[ltp] 2.6 kernel kills the CPU fan

[Bert Haskins]

Helen Borrie wrote:

> At 04:06 AM 8/09/2005 +0100, you wrote:
>
>>>> Helen Borrie wrote:
>>>>
>>>>> Hello all,
>>>>> I'm trying to track down a solution on my TP390E to a problem that 
>>>>> seems rife:  that Linux causes the CPU fan to stop as soon as you 
>>>>> begin to boot the OS. 
>>>>
Normal behavior on all of my 390s is that the fan starts up about 4 seconds
after pushing the on switch.
It runs until post is completed and then stops at the beep.
I think that this is normal for most if not all TPs
as this is post checking the fan.

>>>>>
>>
>>> And, yes, the machine is getting hot enough to put itself into 
>>> suspend (which one can't get out of) and to not want to complete 
>>> POST when rebooted after doing this.  I seem to get about 30 minutes 
>>> "life" before this happens;  and need about 10 mins for cool-down.
>>
If I run memtest-86 the fan will go on after several ( well over ten ) 
minutes and then cycle off for a few minutes
 and then back on for a few minutes for as long as the test is run.

>>> The bottom of the case feels hand-hot when this occurs (i.e. about 
>>> 60-65 Celcius)
>>
Its normal for the bottom of the case to feel very warm even while the 
fan is doing it's
cycling thing.
I have never had a shut down while doing this ( except for memory chip 
problems ) even though
memtest will work the cpu harder that just idling at the prompt.
I also have never had a shutdown even when compiling a kernel.
So I think that you must have some other problem.
Have you run a memory test program?
If so, does it start showing a bunch of errors just before the 390 shuts 
down?

The next test is to have the TP on it's side with the memory cover off 
and a
fan blowing on the memory chips.
I have seen plenty of heat-sensitive memory chips in both desktops and 
laptops.
Some of these will run fine for over an hour and then head for the pits.

>>> and I think I can sometimes smell hot fumes (the latter could be my 
>>> el cheapo gel mousepad, though, it's pretty smelly, even at 5 Celcius).
>>
>>
>> That's definitely a bug. There is no way that this should be deliberate.
>> At least it's an Intel CPU - and so it shuts down when overheated. 
>> AMDs (at least in some cases), just catch fire!
>>
>> However, on reflection, the BIOS shouldn't let the OS mess up the fan 
>> speed anyway. [Eg, what happens when the OS crashes?]
>>
>> If it is the OS, are you running APM or ACPI?  If you run either the 
>> apm or acpi commands, what do you get?
>
>
> Good question.  :-)  At least I know that acpi is installed and 
> running, since I can see it there when I inspect the services.  
> Whether the TP390E supports acpi is another question.  I've never 
> configured acpi before - never needed to, afaik.  My "industrially 
> active" notebook is a thoroughly vanilla step 6 Celeron and it "just 
> works" with Mdk 10.1 distro.  My static server is an AMD Sempron 2200, 
> which runs 24/7 and never runs hot (touches the nearest wooden 
> object), also running Mdk 10.1.  But, yep, I did get the Sempron after 
> frying a mainboard by an unknown technique.  (The AMD Duron 850 
> survived and now runs one of my webservers on a $10 mobo out of 
> eBay.)  I'm not much bothered about power-saving, beyond sleeping when 
> I'm sleeping, since I always run notebooks off the mains and power 
> them down when I'm not using them.  That's not very PC, I know, but I 
> do wash the car with grey water and take canvas bags to the 
> supermarket. :-)
>
> As for the acpi commands...um...I will need to look them up in "1001 
> things you never thought you'd need to know about Linux".
>
> Thanks muchly for the pointers.  I now have a way forward.
>
> cheers,
> Helen
>
>
>
>>> The respective home dirs of the logged in users, i.e. root and one 
>>> other, contained bunches of coredump files.  Unfortunately, I 
>>> wouldn't have a clue about interpreting them.
>>
>>
>> You can ignore them - they won't tell you anything useful. The crash 
>> was caused by an overheated CPU, so the coredump info is irrelevant.
>>
>> However, when testing, it would be safer to:
>>
>> init 1    (or shut down as many services as you can)
>> umount home
>> mount -o remount,ro /
>>
>> That way, when it crashes, it won't harm your data.  Or boot into 
>> single user mode with "linux single" at the lilo command prompt.
>>
>> Running "yes" will put the CPU at full load, to help test.
>>
>>> I will take your advice and get hold of the Knoppix distro.  
>>> Actually, I may be able to test the Knoppix behaviour as I have the 
>>> Ulitmate Boot CD here.
>>
>>
>> Good luck. Knoppix is useful, since it does give you an extra 
>> datapoint for very little effort!
>>
>> Best wishes
>>
>> Richard
>>
>> P.S. You could also check for an updated BIOS.
>> -- 
>> The linux-thinkpad mailing list home page is at:
>> http://mailman.linux-thinkpad.org/mailman/listinfo/linux-thinkpad
>
>