From mboxrd@z Thu Jan 1 00:00:00 1970 From: Michael Tokarev Subject: Re: [Qemu-devel] TSC in qem[-kvm] 1.1+ and in-kernel irqchip Date: Mon, 13 Aug 2012 17:16:22 +0400 Message-ID: <5028FE26.9090504@msgid.tls.msk.ru> References: <5024011F.2020200@msgid.tls.msk.ru> <20120810073346.GD24410@redhat.com> <5024D336.5090108@msgid.tls.msk.ru> <20120812081038.GJ3341@redhat.com> <50277631.2080808@msgid.tls.msk.ru> <5028FC02.4090205@siemens.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Cc: Avi Kivity , Gerd Hoffmann , KVM list , Gleb Natapov , qemu-devel To: Jan Kiszka Return-path: Received: from isrv.corpit.ru ([86.62.121.231]:56650 "EHLO isrv.corpit.ru" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751946Ab2HMNQ0 (ORCPT ); Mon, 13 Aug 2012 09:16:26 -0400 In-Reply-To: <5028FC02.4090205@siemens.com> Sender: kvm-owner@vger.kernel.org List-ID: On 13.08.2012 17:07, Jan Kiszka wrote: [] >> The bisecion leads to this commit: >> >> commit 17ee47418e65b1593defb30edbab33ccd47fc1f8 >> Merge: 13b0496 5d17c0d >> Author: Jan Kiszka >> Date: Tue Apr 10 16:26:23 2012 +0200 >> >> Merge commit '5d17c0d2df4998598e6002b27b8e47e792899a0f' into queues/qemu-merge [] >> Cc'ing Jan for help. The short story: tsc timer calibration >> broke in 1.1+ with in-kernel irqchip (only) for several >> apps (seabios and grub are two examples), the time is ticking >> about 100 times faster. In grub the timer is calibrated >> using pit. The above commit is the result of bisection. > > Did the versions you tested include the commit 0cdd3d1444 (kvm: i8254: > Fix conversion of in-kernel to userspace state)? While bisecting I didn't have this commit applied, since it were applied past (qemu)-1.1. It is included into qemu-kvm 1.1.0 (as 960d355dc60d9), and that version behaves _exactly_ the same - the time in grub is ticking 100 times faster. I mentioned in this thread that the problem persists in current qemu (and qemu-kvm) git too. I can repeat the bisection with this commit applied after the above "bad" commit. Should I? Thanks, /mjt From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([208.118.235.92]:49655) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1T0uVX-0002LQ-Qm for qemu-devel@nongnu.org; Mon, 13 Aug 2012 09:16:37 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1T0uVV-0003LH-Jc for qemu-devel@nongnu.org; Mon, 13 Aug 2012 09:16:35 -0400 Received: from isrv.corpit.ru ([86.62.121.231]:45754) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1T0uVV-0003HI-Bs for qemu-devel@nongnu.org; Mon, 13 Aug 2012 09:16:33 -0400 Message-ID: <5028FE26.9090504@msgid.tls.msk.ru> Date: Mon, 13 Aug 2012 17:16:22 +0400 From: Michael Tokarev MIME-Version: 1.0 References: <5024011F.2020200@msgid.tls.msk.ru> <20120810073346.GD24410@redhat.com> <5024D336.5090108@msgid.tls.msk.ru> <20120812081038.GJ3341@redhat.com> <50277631.2080808@msgid.tls.msk.ru> <5028FC02.4090205@siemens.com> In-Reply-To: <5028FC02.4090205@siemens.com> Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] TSC in qem[-kvm] 1.1+ and in-kernel irqchip List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Jan Kiszka Cc: qemu-devel , Gleb Natapov , Avi Kivity , KVM list , Gerd Hoffmann On 13.08.2012 17:07, Jan Kiszka wrote: [] >> The bisecion leads to this commit: >> >> commit 17ee47418e65b1593defb30edbab33ccd47fc1f8 >> Merge: 13b0496 5d17c0d >> Author: Jan Kiszka >> Date: Tue Apr 10 16:26:23 2012 +0200 >> >> Merge commit '5d17c0d2df4998598e6002b27b8e47e792899a0f' into queues/qemu-merge [] >> Cc'ing Jan for help. The short story: tsc timer calibration >> broke in 1.1+ with in-kernel irqchip (only) for several >> apps (seabios and grub are two examples), the time is ticking >> about 100 times faster. In grub the timer is calibrated >> using pit. The above commit is the result of bisection. > > Did the versions you tested include the commit 0cdd3d1444 (kvm: i8254: > Fix conversion of in-kernel to userspace state)? While bisecting I didn't have this commit applied, since it were applied past (qemu)-1.1. It is included into qemu-kvm 1.1.0 (as 960d355dc60d9), and that version behaves _exactly_ the same - the time in grub is ticking 100 times faster. I mentioned in this thread that the problem persists in current qemu (and qemu-kvm) git too. I can repeat the bisection with this commit applied after the above "bad" commit. Should I? Thanks, /mjt