[PATCH 1/5] arm: provide a mechanism to reserve performance counters

Will Deacon will.deacon at arm.com
Tue Dec 15 12:06:55 EST 2009


* Jamie Iles wrote:

> > I think it's standard Kernel coding style to put the declaration of a function
> > all on one line if it fits. The same goes elsewhere in the patch.
> I couldn't find anything that has this written down and there are plenty of
> other places in the perf code that do this. I personally like it because you
> can grep for "^foo" to find the definition. Unless people have strong
> objections to this I'll leave it as is.

I'm just going by the examples in Documentation/CodingStyle.

> > > +config CPU_HAS_PMU
> > > +	depends on CPU_V6 || CPU_V7 || CPU_XSCALE
> > > +	default y
> > > +	bool
> >
> > I think you should use XSCALE_PMU instead of CPU_XSCALE. Also, this should
> > probably be in the top-level ARM Kconfig instead of the mm/ one.
> Ok, I agree with you about using XSCALE_PMU, but why isn't mm/Kconfig the
> correct one? It's describing what features the CPU has and the PMU *is* a
> feature.

I'd say move it out of mm/Kconfig because the PMU is not related to memory
management and almost everything under mm/ is. oprofile may also want to use
these bools and that is based in the top-level ARM Kconfig.

Will





More information about the linux-arm-kernel mailing list