[Cluster-devel] [PATCH] gfs2: Improve mmap write vs. truncate consistency

[Andreas Gruenbacher]

On filesystems with a block size smaller than PAGE_SIZE, page_mkwrite is
called for each memory-mapped page before that page can be written to.
When such a memory-mapped file is truncated down to size x which is not
a multiple of the page size and then back to a larger size, the page
straddling size x can end up with a partial block mapping.  In that
case, make sure to mark that page read-only so that page_mkwrite will be
called before the page can be written to the next time.

(There is no point in marking the page straddling size x read-only when
truncating down as writing to memory beyond the end of the file will
result in SIGBUS instead of growing the file.)

Fixes xfstests generic/029, generic/030 on filesystems with a block size
smaller than PAGE_SIZE.

Signed-off-by: Andreas Gruenbacher <agruenba at redhat.com>
---
 fs/gfs2/bmap.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index 2043a728f281..9ef543dd38e2 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -2139,7 +2139,7 @@ static int do_grow(struct inode *inode, u64 size)
 	if (error)
 		goto do_end_trans;
 
-	i_size_write(inode, size);
+	truncate_setsize(inode, size);
 	ip->i_inode.i_mtime = ip->i_inode.i_ctime = current_time(&ip->i_inode);
 	gfs2_trans_add_meta(ip->i_gl, dibh);
 	gfs2_dinode_out(ip, dibh->b_data);
-- 
2.20.1