Re: amd64 system powering down

["Steven M. Bellovin" <smb%cs.columbia.edu@localhost>]

On Wed, 2 Jul 2008 15:00:42 -0500 (CDT)
"Jeremy C. Reed" <reed%reedmedia.net@localhost> wrote:

> On Tue, 1 Jul 2008, Jeremy C. Reed wrote:
> 
> > Each time I have tried to build openoffice2 on a NetBSD/amd64
> > 4.99.67 system it has powered down. No logs.
> > 
> > I guess it overheated but don't know (need to watch in real time
> > maybe).
> > 
> > It is a Compaq Presario V2658US (V2000) notebook with AMD Turion 64
> > Mobile Processor.
> > 
> > My dmesg is at 
> > http://mail-index.netbsd.org/port-amd64/2008/06/28/msg000339.html
> > 
> > Anything I should look at or check?
> > 
> > The system ran fine before moving to NetBSD a few days ago.
> > 
> > (I have been testing a sdhc/sdmmc kernel, but I think that is
> > unrelated as it also crashed once without that custom kernel.)
> 
> Any thoughts on this? It is crashing frequently and no logs. I had it
> show the temperature every couple seconds and it last printed:
> 
>                 Current      Max      Min      Avg     Unit
> temperature:     84.000   86.000   81.000   83.667     deg
> 
> So I don't think it was temperature issue. (My critical is 106 I
> think.)

Have you tried memtestplus?  (I'm grasping at straws...)  It's not a
generic problem with dual-core amd64; I do builds on my Thinkpad T61
quite often.

But yes, my first instinct is to guess temperature, too.  I wonder if
there are other temperature sensors you're not seeing.  Here's what
envstat says on my machine:

[acpitz0]
     temperature:     44.000 degC
[acpitz1]
     temperature:     42.000 degC
[coretemp0]
  cpu0 temperature:     42.000 degC
[coretemp1]
  cpu1 temperature:     42.000 degC
[thinkpad0]
              TMP0:     44.000 degC
              TMP1:     39.000 degC
              TMP2:     33.000 degC
              TMP3:        N/A
              TMP4:     33.000 degC
              TMP5:     31.000 degC
              TMP6:     31.000 degC
              TMP7:     31.000 degC

I didn't see the 8 Thinkpad sensors until Jared (I believe) wrote a
special Thinkpad driver.

Hmm -- try booting with 'no acpi' and see what happens.  Of course, I
think that will take away your ability to monitor temperature.
> 
> Is there anything I can enable to log details or debugging? (But
> since it powers down maybe too late.)
> 
Can you ssh in remotely and look at log data there in real-time?  The
ssh windows won't go away for several minutes at least after the crash,
until the kernel is back up and sending RSTs.


                --Steve Bellovin, http://www.cs.columbia.edu/~smb