Re: [PATCH v4] memcg: remove PCG_CACHE page_cgroup flag

[Michal Hocko <mhocko@suse.cz>]

On Tue 24-01-12 17:44:49, Johannes Weiner wrote:
> On Tue, Jan 24, 2012 at 05:01:40PM +0100, Michal Hocko wrote:
> > On Tue 24-01-12 15:54:11, Johannes Weiner wrote:
> > > Hold on, I think this patch is still not complete: end_migration()
> > > directly uses __mem_cgroup_uncharge_common() with the FORCE charge
> > > type.  This will uncharge all migrated anon pages as cache, when it
> > > should decide based on PageAnon(used), which is the page where
> > > ->mapping is intact after migration.
> > 
> > You are right, I've missed that one as well. Anyway
> > MEM_CGROUP_CHARGE_TYPE_FORCE is used only in mem_cgroup_end_migration
> > these days and it got out of sync with its documentation (used by
> > force_empty) quite some time ago (f817ed48). What about something like
> > the following on top of the previous patch?
> > --- 
> > Should be foldet into the previous patch with the updated changelog:
> > 
> > Mapping of the unused page is not touched during migration (see
> 
> used one, not unused.  unused->mapping is globbered during migration.

Yes, you are right:
---
Should be foldet into the previous patch with the updated changelog:

Mapping of the unused page is not touched during migration (see
page_remove_rmap) so we can rely on it and push the correct charge type
down to __mem_cgroup_uncharge_common from end_migration.
The force flag was misleading was abused for skipping the needless
page_mapped() / PageCgroupMigration() check, as we know the unused page
is no longer mapped and cleared the migration flag just a few lines
up.  But doing the checks is no biggie and it's not worth adding another
flag just to skip them.  But I guess this should be mentioned in the
changelog.

[hannes@cmpxchg.org: fix for end_migration with clarification]
---
 mm/memcontrol.c |    7 ++++++-
 1 files changed, 6 insertions(+), 1 deletions(-)

diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index 4d655ee..6a8cc56 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -3195,6 +3195,7 @@ void mem_cgroup_end_migration(struct mem_cgroup *memcg,
 {
 	struct page *used, *unused;
 	struct page_cgroup *pc;
+	bool anon;
 
 	if (!memcg)
 		return;
@@ -3207,6 +3208,7 @@ void mem_cgroup_end_migration(struct mem_cgroup *memcg,
 		used = newpage;
 		unused = oldpage;
 	}
+
 	/*
 	 * We disallowed uncharge of pages under migration because mapcount
 	 * of the page goes down to zero, temporarly.
@@ -3217,7 +3219,10 @@ void mem_cgroup_end_migration(struct mem_cgroup *memcg,
 	ClearPageCgroupMigration(pc);
 	unlock_page_cgroup(pc);
 
-	__mem_cgroup_uncharge_common(unused, MEM_CGROUP_CHARGE_TYPE_FORCE);
+	anon = PageAnon(used);
+	__mem_cgroup_uncharge_common(unused,
+			anon ? MEM_CGROUP_CHARGE_TYPE_MAPPED
+			: MEM_CGROUP_CHARGE_TYPE_CACHE);
 
 	/*
 	 * If a page is a file cache, radix-tree replacement is very atomic
-- 
1.7.8.3


-- 
Michal Hocko
SUSE Labs
SUSE LINUX s.r.o.
Lihovarska 1060/12
190 00 Praha 9    
Czech Republic