[PATCH 2/7] ARM: perf_event: Support percpu irqs for the CPU PMU

[Will Deacon]

Hi Stephen,

On Tue, Jan 14, 2014 at 08:57:53PM +0000, Stephen Boyd wrote:
> On 01/13/14 03:52, Will Deacon wrote:
> > I also don't think that's the right solution, based on the above. It's
> > actually pretty hard to work out what's the right thing to do here...
> 
> Yes it doesn't seem like the right solution.
> 
> >
> > We *could* have a per-cpu pointer to the cpu_pmu_pointer, but then we'd
> > need to update the IRQ handlers, including things like the CCI PMU which
> > really doesn't care about per-cpu stuff. So after all this, the shim we have
> > around the IRQ handler for the U8500 SPI workarounds might be the right
> > thing after all -- it allows us to consolidate the conversion of a pcpu
> > pointer into the relevant instance (actually any instance, since they'd all
> > point at the same thing) for the current CPU.
> >
> > What do you think to having that shim throw away the second level pcpu
> > pointer in the case of a PPI? (probably means we need to revisit that
> > renaming again).
> 
> Ok I think I understand what you're getting at. We pass a per-cpu
> pointer to the cpu_pmu pointer as the dev_id argument to the PPI irq
> handler, and then we check to see if the irq is per-cpu inside the
> armpmu_dispatch_irq() function and throw away the second level of
> pointer, i.e.
> 
> static irqreturn_t armpmu_dispatch_irq(int irq, void *dev)
> {
>         struct arm_pmu *armpmu;
>         struct platform_device *plat_device;
>         struct arm_pmu_platdata *plat;
> 
>         if (irq_is_percpu(irq))
>                 dev = *(struct arm_pmu_cpu **)dev;
>         armpmu = dev;
>         plat_device = armpmu->plat_device;
>         plat = dev_get_platdata(&plat_device->dev);
> 
>         if (plat && plat->handle_irq)
>                 return plat->handle_irq(irq, dev, armpmu->handle_irq);
>         else
>                 return armpmu->handle_irq(irq, dev);
> }

Yup, that's what I was trying to explain (badly). Thanks.

> We still need to make a per-cpu variable to hold the pointer, and assign
> it during cpu_pmu_init like this patch does. Hopefully that is ok.

I think that's ok. The percpu code in genirq requires a pcpu token (for good
reason) and the irq handler needs to get at the pmu structure. The
alternative is adding pointers from something like the pmu_hw_events to the
arm_pmu, but I think that's more ugly.

Will