Weird behaviour (cpufreq-2.4.21-2)

Weird behaviour (cpufreq-2.4.21-2)

[Francesco Poli]

Hello everybody!  :)
I installed a 2.4.21 kernel + cpufreq-2.4.21-2 (downloaded
from http://www.codemonkey.org.uk/cpufreq/) + acpi-20030619-2.4.21
(downloaded from http://acpi.sourceforge.net/) on an Acer Aspire 1302XV
(AMD Mobile Athlon XP 1600+). I manually fixed the vid-fid bug in
powernow-k7.c (as reminded by Carl Thompson in a recent post), activated
the ACPI-related configuration options, compiled and installed.

Here comes the weird!

When I boot all seems right. The CPU is almost 100% idle and
$ cat /proc/cpufreq
          minimum CPU frequency  -  maximum CPU frequency  -  policy
CPU  0       500000 kHz ( 35 %)  -    1400000 kHz (100 %)  - 
performance$ cat /proc/acpi/thermal_zone/THRM/temperature
temperature:              60 C
And this seems to be able to go on (with temperature slowly
decreasing!).

Then I disconnect the laptop from the electrical network: so I run on
battery.
When I go back on AC power, the battery starts charging (as reported by
the LED and by ACPI). And the temperature goes up towards 80 degrees
Celsius; the back fan wakes up thus pushing the temperature down again
to 69 degrees Celsius; then the fan rests and the temperature goes up
again. This thermal cycle goes on and on, even if the CPU is almost
totally idle.

A simple reboot stops this behaviour and, after a while, I get
$ cat /proc/acpi/thermal_zone/THRM/temperature
temperature:              57 C

The question is: what's wrong? What did I messed up with?

-- 
    Francesco Poli       <frx AT firenze.linux.it>
======================================================
You're compiling a program and, all of a sudden, boom!
  -- from APT HOWTO, version 1.8.0

Re: Weird behaviour (cpufreq-2.4.21-2)

[Dominik Brodowski]

On Sat, Jul 19, 2003 at 12:33:02AM +0200, Francesco Poli wrote:
> Then I disconnect the laptop from the electrical network: so I run on
> battery.
> When I go back on AC power, the battery starts charging (as reported by
> the LED and by ACPI). And the temperature goes up towards 80 degrees
> Celsius; the back fan wakes up thus pushing the temperature down again
> to 69 degrees Celsius; then the fan rests and the temperature goes up
> again. This thermal cycle goes on and on, even if the CPU is almost
> totally idle.
> 
> A simple reboot stops this behaviour and, after a while, I get
> $ cat /proc/acpi/thermal_zone/THRM/temperature
> temperature:              57 C
> 
> The question is: what's wrong? What did I messed up with?

Hmmm... might it be that the BIOS meddles with the CPU frequency, too?
You could compare the output of the "bogomips" tool to detect frequency 
changes, for example.

	Dominik

Re: Weird behaviour (cpufreq-2.4.21-2)

[David]

> When I go back on AC power, the battery starts charging (as reported by
> the LED and by ACPI). And the temperature goes up towards 80 degrees
> Celsius; the back fan wakes up thus pushing the temperature down again
> to 69 degrees Celsius; then the fan rests and the temperature goes up
> again. This thermal cycle goes on and on, even if the CPU is almost
> totally idle.

I think it isn't a cpufreq problem. I got the same behaviour without loading 
cpufreq module. Probably it is an ACPI issue.
It happens to me even in another case: if the computer boots in AC power but 
the battery charge is not full, when the battery lever reaches the top, the 
temperature goes up and the fan starts.

I have an Acer Aspire 1304 XC (Athlon XP 1800+)

David

Re: Weird behaviour (cpufreq-2.4.21-2)

[Francesco Poli]

On Sat, 19 Jul 2003 08:45:28 +0200 Dominik Brodowski wrote:

> > The question is: what's wrong? What did I messed up with?
> 
> Hmmm... might it be that the BIOS meddles with the CPU frequency, too?
> You could compare the output of the "bogomips" tool to detect
> frequency changes, for example.

Here's a session log (I used bogomips v1.2 from the Debian Woody
sysutils package)


# alias ETEMP='cat /proc/acpi/thermal_zone/THRM/temperature'
# ETEMP
temperature:             62 C
# ETEMP
temperature:             62 C
# cat /proc/cpufreq
          minimum CPU frequency  -  maximum CPU frequency  -  policy
CPU  0       500000 kHz ( 35 %)  -    1400000 kHz (100 %)  - 
performance# ETEMP
temperature:             62 C
# on_ac_power
# echo $?
0
# on_ac_power
# echo $?
1
# ETEMP
temperature:             62 C
# bogomips
Calibrating delay loop.. ok - 1394.00 BogoMips
# on_ac_power
# echo $?
0
# ETEMP
temperature:             78 C
# ETEMP
temperature:             79 C
# bogomips
Calibrating delay loop.. ok - 1394.00 BogoMips
# ETEMP
temperature:             76 C
# bogomips
Calibrating delay loop.. ok - 1394.00 BogoMips
# ETEMP
temperature:             75 C
# ETEMP
temperature:             74 C
# ETEMP
temperature:             74 C
# bogomips
Calibrating delay loop.. ok - 1394.00 BogoMips
# ETEMP
temperature:             73 C
# ETEMP
temperature:             73 C
# ETEMP
temperature:             73 C
# ETEMP
temperature:             72 C
# bogomips 
Calibrating delay loop.. ok - 1384.00 BogoMips
# ETEMP
temperature:             73 C
# ETEMP
temperature:             72 C
# ETEMP
temperature:             72 C
# ETEMP
temperature:             72 C


It seems that the CPU clock frequency stays pretty fixed, doesn't it?

Any ideas?

-- 
    Francesco Poli       <frx AT firenze.linux.it>
======================================================
You're compiling a program and, all of a sudden, boom!
  -- from APT HOWTO, version 1.8.0

Re: Weird behaviour (cpufreq-2.4.21-2)

[Dominik Brodowski]

On Tue, Jul 22, 2003 at 01:12:20AM +0200, Francesco Poli wrote:
> On Sat, 19 Jul 2003 08:45:28 +0200 Dominik Brodowski wrote:
> 
> > > The question is: what's wrong? What did I messed up with?
> > 
> > Hmmm... might it be that the BIOS meddles with the CPU frequency, too?
> > You could compare the output of the "bogomips" tool to detect
> > frequency changes, for example.
> 
> Here's a session log (I used bogomips v1.2 from the Debian Woody
> sysutils package)

Thanks

> It seems that the CPU clock frequency stays pretty fixed, doesn't it?

Indeed. But, as others pointed out, it might be a non-cpufreq related
problem but more a "battery loading" one. At least, I cannot see why there
should be any additional heat generation by cpufreq as the frequency doesn't
change at all...

	Dominik