Xen devel - Feb 2009 - [PATCH 0/4] ACPI: Backport to make dom0 throttling works with MSR

This patchset backport a series of T-State patches from kernel upstream to make
dom0 ACPI throttling works with MSR( Fixed H/W). Meanwhile, it enable
THERM_CONTROL MSR write for dom0 even cpufreq=xen.

[PATCH 1/4] ACPI: Backport missing part for T-State MSR support
[PATCH 2/4] ACPI: Backport to get correct throtting info after evaluating _PDC
[PATCH 3/4] ACPI: Backport to support ACPI throttling s/w coordination
[PATCH 4/4] ACPI: Enable THERM_CONTROL MSR write for dom0 even cpufreq=xen

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

> This patchset backport a series of T-State patches from kernel upstream
> to make dom0 ACPI throttling works with MSR( Fixed H/W). Meanwhile, it
> enable THERM_CONTROL MSR write for dom0 even cpufreq=xen.
On modern CPUs, is the TSC invariant to T-state throttling? If so, which
generation did this start happening?

On server/desktop systems I believe that T-state throttling is intended to be a
rare event usually due to bad system thermal design. Is this the case with
laptops, or would we expect to see more? Do any of the forthcoming
"turbo" modes change this?

Thanks,
Ian

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

>> This patchset backport a series of T-State patches from kernel upstream
>> to make dom0 ACPI throttling works with MSR( Fixed H/W). Meanwhile, it
>> enable THERM_CONTROL MSR write for dom0 even cpufreq=xen.
> 
> On modern CPUs, is the TSC invariant to T-state throttling? If so, which
> generation did this start happening? 
My perspective is yes. But  I don''t know which generation it started
from yet.

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

>From: Ian Pratt
>Sent: Thursday, February 26, 2009 11:03 PM
>
>> This patchset backport a series of T-State patches from 
>kernel upstream
>> to make dom0 ACPI throttling works with MSR( Fixed H/W). 
>Meanwhile, it
>> enable THERM_CONTROL MSR write for dom0 even cpufreq=xen.
>
>On modern CPUs, is the TSC invariant to T-state throttling? If 
>so, which generation did this start happening?
Yes, TSC is invariant to T-state throttling, i.e. always increment
even when #STPCLK is asserted. There''re two cases where TSC
may vary:

a) On some early processors like Pentium M, TSC is incremented
against internal clock, and thus vary according freq scaling. Then
new processors (Pentium4, core, ...) has constant TSC feature
which allows increment in a constant rate

b) On many processors TSC is halted when external clock is stopped
like in deep C-state. Some new processor (Intel Core-i7) supports
always running TSC feature to solve this issue.
>
>On server/desktop systems I believe that T-state throttling is 
>intended to be a rare event usually due to bad system thermal 
>design. Is this the case with laptops, or would we expect to 
>see more? Do any of the forthcoming "turbo" modes change this?
>
Even on laptop, it''s still possible to see over-heat and then T-state
throttling could be used as one cooling method. Turbo doesn''t break
TDP since the over-clocked core only tries to borrow thermal room
(e.g. saved from cores in deep C-states) under TDP.

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