On Mon, Jul 09, 2012 at 01:01:00AM +0200, Johannes Weiner wrote:

On Sun, Jul 08, 2012 at 10:44:59PM +0800, Fengguang Wu wrote:

quoted

On Tue, Jul 03, 2012 at 03:31:26PM +0900, KAMEZAWA Hiroyuki wrote:

quoted

(2012/06/28 20:05), Sha Zhengju wrote:

quoted

From: Sha Zhengju <redacted>

Similar to dirty page, we add per cgroup writeback pages accounting. The lock
rule still is:
	mem_cgroup_begin_update_page_stat()
	modify page WRITEBACK stat
	mem_cgroup_update_page_stat()
	mem_cgroup_end_update_page_stat()

There're two writeback interface to modify: test_clear/set_page_writeback.

Signed-off-by: Sha Zhengju <redacted>

Seems good to me. BTW, you named macros as MEM_CGROUP_STAT_FILE_XXX
but I wonder these counters will be used for accounting swap-out's dirty pages..

STAT_DIRTY, STAT_WRITEBACK ? do you have better name ?

Perhaps we can follow the established "enum zone_stat_item" names:

        NR_FILE_DIRTY,
        NR_WRITEBACK,

s/NR_/MEM_CGROUP_STAT_/

The names indicate that dirty pages for anonymous pages are not
accounted (by __set_page_dirty_no_writeback()). While the writeback
pages accounting include both the file/anon pages.

Ah then we'll need to update the document in patch 0 accordingly. This
may sound a bit tricky to the users..

We already report the global one as "nr_dirty", though.  Please don't
give the memcg one a different name.

The enum naming is not too critical, but it would be nice to have it
match the public name.

Fair enough. The public name obviously has more weight :)

Thanks,
Fengguang