From mboxrd@z Thu Jan  1 00:00:00 1970
From: Paolo Bonzini <pbonzini@redhat.com>
Subject: Re: [PATCH] KVM/X86: vpmu migration, make perf_event associated with
 vcpu thread
Date: Sat, 30 Nov 2013 12:08:49 +0100
Message-ID: <5299C741.40609@redhat.com>
References: <5298D5FC.8030709@gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Cc: Gleb Natapov <gleb@redhat.com>, kvm <kvm@vger.kernel.org>
To: Wang Hui <john.whui1985@gmail.com>
Return-path: <kvm-owner@vger.kernel.org>
Received: from mail-qe0-f45.google.com ([209.85.128.45]:37680 "EHLO
	mail-qe0-f45.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S1751624Ab3K3LIz (ORCPT <rfc822;kvm@vger.kernel.org>);
	Sat, 30 Nov 2013 06:08:55 -0500
Received: by mail-qe0-f45.google.com with SMTP id 6so11134607qea.4
        for <kvm@vger.kernel.org>; Sat, 30 Nov 2013 03:08:54 -0800 (PST)
In-Reply-To: <5298D5FC.8030709@gmail.com>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>

Il 29/11/2013 18:59, Wang Hui ha scritto:
> But when I wrote a test module to make IA32_PMC1 to count the event of unhalted
> cpu-cycles, after migration the value of IA32_PMC1 never grows up again. I found
> that after migration perf_event was created exactly, but when it was created,
> "current" is qemu's main thread which won't enter no-root mode, so the count of
> perf_event will never increase.

I think the fix is to apply the MSRs in QEMU's VCPU thread, through
run_on_cpu.  Interesting that I could not reproduce this when testing my
patches.

Paolo