From mboxrd@z Thu Jan  1 00:00:00 1970
From: Takuya Yoshikawa <takuya.yoshikawa@gmail.com>
Subject: Re: [PATCH 3/4] KVM: Count the number of dirty pages for dirty
 logging
Date: Wed, 28 Dec 2011 00:03:22 +0900
Message-ID: <20111228000322.33585bc7450433775326292e@gmail.com>
References: <20111114182041.43570cdf.yoshikawa.takuya@oss.ntt.co.jp>
	<20111114182334.f57fbeae.yoshikawa.takuya@oss.ntt.co.jp>
	<4EC0E85F.80004@redhat.com>
	<4EF00F2F.6080305@oss.ntt.co.jp>
	<20111223111451.GD24308@amt.cnet>
	<20111224115251.2aaa9549cb0a98de79f7d918@gmail.com>
	<20111227135017.GB23270@amt.cnet>
	<4EF9D024.8050908@redhat.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Cc: Marcelo Tosatti <mtosatti@redhat.com>,
	Takuya Yoshikawa <yoshikawa.takuya@oss.ntt.co.jp>,
	kvm@vger.kernel.org
To: Avi Kivity <avi@redhat.com>
Return-path: <kvm-owner@vger.kernel.org>
Received: from mail-iy0-f174.google.com ([209.85.210.174]:50667 "EHLO
	mail-iy0-f174.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S1754445Ab1L0PD1 (ORCPT <rfc822;kvm@vger.kernel.org>);
	Tue, 27 Dec 2011 10:03:27 -0500
Received: by iaeh11 with SMTP id h11so20505462iae.19
        for <kvm@vger.kernel.org>; Tue, 27 Dec 2011 07:03:26 -0800 (PST)
In-Reply-To: <4EF9D024.8050908@redhat.com>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>

Avi Kivity <avi@redhat.com> wrote:

> > It should not be necessary, atomicity of updates to
> > memslot->dirty_bitmap, because of RCU updates to the memslot pointer
> > (note memslot = gfn_to_memslot(kvm, gfn); in mark_page_dirty).
> >
> > The order is:
> >
> > - update page data.
> > - grab memslot pointer.
> > - set page bit in memslot.
> >
> 
> What if both grab the same memslot pointer, but set different bits?

(Sorry for my bad wording, that is what I wanted to explain.)

There seems to be possible concurrent __set_bit()'s to the same word
which will result in bit loss.

I was thinking to change the kvm-unit-tests' dirty log test to write
to two consecutive pages concurrently.  Is this good for debugging this?

	Takuya