From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from [140.186.70.92] (port=60880 helo=eggs.gnu.org)
	by lists.gnu.org with esmtp (Exim 4.43) id 1Pi6JY-0003em-EU
	for qemu-devel@nongnu.org; Wed, 26 Jan 2011 09:25:42 -0500
Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71)
	(envelope-from <jan.kiszka@siemens.com>) id 1Pi6JS-0006QS-Si
	for qemu-devel@nongnu.org; Wed, 26 Jan 2011 09:25:40 -0500
Received: from goliath.siemens.de ([192.35.17.28]:24831)
	by eggs.gnu.org with esmtp (Exim 4.71)
	(envelope-from <jan.kiszka@siemens.com>) id 1Pi6JS-0006Pf-K6
	for qemu-devel@nongnu.org; Wed, 26 Jan 2011 09:25:34 -0500
Message-ID: <4D402ED6.2020300@siemens.com>
Date: Wed, 26 Jan 2011 15:25:26 +0100
From: Jan Kiszka <jan.kiszka@siemens.com>
MIME-Version: 1.0
Subject: Re: [Qemu-devel] Re: [PATCH] kvm: Prevent dynticks race condition
	for !CONFIG_IOTHREAD
References: <1296034784-6513-1-git-send-email-stefanha@linux.vnet.ibm.com>	<4D401DA4.9020307@siemens.com>	<4D4027FE.8090808@siemens.com>
	<AANLkTi=Nr9V1RFUAj03d7O01HMNT=uEC7BnGrVY25VxV@mail.gmail.com>
In-Reply-To: <AANLkTi=Nr9V1RFUAj03d7O01HMNT=uEC7BnGrVY25VxV@mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
List-Id: qemu-devel.nongnu.org
List-Unsubscribe: <http://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=unsubscribe>
List-Archive: <http://lists.nongnu.org/archive/html/qemu-devel>
List-Post: <mailto:qemu-devel@nongnu.org>
List-Help: <mailto:qemu-devel-request@nongnu.org?subject=help>
List-Subscribe: <http://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=subscribe>
To: Stefan Hajnoczi <stefanha@gmail.com>
Cc: Avi Kivity <avi@redhat.com>, Marcelo Tosatti <mtosatti@redhat.com>, Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>, "qemu-devel@nongnu.org" <qemu-devel@nongnu.org>

On 2011-01-26 15:01, Stefan Hajnoczi wrote:
> On Wed, Jan 26, 2011 at 1:56 PM, Jan Kiszka <jan.kiszka@siemens.com> wrote:
>> Looks like this should also include SIGIO in case hpet is selected as
>> host timer.
> 
> Yeah that can't hurt although since hpet is periodic we don't risk
> forgetting to rearm the timer.
> 
> Can you explain the signalfd idea a little more?  I'm not sure that
> model would work for !CONFIG_IOTHREAD.  Maybe if QEMU itself uses
> signalfd but KVM sigmask is used to unblock the signals during guest
> code execution (and a solution for TCG too).

Will push a version I would like you to test as well in a few minutes.
This is already the concept: As under IOTHREAD, we will block SIGIO and
SIGALARM the whole time - except during KVM_RUN - and catch those
signals via qemu_signalfd in the main select loop. Of course, this
includes setting up the kvm signal mask properly as well (was already
part of my previously posted series for other reasons).

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux