summaryrefslogtreecommitdiffstats
path: root/arch/x86/kernel/cpu/perf_event_intel.c
diff options
context:
space:
mode:
Diffstat (limited to 'arch/x86/kernel/cpu/perf_event_intel.c')
-rw-r--r--arch/x86/kernel/cpu/perf_event_intel.c105
1 files changed, 70 insertions, 35 deletions
diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
index 121f1be4da1..61d4f79a550 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -28,6 +28,7 @@ static u64 intel_perfmon_event_map[PERF_COUNT_HW_MAX] __read_mostly =
[PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = 0x00c4,
[PERF_COUNT_HW_BRANCH_MISSES] = 0x00c5,
[PERF_COUNT_HW_BUS_CYCLES] = 0x013c,
+ [PERF_COUNT_HW_REF_CPU_CYCLES] = 0x0300, /* pseudo-encoding */
};
static struct event_constraint intel_core_event_constraints[] __read_mostly =
@@ -45,12 +46,7 @@ static struct event_constraint intel_core2_event_constraints[] __read_mostly =
{
FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */
FIXED_EVENT_CONSTRAINT(0x003c, 1), /* CPU_CLK_UNHALTED.CORE */
- /*
- * Core2 has Fixed Counter 2 listed as CPU_CLK_UNHALTED.REF and event
- * 0x013c as CPU_CLK_UNHALTED.BUS and specifies there is a fixed
- * ratio between these counters.
- */
- /* FIXED_EVENT_CONSTRAINT(0x013c, 2), CPU_CLK_UNHALTED.REF */
+ FIXED_EVENT_CONSTRAINT(0x0300, 2), /* CPU_CLK_UNHALTED.REF */
INTEL_EVENT_CONSTRAINT(0x10, 0x1), /* FP_COMP_OPS_EXE */
INTEL_EVENT_CONSTRAINT(0x11, 0x2), /* FP_ASSIST */
INTEL_EVENT_CONSTRAINT(0x12, 0x2), /* MUL */
@@ -68,7 +64,7 @@ static struct event_constraint intel_nehalem_event_constraints[] __read_mostly =
{
FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */
FIXED_EVENT_CONSTRAINT(0x003c, 1), /* CPU_CLK_UNHALTED.CORE */
- /* FIXED_EVENT_CONSTRAINT(0x013c, 2), CPU_CLK_UNHALTED.REF */
+ FIXED_EVENT_CONSTRAINT(0x0300, 2), /* CPU_CLK_UNHALTED.REF */
INTEL_EVENT_CONSTRAINT(0x40, 0x3), /* L1D_CACHE_LD */
INTEL_EVENT_CONSTRAINT(0x41, 0x3), /* L1D_CACHE_ST */
INTEL_EVENT_CONSTRAINT(0x42, 0x3), /* L1D_CACHE_LOCK */
@@ -90,7 +86,7 @@ static struct event_constraint intel_westmere_event_constraints[] __read_mostly
{
FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */
FIXED_EVENT_CONSTRAINT(0x003c, 1), /* CPU_CLK_UNHALTED.CORE */
- /* FIXED_EVENT_CONSTRAINT(0x013c, 2), CPU_CLK_UNHALTED.REF */
+ FIXED_EVENT_CONSTRAINT(0x0300, 2), /* CPU_CLK_UNHALTED.REF */
INTEL_EVENT_CONSTRAINT(0x51, 0x3), /* L1D */
INTEL_EVENT_CONSTRAINT(0x60, 0x1), /* OFFCORE_REQUESTS_OUTSTANDING */
INTEL_EVENT_CONSTRAINT(0x63, 0x3), /* CACHE_LOCK_CYCLES */
@@ -102,7 +98,7 @@ static struct event_constraint intel_snb_event_constraints[] __read_mostly =
{
FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */
FIXED_EVENT_CONSTRAINT(0x003c, 1), /* CPU_CLK_UNHALTED.CORE */
- /* FIXED_EVENT_CONSTRAINT(0x013c, 2), CPU_CLK_UNHALTED.REF */
+ FIXED_EVENT_CONSTRAINT(0x0300, 2), /* CPU_CLK_UNHALTED.REF */
INTEL_EVENT_CONSTRAINT(0x48, 0x4), /* L1D_PEND_MISS.PENDING */
INTEL_UEVENT_CONSTRAINT(0x01c0, 0x2), /* INST_RETIRED.PREC_DIST */
INTEL_EVENT_CONSTRAINT(0xcd, 0x8), /* MEM_TRANS_RETIRED.LOAD_LATENCY */
@@ -125,7 +121,7 @@ static struct event_constraint intel_gen_event_constraints[] __read_mostly =
{
FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */
FIXED_EVENT_CONSTRAINT(0x003c, 1), /* CPU_CLK_UNHALTED.CORE */
- /* FIXED_EVENT_CONSTRAINT(0x013c, 2), CPU_CLK_UNHALTED.REF */
+ FIXED_EVENT_CONSTRAINT(0x0300, 2), /* CPU_CLK_UNHALTED.REF */
EVENT_CONSTRAINT_END
};
@@ -389,14 +385,15 @@ static __initconst const u64 westmere_hw_cache_event_ids
#define NHM_LOCAL_DRAM (1 << 14)
#define NHM_NON_DRAM (1 << 15)
-#define NHM_ALL_DRAM (NHM_REMOTE_DRAM|NHM_LOCAL_DRAM)
+#define NHM_LOCAL (NHM_LOCAL_DRAM|NHM_REMOTE_CACHE_FWD)
+#define NHM_REMOTE (NHM_REMOTE_DRAM)
#define NHM_DMND_READ (NHM_DMND_DATA_RD)
#define NHM_DMND_WRITE (NHM_DMND_RFO|NHM_DMND_WB)
#define NHM_DMND_PREFETCH (NHM_PF_DATA_RD|NHM_PF_DATA_RFO)
#define NHM_L3_HIT (NHM_UNCORE_HIT|NHM_OTHER_CORE_HIT_SNP|NHM_OTHER_CORE_HITM)
-#define NHM_L3_MISS (NHM_NON_DRAM|NHM_ALL_DRAM|NHM_REMOTE_CACHE_FWD)
+#define NHM_L3_MISS (NHM_NON_DRAM|NHM_LOCAL_DRAM|NHM_REMOTE_DRAM|NHM_REMOTE_CACHE_FWD)
#define NHM_L3_ACCESS (NHM_L3_HIT|NHM_L3_MISS)
static __initconst const u64 nehalem_hw_cache_extra_regs
@@ -420,16 +417,16 @@ static __initconst const u64 nehalem_hw_cache_extra_regs
},
[ C(NODE) ] = {
[ C(OP_READ) ] = {
- [ C(RESULT_ACCESS) ] = NHM_DMND_READ|NHM_ALL_DRAM,
- [ C(RESULT_MISS) ] = NHM_DMND_READ|NHM_REMOTE_DRAM,
+ [ C(RESULT_ACCESS) ] = NHM_DMND_READ|NHM_LOCAL|NHM_REMOTE,
+ [ C(RESULT_MISS) ] = NHM_DMND_READ|NHM_REMOTE,
},
[ C(OP_WRITE) ] = {
- [ C(RESULT_ACCESS) ] = NHM_DMND_WRITE|NHM_ALL_DRAM,
- [ C(RESULT_MISS) ] = NHM_DMND_WRITE|NHM_REMOTE_DRAM,
+ [ C(RESULT_ACCESS) ] = NHM_DMND_WRITE|NHM_LOCAL|NHM_REMOTE,
+ [ C(RESULT_MISS) ] = NHM_DMND_WRITE|NHM_REMOTE,
},
[ C(OP_PREFETCH) ] = {
- [ C(RESULT_ACCESS) ] = NHM_DMND_PREFETCH|NHM_ALL_DRAM,
- [ C(RESULT_MISS) ] = NHM_DMND_PREFETCH|NHM_REMOTE_DRAM,
+ [ C(RESULT_ACCESS) ] = NHM_DMND_PREFETCH|NHM_LOCAL|NHM_REMOTE,
+ [ C(RESULT_MISS) ] = NHM_DMND_PREFETCH|NHM_REMOTE,
},
},
};
@@ -1519,7 +1516,7 @@ static __initconst const struct x86_pmu intel_pmu = {
.guest_get_msrs = intel_guest_get_msrs,
};
-static void intel_clovertown_quirks(void)
+static __init void intel_clovertown_quirk(void)
{
/*
* PEBS is unreliable due to:
@@ -1545,19 +1542,60 @@ static void intel_clovertown_quirks(void)
x86_pmu.pebs_constraints = NULL;
}
-static void intel_sandybridge_quirks(void)
+static __init void intel_sandybridge_quirk(void)
{
printk(KERN_WARNING "PEBS disabled due to CPU errata.\n");
x86_pmu.pebs = 0;
x86_pmu.pebs_constraints = NULL;
}
+static const struct { int id; char *name; } intel_arch_events_map[] __initconst = {
+ { PERF_COUNT_HW_CPU_CYCLES, "cpu cycles" },
+ { PERF_COUNT_HW_INSTRUCTIONS, "instructions" },
+ { PERF_COUNT_HW_BUS_CYCLES, "bus cycles" },
+ { PERF_COUNT_HW_CACHE_REFERENCES, "cache references" },
+ { PERF_COUNT_HW_CACHE_MISSES, "cache misses" },
+ { PERF_COUNT_HW_BRANCH_INSTRUCTIONS, "branch instructions" },
+ { PERF_COUNT_HW_BRANCH_MISSES, "branch misses" },
+};
+
+static __init void intel_arch_events_quirk(void)
+{
+ int bit;
+
+ /* disable event that reported as not presend by cpuid */
+ for_each_set_bit(bit, x86_pmu.events_mask, ARRAY_SIZE(intel_arch_events_map)) {
+ intel_perfmon_event_map[intel_arch_events_map[bit].id] = 0;
+ printk(KERN_WARNING "CPUID marked event: \'%s\' unavailable\n",
+ intel_arch_events_map[bit].name);
+ }
+}
+
+static __init void intel_nehalem_quirk(void)
+{
+ union cpuid10_ebx ebx;
+
+ ebx.full = x86_pmu.events_maskl;
+ if (ebx.split.no_branch_misses_retired) {
+ /*
+ * Erratum AAJ80 detected, we work it around by using
+ * the BR_MISP_EXEC.ANY event. This will over-count
+ * branch-misses, but it's still much better than the
+ * architectural event which is often completely bogus:
+ */
+ intel_perfmon_event_map[PERF_COUNT_HW_BRANCH_MISSES] = 0x7f89;
+ ebx.split.no_branch_misses_retired = 0;
+ x86_pmu.events_maskl = ebx.full;
+ printk(KERN_INFO "CPU erratum AAJ80 worked around\n");
+ }
+}
+
__init int intel_pmu_init(void)
{
union cpuid10_edx edx;
union cpuid10_eax eax;
+ union cpuid10_ebx ebx;
unsigned int unused;
- unsigned int ebx;
int version;
if (!cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON)) {
@@ -1574,8 +1612,8 @@ __init int intel_pmu_init(void)
* Check whether the Architectural PerfMon supports
* Branch Misses Retired hw_event or not.
*/
- cpuid(10, &eax.full, &ebx, &unused, &edx.full);
- if (eax.split.mask_length <= ARCH_PERFMON_BRANCH_MISSES_RETIRED)
+ cpuid(10, &eax.full, &ebx.full, &unused, &edx.full);
+ if (eax.split.mask_length < ARCH_PERFMON_EVENTS_COUNT)
return -ENODEV;
version = eax.split.version_id;
@@ -1589,6 +1627,9 @@ __init int intel_pmu_init(void)
x86_pmu.cntval_bits = eax.split.bit_width;
x86_pmu.cntval_mask = (1ULL << eax.split.bit_width) - 1;
+ x86_pmu.events_maskl = ebx.full;
+ x86_pmu.events_mask_len = eax.split.mask_length;
+
/*
* Quirk: v2 perfmon does not report fixed-purpose events, so
* assume at least 3 events:
@@ -1608,6 +1649,8 @@ __init int intel_pmu_init(void)
intel_ds_init();
+ x86_add_quirk(intel_arch_events_quirk); /* Install first, so it runs last */
+
/*
* Install the hw-cache-events table:
*/
@@ -1617,7 +1660,7 @@ __init int intel_pmu_init(void)
break;
case 15: /* original 65 nm celeron/pentium/core2/xeon, "Merom"/"Conroe" */
- x86_pmu.quirks = intel_clovertown_quirks;
+ x86_add_quirk(intel_clovertown_quirk);
case 22: /* single-core 65 nm celeron/core2solo "Merom-L"/"Conroe-L" */
case 23: /* current 45 nm celeron/core2/xeon "Penryn"/"Wolfdale" */
case 29: /* six-core 45 nm xeon "Dunnington" */
@@ -1651,17 +1694,8 @@ __init int intel_pmu_init(void)
/* UOPS_EXECUTED.CORE_ACTIVE_CYCLES,c=1,i=1 */
intel_perfmon_event_map[PERF_COUNT_HW_STALLED_CYCLES_BACKEND] = 0x1803fb1;
- if (ebx & 0x40) {
- /*
- * Erratum AAJ80 detected, we work it around by using
- * the BR_MISP_EXEC.ANY event. This will over-count
- * branch-misses, but it's still much better than the
- * architectural event which is often completely bogus:
- */
- intel_perfmon_event_map[PERF_COUNT_HW_BRANCH_MISSES] = 0x7f89;
+ x86_add_quirk(intel_nehalem_quirk);
- pr_cont("erratum AAJ80 worked around, ");
- }
pr_cont("Nehalem events, ");
break;
@@ -1701,7 +1735,7 @@ __init int intel_pmu_init(void)
break;
case 42: /* SandyBridge */
- x86_pmu.quirks = intel_sandybridge_quirks;
+ x86_add_quirk(intel_sandybridge_quirk);
case 45: /* SandyBridge, "Romely-EP" */
memcpy(hw_cache_event_ids, snb_hw_cache_event_ids,
sizeof(hw_cache_event_ids));
@@ -1738,5 +1772,6 @@ __init int intel_pmu_init(void)
break;
}
}
+
return 0;
}
generated by cgit v1.2.3-70-g09d2 (git 2.46.0) at 2025-06-04 20:23:31 +0000