>
Cc: Mark Fasheh <mfasheh@suse.com>
Cc: Joel Becker <jlbec@evilplan.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---

 fs/ocfs2/aops.c |   18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)

Comments

Mark Fasheh Aug. 28, 2015, 11:27 p.m. UTC | #1
On Wed, Aug 26, 2015 at 03:11:40PM -0700, Andrew Morton wrote:
> From: WeiWei Wang <wangww631@huawei.com>
> Subject: ocfs2: add ip_alloc_sem in direct IO to protect allocation changes
> 
> In ocfs2, ip_alloc_sem is used to protect allocation changes on the node. 
> In direct IO, we add ip_alloc_sem to protect date consistent between
> direct-io and ocfs2_truncate_file race (buffer io use ip_alloc_sem
> already).  Although inode->i_mutex lock is used to avoid concurrency of
> above situation, i think ip_alloc_sem is still needed because protect
> allocation changes is significant.
> 
> Other filesystem like ext4 also uses rw_semaphore to protect data
> consistent between get_block-vs-truncate race by other means, So
> ip_alloc_sem in ocfs2 direct io is needed.
> 
> Signed-off-by: Weiwei Wang <wangww631@huawei.com>
> Cc: Mark Fasheh <mfasheh@suse.com>
> Cc: Joel Becker <jlbec@evilplan.org>
> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

Signed-off-by: Mark Fasheh <mfasheh@suse.de>

--
Mark Fasheh

Patch
diff mbox

diff -puN fs/ocfs2/aops.c~ocfs2-add-ip_alloc_sem-in-direct-io-to-protect-allocation-changes fs/ocfs2/aops.c
--- a/fs/ocfs2/aops.c~ocfs2-add-ip_alloc_sem-in-direct-io-to-protect-allocation-changes
+++ a/fs/ocfs2/aops.c
@@ -533,10 +533,14 @@  static int ocfs2_direct_IO_get_blocks(st
 
 	inode_blocks = ocfs2_blocks_for_bytes(inode->i_sb, i_size_read(inode));
 
+	down_read(&OCFS2_I(inode)->ip_alloc_sem);
+
 	/* This figures out the size of the next contiguous block, and
 	 * our logical offset */
 	ret = ocfs2_extent_map_get_blocks(inode, iblock, &p_blkno,
 					  &contig_blocks, &ext_flags);
+	up_read(&OCFS2_I(inode)->ip_alloc_sem);
+
 	if (ret) {
 		mlog(ML_ERROR, "get_blocks() failed iblock=%llu\n",
 		     (unsigned long long)iblock);
@@ -557,6 +561,8 @@  static int ocfs2_direct_IO_get_blocks(st
 
 		alloc_locked = 1;
 
+		down_write(&OCFS2_I(inode)->ip_alloc_sem);
+
 		/* fill hole, allocate blocks can't be larger than the size
 		 * of the hole */
 		clusters_to_alloc = ocfs2_clusters_for_bytes(inode->i_sb, len);
@@ -569,6 +575,7 @@  static int ocfs2_direct_IO_get_blocks(st
 		ret = ocfs2_extend_allocation(inode, cpos,
 				clusters_to_alloc, 0);
 		if (ret < 0) {
+			up_write(&OCFS2_I(inode)->ip_alloc_sem);
 			mlog_errno(ret);
 			goto bail;
 		}
@@ -576,11 +583,13 @@  static int ocfs2_direct_IO_get_blocks(st
 		ret = ocfs2_extent_map_get_blocks(inode, iblock, &p_blkno,
 				&contig_blocks, &ext_flags);
 		if (ret < 0) {
+			up_write(&OCFS2_I(inode)->ip_alloc_sem);
 			mlog(ML_ERROR, "get_blocks() failed iblock=%llu\n",
 					(unsigned long long)iblock);
 			ret = -EIO;
 			goto bail;
 		}
+		up_write(&OCFS2_I(inode)->ip_alloc_sem);
 	}
 
 	/*
@@ -835,12 +844,17 @@  static ssize_t ocfs2_direct_IO_write(str
 
 		/* zeroing out the previously allocated cluster tail
 		 * that but not zeroed */
-		if (ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb)))
+		if (ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))) {
+			down_read(&OCFS2_I(inode)->ip_alloc_sem);
 			ret = ocfs2_direct_IO_zero_extend(osb, inode, offset,
 					zero_len_tail, cluster_align_tail);
-		else
+			up_read(&OCFS2_I(inode)->ip_alloc_sem);
+		} else {
+			down_write(&OCFS2_I(inode)->ip_alloc_sem);
 			ret = ocfs2_direct_IO_extend_no_holes(osb, inode,
 					offset);
+			up_write(&OCFS2_I(inode)->ip_alloc_sem);
+		}
 		if (ret < 0) {
 			mlog_errno(ret);
 			ocfs2_inode_unlock(inode, 1);