perf cpumap: Give CPUs their own type

A common problem is confusing CPU map indices with the CPU, by wrapping
the CPU with a struct then this is avoided. This approach is similar to
atomic_t.

Committer notes:

To make it build with BUILD_BPF_SKEL=1 these files needed the
conversions to 'struct perf_cpu' usage:

  tools/perf/util/bpf_counter.c
  tools/perf/util/bpf_counter_cgroup.c
  tools/perf/util/bpf_ftrace.c

Also perf_env__get_cpu() was removed back in "perf cpumap: Switch
cpu_map__build_map to cpu function".

Additionally these needed to be fixed for the ARM builds to complete:

  tools/perf/arch/arm/util/cs-etm.c
  tools/perf/arch/arm64/util/pmu.c

Suggested-by: John Garry <[email protected]>
Signed-off-by: Ian Rogers <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kajol Jain <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Leo Yan <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Mathieu Poirier <[email protected]>
Cc: Mike Leach <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Paul Clarke <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Riccardo Mancini <[email protected]>
Cc: Stephane Eranian <[email protected]>
Cc: Suzuki Poulouse <[email protected]>
Cc: Vineet Singh <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

tools/lib/perf/cpumap.c

@@ -10,15 +10,24 @@
 #include <ctype.h>
 #include <limits.h>
 
-struct perf_cpu_map *perf_cpu_map__dummy_new(void)
+static struct perf_cpu_map *perf_cpu_map__alloc(int nr_cpus)
 {
-	struct perf_cpu_map *cpus = malloc(sizeof(*cpus) + sizeof(int));
+	struct perf_cpu_map *cpus = malloc(sizeof(*cpus) + sizeof(struct perf_cpu) * nr_cpus);
 
 	if (cpus != NULL) {
-		cpus->nr = 1;
-		cpus->map[0] = -1;
+		cpus->nr = nr_cpus;
 		refcount_set(&cpus->refcnt, 1);
+
 	}
+	return cpus;
+}
+
+struct perf_cpu_map *perf_cpu_map__dummy_new(void)
+{
+	struct perf_cpu_map *cpus = perf_cpu_map__alloc(1);
+
+	if (cpus)
+		cpus->map[0].cpu = -1;
 
 	return cpus;
 }
@@ -54,15 +63,12 @@ static struct perf_cpu_map *cpu_map__default_new(void)
 	if (nr_cpus < 0)
 		return NULL;
 
-	cpus = malloc(sizeof(*cpus) + nr_cpus * sizeof(int));
+	cpus = perf_cpu_map__alloc(nr_cpus);
 	if (cpus != NULL) {
 		int i;
 
 		for (i = 0; i < nr_cpus; ++i)
-			cpus->map[i] = i;
-
-		cpus->nr = nr_cpus;
-		refcount_set(&cpus->refcnt, 1);
+			cpus->map[i].cpu = i;
 	}
 
 	return cpus;
@@ -73,39 +79,40 @@ struct perf_cpu_map *perf_cpu_map__default_new(void)
 	return cpu_map__default_new();
 }
 
-static int cmp_int(const void *a, const void *b)
+
+static int cmp_cpu(const void *a, const void *b)
 {
-	return *(const int *)a - *(const int*)b;
+	const struct perf_cpu *cpu_a = a, *cpu_b = b;
+
+	return cpu_a->cpu - cpu_b->cpu;
 }
 
-static struct perf_cpu_map *cpu_map__trim_new(int nr_cpus, int *tmp_cpus)
+static struct perf_cpu_map *cpu_map__trim_new(int nr_cpus, const struct perf_cpu *tmp_cpus)
 {
-	size_t payload_size = nr_cpus * sizeof(int);
-	struct perf_cpu_map *cpus = malloc(sizeof(*cpus) + payload_size);
+	size_t payload_size = nr_cpus * sizeof(struct perf_cpu);
+	struct perf_cpu_map *cpus = perf_cpu_map__alloc(nr_cpus);
 	int i, j;
 
 	if (cpus != NULL) {
 		memcpy(cpus->map, tmp_cpus, payload_size);
-		qsort(cpus->map, nr_cpus, sizeof(int), cmp_int);
+		qsort(cpus->map, nr_cpus, sizeof(struct perf_cpu), cmp_cpu);
 		/* Remove dups */
 		j = 0;
 		for (i = 0; i < nr_cpus; i++) {
-			if (i == 0 || cpus->map[i] != cpus->map[i - 1])
-				cpus->map[j++] = cpus->map[i];
+			if (i == 0 || cpus->map[i].cpu != cpus->map[i - 1].cpu)
+				cpus->map[j++].cpu = cpus->map[i].cpu;
 		}
 		cpus->nr = j;
 		assert(j <= nr_cpus);
-		refcount_set(&cpus->refcnt, 1);
 	}
-
 	return cpus;
 }
 
 struct perf_cpu_map *perf_cpu_map__read(FILE *file)
 {
 	struct perf_cpu_map *cpus = NULL;
 	int nr_cpus = 0;
-	int *tmp_cpus = NULL, *tmp;
+	struct perf_cpu *tmp_cpus = NULL, *tmp;
 	int max_entries = 0;
 	int n, cpu, prev;
 	char sep;
@@ -124,24 +131,24 @@ struct perf_cpu_map *perf_cpu_map__read(FILE *file)
 
 			if (new_max >= max_entries) {
 				max_entries = new_max + MAX_NR_CPUS / 2;
-				tmp = realloc(tmp_cpus, max_entries * sizeof(int));
+				tmp = realloc(tmp_cpus, max_entries * sizeof(struct perf_cpu));
 				if (tmp == NULL)
 					goto out_free_tmp;
 				tmp_cpus = tmp;
 			}
 
 			while (++prev < cpu)
-				tmp_cpus[nr_cpus++] = prev;
+				tmp_cpus[nr_cpus++].cpu = prev;
 		}
 		if (nr_cpus == max_entries) {
 			max_entries += MAX_NR_CPUS;
-			tmp = realloc(tmp_cpus, max_entries * sizeof(int));
+			tmp = realloc(tmp_cpus, max_entries * sizeof(struct perf_cpu));
 			if (tmp == NULL)
 				goto out_free_tmp;
 			tmp_cpus = tmp;
 		}
 
-		tmp_cpus[nr_cpus++] = cpu;
+		tmp_cpus[nr_cpus++].cpu = cpu;
 		if (n == 2 && sep == '-')
 			prev = cpu;
 		else
@@ -179,7 +186,7 @@ struct perf_cpu_map *perf_cpu_map__new(const char *cpu_list)
 	unsigned long start_cpu, end_cpu = 0;
 	char *p = NULL;
 	int i, nr_cpus = 0;
-	int *tmp_cpus = NULL, *tmp;
+	struct perf_cpu *tmp_cpus = NULL, *tmp;
 	int max_entries = 0;
 
 	if (!cpu_list)
@@ -220,17 +227,17 @@ struct perf_cpu_map *perf_cpu_map__new(const char *cpu_list)
 		for (; start_cpu <= end_cpu; start_cpu++) {
 			/* check for duplicates */
 			for (i = 0; i < nr_cpus; i++)
-				if (tmp_cpus[i] == (int)start_cpu)
+				if (tmp_cpus[i].cpu == (int)start_cpu)
 					goto invalid;
 
 			if (nr_cpus == max_entries) {
 				max_entries += MAX_NR_CPUS;
-				tmp = realloc(tmp_cpus, max_entries * sizeof(int));
+				tmp = realloc(tmp_cpus, max_entries * sizeof(struct perf_cpu));
 				if (tmp == NULL)
 					goto invalid;
 				tmp_cpus = tmp;
 			}
-			tmp_cpus[nr_cpus++] = (int)start_cpu;
+			tmp_cpus[nr_cpus++].cpu = (int)start_cpu;
 		}
 		if (*p)
 			++p;
@@ -250,12 +257,16 @@ struct perf_cpu_map *perf_cpu_map__new(const char *cpu_list)
 	return cpus;
 }
 
-int perf_cpu_map__cpu(const struct perf_cpu_map *cpus, int idx)
+struct perf_cpu perf_cpu_map__cpu(const struct perf_cpu_map *cpus, int idx)
 {
+	struct perf_cpu result = {
+		.cpu = -1
+	};
+
 	if (cpus && idx < cpus->nr)
 		return cpus->map[idx];
 
-	return -1;
+	return result;
 }
 
 int perf_cpu_map__nr(const struct perf_cpu_map *cpus)
@@ -265,10 +276,10 @@ int perf_cpu_map__nr(const struct perf_cpu_map *cpus)
 
 bool perf_cpu_map__empty(const struct perf_cpu_map *map)
 {
-	return map ? map->map[0] == -1 : true;
+	return map ? map->map[0].cpu == -1 : true;
 }
 
-int perf_cpu_map__idx(const struct perf_cpu_map *cpus, int cpu)
+int perf_cpu_map__idx(const struct perf_cpu_map *cpus, struct perf_cpu cpu)
 {
 	int low, high;
 
@@ -278,13 +289,13 @@ int perf_cpu_map__idx(const struct perf_cpu_map *cpus, int cpu)
 	low = 0;
 	high = cpus->nr;
 	while (low < high) {
-		int idx = (low + high) / 2,
-		    cpu_at_idx = cpus->map[idx];
+		int idx = (low + high) / 2;
+		struct perf_cpu cpu_at_idx = cpus->map[idx];
 
-		if (cpu_at_idx == cpu)
+		if (cpu_at_idx.cpu == cpu.cpu)
 			return idx;
 
-		if (cpu_at_idx > cpu)
+		if (cpu_at_idx.cpu > cpu.cpu)
 			high = idx;
 		else
 			low = idx + 1;
@@ -293,15 +304,19 @@ int perf_cpu_map__idx(const struct perf_cpu_map *cpus, int cpu)
 	return -1;
 }
 
-bool perf_cpu_map__has(const struct perf_cpu_map *cpus, int cpu)
+bool perf_cpu_map__has(const struct perf_cpu_map *cpus, struct perf_cpu cpu)
 {
 	return perf_cpu_map__idx(cpus, cpu) != -1;
 }
 
-int perf_cpu_map__max(struct perf_cpu_map *map)
+struct perf_cpu perf_cpu_map__max(struct perf_cpu_map *map)
 {
+	struct perf_cpu result = {
+		.cpu = -1
+	};
+
 	// cpu_map__trim_new() qsort()s it, cpu_map__default_new() sorts it as well.
-	return map->nr > 0 ? map->map[map->nr - 1] : -1;
+	return map->nr > 0 ? map->map[map->nr - 1] : result;
 }
 
 /*
@@ -315,7 +330,7 @@ int perf_cpu_map__max(struct perf_cpu_map *map)
 struct perf_cpu_map *perf_cpu_map__merge(struct perf_cpu_map *orig,
 					 struct perf_cpu_map *other)
 {
-	int *tmp_cpus;
+	struct perf_cpu *tmp_cpus;
 	int tmp_len;
 	int i, j, k;
 	struct perf_cpu_map *merged;
@@ -329,19 +344,19 @@ struct perf_cpu_map *perf_cpu_map__merge(struct perf_cpu_map *orig,
 	if (!other)
 		return orig;
 	if (orig->nr == other->nr &&
-	    !memcmp(orig->map, other->map, orig->nr * sizeof(int)))
+	    !memcmp(orig->map, other->map, orig->nr * sizeof(struct perf_cpu)))
 		return orig;
 
 	tmp_len = orig->nr + other->nr;
-	tmp_cpus = malloc(tmp_len * sizeof(int));
+	tmp_cpus = malloc(tmp_len * sizeof(struct perf_cpu));
 	if (!tmp_cpus)
 		return NULL;
 
 	/* Standard merge algorithm from wikipedia */
 	i = j = k = 0;
 	while (i < orig->nr && j < other->nr) {
-		if (orig->map[i] <= other->map[j]) {
-			if (orig->map[i] == other->map[j])
+		if (orig->map[i].cpu <= other->map[j].cpu) {
+			if (orig->map[i].cpu == other->map[j].cpu)
 				j++;
 			tmp_cpus[k++] = orig->map[i++];
 		} else

tools/lib/perf/evlist.c

@@ -407,7 +407,7 @@ perf_evlist__mmap_cb_get(struct perf_evlist *evlist, bool overwrite, int idx)
 
 static int
 perf_evlist__mmap_cb_mmap(struct perf_mmap *map, struct perf_mmap_param *mp,
-			  int output, int cpu)
+			  int output, struct perf_cpu cpu)
 {
 	return perf_mmap__mmap(map, mp, output, cpu);
 }
@@ -426,7 +426,7 @@ mmap_per_evsel(struct perf_evlist *evlist, struct perf_evlist_mmap_ops *ops,
 	       int idx, struct perf_mmap_param *mp, int cpu_idx,
 	       int thread, int *_output, int *_output_overwrite)
 {
-	int evlist_cpu = perf_cpu_map__cpu(evlist->cpus, cpu_idx);
+	struct perf_cpu evlist_cpu = perf_cpu_map__cpu(evlist->cpus, cpu_idx);
 	struct perf_evsel *evsel;
 	int revent;
 

tools/lib/perf/evsel.c

@@ -78,10 +78,10 @@ static int perf_evsel__alloc_mmap(struct perf_evsel *evsel, int ncpus, int nthre
 
 static int
 sys_perf_event_open(struct perf_event_attr *attr,
-		    pid_t pid, int cpu, int group_fd,
+		    pid_t pid, struct perf_cpu cpu, int group_fd,
 		    unsigned long flags)
 {
-	return syscall(__NR_perf_event_open, attr, pid, cpu, group_fd, flags);
+	return syscall(__NR_perf_event_open, attr, pid, cpu.cpu, group_fd, flags);
 }
 
 static int get_group_fd(struct perf_evsel *evsel, int cpu_map_idx, int thread, int *group_fd)
@@ -113,7 +113,8 @@ static int get_group_fd(struct perf_evsel *evsel, int cpu_map_idx, int thread, i
 int perf_evsel__open(struct perf_evsel *evsel, struct perf_cpu_map *cpus,
 		     struct perf_thread_map *threads)
 {
-	int cpu, idx, thread, err = 0;
+	struct perf_cpu cpu;
+	int idx, thread, err = 0;
 
 	if (cpus == NULL) {
 		static struct perf_cpu_map *empty_cpu_map;
@@ -252,7 +253,7 @@ int perf_evsel__mmap(struct perf_evsel *evsel, int pages)
 		for (thread = 0; thread < xyarray__max_y(evsel->fd); thread++) {
 			int *fd = FD(evsel, idx, thread);
 			struct perf_mmap *map;
-			int cpu = perf_cpu_map__cpu(evsel->cpus, idx);
+			struct perf_cpu cpu = perf_cpu_map__cpu(evsel->cpus, idx);
 
 			if (fd == NULL || *fd < 0)
 				continue;

tools/lib/perf/include/internal/cpumap.h

@@ -4,6 +4,11 @@
 
 #include <linux/refcount.h>
 
+/** A wrapper around a CPU to avoid confusion with the perf_cpu_map's map's indices. */
+struct perf_cpu {
+	int cpu;
+};
+
 /**
  * A sized, reference counted, sorted array of integers representing CPU
  * numbers. This is commonly used to capture which CPUs a PMU is associated
@@ -16,13 +21,13 @@ struct perf_cpu_map {
 	/** Length of the map array. */
 	int		nr;
 	/** The CPU values. */
-	int		map[];
+	struct perf_cpu	map[];
 };
 
 #ifndef MAX_NR_CPUS
 #define MAX_NR_CPUS	2048
 #endif
 
-int perf_cpu_map__idx(const struct perf_cpu_map *cpus, int cpu);
+int perf_cpu_map__idx(const struct perf_cpu_map *cpus, struct perf_cpu cpu);
 
 #endif /* __LIBPERF_INTERNAL_CPUMAP_H */

tools/lib/perf/include/internal/evlist.h

@@ -4,6 +4,7 @@
 
 #include <linux/list.h>
 #include <api/fd/array.h>
+#include <internal/cpumap.h>
 #include <internal/evsel.h>
 
 #define PERF_EVLIST__HLIST_BITS 8
@@ -36,7 +37,7 @@ typedef void
 typedef struct perf_mmap*
 (*perf_evlist_mmap__cb_get_t)(struct perf_evlist*, bool, int);
 typedef int
-(*perf_evlist_mmap__cb_mmap_t)(struct perf_mmap*, struct perf_mmap_param*, int, int);
+(*perf_evlist_mmap__cb_mmap_t)(struct perf_mmap*, struct perf_mmap_param*, int, struct perf_cpu);
 
 struct perf_evlist_mmap_ops {
 	perf_evlist_mmap__cb_idx_t	idx;

tools/lib/perf/include/internal/evsel.h

@@ -6,8 +6,8 @@
 #include <linux/perf_event.h>
 #include <stdbool.h>
 #include <sys/types.h>
+#include <internal/cpumap.h>
 
-struct perf_cpu_map;
 struct perf_thread_map;
 struct xyarray;
 
@@ -27,7 +27,7 @@ struct perf_sample_id {
 	* queue number.
 	*/
 	int			 idx;
-	int			 cpu;
+	struct perf_cpu		 cpu;
 	pid_t			 tid;
 
 	/* Holds total ID period value for PERF_SAMPLE_READ processing. */