From mboxrd@z Thu Jan 1 00:00:00 1970 From: Avi Kivity Subject: Re: [PATCH v1 0/5] KVM in-guest performance monitoring Date: Thu, 12 May 2011 12:53:56 +0300 Message-ID: <4DCBAE34.2000203@redhat.com> References: <1305129333-7456-1-git-send-email-avi@redhat.com> <20110512093309.GD8707@8bytes.org> <4DCBACC7.8080000@siemens.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: Joerg Roedel , linux-kernel@vger.kernel.org, kvm@vger.kernel.org, Ingo Molnar , Peter Zijlstra , Arnaldo Carvalho de Melo To: Jan Kiszka Return-path: Received: from mx1.redhat.com ([209.132.183.28]:1025 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755580Ab1ELJyN (ORCPT ); Thu, 12 May 2011 05:54:13 -0400 In-Reply-To: <4DCBACC7.8080000@siemens.com> Sender: kvm-owner@vger.kernel.org List-ID: On 05/12/2011 12:47 PM, Jan Kiszka wrote: > > > > Anyway, I thought about a paravirt-approach instead of implementing a > > real PMU... But there are certainly good reasons for both. > > Paravirt is taking away the pressure from CPU vendors to do their virt > extensions properly - and doesn't help with unmodifiable OSes. Yes. In the case of the PMU things are less clear, since they are very model specific. In the case of Linux, you have to lie to the guest and present a model number that it doesn't recognize, otherwise it prefers the model-specific PMU to the architectural PMU. I though of adding a kvm cpuid bit that says "prefer the architectural pmu" to work around this issue. -- error compiling committee.c: too many arguments to function