Re: [PATCH] perfcounters: IRQ and NMI support on AMD CPUs

[Mike Galbraith <efault@gmx.de>]

On Fri, 2009-03-06 at 01:31 +0530, Jaswinder Singh Rajput wrote:
> On Thu, 2009-03-05 at 20:41 +0100, Ingo Molnar wrote:
> > * Peter Zijlstra <peterz@infradead.org> wrote:
> > 
> > > On Fri, 2009-03-06 at 00:26 +0530, Jaswinder Singh Rajput wrote:
> > > 
> > > > But kerneltop is still not working and I am not getting any NMI Interrupts.
> > > 
> > > Does the below work for you?
> > > 
> > > The BKGD suggests its 48 for all existing machines (it 
> > > certainly is for mine).
> > 
> > Yeah, this did the trick on AMD/Barcelona!
> > 
> 
> Nice.
> 
> but on my AMD box I got interrupts and NMI for some time:
> 
> NMI:        100    1108209   Non-maskable interrupts
> LOC:        213     192815   Local timer interrupts
> CNT:        100    1108206   Performance counter interrupts
> RES:      21132      20397   Rescheduling interrupts
> 
> but kerneltop does not show anything, it only shows # irqs/sec for some
> time and then irqs also become 0.

Hm.  Kerneltop on my Q6600 has ceased to function with n 1 as well, but
n 0 still works.  Zero NMIs.

Perfstat has also gone south.
perfstat: perfstat.c:423: main: Assertion `res == sizeof(single_count)' failed.

	-Mike