From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753334Ab1KRGnL (ORCPT ); Fri, 18 Nov 2011 01:43:11 -0500 Received: from mail-bw0-f46.google.com ([209.85.214.46]:59639 "EHLO mail-bw0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750868Ab1KRGnK (ORCPT ); Fri, 18 Nov 2011 01:43:10 -0500 Message-ID: <4EC5FE6A.3080003@openvz.org> Date: Fri, 18 Nov 2011 10:42:50 +0400 From: Konstantin Khlebnikov User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.1.19) Gecko/20111010 Iceape/2.0.14 MIME-Version: 1.0 To: Andrew Morton CC: "linux-mm@kvack.org" , "linux-kernel@vger.kernel.org" , Dave Chinner Subject: Re: [PATCH] mm: account reaped page cache on inode cache pruning References: <20111116134713.8933.34389.stgit@zurg> <20111117162322.1c3e3d05.akpm@linux-foundation.org> In-Reply-To: <20111117162322.1c3e3d05.akpm@linux-foundation.org> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Andrew Morton wrote: > On Wed, 16 Nov 2011 17:47:13 +0300 > Konstantin Khlebnikov wrote: > >> Inode cache pruning indirectly reclaims page-cache by invalidating mapping pages. >> Let's account them into reclaim-state to notice this progress in memory reclaimer. >> >> Signed-off-by: Konstantin Khlebnikov >> --- >> fs/inode.c | 2 ++ >> 1 files changed, 2 insertions(+), 0 deletions(-) >> >> diff --git a/fs/inode.c b/fs/inode.c >> index ee4e66b..1f6c48d 100644 >> --- a/fs/inode.c >> +++ b/fs/inode.c >> @@ -692,6 +692,8 @@ void prune_icache_sb(struct super_block *sb, int nr_to_scan) >> else >> __count_vm_events(PGINODESTEAL, reap); >> spin_unlock(&sb->s_inode_lru_lock); >> + if (current->reclaim_state) >> + current->reclaim_state->reclaimed_slab += reap; >> >> dispose_list(&freeable); >> } > > hm, yes, I suppose we should. > > It seems to be cheating to use the "reclaimed_slab" field for this. > Perhaps it would be cleaner to add an additional field to reclaim_state > for non-slab pages which were also reclaimed. That's a cosmetic thing > and I guess we don't need to go that far, not sure... Do we really need separate on-stack reclaim_state structure with single field? Maybe replace it with single long (or even unsigned int) .reclaimed_pages field on task_struct and account reclaimed pages unconditionally. > >