diff options
Diffstat (limited to 'tools/lib/perf/include/internal')
| -rw-r--r-- | tools/lib/perf/include/internal/cpumap.h | 14 | ||||
| -rw-r--r-- | tools/lib/perf/include/internal/evlist.h | 7 | ||||
| -rw-r--r-- | tools/lib/perf/include/internal/evsel.h | 81 | ||||
| -rw-r--r-- | tools/lib/perf/include/internal/mmap.h | 3 | ||||
| -rw-r--r-- | tools/lib/perf/include/internal/rc_check.h | 113 |
5 files changed, 203 insertions, 15 deletions
diff --git a/tools/lib/perf/include/internal/cpumap.h b/tools/lib/perf/include/internal/cpumap.h index 35dd29642296..e2be2d17c32b 100644 --- a/tools/lib/perf/include/internal/cpumap.h +++ b/tools/lib/perf/include/internal/cpumap.h @@ -4,6 +4,7 @@ #include <linux/refcount.h> #include <perf/cpumap.h> +#include <internal/rc_check.h> /** * A sized, reference counted, sorted array of integers representing CPU @@ -12,7 +13,7 @@ * gaps if CPU numbers were used. For events associated with a pid, rather than * a CPU, a single dummy map with an entry of -1 is used. */ -struct perf_cpu_map { +DECLARE_RC_STRUCT(perf_cpu_map) { refcount_t refcnt; /** Length of the map array. */ int nr; @@ -20,11 +21,14 @@ struct perf_cpu_map { struct perf_cpu map[]; }; -#ifndef MAX_NR_CPUS -#define MAX_NR_CPUS 2048 -#endif - +struct perf_cpu_map *perf_cpu_map__alloc(int nr_cpus); int perf_cpu_map__idx(const struct perf_cpu_map *cpus, struct perf_cpu cpu); bool perf_cpu_map__is_subset(const struct perf_cpu_map *a, const struct perf_cpu_map *b); +void perf_cpu_map__set_nr(struct perf_cpu_map *map, int nr_cpus); + +static inline refcount_t *perf_cpu_map__refcnt(struct perf_cpu_map *map) +{ + return &RC_CHK_ACCESS(map)->refcnt; +} #endif /* __LIBPERF_INTERNAL_CPUMAP_H */ diff --git a/tools/lib/perf/include/internal/evlist.h b/tools/lib/perf/include/internal/evlist.h index 850f07070036..f43bdb9b6227 100644 --- a/tools/lib/perf/include/internal/evlist.h +++ b/tools/lib/perf/include/internal/evlist.h @@ -17,7 +17,6 @@ struct perf_mmap_param; struct perf_evlist { struct list_head entries; int nr_entries; - int nr_groups; bool has_user_cpus; bool needs_map_propagation; /** @@ -127,13 +126,15 @@ u64 perf_evlist__read_format(struct perf_evlist *evlist); void perf_evlist__id_add(struct perf_evlist *evlist, struct perf_evsel *evsel, - int cpu, int thread, u64 id); + int cpu_map_idx, int thread, u64 id); int perf_evlist__id_add_fd(struct perf_evlist *evlist, struct perf_evsel *evsel, - int cpu, int thread, int fd); + int cpu_map_idx, int thread, int fd); void perf_evlist__reset_id_hash(struct perf_evlist *evlist); void __perf_evlist__set_leader(struct list_head *list, struct perf_evsel *leader); + +void perf_evlist__go_system_wide(struct perf_evlist *evlist, struct perf_evsel *evsel); #endif /* __LIBPERF_INTERNAL_EVLIST_H */ diff --git a/tools/lib/perf/include/internal/evsel.h b/tools/lib/perf/include/internal/evsel.h index a99a75d9e78f..fefe64ba5e26 100644 --- a/tools/lib/perf/include/internal/evsel.h +++ b/tools/lib/perf/include/internal/evsel.h @@ -11,6 +11,32 @@ struct perf_thread_map; struct xyarray; +/** + * The per-thread accumulated period storage node. + */ +struct perf_sample_id_period { + struct list_head node; + struct hlist_node hnode; + /* Holds total ID period value for PERF_SAMPLE_READ processing. */ + u64 period; + /* The TID that the values belongs to */ + u32 tid; +}; + +/** + * perf_evsel_for_each_per_thread_period_safe - safely iterate thru all the + * per_stream_periods + * @evlist:perf_evsel instance to iterate + * @item: struct perf_sample_id_period iterator + * @tmp: struct perf_sample_id_period temp iterator + */ +#define perf_evsel_for_each_per_thread_period_safe(evsel, tmp, item) \ + list_for_each_entry_safe(item, tmp, &(evsel)->per_stream_periods, node) + + +#define PERF_SAMPLE_ID__HLIST_BITS 4 +#define PERF_SAMPLE_ID__HLIST_SIZE (1 << PERF_SAMPLE_ID__HLIST_BITS) + /* * Per fd, to map back from PERF_SAMPLE_ID to evsel, only used when there are * more than one entry in the evlist. @@ -34,15 +60,46 @@ struct perf_sample_id { pid_t machine_pid; struct perf_cpu vcpu; - /* Holds total ID period value for PERF_SAMPLE_READ processing. */ - u64 period; + /* + * Per-thread, and global event counts are mutually exclusive: + * Whilst it is possible to combine events into a group with differing + * values of PERF_SAMPLE_READ, it is not valid to have inconsistent + * values for `inherit`. Therefore it is not possible to have a + * situation where a per-thread event is sampled as a global event; + * all !inherit groups are global, and all groups where the sampling + * event is inherit + PERF_SAMPLE_READ will be per-thread. Any event + * that is part of such a group that is inherit but not PERF_SAMPLE_READ + * will be read as per-thread. If such an event can also trigger a + * sample (such as with sample_period > 0) then it will not cause + * `read_format` to be included in its PERF_RECORD_SAMPLE, and + * therefore will not expose the per-thread group members as global. + */ + union { + /* + * Holds total ID period value for PERF_SAMPLE_READ processing + * (when period is not per-thread). + */ + u64 period; + /* + * Holds total ID period value for PERF_SAMPLE_READ processing + * (when period is per-thread). + */ + struct hlist_head periods[PERF_SAMPLE_ID__HLIST_SIZE]; + }; }; struct perf_evsel { struct list_head node; struct perf_event_attr attr; + /** The commonly used cpu map of CPUs the event should be opened upon, etc. */ struct perf_cpu_map *cpus; - struct perf_cpu_map *own_cpus; + /** + * The cpu map read from the PMU. For core PMUs this is the list of all + * CPUs the event can be opened upon. For other PMUs this is the default + * cpu map for opening the event on, for example, the first CPU on a + * socket for an uncore event. + */ + struct perf_cpu_map *pmu_cpus; struct perf_thread_map *threads; struct xyarray *fd; struct xyarray *mmap; @@ -51,13 +108,17 @@ struct perf_evsel { u32 ids; struct perf_evsel *leader; + /* For events where the read_format value is per-thread rather than + * global, stores the per-thread cumulative period */ + struct list_head per_stream_periods; + /* parse modifier helper */ int nr_members; /* * system_wide is for events that need to be on every CPU, irrespective - * of user requested CPUs or threads. Map propagation will set cpus to - * this event's own_cpus, whereby they will contribute to evlist - * all_cpus. + * of user requested CPUs or threads. Tha main example of this is the + * dummy event. Map propagation will set cpus for this event to all CPUs + * as software PMU events like dummy, have a CPU map that is empty. */ bool system_wide; /* @@ -65,11 +126,14 @@ struct perf_evsel { * i.e. it cannot be the 'any CPU' value of -1. */ bool requires_cpu; + /** Is the PMU for the event a core one? Effects the handling of own_cpus. */ + bool is_pmu_core; int idx; }; void perf_evsel__init(struct perf_evsel *evsel, struct perf_event_attr *attr, int idx); +void perf_evsel__exit(struct perf_evsel *evsel); int perf_evsel__alloc_fd(struct perf_evsel *evsel, int ncpus, int nthreads); void perf_evsel__close_fd(struct perf_evsel *evsel); void perf_evsel__free_fd(struct perf_evsel *evsel); @@ -79,4 +143,9 @@ int perf_evsel__apply_filter(struct perf_evsel *evsel, const char *filter); int perf_evsel__alloc_id(struct perf_evsel *evsel, int ncpus, int nthreads); void perf_evsel__free_id(struct perf_evsel *evsel); +bool perf_evsel__attr_has_per_thread_sample_period(struct perf_evsel *evsel); + +u64 *perf_sample_id__get_period_storage(struct perf_sample_id *sid, u32 tid, + bool per_thread); + #endif /* __LIBPERF_INTERNAL_EVSEL_H */ diff --git a/tools/lib/perf/include/internal/mmap.h b/tools/lib/perf/include/internal/mmap.h index 5a062af8e9d8..5f08cab61ece 100644 --- a/tools/lib/perf/include/internal/mmap.h +++ b/tools/lib/perf/include/internal/mmap.h @@ -33,7 +33,8 @@ struct perf_mmap { bool overwrite; u64 flush; libperf_unmap_cb_t unmap_cb; - char event_copy[PERF_SAMPLE_MAX_SIZE] __aligned(8); + void *event_copy; + size_t event_copy_sz; struct perf_mmap *next; }; diff --git a/tools/lib/perf/include/internal/rc_check.h b/tools/lib/perf/include/internal/rc_check.h new file mode 100644 index 000000000000..f80ddfc80129 --- /dev/null +++ b/tools/lib/perf/include/internal/rc_check.h @@ -0,0 +1,113 @@ +/* SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause) */ +#ifndef __LIBPERF_INTERNAL_RC_CHECK_H +#define __LIBPERF_INTERNAL_RC_CHECK_H + +#include <stdlib.h> +#include <linux/zalloc.h> + +/* + * Enable reference count checking implicitly with leak checking, which is + * integrated into address sanitizer. + */ +#if defined(__SANITIZE_ADDRESS__) || defined(LEAK_SANITIZER) || defined(ADDRESS_SANITIZER) +#define REFCNT_CHECKING 1 +#elif defined(__has_feature) +#if __has_feature(address_sanitizer) || __has_feature(leak_sanitizer) +#define REFCNT_CHECKING 1 +#endif +#endif + +/* + * Shared reference count checking macros. + * + * Reference count checking is an approach to sanitizing the use of reference + * counted structs. It leverages address and leak sanitizers to make sure gets + * are paired with a put. Reference count checking adds a malloc-ed layer of + * indirection on a get, and frees it on a put. A missed put will be reported as + * a memory leak. A double put will be reported as a double free. Accessing + * after a put will cause a use-after-free and/or a segfault. + */ + +#ifndef REFCNT_CHECKING +/* Replaces "struct foo" so that the pointer may be interposed. */ +#define DECLARE_RC_STRUCT(struct_name) \ + struct struct_name + +/* Declare a reference counted struct variable. */ +#define RC_STRUCT(struct_name) struct struct_name + +/* + * Interpose the indirection. Result will hold the indirection and object is the + * reference counted struct. + */ +#define ADD_RC_CHK(result, object) (result = object, object) + +/* Strip the indirection layer. */ +#define RC_CHK_ACCESS(object) object + +/* Frees the object and the indirection layer. */ +#define RC_CHK_FREE(object) free(object) + +/* A get operation adding the indirection layer. */ +#define RC_CHK_GET(result, object) ADD_RC_CHK(result, object) + +/* A put operation removing the indirection layer. */ +#define RC_CHK_PUT(object) {} + +/* Pointer equality when the indirection may or may not be there. */ +#define RC_CHK_EQUAL(object1, object2) (object1 == object2) + +#else + +/* Replaces "struct foo" so that the pointer may be interposed. */ +#define DECLARE_RC_STRUCT(struct_name) \ + struct original_##struct_name; \ + struct struct_name { \ + struct original_##struct_name *orig; \ + }; \ + struct original_##struct_name + +/* Declare a reference counted struct variable. */ +#define RC_STRUCT(struct_name) struct original_##struct_name + +/* + * Interpose the indirection. Result will hold the indirection and object is the + * reference counted struct. + */ +#define ADD_RC_CHK(result, object) \ + ( \ + object ? (result = malloc(sizeof(*result)), \ + result ? (result->orig = object, result) \ + : (result = NULL, NULL)) \ + : (result = NULL, NULL) \ + ) + +/* Strip the indirection layer. */ +#define RC_CHK_ACCESS(object) object->orig + +/* Frees the object and the indirection layer. */ +#define RC_CHK_FREE(object) \ + do { \ + zfree(&object->orig); \ + free(object); \ + } while(0) + +/* A get operation adding the indirection layer. */ +#define RC_CHK_GET(result, object) ADD_RC_CHK(result, (object ? object->orig : NULL)) + +/* A put operation removing the indirection layer. */ +#define RC_CHK_PUT(object) \ + do { \ + if (object) { \ + object->orig = NULL; \ + free(object); \ + } \ + } while(0) + +/* Pointer equality when the indirection may or may not be there. */ +#define RC_CHK_EQUAL(object1, object2) (object1 == object2 || \ + (object1 && object2 && object1->orig == object2->orig)) + +#endif + +#endif /* __LIBPERF_INTERNAL_RC_CHECK_H */ |