* [kvm-unit-tests PATCH v2 0/4] arm: pmu: Add support for PMUv3p5
@ 2022-12-20 3:10 ` Ricardo Koller
0 siblings, 0 replies; 13+ messages in thread
From: Ricardo Koller @ 2022-12-20 3:10 UTC (permalink / raw)
To: kvm, kvmarm, andrew.jones
Cc: maz, alexandru.elisei, eric.auger, oliver.upton, reijiw, Ricardo Koller
The first commit fixes the tests when running on PMUv3p5. The issue is that
PMUv3p5 uses 64-bit counters irrespective of whether the PMU is configured
for overflowing at 32 or 64-bits. Tests are currently failing [0] on
PMUv3p5 because of this. They wrongly assume that values will be wrapped
around 32-bits, but they overflow into the other half of the 64-bit
counters.
The second and third commits add new tests for 64-bit overflows, a feature
added with PMUv3p5 (PMCR_EL0.LP == 1). This is done by running all
overflow-related tests in two modes: with 32-bit and 64-bit overflows.
The fourt commit changes the value reporting to use %lx instead of %ld.
This series was tested on PMUv3p5 and PMUv3p4 using the ARM Fast Model and
kvmtool. All tests pass on both PMUv3p5 and PMUv3p4 when using Marc's
PMUv3p5 series [0], plus the suggestion made at [1]. Didn't test AArch32.
Changes from v1 (all suggested by Alexandru):
- report counter values in hexadecimal
- s/overflow_at_64bits/unused for all chained tests
- check that odd counters do not increment when using overflow_at_64bits
(pmcr.LP=1)
- test 64-bit odd counters overflows
- switch confusing var names in test_chained_sw_incr(): cntr0 <-> cntr1
[0] https://lore.kernel.org/kvmarm/20221113163832.3154370-1-maz@kernel.org/
[1] https://lore.kernel.org/kvmarm/Y4jasyxvFRNvvmox@google.com/
Ricardo Koller (4):
arm: pmu: Fix overflow checks for PMUv3p5 long counters
arm: pmu: Prepare for testing 64-bit overflows
arm: pmu: Add tests for 64-bit overflows
arm: pmu: Print counter values as hexadecimals
arm/pmu.c | 240 +++++++++++++++++++++++++++++++++---------------------
1 file changed, 149 insertions(+), 91 deletions(-)
--
2.39.0.314.g84b9a713c41-goog
^ permalink raw reply [flat|nested] 13+ messages in thread
* [kvm-unit-tests PATCH v2 0/4] arm: pmu: Add support for PMUv3p5
@ 2022-12-20 3:10 ` Ricardo Koller
0 siblings, 0 replies; 13+ messages in thread
From: Ricardo Koller @ 2022-12-20 3:10 UTC (permalink / raw)
To: kvm, kvmarm, andrew.jones; +Cc: maz
The first commit fixes the tests when running on PMUv3p5. The issue is that
PMUv3p5 uses 64-bit counters irrespective of whether the PMU is configured
for overflowing at 32 or 64-bits. Tests are currently failing [0] on
PMUv3p5 because of this. They wrongly assume that values will be wrapped
around 32-bits, but they overflow into the other half of the 64-bit
counters.
The second and third commits add new tests for 64-bit overflows, a feature
added with PMUv3p5 (PMCR_EL0.LP == 1). This is done by running all
overflow-related tests in two modes: with 32-bit and 64-bit overflows.
The fourt commit changes the value reporting to use %lx instead of %ld.
This series was tested on PMUv3p5 and PMUv3p4 using the ARM Fast Model and
kvmtool. All tests pass on both PMUv3p5 and PMUv3p4 when using Marc's
PMUv3p5 series [0], plus the suggestion made at [1]. Didn't test AArch32.
Changes from v1 (all suggested by Alexandru):
- report counter values in hexadecimal
- s/overflow_at_64bits/unused for all chained tests
- check that odd counters do not increment when using overflow_at_64bits
(pmcr.LP=1)
- test 64-bit odd counters overflows
- switch confusing var names in test_chained_sw_incr(): cntr0 <-> cntr1
[0] https://lore.kernel.org/kvmarm/20221113163832.3154370-1-maz@kernel.org/
[1] https://lore.kernel.org/kvmarm/Y4jasyxvFRNvvmox@google.com/
Ricardo Koller (4):
arm: pmu: Fix overflow checks for PMUv3p5 long counters
arm: pmu: Prepare for testing 64-bit overflows
arm: pmu: Add tests for 64-bit overflows
arm: pmu: Print counter values as hexadecimals
arm/pmu.c | 240 +++++++++++++++++++++++++++++++++---------------------
1 file changed, 149 insertions(+), 91 deletions(-)
--
2.39.0.314.g84b9a713c41-goog
_______________________________________________
kvmarm mailing list
kvmarm@lists.cs.columbia.edu
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
^ permalink raw reply [flat|nested] 13+ messages in thread
* [kvm-unit-tests PATCH v2 1/4] arm: pmu: Fix overflow checks for PMUv3p5 long counters
2022-12-20 3:10 ` Ricardo Koller
@ 2022-12-20 3:10 ` Ricardo Koller
-1 siblings, 0 replies; 13+ messages in thread
From: Ricardo Koller @ 2022-12-20 3:10 UTC (permalink / raw)
To: kvm, kvmarm, andrew.jones
Cc: maz, alexandru.elisei, eric.auger, oliver.upton, reijiw, Ricardo Koller
PMUv3p5 uses 64-bit counters irrespective of whether the PMU is configured
for overflowing at 32 or 64-bits. The consequence is that tests that check
the counter values after overflowing should not assume that values will be
wrapped around 32-bits: they overflow into the other half of the 64-bit
counters on PMUv3p5.
Fix tests by correctly checking overflowing-counters against the expected
64-bit value.
Signed-off-by: Ricardo Koller <ricarkol@google.com>
---
arm/pmu.c | 37 +++++++++++++++++++++++++------------
1 file changed, 25 insertions(+), 12 deletions(-)
diff --git a/arm/pmu.c b/arm/pmu.c
index cd47b14..1b55e20 100644
--- a/arm/pmu.c
+++ b/arm/pmu.c
@@ -54,10 +54,13 @@
#define EXT_COMMON_EVENTS_LOW 0x4000
#define EXT_COMMON_EVENTS_HIGH 0x403F
-#define ALL_SET 0xFFFFFFFF
-#define ALL_CLEAR 0x0
-#define PRE_OVERFLOW 0xFFFFFFF0
-#define PRE_OVERFLOW2 0xFFFFFFDC
+#define ALL_SET 0x00000000FFFFFFFFULL
+#define ALL_SET_64 0xFFFFFFFFFFFFFFFFULL
+#define ALL_CLEAR 0x0000000000000000ULL
+#define PRE_OVERFLOW 0x00000000FFFFFFF0ULL
+#define PRE_OVERFLOW2 0x00000000FFFFFFDCULL
+
+#define ALL_SET_AT(_64b) (_64b ? ALL_SET_64 : ALL_SET)
#define PMU_PPI 23
@@ -538,6 +541,7 @@ static void test_mem_access(void)
static void test_sw_incr(void)
{
uint32_t events[] = {SW_INCR, SW_INCR};
+ uint64_t cntr0;
int i;
if (!satisfy_prerequisites(events, ARRAY_SIZE(events)))
@@ -572,9 +576,11 @@ static void test_sw_incr(void)
write_sysreg(0x3, pmswinc_el0);
isb();
- report(read_regn_el0(pmevcntr, 0) == 84, "counter #1 after + 100 SW_INCR");
- report(read_regn_el0(pmevcntr, 1) == 100,
- "counter #0 after + 100 SW_INCR");
+ cntr0 = (pmu.version < ID_DFR0_PMU_V3_8_5) ?
+ (uint32_t)PRE_OVERFLOW + 100 :
+ (uint64_t)PRE_OVERFLOW + 100;
+ report(read_regn_el0(pmevcntr, 0) == cntr0, "counter #0 after + 100 SW_INCR");
+ report(read_regn_el0(pmevcntr, 1) == 100, "counter #1 after + 100 SW_INCR");
report_info("counter values after 100 SW_INCR #0=%ld #1=%ld",
read_regn_el0(pmevcntr, 0), read_regn_el0(pmevcntr, 1));
report(read_sysreg(pmovsclr_el0) == 0x1,
@@ -584,6 +590,7 @@ static void test_sw_incr(void)
static void test_chained_counters(void)
{
uint32_t events[] = {CPU_CYCLES, CHAIN};
+ uint64_t all_set = ALL_SET_AT(pmu.version >= ID_DFR0_PMU_V3_8_5);
if (!satisfy_prerequisites(events, ARRAY_SIZE(events)))
return;
@@ -614,17 +621,19 @@ static void test_chained_counters(void)
report(read_sysreg(pmovsclr_el0) == 0x1, "overflow recorded for chained incr #2");
write_regn_el0(pmevcntr, 0, PRE_OVERFLOW);
- write_regn_el0(pmevcntr, 1, ALL_SET);
+ write_regn_el0(pmevcntr, 1, all_set);
precise_instrs_loop(22, pmu.pmcr_ro | PMU_PMCR_E);
report_info("overflow reg = 0x%lx", read_sysreg(pmovsclr_el0));
- report(!read_regn_el0(pmevcntr, 1), "CHAIN counter #1 wrapped");
+ report(read_regn_el0(pmevcntr, 1) == 0, "CHAIN counter #1 wrapped");
+
report(read_sysreg(pmovsclr_el0) == 0x3, "overflow on even and odd counters");
}
static void test_chained_sw_incr(void)
{
uint32_t events[] = {SW_INCR, CHAIN};
+ uint64_t cntr0, cntr1;
int i;
if (!satisfy_prerequisites(events, ARRAY_SIZE(events)))
@@ -665,10 +674,14 @@ static void test_chained_sw_incr(void)
write_sysreg(0x1, pmswinc_el0);
isb();
+ cntr0 = (pmu.version < ID_DFR0_PMU_V3_8_5) ?
+ (uint32_t)PRE_OVERFLOW + 100 :
+ (uint64_t)PRE_OVERFLOW + 100;
+ cntr1 = (pmu.version < ID_DFR0_PMU_V3_8_5) ? 0 : ALL_SET + 1;
report((read_sysreg(pmovsclr_el0) == 0x3) &&
- (read_regn_el0(pmevcntr, 1) == 0) &&
- (read_regn_el0(pmevcntr, 0) == 84),
- "expected overflows and values after 100 SW_INCR/CHAIN");
+ (read_regn_el0(pmevcntr, 0) == cntr0) &&
+ (read_regn_el0(pmevcntr, 1) == cntr1),
+ "expected overflows and values after 100 SW_INCR/CHAIN");
report_info("overflow=0x%lx, #0=%ld #1=%ld", read_sysreg(pmovsclr_el0),
read_regn_el0(pmevcntr, 0), read_regn_el0(pmevcntr, 1));
}
--
2.39.0.314.g84b9a713c41-goog
^ permalink raw reply related [flat|nested] 13+ messages in thread
* [kvm-unit-tests PATCH v2 1/4] arm: pmu: Fix overflow checks for PMUv3p5 long counters
@ 2022-12-20 3:10 ` Ricardo Koller
0 siblings, 0 replies; 13+ messages in thread
From: Ricardo Koller @ 2022-12-20 3:10 UTC (permalink / raw)
To: kvm, kvmarm, andrew.jones; +Cc: maz
PMUv3p5 uses 64-bit counters irrespective of whether the PMU is configured
for overflowing at 32 or 64-bits. The consequence is that tests that check
the counter values after overflowing should not assume that values will be
wrapped around 32-bits: they overflow into the other half of the 64-bit
counters on PMUv3p5.
Fix tests by correctly checking overflowing-counters against the expected
64-bit value.
Signed-off-by: Ricardo Koller <ricarkol@google.com>
---
arm/pmu.c | 37 +++++++++++++++++++++++++------------
1 file changed, 25 insertions(+), 12 deletions(-)
diff --git a/arm/pmu.c b/arm/pmu.c
index cd47b14..1b55e20 100644
--- a/arm/pmu.c
+++ b/arm/pmu.c
@@ -54,10 +54,13 @@
#define EXT_COMMON_EVENTS_LOW 0x4000
#define EXT_COMMON_EVENTS_HIGH 0x403F
-#define ALL_SET 0xFFFFFFFF
-#define ALL_CLEAR 0x0
-#define PRE_OVERFLOW 0xFFFFFFF0
-#define PRE_OVERFLOW2 0xFFFFFFDC
+#define ALL_SET 0x00000000FFFFFFFFULL
+#define ALL_SET_64 0xFFFFFFFFFFFFFFFFULL
+#define ALL_CLEAR 0x0000000000000000ULL
+#define PRE_OVERFLOW 0x00000000FFFFFFF0ULL
+#define PRE_OVERFLOW2 0x00000000FFFFFFDCULL
+
+#define ALL_SET_AT(_64b) (_64b ? ALL_SET_64 : ALL_SET)
#define PMU_PPI 23
@@ -538,6 +541,7 @@ static void test_mem_access(void)
static void test_sw_incr(void)
{
uint32_t events[] = {SW_INCR, SW_INCR};
+ uint64_t cntr0;
int i;
if (!satisfy_prerequisites(events, ARRAY_SIZE(events)))
@@ -572,9 +576,11 @@ static void test_sw_incr(void)
write_sysreg(0x3, pmswinc_el0);
isb();
- report(read_regn_el0(pmevcntr, 0) == 84, "counter #1 after + 100 SW_INCR");
- report(read_regn_el0(pmevcntr, 1) == 100,
- "counter #0 after + 100 SW_INCR");
+ cntr0 = (pmu.version < ID_DFR0_PMU_V3_8_5) ?
+ (uint32_t)PRE_OVERFLOW + 100 :
+ (uint64_t)PRE_OVERFLOW + 100;
+ report(read_regn_el0(pmevcntr, 0) == cntr0, "counter #0 after + 100 SW_INCR");
+ report(read_regn_el0(pmevcntr, 1) == 100, "counter #1 after + 100 SW_INCR");
report_info("counter values after 100 SW_INCR #0=%ld #1=%ld",
read_regn_el0(pmevcntr, 0), read_regn_el0(pmevcntr, 1));
report(read_sysreg(pmovsclr_el0) == 0x1,
@@ -584,6 +590,7 @@ static void test_sw_incr(void)
static void test_chained_counters(void)
{
uint32_t events[] = {CPU_CYCLES, CHAIN};
+ uint64_t all_set = ALL_SET_AT(pmu.version >= ID_DFR0_PMU_V3_8_5);
if (!satisfy_prerequisites(events, ARRAY_SIZE(events)))
return;
@@ -614,17 +621,19 @@ static void test_chained_counters(void)
report(read_sysreg(pmovsclr_el0) == 0x1, "overflow recorded for chained incr #2");
write_regn_el0(pmevcntr, 0, PRE_OVERFLOW);
- write_regn_el0(pmevcntr, 1, ALL_SET);
+ write_regn_el0(pmevcntr, 1, all_set);
precise_instrs_loop(22, pmu.pmcr_ro | PMU_PMCR_E);
report_info("overflow reg = 0x%lx", read_sysreg(pmovsclr_el0));
- report(!read_regn_el0(pmevcntr, 1), "CHAIN counter #1 wrapped");
+ report(read_regn_el0(pmevcntr, 1) == 0, "CHAIN counter #1 wrapped");
+
report(read_sysreg(pmovsclr_el0) == 0x3, "overflow on even and odd counters");
}
static void test_chained_sw_incr(void)
{
uint32_t events[] = {SW_INCR, CHAIN};
+ uint64_t cntr0, cntr1;
int i;
if (!satisfy_prerequisites(events, ARRAY_SIZE(events)))
@@ -665,10 +674,14 @@ static void test_chained_sw_incr(void)
write_sysreg(0x1, pmswinc_el0);
isb();
+ cntr0 = (pmu.version < ID_DFR0_PMU_V3_8_5) ?
+ (uint32_t)PRE_OVERFLOW + 100 :
+ (uint64_t)PRE_OVERFLOW + 100;
+ cntr1 = (pmu.version < ID_DFR0_PMU_V3_8_5) ? 0 : ALL_SET + 1;
report((read_sysreg(pmovsclr_el0) == 0x3) &&
- (read_regn_el0(pmevcntr, 1) == 0) &&
- (read_regn_el0(pmevcntr, 0) == 84),
- "expected overflows and values after 100 SW_INCR/CHAIN");
+ (read_regn_el0(pmevcntr, 0) == cntr0) &&
+ (read_regn_el0(pmevcntr, 1) == cntr1),
+ "expected overflows and values after 100 SW_INCR/CHAIN");
report_info("overflow=0x%lx, #0=%ld #1=%ld", read_sysreg(pmovsclr_el0),
read_regn_el0(pmevcntr, 0), read_regn_el0(pmevcntr, 1));
}
--
2.39.0.314.g84b9a713c41-goog
_______________________________________________
kvmarm mailing list
kvmarm@lists.cs.columbia.edu
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
^ permalink raw reply related [flat|nested] 13+ messages in thread
* [kvm-unit-tests PATCH v2 2/4] arm: pmu: Prepare for testing 64-bit overflows
2022-12-20 3:10 ` Ricardo Koller
@ 2022-12-20 3:10 ` Ricardo Koller
-1 siblings, 0 replies; 13+ messages in thread
From: Ricardo Koller @ 2022-12-20 3:10 UTC (permalink / raw)
To: kvm, kvmarm, andrew.jones
Cc: maz, alexandru.elisei, eric.auger, oliver.upton, reijiw, Ricardo Koller
PMUv3p5 adds a knob, PMCR_EL0.LP == 1, that allows overflowing at 64-bits
instead of 32. Prepare by doing these 3 things:
1. Add a "bool overflow_at_64bits" argument to all tests checking
overflows.
2. Extend satisfy_prerequisites() to check if the machine supports
"overflow_at_64bits".
3. Refactor the test invocations to use the new "run_test()" which adds a
report prefix indicating whether the test uses 64 or 32-bit overflows.
A subsequent commit will actually add the 64-bit overflow tests.
Signed-off-by: Ricardo Koller <ricarkol@google.com>
---
arm/pmu.c | 92 ++++++++++++++++++++++++++++++++-----------------------
1 file changed, 53 insertions(+), 39 deletions(-)
diff --git a/arm/pmu.c b/arm/pmu.c
index 1b55e20..4cd3790 100644
--- a/arm/pmu.c
+++ b/arm/pmu.c
@@ -167,13 +167,13 @@ static void pmu_reset(void)
/* event counter tests only implemented for aarch64 */
static void test_event_introspection(void) {}
static void test_event_counter_config(void) {}
-static void test_basic_event_count(void) {}
-static void test_mem_access(void) {}
-static void test_sw_incr(void) {}
-static void test_chained_counters(void) {}
-static void test_chained_sw_incr(void) {}
-static void test_chain_promotion(void) {}
-static void test_overflow_interrupt(void) {}
+static void test_basic_event_count(bool overflow_at_64bits) {}
+static void test_mem_access(bool overflow_at_64bits) {}
+static void test_sw_incr(bool overflow_at_64bits) {}
+static void test_chained_counters(bool unused) {}
+static void test_chained_sw_incr(bool unused) {}
+static void test_chain_promotion(bool unused) {}
+static void test_overflow_interrupt(bool overflow_at_64bits) {}
#elif defined(__aarch64__)
#define ID_AA64DFR0_PERFMON_SHIFT 8
@@ -419,16 +419,28 @@ static bool satisfy_prerequisites(uint32_t *events, unsigned int nb_events)
return false;
}
}
+
+ return true;
+}
+
+static bool check_overflow_prerequisites(bool overflow_at_64bits)
+{
+ if (overflow_at_64bits && pmu.version < ID_DFR0_PMU_V3_8_5) {
+ report_skip("Skip test as 64 overflows need FEAT_PMUv3p5");
+ return false;
+ }
+
return true;
}
-static void test_basic_event_count(void)
+static void test_basic_event_count(bool overflow_at_64bits)
{
uint32_t implemented_counter_mask, non_implemented_counter_mask;
uint32_t counter_mask;
uint32_t events[] = {CPU_CYCLES, INST_RETIRED};
- if (!satisfy_prerequisites(events, ARRAY_SIZE(events)))
+ if (!satisfy_prerequisites(events, ARRAY_SIZE(events)) ||
+ !check_overflow_prerequisites(overflow_at_64bits))
return;
implemented_counter_mask = BIT(pmu.nb_implemented_counters) - 1;
@@ -502,12 +514,13 @@ static void test_basic_event_count(void)
"check overflow happened on #0 only");
}
-static void test_mem_access(void)
+static void test_mem_access(bool overflow_at_64bits)
{
void *addr = malloc(PAGE_SIZE);
uint32_t events[] = {MEM_ACCESS, MEM_ACCESS};
- if (!satisfy_prerequisites(events, ARRAY_SIZE(events)))
+ if (!satisfy_prerequisites(events, ARRAY_SIZE(events)) ||
+ !check_overflow_prerequisites(overflow_at_64bits))
return;
pmu_reset();
@@ -538,13 +551,14 @@ static void test_mem_access(void)
read_sysreg(pmovsclr_el0));
}
-static void test_sw_incr(void)
+static void test_sw_incr(bool overflow_at_64bits)
{
uint32_t events[] = {SW_INCR, SW_INCR};
uint64_t cntr0;
int i;
- if (!satisfy_prerequisites(events, ARRAY_SIZE(events)))
+ if (!satisfy_prerequisites(events, ARRAY_SIZE(events)) ||
+ !check_overflow_prerequisites(overflow_at_64bits))
return;
pmu_reset();
@@ -587,7 +601,7 @@ static void test_sw_incr(void)
"overflow on counter #0 after 100 SW_INCR");
}
-static void test_chained_counters(void)
+static void test_chained_counters(bool unused)
{
uint32_t events[] = {CPU_CYCLES, CHAIN};
uint64_t all_set = ALL_SET_AT(pmu.version >= ID_DFR0_PMU_V3_8_5);
@@ -630,7 +644,7 @@ static void test_chained_counters(void)
report(read_sysreg(pmovsclr_el0) == 0x3, "overflow on even and odd counters");
}
-static void test_chained_sw_incr(void)
+static void test_chained_sw_incr(bool unused)
{
uint32_t events[] = {SW_INCR, CHAIN};
uint64_t cntr0, cntr1;
@@ -686,7 +700,7 @@ static void test_chained_sw_incr(void)
read_regn_el0(pmevcntr, 0), read_regn_el0(pmevcntr, 1));
}
-static void test_chain_promotion(void)
+static void test_chain_promotion(bool unused)
{
uint32_t events[] = {MEM_ACCESS, CHAIN};
void *addr = malloc(PAGE_SIZE);
@@ -835,13 +849,14 @@ static bool expect_interrupts(uint32_t bitmap)
return true;
}
-static void test_overflow_interrupt(void)
+static void test_overflow_interrupt(bool overflow_at_64bits)
{
uint32_t events[] = {MEM_ACCESS, SW_INCR};
void *addr = malloc(PAGE_SIZE);
int i;
- if (!satisfy_prerequisites(events, ARRAY_SIZE(events)))
+ if (!satisfy_prerequisites(events, ARRAY_SIZE(events)) ||
+ !check_overflow_prerequisites(overflow_at_64bits))
return;
gic_enable_defaults();
@@ -1065,6 +1080,19 @@ static bool pmu_probe(void)
return true;
}
+static void run_test(char *name, void (*test)(bool), bool overflow_at_64bits)
+{
+ const char *prefix = overflow_at_64bits ? "64-bit overflows" : "32-bit overflows";
+
+ report_prefix_push(name);
+ report_prefix_push(prefix);
+
+ test(overflow_at_64bits);
+
+ report_prefix_pop();
+ report_prefix_pop();
+}
+
int main(int argc, char *argv[])
{
int cpi = 0;
@@ -1097,33 +1125,19 @@ int main(int argc, char *argv[])
test_event_counter_config();
report_prefix_pop();
} else if (strcmp(argv[1], "pmu-basic-event-count") == 0) {
- report_prefix_push(argv[1]);
- test_basic_event_count();
- report_prefix_pop();
+ run_test(argv[1], test_basic_event_count, false);
} else if (strcmp(argv[1], "pmu-mem-access") == 0) {
- report_prefix_push(argv[1]);
- test_mem_access();
- report_prefix_pop();
+ run_test(argv[1], test_mem_access, false);
} else if (strcmp(argv[1], "pmu-sw-incr") == 0) {
- report_prefix_push(argv[1]);
- test_sw_incr();
- report_prefix_pop();
+ run_test(argv[1], test_sw_incr, false);
} else if (strcmp(argv[1], "pmu-chained-counters") == 0) {
- report_prefix_push(argv[1]);
- test_chained_counters();
- report_prefix_pop();
+ run_test(argv[1], test_chained_counters, false);
} else if (strcmp(argv[1], "pmu-chained-sw-incr") == 0) {
- report_prefix_push(argv[1]);
- test_chained_sw_incr();
- report_prefix_pop();
+ run_test(argv[1], test_chained_sw_incr, false);
} else if (strcmp(argv[1], "pmu-chain-promotion") == 0) {
- report_prefix_push(argv[1]);
- test_chain_promotion();
- report_prefix_pop();
+ run_test(argv[1], test_chain_promotion, false);
} else if (strcmp(argv[1], "pmu-overflow-interrupt") == 0) {
- report_prefix_push(argv[1]);
- test_overflow_interrupt();
- report_prefix_pop();
+ run_test(argv[1], test_overflow_interrupt, false);
} else {
report_abort("Unknown sub-test '%s'", argv[1]);
}
--
2.39.0.314.g84b9a713c41-goog
^ permalink raw reply related [flat|nested] 13+ messages in thread
* [kvm-unit-tests PATCH v2 2/4] arm: pmu: Prepare for testing 64-bit overflows
@ 2022-12-20 3:10 ` Ricardo Koller
0 siblings, 0 replies; 13+ messages in thread
From: Ricardo Koller @ 2022-12-20 3:10 UTC (permalink / raw)
To: kvm, kvmarm, andrew.jones; +Cc: maz
PMUv3p5 adds a knob, PMCR_EL0.LP == 1, that allows overflowing at 64-bits
instead of 32. Prepare by doing these 3 things:
1. Add a "bool overflow_at_64bits" argument to all tests checking
overflows.
2. Extend satisfy_prerequisites() to check if the machine supports
"overflow_at_64bits".
3. Refactor the test invocations to use the new "run_test()" which adds a
report prefix indicating whether the test uses 64 or 32-bit overflows.
A subsequent commit will actually add the 64-bit overflow tests.
Signed-off-by: Ricardo Koller <ricarkol@google.com>
---
arm/pmu.c | 92 ++++++++++++++++++++++++++++++++-----------------------
1 file changed, 53 insertions(+), 39 deletions(-)
diff --git a/arm/pmu.c b/arm/pmu.c
index 1b55e20..4cd3790 100644
--- a/arm/pmu.c
+++ b/arm/pmu.c
@@ -167,13 +167,13 @@ static void pmu_reset(void)
/* event counter tests only implemented for aarch64 */
static void test_event_introspection(void) {}
static void test_event_counter_config(void) {}
-static void test_basic_event_count(void) {}
-static void test_mem_access(void) {}
-static void test_sw_incr(void) {}
-static void test_chained_counters(void) {}
-static void test_chained_sw_incr(void) {}
-static void test_chain_promotion(void) {}
-static void test_overflow_interrupt(void) {}
+static void test_basic_event_count(bool overflow_at_64bits) {}
+static void test_mem_access(bool overflow_at_64bits) {}
+static void test_sw_incr(bool overflow_at_64bits) {}
+static void test_chained_counters(bool unused) {}
+static void test_chained_sw_incr(bool unused) {}
+static void test_chain_promotion(bool unused) {}
+static void test_overflow_interrupt(bool overflow_at_64bits) {}
#elif defined(__aarch64__)
#define ID_AA64DFR0_PERFMON_SHIFT 8
@@ -419,16 +419,28 @@ static bool satisfy_prerequisites(uint32_t *events, unsigned int nb_events)
return false;
}
}
+
+ return true;
+}
+
+static bool check_overflow_prerequisites(bool overflow_at_64bits)
+{
+ if (overflow_at_64bits && pmu.version < ID_DFR0_PMU_V3_8_5) {
+ report_skip("Skip test as 64 overflows need FEAT_PMUv3p5");
+ return false;
+ }
+
return true;
}
-static void test_basic_event_count(void)
+static void test_basic_event_count(bool overflow_at_64bits)
{
uint32_t implemented_counter_mask, non_implemented_counter_mask;
uint32_t counter_mask;
uint32_t events[] = {CPU_CYCLES, INST_RETIRED};
- if (!satisfy_prerequisites(events, ARRAY_SIZE(events)))
+ if (!satisfy_prerequisites(events, ARRAY_SIZE(events)) ||
+ !check_overflow_prerequisites(overflow_at_64bits))
return;
implemented_counter_mask = BIT(pmu.nb_implemented_counters) - 1;
@@ -502,12 +514,13 @@ static void test_basic_event_count(void)
"check overflow happened on #0 only");
}
-static void test_mem_access(void)
+static void test_mem_access(bool overflow_at_64bits)
{
void *addr = malloc(PAGE_SIZE);
uint32_t events[] = {MEM_ACCESS, MEM_ACCESS};
- if (!satisfy_prerequisites(events, ARRAY_SIZE(events)))
+ if (!satisfy_prerequisites(events, ARRAY_SIZE(events)) ||
+ !check_overflow_prerequisites(overflow_at_64bits))
return;
pmu_reset();
@@ -538,13 +551,14 @@ static void test_mem_access(void)
read_sysreg(pmovsclr_el0));
}
-static void test_sw_incr(void)
+static void test_sw_incr(bool overflow_at_64bits)
{
uint32_t events[] = {SW_INCR, SW_INCR};
uint64_t cntr0;
int i;
- if (!satisfy_prerequisites(events, ARRAY_SIZE(events)))
+ if (!satisfy_prerequisites(events, ARRAY_SIZE(events)) ||
+ !check_overflow_prerequisites(overflow_at_64bits))
return;
pmu_reset();
@@ -587,7 +601,7 @@ static void test_sw_incr(void)
"overflow on counter #0 after 100 SW_INCR");
}
-static void test_chained_counters(void)
+static void test_chained_counters(bool unused)
{
uint32_t events[] = {CPU_CYCLES, CHAIN};
uint64_t all_set = ALL_SET_AT(pmu.version >= ID_DFR0_PMU_V3_8_5);
@@ -630,7 +644,7 @@ static void test_chained_counters(void)
report(read_sysreg(pmovsclr_el0) == 0x3, "overflow on even and odd counters");
}
-static void test_chained_sw_incr(void)
+static void test_chained_sw_incr(bool unused)
{
uint32_t events[] = {SW_INCR, CHAIN};
uint64_t cntr0, cntr1;
@@ -686,7 +700,7 @@ static void test_chained_sw_incr(void)
read_regn_el0(pmevcntr, 0), read_regn_el0(pmevcntr, 1));
}
-static void test_chain_promotion(void)
+static void test_chain_promotion(bool unused)
{
uint32_t events[] = {MEM_ACCESS, CHAIN};
void *addr = malloc(PAGE_SIZE);
@@ -835,13 +849,14 @@ static bool expect_interrupts(uint32_t bitmap)
return true;
}
-static void test_overflow_interrupt(void)
+static void test_overflow_interrupt(bool overflow_at_64bits)
{
uint32_t events[] = {MEM_ACCESS, SW_INCR};
void *addr = malloc(PAGE_SIZE);
int i;
- if (!satisfy_prerequisites(events, ARRAY_SIZE(events)))
+ if (!satisfy_prerequisites(events, ARRAY_SIZE(events)) ||
+ !check_overflow_prerequisites(overflow_at_64bits))
return;
gic_enable_defaults();
@@ -1065,6 +1080,19 @@ static bool pmu_probe(void)
return true;
}
+static void run_test(char *name, void (*test)(bool), bool overflow_at_64bits)
+{
+ const char *prefix = overflow_at_64bits ? "64-bit overflows" : "32-bit overflows";
+
+ report_prefix_push(name);
+ report_prefix_push(prefix);
+
+ test(overflow_at_64bits);
+
+ report_prefix_pop();
+ report_prefix_pop();
+}
+
int main(int argc, char *argv[])
{
int cpi = 0;
@@ -1097,33 +1125,19 @@ int main(int argc, char *argv[])
test_event_counter_config();
report_prefix_pop();
} else if (strcmp(argv[1], "pmu-basic-event-count") == 0) {
- report_prefix_push(argv[1]);
- test_basic_event_count();
- report_prefix_pop();
+ run_test(argv[1], test_basic_event_count, false);
} else if (strcmp(argv[1], "pmu-mem-access") == 0) {
- report_prefix_push(argv[1]);
- test_mem_access();
- report_prefix_pop();
+ run_test(argv[1], test_mem_access, false);
} else if (strcmp(argv[1], "pmu-sw-incr") == 0) {
- report_prefix_push(argv[1]);
- test_sw_incr();
- report_prefix_pop();
+ run_test(argv[1], test_sw_incr, false);
} else if (strcmp(argv[1], "pmu-chained-counters") == 0) {
- report_prefix_push(argv[1]);
- test_chained_counters();
- report_prefix_pop();
+ run_test(argv[1], test_chained_counters, false);
} else if (strcmp(argv[1], "pmu-chained-sw-incr") == 0) {
- report_prefix_push(argv[1]);
- test_chained_sw_incr();
- report_prefix_pop();
+ run_test(argv[1], test_chained_sw_incr, false);
} else if (strcmp(argv[1], "pmu-chain-promotion") == 0) {
- report_prefix_push(argv[1]);
- test_chain_promotion();
- report_prefix_pop();
+ run_test(argv[1], test_chain_promotion, false);
} else if (strcmp(argv[1], "pmu-overflow-interrupt") == 0) {
- report_prefix_push(argv[1]);
- test_overflow_interrupt();
- report_prefix_pop();
+ run_test(argv[1], test_overflow_interrupt, false);
} else {
report_abort("Unknown sub-test '%s'", argv[1]);
}
--
2.39.0.314.g84b9a713c41-goog
_______________________________________________
kvmarm mailing list
kvmarm@lists.cs.columbia.edu
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
^ permalink raw reply related [flat|nested] 13+ messages in thread
* [kvm-unit-tests PATCH v2 3/4] arm: pmu: Add tests for 64-bit overflows
2022-12-20 3:10 ` Ricardo Koller
@ 2022-12-20 3:10 ` Ricardo Koller
-1 siblings, 0 replies; 13+ messages in thread
From: Ricardo Koller @ 2022-12-20 3:10 UTC (permalink / raw)
To: kvm, kvmarm, andrew.jones
Cc: maz, alexandru.elisei, eric.auger, oliver.upton, reijiw, Ricardo Koller
Modify all tests checking overflows to support both 32 (PMCR_EL0.LP == 0)
and 64-bit overflows (PMCR_EL0.LP == 1). 64-bit overflows are only
supported on PMUv3p5.
Note that chained tests do not implement "overflow_at_64bits == true".
That's because there are no CHAIN events when "PMCR_EL0.LP == 1" (for more
details see AArch64.IncrementEventCounter() pseudocode in the ARM ARM DDI
0487H.a, J1.1.1 "aarch64/debug").
Signed-off-by: Ricardo Koller <ricarkol@google.com>
---
arm/pmu.c | 97 ++++++++++++++++++++++++++++++++++++-------------------
1 file changed, 64 insertions(+), 33 deletions(-)
diff --git a/arm/pmu.c b/arm/pmu.c
index 4cd3790..680623d 100644
--- a/arm/pmu.c
+++ b/arm/pmu.c
@@ -28,6 +28,7 @@
#define PMU_PMCR_X (1 << 4)
#define PMU_PMCR_DP (1 << 5)
#define PMU_PMCR_LC (1 << 6)
+#define PMU_PMCR_LP (1 << 7)
#define PMU_PMCR_N_SHIFT 11
#define PMU_PMCR_N_MASK 0x1f
#define PMU_PMCR_ID_SHIFT 16
@@ -58,9 +59,11 @@
#define ALL_SET_64 0xFFFFFFFFFFFFFFFFULL
#define ALL_CLEAR 0x0000000000000000ULL
#define PRE_OVERFLOW 0x00000000FFFFFFF0ULL
+#define PRE_OVERFLOW_64 0xFFFFFFFFFFFFFFF0ULL
#define PRE_OVERFLOW2 0x00000000FFFFFFDCULL
-#define ALL_SET_AT(_64b) (_64b ? ALL_SET_64 : ALL_SET)
+#define PRE_OVERFLOW_AT(_64b) (_64b ? PRE_OVERFLOW_64 : PRE_OVERFLOW)
+#define ALL_SET_AT(_64b) (_64b ? ALL_SET_64 : ALL_SET)
#define PMU_PPI 23
@@ -436,8 +439,10 @@ static bool check_overflow_prerequisites(bool overflow_at_64bits)
static void test_basic_event_count(bool overflow_at_64bits)
{
uint32_t implemented_counter_mask, non_implemented_counter_mask;
- uint32_t counter_mask;
+ uint64_t pre_overflow = PRE_OVERFLOW_AT(overflow_at_64bits);
+ uint64_t pmcr_lp = overflow_at_64bits ? PMU_PMCR_LP : 0;
uint32_t events[] = {CPU_CYCLES, INST_RETIRED};
+ uint32_t counter_mask;
if (!satisfy_prerequisites(events, ARRAY_SIZE(events)) ||
!check_overflow_prerequisites(overflow_at_64bits))
@@ -459,13 +464,13 @@ static void test_basic_event_count(bool overflow_at_64bits)
* clear cycle and all event counters and allow counter enablement
* through PMCNTENSET. LC is RES1.
*/
- set_pmcr(pmu.pmcr_ro | PMU_PMCR_LC | PMU_PMCR_C | PMU_PMCR_P);
+ set_pmcr(pmu.pmcr_ro | PMU_PMCR_LC | PMU_PMCR_C | PMU_PMCR_P | pmcr_lp);
isb();
- report(get_pmcr() == (pmu.pmcr_ro | PMU_PMCR_LC), "pmcr: reset counters");
+ report(get_pmcr() == (pmu.pmcr_ro | PMU_PMCR_LC | pmcr_lp), "pmcr: reset counters");
/* Preset counter #0 to pre overflow value to trigger an overflow */
- write_regn_el0(pmevcntr, 0, PRE_OVERFLOW);
- report(read_regn_el0(pmevcntr, 0) == PRE_OVERFLOW,
+ write_regn_el0(pmevcntr, 0, pre_overflow);
+ report(read_regn_el0(pmevcntr, 0) == pre_overflow,
"counter #0 preset to pre-overflow value");
report(!read_regn_el0(pmevcntr, 1), "counter #1 is 0");
@@ -518,6 +523,8 @@ static void test_mem_access(bool overflow_at_64bits)
{
void *addr = malloc(PAGE_SIZE);
uint32_t events[] = {MEM_ACCESS, MEM_ACCESS};
+ uint64_t pre_overflow = PRE_OVERFLOW_AT(overflow_at_64bits);
+ uint64_t pmcr_lp = overflow_at_64bits ? PMU_PMCR_LP : 0;
if (!satisfy_prerequisites(events, ARRAY_SIZE(events)) ||
!check_overflow_prerequisites(overflow_at_64bits))
@@ -529,7 +536,7 @@ static void test_mem_access(bool overflow_at_64bits)
write_regn_el0(pmevtyper, 1, MEM_ACCESS | PMEVTYPER_EXCLUDE_EL0);
write_sysreg_s(0x3, PMCNTENSET_EL0);
isb();
- mem_access_loop(addr, 20, pmu.pmcr_ro | PMU_PMCR_E);
+ mem_access_loop(addr, 20, pmu.pmcr_ro | PMU_PMCR_E | pmcr_lp);
report_info("counter #0 is %ld (MEM_ACCESS)", read_regn_el0(pmevcntr, 0));
report_info("counter #1 is %ld (MEM_ACCESS)", read_regn_el0(pmevcntr, 1));
/* We may measure more than 20 mem access depending on the core */
@@ -539,11 +546,11 @@ static void test_mem_access(bool overflow_at_64bits)
pmu_reset();
- write_regn_el0(pmevcntr, 0, PRE_OVERFLOW);
- write_regn_el0(pmevcntr, 1, PRE_OVERFLOW);
+ write_regn_el0(pmevcntr, 0, pre_overflow);
+ write_regn_el0(pmevcntr, 1, pre_overflow);
write_sysreg_s(0x3, PMCNTENSET_EL0);
isb();
- mem_access_loop(addr, 20, pmu.pmcr_ro | PMU_PMCR_E);
+ mem_access_loop(addr, 20, pmu.pmcr_ro | PMU_PMCR_E | pmcr_lp);
report(read_sysreg(pmovsclr_el0) == 0x3,
"Ran 20 mem accesses with expected overflows on both counters");
report_info("cnt#0 = %ld cnt#1=%ld overflow=0x%lx",
@@ -553,6 +560,8 @@ static void test_mem_access(bool overflow_at_64bits)
static void test_sw_incr(bool overflow_at_64bits)
{
+ uint64_t pre_overflow = PRE_OVERFLOW_AT(overflow_at_64bits);
+ uint64_t pmcr_lp = overflow_at_64bits ? PMU_PMCR_LP : 0;
uint32_t events[] = {SW_INCR, SW_INCR};
uint64_t cntr0;
int i;
@@ -568,7 +577,7 @@ static void test_sw_incr(bool overflow_at_64bits)
/* enable counters #0 and #1 */
write_sysreg_s(0x3, PMCNTENSET_EL0);
- write_regn_el0(pmevcntr, 0, PRE_OVERFLOW);
+ write_regn_el0(pmevcntr, 0, pre_overflow);
isb();
for (i = 0; i < 100; i++)
@@ -576,14 +585,14 @@ static void test_sw_incr(bool overflow_at_64bits)
isb();
report_info("SW_INCR counter #0 has value %ld", read_regn_el0(pmevcntr, 0));
- report(read_regn_el0(pmevcntr, 0) == PRE_OVERFLOW,
+ report(read_regn_el0(pmevcntr, 0) == pre_overflow,
"PWSYNC does not increment if PMCR.E is unset");
pmu_reset();
- write_regn_el0(pmevcntr, 0, PRE_OVERFLOW);
+ write_regn_el0(pmevcntr, 0, pre_overflow);
write_sysreg_s(0x3, PMCNTENSET_EL0);
- set_pmcr(pmu.pmcr_ro | PMU_PMCR_E);
+ set_pmcr(pmu.pmcr_ro | PMU_PMCR_E | pmcr_lp);
isb();
for (i = 0; i < 100; i++)
@@ -591,8 +600,8 @@ static void test_sw_incr(bool overflow_at_64bits)
isb();
cntr0 = (pmu.version < ID_DFR0_PMU_V3_8_5) ?
- (uint32_t)PRE_OVERFLOW + 100 :
- (uint64_t)PRE_OVERFLOW + 100;
+ (uint32_t)pre_overflow + 100 :
+ (uint64_t)pre_overflow + 100;
report(read_regn_el0(pmevcntr, 0) == cntr0, "counter #0 after + 100 SW_INCR");
report(read_regn_el0(pmevcntr, 1) == 100, "counter #1 after + 100 SW_INCR");
report_info("counter values after 100 SW_INCR #0=%ld #1=%ld",
@@ -851,6 +860,9 @@ static bool expect_interrupts(uint32_t bitmap)
static void test_overflow_interrupt(bool overflow_at_64bits)
{
+ uint64_t pre_overflow = PRE_OVERFLOW_AT(overflow_at_64bits);
+ uint64_t all_set = ALL_SET_AT(overflow_at_64bits);
+ uint64_t pmcr_lp = overflow_at_64bits ? PMU_PMCR_LP : 0;
uint32_t events[] = {MEM_ACCESS, SW_INCR};
void *addr = malloc(PAGE_SIZE);
int i;
@@ -869,16 +881,16 @@ static void test_overflow_interrupt(bool overflow_at_64bits)
write_regn_el0(pmevtyper, 0, MEM_ACCESS | PMEVTYPER_EXCLUDE_EL0);
write_regn_el0(pmevtyper, 1, SW_INCR | PMEVTYPER_EXCLUDE_EL0);
write_sysreg_s(0x3, PMCNTENSET_EL0);
- write_regn_el0(pmevcntr, 0, PRE_OVERFLOW);
- write_regn_el0(pmevcntr, 1, PRE_OVERFLOW);
+ write_regn_el0(pmevcntr, 0, pre_overflow);
+ write_regn_el0(pmevcntr, 1, pre_overflow);
isb();
/* interrupts are disabled (PMINTENSET_EL1 == 0) */
- mem_access_loop(addr, 200, pmu.pmcr_ro | PMU_PMCR_E);
+ mem_access_loop(addr, 20, pmu.pmcr_ro | PMU_PMCR_E | pmcr_lp);
report(expect_interrupts(0), "no overflow interrupt after preset");
- set_pmcr(pmu.pmcr_ro | PMU_PMCR_E);
+ set_pmcr(pmu.pmcr_ro | PMU_PMCR_E | pmcr_lp);
isb();
for (i = 0; i < 100; i++)
@@ -893,12 +905,12 @@ static void test_overflow_interrupt(bool overflow_at_64bits)
pmu_reset_stats();
- write_regn_el0(pmevcntr, 0, PRE_OVERFLOW);
- write_regn_el0(pmevcntr, 1, PRE_OVERFLOW);
+ write_regn_el0(pmevcntr, 0, pre_overflow);
+ write_regn_el0(pmevcntr, 1, pre_overflow);
write_sysreg(ALL_SET, pmintenset_el1);
isb();
- mem_access_loop(addr, 200, pmu.pmcr_ro | PMU_PMCR_E);
+ mem_access_loop(addr, 200, pmu.pmcr_ro | PMU_PMCR_E | pmcr_lp);
for (i = 0; i < 100; i++)
write_sysreg(0x3, pmswinc_el0);
@@ -907,25 +919,40 @@ static void test_overflow_interrupt(bool overflow_at_64bits)
report(expect_interrupts(0x3),
"overflow interrupts expected on #0 and #1");
- /* promote to 64-b */
+ /*
+ * promote to 64-b:
+ *
+ * This only applies to the !overflow_at_64bits case, as
+ * overflow_at_64bits doesn't implement CHAIN events. The
+ * overflow_at_64bits case just checks that chained counters are
+ * not incremented when PMCR.LP == 1.
+ */
pmu_reset_stats();
write_regn_el0(pmevtyper, 1, CHAIN | PMEVTYPER_EXCLUDE_EL0);
- write_regn_el0(pmevcntr, 0, PRE_OVERFLOW);
+ write_regn_el0(pmevcntr, 0, pre_overflow);
isb();
- mem_access_loop(addr, 200, pmu.pmcr_ro | PMU_PMCR_E);
- report(expect_interrupts(0x1),
- "expect overflow interrupt on 32b boundary");
+ mem_access_loop(addr, 200, pmu.pmcr_ro | PMU_PMCR_E | pmcr_lp);
+ report(expect_interrupts(0x1), "expect overflow interrupt");
/* overflow on odd counter */
pmu_reset_stats();
- write_regn_el0(pmevcntr, 0, PRE_OVERFLOW);
- write_regn_el0(pmevcntr, 1, ALL_SET);
+ write_regn_el0(pmevcntr, 0, pre_overflow);
+ write_regn_el0(pmevcntr, 1, all_set);
isb();
- mem_access_loop(addr, 400, pmu.pmcr_ro | PMU_PMCR_E);
- report(expect_interrupts(0x3),
- "expect overflow interrupt on even and odd counter");
+ mem_access_loop(addr, 400, pmu.pmcr_ro | PMU_PMCR_E | pmcr_lp);
+ if (overflow_at_64bits) {
+ report(expect_interrupts(0x1),
+ "expect overflow interrupt on even counter");
+ report(read_regn_el0(pmevcntr, 1) == all_set,
+ "Odd counter did not change");
+ } else {
+ report(expect_interrupts(0x3),
+ "expect overflow interrupt on even and odd counter");
+ report(read_regn_el0(pmevcntr, 1) != all_set,
+ "Odd counter wrapped");
+ }
}
#endif
@@ -1126,10 +1153,13 @@ int main(int argc, char *argv[])
report_prefix_pop();
} else if (strcmp(argv[1], "pmu-basic-event-count") == 0) {
run_test(argv[1], test_basic_event_count, false);
+ run_test(argv[1], test_basic_event_count, true);
} else if (strcmp(argv[1], "pmu-mem-access") == 0) {
run_test(argv[1], test_mem_access, false);
+ run_test(argv[1], test_mem_access, true);
} else if (strcmp(argv[1], "pmu-sw-incr") == 0) {
run_test(argv[1], test_sw_incr, false);
+ run_test(argv[1], test_sw_incr, true);
} else if (strcmp(argv[1], "pmu-chained-counters") == 0) {
run_test(argv[1], test_chained_counters, false);
} else if (strcmp(argv[1], "pmu-chained-sw-incr") == 0) {
@@ -1138,6 +1168,7 @@ int main(int argc, char *argv[])
run_test(argv[1], test_chain_promotion, false);
} else if (strcmp(argv[1], "pmu-overflow-interrupt") == 0) {
run_test(argv[1], test_overflow_interrupt, false);
+ run_test(argv[1], test_overflow_interrupt, true);
} else {
report_abort("Unknown sub-test '%s'", argv[1]);
}
--
2.39.0.314.g84b9a713c41-goog
^ permalink raw reply related [flat|nested] 13+ messages in thread
* [kvm-unit-tests PATCH v2 3/4] arm: pmu: Add tests for 64-bit overflows
@ 2022-12-20 3:10 ` Ricardo Koller
0 siblings, 0 replies; 13+ messages in thread
From: Ricardo Koller @ 2022-12-20 3:10 UTC (permalink / raw)
To: kvm, kvmarm, andrew.jones; +Cc: maz
Modify all tests checking overflows to support both 32 (PMCR_EL0.LP == 0)
and 64-bit overflows (PMCR_EL0.LP == 1). 64-bit overflows are only
supported on PMUv3p5.
Note that chained tests do not implement "overflow_at_64bits == true".
That's because there are no CHAIN events when "PMCR_EL0.LP == 1" (for more
details see AArch64.IncrementEventCounter() pseudocode in the ARM ARM DDI
0487H.a, J1.1.1 "aarch64/debug").
Signed-off-by: Ricardo Koller <ricarkol@google.com>
---
arm/pmu.c | 97 ++++++++++++++++++++++++++++++++++++-------------------
1 file changed, 64 insertions(+), 33 deletions(-)
diff --git a/arm/pmu.c b/arm/pmu.c
index 4cd3790..680623d 100644
--- a/arm/pmu.c
+++ b/arm/pmu.c
@@ -28,6 +28,7 @@
#define PMU_PMCR_X (1 << 4)
#define PMU_PMCR_DP (1 << 5)
#define PMU_PMCR_LC (1 << 6)
+#define PMU_PMCR_LP (1 << 7)
#define PMU_PMCR_N_SHIFT 11
#define PMU_PMCR_N_MASK 0x1f
#define PMU_PMCR_ID_SHIFT 16
@@ -58,9 +59,11 @@
#define ALL_SET_64 0xFFFFFFFFFFFFFFFFULL
#define ALL_CLEAR 0x0000000000000000ULL
#define PRE_OVERFLOW 0x00000000FFFFFFF0ULL
+#define PRE_OVERFLOW_64 0xFFFFFFFFFFFFFFF0ULL
#define PRE_OVERFLOW2 0x00000000FFFFFFDCULL
-#define ALL_SET_AT(_64b) (_64b ? ALL_SET_64 : ALL_SET)
+#define PRE_OVERFLOW_AT(_64b) (_64b ? PRE_OVERFLOW_64 : PRE_OVERFLOW)
+#define ALL_SET_AT(_64b) (_64b ? ALL_SET_64 : ALL_SET)
#define PMU_PPI 23
@@ -436,8 +439,10 @@ static bool check_overflow_prerequisites(bool overflow_at_64bits)
static void test_basic_event_count(bool overflow_at_64bits)
{
uint32_t implemented_counter_mask, non_implemented_counter_mask;
- uint32_t counter_mask;
+ uint64_t pre_overflow = PRE_OVERFLOW_AT(overflow_at_64bits);
+ uint64_t pmcr_lp = overflow_at_64bits ? PMU_PMCR_LP : 0;
uint32_t events[] = {CPU_CYCLES, INST_RETIRED};
+ uint32_t counter_mask;
if (!satisfy_prerequisites(events, ARRAY_SIZE(events)) ||
!check_overflow_prerequisites(overflow_at_64bits))
@@ -459,13 +464,13 @@ static void test_basic_event_count(bool overflow_at_64bits)
* clear cycle and all event counters and allow counter enablement
* through PMCNTENSET. LC is RES1.
*/
- set_pmcr(pmu.pmcr_ro | PMU_PMCR_LC | PMU_PMCR_C | PMU_PMCR_P);
+ set_pmcr(pmu.pmcr_ro | PMU_PMCR_LC | PMU_PMCR_C | PMU_PMCR_P | pmcr_lp);
isb();
- report(get_pmcr() == (pmu.pmcr_ro | PMU_PMCR_LC), "pmcr: reset counters");
+ report(get_pmcr() == (pmu.pmcr_ro | PMU_PMCR_LC | pmcr_lp), "pmcr: reset counters");
/* Preset counter #0 to pre overflow value to trigger an overflow */
- write_regn_el0(pmevcntr, 0, PRE_OVERFLOW);
- report(read_regn_el0(pmevcntr, 0) == PRE_OVERFLOW,
+ write_regn_el0(pmevcntr, 0, pre_overflow);
+ report(read_regn_el0(pmevcntr, 0) == pre_overflow,
"counter #0 preset to pre-overflow value");
report(!read_regn_el0(pmevcntr, 1), "counter #1 is 0");
@@ -518,6 +523,8 @@ static void test_mem_access(bool overflow_at_64bits)
{
void *addr = malloc(PAGE_SIZE);
uint32_t events[] = {MEM_ACCESS, MEM_ACCESS};
+ uint64_t pre_overflow = PRE_OVERFLOW_AT(overflow_at_64bits);
+ uint64_t pmcr_lp = overflow_at_64bits ? PMU_PMCR_LP : 0;
if (!satisfy_prerequisites(events, ARRAY_SIZE(events)) ||
!check_overflow_prerequisites(overflow_at_64bits))
@@ -529,7 +536,7 @@ static void test_mem_access(bool overflow_at_64bits)
write_regn_el0(pmevtyper, 1, MEM_ACCESS | PMEVTYPER_EXCLUDE_EL0);
write_sysreg_s(0x3, PMCNTENSET_EL0);
isb();
- mem_access_loop(addr, 20, pmu.pmcr_ro | PMU_PMCR_E);
+ mem_access_loop(addr, 20, pmu.pmcr_ro | PMU_PMCR_E | pmcr_lp);
report_info("counter #0 is %ld (MEM_ACCESS)", read_regn_el0(pmevcntr, 0));
report_info("counter #1 is %ld (MEM_ACCESS)", read_regn_el0(pmevcntr, 1));
/* We may measure more than 20 mem access depending on the core */
@@ -539,11 +546,11 @@ static void test_mem_access(bool overflow_at_64bits)
pmu_reset();
- write_regn_el0(pmevcntr, 0, PRE_OVERFLOW);
- write_regn_el0(pmevcntr, 1, PRE_OVERFLOW);
+ write_regn_el0(pmevcntr, 0, pre_overflow);
+ write_regn_el0(pmevcntr, 1, pre_overflow);
write_sysreg_s(0x3, PMCNTENSET_EL0);
isb();
- mem_access_loop(addr, 20, pmu.pmcr_ro | PMU_PMCR_E);
+ mem_access_loop(addr, 20, pmu.pmcr_ro | PMU_PMCR_E | pmcr_lp);
report(read_sysreg(pmovsclr_el0) == 0x3,
"Ran 20 mem accesses with expected overflows on both counters");
report_info("cnt#0 = %ld cnt#1=%ld overflow=0x%lx",
@@ -553,6 +560,8 @@ static void test_mem_access(bool overflow_at_64bits)
static void test_sw_incr(bool overflow_at_64bits)
{
+ uint64_t pre_overflow = PRE_OVERFLOW_AT(overflow_at_64bits);
+ uint64_t pmcr_lp = overflow_at_64bits ? PMU_PMCR_LP : 0;
uint32_t events[] = {SW_INCR, SW_INCR};
uint64_t cntr0;
int i;
@@ -568,7 +577,7 @@ static void test_sw_incr(bool overflow_at_64bits)
/* enable counters #0 and #1 */
write_sysreg_s(0x3, PMCNTENSET_EL0);
- write_regn_el0(pmevcntr, 0, PRE_OVERFLOW);
+ write_regn_el0(pmevcntr, 0, pre_overflow);
isb();
for (i = 0; i < 100; i++)
@@ -576,14 +585,14 @@ static void test_sw_incr(bool overflow_at_64bits)
isb();
report_info("SW_INCR counter #0 has value %ld", read_regn_el0(pmevcntr, 0));
- report(read_regn_el0(pmevcntr, 0) == PRE_OVERFLOW,
+ report(read_regn_el0(pmevcntr, 0) == pre_overflow,
"PWSYNC does not increment if PMCR.E is unset");
pmu_reset();
- write_regn_el0(pmevcntr, 0, PRE_OVERFLOW);
+ write_regn_el0(pmevcntr, 0, pre_overflow);
write_sysreg_s(0x3, PMCNTENSET_EL0);
- set_pmcr(pmu.pmcr_ro | PMU_PMCR_E);
+ set_pmcr(pmu.pmcr_ro | PMU_PMCR_E | pmcr_lp);
isb();
for (i = 0; i < 100; i++)
@@ -591,8 +600,8 @@ static void test_sw_incr(bool overflow_at_64bits)
isb();
cntr0 = (pmu.version < ID_DFR0_PMU_V3_8_5) ?
- (uint32_t)PRE_OVERFLOW + 100 :
- (uint64_t)PRE_OVERFLOW + 100;
+ (uint32_t)pre_overflow + 100 :
+ (uint64_t)pre_overflow + 100;
report(read_regn_el0(pmevcntr, 0) == cntr0, "counter #0 after + 100 SW_INCR");
report(read_regn_el0(pmevcntr, 1) == 100, "counter #1 after + 100 SW_INCR");
report_info("counter values after 100 SW_INCR #0=%ld #1=%ld",
@@ -851,6 +860,9 @@ static bool expect_interrupts(uint32_t bitmap)
static void test_overflow_interrupt(bool overflow_at_64bits)
{
+ uint64_t pre_overflow = PRE_OVERFLOW_AT(overflow_at_64bits);
+ uint64_t all_set = ALL_SET_AT(overflow_at_64bits);
+ uint64_t pmcr_lp = overflow_at_64bits ? PMU_PMCR_LP : 0;
uint32_t events[] = {MEM_ACCESS, SW_INCR};
void *addr = malloc(PAGE_SIZE);
int i;
@@ -869,16 +881,16 @@ static void test_overflow_interrupt(bool overflow_at_64bits)
write_regn_el0(pmevtyper, 0, MEM_ACCESS | PMEVTYPER_EXCLUDE_EL0);
write_regn_el0(pmevtyper, 1, SW_INCR | PMEVTYPER_EXCLUDE_EL0);
write_sysreg_s(0x3, PMCNTENSET_EL0);
- write_regn_el0(pmevcntr, 0, PRE_OVERFLOW);
- write_regn_el0(pmevcntr, 1, PRE_OVERFLOW);
+ write_regn_el0(pmevcntr, 0, pre_overflow);
+ write_regn_el0(pmevcntr, 1, pre_overflow);
isb();
/* interrupts are disabled (PMINTENSET_EL1 == 0) */
- mem_access_loop(addr, 200, pmu.pmcr_ro | PMU_PMCR_E);
+ mem_access_loop(addr, 20, pmu.pmcr_ro | PMU_PMCR_E | pmcr_lp);
report(expect_interrupts(0), "no overflow interrupt after preset");
- set_pmcr(pmu.pmcr_ro | PMU_PMCR_E);
+ set_pmcr(pmu.pmcr_ro | PMU_PMCR_E | pmcr_lp);
isb();
for (i = 0; i < 100; i++)
@@ -893,12 +905,12 @@ static void test_overflow_interrupt(bool overflow_at_64bits)
pmu_reset_stats();
- write_regn_el0(pmevcntr, 0, PRE_OVERFLOW);
- write_regn_el0(pmevcntr, 1, PRE_OVERFLOW);
+ write_regn_el0(pmevcntr, 0, pre_overflow);
+ write_regn_el0(pmevcntr, 1, pre_overflow);
write_sysreg(ALL_SET, pmintenset_el1);
isb();
- mem_access_loop(addr, 200, pmu.pmcr_ro | PMU_PMCR_E);
+ mem_access_loop(addr, 200, pmu.pmcr_ro | PMU_PMCR_E | pmcr_lp);
for (i = 0; i < 100; i++)
write_sysreg(0x3, pmswinc_el0);
@@ -907,25 +919,40 @@ static void test_overflow_interrupt(bool overflow_at_64bits)
report(expect_interrupts(0x3),
"overflow interrupts expected on #0 and #1");
- /* promote to 64-b */
+ /*
+ * promote to 64-b:
+ *
+ * This only applies to the !overflow_at_64bits case, as
+ * overflow_at_64bits doesn't implement CHAIN events. The
+ * overflow_at_64bits case just checks that chained counters are
+ * not incremented when PMCR.LP == 1.
+ */
pmu_reset_stats();
write_regn_el0(pmevtyper, 1, CHAIN | PMEVTYPER_EXCLUDE_EL0);
- write_regn_el0(pmevcntr, 0, PRE_OVERFLOW);
+ write_regn_el0(pmevcntr, 0, pre_overflow);
isb();
- mem_access_loop(addr, 200, pmu.pmcr_ro | PMU_PMCR_E);
- report(expect_interrupts(0x1),
- "expect overflow interrupt on 32b boundary");
+ mem_access_loop(addr, 200, pmu.pmcr_ro | PMU_PMCR_E | pmcr_lp);
+ report(expect_interrupts(0x1), "expect overflow interrupt");
/* overflow on odd counter */
pmu_reset_stats();
- write_regn_el0(pmevcntr, 0, PRE_OVERFLOW);
- write_regn_el0(pmevcntr, 1, ALL_SET);
+ write_regn_el0(pmevcntr, 0, pre_overflow);
+ write_regn_el0(pmevcntr, 1, all_set);
isb();
- mem_access_loop(addr, 400, pmu.pmcr_ro | PMU_PMCR_E);
- report(expect_interrupts(0x3),
- "expect overflow interrupt on even and odd counter");
+ mem_access_loop(addr, 400, pmu.pmcr_ro | PMU_PMCR_E | pmcr_lp);
+ if (overflow_at_64bits) {
+ report(expect_interrupts(0x1),
+ "expect overflow interrupt on even counter");
+ report(read_regn_el0(pmevcntr, 1) == all_set,
+ "Odd counter did not change");
+ } else {
+ report(expect_interrupts(0x3),
+ "expect overflow interrupt on even and odd counter");
+ report(read_regn_el0(pmevcntr, 1) != all_set,
+ "Odd counter wrapped");
+ }
}
#endif
@@ -1126,10 +1153,13 @@ int main(int argc, char *argv[])
report_prefix_pop();
} else if (strcmp(argv[1], "pmu-basic-event-count") == 0) {
run_test(argv[1], test_basic_event_count, false);
+ run_test(argv[1], test_basic_event_count, true);
} else if (strcmp(argv[1], "pmu-mem-access") == 0) {
run_test(argv[1], test_mem_access, false);
+ run_test(argv[1], test_mem_access, true);
} else if (strcmp(argv[1], "pmu-sw-incr") == 0) {
run_test(argv[1], test_sw_incr, false);
+ run_test(argv[1], test_sw_incr, true);
} else if (strcmp(argv[1], "pmu-chained-counters") == 0) {
run_test(argv[1], test_chained_counters, false);
} else if (strcmp(argv[1], "pmu-chained-sw-incr") == 0) {
@@ -1138,6 +1168,7 @@ int main(int argc, char *argv[])
run_test(argv[1], test_chain_promotion, false);
} else if (strcmp(argv[1], "pmu-overflow-interrupt") == 0) {
run_test(argv[1], test_overflow_interrupt, false);
+ run_test(argv[1], test_overflow_interrupt, true);
} else {
report_abort("Unknown sub-test '%s'", argv[1]);
}
--
2.39.0.314.g84b9a713c41-goog
_______________________________________________
kvmarm mailing list
kvmarm@lists.cs.columbia.edu
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
^ permalink raw reply related [flat|nested] 13+ messages in thread
* [kvm-unit-tests PATCH v2 4/4] arm: pmu: Print counter values as hexadecimals
2022-12-20 3:10 ` Ricardo Koller
@ 2022-12-20 3:10 ` Ricardo Koller
-1 siblings, 0 replies; 13+ messages in thread
From: Ricardo Koller @ 2022-12-20 3:10 UTC (permalink / raw)
To: kvm, kvmarm, andrew.jones
Cc: maz, alexandru.elisei, eric.auger, oliver.upton, reijiw, Ricardo Koller
The arm/pmu test prints the value of counters as %ld. Most tests start
with counters around 0 or UINT_MAX, so having something like -16 instead of
0xffff_fff0 is not very useful.
Report counter values as hexadecimals.
Reported-by: Alexandru Elisei <alexandru.elisei@arm.com>
Signed-off-by: Ricardo Koller <ricarkol@google.com>
---
arm/pmu.c | 20 ++++++++++----------
1 file changed, 10 insertions(+), 10 deletions(-)
diff --git a/arm/pmu.c b/arm/pmu.c
index 680623d..b6b2871 100644
--- a/arm/pmu.c
+++ b/arm/pmu.c
@@ -537,8 +537,8 @@ static void test_mem_access(bool overflow_at_64bits)
write_sysreg_s(0x3, PMCNTENSET_EL0);
isb();
mem_access_loop(addr, 20, pmu.pmcr_ro | PMU_PMCR_E | pmcr_lp);
- report_info("counter #0 is %ld (MEM_ACCESS)", read_regn_el0(pmevcntr, 0));
- report_info("counter #1 is %ld (MEM_ACCESS)", read_regn_el0(pmevcntr, 1));
+ report_info("counter #0 is 0x%lx (MEM_ACCESS)", read_regn_el0(pmevcntr, 0));
+ report_info("counter #1 is 0x%lx (MEM_ACCESS)", read_regn_el0(pmevcntr, 1));
/* We may measure more than 20 mem access depending on the core */
report((read_regn_el0(pmevcntr, 0) == read_regn_el0(pmevcntr, 1)) &&
(read_regn_el0(pmevcntr, 0) >= 20) && !read_sysreg(pmovsclr_el0),
@@ -553,7 +553,7 @@ static void test_mem_access(bool overflow_at_64bits)
mem_access_loop(addr, 20, pmu.pmcr_ro | PMU_PMCR_E | pmcr_lp);
report(read_sysreg(pmovsclr_el0) == 0x3,
"Ran 20 mem accesses with expected overflows on both counters");
- report_info("cnt#0 = %ld cnt#1=%ld overflow=0x%lx",
+ report_info("cnt#0=0x%lx cnt#1=0x%lx overflow=0x%lx",
read_regn_el0(pmevcntr, 0), read_regn_el0(pmevcntr, 1),
read_sysreg(pmovsclr_el0));
}
@@ -584,7 +584,7 @@ static void test_sw_incr(bool overflow_at_64bits)
write_sysreg(0x1, pmswinc_el0);
isb();
- report_info("SW_INCR counter #0 has value %ld", read_regn_el0(pmevcntr, 0));
+ report_info("SW_INCR counter #0 has value 0x%lx", read_regn_el0(pmevcntr, 0));
report(read_regn_el0(pmevcntr, 0) == pre_overflow,
"PWSYNC does not increment if PMCR.E is unset");
@@ -604,7 +604,7 @@ static void test_sw_incr(bool overflow_at_64bits)
(uint64_t)pre_overflow + 100;
report(read_regn_el0(pmevcntr, 0) == cntr0, "counter #0 after + 100 SW_INCR");
report(read_regn_el0(pmevcntr, 1) == 100, "counter #1 after + 100 SW_INCR");
- report_info("counter values after 100 SW_INCR #0=%ld #1=%ld",
+ report_info("counter values after 100 SW_INCR #0=0x%lx #1=0x%lx",
read_regn_el0(pmevcntr, 0), read_regn_el0(pmevcntr, 1));
report(read_sysreg(pmovsclr_el0) == 0x1,
"overflow on counter #0 after 100 SW_INCR");
@@ -680,7 +680,7 @@ static void test_chained_sw_incr(bool unused)
report((read_sysreg(pmovsclr_el0) == 0x1) &&
(read_regn_el0(pmevcntr, 1) == 1),
"overflow and chain counter incremented after 100 SW_INCR/CHAIN");
- report_info("overflow=0x%lx, #0=%ld #1=%ld", read_sysreg(pmovsclr_el0),
+ report_info("overflow=0x%lx, #0=0x%lx #1=0x%lx", read_sysreg(pmovsclr_el0),
read_regn_el0(pmevcntr, 0), read_regn_el0(pmevcntr, 1));
/* 64b SW_INCR and overflow on CHAIN counter*/
@@ -705,7 +705,7 @@ static void test_chained_sw_incr(bool unused)
(read_regn_el0(pmevcntr, 0) == cntr0) &&
(read_regn_el0(pmevcntr, 1) == cntr1),
"expected overflows and values after 100 SW_INCR/CHAIN");
- report_info("overflow=0x%lx, #0=%ld #1=%ld", read_sysreg(pmovsclr_el0),
+ report_info("overflow=0x%lx, #0=0x%lx #1=0x%lx", read_sysreg(pmovsclr_el0),
read_regn_el0(pmevcntr, 0), read_regn_el0(pmevcntr, 1));
}
@@ -737,11 +737,11 @@ static void test_chain_promotion(bool unused)
mem_access_loop(addr, 20, pmu.pmcr_ro | PMU_PMCR_E);
report(!read_regn_el0(pmevcntr, 1) && (read_sysreg(pmovsclr_el0) == 0x1),
"odd counter did not increment on overflow if disabled");
- report_info("MEM_ACCESS counter #0 has value %ld",
+ report_info("MEM_ACCESS counter #0 has value 0x%lx",
read_regn_el0(pmevcntr, 0));
- report_info("CHAIN counter #1 has value %ld",
+ report_info("CHAIN counter #1 has value 0x%lx",
read_regn_el0(pmevcntr, 1));
- report_info("overflow counter %ld", read_sysreg(pmovsclr_el0));
+ report_info("overflow counter 0x%lx", read_sysreg(pmovsclr_el0));
/* start at 0xFFFFFFDC, +20 with CHAIN enabled, +20 with CHAIN disabled */
pmu_reset();
--
2.39.0.314.g84b9a713c41-goog
^ permalink raw reply related [flat|nested] 13+ messages in thread
* [kvm-unit-tests PATCH v2 4/4] arm: pmu: Print counter values as hexadecimals
@ 2022-12-20 3:10 ` Ricardo Koller
0 siblings, 0 replies; 13+ messages in thread
From: Ricardo Koller @ 2022-12-20 3:10 UTC (permalink / raw)
To: kvm, kvmarm, andrew.jones; +Cc: maz
The arm/pmu test prints the value of counters as %ld. Most tests start
with counters around 0 or UINT_MAX, so having something like -16 instead of
0xffff_fff0 is not very useful.
Report counter values as hexadecimals.
Reported-by: Alexandru Elisei <alexandru.elisei@arm.com>
Signed-off-by: Ricardo Koller <ricarkol@google.com>
---
arm/pmu.c | 20 ++++++++++----------
1 file changed, 10 insertions(+), 10 deletions(-)
diff --git a/arm/pmu.c b/arm/pmu.c
index 680623d..b6b2871 100644
--- a/arm/pmu.c
+++ b/arm/pmu.c
@@ -537,8 +537,8 @@ static void test_mem_access(bool overflow_at_64bits)
write_sysreg_s(0x3, PMCNTENSET_EL0);
isb();
mem_access_loop(addr, 20, pmu.pmcr_ro | PMU_PMCR_E | pmcr_lp);
- report_info("counter #0 is %ld (MEM_ACCESS)", read_regn_el0(pmevcntr, 0));
- report_info("counter #1 is %ld (MEM_ACCESS)", read_regn_el0(pmevcntr, 1));
+ report_info("counter #0 is 0x%lx (MEM_ACCESS)", read_regn_el0(pmevcntr, 0));
+ report_info("counter #1 is 0x%lx (MEM_ACCESS)", read_regn_el0(pmevcntr, 1));
/* We may measure more than 20 mem access depending on the core */
report((read_regn_el0(pmevcntr, 0) == read_regn_el0(pmevcntr, 1)) &&
(read_regn_el0(pmevcntr, 0) >= 20) && !read_sysreg(pmovsclr_el0),
@@ -553,7 +553,7 @@ static void test_mem_access(bool overflow_at_64bits)
mem_access_loop(addr, 20, pmu.pmcr_ro | PMU_PMCR_E | pmcr_lp);
report(read_sysreg(pmovsclr_el0) == 0x3,
"Ran 20 mem accesses with expected overflows on both counters");
- report_info("cnt#0 = %ld cnt#1=%ld overflow=0x%lx",
+ report_info("cnt#0=0x%lx cnt#1=0x%lx overflow=0x%lx",
read_regn_el0(pmevcntr, 0), read_regn_el0(pmevcntr, 1),
read_sysreg(pmovsclr_el0));
}
@@ -584,7 +584,7 @@ static void test_sw_incr(bool overflow_at_64bits)
write_sysreg(0x1, pmswinc_el0);
isb();
- report_info("SW_INCR counter #0 has value %ld", read_regn_el0(pmevcntr, 0));
+ report_info("SW_INCR counter #0 has value 0x%lx", read_regn_el0(pmevcntr, 0));
report(read_regn_el0(pmevcntr, 0) == pre_overflow,
"PWSYNC does not increment if PMCR.E is unset");
@@ -604,7 +604,7 @@ static void test_sw_incr(bool overflow_at_64bits)
(uint64_t)pre_overflow + 100;
report(read_regn_el0(pmevcntr, 0) == cntr0, "counter #0 after + 100 SW_INCR");
report(read_regn_el0(pmevcntr, 1) == 100, "counter #1 after + 100 SW_INCR");
- report_info("counter values after 100 SW_INCR #0=%ld #1=%ld",
+ report_info("counter values after 100 SW_INCR #0=0x%lx #1=0x%lx",
read_regn_el0(pmevcntr, 0), read_regn_el0(pmevcntr, 1));
report(read_sysreg(pmovsclr_el0) == 0x1,
"overflow on counter #0 after 100 SW_INCR");
@@ -680,7 +680,7 @@ static void test_chained_sw_incr(bool unused)
report((read_sysreg(pmovsclr_el0) == 0x1) &&
(read_regn_el0(pmevcntr, 1) == 1),
"overflow and chain counter incremented after 100 SW_INCR/CHAIN");
- report_info("overflow=0x%lx, #0=%ld #1=%ld", read_sysreg(pmovsclr_el0),
+ report_info("overflow=0x%lx, #0=0x%lx #1=0x%lx", read_sysreg(pmovsclr_el0),
read_regn_el0(pmevcntr, 0), read_regn_el0(pmevcntr, 1));
/* 64b SW_INCR and overflow on CHAIN counter*/
@@ -705,7 +705,7 @@ static void test_chained_sw_incr(bool unused)
(read_regn_el0(pmevcntr, 0) == cntr0) &&
(read_regn_el0(pmevcntr, 1) == cntr1),
"expected overflows and values after 100 SW_INCR/CHAIN");
- report_info("overflow=0x%lx, #0=%ld #1=%ld", read_sysreg(pmovsclr_el0),
+ report_info("overflow=0x%lx, #0=0x%lx #1=0x%lx", read_sysreg(pmovsclr_el0),
read_regn_el0(pmevcntr, 0), read_regn_el0(pmevcntr, 1));
}
@@ -737,11 +737,11 @@ static void test_chain_promotion(bool unused)
mem_access_loop(addr, 20, pmu.pmcr_ro | PMU_PMCR_E);
report(!read_regn_el0(pmevcntr, 1) && (read_sysreg(pmovsclr_el0) == 0x1),
"odd counter did not increment on overflow if disabled");
- report_info("MEM_ACCESS counter #0 has value %ld",
+ report_info("MEM_ACCESS counter #0 has value 0x%lx",
read_regn_el0(pmevcntr, 0));
- report_info("CHAIN counter #1 has value %ld",
+ report_info("CHAIN counter #1 has value 0x%lx",
read_regn_el0(pmevcntr, 1));
- report_info("overflow counter %ld", read_sysreg(pmovsclr_el0));
+ report_info("overflow counter 0x%lx", read_sysreg(pmovsclr_el0));
/* start at 0xFFFFFFDC, +20 with CHAIN enabled, +20 with CHAIN disabled */
pmu_reset();
--
2.39.0.314.g84b9a713c41-goog
_______________________________________________
kvmarm mailing list
kvmarm@lists.cs.columbia.edu
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
^ permalink raw reply related [flat|nested] 13+ messages in thread
* Re: [kvm-unit-tests PATCH v2 1/4] arm: pmu: Fix overflow checks for PMUv3p5 long counters
2022-12-20 3:10 ` Ricardo Koller
(?)
@ 2023-01-06 19:28 ` Oliver Upton
2023-01-09 16:10 ` Ricardo Koller
2023-01-09 20:32 ` Ricardo Koller
-1 siblings, 2 replies; 13+ messages in thread
From: Oliver Upton @ 2023-01-06 19:28 UTC (permalink / raw)
To: Ricardo Koller
Cc: kvm, kvmarm, andrew.jones, maz, alexandru.elisei, eric.auger, reijiw
Hi Ricardo,
On Tue, Dec 20, 2022 at 03:10:29AM +0000, Ricardo Koller wrote:
> PMUv3p5 uses 64-bit counters irrespective of whether the PMU is configured
> for overflowing at 32 or 64-bits. The consequence is that tests that check
> the counter values after overflowing should not assume that values will be
> wrapped around 32-bits: they overflow into the other half of the 64-bit
> counters on PMUv3p5.
>
> Fix tests by correctly checking overflowing-counters against the expected
> 64-bit value.
>
> Signed-off-by: Ricardo Koller <ricarkol@google.com>
> ---
> arm/pmu.c | 37 +++++++++++++++++++++++++------------
> 1 file changed, 25 insertions(+), 12 deletions(-)
>
> diff --git a/arm/pmu.c b/arm/pmu.c
> index cd47b14..1b55e20 100644
> --- a/arm/pmu.c
> +++ b/arm/pmu.c
> @@ -54,10 +54,13 @@
> #define EXT_COMMON_EVENTS_LOW 0x4000
> #define EXT_COMMON_EVENTS_HIGH 0x403F
>
> -#define ALL_SET 0xFFFFFFFF
> -#define ALL_CLEAR 0x0
> -#define PRE_OVERFLOW 0xFFFFFFF0
> -#define PRE_OVERFLOW2 0xFFFFFFDC
> +#define ALL_SET 0x00000000FFFFFFFFULL
> +#define ALL_SET_64 0xFFFFFFFFFFFFFFFFULL
> +#define ALL_CLEAR 0x0000000000000000ULL
> +#define PRE_OVERFLOW 0x00000000FFFFFFF0ULL
> +#define PRE_OVERFLOW2 0x00000000FFFFFFDCULL
> +
> +#define ALL_SET_AT(_64b) (_64b ? ALL_SET_64 : ALL_SET)
AFAICT, ALL_SET is mostly used to toggle all PMCs in a configuration
register. Using it for PMEVCNTR<n> seems a bit odd to me. How about
introducing a helper for getting the counter mask to avoid the
open-coded version check?
static uint64_t pmevcntr_mask(void)
{
/*
* Bits [63:0] are always incremented for 64-bit counters,
* even if the PMU is configured to generate an overflow at
* bits [31:0]
*
* See DDI0487I.a, section D11.3 ("Behavior on overflow") for
* more details.
*/
if (pmu.version >= ID_DFR0_PMU_V3_8_5)
return ~0;
return (uint32_t)~0;
}
I've always found the PMU documentation to be a bit difficult to grok,
and the above citation only mentions the intended behavior in passing.
Please feel free to update with a better citation if it exists.
> #define PMU_PPI 23
>
> @@ -538,6 +541,7 @@ static void test_mem_access(void)
> static void test_sw_incr(void)
> {
> uint32_t events[] = {SW_INCR, SW_INCR};
> + uint64_t cntr0;
> int i;
>
> if (!satisfy_prerequisites(events, ARRAY_SIZE(events)))
> @@ -572,9 +576,11 @@ static void test_sw_incr(void)
> write_sysreg(0x3, pmswinc_el0);
>
> isb();
> - report(read_regn_el0(pmevcntr, 0) == 84, "counter #1 after + 100 SW_INCR");
> - report(read_regn_el0(pmevcntr, 1) == 100,
> - "counter #0 after + 100 SW_INCR");
> + cntr0 = (pmu.version < ID_DFR0_PMU_V3_8_5) ?
> + (uint32_t)PRE_OVERFLOW + 100 :
> + (uint64_t)PRE_OVERFLOW + 100;
With the above suggestion, it would be nice to rewrite like so:
cntr0 = (PRE_OVERFLOW + 100) & pmevcntr_mask();
If you do go this route, then you'll probably want to drop all the other
open-coded PMUv3p5 checks in favor of the helper.
--
Thanks,
Oliver
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [kvm-unit-tests PATCH v2 1/4] arm: pmu: Fix overflow checks for PMUv3p5 long counters
2023-01-06 19:28 ` Oliver Upton
@ 2023-01-09 16:10 ` Ricardo Koller
2023-01-09 20:32 ` Ricardo Koller
1 sibling, 0 replies; 13+ messages in thread
From: Ricardo Koller @ 2023-01-09 16:10 UTC (permalink / raw)
To: Oliver Upton
Cc: kvm, kvmarm, andrew.jones, maz, alexandru.elisei, eric.auger, reijiw
On Fri, Jan 06, 2023 at 07:28:57PM +0000, Oliver Upton wrote:
> Hi Ricardo,
>
> On Tue, Dec 20, 2022 at 03:10:29AM +0000, Ricardo Koller wrote:
> > PMUv3p5 uses 64-bit counters irrespective of whether the PMU is configured
> > for overflowing at 32 or 64-bits. The consequence is that tests that check
> > the counter values after overflowing should not assume that values will be
> > wrapped around 32-bits: they overflow into the other half of the 64-bit
> > counters on PMUv3p5.
> >
> > Fix tests by correctly checking overflowing-counters against the expected
> > 64-bit value.
> >
> > Signed-off-by: Ricardo Koller <ricarkol@google.com>
> > ---
> > arm/pmu.c | 37 +++++++++++++++++++++++++------------
> > 1 file changed, 25 insertions(+), 12 deletions(-)
> >
> > diff --git a/arm/pmu.c b/arm/pmu.c
> > index cd47b14..1b55e20 100644
> > --- a/arm/pmu.c
> > +++ b/arm/pmu.c
> > @@ -54,10 +54,13 @@
> > #define EXT_COMMON_EVENTS_LOW 0x4000
> > #define EXT_COMMON_EVENTS_HIGH 0x403F
> >
> > -#define ALL_SET 0xFFFFFFFF
> > -#define ALL_CLEAR 0x0
> > -#define PRE_OVERFLOW 0xFFFFFFF0
> > -#define PRE_OVERFLOW2 0xFFFFFFDC
> > +#define ALL_SET 0x00000000FFFFFFFFULL
> > +#define ALL_SET_64 0xFFFFFFFFFFFFFFFFULL
> > +#define ALL_CLEAR 0x0000000000000000ULL
> > +#define PRE_OVERFLOW 0x00000000FFFFFFF0ULL
> > +#define PRE_OVERFLOW2 0x00000000FFFFFFDCULL
> > +
> > +#define ALL_SET_AT(_64b) (_64b ? ALL_SET_64 : ALL_SET)
>
> AFAICT, ALL_SET is mostly used to toggle all PMCs in a configuration
> register.
It's also used as a pre-overflow counter value. And that's mainly the use of
the newly introduced ALL_SET_AT(), and PRE_OVERFLOW_AT(). The ALL_SET_AT()
changes in this commit should be moved to commit 3; I will do that plus use
pmevcntr_mask() instead of ALL_SET_AT():
uint64_t all_set = pmevcntr_mask();
...
/* overflow on odd counter */
write_regn_el0(pmevcntr, 0, pre_overflow);
write_regn_el0(pmevcntr, 1, all_set);
> Using it for PMEVCNTR<n> seems a bit odd to me. How about
> introducing a helper for getting the counter mask to avoid the
> open-coded version check?
>
> static uint64_t pmevcntr_mask(void)
> {
> /*
> * Bits [63:0] are always incremented for 64-bit counters,
> * even if the PMU is configured to generate an overflow at
> * bits [31:0]
> *
> * See DDI0487I.a, section D11.3 ("Behavior on overflow") for
> * more details.
> */
> if (pmu.version >= ID_DFR0_PMU_V3_8_5)
> return ~0;
>
> return (uint32_t)~0;
> }
>
> I've always found the PMU documentation to be a bit difficult to grok,
> and the above citation only mentions the intended behavior in passing.
> Please feel free to update with a better citation if it exists.
>
> > #define PMU_PPI 23
> >
> > @@ -538,6 +541,7 @@ static void test_mem_access(void)
> > static void test_sw_incr(void)
> > {
> > uint32_t events[] = {SW_INCR, SW_INCR};
> > + uint64_t cntr0;
> > int i;
> >
> > if (!satisfy_prerequisites(events, ARRAY_SIZE(events)))
> > @@ -572,9 +576,11 @@ static void test_sw_incr(void)
> > write_sysreg(0x3, pmswinc_el0);
> >
> > isb();
> > - report(read_regn_el0(pmevcntr, 0) == 84, "counter #1 after + 100 SW_INCR");
> > - report(read_regn_el0(pmevcntr, 1) == 100,
> > - "counter #0 after + 100 SW_INCR");
> > + cntr0 = (pmu.version < ID_DFR0_PMU_V3_8_5) ?
> > + (uint32_t)PRE_OVERFLOW + 100 :
> > + (uint64_t)PRE_OVERFLOW + 100;
>
> With the above suggestion, it would be nice to rewrite like so:
>
> cntr0 = (PRE_OVERFLOW + 100) & pmevcntr_mask();
>
> If you do go this route, then you'll probably want to drop all the other
> open-coded PMUv3p5 checks in favor of the helper.
Will also use this instead of the uintxx_t casting.
Thanks~
Ricardo
>
> --
> Thanks,
> Oliver
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [kvm-unit-tests PATCH v2 1/4] arm: pmu: Fix overflow checks for PMUv3p5 long counters
2023-01-06 19:28 ` Oliver Upton
2023-01-09 16:10 ` Ricardo Koller
@ 2023-01-09 20:32 ` Ricardo Koller
1 sibling, 0 replies; 13+ messages in thread
From: Ricardo Koller @ 2023-01-09 20:32 UTC (permalink / raw)
To: Oliver Upton
Cc: kvm, kvmarm, andrew.jones, maz, alexandru.elisei, eric.auger, reijiw
On Fri, Jan 06, 2023 at 07:28:57PM +0000, Oliver Upton wrote:
> Hi Ricardo,
>
> On Tue, Dec 20, 2022 at 03:10:29AM +0000, Ricardo Koller wrote:
> > PMUv3p5 uses 64-bit counters irrespective of whether the PMU is configured
> > for overflowing at 32 or 64-bits. The consequence is that tests that check
> > the counter values after overflowing should not assume that values will be
> > wrapped around 32-bits: they overflow into the other half of the 64-bit
> > counters on PMUv3p5.
> >
> > Fix tests by correctly checking overflowing-counters against the expected
> > 64-bit value.
> >
> > Signed-off-by: Ricardo Koller <ricarkol@google.com>
> > ---
> > arm/pmu.c | 37 +++++++++++++++++++++++++------------
> > 1 file changed, 25 insertions(+), 12 deletions(-)
> >
> > diff --git a/arm/pmu.c b/arm/pmu.c
> > index cd47b14..1b55e20 100644
> > --- a/arm/pmu.c
> > +++ b/arm/pmu.c
> > @@ -54,10 +54,13 @@
> > #define EXT_COMMON_EVENTS_LOW 0x4000
> > #define EXT_COMMON_EVENTS_HIGH 0x403F
> >
> > -#define ALL_SET 0xFFFFFFFF
> > -#define ALL_CLEAR 0x0
> > -#define PRE_OVERFLOW 0xFFFFFFF0
> > -#define PRE_OVERFLOW2 0xFFFFFFDC
> > +#define ALL_SET 0x00000000FFFFFFFFULL
> > +#define ALL_SET_64 0xFFFFFFFFFFFFFFFFULL
> > +#define ALL_CLEAR 0x0000000000000000ULL
> > +#define PRE_OVERFLOW 0x00000000FFFFFFF0ULL
> > +#define PRE_OVERFLOW2 0x00000000FFFFFFDCULL
> > +
> > +#define ALL_SET_AT(_64b) (_64b ? ALL_SET_64 : ALL_SET)
>
> AFAICT, ALL_SET is mostly used to toggle all PMCs in a configuration
> register. Using it for PMEVCNTR<n> seems a bit odd to me. How about
> introducing a helper for getting the counter mask to avoid the
> open-coded version check?
>
> static uint64_t pmevcntr_mask(void)
> {
> /*
> * Bits [63:0] are always incremented for 64-bit counters,
> * even if the PMU is configured to generate an overflow at
> * bits [31:0]
> *
> * See DDI0487I.a, section D11.3 ("Behavior on overflow") for
> * more details.
> */
> if (pmu.version >= ID_DFR0_PMU_V3_8_5)
> return ~0;
>
> return (uint32_t)~0;
> }
>
> I've always found the PMU documentation to be a bit difficult to grok,
> and the above citation only mentions the intended behavior in passing.
> Please feel free to update with a better citation if it exists.
For this particular case, "Bits [63:0] are always incremented for 64-bit
counters", I find AArch64.IncrementEventCounter() a bit easier.
>
> > #define PMU_PPI 23
> >
> > @@ -538,6 +541,7 @@ static void test_mem_access(void)
> > static void test_sw_incr(void)
> > {
> > uint32_t events[] = {SW_INCR, SW_INCR};
> > + uint64_t cntr0;
> > int i;
> >
> > if (!satisfy_prerequisites(events, ARRAY_SIZE(events)))
> > @@ -572,9 +576,11 @@ static void test_sw_incr(void)
> > write_sysreg(0x3, pmswinc_el0);
> >
> > isb();
> > - report(read_regn_el0(pmevcntr, 0) == 84, "counter #1 after + 100 SW_INCR");
> > - report(read_regn_el0(pmevcntr, 1) == 100,
> > - "counter #0 after + 100 SW_INCR");
> > + cntr0 = (pmu.version < ID_DFR0_PMU_V3_8_5) ?
> > + (uint32_t)PRE_OVERFLOW + 100 :
> > + (uint64_t)PRE_OVERFLOW + 100;
>
> With the above suggestion, it would be nice to rewrite like so:
>
> cntr0 = (PRE_OVERFLOW + 100) & pmevcntr_mask();
>
> If you do go this route, then you'll probably want to drop all the other
> open-coded PMUv3p5 checks in favor of the helper.
>
> --
> Thanks,
> Oliver
^ permalink raw reply [flat|nested] 13+ messages in thread
end of thread, other threads:[~2023-01-09 20:32 UTC | newest]
Thread overview: 13+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-12-20 3:10 [kvm-unit-tests PATCH v2 0/4] arm: pmu: Add support for PMUv3p5 Ricardo Koller
2022-12-20 3:10 ` Ricardo Koller
2022-12-20 3:10 ` [kvm-unit-tests PATCH v2 1/4] arm: pmu: Fix overflow checks for PMUv3p5 long counters Ricardo Koller
2022-12-20 3:10 ` Ricardo Koller
2023-01-06 19:28 ` Oliver Upton
2023-01-09 16:10 ` Ricardo Koller
2023-01-09 20:32 ` Ricardo Koller
2022-12-20 3:10 ` [kvm-unit-tests PATCH v2 2/4] arm: pmu: Prepare for testing 64-bit overflows Ricardo Koller
2022-12-20 3:10 ` Ricardo Koller
2022-12-20 3:10 ` [kvm-unit-tests PATCH v2 3/4] arm: pmu: Add tests for " Ricardo Koller
2022-12-20 3:10 ` Ricardo Koller
2022-12-20 3:10 ` [kvm-unit-tests PATCH v2 4/4] arm: pmu: Print counter values as hexadecimals Ricardo Koller
2022-12-20 3:10 ` Ricardo Koller
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.