From mboxrd@z Thu Jan  1 00:00:00 1970
From: Stefan Hajnoczi <stefanha@gmail.com>
Subject: Re: raw disks no longer work in latest kvm (kvm-88 was fine)
Date: Sat, 29 May 2010 10:55:18 +0100
Message-ID: <AANLkTikL_3uDvel-bcwslgElk27ly3Sz9cH3JyPwSaDD@mail.gmail.com>
References: <4B92BF97.4040001@nagafix.co.uk>
	<4BF7BD33.2090700@msgid.tls.msk.ru>
	<4BF7C172.2000508@nagafix.co.uk>
	<4BF8ED0C.2070004@nagafix.co.uk>
	<4BF91844.8030903@redhat.com>
	<4BF936BB.5000401@nagafix.co.uk>
	<4BF9391A.1080302@redhat.com>
	<4BF93EF5.5090106@nagafix.co.uk>
	<4BF945E4.8080209@redhat.com>
	<4BF9554D.3000100@nagafix.co.uk>
	<AANLkTimL6bE4XxxKDBq-n4PQo_gYD7utNF_nVU_068ce@mail.gmail.com>
	<4C00E1A3.50109@nagafix.co.uk>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: QUOTED-PRINTABLE
Cc: Avi Kivity <avi@redhat.com>, Michael Tokarev <mjt@tls.msk.ru>,
	Christoph Hellwig <hch@infradead.org>, kvm@vger.kernel.org
To: Antoine Martin <antoine@nagafix.co.uk>
Return-path: <kvm-owner@vger.kernel.org>
Received: from mail-vw0-f46.google.com ([209.85.212.46]:52568 "EHLO
	mail-vw0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S1751173Ab0E2JzT convert rfc822-to-8bit (ORCPT
	<rfc822;kvm@vger.kernel.org>); Sat, 29 May 2010 05:55:19 -0400
Received: by vws9 with SMTP id 9so2113636vws.19
        for <kvm@vger.kernel.org>; Sat, 29 May 2010 02:55:18 -0700 (PDT)
In-Reply-To: <4C00E1A3.50109@nagafix.co.uk>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>

On Sat, May 29, 2010 at 10:42 AM, Antoine Martin <antoine@nagafix.co.uk=
> wrote:
> Can someone explain the aio options?
> All I can find is this:
> # qemu-system-x86_64 -h | grep -i aio
> =A0 =A0 =A0 [,addr=3DA][,id=3Dname][,aio=3Dthreads|native]
> I assume it means the aio=3Dthreads emulates the kernel's aio with
> separate threads? And is therefore likely to be slower, right?
> Is there a reason why aio=3Dnative is not the default? Shouldn't
> aio=3Dthreads be the fallback?

aio=3Dthreads uses posix-aio-compat.c, a POSIX AIO-like implementation
using a thread pool.  Each thread services queued I/O requests using
blocking syscalls (e.g. preadv()/pwritev()).

aio=3Dnative uses Linux libaio, the native (non-POSIX) AIO interface.

I would expect that aio=3Dnative is faster but benchmarks show that thi=
s
isn't true for all workloads.

Stefan