thr3ads.net - Xen devel - [Xen-devel] Weird NOW() values after host S3 resume [May 2008]

If this information is useful, please help other people find it:
Share via:

Guy Zana

2008-May-21 16:09 UTC

[Xen-devel] Weird NOW() values after host S3 resume

Hi all,

 

It seems that something is messed up with the calibration of the TSCs after
resuming from sleep.

 

First, after resume NOW() issued from CPU0 gives a value that is much lower than
before sleep, and it is probably returned to the same (default?) value after
each resume.

Second, after each resume there is an increased differences between CPU0 and
CPU1 NOW() values, I''m not really sure about this one because I
didn''t tried to sync my test against the HPET so take that into
consideration.

 

Can you check this out please?

 

The chipset is Intel GM965/ICH8, the cpu is T7300.

Observed in Xen 3.2 stable release.

 

Thanks,

Guy.



_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Xu, Dongxiao

2008-May-22 02:49 UTC

head link

[Xen-devel] RE: Weird NOW() values after host S3 resume

Hi, Guy
    Could you have a try of the latest Xen-unstable tree to see if this issue
still exists? I remember some S3 time patches are added after Xen-3.2 stable
release.

Thanks!
-- Dongxiao

________________________________________
From: Guy Zana [mailto:guy@neocleus.com] 
Sent: 2008年5月22日 0:09
To: xen-devel@lists.xensource.com
Cc: Tian, Kevin; Xu, Dongxiao
Subject: Weird NOW() values after host S3 resume

Hi all,

It seems that something is messed up with the calibration of the TSCs after
resuming from sleep.

First, after resume NOW() issued from CPU0 gives a value that is much lower than
before sleep, and it is probably returned to the same (default?) value after
each resume.
Second, after each resume there is an increased differences between CPU0 and
CPU1 NOW() values, I''m not really sure about this one because I
didn''t tried to sync my test against the HPET so take that into
consideration.

Can you check this out please?

The chipset is Intel GM965/ICH8, the cpu is T7300.
Observed in Xen 3.2 stable release.

Thanks,
Guy.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Xu, Dongxiao

2008-May-22 07:11 UTC

head link

RE: [Xen-devel] RE: Weird NOW() values after host S3 resume

Or you may port Xen C/S 16688 and 16882 to your tree and see if this issue still
exists. :-)

Best regards,
-- Dongxiao

-----Original Message-----
From: xen-devel-bounces@lists.xensource.com
[mailto:xen-devel-bounces@lists.xensource.com] On Behalf Of Xu, Dongxiao
Sent: 2008年5月22日 10:50
To: Guy Zana; xen-devel@lists.xensource.com
Cc: Tian, Kevin
Subject: [Xen-devel] RE: Weird NOW() values after host S3 resume

Hi, Guy
    Could you have a try of the latest Xen-unstable tree to see if this issue
still exists? I remember some S3 time patches are added after Xen-3.2 stable
release.

Thanks!
-- Dongxiao

________________________________________
From: Guy Zana [mailto:guy@neocleus.com] 
Sent: 2008年5月22日 0:09
To: xen-devel@lists.xensource.com
Cc: Tian, Kevin; Xu, Dongxiao
Subject: Weird NOW() values after host S3 resume

Hi all,

It seems that something is messed up with the calibration of the TSCs after
resuming from sleep.

First, after resume NOW() issued from CPU0 gives a value that is much lower than
before sleep, and it is probably returned to the same (default?) value after
each resume.
Second, after each resume there is an increased differences between CPU0 and
CPU1 NOW() values, I''m not really sure about this one because I
didn''t tried to sync my test against the HPET so take that into
consideration.

Can you check this out please?

The chipset is Intel GM965/ICH8, the cpu is T7300.
Observed in Xen 3.2 stable release.

Thanks,
Guy.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Keir Fraser

2008-May-22 07:15 UTC

head link

Re: [Xen-devel] RE: Weird NOW() values after host S3 resume

Xen 3.2.0 actually has the first of those changesets. If you upgrade to Xen
3.2.1 then you have the second of them also.

 -- Keir

On 22/5/08 08:11, "Xu, Dongxiao" <dongxiao.xu@intel.com> wrote:
> Or you may port Xen C/S 16688 and 16882 to your tree and see if this issue
> still exists. :-)
> 
> Best regards,
> -- Dongxiao
> 
> -----Original Message-----
> From: xen-devel-bounces@lists.xensource.com
> [mailto:xen-devel-bounces@lists.xensource.com] On Behalf Of Xu, Dongxiao
> Sent: 2008年5月22日 10:50
> To: Guy Zana; xen-devel@lists.xensource.com
> Cc: Tian, Kevin
> Subject: [Xen-devel] RE: Weird NOW() values after host S3 resume
> 
> Hi, Guy
>     Could you have a try of the latest Xen-unstable tree to see if this
issue
> still exists? I remember some S3 time patches are added after Xen-3.2
stable
> release. 
> 
> Thanks!
> -- Dongxiao
> 
> ________________________________________
> From: Guy Zana [mailto:guy@neocleus.com]
> Sent: 2008年5月22日 0:09
> To: xen-devel@lists.xensource.com
> Cc: Tian, Kevin; Xu, Dongxiao
> Subject: Weird NOW() values after host S3 resume
> 
> Hi all,
> 
> It seems that something is messed up with the calibration of the TSCs after
> resuming from sleep.
> 
> First, after resume NOW() issued from CPU0 gives a value that is much lower
> than before sleep, and it is probably returned to the same (default?) value
> after each resume.
> Second, after each resume there is an increased differences between CPU0
and
> CPU1 NOW() values, I''m not really sure about this one because I
didn''t tried
> to sync my test against the HPET so take that into consideration.
> 
> Can you check this out please?
> 
> The chipset is Intel GM965/ICH8, the cpu is T7300.
> Observed in Xen 3.2 stable release.
> 
> Thanks,
> Guy.
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Guy Zana

2008-May-25 08:35 UTC

head link

RE: [Xen-devel] RE: Weird NOW() values after host S3 resume

Sorry for the late replay,

xen-unstable.hg suffer from a time went backwards bug after resume, I
didn''t tried to investigate any further.
in xen-3.2.0 it is sufficient to re-enable the hpet to solve most of the
problems, but still NOW() returns a lower value than before sleep...
> -----Original Message-----
> From: Keir Fraser [mailto:keir.fraser@eu.citrix.com]
> Sent: Thursday, May 22, 2008 10:15 AM
> To: Xu, Dongxiao; Guy Zana; xen-devel@lists.xensource.com
> Cc: Tian, Kevin
> Subject: Re: [Xen-devel] RE: Weird NOW() values after host S3 resume
> 
> Xen 3.2.0 actually has the first of those changesets. If you upgrade to
> Xen
> 3.2.1 then you have the second of them also.
> 
>  -- Keir
> 
> On 22/5/08 08:11, "Xu, Dongxiao" <dongxiao.xu@intel.com>
wrote:
> 
> > Or you may port Xen C/S 16688 and 16882 to your tree and see if this
> issue
> > still exists. :-)
> >
> > Best regards,
> > -- Dongxiao
> >
> > -----Original Message-----
> > From: xen-devel-bounces@lists.xensource.com
> > [mailto:xen-devel-bounces@lists.xensource.com] On Behalf Of Xu,
> Dongxiao
> > Sent: 2008年5月22日 10:50
> > To: Guy Zana; xen-devel@lists.xensource.com
> > Cc: Tian, Kevin
> > Subject: [Xen-devel] RE: Weird NOW() values after host S3 resume
> >
> > Hi, Guy
> >     Could you have a try of the latest Xen-unstable tree to see if
> this issue
> > still exists? I remember some S3 time patches are added after Xen-3.2
> stable
> > release.
> >
> > Thanks!
> > -- Dongxiao
> >
> > ________________________________________
> > From: Guy Zana [mailto:guy@neocleus.com]
> > Sent: 2008年5月22日 0:09
> > To: xen-devel@lists.xensource.com
> > Cc: Tian, Kevin; Xu, Dongxiao
> > Subject: Weird NOW() values after host S3 resume
> >
> > Hi all,
> >
> > It seems that something is messed up with the calibration of the TSCs
> after
> > resuming from sleep.
> >
> > First, after resume NOW() issued from CPU0 gives a value that is much
> lower
> > than before sleep, and it is probably returned to the same (default?)
> value
> > after each resume.
> > Second, after each resume there is an increased differences between
> CPU0 and
> > CPU1 NOW() values, I''m not really sure about this one because
I
> didn''t tried
> > to sync my test against the HPET so take that into consideration.
> >
> > Can you check this out please?
> >
> > The chipset is Intel GM965/ICH8, the cpu is T7300.
> > Observed in Xen 3.2 stable release.
> >
> > Thanks,
> > Guy.
> >
> > _______________________________________________
> > Xen-devel mailing list
> > Xen-devel@lists.xensource.com
> > http://lists.xensource.com/xen-devel
> >
> > _______________________________________________
> > Xen-devel mailing list
> > Xen-devel@lists.xensource.com
> > http://lists.xensource.com/xen-devel
> 

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Xen devel - May 2008 - Weird NOW() values after host S3 resume

[Xen-devel] Weird NOW() values after host S3 resume

[Xen-devel] RE: Weird NOW() values after host S3 resume

RE: [Xen-devel] RE: Weird NOW() values after host S3 resume

Re: [Xen-devel] RE: Weird NOW() values after host S3 resume

RE: [Xen-devel] RE: Weird NOW() values after host S3 resume