From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1034303AbdAEUR0 (ORCPT ); Thu, 5 Jan 2017 15:17:26 -0500 Received: from mx1.redhat.com ([209.132.183.28]:54222 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753722AbdAEURR (ORCPT ); Thu, 5 Jan 2017 15:17:17 -0500 From: Waiman Long To: Thomas Gleixner , Andrew Morton Cc: linux-kernel@vger.kernel.org, "Du Changbin" , Christian Borntraeger , Jan Stancek , Waiman Long Subject: [RESEND PATCH v2 0/3] debugobjects: Reduce global pool_lock contention Date: Thu, 5 Jan 2017 15:17:02 -0500 Message-Id: <1483647425-4135-1-git-send-email-longman@redhat.com> X-Greylist: Sender IP whitelisted, not delayed by milter-greylist-4.5.16 (mx1.redhat.com [10.5.110.28]); Thu, 05 Jan 2017 20:17:18 +0000 (UTC) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org v1->v2: - Move patch 2 in front of patch 1. - Fix merge conflict with linux-next. This patchset aims to reduce contention of the global pool_lock while improving performance at the same time. It is done to resolve the following soft lockup problem with a debug kernel in some of the large SMP systems: NMI watchdog: BUG: soft lockup - CPU#35 stuck for 22s! [rcuos/1:21] ... RIP: 0010:[] [] _raw_spin_unlock_irqrestore+0x3b/0x60 ... Call Trace: [] free_object+0x81/0xb0 [] debug_check_no_obj_freed+0x193/0x220 [] ? trace_hardirqs_on_caller+0xf9/0x1c0 [] ? file_free_rcu+0x36/0x60 [] kmem_cache_free+0xd2/0x380 [] ? fput+0x90/0x90 [] file_free_rcu+0x36/0x60 [] rcu_nocb_kthread+0x1b3/0x550 [] ? rcu_nocb_kthread+0x101/0x550 [] ? sync_exp_work_done.constprop.63+0x50/0x50 [] kthread+0x101/0x120 [] ? trace_hardirqs_on_caller+0xf9/0x1c0 [] ret_from_fork+0x22/0x50 On a 8-socket IvyBridge-EX system (120 cores, 240 threads), the elapsed time of a 4.9-rc7 kernel parallel build (make -j 240) was reduced from 7m57s to 7m19s with a patched 4.9-rc7 kernel. There was also about a 10X reduction in the number of debug objects being allocated from or freed to the kmemcache during the kernel build. Waiman Long (3): debugobjects: Track number of kmem_cache_alloc/kmem_cache_free done debugobjects: Scale thresholds with # of CPUs debugobjects: Reduce contention on the global pool_lock lib/debugobjects.c | 57 ++++++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 45 insertions(+), 12 deletions(-) -- 1.8.3.1