[PATCH] KVM: arm64: Replace KVM_ARM_PMU with HW_PERF_EVENTS

[Marc Zyngier]

Hi Alex,

On 2021-01-05 15:49, Alexandru Elisei wrote:
> Hi Marc,
> 
> On 1/4/21 5:27 PM, Marc Zyngier wrote:
>> KVM_ARM_PMU only existed for the benefit of 32bit ARM hosts,
>> and makes no sense now that we are 64bit only. Get rid of it.
>> 
>> Signed-off-by: Marc Zyngier <maz at kernel.org>
>> ---
>>  arch/arm64/kvm/Kconfig  | 8 --------
>>  arch/arm64/kvm/Makefile | 2 +-
>>  include/kvm/arm_pmu.h   | 2 +-
>>  3 files changed, 2 insertions(+), 10 deletions(-)
>> 
>> diff --git a/arch/arm64/kvm/Kconfig b/arch/arm64/kvm/Kconfig
>> index 043756db8f6e..3964acf5451e 100644
>> --- a/arch/arm64/kvm/Kconfig
>> +++ b/arch/arm64/kvm/Kconfig
>> @@ -49,14 +49,6 @@ if KVM
>> 
>>  source "virt/kvm/Kconfig"
>> 
>> -config KVM_ARM_PMU
>> -	bool "Virtual Performance Monitoring Unit (PMU) support"
>> -	depends on HW_PERF_EVENTS
>> -	default y
>> -	help
>> -	  Adds support for a virtual Performance Monitoring Unit (PMU) in
>> -	  virtual machines.
>> -
>>  endif # KVM
>> 
>>  endif # VIRTUALIZATION
>> diff --git a/arch/arm64/kvm/Makefile b/arch/arm64/kvm/Makefile
>> index 60fd181df624..13b017284bf9 100644
>> --- a/arch/arm64/kvm/Makefile
>> +++ b/arch/arm64/kvm/Makefile
>> @@ -24,4 +24,4 @@ kvm-y := $(KVM)/kvm_main.o $(KVM)/coalesced_mmio.o 
>> $(KVM)/eventfd.o \
>>  	 vgic/vgic-mmio-v3.o vgic/vgic-kvm-device.o \
>>  	 vgic/vgic-its.o vgic/vgic-debug.o
>> 
>> -kvm-$(CONFIG_KVM_ARM_PMU)  += pmu-emul.o
>> +kvm-$(CONFIG_HW_PERF_EVENTS)  += pmu-emul.o
>> diff --git a/include/kvm/arm_pmu.h b/include/kvm/arm_pmu.h
>> index fc85f50fa0e9..8dcb3e1477bc 100644
>> --- a/include/kvm/arm_pmu.h
>> +++ b/include/kvm/arm_pmu.h
>> @@ -13,7 +13,7 @@
>>  #define ARMV8_PMU_CYCLE_IDX		(ARMV8_PMU_MAX_COUNTERS - 1)
>>  #define ARMV8_PMU_MAX_COUNTER_PAIRS	((ARMV8_PMU_MAX_COUNTERS + 1) >> 
>> 1)
>> 
>> -#ifdef CONFIG_KVM_ARM_PMU
>> +#ifdef CONFIG_HW_PERF_EVENTS
>> 
>>  struct kvm_pmc {
>>  	u8 idx;	/* index into the pmu->pmc array */
> 
> I grep'ed for KVM_ARM_PMU in the Linux sources, this patch takes care 
> of all its
> occurrences.
> 
> A few things popped into my mind when I saw the patch.
> 
> 1. Replacing KVM_ARM_PMU with CONFIG_HW_PERF_EVENTS means it's not 
> possible
> anymore for the host to have perf support while KVM does not support 
> emulating a
> PMU. In this scenario, functions which would have been empty functions 
> if
> KVM_ARM_PMU was still around (I only found kvm_pmu_flush_hwstate() and
> kvm_pmu_sync_hwstate() on the KVM_RUN path) will now be called and 
> return early
> after kvm_vcpu_has_pmu() returns 0. The overhead looks negligible to 
> me, and I
> don't think this configuration was common (especially since the default 
> was y).

I don't think this is either common nor useful. If the kernel supports
the PMU, then finding a PMU enables all the uses of the PMU, including
for guests. And userspace is still in control of what it exposes to
the guest. Yes, it's a tiny more overhead (one extra load) on the
exit/entry path. Should we care? I don't think so.

> 2. I did a grep for the files that include arm_pmu.h, and all the files 
> were in
> arch/arm64. I suppose arm_pmu.h exists in include/kvm instead of
> arch/arm64/include/asm because it was shared with KVM/arm when it was 
> still
> around, right? Or is there another reason for that?

No, that's basically the only reason. That was the easy landing spot
for anything shared (including things like GIC, timers and co).
I'm not sure it is worth the move, TBH...

> [1] https://www.spinics.net/lists/kvm-arm/msg44184.html

Yuo, and that's the reason I posted this patch. I have a couple more
that I'll post by the end of the day.

Thanks,

         M.
-- 
Jazz is not dead. It just smells funny...