From 6a600a8b8749566a7d81ad75dcb8bf5342b5a39a Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Tue, 15 Nov 2011 10:51:15 +0100 Subject: perf, x86: Disable PEBS on SandyBridge chips Cc: Stephane Eranian Cc: stable@kernel.org Signed-off-by: Peter Zijlstra Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event_intel.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'arch/x86/kernel/cpu/perf_event_intel.c') diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c index 2be5ebe99872..8d601b18bf9f 100644 --- a/arch/x86/kernel/cpu/perf_event_intel.c +++ b/arch/x86/kernel/cpu/perf_event_intel.c @@ -1545,6 +1545,13 @@ static void intel_clovertown_quirks(void) x86_pmu.pebs_constraints = NULL; } +static void intel_sandybridge_quirks(void) +{ + printk(KERN_WARNING "PEBS disabled due to CPU errata.\n"); + x86_pmu.pebs = 0; + x86_pmu.pebs_constraints = NULL; +} + __init int intel_pmu_init(void) { union cpuid10_edx edx; @@ -1694,6 +1701,7 @@ __init int intel_pmu_init(void) break; case 42: /* SandyBridge */ + x86_pmu.quirks = intel_sandybridge_quirks; case 45: /* SandyBridge, "Romely-EP" */ memcpy(hw_cache_event_ids, snb_hw_cache_event_ids, sizeof(hw_cache_event_ids)); -- cgit v1.2.3 From ffb871bc9156ee2e5cf442f61250c5bd6aad17e3 Mon Sep 17 00:00:00 2001 From: Gleb Natapov Date: Thu, 10 Nov 2011 14:57:26 +0200 Subject: x86, perf: Disable non available architectural events Intel CPUs report non-available architectural events in cpuid leaf 0AH.EBX. Use it to disable events that are not available according to CPU. Signed-off-by: Gleb Natapov Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/1320929850-10480-7-git-send-email-gleb@redhat.com Signed-off-by: Ingo Molnar --- arch/x86/include/asm/perf_event.h | 14 ++++++++++++++ arch/x86/kernel/cpu/perf_event.h | 5 +++++ arch/x86/kernel/cpu/perf_event_intel.c | 28 +++++++++++++++++++++++----- 3 files changed, 42 insertions(+), 5 deletions(-) (limited to 'arch/x86/kernel/cpu/perf_event_intel.c') diff --git a/arch/x86/include/asm/perf_event.h b/arch/x86/include/asm/perf_event.h index f61c62f7d5d8..c6998bc75456 100644 --- a/arch/x86/include/asm/perf_event.h +++ b/arch/x86/include/asm/perf_event.h @@ -57,6 +57,7 @@ (1 << (ARCH_PERFMON_UNHALTED_CORE_CYCLES_INDEX)) #define ARCH_PERFMON_BRANCH_MISSES_RETIRED 6 +#define ARCH_PERFMON_EVENTS_COUNT 7 /* * Intel "Architectural Performance Monitoring" CPUID @@ -72,6 +73,19 @@ union cpuid10_eax { unsigned int full; }; +union cpuid10_ebx { + struct { + unsigned int no_unhalted_core_cycles:1; + unsigned int no_instructions_retired:1; + unsigned int no_unhalted_reference_cycles:1; + unsigned int no_llc_reference:1; + unsigned int no_llc_misses:1; + unsigned int no_branch_instruction_retired:1; + unsigned int no_branch_misses_retired:1; + } split; + unsigned int full; +}; + union cpuid10_edx { struct { unsigned int num_counters_fixed:5; diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h index 51a985cbc12f..f49c5c21085c 100644 --- a/arch/x86/kernel/cpu/perf_event.h +++ b/arch/x86/kernel/cpu/perf_event.h @@ -285,6 +285,11 @@ struct x86_pmu { int num_counters_fixed; int cntval_bits; u64 cntval_mask; + union { + unsigned long events_maskl; + unsigned long events_mask[BITS_TO_LONGS(ARCH_PERFMON_EVENTS_COUNT)]; + }; + int events_mask_len; int apic; u64 max_period; struct event_constraint * diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c index 8d601b18bf9f..201156b80a37 100644 --- a/arch/x86/kernel/cpu/perf_event_intel.c +++ b/arch/x86/kernel/cpu/perf_event_intel.c @@ -1552,13 +1552,23 @@ static void intel_sandybridge_quirks(void) x86_pmu.pebs_constraints = NULL; } +static const int intel_event_id_to_hw_id[] __initconst = { + PERF_COUNT_HW_CPU_CYCLES, + PERF_COUNT_HW_INSTRUCTIONS, + PERF_COUNT_HW_BUS_CYCLES, + PERF_COUNT_HW_CACHE_REFERENCES, + PERF_COUNT_HW_CACHE_MISSES, + PERF_COUNT_HW_BRANCH_INSTRUCTIONS, + PERF_COUNT_HW_BRANCH_MISSES, +}; + __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; + int version, bit; if (!cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON)) { switch (boot_cpu_data.x86) { @@ -1574,8 +1584,8 @@ __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; @@ -1651,7 +1661,7 @@ __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) { + 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 @@ -1659,6 +1669,7 @@ __init int intel_pmu_init(void) * architectural event which is often completely bogus: */ intel_perfmon_event_map[PERF_COUNT_HW_BRANCH_MISSES] = 0x7f89; + ebx.split.no_branch_misses_retired = 0; pr_cont("erratum AAJ80 worked around, "); } @@ -1738,5 +1749,12 @@ __init int intel_pmu_init(void) break; } } + x86_pmu.events_maskl = ebx.full; + x86_pmu.events_mask_len = eax.split.mask_length; + + /* disable event that reported as not presend by cpuid */ + for_each_set_bit(bit, x86_pmu.events_mask, ARRAY_SIZE(intel_event_id_to_hw_id)) + intel_perfmon_event_map[intel_event_id_to_hw_id[bit]] = 0; + return 0; } -- cgit v1.2.3 From c1d6f42f1a42c721513e2f388c208e5348004f64 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Tue, 6 Dec 2011 14:07:15 +0100 Subject: perf, x86: Implement arch event mask as quirk Implement the disabling of arch events as a quirk so that we can print a message along with it. This creates some visibility into the problem space and could allow us to work on adding more work-around like the AAJ80 one. Requested-by: Ingo Molnar Cc: Gleb Natapov Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/n/tip-wcja2z48wklzu1b0nkz0a5y7@git.kernel.org Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event.c | 5 ++- arch/x86/kernel/cpu/perf_event.h | 16 ++++++- arch/x86/kernel/cpu/perf_event_intel.c | 80 +++++++++++++++++++++------------- 3 files changed, 68 insertions(+), 33 deletions(-) (limited to 'arch/x86/kernel/cpu/perf_event_intel.c') diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index 66f8ba9a67f9..55889e0b1452 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -1248,6 +1248,7 @@ static void __init pmu_check_apic(void) static int __init init_hw_perf_events(void) { + struct x86_pmu_quirk *quirk; struct event_constraint *c; int err; @@ -1276,8 +1277,8 @@ static int __init init_hw_perf_events(void) pr_cont("%s PMU driver.\n", x86_pmu.name); - if (x86_pmu.quirks) - x86_pmu.quirks(); + for (quirk = x86_pmu.quirks; quirk; quirk = quirk->next) + quirk->func(); if (x86_pmu.num_counters > X86_PMC_MAX_GENERIC) { WARN(1, KERN_ERR "hw perf events %d > max(%d), clipping!", diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h index f49c5c21085c..8944062f46e2 100644 --- a/arch/x86/kernel/cpu/perf_event.h +++ b/arch/x86/kernel/cpu/perf_event.h @@ -261,6 +261,11 @@ union perf_capabilities { u64 capabilities; }; +struct x86_pmu_quirk { + struct x86_pmu_quirk *next; + void (*func)(void); +}; + /* * struct x86_pmu - generic x86 pmu */ @@ -299,7 +304,7 @@ struct x86_pmu { void (*put_event_constraints)(struct cpu_hw_events *cpuc, struct perf_event *event); struct event_constraint *event_constraints; - void (*quirks)(void); + struct x86_pmu_quirk *quirks; int perfctr_second_write; int (*cpu_prepare)(int cpu); @@ -340,6 +345,15 @@ struct x86_pmu { struct perf_guest_switch_msr *(*guest_get_msrs)(int *nr); }; +#define x86_add_quirk(func_) \ +do { \ + static struct x86_pmu_quirk __quirk __initdata = { \ + .func = func_, \ + }; \ + __quirk.next = x86_pmu.quirks; \ + x86_pmu.quirks = &__quirk; \ +} while (0) + #define ERF_NO_HT_SHARING 1 #define ERF_HAS_RSP_1 2 diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c index 201156b80a37..2c3bf53d0302 100644 --- a/arch/x86/kernel/cpu/perf_event_intel.c +++ b/arch/x86/kernel/cpu/perf_event_intel.c @@ -1519,7 +1519,7 @@ 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: @@ -1545,30 +1545,61 @@ 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 int intel_event_id_to_hw_id[] __initconst = { - PERF_COUNT_HW_CPU_CYCLES, - PERF_COUNT_HW_INSTRUCTIONS, - PERF_COUNT_HW_BUS_CYCLES, - PERF_COUNT_HW_CACHE_REFERENCES, - PERF_COUNT_HW_CACHE_MISSES, - PERF_COUNT_HW_BRANCH_INSTRUCTIONS, - PERF_COUNT_HW_BRANCH_MISSES, +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; - int version, bit; + int version; if (!cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON)) { switch (boot_cpu_data.x86) { @@ -1599,6 +1630,9 @@ __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: @@ -1618,6 +1652,8 @@ __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: */ @@ -1627,7 +1663,7 @@ __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" */ @@ -1661,18 +1697,8 @@ __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.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_add_quirk(intel_nehalem_quirk); - pr_cont("erratum AAJ80 worked around, "); - } pr_cont("Nehalem events, "); break; @@ -1712,7 +1738,7 @@ __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)); @@ -1749,12 +1775,6 @@ __init int intel_pmu_init(void) break; } } - x86_pmu.events_maskl = ebx.full; - x86_pmu.events_mask_len = eax.split.mask_length; - - /* disable event that reported as not presend by cpuid */ - for_each_set_bit(bit, x86_pmu.events_mask, ARRAY_SIZE(intel_event_id_to_hw_id)) - intel_perfmon_event_map[intel_event_id_to_hw_id[bit]] = 0; return 0; } -- cgit v1.2.3 From cd09c0c40a971549800ce6a7e53c63f5139dd175 Mon Sep 17 00:00:00 2001 From: Stephane Eranian Date: Sun, 11 Dec 2011 00:28:51 +0100 Subject: perf events: Enable raw event support for Intel unhalted_reference_cycles event This patch adds the encoding and definitions necessary for the unhalted_reference_cycles event avaialble since Intel Core 2 processors. Signed-off-by: Stephane Eranian Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/1323559734-3488-2-git-send-email-eranian@google.com Signed-off-by: Ingo Molnar --- arch/x86/include/asm/perf_event.h | 15 ++++++++------- arch/x86/kernel/cpu/perf_event.c | 8 +++++++- arch/x86/kernel/cpu/perf_event_intel.c | 15 +++++---------- 3 files changed, 20 insertions(+), 18 deletions(-) (limited to 'arch/x86/kernel/cpu/perf_event_intel.c') diff --git a/arch/x86/include/asm/perf_event.h b/arch/x86/include/asm/perf_event.h index b50e9d15aae0..096c975e099f 100644 --- a/arch/x86/include/asm/perf_event.h +++ b/arch/x86/include/asm/perf_event.h @@ -112,23 +112,24 @@ struct x86_pmu_capability { /* * All 3 fixed-mode PMCs are configured via this single MSR: */ -#define MSR_ARCH_PERFMON_FIXED_CTR_CTRL 0x38d +#define MSR_ARCH_PERFMON_FIXED_CTR_CTRL 0x38d /* * The counts are available in three separate MSRs: */ /* Instr_Retired.Any: */ -#define MSR_ARCH_PERFMON_FIXED_CTR0 0x309 -#define X86_PMC_IDX_FIXED_INSTRUCTIONS (X86_PMC_IDX_FIXED + 0) +#define MSR_ARCH_PERFMON_FIXED_CTR0 0x309 +#define X86_PMC_IDX_FIXED_INSTRUCTIONS (X86_PMC_IDX_FIXED + 0) /* CPU_CLK_Unhalted.Core: */ -#define MSR_ARCH_PERFMON_FIXED_CTR1 0x30a -#define X86_PMC_IDX_FIXED_CPU_CYCLES (X86_PMC_IDX_FIXED + 1) +#define MSR_ARCH_PERFMON_FIXED_CTR1 0x30a +#define X86_PMC_IDX_FIXED_CPU_CYCLES (X86_PMC_IDX_FIXED + 1) /* CPU_CLK_Unhalted.Ref: */ -#define MSR_ARCH_PERFMON_FIXED_CTR2 0x30b -#define X86_PMC_IDX_FIXED_BUS_CYCLES (X86_PMC_IDX_FIXED + 2) +#define MSR_ARCH_PERFMON_FIXED_CTR2 0x30b +#define X86_PMC_IDX_FIXED_REF_CYCLES (X86_PMC_IDX_FIXED + 2) +#define X86_PMC_MSK_FIXED_REF_CYCLES (1ULL << X86_PMC_IDX_FIXED_REF_CYCLES) /* * We model BTS tracing as another fixed-mode PMC. diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index 930fe4879542..5adce1040b11 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -1304,9 +1304,15 @@ static int __init init_hw_perf_events(void) 0, x86_pmu.num_counters, 0); if (x86_pmu.event_constraints) { + /* + * event on fixed counter2 (REF_CYCLES) only works on this + * counter, so do not extend mask to generic counters + */ for_each_event_constraint(c, x86_pmu.event_constraints) { - if (c->cmask != X86_RAW_EVENT_MASK) + if (c->cmask != X86_RAW_EVENT_MASK + || c->idxmsk64 == X86_PMC_MSK_FIXED_REF_CYCLES) { continue; + } c->idxmsk64 |= (1ULL << x86_pmu.num_counters) - 1; c->weight += x86_pmu.num_counters; diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c index 2c3bf53d0302..61f865f947b3 100644 --- a/arch/x86/kernel/cpu/perf_event_intel.c +++ b/arch/x86/kernel/cpu/perf_event_intel.c @@ -45,12 +45,7 @@ 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 +63,7 @@ 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 +85,7 @@ 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 +97,7 @@ 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 +120,7 @@ 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 }; -- cgit v1.2.3 From 9c1497ea591b25d491f8e795f90a1405100b75ef Mon Sep 17 00:00:00 2001 From: Stephane Eranian Date: Sun, 11 Dec 2011 00:28:53 +0100 Subject: perf events: Add Intel x86 mapping for PERF_COUNT_HW_REF_CPU_CYCLES Add event maps for Intel x86 processors (with architected PMU v2 or later). On AMD, there is frequency scaling but no Turbo. There is no core cycle event not subject to frequency scaling, therefore we do not provide a mapping. Signed-off-by: Stephane Eranian Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/1323559734-3488-4-git-send-email-eranian@google.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event_intel.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch/x86/kernel/cpu/perf_event_intel.c') diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c index 61f865f947b3..cbfaaa2475ea 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 = -- cgit v1.2.3 From 2e64694de21a812d637dcbea4471ad1f7897b049 Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Fri, 23 Dec 2011 14:24:25 +0100 Subject: perf/x86: Fix raw_spin_unlock_irqrestore() usage Use raw_spin_unlock_irqrestore() as equivalent to raw_spin_lock_irqsave(). Signed-off-by: Robert Richter Cc: Stephane Eranian Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1324646665-13334-1-git-send-email-robert.richter@amd.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event_intel.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86/kernel/cpu/perf_event_intel.c') diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c index 8d601b18bf9f..121f1be4da19 100644 --- a/arch/x86/kernel/cpu/perf_event_intel.c +++ b/arch/x86/kernel/cpu/perf_event_intel.c @@ -1169,7 +1169,7 @@ again: */ c = &unconstrained; } else if (intel_try_alt_er(event, orig_idx)) { - raw_spin_unlock(&era->lock); + raw_spin_unlock_irqrestore(&era->lock, flags); goto again; } raw_spin_unlock_irqrestore(&era->lock, flags); -- cgit v1.2.3