Re: [PATCH/RFC v1 0/2] Human readable performance event description in sysfs

[Russell King - ARM Linux <linux@arm.linux.org.uk>]

On Wed, Jan 20, 2010 at 03:03:03PM +0000, Jamie Iles wrote:
> On Wed, Jan 20, 2010 at 02:41:40PM +0000, Russell King - ARM Linux wrote:
> > Given that history has shown that identification schemes on ARM change
> > in extremely annoying ways, I don't think decoding these registers to
> > some kind of textual representation for /proc/cpuinfo is the right
> > approach.  It might instead make more sense to just export the entire
> > set of CPU ID registers to userspace, and let userspace grapple with
> > the complexities of decoding the information it wants from them.
> Yes, this would probably be the best generic solution, but in the specific
> case of ARM perfevents, the kernel code already has to decode some of the CPU
> ID registers to work out what set of events to use. Why make userspace do all
> of this decoding again? The x86 code sets up the x86_pmu depending on CPU type
> so this is doing a similar thing (although it is easier for x86).

If you're referring to reading the main CPU ID register and relying
on the part number telling you what CPU you're running on, that's
unreliable if you're only checking the part number - you at least
need to check the implementer.

If you want to do ID checking via the main ID register, there are
some clashes even if you take the implementer field into account.