Re: [PATCH 1/2] perfcounter: Ignore the nmi call frames in the x86-64 backtraces

[Frederic Weisbecker <fweisbec@gmail.com>]

On Mon, Jul 06, 2009 at 02:17:48PM +0200, Peter Zijlstra wrote:
> On Wed, 2009-07-01 at 21:02 +0200, Frederic Weisbecker wrote:
> 
> > +int x86_is_stack_id(int id, char *name)
> > +{
> > +	return x86_stack_ids[id - 1] == name;
> > +}
> 
> OK, this bit would need a bit of a comment explaining why this works, as
> I don't think the C language guarantees de-duplication of constant
> strings.
> 
> Therefore the above only works correctly if its passed a pointer from
> the x86_stack_ids[] array -- as it indeed is, since the string passed is
> obtained through the below method:


Right, that need some comments, I will do that with x86-32 support at the same
time.

Thanks.


 
> > +static unsigned long *in_exception_stack(unsigned cpu, unsigned long stack,
> > +					unsigned *usedp, char **idp)
> > +{
> >  	unsigned k;
> >  
> >  	/*
> > @@ -61,7 +68,7 @@ static unsigned long *in_exception_stack(unsigned cpu, unsigned long stack,
> >  			if (*usedp & (1U << k))
> >  				break;
> >  			*usedp |= 1U << k;
> > -			*idp = ids[k];
> > +			*idp = x86_stack_ids[k];
> >  			return (unsigned long *)end;
> >  		}
> >  		/*
> 
>