Temperature above threshold loop with 2.6.31.1

Temperature above threshold loop with 2.6.31.1

[Trevor Hemsley]

Hi

I just downloaded and installed the latest 2.6.31.1 kernel on my
machine here and I think I found a small bug. Shortly after boot I
start to receive messages like 

CPU0: Temperature above threshold, cpu clock throttled (total events =
21672)

This is on an Intel D975XBX2 motherboard with an Intel Xeon X3220
2.4GHz quad core chip installed (the Xeon equivalent of a Q6600). The
BIOS reports the cpu temperature as being consistently 51C which may be
a bit on the high side but not dangerously so.

And, yes, I'm aware that this might be a valid temperature warning and
I aim to dismantle the machaine and check everything out over the
weekend but...

More importantly, I get approximately 100,000 of these messages per
minute and the machine is completely unusable. All of these are for
CPU0 - at least all the ones that get written to /var/log/messages. 

1,666 notifications a second seems a little on the 'too frequent' side
of things to me :-)

Please cc me on any replies as I'm not subscribed to the list. Thanks.


-- 
Trevor Hemsley, Brighton, UK
Trevor dot Hemsley at ntlworld dot com

Re: Temperature above threshold loop with 2.6.31.1

[Frans Pop]

Trevor Hemsley wrote:
> I just downloaded and installed the latest 2.6.31.1 kernel on my
> machine here and I think I found a small bug. Shortly after boot I
> start to receive messages like
> 
> CPU0: Temperature above threshold, cpu clock throttled (total events =
> 21672)
[...] 
> More importantly, I get approximately 100,000 of these messages per
> minute and the machine is completely unusable. All of these are for
> CPU0 - at least all the ones that get written to /var/log/messages.
> 
> 1,666 notifications a second seems a little on the 'too frequent' side
> of things to me :-)

Looks like this may already be fixed in mainline by the following commit:
commit b417c9fd8690637f0c91479435ab3e2bf450c038
Author: Ingo Molnar <mingo@elte.hu>
Date:   Tue Sep 22 15:50:24 2009 +0200
    x86: mce: Fix thermal throttling message storm

Can you confirm that please, either by compiling current git or by applying 
that commit on top of 2.6.31.1?

Ingo: is that patch already scheduled for stable?

Cheers,
FJP

Re: Temperature above threshold loop with 2.6.31.1

[Ingo Molnar]

* Frans Pop <elendil@planet.nl> wrote:

> Trevor Hemsley wrote:
> > I just downloaded and installed the latest 2.6.31.1 kernel on my
> > machine here and I think I found a small bug. Shortly after boot I
> > start to receive messages like
> > 
> > CPU0: Temperature above threshold, cpu clock throttled (total events =
> > 21672)
> [...] 
> > More importantly, I get approximately 100,000 of these messages per
> > minute and the machine is completely unusable. All of these are for
> > CPU0 - at least all the ones that get written to /var/log/messages.
> > 
> > 1,666 notifications a second seems a little on the 'too frequent' side
> > of things to me :-)
> 
> Looks like this may already be fixed in mainline by the following commit:
> commit b417c9fd8690637f0c91479435ab3e2bf450c038
> Author: Ingo Molnar <mingo@elte.hu>
> Date:   Tue Sep 22 15:50:24 2009 +0200
>     x86: mce: Fix thermal throttling message storm
> 
> Can you confirm that please, either by compiling current git or by applying 
> that commit on top of 2.6.31.1?
> 
> Ingo: is that patch already scheduled for stable?

Not yet - i just forwarded it. Thanks for pointing it out,

	Ingo

Re: Temperature above threshold loop with 2.6.31.1

[Frans Pop]

On Friday 02 October 2009, Frans Pop wrote:
> Trevor Hemsley wrote:
> > 1,666 notifications a second seems a little on the 'too frequent' side
> > of things to me :-)
>
> Looks like this may already be fixed in mainline by the following
> commit: commit b417c9fd8690637f0c91479435ab3e2bf450c038
> Author: Ingo Molnar <mingo@elte.hu>
> Date:   Tue Sep 22 15:50:24 2009 +0200
>     x86: mce: Fix thermal throttling message storm
>
> Can you confirm that please, either by compiling current git or by
> applying that commit on top of 2.6.31.1?

If you want to apply it to .31.1 you'll also need this commit:
commit 3967684006f30c253bc6d4a6604d1bad4a7fc672
Author: Ingo Molnar <mingo@elte.hu>
Date:   Tue Sep 22 15:50:24 2009 +0200
    x86: mce: Clean up thermal throttling state tracking code

Re: Temperature above threshold loop with 2.6.31.1

[Ingo Molnar]

* Frans Pop <elendil@planet.nl> wrote:

> On Friday 02 October 2009, Frans Pop wrote:
> > Trevor Hemsley wrote:
> > > 1,666 notifications a second seems a little on the 'too frequent' side
> > > of things to me :-)
> >
> > Looks like this may already be fixed in mainline by the following
> > commit: commit b417c9fd8690637f0c91479435ab3e2bf450c038
> > Author: Ingo Molnar <mingo@elte.hu>
> > Date:   Tue Sep 22 15:50:24 2009 +0200
> >     x86: mce: Fix thermal throttling message storm
> >
> > Can you confirm that please, either by compiling current git or by
> > applying that commit on top of 2.6.31.1?
> 
> If you want to apply it to .31.1 you'll also need this commit:
> commit 3967684006f30c253bc6d4a6604d1bad4a7fc672
> Author: Ingo Molnar <mingo@elte.hu>
> Date:   Tue Sep 22 15:50:24 2009 +0200
>     x86: mce: Clean up thermal throttling state tracking code

Yes. The way to test this is to do this on top of a .31.1 tree:

 git cherry-pick 3967684006f30c253bc6d4a6604d1bad4a7fc672
 git cherry-pick b417c9fd8690637f0c91479435ab3e2bf450c038

	Ingo

Re: Temperature above threshold loop with 2.6.31.1

[Trevor Hemsley]

On Fri, 2 Oct 2009 19:57:45 +0200, Ingo Molnar wrote:

>Yes. The way to test this is to do this on top of a .31.1 tree:
>
> git cherry-pick 3967684006f30c253bc6d4a6604d1bad4a7fc672
> git cherry-pick b417c9fd8690637f0c91479435ab3e2bf450c038

The fix was tested and the fix was good.

It's still hot and throttling but the messages are only issued every ~5
mins now.

Thanks.
-- 
Trevor Hemsley, Brighton, UK
Trevor dot Hemsley at ntlworld dot com