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

[Andreas Gruenbacher <agruenba@redhat.com>]

When punching a hole in a file, use filemap_write_and_wait_range to
write back any dirty pages in the range of the hole.  As a side effect,
if the hole isn't page aligned, this marks unaligned pages at the
beginning and the end of the hole read-only.  This is required when the
block size is smaller than the page size: when those pages are written
to again after the hole punching, we must make sure that page_mkwrite is
called for those pages so that the page will be fully allocated and any
blocks turned into holes from the hole punching will be reallocated.
(If a page is writably mapped, page_mkwrite won't be called.)

Fixes xfstest generic/567.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
---
 fs/gfs2/bmap.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index f63df54a08c6..bb0113a0b0f4 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -2440,8 +2440,16 @@ int __gfs2_punch_hole(struct file *file, loff_t offset, loff_t length)
 	struct inode *inode = file_inode(file);
 	struct gfs2_inode *ip = GFS2_I(inode);
 	struct gfs2_sbd *sdp = GFS2_SB(inode);
+	unsigned int blocksize = i_blocksize(inode);
+	loff_t start, end;
 	int error;
 
+	start = round_down(offset, blocksize);
+	end = round_up(offset + length, blocksize) - 1;
+	error = filemap_write_and_wait_range(inode->i_mapping, start, end);
+	if (error)
+		return error;
+
 	if (gfs2_is_jdata(ip))
 		error = gfs2_trans_begin(sdp, RES_DINODE + 2 * RES_JDATA,
 					 GFS2_JTRUNC_REVOKES);
@@ -2455,9 +2463,8 @@ int __gfs2_punch_hole(struct file *file, loff_t offset, loff_t length)
 		if (error)
 			goto out;
 	} else {
-		unsigned int start_off, end_len, blocksize;
+		unsigned int start_off, end_len;
 
-		blocksize = i_blocksize(inode);
 		start_off = offset & (blocksize - 1);
 		end_len = (offset + length) & (blocksize - 1);
 		if (start_off) {
-- 
2.20.1