From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S933187AbXGZRMc (ORCPT ); Thu, 26 Jul 2007 13:12:32 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1762547AbXGZRMY (ORCPT ); Thu, 26 Jul 2007 13:12:24 -0400 Received: from extu-mxob-1.symantec.com ([216.10.194.28]:60809 "EHLO extu-mxob-1.symantec.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1761481AbXGZRMX (ORCPT ); Thu, 26 Jul 2007 13:12:23 -0400 Date: Thu, 26 Jul 2007 18:11:32 +0100 (BST) From: Hugh Dickins X-X-Sender: hugh@blonde.wat.veritas.com To: Benjamin Herrenschmidt cc: Andi Kleen , Peter Zijlstra , Andrew Morton , Ingo Molnar , linux-kernel@vger.kernel.org Subject: Re: [patch] mm: reduce pagetable-freeing latencies In-Reply-To: <1185360414.5439.335.camel@localhost.localdomain> Message-ID: References: <20070724083855.GA858@elte.hu> <20070724015441.8604d85d.akpm@linux-foundation.org> <1185270045.5439.249.camel@localhost.localdomain> <1185312559.5439.276.camel@localhost.localdomain> <1185345850.8197.64.camel@twins> <20070725094605.GA9045@one.firstfloor.org> <1185358093.5439.325.camel@localhost.localdomain> <20070725102608.GB9045@one.firstfloor.org> <1185360414.5439.335.camel@localhost.localdomain> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-Brightmail-Verdict: VlJEQwAAAAIAAAABAAAAAAAAAAEAAAAAAAAABmluYm94AGxpbnV4LWtlcm5lbEB2Z2VyLmtlcm5lbC5vcmcAYWtwbUBsaW51eC1mb3VuZGF0aW9uLm9yZwBiZW5oQGtlcm5lbC5jcmFzaGluZy5vcmcAcGV0ZXJ6QGluZnJhZGVhZC5vcmcAYW5kaUBmaXJzdGZsb29yLm9yZwBtaW5nb0BlbHRlLmh1AA== X-Brightmail-Tracker: AAAAAA== Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Wed, 25 Jul 2007, Benjamin Herrenschmidt wrote: > > In fact, i discussed with peter and I think the best is to only have one > entry in the stack mmu_gather structure. Four or eight would be much nicer than one, but you're right there's a stackdepth worry: good spotting of those unmap_mapping_range cases, I'd certainly been thinking we were fairly near the top of the stack in unmap_vmas, which is not so (or not obviously so) in those cases. > If we fail to allocate a page > to batch entries, we flush that one and steal it :-) Then we have a page > to gather more. That will often be the case for anonymous pages, but not for pagecache pages. I had wondered whether to make exit_mmap unmap the stack first (rather than text first as usually happens), but seems rather a hack. Hugh