On Tue, 17 Jan 2012 17:46:05 +0100 Michal Hocko <mhocko@xxxxxxx> wrote: > On Fri 13-01-12 17:40:19, KAMEZAWA Hiroyuki wrote: > > > > From 1008e84d94245b1e7c4d237802ff68ff00757736 Mon Sep 17 00:00:00 2001 > > From: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx> > > Date: Thu, 12 Jan 2012 15:53:24 +0900 > > Subject: [PATCH 3/7] memcg: remove PCG_MOVE_LOCK flag from pc->flags. > > > > PCG_MOVE_LOCK bit is used for bit spinlock for avoiding race between > > memcg's account moving and page state statistics updates. > > > > Considering page-statistics update, very hot path, this lock is > > taken only when someone is moving account (or PageTransHuge()) > > Outdated comment? THP is not an issue here. > Ah, sorry. I reorderd patches. > > And, now, all moving-account between memcgroups (by task-move) > > are serialized. > > > > So, it seems too costly to have 1bit per page for this purpose. > > > > This patch removes PCG_MOVE_LOCK and add hashed rwlock array > > instead of it. This works well enough. Even when we need to > > take the lock, > > Hmmm, rwlocks are not popular these days very much. > Anyway, can we rather make it (source) memcg (bit)spinlock instead. We > would reduce false sharing this way and would penalize only pages from > the moving group. > per-memcg spinlock ? The reason I used rwlock() is to avoid disabling IRQ. This routine will be called by IRQ context (for dirty ratio support). So, IRQ disable will be required if we use spinlock. Thanks, -Kame -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. 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@xxxxxxxxx";> email@xxxxxxxxx </a>