From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754002Ab2GAQ4y (ORCPT ); Sun, 1 Jul 2012 12:56:54 -0400 Received: from mx1.redhat.com ([209.132.183.28]:17158 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753178Ab2GAQ4y (ORCPT ); Sun, 1 Jul 2012 12:56:54 -0400 Message-ID: <4FF08154.3050407@redhat.com> Date: Sun, 01 Jul 2012 12:56:52 -0400 From: Prarit Bhargava User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.2.17) Gecko/20110419 Red Hat/3.1.10-1.el6_0 Thunderbird/3.1.10 MIME-Version: 1.0 To: john stultz , Linux Kernel Subject: Re: [PATCH] [RFC] Potential fix for leapsecond caused futex related load spikes References: <4FF06CAB.9020800@redhat.com> In-Reply-To: <4FF06CAB.9020800@redhat.com> Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 07/01/2012 11:28 AM, Prarit Bhargava wrote: > John, > > I was hit by the futex issue as well. I saw your patch and quickly did a test > with top-of-tree + your patch using your reproducer. I end up with warnings > from the smp_call_function code followed by all sorts of deadlocks, etc. > > I haven't had a chance to debug and will start doing so shortly ... > > intel-canoepass-02 login: [ 108.479555] Clock: inserting leap second 23:59:60 UTC > [ 108.485199] ------------[ cut here ]------------ > [ 108.490368] WARNING: at kernel/smp.c:461 smp_call_function_many+0xbd/0x260() > [ 108.498236] Hardware name: S2600CP > [ 108.502060] Modules linked in: nfs nfs_acl auth_rpcgss fscache lockd sunrpc > kvm_intel igb coretemp kvm ixgbe ptp pps_core ioatdma mdio tpm_tis crc32c_intel > wmi joydev dca tpm lpc_ich ghash_clmulni_intel sb_edac mfd_core edac_core > i2c_i801 microcode pcspkr tpm_bios hid_generic isci libsas scsi_transport_sas > mgag200 i2c_algo_bit drm_kms_helper ttm drm i2c_core [last unloaded: scsi_wait_scan] > [ 108.540561] Pid: 1328, comm: leaptest Not tainted 3.5.0-rc4+ #4 > [ 108.547169] Hypervisor: no hypervisor > [ 108.551273] Call Trace: > [ 108.554019] [] warn_slowpath_common+0x7f/0xc0 > [ 108.561398] [] warn_slowpath_null+0x1a/0x20 > [ 108.567911] [] smp_call_function_many+0xbd/0x260 > [ 108.574931] [] ? hrtimer_wakeup+0x30/0x30 > [ 108.581242] [] ? hrtimer_wakeup+0x30/0x30 > [ 108.587560] [] smp_call_function+0x22/0x30 > [ 108.593982] [] on_each_cpu+0x28/0x70 > [ 108.599825] [] clock_was_set+0x1c/0x30 John, the issue is that clock_was_set calls on_each_cpu() -- which cannot be called from interrupt context as it calls smp_call_function_many(). I don't think you can call call_was_set() from update_wall_time() as update_wall_time() is called in interrupt context. Looking into it more ... P.