On Wed, 19 Jan 2011 13:10:43 +0100
Johannes Weiner <hannes@xxxxxxxxxxx> wrote:
> Hello KAMEZAWA-san,
>
> On Tue, Jan 18, 2011 at 11:40:49AM +0900, KAMEZAWA Hiroyuki wrote:
> > +void mem_cgroup_split_huge_fixup(struct page *head, struct page *tail)
> > +{
> > + struct page_cgroup *head_pc = lookup_page_cgroup(head);
> > + struct page_cgroup *tail_pc = lookup_page_cgroup(tail);
> > + unsigned long flags;
> > +
> > + /*
> > + * We have no races witch charge/uncharge but will have races with
> > + * page state accounting.
> > + */
> > + move_lock_page_cgroup(head_pc, &flags);
> > +
> > + tail_pc->mem_cgroup = head_pc->mem_cgroup;
> > + smp_wmb(); /* see __commit_charge() */
>
> I thought the barriers were needed because charging does not hold the
> lru lock. But here we do, and all the 'lockless' read-sides do so as
> well. Am I missing something or can this barrier be removed?
>
Hmm. I think this can be removed. But it's just because there is only one
referer of lockless access to pc->mem_cgroup. I think it's ok to remove
this but it should be done by independent patch with enough patch
clarification. IOW, I'll do later in another patch.
Thanks,
-Kame
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxxx For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom policy in Canada: sign http://dissolvethecrtc.ca/
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>