2009-11-11 15:06:35

by Justin Piszcz

[permalink] [raw]
Subject: Kernel: 2.6.31.5: CE: hpet increasing min_delta_ns: What do these messages mean?

Hello,

System: Intel DP55KG
CPU: Intel Core i870

I see these messages in my kernel log: (2.6.31.5)

[ 48.886129] CE: hpet increasing min_delta_ns to 15000 nsec
[ 3258.107039] CE: hpet increasing min_delta_ns to 22500 nsec
[53705.451463] CE: hpet increasing min_delta_ns to 33750 nsec

I only see them when ACPI_FREQ and NO_HZ are enabled, to take advantage
of turbo boost.

It is not the result of throttling:

# cat /sys/devices/system/cpu/*/thermal_throttle/count
0
0
0
0
0
0
0
0

Temps are also good, 1 core @ 100% CPU:

Core 0: +41.0?C (high = +84.0?C, crit = +100.0?C)
Core 1: +41.0?C (high = +84.0?C, crit = +100.0?C)
Core 2: +54.0?C (high = +84.0?C, crit = +100.0?C)
Core 3: +54.0?C (high = +84.0?C, crit = +100.0?C)
Core 4: +42.0?C (high = +84.0?C, crit = +100.0?C)
Core 5: +42.0?C (high = +84.0?C, crit = +100.0?C)
Core 6: +42.0?C (high = +84.0?C, crit = +100.0?C)
Core 7: +42.0?C (high = +84.0?C, crit = +100.0?C)

I see various discussions on the meaning of these messages but no clear
answer.

Also in reference to loading ACPI_FREQ I mentioned before that when this
module is loaded and there is no CPU activity, there is a high pitch noise
from the CPU/VRM presumably, when ACPI_FREQ is not loaded, this does not
occur, is this expected?

Justin.


2009-11-11 20:07:02

by john stultz

[permalink] [raw]
Subject: Re: Kernel: 2.6.31.5: CE: hpet increasing min_delta_ns: What do these messages mean?

On Wed, Nov 11, 2009 at 7:06 AM, Justin Piszcz <[email protected]> wrote:
> I see these messages in my kernel log: (2.6.31.5)
>
> [ ? 48.886129] CE: hpet increasing min_delta_ns to 15000 nsec
> [ 3258.107039] CE: hpet increasing min_delta_ns to 22500 nsec
> [53705.451463] CE: hpet increasing min_delta_ns to 33750 nsec

Basically it means setting the hpet timer interrupt failed because the
time it took to set it was more then the time requested. Why its
taking so long is an open question, but we're looking into it in the
following bug:

http://bugzilla.kernel.org/show_bug.cgi?id=14426

thanks
-john