ocfs2: add ip_alloc_sem in direct IO to protect allocation changes
authorWeiWei Wang <wangww631@huawei.com>
Fri, 4 Sep 2015 22:44:23 +0000 (15:44 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 4 Sep 2015 23:54:41 +0000 (16:54 -0700)
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>
Signed-off-by: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/ocfs2/aops.c

index b36dcad3a14058320b9112f13d3bbe3b93882405..a7ab145e29014d3e3ee7d6a127de0967035e0eac 100644 (file)
@@ -533,10 +533,14 @@ static int ocfs2_direct_IO_get_blocks(struct inode *inode, sector_t iblock,
 
        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(struct inode *inode, sector_t iblock,
 
                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(struct inode *inode, sector_t iblock,
                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(struct inode *inode, sector_t iblock,
                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(struct kiocb *iocb,
 
                /* 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);