Traduisez - Übersetzen - Traduzca - Traduza - Tradurre - Translate

Re: WARNING: New K6 bug

Clemmitt Sigler (siglercm@alphamb2.phys.vt.edu)
Thu, 11 Jun 1998 10:28:41 -0400 (EDT)

On Thu, 11 Jun 1998, Mike Montour wrote:
> > Could you send us the output of /proc/cpuinfo, please?
>
> (running 2.1.105 kernel)[root@elric mmontour]# cat /proc/cpuinfo
> processor : 0
> cpu family : 5
> model : K6 (166 - 266)
> vendor_id : AuthenticAMD
> stepping : 2
> fdiv_bug : no
> hlt_bug : no
> sep_bug : no
> f00f_bug : no
> fpu : yes
> fpu_exception : yes
> cpuid level : 1
> wp : yes
> flags : fpu vme de pse tsc msr mce cx8 mmx
> bogomips : 466.94

I have the newest K6-266 2.2V CPU, running 2.0.30 kernel. I get from
/proc/cpuinfo:

processor : 0
cpu : 586
model : 7
vendor_id : AuthenticAMD
stepping : unknown
fdiv_bug : no
hlt_bug : no
fpu : yes
fpu_exception : yes
cpuid : yes
wp : yes
flags : fpu vme de pse tsc msr mce cx8 mmx
bogomips : 598.02

In case I didn't mention, I'm running the Asus P/I-P55T2P4 motherboard,
Rev. 3.1, running the K6-266 at 300MHz (75MHzx4.0) at 2.1V core voltage.

> Now that I look at it, the "Stepping: 2" line seems a bit odd - wouldn't a
> "C" stepping (from the serial number) show up as a "3" or something? [
> I suppose it's remotely possible that this chip is a counterfeit re-labelled
> product, but I've never heard of that happening on non-Intel chips]. Or does
> 'stepping' start counting at 0?

I seriously doubt it's a counterfeit/remarked. You may be right, the
first stepping might be 0. That'd be the way a C programmer would do
it :^)

> > Also, you might
> > try running the program cpuspd3e that you can download from the AMD
> > web site (http://www.amd.com/K6/k6docs/cpuspd3e.exe) on your chip and
> > posting the results.
>
> I have downloaded this, and will run it later (I will have to reboot to
> MS-DOG).

I will run it as well and post my results for the CPU described above.

> > You could also try compiling the crashme program and running it with
> > the seed values mentioned in some of the posts to linux-kernel to
> > see if you can reproduce their hard lock-ups.
>
> I've run it a couple of times, but no conclusive results yet. I will keep you
> posted.

OK. If you try the crashme parameters documented in the linux-kernel
mailing list and it doesn't crash the kernel/CPU, I'd say you're safe
and your problem most likely lies elsewhere. Like I said, linux-kernel
claimed that this AMD CPU bug/erratum was (unintentionally) worked around
by changes in the kernel memory handling in kernels >=2.1.43.

Clemmitt Sigler
Va. Tech Physics Dept.