Re: [PATCH v7 2/3] KVM: selftests: aarch64: Introduce pmu_event_filter_test

[Shaoqin Huang <shahuang@redhat.com>]

Hi Eric,

On 5/7/24 16:45, Eric Auger wrote:
> Hi Shaoqin,
> 
> On 4/9/24 05:03, Shaoqin Huang wrote:
>> Introduce pmu_event_filter_test for arm64 platforms. The test configures
>> PMUv3 for a vCPU, and sets different pmu event filters for the vCPU, and
>> check if the guest can see those events which user allow and can't use
>> those events which use deny.
>>
>> This test refactor the create_vpmu_vm() and make it a wrapper for
>> __create_vpmu_vm(), which allows some extra init code before
>> KVM_ARM_VCPU_PMU_V3_INIT.
>>
>> And this test use the KVM_ARM_VCPU_PMU_V3_FILTER attribute to set the
>> pmu event filter in KVM. And choose to filter two common event
>> branches_retired and instructions_retired, and let the guest to check if
>> it see the right pmceid register.
>>
>> Signed-off-by: Shaoqin Huang <shahuang@redhat.com>
>> ---
>>   tools/testing/selftests/kvm/Makefile          |   1 +
>>   .../kvm/aarch64/pmu_event_filter_test.c       | 298 ++++++++++++++++++
>>   2 files changed, 299 insertions(+)
>>   create mode 100644 tools/testing/selftests/kvm/aarch64/pmu_event_filter_test.c
>>
>> diff --git a/tools/testing/selftests/kvm/Makefile b/tools/testing/selftests/kvm/Makefile
>> index 741c7dc16afc..9745be534df3 100644
>> --- a/tools/testing/selftests/kvm/Makefile
>> +++ b/tools/testing/selftests/kvm/Makefile
>> @@ -151,6 +151,7 @@ TEST_GEN_PROGS_aarch64 += aarch64/aarch32_id_regs
>>   TEST_GEN_PROGS_aarch64 += aarch64/debug-exceptions
>>   TEST_GEN_PROGS_aarch64 += aarch64/hypercalls
>>   TEST_GEN_PROGS_aarch64 += aarch64/page_fault_test
>> +TEST_GEN_PROGS_aarch64 += aarch64/pmu_event_filter_test
>>   TEST_GEN_PROGS_aarch64 += aarch64/psci_test
>>   TEST_GEN_PROGS_aarch64 += aarch64/set_id_regs
>>   TEST_GEN_PROGS_aarch64 += aarch64/smccc_filter
>> diff --git a/tools/testing/selftests/kvm/aarch64/pmu_event_filter_test.c b/tools/testing/selftests/kvm/aarch64/pmu_event_filter_test.c
>> new file mode 100644
>> index 000000000000..972384e81067
>> --- /dev/null
>> +++ b/tools/testing/selftests/kvm/aarch64/pmu_event_filter_test.c
>> @@ -0,0 +1,298 @@
>> +
>> +// SPDX-License-Identifier: GPL-2.0
>> +/*
>> + * pmu_event_filter_test - Test user limit pmu event for guest.
>> + *
>> + * Copyright (c) 2023 Red Hat, Inc.> + *
>> + * This test checks if the guest only see the limited pmu event that userspace> + * sets, if the guest can use those events which user allow, and if
> the guest
>> + * can't use those events which user deny.
>> + * This test runs only when KVM_CAP_ARM_PMU_V3, KVM_ARM_VCPU_PMU_V3_FILTER> + * is supported on the host.
>> + */
>> +#include <kvm_util.h>
>> +#include <processor.h>
>> +#include <vgic.h>
>> +#include <vpmu.h>
>> +#include <test_util.h>
>> +#include <perf/arm_pmuv3.h>
>> +
>> +struct pmu_common_event_ids {
>> +	uint64_t pmceid0;
>> +	uint64_t pmceid1;
>> +} max_pmce, expected_pmce;
>> +
>> +struct vpmu_vm {
>> +	struct kvm_vm *vm;
>> +	struct kvm_vcpu *vcpu;
>> +	int gic_fd;
>> +};
>> +
>> +static struct vpmu_vm vpmu_vm;
>> +
>> +#define FILTER_NR 10
>> +
>> +struct test_desc {
>> +	const char *name;
>> +	struct kvm_pmu_event_filter filter[FILTER_NR];
>> +};
>> +
>> +#define __DEFINE_FILTER(base, num, act)		\
>> +	((struct kvm_pmu_event_filter) {	\
>> +		.base_event	= base,		\
>> +		.nevents	= num,		\
>> +		.action		= act,		\
>> +	})
>> +
>> +#define DEFINE_FILTER(base, act) __DEFINE_FILTER(base, 1, act)
>> +
>> +static void guest_code(void)
>> +{
>> +	uint64_t pmceid0 = read_sysreg(pmceid0_el0);
>> +	uint64_t pmceid1 = read_sysreg(pmceid1_el0);
>> +
>> +	GUEST_ASSERT_EQ(expected_pmce.pmceid0, pmceid0);
>> +	GUEST_ASSERT_EQ(expected_pmce.pmceid1, pmceid1);
>> +
>> +	GUEST_DONE();
>> +}
>> +
>> +static void guest_get_pmceid(void)
>> +{
>> +	max_pmce.pmceid0 = read_sysreg(pmceid0_el0);
>> +	max_pmce.pmceid1 = read_sysreg(pmceid1_el0);
>> +
>> +	GUEST_DONE();
>> +}
>> +
>> +static void run_vcpu(struct kvm_vcpu *vcpu)
>> +{
>> +	struct ucall uc;
>> +
>> +	while (1) {
>> +		vcpu_run(vcpu);
>> +		switch (get_ucall(vcpu, &uc)) {
>> +		case UCALL_DONE:
>> +			return;
>> +		case UCALL_ABORT:
>> +			REPORT_GUEST_ASSERT(uc);
>> +			break;
>> +		default:
>> +			TEST_FAIL("Unknown ucall %lu", uc.cmd);
>> +		}
>> +	}
>> +}
>> +
>> +static void set_pmce(struct pmu_common_event_ids *pmce, int action, int event)
>> +{
>> +	int base = 0;
>> +	uint64_t *pmceid = NULL;
>> +
>> +	if (event >= 0x4000) {
>> +		event -= 0x4000;
>> +		base = 32;
>> +	}
>> +
>> +	if (event >= 0 && event <= 0x1F) {
>> +		pmceid = &pmce->pmceid0;
>> +	} else if (event >= 0x20 && event <= 0x3F) {
>> +		event -= 0x20;
>> +		pmceid = &pmce->pmceid1;
>> +	} else {
>> +		return;
>> +	}
>> +
>> +	event += base;
>> +	if (action == KVM_PMU_EVENT_ALLOW)
>> +		*pmceid |= BIT(event);
>> +	else
>> +		*pmceid &= ~BIT(event);
>> +}
>> +
>> +static void prepare_expected_pmce(struct kvm_pmu_event_filter *filter)
>> +{
>> +	struct pmu_common_event_ids pmce_mask = { ~0, ~0 };
>> +	bool first_filter = true;
>> +	int i;
>> +
>> +	while (filter && filter->nevents != 0) {
>> +		if (first_filter) {
>> +			if (filter->action == KVM_PMU_EVENT_ALLOW)
>> +				memset(&pmce_mask, 0, sizeof(pmce_mask));
>> +			first_filter = false;
>> +		}
>> +
>> +		for (i = 0; i < filter->nevents; i++)
>> +			set_pmce(&pmce_mask, filter->action,
>> +				 filter->base_event + i);
>> +
>> +		filter++;
>> +	}
>> +
>> +	expected_pmce.pmceid0 = max_pmce.pmceid0 & pmce_mask.pmceid0;
>> +	expected_pmce.pmceid1 = max_pmce.pmceid1 & pmce_mask.pmceid1;
>> +}
>> +
>> +static void pmu_event_filter_init(struct kvm_pmu_event_filter *filter)
>> +{
>> +	while (filter && filter->nevents != 0) {
>> +		kvm_device_attr_set(vpmu_vm.vcpu->fd,
>> +				    KVM_ARM_VCPU_PMU_V3_CTRL,
>> +				    KVM_ARM_VCPU_PMU_V3_FILTER,
>> +				    filter);
>> +		filter++;
>> +	}
>> +}
>> +
>> +#define GICD_BASE_GPA	0x8000000ULL
>> +#define GICR_BASE_GPA	0x80A0000ULL
> in v4 Oliver suggested "Shouldn't a standardized layout of the GIC
> frames go with the rest of the GIC stuff?"
> 

Oliver replied there is another commits did that, so I will remove them 
when I update it.

>> +
>> +/* Create a VM that has one vCPU with PMUv3 configured. */
>> +static void create_vpmu_vm_with_filter(void *guest_code,
>> +				       struct kvm_pmu_event_filter *filter)
>> +{
>> +	uint64_t irq = 23;
>> +
>> +	/* The test creates the vpmu_vm multiple times. Ensure a clean state */
>> +	memset(&vpmu_vm, 0, sizeof(vpmu_vm));
>> +
>> +	vpmu_vm.vm = vm_create(1);
>> +	vpmu_vm.vcpu = vm_vcpu_add_with_vpmu(vpmu_vm.vm, 0, guest_code);
>> +	vpmu_vm.gic_fd = vgic_v3_setup(vpmu_vm.vm, 1, 64,
>> +					GICD_BASE_GPA, GICR_BASE_GPA);
>> +	__TEST_REQUIRE(vpmu_vm.gic_fd >= 0,
>> +		       "Failed to create vgic-v3, skipping");
>> +
>> +	pmu_event_filter_init(filter);
>> +
>> +	/* Initialize vPMU */
>> +	vpmu_set_irq(vpmu_vm.vcpu, irq);
>> +	vpmu_init(vpmu_vm.vcpu);
>> +}
>> +
>> +static void create_vpmu_vm(void *guest_code)
>> +{
>> +	create_vpmu_vm_with_filter(guest_code, NULL);
>> +}
>> +
>> +static void destroy_vpmu_vm(void)
>> +{
>> +	close(vpmu_vm.gic_fd);
>> +	kvm_vm_free(vpmu_vm.vm);
>> +}
>> +
>> +static void run_test(struct test_desc *t)
>> +{
>> +	pr_info("Test: %s\n", t->name);
>> +
>> +	create_vpmu_vm_with_filter(guest_code, t->filter);
>> +	prepare_expected_pmce(t->filter);
>> +	sync_global_to_guest(vpmu_vm.vm, expected_pmce);
>> +
>> +	run_vcpu(vpmu_vm.vcpu);
>> +
>> +	destroy_vpmu_vm();
>> +}
>> +
>> +static struct test_desc tests[] = {
>> +	{
>> +		.name = "without_filter",
>> +		.filter = {
>> +			{ 0 }
>> +		},
>> +	},
>> +	{
>> +		.name = "member_allow_filter",
>> +		.filter = {
>> +			DEFINE_FILTER(ARMV8_PMUV3_PERFCTR_SW_INCR, 0),
>> +			DEFINE_FILTER(ARMV8_PMUV3_PERFCTR_INST_RETIRED, 0),
>> +			DEFINE_FILTER(ARMV8_PMUV3_PERFCTR_BR_RETIRED, 0),
>> +			{ 0 },
>> +		},
>> +	},
>> +	{
>> +		.name = "member_deny_filter",
>> +		.filter = {
>> +			DEFINE_FILTER(ARMV8_PMUV3_PERFCTR_SW_INCR, 1),
>> +			DEFINE_FILTER(ARMV8_PMUV3_PERFCTR_INST_RETIRED, 1),
>> +			DEFINE_FILTERShouldn't a standardized layout of the GIC frames go with the rest of
> the GIC stuff?(ARMV8_PMUV3_PERFCTR_BR_RETIRED, 1),
>> +			{ 0 },
>> +		},
>> +	},
>> +	{
>> +		.name = "not_member_deny_filter",
>> +		.filter = {
>> +			DEFINE_FILTER(ARMV8_PMUV3_PERFCTR_SW_INCR, 1),
>> +			{ 0 },
>> +		},
>> +	},
>> +	{
>> +		.name = "not_member_allow_filter",
>> +		.filter = {
>> +			DEFINE_FILTER(ARMV8_PMUV3_PERFCTR_SW_INCR, 0),
>> +			{ 0 },
>> +		},
>> +	},
>> +	{
>> +		.name = "deny_chain_filter",
>> +		.filter = {
>> +			DEFINE_FILTER(ARMV8_PMUV3_PERFCTR_CHAIN, 1),
>> +			{ 0 },
>> +		},
>> +	},
>> +	{
>> +		.name = "deny_cpu_cycles_filter",
>> +		.filter = {
>> +			DEFINE_FILTER(ARMV8_PMUV3_PERFCTR_CPU_CYCLES, 1),
>> +			{ 0 },
>> +		},
>> +	},
>> +	{
>> +		.name = "cancel_filter",
>> +		.filter = {
>> +			DEFINE_FILTER(ARMV8_PMUV3_PERFCTR_CPU_CYCLES, 0),
>> +			DEFINE_FILTER(ARMV8_PMUV3_PERFCTR_CPU_CYCLES, 1),
>> +		},
>> +	},
>> +	{
>> +		.name = "multiple_filter",
>> +		.filter = {
>> +			__DEFINE_FILTER(0x0, 0x10, 0),
>> +			__DEFINE_FILTER(0x6, 0x3, 1),
>> +		},
>> +	},
>> +	{ 0 }
>> +};
>> +
>> +static void run_tests(void)
>> +{
>> +	struct test_desc *t;
>> +
>> +	for (t = &tests[0]; t->name; t++)
>> +		run_test(t);
>> +}
>> +
>> +static bool kvm_pmu_support_events(void)
>> +{
>> +	create_vpmu_vm(guest_get_pmceid);
>> +
>> +	memset(&max_pmce, 0, sizeof(max_pmce));
>> +	sync_global_to_guest(vpmu_vm.vm, max_pmce);
>> +	run_vcpu(vpmu_vm.vcpu);
>> +	sync_global_from_guest(vpmu_vm.vm, max_pmce);
>> +	destroy_vpmu_vm();
>> +
>> +	return max_pmce.pmceid0 &
>> +	       (ARMV8_PMUV3_PERFCTR_BR_RETIRED |
>> +	       ARMV8_PMUV3_PERFCTR_INST_RETIRED |
>> +	       ARMV8_PMUV3_PERFCTR_CHAIN);
> those are not bit masks but bit shifts. Also don't you want to test that
> all of them are supported?

Thanks for catching this bug. Yes I want to test all of them are 
supported but wrongly checking the bit masks. I will fix them.

> 
> BR_RETIRED is 0x21 so doesn't it belong to pmceid1?

Yes, it should belong to pmceid1. But my wrong checking didn't help me 
find it. Thanks a lot.

> 
> 
> in v4 Oliver suggested to use sysfs instead of spawning a scratch VM.

In that version I changed to function name to kvm_pmu_support_events, 
which means I want to detect what the KVM supports about the PMU events 
rather than the host supportted PMU events. I think that would be more 
suitable if we test KVM.

Thanks,
Shaoqin

>> +}
>> +
>> +int main(void)
>> +{
>> +	TEST_REQUIRE(kvm_has_cap(KVM_CAP_ARM_PMU_V3));
>> +	TEST_REQUIRE(kvm_pmu_support_events());
>> +
>> +	run_tests();
>> +}
> Eric
> 

-- 
Shaoqin