From 4dca4fadb4fcdab59b71f380558030f31f75fa2c Mon Sep 17 00:00:00 2001 From: Kan Liang Date: Wed, 25 Oct 2023 13:16:19 -0700 Subject: [PATCH 1/5] perf: Add branch stack counters Upstream commit: 571d91dcadfa3cef499010b4eddb9b58b0da4d24 Conflict: none Currently, the additional information of a branch entry is stored in a u64 space. With more and more information added, the space is running out. For example, the information of occurrences of events will be added for each branch. Two places were suggested to append the counters. https://lore.kernel.org/lkml/20230802215814.GH231007@hirez.programming.kicks-ass.net/ One place is right after the flags of each branch entry. It changes the existing struct perf_branch_entry. The later ARCH specific implementation has to be really careful to consistently pick the right struct. The other place is right after the entire struct perf_branch_stack. The disadvantage is that the pointer of the extra space has to be recorded. The common interface perf_sample_save_brstack() has to be updated. The latter is much straightforward, and should be easily understood and maintained. It is implemented in the patch. Add a new branch sample type, PERF_SAMPLE_BRANCH_COUNTERS, to indicate the event which is recorded in the branch info. The "u64 counters" may store the occurrences of several events. The information regarding the number of events/counters and the width of each counter should be exposed via sysfs as a reference for the perf tool. Define the branch_counter_nr and branch_counter_width ABI here. The support will be implemented later in the Intel-specific patch. Intel-SIG: commit 571d91dcadfa perf: Add branch stack counters Backport LBR branch counter support to kernel v6.6. Suggested-by: Peter Zijlstra (Intel) Signed-off-by: Kan Liang Signed-off-by: Peter Zijlstra (Intel) Link: https://lkml.kernel.org/r/20231025201626.3000228-1-kan.liang@linux.intel.com [ Yunying Sun: amend commit log ] Signed-off-by: Yunying Sun --- .../testing/sysfs-bus-event_source-devices-caps | 6 ++++++ arch/powerpc/perf/core-book3s.c | 2 +- arch/x86/events/amd/core.c | 2 +- arch/x86/events/core.c | 2 +- arch/x86/events/intel/core.c | 2 +- arch/x86/events/intel/ds.c | 4 ++-- include/linux/perf_event.h | 17 ++++++++++++++++- include/uapi/linux/perf_event.h | 10 ++++++++++ kernel/events/core.c | 8 ++++++++ 9 files changed, 46 insertions(+), 7 deletions(-) diff --git a/Documentation/ABI/testing/sysfs-bus-event_source-devices-caps b/Documentation/ABI/testing/sysfs-bus-event_source-devices-caps index 8757dcf41c08..a5f506f7d481 100644 --- a/Documentation/ABI/testing/sysfs-bus-event_source-devices-caps +++ b/Documentation/ABI/testing/sysfs-bus-event_source-devices-caps @@ -16,3 +16,9 @@ Description: Example output in powerpc: grep . /sys/bus/event_source/devices/cpu/caps/* /sys/bus/event_source/devices/cpu/caps/pmu_name:POWER9 + + The "branch_counter_nr" in the supported platform exposes the + maximum number of counters which can be shown in the u64 counters + of PERF_SAMPLE_BRANCH_COUNTERS, while the "branch_counter_width" + exposes the width of each counter. Both of them can be used by + the perf tool to parse the logged counters in each branch. diff --git a/arch/powerpc/perf/core-book3s.c b/arch/powerpc/perf/core-book3s.c index 10b946e9c6e7..b7ff680cde96 100644 --- a/arch/powerpc/perf/core-book3s.c +++ b/arch/powerpc/perf/core-book3s.c @@ -2312,7 +2312,7 @@ static void record_and_restart(struct perf_event *event, unsigned long val, struct cpu_hw_events *cpuhw; cpuhw = this_cpu_ptr(&cpu_hw_events); power_pmu_bhrb_read(event, cpuhw); - perf_sample_save_brstack(&data, event, &cpuhw->bhrb_stack); + perf_sample_save_brstack(&data, event, &cpuhw->bhrb_stack, NULL); } if (event->attr.sample_type & PERF_SAMPLE_DATA_SRC && diff --git a/arch/x86/events/amd/core.c b/arch/x86/events/amd/core.c index e24976593a29..4ee6390b45c9 100644 --- a/arch/x86/events/amd/core.c +++ b/arch/x86/events/amd/core.c @@ -940,7 +940,7 @@ static int amd_pmu_v2_handle_irq(struct pt_regs *regs) continue; if (has_branch_stack(event)) - perf_sample_save_brstack(&data, event, &cpuc->lbr_stack); + perf_sample_save_brstack(&data, event, &cpuc->lbr_stack, NULL); if (perf_event_overflow(event, &data, regs)) x86_pmu_stop(event, 0); diff --git a/arch/x86/events/core.c b/arch/x86/events/core.c index 185f902e5f28..2919bb5a53a0 100644 --- a/arch/x86/events/core.c +++ b/arch/x86/events/core.c @@ -1702,7 +1702,7 @@ int x86_pmu_handle_irq(struct pt_regs *regs) perf_sample_data_init(&data, 0, event->hw.last_period); if (has_branch_stack(event)) - perf_sample_save_brstack(&data, event, &cpuc->lbr_stack); + perf_sample_save_brstack(&data, event, &cpuc->lbr_stack, NULL); if (perf_event_overflow(event, &data, regs)) x86_pmu_stop(event, 0); diff --git a/arch/x86/events/intel/core.c b/arch/x86/events/intel/core.c index fa355d3658a6..71444d133e71 100644 --- a/arch/x86/events/intel/core.c +++ b/arch/x86/events/intel/core.c @@ -3058,7 +3058,7 @@ static int handle_pmi_common(struct pt_regs *regs, u64 status) perf_sample_data_init(&data, 0, event->hw.last_period); if (has_branch_stack(event)) - perf_sample_save_brstack(&data, event, &cpuc->lbr_stack); + perf_sample_save_brstack(&data, event, &cpuc->lbr_stack, NULL); if (perf_event_overflow(event, &data, regs)) x86_pmu_stop(event, 0); diff --git a/arch/x86/events/intel/ds.c b/arch/x86/events/intel/ds.c index eb8dd8b8a1e8..7566190389f0 100644 --- a/arch/x86/events/intel/ds.c +++ b/arch/x86/events/intel/ds.c @@ -1755,7 +1755,7 @@ static void setup_pebs_fixed_sample_data(struct perf_event *event, setup_pebs_time(event, data, pebs->tsc); if (has_branch_stack(event)) - perf_sample_save_brstack(data, event, &cpuc->lbr_stack); + perf_sample_save_brstack(data, event, &cpuc->lbr_stack, NULL); } static void adaptive_pebs_save_regs(struct pt_regs *regs, @@ -1912,7 +1912,7 @@ static void setup_pebs_adaptive_sample_data(struct perf_event *event, if (has_branch_stack(event)) { intel_pmu_store_pebs_lbrs(lbr); - perf_sample_save_brstack(data, event, &cpuc->lbr_stack); + perf_sample_save_brstack(data, event, &cpuc->lbr_stack, NULL); } } diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index da572604d83e..1557d33e8016 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -1148,6 +1148,10 @@ static inline bool branch_sample_priv(const struct perf_event *event) return event->attr.branch_sample_type & PERF_SAMPLE_BRANCH_PRIV_SAVE; } +static inline bool branch_sample_counters(const struct perf_event *event) +{ + return event->attr.branch_sample_type & PERF_SAMPLE_BRANCH_COUNTERS; +} struct perf_sample_data { /* @@ -1182,6 +1186,7 @@ struct perf_sample_data { struct perf_callchain_entry *callchain; struct perf_raw_record *raw; struct perf_branch_stack *br_stack; + u64 *br_stack_cntr; union perf_sample_weight weight; union perf_mem_data_src data_src; u64 txn; @@ -1259,7 +1264,8 @@ static inline void perf_sample_save_raw_data(struct perf_sample_data *data, static inline void perf_sample_save_brstack(struct perf_sample_data *data, struct perf_event *event, - struct perf_branch_stack *brs) + struct perf_branch_stack *brs, + u64 *brs_cntr) { int size = sizeof(u64); /* nr */ @@ -1267,7 +1273,16 @@ static inline void perf_sample_save_brstack(struct perf_sample_data *data, size += sizeof(u64); size += brs->nr * sizeof(struct perf_branch_entry); + /* + * The extension space for counters is appended after the + * struct perf_branch_stack. It is used to store the occurrences + * of events of each branch. + */ + if (brs_cntr) + size += brs->nr * sizeof(u64); + data->br_stack = brs; + data->br_stack_cntr = brs_cntr; data->dyn_size += size; data->sample_flags |= PERF_SAMPLE_BRANCH_STACK; } diff --git a/include/uapi/linux/perf_event.h b/include/uapi/linux/perf_event.h index 39c6a250dd1b..4461f380425b 100644 --- a/include/uapi/linux/perf_event.h +++ b/include/uapi/linux/perf_event.h @@ -204,6 +204,8 @@ enum perf_branch_sample_type_shift { PERF_SAMPLE_BRANCH_PRIV_SAVE_SHIFT = 18, /* save privilege mode */ + PERF_SAMPLE_BRANCH_COUNTERS_SHIFT = 19, /* save occurrences of events on a branch */ + PERF_SAMPLE_BRANCH_MAX_SHIFT /* non-ABI */ }; @@ -235,6 +237,8 @@ enum perf_branch_sample_type { PERF_SAMPLE_BRANCH_PRIV_SAVE = 1U << PERF_SAMPLE_BRANCH_PRIV_SAVE_SHIFT, + PERF_SAMPLE_BRANCH_COUNTERS = 1U << PERF_SAMPLE_BRANCH_COUNTERS_SHIFT, + PERF_SAMPLE_BRANCH_MAX = 1U << PERF_SAMPLE_BRANCH_MAX_SHIFT, }; @@ -982,6 +986,12 @@ enum perf_event_type { * { u64 nr; * { u64 hw_idx; } && PERF_SAMPLE_BRANCH_HW_INDEX * { u64 from, to, flags } lbr[nr]; + * # + * # The format of the counters is decided by the + * # "branch_counter_nr" and "branch_counter_width", + * # which are defined in the ABI. + * # + * { u64 counters; } cntr[nr] && PERF_SAMPLE_BRANCH_COUNTERS * } && PERF_SAMPLE_BRANCH_STACK * * { u64 abi; # enum perf_sample_regs_abi diff --git a/kernel/events/core.c b/kernel/events/core.c index 6dbb03c53237..c252302b7c35 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -7360,6 +7360,14 @@ void perf_output_sample(struct perf_output_handle *handle, if (branch_sample_hw_index(event)) perf_output_put(handle, data->br_stack->hw_idx); perf_output_copy(handle, data->br_stack->entries, size); + /* + * Add the extension space which is appended + * right after the struct perf_branch_stack. + */ + if (data->br_stack_cntr) { + size = data->br_stack->nr * sizeof(u64); + perf_output_copy(handle, data->br_stack_cntr, size); + } } else { /* * we always store at least the value of nr -- Gitee From 8a495d933db6aefd9308f99af9d1225a3d573749 Mon Sep 17 00:00:00 2001 From: Kan Liang Date: Wed, 25 Oct 2023 13:16:20 -0700 Subject: [PATCH 2/5] perf/x86: Add PERF_X86_EVENT_NEEDS_BRANCH_STACK flag Upstream commit: 85846b27072defc7ab3dcee7ff36563a040079dc Conflict: none Currently, branch_sample_type !=0 is used to check whether a branch stack setup is required. But it doesn't check the sample type, unnecessary branch stack setup may be done for a counting event. E.g., perf record -e "{branch-instructions,branch-misses}:S" -j any Also, the event only with the new PERF_SAMPLE_BRANCH_COUNTERS branch sample type may not require a branch stack setup either. Add a new flag NEEDS_BRANCH_STACK to indicate whether the event requires a branch stack setup. Replace the needs_branch_stack() by checking the new flag. The counting event check is implemented here. The later patch will take the new PERF_SAMPLE_BRANCH_COUNTERS into account. Intel-SIG: commit 85846b27072d perf/x86: Add PERF_X86_EVENT_NEEDS_BRANCH_STACK flag Backport LBR branch counter support to kernel v6.6. Signed-off-by: Kan Liang Signed-off-by: Peter Zijlstra (Intel) Link: https://lkml.kernel.org/r/20231025201626.3000228-2-kan.liang@linux.intel.com [ Yunying Sun: amend commit log ] Signed-off-by: Yunying Sun --- arch/x86/events/intel/core.c | 14 +++++++++++--- arch/x86/events/perf_event_flags.h | 1 + 2 files changed, 12 insertions(+), 3 deletions(-) diff --git a/arch/x86/events/intel/core.c b/arch/x86/events/intel/core.c index 71444d133e71..66ee69ab5bfd 100644 --- a/arch/x86/events/intel/core.c +++ b/arch/x86/events/intel/core.c @@ -2518,9 +2518,14 @@ static void intel_pmu_assign_event(struct perf_event *event, int idx) perf_report_aux_output_id(event, idx); } +static __always_inline bool intel_pmu_needs_branch_stack(struct perf_event *event) +{ + return event->hw.flags & PERF_X86_EVENT_NEEDS_BRANCH_STACK; +} + static void intel_pmu_del_event(struct perf_event *event) { - if (needs_branch_stack(event)) + if (intel_pmu_needs_branch_stack(event)) intel_pmu_lbr_del(event); if (event->attr.precise_ip) intel_pmu_pebs_del(event); @@ -2831,7 +2836,7 @@ static void intel_pmu_add_event(struct perf_event *event) { if (event->attr.precise_ip) intel_pmu_pebs_add(event); - if (needs_branch_stack(event)) + if (intel_pmu_needs_branch_stack(event)) intel_pmu_lbr_add(event); } @@ -3908,7 +3913,10 @@ static int intel_pmu_hw_config(struct perf_event *event) x86_pmu.pebs_aliases(event); } - if (needs_branch_stack(event)) { + if (needs_branch_stack(event) && is_sampling_event(event)) + event->hw.flags |= PERF_X86_EVENT_NEEDS_BRANCH_STACK; + + if (intel_pmu_needs_branch_stack(event)) { ret = intel_pmu_setup_lbr_filter(event); if (ret) return ret; diff --git a/arch/x86/events/perf_event_flags.h b/arch/x86/events/perf_event_flags.h index 1dc19b9b4426..a1685981c520 100644 --- a/arch/x86/events/perf_event_flags.h +++ b/arch/x86/events/perf_event_flags.h @@ -20,3 +20,4 @@ PERF_ARCH(TOPDOWN, 0x04000) /* Count Topdown slots/metrics events */ PERF_ARCH(PEBS_STLAT, 0x08000) /* st+stlat data address sampling */ PERF_ARCH(AMD_BRS, 0x10000) /* AMD Branch Sampling */ PERF_ARCH(PEBS_LAT_HYBRID, 0x20000) /* ld and st lat for hybrid */ +PERF_ARCH(NEEDS_BRANCH_STACK, 0x40000) /* require branch stack setup */ -- Gitee From 8459fa7dfd4d56c3cbb66d5328897e7798194fd1 Mon Sep 17 00:00:00 2001 From: Kan Liang Date: Wed, 25 Oct 2023 13:16:21 -0700 Subject: [PATCH 3/5] perf: Add branch_sample_call_stack Upstream commit: 1f2376cd03dd3b965d130ed46a7c92769d614ba1 Conflict: none Add a helper function to check call stack sample type. The later patch will invoke the function in several places. Intel-SIG: commit 1f2376cd03dd perf: Add branch_sample_call_stack Backport LBR branch counter support to kernel v6.6. Signed-off-by: Kan Liang Signed-off-by: Peter Zijlstra (Intel) Link: https://lkml.kernel.org/r/20231025201626.3000228-3-kan.liang@linux.intel.com [ Yunying Sun: amend commit log ] Signed-off-by: Yunying Sun --- arch/x86/events/core.c | 2 +- include/linux/perf_event.h | 5 +++++ 2 files changed, 6 insertions(+), 1 deletion(-) diff --git a/arch/x86/events/core.c b/arch/x86/events/core.c index 2919bb5a53a0..b56e9c3a0cc8 100644 --- a/arch/x86/events/core.c +++ b/arch/x86/events/core.c @@ -601,7 +601,7 @@ int x86_pmu_hw_config(struct perf_event *event) } } - if (event->attr.branch_sample_type & PERF_SAMPLE_BRANCH_CALL_STACK) + if (branch_sample_call_stack(event)) event->attach_state |= PERF_ATTACH_TASK_DATA; /* diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 1557d33e8016..3747004299b5 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -1153,6 +1153,11 @@ static inline bool branch_sample_counters(const struct perf_event *event) return event->attr.branch_sample_type & PERF_SAMPLE_BRANCH_COUNTERS; } +static inline bool branch_sample_call_stack(const struct perf_event *event) +{ + return event->attr.branch_sample_type & PERF_SAMPLE_BRANCH_CALL_STACK; +} + struct perf_sample_data { /* * Fields set by perf_sample_data_init() unconditionally, -- Gitee From ca4452a06006c7ee2edde7ce64e932dcb6851d2b Mon Sep 17 00:00:00 2001 From: Kan Liang Date: Wed, 25 Oct 2023 13:16:22 -0700 Subject: [PATCH 4/5] perf/x86/intel: Reorganize attrs and is_visible Upstream commit: 318c4985911245508f7e0bab5265e208a38b5f18 Conflict: none Some attrs and is_visible implementations are rather far away from one another which makes the whole thing hard to interpret. There are only two attribute groups which have both .attrs and .is_visible, group_default and group_caps_lbr. Move them together. No functional changes. Intel-SIG: commit 318c49859112 perf/x86/intel: Reorganize attrs and is_visible Backport LBR branch counter support to kernel v6.6. Suggested-by: Peter Zijlstra (Intel) Signed-off-by: Kan Liang Signed-off-by: Peter Zijlstra (Intel) Link: https://lkml.kernel.org/r/20231025201626.3000228-4-kan.liang@linux.intel.com [ Yunying Sun: amend commit log ] Signed-off-by: Yunying Sun --- arch/x86/events/intel/core.c | 30 +++++++++++++++--------------- 1 file changed, 15 insertions(+), 15 deletions(-) diff --git a/arch/x86/events/intel/core.c b/arch/x86/events/intel/core.c index 66ee69ab5bfd..6a8c50bd0608 100644 --- a/arch/x86/events/intel/core.c +++ b/arch/x86/events/intel/core.c @@ -5476,6 +5476,12 @@ static struct attribute *lbr_attrs[] = { NULL }; +static umode_t +lbr_is_visible(struct kobject *kobj, struct attribute *attr, int i) +{ + return x86_pmu.lbr_nr ? attr->mode : 0; +} + static char pmu_name_str[30]; static ssize_t pmu_name_show(struct device *cdev, @@ -5502,6 +5508,15 @@ static struct attribute *intel_pmu_attrs[] = { NULL, }; +static umode_t +default_is_visible(struct kobject *kobj, struct attribute *attr, int i) +{ + if (attr == &dev_attr_allow_tsx_force_abort.attr) + return x86_pmu.flags & PMU_FL_TFA ? attr->mode : 0; + + return attr->mode; +} + static umode_t tsx_is_visible(struct kobject *kobj, struct attribute *attr, int i) { @@ -5523,27 +5538,12 @@ mem_is_visible(struct kobject *kobj, struct attribute *attr, int i) return pebs_is_visible(kobj, attr, i); } -static umode_t -lbr_is_visible(struct kobject *kobj, struct attribute *attr, int i) -{ - return x86_pmu.lbr_nr ? attr->mode : 0; -} - static umode_t exra_is_visible(struct kobject *kobj, struct attribute *attr, int i) { return x86_pmu.version >= 2 ? attr->mode : 0; } -static umode_t -default_is_visible(struct kobject *kobj, struct attribute *attr, int i) -{ - if (attr == &dev_attr_allow_tsx_force_abort.attr) - return x86_pmu.flags & PMU_FL_TFA ? attr->mode : 0; - - return attr->mode; -} - static struct attribute_group group_events_td = { .name = "events", }; -- Gitee From b72acbab048730d14734fed8cb1d325b0cb4c7d2 Mon Sep 17 00:00:00 2001 From: Kan Liang Date: Wed, 25 Oct 2023 13:16:23 -0700 Subject: [PATCH 5/5] perf/x86/intel: Support branch counters logging Upstream commit: 33744916196b4ed7a50f6f47af7c3ad46b730ce6 Conflict: none The branch counters logging (A.K.A LBR event logging) introduces a per-counter indication of precise event occurrences in LBRs. It can provide a means to attribute exposed retirement latency to combinations of events across a block of instructions. It also provides a means of attributing Timed LBR latencies to events. The feature is first introduced on SRF/GRR. It is an enhancement of the ARCH LBR. It adds new fields in the LBR_INFO MSRs to log the occurrences of events on the GP counters. The information is displayed by the order of counters. The design proposed in this patch requires that the events which are logged must be in a group with the event that has LBR. If there are more than one LBR group, the counters logging information only from the current group (overflowed) are stored for the perf tool, otherwise the perf tool cannot know which and when other groups are scheduled especially when multiplexing is triggered. The user can ensure it uses the maximum number of counters that support LBR info (4 by now) by making the group large enough. The HW only logs events by the order of counters. The order may be different from the order of enabling which the perf tool can understand. When parsing the information of each branch entry, convert the counter order to the enabled order, and store the enabled order in the extension space. Unconditionally reset LBRs for an LBR event group when it's deleted. The logged counter information is only valid for the current LBR group. If another LBR group is scheduled later, the information from the stale LBRs would be otherwise wrongly interpreted. Add a sanity check in intel_pmu_hw_config(). Disable the feature if other counter filters (inv, cmask, edge, in_tx) are set or LBR call stack mode is enabled. (For the LBR call stack mode, we cannot simply flush the LBR, since it will break the call stack. Also, there is no obvious usage with the call stack mode for now.) Only applying the PERF_SAMPLE_BRANCH_COUNTERS doesn't require any branch stack setup. Expose the maximum number of supported counters and the width of the counters into the sysfs. The perf tool can use the information to parse the logged counters in each branch. Intel-SIG: commit 33744916196b perf/x86/intel: Support branch counters logging Backport LBR branch counter support to kernel v6.6. Signed-off-by: Kan Liang Signed-off-by: Peter Zijlstra (Intel) Link: https://lkml.kernel.org/r/20231025201626.3000228-5-kan.liang@linux.intel.com [ Yunying Sun: amend commit log ] Signed-off-by: Yunying Sun --- arch/x86/events/intel/core.c | 103 +++++++++++++++++++++++++++-- arch/x86/events/intel/ds.c | 2 +- arch/x86/events/intel/lbr.c | 85 +++++++++++++++++++++++- arch/x86/events/perf_event.h | 12 ++++ arch/x86/events/perf_event_flags.h | 1 + arch/x86/include/asm/msr-index.h | 5 ++ arch/x86/include/asm/perf_event.h | 4 ++ include/uapi/linux/perf_event.h | 3 + 8 files changed, 207 insertions(+), 8 deletions(-) diff --git a/arch/x86/events/intel/core.c b/arch/x86/events/intel/core.c index 6a8c50bd0608..5f506d02caf7 100644 --- a/arch/x86/events/intel/core.c +++ b/arch/x86/events/intel/core.c @@ -2803,6 +2803,7 @@ static void intel_pmu_enable_fixed(struct perf_event *event) static void intel_pmu_enable_event(struct perf_event *event) { + u64 enable_mask = ARCH_PERFMON_EVENTSEL_ENABLE; struct hw_perf_event *hwc = &event->hw; int idx = hwc->idx; @@ -2811,8 +2812,10 @@ static void intel_pmu_enable_event(struct perf_event *event) switch (idx) { case 0 ... INTEL_PMC_IDX_FIXED - 1: + if (branch_sample_counters(event)) + enable_mask |= ARCH_PERFMON_EVENTSEL_BR_CNTR; intel_set_masks(event, idx); - __x86_pmu_enable_event(hwc, ARCH_PERFMON_EVENTSEL_ENABLE); + __x86_pmu_enable_event(hwc, enable_mask); break; case INTEL_PMC_IDX_FIXED ... INTEL_PMC_IDX_FIXED_BTS - 1: case INTEL_PMC_IDX_METRIC_BASE ... INTEL_PMC_IDX_METRIC_END: @@ -3063,7 +3066,7 @@ static int handle_pmi_common(struct pt_regs *regs, u64 status) perf_sample_data_init(&data, 0, event->hw.last_period); if (has_branch_stack(event)) - perf_sample_save_brstack(&data, event, &cpuc->lbr_stack, NULL); + intel_pmu_lbr_save_brstack(&data, cpuc, event); if (perf_event_overflow(event, &data, regs)) x86_pmu_stop(event, 0); @@ -3628,6 +3631,13 @@ intel_get_event_constraints(struct cpu_hw_events *cpuc, int idx, if (cpuc->excl_cntrs) return intel_get_excl_constraints(cpuc, event, idx, c2); + /* Not all counters support the branch counter feature. */ + if (branch_sample_counters(event)) { + c2 = dyn_constraint(cpuc, c2, idx); + c2->idxmsk64 &= x86_pmu.lbr_counters; + c2->weight = hweight64(c2->idxmsk64); + } + return c2; } @@ -3916,6 +3926,58 @@ static int intel_pmu_hw_config(struct perf_event *event) if (needs_branch_stack(event) && is_sampling_event(event)) event->hw.flags |= PERF_X86_EVENT_NEEDS_BRANCH_STACK; + if (branch_sample_counters(event)) { + struct perf_event *leader, *sibling; + int num = 0; + + if (!(x86_pmu.flags & PMU_FL_BR_CNTR) || + (event->attr.config & ~INTEL_ARCH_EVENT_MASK)) + return -EINVAL; + + /* + * The branch counter logging is not supported in the call stack + * mode yet, since we cannot simply flush the LBR during e.g., + * multiplexing. Also, there is no obvious usage with the call + * stack mode. Simply forbids it for now. + * + * If any events in the group enable the branch counter logging + * feature, the group is treated as a branch counter logging + * group, which requires the extra space to store the counters. + */ + leader = event->group_leader; + if (branch_sample_call_stack(leader)) + return -EINVAL; + if (branch_sample_counters(leader)) + num++; + leader->hw.flags |= PERF_X86_EVENT_BRANCH_COUNTERS; + + for_each_sibling_event(sibling, leader) { + if (branch_sample_call_stack(sibling)) + return -EINVAL; + if (branch_sample_counters(sibling)) + num++; + } + + if (num > fls(x86_pmu.lbr_counters)) + return -EINVAL; + /* + * Only applying the PERF_SAMPLE_BRANCH_COUNTERS doesn't + * require any branch stack setup. + * Clear the bit to avoid unnecessary branch stack setup. + */ + if (0 == (event->attr.branch_sample_type & + ~(PERF_SAMPLE_BRANCH_PLM_ALL | + PERF_SAMPLE_BRANCH_COUNTERS))) + event->hw.flags &= ~PERF_X86_EVENT_NEEDS_BRANCH_STACK; + + /* + * Force the leader to be a LBR event. So LBRs can be reset + * with the leader event. See intel_pmu_lbr_del() for details. + */ + if (!intel_pmu_needs_branch_stack(leader)) + return -EINVAL; + } + if (intel_pmu_needs_branch_stack(event)) { ret = intel_pmu_setup_lbr_filter(event); if (ret) @@ -4394,8 +4456,13 @@ cmt_get_event_constraints(struct cpu_hw_events *cpuc, int idx, */ if (event->attr.precise_ip == 3) { /* Force instruction:ppp on PMC0, 1 and Fixed counter 0 */ - if (constraint_match(&fixed0_constraint, event->hw.config)) - return &fixed0_counter0_1_constraint; + if (constraint_match(&fixed0_constraint, event->hw.config)) { + /* The fixed counter 0 doesn't support LBR event logging. */ + if (branch_sample_counters(event)) + return &counter0_1_constraint; + else + return &fixed0_counter0_1_constraint; + } switch (c->idxmsk64 & 0x3ull) { case 0x1: @@ -4574,7 +4641,7 @@ int intel_cpuc_prepare(struct cpu_hw_events *cpuc, int cpu) goto err; } - if (x86_pmu.flags & (PMU_FL_EXCL_CNTRS | PMU_FL_TFA)) { + if (x86_pmu.flags & (PMU_FL_EXCL_CNTRS | PMU_FL_TFA | PMU_FL_BR_CNTR)) { size_t sz = X86_PMC_IDX_MAX * sizeof(struct event_constraint); cpuc->constraint_list = kzalloc_node(sz, GFP_KERNEL, cpu_to_node(cpu)); @@ -5471,15 +5538,39 @@ static ssize_t branches_show(struct device *cdev, static DEVICE_ATTR_RO(branches); +static ssize_t branch_counter_nr_show(struct device *cdev, + struct device_attribute *attr, + char *buf) +{ + return snprintf(buf, PAGE_SIZE, "%d\n", fls(x86_pmu.lbr_counters)); +} + +static DEVICE_ATTR_RO(branch_counter_nr); + +static ssize_t branch_counter_width_show(struct device *cdev, + struct device_attribute *attr, + char *buf) +{ + return snprintf(buf, PAGE_SIZE, "%d\n", LBR_INFO_BR_CNTR_BITS); +} + +static DEVICE_ATTR_RO(branch_counter_width); + static struct attribute *lbr_attrs[] = { &dev_attr_branches.attr, + &dev_attr_branch_counter_nr.attr, + &dev_attr_branch_counter_width.attr, NULL }; static umode_t lbr_is_visible(struct kobject *kobj, struct attribute *attr, int i) { - return x86_pmu.lbr_nr ? attr->mode : 0; + /* branches */ + if (i == 0) + return x86_pmu.lbr_nr ? attr->mode : 0; + + return (x86_pmu.flags & PMU_FL_BR_CNTR) ? attr->mode : 0; } static char pmu_name_str[30]; diff --git a/arch/x86/events/intel/ds.c b/arch/x86/events/intel/ds.c index 7566190389f0..5ff81dbf8aa3 100644 --- a/arch/x86/events/intel/ds.c +++ b/arch/x86/events/intel/ds.c @@ -1912,7 +1912,7 @@ static void setup_pebs_adaptive_sample_data(struct perf_event *event, if (has_branch_stack(event)) { intel_pmu_store_pebs_lbrs(lbr); - perf_sample_save_brstack(data, event, &cpuc->lbr_stack, NULL); + intel_pmu_lbr_save_brstack(data, cpuc, event); } } diff --git a/arch/x86/events/intel/lbr.c b/arch/x86/events/intel/lbr.c index c3b0d15a9841..78cd5084104e 100644 --- a/arch/x86/events/intel/lbr.c +++ b/arch/x86/events/intel/lbr.c @@ -676,6 +676,25 @@ void intel_pmu_lbr_del(struct perf_event *event) WARN_ON_ONCE(cpuc->lbr_users < 0); WARN_ON_ONCE(cpuc->lbr_pebs_users < 0); perf_sched_cb_dec(event->pmu); + + /* + * The logged occurrences information is only valid for the + * current LBR group. If another LBR group is scheduled in + * later, the information from the stale LBRs will be wrongly + * interpreted. Reset the LBRs here. + * + * Only clear once for a branch counter group with the leader + * event. Because + * - Cannot simply reset the LBRs with the !cpuc->lbr_users. + * Because it's possible that the last LBR user is not in a + * branch counter group, e.g., a branch_counters group + + * several normal LBR events. + * - The LBR reset can be done with any one of the events in a + * branch counter group, since they are always scheduled together. + * It's easy to force the leader event an LBR event. + */ + if (is_branch_counters_group(event) && event == event->group_leader) + intel_pmu_lbr_reset(); } static inline bool vlbr_exclude_host(void) @@ -866,6 +885,8 @@ static __always_inline u16 get_lbr_cycles(u64 info) return cycles; } +static_assert((64 - PERF_BRANCH_ENTRY_INFO_BITS_MAX) > LBR_INFO_BR_CNTR_NUM * LBR_INFO_BR_CNTR_BITS); + static void intel_pmu_store_lbr(struct cpu_hw_events *cpuc, struct lbr_entry *entries) { @@ -898,11 +919,67 @@ static void intel_pmu_store_lbr(struct cpu_hw_events *cpuc, e->abort = !!(info & LBR_INFO_ABORT); e->cycles = get_lbr_cycles(info); e->type = get_lbr_br_type(info); + + /* + * Leverage the reserved field of cpuc->lbr_entries[i] to + * temporarily store the branch counters information. + * The later code will decide what content can be disclosed + * to the perf tool. Pleae see intel_pmu_lbr_counters_reorder(). + */ + e->reserved = (info >> LBR_INFO_BR_CNTR_OFFSET) & LBR_INFO_BR_CNTR_FULL_MASK; } cpuc->lbr_stack.nr = i; } +/* + * The enabled order may be different from the counter order. + * Update the lbr_counters with the enabled order. + */ +static void intel_pmu_lbr_counters_reorder(struct cpu_hw_events *cpuc, + struct perf_event *event) +{ + int i, j, pos = 0, order[X86_PMC_IDX_MAX]; + struct perf_event *leader, *sibling; + u64 src, dst, cnt; + + leader = event->group_leader; + if (branch_sample_counters(leader)) + order[pos++] = leader->hw.idx; + + for_each_sibling_event(sibling, leader) { + if (!branch_sample_counters(sibling)) + continue; + order[pos++] = sibling->hw.idx; + } + + WARN_ON_ONCE(!pos); + + for (i = 0; i < cpuc->lbr_stack.nr; i++) { + src = cpuc->lbr_entries[i].reserved; + dst = 0; + for (j = 0; j < pos; j++) { + cnt = (src >> (order[j] * LBR_INFO_BR_CNTR_BITS)) & LBR_INFO_BR_CNTR_MASK; + dst |= cnt << j * LBR_INFO_BR_CNTR_BITS; + } + cpuc->lbr_counters[i] = dst; + cpuc->lbr_entries[i].reserved = 0; + } +} + +void intel_pmu_lbr_save_brstack(struct perf_sample_data *data, + struct cpu_hw_events *cpuc, + struct perf_event *event) +{ + if (is_branch_counters_group(event)) { + intel_pmu_lbr_counters_reorder(cpuc, event); + perf_sample_save_brstack(data, event, &cpuc->lbr_stack, cpuc->lbr_counters); + return; + } + + perf_sample_save_brstack(data, event, &cpuc->lbr_stack, NULL); +} + static void intel_pmu_arch_lbr_read(struct cpu_hw_events *cpuc) { intel_pmu_store_lbr(cpuc, NULL); @@ -1173,8 +1250,10 @@ intel_pmu_lbr_filter(struct cpu_hw_events *cpuc) for (i = 0; i < cpuc->lbr_stack.nr; ) { if (!cpuc->lbr_entries[i].from) { j = i; - while (++j < cpuc->lbr_stack.nr) + while (++j < cpuc->lbr_stack.nr) { cpuc->lbr_entries[j-1] = cpuc->lbr_entries[j]; + cpuc->lbr_counters[j-1] = cpuc->lbr_counters[j]; + } cpuc->lbr_stack.nr--; if (!cpuc->lbr_entries[i].from) continue; @@ -1525,8 +1604,12 @@ void __init intel_pmu_arch_lbr_init(void) x86_pmu.lbr_mispred = ecx.split.lbr_mispred; x86_pmu.lbr_timed_lbr = ecx.split.lbr_timed_lbr; x86_pmu.lbr_br_type = ecx.split.lbr_br_type; + x86_pmu.lbr_counters = ecx.split.lbr_counters; x86_pmu.lbr_nr = lbr_nr; + if (!!x86_pmu.lbr_counters) + x86_pmu.flags |= PMU_FL_BR_CNTR; + if (x86_pmu.lbr_mispred) static_branch_enable(&x86_lbr_mispred); if (x86_pmu.lbr_timed_lbr) diff --git a/arch/x86/events/perf_event.h b/arch/x86/events/perf_event.h index c8ba2be7585d..b8a2d3ba4ccd 100644 --- a/arch/x86/events/perf_event.h +++ b/arch/x86/events/perf_event.h @@ -110,6 +110,11 @@ static inline bool is_topdown_event(struct perf_event *event) return is_metric_event(event) || is_slots_event(event); } +static inline bool is_branch_counters_group(struct perf_event *event) +{ + return event->group_leader->hw.flags & PERF_X86_EVENT_BRANCH_COUNTERS; +} + struct amd_nb { int nb_id; /* NorthBridge id */ int refcnt; /* reference count */ @@ -283,6 +288,7 @@ struct cpu_hw_events { int lbr_pebs_users; struct perf_branch_stack lbr_stack; struct perf_branch_entry lbr_entries[MAX_LBR_ENTRIES]; + u64 lbr_counters[MAX_LBR_ENTRIES]; /* branch stack extra */ union { struct er_account *lbr_sel; struct er_account *lbr_ctl; @@ -881,6 +887,7 @@ struct x86_pmu { unsigned int lbr_mispred:1; unsigned int lbr_timed_lbr:1; unsigned int lbr_br_type:1; + unsigned int lbr_counters:4; void (*lbr_reset)(void); void (*lbr_read)(struct cpu_hw_events *cpuc); @@ -1005,6 +1012,7 @@ do { \ #define PMU_FL_INSTR_LATENCY 0x80 /* Support Instruction Latency in PEBS Memory Info Record */ #define PMU_FL_MEM_LOADS_AUX 0x100 /* Require an auxiliary event for the complete memory info */ #define PMU_FL_RETIRE_LATENCY 0x200 /* Support Retire Latency in PEBS */ +#define PMU_FL_BR_CNTR 0x400 /* Support branch counter logging */ #define EVENT_VAR(_id) event_attr_##_id #define EVENT_PTR(_id) &event_attr_##_id.attr.attr @@ -1545,6 +1553,10 @@ void intel_pmu_store_pebs_lbrs(struct lbr_entry *lbr); void intel_ds_init(void); +void intel_pmu_lbr_save_brstack(struct perf_sample_data *data, + struct cpu_hw_events *cpuc, + struct perf_event *event); + void intel_pmu_lbr_swap_task_ctx(struct perf_event_pmu_context *prev_epc, struct perf_event_pmu_context *next_epc); diff --git a/arch/x86/events/perf_event_flags.h b/arch/x86/events/perf_event_flags.h index a1685981c520..6c977c19f2cd 100644 --- a/arch/x86/events/perf_event_flags.h +++ b/arch/x86/events/perf_event_flags.h @@ -21,3 +21,4 @@ PERF_ARCH(PEBS_STLAT, 0x08000) /* st+stlat data address sampling */ PERF_ARCH(AMD_BRS, 0x10000) /* AMD Branch Sampling */ PERF_ARCH(PEBS_LAT_HYBRID, 0x20000) /* ld and st lat for hybrid */ PERF_ARCH(NEEDS_BRANCH_STACK, 0x40000) /* require branch stack setup */ +PERF_ARCH(BRANCH_COUNTERS, 0x80000) /* logs the counters in the extra space of each branch */ diff --git a/arch/x86/include/asm/msr-index.h b/arch/x86/include/asm/msr-index.h index 389f9594746e..92115b1c28eb 100644 --- a/arch/x86/include/asm/msr-index.h +++ b/arch/x86/include/asm/msr-index.h @@ -236,6 +236,11 @@ #define LBR_INFO_CYCLES 0xffff #define LBR_INFO_BR_TYPE_OFFSET 56 #define LBR_INFO_BR_TYPE (0xfull << LBR_INFO_BR_TYPE_OFFSET) +#define LBR_INFO_BR_CNTR_OFFSET 32 +#define LBR_INFO_BR_CNTR_NUM 4 +#define LBR_INFO_BR_CNTR_BITS 2 +#define LBR_INFO_BR_CNTR_MASK GENMASK_ULL(LBR_INFO_BR_CNTR_BITS - 1, 0) +#define LBR_INFO_BR_CNTR_FULL_MASK GENMASK_ULL(LBR_INFO_BR_CNTR_NUM * LBR_INFO_BR_CNTR_BITS - 1, 0) #define MSR_ARCH_LBR_CTL 0x000014ce #define ARCH_LBR_CTL_LBREN BIT(0) diff --git a/arch/x86/include/asm/perf_event.h b/arch/x86/include/asm/perf_event.h index 85a9fd5a3ec3..4decedf2c077 100644 --- a/arch/x86/include/asm/perf_event.h +++ b/arch/x86/include/asm/perf_event.h @@ -31,6 +31,7 @@ #define ARCH_PERFMON_EVENTSEL_ENABLE (1ULL << 22) #define ARCH_PERFMON_EVENTSEL_INV (1ULL << 23) #define ARCH_PERFMON_EVENTSEL_CMASK 0xFF000000ULL +#define ARCH_PERFMON_EVENTSEL_BR_CNTR (1ULL << 35) #define INTEL_FIXED_BITS_MASK 0xFULL #define INTEL_FIXED_BITS_STRIDE 4 @@ -216,6 +217,9 @@ union cpuid28_ecx { unsigned int lbr_timed_lbr:1; /* Branch Type Field Supported */ unsigned int lbr_br_type:1; + unsigned int reserved:13; + /* Branch counters (Event Logging) Supported */ + unsigned int lbr_counters:4; } split; unsigned int full; }; diff --git a/include/uapi/linux/perf_event.h b/include/uapi/linux/perf_event.h index 4461f380425b..3a64499b0f5d 100644 --- a/include/uapi/linux/perf_event.h +++ b/include/uapi/linux/perf_event.h @@ -1437,6 +1437,9 @@ struct perf_branch_entry { reserved:31; }; +/* Size of used info bits in struct perf_branch_entry */ +#define PERF_BRANCH_ENTRY_INFO_BITS_MAX 33 + union perf_sample_weight { __u64 full; #if defined(__LITTLE_ENDIAN_BITFIELD) -- Gitee