NMI watchdog not ticking at the right intervals

NMI watchdog not ticking at the right intervals

[Corey Minyard]

As I have been working on my NMI patch, I have noticed that the NMI 
watchdog does not seem to be ticking correctly.  I've tried 2.4 and 2.5 
kernels, and I get the same results.  From my reading of the code, it 
should tick once a second.  However, I have had the time between ticks 
vary from around 33 to over 100 seconds.  Tbe time between ticks is 
different on every boot, but is consistent once booted.  Is there some 
divider register that's not getting initialized?

Here's my cpuinfo:

processor    : 0
cpu_package    : 0
vendor_id    : GenuineIntel
cpu family    : 6
model        : 11
model name    : Intel(R) Pentium(R) III Mobile CPU      1066MHz
stepping    : 1
cpu MHz        : 730.601
cache size    : 512 KB
fdiv_bug    : no
hlt_bug        : no
f00f_bug    : no
coma_bug    : no
fpu        : yes
fpu_exception    : yes
cpuid level    : 2
wp        : yes
flags        : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca 
cmov pat pse36 mmx fxsr sse
bogomips    : 1458.17


-Corey

Re: NMI watchdog not ticking at the right intervals

[george anzinger]

Corey Minyard wrote:
> 
> As I have been working on my NMI patch, I have noticed that the NMI
> watchdog does not seem to be ticking correctly.  I've tried 2.4 and 2.5
> kernels, and I get the same results.  From my reading of the code, it
> should tick once a second.  However, I have had the time between ticks
> vary from around 33 to over 100 seconds.  Tbe time between ticks is
> different on every boot, but is consistent once booted.  Is there some
> divider register that's not getting initialized?

That seems unlikely.  It is driven by the performance
counter stuff which is clocked at TSC rates.  It could be
that it is not being set up properly to count every event.

-g
> 
> Here's my cpuinfo:
> 
> processor    : 0
> cpu_package    : 0
> vendor_id    : GenuineIntel
> cpu family    : 6
> model        : 11
> model name    : Intel(R) Pentium(R) III Mobile CPU      1066MHz
> stepping    : 1
> cpu MHz        : 730.601
> cache size    : 512 KB
> fdiv_bug    : no
> hlt_bug        : no
> f00f_bug    : no
> coma_bug    : no
> fpu        : yes
> fpu_exception    : yes
> cpuid level    : 2
> wp        : yes
> flags        : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca
> cmov pat pse36 mmx fxsr sse
> bogomips    : 1458.17
> 
> -Corey
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

-- 
George Anzinger   george@mvista.com
High-res-timers: 
http://sourceforge.net/projects/high-res-timers/
Preemption patch:
http://www.kernel.org/pub/linux/kernel/people/rml

Re: NMI watchdog not ticking at the right intervals

[Mikael Pettersson]

On Fri, 25 Oct 2002 09:09:10 -0500, Corey Minyard wrote:
>As I have been working on my NMI patch, I have noticed that the NMI 
>watchdog does not seem to be ticking correctly.  I've tried 2.4 and 2.5 
>kernels, and I get the same results.  From my reading of the code, it 
>should tick once a second.  However, I have had the time between ticks 
>vary from around 33 to over 100 seconds.  Tbe time between ticks is 
>different on every boot, but is consistent once booted.  Is there some 
>divider register that's not getting initialized?
>
>Here's my cpuinfo:
>
>processor    : 0
>cpu_package    : 0
>vendor_id    : GenuineIntel
>cpu family    : 6
>model        : 11
>model name    : Intel(R) Pentium(R) III Mobile CPU      1066MHz

(Me thinks "speedstep?")

Do you boot with nmi_watchdog=1 or 2?
The perfctr + local-APIC driven NMI watchdog is dependent
on the CPU's clock frequency. If this changes, the NMI rate
will change accordingly.

The NMI rate may also be affected by APM/ACPI and how often
the kernel executes HLT.

/Mikael

Re: NMI watchdog not ticking at the right intervals

[Corey Minyard]

Mikael Pettersson wrote:

>On Fri, 25 Oct 2002 09:09:10 -0500, Corey Minyard wrote:
>  
>
>>As I have been working on my NMI patch, I have noticed that the NMI 
>>watchdog does not seem to be ticking correctly.  I've tried 2.4 and 2.5 
>>kernels, and I get the same results.  From my reading of the code, it 
>>should tick once a second.  However, I have had the time between ticks 
>>vary from around 33 to over 100 seconds.  Tbe time between ticks is 
>>different on every boot, but is consistent once booted.  Is there some 
>>divider register that's not getting initialized?
>>
>>Here's my cpuinfo:
>>
>>processor    : 0
>>cpu_package    : 0
>>vendor_id    : GenuineIntel
>>cpu family    : 6
>>model        : 11
>>model name    : Intel(R) Pentium(R) III Mobile CPU      1066MHz
>>    
>>
>
>(Me thinks "speedstep?")
>
>Do you boot with nmi_watchdog=1 or 2?
>
It's set to 2 (local APIC).  Actually, when I set it to 1 (in 2.5) the 
code overrides it and sets it to 2.  But 2 is what it is running with now.

>The perfctr + local-APIC driven NMI watchdog is dependent
>on the CPU's clock frequency. If this changes, the NMI rate
>will change accordingly.
>
>The NMI rate may also be affected by APM/ACPI and how often
>the kernel executes HLT.
>
This board is not doing speed stepping.  It's running derated for 
reliability.

But that makes me think of something...

If I put a program into an infinite loop, then it will step every second 
like it is supposed to.  That makes sense now.

Thanks,

-Corey