Skip to contentSkip to conversation

exofs: simple_write_end does not mark_inode_dirty

PATCH In review Boaz Harrosh 2 participants 3 messages Dec 27, 2009 – Jan 5, 2010 VFS

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 exofs: simple_write_end does not mark_inode_dirty

exofs uses simple_write_end() for it's .write_end handler. But it is not enough because simple_write_end() does not call mark_inode_dirty() when it extends i_size. So even if we do call mark_inode_dirty at beginning of write out, with a very long IO and a saturated system we might get the .write_inode() called while still extend-writing to file and miss out on the last i_size updates.

So override .write_end, call simple_write_end(), and afterwords if i_size was changed call mark_inode_dirty().

It stands to logic that since simple_write_end() was the one extending i_size it should also call mark_inode_dirty(). But it looks like all users of simple_write_end() are memory-bound pseudo filesystems, who could careless about mark_inode_dirty(). I might submit a warning-comment patch to simple_write_end() in future.

---
 fs/exofs/inode.c |   17 ++++++++++++++++-
 1 files changed, 16 insertions(+), 1 deletions(-)
fs/exofs/inode.c +16 -1
index 0b6f318..a8de6ae 100644
@@ -738,13 +738,28 @@ static int exofs_write_begin_export(struct file *file,
 					fsdata);
 }
 
+static int exofs_write_end(struct file *file, struct address_space *mapping,
+			loff_t pos, unsigned len, unsigned copied,
+			struct page *page, void *fsdata)
+{
+	struct inode *inode = mapping->host;
+	/* According to comment in simple_write_end i_mutex is held */
+	loff_t i_size = inode->i_size;
+	int ret;
+
+	ret = simple_write_end(file, mapping,pos, len, copied, page, fsdata);
+	if (i_size != inode->i_size)
+		mark_inode_dirty(inode);
+	return ret;
+}
+
 const struct address_space_operations exofs_aops = {
 	.readpage	= exofs_readpage,
 	.readpages	= exofs_readpages,
 	.writepage	= exofs_writepage,
 	.writepages	= exofs_writepages,
 	.write_begin	= exofs_write_begin_export,
-	.write_end	= simple_write_end,
+	.write_end	= exofs_write_end,
 };
 
 /******************************************************************************
#1 lore raw
Quoted 12 lines · On Sun, 27 Dec 2009 18:44:21 +0200 Boaz Harrosh <[email protected]> wrote:

exofs uses simple_write_end() for it's .write_end handler. But it is not enough because simple_write_end() does not call mark_inode_dirty() when it extends i_size. So even if we do call mark_inode_dirty at beginning of write out, with a very long IO and a saturated system we might get the .write_inode() called while still extend-writing to file and miss out on the last i_size updates.

So override .write_end, call simple_write_end(), and afterwords if i_size was changed call mark_inode_dirty().

Quoted 7 lines

It stands to logic that since simple_write_end() was the one extending i_size it should also call mark_inode_dirty(). But it looks like all users of simple_write_end() are memory-bound pseudo filesystems, who could careless about mark_inode_dirty(). I might submit a warning-comment patch to simple_write_end() in future.

You added this to linux-next without a "Cc: <[email protected]>" to the changelog, so it won't be backported. However this patch should be backported, IMO.

#2 lore raw
Quoted 15 lines · Andrew Morton
Quoted 12 lines · On Sun, 27 Dec 2009 18:44:21 +0200 Boaz Harrosh <[email protected]> wrote:

exofs uses simple_write_end() for it's .write_end handler. But it is not enough because simple_write_end() does not call mark_inode_dirty() when it extends i_size. So even if we do call mark_inode_dirty at beginning of write out, with a very long IO and a saturated system we might get the .write_inode() called while still extend-writing to file and miss out on the last i_size updates.

So override .write_end, call simple_write_end(), and afterwords if i_size was changed call mark_inode_dirty().

Quoted 8 lines
Quoted 7 lines

It stands to logic that since simple_write_end() was the one extending i_size it should also call mark_inode_dirty(). But it looks like all users of simple_write_end() are memory-bound pseudo filesystems, who could careless about mark_inode_dirty(). I might submit a warning-comment patch to simple_write_end() in future.

OK, Thanks Andrew
I'll redo it ASAP before pushing to Linus

Quoted 5 lines

You added this to linux-next without a "Cc: <[email protected]>" to the changelog, so it won't be backported. However this patch should be backported, IMO.

Boaz