Re: [PATCH v2 00/11] perf: Intel Cache QoS Monitoring support

[Peter Zijlstra <peterz@infradead.org>]

On Thu, Oct 02, 2014 at 07:51:24AM +0100, Matt Fleming wrote:
> On Wed, 01 Oct, at 02:42:05PM, Andi Kleen wrote:
> > Matt Fleming <matt@console-pimps.org> writes:
> > >
> > > Changes in v2:
> > >
> > >  - Added Jiri Olsa's Acked-by to PATCH 02/11.
> > >
> > >  - Use x86_match_cpu() in intel_cqm_init() and make sure we grab the
> > >    hotplug lock to prevent races as pointed out by Andi Kleen in
> > >    PATCH 08/11.
> > >
> > >  - Delete a stale comment in commit message of PATCH 10/11.
> > 
> > No support for a modular driver?
> 
> I didn't include support for that because none of the other x86 pmu
> drivers are buildable as modules. Is there a reason for that?

Not really, the rapl driver could easily be a module, and I think there
were osme patches to make the uncore driver a module.

> It's not clear to me that it would be a worthwhile change to make this
> buildable as a module, since that would require a Kconfig entry and the
> user would be required to make a decision regarding whether or not they
> want to build CQM support.
> 
> It's much simpler to just build it in and enable it based on
> CONFIG_PERF_EVENTS + CONFIG_INTEL_CPU. In fact, that's one of the things
> I've always liked about the pmu drivers, if I move to a new machine and
> turn on PERF_EVENTS the correct drivers will be run.

Some people worry about the memory footprint of all that.. That said, I
think we can do without the Kconfig stuff if we allow CONFIG_PERF_EVENTS
to be a tristate, in that case M would build everything that can be a
module as a module and keep the rest builtin.