Re: [PATCH 47/47] KVM: arm64: nv: Add trap forwarding for FEAT_FGT2 described registers

[Anshuman Khandual]

On 10/1/24 20:16, Oliver Upton wrote:
> Hi Anshuman,
> 
> On Tue, Oct 01, 2024 at 08:13:56AM +0530, Anshuman Khandual wrote:
>> +#define check_cntr_accessible(num)						\
>> +static enum trap_behaviour check_cntr_accessible_##num(struct kvm_vcpu *vcpu)	\
>> +{										\
>> +	u64 mdcr_el2 = __vcpu_sys_reg(vcpu, MDCR_EL2);				\
>> +	int cntr = FIELD_GET(MDCR_EL2_HPMN_MASK, mdcr_el2);			\
>> +										\
>> +	if (num >= cntr)							\
>> +		return BEHAVE_FORWARD_ANY;					\
>> +	return BEHAVE_HANDLE_LOCALLY;						\
>> +}										\
>> +
>> +check_cntr_accessible(0)
>> +check_cntr_accessible(1)
>> +check_cntr_accessible(2)
>> +check_cntr_accessible(3)
>> +check_cntr_accessible(4)
>> +check_cntr_accessible(5)
>> +check_cntr_accessible(6)
>> +check_cntr_accessible(7)
>> +check_cntr_accessible(8)
>> +check_cntr_accessible(9)
>> +check_cntr_accessible(10)
>> +check_cntr_accessible(11)
>> +check_cntr_accessible(12)
>> +check_cntr_accessible(13)
>> +check_cntr_accessible(14)
>> +check_cntr_accessible(15)
>> +check_cntr_accessible(16)
>> +check_cntr_accessible(17)
>> +check_cntr_accessible(18)
>> +check_cntr_accessible(19)
>> +check_cntr_accessible(20)
>> +check_cntr_accessible(21)
>> +check_cntr_accessible(22)
>> +check_cntr_accessible(23)
>> +check_cntr_accessible(24)
>> +check_cntr_accessible(25)
>> +check_cntr_accessible(26)
>> +check_cntr_accessible(27)
>> +check_cntr_accessible(28)
>> +check_cntr_accessible(29)
>> +check_cntr_accessible(30)
> 
> I'd rather we not use templates for this problem. It bloats the kernel text
> as well as the trap encoding space.

Alright, fair point.

> 
> I have a patch in the nested PMU series that uses a single complex trap
> ID to evaluate HPMN, and derives the index from ESR_EL2. I think it
> could also be extended to the PMEVCNTSVR<n> range as well.

Just for reference - the mentioned complex trap ID function from the
given link below.

static enum trap_behaviour check_mdcr_hpmn(struct kvm_vcpu *vcpu)
{
	u32 sysreg = esr_sys64_to_sysreg(kvm_vcpu_get_esr(vcpu));
	u64 mask = kvm_pmu_accessible_counter_mask(vcpu);
	unsigned int idx;


	switch (sysreg) {
	case SYS_PMEVTYPERn_EL0(0) ... SYS_PMEVTYPERn_EL0(30):
	case SYS_PMEVCNTRn_EL0(0) ... SYS_PMEVCNTRn_EL0(30):

---------------------------------------------------------------------
Just add the new system register range here ?

+	case SYS_PMEVCNTSVR_EL1(0)... SYS_PMEVCNTSVR_EL1(31):
---------------------------------------------------------------------

		idx = (sys_reg_CRm(sysreg) & 0x3) << 3 | sys_reg_Op2(sysreg);
		break;
	case SYS_PMXEVTYPER_EL0:
	case SYS_PMXEVCNTR_EL0:
		idx = __vcpu_sys_reg(vcpu, PMSELR_EL0) & ARMV8_PMU_COUNTER_MASK;
		break;
	default:
		/* Someone used this trap helper for something else... */
		KVM_BUG_ON(1, vcpu->kvm);
		return BEHAVE_HANDLE_LOCALLY;
	}

	/*
	 * Programming HPMN=0 is CONSTRAINED UNPREDICTABLE if FEAT_HPMN0 isn't
	 * implemented. Since KVM's ability to emulate HPMN=0 does not directly
	 * depend on hardware (all PMU registers are trapped), make the
	 * implementation choice that all counters are included in the second
	 * range reserved for EL2/EL3.
	 */
	return !(BIT(idx) & mask) ? (BEHAVE_FORWARD_RW | BEHAVE_IN_HOST_EL0) :
			BEHAVE_HANDLE_LOCALLY;
}

> 
> Also, keep in mind that the HPMN trap is annoying since it affects Host
> EL0 in addition to 'guest' ELs.

Does this require any more special handling other than the above complex trap
ID function ?

> 
> [*]: https://lore.kernel.org/kvmarm/20240827002235.1753237-9-oliver.upton@xxxxxxxxx/
>