This series adds a fix for KVM PMU code and improves the pmu selftest by allowing generating precise number of interrupts. It also provided another additional option to the overflow test that allows user to generate custom number of LCOFI interrupts.
Signed-off-by: Atish Patra atishp@rivosinc.com --- Changes in v2: - Initialized the local overflow irq variable to 0 indicate that it's not a allowed value. - Moved the introduction of argument option `n` to the last patch. - Link to v1: https://lore.kernel.org/r/20250226-kvm_pmu_improve-v1-0-74c058c2bf6d@rivosin...
--- Atish Patra (4): RISC-V: KVM: Disable the kernel perf counter during configure KVM: riscv: selftests: Do not start the counter in the overflow handler KVM: riscv: selftests: Change command line option KVM: riscv: selftests: Allow number of interrupts to be configurable
arch/riscv/kvm/vcpu_pmu.c | 1 + tools/testing/selftests/kvm/riscv/sbi_pmu_test.c | 81 ++++++++++++++++-------- 2 files changed, 57 insertions(+), 25 deletions(-) --- base-commit: 0ad2507d5d93f39619fc42372c347d6006b64319 change-id: 20250225-kvm_pmu_improve-fffd038b2404 -- Regards, Atish patra
The perf event should be marked disabled during the creation as it is not ready to be scheduled until there is SBI PMU start call or config matching is called with auto start. Otherwise, event add/start gets called during perf_event_create_kernel_counter function. It will be enabled and scheduled to run via perf_event_enable during either the above mentioned scenario.
Fixes: 0cb74b65d2e5 ("RISC-V: KVM: Implement perf support without sampling")
Reviewed-by: Andrew Jones ajones@ventanamicro.com Signed-off-by: Atish Patra atishp@rivosinc.com --- arch/riscv/kvm/vcpu_pmu.c | 1 + 1 file changed, 1 insertion(+)
diff --git a/arch/riscv/kvm/vcpu_pmu.c b/arch/riscv/kvm/vcpu_pmu.c index 2707a51b082c..78ac3216a54d 100644 --- a/arch/riscv/kvm/vcpu_pmu.c +++ b/arch/riscv/kvm/vcpu_pmu.c @@ -666,6 +666,7 @@ int kvm_riscv_vcpu_pmu_ctr_cfg_match(struct kvm_vcpu *vcpu, unsigned long ctr_ba .type = etype, .size = sizeof(struct perf_event_attr), .pinned = true, + .disabled = true, /* * It should never reach here if the platform doesn't support the sscofpmf * extension as mode filtering won't work without it.
There is no need to start the counter in the overflow handler as we intend to trigger precise number of LCOFI interrupts through these tests. The overflow irq handler has already stopped the counter. As a result, the stop call from the test function may return already stopped error which is fine as well.
Reviewed-by: Andrew Jones ajones@ventanamicro.com Signed-off-by: Atish Patra atishp@rivosinc.com --- tools/testing/selftests/kvm/riscv/sbi_pmu_test.c | 9 ++------- 1 file changed, 2 insertions(+), 7 deletions(-)
diff --git a/tools/testing/selftests/kvm/riscv/sbi_pmu_test.c b/tools/testing/selftests/kvm/riscv/sbi_pmu_test.c index f45c0ecc902d..284bc80193bd 100644 --- a/tools/testing/selftests/kvm/riscv/sbi_pmu_test.c +++ b/tools/testing/selftests/kvm/riscv/sbi_pmu_test.c @@ -118,8 +118,8 @@ static void stop_counter(unsigned long counter, unsigned long stop_flags)
ret = sbi_ecall(SBI_EXT_PMU, SBI_EXT_PMU_COUNTER_STOP, counter, 1, stop_flags, 0, 0, 0); - __GUEST_ASSERT(ret.error == 0, "Unable to stop counter %ld error %ld\n", - counter, ret.error); + __GUEST_ASSERT(ret.error == 0 || ret.error == SBI_ERR_ALREADY_STOPPED, + "Unable to stop counter %ld error %ld\n", counter, ret.error); }
static void guest_illegal_exception_handler(struct ex_regs *regs) @@ -137,7 +137,6 @@ static void guest_irq_handler(struct ex_regs *regs) unsigned int irq_num = regs->cause & ~CAUSE_IRQ_FLAG; struct riscv_pmu_snapshot_data *snapshot_data = snapshot_gva; unsigned long overflown_mask; - unsigned long counter_val = 0;
/* Validate that we are in the correct irq handler */ GUEST_ASSERT_EQ(irq_num, IRQ_PMU_OVF); @@ -151,10 +150,6 @@ static void guest_irq_handler(struct ex_regs *regs) GUEST_ASSERT(overflown_mask & 0x01);
WRITE_ONCE(vcpu_shared_irq_count, vcpu_shared_irq_count+1); - - counter_val = READ_ONCE(snapshot_data->ctr_values[0]); - /* Now start the counter to mimick the real driver behavior */ - start_counter(counter_in_use, SBI_PMU_START_FLAG_SET_INIT_VALUE, counter_val); }
static unsigned long get_counter_index(unsigned long cbase, unsigned long cmask,
The PMU test commandline option takes an argument to disable a certain test. The initial assumption behind this was a common use case is just to run all the test most of the time. However, running a single test seems more useful instead. Especially, the overflow test has been helpful to validate PMU virtualizaiton interrupt changes.
Switching the command line option to run a single test instead of disabling a single test also allows to provide additional test specific arguments to the test. The default without any options remains unchanged which continues to run all the tests.
Reviewed-by: Andrew Jones ajones@ventanamicro.com Signed-off-by: Atish Patra atishp@rivosinc.com --- tools/testing/selftests/kvm/riscv/sbi_pmu_test.c | 40 +++++++++++++++--------- 1 file changed, 26 insertions(+), 14 deletions(-)
diff --git a/tools/testing/selftests/kvm/riscv/sbi_pmu_test.c b/tools/testing/selftests/kvm/riscv/sbi_pmu_test.c index 284bc80193bd..de66099235d9 100644 --- a/tools/testing/selftests/kvm/riscv/sbi_pmu_test.c +++ b/tools/testing/selftests/kvm/riscv/sbi_pmu_test.c @@ -39,7 +39,11 @@ static bool illegal_handler_invoked; #define SBI_PMU_TEST_SNAPSHOT BIT(2) #define SBI_PMU_TEST_OVERFLOW BIT(3)
-static int disabled_tests; +struct test_args { + int disabled_tests; +}; + +static struct test_args targs;
unsigned long pmu_csr_read_num(int csr_num) { @@ -604,7 +608,11 @@ static void test_vm_events_overflow(void *guest_code) vcpu_init_vector_tables(vcpu); /* Initialize guest timer frequency. */ timer_freq = vcpu_get_reg(vcpu, RISCV_TIMER_REG(frequency)); + + /* Export the shared variables to the guest */ sync_global_to_guest(vm, timer_freq); + sync_global_to_guest(vm, vcpu_shared_irq_count); + sync_global_to_guest(vm, targs);
run_vcpu(vcpu);
@@ -613,28 +621,30 @@ static void test_vm_events_overflow(void *guest_code)
static void test_print_help(char *name) { - pr_info("Usage: %s [-h] [-d <test name>]\n", name); - pr_info("\t-d: Test to disable. Available tests are 'basic', 'events', 'snapshot', 'overflow'\n"); + pr_info("Usage: %s [-h] [-t <test name>]\n", name); + pr_info("\t-t: Test to run (default all). Available tests are 'basic', 'events', 'snapshot', 'overflow'\n"); pr_info("\t-h: print this help screen\n"); }
static bool parse_args(int argc, char *argv[]) { int opt; - - while ((opt = getopt(argc, argv, "hd:")) != -1) { + int temp_disabled_tests = SBI_PMU_TEST_BASIC | SBI_PMU_TEST_EVENTS | SBI_PMU_TEST_SNAPSHOT | + SBI_PMU_TEST_OVERFLOW; + while ((opt = getopt(argc, argv, "ht:")) != -1) { switch (opt) { - case 'd': + case 't': if (!strncmp("basic", optarg, 5)) - disabled_tests |= SBI_PMU_TEST_BASIC; + temp_disabled_tests &= ~SBI_PMU_TEST_BASIC; else if (!strncmp("events", optarg, 6)) - disabled_tests |= SBI_PMU_TEST_EVENTS; + temp_disabled_tests &= ~SBI_PMU_TEST_EVENTS; else if (!strncmp("snapshot", optarg, 8)) - disabled_tests |= SBI_PMU_TEST_SNAPSHOT; + temp_disabled_tests &= ~SBI_PMU_TEST_SNAPSHOT; else if (!strncmp("overflow", optarg, 8)) - disabled_tests |= SBI_PMU_TEST_OVERFLOW; + temp_disabled_tests &= ~SBI_PMU_TEST_OVERFLOW; else goto done; + targs.disabled_tests = temp_disabled_tests; break; case 'h': default: @@ -650,25 +660,27 @@ static bool parse_args(int argc, char *argv[])
int main(int argc, char *argv[]) { + targs.disabled_tests = 0; + if (!parse_args(argc, argv)) exit(KSFT_SKIP);
- if (!(disabled_tests & SBI_PMU_TEST_BASIC)) { + if (!(targs.disabled_tests & SBI_PMU_TEST_BASIC)) { test_vm_basic_test(test_pmu_basic_sanity); pr_info("SBI PMU basic test : PASS\n"); }
- if (!(disabled_tests & SBI_PMU_TEST_EVENTS)) { + if (!(targs.disabled_tests & SBI_PMU_TEST_EVENTS)) { test_vm_events_test(test_pmu_events); pr_info("SBI PMU event verification test : PASS\n"); }
- if (!(disabled_tests & SBI_PMU_TEST_SNAPSHOT)) { + if (!(targs.disabled_tests & SBI_PMU_TEST_SNAPSHOT)) { test_vm_events_snapshot_test(test_pmu_events_snaphost); pr_info("SBI PMU event verification with snapshot test : PASS\n"); }
- if (!(disabled_tests & SBI_PMU_TEST_OVERFLOW)) { + if (!(targs.disabled_tests & SBI_PMU_TEST_OVERFLOW)) { test_vm_events_overflow(test_pmu_events_overflow); pr_info("SBI PMU event verification with overflow test : PASS\n"); }
It is helpful to vary the number of the LCOFI interrupts generated by the overflow test. Allow additional argument for overflow test to accommodate that. It can be easily cross-validated with /proc/interrupts output in the host.
Signed-off-by: Atish Patra atishp@rivosinc.com --- tools/testing/selftests/kvm/riscv/sbi_pmu_test.c | 38 +++++++++++++++++++----- 1 file changed, 31 insertions(+), 7 deletions(-)
diff --git a/tools/testing/selftests/kvm/riscv/sbi_pmu_test.c b/tools/testing/selftests/kvm/riscv/sbi_pmu_test.c index de66099235d9..03406de4989d 100644 --- a/tools/testing/selftests/kvm/riscv/sbi_pmu_test.c +++ b/tools/testing/selftests/kvm/riscv/sbi_pmu_test.c @@ -39,8 +39,10 @@ static bool illegal_handler_invoked; #define SBI_PMU_TEST_SNAPSHOT BIT(2) #define SBI_PMU_TEST_OVERFLOW BIT(3)
+#define SBI_PMU_OVERFLOW_IRQNUM_DEFAULT 5 struct test_args { int disabled_tests; + int overflow_irqnum; };
static struct test_args targs; @@ -478,7 +480,7 @@ static void test_pmu_events_snaphost(void)
static void test_pmu_events_overflow(void) { - int num_counters = 0; + int num_counters = 0, i = 0;
/* Verify presence of SBI PMU and minimum requrired SBI version */ verify_sbi_requirement_assert(); @@ -495,11 +497,15 @@ static void test_pmu_events_overflow(void) * Qemu supports overflow for cycle/instruction. * This test may fail on any platform that do not support overflow for these two events. */ - test_pmu_event_overflow(SBI_PMU_HW_CPU_CYCLES); - GUEST_ASSERT_EQ(vcpu_shared_irq_count, 1); + for (i = 0; i < targs.overflow_irqnum; i++) + test_pmu_event_overflow(SBI_PMU_HW_CPU_CYCLES); + GUEST_ASSERT_EQ(vcpu_shared_irq_count, targs.overflow_irqnum); + + vcpu_shared_irq_count = 0;
- test_pmu_event_overflow(SBI_PMU_HW_INSTRUCTIONS); - GUEST_ASSERT_EQ(vcpu_shared_irq_count, 2); + for (i = 0; i < targs.overflow_irqnum; i++) + test_pmu_event_overflow(SBI_PMU_HW_INSTRUCTIONS); + GUEST_ASSERT_EQ(vcpu_shared_irq_count, targs.overflow_irqnum);
GUEST_DONE(); } @@ -621,8 +627,11 @@ static void test_vm_events_overflow(void *guest_code)
static void test_print_help(char *name) { - pr_info("Usage: %s [-h] [-t <test name>]\n", name); + pr_info("Usage: %s [-h] [-t <test name>] [-n <number of LCOFI interrupt for overflow test>]\n", + name); pr_info("\t-t: Test to run (default all). Available tests are 'basic', 'events', 'snapshot', 'overflow'\n"); + pr_info("\t-n: Number of LCOFI interrupt to trigger for each event in overflow test (default: %d)\n", + SBI_PMU_OVERFLOW_IRQNUM_DEFAULT); pr_info("\t-h: print this help screen\n"); }
@@ -631,7 +640,9 @@ static bool parse_args(int argc, char *argv[]) int opt; int temp_disabled_tests = SBI_PMU_TEST_BASIC | SBI_PMU_TEST_EVENTS | SBI_PMU_TEST_SNAPSHOT | SBI_PMU_TEST_OVERFLOW; - while ((opt = getopt(argc, argv, "ht:")) != -1) { + int overflow_interrupts = 0; + + while ((opt = getopt(argc, argv, "ht:n:")) != -1) { switch (opt) { case 't': if (!strncmp("basic", optarg, 5)) @@ -646,12 +657,24 @@ static bool parse_args(int argc, char *argv[]) goto done; targs.disabled_tests = temp_disabled_tests; break; + case 'n': + overflow_interrupts = atoi_positive("Number of LCOFI", optarg); + break; case 'h': default: goto done; } }
+ if (overflow_interrupts > 0) { + if (targs.disabled_tests & SBI_PMU_TEST_OVERFLOW) { + pr_info("-n option is only available for overflow test\n"); + goto done; + } else { + targs.overflow_irqnum = overflow_interrupts; + } + } + return true; done: test_print_help(argv[0]); @@ -661,6 +684,7 @@ static bool parse_args(int argc, char *argv[]) int main(int argc, char *argv[]) { targs.disabled_tests = 0; + targs.overflow_irqnum = SBI_PMU_OVERFLOW_IRQNUM_DEFAULT;
if (!parse_args(argc, argv)) exit(KSFT_SKIP);
On Mon, Mar 03, 2025 at 02:53:09PM -0800, Atish Patra wrote:
It is helpful to vary the number of the LCOFI interrupts generated by the overflow test. Allow additional argument for overflow test to accommodate that. It can be easily cross-validated with /proc/interrupts output in the host.
Signed-off-by: Atish Patra atishp@rivosinc.com
tools/testing/selftests/kvm/riscv/sbi_pmu_test.c | 38 +++++++++++++++++++----- 1 file changed, 31 insertions(+), 7 deletions(-)
Reviewed-by: Andrew Jones ajones@ventanamicro.com
On Tue, Mar 4, 2025 at 4:23 AM Atish Patra atishp@rivosinc.com wrote:
This series adds a fix for KVM PMU code and improves the pmu selftest by allowing generating precise number of interrupts. It also provided another additional option to the overflow test that allows user to generate custom number of LCOFI interrupts.
Signed-off-by: Atish Patra atishp@rivosinc.com
Changes in v2:
- Initialized the local overflow irq variable to 0 indicate that it's not a allowed value.
- Moved the introduction of argument option `n` to the last patch.
- Link to v1: https://lore.kernel.org/r/20250226-kvm_pmu_improve-v1-0-74c058c2bf6d@rivosin...
Atish Patra (4): RISC-V: KVM: Disable the kernel perf counter during configure KVM: riscv: selftests: Do not start the counter in the overflow handler KVM: riscv: selftests: Change command line option KVM: riscv: selftests: Allow number of interrupts to be configurable
Queued this series for Linux-6.15.
Thanks, Anup
arch/riscv/kvm/vcpu_pmu.c | 1 + tools/testing/selftests/kvm/riscv/sbi_pmu_test.c | 81 ++++++++++++++++-------- 2 files changed, 57 insertions(+), 25 deletions(-)
base-commit: 0ad2507d5d93f39619fc42372c347d6006b64319 change-id: 20250225-kvm_pmu_improve-fffd038b2404 -- Regards, Atish patra
linux-kselftest-mirror@lists.linaro.org
-
Andrew Jones -
Anup Patel -
Atish Patra