Subject: Re: "running without thermal monitor"?
To: Quentin Garnier <cube@cubidou.net>
From: Steven M. Bellovin <smb@cs.columbia.edu>
List: netbsd-users
Date: 01/07/2005 11:41:47
In message <20050107090428.GL26974@gallia.cubidou.net>, Quentin Garnier writes:
>
>--VBq/nvTu32OVLBUP
>Content-Type: text/plain; charset=us-ascii
>Content-Disposition: inline
>Content-Transfer-Encoding: quoted-printable
>
>On Fri, Jan 07, 2005 at 09:23:23AM +0100, Florian Stoehr wrote:
>> On Thu, 6 Jan 2005, Steven M. Bellovin wrote:
>>=20
>> >On a new P4, this message shows up very early in the boot process:
>> >
>> >cpu0: running without thermal monitor!
>[...]
>> >acpi0: X/RSDT: OemId <DELL  , 8400   ,00000007>, AslId <ASL ,00000061>
>[...]
>> is it a HP server? A "Pizzabox"?
>
>It's obviously not a HP.
>
>Anyway, the message Steve is seeing just means that the BIOS doesn't
>activate the Thermal Monitor feature on the processor.  It might just
>not actually have the thermal chip, just the bit set in the feature
>register, just like all Intel processors have the HTT bit set even when
>they don't have a hyperthreading core.
>
>It's not like NetBSD does anything with it when the feature is active.
>It's not really supposed to either according to the documentation.
>
>As for what happens when the processor overheats, I guess it stops
>or reduce its voltage/frequency to static values instead of "smartly"
>lowering specs.
>
>In conclusion, the message in dmesg might be slightly too alarming :)
>
It was the exclamation point that alarmed me...

It was also messages like this, from about a year ago:

  Date:    Mon, 03 Nov 2003 08:30:07 EST
  To:      Steve Woodford <scw@NetBSD.org>
  cc:      current-users@NetBSD.org
  From:    "Perry E.Metzger" <perry@piermont.com>
  Subject: Re: ACPI warning
  
  Steve Woodford <scw@netbsd.org> writes:
  > This is just a heads up for folks who are running acpi-enabled -current 
  > on laptops for the first time.
  >
  > Some laptop BIOSs, including my HP Omnibook, hand *complete* control of 
  > thermal management (i.e. fan control) over to the OS if ACPI is 
  > utilised; with no failsafe mechanism in hardware. This is a seriously 
  > dumb hardware design decision, IMHO. But that's just how it is.
  >
  > So, after enabling ACPI, I suggest checking if your laptop's fan still 
  > operates as before. Mine certainly doesn't switch on at all with ACPI 
  > in the kernel, so I've had to revert to non-ACPI rather than risk 
  > frying the cpu. :(
  
  FYI, the FreeBSD folks have a pretty clean thermal control/fan control
  module -- someone should steal it for NetBSD so this problem goes away.

I'll run some fan tests; if I don't hear changes in speed when I load 
the machine, I'll turn off ACPI (and hence give up HT).  (Until this 
morning, I had an HP dcf5000 where the hardware handled the fan speed 
quite nicely, even with ACPI on.  In fact, I should have paid a bit 
more attention to it -- it seemed to be running too fast when the 
machine was idle.  It was days later that I found a run-away OpenOffice 
binary....)

		--Prof. Steven M. Bellovin, http://www.cs.columbia.edu/~smb