From mboxrd@z Thu Jan  1 00:00:00 1970
From: Ingo Molnar <mingo@elte.hu>
Subject: Re: [PATCH 1/2] kvm tools: Respect ISR status in virtio header
Date: Sat, 7 May 2011 16:55:21 +0200
Message-ID: <20110507145521.GG2859@elte.hu>
References: <1304735660-10844-1-git-send-email-asias.hejun@gmail.com>
 <20110507093027.GD27657@elte.hu>
 <4DC545BA.3030501@us.ibm.com>
 <20110507140251.GB2859@elte.hu>
 <4DC55562.4070304@codemonkey.ws>
 <20110507144735.GD2859@elte.hu>
 <1304779960.10621.16.camel@jaguar>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Cc: Anthony Liguori <anthony@codemonkey.ws>,
	Asias He <asias.hejun@gmail.com>,
	"Michael S. Tsirkin" <mst@redhat.com>,
	Rusty Russell <rusty@rustcorp.com.au>,
	Mark McLoughlin <markmc@redhat.com>,
	Cyrill Gorcunov <gorcunov@gmail.com>,
	Sasha Levin <levinsasha928@gmail.com>,
	Prasad Joshi <prasadjoshi124@gmail.com>, kvm@vger.kernel.org
To: Pekka Enberg <penberg@kernel.org>
Return-path: <kvm-owner@vger.kernel.org>
Received: from mx3.mail.elte.hu ([157.181.1.138]:43692 "EHLO mx3.mail.elte.hu"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1755158Ab1EGOzf (ORCPT <rfc822;kvm@vger.kernel.org>);
	Sat, 7 May 2011 10:55:35 -0400
Content-Disposition: inline
In-Reply-To: <1304779960.10621.16.camel@jaguar>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>


* Pekka Enberg <penberg@kernel.org> wrote:

> On Sat, 2011-05-07 at 16:47 +0200, Ingo Molnar wrote:
> > Can you anything in the virtio protocol implementation that would explain 
> > networking lags, which seem to be caused by guest notifications either not be 
> > sent or being missed?
> > 
> > In particular this sequence:
> > 
> > > while pending_requests:
> > >    a = get_next_request();
> > >    process_next_request(a);
> > 
> > is apparently not what Qemu uses - so maybe there's some latent bug or some 
> > silly oversight somewhere.
> > 
> > It is suboptimal and i agree with you that the better sequence should be 
> > implemented, but the above *should* work, yet it does not.
> 
> Yes, so the performance benefits of Asias' patch aren't the interesting
> part but the fact that it fixes a real bug in our tool.

It could be the same like the mutex_lock() change: that too seemed to 'fix' the 
latency bug but we still do not understand the root cause of the 'stuck 
ring-buffer' situation.

I.e. some sort of timing related condition which goes away spuriously when 
unrelated but timing-relevant changes are done to the code.

And we'll continue to see these problems on and off, in probably all virtio 
drivers. virtio-console might be suffering from it, virtio-blk, etc. etc.

I'd suggest freezing changes to this driver until this bug is analyzed 
correctly...

Thanks,

	Ingo