Skip to contentSkip to conversation

2.6.24 Temperature/speed _not_ normal - no thermal throttling?

DISCUSSION Ron Rechenmacher 5 participants 5 messages Feb 20, 2008 – Feb 26, 2008 ACPI

lore
Share & tools
lore

Ron Rechenmacher

author
#0 lore raw

Hi,
I believe I am having a critical thermal problem. I do not know if it
is limited to the 2.6.24.2 kernel which I am running. I do see there has been some discussion about thermal zones and throttling on the list, but I can not tell if it means that thermal throttling is not working in 2.6.24.2

When I try to build several kernel source rpms, my dell d830 laptop seems to over heat and hang. It's happened 3 times now and I would like to learn what's going on and not let it happen again.

I'm a newbie (and have had problems trying to post :), so I do apologize if I've missing something relatively simple or if this is post is not appropriate in any way.

I'm running a Scientific Linux 5 (based on RHEL5) distribution and am just running a cpuspeed user space utility --- and therefor do not believe I have any user space process watching temperature. However, in the earlier kernels, I use to be able to (manually) write to /proc/acpi/processor/CPU0/throttling and see a change when read back, but now the write does not seem to do anything. This might be OK as I 'm thinking the kernel and/or the hardware itself might now suppose to be doing the throttling?

Anyway, in 3 windows, I run:
win1: stress --cpu 8 --io 4 --vm 2 --vm-bytes 128M --timeout 180s
win2: while sleep 1;do cat /proc/acpi/thermal_zone/THM/temperature;done
win3: tail -f /var/log/messages
win4; while sleep 1;do cat /proc/acpi/processor/CPU0/throttling;done

In win2, I see the temperature go from 50 C to over 86 C.
In win3, before, the temp in win2 reaches 70 C, I see "kernel: CPU0: Temperature/speed normal" (and also CPU1) and "kernel: Machine check events logged" The temperature would probably just continue to climb if I ran the test for longer that 180 seconds (the kernel rpms take much longer and do not complete before the system hangs :(

In /var/log/mcelog, (running mcelog-0.8pre), I only see "Processor core below trip temperature. Throttling disabled" messages. This is strange because it seems to be being disabling after never being enabled. (Is there a newer mcelog I should be running?)

The fan speed does increase, but the throttling state indication never changes (it's always "T0: 100%"). It seems that when I build the kernel rpms, the increased fan speed is not enough to keep the temperature form running away. It seems that thermal throttling would be required and is not happening. Should I be doing something from user space? Can I do something from user space?

Thanks,
Ron

#1 lore raw

Hi Ron,

Throttling is meant as a last line of defense before powering-off machine, and not a thermal regulation feature. Please check if you have cpufreq compiled in and able to change frequency.
Please open a bug report at bugzilla.kernel.org against ACPI/Thermal.
Please attach dmesg output and 'grep . /proc/acpi/thermal/*/*'

Thanks,
Alex.

Quoted 33 lines · Ron Rechenmacher

Hi,
I believe I am having a critical thermal problem. I do not know if it
is limited to the 2.6.24.2 kernel which I am running. I do see there has been some discussion about thermal zones and throttling on the list, but I can not tell if it means that thermal throttling is not working in 2.6.24.2

When I try to build several kernel source rpms, my dell d830 laptop seems to over heat and hang. It's happened 3 times now and I would like to learn what's going on and not let it happen again.

I'm a newbie (and have had problems trying to post :), so I do apologize if I've missing something relatively simple or if this is post is not appropriate in any way.

I'm running a Scientific Linux 5 (based on RHEL5) distribution and am just running a cpuspeed user space utility --- and therefor do not believe I have any user space process watching temperature. However, in the earlier kernels, I use to be able to (manually) write to /proc/acpi/processor/CPU0/throttling and see a change when read back, but now the write does not seem to do anything. This might be OK as I 'm thinking the kernel and/or the hardware itself might now suppose to be doing the throttling?

Anyway, in 3 windows, I run:
win1: stress --cpu 8 --io 4 --vm 2 --vm-bytes 128M --timeout 180s
win2: while sleep 1;do cat /proc/acpi/thermal_zone/THM/temperature;done
win3: tail -f /var/log/messages
win4; while sleep 1;do cat /proc/acpi/processor/CPU0/throttling;done

In win2, I see the temperature go from 50 C to over 86 C.
In win3, before, the temp in win2 reaches 70 C, I see "kernel: CPU0: Temperature/speed normal" (and also CPU1) and "kernel: Machine check events logged" The temperature would probably just continue to climb if I ran the test for longer that 180 seconds (the kernel rpms take much longer and do not complete before the system hangs :(

In /var/log/mcelog, (running mcelog-0.8pre), I only see "Processor core below trip temperature. Throttling disabled" messages. This is strange because it seems to be being disabling after never being enabled. (Is there a newer mcelog I should be running?)

The fan speed does increase, but the throttling state indication never changes (it's always "T0: 100%"). It seems that when I build the kernel rpms, the increased fan speed is not enough to keep the temperature form running away. It seems that thermal throttling would be required and is not happening. Should I be doing something from user space? Can I do something from user space?

Thanks,
Ron

-
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

reviewer
#2 lore raw
In reply to Ron Rechenmacher
Quoted 1 line · Ron Rechenmacher

my dell d830 laptop seems to over heat and hang.

Ron,
see "Thermal Issues" here:

http://www.lesswatts.org/projects/acpi/debug.php

My guess is that ACPI (and thus Linux) have no control over the fans on this system (as I've never seen a Dell with OS controlled fans) If the fans are spinning fast when you heat up the machine, then they are probably clogged with dust or there a mechanical issue with the thermal solution.

cheers,
-Len

#3 lore raw
In reply to Ron Rechenmacher
Quoted 7 lines · Ron Rechenmacher

Hi,
I believe I am having a critical thermal problem. I do not know if it is limited to the 2.6.24.2 kernel which I am running. I do see there has been some discussion about thermal zones and throttling on the list, but I can not tell if it means that thermal throttling is not working in 2.6.24.2

What does /proc/interrupts say about thermal event interrupts?

#4 lore raw
In reply to Ron Rechenmacher
Quoted 43 lines · Ron Rechenmacher

Hi,
I believe I am having a critical thermal problem. I do not know if it is limited to the 2.6.24.2 kernel which I am running. I do see there has been some discussion about thermal zones and throttling on the list, but I can not tell if it means that thermal throttling is not working in 2.6.24.2

When I try to build several kernel source rpms, my dell d830 laptop seems to over heat and hang. It's happened 3 times now and I would like to learn what's going on and not let it happen again.

I'm a newbie (and have had problems trying to post :), so I do apologize if I've missing something relatively simple or if this is post is not appropriate in any way.

I'm running a Scientific Linux 5 (based on RHEL5) distribution and am just running a cpuspeed user space utility --- and therefor do not believe I have any user space process watching temperature. However, in the earlier kernels, I use to be able to (manually) write to /proc/acpi/processor/CPU0/throttling and see a change when read back, but now the write does not seem to do anything. This might be OK as I 'm thinking the kernel and/or the hardware itself might now suppose to be doing the throttling?

Anyway, in 3 windows, I run:
win1: stress --cpu 8 --io 4 --vm 2 --vm-bytes 128M --timeout 180s
win2: while sleep 1;do cat /proc/acpi/thermal_zone/THM/temperature;done
win3: tail -f /var/log/messages
win4; while sleep 1;do cat /proc/acpi/processor/CPU0/throttling;done

In win2, I see the temperature go from 50 C to over 86 C. In win3, before, the temp in win2 reaches 70 C, I see "kernel: CPU0:
Temperature/speed normal" (and also CPU1) and "kernel: Machine check events logged"
The temperature would probably just continue to climb if I ran the test for longer that 180 seconds (the kernel rpms take much longer and do not complete before the system hangs :(

In /var/log/mcelog, (running mcelog-0.8pre), I only see "Processor core below trip temperature. Throttling disabled" messages. This is strange because it seems to be being disabling after never being enabled. (Is there a newer mcelog I should be running?)

Quoted 7 lines

The fan speed does increase, but the throttling state indication never changes (it's always "T0: 100%"). It seems that when I build the kernel rpms, the increased fan speed is not enough to keep the temperature form running away. It seems that thermal throttling would be required and is not happening.
Should I be doing something from user space? Can I do something from user space?

Does cleaning the fan slots help?

If not it might be related to this one:
https://bugzilla.novell.com/show_bug.cgi?id=333043
Even these are ThinkPads (and the temperature for some reason seem to be 20 C higher than on others - same model, very weird...), those are running at the very edge of passive and critical thermal trip points. It seems a kernel change came in some time ago which makes them shutdown because the thermal notification for the passive trip point is not executed and passed fast enough to the cpufreq layer.
On the ThinkPads it is easily reproducable by starting several CPU intensive tasks (e.g. a one thread kernel compile works, CPU is lowered, make -j5 will hang).

This could only be the case if your machine defines a passive trip point and supports cpufreq:
/proc/acpi/thermal_zone/*/trip_points

Thomas