From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754838Ab1HWBrE (ORCPT ); Mon, 22 Aug 2011 21:47:04 -0400 Received: from mga02.intel.com ([134.134.136.20]:30298 "EHLO mga02.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752888Ab1HWBrD (ORCPT ); Mon, 22 Aug 2011 21:47:03 -0400 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="4.67,352,1309762800"; d="scan'208";a="41154897" Subject: Re: [patch 1/2]slub: add slab with one free object to partial list tail From: Shaohua Li To: Andrew Morton Cc: linux-mm , lkml , cl@linux.com, penberg@kernel.org, "Shi, Alex" , "Chen, Tim C" In-Reply-To: <1314059819.29510.18.camel@sli10-conroe> References: <1314059819.29510.18.camel@sli10-conroe> Content-Type: text/plain; charset="UTF-8" Date: Tue, 23 Aug 2011 09:48:18 +0800 Message-ID: <1314064098.29510.23.camel@sli10-conroe> Mime-Version: 1.0 X-Mailer: Evolution 2.32.2 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tue, 2011-08-23 at 08:36 +0800, Shaohua Li wrote: > The slab has just one free object, adding it to partial list head doesn't make > sense. And it can cause lock contentation. For example, > 1. CPU takes the slab from partial list > 2. fetch an object > 3. switch to another slab > 4. free an object, then the slab is added to partial list again > In this way n->list_lock will be heavily contended. > In fact, Alex had a hackbench regression. 3.1-rc1 performance drops about 70% > against 3.0. This patch fixes it. updated the change log a little bit. My apologize to not mention Alex's credit. Thanks, Shaohua The slab has just one free object, adding it to partial list head doesn't make sense. And it can cause lock contentation. For example, 1. CPU takes the slab from partial list 2. fetch an object 3. switch to another slab 4. free an object, then the slab is added to partial list again In this way n->list_lock will be heavily contended. In fact, Alex had a hackbench regression. 3.1-rc1 performance drops about 70% against 3.0. This patch fixes it. Thanks Alex to bisect the issue to be a slub regression and collect perf data. Reported-by: Alex Shi Signed-off-by: Shaohua Li Signed-off-by: Shaohua Li --- mm/slub.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) Index: linux/mm/slub.c =================================================================== --- linux.orig/mm/slub.c 2011-08-15 09:55:21.000000000 +0800 +++ linux/mm/slub.c 2011-08-23 08:13:54.000000000 +0800 @@ -2377,7 +2377,7 @@ static void __slab_free(struct kmem_cach */ if (unlikely(!prior)) { remove_full(s, page); - add_partial(n, page, 0); + add_partial(n, page, 1); stat(s, FREE_ADD_PARTIAL); } }