tools/bpftool: switch map event_pipe to libbpf's perf_buffer

Switch event_pipe implementation to rely on new libbpf perf buffer API
(it's raw low-level variant).

Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Acked-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

tools/bpf/bpftool/map_perf_ring.c

@@ -28,7 +28,7 @@
 
 #define MMAP_PAGE_CNT	16
 
-static bool stop;
+static volatile bool stop;
 
 struct event_ring_info {
 	int fd;
@@ -44,32 +44,44 @@ struct perf_event_sample {
 	unsigned char data[];
 };
 
+struct perf_event_lost {
+	struct perf_event_header header;
+	__u64 id;
+	__u64 lost;
+};
+
 static void int_exit(int signo)
 {
 	fprintf(stderr, "Stopping...\n");
 	stop = true;
 }
 
+struct event_pipe_ctx {
+	bool all_cpus;
+	int cpu;
+	int idx;
+};
+
 static enum bpf_perf_event_ret
-print_bpf_output(struct perf_event_header *event, void *private_data)
+print_bpf_output(void *private_data, int cpu, struct perf_event_header *event)
 {
-	struct perf_event_sample *e = container_of(event, struct perf_event_sample,
+	struct perf_event_sample *e = container_of(event,
+						   struct perf_event_sample,
 						   header);
-	struct event_ring_info *ring = private_data;
+	struct perf_event_lost *lost = container_of(event,
-	struct {
+						    struct perf_event_lost,
-		struct perf_event_header header;
+						    header);
-		__u64 id;
+	struct event_pipe_ctx *ctx = private_data;
-		__u64 lost;
+	int idx = ctx->all_cpus ? cpu : ctx->idx;
-	} *lost = (typeof(lost))event;
 
 	if (json_output) {
 		jsonw_start_object(json_wtr);
 		jsonw_name(json_wtr, "type");
 		jsonw_uint(json_wtr, e->header.type);
 		jsonw_name(json_wtr, "cpu");
-		jsonw_uint(json_wtr, ring->cpu);
+		jsonw_uint(json_wtr, cpu);
 		jsonw_name(json_wtr, "index");
-		jsonw_uint(json_wtr, ring->key);
+		jsonw_uint(json_wtr, idx);
 		if (e->header.type == PERF_RECORD_SAMPLE) {
 			jsonw_name(json_wtr, "timestamp");
 			jsonw_uint(json_wtr, e->time);
@@ -89,7 +101,7 @@ print_bpf_output(struct perf_event_header *event, void *private_data)
 		if (e->header.type == PERF_RECORD_SAMPLE) {
 			printf("== @%lld.%09lld CPU: %d index: %d =====\n",
 			       e->time / 1000000000ULL, e->time % 1000000000ULL,
-			       ring->cpu, ring->key);
+			       cpu, idx);
 			fprint_hex(stdout, e->data, e->size, " ");
 			printf("\n");
 		} else if (e->header.type == PERF_RECORD_LOST) {
@@ -103,87 +115,25 @@ print_bpf_output(struct perf_event_header *event, void *private_data)
 	return LIBBPF_PERF_EVENT_CONT;
 }
 
-static void
+int do_event_pipe(int argc, char **argv)
-perf_event_read(struct event_ring_info *ring, void **buf, size_t *buf_len)
 {
-	enum bpf_perf_event_ret ret;
+	struct perf_event_attr perf_attr = {
-
-	ret = bpf_perf_event_read_simple(ring->mem,
-					 MMAP_PAGE_CNT * get_page_size(),
-					 get_page_size(), buf, buf_len,
-					 print_bpf_output, ring);
-	if (ret != LIBBPF_PERF_EVENT_CONT) {
-		fprintf(stderr, "perf read loop failed with %d\n", ret);
-		stop = true;
-	}
-}
-
-static int perf_mmap_size(void)
-{
-	return get_page_size() * (MMAP_PAGE_CNT + 1);
-}
-
-static void *perf_event_mmap(int fd)
-{
-	int mmap_size = perf_mmap_size();
-	void *base;
-
-	base = mmap(NULL, mmap_size, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
-	if (base == MAP_FAILED) {
-		p_err("event mmap failed: %s\n", strerror(errno));
-		return NULL;
-	}
-
-	return base;
-}
-
-static void perf_event_unmap(void *mem)
-{
-	if (munmap(mem, perf_mmap_size()))
-		fprintf(stderr, "Can't unmap ring memory!\n");
-}
-
-static int bpf_perf_event_open(int map_fd, int key, int cpu)
-{
-	struct perf_event_attr attr = {
 		.sample_type = PERF_SAMPLE_RAW | PERF_SAMPLE_TIME,
 		.type = PERF_TYPE_SOFTWARE,
 		.config = PERF_COUNT_SW_BPF_OUTPUT,
+		.sample_period = 1,
+		.wakeup_events = 1,
 	};
-	int pmu_fd;
-
-	pmu_fd = sys_perf_event_open(&attr, -1, cpu, -1, 0);
-	if (pmu_fd < 0) {
-		p_err("failed to open perf event %d for CPU %d", key, cpu);
-		return -1;
-	}
-
-	if (bpf_map_update_elem(map_fd, &key, &pmu_fd, BPF_ANY)) {
-		p_err("failed to update map for event %d for CPU %d", key, cpu);
-		goto err_close;
-	}
-	if (ioctl(pmu_fd, PERF_EVENT_IOC_ENABLE, 0)) {
-		p_err("failed to enable event %d for CPU %d", key, cpu);
-		goto err_close;
-	}
-
-	return pmu_fd;
-
-err_close:
-	close(pmu_fd);
-	return -1;
-}
-
-int do_event_pipe(int argc, char **argv)
-{
-	int i, nfds, map_fd, index = -1, cpu = -1;
 	struct bpf_map_info map_info = {};
-	struct event_ring_info *rings;
+	struct perf_buffer_raw_opts opts = {};
-	size_t tmp_buf_sz = 0;
+	struct event_pipe_ctx ctx = {
-	void *tmp_buf = NULL;
+		.all_cpus = true,
-	struct pollfd *pfds;
+		.cpu = -1,
+		.idx = -1,
+	};
+	struct perf_buffer *pb;
 	__u32 map_info_len;
-	bool do_all = true;
+	int err, map_fd;
 
 	map_info_len = sizeof(map_info);
 	map_fd = map_parse_fd_and_info(&argc, &argv, &map_info, &map_info_len);
@@ -205,7 +155,7 @@ int do_event_pipe(int argc, char **argv)
 			char *endptr;
 
 			NEXT_ARG();
-			cpu = strtoul(*argv, &endptr, 0);
+			ctx.cpu = strtoul(*argv, &endptr, 0);
 			if (*endptr) {
 				p_err("can't parse %s as CPU ID", **argv);
 				goto err_close_map;
@@ -216,7 +166,7 @@ int do_event_pipe(int argc, char **argv)
 			char *endptr;
 
 			NEXT_ARG();
-			index = strtoul(*argv, &endptr, 0);
+			ctx.idx = strtoul(*argv, &endptr, 0);
 			if (*endptr) {
 				p_err("can't parse %s as index", **argv);
 				goto err_close_map;
@@ -228,45 +178,32 @@ int do_event_pipe(int argc, char **argv)
 			goto err_close_map;
 		}
 
-		do_all = false;
+		ctx.all_cpus = false;
 	}
 
-	if (!do_all) {
+	if (!ctx.all_cpus) {
-		if (index == -1 || cpu == -1) {
+		if (ctx.idx == -1 || ctx.cpu == -1) {
 			p_err("cpu and index must be specified together");
 			goto err_close_map;
 		}
-
-		nfds = 1;
 	} else {
-		nfds = min(get_possible_cpus(), map_info.max_entries);
+		ctx.cpu = 0;
-		cpu = 0;
+		ctx.idx = 0;
-		index = 0;
 	}
 
-	rings = calloc(nfds, sizeof(rings[0]));
+	opts.attr = &perf_attr;
-	if (!rings)
+	opts.event_cb = print_bpf_output;
+	opts.ctx = &ctx;
+	opts.cpu_cnt = ctx.all_cpus ? 0 : 1;
+	opts.cpus = &ctx.cpu;
+	opts.map_keys = &ctx.idx;
+
+	pb = perf_buffer__new_raw(map_fd, MMAP_PAGE_CNT, &opts);
+	err = libbpf_get_error(pb);
+	if (err) {
+		p_err("failed to create perf buffer: %s (%d)",
+		      strerror(err), err);
 		goto err_close_map;
-
-	pfds = calloc(nfds, sizeof(pfds[0]));
-	if (!pfds)
-		goto err_free_rings;
-
-	for (i = 0; i < nfds; i++) {
-		rings[i].cpu = cpu + i;
-		rings[i].key = index + i;
-
-		rings[i].fd = bpf_perf_event_open(map_fd, rings[i].key,
-						  rings[i].cpu);
-		if (rings[i].fd < 0)
-			goto err_close_fds_prev;
-
-		rings[i].mem = perf_event_mmap(rings[i].fd);
-		if (!rings[i].mem)
-			goto err_close_fds_current;
-
-		pfds[i].fd = rings[i].fd;
-		pfds[i].events = POLLIN;
 	}
 
 	signal(SIGINT, int_exit);
@@ -277,34 +214,24 @@ int do_event_pipe(int argc, char **argv)
 		jsonw_start_array(json_wtr);
 
 	while (!stop) {
-		poll(pfds, nfds, 200);
+		err = perf_buffer__poll(pb, 200);
-		for (i = 0; i < nfds; i++)
+		if (err < 0 && err != -EINTR) {
-			perf_event_read(&rings[i], &tmp_buf, &tmp_buf_sz);
+			p_err("perf buffer polling failed: %s (%d)",
+			      strerror(err), err);
+			goto err_close_pb;
+		}
 	}
-	free(tmp_buf);
 
 	if (json_output)
 		jsonw_end_array(json_wtr);
 
-	for (i = 0; i < nfds; i++) {
+	perf_buffer__free(pb);
-		perf_event_unmap(rings[i].mem);
-		close(rings[i].fd);
-	}
-	free(pfds);
-	free(rings);
 	close(map_fd);
 
 	return 0;
 
-err_close_fds_prev:
+err_close_pb:
-	while (i--) {
+	perf_buffer__free(pb);
-		perf_event_unmap(rings[i].mem);
-err_close_fds_current:
-		close(rings[i].fd);
-	}
-	free(pfds);
-err_free_rings:
-	free(rings);
 err_close_map:
 	close(map_fd);
 	return -1;