ocfs2: fix shift left overflow
authorJoseph Qi <joseph.qi@huawei.com>
Thu, 6 Aug 2015 22:46:48 +0000 (15:46 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 7 Aug 2015 01:39:41 +0000 (04:39 +0300)
When using a large volume, for example 9T volume with 2T already used,
frequent creation of small files with O_DIRECT when the IO is not
cluster aligned may clear sectors in the wrong place.  This will cause
filesystem corruption.

This is because p_cpos is a u32.  When calculating the corresponding
sector it should be converted to u64 first, otherwise it may overflow.

Signed-off-by: Joseph Qi <joseph.qi@huawei.com>
Cc: Mark Fasheh <mfasheh@suse.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: <stable@vger.kernel.org> [4.0+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/ocfs2/aops.c

index 1a35c6139656344516aacd59c7120f6fb877f2a2..0f5fd9db8194ef5d135f1896f6e2645a5f059cd8 100644 (file)
@@ -685,7 +685,7 @@ static int ocfs2_direct_IO_zero_extend(struct ocfs2_super *osb,
 
        if (p_cpos && !(ext_flags & OCFS2_EXT_UNWRITTEN)) {
                u64 s = i_size_read(inode);
-               sector_t sector = (p_cpos << (osb->s_clustersize_bits - 9)) +
+               sector_t sector = ((u64)p_cpos << (osb->s_clustersize_bits - 9)) +
                        (do_div(s, osb->s_clustersize) >> 9);
 
                ret = blkdev_issue_zeroout(osb->sb->s_bdev, sector,
@@ -910,7 +910,7 @@ static ssize_t ocfs2_direct_IO_write(struct kiocb *iocb,
                BUG_ON(!p_cpos || (ext_flags & OCFS2_EXT_UNWRITTEN));
 
                ret = blkdev_issue_zeroout(osb->sb->s_bdev,
-                               p_cpos << (osb->s_clustersize_bits - 9),
+                               (u64)p_cpos << (osb->s_clustersize_bits - 9),
                                zero_len_head >> 9, GFP_NOFS, false);
                if (ret < 0)
                        mlog_errno(ret);