XEN increases idle power consumption by 25%

XEN increases idle power consumption by 25%

[Martin Meier]

[-- Attachment #1.1: Type: text/plain, Size: 644 bytes --]

Hi,

I'm seeing a change in AC idle power consumption from ~23W to ~29W
when changing form 3.1-rc2 to 3.1-rc2 on top of xen-4.2-unstable  with no
domUs running.

It seems that the CPU never gets into the low power idle state for the whole
CPU-package 'PC6'
and the Vcore voltage never gets below +1.06 V (It get down to 0.7V w/o
xen).


(XEN) 'c' pressed -> printing ACPI Cx structures
(XEN) ==cpu0==
(XEN) active state:
C-1


What does the state 'C-1' mean?

Used hardware:
Board: Intel DQ67SW
CPU: Intel i5 2400

Is this whole CPU-package PC6 state supposed to work in xen/dom0?

Should I file a bug report for this problem?


Thanks,
Martin

[-- Attachment #1.2: Type: text/html, Size: 883 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: XEN increases idle power consumption by 25%

[Jan Beulich]

>>> On 24.08.11 at 14:03, Martin Meier <martin4meier@googlemail.com> wrote:
> Hi,
> 
> I'm seeing a change in AC idle power consumption from ~23W to ~29W
> when changing form 3.1-rc2 to 3.1-rc2 on top of xen-4.2-unstable  with no
> domUs running.
> 
> It seems that the CPU never gets into the low power idle state for the whole
> CPU-package 'PC6'
> and the Vcore voltage never gets below +1.06 V (It get down to 0.7V w/o
> xen).

This has been discussed recently; see e.g. the thread starting at
http://lists.xensource.com/archives/html/xen-devel/2011-08/msg00511.html.

Jan

> (XEN) 'c' pressed -> printing ACPI Cx structures
> (XEN) ==cpu0==
> (XEN) active state:
> C-1
> 
> 
> What does the state 'C-1' mean?
> 
> Used hardware:
> Board: Intel DQ67SW
> CPU: Intel i5 2400
> 
> Is this whole CPU-package PC6 state supposed to work in xen/dom0?
> 
> Should I file a bug report for this problem?
> 
> 
> Thanks,
> Martin

Re: XEN increases idle power consumption by 25%

[Jeremy Fitzhardinge]

On 08/24/2011 06:29 AM, Jan Beulich wrote:
>>>> On 24.08.11 at 14:03, Martin Meier <martin4meier@googlemail.com> wrote:
>> Hi,
>>
>> I'm seeing a change in AC idle power consumption from ~23W to ~29W
>> when changing form 3.1-rc2 to 3.1-rc2 on top of xen-4.2-unstable  with no
>> domUs running.
>>
>> It seems that the CPU never gets into the low power idle state for the whole
>> CPU-package 'PC6'
>> and the Vcore voltage never gets below +1.06 V (It get down to 0.7V w/o
>> xen).
> This has been discussed recently; see e.g. the thread starting at
> http://lists.xensource.com/archives/html/xen-devel/2011-08/msg00511.html.

Also, upstream Linux doesn't have the patches to pass the power state
stuff through to Xen, which I guess it part of it?

    J

Re: XEN increases idle power consumption by 25%

[Konrad Rzeszutek Wilk]

On Wed, Aug 24, 2011 at 12:33:01PM -0700, Jeremy Fitzhardinge wrote:
> On 08/24/2011 06:29 AM, Jan Beulich wrote:
> >>>> On 24.08.11 at 14:03, Martin Meier <martin4meier@googlemail.com> wrote:
> >> Hi,
> >>
> >> I'm seeing a change in AC idle power consumption from ~23W to ~29W
> >> when changing form 3.1-rc2 to 3.1-rc2 on top of xen-4.2-unstable  with no

Um, from the same kernel to the same kernel?

> >> domUs running.
> >>
> >> It seems that the CPU never gets into the low power idle state for the whole
> >> CPU-package 'PC6'
> >> and the Vcore voltage never gets below +1.06 V (It get down to 0.7V w/o
> >> xen).
> > This has been discussed recently; see e.g. the thread starting at
> > http://lists.xensource.com/archives/html/xen-devel/2011-08/msg00511.html.
> 
> Also, upstream Linux doesn't have the patches to pass the power state
> stuff through to Xen, which I guess it part of it?

And you can test the xen-cpu-idle patches in #testing if you are up for
seeing if that is the cause of your power consumption issues.

git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen.git #testing

specifically.

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

Re: XEN increases idle power consumption by 25%

[Jan Beulich]

>>> On 24.08.11 at 21:33, Jeremy Fitzhardinge <jeremy@goop.org> wrote:
> On 08/24/2011 06:29 AM, Jan Beulich wrote:
>>>>> On 24.08.11 at 14:03, Martin Meier <martin4meier@googlemail.com> wrote:
>>> Hi,
>>>
>>> I'm seeing a change in AC idle power consumption from ~23W to ~29W
>>> when changing form 3.1-rc2 to 3.1-rc2 on top of xen-4.2-unstable  with no
>>> domUs running.
>>>
>>> It seems that the CPU never gets into the low power idle state for the whole
>>> CPU-package 'PC6'
>>> and the Vcore voltage never gets below +1.06 V (It get down to 0.7V w/o
>>> xen).
>> This has been discussed recently; see e.g. the thread starting at
>> http://lists.xensource.com/archives/html/xen-devel/2011-08/msg00511.html.
> 
> Also, upstream Linux doesn't have the patches to pass the power state
> stuff through to Xen, which I guess it part of it?

Oh yes, of course.

Jan

Re: XEN increases idle power consumption by 25%

[Martin Meier]

On Thu, Aug 25, 2011 at 12:25 AM, Konrad Rzeszutek Wilk
<konrad.wilk@oracle.com> wrote:

> And you can test the xen-cpu-idle patches in #testing if you are up for
> seeing if that is the cause of your power consumption issues.
>
> git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen.git #testing

OK, I switched from 3.1-rc2 to your 3.0.3 kernel - I now see ~23W w/
or w/o xen :-)

Thanks!


Martin

Re: XEN increases idle power consumption by 25%

[Konrad Rzeszutek Wilk]

On Mon, Aug 29, 2011 at 05:30:58PM +0200, Martin Meier wrote:
> On Thu, Aug 25, 2011 at 12:25 AM, Konrad Rzeszutek Wilk
> <konrad.wilk@oracle.com> wrote:
> 
> > And you can test the xen-cpu-idle patches in #testing if you are up for
> > seeing if that is the cause of your power consumption issues.
> >
> > git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen.git #testing
> 
> OK, I switched from 3.1-rc2 to your 3.0.3 kernel - I now see ~23W w/
> or w/o xen :-)

Ok, so that is the #master branch right? And ~23W was what you wanted
right?

Re: XEN increases idle power consumption by 25%

[Martin Meier]

On Mon, Aug 29, 2011 at 10:09 PM, Konrad Rzeszutek Wilk
<konrad.wilk@oracle.com> wrote:
> On Mon, Aug 29, 2011 at 05:30:58PM +0200, Martin Meier wrote:
>> On Thu, Aug 25, 2011 at 12:25 AM, Konrad Rzeszutek Wilk
>> <konrad.wilk@oracle.com> wrote:
>>
>> > And you can test the xen-cpu-idle patches in #testing if you are up for
>> > seeing if that is the cause of your power consumption issues.
>> >
>> > git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen.git #testing
>>
>> OK, I switched from 3.1-rc2 to your 3.0.3 kernel - I now see ~23W w/
>> or w/o xen :-)
>
> Ok, so that is the #master branch right? And ~23W was what you wanted
> right?

Yes, and yes.

I tried #testing today and it worked OK, too.

FYI, after a minor change to the kernel .config:

-CONFIG_ACPI_PROCESSOR=y
+CONFIG_ACPI_PROCESSOR=m

-CONFIG_ACPI_PROCESSOR_XEN=y
+CONFIG_ACPI_PROCESSOR_XEN=m

I noticed this:

[   11.698368] ------------[ cut here ]------------
[   11.698386] WARNING: at drivers/cpuidle/driver.c:62
cpuidle_unregister_driver+0x2a/0x61()
[   11.698392] Hardware name:
[   11.698396] invalid cpuidle_unregister_driver(acpi_idle)
[   11.698400] Modules linked in: processor(+) snd evdev soundcore
i2c_i801 rtc_cmos psmouse pcspkr serio_raw acpi_processor
snd_page_all]
[   11.698451] Pid: 560, comm: modprobe Tainted: G         C
3.1.0-rc4-konrad-1+ #17
[   11.698456] Call Trace:
[   11.698465]  [<ffffffff810587c3>] warn_slowpath_common+0x80/0x98
[   11.698471]  [<ffffffffa0117000>] ? 0xffffffffa0116fff
[   11.698501]  [<ffffffff8105886f>] warn_slowpath_fmt+0x41/0x43
[   11.698508]  [<ffffffff8132959b>] cpuidle_unregister_driver+0x2a/0x61
[   11.698516]  [<ffffffffa01170e3>] acpi_processor_init+0xe3/0xe8 [processor]
[   11.698524]  [<ffffffff8100207a>] do_one_initcall+0x7a/0x150
[   11.698531]  [<ffffffff8108d968>] sys_init_module+0x7e/0x1c7
[   11.698539]  [<ffffffff813de7d2>] system_call_fastpath+0x16/0x1b
[   11.698543] ---[ end trace f1b3f5c476845ee5 ]---

and a high idle power consumption again.

Re: XEN increases idle power consumption by 25%

[Konrad Rzeszutek Wilk]

On Tue, Aug 30, 2011 at 02:19:17PM +0200, Martin Meier wrote:
> On Mon, Aug 29, 2011 at 10:09 PM, Konrad Rzeszutek Wilk
> <konrad.wilk@oracle.com> wrote:
> > On Mon, Aug 29, 2011 at 05:30:58PM +0200, Martin Meier wrote:
> >> On Thu, Aug 25, 2011 at 12:25 AM, Konrad Rzeszutek Wilk
> >> <konrad.wilk@oracle.com> wrote:
> >>
> >> > And you can test the xen-cpu-idle patches in #testing if you are up for
> >> > seeing if that is the cause of your power consumption issues.
> >> >
> >> > git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen.git #testing
> >>
> >> OK, I switched from 3.1-rc2 to your 3.0.3 kernel - I now see ~23W w/
> >> or w/o xen :-)
> >
> > Ok, so that is the #master branch right? And ~23W was what you wanted
> > right?
> 
> Yes, and yes.
> 
> I tried #testing today and it worked OK, too.
> 
> FYI, after a minor change to the kernel .config:
> 
> -CONFIG_ACPI_PROCESSOR=y
> +CONFIG_ACPI_PROCESSOR=m
> 
> -CONFIG_ACPI_PROCESSOR_XEN=y
> +CONFIG_ACPI_PROCESSOR_XEN=m

Ah, so the baremetal processor gets in first and then the Xen
can't register. 
> 
> I noticed this:
> 
> [   11.698368] ------------[ cut here ]------------
> [   11.698386] WARNING: at drivers/cpuidle/driver.c:62
> cpuidle_unregister_driver+0x2a/0x61()
> [   11.698392] Hardware name:
> [   11.698396] invalid cpuidle_unregister_driver(acpi_idle)
> [   11.698400] Modules linked in: processor(+) snd evdev soundcore
> i2c_i801 rtc_cmos psmouse pcspkr serio_raw acpi_processor
> snd_page_all]
> [   11.698451] Pid: 560, comm: modprobe Tainted: G         C
> 3.1.0-rc4-konrad-1+ #17
> [   11.698456] Call Trace:
> [   11.698465]  [<ffffffff810587c3>] warn_slowpath_common+0x80/0x98
> [   11.698471]  [<ffffffffa0117000>] ? 0xffffffffa0116fff
> [   11.698501]  [<ffffffff8105886f>] warn_slowpath_fmt+0x41/0x43
> [   11.698508]  [<ffffffff8132959b>] cpuidle_unregister_driver+0x2a/0x61
> [   11.698516]  [<ffffffffa01170e3>] acpi_processor_init+0xe3/0xe8 [processor]
> [   11.698524]  [<ffffffff8100207a>] do_one_initcall+0x7a/0x150
> [   11.698531]  [<ffffffff8108d968>] sys_init_module+0x7e/0x1c7
> [   11.698539]  [<ffffffff813de7d2>] system_call_fastpath+0x16/0x1b
> [   11.698543] ---[ end trace f1b3f5c476845ee5 ]---
> 
> and a high idle power consumption again.

Yup. That would be happen. Thanks for testing that. Will have to
find some way of making the Xen one run first. Or deregistering
the baremetal one perhaps.