From mboxrd@z Thu Jan 1 00:00:00 1970 From: Anthony Liguori Subject: Re: [Qemu-devel] Re: [RFC: 0/2] patch for QEMU HPET periodic timer emulation to alleviate time drift Date: Mon, 07 Feb 2011 09:18:35 -0600 Message-ID: <4D500D4B.4070701@codemonkey.ws> References: <480481933.225059.1296734409954.JavaMail.root@zmail07.collab.prod.int.phx2.redhat.com> <1375835067.226263.1296740625327.JavaMail.root@zmail07.collab.prod.int.phx2.redhat.com> <4D4AC99A.2070803@siemens.com> <4D4B0B07.2040904@codemonkey.ws> <4D4B1CF8.8040800@web.de> <4D4B5F23.7040801@codemonkey.ws> <4D4BBF55.9060000@web.de> <4D4FE6BF.5080502@redhat.com> <4D4FEF81.1040603@codemonkey.ws> <4D4FF02F.2030309@redhat.com> <4D4FF24A.7000004@codemonkey.ws> <4D4FFD3B.2030903@siemens.com> <4D5001A0.8020503@codemonkey.ws> <4D5004FC.80000@siemens.com> <4D5007B9.7060806@codemonkey.ws> <4D5008B3.90505@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: Jan Kiszka , Glauber Costa , Ulrich Obergfell , kvm , qemu-devel To: Avi Kivity Return-path: Received: from mail-bw0-f46.google.com ([209.85.214.46]:42526 "EHLO mail-bw0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753616Ab1BGPSk (ORCPT ); Mon, 7 Feb 2011 10:18:40 -0500 Received: by bwz15 with SMTP id 15so4995657bwz.19 for ; Mon, 07 Feb 2011 07:18:39 -0800 (PST) In-Reply-To: <4D5008B3.90505@redhat.com> Sender: kvm-owner@vger.kernel.org List-ID: On 02/07/2011 08:58 AM, Avi Kivity wrote: > On 02/07/2011 04:54 PM, Anthony Liguori wrote: >> >>> Why the accumulated_ticks argument? >> >> Then the missing ticks is stored in the PeriodicTimer instead of >> storing it in the device state. That means we won't forget to save >> it in vmstate. >> >> It's convenient because then if we lose ticks in the PeriodicTimer >> layer, the devices have instance access to that info. When you do a >> read() from timerfd, it returns the number of coalesced events. >> That's the interface I had in my mind. >> >> We could just add a getter for PeriodicTimer and it would serve the >> same purpose. > > If a drift compensation policy is in effect, you don't need the missed > ticks, since you will get one callback for each (delayed) tick. If > there is no drift compensation policy, presumably you aren't > interested in lost ticks. So the ticks argument isn't very useful. > > On the other hand, we need a way to inject lost ticks into a > PeriodicTimer. If interrupt injection detects that an interrupt was > coalesced, we want the timer to schedule a new tick for us. As an optimization, if you can do something useful with the knowledge that we missed 20 ticks instead of waiting for 20 callbacks, it's useful to have. But adding a getter makes that possible and I agree that it clutters the interface for an edge use-case. Regards, Anthony Liguori