Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753845AbYJaCAm (ORCPT ); Thu, 30 Oct 2008 22:00:42 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1753039AbYJaCAd (ORCPT ); Thu, 30 Oct 2008 22:00:33 -0400 Received: from mga09.intel.com ([134.134.136.24]:8807 "EHLO mga09.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753009AbYJaCAc (ORCPT ); Thu, 30 Oct 2008 22:00:32 -0400 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="4.33,519,1220252400"; d="scan'208";a="354249251" Date: Thu, 30 Oct 2008 19:00:31 -0700 From: Venki Pallipadi To: Thomas Gleixner Cc: linux-kernel Subject: nohz regression with upstream git Message-ID: <20081031020030.GA4506@linux-os.sc.intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.4.1i Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 5072 Lines: 124 Thomas, I am seeing a nohz regression on one of my dual core mobile test boxes. With 2.6.27 tickless idle works fine. With latest git, I see ~2000 interrupts per second (HZ=1000), even when idle. git bisect points towards commit fb02fbc14d17837b4b7b02dbb36142c16a7bf208 Author: Thomas Gleixner Date: Fri Oct 17 10:01:23 2008 +0200 NOHZ: restart tick device from irq_enter() And good powertop data PowerTOP 1.10 (C) 2007, 2008 Intel Corporation Collecting data for 15 seconds Cn Avg residency C0 (cpu running) ( 0.1%) polling 0.0ms ( 0.0%) C1 0.0ms ( 0.0%) C3 18.6ms (99.9%) P-states (frequencies) 2.21 Ghz 0.0% 2.21 Ghz 0.0% 1.60 Ghz 0.0% 800 Mhz 100.0% Wakeups-from-idle per second : 53.9 interval: 15.0s no ACPI power usage estimate available Top causes for wakeups: 65.5% ( 32.6) : extra timer interrupt 16.7% ( 8.3) : Rescheduling interrupts 3.3% ( 1.7) kacpid : schedule_timeout (process_timeout) 2.8% ( 1.4) : ata_piix, uhci_hcd:usb4 2.3% ( 1.1) : eth1 2.0% ( 1.0) : enqueue_task_rt (sched_rt_period_timer) 1.9% ( 0.9) : schedule_delayed_work_on (delayed_work_timer_fn) 1.1% ( 0.5) : neigh_table_init_no_netlink (neigh_periodic_timer) 1.1% ( 0.5) : e1000_intr_msi (e1000_watchdog) 0.7% ( 0.3) : Function call interrupts 0.4% ( 0.2) events/1 : __netdev_watchdog_up (dev_watchdog) 0.4% ( 0.2) sendmail : schedule_timeout (process_timeout) 0.4% ( 0.2) : page_writeback_init (wb_timer_fn) 0.3% ( 0.1) kerneloops : schedule_timeout (process_timeout) 0.3% ( 0.1) kerneloops : do_nanosleep (hrtimer_wakeup) 0.1% ( 0.1) sshd : sk_reset_timer (tcp_write_timer) 0.1% ( 0.1) events/1 : queue_delayed_work (delayed_work_timer_fn) 0.1% ( 0.1) mount : start_this_handle (commit_timeout) 0.1% ( 0.1) : queue_delayed_work (delayed_work_timer_fn) 0.1% ( 0.1) hald-addon-stor : schedule_timeout (process_timeout) 0.1% ( 0.1) : neigh_add_timer (neigh_timer_handler) 0.1% ( 0.1) : sk_reset_timer (tcp_delack_timer) and bad powertop data PowerTOP 1.10 (C) 2007, 2008 Intel Corporation Collecting data for 15 seconds Cn Avg residency C0 (cpu running) ( 5.6%) polling 0.0ms ( 0.0%) C1 1.2ms ( 0.0%) C3 0.5ms (94.4%) P-states (frequencies) 2.21 Ghz 0.0% 2.21 Ghz 0.0% 1.60 Ghz 0.0% 800 Mhz 100.0% Wakeups-from-idle per second : 1919.6 interval: 15.0s no ACPI power usage estimate available Top causes for wakeups: 99.0% (1902.4) : extra timer interrupt 0.3% ( 6.1) : Rescheduling interrupts 0.3% ( 5.6) : acpi 0.1% ( 1.4) : ata_piix, uhci_hcd:usb4 0.1% ( 1.1) : ata_piix, ehci_hcd:usb1, uhci_hcd:usb6 0.1% ( 1.1) : eth1 0.1% ( 1.0) : enqueue_task_rt (sched_rt_period_timer) 0.0% ( 0.9) : schedule_delayed_work_on (delayed_work_timer_fn) 0.0% ( 0.5) NetworkManager : e1000_intr_msi (e1000_watchdog) 0.0% ( 0.5) : neigh_table_init_no_netlink (neigh_periodic_timer) 0.0% ( 0.2) : page_writeback_init (wb_timer_fn) 0.0% ( 0.2) events/1 : __netdev_watchdog_up (dev_watchdog) 0.0% ( 0.2) sendmail : schedule_timeout (process_timeout) 0.0% ( 0.1) kerneloops : schedule_timeout (process_timeout) 0.0% ( 0.1) events/0 : sk_reset_timer (tcp_delack_timer) 0.0% ( 0.1) sshd : sk_reset_timer (tcp_write_timer) 0.0% ( 0.1) events/0 : neigh_add_timer (neigh_timer_handler) 0.0% ( 0.1) rsyslogd : schedule_timeout (process_timeout) 0.0% ( 0.1) rpcbind : schedule_timeout (process_timeout) 0.0% ( 0.1) mount : start_this_handle (commit_timeout) 0.0% ( 0.1) hald : schedule_timeout (process_timeout) 0.0% ( 0.1) crond : do_nanosleep (hrtimer_wakeup) 0.0% ( 0.1) : queue_delayed_work (delayed_work_timer_fn) 0.0% ( 0.1) hald-addon-stor : schedule_timeout (process_timeout) Not sure what is so specific about this platform that causes this problem. Its a regular Centrino 2 development box. I will poke more at this tomorrow. But, wanted to pass the info on to you before calling it a night. Let me know if you need any more info on this. Thanks, Venki -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/