diff options
| -rw-r--r-- | Documentation/ABI/testing/sysfs-devices-mapping | 30 | ||||
| -rw-r--r-- | arch/arm64/kernel/perf_event.c | 18 | ||||
| -rw-r--r-- | arch/powerpc/perf/core-book3s.c | 8 | ||||
| -rw-r--r-- | arch/s390/kernel/perf_pai_crypto.c | 2 | ||||
| -rw-r--r-- | arch/s390/kernel/perf_pai_ext.c | 2 | ||||
| -rw-r--r-- | arch/x86/events/amd/brs.c | 2 | ||||
| -rw-r--r-- | arch/x86/events/amd/ibs.c | 4 | ||||
| -rw-r--r-- | arch/x86/events/amd/lbr.c | 6 | ||||
| -rw-r--r-- | arch/x86/events/core.c | 48 | ||||
| -rw-r--r-- | arch/x86/events/intel/core.c | 23 | ||||
| -rw-r--r-- | arch/x86/events/intel/ds.c | 4 | ||||
| -rw-r--r-- | arch/x86/events/intel/lbr.c | 30 | ||||
| -rw-r--r-- | arch/x86/events/intel/uncore.h | 24 | ||||
| -rw-r--r-- | arch/x86/events/intel/uncore_snb.c | 3 | ||||
| -rw-r--r-- | arch/x86/events/intel/uncore_snbep.c | 495 | ||||
| -rw-r--r-- | arch/x86/events/perf_event.h | 31 | ||||
| -rw-r--r-- | drivers/perf/arm_pmu.c | 16 | ||||
| -rw-r--r-- | include/linux/perf/arm_pmu.h | 2 | ||||
| -rw-r--r-- | include/linux/perf_event.h | 125 | ||||
| -rw-r--r-- | include/linux/sched.h | 2 | ||||
| -rw-r--r-- | kernel/events/core.c | 2100 |
21 files changed, 1755 insertions, 1220 deletions
diff --git a/Documentation/ABI/testing/sysfs-devices-mapping b/Documentation/ABI/testing/sysfs-devices-mapping index 8d202bac9394..2eee1446ad4c 100644 --- a/Documentation/ABI/testing/sysfs-devices-mapping +++ b/Documentation/ABI/testing/sysfs-devices-mapping @@ -1,6 +1,6 @@ What: /sys/devices/uncore_iio_x/dieX Date: February 2020 -Contact: Roman Sudarikov <roman.sudarikov@linux.intel.com> +Contact: Alexander Antonov <alexander.antonov@linux.intel.com> Description: Each IIO stack (PCIe root port) has its own IIO PMON block, so each dieX file (where X is die number) holds "Segment:Root Bus" @@ -32,3 +32,31 @@ Description: IIO PMU 0 on die 1 belongs to PCI RP on bus 0x40, domain 0x0000 IIO PMU 0 on die 2 belongs to PCI RP on bus 0x80, domain 0x0000 IIO PMU 0 on die 3 belongs to PCI RP on bus 0xc0, domain 0x0000 + +What: /sys/devices/uncore_upi_x/dieX +Date: March 2022 +Contact: Alexander Antonov <alexander.antonov@linux.intel.com> +Description: + Each /sys/devices/uncore_upi_X/dieY file holds "upi_Z,die_W" + value that means UPI link number X on die Y is connected to UPI + link Z on die W and this link between sockets can be monitored + by UPI PMON block. + For example, 4-die Sapphire Rapids platform has the following + UPI 0 topology:: + + # tail /sys/devices/uncore_upi_0/die* + ==> /sys/devices/uncore_upi_0/die0 <== + upi_1,die_1 + ==> /sys/devices/uncore_upi_0/die1 <== + upi_0,die_3 + ==> /sys/devices/uncore_upi_0/die2 <== + upi_1,die_3 + ==> /sys/devices/uncore_upi_0/die3 <== + upi_0,die_1 + + Which means:: + + UPI link 0 on die 0 is connected to UPI link 1 on die 1 + UPI link 0 on die 1 is connected to UPI link 0 on die 3 + UPI link 0 on die 2 is connected to UPI link 1 on die 3 + UPI link 0 on die 3 is connected to UPI link 0 on die 1
\ No newline at end of file diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c index a15b3c1d15d9..a5193f2146a6 100644 --- a/arch/arm64/kernel/perf_event.c +++ b/arch/arm64/kernel/perf_event.c @@ -806,10 +806,14 @@ static void armv8pmu_disable_event(struct perf_event *event) static void armv8pmu_start(struct arm_pmu *cpu_pmu) { - struct perf_event_context *task_ctx = - this_cpu_ptr(cpu_pmu->pmu.pmu_cpu_context)->task_ctx; + struct perf_event_context *ctx; + int nr_user = 0; - if (sysctl_perf_user_access && task_ctx && task_ctx->nr_user) + ctx = perf_cpu_task_ctx(); + if (ctx) + nr_user = ctx->nr_user; + + if (sysctl_perf_user_access && nr_user) armv8pmu_enable_user_access(cpu_pmu); else armv8pmu_disable_user_access(); @@ -1019,10 +1023,10 @@ static int armv8pmu_set_event_filter(struct hw_perf_event *event, return 0; } -static int armv8pmu_filter_match(struct perf_event *event) +static bool armv8pmu_filter(struct pmu *pmu, int cpu) { - unsigned long evtype = event->hw.config_base & ARMV8_PMU_EVTYPE_EVENT; - return evtype != ARMV8_PMUV3_PERFCTR_CHAIN; + struct arm_pmu *armpmu = to_arm_pmu(pmu); + return !cpumask_test_cpu(smp_processor_id(), &armpmu->supported_cpus); } static void armv8pmu_reset(void *info) @@ -1254,7 +1258,7 @@ static int armv8_pmu_init(struct arm_pmu *cpu_pmu, char *name, cpu_pmu->stop = armv8pmu_stop; cpu_pmu->reset = armv8pmu_reset; cpu_pmu->set_event_filter = armv8pmu_set_event_filter; - cpu_pmu->filter_match = armv8pmu_filter_match; + cpu_pmu->filter = armv8pmu_filter; cpu_pmu->pmu.event_idx = armv8pmu_user_event_idx; diff --git a/arch/powerpc/perf/core-book3s.c b/arch/powerpc/perf/core-book3s.c index 942aa830e110..bf318dd9b709 100644 --- a/arch/powerpc/perf/core-book3s.c +++ b/arch/powerpc/perf/core-book3s.c @@ -132,7 +132,7 @@ static unsigned long ebb_switch_in(bool ebb, struct cpu_hw_events *cpuhw) static inline void power_pmu_bhrb_enable(struct perf_event *event) {} static inline void power_pmu_bhrb_disable(struct perf_event *event) {} -static void power_pmu_sched_task(struct perf_event_context *ctx, bool sched_in) {} +static void power_pmu_sched_task(struct perf_event_pmu_context *pmu_ctx, bool sched_in) {} static inline void power_pmu_bhrb_read(struct perf_event *event, struct cpu_hw_events *cpuhw) {} static void pmao_restore_workaround(bool ebb) { } #endif /* CONFIG_PPC32 */ @@ -424,7 +424,7 @@ static void power_pmu_bhrb_enable(struct perf_event *event) cpuhw->bhrb_context = event->ctx; } cpuhw->bhrb_users++; - perf_sched_cb_inc(event->ctx->pmu); + perf_sched_cb_inc(event->pmu); } static void power_pmu_bhrb_disable(struct perf_event *event) @@ -436,7 +436,7 @@ static void power_pmu_bhrb_disable(struct perf_event *event) WARN_ON_ONCE(!cpuhw->bhrb_users); cpuhw->bhrb_users--; - perf_sched_cb_dec(event->ctx->pmu); + perf_sched_cb_dec(event->pmu); if (!cpuhw->disabled && !cpuhw->bhrb_users) { /* BHRB cannot be turned off when other @@ -451,7 +451,7 @@ static void power_pmu_bhrb_disable(struct perf_event *event) /* Called from ctxsw to prevent one process's branch entries to * mingle with the other process's entries during context switch. */ -static void power_pmu_sched_task(struct perf_event_context *ctx, bool sched_in) +static void power_pmu_sched_task(struct perf_event_pmu_context *pmu_ctx, bool sched_in) { if (!ppmu->bhrb_nr) return; diff --git a/arch/s390/kernel/perf_pai_crypto.c b/arch/s390/kernel/perf_pai_crypto.c index 529a2fee4ea5..985e243a2ed8 100644 --- a/arch/s390/kernel/perf_pai_crypto.c +++ b/arch/s390/kernel/perf_pai_crypto.c @@ -377,7 +377,7 @@ static int paicrypt_push_sample(void) /* Called on schedule-in and schedule-out. No access to event structure, * but for sampling only event CRYPTO_ALL is allowed. */ -static void paicrypt_sched_task(struct perf_event_context *ctx, bool sched_in) +static void paicrypt_sched_task(struct perf_event_pmu_context *pmu_ctx, bool sched_in) { /* We started with a clean page on event installation. So read out * results on schedule_out and if page was dirty, clear values. diff --git a/arch/s390/kernel/perf_pai_ext.c b/arch/s390/kernel/perf_pai_ext.c index a46cd7406b20..1138f57baae3 100644 --- a/arch/s390/kernel/perf_pai_ext.c +++ b/arch/s390/kernel/perf_pai_ext.c @@ -466,7 +466,7 @@ static int paiext_push_sample(void) /* Called on schedule-in and schedule-out. No access to event structure, * but for sampling only event NNPA_ALL is allowed. */ -static void paiext_sched_task(struct perf_event_context *ctx, bool sched_in) +static void paiext_sched_task(struct perf_event_pmu_context *pmu_ctx, bool sched_in) { /* We started with a clean page on event installation. So read out * results on schedule_out and if page was dirty, clear values. diff --git a/arch/x86/events/amd/brs.c b/arch/x86/events/amd/brs.c index f1bff153d945..58461fa18b6f 100644 --- a/arch/x86/events/amd/brs.c +++ b/arch/x86/events/amd/brs.c @@ -384,7 +384,7 @@ static void amd_brs_poison_buffer(void) * On ctxswin, sched_in = true, called after the PMU has started * On ctxswout, sched_in = false, called before the PMU is stopped */ -void amd_pmu_brs_sched_task(struct perf_event_context *ctx, bool sched_in) +void amd_pmu_brs_sched_task(struct perf_event_pmu_context *pmu_ctx, bool sched_in) { struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events); diff --git a/arch/x86/events/amd/ibs.c b/arch/x86/events/amd/ibs.c index 4cb710efbdd9..da3f5ebac4e1 100644 --- a/arch/x86/events/amd/ibs.c +++ b/arch/x86/events/amd/ibs.c @@ -631,7 +631,7 @@ static const struct attribute_group *op_attr_update[] = { static struct perf_ibs perf_ibs_fetch = { .pmu = { - .task_ctx_nr = perf_invalid_context, + .task_ctx_nr = perf_hw_context, .event_init = perf_ibs_init, .add = perf_ibs_add, @@ -655,7 +655,7 @@ static struct perf_ibs perf_ibs_fetch = { static struct perf_ibs perf_ibs_op = { .pmu = { - .task_ctx_nr = perf_invalid_context, + .task_ctx_nr = perf_hw_context, .event_init = perf_ibs_init, .add = perf_ibs_add, diff --git a/arch/x86/events/amd/lbr.c b/arch/x86/events/amd/lbr.c index 38a75216c12c..eb31f850841a 100644 --- a/arch/x86/events/amd/lbr.c +++ b/arch/x86/events/amd/lbr.c @@ -352,7 +352,7 @@ void amd_pmu_lbr_add(struct perf_event *event) cpuc->br_sel = reg->reg; } - perf_sched_cb_inc(event->ctx->pmu); + perf_sched_cb_inc(event->pmu); if (!cpuc->lbr_users++ && !event->total_time_running) amd_pmu_lbr_reset(); @@ -370,10 +370,10 @@ void amd_pmu_lbr_del(struct perf_event *event) cpuc->lbr_users--; WARN_ON_ONCE(cpuc->lbr_users < 0); - perf_sched_cb_dec(event->ctx->pmu); + perf_sched_cb_dec(event->pmu); } -void amd_pmu_lbr_sched_task(struct perf_event_context *ctx, bool sched_in) +void amd_pmu_lbr_sched_task(struct perf_event_pmu_context *pmu_ctx, bool sched_in) { struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events); diff --git a/arch/x86/events/core.c b/arch/x86/events/core.c index b30b8bbcd1e2..85a63a41c471 100644 --- a/arch/x86/events/core.c +++ b/arch/x86/events/core.c @@ -90,6 +90,8 @@ DEFINE_STATIC_CALL_NULL(x86_pmu_swap_task_ctx, *x86_pmu.swap_task_ctx); DEFINE_STATIC_CALL_NULL(x86_pmu_drain_pebs, *x86_pmu.drain_pebs); DEFINE_STATIC_CALL_NULL(x86_pmu_pebs_aliases, *x86_pmu.pebs_aliases); +DEFINE_STATIC_CALL_NULL(x86_pmu_filter, *x86_pmu.filter); + /* * This one is magic, it will get called even when PMU init fails (because * there is no PMU), in which case it should simply return NULL. @@ -2031,6 +2033,7 @@ static void x86_pmu_static_call_update(void) static_call_update(x86_pmu_pebs_aliases, x86_pmu.pebs_aliases); static_call_update(x86_pmu_guest_get_msrs, x86_pmu.guest_get_msrs); + static_call_update(x86_pmu_filter, x86_pmu.filter); } static void _x86_pmu_read(struct perf_event *event) @@ -2052,23 +2055,6 @@ void x86_pmu_show_pmu_cap(int num_counters, int num_counters_fixed, pr_info("... event mask: %016Lx\n", intel_ctrl); } -/* - * The generic code is not hybrid friendly. The hybrid_pmu->pmu - * of the first registered PMU is unconditionally assigned to - * each possible cpuctx->ctx.pmu. - * Update the correct hybrid PMU to the cpuctx->ctx.pmu. - */ -void x86_pmu_update_cpu_context(struct pmu *pmu, int cpu) -{ - struct perf_cpu_context *cpuctx; - - if (!pmu->pmu_cpu_context) - return; - - cpuctx = per_cpu_ptr(pmu->pmu_cpu_context, cpu); - cpuctx->ctx.pmu = pmu; -} - static int __init init_hw_perf_events(void) { struct x86_pmu_quirk *quirk; @@ -2175,13 +2161,9 @@ static int __init init_hw_perf_events(void) if (err) goto out2; } else { - u8 cpu_type = get_this_hybrid_cpu_type(); struct x86_hybrid_pmu *hybrid_pmu; int i, j; - if (!cpu_type && x86_pmu.get_hybrid_cpu_type) - cpu_type = x86_pmu.get_hybrid_cpu_type(); - for (i = 0; i < x86_pmu.num_hybrid_pmus; i++) { hybrid_pmu = &x86_pmu.hybrid_pmu[i]; @@ -2195,9 +2177,6 @@ static int __init init_hw_perf_events(void) (hybrid_pmu->cpu_type == hybrid_big) ? PERF_TYPE_RAW : -1); if (err) break; - - if (cpu_type == hybrid_pmu->cpu_type) - x86_pmu_update_cpu_context(&hybrid_pmu->pmu, raw_smp_processor_id()); } if (i < x86_pmu.num_hybrid_pmus) { @@ -2646,15 +2625,15 @@ static const struct attribute_group *x86_pmu_attr_groups[] = { NULL, }; -static void x86_pmu_sched_task(struct perf_event_context *ctx, bool sched_in) +static void x86_pmu_sched_task(struct perf_event_pmu_context *pmu_ctx, bool sched_in) { - static_call_cond(x86_pmu_sched_task)(ctx, sched_in); + static_call_cond(x86_pmu_sched_task)(pmu_ctx, sched_in); } -static void x86_pmu_swap_task_ctx(struct perf_event_context *prev, - struct perf_event_context *next) +static void x86_pmu_swap_task_ctx(struct perf_event_pmu_context *prev_epc, + struct perf_event_pmu_context *next_epc) { - static_call_cond(x86_pmu_swap_task_ctx)(prev, next); + static_call_cond(x86_pmu_swap_task_ctx)(prev_epc, next_epc); } void perf_check_microcode(void) @@ -2689,12 +2668,13 @@ static int x86_pmu_aux_output_match(struct perf_event *event) return 0; } -static int x86_pmu_filter_match(struct perf_event *event) +static bool x86_pmu_filter(struct pmu *pmu, int cpu) { - if (x86_pmu.filter_match) - return x86_pmu.filter_match(event); + bool ret = false; - return 1; + static_call_cond(x86_pmu_filter)(pmu, cpu, &ret); + + return ret; } static struct pmu pmu = { @@ -2725,7 +2705,7 @@ static struct pmu pmu = { .aux_output_match = x86_pmu_aux_output_match, - .filter_match = x86_pmu_filter_match, + .filter = x86_pmu_filter, }; void arch_perf_update_userpage(struct perf_event *event, diff --git a/arch/x86/events/intel/core.c b/arch/x86/events/intel/core.c index 1b92bf05fd65..dfd2c124cdf8 100644 --- a/arch/x86/events/intel/core.c +++ b/arch/x86/events/intel/core.c @@ -4536,8 +4536,6 @@ end: cpumask_set_cpu(cpu, &pmu->supported_cpus); cpuc->pmu = &pmu->pmu; - x86_pmu_update_cpu_context(&pmu->pmu, cpu); - return true; } @@ -4671,17 +4669,17 @@ static void intel_pmu_cpu_dead(int cpu) cpumask_clear_cpu(cpu, &hybrid_pmu(cpuc->pmu)->supported_cpus); } -static void intel_pmu_sched_task(struct perf_event_context *ctx, +static void intel_pmu_sched_task(struct perf_event_pmu_context *pmu_ctx, bool sched_in) { - intel_pmu_pebs_sched_task(ctx, sched_in); - intel_pmu_lbr_sched_task(ctx, sched_in); + intel_pmu_pebs_sched_task(pmu_ctx, sched_in); + intel_pmu_lbr_sched_task(pmu_ctx, sched_in); } -static void intel_pmu_swap_task_ctx(struct perf_event_context *prev, - struct perf_event_context *next) +static void intel_pmu_swap_task_ctx(struct perf_event_pmu_context *prev_epc, + struct perf_event_pmu_context *next_epc) { - intel_pmu_lbr_swap_task_ctx(prev, next); + intel_pmu_lbr_swap_task_ctx(prev_epc, next_epc); } static int intel_pmu_check_period(struct perf_event *event, u64 value) @@ -4705,12 +4703,11 @@ static int intel_pmu_aux_output_match(struct perf_event *event) return is_intel_pt_event(event); } -static int intel_pmu_filter_match(struct perf_event *event) +static void intel_pmu_filter(struct pmu *pmu, int cpu, bool *ret) { - struct x86_hybrid_pmu *pmu = hybrid_pmu(event->pmu); - unsigned int cpu = smp_processor_id(); + struct x86_hybrid_pmu *hpmu = hybrid_pmu(pmu); - return cpumask_test_cpu(cpu, &pmu->supported_cpus); + *ret = !cpumask_test_cpu(cpu, &hpmu->supported_cpus); } PMU_FORMAT_ATTR(offcore_rsp, "config1:0-63"); @@ -6413,7 +6410,7 @@ __init int intel_pmu_init(void) static_call_update(intel_pmu_set_topdown_event_period, &adl_set_topdown_event_period); - x86_pmu.filter_match = intel_pmu_filter_match; + x86_pmu.filter = intel_pmu_filter; x86_pmu.get_event_constraints = adl_get_event_constraints; x86_pmu.hw_config = adl_hw_config; x86_pmu.limit_period = spr_limit_period; diff --git a/arch/x86/events/intel/ds.c b/arch/x86/events/intel/ds.c index 446d2833efa7..88e58b6ee73c 100644 --- a/arch/x86/events/intel/ds.c +++ b/arch/x86/events/intel/ds.c @@ -1069,7 +1069,7 @@ static inline bool pebs_needs_sched_cb(struct cpu_hw_events *cpuc) return cpuc->n_pebs && (cpuc->n_pebs == cpuc->n_large_pebs); } -void intel_pmu_pebs_sched_task(struct perf_event_context *ctx, bool sched_in) +void intel_pmu_pebs_sched_task(struct perf_event_pmu_context *pmu_ctx, bool sched_in) { struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events); @@ -1177,7 +1177,7 @@ static void pebs_update_state(bool needed_cb, struct cpu_hw_events *cpuc, struct perf_event *event, bool add) { - struct pmu *pmu = event->ctx->pmu; + struct pmu *pmu = event->pmu; /* * Make sure we get updated with the first PEBS * event. It will trigger also during removal, but diff --git a/arch/x86/events/intel/lbr.c b/arch/x86/events/intel/lbr.c index 8259d725054d..017baba56b01 100644 --- a/arch/x86/events/intel/lbr.c +++ b/arch/x86/events/intel/lbr.c @@ -515,21 +515,21 @@ static void __intel_pmu_lbr_save(void *ctx) cpuc->last_log_id = ++task_context_opt(ctx)->log_id; } -void intel_pmu_lbr_swap_task_ctx(struct perf_event_context *prev, - struct perf_event_context *next) +void intel_pmu_lbr_swap_task_ctx(struct perf_event_pmu_context *prev_epc, + struct perf_event_pmu_context *next_epc) { void *prev_ctx_data, *next_ctx_data; - swap(prev->task_ctx_data, next->task_ctx_data); + swap(prev_epc->task_ctx_data, next_epc->task_ctx_data); /* - * Architecture specific synchronization makes sense in - * case both prev->task_ctx_data and next->task_ctx_data + * Architecture specific synchronization makes sense in case + * both prev_epc->task_ctx_data and next_epc->task_ctx_data * pointers are allocated. */ - prev_ctx_data = next->task_ctx_data; - next_ctx_data = prev->task_ctx_data; + prev_ctx_data = next_epc->task_ctx_data; + next_ctx_data = prev_epc->task_ctx_data; if (!prev_ctx_data || !next_ctx_data) return; @@ -538,7 +538,7 @@ void intel_pmu_lbr_swap_task_ctx(struct perf_event_context *prev, task_context_opt(next_ctx_data)->lbr_callstack_users); } -void intel_pmu_lbr_sched_task(struct perf_event_context *ctx, bool sched_in) +void intel_pmu_lbr_sched_task(struct perf_event_pmu_context *pmu_ctx, bool sched_in) { struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events); void *task_ctx; @@ -551,7 +551,7 @@ void intel_pmu_lbr_sched_task(struct perf_event_context *ctx, bool sched_in) * the task was scheduled out, restore the stack. Otherwise flush * the LBR stack. */ - task_ctx = ctx ? ctx->task_ctx_data : NULL; + task_ctx = pmu_ctx ? pmu_ctx->task_ctx_data : NULL; if (task_ctx) { if (sched_in) __intel_pmu_lbr_restore(task_ctx); @@ -587,8 +587,8 @@ void intel_pmu_lbr_add(struct perf_event *event) cpuc->br_sel = event->hw.branch_reg.reg; - if (branch_user_callstack(cpuc->br_sel) && event->ctx->task_ctx_data) - task_context_opt(event->ctx->task_ctx_data)->lbr_callstack_users++; + if (branch_user_callstack(cpuc->br_sel) && event->pmu_ctx->task_ctx_data) + task_context_opt(event->pmu_ctx->task_ctx_data)->lbr_callstack_users++; /* * Request pmu::sched_task() callback, which will fire inside the @@ -611,7 +611,7 @@ void intel_pmu_lbr_add(struct perf_event *event) */ if (x86_pmu.intel_cap.pebs_baseline && event->attr.precise_ip > 0) cpuc->lbr_pebs_users++; - perf_sched_cb_inc(event->ctx->pmu); + perf_sched_cb_inc(event->pmu); if (!cpuc->lbr_users++ && !event->total_time_running) intel_pmu_lbr_reset(); } @@ -664,8 +664,8 @@ void intel_pmu_lbr_del(struct perf_event *event) return; if (branch_user_callstack(cpuc->br_sel) && - event->ctx->task_ctx_data) - task_context_opt(event->ctx->task_ctx_data)->lbr_callstack_users--; + event->pmu_ctx->task_ctx_data) + task_context_opt(event->pmu_ctx->task_ctx_data)->lbr_callstack_users--; if (event->hw.flags & PERF_X86_EVENT_LBR_SELECT) cpuc->lbr_select = 0; @@ -675,7 +675,7 @@ void intel_pmu_lbr_del(struct perf_event *event) cpuc->lbr_users--; WARN_ON_ONCE(cpuc->lbr_users < 0); WARN_ON_ONCE(cpuc->lbr_pebs_users < 0); - perf_sched_cb_dec(event->ctx->pmu); + perf_sched_cb_dec(event->pmu); } static inline bool vlbr_exclude_host(void) diff --git a/arch/x86/events/intel/uncore.h b/arch/x86/events/intel/uncore.h index 2adeaf4de4df..e278e2e7c051 100644 --- a/arch/x86/events/intel/uncore.h +++ b/arch/x86/events/intel/uncore.h @@ -2,6 +2,7 @@ #include <linux/slab.h> #include <linux/pci.h> #include <asm/apicdef.h> +#include <asm/intel-family.h> #include <linux/io-64-nonatomic-lo-hi.h> #include <linux/perf_event.h> @@ -88,12 +89,12 @@ struct intel_uncore_type { * to identify which platform component each PMON block of that type is * supposed to monitor. */ - struct intel_uncore_topology *topology; + struct intel_uncore_topology **topology; /* * Optional callbacks for managing mapping of Uncore units to PMONs */ int (*get_topology)(struct intel_uncore_type *type); - int (*set_mapping)(struct intel_uncore_type *type); + void (*set_mapping)(struct intel_uncore_type *type); void (*cleanup_mapping)(struct intel_uncore_type *type); }; @@ -178,11 +179,26 @@ struct freerunning_counters { unsigned *box_offsets; }; -struct intel_uncore_topology { - u64 configuration; +struct uncore_iio_topology { + int pci_bus_no; int segment; }; +struct uncore_upi_topology { + int die_to; + int pmu_idx_to; + int enabled; +}; + +struct intel_uncore_topology { + int pmu_idx; + union { + void *untyped; + struct uncore_iio_topology *iio; + struct uncore_upi_topology *upi; + }; +}; + struct pci2phy_map { struct list_head list; int segment; diff --git a/arch/x86/events/intel/uncore_snb.c b/arch/x86/events/intel/uncore_snb.c index 1ef4f7861e2e..1f4869227efb 100644 --- a/arch/x86/events/intel/uncore_snb.c +++ b/arch/x86/events/intel/uncore_snb.c @@ -1338,6 +1338,7 @@ static void __uncore_imc_init_box(struct intel_uncore_box *box, /* MCHBAR is disabled */ if (!(mch_bar & BIT(0))) { pr_warn("perf uncore: MCHBAR is disabled. Failed to map IMC free-running counters.\n"); + pci_dev_put(pdev); return; } mch_bar &= ~BIT(0); @@ -1352,6 +1353,8 @@ static void __uncore_imc_init_box(struct intel_uncore_box *box, box->io_addr = ioremap(addr, type->mmio_map_size); if (!box->io_addr) pr_warn("perf uncore: Failed to ioremap for %s.\n", type->name); + + pci_dev_put(pdev); } static void tgl_uncore_imc_freerunning_init_box(struct intel_uncore_box *box) diff --git a/arch/x86/events/intel/uncore_snbep.c b/arch/x86/events/intel/uncore_snbep.c index ed869443efb2..44c2f879f708 100644 --- a/arch/x86/events/intel/uncore_snbep.c +++ b/arch/x86/events/intel/uncore_snbep.c @@ -445,6 +445,7 @@ #define ICX_UPI_PCI_PMON_CTR0 0x320 #define ICX_UPI_PCI_PMON_BOX_CTL 0x318 #define ICX_UPI_CTL_UMASK_EXT 0xffffff +#define ICX_UBOX_DID 0x3450 /* ICX M3UPI*/ #define ICX_M3UPI_PCI_PMON_CTL0 0xd8 @@ -457,6 +458,7 @@ /* SPR */ #define SPR_RAW_EVENT_MASK_EXT 0xffffff +#define SPR_UBOX_DID 0x3250 /* SPR CHA */ #define SPR_CHA_PMON_CTL_TID_EN (1 << 16) @@ -1372,6 +1374,28 @@ static struct pci_driver snbep_uncore_pci_driver = { #define NODE_ID_MASK 0x7 +/* Each three bits from 0 to 23 of GIDNIDMAP register correspond Node ID. */ +#define GIDNIDMAP(config, id) (((config) >> (3 * (id))) & 0x7) + +static int upi_nodeid_groupid(struct pci_dev *ubox_dev, int nodeid_loc, int idmap_loc, + int *nodeid, int *groupid) +{ + int ret; + + /* get the Node ID of the local register */ + ret = pci_read_config_dword(ubox_dev, nodeid_loc, nodeid); + if (ret) + goto err; + + *nodeid = *nodeid & NODE_ID_MASK; + /* get the Node ID mapping */ + ret = pci_read_config_dword(ubox_dev, idmap_loc, groupid); + if (ret) + goto err; +err: + return ret; +} + /* * build pci bus to socket mapping */ @@ -1397,13 +1421,8 @@ static int snbep_pci2phy_map_init(int devid, int nodeid_loc, int idmap_loc, bool * the topology. */ if (nr_node_ids <= 8) { - /* get the Node ID of the local register */ - err = pci_read_config_dword(ubox_dev, nodeid_loc, &config); - if (err) - break; - nodeid = config & NODE_ID_MASK; - /* get the Node ID mapping */ - err = pci_read_config_dword(ubox_dev, idmap_loc, &config); + err = upi_nodeid_groupid(ubox_dev, nodeid_loc, idmap_loc, + &nodeid, &config); if (err) break; @@ -1421,7 +1440,7 @@ static int snbep_pci2phy_map_init(int devid, int nodeid_loc, int idmap_loc, bool * to a particular node. */ for (i = 0; i < 8; i++) { - if (nodeid == ((config >> (3 * i)) & 0x7)) { + if (nodeid == GIDNIDMAP(config, i)) { if (topology_max_die_per_package() > 1) die_id = i; else @@ -2891,6 +2910,7 @@ static bool hswep_has_limit_sbox(unsigned int device) return false; pci_read_config_dword(dev, HSWEP_PCU_CAPID4_OFFET, &capid4); + pci_dev_put(dev); if (!hswep_get_chop(capid4)) return true; @@ -3699,10 +3719,16 @@ static struct intel_uncore_ops skx_uncore_iio_ops = { .read_counter = uncore_msr_read_counter, }; -static inline u8 skx_iio_stack(struct intel_uncore_pmu *pmu, int die) +static struct intel_uncore_topology *pmu_topology(struct intel_uncore_pmu *pmu, int die) { - return pmu->type->topology[die].configuration >> - (pmu->pmu_idx * BUS_NUM_STRIDE); + int idx; + + for (idx = 0; idx < pmu->type->num_boxes; idx++) { + if (pmu->type->topology[die][idx].pmu_idx == pmu->pmu_idx) + return &pmu->type->topology[die][idx]; + } + + return NULL; } static umode_t @@ -3710,8 +3736,9 @@ pmu_iio_mapping_visible(struct kobject *kobj, struct attribute *attr, int die, int zero_bus_pmu) { struct intel_uncore_pmu *pmu = dev_to_uncore_pmu(kobj_to_dev(kobj)); + struct intel_uncore_topology *pmut = pmu_topology(pmu, die); - return (!skx_iio_stack(pmu, die) && pmu->pmu_idx != zero_bus_pmu) ? 0 : attr->mode; + return (pmut && !pmut->iio->pci_bus_no && pmu->pmu_idx != zero_bus_pmu) ? 0 : attr->mode; } static umode_t @@ -3727,9 +3754,10 @@ static ssize_t skx_iio_mapping_show(struct device *dev, struct intel_uncore_pmu *pmu = dev_to_uncore_pmu(dev); struct dev_ext_attribute *ea = to_dev_ext_attribute(attr); long die = (long)ea->var; + struct intel_uncore_topology *pmut = pmu_topology(pmu, die); - return sprintf(buf, "%04x:%02x\n", pmu->type->topology[die].segment, - skx_iio_stack(pmu, die)); + return sprintf(buf, "%04x:%02x\n", pmut ? pmut->iio->segment : 0, + pmut ? pmut->iio->pci_bus_no : 0); } static int skx_msr_cpu_bus_read(int cpu, u64 *topology) @@ -3764,18 +3792,79 @@ static int die_to_cpu(int die) return res; } -static int skx_iio_get_topology(struct intel_uncore_type *type) +enum { + IIO_TOPOLOGY_TYPE, + UPI_TOPOLOGY_TYPE, + TOPOLOGY_MAX +}; + +static const size_t topology_size[TOPOLOGY_MAX] = { + sizeof(*((struct intel_uncore_topology *)NULL)->iio), + sizeof(*((struct intel_uncore_topology *)NULL)->upi) +}; + +static int pmu_alloc_topology(struct intel_uncore_type *type, int topology_type) { - int die, ret = -EPERM; + int die, idx; + struct intel_uncore_topology **topology; + + if (!type->num_boxes) + return -EPERM; - type->topology = kcalloc(uncore_max_dies(), sizeof(*type->topology), - GFP_KERNEL); - if (!type->topology) - return -ENOMEM; + topology = kcalloc(uncore_max_dies(), sizeof(*topology), GFP_KERNEL); + if (!topology) + goto err; for (die = 0; die < uncore_max_dies(); die++) { - ret = skx_msr_cpu_bus_read(die_to_cpu(die), - &type->topology[die].configuration); + topology[die] = kcalloc(type->num_boxes, sizeof(**topology), GFP_KERNEL); + if (!topology[die]) + goto clear; + for (idx = 0; idx < type->num_boxes; idx++) { + topology[die][idx].untyped = kcalloc(type->num_boxes, + topology_size[topology_type], + GFP_KERNEL); + if (!topology[die][idx].untyped) + goto clear; + } + } + + type->topology = topology; + + return 0; +clear: + for (; die >= 0; die--) { + for (idx = 0; idx < type->num_boxes; idx++) + kfree(topology[die][idx].untyped); + kfree(topology[die]); + } + kfree(topology); +err: + return -ENOMEM; +} + +static void pmu_free_topology(struct intel_uncore_type *type) +{ + int die, idx; + + if (type->topology) { + for (die = 0; die < uncore_max_dies(); die++) { + for (idx = 0; idx < type->num_boxes; idx++) + kfree(type->topology[die][idx].untyped); + kfree(type->topology[die]); + } + kfree(type->topology); + type->topology = NULL; + } +} + +static int skx_pmu_get_topology(struct intel_uncore_type *type, + int (*topology_cb)(struct intel_uncore_type*, int, int, u64)) +{ + int die, ret = -EPERM; + u64 cpu_bus_msr; + + for (die = 0; die < uncore_max_dies(); die++) { + ret = skx_msr_cpu_bus_read(die_to_cpu(die), &cpu_bus_msr); if (ret) break; @@ -3783,15 +3872,33 @@ static int skx_iio_get_topology(struct intel_uncore_type *type) if (ret < 0) break; - type->topology[die].segment = ret; + ret = topology_cb(type, ret, die, cpu_bus_msr); + if (ret) + break; } - if (ret < 0) { - kfree(type->topology); - type->topology = NULL; + return ret; +} + +static int skx_iio_topology_cb(struct intel_uncore_type *type, int segment, + int die, u64 cpu_bus_msr) +{ + int idx; + struct intel_uncore_topology *t; + + for (idx = 0; idx < type->num_boxes; idx++) { + t = &type->topology[die][idx]; + t->pmu_idx = idx; + t->iio->segment = segment; + t->iio->pci_bus_no = (cpu_bus_msr >> (idx * BUS_NUM_STRIDE)) & 0xff; } - return ret; + return 0; +} + +static int skx_iio_get_topology(struct intel_uncore_type *type) +{ + return skx_pmu_get_topology(type, skx_iio_topology_cb); } static struct attribute_group skx_iio_mapping_group = { @@ -3803,8 +3910,25 @@ static const struct attribute_group *skx_iio_attr_update[] = { NULL, }; -static int -pmu_iio_set_mapping(struct intel_uncore_type *type, struct attribute_group *ag) +static void pmu_clear_mapping_attr(const struct attribute_group **groups, + struct attribute_group *ag) +{ + int i; + + for (i = 0; groups[i]; i++) { + if (groups[i] == ag) { + for (i++; groups[i]; i++) + groups[i - 1] = groups[i]; + groups[i - 1] = NULL; + break; + } + } +} + +static void +pmu_set_mapping(struct intel_uncore_type *type, struct attribute_group *ag, + ssize_t (*show)(struct device*, struct device_attribute*, char*), + int topology_type) { char buf[64]; int ret; @@ -3812,11 +3936,13 @@ pmu_iio_set_mapping(struct intel_uncore_type *type, struct attribute_group *ag) struct attribute **attrs = NULL; struct dev_ext_attribute *eas = NULL; - ret = type->get_topology(type); + ret = pmu_alloc_topology(type, topology_type); if (ret < 0) goto clear_attr_update; - ret = -ENOMEM; + ret = type->get_topology(type); + if (ret < 0) + goto clear_topology; /* One more for NULL. */ attrs = kcalloc((uncore_max_dies() + 1), sizeof(*attrs), GFP_KERNEL); @@ -3828,20 +3954,20 @@ pmu_iio_set_mapping(struct intel_uncore_type *type, struct attribute_group *ag) goto clear_attrs; for (die = 0; die < uncore_max_dies(); die++) { - sprintf(buf, "die%ld", die); + snprintf(buf, sizeof(buf), "die%ld", die); sysfs_attr_init(&eas[die].attr.attr); eas[die].attr.attr.name = kstrdup(buf, GFP_KERNEL); if (!eas[die].attr.attr.name) goto err; eas[die].attr.attr.mode = 0444; - eas[die].attr.show = skx_iio_mapping_show; + eas[die].attr.show = show; eas[die].attr.store = NULL; eas[die].var = (void *)die; attrs[die] = &eas[die].attr.attr; } ag->attrs = attrs; - return 0; + return; err: for (; die >= 0; die--) kfree(eas[die].attr.attr.name); @@ -3849,14 +3975,13 @@ err: clear_attrs: kfree(attrs); clear_topology: - kfree(type->topology); + pmu_free_topology(type); clear_attr_update: - type->attr_update = NULL; - return ret; + pmu_clear_mapping_attr(type->attr_update, ag); } static void -pmu_iio_cleanup_mapping(struct intel_uncore_type *type, struct attribute_group *ag) +pmu_cleanup_mapping(struct intel_uncore_type *type, struct attribute_group *ag) { struct attribute **attr = ag->attrs; @@ -3868,17 +3993,23 @@ pmu_iio_cleanup_mapping(struct intel_uncore_type *type, struct attribute_group * kfree(attr_to_ext_attr(*ag->attrs)); kfree(ag->attrs); ag->attrs = NULL; - kfree(type->topology); + pmu_free_topology(type); +} + +static void +pmu_iio_set_mapping(struct intel_uncore_type *type, struct attribute_group *ag) +{ + pmu_set_mapping(type, ag, skx_iio_mapping_show, IIO_TOPOLOGY_TYPE); } -static int skx_iio_set_mapping(struct intel_uncore_type *type) +static void skx_iio_set_mapping(struct intel_uncore_type *type) { - return pmu_iio_set_mapping(type, &skx_iio_mapping_group); + pmu_iio_set_mapping(type, &skx_iio_mapping_group); } static void skx_iio_cleanup_mapping(struct intel_uncore_type *type) { - pmu_iio_cleanup_mapping(type, &skx_iio_mapping_group); + pmu_cleanup_mapping(type, &skx_iio_mapping_group); } static struct intel_uncore_type skx_uncore_iio = { @@ -4139,6 +4270,132 @@ static struct intel_uncore_ops skx_upi_uncore_pci_ops = { .read_counter = snbep_uncore_pci_read_counter, }; +static umode_t +skx_upi_mapping_visible(struct kobject *kobj, struct attribute *attr, int die) +{ + struct intel_uncore_pmu *pmu = dev_to_uncore_pmu(kobj_to_dev(kobj)); + + return pmu->type->topology[die][pmu->pmu_idx].upi->enabled ? attr->mode : 0; +} + +static ssize_t skx_upi_mapping_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct intel_uncore_pmu *pmu = dev_to_uncore_pmu(dev); + struct dev_ext_attribute *ea = to_dev_ext_attribute(attr); + long die = (long)ea->var; + struct uncore_upi_topology *upi = pmu->type->topology[die][pmu->pmu_idx].upi; + + return sysfs_emit(buf, "upi_%d,die_%d\n", upi->pmu_idx_to, upi->die_to); +} + +#define SKX_UPI_REG_DID 0x2058 +#define SKX_UPI_REGS_ADDR_DEVICE_LINK0 0x0e +#define SKX_UPI_REGS_ADDR_FUNCTION 0x00 + +/* + * UPI Link Parameter 0 + * | Bit | Default | Description + * | 19:16 | 0h | base_nodeid - The NodeID of the sending socket. + * | 12:8 | 00h | sending_port - The processor die port number of the sending port. + */ +#define SKX_KTILP0_OFFSET 0x94 + +/* + * UPI Pcode Status. This register is used by PCode to store the link training status. + * | Bit | Default | Description + * | 4 | 0h | ll_status_valid — Bit indicates the valid training status + * logged from PCode to the BIOS. + */ +#define SKX_KTIPCSTS_OFFSET 0x120 + +static int upi_fill_topology(struct pci_dev *dev, struct intel_uncore_topology *tp, + int pmu_idx) +{ + int ret; + u32 upi_conf; + struct uncore_upi_topology *upi = tp->upi; + + tp->pmu_idx = pmu_idx; + ret = pci_read_config_dword(dev, SKX_KTIPCSTS_OFFSET, &upi_conf); + if (ret) { + ret = pcibios_err_to_errno(ret); + goto err; + } + upi->enabled = (upi_conf >> 4) & 1; + if (upi->enabled) { + ret = pci_read_config_dword(dev, SKX_KTILP0_OFFSET, + &upi_conf); + if (ret) { + ret = pcibios_err_to_errno(ret); + goto err; + } + upi->die_to = (upi_conf >> 16) & 0xf; + upi->pmu_idx_to = (upi_conf >> 8) & 0x1f; + } +err: + return ret; +} + +static int skx_upi_topology_cb(struct intel_uncore_type *type, int segment, + int die, u64 cpu_bus_msr) +{ + int idx, ret; + struct intel_uncore_topology *upi; + unsigned int devfn; + struct pci_dev *dev = NULL; + u8 bus = cpu_bus_msr >> (3 * BUS_NUM_STRIDE); + + for (idx = 0; idx < type->num_boxes; idx++) { + upi = &type->topology[die][idx]; + devfn = PCI_DEVFN(SKX_UPI_REGS_ADDR_DEVICE_LINK0 + idx, + SKX_UPI_REGS_ADDR_FUNCTION); + dev = pci_get_domain_bus_and_slot(segment, bus, devfn); + if (dev) { + ret = upi_fill_topology(dev, upi, idx); + if (ret) + break; + } + } + + pci_dev_put(dev); + return ret; +} + +static int skx_upi_get_topology(struct intel_uncore_type *type) +{ + /* CPX case is not supported */ + if (boot_cpu_data.x86_stepping == 11) + return -EPERM; + + return skx_pmu_get_topology(type, skx_upi_topology_cb); +} + +static struct attribute_group skx_upi_mapping_group = { + .is_visible = skx_upi_mapping_visible, +}; + +static const struct attribute_group *skx_upi_attr_update[] = { + &skx_upi_mapping_group, + NULL +}; + +static void +pmu_upi_set_mapping(struct intel_uncore_type *type, struct attribute_group *ag) +{ + pmu_set_mapping(type, ag, skx_upi_mapping_show, UPI_TOPOLOGY_TYPE); +} + +static void skx_upi_set_mapping(struct intel_uncore_type *type) +{ + pmu_upi_set_mapping(type, &skx_upi_mapping_group); +} + +static void skx_upi_cleanup_mapping(struct intel_uncore_type *type) +{ + pmu_cleanup_mapping(type, &skx_upi_mapping_group); +} + static struct intel_uncore_type skx_uncore_upi = { .name = "upi", .num_counters = 4, @@ -4151,6 +4408,10 @@ static struct intel_uncore_type skx_uncore_upi = { .box_ctl = SKX_UPI_PCI_PMON_BOX_CTL, .ops = &skx_upi_uncore_pci_ops, .format_group = &skx_upi_uncore_format_group, + .attr_update = skx_upi_attr_update, + .get_topology = skx_upi_get_topology, + .set_mapping = skx_upi_set_mapping, + .cleanup_mapping = skx_upi_cleanup_mapping, }; static void skx_m2m_uncore_pci_init_box(struct intel_uncore_box *box) @@ -4461,11 +4722,6 @@ static int sad_cfg_iio_topology(struct intel_uncore_type *type, u8 *sad_pmon_map int die, stack_id, ret = -EPERM; struct pci_dev *dev = NULL; - type->topology = kcalloc(uncore_max_dies(), sizeof(*type->topology), - GFP_KERNEL); - if (!type->topology) - return -ENOMEM; - while ((dev = pci_get_device(PCI_VENDOR_ID_INTEL, SNR_ICX_MESH2IIO_MMAP_DID, dev))) { ret = pci_read_config_dword(dev, SNR_ICX_SAD_CONTROL_CFG, &sad_cfg); if (ret) { @@ -4483,14 +4739,12 @@ static int sad_cfg_iio_topology(struct intel_uncore_type *type, u8 *sad_pmon_map /* Convert stack id from SAD_CONTROL to PMON notation. */ stack_id = sad_pmon_mapping[stack_id]; - ((u8 *)&(type->topology[die].configuration))[stack_id] = dev->bus->number; - type->topology[die].segment = pci_domain_nr(dev->bus); + type->topology[die][stack_id].iio->segment = pci_domain_nr(dev->bus); + type->topology[die][stack_id].pmu_idx = stack_id; + type->topology[die][stack_id].iio->pci_bus_no = dev->bus->number; } - if (ret) { - kfree(type->topology); - type->topology = NULL; - } + pci_dev_put(dev); return ret; } @@ -4519,14 +4773,14 @@ static int snr_iio_get_topology(struct intel_uncore_type *type) return sad_cfg_iio_topology(type, snr_sad_pmon_mapping); } -static int snr_iio_set_mapping(struct intel_uncore_type *type) +static void snr_iio_set_mapping(struct intel_uncore_type *type) { - return pmu_iio_set_mapping(type, &snr_iio_mapping_group); + pmu_iio_set_mapping(type, &snr_iio_mapping_group); } static void snr_iio_cleanup_mapping(struct intel_uncore_type *type) { - pmu_iio_cleanup_mapping(type, &snr_iio_mapping_group); + pmu_cleanup_mapping(type, &snr_iio_mapping_group); } static struct event_constraint snr_uncore_iio_constraints[] = { @@ -4857,6 +5111,8 @@ static int snr_uncore_mmio_map(struct intel_uncore_box *box, addr += box_ctl; + pci_dev_put(pdev); + box->io_addr = ioremap(addr, type->mmio_map_size); if (!box->io_addr) { pr_warn("perf uncore: Failed to ioremap for %s.\n", type->name); @@ -5137,14 +5393,19 @@ static int icx_iio_get_topology(struct intel_uncore_type *type) return sad_cfg_iio_topology(type, icx_sad_pmon_mapping); } -static int icx_iio_set_mapping(struct intel_uncore_type *type) +static void icx_iio_set_mapping(struct intel_uncore_type *type) { - return pmu_iio_set_mapping(type, &icx_iio_mapping_group); + /* Detect ICX-D system. This case is not supported */ + if (boot_cpu_data.x86_model == INTEL_FAM6_ICELAKE_D) { + pmu_clear_mapping_attr(type->attr_update, &icx_iio_mapping_group); + return; + } + pmu_iio_set_mapping(type, &icx_iio_mapping_group); } static void icx_iio_cleanup_mapping(struct intel_uncore_type *type) { - pmu_iio_cleanup_mapping(type, &icx_iio_mapping_group); + pmu_cleanup_mapping(type, &icx_iio_mapping_group); } static struct intel_uncore_type icx_uncore_iio = { @@ -5337,6 +5598,76 @@ static const struct attribute_group icx_upi_uncore_format_group = { .attrs = icx_upi_uncore_formats_attr, }; +#define ICX_UPI_REGS_ADDR_DEVICE_LINK0 0x02 +#define ICX_UPI_REGS_ADDR_FUNCTION 0x01 + +static int discover_upi_topology(struct intel_uncore_type *type, int ubox_did, int dev_link0) +{ + struct pci_dev *ubox = NULL; + struct pci_dev *dev = NULL; + u32 nid, gid; + int i, idx, ret = -EPERM; + struct intel_uncore_topology *upi; + unsigned int devfn; + + /* GIDNIDMAP method supports machines which have less than 8 sockets. */ + if (uncore_max_dies() > 8) + goto err; + + while ((ubox = pci_get_device(PCI_VENDOR_ID_INTEL, ubox_did, ubox))) { + ret = upi_nodeid_groupid(ubox, SKX_CPUNODEID, SKX_GIDNIDMAP, &nid, &gid); + if (ret) { + ret = pcibios_err_to_errno(ret); + break; + } + + for (i = 0; i < 8; i++) { + if (nid != GIDNIDMAP(gid, i)) + continue; + for (idx = 0; idx < type->num_boxes; idx++) { + upi = &type->topology[nid][idx]; + devfn = PCI_DEVFN(dev_link0 + idx, ICX_UPI_REGS_ADDR_FUNCTION); + dev = pci_get_domain_bus_and_slot(pci_domain_nr(ubox->bus), + ubox->bus->number, + devfn); + if (dev) { + ret = upi_fill_topology(dev, upi, idx); + if (ret) + goto err; + } + } + } + } +err: + pci_dev_put(ubox); + pci_dev_put(dev); + return ret; +} + +static int icx_upi_get_topology(struct intel_uncore_type *type) +{ + return discover_upi_topology(type, ICX_UBOX_DID, ICX_UPI_REGS_ADDR_DEVICE_LINK0); +} + +static struct attribute_group icx_upi_mapping_group = { + .is_visible = skx_upi_mapping_visible, +}; + +static const struct attribute_group *icx_upi_attr_update[] = { + &icx_upi_mapping_group, + NULL +}; + +static void icx_upi_set_mapping(struct intel_uncore_type *type) +{ + pmu_upi_set_mapping(type, &icx_upi_mapping_group); +} + +static void icx_upi_cleanup_mapping(struct intel_uncore_type *type) +{ + pmu_cleanup_mapping(type, &icx_upi_mapping_group); +} + static struct intel_uncore_type icx_uncore_upi = { .name = "upi", .num_counters = 4, @@ -5349,6 +5680,10 @@ static struct intel_uncore_type icx_uncore_upi = { .box_ctl = ICX_UPI_PCI_PMON_BOX_CTL, .ops = &skx_upi_uncore_pci_ops, .format_group = &icx_upi_uncore_format_group, + .attr_update = icx_upi_attr_update, + .get_topology = icx_upi_get_topology, + .set_mapping = icx_upi_set_mapping, + .cleanup_mapping = icx_upi_cleanup_mapping, }; static struct event_constraint icx_uncore_m3upi_constraints[] = { @@ -5780,9 +6115,43 @@ static struct intel_uncore_type spr_uncore_m2m = { .name = "m2m", }; +static struct attribute_group spr_upi_mapping_group = { + .is_visible = skx_upi_mapping_visible, +}; + +static const struct attribute_group *spr_upi_attr_update[] = { + &uncore_alias_group, + &spr_upi_mapping_group, + NULL +}; + +#define SPR_UPI_REGS_ADDR_DEVICE_LINK0 0x01 + +static void spr_upi_set_mapping(struct intel_uncore_type *type) +{ + pmu_upi_set_mapping(type, &spr_upi_mapping_group); +} + +static void spr_upi_cleanup_mapping(struct intel_uncore_type *type) +{ + pmu_cleanup_mapping(type, &spr_upi_mapping_group); +} + +static int spr_upi_get_topology(struct intel_uncore_type *type) +{ + return discover_upi_topology(type, SPR_UBOX_DID, SPR_UPI_REGS_ADDR_DEVICE_LINK0); +} + static struct intel_uncore_type spr_uncore_upi = { - SPR_UNCORE_PCI_COMMON_FORMAT(), + .event_mask = SNBEP_PMON_RAW_EVENT_MASK, + .event_mask_ext = SPR_RAW_EVENT_MASK_EXT, + .format_group = &spr_uncore_raw_format_group, + .ops = &spr_uncore_pci_ops, .name = "upi", + .attr_update = spr_upi_attr_update, + .get_topology = spr_upi_get_topology, + .set_mapping = spr_upi_set_mapping, + .cleanup_mapping = spr_upi_cleanup_mapping, }; static struct intel_uncore_type spr_uncore_m3upi = { @@ -5986,6 +6355,12 @@ static void uncore_type_customized_copy(struct intel_uncore_type *to_type, to_type->format_group = from_type->format_group; if (from_type->attr_update) to_type->attr_update = from_type->attr_update; + if (from_type->set_mapping) + to_type->set_mapping = from_type->set_mapping; + if (from_type->get_topology) + to_type->get_topology = from_type->get_topology; + if (from_type->cleanup_mapping) + to_type->cleanup_mapping = from_type->cleanup_mapping; } static struct intel_uncore_type ** diff --git a/arch/x86/events/perf_event.h b/arch/x86/events/perf_event.h index 332d2e6d8ae4..0e849f28a5c1 100644 --- a/arch/x86/events/perf_event.h +++ b/arch/x86/events/perf_event.h @@ -811,7 +811,7 @@ struct x86_pmu { void (*cpu_dead)(int cpu); void (*check_microcode)(void); - void (*sched_task)(struct perf_event_context *ctx, + void (*sched_task)(struct perf_event_pmu_context *pmu_ctx, bool sched_in); /* @@ -894,12 +894,12 @@ struct x86_pmu { int num_topdown_events; /* - * perf task context (i.e. struct perf_event_context::task_ctx_data) + * perf task context (i.e. struct perf_event_pmu_context::task_ctx_data) * switch helper to bridge calls from perf/core to perf/x86. * See struct pmu::swap_task_ctx() usage for examples; */ - void (*swap_task_ctx)(struct perf_event_context *prev, - struct perf_event_context *next); + void (*swap_task_ctx)(struct perf_event_pmu_context *prev_epc, + struct perf_event_pmu_context *next_epc); /* * AMD bits @@ -925,7 +925,7 @@ struct x86_pmu { int (*aux_output_match) (struct perf_event *event); - int (*filter_match)(struct perf_event *event); + void (*filter)(struct pmu *pmu, int cpu, bool *ret); /* * Hybrid support * @@ -1180,8 +1180,6 @@ int x86_pmu_handle_irq(struct pt_regs *regs); void x86_pmu_show_pmu_cap(int num_counters, int num_counters_fixed, u64 intel_ctrl); -void x86_pmu_update_cpu_context(struct pmu *pmu, int cpu); - extern struct event_constraint emptyconstraint; extern struct event_constraint unconstrained; @@ -1306,7 +1304,7 @@ void amd_pmu_lbr_reset(void); void amd_pmu_lbr_read(void); void amd_pmu_lbr_add(struct perf_event *event); void amd_pmu_lbr_del(struct perf_event *event); -void amd_pmu_lbr_sched_task(struct perf_event_context *ctx, bool sched_in); +void amd_pmu_lbr_sched_task(struct perf_event_pmu_context *pmu_ctx, bool sched_in); void amd_pmu_lbr_enable_all(void); void amd_pmu_lbr_disable_all(void); int amd_pmu_lbr_hw_config(struct perf_event *event); @@ -1322,7 +1320,6 @@ void amd_brs_enable_all(void); void amd_brs_disable_all(void); void amd_brs_drain(void); void amd_brs_lopwr_init(void); -void amd_brs_disable_all(void); int amd_brs_hw_config(struct perf_event *event); void amd_brs_reset(void); @@ -1330,7 +1327,7 @@ static inline void amd_pmu_brs_add(struct perf_event *event) { struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events); - perf_sched_cb_inc(event->ctx->pmu); + perf_sched_cb_inc(event->pmu); cpuc->lbr_users++; /* * No need to reset BRS because it is reset @@ -1345,10 +1342,10 @@ static inline void amd_pmu_brs_del(struct perf_event *event) cpuc->lbr_users--; WARN_ON_ONCE(cpuc->lbr_users < 0); - perf_sched_cb_dec(event->ctx->pmu); + perf_sched_cb_dec(event->pmu); } -void amd_pmu_brs_sched_task(struct perf_event_context *ctx, bool sched_in); +void amd_pmu_brs_sched_task(struct perf_event_pmu_context *pmu_ctx, bool sched_in); #else static inline int amd_brs_init(void) { @@ -1373,7 +1370,7 @@ static inline void amd_pmu_brs_del(struct perf_event *event) { } -static inline void amd_pmu_brs_sched_task(struct perf_event_context *ctx, bool sched_in) +static inline void amd_pmu_brs_sched_task(struct perf_event_pmu_context *pmu_ctx, bool sched_in) { } @@ -1533,7 +1530,7 @@ void intel_pmu_pebs_enable_all(void); void intel_pmu_pebs_disable_all(void); -void intel_pmu_pebs_sched_task(struct perf_event_context *ctx, bool sched_in); +void intel_pmu_pebs_sched_task(struct perf_event_pmu_context *pmu_ctx, bool sched_in); void intel_pmu_auto_reload_read(struct perf_event *event); @@ -1541,10 +1538,10 @@ void intel_pmu_store_pebs_lbrs(struct lbr_entry *lbr); void intel_ds_init(void); -void intel_pmu_lbr_swap_task_ctx(struct perf_event_context *prev, - struct perf_event_context *next); +void intel_pmu_lbr_swap_task_ctx(struct perf_event_pmu_context *prev_epc, + struct perf_event_pmu_context *next_epc); -void intel_pmu_lbr_sched_task(struct perf_event_context *ctx, bool sched_in); +void intel_pmu_lbr_sched_task(struct perf_event_pmu_context *pmu_ctx, bool sched_in); u64 lbr_from_signext_quirk_wr(u64 val); diff --git a/drivers/perf/arm_pmu.c b/drivers/perf/arm_pmu.c index bb56676f50ef..9b593f985805 100644 --- a/drivers/perf/arm_pmu.c +++ b/drivers/perf/arm_pmu.c @@ -547,15 +547,14 @@ static void armpmu_disable(struct pmu *pmu) * microarchitecture, and aren't suitable for another. Thus, only match CPUs of * the same microarchitecture. */ -static int armpmu_filter_match(struct perf_event *event) +static bool armpmu_filter(struct pmu *pmu, int cpu) { - struct arm_pmu *armpmu = to_arm_pmu(event->pmu); - unsigned int cpu = smp_processor_id(); - int ret; + struct arm_pmu *armpmu = to_arm_pmu(pmu); + bool ret; ret = cpumask_test_cpu(cpu, &armpmu->supported_cpus); - if (ret && armpmu->filter_match) - return armpmu->filter_match(event); + if (ret && armpmu->filter) + return armpmu->filter(pmu, cpu); return ret; } @@ -882,14 +881,13 @@ struct arm_pmu *armpmu_alloc(void) .start = armpmu_start, .stop = armpmu_stop, .read = armpmu_read, - .filter_match = armpmu_filter_match, + .filter = armpmu_filter, .attr_groups = pmu->attr_groups, /* * This is a CPU PMU potentially in a heterogeneous * configuration (e.g. big.LITTLE). This is not an uncore PMU, * and we have taken ctx sharing into account (e.g. with our - * pmu::filter_match callback and pmu::event_init group - * validation). + * pmu::filter callback and pmu::event_init group validation). */ .capabilities = PERF_PMU_CAP_HETEROGENEOUS_CPUS | PERF_PMU_CAP_EXTENDED_REGS, }; diff --git a/include/linux/perf/arm_pmu.h b/include/linux/perf/arm_pmu.h index 0c15c5b7f801..ef914a600087 100644 --- a/include/linux/perf/arm_pmu.h +++ b/include/linux/perf/arm_pmu.h @@ -100,7 +100,7 @@ struct arm_pmu { void (*stop)(struct arm_pmu *); void (*reset)(void *); int (*map_event)(struct perf_event *event); - int (*filter_match)(struct perf_event *event); + bool (*filter)(struct pmu *pmu, int cpu); int num_events; bool secure_access; /* 32-bit ARM only */ #define ARMV8_PMUV3_MAX_COMMON_EVENTS 0x40 diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 0031f7b4d9ab..c6a3bac76966 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -266,6 +266,7 @@ struct hw_perf_event { }; struct perf_event; +struct perf_event_pmu_context; /* * Common implementation detail of pmu::{start,commit,cancel}_txn @@ -308,7 +309,7 @@ struct pmu { int capabilities; int __percpu *pmu_disable_count; - struct perf_cpu_context __percpu *pmu_cpu_context; + struct perf_cpu_pmu_context __percpu *cpu_pmu_context; atomic_t exclusive_cnt; /* < 0: cpu; > 0: tsk */ int task_ctx_nr; int hrtimer_interval_ms; @@ -443,7 +444,7 @@ struct pmu { /* * context-switches callback */ - void (*sched_task) (struct perf_event_context *ctx, + void (*sched_task) (struct perf_event_pmu_context *pmu_ctx, bool sched_in); /* @@ -457,8 +458,8 @@ struct pmu { * implementation and Perf core context switch handling callbacks for usage * examples. */ - void (*swap_task_ctx) (struct perf_event_context *prev, - struct perf_event_context *next); + void (*swap_task_ctx) (struct perf_event_pmu_context *prev_epc, + struct perf_event_pmu_context *next_epc); /* optional */ /* @@ -522,9 +523,10 @@ struct pmu { /* optional */ /* - * Filter events for PMU-specific reasons. + * Skip programming this PMU on the given CPU. Typically needed for + * big.LITTLE things. */ - int (*filter_match) (struct perf_event *event); /* optional */ + bool (*filter) (struct pmu *pmu, int cpu); /* optional */ /* * Check period value for PERF_EVENT_IOC_PERIOD ioctl. @@ -695,6 +697,11 @@ struct perf_event { int group_caps; struct perf_event *group_leader; + /* + * event->pmu will always point to pmu in which this event belongs. + * Whereas event->pmu_ctx->pmu may point to other pmu when group of + * different pmu events is created. + */ struct pmu *pmu; void *pmu_private; @@ -720,6 +727,12 @@ struct perf_event { struct hw_perf_event hw; struct perf_event_context *ctx; + /* + * event->pmu_ctx points to perf_event_pmu_context in which the event + * is added. This pmu_ctx can be of other pmu for sw event when that + * sw event is part of a group which also contains non-sw events. + */ + struct perf_event_pmu_context *pmu_ctx; atomic_long_t refcount; /* @@ -812,19 +825,69 @@ struct perf_event { #endif /* CONFIG_PERF_EVENTS */ }; +/* + * ,-----------------------[1:n]----------------------. + * V V + * perf_event_context <-[1:n]-> perf_event_pmu_context <--- perf_event + * ^ ^ | | + * `--------[1:n]---------' `-[n:1]-> pmu <-[1:n]-' + * + * + * struct perf_event_pmu_context lifetime is refcount based and RCU freed + * (similar to perf_event_context). Locking is as if it were a member of + * perf_event_context; specifically: + * + * modification, both: ctx->mutex && ctx->lock + * reading, either: ctx->mutex || ctx->lock + * + * There is one exception to this; namely put_pmu_ctx() isn't always called + * with ctx->mutex held; this means that as long as we can guarantee the epc + * has events the above rules hold. + * + * Specificially, sys_perf_event_open()'s group_leader case depends on + * ctx->mutex pinning the configuration. Since we hold a reference on + * group_leader (through the filedesc) it can't go away, therefore it's + * associated pmu_ctx must exist and cannot change due to ctx->mutex. + */ +struct perf_event_pmu_context { + struct pmu *pmu; + struct perf_event_context *ctx; + + struct list_head pmu_ctx_entry; + + struct list_head pinned_active; + struct list_head flexible_active; + + /* Used to avoid freeing per-cpu perf_event_pmu_context */ + unsigned int embedded : 1; + + unsigned int nr_events; + + atomic_t refcount; /* event <-> epc */ + struct rcu_head rcu_head; + + void *task_ctx_data; /* pmu specific data */ + /* + * Set when one or more (plausibly active) event can't be scheduled + * due to pmu overcommit or pmu constraints, except tolerant to + * events not necessary to be active due to scheduling constraints, + * such as cgroups. + */ + int rotate_necessary; +}; struct perf_event_groups { struct rb_root tree; u64 index; }; + /** * struct perf_event_context - event context structure * * Used as a container for task events and CPU events as well: */ struct perf_event_context { - struct pmu *pmu; /* * Protect the states of the events in the list, * nr_active, and the list: @@ -837,27 +900,21 @@ struct perf_event_context { */ struct mutex mutex; - struct list_head active_ctx_list; + struct list_head pmu_ctx_list; struct perf_event_groups pinned_groups; struct perf_event_groups flexible_groups; struct list_head event_list; - struct list_head pinned_active; - struct list_head flexible_active; - int nr_events; - int nr_active; int nr_user; int is_active; + + int nr_task_data; int nr_stat; int nr_freq; int rotate_disable; - /* - * Set when nr_events != nr_active, except tolerant to events not - * necessary to be active due to scheduling constraints, such as cgroups. - */ - int rotate_necessary; - refcount_t refcount; + + refcount_t refcount; /* event <-> ctx */ struct task_struct *task; /* @@ -878,7 +935,6 @@ struct perf_event_context { #ifdef CONFIG_CGROUP_PERF int nr_cgroups; /* cgroup evts */ #endif - void *task_ctx_data; /* pmu specific data */ struct rcu_head rcu_head; /* @@ -896,12 +952,13 @@ struct perf_event_context { */ #define PERF_NR_CONTEXTS 4 -/** - * struct perf_cpu_context - per cpu event context structure - */ -struct perf_cpu_context { - struct perf_event_context ctx; - struct perf_event_context *task_ctx; +struct perf_cpu_pmu_context { + struct perf_event_pmu_context epc; + struct perf_event_pmu_context *task_epc; + + struct list_head sched_cb_entry; + int sched_cb_usage; + int active_oncpu; int exclusive; @@ -909,16 +966,20 @@ struct perf_cpu_context { struct hrtimer hrtimer; ktime_t hrtimer_interval; unsigned int hrtimer_active; +}; + +/** + * struct perf_event_cpu_context - per cpu event context structure + */ +struct perf_cpu_context { + struct perf_event_context ctx; + struct perf_event_context *task_ctx; + int online; #ifdef CONFIG_CGROUP_PERF struct perf_cgroup *cgrp; - struct list_head cgrp_cpuctx_entry; #endif - struct list_head sched_cb_entry; - int sched_cb_usage; - - int online; /* * Per-CPU storage for iterators used in visit_groups_merge. The default * storage is of size 2 to hold the CPU and any CPU event iterators. @@ -982,6 +1043,8 @@ perf_cgroup_from_task(struct task_struct *task, struct perf_event_context *ctx) #ifdef CONFIG_PERF_EVENTS +extern struct perf_event_context *perf_cpu_task_ctx(void); + extern void *perf_aux_output_begin(struct perf_output_handle *handle, struct perf_event *event); extern void perf_aux_output_end(struct perf_output_handle *handle, @@ -1187,7 +1250,7 @@ static inline int is_software_event(struct perf_event *event) */ static inline int in_software_context(struct perf_event *event) { - return event->ctx->pmu->task_ctx_nr == perf_sw_context; + return event->pmu_ctx->pmu->task_ctx_nr == perf_sw_context; } static inline int is_exclusive_pmu(struct pmu *pmu) diff --git a/include/linux/sched.h b/include/linux/sched.h index ffb6eb55cd13..4e03f1dcbe52 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1243,7 +1243,7 @@ struct task_struct { unsigned int futex_state; #endif #ifdef CONFIG_PERF_EVENTS - struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts]; + struct perf_event_context *perf_event_ctxp; struct mutex perf_event_mutex; struct list_head perf_event_list; #endif diff --git a/kernel/events/core.c b/kernel/events/core.c index 7f04f995c975..e47914ac8732 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -155,12 +155,6 @@ static int cpu_function_call(int cpu, remote_function_f func, void *info) return data.ret; } -static inline struct perf_cpu_context * -__get_cpu_context(struct perf_event_context *ctx) -{ - return this_cpu_ptr(ctx->pmu->pmu_cpu_context); -} - static void perf_ctx_lock(struct perf_cpu_context *cpuctx, struct perf_event_context *ctx) { @@ -184,6 +178,14 @@ static bool is_kernel_event(struct perf_event *event) return READ_ONCE(event->owner) == TASK_TOMBSTONE; } +static DEFINE_PER_CPU(struct perf_cpu_context, perf_cpu_context); + +struct perf_event_context *perf_cpu_task_ctx(void) +{ + lockdep_assert_irqs_disabled(); + return this_cpu_ptr(&perf_cpu_context)->task_ctx; +} + /* * On task ctx scheduling... * @@ -217,7 +219,7 @@ static int event_function(void *info) struct event_function_struct *efs = info; struct perf_event *event = efs->event; struct perf_event_context *ctx = event->ctx; - struct perf_cpu_context *cpuctx = __get_cpu_context(ctx); + struct perf_cpu_context *cpuctx = this_cpu_ptr(&perf_cpu_context); struct perf_event_context *task_ctx = cpuctx->task_ctx; int ret = 0; @@ -314,7 +316,7 @@ again: static void event_function_local(struct perf_event *event, event_f func, void *data) { struct perf_event_context *ctx = event->ctx; - struct perf_cpu_context *cpuctx = __get_cpu_context(ctx); + struct perf_cpu_context *cpuctx = this_cpu_ptr(&perf_cpu_context); struct task_struct *task = READ_ONCE(ctx->task); struct perf_event_context *task_ctx = NULL; @@ -388,7 +390,6 @@ static DEFINE_MUTEX(perf_sched_mutex); static atomic_t perf_sched_count; static DEFINE_PER_CPU(atomic_t, perf_cgroup_events); -static DEFINE_PER_CPU(int, perf_sched_cb_usages); static DEFINE_PER_CPU(struct pmu_event_list, pmu_sb_events); static atomic_t nr_mmap_events __read_mostly; @@ -448,7 +449,7 @@ static void update_perf_cpu_limits(void) WRITE_ONCE(perf_sample_allowed_ns, tmp); } -static bool perf_rotate_context(struct perf_cpu_context *cpuctx); +static bool perf_rotate_context(struct perf_cpu_pmu_context *cpc); int perf_proc_update_handler(struct ctl_table *table, int write, void *buffer, size_t *lenp, loff_t *ppos) @@ -571,12 +572,6 @@ void perf_sample_event_took(u64 sample_len_ns) static atomic64_t perf_event_id; -static void cpu_ctx_sched_out(struct perf_cpu_context *cpuctx, - enum event_type_t event_type); - -static void cpu_ctx_sched_in(struct perf_cpu_context *cpuctx, - enum event_type_t event_type); - static void update_context_time(struct perf_event_context *ctx); static u64 perf_event_time(struct perf_event *event); @@ -691,13 +686,31 @@ do { \ ___p; \ }) +static void perf_ctx_disable(struct perf_event_context *ctx) +{ + struct perf_event_pmu_context *pmu_ctx; + + list_for_each_entry(pmu_ctx, &ctx->pmu_ctx_list, pmu_ctx_entry) + perf_pmu_disable(pmu_ctx->pmu); +} + +static void perf_ctx_enable(struct perf_event_context *ctx) +{ + struct perf_event_pmu_context *pmu_ctx; + + list_for_each_entry(pmu_ctx, &ctx->pmu_ctx_list, pmu_ctx_entry) + perf_pmu_enable(pmu_ctx->pmu); +} + +static void ctx_sched_out(struct perf_event_context *ctx, enum event_type_t event_type); +static void ctx_sched_in(struct perf_event_context *ctx, enum event_type_t event_type); + #ifdef CONFIG_CGROUP_PERF static inline bool perf_cgroup_match(struct perf_event *event) { - struct perf_event_context *ctx = event->ctx; - struct perf_cpu_context *cpuctx = __get_cpu_context(ctx); + struct perf_cpu_context *cpuctx = this_cpu_ptr(&perf_cpu_context); /* @event doesn't care about cgroup */ if (!event->cgrp) @@ -823,54 +836,39 @@ perf_cgroup_set_timestamp(struct perf_cpu_context *cpuctx) } } -static DEFINE_PER_CPU(struct list_head, cgrp_cpuctx_list); - /* * reschedule events based on the cgroup constraint of task. */ static void perf_cgroup_switch(struct task_struct *task) { + struct perf_cpu_context *cpuctx = this_cpu_ptr(&perf_cpu_context); struct perf_cgroup *cgrp; - struct perf_cpu_context *cpuctx, *tmp; - struct list_head *list; - unsigned long flags; - - /* - * Disable interrupts and preemption to avoid this CPU's - * cgrp_cpuctx_entry to change under us. - */ - local_irq_save(flags); cgrp = perf_cgroup_from_task(task, NULL); - list = this_cpu_ptr(&cgrp_cpuctx_list); - list_for_each_entry_safe(cpuctx, tmp, list, cgrp_cpuctx_entry) { - WARN_ON_ONCE(cpuctx->ctx.nr_cgroups == 0); - if (READ_ONCE(cpuctx->cgrp) == cgrp) - continue; - - perf_ctx_lock(cpuctx, cpuctx->task_ctx); - perf_pmu_disable(cpuctx->ctx.pmu); + WARN_ON_ONCE(cpuctx->ctx.nr_cgroups == 0); + if (READ_ONCE(cpuctx->cgrp) == cgrp) + return; - cpu_ctx_sched_out(cpuctx, EVENT_ALL); - /* - * must not be done before ctxswout due - * to update_cgrp_time_from_cpuctx() in - * ctx_sched_out() - */ - cpuctx->cgrp = cgrp; - /* - * set cgrp before ctxsw in to allow - * perf_cgroup_set_timestamp() in ctx_sched_in() - * to not have to pass task around - */ - cpu_ctx_sched_in(cpuctx, EVENT_ALL); + perf_ctx_lock(cpuctx, cpuctx->task_ctx); + perf_ctx_disable(&cpuctx->ctx); - perf_pmu_enable(cpuctx->ctx.pmu); - perf_ctx_unlock(cpuctx, cpuctx->task_ctx); - } + ctx_sched_out(&cpuctx->ctx, EVENT_ALL); + /* + * must not be done before ctxswout due + * to update_cgrp_time_from_cpuctx() in + * ctx_sched_out() + */ + cpuctx->cgrp = cgrp; + /* + * set cgrp before ctxsw in to allow + * perf_cgroup_set_timestamp() in ctx_sched_in() + * to not have to pass task around + */ + ctx_sched_in(&cpuctx->ctx, EVENT_ALL); - local_irq_restore(flags); + perf_ctx_enable(&cpuctx->ctx); + perf_ctx_unlock(cpuctx, cpuctx->task_ctx); } static int perf_cgroup_ensure_storage(struct perf_event *event, @@ -888,7 +886,7 @@ static int perf_cgroup_ensure_storage(struct perf_event *event, heap_size++; for_each_possible_cpu(cpu) { - cpuctx = per_cpu_ptr(event->pmu->pmu_cpu_context, cpu); + cpuctx = per_cpu_ptr(&perf_cpu_context, cpu); if (heap_size <= cpuctx->heap_size) continue; @@ -972,8 +970,6 @@ perf_cgroup_event_enable(struct perf_event *event, struct perf_event_context *ct return; cpuctx->cgrp = perf_cgroup_from_task(current, ctx); - list_add(&cpuctx->cgrp_cpuctx_entry, - per_cpu_ptr(&cgrp_cpuctx_list, event->cpu)); } static inline void @@ -994,7 +990,6 @@ perf_cgroup_event_disable(struct perf_event *event, struct perf_event_context *c return; cpuctx->cgrp = NULL; - list_del(&cpuctx->cgrp_cpuctx_entry); } #else /* !CONFIG_CGROUP_PERF */ @@ -1069,34 +1064,30 @@ static void perf_cgroup_switch(struct task_struct *task) */ static enum hrtimer_restart perf_mux_hrtimer_handler(struct hrtimer *hr) { - struct perf_cpu_context *cpuctx; + struct perf_cpu_pmu_context *cpc; bool rotations; lockdep_assert_irqs_disabled(); - cpuctx = container_of(hr, struct perf_cpu_context, hrtimer); - rotations = perf_rotate_context(cpuctx); + cpc = container_of(hr, struct perf_cpu_pmu_context, hrtimer); + rotations = perf_rotate_context(cpc); - raw_spin_lock(&cpuctx->hrtimer_lock); + raw_spin_lock(&cpc->hrtimer_lock); if (rotations) - hrtimer_forward_now(hr, cpuctx->hrtimer_interval); + hrtimer_forward_now(hr, cpc->hrtimer_interval); else - cpuctx->hrtimer_active = 0; - raw_spin_unlock(&cpuctx->hrtimer_lock); + cpc->hrtimer_active = 0; + raw_spin_unlock(&cpc->hrtimer_lock); return rotations ? HRTIMER_RESTART : HRTIMER_NORESTART; } -static void __perf_mux_hrtimer_init(struct perf_cpu_context *cpuctx, int cpu) +static void __perf_mux_hrtimer_init(struct perf_cpu_pmu_context *cpc, int cpu) { - struct hrtimer *timer = &cpuctx->hrtimer; - struct pmu *pmu = cpuctx->ctx.pmu; + struct hrtimer *timer = &cpc->hrtimer; + struct pmu *pmu = cpc->epc.pmu; u64 interval; - /* no multiplexing needed for SW PMU */ - if (pmu->task_ctx_nr == perf_sw_context) - return; - /* * check default is sane, if not set then force to * default interval (1/tick) @@ -1105,34 +1096,34 @@ static void __perf_mux_hrtimer_init(struct perf_cpu_context *cpuctx, int cpu) if (interval < 1) interval = pmu->hrtimer_interval_ms = PERF_CPU_HRTIMER; - cpuctx->hrtimer_interval = ns_to_ktime(NSEC_PER_MSEC * interval); + cpc->hrtimer_interval = ns_to_ktime(NSEC_PER_MSEC * interval); - raw_spin_lock_init(&cpuctx->hrtimer_lock); + raw_spin_lock_init(&cpc->hrtimer_lock); hrtimer_init(timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS_PINNED_HARD); timer->function = perf_mux_hrtimer_handler; } -static int perf_mux_hrtimer_restart(struct perf_cpu_context *cpuctx) +static int perf_mux_hrtimer_restart(struct perf_cpu_pmu_context *cpc) { - struct hrtimer *timer = &cpuctx->hrtimer; - struct pmu *pmu = cpuctx->ctx.pmu; + struct hrtimer *timer = &cpc->hrtimer; unsigned long flags; - /* not for SW PMU */ - if (pmu->task_ctx_nr == perf_sw_context) - return 0; - - raw_spin_lock_irqsave(&cpuctx->hrtimer_lock, flags); - if (!cpuctx->hrtimer_active) { - cpuctx->hrtimer_active = 1; - hrtimer_forward_now(timer, cpuctx->hrtimer_interval); + raw_spin_lock_irqsave(&cpc->hrtimer_lock, flags); + if (!cpc->hrtimer_active) { + cpc->hrtimer_active = 1; + hrtimer_forward_now(timer, cpc->hrtimer_interval); hrtimer_start_expires(timer, HRTIMER_MODE_ABS_PINNED_HARD); } - raw_spin_unlock_irqrestore(&cpuctx->hrtimer_lock, flags); + raw_spin_unlock_irqrestore(&cpc->hrtimer_lock, flags); return 0; } +static int perf_mux_hrtimer_restart_ipi(void *arg) +{ + return perf_mux_hrtimer_restart(arg); +} + void perf_pmu_disable(struct pmu *pmu) { int *count = this_cpu_ptr(pmu->pmu_disable_count); @@ -1147,32 +1138,9 @@ void perf_pmu_enable(struct pmu *pmu) pmu->pmu_enable(pmu); } -static DEFINE_PER_CPU(struct list_head, active_ctx_list); - -/* - * perf_event_ctx_activate(), perf_event_ctx_deactivate(), and - * perf_event_task_tick() are fully serialized because they're strictly cpu - * affine and perf_event_ctx{activate,deactivate} are called with IRQs - * disabled, while perf_event_task_tick is called from IRQ context. - */ -static void perf_event_ctx_activate(struct perf_event_context *ctx) +static void perf_assert_pmu_disabled(struct pmu *pmu) { - struct list_head *head = this_cpu_ptr(&active_ctx_list); - - lockdep_assert_irqs_disabled(); - - WARN_ON(!list_empty(&ctx->active_ctx_list)); - - list_add(&ctx->active_ctx_list, head); -} - -static void perf_event_ctx_deactivate(struct perf_event_context *ctx) -{ - lockdep_assert_irqs_disabled(); - - WARN_ON(list_empty(&ctx->active_ctx_list)); - - list_del_init(&ctx->active_ctx_list); + WARN_ON_ONCE(*this_cpu_ptr(pmu->pmu_disable_count) == 0); } static void get_ctx(struct perf_event_context *ctx) @@ -1199,7 +1167,6 @@ static void free_ctx(struct rcu_head *head) struct perf_event_context *ctx; ctx = container_of(head, struct perf_event_context, rcu_head); - free_task_ctx_data(ctx->pmu, ctx->task_ctx_data); kfree(ctx); } @@ -1384,7 +1351,7 @@ static u64 primary_event_id(struct perf_event *event) * the context could get moved to another task. */ static struct perf_event_context * -perf_lock_task_context(struct task_struct *task, int ctxn, unsigned long *flags) +perf_lock_task_context(struct task_struct *task, unsigned long *flags) { struct perf_event_context *ctx; @@ -1400,7 +1367,7 @@ retry: */ local_irq_save(*flags); rcu_read_lock(); - ctx = rcu_dereference(task->perf_event_ctxp[ctxn]); + ctx = rcu_dereference(task->perf_event_ctxp); if (ctx) { /* * If this context is a clone of another, it might @@ -1413,7 +1380,7 @@ retry: * can't get swapped on us any more. */ raw_spin_lock(&ctx->lock); - if (ctx != rcu_dereference(task->perf_event_ctxp[ctxn])) { + if (ctx != rcu_dereference(task->perf_event_ctxp)) { raw_spin_unlock(&ctx->lock); rcu_read_unlock(); local_irq_restore(*flags); @@ -1440,12 +1407,12 @@ retry: * reference count so that the context can't get freed. */ static struct perf_event_context * -perf_pin_task_context(struct task_struct *task, int ctxn) +perf_pin_task_context(struct task_struct *task) { struct perf_event_context *ctx; unsigned long flags; - ctx = perf_lock_task_context(task, ctxn, &flags); + ctx = perf_lock_task_context(task, &flags); if (ctx) { ++ctx->pin_count; raw_spin_unlock_irqrestore(&ctx->lock, flags); @@ -1593,14 +1560,22 @@ static inline struct cgroup *event_cgroup(const struct perf_event *event) * which provides ordering when rotating groups for the same CPU. */ static __always_inline int -perf_event_groups_cmp(const int left_cpu, const struct cgroup *left_cgroup, - const u64 left_group_index, const struct perf_event *right) +perf_event_groups_cmp(const int left_cpu, const struct pmu *left_pmu, + const struct cgroup *left_cgroup, const u64 left_group_index, + const struct perf_event *right) { if (left_cpu < right->cpu) return -1; if (left_cpu > right->cpu) return 1; + if (left_pmu) { + if (left_pmu < right->pmu_ctx->pmu) + return -1; + if (left_pmu > right->pmu_ctx->pmu) + return 1; + } + #ifdef CONFIG_CGROUP_PERF { const struct cgroup *right_cgroup = event_cgroup(right); @@ -1643,12 +1618,13 @@ perf_event_groups_cmp(const int left_cpu, const struct cgroup *left_cgroup, static inline bool __group_less(struct rb_node *a, const struct rb_node *b) { struct perf_event *e = __node_2_pe(a); - return perf_event_groups_cmp(e->cpu, event_cgroup(e), e->group_index, - __node_2_pe(b)) < 0; + return perf_event_groups_cmp(e->cpu, e->pmu_ctx->pmu, event_cgroup(e), + e->group_index, __node_2_pe(b)) < 0; } struct __group_key { int cpu; + struct pmu *pmu; struct cgroup *cgroup; }; @@ -1657,14 +1633,25 @@ static inline int __group_cmp(const void *key, const struct rb_node *node) const struct __group_key *a = key; const struct perf_event *b = __node_2_pe(node); - /* partial/subtree match: @cpu, @cgroup; ignore: @group_index */ - return perf_event_groups_cmp(a->cpu, a->cgroup, b->group_index, b); + /* partial/subtree match: @cpu, @pmu, @cgroup; ignore: @group_index */ + return perf_event_groups_cmp(a->cpu, a->pmu, a->cgroup, b->group_index, b); +} + +static inline int +__group_cmp_ignore_cgroup(const void *key, const struct rb_node *node) +{ + const struct __group_key *a = key; + const struct perf_event *b = __node_2_pe(node); + + /* partial/subtree match: @cpu, @pmu, ignore: @cgroup, @group_index */ + return perf_event_groups_cmp(a->cpu, a->pmu, event_cgroup(b), + b->group_index, b); } /* - * Insert @event into @groups' tree; using {@event->cpu, ++@groups->index} for - * key (see perf_event_groups_less). This places it last inside the CPU - * subtree. + * Insert @event into @groups' tree; using + * {@event->cpu, @event->pmu_ctx->pmu, event_cgroup(@event), ++@groups->index} + * as key. This places it last inside the {cpu,pmu,cgroup} subtree. */ static void perf_event_groups_insert(struct perf_event_groups *groups, @@ -1714,14 +1701,15 @@ del_event_from_groups(struct perf_event *event, struct perf_event_context *ctx) } /* - * Get the leftmost event in the cpu/cgroup subtree. + * Get the leftmost event in the {cpu,pmu,cgroup} subtree. */ static struct perf_event * perf_event_groups_first(struct perf_event_groups *groups, int cpu, - struct cgroup *cgrp) + struct pmu *pmu, struct cgroup *cgrp) { struct __group_key key = { .cpu = cpu, + .pmu = pmu, .cgroup = cgrp, }; struct rb_node *node; @@ -1733,14 +1721,12 @@ perf_event_groups_first(struct perf_event_groups *groups, int cpu, return NULL; } -/* - * Like rb_entry_next_safe() for the @cpu subtree. - */ static struct perf_event * -perf_event_groups_next(struct perf_event *event) +perf_event_groups_next(struct perf_event *event, struct pmu *pmu) { struct __group_key key = { .cpu = event->cpu, + .pmu = pmu, .cgroup = event_cgroup(event), }; struct rb_node *next; @@ -1752,6 +1738,10 @@ perf_event_groups_next(struct perf_event *event) return NULL; } +#define perf_event_groups_for_cpu_pmu(event, groups, cpu, pmu) \ + for (event = perf_event_groups_first(groups, cpu, pmu, NULL); \ + event; event = perf_event_groups_next(event, pmu)) + /* * Iterate through the whole groups tree. */ @@ -1796,6 +1786,7 @@ list_add_event(struct perf_event *event, struct perf_event_context *ctx) perf_cgroup_event_enable(event, ctx); ctx->generation++; + event->pmu_ctx->nr_events++; } /* @@ -1941,7 +1932,8 @@ static void perf_group_attach(struct perf_event *event) lockdep_assert_held(&event->ctx->lock); /* - * We can have double attach due to group movement in perf_event_open. + * We can have double attach due to group movement (move_group) in + * perf_event_open(). */ if (event->attach_state & PERF_ATTACH_GROUP) return; @@ -2006,6 +1998,7 @@ list_del_event(struct perf_event *event, struct perf_event_context *ctx) } ctx->generation++; + event->pmu_ctx->nr_events--; } static int @@ -2022,13 +2015,11 @@ perf_aux_output_match(struct perf_event *event, struct perf_event *aux_event) static void put_event(struct perf_event *event); static void event_sched_out(struct perf_event *event, - struct perf_cpu_context *cpuctx, struct perf_event_context *ctx); static void perf_put_aux_event(struct perf_event *event) { struct perf_event_context *ctx = event->ctx; - struct perf_cpu_context *cpuctx = __get_cpu_context(ctx); struct perf_event *iter; /* @@ -2057,7 +2048,7 @@ static void perf_put_aux_event(struct perf_event *event) * state so that we don't try to schedule it again. Note * that perf_event_enable() will clear the ERROR status. */ - event_sched_out(iter, cpuctx, ctx); + event_sched_out(iter, ctx); perf_event_set_state(event, PERF_EVENT_STATE_ERROR); } } @@ -2108,8 +2099,8 @@ static int perf_get_aux_event(struct perf_event *event, static inline struct list_head *get_event_list(struct perf_event *event) { - struct perf_event_context *ctx = event->ctx; - return event->attr.pinned ? &ctx->pinned_active : &ctx->flexible_active; + return event->attr.pinned ? &event->pmu_ctx->pinned_active : + &event->pmu_ctx->flexible_active; } /* @@ -2120,10 +2111,7 @@ static inline struct list_head *get_event_list(struct perf_event *event) */ static inline void perf_remove_sibling_event(struct perf_event *event) { - struct perf_event_context *ctx = event->ctx; - struct perf_cpu_context *cpuctx = __get_cpu_context(ctx); - - event_sched_out(event, cpuctx, ctx); + event_sched_out(event, event->ctx); perf_event_set_state(event, PERF_EVENT_STATE_ERROR); } @@ -2212,53 +2200,22 @@ static bool is_orphaned_event(struct perf_event *event) return event->state == PERF_EVENT_STATE_DEAD; } -static inline int __pmu_filter_match(struct perf_event *event) -{ - struct pmu *pmu = event->pmu; - return pmu->filter_match ? pmu->filter_match(event) : 1; -} - -/* - * Check whether we should attempt to schedule an event group based on - * PMU-specific filtering. An event group can consist of HW and SW events, - * potentially with a SW leader, so we must check all the filters, to - * determine whether a group is schedulable: - */ -static inline int pmu_filter_match(struct perf_event *event) -{ - struct perf_event *sibling; - unsigned long flags; - int ret = 1; - - if (!__pmu_filter_match(event)) - return 0; - - local_irq_save(flags); - for_each_sibling_event(sibling, event) { - if (!__pmu_filter_match(sibling)) { - ret = 0; - break; - } - } - local_irq_restore(flags); - - return ret; -} - static inline int event_filter_match(struct perf_event *event) { return (event->cpu == -1 || event->cpu == smp_processor_id()) && - perf_cgroup_match(event) && pmu_filter_match(event); + perf_cgroup_match(event); } static void -event_sched_out(struct perf_event *event, - struct perf_cpu_context *cpuctx, - struct perf_event_context *ctx) +event_sched_out(struct perf_event *event, struct perf_event_context *ctx) { + struct perf_event_pmu_context *epc = event->pmu_ctx; + struct perf_cpu_pmu_context *cpc = this_cpu_ptr(epc->pmu->cpu_pmu_context); enum perf_event_state state = PERF_EVENT_STATE_INACTIVE; + // XXX cpc serialization, probably per-cpu IRQ disabled + WARN_ON_ONCE(event->ctx != ctx); lockdep_assert_held(&ctx->lock); @@ -2301,38 +2258,32 @@ event_sched_out(struct perf_event *event, perf_event_set_state(event, state); if (!is_software_event(event)) - cpuctx->active_oncpu--; - if (!--ctx->nr_active) - perf_event_ctx_deactivate(ctx); + cpc->active_oncpu--; if (event->attr.freq && event->attr.sample_freq) ctx->nr_freq--; - if (event->attr.exclusive || !cpuctx->active_oncpu) - cpuctx->exclusive = 0; + if (event->attr.exclusive || !cpc->active_oncpu) + cpc->exclusive = 0; perf_pmu_enable(event->pmu); } static void -group_sched_out(struct perf_event *group_event, - struct perf_cpu_context *cpuctx, - struct perf_event_context *ctx) +group_sched_out(struct perf_event *group_event, struct perf_event_context *ctx) { struct perf_event *event; if (group_event->state != PERF_EVENT_STATE_ACTIVE) return; - perf_pmu_disable(ctx->pmu); + perf_assert_pmu_disabled(group_event->pmu_ctx->pmu); - event_sched_out(group_event, cpuctx, ctx); + event_sched_out(group_event, ctx); /* * Schedule out siblings (if any): */ for_each_sibling_event(event, group_event) - event_sched_out(event, cpuctx, ctx); - - perf_pmu_enable(ctx->pmu); + event_sched_out(event, ctx); } #define DETACH_GROUP 0x01UL @@ -2351,6 +2302,7 @@ __perf_remove_from_context(struct perf_event *event, struct perf_event_context *ctx, void *info) { + struct perf_event_pmu_context *pmu_ctx = event->pmu_ctx; unsigned long flags = (unsigned long)info; if (ctx->is_active & EVENT_TIME) { @@ -2364,7 +2316,7 @@ __perf_remove_from_context(struct perf_event *event, */ if (flags & DETACH_DEAD) event->pending_disable = 1; - event_sched_out(event, cpuctx, ctx); + event_sched_out(event, ctx); if (flags & DETACH_GROUP) perf_group_detach(event); if (flags & DETACH_CHILD) @@ -2373,12 +2325,23 @@ __perf_remove_from_context(struct perf_event *event, if (flags & DETACH_DEAD) event->state = PERF_EVENT_STATE_DEAD; + if (!pmu_ctx->nr_events) { + pmu_ctx->rotate_necessary = 0; + + if (ctx->task && ctx->is_active) { + struct perf_cpu_pmu_context *cpc; + + cpc = this_cpu_ptr(pmu_ctx->pmu->cpu_pmu_context); + WARN_ON_ONCE(cpc->task_epc && cpc->task_epc != pmu_ctx); + cpc->task_epc = NULL; + } + } + if (!ctx->nr_events && ctx->is_active) { if (ctx == &cpuctx->ctx) update_cgrp_time_from_cpuctx(cpuctx, true); ctx->is_active = 0; - ctx->rotate_necessary = 0; if (ctx->task) { WARN_ON_ONCE(cpuctx->task_ctx != ctx); cpuctx->task_ctx = NULL; @@ -2408,12 +2371,8 @@ static void perf_remove_from_context(struct perf_event *event, unsigned long fla * event_function_call() user. */ raw_spin_lock_irq(&ctx->lock); - /* - * Cgroup events are per-cpu events, and must IPI because of - * cgrp_cpuctx_list. - */ - if (!ctx->is_active && !is_cgroup_event(event)) { - __perf_remove_from_context(event, __get_cpu_context(ctx), + if (!ctx->is_active) { + __perf_remove_from_context(event, this_cpu_ptr(&perf_cpu_context), ctx, (void *)flags); raw_spin_unlock_irq(&ctx->lock); return; @@ -2439,13 +2398,17 @@ static void __perf_event_disable(struct perf_event *event, update_cgrp_time_from_event(event); } + perf_pmu_disable(event->pmu_ctx->pmu); + if (event == event->group_leader) - group_sched_out(event, cpuctx, ctx); + group_sched_out(event, ctx); else - event_sched_out(event, cpuctx, ctx); + event_sched_out(event, ctx); perf_event_set_state(event, PERF_EVENT_STATE_OFF); perf_cgroup_event_disable(event, ctx); + + perf_pmu_enable(event->pmu_ctx->pmu); } /* @@ -2507,10 +2470,10 @@ static void perf_log_throttle(struct perf_event *event, int enable); static void perf_log_itrace_start(struct perf_event *event); static int -event_sched_in(struct perf_event *event, - struct perf_cpu_context *cpuctx, - struct perf_event_context *ctx) +event_sched_in(struct perf_event *event, struct perf_event_context *ctx) { + struct perf_event_pmu_context *epc = event->pmu_ctx; + struct perf_cpu_pmu_context *cpc = this_cpu_ptr(epc->pmu->cpu_pmu_context); int ret = 0; WARN_ON_ONCE(event->ctx != ctx); @@ -2551,14 +2514,12 @@ event_sched_in(struct perf_event *event, } if (!is_software_event(event)) - cpuctx->active_oncpu++; - if (!ctx->nr_active++) - perf_event_ctx_activate(ctx); + cpc->active_oncpu++; if (event->attr.freq && event->attr.sample_freq) ctx->nr_freq++; if (event->attr.exclusive) - cpuctx->exclusive = 1; + cpc->exclusive = 1; out: perf_pmu_enable(event->pmu); @@ -2567,26 +2528,24 @@ out: } static int -group_sched_in(struct perf_event *group_event, - struct perf_cpu_context *cpuctx, - struct perf_event_context *ctx) +group_sched_in(struct perf_event *group_event, struct perf_event_context *ctx) { struct perf_event *event, *partial_group = NULL; - struct pmu *pmu = ctx->pmu; + struct pmu *pmu = group_event->pmu_ctx->pmu; if (group_event->state == PERF_EVENT_STATE_OFF) return 0; pmu->start_txn(pmu, PERF_PMU_TXN_ADD); - if (event_sched_in(group_event, cpuctx, ctx)) + if (event_sched_in(group_event, ctx)) goto error; /* * Schedule in siblings as one group (if any): */ for_each_sibling_event(event, group_event) { - if (event_sched_in(event, cpuctx, ctx)) { + if (event_sched_in(event, ctx)) { partial_group = event; goto group_error; } @@ -2605,9 +2564,9 @@ group_error: if (event == partial_group) break; - event_sched_out(event, cpuctx, ctx); + event_sched_out(event, ctx); } - event_sched_out(group_event, cpuctx, ctx); + event_sched_out(group_event, ctx); error: pmu->cancel_txn(pmu); @@ -2617,10 +2576,11 @@ error: /* * Work out whether we can put this event group on the CPU now. */ -static int group_can_go_on(struct perf_event *event, - struct perf_cpu_context *cpuctx, - int can_add_hw) +static int group_can_go_on(struct perf_event *event, int can_add_hw) { + struct perf_event_pmu_context *epc = event->pmu_ctx; + struct perf_cpu_pmu_context *cpc = this_cpu_ptr(epc->pmu->cpu_pmu_context); + /* * Groups consisting entirely of software events can always go on. */ @@ -2630,7 +2590,7 @@ static int group_can_go_on(struct perf_event *event, * If an exclusive group is already on, no other hardware * events can go on. */ - if (cpuctx->exclusive) + if (cpc->exclusive) return 0; /* * If this group is exclusive and there are already @@ -2652,36 +2612,29 @@ static void add_event_to_ctx(struct perf_event *event, perf_group_attach(event); } -static void ctx_sched_out(struct perf_event_context *ctx, - struct perf_cpu_context *cpuctx, - enum event_type_t event_type); -static void -ctx_sched_in(struct perf_event_context *ctx, - struct perf_cpu_context *cpuctx, - enum event_type_t event_type); - -static void task_ctx_sched_out(struct perf_cpu_context *cpuctx, - struct perf_event_context *ctx, - enum event_type_t event_type) +static void task_ctx_sched_out(struct perf_event_context *ctx, + enum event_type_t event_type) { + struct perf_cpu_context *cpuctx = this_cpu_ptr(&perf_cpu_context); + if (!cpuctx->task_ctx) return; if (WARN_ON_ONCE(ctx != cpuctx->task_ctx)) return; - ctx_sched_out(ctx, cpuctx, event_type); + ctx_sched_out(ctx, event_type); } static void perf_event_sched_in(struct perf_cpu_context *cpuctx, struct perf_event_context *ctx) { - cpu_ctx_sched_in(cpuctx, EVENT_PINNED); + ctx_sched_in(&cpuctx->ctx, EVENT_PINNED); if (ctx) - ctx_sched_in(ctx, cpuctx, EVENT_PINNED); - cpu_ctx_sched_in(cpuctx, EVENT_FLEXIBLE); + ctx_sched_in(ctx, EVENT_PINNED); + ctx_sched_in(&cpuctx->ctx, EVENT_FLEXIBLE); if (ctx) - ctx_sched_in(ctx, cpuctx, EVENT_FLEXIBLE); + ctx_sched_in(ctx, EVENT_FLEXIBLE); } /* @@ -2699,11 +2652,15 @@ static void perf_event_sched_in(struct perf_cpu_context *cpuctx, * event_type is a bit mask of the types of events involved. For CPU events, * event_type is only either EVENT_PINNED or EVENT_FLEXIBLE. */ +/* + * XXX: ctx_resched() reschedule entire perf_event_context while adding new + * event to the context or enabling existing event in the context. We can + * probably optimize it by rescheduling only affected pmu_ctx. + */ static void ctx_resched(struct perf_cpu_context *cpuctx, struct perf_event_context *task_ctx, enum event_type_t event_type) { - enum event_type_t ctx_event_type; bool cpu_event = !!(event_type & EVENT_CPU); /* @@ -2713,11 +2670,13 @@ static void ctx_resched(struct perf_cpu_context *cpuctx, if (event_type & EVENT_PINNED) event_type |= EVENT_FLEXIBLE; - ctx_event_type = event_type & EVENT_ALL; + event_type &= EVENT_ALL; - perf_pmu_disable(cpuctx->ctx.pmu); - if (task_ctx) - task_ctx_sched_out(cpuctx, task_ctx, event_type); + perf_ctx_disable(&cpuctx->ctx); + if (task_ctx) { + perf_ctx_disable(task_ctx); + task_ctx_sched_out(task_ctx, event_type); + } /* * Decide which cpu ctx groups to schedule out based on the types @@ -2727,17 +2686,20 @@ static void ctx_resched(struct perf_cpu_context *cpuctx, * - otherwise, do nothing more. */ if (cpu_event) - cpu_ctx_sched_out(cpuctx, ctx_event_type); - else if (ctx_event_type & EVENT_PINNED) - cpu_ctx_sched_out(cpuctx, EVENT_FLEXIBLE); + ctx_sched_out(&cpuctx->ctx, event_type); + else if (event_type & EVENT_PINNED) + ctx_sched_out(&cpuctx->ctx, EVENT_FLEXIBLE); perf_event_sched_in(cpuctx, task_ctx); - perf_pmu_enable(cpuctx->ctx.pmu); + + perf_ctx_enable(&cpuctx->ctx); + if (task_ctx) + perf_ctx_enable(task_ctx); } void perf_pmu_resched(struct pmu *pmu) { - struct perf_cpu_context *cpuctx = this_cpu_ptr(pmu->pmu_cpu_context); + struct perf_cpu_context *cpuctx = this_cpu_ptr(&perf_cpu_context); struct perf_event_context *task_ctx = cpuctx->task_ctx; perf_ctx_lock(cpuctx, task_ctx); @@ -2755,7 +2717,7 @@ static int __perf_install_in_context(void *info) { struct perf_event *event = info; struct perf_event_context *ctx = event->ctx; - struct perf_cpu_context *cpuctx = __get_cpu_context(ctx); + struct perf_cpu_context *cpuctx = this_cpu_ptr(&perf_cpu_context); struct perf_event_context *task_ctx = cpuctx->task_ctx; bool reprogram = true; int ret = 0; @@ -2797,7 +2759,7 @@ static int __perf_install_in_context(void *info) #endif if (reprogram) { - ctx_sched_out(ctx, cpuctx, EVENT_TIME); + ctx_sched_out(ctx, EVENT_TIME); add_event_to_ctx(event, ctx); ctx_resched(cpuctx, task_ctx, get_event_type(event)); } else { @@ -2830,7 +2792,7 @@ perf_install_in_context(struct perf_event_context *ctx, WARN_ON_ONCE(!exclusive_event_installable(event, ctx)); if (event->cpu != -1) - event->cpu = cpu; + WARN_ON_ONCE(event->cpu != cpu); /* * Ensures that if we can observe event->ctx, both the event and ctx @@ -2842,8 +2804,6 @@ perf_install_in_context(struct perf_event_context *ctx, * perf_event_attr::disabled events will not run and can be initialized * without IPI. Except when this is the first event for the context, in * that case we need the magic of the IPI to set ctx->is_active. - * Similarly, cgroup events for the context also needs the IPI to - * manipulate the cgrp_cpuctx_list. * * The IOC_ENABLE that is sure to follow the creation of a disabled * event will issue the IPI and reprogram the hardware. @@ -2945,7 +2905,7 @@ static void __perf_event_enable(struct perf_event *event, return; if (ctx->is_active) - ctx_sched_out(ctx, cpuctx, EVENT_TIME); + ctx_sched_out(ctx, EVENT_TIME); perf_event_set_state(event, PERF_EVENT_STATE_INACTIVE); perf_cgroup_event_enable(event, ctx); @@ -2954,7 +2914,7 @@ static void __perf_event_enable(struct perf_event *event, return; if (!event_filter_match(event)) { - ctx_sched_in(ctx, cpuctx, EVENT_TIME); + ctx_sched_in(ctx, EVENT_TIME); return; } @@ -2963,7 +2923,7 @@ static void __perf_event_enable(struct perf_event *event, * then don't put it on unless the group is on. */ if (leader != event && leader->state != PERF_EVENT_STATE_ACTIVE) { - ctx_sched_in(ctx, cpuctx, EVENT_TIME); + ctx_sched_in(ctx, EVENT_TIME); return; } @@ -3232,11 +3192,52 @@ out: return err; } -static void ctx_sched_out(struct perf_event_context *ctx, - struct perf_cpu_context *cpuctx, - enum event_type_t event_type) +static void __pmu_ctx_sched_out(struct perf_event_pmu_context *pmu_ctx, + enum event_type_t event_type) { + struct perf_event_context *ctx = pmu_ctx->ctx; struct perf_event *event, *tmp; + struct pmu *pmu = pmu_ctx->pmu; + + if (ctx->task && !ctx->is_active) { + struct perf_cpu_pmu_context *cpc; + + cpc = this_cpu_ptr(pmu->cpu_pmu_context); + WARN_ON_ONCE(cpc->task_epc && cpc->task_epc != pmu_ctx); + cpc->task_epc = NULL; + } + + if (!event_type) + return; + + perf_pmu_disable(pmu); + if (event_type & EVENT_PINNED) { + list_for_each_entry_safe(event, tmp, + &pmu_ctx->pinned_active, + active_list) + group_sched_out(event, ctx); + } + + if (event_type & EVENT_FLEXIBLE) { + list_for_each_entry_safe(event, tmp, + &pmu_ctx->flexible_active, + active_list) + group_sched_out(event, ctx); + /* + * Since we cleared EVENT_FLEXIBLE, also clear + * rotate_necessary, is will be reset by + * ctx_flexible_sched_in() when needed. + */ + pmu_ctx->rotate_necessary = 0; + } + perf_pmu_enable(pmu); +} + +static void +ctx_sched_out(struct perf_event_context *ctx, enum event_type_t event_type) +{ + struct perf_cpu_context *cpuctx = this_cpu_ptr(&perf_cpu_context); + struct perf_event_pmu_context *pmu_ctx; int is_active = ctx->is_active; lockdep_assert_held(&ctx->lock); @@ -3284,27 +3285,8 @@ static void ctx_sched_out(struct perf_event_context *ctx, is_active ^= ctx->is_active; /* changed bits */ - if (!ctx->nr_active || !(is_active & EVENT_ALL)) - return; - - perf_pmu_disable(ctx->pmu); - if (is_active & EVENT_PINNED) { - list_for_each_entry_safe(event, tmp, &ctx->pinned_active, active_list) - group_sched_out(event, cpuctx, ctx); - } - - if (is_active & EVENT_FLEXIBLE) { - list_for_each_entry_safe(event, tmp, &ctx->flexible_active, active_list) - group_sched_out(event, cpuctx, ctx); - - /* - * Since we cleared EVENT_FLEXIBLE, also clear - * rotate_necessary, is will be reset by - * ctx_flexible_sched_in() when needed. - */ - ctx->rotate_necessary = 0; - } - perf_pmu_enable(ctx->pmu); + list_for_each_entry(pmu_ctx, &ctx->pmu_ctx_list, pmu_ctx_entry) + __pmu_ctx_sched_out(pmu_ctx, is_active); } /* @@ -3409,26 +3391,68 @@ static void perf_event_sync_stat(struct perf_event_context *ctx, } } -static void perf_event_context_sched_out(struct task_struct *task, int ctxn, - struct task_struct *next) +#define double_list_for_each_entry(pos1, pos2, head1, head2, member) \ + for (pos1 = list_first_entry(head1, typeof(*pos1), member), \ + pos2 = list_first_entry(head2, typeof(*pos2), member); \ + !list_entry_is_head(pos1, head1, member) && \ + !list_entry_is_head(pos2, head2, member); \ + pos1 = list_next_entry(pos1, member), \ + pos2 = list_next_entry(pos2, member)) + +static void perf_event_swap_task_ctx_data(struct perf_event_context *prev_ctx, + struct perf_event_context *next_ctx) { - struct perf_event_context *ctx = task->perf_event_ctxp[ctxn]; + struct perf_event_pmu_context *prev_epc, *next_epc; + + if (!prev_ctx->nr_task_data) + return; + + double_list_for_each_entry(prev_epc, next_epc, + &prev_ctx->pmu_ctx_list, &next_ctx->pmu_ctx_list, + pmu_ctx_entry) { + + if (WARN_ON_ONCE(prev_epc->pmu != next_epc->pmu)) + continue; + + /* + * PMU specific parts of task perf context can require + * additional synchronization. As an example of such + * synchronization see implementation details of Intel + * LBR call stack data profiling; + */ + if (prev_epc->pmu->swap_task_ctx) + prev_epc->pmu->swap_task_ctx(prev_epc, next_epc); + else + swap(prev_epc->task_ctx_data, next_epc->task_ctx_data); + } +} + +static void perf_ctx_sched_task_cb(struct perf_event_context *ctx, bool sched_in) +{ + struct perf_event_pmu_context *pmu_ctx; + struct perf_cpu_pmu_context *cpc; + + list_for_each_entry(pmu_ctx, &ctx->pmu_ctx_list, pmu_ctx_entry) { + cpc = this_cpu_ptr(pmu_ctx->pmu->cpu_pmu_context); + + if (cpc->sched_cb_usage && pmu_ctx->pmu->sched_task) + pmu_ctx->pmu->sched_task(pmu_ctx, sched_in); + } +} + +static void +perf_event_context_sched_out(struct task_struct *task, struct task_struct *next) +{ + struct perf_event_context *ctx = task->perf_event_ctxp; struct perf_event_context *next_ctx; struct perf_event_context *parent, *next_parent; - struct perf_cpu_context *cpuctx; int do_switch = 1; - struct pmu *pmu; if (likely(!ctx)) return; - pmu = ctx->pmu; - cpuctx = __get_cpu_context(ctx); - if (!cpuctx->task_ctx) - return; - rcu_read_lock(); - next_ctx = next->perf_event_ctxp[ctxn]; + next_ctx = rcu_dereference(next->perf_event_ctxp); if (!next_ctx) goto unlock; @@ -3453,7 +3477,7 @@ static void perf_event_context_sched_out(struct task_struct *task, int ctxn, raw_spin_lock_nested(&next_ctx->lock, SINGLE_DEPTH_NESTING); if (context_equiv(ctx, next_ctx)) { - perf_pmu_disable(pmu); + perf_ctx_disable(ctx); /* PMIs are disabled; ctx->nr_pending is stable. */ if (local_read(&ctx->nr_pending) || @@ -3470,21 +3494,10 @@ static void perf_event_context_sched_out(struct task_struct *task, int ctxn, WRITE_ONCE(ctx->task, next); WRITE_ONCE(next_ctx->task, task); - if (cpuctx->sched_cb_usage && pmu->sched_task) - pmu->sched_task(ctx, false); - - /* - * PMU specific parts of task perf context can require - * additional synchronization. As an example of such - * synchronization see implementation details of Intel - * LBR call stack data profiling; - */ - if (pmu->swap_task_ctx) - pmu->swap_task_ctx(ctx, next_ctx); - else - swap(ctx->task_ctx_data, next_ctx->task_ctx_data); + perf_ctx_sched_task_cb(ctx, false); + perf_event_swap_task_ctx_data(ctx, next_ctx); - perf_pmu_enable(pmu); + perf_ctx_enable(ctx); /* * RCU_INIT_POINTER here is safe because we've not @@ -3493,8 +3506,8 @@ static void perf_event_context_sched_out(struct task_struct *task, int ctxn, * since those values are always verified under * ctx->lock which we're now holding. */ - RCU_INIT_POINTER(task->perf_event_ctxp[ctxn], next_ctx); - RCU_INIT_POINTER(next->perf_event_ctxp[ctxn], ctx); + RCU_INIT_POINTER(task->perf_event_ctxp, next_ctx); + RCU_INIT_POINTER(next->perf_event_ctxp, ctx); do_switch = 0; @@ -3508,38 +3521,40 @@ unlock: if (do_switch) { raw_spin_lock(&ctx->lock); - perf_pmu_disable(pmu); + perf_ctx_disable(ctx); inside_switch: - if (cpuctx->sched_cb_usage && pmu->sched_task) - pmu->sched_task(ctx, false); - task_ctx_sched_out(cpuctx, ctx, EVENT_ALL); + perf_ctx_sched_task_cb(ctx, false); + task_ctx_sched_out(ctx, EVENT_ALL); - perf_pmu_enable(pmu); + perf_ctx_enable(ctx); raw_spin_unlock(&ctx->lock); } } static DEFINE_PER_CPU(struct list_head, sched_cb_list); +static DEFINE_PER_CPU(int, perf_sched_cb_usages); void perf_sched_cb_dec(struct pmu *pmu) { - struct perf_cpu_context *cpuctx = this_cpu_ptr(pmu->pmu_cpu_context); + struct perf_cpu_pmu_context *cpc = this_cpu_ptr(pmu->cpu_pmu_context); this_cpu_dec(perf_sched_cb_usages); + barrier(); - if (!--cpuctx->sched_cb_usage) - list_del(&cpuctx->sched_cb_entry); + if (!--cpc->sched_cb_usage) + list_del(&cpc->sched_cb_entry); } void perf_sched_cb_inc(struct pmu *pmu) { - struct perf_cpu_context *cpuctx = this_cpu_ptr(pmu->pmu_cpu_context); + struct perf_cpu_pmu_context *cpc = this_cpu_ptr(pmu->cpu_pmu_context); - if (!cpuctx->sched_cb_usage++) - list_add(&cpuctx->sched_cb_entry, this_cpu_ptr(&sched_cb_list)); + if (!cpc->sched_cb_usage++) + list_add(&cpc->sched_cb_entry, this_cpu_ptr(&sched_cb_list)); + barrier(); this_cpu_inc(perf_sched_cb_usages); } @@ -3551,19 +3566,21 @@ void perf_sched_cb_inc(struct pmu *pmu) * PEBS requires this to provide PID/TID information. This requires we flush * all queued PEBS records before we context switch to a new task. */ -static void __perf_pmu_sched_task(struct perf_cpu_context *cpuctx, bool sched_in) +static void __perf_pmu_sched_task(struct perf_cpu_pmu_context *cpc, bool sched_in) { + struct perf_cpu_context *cpuctx = this_cpu_ptr(&perf_cpu_context); struct pmu *pmu; - pmu = cpuctx->ctx.pmu; /* software PMUs will not have sched_task */ + pmu = cpc->epc.pmu; + /* software PMUs will not have sched_task */ if (WARN_ON_ONCE(!pmu->sched_task)) return; perf_ctx_lock(cpuctx, cpuctx->task_ctx); perf_pmu_disable(pmu); - pmu->sched_task(cpuctx->task_ctx, sched_in); + pmu->sched_task(cpc->task_epc, sched_in); perf_pmu_enable(pmu); perf_ctx_unlock(cpuctx, cpuctx->task_ctx); @@ -3573,26 +3590,20 @@ static void perf_pmu_sched_task(struct task_struct *prev, struct task_struct *next, bool sched_in) { - struct perf_cpu_context *cpuctx; + struct perf_cpu_context *cpuctx = this_cpu_ptr(&perf_cpu_context); + struct perf_cpu_pmu_context *cpc; - if (prev == next) + /* cpuctx->task_ctx will be handled in perf_event_context_sched_in/out */ + if (prev == next || cpuctx->task_ctx) return; - list_for_each_entry(cpuctx, this_cpu_ptr(&sched_cb_list), sched_cb_entry) { - /* will be handled in perf_event_context_sched_in/out */ - if (cpuctx->task_ctx) - continue; - - __perf_pmu_sched_task(cpuctx, sched_in); - } + list_for_each_entry(cpc, this_cpu_ptr(&sched_cb_list), sched_cb_entry) + __perf_pmu_sched_task(cpc, sched_in); } static void perf_event_switch(struct task_struct *task, struct task_struct *next_prev, bool sched_in); -#define for_each_task_context_nr(ctxn) \ - for ((ctxn) = 0; (ctxn) < perf_nr_task_contexts; (ctxn)++) - /* * Called from scheduler to remove the events of the current task, * with interrupts disabled. @@ -3607,16 +3618,13 @@ static void perf_event_switch(struct task_struct *task, void __perf_event_task_sched_out(struct task_struct *task, struct task_struct *next) { - int ctxn; - if (__this_cpu_read(perf_sched_cb_usages)) perf_pmu_sched_task(task, next, false); if (atomic_read(&nr_switch_events)) perf_event_switch(task, next, false); - for_each_task_context_nr(ctxn) - perf_event_context_sched_out(task, ctxn, next); + perf_event_context_sched_out(task, next); /* * if cgroup events exist on this CPU, then we need @@ -3627,15 +3635,6 @@ void __perf_event_task_sched_out(struct task_struct *task, perf_cgroup_switch(next); } -/* - * Called with IRQs disabled - */ -static void cpu_ctx_sched_out(struct perf_cpu_context *cpuctx, - enum event_type_t event_type) -{ - ctx_sched_out(&cpuctx->ctx, cpuctx, event_type); -} - static bool perf_less_group_idx(const void *l, const void *r) { const struct perf_event *le = *(const struct perf_event **)l; @@ -3667,21 +3666,39 @@ static void __heap_add(struct min_heap *heap, struct perf_event *event) } } -static noinline int visit_groups_merge(struct perf_cpu_context *cpuctx, +static void __link_epc(struct perf_event_pmu_context *pmu_ctx) +{ + struct perf_cpu_pmu_context *cpc; + + if (!pmu_ctx->ctx->task) + return; + + cpc = this_cpu_ptr(pmu_ctx->pmu->cpu_pmu_context); + WARN_ON_ONCE(cpc->task_epc && cpc->task_epc != pmu_ctx); + cpc->task_epc = pmu_ctx; +} + +static noinline int visit_groups_merge(struct perf_event_context *ctx, struct perf_event_groups *groups, int cpu, + struct pmu *pmu, int (*func)(struct perf_event *, void *), void *data) { #ifdef CONFIG_CGROUP_PERF struct cgroup_subsys_state *css = NULL; #endif + struct perf_cpu_context *cpuctx = NULL; /* Space for per CPU and/or any CPU event iterators. */ struct perf_event *itrs[2]; struct min_heap event_heap; struct perf_event **evt; int ret; - if (cpuctx) { + if (pmu->filter && pmu->filter(pmu, cpu)) + return 0; + + if (!ctx->task) { + cpuctx = this_cpu_ptr(&perf_cpu_context); event_heap = (struct min_heap){ .data = cpuctx->heap, .nr = 0, @@ -3701,17 +3718,22 @@ static noinline int visit_groups_merge(struct perf_cpu_context *cpuctx, .size = ARRAY_SIZE(itrs), }; /* Events not within a CPU context may be on any CPU. */ - __heap_add(&event_heap, perf_event_groups_first(groups, -1, NULL)); + __heap_add(&event_heap, perf_event_groups_first(groups, -1, pmu, NULL)); } evt = event_heap.data; - __heap_add(&event_heap, perf_event_groups_first(groups, cpu, NULL)); + __heap_add(&event_heap, perf_event_groups_first(groups, cpu, pmu, NULL)); #ifdef CONFIG_CGROUP_PERF for (; css; css = css->parent) - __heap_add(&event_heap, perf_event_groups_first(groups, cpu, css->cgroup)); + __heap_add(&event_heap, perf_event_groups_first(groups, cpu, pmu, css->cgroup)); #endif + if (event_heap.nr) { + __link_epc((*evt)->pmu_ctx); + perf_assert_pmu_disabled((*evt)->pmu_ctx->pmu); + } + min_heapify_all(&event_heap, &perf_min_heap); while (event_heap.nr) { @@ -3719,7 +3741,7 @@ static noinline int visit_groups_merge(struct perf_cpu_context *cpuctx, if (ret) return ret; - *evt = perf_event_groups_next(*evt); + *evt = perf_event_groups_next(*evt, pmu); if (*evt) min_heapify(&event_heap, 0, &perf_min_heap); else @@ -3761,7 +3783,6 @@ static inline void group_update_userpage(struct perf_event *group_event) static int merge_sched_in(struct perf_event *event, void *data) { struct perf_event_context *ctx = event->ctx; - struct perf_cpu_context *cpuctx = __get_cpu_context(ctx); int *can_add_hw = data; if (event->state <= PERF_EVENT_STATE_OFF) @@ -3770,8 +3791,8 @@ static int merge_sched_in(struct perf_event *event, void *data) if (!event_filter_match(event)) return 0; - if (group_can_go_on(event, cpuctx, *can_add_hw)) { - if (!group_sched_in(event, cpuctx, ctx)) + if (group_can_go_on(event, *can_add_hw)) { + if (!group_sched_in(event, ctx)) list_add_tail(&event->active_list, get_event_list(event)); } @@ -3781,8 +3802,11 @@ static int merge_sched_in(struct perf_event *event, void *data) perf_cgroup_event_disable(event, ctx); perf_event_set_state(event, PERF_EVENT_STATE_ERROR); } else { - ctx->rotate_necessary = 1; - perf_mux_hrtimer_restart(cpuctx); + struct perf_cpu_pmu_context *cpc; + + event->pmu_ctx->rotate_necessary = 1; + cpc = this_cpu_ptr(event->pmu_ctx->pmu->cpu_pmu_context); + perf_mux_hrtimer_restart(cpc); group_update_userpage(event); } } @@ -3790,39 +3814,53 @@ static int merge_sched_in(struct perf_event *event, void *data) return 0; } -static void -ctx_pinned_sched_in(struct perf_event_context *ctx, - struct perf_cpu_context *cpuctx) +static void ctx_pinned_sched_in(struct perf_event_context *ctx, struct pmu *pmu) { + struct perf_event_pmu_context *pmu_ctx; int can_add_hw = 1; - if (ctx != &cpuctx->ctx) - cpuctx = NULL; - - visit_groups_merge(cpuctx, &ctx->pinned_groups, - smp_processor_id(), - merge_sched_in, &can_add_hw); + if (pmu) { + visit_groups_merge(ctx, &ctx->pinned_groups, + smp_processor_id(), pmu, + merge_sched_in, &can_add_hw); + } else { + list_for_each_entry(pmu_ctx, &ctx->pmu_ctx_list, pmu_ctx_entry) { + can_add_hw = 1; + visit_groups_merge(ctx, &ctx->pinned_groups, + smp_processor_id(), pmu_ctx->pmu, + merge_sched_in, &can_add_hw); + } + } } -static void -ctx_flexible_sched_in(struct perf_event_context *ctx, - struct perf_cpu_context *cpuctx) +static void ctx_flexible_sched_in(struct perf_event_context *ctx, struct pmu *pmu) { + struct perf_event_pmu_context *pmu_ctx; int can_add_hw = 1; - if (ctx != &cpuctx->ctx) - cpuctx = NULL; + if (pmu) { + visit_groups_merge(ctx, &ctx->flexible_groups, + smp_processor_id(), pmu, + merge_sched_in, &can_add_hw); + } else { + list_for_each_entry(pmu_ctx, &ctx->pmu_ctx_list, pmu_ctx_entry) { + can_add_hw = 1; + visit_groups_merge(ctx, &ctx->flexible_groups, + smp_processor_id(), pmu_ctx->pmu, + merge_sched_in, &can_add_hw); + } + } +} - visit_groups_merge(cpuctx, &ctx->flexible_groups, - smp_processor_id(), - merge_sched_in, &can_add_hw); +static void __pmu_ctx_sched_in(struct perf_event_context *ctx, struct pmu *pmu) +{ + ctx_flexible_sched_in(ctx, pmu); } static void -ctx_sched_in(struct perf_event_context *ctx, - struct perf_cpu_context *cpuctx, - enum event_type_t event_type) +ctx_sched_in(struct perf_event_context *ctx, enum event_type_t event_type) { + struct perf_cpu_context *cpuctx = this_cpu_ptr(&perf_cpu_context); int is_active = ctx->is_active; lockdep_assert_held(&ctx->lock); @@ -3856,39 +3894,32 @@ ctx_sched_in(struct perf_event_context *ctx, * in order to give them the best chance of going on. */ if (is_active & EVENT_PINNED) - ctx_pinned_sched_in(ctx, cpuctx); + ctx_pinned_sched_in(ctx, NULL); /* Then walk through the lower prio flexible groups */ if (is_active & EVENT_FLEXIBLE) - ctx_flexible_sched_in(ctx, cpuctx); + ctx_flexible_sched_in(ctx, NULL); } -static void cpu_ctx_sched_in(struct perf_cpu_context *cpuctx, - enum event_type_t event_type) +static void perf_event_context_sched_in(struct task_struct *task) { - struct perf_event_context *ctx = &cpuctx->ctx; - - ctx_sched_in(ctx, cpuctx, event_type); -} + struct perf_cpu_context *cpuctx = this_cpu_ptr(&perf_cpu_context); + struct perf_event_context *ctx; -static void perf_event_context_sched_in(struct perf_event_context *ctx, - struct task_struct *task) -{ - struct perf_cpu_context *cpuctx; - struct pmu *pmu; + rcu_read_lock(); + ctx = rcu_dereference(task->perf_event_ctxp); + if (!ctx) + goto rcu_unlock; - cpuctx = __get_cpu_context(ctx); + if (cpuctx->task_ctx == ctx) { + perf_ctx_lock(cpuctx, ctx); + perf_ctx_disable(ctx); - /* - * HACK: for HETEROGENEOUS the task context might have switched to a - * different PMU, force (re)set the context, - */ - pmu = ctx->pmu = cpuctx->ctx.pmu; + perf_ctx_sched_task_cb(ctx, true); - if (cpuctx->task_ctx == ctx) { - if (cpuctx->sched_cb_usage) - __perf_pmu_sched_task(cpuctx, true); - return; + perf_ctx_enable(ctx); + perf_ctx_unlock(cpuctx, ctx); + goto rcu_unlock; } perf_ctx_lock(cpuctx, ctx); @@ -3899,7 +3930,7 @@ static void perf_event_context_sched_in(struct perf_event_context *ctx, if (!ctx->nr_events) goto unlock; - perf_pmu_disable(pmu); + perf_ctx_disable(ctx); /* * We want to keep the following priority order: * cpu pinned (that don't need to move), task pinned, @@ -3908,17 +3939,24 @@ static void perf_event_context_sched_in(struct perf_event_context *ctx, * However, if task's ctx is not carrying any pinned * events, no need to flip the cpuctx's events around. */ - if (!RB_EMPTY_ROOT(&ctx->pinned_groups.tree)) - cpu_ctx_sched_out(cpuctx, EVENT_FLEXIBLE); + if (!RB_EMPTY_ROOT(&ctx->pinned_groups.tree)) { + perf_ctx_disable(&cpuctx->ctx); + ctx_sched_out(&cpuctx->ctx, EVENT_FLEXIBLE); + } + perf_event_sched_in(cpuctx, ctx); - if (cpuctx->sched_cb_usage && pmu->sched_task) - pmu->sched_task(cpuctx->task_ctx, true); + perf_ctx_sched_task_cb(cpuctx->task_ctx, true); - perf_pmu_enable(pmu); + if (!RB_EMPTY_ROOT(&ctx->pinned_groups.tree)) + perf_ctx_enable(&cpuctx->ctx); + + perf_ctx_enable(ctx); unlock: perf_ctx_unlock(cpuctx, ctx); +rcu_unlock: + rcu_read_unlock(); } /* @@ -3935,16 +3973,7 @@ unlock: void __perf_event_task_sched_in(struct task_struct *prev, struct task_struct *task) { - struct perf_event_context *ctx; - int ctxn; - - for_each_task_context_nr(ctxn) { - ctx = task->perf_event_ctxp[ctxn]; - if (likely(!ctx)) - continue; - - perf_event_context_sched_in(ctx, task); - } + perf_event_context_sched_in(task); if (atomic_read(&nr_switch_events)) perf_event_switch(task, prev, true); @@ -4063,8 +4092,8 @@ static void perf_adjust_period(struct perf_event *event, u64 nsec, u64 count, bo * events. At the same time, make sure, having freq events does not change * the rate of unthrottling as that would introduce bias. */ -static void perf_adjust_freq_unthr_context(struct perf_event_context *ctx, - int needs_unthr) +static void +perf_adjust_freq_unthr_context(struct perf_event_context *ctx, bool unthrottle) { struct perf_event *event; struct hw_perf_event *hwc; @@ -4076,16 +4105,16 @@ static void perf_adjust_freq_unthr_context(struct perf_event_context *ctx, * - context have events in frequency mode (needs freq adjust) * - there are events to unthrottle on this cpu */ - if (!(ctx->nr_freq || needs_unthr)) + if (!(ctx->nr_freq || unthrottle)) return; raw_spin_lock(&ctx->lock); - perf_pmu_disable(ctx->pmu); list_for_each_entry_rcu(event, &ctx->event_list, event_entry) { if (event->state != PERF_EVENT_STATE_ACTIVE) continue; + // XXX use visit thingy to avoid the -1,cpu match if (!event_filter_match(event)) continue; @@ -4126,7 +4155,6 @@ static void perf_adjust_freq_unthr_context(struct perf_event_context *ctx, perf_pmu_enable(event->pmu); } - perf_pmu_enable(ctx->pmu); raw_spin_unlock(&ctx->lock); } @@ -4148,72 +4176,109 @@ static void rotate_ctx(struct perf_event_context *ctx, struct perf_event *event) /* pick an event from the flexible_groups to rotate */ static inline struct perf_event * -ctx_event_to_rotate(struct perf_event_context *ctx) +ctx_event_to_rotate(struct perf_event_pmu_context *pmu_ctx) { struct perf_event *event; + struct rb_node *node; + struct rb_root *tree; + struct __group_key key = { + .pmu = pmu_ctx->pmu, + }; /* pick the first active flexible event */ - event = list_first_entry_or_null(&ctx->flexible_active, + event = list_first_entry_or_null(&pmu_ctx->flexible_active, struct perf_event, active_list); + if (event) + goto out; /* if no active flexible event, pick the first event */ - if (!event) { - event = rb_entry_safe(rb_first(&ctx->flexible_groups.tree), - typeof(*event), group_node); + tree = &pmu_ctx->ctx->flexible_groups.tree; + + if (!pmu_ctx->ctx->task) { + key.cpu = smp_processor_id(); + + node = rb_find_first(&key, tree, __group_cmp_ignore_cgroup); + if (node) + event = __node_2_pe(node); + goto out; } + key.cpu = -1; + node = rb_find_first(&key, tree, __group_cmp_ignore_cgroup); + if (node) { + event = __node_2_pe(node); + goto out; + } + + key.cpu = smp_processor_id(); + node = rb_find_first(&key, tree, __group_cmp_ignore_cgroup); + if (node) + event = __node_2_pe(node); + +out: /* * Unconditionally clear rotate_necessary; if ctx_flexible_sched_in() * finds there are unschedulable events, it will set it again. */ - ctx->rotate_necessary = 0; + pmu_ctx->rotate_necessary = 0; return event; } -static bool perf_rotate_context(struct perf_cpu_context *cpuctx) +static bool perf_rotate_context(struct perf_cpu_pmu_context *cpc) { + struct perf_cpu_context *cpuctx = this_cpu_ptr(&perf_cpu_context); + struct perf_event_pmu_context *cpu_epc, *task_epc = NULL; struct perf_event *cpu_event = NULL, *task_event = NULL; - struct perf_event_context *task_ctx = NULL; int cpu_rotate, task_rotate; + struct pmu *pmu; /* * Since we run this from IRQ context, nobody can install new * events, thus the event count values are stable. */ - cpu_rotate = cpuctx->ctx.rotate_necessary; - task_ctx = cpuctx->task_ctx; - task_rotate = task_ctx ? task_ctx->rotate_necessary : 0; + cpu_epc = &cpc->epc; + pmu = cpu_epc->pmu; + task_epc = cpc->task_epc; + + cpu_rotate = cpu_epc->rotate_necessary; + task_rotate = task_epc ? task_epc->rotate_necessary : 0; if (!(cpu_rotate || task_rotate)) return false; perf_ctx_lock(cpuctx, cpuctx->task_ctx); - perf_pmu_disable(cpuctx->ctx.pmu); + perf_pmu_disable(pmu); if (task_rotate) - task_event = ctx_event_to_rotate(task_ctx); + task_event = ctx_event_to_rotate(task_epc); if (cpu_rotate) - cpu_event = ctx_event_to_rotate(&cpuctx->ctx); + cpu_event = ctx_event_to_rotate(cpu_epc); /* * As per the order given at ctx_resched() first 'pop' task flexible * and then, if needed CPU flexible. */ - if (task_event || (task_ctx && cpu_event)) - ctx_sched_out(task_ctx, cpuctx, EVENT_FLEXIBLE); - if (cpu_event) - cpu_ctx_sched_out(cpuctx, EVENT_FLEXIBLE); + if (task_event || (task_epc && cpu_event)) { + update_context_time(task_epc->ctx); + __pmu_ctx_sched_out(task_epc, EVENT_FLEXIBLE); + } - if (task_event) - rotate_ctx(task_ctx, task_event); - if (cpu_event) + if (cpu_event) { + update_context_time(&cpuctx->ctx); + __pmu_ctx_sched_out(cpu_epc, EVENT_FLEXIBLE); rotate_ctx(&cpuctx->ctx, cpu_event); + __pmu_ctx_sched_in(&cpuctx->ctx, pmu); + } - perf_event_sched_in(cpuctx, task_ctx); + if (task_event) + rotate_ctx(task_epc->ctx, task_event); + + if (task_event || (task_epc && cpu_event)) + __pmu_ctx_sched_in(task_epc->ctx, pmu); - perf_pmu_enable(cpuctx->ctx.pmu); + perf_pmu_enable(pmu); perf_ctx_unlock(cpuctx, cpuctx->task_ctx); return true; @@ -4221,8 +4286,8 @@ static bool perf_rotate_context(struct perf_cpu_context *cpuctx) void perf_event_task_tick(void) { - struct list_head *head = this_cpu_ptr(&active_ctx_list); - struct perf_event_context *ctx, *tmp; + struct perf_cpu_context *cpuctx = this_cpu_ptr(&perf_cpu_context); + struct perf_event_context *ctx; int throttled; lockdep_assert_irqs_disabled(); @@ -4231,8 +4296,13 @@ void perf_event_task_tick(void) throttled = __this_cpu_xchg(perf_throttled_count, 0); tick_dep_clear_cpu(smp_processor_id(), TICK_DEP_BIT_PERF_EVENTS); - list_for_each_entry_safe(ctx, tmp, head, active_ctx_list) - perf_adjust_freq_unthr_context(ctx, throttled); + perf_adjust_freq_unthr_context(&cpuctx->ctx, !!throttled); + + rcu_read_lock(); + ctx = rcu_dereference(current->perf_event_ctxp); + if (ctx) + perf_adjust_freq_unthr_context(ctx, !!throttled); + rcu_read_unlock(); } static int event_enable_on_exec(struct perf_event *event, @@ -4254,9 +4324,9 @@ static int event_enable_on_exec(struct perf_event *event, * Enable all of a task's events that have been marked enable-on-exec. * This expects task == current. */ -static void perf_event_enable_on_exec(int ctxn) +static void perf_event_enable_on_exec(struct perf_event_context *ctx) { - struct perf_event_context *ctx, *clone_ctx = NULL; + struct perf_event_context *clone_ctx = NULL; enum event_type_t event_type = 0; struct perf_cpu_context *cpuctx; struct perf_event *event; @@ -4264,13 +4334,16 @@ static void perf_event_enable_on_exec(int ctxn) int enabled = 0; local_irq_save(flags); - ctx = current->perf_event_ctxp[ctxn]; - if (!ctx || !ctx->nr_events) + if (WARN_ON_ONCE(current->perf_event_ctxp != ctx)) + goto out; + + if (!ctx->nr_events) goto out; - cpuctx = __get_cpu_context(ctx); + cpuctx = this_cpu_ptr(&perf_cpu_context); perf_ctx_lock(cpuctx, ctx); - ctx_sched_out(ctx, cpuctx, EVENT_TIME); + ctx_sched_out(ctx, EVENT_TIME); + list_for_each_entry(event, &ctx->event_list, event_entry) { enabled |= event_enable_on_exec(event, ctx); event_type |= get_event_type(event); @@ -4283,7 +4356,7 @@ static void perf_event_enable_on_exec(int ctxn) clone_ctx = unclone_ctx(ctx); ctx_resched(cpuctx, ctx, event_type); } else { - ctx_sched_in(ctx, cpuctx, EVENT_TIME); + ctx_sched_in(ctx, EVENT_TIME); } perf_ctx_unlock(cpuctx, ctx); @@ -4302,17 +4375,13 @@ static void perf_event_exit_event(struct perf_event *event, * Removes all events from the current task that have been marked * remove-on-exec, and feeds their values back to parent events. */ -static void perf_event_remove_on_exec(int ctxn) +static void perf_event_remove_on_exec(struct perf_event_context *ctx) { - struct perf_event_context *ctx, *clone_ctx = NULL; + struct perf_event_context *clone_ctx = NULL; struct perf_event *event, *next; unsigned long flags; bool modified = false; - ctx = perf_pin_task_context(current, ctxn); - if (!ctx) - return; - mutex_lock(&ctx->mutex); if (WARN_ON_ONCE(ctx->task != current)) @@ -4333,13 +4402,11 @@ static void perf_event_remove_on_exec(int ctxn) raw_spin_lock_irqsave(&ctx->lock, flags); if (modified) clone_ctx = unclone_ctx(ctx); - --ctx->pin_count; raw_spin_unlock_irqrestore(&ctx->lock, flags); unlock: mutex_unlock(&ctx->mutex); - put_ctx(ctx); if (clone_ctx) put_ctx(clone_ctx); } @@ -4375,7 +4442,7 @@ static void __perf_event_read(void *info) struct perf_read_data *data = info; struct perf_event *sub, *event = data->event; struct perf_event_context *ctx = event->ctx; - struct perf_cpu_context *cpuctx = __get_cpu_context(ctx); + struct perf_cpu_context *cpuctx = this_cpu_ptr(&perf_cpu_context); struct pmu *pmu = event->pmu; /* @@ -4601,17 +4668,25 @@ static void __perf_event_init_context(struct perf_event_context *ctx) { raw_spin_lock_init(&ctx->lock); mutex_init(&ctx->mutex); - INIT_LIST_HEAD(&ctx->active_ctx_list); + INIT_LIST_HEAD(&ctx->pmu_ctx_list); perf_event_groups_init(&ctx->pinned_groups); perf_event_groups_init(&ctx->flexible_groups); INIT_LIST_HEAD(&ctx->event_list); - INIT_LIST_HEAD(&ctx->pinned_active); - INIT_LIST_HEAD(&ctx->flexible_active); refcount_set(&ctx->refcount, 1); } +static void +__perf_init_event_pmu_context(struct perf_event_pmu_context *epc, struct pmu *pmu) +{ + epc->pmu = pmu; + INIT_LIST_HEAD(&epc->pmu_ctx_entry); + INIT_LIST_HEAD(&epc->pinned_active); + INIT_LIST_HEAD(&epc->flexible_active); + atomic_set(&epc->refcount, 1); +} + static struct perf_event_context * -alloc_perf_context(struct pmu *pmu, struct task_struct *task) +alloc_perf_context(struct task_struct *task) { struct perf_event_context *ctx; @@ -4622,7 +4697,6 @@ alloc_perf_context(struct pmu *pmu, struct task_struct *task) __perf_event_init_context(ctx); if (task) ctx->task = get_task_struct(task); - ctx->pmu = pmu; return ctx; } @@ -4651,15 +4725,12 @@ find_lively_task_by_vpid(pid_t vpid) * Returns a matching context with refcount and pincount. */ static struct perf_event_context * -find_get_context(struct pmu *pmu, struct task_struct *task, - struct perf_event *event) +find_get_context(struct task_struct *task, struct perf_event *event) { struct perf_event_context *ctx, *clone_ctx = NULL; struct perf_cpu_context *cpuctx; - void *task_ctx_data = NULL; unsigned long flags; - int ctxn, err; - int cpu = event->cpu; + int err; if (!task) { /* Must be root to operate on a CPU event: */ @@ -4667,7 +4738,7 @@ find_get_context(struct pmu *pmu, struct task_struct *task, if (err) return ERR_PTR(err); - cpuctx = per_cpu_ptr(pmu->pmu_cpu_context, cpu); + cpuctx = per_cpu_ptr(&perf_cpu_context, event->cpu); ctx = &cpuctx->ctx; get_ctx(ctx); raw_spin_lock_irqsave(&ctx->lock, flags); @@ -4678,43 +4749,22 @@ find_get_context(struct pmu *pmu, struct task_struct *task, } err = -EINVAL; - ctxn = pmu->task_ctx_nr; - if (ctxn < 0) - goto errout; - - if (event->attach_state & PERF_ATTACH_TASK_DATA) { - task_ctx_data = alloc_task_ctx_data(pmu); - if (!task_ctx_data) { - err = -ENOMEM; - goto errout; - } - } - retry: - ctx = perf_lock_task_context(task, ctxn, &flags); + ctx = perf_lock_task_context(task, &flags); if (ctx) { clone_ctx = unclone_ctx(ctx); ++ctx->pin_count; - if (task_ctx_data && !ctx->task_ctx_data) { - ctx->task_ctx_data = task_ctx_data; - task_ctx_data = NULL; - } raw_spin_unlock_irqrestore(&ctx->lock, flags); if (clone_ctx) put_ctx(clone_ctx); } else { - ctx = alloc_perf_context(pmu, task); + ctx = alloc_perf_context(task); err = -ENOMEM; if (!ctx) goto errout; - if (task_ctx_data) { - ctx->task_ctx_data = task_ctx_data; - task_ctx_data = NULL; - } - err = 0; mutex_lock(&task->perf_event_mutex); /* @@ -4723,12 +4773,12 @@ retry: */ if (task->flags & PF_EXITING) err = -ESRCH; - else if (task->perf_event_ctxp[ctxn]) + else if (task->perf_event_ctxp) err = -EAGAIN; else { get_ctx(ctx); ++ctx->pin_count; - rcu_assign_pointer(task->perf_event_ctxp[ctxn], ctx); + rcu_assign_pointer(task->perf_event_ctxp, ctx); } mutex_unlock(&task->perf_event_mutex); @@ -4741,21 +4791,146 @@ retry: } } - free_task_ctx_data(pmu, task_ctx_data); return ctx; errout: - free_task_ctx_data(pmu, task_ctx_data); return ERR_PTR(err); } +static struct perf_event_pmu_context * +find_get_pmu_context(struct pmu *pmu, struct perf_event_context *ctx, + struct perf_event *event) +{ + struct perf_event_pmu_context *new = NULL, *epc; + void *task_ctx_data = NULL; + + if (!ctx->task) { + struct perf_cpu_pmu_context *cpc; + + cpc = per_cpu_ptr(pmu->cpu_pmu_context, event->cpu); + epc = &cpc->epc; + + if (!epc->ctx) { + atomic_set(&epc->refcount, 1); + epc->embedded = 1; + raw_spin_lock_irq(&ctx->lock); + list_add(&epc->pmu_ctx_entry, &ctx->pmu_ctx_list); + epc->ctx = ctx; + raw_spin_unlock_irq(&ctx->lock); + } else { + WARN_ON_ONCE(epc->ctx != ctx); + atomic_inc(&epc->refcount); + } + + return epc; + } + + new = kzalloc(sizeof(*epc), GFP_KERNEL); + if (!new) + return ERR_PTR(-ENOMEM); + + if (event->attach_state & PERF_ATTACH_TASK_DATA) { + task_ctx_data = alloc_task_ctx_data(pmu); + if (!task_ctx_data) { + kfree(new); + return ERR_PTR(-ENOMEM); + } + } + + __perf_init_event_pmu_context(new, pmu); + + /* + * XXX + * + * lockdep_assert_held(&ctx->mutex); + * + * can't because perf_event_init_task() doesn't actually hold the + * child_ctx->mutex. + */ + + raw_spin_lock_irq(&ctx->lock); + list_for_each_entry(epc, &ctx->pmu_ctx_list, pmu_ctx_entry) { + if (epc->pmu == pmu) { + WARN_ON_ONCE(epc->ctx != ctx); + atomic_inc(&epc->refcount); + goto found_epc; + } + } + + epc = new; + new = NULL; + + list_add(&epc->pmu_ctx_entry, &ctx->pmu_ctx_list); + epc->ctx = ctx; + +found_epc: + if (task_ctx_data && !epc->task_ctx_data) { + epc->task_ctx_data = task_ctx_data; + task_ctx_data = NULL; + ctx->nr_task_data++; + } + raw_spin_unlock_irq(&ctx->lock); + + free_task_ctx_data(pmu, task_ctx_data); + kfree(new); + + return epc; +} + +static void get_pmu_ctx(struct perf_event_pmu_context *epc) +{ + WARN_ON_ONCE(!atomic_inc_not_zero(&epc->refcount)); +} + +static void free_epc_rcu(struct rcu_head *head) +{ + struct perf_event_pmu_context *epc = container_of(head, typeof(*epc), rcu_head); + + kfree(epc->task_ctx_data); + kfree(epc); +} + +static void put_pmu_ctx(struct perf_event_pmu_context *epc) +{ + unsigned long flags; + + if (!atomic_dec_and_test(&epc->refcount)) + return; + + if (epc->ctx) { + struct perf_event_context *ctx = epc->ctx; + + /* + * XXX + * + * lockdep_assert_held(&ctx->mutex); + * + * can't because of the call-site in _free_event()/put_event() + * which isn't always called under ctx->mutex. + */ + + WARN_ON_ONCE(list_empty(&epc->pmu_ctx_entry)); + raw_spin_lock_irqsave(&ctx->lock, flags); + list_del_init(&epc->pmu_ctx_entry); + epc->ctx = NULL; + raw_spin_unlock_irqrestore(&ctx->lock, flags); + } + + WARN_ON_ONCE(!list_empty(&epc->pinned_active)); + WARN_ON_ONCE(!list_empty(&epc->flexible_active)); + + if (epc->embedded) + return; + + call_rcu(&epc->rcu_head, free_epc_rcu); +} + static void perf_event_free_filter(struct perf_event *event); static void free_event_rcu(struct rcu_head *head) { - struct perf_event *event; + struct perf_event *event = container_of(head, typeof(*event), rcu_head); - event = container_of(head, struct perf_event, rcu_head); if (event->ns) put_pid_ns(event->ns); perf_event_free_filter(event); @@ -4893,7 +5068,7 @@ static void perf_sched_delayed(struct work_struct *work) * * 1) cpu-wide events in the presence of per-task events, * 2) per-task events in the presence of cpu-wide events, - * 3) two matching events on the same context. + * 3) two matching events on the same perf_event_context. * * The former two cases are handled in the allocation path (perf_event_alloc(), * _free_event()), the latter -- before the first perf_install_in_context(). @@ -5017,6 +5192,9 @@ static void _free_event(struct perf_event *event) if (event->hw.target) put_task_struct(event->hw.target); + if (event->pmu_ctx) + put_pmu_ctx(event->pmu_ctx); + /* * perf_event_free_task() relies on put_ctx() being 'last', in particular * all task references must be cleaned up. @@ -5117,8 +5295,8 @@ int perf_event_release_kernel(struct perf_event *event) LIST_HEAD(free_list); /* - * If we got here through err_file: fput(event_file); we will not have - * attached to a context yet. + * If we got here through err_alloc: free_event(event); we will not + * have attached to a context yet. */ if (!ctx) { WARN_ON_ONCE(event->attach_state & @@ -5550,7 +5728,7 @@ static void __perf_event_period(struct perf_event *event, active = (event->state == PERF_EVENT_STATE_ACTIVE); if (active) { - perf_pmu_disable(ctx->pmu); + perf_pmu_disable(event->pmu); /* * We could be throttled; unthrottle now to avoid the tick * trying to unthrottle while we already re-started the event. @@ -5566,7 +5744,7 @@ static void __perf_event_period(struct perf_event *event, if (active) { event->pmu->start(event, PERF_EF_RELOAD); - perf_pmu_enable(ctx->pmu); + perf_pmu_enable(event->pmu); } } @@ -7729,7 +7907,6 @@ perf_iterate_sb(perf_iterate_f output, void *data, struct perf_event_context *task_ctx) { struct perf_event_context *ctx; - int ctxn; rcu_read_lock(); preempt_disable(); @@ -7746,11 +7923,9 @@ perf_iterate_sb(perf_iterate_f output, void *data, perf_iterate_sb_cpu(output, data); - for_each_task_context_nr(ctxn) { - ctx = rcu_dereference(current->perf_event_ctxp[ctxn]); - if (ctx) - perf_iterate_ctx(ctx, output, data, false); - } + ctx = rcu_dereference(current->perf_event_ctxp); + if (ctx) + perf_iterate_ctx(ctx, output, data, false); done: preempt_enable(); rcu_read_unlock(); @@ -7792,20 +7967,17 @@ static void perf_event_addr_filters_exec(struct perf_event *event, void *data) void perf_event_exec(void) { struct perf_event_context *ctx; - int ctxn; - for_each_task_context_nr(ctxn) { - perf_event_enable_on_exec(ctxn); - perf_event_remove_on_exec(ctxn); + ctx = perf_pin_task_context(current); + if (!ctx) + return; - rcu_read_lock(); - ctx = rcu_dereference(current->perf_event_ctxp[ctxn]); - if (ctx) { - perf_iterate_ctx(ctx, perf_event_addr_filters_exec, - NULL, true); - } - rcu_read_unlock(); - } + perf_event_enable_on_exec(ctx); + perf_event_remove_on_exec(ctx); + perf_iterate_ctx(ctx, perf_event_addr_filters_exec, NULL, true); + + perf_unpin_context(ctx); + put_ctx(ctx); } struct remote_output { @@ -7845,8 +8017,7 @@ static void __perf_event_output_stop(struct perf_event *event, void *data) static int __perf_pmu_output_stop(void *info) { struct perf_event *event = info; - struct pmu *pmu = event->ctx->pmu; - struct perf_cpu_context *cpuctx = this_cpu_ptr(pmu->pmu_cpu_context); + struct perf_cpu_context *cpuctx = this_cpu_ptr(&perf_cpu_context); struct remote_output ro = { .rb = event->rb, }; @@ -8635,7 +8806,6 @@ static void __perf_addr_filters_adjust(struct perf_event *event, void *data) static void perf_addr_filters_adjust(struct vm_area_struct *vma) { struct perf_event_context *ctx; - int ctxn; /* * Data tracing isn't supported yet and as such there is no need @@ -8645,13 +8815,9 @@ static void perf_addr_filters_adjust(struct vm_area_struct *vma) return; rcu_read_lock(); - for_each_task_context_nr(ctxn) { - ctx = rcu_dereference(current->perf_event_ctxp[ctxn]); - if (!ctx) - continue; - + ctx = rcu_dereference(current->perf_event_ctxp); + if (ctx) perf_iterate_ctx(ctx, __perf_addr_filters_adjust, vma, true); - } rcu_read_unlock(); } @@ -9826,6 +9992,44 @@ static struct pmu perf_swevent = { #ifdef CONFIG_EVENT_TRACING +static void tp_perf_event_destroy(struct perf_event *event) +{ + perf_trace_destroy(event); +} + +static int perf_tp_event_init(struct perf_event *event) +{ + int err; + + if (event->attr.type != PERF_TYPE_TRACEPOINT) + return -ENOENT; + + /* + * no branch sampling for tracepoint events + */ + if (has_branch_stack(event)) + return -EOPNOTSUPP; + + err = perf_trace_init(event); + if (err) + return err; + + event->destroy = tp_perf_event_destroy; + + return 0; +} + +static struct pmu perf_tracepoint = { + .task_ctx_nr = perf_sw_context, + + .event_init = perf_tp_event_init, + .add = perf_trace_add, + .del = perf_trace_del, + .start = perf_swevent_start, + .stop = perf_swevent_stop, + .read = perf_swevent_read, +}; + static int perf_tp_filter_match(struct perf_event *event, struct perf_sample_data *data) { @@ -9875,6 +10079,44 @@ void perf_trace_run_bpf_submit(void *raw_data, int size, int rctx, } EXPORT_SYMBOL_GPL(perf_trace_run_bpf_submit); +static void __perf_tp_event_target_task(u64 count, void *record, + struct pt_regs *regs, + struct perf_sample_data *data, + struct perf_event *event) +{ + struct trace_entry *entry = record; + + if (event->attr.config != entry->type) + return; + /* Cannot deliver synchronous signal to other task. */ + if (event->attr.sigtrap) + return; + if (perf_tp_event_match(event, data, regs)) + perf_swevent_event(event, count, data, regs); +} + +static void perf_tp_event_target_task(u64 count, void *record, + struct pt_regs *regs, + struct perf_sample_data *data, + struct perf_event_context *ctx) +{ + unsigned int cpu = smp_processor_id(); + struct pmu *pmu = &perf_tracepoint; + struct perf_event *event, *sibling; + + perf_event_groups_for_cpu_pmu(event, &ctx->pinned_groups, cpu, pmu) { + __perf_tp_event_target_task(count, record, regs, data, event); + for_each_sibling_event(sibling, event) + __perf_tp_event_target_task(count, record, regs, data, sibling); + } + + perf_event_groups_for_cpu_pmu(event, &ctx->flexible_groups, cpu, pmu) { + __perf_tp_event_target_task(count, record, regs, data, event); + for_each_sibling_event(sibling, event) + __perf_tp_event_target_task(count, record, regs, data, sibling); + } +} + void perf_tp_event(u16 event_type, u64 count, void *record, int entry_size, struct pt_regs *regs, struct hlist_head *head, int rctx, struct task_struct *task) @@ -9906,26 +10148,15 @@ void perf_tp_event(u16 event_type, u64 count, void *record, int entry_size, */ if (task && task != current) { struct perf_event_context *ctx; - struct trace_entry *entry = record; rcu_read_lock(); - ctx = rcu_dereference(task->perf_event_ctxp[perf_sw_context]); + ctx = rcu_dereference(task->perf_event_ctxp); if (!ctx) goto unlock; - list_for_each_entry_rcu(event, &ctx->event_list, event_entry) { - if (event->cpu != smp_processor_id()) - continue; - if (event->attr.type != PERF_TYPE_TRACEPOINT) - continue; - if (event->attr.config != entry->type) - continue; - /* Cannot deliver synchronous signal to other task. */ - if (event->attr.sigtrap) - continue; - if (perf_tp_event_match(event, &data, regs)) - perf_swevent_event(event, count, &data, regs); - } + raw_spin_lock(&ctx->lock); + perf_tp_event_target_task(count, record, regs, &data, ctx); + raw_spin_unlock(&ctx->lock); unlock: rcu_read_unlock(); } @@ -9934,44 +10165,6 @@ unlock: } EXPORT_SYMBOL_GPL(perf_tp_event); -static void tp_perf_event_destroy(struct perf_event *event) -{ - perf_trace_destroy(event); -} - -static int perf_tp_event_init(struct perf_event *event) -{ - int err; - - if (event->attr.type != PERF_TYPE_TRACEPOINT) - return -ENOENT; - - /* - * no branch sampling for tracepoint events - */ - if (has_branch_stack(event)) - return -EOPNOTSUPP; - - err = perf_trace_init(event); - if (err) - return err; - - event->destroy = tp_perf_event_destroy; - - return 0; -} - -static struct pmu perf_tracepoint = { - .task_ctx_nr = perf_sw_context, - - .event_init = perf_tp_event_init, - .add = perf_trace_add, - .del = perf_trace_del, - .start = perf_swevent_start, - .stop = perf_swevent_stop, - .read = perf_swevent_read, -}; - #if defined(CONFIG_KPROBE_EVENTS) || defined(CONFIG_UPROBE_EVENTS) /* * Flags in config, used by dynamic PMU kprobe and uprobe @@ -11058,36 +11251,9 @@ static int perf_event_idx_default(struct perf_event *event) return 0; } -/* - * Ensures all contexts with the same task_ctx_nr have the same - * pmu_cpu_context too. - */ -static struct perf_cpu_context __percpu *find_pmu_context(int ctxn) -{ - struct pmu *pmu; - - if (ctxn < 0) - return NULL; - - list_for_each_entry(pmu, &pmus, entry) { - if (pmu->task_ctx_nr == ctxn) - return pmu->pmu_cpu_context; - } - - return NULL; -} - static void free_pmu_context(struct pmu *pmu) { - /* - * Static contexts such as perf_sw_context have a global lifetime - * and may be shared between different PMUs. Avoid freeing them - * when a single PMU is going away. - */ - if (pmu->task_ctx_nr > perf_invalid_context) - return; - - free_percpu(pmu->pmu_cpu_context); + free_percpu(pmu->cpu_pmu_context); } /* @@ -11151,12 +11317,11 @@ perf_event_mux_interval_ms_store(struct device *dev, /* update all cpuctx for this PMU */ cpus_read_lock(); for_each_online_cpu(cpu) { - struct perf_cpu_context *cpuctx; - cpuctx = per_cpu_ptr(pmu->pmu_cpu_context, cpu); - cpuctx->hrtimer_interval = ns_to_ktime(NSEC_PER_MSEC * timer); + struct perf_cpu_pmu_context *cpc; + cpc = per_cpu_ptr(pmu->cpu_pmu_context, cpu); + cpc->hrtimer_interval = ns_to_ktime(NSEC_PER_MSEC * timer); - cpu_function_call(cpu, - (remote_function_f)perf_mux_hrtimer_restart, cpuctx); + cpu_function_call(cpu, perf_mux_hrtimer_restart_ipi, cpc); } cpus_read_unlock(); mutex_unlock(&mux_interval_mutex); @@ -11193,13 +11358,15 @@ static int pmu_dev_alloc(struct pmu *pmu) pmu->dev->groups = pmu->attr_groups; device_initialize(pmu->dev); - ret = dev_set_name(pmu->dev, "%s", pmu->name); - if (ret) - goto free_dev; dev_set_drvdata(pmu->dev, pmu); pmu->dev->bus = &pmu_bus; pmu->dev->release = pmu_dev_release; + + ret = dev_set_name(pmu->dev, "%s", pmu->name); + if (ret) + goto free_dev; + ret = device_add(pmu->dev); if (ret) goto free_dev; @@ -11267,47 +11434,19 @@ int perf_pmu_register(struct pmu *pmu, const char *name, int type) } skip_type: - if (pmu->task_ctx_nr == perf_hw_context) { - static int hw_context_taken = 0; - - /* - * Other than systems with heterogeneous CPUs, it never makes - * sense for two PMUs to share perf_hw_context. PMUs which are - * uncore must use perf_invalid_context. - */ - if (WARN_ON_ONCE(hw_context_taken && - !(pmu->capabilities & PERF_PMU_CAP_HETEROGENEOUS_CPUS))) - pmu->task_ctx_nr = perf_invalid_context; - - hw_context_taken = 1; - } - - pmu->pmu_cpu_context = find_pmu_context(pmu->task_ctx_nr); - if (pmu->pmu_cpu_context) - goto got_cpu_context; - ret = -ENOMEM; - pmu->pmu_cpu_context = alloc_percpu(struct perf_cpu_context); - if (!pmu->pmu_cpu_context) + pmu->cpu_pmu_context = alloc_percpu(struct perf_cpu_pmu_context); + if (!pmu->cpu_pmu_context) goto free_dev; for_each_possible_cpu(cpu) { - struct perf_cpu_context *cpuctx; + struct perf_cpu_pmu_context *cpc; - cpuctx = per_cpu_ptr(pmu->pmu_cpu_context, cpu); - __perf_event_init_context(&cpuctx->ctx); - lockdep_set_class(&cpuctx->ctx.mutex, &cpuctx_mutex); - lockdep_set_class(&cpuctx->ctx.lock, &cpuctx_lock); - cpuctx->ctx.pmu = pmu; - cpuctx->online = cpumask_test_cpu(cpu, perf_online_mask); - - __perf_mux_hrtimer_init(cpuctx, cpu); - - cpuctx->heap_size = ARRAY_SIZE(cpuctx->heap_default); - cpuctx->heap = cpuctx->heap_default; + cpc = per_cpu_ptr(pmu->cpu_pmu_context, cpu); + __perf_init_event_pmu_context(&cpc->epc, pmu); + __perf_mux_hrtimer_init(cpc, cpu); } -got_cpu_context: if (!pmu->start_txn) { if (pmu->pmu_enable) { /* @@ -11786,10 +11925,11 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu, } /* - * Disallow uncore-cgroup events, they don't make sense as the cgroup will - * be different on other CPUs in the uncore mask. + * Disallow uncore-task events. Similarly, disallow uncore-cgroup + * events (they don't make sense as the cgroup will be different + * on other CPUs in the uncore mask). */ - if (pmu->task_ctx_nr == perf_invalid_context && cgroup_fd != -1) { + if (pmu->task_ctx_nr == perf_invalid_context && (task || cgroup_fd != -1)) { err = -EINVAL; goto err_pmu; } @@ -12136,37 +12276,6 @@ static int perf_event_set_clock(struct perf_event *event, clockid_t clk_id) return 0; } -/* - * Variation on perf_event_ctx_lock_nested(), except we take two context - * mutexes. - */ -static struct perf_event_context * -__perf_event_ctx_lock_double(struct perf_event *group_leader, - struct perf_event_context *ctx) -{ - struct perf_event_context *gctx; - -again: - rcu_read_lock(); - gctx = READ_ONCE(group_leader->ctx); - if (!refcount_inc_not_zero(&gctx->refcount)) { - rcu_read_unlock(); - goto again; - } - rcu_read_unlock(); - - mutex_lock_double(&gctx->mutex, &ctx->mutex); - - if (group_leader->ctx != gctx) { - mutex_unlock(&ctx->mutex); - mutex_unlock(&gctx->mutex); - put_ctx(gctx); - goto again; - } - - return gctx; -} - static bool perf_check_permission(struct perf_event_attr *attr, struct task_struct *task) { @@ -12212,9 +12321,10 @@ SYSCALL_DEFINE5(perf_event_open, pid_t, pid, int, cpu, int, group_fd, unsigned long, flags) { struct perf_event *group_leader = NULL, *output_event = NULL; + struct perf_event_pmu_context *pmu_ctx; struct perf_event *event, *sibling; struct perf_event_attr attr; - struct perf_event_context *ctx, *gctx; + struct perf_event_context *ctx; struct file *event_file = NULL; struct fd group = {NULL, 0}; struct task_struct *task = NULL; @@ -12344,42 +12454,53 @@ SYSCALL_DEFINE5(perf_event_open, if (pmu->task_ctx_nr == perf_sw_context) event->event_caps |= PERF_EV_CAP_SOFTWARE; - if (group_leader) { - if (is_software_event(event) && - !in_software_context(group_leader)) { - /* - * If the event is a sw event, but the group_leader - * is on hw context. - * - * Allow the addition of software events to hw - * groups, this is safe because software events - * never fail to schedule. - */ - pmu = group_leader->ctx->pmu; - } else if (!is_software_event(event) && - is_software_event(group_leader) && - (group_leader->group_caps & PERF_EV_CAP_SOFTWARE)) { - /* - * In case the group is a pure software group, and we - * try to add a hardware event, move the whole group to - * the hardware context. - */ - move_group = 1; - } + if (task) { + err = down_read_interruptible(&task->signal->exec_update_lock); + if (err) + goto err_alloc; + + /* + * We must hold exec_update_lock across this and any potential + * perf_install_in_context() call for this new event to + * serialize against exec() altering our credentials (and the + * perf_event_exit_task() that could imply). + */ + err = -EACCES; + if (!perf_check_permission(&attr, task)) + goto err_cred; } /* * Get the target context (task or percpu): */ - ctx = find_get_context(pmu, task, event); + ctx = find_get_context(task, event); if (IS_ERR(ctx)) { err = PTR_ERR(ctx); - goto err_alloc; + goto err_cred; + } + + mutex_lock(&ctx->mutex); + + if (ctx->task == TASK_TOMBSTONE) { + err = -ESRCH; + goto err_locked; + } + + if (!task) { + /* + * Check if the @cpu we're creating an event for is online. + * + * We use the perf_cpu_context::ctx::mutex to serialize against + * the hotplug notifiers. See perf_event_{init,exit}_cpu(). + */ + struct perf_cpu_context *cpuctx = per_cpu_ptr(&perf_cpu_context, event->cpu); + + if (!cpuctx->online) { + err = -ENODEV; + goto err_locked; + } } - /* - * Look up the group leader (we will attach this event to it): - */ if (group_leader) { err = -EINVAL; @@ -12388,11 +12509,11 @@ SYSCALL_DEFINE5(perf_event_open, * becoming part of another group-sibling): */ if (group_leader->group_leader != group_leader) - goto err_context; + goto err_locked; /* All events in a group should have the same clock */ if (group_leader->clock != event->clock) - goto err_context; + goto err_locked; /* * Make sure we're both events for the same CPU; @@ -12400,145 +12521,76 @@ SYSCALL_DEFINE5(perf_event_open, * you can never concurrently schedule them anyhow. */ if (group_leader->cpu != event->cpu) - goto err_context; - - /* - * Make sure we're both on the same task, or both - * per-CPU events. - */ - if (group_leader->ctx->task != ctx->task) - goto err_context; + goto err_locked; /* - * Do not allow to attach to a group in a different task - * or CPU context. If we're moving SW events, we'll fix - * this up later, so allow that. - * - * Racy, not holding group_leader->ctx->mutex, see comment with - * perf_event_ctx_lock(). + * Make sure we're both on the same context; either task or cpu. */ - if (!move_group && group_leader->ctx != ctx) - goto err_context; + if (group_leader->ctx != ctx) + goto err_locked; /* * Only a group leader can be exclusive or pinned */ if (attr.exclusive || attr.pinned) - goto err_context; - } - - if (output_event) { - err = perf_event_set_output(event, output_event); - if (err) - goto err_context; - } - - event_file = anon_inode_getfile("[perf_event]", &perf_fops, event, - f_flags); - if (IS_ERR(event_file)) { - err = PTR_ERR(event_file); - event_file = NULL; - goto err_context; - } - - if (task) { - err = down_read_interruptible(&task->signal->exec_update_lock); - if (err) - goto err_file; - - /* - * We must hold exec_update_lock across this and any potential - * perf_install_in_context() call for this new event to - * serialize against exec() altering our credentials (and the - * perf_event_exit_task() that could imply). - */ - err = -EACCES; - if (!perf_check_permission(&attr, task)) - goto err_cred; - } - - if (move_group) { - gctx = __perf_event_ctx_lock_double(group_leader, ctx); - - if (gctx->task == TASK_TOMBSTONE) { - err = -ESRCH; goto err_locked; - } - /* - * Check if we raced against another sys_perf_event_open() call - * moving the software group underneath us. - */ - if (!(group_leader->group_caps & PERF_EV_CAP_SOFTWARE)) { + if (is_software_event(event) && + !in_software_context(group_leader)) { /* - * If someone moved the group out from under us, check - * if this new event wound up on the same ctx, if so - * its the regular !move_group case, otherwise fail. + * If the event is a sw event, but the group_leader + * is on hw context. + * + * Allow the addition of software events to hw + * groups, this is safe because software events + * never fail to schedule. + * + * Note the comment that goes with struct + * perf_event_pmu_context. */ - if (gctx != ctx) { - err = -EINVAL; - goto err_locked; - } else { - perf_event_ctx_unlock(group_leader, gctx); - move_group = 0; - goto not_move_group; + pmu = group_leader->pmu_ctx->pmu; + } else if (!is_software_event(event)) { + if (is_software_event(group_leader) && + (group_leader->group_caps & PERF_EV_CAP_SOFTWARE)) { + /* + * In case the group is a pure software group, and we + * try to add a hardware event, move the whole group to + * the hardware context. + */ + move_group = 1; } - } - - /* - * Failure to create exclusive events returns -EBUSY. - */ - err = -EBUSY; - if (!exclusive_event_installable(group_leader, ctx)) - goto err_locked; - for_each_sibling_event(sibling, group_leader) { - if (!exclusive_event_installable(sibling, ctx)) + /* Don't allow group of multiple hw events from different pmus */ + if (!in_software_context(group_leader) && + group_leader->pmu_ctx->pmu != pmu) goto err_locked; } - } else { - mutex_lock(&ctx->mutex); - - /* - * Now that we hold ctx->lock, (re)validate group_leader->ctx == ctx, - * see the group_leader && !move_group test earlier. - */ - if (group_leader && group_leader->ctx != ctx) { - err = -EINVAL; - goto err_locked; - } } -not_move_group: - if (ctx->task == TASK_TOMBSTONE) { - err = -ESRCH; + /* + * Now that we're certain of the pmu; find the pmu_ctx. + */ + pmu_ctx = find_get_pmu_context(pmu, ctx, event); + if (IS_ERR(pmu_ctx)) { + err = PTR_ERR(pmu_ctx); goto err_locked; } + event->pmu_ctx = pmu_ctx; - if (!perf_event_validate_size(event)) { - err = -E2BIG; - goto err_locked; + if (output_event) { + err = perf_event_set_output(event, output_event); + if (err) + goto err_context; } - if (!task) { - /* - * Check if the @cpu we're creating an event for is online. - * - * We use the perf_cpu_context::ctx::mutex to serialize against - * the hotplug notifiers. See perf_event_{init,exit}_cpu(). - */ - struct perf_cpu_context *cpuctx = - container_of(ctx, struct perf_cpu_context, ctx); - - if (!cpuctx->online) { - err = -ENODEV; - goto err_locked; - } + if (!perf_event_validate_size(event)) { + err = -E2BIG; + goto err_context; } if (perf_need_aux_event(event) && !perf_get_aux_event(event, group_leader)) { err = -EINVAL; - goto err_locked; + goto err_context; } /* @@ -12547,36 +12599,33 @@ not_move_group: */ if (!exclusive_event_installable(event, ctx)) { err = -EBUSY; - goto err_locked; + goto err_context; } WARN_ON_ONCE(ctx->parent_ctx); + event_file = anon_inode_getfile("[perf_event]", &perf_fops, event, f_flags); + if (IS_ERR(event_file)) { + err = PTR_ERR(event_file); + event_file = NULL; + goto err_context; + } + /* * This is the point on no return; we cannot fail hereafter. This is * where we start modifying current state. */ if (move_group) { - /* - * See perf_event_ctx_lock() for comments on the details - * of swizzling perf_event::ctx. - */ perf_remove_from_context(group_leader, 0); - put_ctx(gctx); + put_pmu_ctx(group_leader->pmu_ctx); for_each_sibling_event(sibling, group_leader) { perf_remove_from_context(sibling, 0); - put_ctx(gctx); + put_pmu_ctx(sibling->pmu_ctx); } /* - * Wait for everybody to stop referencing the events through - * the old lists, before installing it on new lists. - */ - synchronize_rcu(); - - /* * Install the group siblings before the group leader. * * Because a group leader will try and install the entire group @@ -12587,9 +12636,10 @@ not_move_group: * reachable through the group lists. */ for_each_sibling_event(sibling, group_leader) { + sibling->pmu_ctx = pmu_ctx; + get_pmu_ctx(pmu_ctx); perf_event__state_init(sibling); perf_install_in_context(ctx, sibling, sibling->cpu); - get_ctx(ctx); } /* @@ -12597,9 +12647,10 @@ not_move_group: * event. What we want here is event in the initial * startup state, ready to be add into new context. */ + group_leader->pmu_ctx = pmu_ctx; + get_pmu_ctx(pmu_ctx); perf_event__state_init(group_leader); perf_install_in_context(ctx, group_leader, group_leader->cpu); - get_ctx(ctx); } /* @@ -12616,8 +12667,6 @@ not_move_group: perf_install_in_context(ctx, event, event->cpu); perf_unpin_context(ctx); - if (move_group) - perf_event_ctx_unlock(group_leader, gctx); mutex_unlock(&ctx->mutex); if (task) { @@ -12639,25 +12688,17 @@ not_move_group: fd_install(event_fd, event_file); return event_fd; +err_context: + /* event->pmu_ctx freed by free_event() */ err_locked: - if (move_group) - perf_event_ctx_unlock(group_leader, gctx); mutex_unlock(&ctx->mutex); + perf_unpin_context(ctx); + put_ctx(ctx); err_cred: if (task) up_read(&task->signal->exec_update_lock); -err_file: - fput(event_file); -err_context: - perf_unpin_context(ctx); - put_ctx(ctx); err_alloc: - /* - * If event_file is set, the fput() above will have called ->release() - * and that will take care of freeing the event. - */ - if (!event_file) - free_event(event); + free_event(event); err_task: if (task) put_task_struct(task); @@ -12683,8 +12724,10 @@ perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu, perf_overflow_handler_t overflow_handler, void *context) { + struct perf_event_pmu_context *pmu_ctx; struct perf_event_context *ctx; struct perf_event *event; + struct pmu *pmu; int err; /* @@ -12703,14 +12746,18 @@ perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu, /* Mark owner so we could distinguish it from user events. */ event->owner = TASK_TOMBSTONE; + pmu = event->pmu; + + if (pmu->task_ctx_nr == perf_sw_context) + event->event_caps |= PERF_EV_CAP_SOFTWARE; /* * Get the target context (task or percpu): */ - ctx = find_get_context(event->pmu, task, event); + ctx = find_get_context(task, event); if (IS_ERR(ctx)) { err = PTR_ERR(ctx); - goto err_free; + goto err_alloc; } WARN_ON_ONCE(ctx->parent_ctx); @@ -12720,6 +12767,13 @@ perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu, goto err_unlock; } + pmu_ctx = find_get_pmu_context(pmu, ctx, event); + if (IS_ERR(pmu_ctx)) { + err = PTR_ERR(pmu_ctx); + goto err_unlock; + } + event->pmu_ctx = pmu_ctx; + if (!task) { /* * Check if the @cpu we're creating an event for is online. @@ -12731,13 +12785,13 @@ perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu, container_of(ctx, struct perf_cpu_context, ctx); if (!cpuctx->online) { err = -ENODEV; - goto err_unlock; + goto err_pmu_ctx; } } if (!exclusive_event_installable(event, ctx)) { err = -EBUSY; - goto err_unlock; + goto err_pmu_ctx; } perf_install_in_context(ctx, event, event->cpu); @@ -12746,44 +12800,61 @@ perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu, return event; +err_pmu_ctx: + put_pmu_ctx(pmu_ctx); err_unlock: mutex_unlock(&ctx->mutex); perf_unpin_context(ctx); put_ctx(ctx); -err_free: +err_alloc: free_event(event); err: return ERR_PTR(err); } EXPORT_SYMBOL_GPL(perf_event_create_kernel_counter); -void perf_pmu_migrate_context(struct pmu *pmu, int src_cpu, int dst_cpu) +static void __perf_pmu_remove(struct perf_event_context *ctx, + int cpu, struct pmu *pmu, + struct perf_event_groups *groups, + struct list_head *events) { - struct perf_event_context *src_ctx; - struct perf_event_context *dst_ctx; - struct perf_event *event, *tmp; - LIST_HEAD(events); - - src_ctx = &per_cpu_ptr(pmu->pmu_cpu_context, src_cpu)->ctx; - dst_ctx = &per_cpu_ptr(pmu->pmu_cpu_context, dst_cpu)->ctx; + struct perf_event *event, *sibling; - /* - * See perf_event_ctx_lock() for comments on the details - * of swizzling perf_event::ctx. - */ - mutex_lock_double(&src_ctx->mutex, &dst_ctx->mutex); - list_for_each_entry_safe(event, tmp, &src_ctx->event_list, - event_entry) { + perf_event_groups_for_cpu_pmu(event, groups, cpu, pmu) { perf_remove_from_context(event, 0); - unaccount_event_cpu(event, src_cpu); - put_ctx(src_ctx); - list_add(&event->migrate_entry, &events); + unaccount_event_cpu(event, cpu); + put_pmu_ctx(event->pmu_ctx); + list_add(&event->migrate_entry, events); + + for_each_sibling_event(sibling, event) { + perf_remove_from_context(sibling, 0); + unaccount_event_cpu(sibling, cpu); + put_pmu_ctx(sibling->pmu_ctx); + list_add(&sibling->migrate_entry, events); + } } +} - /* - * Wait for the events to quiesce before re-instating them. - */ - synchronize_rcu(); +static void __perf_pmu_install_event(struct pmu *pmu, + struct perf_event_context *ctx, + int cpu, struct perf_event *event) +{ + struct perf_event_pmu_context *epc; + + event->cpu = cpu; + epc = find_get_pmu_context(pmu, ctx, event); + event->pmu_ctx = epc; + + if (event->state >= PERF_EVENT_STATE_OFF) + event->state = PERF_EVENT_STATE_INACTIVE; + account_event_cpu(event, cpu); + perf_install_in_context(ctx, event, cpu); +} + +static void __perf_pmu_install(struct perf_event_context *ctx, + int cpu, struct pmu *pmu, struct list_head *events) +{ + struct perf_event *event, *tmp; /* * Re-instate events in 2 passes. @@ -12793,30 +12864,48 @@ void perf_pmu_migrate_context(struct pmu *pmu, int src_cpu, int dst_cpu) * leader will enable its siblings, even if those are still on the old * context. */ - list_for_each_entry_safe(event, tmp, &events, migrate_entry) { + list_for_each_entry_safe(event, tmp, events, migrate_entry) { if (event->group_leader == event) continue; list_del(&event->migrate_entry); - if (event->state >= PERF_EVENT_STATE_OFF) - event->state = PERF_EVENT_STATE_INACTIVE; - account_event_cpu(event, dst_cpu); - perf_install_in_context(dst_ctx, event, dst_cpu); - get_ctx(dst_ctx); + __perf_pmu_install_event(pmu, ctx, cpu, event); } /* * Once all the siblings are setup properly, install the group leaders * to make it go. */ - list_for_each_entry_safe(event, tmp, &events, migrate_entry) { + list_for_each_entry_safe(event, tmp, events, migrate_entry) { list_del(&event->migrate_entry); - if (event->state >= PERF_EVENT_STATE_OFF) - event->state = PERF_EVENT_STATE_INACTIVE; - account_event_cpu(event, dst_cpu); - perf_install_in_context(dst_ctx, event, dst_cpu); - get_ctx(dst_ctx); + __perf_pmu_install_event(pmu, ctx, cpu, event); } +} + +void perf_pmu_migrate_context(struct pmu *pmu, int src_cpu, int dst_cpu) +{ + struct perf_event_context *src_ctx, *dst_ctx; + LIST_HEAD(events); + + src_ctx = &per_cpu_ptr(&perf_cpu_context, src_cpu)->ctx; + dst_ctx = &per_cpu_ptr(&perf_cpu_context, dst_cpu)->ctx; + + /* + * See perf_event_ctx_lock() for comments on the details + * of swizzling perf_event::ctx. + */ + mutex_lock_double(&src_ctx->mutex, &dst_ctx->mutex); + + __perf_pmu_remove(src_ctx, src_cpu, pmu, &src_ctx->pinned_groups, &events); + __perf_pmu_remove(src_ctx, src_cpu, pmu, &src_ctx->flexible_groups, &events); + + /* + * Wait for the events to quiesce before re-instating them. + */ + synchronize_rcu(); + + __perf_pmu_install(dst_ctx, dst_cpu, pmu, &events); + mutex_unlock(&dst_ctx->mutex); mutex_unlock(&src_ctx->mutex); } @@ -12896,14 +12985,14 @@ perf_event_exit_event(struct perf_event *event, struct perf_event_context *ctx) perf_event_wakeup(event); } -static void perf_event_exit_task_context(struct task_struct *child, int ctxn) +static void perf_event_exit_task_context(struct task_struct *child) { struct perf_event_context *child_ctx, *clone_ctx = NULL; struct perf_event *child_event, *next; WARN_ON_ONCE(child != current); - child_ctx = perf_pin_task_context(child, ctxn); + child_ctx = perf_pin_task_context(child); if (!child_ctx) return; @@ -12925,13 +13014,13 @@ static void perf_event_exit_task_context(struct task_struct *child, int ctxn) * in. */ raw_spin_lock_irq(&child_ctx->lock); - task_ctx_sched_out(__get_cpu_context(child_ctx), child_ctx, EVENT_ALL); + task_ctx_sched_out(child_ctx, EVENT_ALL); /* * Now that the context is inactive, destroy the task <-> ctx relation * and mark the context dead. */ - RCU_INIT_POINTER(child->perf_event_ctxp[ctxn], NULL); + RCU_INIT_POINTER(child->perf_event_ctxp, NULL); put_ctx(child_ctx); /* cannot be last */ WRITE_ONCE(child_ctx->task, TASK_TOMBSTONE); put_task_struct(current); /* cannot be last */ @@ -12966,7 +13055,6 @@ static void perf_event_exit_task_context(struct task_struct *child, int ctxn) void perf_event_exit_task(struct task_struct *child) { struct perf_event *event, *tmp; - int ctxn; mutex_lock(&child->perf_event_mutex); list_for_each_entry_safe(event, tmp, &child->perf_event_list, @@ -12982,8 +13070,7 @@ void perf_event_exit_task(struct task_struct *child) } mutex_unlock(&child->perf_event_mutex); - for_each_task_context_nr(ctxn) - perf_event_exit_task_context(child, ctxn); + perf_event_exit_task_context(child); /* * The perf_event_exit_task_context calls perf_event_task @@ -13026,56 +13113,51 @@ void perf_event_free_task(struct task_struct *task) { struct perf_event_context *ctx; struct perf_event *event, *tmp; - int ctxn; - for_each_task_context_nr(ctxn) { - ctx = task->perf_event_ctxp[ctxn]; - if (!ctx) - continue; + ctx = rcu_access_pointer(task->perf_event_ctxp); + if (!ctx) + return; - mutex_lock(&ctx->mutex); - raw_spin_lock_irq(&ctx->lock); - /* - * Destroy the task <-> ctx relation and mark the context dead. - * - * This is important because even though the task hasn't been - * exposed yet the context has been (through child_list). - */ - RCU_INIT_POINTER(task->perf_event_ctxp[ctxn], NULL); - WRITE_ONCE(ctx->task, TASK_TOMBSTONE); - put_task_struct(task); /* cannot be last */ - raw_spin_unlock_irq(&ctx->lock); + mutex_lock(&ctx->mutex); + raw_spin_lock_irq(&ctx->lock); + /* + * Destroy the task <-> ctx relation and mark the context dead. + * + * This is important because even though the task hasn't been + * exposed yet the context has been (through child_list). + */ + RCU_INIT_POINTER(task->perf_event_ctxp, NULL); + WRITE_ONCE(ctx->task, TASK_TOMBSTONE); + put_task_struct(task); /* cannot be last */ + raw_spin_unlock_irq(&ctx->lock); - list_for_each_entry_safe(event, tmp, &ctx->event_list, event_entry) - perf_free_event(event, ctx); - mutex_unlock(&ctx->mutex); + list_for_each_entry_safe(event, tmp, &ctx->event_list, event_entry) + perf_free_event(event, ctx); - /* - * perf_event_release_kernel() could've stolen some of our - * child events and still have them on its free_list. In that - * case we must wait for these events to have been freed (in - * particular all their references to this task must've been - * dropped). - * - * Without this copy_process() will unconditionally free this - * task (irrespective of its reference count) and - * _free_event()'s put_task_struct(event->hw.target) will be a - * use-after-free. - * - * Wait for all events to drop their context reference. - */ - wait_var_event(&ctx->refcount, refcount_read(&ctx->refcount) == 1); - put_ctx(ctx); /* must be last */ - } + mutex_unlock(&ctx->mutex); + + /* + * perf_event_release_kernel() could've stolen some of our + * child events and still have them on its free_list. In that + * case we must wait for these events to have been freed (in + * particular all their references to this task must've been + * dropped). + * + * Without this copy_process() will unconditionally free this + * task (irrespective of its reference count) and + * _free_event()'s put_task_struct(event->hw.target) will be a + * use-after-free. + * + * Wait for all events to drop their context reference. + */ + wait_var_event(&ctx->refcount, refcount_read(&ctx->refcount) == 1); + put_ctx(ctx); /* must be last */ } void perf_event_delayed_put(struct task_struct *task) { - int ctxn; - - for_each_task_context_nr(ctxn) - WARN_ON_ONCE(task->perf_event_ctxp[ctxn]); + WARN_ON_ONCE(task->perf_event_ctxp); } struct file *perf_event_get(unsigned int fd) @@ -13125,6 +13207,7 @@ inherit_event(struct perf_event *parent_event, struct perf_event_context *child_ctx) { enum perf_event_state parent_state = parent_event->state; + struct perf_event_pmu_context *pmu_ctx; struct perf_event *child_event; unsigned long flags; @@ -13145,17 +13228,12 @@ inherit_event(struct perf_event *parent_event, if (IS_ERR(child_event)) return child_event; - - if ((child_event->attach_state & PERF_ATTACH_TASK_DATA) && - !child_ctx->task_ctx_data) { - struct pmu *pmu = child_event->pmu; - - child_ctx->task_ctx_data = alloc_task_ctx_data(pmu); - if (!child_ctx->task_ctx_data) { - free_event(child_event); - return ERR_PTR(-ENOMEM); - } + pmu_ctx = find_get_pmu_context(child_event->pmu, child_ctx, child_event); + if (IS_ERR(pmu_ctx)) { + free_event(child_event); + return NULL; } + child_event->pmu_ctx = pmu_ctx; /* * is_orphaned_event() and list_add_tail(&parent_event->child_list) @@ -13278,11 +13356,11 @@ static int inherit_group(struct perf_event *parent_event, static int inherit_task_group(struct perf_event *event, struct task_struct *parent, struct perf_event_context *parent_ctx, - struct task_struct *child, int ctxn, + struct task_struct *child, u64 clone_flags, int *inherited_all) { - int ret; struct perf_event_context *child_ctx; + int ret; if (!event->attr.inherit || (event->attr.inherit_thread && !(clone_flags & CLONE_THREAD)) || @@ -13292,7 +13370,7 @@ inherit_task_group(struct perf_event *event, struct task_struct *parent, return 0; } - child_ctx = child->perf_event_ctxp[ctxn]; + child_ctx = child->perf_event_ctxp; if (!child_ctx) { /* * This is executed from the parent task context, so @@ -13300,16 +13378,14 @@ inherit_task_group(struct perf_event *event, struct task_struct *parent, * First allocate and initialize a context for the * child. */ - child_ctx = alloc_perf_context(parent_ctx->pmu, child); + child_ctx = alloc_perf_context(child); if (!child_ctx) return -ENOMEM; - child->perf_event_ctxp[ctxn] = child_ctx; + child->perf_event_ctxp = child_ctx; } - ret = inherit_group(event, parent, parent_ctx, - child, child_ctx); - + ret = inherit_group(event, parent, parent_ctx, child, child_ctx); if (ret) *inherited_all = 0; @@ -13319,8 +13395,7 @@ inherit_task_group(struct perf_event *event, struct task_struct *parent, /* * Initialize the perf_event context in task_struct */ -static int perf_event_init_context(struct task_struct *child, int ctxn, - u64 clone_flags) +static int perf_event_init_context(struct task_struct *child, u64 clone_flags) { struct perf_event_context *child_ctx, *parent_ctx; struct perf_event_context *cloned_ctx; @@ -13330,14 +13405,14 @@ static int perf_event_init_context(struct task_struct *child, int ctxn, unsigned long flags; int ret = 0; - if (likely(!parent->perf_event_ctxp[ctxn])) + if (likely(!parent->perf_event_ctxp)) return 0; /* * If the parent's context is a clone, pin it so it won't get * swapped under us. */ - parent_ctx = perf_pin_task_context(parent, ctxn); + parent_ctx = perf_pin_task_context(parent); if (!parent_ctx) return 0; @@ -13360,8 +13435,7 @@ static int perf_event_init_context(struct task_struct *child, int ctxn, */ perf_event_groups_for_each(event, &parent_ctx->pinned_groups) { ret = inherit_task_group(event, parent, parent_ctx, - child, ctxn, clone_flags, - &inherited_all); + child, clone_flags, &inherited_all); if (ret) goto out_unlock; } @@ -13377,8 +13451,7 @@ static int perf_event_init_context(struct task_struct *child, int ctxn, perf_event_groups_for_each(event, &parent_ctx->flexible_groups) { ret = inherit_task_group(event, parent, parent_ctx, - child, ctxn, clone_flags, - &inherited_all); + child, clone_flags, &inherited_all); if (ret) goto out_unlock; } @@ -13386,7 +13459,7 @@ static int perf_event_init_context(struct task_struct *child, int ctxn, raw_spin_lock_irqsave(&parent_ctx->lock, flags); parent_ctx->rotate_disable = 0; - child_ctx = child->perf_event_ctxp[ctxn]; + child_ctx = child->perf_event_ctxp; if (child_ctx && inherited_all) { /* @@ -13422,18 +13495,16 @@ out_unlock: */ int perf_event_init_task(struct task_struct *child, u64 clone_flags) { - int ctxn, ret; + int ret; - memset(child->perf_event_ctxp, 0, sizeof(child->perf_event_ctxp)); + child->perf_event_ctxp = NULL; mutex_init(&child->perf_event_mutex); INIT_LIST_HEAD(&child->perf_event_list); - for_each_task_context_nr(ctxn) { - ret = perf_event_init_context(child, ctxn, clone_flags); - if (ret) { - perf_event_free_task(child); - return ret; - } + ret = perf_event_init_context(child, clone_flags); + if (ret) { + perf_event_free_task(child); + return ret; } return 0; @@ -13442,6 +13513,7 @@ int perf_event_init_task(struct task_struct *child, u64 clone_flags) static void __init perf_event_init_all_cpus(void) { struct swevent_htable *swhash; + struct perf_cpu_context *cpuctx; int cpu; zalloc_cpumask_var(&perf_online_mask, GFP_KERNEL); @@ -13449,15 +13521,19 @@ static void __init perf_event_init_all_cpus(void) for_each_possible_cpu(cpu) { swhash = &per_cpu(swevent_htable, cpu); mutex_init(&swhash->hlist_mutex); - INIT_LIST_HEAD(&per_cpu(active_ctx_list, cpu)); INIT_LIST_HEAD(&per_cpu(pmu_sb_events.list, cpu)); raw_spin_lock_init(&per_cpu(pmu_sb_events.lock, cpu)); -#ifdef CONFIG_CGROUP_PERF - INIT_LIST_HEAD(&per_cpu(cgrp_cpuctx_list, cpu)); -#endif INIT_LIST_HEAD(&per_cpu(sched_cb_list, cpu)); + + cpuctx = per_cpu_ptr(&perf_cpu_context, cpu); + __perf_event_init_context(&cpuctx->ctx); + lockdep_set_class(&cpuctx->ctx.mutex, &cpuctx_mutex); + lockdep_set_class(&cpuctx->ctx.lock, &cpuctx_lock); + cpuctx->online = cpumask_test_cpu(cpu, perf_online_mask); + cpuctx->heap_size = ARRAY_SIZE(cpuctx->heap_default); + cpuctx->heap = cpuctx->heap_default; } } @@ -13479,12 +13555,12 @@ static void perf_swevent_init_cpu(unsigned int cpu) #if defined CONFIG_HOTPLUG_CPU || defined CONFIG_KEXEC_CORE static void __perf_event_exit_context(void *__info) { + struct perf_cpu_context *cpuctx = this_cpu_ptr(&perf_cpu_context); struct perf_event_context *ctx = __info; - struct perf_cpu_context *cpuctx = __get_cpu_context(ctx); struct perf_event *event; raw_spin_lock(&ctx->lock); - ctx_sched_out(ctx, cpuctx, EVENT_TIME); + ctx_sched_out(ctx, EVENT_TIME); list_for_each_entry(event, &ctx->event_list, event_entry) __perf_remove_from_context(event, cpuctx, ctx, (void *)DETACH_GROUP); raw_spin_unlock(&ctx->lock); @@ -13494,18 +13570,16 @@ static void perf_event_exit_cpu_context(int cpu) { struct perf_cpu_context *cpuctx; struct perf_event_context *ctx; - struct pmu *pmu; + // XXX simplify cpuctx->online mutex_lock(&pmus_lock); - list_for_each_entry(pmu, &pmus, entry) { - cpuctx = per_cpu_ptr(pmu->pmu_cpu_context, cpu); - ctx = &cpuctx->ctx; + cpuctx = per_cpu_ptr(&perf_cpu_context, cpu); + ctx = &cpuctx->ctx; - mutex_lock(&ctx->mutex); - smp_call_function_single(cpu, __perf_event_exit_context, ctx, 1); - cpuctx->online = 0; - mutex_unlock(&ctx->mutex); - } + mutex_lock(&ctx->mutex); + smp_call_function_single(cpu, __perf_event_exit_context, ctx, 1); + cpuctx->online = 0; + mutex_unlock(&ctx->mutex); cpumask_clear_cpu(cpu, perf_online_mask); mutex_unlock(&pmus_lock); } @@ -13519,20 +13593,17 @@ int perf_event_init_cpu(unsigned int cpu) { struct perf_cpu_context *cpuctx; struct perf_event_context *ctx; - struct pmu *pmu; perf_swevent_init_cpu(cpu); mutex_lock(&pmus_lock); cpumask_set_cpu(cpu, perf_online_mask); - list_for_each_entry(pmu, &pmus, entry) { - cpuctx = per_cpu_ptr(pmu->pmu_cpu_context, cpu); - ctx = &cpuctx->ctx; + cpuctx = per_cpu_ptr(&perf_cpu_context, cpu); + ctx = &cpuctx->ctx; - mutex_lock(&ctx->mutex); - cpuctx->online = 1; - mutex_unlock(&ctx->mutex); - } + mutex_lock(&ctx->mutex); + cpuctx->online = 1; + mutex_unlock(&ctx->mutex); mutex_unlock(&pmus_lock); return 0; @@ -13669,9 +13740,12 @@ static int perf_cgroup_css_online(struct cgroup_subsys_state *css) static int __perf_cgroup_move(void *info) { struct task_struct *task = info; - rcu_read_lock(); - perf_cgroup_switch(task); - rcu_read_unlock(); + + preempt_disable(); + if (atomic_read(this_cpu_ptr(&perf_cgroup_events))) + perf_cgroup_switch(task); + preempt_enable(); + return 0; } |