From mboxrd@z Thu Jan 1 00:00:00 1970 From: Avi Kivity Subject: Re: Biweekly KVM Test report, kernel 7597f... qemu 1c45e... Date: Tue, 25 Aug 2009 13:20:12 +0300 Message-ID: <4A93BADC.4090106@redhat.com> References: <0463F45F3606F4428ED35AC8C709F92E0897FE903F@pdsmsx502.ccr.corp.intel.com> <4A894A5B.2080901@redhat.com> <0463F45F3606F4428ED35AC8C709F92E0899F3DD99@pdsmsx502.ccr.corp.intel.com> <4A8BB304.6020202@redhat.com> <0463F45F3606F4428ED35AC8C709F92E0899F3E6E2@pdsmsx502.ccr.corp.intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: "'kvm-devel'" To: "Xu, Jiajun" , Marcelo Tosatti Return-path: Received: from mx1.redhat.com ([209.132.183.28]:12397 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751680AbZHYKUU (ORCPT ); Tue, 25 Aug 2009 06:20:20 -0400 In-Reply-To: <0463F45F3606F4428ED35AC8C709F92E0899F3E6E2@pdsmsx502.ccr.corp.intel.com> Sender: kvm-owner@vger.kernel.org List-ID: On 08/21/2009 10:14 AM, Xu, Jiajun wrote: > I found the migration failure is caused by a configuration mistake on our testing machine. Now 64-bit migration works well. > But I found on PAE host, migration will cause host kernel call trace. > > > Pid: 12053, comm: qemu-system-x86 Tainted: G D (2.6.31-rc2 #1) > EIP: 0060:[] EFLAGS: 00210202 CPU: 0 > EIP is at lock_hrtimer_base+0x11/0x33 > EAX: f5d1541c EBX: 00000010 ECX: 000004a9 EDX: f5c1bc7c > ESI: f5d1541c EDI: f5c1bc7c EBP: f5c1bc74 ESP: f5c1bc68 > DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0068 > Process qemu-system-x86 (pid: 12053, ti=f5c1b000 task=f61cb410 > task.ti=f5c1b000) > Stack: > f5d1541c ffffffff 000004a9 f5c1bc8c c043e097 f9b7f7cb f5d1541c 00000000 > <0> 000004a9 f5c1bc98 c043e0f0 f5d153d0 f5c1bcb0 f9b9b4df 00000000 bfd8a102 > <0> f3c1e000 f5d15440 f5c1bcc0 f9b9b56d bfd8a10c f3c1e000 f5c1bda0 f9b8c26b > Call Trace: > [] ? hrtimer_try_to_cancel+0x16/0x62 > [] ? kvm_flush_remote_tlbs+0xd/0x1a [kvm] > [] ? hrtimer_cancel+0xd/0x18 > [] ? pit_load_count+0x98/0x9e [kvm] > [] ? kvm_pit_load_count+0x21/0x35 [kvm] > Marcelo, any idea? Looks like the PIT was reloaded, but the hrtimer wasn't initialized? -- error compiling committee.c: too many arguments to function