diff options
Diffstat (limited to 'arch/x86/kernel/cpu/perf_event.c')
| -rw-r--r-- | arch/x86/kernel/cpu/perf_event.c | 3274 |
1 files changed, 1488 insertions, 1786 deletions
diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index d616c06e99b..2879ecdaac4 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -7,6 +7,7 @@ * Copyright (C) 2009 Advanced Micro Devices, Inc., Robert Richter * Copyright (C) 2008-2009 Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com> * Copyright (C) 2009 Intel Corporation, <markus.t.metzger@intel.com> + * Copyright (C) 2009 Google, Inc., Stephane Eranian * * For licencing details see kernel-base/COPYING */ @@ -20,663 +21,51 @@ #include <linux/kdebug.h> #include <linux/sched.h> #include <linux/uaccess.h> -#include <linux/highmem.h> +#include <linux/slab.h> #include <linux/cpu.h> +#include <linux/bitops.h> +#include <linux/device.h> #include <asm/apic.h> #include <asm/stacktrace.h> #include <asm/nmi.h> +#include <asm/smp.h> +#include <asm/alternative.h> +#include <asm/timer.h> +#include <asm/desc.h> +#include <asm/ldt.h> -static u64 perf_event_mask __read_mostly; +#include "perf_event.h" -/* The maximal number of PEBS events: */ -#define MAX_PEBS_EVENTS 4 +struct x86_pmu x86_pmu __read_mostly; -/* The size of a BTS record in bytes: */ -#define BTS_RECORD_SIZE 24 - -/* The size of a per-cpu BTS buffer in bytes: */ -#define BTS_BUFFER_SIZE (BTS_RECORD_SIZE * 2048) - -/* The BTS overflow threshold in bytes from the end of the buffer: */ -#define BTS_OVFL_TH (BTS_RECORD_SIZE * 128) - - -/* - * Bits in the debugctlmsr controlling branch tracing. - */ -#define X86_DEBUGCTL_TR (1 << 6) -#define X86_DEBUGCTL_BTS (1 << 7) -#define X86_DEBUGCTL_BTINT (1 << 8) -#define X86_DEBUGCTL_BTS_OFF_OS (1 << 9) -#define X86_DEBUGCTL_BTS_OFF_USR (1 << 10) - -/* - * A debug store configuration. - * - * We only support architectures that use 64bit fields. - */ -struct debug_store { - u64 bts_buffer_base; - u64 bts_index; - u64 bts_absolute_maximum; - u64 bts_interrupt_threshold; - u64 pebs_buffer_base; - u64 pebs_index; - u64 pebs_absolute_maximum; - u64 pebs_interrupt_threshold; - u64 pebs_event_reset[MAX_PEBS_EVENTS]; -}; - -struct cpu_hw_events { - struct perf_event *events[X86_PMC_IDX_MAX]; - unsigned long used_mask[BITS_TO_LONGS(X86_PMC_IDX_MAX)]; - unsigned long active_mask[BITS_TO_LONGS(X86_PMC_IDX_MAX)]; - unsigned long interrupts; - int enabled; - struct debug_store *ds; -}; - -struct event_constraint { - unsigned long idxmsk[BITS_TO_LONGS(X86_PMC_IDX_MAX)]; - int code; -}; - -#define EVENT_CONSTRAINT(c, m) { .code = (c), .idxmsk[0] = (m) } -#define EVENT_CONSTRAINT_END { .code = 0, .idxmsk[0] = 0 } - -#define for_each_event_constraint(e, c) \ - for ((e) = (c); (e)->idxmsk[0]; (e)++) - - -/* - * struct x86_pmu - generic x86 pmu - */ -struct x86_pmu { - const char *name; - int version; - int (*handle_irq)(struct pt_regs *); - void (*disable_all)(void); - void (*enable_all)(void); - void (*enable)(struct hw_perf_event *, int); - void (*disable)(struct hw_perf_event *, int); - unsigned eventsel; - unsigned perfctr; - u64 (*event_map)(int); - u64 (*raw_event)(u64); - int max_events; - int num_events; - int num_events_fixed; - int event_bits; - u64 event_mask; - int apic; - u64 max_period; - u64 intel_ctrl; - void (*enable_bts)(u64 config); - void (*disable_bts)(void); - int (*get_event_idx)(struct cpu_hw_events *cpuc, - struct hw_perf_event *hwc); -}; - -static struct x86_pmu x86_pmu __read_mostly; - -static DEFINE_PER_CPU(struct cpu_hw_events, cpu_hw_events) = { +DEFINE_PER_CPU(struct cpu_hw_events, cpu_hw_events) = { .enabled = 1, }; -static const struct event_constraint *event_constraints; - -/* - * Not sure about some of these - */ -static const u64 p6_perfmon_event_map[] = -{ - [PERF_COUNT_HW_CPU_CYCLES] = 0x0079, - [PERF_COUNT_HW_INSTRUCTIONS] = 0x00c0, - [PERF_COUNT_HW_CACHE_REFERENCES] = 0x0f2e, - [PERF_COUNT_HW_CACHE_MISSES] = 0x012e, - [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = 0x00c4, - [PERF_COUNT_HW_BRANCH_MISSES] = 0x00c5, - [PERF_COUNT_HW_BUS_CYCLES] = 0x0062, -}; - -static u64 p6_pmu_event_map(int hw_event) -{ - return p6_perfmon_event_map[hw_event]; -} - -/* - * Event setting that is specified not to count anything. - * We use this to effectively disable a counter. - * - * L2_RQSTS with 0 MESI unit mask. - */ -#define P6_NOP_EVENT 0x0000002EULL - -static u64 p6_pmu_raw_event(u64 hw_event) -{ -#define P6_EVNTSEL_EVENT_MASK 0x000000FFULL -#define P6_EVNTSEL_UNIT_MASK 0x0000FF00ULL -#define P6_EVNTSEL_EDGE_MASK 0x00040000ULL -#define P6_EVNTSEL_INV_MASK 0x00800000ULL -#define P6_EVNTSEL_REG_MASK 0xFF000000ULL - -#define P6_EVNTSEL_MASK \ - (P6_EVNTSEL_EVENT_MASK | \ - P6_EVNTSEL_UNIT_MASK | \ - P6_EVNTSEL_EDGE_MASK | \ - P6_EVNTSEL_INV_MASK | \ - P6_EVNTSEL_REG_MASK) - - return hw_event & P6_EVNTSEL_MASK; -} - -static const struct event_constraint intel_p6_event_constraints[] = -{ - EVENT_CONSTRAINT(0xc1, 0x1), /* FLOPS */ - EVENT_CONSTRAINT(0x10, 0x1), /* FP_COMP_OPS_EXE */ - EVENT_CONSTRAINT(0x11, 0x1), /* FP_ASSIST */ - EVENT_CONSTRAINT(0x12, 0x2), /* MUL */ - EVENT_CONSTRAINT(0x13, 0x2), /* DIV */ - EVENT_CONSTRAINT(0x14, 0x1), /* CYCLES_DIV_BUSY */ - EVENT_CONSTRAINT_END -}; - -/* - * Intel PerfMon v3. Used on Core2 and later. - */ -static const u64 intel_perfmon_event_map[] = -{ - [PERF_COUNT_HW_CPU_CYCLES] = 0x003c, - [PERF_COUNT_HW_INSTRUCTIONS] = 0x00c0, - [PERF_COUNT_HW_CACHE_REFERENCES] = 0x4f2e, - [PERF_COUNT_HW_CACHE_MISSES] = 0x412e, - [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = 0x00c4, - [PERF_COUNT_HW_BRANCH_MISSES] = 0x00c5, - [PERF_COUNT_HW_BUS_CYCLES] = 0x013c, -}; - -static const struct event_constraint intel_core_event_constraints[] = -{ - EVENT_CONSTRAINT(0x10, 0x1), /* FP_COMP_OPS_EXE */ - EVENT_CONSTRAINT(0x11, 0x2), /* FP_ASSIST */ - EVENT_CONSTRAINT(0x12, 0x2), /* MUL */ - EVENT_CONSTRAINT(0x13, 0x2), /* DIV */ - EVENT_CONSTRAINT(0x14, 0x1), /* CYCLES_DIV_BUSY */ - EVENT_CONSTRAINT(0x18, 0x1), /* IDLE_DURING_DIV */ - EVENT_CONSTRAINT(0x19, 0x2), /* DELAYED_BYPASS */ - EVENT_CONSTRAINT(0xa1, 0x1), /* RS_UOPS_DISPATCH_CYCLES */ - EVENT_CONSTRAINT(0xcb, 0x1), /* MEM_LOAD_RETIRED */ - EVENT_CONSTRAINT_END -}; - -static const struct event_constraint intel_nehalem_event_constraints[] = -{ - EVENT_CONSTRAINT(0x40, 0x3), /* L1D_CACHE_LD */ - EVENT_CONSTRAINT(0x41, 0x3), /* L1D_CACHE_ST */ - EVENT_CONSTRAINT(0x42, 0x3), /* L1D_CACHE_LOCK */ - EVENT_CONSTRAINT(0x43, 0x3), /* L1D_ALL_REF */ - EVENT_CONSTRAINT(0x4e, 0x3), /* L1D_PREFETCH */ - EVENT_CONSTRAINT(0x4c, 0x3), /* LOAD_HIT_PRE */ - EVENT_CONSTRAINT(0x51, 0x3), /* L1D */ - EVENT_CONSTRAINT(0x52, 0x3), /* L1D_CACHE_PREFETCH_LOCK_FB_HIT */ - EVENT_CONSTRAINT(0x53, 0x3), /* L1D_CACHE_LOCK_FB_HIT */ - EVENT_CONSTRAINT(0xc5, 0x3), /* CACHE_LOCK_CYCLES */ - EVENT_CONSTRAINT_END -}; - -static u64 intel_pmu_event_map(int hw_event) -{ - return intel_perfmon_event_map[hw_event]; -} - -/* - * Generalized hw caching related hw_event table, filled - * in on a per model basis. A value of 0 means - * 'not supported', -1 means 'hw_event makes no sense on - * this CPU', any other value means the raw hw_event - * ID. - */ - -#define C(x) PERF_COUNT_HW_CACHE_##x - -static u64 __read_mostly hw_cache_event_ids +u64 __read_mostly hw_cache_event_ids [PERF_COUNT_HW_CACHE_MAX] [PERF_COUNT_HW_CACHE_OP_MAX] [PERF_COUNT_HW_CACHE_RESULT_MAX]; - -static __initconst u64 nehalem_hw_cache_event_ids +u64 __read_mostly hw_cache_extra_regs [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) ] = 0x0f40, /* L1D_CACHE_LD.MESI */ - [ C(RESULT_MISS) ] = 0x0140, /* L1D_CACHE_LD.I_STATE */ - }, - [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = 0x0f41, /* L1D_CACHE_ST.MESI */ - [ C(RESULT_MISS) ] = 0x0141, /* L1D_CACHE_ST.I_STATE */ - }, - [ 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) ] = 0x0f40, /* L1D_CACHE_LD.MESI (alias) */ - [ C(RESULT_MISS) ] = 0x0108, /* DTLB_LOAD_MISSES.ANY */ - }, - [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = 0x0f41, /* L1D_CACHE_ST.MESI (alias) */ - [ 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) ] = 0x20c8, /* ITLB_MISS_RETIRED */ - }, - [ 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 core2_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) ] = 0x0f40, /* L1D_CACHE_LD.MESI */ - [ C(RESULT_MISS) ] = 0x0140, /* L1D_CACHE_LD.I_STATE */ - }, - [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = 0x0f41, /* L1D_CACHE_ST.MESI */ - [ C(RESULT_MISS) ] = 0x0141, /* L1D_CACHE_ST.I_STATE */ - }, - [ C(OP_PREFETCH) ] = { - [ C(RESULT_ACCESS) ] = 0x104e, /* L1D_PREFETCH.REQUESTS */ - [ C(RESULT_MISS) ] = 0, - }, - }, - [ C(L1I ) ] = { - [ C(OP_READ) ] = { - [ C(RESULT_ACCESS) ] = 0x0080, /* L1I.READS */ - [ C(RESULT_MISS) ] = 0x0081, /* L1I.MISSES */ - }, - [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = -1, - [ C(RESULT_MISS) ] = -1, - }, - [ C(OP_PREFETCH) ] = { - [ C(RESULT_ACCESS) ] = 0, - [ C(RESULT_MISS) ] = 0, - }, - }, - [ C(LL ) ] = { - [ C(OP_READ) ] = { - [ C(RESULT_ACCESS) ] = 0x4f29, /* L2_LD.MESI */ - [ C(RESULT_MISS) ] = 0x4129, /* L2_LD.ISTATE */ - }, - [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = 0x4f2A, /* L2_ST.MESI */ - [ C(RESULT_MISS) ] = 0x412A, /* L2_ST.ISTATE */ - }, - [ C(OP_PREFETCH) ] = { - [ C(RESULT_ACCESS) ] = 0, - [ C(RESULT_MISS) ] = 0, - }, - }, - [ C(DTLB) ] = { - [ C(OP_READ) ] = { - [ C(RESULT_ACCESS) ] = 0x0f40, /* L1D_CACHE_LD.MESI (alias) */ - [ C(RESULT_MISS) ] = 0x0208, /* DTLB_MISSES.MISS_LD */ - }, - [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = 0x0f41, /* L1D_CACHE_ST.MESI (alias) */ - [ C(RESULT_MISS) ] = 0x0808, /* DTLB_MISSES.MISS_ST */ - }, - [ C(OP_PREFETCH) ] = { - [ C(RESULT_ACCESS) ] = 0, - [ C(RESULT_MISS) ] = 0, - }, - }, - [ C(ITLB) ] = { - [ C(OP_READ) ] = { - [ C(RESULT_ACCESS) ] = 0x00c0, /* INST_RETIRED.ANY_P */ - [ C(RESULT_MISS) ] = 0x1282, /* ITLBMISSES */ - }, - [ 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.ANY */ - [ C(RESULT_MISS) ] = 0x00c5, /* BP_INST_RETIRED.MISPRED */ - }, - [ 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 atom_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) ] = 0x2140, /* L1D_CACHE.LD */ - [ C(RESULT_MISS) ] = 0, - }, - [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = 0x2240, /* L1D_CACHE.ST */ - [ C(RESULT_MISS) ] = 0, - }, - [ C(OP_PREFETCH) ] = { - [ C(RESULT_ACCESS) ] = 0x0, - [ C(RESULT_MISS) ] = 0, - }, - }, - [ 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) ] = 0, - [ C(RESULT_MISS) ] = 0, - }, - }, - [ C(LL ) ] = { - [ C(OP_READ) ] = { - [ C(RESULT_ACCESS) ] = 0x4f29, /* L2_LD.MESI */ - [ C(RESULT_MISS) ] = 0x4129, /* L2_LD.ISTATE */ - }, - [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = 0x4f2A, /* L2_ST.MESI */ - [ C(RESULT_MISS) ] = 0x412A, /* L2_ST.ISTATE */ - }, - [ C(OP_PREFETCH) ] = { - [ C(RESULT_ACCESS) ] = 0, - [ C(RESULT_MISS) ] = 0, - }, - }, - [ C(DTLB) ] = { - [ C(OP_READ) ] = { - [ C(RESULT_ACCESS) ] = 0x2140, /* L1D_CACHE_LD.MESI (alias) */ - [ C(RESULT_MISS) ] = 0x0508, /* DTLB_MISSES.MISS_LD */ - }, - [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = 0x2240, /* L1D_CACHE_ST.MESI (alias) */ - [ C(RESULT_MISS) ] = 0x0608, /* DTLB_MISSES.MISS_ST */ - }, - [ C(OP_PREFETCH) ] = { - [ C(RESULT_ACCESS) ] = 0, - [ C(RESULT_MISS) ] = 0, - }, - }, - [ C(ITLB) ] = { - [ C(OP_READ) ] = { - [ C(RESULT_ACCESS) ] = 0x00c0, /* INST_RETIRED.ANY_P */ - [ C(RESULT_MISS) ] = 0x0282, /* ITLB.MISSES */ - }, - [ 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.ANY */ - [ C(RESULT_MISS) ] = 0x00c5, /* BP_INST_RETIRED.MISPRED */ - }, - [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = -1, - [ C(RESULT_MISS) ] = -1, - }, - [ C(OP_PREFETCH) ] = { - [ C(RESULT_ACCESS) ] = -1, - [ C(RESULT_MISS) ] = -1, - }, - }, -}; - -static u64 intel_pmu_raw_event(u64 hw_event) -{ -#define CORE_EVNTSEL_EVENT_MASK 0x000000FFULL -#define CORE_EVNTSEL_UNIT_MASK 0x0000FF00ULL -#define CORE_EVNTSEL_EDGE_MASK 0x00040000ULL -#define CORE_EVNTSEL_INV_MASK 0x00800000ULL -#define CORE_EVNTSEL_REG_MASK 0xFF000000ULL - -#define CORE_EVNTSEL_MASK \ - (CORE_EVNTSEL_EVENT_MASK | \ - CORE_EVNTSEL_UNIT_MASK | \ - CORE_EVNTSEL_EDGE_MASK | \ - CORE_EVNTSEL_INV_MASK | \ - CORE_EVNTSEL_REG_MASK) - - return hw_event & CORE_EVNTSEL_MASK; -} - -static __initconst u64 amd_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) ] = 0x0040, /* Data Cache Accesses */ - [ C(RESULT_MISS) ] = 0x0041, /* Data Cache Misses */ - }, - [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = 0x0142, /* Data Cache Refills :system */ - [ C(RESULT_MISS) ] = 0, - }, - [ C(OP_PREFETCH) ] = { - [ C(RESULT_ACCESS) ] = 0x0267, /* Data Prefetcher :attempts */ - [ C(RESULT_MISS) ] = 0x0167, /* Data Prefetcher :cancelled */ - }, - }, - [ C(L1I ) ] = { - [ C(OP_READ) ] = { - [ C(RESULT_ACCESS) ] = 0x0080, /* Instruction cache fetches */ - [ C(RESULT_MISS) ] = 0x0081, /* Instruction cache misses */ - }, - [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = -1, - [ C(RESULT_MISS) ] = -1, - }, - [ C(OP_PREFETCH) ] = { - [ C(RESULT_ACCESS) ] = 0x014B, /* Prefetch Instructions :Load */ - [ C(RESULT_MISS) ] = 0, - }, - }, - [ C(LL ) ] = { - [ C(OP_READ) ] = { - [ C(RESULT_ACCESS) ] = 0x037D, /* Requests to L2 Cache :IC+DC */ - [ C(RESULT_MISS) ] = 0x037E, /* L2 Cache Misses : IC+DC */ - }, - [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = 0x017F, /* L2 Fill/Writeback */ - [ C(RESULT_MISS) ] = 0, - }, - [ C(OP_PREFETCH) ] = { - [ C(RESULT_ACCESS) ] = 0, - [ C(RESULT_MISS) ] = 0, - }, - }, - [ C(DTLB) ] = { - [ C(OP_READ) ] = { - [ C(RESULT_ACCESS) ] = 0x0040, /* Data Cache Accesses */ - [ C(RESULT_MISS) ] = 0x0046, /* L1 DTLB and L2 DLTB Miss */ - }, - [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = 0, - [ C(RESULT_MISS) ] = 0, - }, - [ C(OP_PREFETCH) ] = { - [ C(RESULT_ACCESS) ] = 0, - [ C(RESULT_MISS) ] = 0, - }, - }, - [ C(ITLB) ] = { - [ C(OP_READ) ] = { - [ C(RESULT_ACCESS) ] = 0x0080, /* Instruction fecthes */ - [ C(RESULT_MISS) ] = 0x0085, /* Instr. fetch ITLB misses */ - }, - [ 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) ] = 0x00c2, /* Retired Branch Instr. */ - [ C(RESULT_MISS) ] = 0x00c3, /* Retired Mispredicted BI */ - }, - [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = -1, - [ C(RESULT_MISS) ] = -1, - }, - [ C(OP_PREFETCH) ] = { - [ C(RESULT_ACCESS) ] = -1, - [ C(RESULT_MISS) ] = -1, - }, - }, -}; - -/* - * AMD Performance Monitor K7 and later. - */ -static const u64 amd_perfmon_event_map[] = -{ - [PERF_COUNT_HW_CPU_CYCLES] = 0x0076, - [PERF_COUNT_HW_INSTRUCTIONS] = 0x00c0, - [PERF_COUNT_HW_CACHE_REFERENCES] = 0x0080, - [PERF_COUNT_HW_CACHE_MISSES] = 0x0081, - [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = 0x00c4, - [PERF_COUNT_HW_BRANCH_MISSES] = 0x00c5, -}; - -static u64 amd_pmu_event_map(int hw_event) -{ - return amd_perfmon_event_map[hw_event]; -} - -static u64 amd_pmu_raw_event(u64 hw_event) -{ -#define K7_EVNTSEL_EVENT_MASK 0x7000000FFULL -#define K7_EVNTSEL_UNIT_MASK 0x00000FF00ULL -#define K7_EVNTSEL_EDGE_MASK 0x000040000ULL -#define K7_EVNTSEL_INV_MASK 0x000800000ULL -#define K7_EVNTSEL_REG_MASK 0x0FF000000ULL - -#define K7_EVNTSEL_MASK \ - (K7_EVNTSEL_EVENT_MASK | \ - K7_EVNTSEL_UNIT_MASK | \ - K7_EVNTSEL_EDGE_MASK | \ - K7_EVNTSEL_INV_MASK | \ - K7_EVNTSEL_REG_MASK) - - return hw_event & K7_EVNTSEL_MASK; -} + [PERF_COUNT_HW_CACHE_RESULT_MAX]; /* * Propagate event elapsed time into the generic event. * Can only be executed on the CPU where the event is active. * Returns the delta events processed. */ -static u64 -x86_perf_event_update(struct perf_event *event, - struct hw_perf_event *hwc, int idx) +u64 x86_perf_event_update(struct perf_event *event) { - int shift = 64 - x86_pmu.event_bits; + struct hw_perf_event *hwc = &event->hw; + int shift = 64 - x86_pmu.cntval_bits; u64 prev_raw_count, new_raw_count; + int idx = hwc->idx; s64 delta; - if (idx == X86_PMC_IDX_FIXED_BTS) + if (idx == INTEL_PMC_IDX_FIXED_BTS) return 0; /* @@ -687,10 +76,10 @@ x86_perf_event_update(struct perf_event *event, * count to the generic event atomically: */ again: - prev_raw_count = atomic64_read(&hwc->prev_count); - rdmsrl(hwc->event_base + idx, new_raw_count); + prev_raw_count = local64_read(&hwc->prev_count); + rdpmcl(hwc->event_base_rdpmc, new_raw_count); - if (atomic64_cmpxchg(&hwc->prev_count, prev_raw_count, + if (local64_cmpxchg(&hwc->prev_count, prev_raw_count, new_raw_count) != prev_raw_count) goto again; @@ -705,173 +94,165 @@ again: delta = (new_raw_count << shift) - (prev_raw_count << shift); delta >>= shift; - atomic64_add(delta, &event->count); - atomic64_sub(delta, &hwc->period_left); + local64_add(delta, &event->count); + local64_sub(delta, &hwc->period_left); return new_raw_count; } +/* + * Find and validate any extra registers to set up. + */ +static int x86_pmu_extra_regs(u64 config, struct perf_event *event) +{ + struct hw_perf_event_extra *reg; + struct extra_reg *er; + + reg = &event->hw.extra_reg; + + if (!x86_pmu.extra_regs) + return 0; + + for (er = x86_pmu.extra_regs; er->msr; er++) { + if (er->event != (config & er->config_mask)) + continue; + if (event->attr.config1 & ~er->valid_mask) + return -EINVAL; + /* Check if the extra msrs can be safely accessed*/ + if (!er->extra_msr_access) + return -ENXIO; + + reg->idx = er->idx; + reg->config = event->attr.config1; + reg->reg = er->msr; + break; + } + return 0; +} + static atomic_t active_events; static DEFINE_MUTEX(pmc_reserve_mutex); +#ifdef CONFIG_X86_LOCAL_APIC + static bool reserve_pmc_hardware(void) { -#ifdef CONFIG_X86_LOCAL_APIC int i; - if (nmi_watchdog == NMI_LOCAL_APIC) - disable_lapic_nmi_watchdog(); - - for (i = 0; i < x86_pmu.num_events; i++) { - if (!reserve_perfctr_nmi(x86_pmu.perfctr + i)) + for (i = 0; i < x86_pmu.num_counters; i++) { + if (!reserve_perfctr_nmi(x86_pmu_event_addr(i))) goto perfctr_fail; } - for (i = 0; i < x86_pmu.num_events; i++) { - if (!reserve_evntsel_nmi(x86_pmu.eventsel + i)) + for (i = 0; i < x86_pmu.num_counters; i++) { + if (!reserve_evntsel_nmi(x86_pmu_config_addr(i))) goto eventsel_fail; } -#endif return true; -#ifdef CONFIG_X86_LOCAL_APIC eventsel_fail: for (i--; i >= 0; i--) - release_evntsel_nmi(x86_pmu.eventsel + i); + release_evntsel_nmi(x86_pmu_config_addr(i)); - i = x86_pmu.num_events; + i = x86_pmu.num_counters; perfctr_fail: for (i--; i >= 0; i--) - release_perfctr_nmi(x86_pmu.perfctr + i); - - if (nmi_watchdog == NMI_LOCAL_APIC) - enable_lapic_nmi_watchdog(); + release_perfctr_nmi(x86_pmu_event_addr(i)); return false; -#endif } static void release_pmc_hardware(void) { -#ifdef CONFIG_X86_LOCAL_APIC int i; - for (i = 0; i < x86_pmu.num_events; i++) { - release_perfctr_nmi(x86_pmu.perfctr + i); - release_evntsel_nmi(x86_pmu.eventsel + i); + for (i = 0; i < x86_pmu.num_counters; i++) { + release_perfctr_nmi(x86_pmu_event_addr(i)); + release_evntsel_nmi(x86_pmu_config_addr(i)); } - - if (nmi_watchdog == NMI_LOCAL_APIC) - enable_lapic_nmi_watchdog(); -#endif } -static inline bool bts_available(void) -{ - return x86_pmu.enable_bts != NULL; -} - -static inline void init_debug_store_on_cpu(int cpu) -{ - struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds; - - if (!ds) - return; - - wrmsr_on_cpu(cpu, MSR_IA32_DS_AREA, - (u32)((u64)(unsigned long)ds), - (u32)((u64)(unsigned long)ds >> 32)); -} +#else -static inline void fini_debug_store_on_cpu(int cpu) -{ - if (!per_cpu(cpu_hw_events, cpu).ds) - return; +static bool reserve_pmc_hardware(void) { return true; } +static void release_pmc_hardware(void) {} - wrmsr_on_cpu(cpu, MSR_IA32_DS_AREA, 0, 0); -} +#endif -static void release_bts_hardware(void) +static bool check_hw_exists(void) { - int cpu; - - if (!bts_available()) - return; - - get_online_cpus(); - - for_each_online_cpu(cpu) - fini_debug_store_on_cpu(cpu); - - for_each_possible_cpu(cpu) { - struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds; - - if (!ds) - continue; - - per_cpu(cpu_hw_events, cpu).ds = NULL; + u64 val, val_fail, val_new= ~0; + int i, reg, reg_fail, ret = 0; + int bios_fail = 0; - kfree((void *)(unsigned long)ds->bts_buffer_base); - kfree(ds); + /* + * Check to see if the BIOS enabled any of the counters, if so + * complain and bail. + */ + for (i = 0; i < x86_pmu.num_counters; i++) { + reg = x86_pmu_config_addr(i); + ret = rdmsrl_safe(reg, &val); + if (ret) + goto msr_fail; + if (val & ARCH_PERFMON_EVENTSEL_ENABLE) { + bios_fail = 1; + val_fail = val; + reg_fail = reg; + } } - put_online_cpus(); -} - -static int reserve_bts_hardware(void) -{ - int cpu, err = 0; - - if (!bts_available()) - return 0; - - get_online_cpus(); - - for_each_possible_cpu(cpu) { - struct debug_store *ds; - void *buffer; - - err = -ENOMEM; - buffer = kzalloc(BTS_BUFFER_SIZE, GFP_KERNEL); - if (unlikely(!buffer)) - break; - - ds = kzalloc(sizeof(*ds), GFP_KERNEL); - if (unlikely(!ds)) { - kfree(buffer); - break; + if (x86_pmu.num_counters_fixed) { + reg = MSR_ARCH_PERFMON_FIXED_CTR_CTRL; + ret = rdmsrl_safe(reg, &val); + if (ret) + goto msr_fail; + for (i = 0; i < x86_pmu.num_counters_fixed; i++) { + if (val & (0x03 << i*4)) { + bios_fail = 1; + val_fail = val; + reg_fail = reg; + } } + } - ds->bts_buffer_base = (u64)(unsigned long)buffer; - ds->bts_index = ds->bts_buffer_base; - ds->bts_absolute_maximum = - ds->bts_buffer_base + BTS_BUFFER_SIZE; - ds->bts_interrupt_threshold = - ds->bts_absolute_maximum - BTS_OVFL_TH; + /* + * Read the current value, change it and read it back to see if it + * matches, this is needed to detect certain hardware emulators + * (qemu/kvm) that don't trap on the MSR access and always return 0s. + */ + reg = x86_pmu_event_addr(0); + if (rdmsrl_safe(reg, &val)) + goto msr_fail; + val ^= 0xffffUL; + ret = wrmsrl_safe(reg, val); + ret |= rdmsrl_safe(reg, &val_new); + if (ret || val != val_new) + goto msr_fail; - per_cpu(cpu_hw_events, cpu).ds = ds; - err = 0; + /* + * We still allow the PMU driver to operate: + */ + if (bios_fail) { + printk(KERN_CONT "Broken BIOS detected, complain to your hardware vendor.\n"); + printk(KERN_ERR FW_BUG "the BIOS has corrupted hw-PMU resources (MSR %x is %Lx)\n", reg_fail, val_fail); } - if (err) - release_bts_hardware(); - else { - for_each_online_cpu(cpu) - init_debug_store_on_cpu(cpu); - } + return true; - put_online_cpus(); +msr_fail: + printk(KERN_CONT "Broken PMU hardware detected, using software events only.\n"); + printk(KERN_ERR "Failed to access perfctr msr (MSR %x is %Lx)\n", reg, val_new); - return err; + return false; } static void hw_perf_event_destroy(struct perf_event *event) { if (atomic_dec_and_mutex_lock(&active_events, &pmc_reserve_mutex)) { release_pmc_hardware(); - release_bts_hardware(); + release_ds_buffers(); mutex_unlock(&pmc_reserve_mutex); } } @@ -882,8 +263,9 @@ static inline int x86_pmu_initialized(void) } static inline int -set_ext_hw_attr(struct hw_perf_event *hwc, struct perf_event_attr *attr) +set_ext_hw_attr(struct hw_perf_event *hwc, struct perf_event *event) { + struct perf_event_attr *attr = &event->attr; unsigned int cache_type, cache_op, cache_result; u64 config, val; @@ -910,54 +292,165 @@ set_ext_hw_attr(struct hw_perf_event *hwc, struct perf_event_attr *attr) return -EINVAL; hwc->config |= val; + attr->config1 = hw_cache_extra_regs[cache_type][cache_op][cache_result]; + return x86_pmu_extra_regs(val, event); +} + +int x86_setup_perfctr(struct perf_event *event) +{ + struct perf_event_attr *attr = &event->attr; + struct hw_perf_event *hwc = &event->hw; + u64 config; + + if (!is_sampling_event(event)) { + hwc->sample_period = x86_pmu.max_period; + hwc->last_period = hwc->sample_period; + local64_set(&hwc->period_left, hwc->sample_period); + } + + if (attr->type == PERF_TYPE_RAW) + return x86_pmu_extra_regs(event->attr.config, event); + + if (attr->type == PERF_TYPE_HW_CACHE) + return set_ext_hw_attr(hwc, event); + + if (attr->config >= x86_pmu.max_events) + return -EINVAL; + + /* + * The generic map: + */ + config = x86_pmu.event_map(attr->config); + + if (config == 0) + return -ENOENT; + + if (config == -1LL) + return -EINVAL; + + /* + * Branch tracing: + */ + if (attr->config == PERF_COUNT_HW_BRANCH_INSTRUCTIONS && + !attr->freq && hwc->sample_period == 1) { + /* BTS is not supported by this architecture. */ + if (!x86_pmu.bts_active) + return -EOPNOTSUPP; + + /* BTS is currently only allowed for user-mode. */ + if (!attr->exclude_kernel) + return -EOPNOTSUPP; + } + + hwc->config |= config; return 0; } -static void intel_pmu_enable_bts(u64 config) +/* + * check that branch_sample_type is compatible with + * settings needed for precise_ip > 1 which implies + * using the LBR to capture ALL taken branches at the + * priv levels of the measurement + */ +static inline int precise_br_compat(struct perf_event *event) { - unsigned long debugctlmsr; + u64 m = event->attr.branch_sample_type; + u64 b = 0; - debugctlmsr = get_debugctlmsr(); + /* must capture all branches */ + if (!(m & PERF_SAMPLE_BRANCH_ANY)) + return 0; - debugctlmsr |= X86_DEBUGCTL_TR; - debugctlmsr |= X86_DEBUGCTL_BTS; - debugctlmsr |= X86_DEBUGCTL_BTINT; + m &= PERF_SAMPLE_BRANCH_KERNEL | PERF_SAMPLE_BRANCH_USER; - if (!(config & ARCH_PERFMON_EVENTSEL_OS)) - debugctlmsr |= X86_DEBUGCTL_BTS_OFF_OS; + if (!event->attr.exclude_user) + b |= PERF_SAMPLE_BRANCH_USER; - if (!(config & ARCH_PERFMON_EVENTSEL_USR)) - debugctlmsr |= X86_DEBUGCTL_BTS_OFF_USR; + if (!event->attr.exclude_kernel) + b |= PERF_SAMPLE_BRANCH_KERNEL; - update_debugctlmsr(debugctlmsr); + /* + * ignore PERF_SAMPLE_BRANCH_HV, not supported on x86 + */ + + return m == b; } -static void intel_pmu_disable_bts(void) +int x86_pmu_hw_config(struct perf_event *event) { - struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); - unsigned long debugctlmsr; + if (event->attr.precise_ip) { + int precise = 0; - if (!cpuc->ds) - return; + /* Support for constant skid */ + if (x86_pmu.pebs_active && !x86_pmu.pebs_broken) { + precise++; + + /* Support for IP fixup */ + if (x86_pmu.lbr_nr) + precise++; + } - debugctlmsr = get_debugctlmsr(); + if (event->attr.precise_ip > precise) + return -EOPNOTSUPP; + /* + * check that PEBS LBR correction does not conflict with + * whatever the user is asking with attr->branch_sample_type + */ + if (event->attr.precise_ip > 1 && + x86_pmu.intel_cap.pebs_format < 2) { + u64 *br_type = &event->attr.branch_sample_type; + + if (has_branch_stack(event)) { + if (!precise_br_compat(event)) + return -EOPNOTSUPP; + + /* branch_sample_type is compatible */ + + } else { + /* + * user did not specify branch_sample_type + * + * For PEBS fixups, we capture all + * the branches at the priv level of the + * event. + */ + *br_type = PERF_SAMPLE_BRANCH_ANY; + + if (!event->attr.exclude_user) + *br_type |= PERF_SAMPLE_BRANCH_USER; + + if (!event->attr.exclude_kernel) + *br_type |= PERF_SAMPLE_BRANCH_KERNEL; + } + } + } - debugctlmsr &= - ~(X86_DEBUGCTL_TR | X86_DEBUGCTL_BTS | X86_DEBUGCTL_BTINT | - X86_DEBUGCTL_BTS_OFF_OS | X86_DEBUGCTL_BTS_OFF_USR); + /* + * Generate PMC IRQs: + * (keep 'enabled' bit clear for now) + */ + event->hw.config = ARCH_PERFMON_EVENTSEL_INT; - update_debugctlmsr(debugctlmsr); + /* + * Count user and OS events unless requested not to + */ + if (!event->attr.exclude_user) + event->hw.config |= ARCH_PERFMON_EVENTSEL_USR; + if (!event->attr.exclude_kernel) + event->hw.config |= ARCH_PERFMON_EVENTSEL_OS; + + if (event->attr.type == PERF_TYPE_RAW) + event->hw.config |= event->attr.config & X86_RAW_EVENT_MASK; + + return x86_setup_perfctr(event); } /* * Setup the hardware configuration for a given attr_type */ -static int __hw_perf_event_init(struct perf_event *event) +static int __x86_pmu_event_init(struct perf_event *event) { - struct perf_event_attr *attr = &event->attr; - struct hw_perf_event *hwc = &event->hw; - u64 config; int err; if (!x86_pmu_initialized()) @@ -970,7 +463,7 @@ static int __hw_perf_event_init(struct perf_event *event) if (!reserve_pmc_hardware()) err = -EBUSY; else - err = reserve_bts_hardware(); + reserve_ds_buffers(); } if (!err) atomic_inc(&active_events); @@ -981,291 +474,469 @@ static int __hw_perf_event_init(struct perf_event *event) event->destroy = hw_perf_event_destroy; - /* - * Generate PMC IRQs: - * (keep 'enabled' bit clear for now) - */ - hwc->config = ARCH_PERFMON_EVENTSEL_INT; - - hwc->idx = -1; - - /* - * Count user and OS events unless requested not to. - */ - if (!attr->exclude_user) - hwc->config |= ARCH_PERFMON_EVENTSEL_USR; - if (!attr->exclude_kernel) - hwc->config |= ARCH_PERFMON_EVENTSEL_OS; - - if (!hwc->sample_period) { - hwc->sample_period = x86_pmu.max_period; - hwc->last_period = hwc->sample_period; - atomic64_set(&hwc->period_left, hwc->sample_period); - } else { - /* - * If we have a PMU initialized but no APIC - * interrupts, we cannot sample hardware - * events (user-space has to fall back and - * sample via a hrtimer based software event): - */ - if (!x86_pmu.apic) - return -EOPNOTSUPP; - } - - /* - * Raw hw_event type provide the config in the hw_event structure - */ - if (attr->type == PERF_TYPE_RAW) { - hwc->config |= x86_pmu.raw_event(attr->config); - return 0; - } - - if (attr->type == PERF_TYPE_HW_CACHE) - return set_ext_hw_attr(hwc, attr); + event->hw.idx = -1; + event->hw.last_cpu = -1; + event->hw.last_tag = ~0ULL; - if (attr->config >= x86_pmu.max_events) - return -EINVAL; + /* mark unused */ + event->hw.extra_reg.idx = EXTRA_REG_NONE; + event->hw.branch_reg.idx = EXTRA_REG_NONE; - /* - * The generic map: - */ - config = x86_pmu.event_map(attr->config); - - if (config == 0) - return -ENOENT; - - if (config == -1LL) - return -EINVAL; - - /* - * Branch tracing: - */ - if ((attr->config == PERF_COUNT_HW_BRANCH_INSTRUCTIONS) && - (hwc->sample_period == 1)) { - /* BTS is not supported by this architecture. */ - if (!bts_available()) - return -EOPNOTSUPP; - - /* BTS is currently only allowed for user-mode. */ - if (hwc->config & ARCH_PERFMON_EVENTSEL_OS) - return -EOPNOTSUPP; - } - - hwc->config |= config; - - return 0; + return x86_pmu.hw_config(event); } -static void p6_pmu_disable_all(void) +void x86_pmu_disable_all(void) { struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); - u64 val; - - if (!cpuc->enabled) - return; + int idx; - cpuc->enabled = 0; - barrier(); + for (idx = 0; idx < x86_pmu.num_counters; idx++) { + u64 val; - /* p6 only has one enable register */ - rdmsrl(MSR_P6_EVNTSEL0, val); - val &= ~ARCH_PERFMON_EVENTSEL0_ENABLE; - wrmsrl(MSR_P6_EVNTSEL0, val); + if (!test_bit(idx, cpuc->active_mask)) + continue; + rdmsrl(x86_pmu_config_addr(idx), val); + if (!(val & ARCH_PERFMON_EVENTSEL_ENABLE)) + continue; + val &= ~ARCH_PERFMON_EVENTSEL_ENABLE; + wrmsrl(x86_pmu_config_addr(idx), val); + } } -static void intel_pmu_disable_all(void) +static void x86_pmu_disable(struct pmu *pmu) { struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + if (!x86_pmu_initialized()) + return; + if (!cpuc->enabled) return; + cpuc->n_added = 0; 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(); + x86_pmu.disable_all(); } -static void amd_pmu_disable_all(void) +void x86_pmu_enable_all(int added) { 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; + for (idx = 0; idx < x86_pmu.num_counters; idx++) { + struct hw_perf_event *hwc = &cpuc->events[idx]->hw; if (!test_bit(idx, cpuc->active_mask)) continue; - rdmsrl(MSR_K7_EVNTSEL0 + idx, val); - if (!(val & ARCH_PERFMON_EVENTSEL0_ENABLE)) - continue; - val &= ~ARCH_PERFMON_EVENTSEL0_ENABLE; - wrmsrl(MSR_K7_EVNTSEL0 + idx, val); + + __x86_pmu_enable_event(hwc, ARCH_PERFMON_EVENTSEL_ENABLE); } } -void hw_perf_disable(void) +static struct pmu pmu; + +static inline int is_x86_event(struct perf_event *event) { - if (!x86_pmu_initialized()) - return; - return x86_pmu.disable_all(); + return event->pmu == &pmu; } -static void p6_pmu_enable_all(void) +/* + * Event scheduler state: + * + * Assign events iterating over all events and counters, beginning + * with events with least weights first. Keep the current iterator + * state in struct sched_state. + */ +struct sched_state { + int weight; + int event; /* event index */ + int counter; /* counter index */ + int unassigned; /* number of events to be assigned left */ + unsigned long used[BITS_TO_LONGS(X86_PMC_IDX_MAX)]; +}; + +/* Total max is X86_PMC_IDX_MAX, but we are O(n!) limited */ +#define SCHED_STATES_MAX 2 + +struct perf_sched { + int max_weight; + int max_events; + struct perf_event **events; + struct sched_state state; + int saved_states; + struct sched_state saved[SCHED_STATES_MAX]; +}; + +/* + * Initialize interator that runs through all events and counters. + */ +static void perf_sched_init(struct perf_sched *sched, struct perf_event **events, + int num, int wmin, int wmax) { - struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); - unsigned long val; + int idx; - if (cpuc->enabled) - return; + memset(sched, 0, sizeof(*sched)); + sched->max_events = num; + sched->max_weight = wmax; + sched->events = events; - cpuc->enabled = 1; - barrier(); + for (idx = 0; idx < num; idx++) { + if (events[idx]->hw.constraint->weight == wmin) + break; + } - /* p6 only has one enable register */ - rdmsrl(MSR_P6_EVNTSEL0, val); - val |= ARCH_PERFMON_EVENTSEL0_ENABLE; - wrmsrl(MSR_P6_EVNTSEL0, val); + sched->state.event = idx; /* start with min weight */ + sched->state.weight = wmin; + sched->state.unassigned = num; } -static void intel_pmu_enable_all(void) +static void perf_sched_save_state(struct perf_sched *sched) { - struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); - - if (cpuc->enabled) + if (WARN_ON_ONCE(sched->saved_states >= SCHED_STATES_MAX)) return; - cpuc->enabled = 1; - barrier(); + sched->saved[sched->saved_states] = sched->state; + sched->saved_states++; +} - wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL, x86_pmu.intel_ctrl); +static bool perf_sched_restore_state(struct perf_sched *sched) +{ + if (!sched->saved_states) + return false; - if (test_bit(X86_PMC_IDX_FIXED_BTS, cpuc->active_mask)) { - struct perf_event *event = - cpuc->events[X86_PMC_IDX_FIXED_BTS]; + sched->saved_states--; + sched->state = sched->saved[sched->saved_states]; - if (WARN_ON_ONCE(!event)) - return; + /* continue with next counter: */ + clear_bit(sched->state.counter++, sched->state.used); - intel_pmu_enable_bts(event->hw.config); - } + return true; } -static void amd_pmu_enable_all(void) +/* + * Select a counter for the current event to schedule. Return true on + * success. + */ +static bool __perf_sched_find_counter(struct perf_sched *sched) { - struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + struct event_constraint *c; int idx; - if (cpuc->enabled) - return; + if (!sched->state.unassigned) + return false; - cpuc->enabled = 1; - barrier(); + if (sched->state.event >= sched->max_events) + return false; - for (idx = 0; idx < x86_pmu.num_events; idx++) { - struct perf_event *event = cpuc->events[idx]; - u64 val; + c = sched->events[sched->state.event]->hw.constraint; + /* Prefer fixed purpose counters */ + if (c->idxmsk64 & (~0ULL << INTEL_PMC_IDX_FIXED)) { + idx = INTEL_PMC_IDX_FIXED; + for_each_set_bit_from(idx, c->idxmsk, X86_PMC_IDX_MAX) { + if (!__test_and_set_bit(idx, sched->state.used)) + goto done; + } + } + /* Grab the first unused counter starting with idx */ + idx = sched->state.counter; + for_each_set_bit_from(idx, c->idxmsk, INTEL_PMC_IDX_FIXED) { + if (!__test_and_set_bit(idx, sched->state.used)) + goto done; + } - if (!test_bit(idx, cpuc->active_mask)) - continue; + return false; - val = event->hw.config; - val |= ARCH_PERFMON_EVENTSEL0_ENABLE; - wrmsrl(MSR_K7_EVNTSEL0 + idx, val); - } +done: + sched->state.counter = idx; + + if (c->overlap) + perf_sched_save_state(sched); + + return true; } -void hw_perf_enable(void) +static bool perf_sched_find_counter(struct perf_sched *sched) { - if (!x86_pmu_initialized()) - return; - x86_pmu.enable_all(); + while (!__perf_sched_find_counter(sched)) { + if (!perf_sched_restore_state(sched)) + return false; + } + + return true; } -static inline u64 intel_pmu_get_status(void) +/* + * Go through all unassigned events and find the next one to schedule. + * Take events with the least weight first. Return true on success. + */ +static bool perf_sched_next_event(struct perf_sched *sched) { - u64 status; + struct event_constraint *c; - rdmsrl(MSR_CORE_PERF_GLOBAL_STATUS, status); + if (!sched->state.unassigned || !--sched->state.unassigned) + return false; - return status; -} + do { + /* next event */ + sched->state.event++; + if (sched->state.event >= sched->max_events) { + /* next weight */ + sched->state.event = 0; + sched->state.weight++; + if (sched->state.weight > sched->max_weight) + return false; + } + c = sched->events[sched->state.event]->hw.constraint; + } while (c->weight != sched->state.weight); -static inline void intel_pmu_ack_status(u64 ack) -{ - wrmsrl(MSR_CORE_PERF_GLOBAL_OVF_CTRL, ack); + sched->state.counter = 0; /* start with first counter */ + + return true; } -static inline void x86_pmu_enable_event(struct hw_perf_event *hwc, int idx) +/* + * Assign a counter for each event. + */ +int perf_assign_events(struct perf_event **events, int n, + int wmin, int wmax, int *assign) { - (void)checking_wrmsrl(hwc->config_base + idx, - hwc->config | ARCH_PERFMON_EVENTSEL0_ENABLE); + struct perf_sched sched; + + perf_sched_init(&sched, events, n, wmin, wmax); + + do { + if (!perf_sched_find_counter(&sched)) + break; /* failed */ + if (assign) + assign[sched.state.event] = sched.state.counter; + } while (perf_sched_next_event(&sched)); + + return sched.state.unassigned; } +EXPORT_SYMBOL_GPL(perf_assign_events); -static inline void x86_pmu_disable_event(struct hw_perf_event *hwc, int idx) +int x86_schedule_events(struct cpu_hw_events *cpuc, int n, int *assign) { - (void)checking_wrmsrl(hwc->config_base + idx, hwc->config); + struct event_constraint *c; + unsigned long used_mask[BITS_TO_LONGS(X86_PMC_IDX_MAX)]; + struct perf_event *e; + int i, wmin, wmax, num = 0; + struct hw_perf_event *hwc; + + bitmap_zero(used_mask, X86_PMC_IDX_MAX); + + for (i = 0, wmin = X86_PMC_IDX_MAX, wmax = 0; i < n; i++) { + hwc = &cpuc->event_list[i]->hw; + c = x86_pmu.get_event_constraints(cpuc, cpuc->event_list[i]); + hwc->constraint = c; + + wmin = min(wmin, c->weight); + wmax = max(wmax, c->weight); + } + + /* + * fastpath, try to reuse previous register + */ + for (i = 0; i < n; i++) { + hwc = &cpuc->event_list[i]->hw; + c = hwc->constraint; + + /* never assigned */ + if (hwc->idx == -1) + break; + + /* constraint still honored */ + if (!test_bit(hwc->idx, c->idxmsk)) + break; + + /* not already used */ + if (test_bit(hwc->idx, used_mask)) + break; + + __set_bit(hwc->idx, used_mask); + if (assign) + assign[i] = hwc->idx; + } + + /* slow path */ + if (i != n) + num = perf_assign_events(cpuc->event_list, n, wmin, + wmax, assign); + + /* + * Mark the event as committed, so we do not put_constraint() + * in case new events are added and fail scheduling. + */ + if (!num && assign) { + for (i = 0; i < n; i++) { + e = cpuc->event_list[i]; + e->hw.flags |= PERF_X86_EVENT_COMMITTED; + } + } + /* + * scheduling failed or is just a simulation, + * free resources if necessary + */ + if (!assign || num) { + for (i = 0; i < n; i++) { + e = cpuc->event_list[i]; + /* + * do not put_constraint() on comitted events, + * because they are good to go + */ + if ((e->hw.flags & PERF_X86_EVENT_COMMITTED)) + continue; + + if (x86_pmu.put_event_constraints) + x86_pmu.put_event_constraints(cpuc, e); + } + } + return num ? -EINVAL : 0; } -static inline void -intel_pmu_disable_fixed(struct hw_perf_event *hwc, int __idx) +/* + * dogrp: true if must collect siblings events (group) + * returns total number of events and error code + */ +static int collect_events(struct cpu_hw_events *cpuc, struct perf_event *leader, bool dogrp) { - int idx = __idx - X86_PMC_IDX_FIXED; - u64 ctrl_val, mask; + struct perf_event *event; + int n, max_count; - mask = 0xfULL << (idx * 4); + max_count = x86_pmu.num_counters + x86_pmu.num_counters_fixed; - rdmsrl(hwc->config_base, ctrl_val); - ctrl_val &= ~mask; - (void)checking_wrmsrl(hwc->config_base, ctrl_val); + /* current number of events already accepted */ + n = cpuc->n_events; + + if (is_x86_event(leader)) { + if (n >= max_count) + return -EINVAL; + cpuc->event_list[n] = leader; + n++; + } + if (!dogrp) + return n; + + list_for_each_entry(event, &leader->sibling_list, group_entry) { + if (!is_x86_event(event) || + event->state <= PERF_EVENT_STATE_OFF) + continue; + + if (n >= max_count) + return -EINVAL; + + cpuc->event_list[n] = event; + n++; + } + return n; } -static inline void -p6_pmu_disable_event(struct hw_perf_event *hwc, int idx) +static inline void x86_assign_hw_event(struct perf_event *event, + struct cpu_hw_events *cpuc, int i) { - struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); - u64 val = P6_NOP_EVENT; + struct hw_perf_event *hwc = &event->hw; - if (cpuc->enabled) - val |= ARCH_PERFMON_EVENTSEL0_ENABLE; + hwc->idx = cpuc->assign[i]; + hwc->last_cpu = smp_processor_id(); + hwc->last_tag = ++cpuc->tags[i]; - (void)checking_wrmsrl(hwc->config_base + idx, val); + if (hwc->idx == INTEL_PMC_IDX_FIXED_BTS) { + hwc->config_base = 0; + hwc->event_base = 0; + } else if (hwc->idx >= INTEL_PMC_IDX_FIXED) { + hwc->config_base = MSR_ARCH_PERFMON_FIXED_CTR_CTRL; + hwc->event_base = MSR_ARCH_PERFMON_FIXED_CTR0 + (hwc->idx - INTEL_PMC_IDX_FIXED); + hwc->event_base_rdpmc = (hwc->idx - INTEL_PMC_IDX_FIXED) | 1<<30; + } else { + hwc->config_base = x86_pmu_config_addr(hwc->idx); + hwc->event_base = x86_pmu_event_addr(hwc->idx); + hwc->event_base_rdpmc = x86_pmu_rdpmc_index(hwc->idx); + } } -static inline void -intel_pmu_disable_event(struct hw_perf_event *hwc, int idx) +static inline int match_prev_assignment(struct hw_perf_event *hwc, + struct cpu_hw_events *cpuc, + int i) { - if (unlikely(idx == X86_PMC_IDX_FIXED_BTS)) { - intel_pmu_disable_bts(); + return hwc->idx == cpuc->assign[i] && + hwc->last_cpu == smp_processor_id() && + hwc->last_tag == cpuc->tags[i]; +} + +static void x86_pmu_start(struct perf_event *event, int flags); + +static void x86_pmu_enable(struct pmu *pmu) +{ + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + struct perf_event *event; + struct hw_perf_event *hwc; + int i, added = cpuc->n_added; + + if (!x86_pmu_initialized()) return; - } - if (unlikely(hwc->config_base == MSR_ARCH_PERFMON_FIXED_CTR_CTRL)) { - intel_pmu_disable_fixed(hwc, idx); + if (cpuc->enabled) return; + + if (cpuc->n_added) { + int n_running = cpuc->n_events - cpuc->n_added; + /* + * apply assignment obtained either from + * hw_perf_group_sched_in() or x86_pmu_enable() + * + * step1: save events moving to new counters + */ + for (i = 0; i < n_running; i++) { + event = cpuc->event_list[i]; + hwc = &event->hw; + + /* + * we can avoid reprogramming counter if: + * - assigned same counter as last time + * - running on same CPU as last time + * - no other event has used the counter since + */ + if (hwc->idx == -1 || + match_prev_assignment(hwc, cpuc, i)) + continue; + + /* + * Ensure we don't accidentally enable a stopped + * counter simply because we rescheduled. + */ + if (hwc->state & PERF_HES_STOPPED) + hwc->state |= PERF_HES_ARCH; + + x86_pmu_stop(event, PERF_EF_UPDATE); + } + + /* + * step2: reprogram moved events into new counters + */ + for (i = 0; i < cpuc->n_events; i++) { + event = cpuc->event_list[i]; + hwc = &event->hw; + + if (!match_prev_assignment(hwc, cpuc, i)) + x86_assign_hw_event(event, cpuc, i); + else if (i < n_running) + continue; + + if (hwc->state & PERF_HES_ARCH) + continue; + + x86_pmu_start(event, PERF_EF_RELOAD); + } + cpuc->n_added = 0; + perf_events_lapic_init(); } - x86_pmu_disable_event(hwc, idx); -} + cpuc->enabled = 1; + barrier(); -static inline void -amd_pmu_disable_event(struct hw_perf_event *hwc, int idx) -{ - x86_pmu_disable_event(hwc, idx); + x86_pmu.enable_all(added); } static DEFINE_PER_CPU(u64 [X86_PMC_IDX_MAX], pmc_prev_left); @@ -1274,15 +945,14 @@ static DEFINE_PER_CPU(u64 [X86_PMC_IDX_MAX], pmc_prev_left); * Set the next IRQ period, based on the hwc->period_left value. * To be called with the event disabled in hw: */ -static int -x86_perf_event_set_period(struct perf_event *event, - struct hw_perf_event *hwc, int idx) +int x86_perf_event_set_period(struct perf_event *event) { - s64 left = atomic64_read(&hwc->period_left); + struct hw_perf_event *hwc = &event->hw; + s64 left = local64_read(&hwc->period_left); s64 period = hwc->sample_period; - int err, ret = 0; + int ret = 0, idx = hwc->idx; - if (idx == X86_PMC_IDX_FIXED_BTS) + if (idx == INTEL_PMC_IDX_FIXED_BTS) return 0; /* @@ -1290,14 +960,14 @@ x86_perf_event_set_period(struct perf_event *event, */ if (unlikely(left <= -period)) { left = period; - atomic64_set(&hwc->period_left, left); + local64_set(&hwc->period_left, left); hwc->last_period = period; ret = 1; } if (unlikely(left <= 0)) { left += period; - atomic64_set(&hwc->period_left, left); + local64_set(&hwc->period_left, left); hwc->last_period = period; ret = 1; } @@ -1316,247 +986,123 @@ x86_perf_event_set_period(struct perf_event *event, * The hw event starts counting from this event offset, * mark it to be able to extra future deltas: */ - atomic64_set(&hwc->prev_count, (u64)-left); + local64_set(&hwc->prev_count, (u64)-left); - err = checking_wrmsrl(hwc->event_base + idx, - (u64)(-left) & x86_pmu.event_mask); - - perf_event_update_userpage(event); - - return ret; -} - -static inline void -intel_pmu_enable_fixed(struct hw_perf_event *hwc, int __idx) -{ - int idx = __idx - X86_PMC_IDX_FIXED; - u64 ctrl_val, bits, mask; - int err; + wrmsrl(hwc->event_base, (u64)(-left) & x86_pmu.cntval_mask); /* - * Enable IRQ generation (0x8), - * and enable ring-3 counting (0x2) and ring-0 counting (0x1) - * if requested: + * Due to erratum on certan cpu we need + * a second write to be sure the register + * is updated properly */ - bits = 0x8ULL; - if (hwc->config & ARCH_PERFMON_EVENTSEL_USR) - bits |= 0x2; - if (hwc->config & ARCH_PERFMON_EVENTSEL_OS) - bits |= 0x1; - bits <<= (idx * 4); - mask = 0xfULL << (idx * 4); - - rdmsrl(hwc->config_base, ctrl_val); - ctrl_val &= ~mask; - ctrl_val |= bits; - err = checking_wrmsrl(hwc->config_base, ctrl_val); -} - -static void p6_pmu_enable_event(struct hw_perf_event *hwc, int idx) -{ - struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); - u64 val; - - val = hwc->config; - if (cpuc->enabled) - val |= ARCH_PERFMON_EVENTSEL0_ENABLE; - - (void)checking_wrmsrl(hwc->config_base + idx, val); -} - - -static void intel_pmu_enable_event(struct hw_perf_event *hwc, int idx) -{ - if (unlikely(idx == X86_PMC_IDX_FIXED_BTS)) { - if (!__get_cpu_var(cpu_hw_events).enabled) - return; - - intel_pmu_enable_bts(hwc->config); - return; + if (x86_pmu.perfctr_second_write) { + wrmsrl(hwc->event_base, + (u64)(-left) & x86_pmu.cntval_mask); } - if (unlikely(hwc->config_base == MSR_ARCH_PERFMON_FIXED_CTR_CTRL)) { - intel_pmu_enable_fixed(hwc, idx); - return; - } + perf_event_update_userpage(event); - x86_pmu_enable_event(hwc, idx); + return ret; } -static void amd_pmu_enable_event(struct hw_perf_event *hwc, int idx) +void x86_pmu_enable_event(struct perf_event *event) { - struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); - - if (cpuc->enabled) - x86_pmu_enable_event(hwc, idx); + if (__this_cpu_read(cpu_hw_events.enabled)) + __x86_pmu_enable_event(&event->hw, + ARCH_PERFMON_EVENTSEL_ENABLE); } -static int fixed_mode_idx(struct hw_perf_event *hwc) +/* + * Add a single event to the PMU. + * + * The event is added to the group of enabled events + * but only if it can be scehduled with existing events. + */ +static int x86_pmu_add(struct perf_event *event, int flags) { - unsigned int hw_event; + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + struct hw_perf_event *hwc; + int assign[X86_PMC_IDX_MAX]; + int n, n0, ret; - hw_event = hwc->config & ARCH_PERFMON_EVENT_MASK; + hwc = &event->hw; - if (unlikely((hw_event == - x86_pmu.event_map(PERF_COUNT_HW_BRANCH_INSTRUCTIONS)) && - (hwc->sample_period == 1))) - return X86_PMC_IDX_FIXED_BTS; + perf_pmu_disable(event->pmu); + n0 = cpuc->n_events; + ret = n = collect_events(cpuc, event, false); + if (ret < 0) + goto out; - if (!x86_pmu.num_events_fixed) - return -1; + hwc->state = PERF_HES_UPTODATE | PERF_HES_STOPPED; + if (!(flags & PERF_EF_START)) + hwc->state |= PERF_HES_ARCH; /* - * fixed counters do not take all possible filters + * If group events scheduling transaction was started, + * skip the schedulability test here, it will be performed + * at commit time (->commit_txn) as a whole. */ - if (hwc->config & ARCH_PERFMON_EVENT_FILTER_MASK) - return -1; - - if (unlikely(hw_event == x86_pmu.event_map(PERF_COUNT_HW_INSTRUCTIONS))) - return X86_PMC_IDX_FIXED_INSTRUCTIONS; - if (unlikely(hw_event == x86_pmu.event_map(PERF_COUNT_HW_CPU_CYCLES))) - return X86_PMC_IDX_FIXED_CPU_CYCLES; - if (unlikely(hw_event == x86_pmu.event_map(PERF_COUNT_HW_BUS_CYCLES))) - return X86_PMC_IDX_FIXED_BUS_CYCLES; - - return -1; -} - -/* - * generic counter allocator: get next free counter - */ -static int -gen_get_event_idx(struct cpu_hw_events *cpuc, struct hw_perf_event *hwc) -{ - int idx; - - idx = find_first_zero_bit(cpuc->used_mask, x86_pmu.num_events); - return idx == x86_pmu.num_events ? -1 : idx; -} - -/* - * intel-specific counter allocator: check event constraints - */ -static int -intel_get_event_idx(struct cpu_hw_events *cpuc, struct hw_perf_event *hwc) -{ - const struct event_constraint *event_constraint; - int i, code; + if (cpuc->group_flag & PERF_EVENT_TXN) + goto done_collect; - if (!event_constraints) - goto skip; + ret = x86_pmu.schedule_events(cpuc, n, assign); + if (ret) + goto out; + /* + * copy new assignment, now we know it is possible + * will be used by hw_perf_enable() + */ + memcpy(cpuc->assign, assign, n*sizeof(int)); - code = hwc->config & CORE_EVNTSEL_EVENT_MASK; +done_collect: + /* + * Commit the collect_events() state. See x86_pmu_del() and + * x86_pmu_*_txn(). + */ + cpuc->n_events = n; + cpuc->n_added += n - n0; + cpuc->n_txn += n - n0; - for_each_event_constraint(event_constraint, event_constraints) { - if (code == event_constraint->code) { - for_each_bit(i, event_constraint->idxmsk, X86_PMC_IDX_MAX) { - if (!test_and_set_bit(i, cpuc->used_mask)) - return i; - } - return -1; - } - } -skip: - return gen_get_event_idx(cpuc, hwc); + ret = 0; +out: + perf_pmu_enable(event->pmu); + return ret; } -static int -x86_schedule_event(struct cpu_hw_events *cpuc, struct hw_perf_event *hwc) +static void x86_pmu_start(struct perf_event *event, int flags) { - int idx; + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + int idx = event->hw.idx; - idx = fixed_mode_idx(hwc); - if (idx == X86_PMC_IDX_FIXED_BTS) { - /* BTS is already occupied. */ - if (test_and_set_bit(idx, cpuc->used_mask)) - return -EAGAIN; + if (WARN_ON_ONCE(!(event->hw.state & PERF_HES_STOPPED))) + return; - hwc->config_base = 0; - hwc->event_base = 0; - hwc->idx = idx; - } else if (idx >= 0) { - /* - * Try to get the fixed event, if that is already taken - * then try to get a generic event: - */ - if (test_and_set_bit(idx, cpuc->used_mask)) - goto try_generic; + if (WARN_ON_ONCE(idx == -1)) + return; - hwc->config_base = MSR_ARCH_PERFMON_FIXED_CTR_CTRL; - /* - * We set it so that event_base + idx in wrmsr/rdmsr maps to - * MSR_ARCH_PERFMON_FIXED_CTR0 ... CTR2: - */ - hwc->event_base = - MSR_ARCH_PERFMON_FIXED_CTR0 - X86_PMC_IDX_FIXED; - hwc->idx = idx; - } else { - idx = hwc->idx; - /* Try to get the previous generic event again */ - if (idx == -1 || test_and_set_bit(idx, cpuc->used_mask)) { -try_generic: - idx = x86_pmu.get_event_idx(cpuc, hwc); - if (idx == -1) - return -EAGAIN; - - set_bit(idx, cpuc->used_mask); - hwc->idx = idx; - } - hwc->config_base = x86_pmu.eventsel; - hwc->event_base = x86_pmu.perfctr; + if (flags & PERF_EF_RELOAD) { + WARN_ON_ONCE(!(event->hw.state & PERF_HES_UPTODATE)); + x86_perf_event_set_period(event); } - return idx; -} - -/* - * Find a PMC slot for the freshly enabled / scheduled in event: - */ -static int x86_pmu_enable(struct perf_event *event) -{ - struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); - struct hw_perf_event *hwc = &event->hw; - int idx; - - idx = x86_schedule_event(cpuc, hwc); - if (idx < 0) - return idx; - - perf_events_lapic_init(); - - x86_pmu.disable(hwc, idx); + event->hw.state = 0; cpuc->events[idx] = event; - set_bit(idx, cpuc->active_mask); - - x86_perf_event_set_period(event, hwc, idx); - x86_pmu.enable(hwc, idx); - + __set_bit(idx, cpuc->active_mask); + __set_bit(idx, cpuc->running); + x86_pmu.enable(event); perf_event_update_userpage(event); - - return 0; -} - -static void x86_pmu_unthrottle(struct perf_event *event) -{ - struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); - struct hw_perf_event *hwc = &event->hw; - - if (WARN_ON_ONCE(hwc->idx >= X86_PMC_IDX_MAX || - cpuc->events[hwc->idx] != event)) - return; - - x86_pmu.enable(hwc, hwc->idx); } void perf_event_print_debug(void) { u64 ctrl, status, overflow, pmc_ctrl, pmc_count, prev_left, fixed; + u64 pebs; struct cpu_hw_events *cpuc; unsigned long flags; int cpu, idx; - if (!x86_pmu.num_events) + if (!x86_pmu.num_counters) return; local_irq_save(flags); @@ -1569,18 +1115,20 @@ void perf_event_print_debug(void) rdmsrl(MSR_CORE_PERF_GLOBAL_STATUS, status); rdmsrl(MSR_CORE_PERF_GLOBAL_OVF_CTRL, overflow); rdmsrl(MSR_ARCH_PERFMON_FIXED_CTR_CTRL, fixed); + rdmsrl(MSR_IA32_PEBS_ENABLE, pebs); pr_info("\n"); pr_info("CPU#%d: ctrl: %016llx\n", cpu, ctrl); pr_info("CPU#%d: status: %016llx\n", cpu, status); pr_info("CPU#%d: overflow: %016llx\n", cpu, overflow); pr_info("CPU#%d: fixed: %016llx\n", cpu, fixed); + pr_info("CPU#%d: pebs: %016llx\n", cpu, pebs); } - pr_info("CPU#%d: used: %016llx\n", cpu, *(u64 *)cpuc->used_mask); + pr_info("CPU#%d: active: %016llx\n", cpu, *(u64 *)cpuc->active_mask); - for (idx = 0; idx < x86_pmu.num_events; idx++) { - rdmsrl(x86_pmu.eventsel + idx, pmc_ctrl); - rdmsrl(x86_pmu.perfctr + idx, pmc_count); + for (idx = 0; idx < x86_pmu.num_counters; idx++) { + rdmsrl(x86_pmu_config_addr(idx), pmc_ctrl); + rdmsrl(x86_pmu_event_addr(idx), pmc_count); prev_left = per_cpu(pmc_prev_left[idx], cpu); @@ -1591,7 +1139,7 @@ void perf_event_print_debug(void) pr_info("CPU#%d: gen-PMC%d left: %016llx\n", cpu, idx, prev_left); } - for (idx = 0; idx < x86_pmu.num_events_fixed; idx++) { + for (idx = 0; idx < x86_pmu.num_counters_fixed; idx++) { rdmsrl(MSR_ARCH_PERFMON_FIXED_CTR0 + idx, pmc_count); pr_info("CPU#%d: fixed-PMC%d count: %016llx\n", @@ -1600,182 +1148,126 @@ void perf_event_print_debug(void) local_irq_restore(flags); } -static void intel_pmu_drain_bts_buffer(struct cpu_hw_events *cpuc) +void x86_pmu_stop(struct perf_event *event, int flags) { - struct debug_store *ds = cpuc->ds; - struct bts_record { - u64 from; - u64 to; - u64 flags; - }; - struct perf_event *event = cpuc->events[X86_PMC_IDX_FIXED_BTS]; - struct bts_record *at, *top; - struct perf_output_handle handle; - struct perf_event_header header; - struct perf_sample_data data; - struct pt_regs regs; - - if (!event) - return; - - if (!ds) - return; - - at = (struct bts_record *)(unsigned long)ds->bts_buffer_base; - top = (struct bts_record *)(unsigned long)ds->bts_index; - - if (top <= at) - return; - - ds->bts_index = ds->bts_buffer_base; - - - data.period = event->hw.last_period; - data.addr = 0; - data.raw = NULL; - regs.ip = 0; - - /* - * Prepare a generic sample, i.e. fill in the invariant fields. - * We will overwrite the from and to address before we output - * the sample. - */ - perf_prepare_sample(&header, &data, event, ®s); - - if (perf_output_begin(&handle, event, - header.size * (top - at), 1, 1)) - return; - - for (; at < top; at++) { - data.ip = at->from; - data.addr = at->to; + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + struct hw_perf_event *hwc = &event->hw; - perf_output_sample(&handle, &header, &data, event); + if (__test_and_clear_bit(hwc->idx, cpuc->active_mask)) { + x86_pmu.disable(event); + cpuc->events[hwc->idx] = NULL; + WARN_ON_ONCE(hwc->state & PERF_HES_STOPPED); + hwc->state |= PERF_HES_STOPPED; } - perf_output_end(&handle); - - /* There's new data available. */ - event->hw.interrupts++; - event->pending_kill = POLL_IN; + if ((flags & PERF_EF_UPDATE) && !(hwc->state & PERF_HES_UPTODATE)) { + /* + * Drain the remaining delta count out of a event + * that we are disabling: + */ + x86_perf_event_update(event); + hwc->state |= PERF_HES_UPTODATE; + } } -static void x86_pmu_disable(struct perf_event *event) +static void x86_pmu_del(struct perf_event *event, int flags) { struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); - struct hw_perf_event *hwc = &event->hw; - int idx = hwc->idx; + int i; /* - * Must be done before we disable, otherwise the nmi handler - * could reenable again: + * event is descheduled */ - clear_bit(idx, cpuc->active_mask); - x86_pmu.disable(hwc, idx); + event->hw.flags &= ~PERF_X86_EVENT_COMMITTED; /* - * Make sure the cleared pointer becomes visible before we - * (potentially) free the event: + * If we're called during a txn, we don't need to do anything. + * The events never got scheduled and ->cancel_txn will truncate + * the event_list. + * + * XXX assumes any ->del() called during a TXN will only be on + * an event added during that same TXN. */ - barrier(); + if (cpuc->group_flag & PERF_EVENT_TXN) + return; /* - * Drain the remaining delta count out of a event - * that we are disabling: + * Not a TXN, therefore cleanup properly. */ - x86_perf_event_update(event, hwc, idx); - - /* Drain the remaining BTS records. */ - if (unlikely(idx == X86_PMC_IDX_FIXED_BTS)) - intel_pmu_drain_bts_buffer(cpuc); - - cpuc->events[idx] = NULL; - clear_bit(idx, cpuc->used_mask); - - perf_event_update_userpage(event); -} - -/* - * Save and restart an expired event. Called by NMI contexts, - * so it has to be careful about preempting normal event ops: - */ -static int intel_pmu_save_and_restart(struct perf_event *event) -{ - struct hw_perf_event *hwc = &event->hw; - int idx = hwc->idx; - int ret; + x86_pmu_stop(event, PERF_EF_UPDATE); - x86_perf_event_update(event, hwc, idx); - ret = x86_perf_event_set_period(event, hwc, idx); - - if (event->state == PERF_EVENT_STATE_ACTIVE) - intel_pmu_enable_event(hwc, idx); - - return ret; -} - -static void intel_pmu_reset(void) -{ - struct debug_store *ds = __get_cpu_var(cpu_hw_events).ds; - unsigned long flags; - int idx; + for (i = 0; i < cpuc->n_events; i++) { + if (event == cpuc->event_list[i]) + break; + } - if (!x86_pmu.num_events) + if (WARN_ON_ONCE(i == cpuc->n_events)) /* called ->del() without ->add() ? */ return; - local_irq_save(flags); + /* If we have a newly added event; make sure to decrease n_added. */ + if (i >= cpuc->n_events - cpuc->n_added) + --cpuc->n_added; - printk("clearing PMU state on CPU#%d\n", smp_processor_id()); + if (x86_pmu.put_event_constraints) + x86_pmu.put_event_constraints(cpuc, event); - for (idx = 0; idx < x86_pmu.num_events; idx++) { - checking_wrmsrl(x86_pmu.eventsel + idx, 0ull); - checking_wrmsrl(x86_pmu.perfctr + idx, 0ull); - } - for (idx = 0; idx < x86_pmu.num_events_fixed; idx++) { - checking_wrmsrl(MSR_ARCH_PERFMON_FIXED_CTR0 + idx, 0ull); - } - if (ds) - ds->bts_index = ds->bts_buffer_base; + /* Delete the array entry. */ + while (++i < cpuc->n_events) + cpuc->event_list[i-1] = cpuc->event_list[i]; + --cpuc->n_events; - local_irq_restore(flags); + perf_event_update_userpage(event); } -static int p6_pmu_handle_irq(struct pt_regs *regs) +int x86_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)) + /* + * Some chipsets need to unmask the LVTPC in a particular spot + * inside the nmi handler. As a result, the unmasking was pushed + * into all the nmi handlers. + * + * This generic handler doesn't seem to have any issues where the + * unmasking occurs so it was left at the top. + */ + apic_write(APIC_LVTPC, APIC_DM_NMI); + + for (idx = 0; idx < x86_pmu.num_counters; idx++) { + if (!test_bit(idx, cpuc->active_mask)) { + /* + * Though we deactivated the counter some cpus + * might still deliver spurious interrupts still + * in flight. Catch them: + */ + if (__test_and_clear_bit(idx, cpuc->running)) + handled++; continue; + } event = cpuc->events[idx]; - hwc = &event->hw; - val = x86_perf_event_update(event, hwc, idx); - if (val & (1ULL << (x86_pmu.event_bits - 1))) + val = x86_perf_event_update(event); + if (val & (1ULL << (x86_pmu.cntval_bits - 1))) continue; /* * event overflow */ - handled = 1; - data.period = event->hw.last_period; + handled++; + perf_sample_data_init(&data, 0, event->hw.last_period); - if (!x86_perf_event_set_period(event, hwc, idx)) + if (!x86_perf_event_set_period(event)) continue; - if (perf_event_overflow(event, 1, &data, regs)) - p6_pmu_disable_event(hwc, idx); + if (perf_event_overflow(event, &data, regs)) + x86_pmu_stop(event, 0); } if (handled) @@ -1784,391 +1276,248 @@ static int p6_pmu_handle_irq(struct pt_regs *regs) return handled; } -/* - * This handler is triggered by the local APIC, so the APIC IRQ handling - * rules apply: - */ -static int intel_pmu_handle_irq(struct pt_regs *regs) +void perf_events_lapic_init(void) { - struct perf_sample_data data; - struct cpu_hw_events *cpuc; - int bit, loops; - u64 ack, status; - - data.addr = 0; - data.raw = NULL; - - cpuc = &__get_cpu_var(cpu_hw_events); - - perf_disable(); - intel_pmu_drain_bts_buffer(cpuc); - status = intel_pmu_get_status(); - if (!status) { - perf_enable(); - return 0; - } - - loops = 0; -again: - if (++loops > 100) { - WARN_ONCE(1, "perfevents: irq loop stuck!\n"); - perf_event_print_debug(); - intel_pmu_reset(); - perf_enable(); - return 1; - } - - inc_irq_stat(apic_perf_irqs); - ack = status; - for_each_bit(bit, (unsigned long *)&status, X86_PMC_IDX_MAX) { - struct perf_event *event = cpuc->events[bit]; - - clear_bit(bit, (unsigned long *) &status); - if (!test_bit(bit, cpuc->active_mask)) - continue; - - if (!intel_pmu_save_and_restart(event)) - continue; - - data.period = event->hw.last_period; - - if (perf_event_overflow(event, 1, &data, regs)) - intel_pmu_disable_event(&event->hw, bit); - } - - intel_pmu_ack_status(ack); + if (!x86_pmu.apic || !x86_pmu_initialized()) + return; /* - * Repeat if there is more work to be done: + * Always use NMI for PMU */ - status = intel_pmu_get_status(); - if (status) - goto again; - - perf_enable(); - - return 1; + apic_write(APIC_LVTPC, APIC_DM_NMI); } -static int amd_pmu_handle_irq(struct pt_regs *regs) +static int +perf_event_nmi_handler(unsigned int cmd, 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; + u64 start_clock; + u64 finish_clock; + int ret; - data.addr = 0; - data.raw = NULL; + if (!atomic_read(&active_events)) + return NMI_DONE; - cpuc = &__get_cpu_var(cpu_hw_events); + start_clock = sched_clock(); + ret = x86_pmu.handle_irq(regs); + finish_clock = sched_clock(); - for (idx = 0; idx < x86_pmu.num_events; idx++) { - if (!test_bit(idx, cpuc->active_mask)) - continue; + perf_sample_event_took(finish_clock - start_clock); - event = cpuc->events[idx]; - hwc = &event->hw; + return ret; +} +NOKPROBE_SYMBOL(perf_event_nmi_handler); - val = x86_perf_event_update(event, hwc, idx); - if (val & (1ULL << (x86_pmu.event_bits - 1))) - continue; +struct event_constraint emptyconstraint; +struct event_constraint unconstrained; - /* - * event overflow - */ - handled = 1; - data.period = event->hw.last_period; +static int +x86_pmu_notifier(struct notifier_block *self, unsigned long action, void *hcpu) +{ + unsigned int cpu = (long)hcpu; + struct cpu_hw_events *cpuc = &per_cpu(cpu_hw_events, cpu); + int ret = NOTIFY_OK; - if (!x86_perf_event_set_period(event, hwc, idx)) - continue; + switch (action & ~CPU_TASKS_FROZEN) { + case CPU_UP_PREPARE: + cpuc->kfree_on_online = NULL; + if (x86_pmu.cpu_prepare) + ret = x86_pmu.cpu_prepare(cpu); + break; - if (perf_event_overflow(event, 1, &data, regs)) - amd_pmu_disable_event(hwc, idx); - } + case CPU_STARTING: + if (x86_pmu.attr_rdpmc) + set_in_cr4(X86_CR4_PCE); + if (x86_pmu.cpu_starting) + x86_pmu.cpu_starting(cpu); + break; - if (handled) - inc_irq_stat(apic_perf_irqs); + case CPU_ONLINE: + kfree(cpuc->kfree_on_online); + break; - return handled; -} + case CPU_DYING: + if (x86_pmu.cpu_dying) + x86_pmu.cpu_dying(cpu); + break; -void smp_perf_pending_interrupt(struct pt_regs *regs) -{ - irq_enter(); - ack_APIC_irq(); - inc_irq_stat(apic_pending_irqs); - perf_event_do_pending(); - irq_exit(); -} + case CPU_UP_CANCELED: + case CPU_DEAD: + if (x86_pmu.cpu_dead) + x86_pmu.cpu_dead(cpu); + break; -void set_perf_event_pending(void) -{ -#ifdef CONFIG_X86_LOCAL_APIC - if (!x86_pmu.apic || !x86_pmu_initialized()) - return; + default: + break; + } - apic->send_IPI_self(LOCAL_PENDING_VECTOR); -#endif + return ret; } -void perf_events_lapic_init(void) +static void __init pmu_check_apic(void) { -#ifdef CONFIG_X86_LOCAL_APIC - if (!x86_pmu.apic || !x86_pmu_initialized()) + if (cpu_has_apic) return; + x86_pmu.apic = 0; + pr_info("no APIC, boot with the \"lapic\" boot parameter to force-enable it.\n"); + pr_info("no hardware sampling interrupt available.\n"); + /* - * Always use NMI for PMU + * If we have a PMU initialized but no APIC + * interrupts, we cannot sample hardware + * events (user-space has to fall back and + * sample via a hrtimer based software event): */ - apic_write(APIC_LVTPC, APIC_DM_NMI); -#endif + pmu.capabilities |= PERF_PMU_CAP_NO_INTERRUPT; + } -static int __kprobes -perf_event_nmi_handler(struct notifier_block *self, - unsigned long cmd, void *__args) +static struct attribute_group x86_pmu_format_group = { + .name = "format", + .attrs = NULL, +}; + +/* + * Remove all undefined events (x86_pmu.event_map(id) == 0) + * out of events_attr attributes. + */ +static void __init filter_events(struct attribute **attrs) { - struct die_args *args = __args; - struct pt_regs *regs; + struct device_attribute *d; + struct perf_pmu_events_attr *pmu_attr; + int i, j; - if (!atomic_read(&active_events)) - return NOTIFY_DONE; + for (i = 0; attrs[i]; i++) { + d = (struct device_attribute *)attrs[i]; + pmu_attr = container_of(d, struct perf_pmu_events_attr, attr); + /* str trumps id */ + if (pmu_attr->event_str) + continue; + if (x86_pmu.event_map(i)) + continue; - switch (cmd) { - case DIE_NMI: - case DIE_NMI_IPI: - break; + for (j = i; attrs[j]; j++) + attrs[j] = attrs[j + 1]; - default: - return NOTIFY_DONE; + /* Check the shifted attr. */ + i--; } - - regs = args->regs; - -#ifdef CONFIG_X86_LOCAL_APIC - apic_write(APIC_LVTPC, APIC_DM_NMI); -#endif - /* - * Can't rely on the handled return value to say it was our NMI, two - * events could trigger 'simultaneously' raising two back-to-back NMIs. - * - * If the first NMI handles both, the latter will be empty and daze - * the CPU. - */ - x86_pmu.handle_irq(regs); - - return NOTIFY_STOP; } -static __read_mostly struct notifier_block perf_event_nmi_notifier = { - .notifier_call = perf_event_nmi_handler, - .next = NULL, - .priority = 1 -}; - -static __initconst struct x86_pmu p6_pmu = { - .name = "p6", - .handle_irq = p6_pmu_handle_irq, - .disable_all = p6_pmu_disable_all, - .enable_all = p6_pmu_enable_all, - .enable = p6_pmu_enable_event, - .disable = p6_pmu_disable_event, - .eventsel = MSR_P6_EVNTSEL0, - .perfctr = MSR_P6_PERFCTR0, - .event_map = p6_pmu_event_map, - .raw_event = p6_pmu_raw_event, - .max_events = ARRAY_SIZE(p6_perfmon_event_map), - .apic = 1, - .max_period = (1ULL << 31) - 1, - .version = 0, - .num_events = 2, - /* - * Events have 40 bits implemented. However they are designed such - * that bits [32-39] are sign extensions of bit 31. As such the - * effective width of a event for P6-like PMU is 32 bits only. - * - * See IA-32 Intel Architecture Software developer manual Vol 3B - */ - .event_bits = 32, - .event_mask = (1ULL << 32) - 1, - .get_event_idx = intel_get_event_idx, -}; - -static __initconst struct x86_pmu intel_pmu = { - .name = "Intel", - .handle_irq = intel_pmu_handle_irq, - .disable_all = intel_pmu_disable_all, - .enable_all = intel_pmu_enable_all, - .enable = intel_pmu_enable_event, - .disable = intel_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, - .enable_bts = intel_pmu_enable_bts, - .disable_bts = intel_pmu_disable_bts, - .get_event_idx = intel_get_event_idx, -}; +/* Merge two pointer arrays */ +static __init struct attribute **merge_attr(struct attribute **a, struct attribute **b) +{ + struct attribute **new; + int j, i; -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, - .eventsel = MSR_K7_EVNTSEL0, - .perfctr = MSR_K7_PERFCTR0, - .event_map = amd_pmu_event_map, - .raw_event = amd_pmu_raw_event, - .max_events = ARRAY_SIZE(amd_perfmon_event_map), - .num_events = 4, - .event_bits = 48, - .event_mask = (1ULL << 48) - 1, - .apic = 1, - /* use highest bit to detect overflow */ - .max_period = (1ULL << 47) - 1, - .get_event_idx = gen_get_event_idx, -}; + for (j = 0; a[j]; j++) + ; + for (i = 0; b[i]; i++) + j++; + j++; -static __init int p6_pmu_init(void) -{ - switch (boot_cpu_data.x86_model) { - case 1: - case 3: /* Pentium Pro */ - case 5: - case 6: /* Pentium II */ - case 7: - case 8: - case 11: /* Pentium III */ - event_constraints = intel_p6_event_constraints; - break; - case 9: - case 13: - /* Pentium M */ - event_constraints = intel_p6_event_constraints; - break; - default: - pr_cont("unsupported p6 CPU model %d ", - boot_cpu_data.x86_model); - return -ENODEV; - } + new = kmalloc(sizeof(struct attribute *) * j, GFP_KERNEL); + if (!new) + return NULL; - x86_pmu = p6_pmu; + j = 0; + for (i = 0; a[i]; i++) + new[j++] = a[i]; + for (i = 0; b[i]; i++) + new[j++] = b[i]; + new[j] = NULL; - return 0; + return new; } -static __init int intel_pmu_init(void) +ssize_t events_sysfs_show(struct device *dev, struct device_attribute *attr, + char *page) { - union cpuid10_edx edx; - union cpuid10_eax eax; - unsigned int unused; - unsigned int ebx; - int version; + struct perf_pmu_events_attr *pmu_attr = \ + container_of(attr, struct perf_pmu_events_attr, attr); + u64 config = x86_pmu.event_map(pmu_attr->id); - if (!cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON)) { - /* check for P6 processor family */ - if (boot_cpu_data.x86 == 6) { - return p6_pmu_init(); - } else { - return -ENODEV; - } - } + /* string trumps id */ + if (pmu_attr->event_str) + return sprintf(page, "%s", pmu_attr->event_str); - /* - * 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) - return -ENODEV; + return x86_pmu.events_sysfs_show(page, config); +} - version = eax.split.version_id; - if (version < 2) - return -ENODEV; +EVENT_ATTR(cpu-cycles, CPU_CYCLES ); +EVENT_ATTR(instructions, INSTRUCTIONS ); +EVENT_ATTR(cache-references, CACHE_REFERENCES ); +EVENT_ATTR(cache-misses, CACHE_MISSES ); +EVENT_ATTR(branch-instructions, BRANCH_INSTRUCTIONS ); +EVENT_ATTR(branch-misses, BRANCH_MISSES ); +EVENT_ATTR(bus-cycles, BUS_CYCLES ); +EVENT_ATTR(stalled-cycles-frontend, STALLED_CYCLES_FRONTEND ); +EVENT_ATTR(stalled-cycles-backend, STALLED_CYCLES_BACKEND ); +EVENT_ATTR(ref-cycles, REF_CPU_CYCLES ); - x86_pmu = intel_pmu; - x86_pmu.version = version; - x86_pmu.num_events = eax.split.num_events; - x86_pmu.event_bits = eax.split.bit_width; - x86_pmu.event_mask = (1ULL << eax.split.bit_width) - 1; +static struct attribute *empty_attrs; - /* - * 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); +static struct attribute *events_attr[] = { + EVENT_PTR(CPU_CYCLES), + EVENT_PTR(INSTRUCTIONS), + EVENT_PTR(CACHE_REFERENCES), + EVENT_PTR(CACHE_MISSES), + EVENT_PTR(BRANCH_INSTRUCTIONS), + EVENT_PTR(BRANCH_MISSES), + EVENT_PTR(BUS_CYCLES), + EVENT_PTR(STALLED_CYCLES_FRONTEND), + EVENT_PTR(STALLED_CYCLES_BACKEND), + EVENT_PTR(REF_CPU_CYCLES), + NULL, +}; + +static struct attribute_group x86_pmu_events_group = { + .name = "events", + .attrs = events_attr, +}; + +ssize_t x86_event_sysfs_show(char *page, u64 config, u64 event) +{ + u64 umask = (config & ARCH_PERFMON_EVENTSEL_UMASK) >> 8; + u64 cmask = (config & ARCH_PERFMON_EVENTSEL_CMASK) >> 24; + bool edge = (config & ARCH_PERFMON_EVENTSEL_EDGE); + bool pc = (config & ARCH_PERFMON_EVENTSEL_PIN_CONTROL); + bool any = (config & ARCH_PERFMON_EVENTSEL_ANY); + bool inv = (config & ARCH_PERFMON_EVENTSEL_INV); + ssize_t ret; /* - * Install the hw-cache-events table: - */ - switch (boot_cpu_data.x86_model) { - 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" */ - case 29: /* six-core 45 nm xeon "Dunnington" */ - memcpy(hw_cache_event_ids, core2_hw_cache_event_ids, - sizeof(hw_cache_event_ids)); - - pr_cont("Core2 events, "); - event_constraints = intel_core_event_constraints; - break; - default: - case 26: - memcpy(hw_cache_event_ids, nehalem_hw_cache_event_ids, - sizeof(hw_cache_event_ids)); + * We have whole page size to spend and just little data + * to write, so we can safely use sprintf. + */ + ret = sprintf(page, "event=0x%02llx", event); - event_constraints = intel_nehalem_event_constraints; - pr_cont("Nehalem/Corei7 events, "); - break; - case 28: - memcpy(hw_cache_event_ids, atom_hw_cache_event_ids, - sizeof(hw_cache_event_ids)); + if (umask) + ret += sprintf(page + ret, ",umask=0x%02llx", umask); - pr_cont("Atom events, "); - break; - } - return 0; -} + if (edge) + ret += sprintf(page + ret, ",edge"); -static __init int amd_pmu_init(void) -{ - /* Performance-monitoring supported from K7 and later: */ - if (boot_cpu_data.x86 < 6) - return -ENODEV; + if (pc) + ret += sprintf(page + ret, ",pc"); - x86_pmu = amd_pmu; + if (any) + ret += sprintf(page + ret, ",any"); - /* Events are common for all AMDs */ - memcpy(hw_cache_event_ids, amd_hw_cache_event_ids, - sizeof(hw_cache_event_ids)); + if (inv) + ret += sprintf(page + ret, ",inv"); - return 0; -} + if (cmask) + ret += sprintf(page + ret, ",cmask=0x%02llx", cmask); -static void __init pmu_check_apic(void) -{ - if (cpu_has_apic) - return; + ret += sprintf(page + ret, "\n"); - x86_pmu.apic = 0; - pr_info("no APIC, boot with the \"lapic\" boot parameter to force-enable it.\n"); - pr_info("no hardware sampling interrupt available.\n"); + return ret; } -void __init init_hw_perf_events(void) +static int __init init_hw_perf_events(void) { + struct x86_pmu_quirk *quirk; int err; pr_info("Performance Events: "); @@ -2181,142 +1530,419 @@ void __init init_hw_perf_events(void) err = amd_pmu_init(); break; default: - return; + err = -ENOTSUPP; } if (err != 0) { pr_cont("no PMU driver, software events only.\n"); - return; + return 0; } pmu_check_apic(); + /* sanity check that the hardware exists or is emulated */ + if (!check_hw_exists()) + return 0; + pr_cont("%s PMU driver.\n", x86_pmu.name); - if (x86_pmu.num_events > X86_PMC_MAX_GENERIC) { - WARN(1, KERN_ERR "hw perf events %d > max(%d), clipping!", - x86_pmu.num_events, X86_PMC_MAX_GENERIC); - x86_pmu.num_events = X86_PMC_MAX_GENERIC; - } - perf_event_mask = (1 << x86_pmu.num_events) - 1; - perf_max_events = x86_pmu.num_events; + x86_pmu.attr_rdpmc = 1; /* enable userspace RDPMC usage by default */ - if (x86_pmu.num_events_fixed > X86_PMC_MAX_FIXED) { - WARN(1, KERN_ERR "hw perf events fixed %d > max(%d), clipping!", - x86_pmu.num_events_fixed, X86_PMC_MAX_FIXED); - x86_pmu.num_events_fixed = X86_PMC_MAX_FIXED; - } + for (quirk = x86_pmu.quirks; quirk; quirk = quirk->next) + quirk->func(); - perf_event_mask |= - ((1LL << x86_pmu.num_events_fixed)-1) << X86_PMC_IDX_FIXED; - x86_pmu.intel_ctrl = perf_event_mask; + if (!x86_pmu.intel_ctrl) + x86_pmu.intel_ctrl = (1 << x86_pmu.num_counters) - 1; perf_events_lapic_init(); - register_die_notifier(&perf_event_nmi_notifier); + register_nmi_handler(NMI_LOCAL, perf_event_nmi_handler, 0, "PMI"); + + unconstrained = (struct event_constraint) + __EVENT_CONSTRAINT(0, (1ULL << x86_pmu.num_counters) - 1, + 0, x86_pmu.num_counters, 0, 0); + + x86_pmu_format_group.attrs = x86_pmu.format_attrs; + + if (x86_pmu.event_attrs) + x86_pmu_events_group.attrs = x86_pmu.event_attrs; + + if (!x86_pmu.events_sysfs_show) + x86_pmu_events_group.attrs = &empty_attrs; + else + filter_events(x86_pmu_events_group.attrs); + + if (x86_pmu.cpu_events) { + struct attribute **tmp; + + tmp = merge_attr(x86_pmu_events_group.attrs, x86_pmu.cpu_events); + if (!WARN_ON(!tmp)) + x86_pmu_events_group.attrs = tmp; + } pr_info("... version: %d\n", x86_pmu.version); - pr_info("... bit width: %d\n", x86_pmu.event_bits); - pr_info("... generic registers: %d\n", x86_pmu.num_events); - pr_info("... value mask: %016Lx\n", x86_pmu.event_mask); + pr_info("... bit width: %d\n", x86_pmu.cntval_bits); + pr_info("... generic registers: %d\n", x86_pmu.num_counters); + pr_info("... value mask: %016Lx\n", x86_pmu.cntval_mask); pr_info("... max period: %016Lx\n", x86_pmu.max_period); - pr_info("... fixed-purpose events: %d\n", x86_pmu.num_events_fixed); - pr_info("... event mask: %016Lx\n", perf_event_mask); + pr_info("... fixed-purpose events: %d\n", x86_pmu.num_counters_fixed); + pr_info("... event mask: %016Lx\n", x86_pmu.intel_ctrl); + + perf_pmu_register(&pmu, "cpu", PERF_TYPE_RAW); + perf_cpu_notifier(x86_pmu_notifier); + + return 0; } +early_initcall(init_hw_perf_events); static inline void x86_pmu_read(struct perf_event *event) { - x86_perf_event_update(event, &event->hw, event->hw.idx); + x86_perf_event_update(event); } -static const struct pmu pmu = { - .enable = x86_pmu_enable, - .disable = x86_pmu_disable, - .read = x86_pmu_read, - .unthrottle = x86_pmu_unthrottle, -}; +/* + * Start group events scheduling transaction + * Set the flag to make pmu::enable() not perform the + * schedulability test, it will be performed at commit time + */ +static void x86_pmu_start_txn(struct pmu *pmu) +{ + perf_pmu_disable(pmu); + __this_cpu_or(cpu_hw_events.group_flag, PERF_EVENT_TXN); + __this_cpu_write(cpu_hw_events.n_txn, 0); +} -static int -validate_event(struct cpu_hw_events *cpuc, struct perf_event *event) +/* + * Stop group events scheduling transaction + * Clear the flag and pmu::enable() will perform the + * schedulability test. + */ +static void x86_pmu_cancel_txn(struct pmu *pmu) { - struct hw_perf_event fake_event = event->hw; + __this_cpu_and(cpu_hw_events.group_flag, ~PERF_EVENT_TXN); + /* + * Truncate collected array by the number of events added in this + * transaction. See x86_pmu_add() and x86_pmu_*_txn(). + */ + __this_cpu_sub(cpu_hw_events.n_added, __this_cpu_read(cpu_hw_events.n_txn)); + __this_cpu_sub(cpu_hw_events.n_events, __this_cpu_read(cpu_hw_events.n_txn)); + perf_pmu_enable(pmu); +} - if (event->pmu && event->pmu != &pmu) - return 0; +/* + * Commit group events scheduling transaction + * Perform the group schedulability test as a whole + * Return 0 if success + * + * Does not cancel the transaction on failure; expects the caller to do this. + */ +static int x86_pmu_commit_txn(struct pmu *pmu) +{ + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + int assign[X86_PMC_IDX_MAX]; + int n, ret; + + n = cpuc->n_events; + + if (!x86_pmu_initialized()) + return -EAGAIN; + + ret = x86_pmu.schedule_events(cpuc, n, assign); + if (ret) + return ret; + + /* + * copy new assignment, now we know it is possible + * will be used by hw_perf_enable() + */ + memcpy(cpuc->assign, assign, n*sizeof(int)); + + cpuc->group_flag &= ~PERF_EVENT_TXN; + perf_pmu_enable(pmu); + return 0; +} +/* + * a fake_cpuc is used to validate event groups. Due to + * the extra reg logic, we need to also allocate a fake + * per_core and per_cpu structure. Otherwise, group events + * using extra reg may conflict without the kernel being + * able to catch this when the last event gets added to + * the group. + */ +static void free_fake_cpuc(struct cpu_hw_events *cpuc) +{ + kfree(cpuc->shared_regs); + kfree(cpuc); +} + +static struct cpu_hw_events *allocate_fake_cpuc(void) +{ + struct cpu_hw_events *cpuc; + int cpu = raw_smp_processor_id(); + + cpuc = kzalloc(sizeof(*cpuc), GFP_KERNEL); + if (!cpuc) + return ERR_PTR(-ENOMEM); + + /* only needed, if we have extra_regs */ + if (x86_pmu.extra_regs) { + cpuc->shared_regs = allocate_shared_regs(cpu); + if (!cpuc->shared_regs) + goto error; + } + cpuc->is_fake = 1; + return cpuc; +error: + free_fake_cpuc(cpuc); + return ERR_PTR(-ENOMEM); +} + +/* + * validate that we can schedule this event + */ +static int validate_event(struct perf_event *event) +{ + struct cpu_hw_events *fake_cpuc; + struct event_constraint *c; + int ret = 0; + + fake_cpuc = allocate_fake_cpuc(); + if (IS_ERR(fake_cpuc)) + return PTR_ERR(fake_cpuc); + + c = x86_pmu.get_event_constraints(fake_cpuc, event); + + if (!c || !c->weight) + ret = -EINVAL; + + if (x86_pmu.put_event_constraints) + x86_pmu.put_event_constraints(fake_cpuc, event); + + free_fake_cpuc(fake_cpuc); - return x86_schedule_event(cpuc, &fake_event) >= 0; + return ret; } +/* + * validate a single event group + * + * validation include: + * - check events are compatible which each other + * - events do not compete for the same counter + * - number of events <= number of counters + * + * validation ensures the group can be loaded onto the + * PMU if it was the only group available. + */ static int validate_group(struct perf_event *event) { - struct perf_event *sibling, *leader = event->group_leader; - struct cpu_hw_events fake_pmu; + struct perf_event *leader = event->group_leader; + struct cpu_hw_events *fake_cpuc; + int ret = -EINVAL, n; - memset(&fake_pmu, 0, sizeof(fake_pmu)); + fake_cpuc = allocate_fake_cpuc(); + if (IS_ERR(fake_cpuc)) + return PTR_ERR(fake_cpuc); + /* + * the event is not yet connected with its + * siblings therefore we must first collect + * existing siblings, then add the new event + * before we can simulate the scheduling + */ + n = collect_events(fake_cpuc, leader, true); + if (n < 0) + goto out; - if (!validate_event(&fake_pmu, leader)) - return -ENOSPC; + fake_cpuc->n_events = n; + n = collect_events(fake_cpuc, event, false); + if (n < 0) + goto out; - list_for_each_entry(sibling, &leader->sibling_list, group_entry) { - if (!validate_event(&fake_pmu, sibling)) - return -ENOSPC; - } + fake_cpuc->n_events = n; - if (!validate_event(&fake_pmu, event)) - return -ENOSPC; + ret = x86_pmu.schedule_events(fake_cpuc, n, NULL); - return 0; +out: + free_fake_cpuc(fake_cpuc); + return ret; } -const struct pmu *hw_perf_event_init(struct perf_event *event) +static int x86_pmu_event_init(struct perf_event *event) { + struct pmu *tmp; int err; - err = __hw_perf_event_init(event); + switch (event->attr.type) { + case PERF_TYPE_RAW: + case PERF_TYPE_HARDWARE: + case PERF_TYPE_HW_CACHE: + break; + + default: + return -ENOENT; + } + + err = __x86_pmu_event_init(event); if (!err) { + /* + * we temporarily connect event to its pmu + * such that validate_group() can classify + * it as an x86 event using is_x86_event() + */ + tmp = event->pmu; + event->pmu = &pmu; + if (event->group_leader != event) err = validate_group(event); + else + err = validate_event(event); + + event->pmu = tmp; } if (err) { if (event->destroy) event->destroy(event); - return ERR_PTR(err); } - return &pmu; + return err; } -/* - * callchain support - */ +static int x86_pmu_event_idx(struct perf_event *event) +{ + int idx = event->hw.idx; + + if (!x86_pmu.attr_rdpmc) + return 0; + + if (x86_pmu.num_counters_fixed && idx >= INTEL_PMC_IDX_FIXED) { + idx -= INTEL_PMC_IDX_FIXED; + idx |= 1 << 30; + } + + return idx + 1; +} -static inline -void callchain_store(struct perf_callchain_entry *entry, u64 ip) +static ssize_t get_attr_rdpmc(struct device *cdev, + struct device_attribute *attr, + char *buf) { - if (entry->nr < PERF_MAX_STACK_DEPTH) - entry->ip[entry->nr++] = ip; + return snprintf(buf, 40, "%d\n", x86_pmu.attr_rdpmc); } -static DEFINE_PER_CPU(struct perf_callchain_entry, pmc_irq_entry); -static DEFINE_PER_CPU(struct perf_callchain_entry, pmc_nmi_entry); -static DEFINE_PER_CPU(int, in_ignored_frame); +static void change_rdpmc(void *info) +{ + bool enable = !!(unsigned long)info; + if (enable) + set_in_cr4(X86_CR4_PCE); + else + clear_in_cr4(X86_CR4_PCE); +} -static void -backtrace_warning_symbol(void *data, char *msg, unsigned long symbol) +static ssize_t set_attr_rdpmc(struct device *cdev, + struct device_attribute *attr, + const char *buf, size_t count) { - /* Ignore warnings */ + unsigned long val; + ssize_t ret; + + ret = kstrtoul(buf, 0, &val); + if (ret) + return ret; + + if (x86_pmu.attr_rdpmc_broken) + return -ENOTSUPP; + + if (!!val != !!x86_pmu.attr_rdpmc) { + x86_pmu.attr_rdpmc = !!val; + on_each_cpu(change_rdpmc, (void *)val, 1); + } + + return count; } -static void backtrace_warning(void *data, char *msg) +static DEVICE_ATTR(rdpmc, S_IRUSR | S_IWUSR, get_attr_rdpmc, set_attr_rdpmc); + +static struct attribute *x86_pmu_attrs[] = { + &dev_attr_rdpmc.attr, + NULL, +}; + +static struct attribute_group x86_pmu_attr_group = { + .attrs = x86_pmu_attrs, +}; + +static const struct attribute_group *x86_pmu_attr_groups[] = { + &x86_pmu_attr_group, + &x86_pmu_format_group, + &x86_pmu_events_group, + NULL, +}; + +static void x86_pmu_flush_branch_stack(void) { - /* Ignore warnings */ + if (x86_pmu.flush_branch_stack) + x86_pmu.flush_branch_stack(); } -static int backtrace_stack(void *data, char *name) +void perf_check_microcode(void) { - per_cpu(in_ignored_frame, smp_processor_id()) = - x86_is_stack_id(NMI_STACK, name) || - x86_is_stack_id(DEBUG_STACK, name); + if (x86_pmu.check_microcode) + x86_pmu.check_microcode(); +} +EXPORT_SYMBOL_GPL(perf_check_microcode); + +static struct pmu pmu = { + .pmu_enable = x86_pmu_enable, + .pmu_disable = x86_pmu_disable, + + .attr_groups = x86_pmu_attr_groups, + + .event_init = x86_pmu_event_init, + + .add = x86_pmu_add, + .del = x86_pmu_del, + .start = x86_pmu_start, + .stop = x86_pmu_stop, + .read = x86_pmu_read, + .start_txn = x86_pmu_start_txn, + .cancel_txn = x86_pmu_cancel_txn, + .commit_txn = x86_pmu_commit_txn, + + .event_idx = x86_pmu_event_idx, + .flush_branch_stack = x86_pmu_flush_branch_stack, +}; + +void arch_perf_update_userpage(struct perf_event_mmap_page *userpg, u64 now) +{ + struct cyc2ns_data *data; + + userpg->cap_user_time = 0; + userpg->cap_user_time_zero = 0; + userpg->cap_user_rdpmc = x86_pmu.attr_rdpmc; + userpg->pmc_width = x86_pmu.cntval_bits; + + if (!sched_clock_stable()) + return; + + data = cyc2ns_read_begin(); + + userpg->cap_user_time = 1; + userpg->time_mult = data->cyc2ns_mul; + userpg->time_shift = data->cyc2ns_shift; + userpg->time_offset = data->cyc2ns_offset - now; + + userpg->cap_user_time_zero = 1; + userpg->time_zero = data->cyc2ns_offset; + + cyc2ns_read_end(data); +} + +/* + * callchain support + */ + +static int backtrace_stack(void *data, char *name) +{ return 0; } @@ -2324,145 +1950,221 @@ static void backtrace_address(void *data, unsigned long addr, int reliable) { struct perf_callchain_entry *entry = data; - if (per_cpu(in_ignored_frame, smp_processor_id())) - return; - - if (reliable) - callchain_store(entry, addr); + perf_callchain_store(entry, addr); } static const struct stacktrace_ops backtrace_ops = { - .warning = backtrace_warning, - .warning_symbol = backtrace_warning_symbol, .stack = backtrace_stack, .address = backtrace_address, .walk_stack = print_context_stack_bp, }; -#include "../dumpstack.h" - -static void -perf_callchain_kernel(struct pt_regs *regs, struct perf_callchain_entry *entry) +void +perf_callchain_kernel(struct perf_callchain_entry *entry, struct pt_regs *regs) { - callchain_store(entry, PERF_CONTEXT_KERNEL); - callchain_store(entry, regs->ip); + if (perf_guest_cbs && perf_guest_cbs->is_in_guest()) { + /* TODO: We don't support guest os callchain now */ + return; + } + + perf_callchain_store(entry, regs->ip); - dump_trace(NULL, regs, NULL, regs->bp, &backtrace_ops, entry); + dump_trace(NULL, regs, NULL, 0, &backtrace_ops, entry); } -/* - * best effort, GUP based copy_from_user() that assumes IRQ or NMI context - */ -static unsigned long -copy_from_user_nmi(void *to, const void __user *from, unsigned long n) -{ - unsigned long offset, addr = (unsigned long)from; - int type = in_nmi() ? KM_NMI : KM_IRQ0; - unsigned long size, len = 0; - struct page *page; - void *map; - int ret; +static inline int +valid_user_frame(const void __user *fp, unsigned long size) +{ + return (__range_not_ok(fp, size, TASK_SIZE) == 0); +} - do { - ret = __get_user_pages_fast(addr, 1, 0, &page); - if (!ret) - break; +static unsigned long get_segment_base(unsigned int segment) +{ + struct desc_struct *desc; + int idx = segment >> 3; - offset = addr & (PAGE_SIZE - 1); - size = min(PAGE_SIZE - offset, n - len); + if ((segment & SEGMENT_TI_MASK) == SEGMENT_LDT) { + if (idx > LDT_ENTRIES) + return 0; - map = kmap_atomic(page, type); - memcpy(to, map+offset, size); - kunmap_atomic(map, type); - put_page(page); + if (idx > current->active_mm->context.size) + return 0; - len += size; - to += size; - addr += size; + desc = current->active_mm->context.ldt; + } else { + if (idx > GDT_ENTRIES) + return 0; - } while (len < n); + desc = __this_cpu_ptr(&gdt_page.gdt[0]); + } - return len; + return get_desc_base(desc + idx); } -static int copy_stack_frame(const void __user *fp, struct stack_frame *frame) +#ifdef CONFIG_COMPAT + +#include <asm/compat.h> + +static inline int +perf_callchain_user32(struct pt_regs *regs, struct perf_callchain_entry *entry) { - unsigned long bytes; + /* 32-bit process in 64-bit kernel. */ + unsigned long ss_base, cs_base; + struct stack_frame_ia32 frame; + const void __user *fp; + + if (!test_thread_flag(TIF_IA32)) + return 0; + + cs_base = get_segment_base(regs->cs); + ss_base = get_segment_base(regs->ss); + + fp = compat_ptr(ss_base + regs->bp); + while (entry->nr < PERF_MAX_STACK_DEPTH) { + unsigned long bytes; + frame.next_frame = 0; + frame.return_address = 0; + + bytes = copy_from_user_nmi(&frame, fp, sizeof(frame)); + if (bytes != 0) + break; - bytes = copy_from_user_nmi(frame, fp, sizeof(*frame)); + if (!valid_user_frame(fp, sizeof(frame))) + break; - return bytes == sizeof(*frame); + perf_callchain_store(entry, cs_base + frame.return_address); + fp = compat_ptr(ss_base + frame.next_frame); + } + return 1; } +#else +static inline int +perf_callchain_user32(struct pt_regs *regs, struct perf_callchain_entry *entry) +{ + return 0; +} +#endif -static void -perf_callchain_user(struct pt_regs *regs, struct perf_callchain_entry *entry) +void +perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs) { struct stack_frame frame; const void __user *fp; - if (!user_mode(regs)) - regs = task_pt_regs(current); + if (perf_guest_cbs && perf_guest_cbs->is_in_guest()) { + /* TODO: We don't support guest os callchain now */ + return; + } + + /* + * We don't know what to do with VM86 stacks.. ignore them for now. + */ + if (regs->flags & (X86_VM_MASK | PERF_EFLAGS_VM)) + return; fp = (void __user *)regs->bp; - callchain_store(entry, PERF_CONTEXT_USER); - callchain_store(entry, regs->ip); + perf_callchain_store(entry, regs->ip); + + if (!current->mm) + return; + + if (perf_callchain_user32(regs, entry)) + return; while (entry->nr < PERF_MAX_STACK_DEPTH) { + unsigned long bytes; frame.next_frame = NULL; frame.return_address = 0; - if (!copy_stack_frame(fp, &frame)) + bytes = copy_from_user_nmi(&frame, fp, sizeof(frame)); + if (bytes != 0) break; - if ((unsigned long)fp < regs->sp) + if (!valid_user_frame(fp, sizeof(frame))) break; - callchain_store(entry, frame.return_address); + perf_callchain_store(entry, frame.return_address); fp = frame.next_frame; } } -static void -perf_do_callchain(struct pt_regs *regs, struct perf_callchain_entry *entry) +/* + * Deal with code segment offsets for the various execution modes: + * + * VM86 - the good olde 16 bit days, where the linear address is + * 20 bits and we use regs->ip + 0x10 * regs->cs. + * + * IA32 - Where we need to look at GDT/LDT segment descriptor tables + * to figure out what the 32bit base address is. + * + * X32 - has TIF_X32 set, but is running in x86_64 + * + * X86_64 - CS,DS,SS,ES are all zero based. + */ +static unsigned long code_segment_base(struct pt_regs *regs) { - int is_user; - - if (!regs) - return; - - is_user = user_mode(regs); - - if (!current || current->pid == 0) - return; + /* + * If we are in VM86 mode, add the segment offset to convert to a + * linear address. + */ + if (regs->flags & X86_VM_MASK) + return 0x10 * regs->cs; - if (is_user && current->state != TASK_RUNNING) - return; + /* + * For IA32 we look at the GDT/LDT segment base to convert the + * effective IP to a linear address. + */ +#ifdef CONFIG_X86_32 + if (user_mode(regs) && regs->cs != __USER_CS) + return get_segment_base(regs->cs); +#else + if (test_thread_flag(TIF_IA32)) { + if (user_mode(regs) && regs->cs != __USER32_CS) + return get_segment_base(regs->cs); + } +#endif + return 0; +} - if (!is_user) - perf_callchain_kernel(regs, entry); +unsigned long perf_instruction_pointer(struct pt_regs *regs) +{ + if (perf_guest_cbs && perf_guest_cbs->is_in_guest()) + return perf_guest_cbs->get_guest_ip(); - if (current->mm) - perf_callchain_user(regs, entry); + return regs->ip + code_segment_base(regs); } -struct perf_callchain_entry *perf_callchain(struct pt_regs *regs) +unsigned long perf_misc_flags(struct pt_regs *regs) { - struct perf_callchain_entry *entry; + int misc = 0; - if (in_nmi()) - entry = &__get_cpu_var(pmc_nmi_entry); - else - entry = &__get_cpu_var(pmc_irq_entry); - - entry->nr = 0; + if (perf_guest_cbs && perf_guest_cbs->is_in_guest()) { + if (perf_guest_cbs->is_user_mode()) + misc |= PERF_RECORD_MISC_GUEST_USER; + else + misc |= PERF_RECORD_MISC_GUEST_KERNEL; + } else { + if (user_mode(regs)) + misc |= PERF_RECORD_MISC_USER; + else + misc |= PERF_RECORD_MISC_KERNEL; + } - perf_do_callchain(regs, entry); + if (regs->flags & PERF_EFLAGS_EXACT) + misc |= PERF_RECORD_MISC_EXACT_IP; - return entry; + return misc; } -void hw_perf_event_setup_online(int cpu) +void perf_get_x86_pmu_capability(struct x86_pmu_capability *cap) { - init_debug_store_on_cpu(cpu); + cap->version = x86_pmu.version; + cap->num_counters_gp = x86_pmu.num_counters; + cap->num_counters_fixed = x86_pmu.num_counters_fixed; + cap->bit_width_gp = x86_pmu.cntval_bits; + cap->bit_width_fixed = x86_pmu.cntval_bits; + cap->events_mask = (unsigned int)x86_pmu.events_maskl; + cap->events_mask_len = x86_pmu.events_mask_len; } +EXPORT_SYMBOL_GPL(perf_get_x86_pmu_capability); |