From mboxrd@z Thu Jan  1 00:00:00 1970
From: "Jan Beulich" <JBeulich@novell.com>
Subject: Re: Re: DomU lockups after resume from S3 on Core
	i5 processors
Date: Tue, 06 Jul 2010 10:57:50 +0100
Message-ID: <4C331A3E0200007800009B3A@vpn.id2.novell.com>
References: <4C31B629.7070601@invisiblethingslab.com>
	<4C324E8C.4030305@invisiblethingslab.com>
	<4C3257B2.1040002@invisiblethingslab.com> <4C32602A.8070305@goop.org>
	<4C326241.2030503@invisiblethingslab.com> <4C3267FB.3070202@goop.org>
	<4C33085E0200007800009AE1@vpn.id2.novell.com>
	<4C32F078.2080009@invisiblethingslab.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: quoted-printable
Return-path: <xen-devel-bounces@lists.xensource.com>
In-Reply-To: <4C32F078.2080009@invisiblethingslab.com>
Content-Disposition: inline
List-Unsubscribe: <http://lists.xensource.com/mailman/listinfo/xen-devel>,
	<mailto:xen-devel-request@lists.xensource.com?subject=unsubscribe>
List-Post: <mailto:xen-devel@lists.xensource.com>
List-Help: <mailto:xen-devel-request@lists.xensource.com?subject=help>
List-Subscribe: <http://lists.xensource.com/mailman/listinfo/xen-devel>,
	<mailto:xen-devel-request@lists.xensource.com?subject=subscribe>
Sender: xen-devel-bounces@lists.xensource.com
Errors-To: xen-devel-bounces@lists.xensource.com
To: Joanna Rutkowska <joanna@invisiblethingslab.com>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>, xen-devel@lists.xensource.com
List-Id: xen-devel@lists.xenproject.org

>>> On 06.07.10 at 10:59, Joanna Rutkowska <joanna@invisiblethingslab.com> =
wrote:
> On 07/06/10 10:41, Jan Beulich wrote:
>> Did either of you try disabling the setting of sched_clock_stable in
>> arch/x86/kernel/cpu/intel.c:early_init_intel()? I found this to be a
>> requirement in our pv kernels (though in connection with the use of
>> C-states, not with S3).
>>=20
> Before I try it -- can you explain what would be the theory behind it,
> specifically how this would be related to HT? Clearly it is a HT
> problem, and intuitively, I would expect this to be a Xen-side problem,
> rather than DomU-side?

The HT connection is only a vague one, as Jeremy also hinted at in
his reply. The issue is that with sched_clock_stable set,
sched_clock_cpu() (and cpu_clock()) gets short cut to sched_clock(),
hence becoming susceptible to eventual non-monotonic behavior of
that function.

In any case, it's a wild guess only, attributed to the partially-hung
observations you made matching my observations (on new Intel
CPUs only) prior to addressing this issue.

Jan