From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1161647AbcFGPOs (ORCPT ); Tue, 7 Jun 2016 11:14:48 -0400 Received: from e31.co.us.ibm.com ([32.97.110.149]:50935 "EHLO e31.co.us.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1161110AbcFGPOp (ORCPT ); Tue, 7 Jun 2016 11:14:45 -0400 X-IBM-Helo: d03dlp02.boulder.ibm.com X-IBM-MailFrom: ego@linux.vnet.ibm.com X-IBM-RcptTo: mpe@ellerman.id.au;htejun@gmail.com;peterz@infradead.org;tglx@linutronix.de;linuxppc-dev@lists.ozlabs.org;linux-kernel@vger.kernel.org From: "Gautham R. Shenoy" To: Peter Zijlstra , Thomas Gleixner , Tejun Heo , Michael Ellerman , Abdul Haleem , Aneesh Kumar Cc: linuxppc-dev@lists.ozlabs.org, linux-kernel@vger.kernel.org, "Gautham R. Shenoy" Subject: [PATCH 1/2] workqueue: Move wq_update_unbound_numa() to the beginning of CPU_ONLINE Date: Tue, 7 Jun 2016 20:44:02 +0530 Message-Id: <6b3c7059ec5d2d6157d23d619e4507692a42a5bd.1465311052.git.ego@linux.vnet.ibm.com> X-Mailer: git-send-email 1.9.3 In-Reply-To: References: <5756BE20.3040208@linux.vnet.ibm.com> In-Reply-To: References: X-TM-AS-GCONF: 00 X-Content-Scanned: Fidelis XPS MAILER x-cbid: 16060715-8236-0000-0000-000031FE91F4 X-IBM-AV-DETECTION: SAVI=unused REMOTE=unused XFE=unused Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Currently in the CPU_ONLINE workqueue handler, the restore_unbound_workers_cpumask() will never call set_cpus_allowed_ptr() for a newly created unbound worker thread. This is because the function which creates a new unbound worker thread when the first CPU in the node comes online [wq_update_unbound_numa()] is invoked after the call to restore_unbound_workers_cpumask(). Thus the affinity is never set for this worker thread when the first CPU in the node comes online. Furthermore, due to an optimization in restore_unbound_workers_cpumask(), set_cpus_allowed_ptr() is not called when subsequent CPUs in the node come online since it assumes that the affinity would have been set when the first CPU has come online. This patch fixes this issue by invoking wq_update_unbound_numa() before the calling restore_unbound_workers_cpumask(). Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: Tejun Heo Cc: Michael Ellerman Signed-off-by: Gautham R. Shenoy --- kernel/workqueue.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/kernel/workqueue.c b/kernel/workqueue.c index e1c0e99..e412794 100644 --- a/kernel/workqueue.c +++ b/kernel/workqueue.c @@ -4638,6 +4638,10 @@ static int workqueue_cpu_up_callback(struct notifier_block *nfb, case CPU_ONLINE: mutex_lock(&wq_pool_mutex); + /* update NUMA affinity of unbound workqueues */ + list_for_each_entry(wq, &workqueues, list) + wq_update_unbound_numa(wq, cpu, true); + for_each_pool(pool, pi) { mutex_lock(&pool->attach_mutex); @@ -4649,10 +4653,6 @@ static int workqueue_cpu_up_callback(struct notifier_block *nfb, mutex_unlock(&pool->attach_mutex); } - /* update NUMA affinity of unbound workqueues */ - list_for_each_entry(wq, &workqueues, list) - wq_update_unbound_numa(wq, cpu, true); - mutex_unlock(&wq_pool_mutex); break; } -- 1.9.3