From 97c79a38cd454602645f0470ffb444b3b75ce574 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Thu, 28 Apr 2016 13:16:33 -0300 Subject: perf core: Per event callchain limit Additionally to being able to control the system wide maximum depth via /proc/sys/kernel/perf_event_max_stack, now we are able to ask for different depths per event, using perf_event_attr.sample_max_stack for that. This uses an u16 hole at the end of perf_event_attr, that, when perf_event_attr.sample_type has the PERF_SAMPLE_CALLCHAIN, if sample_max_stack is zero, means use perf_event_max_stack, otherwise it'll be bounds checked under callchain_mutex. Cc: Adrian Hunter Cc: Alexander Shishkin Cc: Alexei Starovoitov Cc: Brendan Gregg Cc: David Ahern Cc: Frederic Weisbecker Cc: He Kuang Cc: Jiri Olsa Cc: Linus Torvalds Cc: Masami Hiramatsu Cc: Milian Wolff Cc: Namhyung Kim Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Vince Weaver Cc: Wang Nan Cc: Zefan Li Link: http://lkml.kernel.org/n/tip-kolmn1yo40p7jhswxwrc7rrd@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- include/linux/perf_event.h | 2 +- include/uapi/linux/perf_event.h | 6 +++++- 2 files changed, 6 insertions(+), 2 deletions(-) (limited to 'include') diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 6b87be908790..0e43355c7aad 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -1076,7 +1076,7 @@ extern void perf_callchain_kernel(struct perf_callchain_entry_ctx *entry, struct extern struct perf_callchain_entry * get_perf_callchain(struct pt_regs *regs, u32 init_nr, bool kernel, bool user, u32 max_stack, bool crosstask, bool add_mark); -extern int get_callchain_buffers(void); +extern int get_callchain_buffers(int max_stack); extern void put_callchain_buffers(void); extern int sysctl_perf_event_max_stack; diff --git a/include/uapi/linux/perf_event.h b/include/uapi/linux/perf_event.h index 36ce552cf6a9..c66a485a24ac 100644 --- a/include/uapi/linux/perf_event.h +++ b/include/uapi/linux/perf_event.h @@ -276,6 +276,9 @@ enum perf_event_read_format { /* * Hardware event_id to monitor via a performance monitoring event: + * + * @sample_max_stack: Max number of frame pointers in a callchain, + * should be < /proc/sys/kernel/perf_event_max_stack */ struct perf_event_attr { @@ -385,7 +388,8 @@ struct perf_event_attr { * Wakeup watermark for AUX area */ __u32 aux_watermark; - __u32 __reserved_2; /* align to __u64 */ + __u16 sample_max_stack; + __u16 __reserved_2; /* align to __u64 */ }; #define perf_flags(attr) (*(&(attr)->read_format + 1)) -- cgit v1.2.3 From f2fb6bef92514432398a653df1c2f1041d79ac46 Mon Sep 17 00:00:00 2001 From: Kan Liang Date: Wed, 23 Mar 2016 11:24:37 -0700 Subject: perf/core: Optimize side-band event delivery The perf_event_aux() function iterates all PMUs and all events in their respective per-CPU contexts to find the events to deliver side-band records to. For example, the brk test case in lkp triggers many mmap() operations, which, if we're also running perf, results in many perf_event_aux() invocations. If we enable uncore PMU support (even when uncore events are not used), dozens of uncore PMUs will be iterated, which can significantly decrease brk_test's throughput. For example, the brk throughput: without uncore PMUs: 2647573 ops_per_sec with uncore PMUs: 1768444 ops_per_sec ... a 33% reduction. To get at the per-CPU events that need side-band records, this patch puts these events on a per-CPU list, this avoids iterating the PMUs and any events that do not need side-band records. Per task events are unchanged to avoid extra overhead on the context switch paths. Suggested-by: Peter Zijlstra (Intel) Reported-by: Huang, Ying Signed-off-by: Kan Liang Signed-off-by: Peter Zijlstra (Intel) Cc: Alexander Shishkin Cc: Arnaldo Carvalho de Melo Cc: Jiri Olsa Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Vince Weaver Link: http://lkml.kernel.org/r/1458757477-3781-1-git-send-email-kan.liang@intel.com Signed-off-by: Ingo Molnar --- include/linux/perf_event.h | 6 ++++ kernel/events/core.c | 85 +++++++++++++++++++++++++++++++++++++++------- 2 files changed, 79 insertions(+), 12 deletions(-) (limited to 'include') diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 0e43355c7aad..92e9ce737432 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -517,6 +517,11 @@ struct swevent_hlist { struct perf_cgroup; struct ring_buffer; +struct pmu_event_list { + raw_spinlock_t lock; + struct list_head list; +}; + /** * struct perf_event - performance event kernel representation: */ @@ -675,6 +680,7 @@ struct perf_event { int cgrp_defer_enabled; #endif + struct list_head sb_list; #endif /* CONFIG_PERF_EVENTS */ }; diff --git a/kernel/events/core.c b/kernel/events/core.c index 79363f298445..6615c8922ee3 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -335,6 +335,7 @@ 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; static atomic_t nr_comm_events __read_mostly; @@ -3665,6 +3666,26 @@ static void free_event_rcu(struct rcu_head *head) static void ring_buffer_attach(struct perf_event *event, struct ring_buffer *rb); +static void detach_sb_event(struct perf_event *event) +{ + struct pmu_event_list *pel = per_cpu_ptr(&pmu_sb_events, event->cpu); + + raw_spin_lock(&pel->lock); + list_del_rcu(&event->sb_list); + raw_spin_unlock(&pel->lock); +} + +static void unaccount_pmu_sb_event(struct perf_event *event) +{ + if (event->parent) + return; + + if (event->attach_state & PERF_ATTACH_TASK) + return; + + detach_sb_event(event); +} + static void unaccount_event_cpu(struct perf_event *event, int cpu) { if (event->parent) @@ -3728,6 +3749,8 @@ static void unaccount_event(struct perf_event *event) } unaccount_event_cpu(event, event->cpu); + + unaccount_pmu_sb_event(event); } static void perf_sched_delayed(struct work_struct *work) @@ -5888,13 +5911,25 @@ perf_event_aux_task_ctx(perf_event_aux_output_cb output, void *data, rcu_read_unlock(); } +static void perf_event_sb_iterate(perf_event_aux_output_cb output, void *data) +{ + struct pmu_event_list *pel = this_cpu_ptr(&pmu_sb_events); + struct perf_event *event; + + list_for_each_entry_rcu(event, &pel->list, sb_list) { + if (event->state < PERF_EVENT_STATE_INACTIVE) + continue; + if (!event_filter_match(event)) + continue; + output(event, data); + } +} + static void perf_event_aux(perf_event_aux_output_cb output, void *data, struct perf_event_context *task_ctx) { - struct perf_cpu_context *cpuctx; struct perf_event_context *ctx; - struct pmu *pmu; int ctxn; /* @@ -5909,20 +5944,15 @@ perf_event_aux(perf_event_aux_output_cb output, void *data, } rcu_read_lock(); - list_for_each_entry_rcu(pmu, &pmus, entry) { - cpuctx = get_cpu_ptr(pmu->pmu_cpu_context); - if (cpuctx->unique_pmu != pmu) - goto next; - perf_event_aux_ctx(&cpuctx->ctx, output, data, false); - ctxn = pmu->task_ctx_nr; - if (ctxn < 0) - goto next; + preempt_disable(); + perf_event_sb_iterate(output, data); + + for_each_task_context_nr(ctxn) { ctx = rcu_dereference(current->perf_event_ctxp[ctxn]); if (ctx) perf_event_aux_ctx(ctx, output, data, false); -next: - put_cpu_ptr(pmu->pmu_cpu_context); } + preempt_enable(); rcu_read_unlock(); } @@ -8615,6 +8645,32 @@ unlock: return pmu; } +static void attach_sb_event(struct perf_event *event) +{ + struct pmu_event_list *pel = per_cpu_ptr(&pmu_sb_events, event->cpu); + + raw_spin_lock(&pel->lock); + list_add_rcu(&event->sb_list, &pel->list); + raw_spin_unlock(&pel->lock); +} + +static void account_pmu_sb_event(struct perf_event *event) +{ + struct perf_event_attr *attr = &event->attr; + + if (event->parent) + return; + + if (event->attach_state & PERF_ATTACH_TASK) + return; + + if (attr->mmap || attr->mmap_data || attr->mmap2 || + attr->comm || attr->comm_exec || + attr->task || + attr->context_switch) + attach_sb_event(event); +} + static void account_event_cpu(struct perf_event *event, int cpu) { if (event->parent) @@ -8695,6 +8751,8 @@ static void account_event(struct perf_event *event) enabled: account_event_cpu(event, event->cpu); + + account_pmu_sb_event(event); } /* @@ -10203,6 +10261,9 @@ static void __init perf_event_init_all_cpus(void) 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)); } } -- cgit v1.2.3 From fc07e9f983b4b11922c22b6cccadc1f342f05a4c Mon Sep 17 00:00:00 2001 From: Andi Kleen Date: Thu, 19 May 2016 17:09:56 -0700 Subject: perf/x86: Support sysfs files depending on SMT status Add a way to show different sysfs events attributes depending on HyperThreading is on or off. This is difficult to determine early at boot, so we just do it dynamically when the sysfs attribute is read. Signed-off-by: Andi Kleen Signed-off-by: Peter Zijlstra (Intel) Cc: Alexander Shishkin Cc: Arnaldo Carvalho de Melo Cc: Jiri Olsa Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Vince Weaver Cc: acme@kernel.org Cc: jolsa@kernel.org Link: http://lkml.kernel.org/r/1463703002-19686-3-git-send-email-andi@firstfloor.org Signed-off-by: Ingo Molnar --- arch/x86/events/core.c | 23 +++++++++++++++++++++++ arch/x86/events/perf_event.h | 10 ++++++++++ include/linux/perf_event.h | 7 +++++++ 3 files changed, 40 insertions(+) (limited to 'include') diff --git a/arch/x86/events/core.c b/arch/x86/events/core.c index 33787ee817f0..929655db5084 100644 --- a/arch/x86/events/core.c +++ b/arch/x86/events/core.c @@ -1622,6 +1622,29 @@ ssize_t events_sysfs_show(struct device *dev, struct device_attribute *attr, cha } EXPORT_SYMBOL_GPL(events_sysfs_show); +ssize_t events_ht_sysfs_show(struct device *dev, struct device_attribute *attr, + char *page) +{ + struct perf_pmu_events_ht_attr *pmu_attr = + container_of(attr, struct perf_pmu_events_ht_attr, attr); + + /* + * Report conditional events depending on Hyper-Threading. + * + * This is overly conservative as usually the HT special + * handling is not needed if the other CPU thread is idle. + * + * Note this does not (and cannot) handle the case when thread + * siblings are invisible, for example with virtualization + * if they are owned by some other guest. The user tool + * has to re-read when a thread sibling gets onlined later. + */ + return sprintf(page, "%s", + topology_max_smt_threads() > 1 ? + pmu_attr->event_str_ht : + pmu_attr->event_str_noht); +} + EVENT_ATTR(cpu-cycles, CPU_CYCLES ); EVENT_ATTR(instructions, INSTRUCTIONS ); EVENT_ATTR(cache-references, CACHE_REFERENCES ); diff --git a/arch/x86/events/perf_event.h b/arch/x86/events/perf_event.h index 8bd764df815d..e2d7285a2dac 100644 --- a/arch/x86/events/perf_event.h +++ b/arch/x86/events/perf_event.h @@ -668,6 +668,14 @@ static struct perf_pmu_events_attr event_attr_##v = { \ .event_str = str, \ }; +#define EVENT_ATTR_STR_HT(_name, v, noht, ht) \ +static struct perf_pmu_events_ht_attr event_attr_##v = { \ + .attr = __ATTR(_name, 0444, events_ht_sysfs_show, NULL),\ + .id = 0, \ + .event_str_noht = noht, \ + .event_str_ht = ht, \ +} + extern struct x86_pmu x86_pmu __read_mostly; static inline bool x86_pmu_has_lbr_callstack(void) @@ -803,6 +811,8 @@ struct attribute **merge_attr(struct attribute **a, struct attribute **b); ssize_t events_sysfs_show(struct device *dev, struct device_attribute *attr, char *page); +ssize_t events_ht_sysfs_show(struct device *dev, struct device_attribute *attr, + char *page); #ifdef CONFIG_CPU_SUP_AMD diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 92e9ce737432..a7593d653b40 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -1334,6 +1334,13 @@ struct perf_pmu_events_attr { const char *event_str; }; +struct perf_pmu_events_ht_attr { + struct device_attribute attr; + u64 id; + const char *event_str_ht; + const char *event_str_noht; +}; + ssize_t perf_event_sysfs_show(struct device *dev, struct device_attribute *attr, char *page); -- cgit v1.2.3 From 2c95afc1e83d93fac3be6923465e1753c2c53b0a Mon Sep 17 00:00:00 2001 From: Andi Kleen Date: Thu, 9 Jun 2016 06:14:38 -0700 Subject: perf/x86/intel, watchdog: Switch NMI watchdog to ref cycles on x86 The NMI watchdog uses either the fixed cycles or a generic cycles counter. This causes a lot of conflicts with users of the PMU who want to run a full group including the cycles fixed counter, for example the --topdown support recently added to perf stat. The code needs to fall back to not use groups, which can cause measurement inaccuracy due to multiplexing errors. This patch switches the NMI watchdog to use reference cycles on Intel systems. This is actually more accurate than cycles, because cycles can tick faster than the measured CPU Frequency due to Turbo mode. The ref cycles always tick at their frequency, or slower when the system is idling. That means the NMI watchdog can never expire too early, unlike with cycles. The reference cycles tick roughly at the frequency of the TSC, so the same period computation can be used. Signed-off-by: Andi Kleen Signed-off-by: Peter Zijlstra (Intel) Cc: Alexander Shishkin Cc: Arnaldo Carvalho de Melo Cc: Jiri Olsa Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Vince Weaver Cc: acme@kernel.org Cc: jolsa@kernel.org Link: http://lkml.kernel.org/r/1465478079-19993-1-git-send-email-andi@firstfloor.org Signed-off-by: Ingo Molnar --- arch/x86/kernel/apic/hw_nmi.c | 8 ++++++++ include/linux/nmi.h | 1 + kernel/watchdog.c | 7 +++++++ 3 files changed, 16 insertions(+) (limited to 'include') diff --git a/arch/x86/kernel/apic/hw_nmi.c b/arch/x86/kernel/apic/hw_nmi.c index 7788ce643bf4..016f4263fad4 100644 --- a/arch/x86/kernel/apic/hw_nmi.c +++ b/arch/x86/kernel/apic/hw_nmi.c @@ -18,8 +18,16 @@ #include #include #include +#include #ifdef CONFIG_HARDLOCKUP_DETECTOR +int hw_nmi_get_event(void) +{ + if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL) + return PERF_COUNT_HW_REF_CPU_CYCLES; + return PERF_COUNT_HW_CPU_CYCLES; +} + u64 hw_nmi_get_sample_period(int watchdog_thresh) { return (u64)(cpu_khz) * 1000 * watchdog_thresh; diff --git a/include/linux/nmi.h b/include/linux/nmi.h index 4630eeae18e0..79858af27209 100644 --- a/include/linux/nmi.h +++ b/include/linux/nmi.h @@ -66,6 +66,7 @@ static inline bool trigger_allbutself_cpu_backtrace(void) #ifdef CONFIG_LOCKUP_DETECTOR u64 hw_nmi_get_sample_period(int watchdog_thresh); +int hw_nmi_get_event(void); extern int nmi_watchdog_enabled; extern int soft_watchdog_enabled; extern int watchdog_user_enabled; diff --git a/kernel/watchdog.c b/kernel/watchdog.c index 9acb29f280ec..8dd30fcd91be 100644 --- a/kernel/watchdog.c +++ b/kernel/watchdog.c @@ -315,6 +315,12 @@ static int is_softlockup(unsigned long touch_ts) #ifdef CONFIG_HARDLOCKUP_DETECTOR +/* Can be overriden by architecture */ +__weak int hw_nmi_get_event(void) +{ + return PERF_COUNT_HW_CPU_CYCLES; +} + static struct perf_event_attr wd_hw_attr = { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CPU_CYCLES, @@ -604,6 +610,7 @@ static int watchdog_nmi_enable(unsigned int cpu) wd_attr = &wd_hw_attr; wd_attr->sample_period = hw_nmi_get_sample_period(watchdog_thresh); + wd_attr->config = hw_nmi_get_event(); /* Try to register using hardware perf events */ event = perf_event_create_kernel_counter(wd_attr, cpu, NULL, watchdog_overflow_callback, NULL); -- cgit v1.2.3 From 44530d588e142a96cf0cd345a7cb8911c4f88720 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Sun, 10 Jul 2016 20:58:36 +0200 Subject: Revert "perf/x86/intel, watchdog: Switch NMI watchdog to ref cycles on x86" This reverts commit 2c95afc1e83d93fac3be6923465e1753c2c53b0a. Stephane reported the following regression: > Since Andi added: > > commit 2c95afc1e83d93fac3be6923465e1753c2c53b0a > Author: Andi Kleen > Date: Thu Jun 9 06:14:38 2016 -0700 > > perf/x86/intel, watchdog: Switch NMI watchdog to ref cycles on x86 > > $ perf stat -e ref-cycles ls > .... > > fails systematically because the ref-cycles is now used by the > watchdog and given this is a system-wide pinned event, it monopolizes > the fixed counter 2 which is the only counter able to measure this event. Since the next merge window is near, fix the regression for now by reverting the commit. Reported-by: Stephane Eranian Cc: Andi Kleen Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Jiri Olsa Cc: Vince Weaver Cc: Alexander Shishkin Cc: Linus Torvalds Cc: Thomas Gleixner Cc: linux-kernel@vger.kernel.org Signed-off-by: Ingo Molnar --- arch/x86/kernel/apic/hw_nmi.c | 8 -------- include/linux/nmi.h | 1 - kernel/watchdog.c | 7 ------- 3 files changed, 16 deletions(-) (limited to 'include') diff --git a/arch/x86/kernel/apic/hw_nmi.c b/arch/x86/kernel/apic/hw_nmi.c index 016f4263fad4..7788ce643bf4 100644 --- a/arch/x86/kernel/apic/hw_nmi.c +++ b/arch/x86/kernel/apic/hw_nmi.c @@ -18,16 +18,8 @@ #include #include #include -#include #ifdef CONFIG_HARDLOCKUP_DETECTOR -int hw_nmi_get_event(void) -{ - if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL) - return PERF_COUNT_HW_REF_CPU_CYCLES; - return PERF_COUNT_HW_CPU_CYCLES; -} - u64 hw_nmi_get_sample_period(int watchdog_thresh) { return (u64)(cpu_khz) * 1000 * watchdog_thresh; diff --git a/include/linux/nmi.h b/include/linux/nmi.h index 79858af27209..4630eeae18e0 100644 --- a/include/linux/nmi.h +++ b/include/linux/nmi.h @@ -66,7 +66,6 @@ static inline bool trigger_allbutself_cpu_backtrace(void) #ifdef CONFIG_LOCKUP_DETECTOR u64 hw_nmi_get_sample_period(int watchdog_thresh); -int hw_nmi_get_event(void); extern int nmi_watchdog_enabled; extern int soft_watchdog_enabled; extern int watchdog_user_enabled; diff --git a/kernel/watchdog.c b/kernel/watchdog.c index 8dd30fcd91be..9acb29f280ec 100644 --- a/kernel/watchdog.c +++ b/kernel/watchdog.c @@ -315,12 +315,6 @@ static int is_softlockup(unsigned long touch_ts) #ifdef CONFIG_HARDLOCKUP_DETECTOR -/* Can be overriden by architecture */ -__weak int hw_nmi_get_event(void) -{ - return PERF_COUNT_HW_CPU_CYCLES; -} - static struct perf_event_attr wd_hw_attr = { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CPU_CYCLES, @@ -610,7 +604,6 @@ static int watchdog_nmi_enable(unsigned int cpu) wd_attr = &wd_hw_attr; wd_attr->sample_period = hw_nmi_get_sample_period(watchdog_thresh); - wd_attr->config = hw_nmi_get_event(); /* Try to register using hardware perf events */ event = perf_event_create_kernel_counter(wd_attr, cpu, NULL, watchdog_overflow_callback, NULL); -- cgit v1.2.3