struct perf_event *event_list[X86_PMC_IDX_MAX]; /* in enabled order */
};
-#define EVENT_CONSTRAINT(c, n, m) { \
+#define __EVENT_CONSTRAINT(c, n, m, w) {\
{ .idxmsk64[0] = (n) }, \
.code = (c), \
.cmask = (m), \
- .weight = HWEIGHT64((u64)(n)), \
+ .weight = (w), \
}
-#define INTEL_EVENT_CONSTRAINT(c, n) EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK)
-#define FIXED_EVENT_CONSTRAINT(c, n) EVENT_CONSTRAINT(c, n, INTEL_ARCH_FIXED_MASK)
+#define EVENT_CONSTRAINT(c, n, m) \
+ __EVENT_CONSTRAINT(c, n, m, HWEIGHT(n))
-#define EVENT_CONSTRAINT_END EVENT_CONSTRAINT(0, 0, 0)
+#define INTEL_EVENT_CONSTRAINT(c, n) \
+ EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVTSEL_MASK)
-#define for_each_event_constraint(e, c) for ((e) = (c); (e)->cmask; (e)++)
+#define FIXED_EVENT_CONSTRAINT(c, n) \
+ EVENT_CONSTRAINT(c, n, INTEL_ARCH_FIXED_MASK)
+
+#define EVENT_CONSTRAINT_END \
+ EVENT_CONSTRAINT(0, 0, 0)
+
+#define for_each_event_constraint(e, c) \
+ for ((e) = (c); (e)->cmask; (e)++)
/*
* struct x86_pmu - generic x86 pmu
};
static struct event_constraint intel_core_event_constraints[] =
+{
+ INTEL_EVENT_CONSTRAINT(0x11, 0x2), /* FP_ASSIST */
+ INTEL_EVENT_CONSTRAINT(0x12, 0x2), /* MUL */
+ INTEL_EVENT_CONSTRAINT(0x13, 0x2), /* DIV */
+ INTEL_EVENT_CONSTRAINT(0x14, 0x1), /* CYCLES_DIV_BUSY */
+ INTEL_EVENT_CONSTRAINT(0x19, 0x2), /* DELAYED_BYPASS */
+ INTEL_EVENT_CONSTRAINT(0xc1, 0x1), /* FP_COMP_INSTR_RET */
+ EVENT_CONSTRAINT_END
+};
+
+static struct event_constraint intel_core2_event_constraints[] =
{
FIXED_EVENT_CONSTRAINT(0xc0, (0x3|(1ULL<<32))), /* INSTRUCTIONS_RETIRED */
FIXED_EVENT_CONSTRAINT(0x3c, (0x3|(1ULL<<33))), /* UNHALTED_CORE_CYCLES */
static struct event_constraint intel_nehalem_event_constraints[] =
{
- FIXED_EVENT_CONSTRAINT(0xc0, (0x3|(1ULL<<32))), /* INSTRUCTIONS_RETIRED */
- FIXED_EVENT_CONSTRAINT(0x3c, (0x3|(1ULL<<33))), /* UNHALTED_CORE_CYCLES */
+ FIXED_EVENT_CONSTRAINT(0xc0, (0xf|(1ULL<<32))), /* INSTRUCTIONS_RETIRED */
+ FIXED_EVENT_CONSTRAINT(0x3c, (0xf|(1ULL<<33))), /* UNHALTED_CORE_CYCLES */
INTEL_EVENT_CONSTRAINT(0x40, 0x3), /* L1D_CACHE_LD */
INTEL_EVENT_CONSTRAINT(0x41, 0x3), /* L1D_CACHE_ST */
INTEL_EVENT_CONSTRAINT(0x42, 0x3), /* L1D_CACHE_LOCK */
INTEL_EVENT_CONSTRAINT(0x43, 0x3), /* L1D_ALL_REF */
+ INTEL_EVENT_CONSTRAINT(0x48, 0x3), /* L1D_PEND_MISS */
INTEL_EVENT_CONSTRAINT(0x4e, 0x3), /* L1D_PREFETCH */
- INTEL_EVENT_CONSTRAINT(0x4c, 0x3), /* LOAD_HIT_PRE */
INTEL_EVENT_CONSTRAINT(0x51, 0x3), /* L1D */
- INTEL_EVENT_CONSTRAINT(0x52, 0x3), /* L1D_CACHE_PREFETCH_LOCK_FB_HIT */
- INTEL_EVENT_CONSTRAINT(0x53, 0x3), /* L1D_CACHE_LOCK_FB_HIT */
- INTEL_EVENT_CONSTRAINT(0xc5, 0x3), /* CACHE_LOCK_CYCLES */
+ INTEL_EVENT_CONSTRAINT(0x63, 0x3), /* CACHE_LOCK_CYCLES */
+ EVENT_CONSTRAINT_END
+};
+
+static struct event_constraint intel_westmere_event_constraints[] =
+{
+ FIXED_EVENT_CONSTRAINT(0xc0, (0xf|(1ULL<<32))), /* INSTRUCTIONS_RETIRED */
+ FIXED_EVENT_CONSTRAINT(0x3c, (0xf|(1ULL<<33))), /* UNHALTED_CORE_CYCLES */
+ INTEL_EVENT_CONSTRAINT(0x51, 0x3), /* L1D */
+ INTEL_EVENT_CONSTRAINT(0x60, 0x1), /* OFFCORE_REQUESTS_OUTSTANDING */
+ INTEL_EVENT_CONSTRAINT(0x63, 0x3), /* CACHE_LOCK_CYCLES */
EVENT_CONSTRAINT_END
};
[PERF_COUNT_HW_CACHE_OP_MAX]
[PERF_COUNT_HW_CACHE_RESULT_MAX];
+static __initconst u64 westmere_hw_cache_event_ids
+ [PERF_COUNT_HW_CACHE_MAX]
+ [PERF_COUNT_HW_CACHE_OP_MAX]
+ [PERF_COUNT_HW_CACHE_RESULT_MAX] =
+{
+ [ C(L1D) ] = {
+ [ C(OP_READ) ] = {
+ [ C(RESULT_ACCESS) ] = 0x010b, /* MEM_INST_RETIRED.LOADS */
+ [ C(RESULT_MISS) ] = 0x0151, /* L1D.REPL */
+ },
+ [ C(OP_WRITE) ] = {
+ [ C(RESULT_ACCESS) ] = 0x020b, /* MEM_INST_RETURED.STORES */
+ [ C(RESULT_MISS) ] = 0x0251, /* L1D.M_REPL */
+ },
+ [ C(OP_PREFETCH) ] = {
+ [ C(RESULT_ACCESS) ] = 0x014e, /* L1D_PREFETCH.REQUESTS */
+ [ C(RESULT_MISS) ] = 0x024e, /* L1D_PREFETCH.MISS */
+ },
+ },
+ [ C(L1I ) ] = {
+ [ C(OP_READ) ] = {
+ [ C(RESULT_ACCESS) ] = 0x0380, /* L1I.READS */
+ [ C(RESULT_MISS) ] = 0x0280, /* L1I.MISSES */
+ },
+ [ C(OP_WRITE) ] = {
+ [ C(RESULT_ACCESS) ] = -1,
+ [ C(RESULT_MISS) ] = -1,
+ },
+ [ C(OP_PREFETCH) ] = {
+ [ C(RESULT_ACCESS) ] = 0x0,
+ [ C(RESULT_MISS) ] = 0x0,
+ },
+ },
+ [ C(LL ) ] = {
+ [ C(OP_READ) ] = {
+ [ C(RESULT_ACCESS) ] = 0x0324, /* L2_RQSTS.LOADS */
+ [ C(RESULT_MISS) ] = 0x0224, /* L2_RQSTS.LD_MISS */
+ },
+ [ C(OP_WRITE) ] = {
+ [ C(RESULT_ACCESS) ] = 0x0c24, /* L2_RQSTS.RFOS */
+ [ C(RESULT_MISS) ] = 0x0824, /* L2_RQSTS.RFO_MISS */
+ },
+ [ C(OP_PREFETCH) ] = {
+ [ C(RESULT_ACCESS) ] = 0x4f2e, /* LLC Reference */
+ [ C(RESULT_MISS) ] = 0x412e, /* LLC Misses */
+ },
+ },
+ [ C(DTLB) ] = {
+ [ C(OP_READ) ] = {
+ [ C(RESULT_ACCESS) ] = 0x010b, /* MEM_INST_RETIRED.LOADS */
+ [ C(RESULT_MISS) ] = 0x0108, /* DTLB_LOAD_MISSES.ANY */
+ },
+ [ C(OP_WRITE) ] = {
+ [ C(RESULT_ACCESS) ] = 0x020b, /* MEM_INST_RETURED.STORES */
+ [ C(RESULT_MISS) ] = 0x010c, /* MEM_STORE_RETIRED.DTLB_MISS */
+ },
+ [ C(OP_PREFETCH) ] = {
+ [ C(RESULT_ACCESS) ] = 0x0,
+ [ C(RESULT_MISS) ] = 0x0,
+ },
+ },
+ [ C(ITLB) ] = {
+ [ C(OP_READ) ] = {
+ [ C(RESULT_ACCESS) ] = 0x01c0, /* INST_RETIRED.ANY_P */
+ [ C(RESULT_MISS) ] = 0x0185, /* ITLB_MISSES.ANY */
+ },
+ [ C(OP_WRITE) ] = {
+ [ C(RESULT_ACCESS) ] = -1,
+ [ C(RESULT_MISS) ] = -1,
+ },
+ [ C(OP_PREFETCH) ] = {
+ [ C(RESULT_ACCESS) ] = -1,
+ [ C(RESULT_MISS) ] = -1,
+ },
+ },
+ [ C(BPU ) ] = {
+ [ C(OP_READ) ] = {
+ [ C(RESULT_ACCESS) ] = 0x00c4, /* BR_INST_RETIRED.ALL_BRANCHES */
+ [ C(RESULT_MISS) ] = 0x03e8, /* BPU_CLEARS.ANY */
+ },
+ [ C(OP_WRITE) ] = {
+ [ C(RESULT_ACCESS) ] = -1,
+ [ C(RESULT_MISS) ] = -1,
+ },
+ [ C(OP_PREFETCH) ] = {
+ [ C(RESULT_ACCESS) ] = -1,
+ [ C(RESULT_MISS) ] = -1,
+ },
+ },
+};
+
static __initconst u64 nehalem_hw_cache_event_ids
[PERF_COUNT_HW_CACHE_MAX]
[PERF_COUNT_HW_CACHE_OP_MAX]
static void p6_pmu_disable_all(void)
{
- struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
u64 val;
- if (!cpuc->enabled)
- return;
-
- cpuc->enabled = 0;
- barrier();
-
/* p6 only has one enable register */
rdmsrl(MSR_P6_EVNTSEL0, val);
val &= ~ARCH_PERFMON_EVENTSEL0_ENABLE;
{
struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
- if (!cpuc->enabled)
- return;
-
- cpuc->enabled = 0;
- barrier();
-
wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL, 0);
if (test_bit(X86_PMC_IDX_FIXED_BTS, cpuc->active_mask))
intel_pmu_disable_bts();
}
-static void amd_pmu_disable_all(void)
+static void x86_pmu_disable_all(void)
{
struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
int idx;
- if (!cpuc->enabled)
- return;
-
- cpuc->enabled = 0;
- /*
- * ensure we write the disable before we start disabling the
- * events proper, so that amd_pmu_enable_event() does the
- * right thing.
- */
- barrier();
-
for (idx = 0; idx < x86_pmu.num_events; idx++) {
u64 val;
if (!test_bit(idx, cpuc->active_mask))
continue;
- rdmsrl(MSR_K7_EVNTSEL0 + idx, val);
+ rdmsrl(x86_pmu.eventsel + idx, val);
if (!(val & ARCH_PERFMON_EVENTSEL0_ENABLE))
continue;
val &= ~ARCH_PERFMON_EVENTSEL0_ENABLE;
- wrmsrl(MSR_K7_EVNTSEL0 + idx, val);
+ wrmsrl(x86_pmu.eventsel + idx, val);
}
}
if (!x86_pmu_initialized())
return;
- if (cpuc->enabled)
- cpuc->n_added = 0;
+ if (!cpuc->enabled)
+ return;
+
+ cpuc->n_added = 0;
+ cpuc->enabled = 0;
+ barrier();
x86_pmu.disable_all();
}
static void p6_pmu_enable_all(void)
{
- struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
unsigned long val;
- if (cpuc->enabled)
- return;
-
- cpuc->enabled = 1;
- barrier();
-
/* p6 only has one enable register */
rdmsrl(MSR_P6_EVNTSEL0, val);
val |= ARCH_PERFMON_EVENTSEL0_ENABLE;
{
struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
- if (cpuc->enabled)
- return;
-
- cpuc->enabled = 1;
- barrier();
-
wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL, x86_pmu.intel_ctrl);
if (test_bit(X86_PMC_IDX_FIXED_BTS, cpuc->active_mask)) {
}
}
-static void amd_pmu_enable_all(void)
+static void x86_pmu_enable_all(void)
{
struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
int idx;
- if (cpuc->enabled)
- return;
-
- cpuc->enabled = 1;
- barrier();
-
for (idx = 0; idx < x86_pmu.num_events; idx++) {
struct perf_event *event = cpuc->events[idx];
u64 val;
val = event->hw.config;
val |= ARCH_PERFMON_EVENTSEL0_ENABLE;
- wrmsrl(MSR_K7_EVNTSEL0 + idx, val);
+ wrmsrl(x86_pmu.eventsel + idx, val);
}
}
if (!x86_pmu_initialized())
return;
+
+ if (cpuc->enabled)
+ return;
+
if (cpuc->n_added) {
/*
* apply assignment obtained either from
cpuc->n_added = 0;
perf_events_lapic_init();
}
+
+ cpuc->enabled = 1;
+ barrier();
+
x86_pmu.enable_all();
}
wrmsrl(MSR_CORE_PERF_GLOBAL_OVF_CTRL, ack);
}
-static inline void x86_pmu_enable_event(struct hw_perf_event *hwc, int idx)
+static inline void __x86_pmu_enable_event(struct hw_perf_event *hwc, int idx)
{
(void)checking_wrmsrl(hwc->config_base + idx,
hwc->config | ARCH_PERFMON_EVENTSEL0_ENABLE);
x86_pmu_disable_event(hwc, idx);
}
-static inline void
-amd_pmu_disable_event(struct hw_perf_event *hwc, int idx)
-{
- x86_pmu_disable_event(hwc, idx);
-}
-
static DEFINE_PER_CPU(u64 [X86_PMC_IDX_MAX], pmc_prev_left);
/*
bits |= 0x2;
if (hwc->config & ARCH_PERFMON_EVENTSEL_OS)
bits |= 0x1;
+
+ /*
+ * ANY bit is supported in v3 and up
+ */
+ if (x86_pmu.version > 2 && hwc->config & ARCH_PERFMON_EVENTSEL_ANY)
+ bits |= 0x4;
+
bits <<= (idx * 4);
mask = 0xfULL << (idx * 4);
return;
}
- x86_pmu_enable_event(hwc, idx);
+ __x86_pmu_enable_event(hwc, idx);
}
-static void amd_pmu_enable_event(struct hw_perf_event *hwc, int idx)
+static void x86_pmu_enable_event(struct hw_perf_event *hwc, int idx)
{
struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
-
if (cpuc->enabled)
- x86_pmu_enable_event(hwc, idx);
+ __x86_pmu_enable_event(hwc, idx);
}
/*
cpuc->n_events = n;
cpuc->n_added = n - n0;
- if (hwc->idx != -1)
- x86_perf_event_set_period(event, hwc, hwc->idx);
-
return 0;
}
local_irq_restore(flags);
}
-static int p6_pmu_handle_irq(struct pt_regs *regs)
-{
- struct perf_sample_data data;
- struct cpu_hw_events *cpuc;
- struct perf_event *event;
- struct hw_perf_event *hwc;
- int idx, handled = 0;
- u64 val;
-
- data.addr = 0;
- data.raw = NULL;
-
- cpuc = &__get_cpu_var(cpu_hw_events);
-
- for (idx = 0; idx < x86_pmu.num_events; idx++) {
- if (!test_bit(idx, cpuc->active_mask))
- continue;
-
- event = cpuc->events[idx];
- hwc = &event->hw;
-
- val = x86_perf_event_update(event, hwc, idx);
- if (val & (1ULL << (x86_pmu.event_bits - 1)))
- continue;
-
- /*
- * event overflow
- */
- handled = 1;
- data.period = event->hw.last_period;
-
- if (!x86_perf_event_set_period(event, hwc, idx))
- continue;
-
- if (perf_event_overflow(event, 1, &data, regs))
- p6_pmu_disable_event(hwc, idx);
- }
-
- if (handled)
- inc_irq_stat(apic_perf_irqs);
-
- return handled;
-}
-
/*
* This handler is triggered by the local APIC, so the APIC IRQ handling
* rules apply:
return 1;
}
-static int amd_pmu_handle_irq(struct pt_regs *regs)
+static int x86_pmu_handle_irq(struct pt_regs *regs)
{
struct perf_sample_data data;
struct cpu_hw_events *cpuc;
continue;
if (perf_event_overflow(event, 1, &data, regs))
- amd_pmu_disable_event(hwc, idx);
+ x86_pmu.disable(hwc, idx);
}
if (handled)
static __initconst struct x86_pmu p6_pmu = {
.name = "p6",
- .handle_irq = p6_pmu_handle_irq,
+ .handle_irq = x86_pmu_handle_irq,
.disable_all = p6_pmu_disable_all,
.enable_all = p6_pmu_enable_all,
.enable = p6_pmu_enable_event,
.event_constraints = intel_p6_event_constraints
};
+static __initconst struct x86_pmu core_pmu = {
+ .name = "core",
+ .handle_irq = x86_pmu_handle_irq,
+ .disable_all = x86_pmu_disable_all,
+ .enable_all = x86_pmu_enable_all,
+ .enable = x86_pmu_enable_event,
+ .disable = x86_pmu_disable_event,
+ .eventsel = MSR_ARCH_PERFMON_EVENTSEL0,
+ .perfctr = MSR_ARCH_PERFMON_PERFCTR0,
+ .event_map = intel_pmu_event_map,
+ .raw_event = intel_pmu_raw_event,
+ .max_events = ARRAY_SIZE(intel_perfmon_event_map),
+ .apic = 1,
+ /*
+ * Intel PMCs cannot be accessed sanely above 32 bit width,
+ * so we install an artificial 1<<31 period regardless of
+ * the generic event period:
+ */
+ .max_period = (1ULL << 31) - 1,
+ .get_event_constraints = intel_get_event_constraints,
+ .event_constraints = intel_core_event_constraints,
+};
+
static __initconst struct x86_pmu intel_pmu = {
.name = "Intel",
.handle_irq = intel_pmu_handle_irq,
static __initconst struct x86_pmu amd_pmu = {
.name = "AMD",
- .handle_irq = amd_pmu_handle_irq,
- .disable_all = amd_pmu_disable_all,
- .enable_all = amd_pmu_enable_all,
- .enable = amd_pmu_enable_event,
- .disable = amd_pmu_disable_event,
+ .handle_irq = x86_pmu_handle_irq,
+ .disable_all = x86_pmu_disable_all,
+ .enable_all = x86_pmu_enable_all,
+ .enable = x86_pmu_enable_event,
+ .disable = x86_pmu_disable_event,
.eventsel = MSR_K7_EVNTSEL0,
.perfctr = MSR_K7_PERFCTR0,
.event_map = amd_pmu_event_map,
version = eax.split.version_id;
if (version < 2)
- return -ENODEV;
+ x86_pmu = core_pmu;
+ else
+ x86_pmu = intel_pmu;
- x86_pmu = intel_pmu;
x86_pmu.version = version;
x86_pmu.num_events = eax.split.num_events;
x86_pmu.event_bits = eax.split.bit_width;
* Quirk: v2 perfmon does not report fixed-purpose events, so
* assume at least 3 events:
*/
- x86_pmu.num_events_fixed = max((int)edx.split.num_events_fixed, 3);
+ if (version > 1)
+ x86_pmu.num_events_fixed = max((int)edx.split.num_events_fixed, 3);
/*
* Install the hw-cache-events table:
*/
switch (boot_cpu_data.x86_model) {
+ case 14: /* 65 nm core solo/duo, "Yonah" */
+ pr_cont("Core events, ");
+ break;
+
case 15: /* original 65 nm celeron/pentium/core2/xeon, "Merom"/"Conroe" */
case 22: /* single-core 65 nm celeron/core2solo "Merom-L"/"Conroe-L" */
case 23: /* current 45 nm celeron/core2/xeon "Penryn"/"Wolfdale" */
memcpy(hw_cache_event_ids, core2_hw_cache_event_ids,
sizeof(hw_cache_event_ids));
- x86_pmu.event_constraints = intel_core_event_constraints;
+ x86_pmu.event_constraints = intel_core2_event_constraints;
pr_cont("Core2 events, ");
break;
- case 26:
+
+ case 26: /* 45 nm nehalem, "Bloomfield" */
+ case 30: /* 45 nm nehalem, "Lynnfield" */
memcpy(hw_cache_event_ids, nehalem_hw_cache_event_ids,
sizeof(hw_cache_event_ids));
x86_pmu.event_constraints = intel_gen_event_constraints;
pr_cont("Atom events, ");
break;
+
+ case 37: /* 32 nm nehalem, "Clarkdale" */
+ case 44: /* 32 nm nehalem, "Gulftown" */
+ memcpy(hw_cache_event_ids, westmere_hw_cache_event_ids,
+ sizeof(hw_cache_event_ids));
+
+ x86_pmu.event_constraints = intel_westmere_event_constraints;
+ pr_cont("Westmere events, ");
+ break;
default:
/*
* default constraints for v2 and up
register_die_notifier(&perf_event_nmi_notifier);
unconstrained = (struct event_constraint)
- EVENT_CONSTRAINT(0, (1ULL << x86_pmu.num_events) - 1, 0);
+ __EVENT_CONSTRAINT(0, (1ULL << x86_pmu.num_events) - 1,
+ 0, x86_pmu.num_events);
pr_info("... version: %d\n", x86_pmu.version);
pr_info("... bit width: %d\n", x86_pmu.event_bits);