From mboxrd@z Thu Jan 1 00:00:00 1970 From: Priya Subject: Re: Domain-Virtual time Date: Fri, 26 Feb 2010 13:52:50 -0500 Message-ID: <5c3550fe1002261052y45bcbea9i813126ff91bff38c@mail.gmail.com> References: <5c3550fe1002250803q21decea2j101c53fe390b856c@mail.gmail.com> <5c3550fe1002260846s6775ac4dsabf90bba503d7b@mail.gmail.com> <5c3550fe1002260920n77c43e2cmbfd6ccc29be527c9@mail.gmail.com> Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="===============1188424291==" Return-path: In-Reply-To: List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Dan Magenheimer Cc: xen-devel@lists.xensource.com, Tim Deegan , xen-users@lists.xensource.com List-Id: xen-devel@lists.xenproject.org --===============1188424291== Content-Type: multipart/alternative; boundary=00504502c62efcde020480856745 --00504502c62efcde020480856745 Content-Type: text/plain; charset=ISO-8859-1 Thanks guys ! I am really grateful ! On Fri, Feb 26, 2010 at 12:46 PM, Dan Magenheimer < dan.magenheimer@oracle.com> wrote: > The timer_mode=X can be set in the "VM configuration" file > in which you specify things like the disk, memory, etc. > > Setting the Xen time source on your machine is done by > adding clocksource=xxxxx to your Xen boot line, but unless > your system has one broken clocksource and another is fine, > this is rarely the source of extreme drift in an HVM. > (And, by the way, doing something like enabling HPET in > your BIOS may cause Xen to choose a different clocksource. > Xen tries to choose what it deems to be the best, and > the clocksource= is just a manual override.) > > There are similar Linux kernel parameters that must > be specified within the boot-time grub file for the HVM. > These parameters vary from release to release (and > between 32-bit and 64-bit). Many more recent kernels > allow you to change the clocksource dynamically > using a sysfs file. > > Be aware that in some cases, a misspelled option name > or one that works fine on one version but not on the > currently booting Xen/kernel, may be completely and silently > ignored. > > And finally, time has been a moving target on Xen, > Linux, VMware, etc, so problems that appear in older > versions may have been resolved in later versions. > ANY time you post questions/problems regarding time > (or for that matter even non-time-related problems), > it is best to describe what version of Xen and kernels > you are using. > > ====== > > From: Priya [mailto:pbhat@acis.ufl.edu] > Sent: Friday, February 26, 2010 10:20 AM > To: Tim Deegan > Cc: Dan Magenheimer; xen-devel@lists.xensource.com; > xen-users@lists.xensource.com > Subject: Re: [Xen-devel] Domain-Virtual time > > Thanks Tim! > > I have some follow-up questions: > 1. How can I check/change the timer_mode on a XenServer HVM and on a Xen > HVM? > 2. How can I check/change the time-source on my machines? > 3. No; they are showing system time modified according to the timer_mode > setting, then extrapolated into a collection of virtual timers and > reconstituted by the linux kernel. > Can you please elaborate a little bit. Or please point me to a reference. > Thanks again ! > Pr > > > On Fri, Feb 26, 2010 at 12:01 PM, Tim Deegan > wrote: > Hi, > > At 16:46 +0000 on 26 Feb (1267202798), Priya wrote: > > The funny thing is that NTP is measuring a very different drift on my > > three machines (-189.206, -108.373 and -71.321 parts per million). The > > drift reported on Domain-0 is -11.393. So I don't think my machines > > are showing the system time. > No; they are showing system time modified according to the timer_mode > setting, then extrapolated into a collection of virtual timers and > reconstituted by the linux kernel. > > I don't know a lot about linux HVM because I always run Linux with PV > kernels, but on Windows I've found that the (virtual) ACPI PM timer is a > better time-source than the HPET or RTC. > > > In addition, the negative sign on the drift means that my machines are > > running faster that the real time, which is again puzzling. > Xen system time generally does drift forward, IIRC that's because xen > always tends to catch up to the fastest CPU, so cross-CPU jitter turns > into forward drift. Linux may be doing the same thing inside the HVM > VM, where the jitter is higher. > > Tim. > > -- > Tim Deegan > Principal Software Engineer, XenServer Engineering > Citrix Systems UK Ltd. (Company #02937203, SL9 0BG) > > --00504502c62efcde020480856745 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Thanks guys ! I am really grateful !

On F= ri, Feb 26, 2010 at 12:46 PM, Dan Magenheimer <dan.magenheimer@oracle.com> wrote:
The timer_mode=3D= X can be set in the "VM configuration" file
in which you specify things like the disk, memory, etc.

Setting the Xen time source on your machine is done by
adding clocksource=3Dxxxxx to your Xen boot line, but unless
your system has one broken clocksource and another is fine,
this is rarely the source of extreme drift in an HVM.
(And, by the way, doing something like enabling HPET in
your BIOS may cause Xen to choose a different clocksource.
Xen tries to choose what it deems to be the best, and
the clocksource=3D is just a manual override.)

There are similar Linux kernel parameters that must
be specified within the boot-time grub file for the HVM.
These parameters vary from release to release (and
between 32-bit and 64-bit). =A0Many more recent kernels
allow you to change the clocksource dynamically
using a sysfs file.

Be aware that in some cases, a misspelled option name
or one that works fine on one version but not on the
currently booting Xen/kernel, may be completely and silently
ignored.

And finally, time has been a moving target on Xen,
Linux, VMware, etc, so problems that appear in older
versions may have been resolved in later versions.
ANY time you post questions/problems regarding time
(or for that matter even non-time-related problems),
it is best to describe what version of Xen and kernels
you are using.

=3D=3D=3D=3D=3D=3D

From: Priya [mailto:pbhat@acis.ufl.ed= u]
Sent: Friday, February 26, 2010 10:20 AM
To: Tim Deegan
Cc: Dan Magenheimer; xen-d= evel@lists.xensource.com; xen-users@lists.xensource.com
Subject: Re: [Xen-devel] Domain-Virtual time

Thanks Tim!

I have some follow-up questions:
1. How can I check/change the timer_mode on a XenServer HVM and on a = Xen HVM?
2. How can I check/change the time-source on my machines?
3. No; they are showing system time modified according to the timer_mode
setting, then extrapolated into a collect= ion of virtual timers and
reconstituted by the linux kernel.
Can you please elaborate a little bit. Or please point me to a reference. Thanks again !
Pr


On Fri, Feb 26, 2010 at 12:01 PM, Tim Deegan <Tim.Deegan@citrix.com> wrote:
Hi,

At 16:46 +0000 on 26 Feb (1267202798), Priya wrote:
> The funny thing is that NTP is measuring a very different drift on my<= br> > three machines (-189.206, -108.373 and -71.321 parts per million). The=
> drift reported on Domain-0 is -11.393. So I don't think my machine= s
> are showing the system time.
No; they are showing system time modified according to the timer_mode
setting, then extrapolated into a collection of virtual timers and
reconstituted by the linux kernel.

I don't know a lot about linux HVM because I always run Linux with PV kernels, but on Windows I've found that the (virtual) ACPI PM timer is = a
better time-source than the HPET or RTC.

> In addition, the negative sign on the drift means that my machines are=
> running faster that the real time, which is again puzzling.
Xen system time generally does drift forward, IIRC that's because xen always tends to catch up to the fastest CPU, so cross-CPU jitter turns
into forward drift. =A0Linux may be doing the same thing inside the HVM
VM, where the jitter is higher.

Tim.

--
Tim Deegan <Tim.Deegan@citrix.c= om>
Principal Software Engineer, XenServer Engineering
Citrix Systems UK Ltd. =A0(Company #02937203, SL9 0BG)


--00504502c62efcde020480856745-- --===============1188424291== Content-Type: text/plain; charset="us-ascii" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit Content-Disposition: inline _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel --===============1188424291==--