[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <f488382f0910042203x38365b9amd557803872cc1435@mail.gmail.com>
Date: Sun, 4 Oct 2009 22:03:16 -0700
From: Steven Noonan <steven@...inklabs.net>
To: linux-kernel@...r.kernel.org
Subject: Re: cpufreq scaling strangely, system feels warm
On Sun, Oct 4, 2009 at 7:24 PM, Steven Noonan <steven@...inklabs.net> wrote:
> On Sat, Oct 3, 2009 at 5:54 PM, Steven Noonan <steven@...inklabs.net> wrote:
>> I have a system that's behaving as though it's under a moderate load,
>> but nothing's happening and the load average is zero. If I change my
>> cpufreq governor to 'conservative' or 'ondemand', the CPU immediately
>> scales up to the maximum clock frequency (as though something was
>> demanding that much power). The system is also very warm, which is
>> abnormal even when the machine is always clocked at its maximum
>> frequency.
>>
>> Top shows that there are no processes which are hogging the CPU, and
>> /proc/interrupts doesn't reveal anything interesting (i.e. an
>> interrupt storm). The only way to get my system to stay cool is to
>> change the cpufreq governor to 'powersave' or 'userspace' (and then
>> clock it at the lowest, manually).
>>
>
> Scratch that. It's probably an interrupt storm. 'perf top' shows
> 19,000 IRQs per second. And I recall that 500-1500 per second was
> normal for this machine. Now the problem is figuring out where the
> interrupt storm is coming from...
>
Went through my kernel config and found some odd settings, fixed them
and recompiled. Now 'ondemand' works as it should, but 'conservative'
always clocks up to the maximum frequency. I can't figure out why.
Is there a way to get cpufreq to print out some explanations in dmesg
as to why 'conservative' is doing what it does?
- Steven
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists