From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753084AbdHJTD5 (ORCPT ); Thu, 10 Aug 2017 15:03:57 -0400 Received: from mga04.intel.com ([192.55.52.120]:16831 "EHLO mga04.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752728AbdHJTD4 (ORCPT ); Thu, 10 Aug 2017 15:03:56 -0400 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.41,354,1498546800"; d="scan'208";a="136024595" Date: Thu, 10 Aug 2017 12:03:55 -0700 From: Andi Kleen To: Mark Rutland Cc: linux-kernel@vger.kernel.org, Peter Zijlstra , Alexander Shishkin , Arnaldo Carvalho de Melo , Jiri Olsa , Stephane Eranian , Thomas Gleixner , Vince Weaver , Ingo Molnar Subject: Re: pmu::read() called erroneously in v4.13-rc{3,4} Message-ID: <20170810190355.GL3946@tassilo.jf.intel.com> References: <20170810173551.GD12812@leverpostej> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20170810173551.GD12812@leverpostej> User-Agent: Mutt/1.8.3 (2017-05-23) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org > [ 230.909172] ------------[ cut here ]------------ > [ 230.913796] WARNING: CPU: 1 PID: 3637 at arch/x86/events/core.c:75 x86_perf_event_update+0xbd/0xd0 > [ 230.922739] Modules linked in: > [ 230.925798] CPU: 1 PID: 3637 Comm: wrong-counter Tainted: G W 4.13.0-rc4+ #230 > [ 230.934132] Hardware name: LENOVO 7484A3G/LENOVO, BIOS 5CKT54AUS 09/07/2009 > [ 230.941080] task: ffff9ad5b0232640 task.stack: ffffb2f001aac000 > [ 230.946989] RIP: 0010:x86_perf_event_update+0xbd/0xd0 > [ 230.952038] RSP: 0018:ffffb2f001aaf978 EFLAGS: 00010282 > [ 230.957257] RAX: 0000000000000020 RBX: ffffb2f001aafa48 RCX: 0000000000000000 > [ 230.964378] RDX: 0000000000000001 RSI: ffff9ad5a9d33400 RDI: ffff9ad5a9d32800 > [ 230.971498] RBP: ffffb2f001aaf9a0 R08: 0000000000000000 R09: 0000000000000028 > [ 230.978622] R10: ffff9ad5b0232640 R11: ffff9ad5a9cd4840 R12: 0000000000000002 > [ 230.985745] R13: 0000000000000018 R14: 0000000000000001 R15: ffff9ad5a9d32800 > [ 230.992866] FS: 0000000000000000(0000) GS:ffff9ad5bec80000(0000) knlGS:0000000000000000 > [ 231.000939] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 > [ 231.006680] CR2: 00007f26cba069d0 CR3: 0000000085a0a000 CR4: 00000000000406e0 > [ 231.013802] Call Trace: > [ 231.016254] x86_pmu_read+0x9/0x10 > [ 231.019656] perf_output_read+0x19f/0x410 > [ 231.023665] ? __alloc_pages_nodemask+0xe8/0x1f0 > [ 231.028274] perf_event_read_event+0xd2/0x110 > [ 231.032628] ? x86_pmu_del+0x3e/0x130 > [ 231.036290] ? __intel_pmu_enable_all.isra.12+0x27/0x90 > [ 231.041512] ? intel_pmu_enable_all+0xb/0x10 > [ 231.045779] ? x86_pmu_enable+0x25e/0x2f0 > [ 231.049787] ? group_sched_out+0x8c/0xd0 > [ 231.053708] ? perf_pmu_enable+0x22/0x30 > [ 231.057629] ? update_group_times+0x13/0x40 > [ 231.061812] ? list_del_event+0x66/0xc0 > [ 231.065647] perf_event_exit_task+0x2fa/0x380 I suspect the problem is that exit schedules before it calls perf_event_exit_task. We could just move it up to the beginning of exit. Untested patch here. Does that help? diff --git a/kernel/exit.c b/kernel/exit.c index 516acdb0e0ec..9a6f5f299588 100644 --- a/kernel/exit.c +++ b/kernel/exit.c @@ -779,6 +779,14 @@ void __noreturn do_exit(long code) int group_dead; TASKS_RCU(int tasks_rcu_i); + /* + * Flush inherited counters to the parent - before the parent + * gets woken up by child-exit notifications. + * + * because of cgroup mode, must be called before cgroup_exit() + */ + perf_event_exit_task(tsk); + profile_task_exit(tsk); kcov_task_exit(tsk); @@ -878,14 +886,6 @@ void __noreturn do_exit(long code) exit_task_work(tsk); exit_thread(tsk); - /* - * Flush inherited counters to the parent - before the parent - * gets woken up by child-exit notifications. - * - * because of cgroup mode, must be called before cgroup_exit() - */ - perf_event_exit_task(tsk); - sched_autogroup_exit_task(tsk); cgroup_exit(tsk);