From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756301AbZAAWI6 (ORCPT ); Thu, 1 Jan 2009 17:08:58 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1753332AbZAAWIt (ORCPT ); Thu, 1 Jan 2009 17:08:49 -0500 Received: from e4.ny.us.ibm.com ([32.97.182.144]:60632 "EHLO e4.ny.us.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753307AbZAAWIs (ORCPT ); Thu, 1 Jan 2009 17:08:48 -0500 Date: Thu, 1 Jan 2009 14:08:46 -0800 From: "Paul E. McKenney" To: Eric Sesterhenn Cc: linux-kernel@vger.kernel.org, dipankar@in.ibm.com Subject: Re: [BUG] Cpustall with kcore Message-ID: <20090101220846.GB7134@linux.vnet.ibm.com> Reply-To: paulmck@linux.vnet.ibm.com References: <20090101201914.GB6698@alice> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20090101201914.GB6698@alice> User-Agent: Mutt/1.5.15+20070412 (2007-04-11) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Thu, Jan 01, 2009 at 09:19:14PM +0100, Eric Sesterhenn wrote: > hi, > > when i issue a "cat /proc/kcore > /dev/null", the kernel issues the > following warning and i cant kill the cat process anymore. > > [ 356.948019] INFO: RCU detected CPU 0 stall (t=2500 jiffies) > [ 356.948019] Pid: 4156, comm: cat Not tainted 2.6.28 #77 > [ 356.948019] Call Trace: > [ 356.948019] [] ? printk+0x18/0x1a > [ 356.948019] [] __rcu_pending+0x62/0x250 > [ 356.948019] [] ? trace_hardirqs_off+0xb/0x10 > [ 356.948019] [] rcu_pending+0x12/0x40 > [ 356.948019] [] update_process_times+0x2a/0x60 > [ 356.948019] [] tick_sched_timer+0x55/0xc0 > [ 356.948019] [] ? tick_sched_timer+0x0/0xc0 > [ 356.948019] [] __run_hrtimer+0x70/0xe0 > [ 356.948019] [] ? tick_sched_timer+0x0/0xc0 > [ 356.948019] [] hrtimer_interrupt+0xed/0x190 > [ 356.948019] [] timer_interrupt+0x3b/0x50 > [ 356.948019] [] handle_IRQ_event+0x29/0x60 > [ 356.948019] [] handle_level_irq+0x65/0xe0 > [ 356.948019] [] ? handle_level_irq+0x0/0xe0 > [ 356.948019] [] ? > trace_hardirqs_on_caller+0x151/0x1c0 > [ 356.948019] [] ? common_interrupt+0x2c/0x34 > [ 356.948019] [] ? setup_per_zone_pages_min+0x11b/0x270 > [ 356.948019] [] ? copy_to_user+0x112/0x120 > [ 356.948019] [] ? read_kcore+0x385/0x440 > [ 356.948019] [] ? read_kcore+0x0/0x440 > [ 356.948019] [] ? proc_reg_read+0x5d/0x90 > [ 356.948019] [] ? vfs_read+0x99/0x160 > [ 356.948019] [] ? sysenter_exit+0xf/0x16 > [ 356.948019] [] ? proc_reg_read+0x0/0x90 > [ 356.948019] [] ? sys_read+0x3d/0x70 > [ 356.948019] [] ? sysenter_do_call+0x12/0x31 This message indicates that RCU is no longer getting any response from CPU 0. This is most likely because CPU 0 is spinning in an infinite loop with interrupts disabled -- otherwise CPU 0 would have reported itself and given a stack trace. So, do you have any way to collect CPU 0's stack trace? Kernel debugger or some such? > RCU options are: > # CONFIG_CLASSIC_RCU is not set > # CONFIG_TREE_RCU is not set > CONFIG_PREEMPT_RCU=y > CONFIG_RCU_TRACE=y > # CONFIG_TREE_RCU_TRACE is not set > CONFIG_PREEMPT_RCU_TRACE=y > CONFIG_RCU_TORTURE_TEST=m > > I tried this on another machine, which completely hangs after the cat, > i cant log in anymore via ssh. This one has > > # CONFIG_CLASSIC_RCU is not set > CONFIG_TREE_RCU=y > # CONFIG_PREEMPT_RCU is not set > CONFIG_RCU_TRACE=y > CONFIG_RCU_FANOUT=32 > # CONFIG_RCU_FANOUT_EXACT is not set > CONFIG_TREE_RCU_TRACE=y > # CONFIG_PREEMPT_RCU_TRACE is not set > CONFIG_RCU_TORTURE_TEST=m > CONFIG_RCU_CPU_STALL_DETECTOR=y If I am correct about the CPU going out to lunch with interrupts disabled, the type of RCU will be irrelevant. It would nevertheless be interesting to try Classic RCU with CPU stall detection enabled. Thanx, Paul