diff options
Diffstat (limited to 'include/linux/perf_event.h')
| -rw-r--r-- | include/linux/perf_event.h | 221 |
1 files changed, 172 insertions, 49 deletions
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index ee8b9ecdc03b..c6a3bac76966 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -36,6 +36,7 @@ struct perf_guest_info_callbacks { }; #ifdef CONFIG_HAVE_HW_BREAKPOINT +#include <linux/rhashtable-types.h> #include <asm/hw_breakpoint.h> #endif @@ -60,6 +61,7 @@ struct perf_guest_info_callbacks { #include <linux/refcount.h> #include <linux/security.h> #include <linux/static_call.h> +#include <linux/lockdep.h> #include <asm/local.h> struct perf_callchain_entry { @@ -137,9 +139,11 @@ struct hw_perf_event_extra { * PERF_EVENT_FLAG_ARCH bits are reserved for architecture-specific * usage. */ -#define PERF_EVENT_FLAG_ARCH 0x0000ffff +#define PERF_EVENT_FLAG_ARCH 0x000fffff #define PERF_EVENT_FLAG_USER_READ_CNT 0x80000000 +static_assert((PERF_EVENT_FLAG_USER_READ_CNT & PERF_EVENT_FLAG_ARCH) == 0); + /** * struct hw_perf_event - performance event hardware details: */ @@ -178,7 +182,7 @@ struct hw_perf_event { * creation and event initalization. */ struct arch_hw_breakpoint info; - struct list_head bp_list; + struct rhlist_head bp_list; }; #endif struct { /* amd_iommu */ @@ -262,6 +266,7 @@ struct hw_perf_event { }; struct perf_event; +struct perf_event_pmu_context; /* * Common implementation detail of pmu::{start,commit,cancel}_txn @@ -304,7 +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; @@ -439,7 +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); /* @@ -453,8 +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 */ /* @@ -518,9 +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. @@ -631,7 +637,23 @@ struct pmu_event_list { struct list_head list; }; +/* + * event->sibling_list is modified whole holding both ctx->lock and ctx->mutex + * as such iteration must hold either lock. However, since ctx->lock is an IRQ + * safe lock, and is only held by the CPU doing the modification, having IRQs + * disabled is sufficient since it will hold-off the IPIs. + */ +#ifdef CONFIG_PROVE_LOCKING +#define lockdep_assert_event_ctx(event) \ + WARN_ON_ONCE(__lockdep_enabled && \ + (this_cpu_read(hardirqs_enabled) && \ + lockdep_is_held(&(event)->ctx->mutex) != LOCK_STATE_HELD)) +#else +#define lockdep_assert_event_ctx(event) +#endif + #define for_each_sibling_event(sibling, event) \ + lockdep_assert_event_ctx(event); \ if ((event)->group_leader == (event)) \ list_for_each_entry((sibling), &(event)->sibling_list, sibling_list) @@ -675,6 +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; @@ -700,6 +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; /* @@ -736,11 +769,14 @@ struct perf_event { struct fasync_struct *fasync; /* delayed work for NMIs and such */ - int pending_wakeup; - int pending_kill; - int pending_disable; + unsigned int pending_wakeup; + unsigned int pending_kill; + unsigned int pending_disable; + unsigned int pending_sigtrap; unsigned long pending_addr; /* SIGTRAP */ - struct irq_work pending; + struct irq_work pending_irq; + struct callback_head pending_task; + unsigned int pending_work; atomic_t event_limit; @@ -789,19 +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: @@ -814,27 +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; /* @@ -855,8 +935,15 @@ 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; + + /* + * Sum (event->pending_sigtrap + event->pending_work) + * + * The SIGTRAP is targeted at ctx->task, as such it won't do changing + * that until the signal is delivered. + */ + local_t nr_pending; }; /* @@ -865,12 +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; @@ -878,16 +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. @@ -951,6 +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, @@ -1007,18 +1101,20 @@ struct perf_sample_data { * Fields set by perf_sample_data_init(), group so as to * minimize the cachelines touched. */ - u64 addr; - struct perf_raw_record *raw; - struct perf_branch_stack *br_stack; + u64 sample_flags; u64 period; - union perf_sample_weight weight; - u64 txn; - union perf_mem_data_src data_src; /* * The other fields, optionally {set,used} by * perf_{prepare,output}_sample(). */ + struct perf_branch_stack *br_stack; + union perf_sample_weight weight; + union perf_mem_data_src data_src; + u64 txn; + u64 addr; + struct perf_raw_record *raw; + u64 type; u64 ip; struct { @@ -1056,13 +1152,13 @@ static inline void perf_sample_data_init(struct perf_sample_data *data, u64 addr, u64 period) { /* remaining struct members initialized in perf_prepare_sample() */ - data->addr = addr; - data->raw = NULL; - data->br_stack = NULL; + data->sample_flags = PERF_SAMPLE_PERIOD; data->period = period; - data->weight.full = 0; - data->data_src.val = PERF_MEM_NA; - data->txn = 0; + + if (addr) { + data->addr = addr; + data->sample_flags |= PERF_SAMPLE_ADDR; + } } /* @@ -1078,6 +1174,7 @@ static inline void perf_clear_branch_entry_bitfields(struct perf_branch_entry *b br->abort = 0; br->cycles = 0; br->type = 0; + br->spec = PERF_BR_SPEC_NA; br->reserved = 0; } @@ -1153,7 +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) @@ -1684,4 +1781,30 @@ static inline void perf_lopwr_cb(bool mode) } #endif +#ifdef CONFIG_PERF_EVENTS +static inline bool branch_sample_no_flags(const struct perf_event *event) +{ + return event->attr.branch_sample_type & PERF_SAMPLE_BRANCH_NO_FLAGS; +} + +static inline bool branch_sample_no_cycles(const struct perf_event *event) +{ + return event->attr.branch_sample_type & PERF_SAMPLE_BRANCH_NO_CYCLES; +} + +static inline bool branch_sample_type(const struct perf_event *event) +{ + return event->attr.branch_sample_type & PERF_SAMPLE_BRANCH_TYPE_SAVE; +} + +static inline bool branch_sample_hw_index(const struct perf_event *event) +{ + return event->attr.branch_sample_type & PERF_SAMPLE_BRANCH_HW_INDEX; +} + +static inline bool branch_sample_priv(const struct perf_event *event) +{ + return event->attr.branch_sample_type & PERF_SAMPLE_BRANCH_PRIV_SAVE; +} +#endif /* CONFIG_PERF_EVENTS */ #endif /* _LINUX_PERF_EVENT_H */ |