Re: [RFC][PATCH 4/5] memcg: remove PCG_CACHE bit

[KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>]

On Thu, 15 Dec 2011 11:24:42 +0100
Johannes Weiner <hannes@cmpxchg.org> wrote:

> On Thu, Dec 15, 2011 at 03:08:22PM +0900, KAMEZAWA Hiroyuki wrote:
> > >From 577b441ae259728d83a99baba11bf4925b4542d4 Mon Sep 17 00:00:00 2001
> > From: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
> > Date: Thu, 15 Dec 2011 12:09:03 +0900
> > Subject: [PATCH 4/5] memcg: remove PCG_CACHE bit.
> > 
> > This bit can be replaced by PageAnon(page) check.
> 
> You need to make sure that only fully rmapped pages are
> charged/uncharged, which is not the case yet.  E.g. look at the
> newpage_charge() in mm/memory.c::do_anonymous_page() - if the page
> table is updated by someone else, that uncharge_page() in the error
> path will see !PageAnon() because page->mapping is not set yet and
> uncharge it as cache.
> 

Ah, nice catch. I missed the error path.

> What I think is required is to break up the charging and committing
> like we do for swap cache already:
> 
> 	if (!mem_cgroup_try_charge())
> 		goto error;
> 	page_add_new_anon_rmap()
> 	mem_cgroup_commit()
> 
> This will also allow us to even get rid of passing around the charge
> type everywhere...
> 

Thank you. I'll look into.

To be honest, I want to remove 'rss' and 'cache' counter ;(
This doesn't have much meanings after lru was splitted.

Thanks,
-Kame

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>