Re: [PATCH v5 20/24] perf: arm_pmuv3: Handle IRQs for Partitioned PMU guest counters

[Colton Lewis]

Oliver Upton <oup...@kernel.org> writes:

On Tue, Dec 09, 2025 at 08:51:17PM +0000, Colton Lewis wrote:

Because ARM hardware is not yet capable of direct interrupt injection

PPI injection, it can do LPIs just fine.

Clarification noted. I will update the message.

@@ -961,6 +964,12 @@ static irqreturn_t armv8pmu_handle_irq(struct arm_pmu *cpu_pmu)

                 */
                perf_event_overflow(event, &data, regs);
        }
+
+       govf = pmovsr & kvm_pmu_guest_counter_mask(cpu_pmu);
+
+       if (kvm_pmu_is_partitioned(cpu_pmu) && govf)
+               kvm_pmu_handle_guest_irq(govf);
+

The state ownership of this whole interaction is very odd. I would much
rather that KVM have full ownership of the range of counters while the
guest is loaded. By that I mean the PMUv3 driver only clears overflows
on PMCs that it owns and KVM will do the same on the back of the IRQ.

If I'm understanding correctly this is a code location concern, because
the host driver has to handle the interrupt in this function or a callee
of this function.

I will do it that way. It would be duplicating a small amount of logic
in kvm_pmu_handle_guest_irq() but I see your reasoning.

Similarly, KVM should be leaving the "guest" range of counters in a
non-overflow condition at vcpu_put().

Noted from your comments on kvm_pmu_put()

Thanks,
Oliver