[Linux-ia64] Spurious PMU Interrupt

[Linux-ia64] Spurious PMU Interrupt

[Ray Bryant]

Stephane,

I am getting one log message of the following type per CPU every time I
run a system-wide sampling session:

kernel: perfmon: Spurious PMU overflow interrupt on CPUxx: pmc0=0x1
owner\000000000000000 

Any idea what might be causing this?  This is a nuisance error, so if we
can just ignore it I will.

Thanks,
-- 
Best Regards,
Ray
-----------------------------------------------
Ray Bryant               SGI
512-453-9679 (work)      512-507-7807 (cell)
raybry@sgi.com           raybry@austin.rr.com
-----------------------------------------------

Re: [Linux-ia64] Spurious PMU Interrupt

[Stephane Eranian]

Ray,

On Tue, Dec 17, 2002 at 06:53:31PM -0600, Ray Bryant wrote:
> Stephane,
> 
> I am getting one log message of the following type per CPU every time I
> run a system-wide sampling session:
> 
> kernel: perfmon: Spurious PMU overflow interrupt on CPUxx: pmc0=0x1
> owner\000000000000000 
> 

This happens on McKinley only, this is not an error and can be ignored.
In fact in the latest version of the kernel (2.4.20), I removed the printk() 
to avoid confusion. This is due do an erratum in the McKinley PMU. The problem
is in fact a desired feature which got cast as an error due to a bug in  
the Itanium PMU. Basically, if you set the freeze bit when it is not set
you'll get a PMU interrupt. This is what happens at the end of a monitoring
session in the current perfmon codebase for both 2.4 and 2.5. I think
you can see how this "bug" is in fact a feature which could be use during
the context switch, for instance.

So you can ignore the problem. The new perfmon codebase that I am working
on will never set the freeze bit explicitely anymore, so the "problem" will
go away.

-- 
-Stephane