[mirror_ubuntu-jammy-kernel.git] / samples / bpf / trace_event_user.c

// SPDX-License-Identifier: GPL-2.0-only
/* Copyright (c) 2016 Facebook
 */
#include <stdio.h>
#include <unistd.h>
#include <stdlib.h>
#include <stdbool.h>
#include <string.h>
#include <fcntl.h>
#include <poll.h>
#include <sys/ioctl.h>
#include <linux/perf_event.h>
#include <linux/bpf.h>
#include <signal.h>
#include <assert.h>
#include <errno.h>
#include <sys/resource.h>
#include "libbpf.h"
#include "bpf_load.h"
#include "perf-sys.h"
#include "trace_helpers.h"

#define SAMPLE_FREQ 50

static bool sys_read_seen, sys_write_seen;

static void print_ksym(__u64 addr)
{
	struct ksym *sym;

	if (!addr)
		return;
	sym = ksym_search(addr);
	if (!sym) {
		printf("ksym not found. Is kallsyms loaded?\n");
		return;
	}

	printf("%s;", sym->name);
	if (!strcmp(sym->name, "sys_read"))
		sys_read_seen = true;
	else if (!strcmp(sym->name, "sys_write"))
		sys_write_seen = true;
}

static void print_addr(__u64 addr)
{
	if (!addr)
		return;
	printf("%llx;", addr);
}

#define TASK_COMM_LEN 16

struct key_t {
	char comm[TASK_COMM_LEN];
	__u32 kernstack;
	__u32 userstack;
};

static void print_stack(struct key_t *key, __u64 count)
{
	__u64 ip[PERF_MAX_STACK_DEPTH] = {};
	static bool warned;
	int i;

	printf("%3lld %s;", count, key->comm);
	if (bpf_map_lookup_elem(map_fd[1], &key->kernstack, ip) != 0) {
		printf("---;");
	} else {
		for (i = PERF_MAX_STACK_DEPTH - 1; i >= 0; i--)
			print_ksym(ip[i]);
	}
	printf("-;");
	if (bpf_map_lookup_elem(map_fd[1], &key->userstack, ip) != 0) {
		printf("---;");
	} else {
		for (i = PERF_MAX_STACK_DEPTH - 1; i >= 0; i--)
			print_addr(ip[i]);
	}
	if (count < 6)
		printf("\r");
	else
		printf("\n");

	if (key->kernstack == -EEXIST && !warned) {
		printf("stackmap collisions seen. Consider increasing size\n");
		warned = true;
	} else if ((int)key->kernstack < 0 && (int)key->userstack < 0) {
		printf("err stackid %d %d\n", key->kernstack, key->userstack);
	}
}

static void int_exit(int sig)
{
	kill(0, SIGKILL);
	exit(0);
}

static void print_stacks(void)
{
	struct key_t key = {}, next_key;
	__u64 value;
	__u32 stackid = 0, next_id;
	int fd = map_fd[0], stack_map = map_fd[1];

	sys_read_seen = sys_write_seen = false;
	while (bpf_map_get_next_key(fd, &key, &next_key) == 0) {
		bpf_map_lookup_elem(fd, &next_key, &value);
		print_stack(&next_key, value);
		bpf_map_delete_elem(fd, &next_key);
		key = next_key;
	}
	printf("\n");
	if (!sys_read_seen || !sys_write_seen) {
		printf("BUG kernel stack doesn't contain sys_read() and sys_write()\n");
		int_exit(0);
	}

	/* clear stack map */
	while (bpf_map_get_next_key(stack_map, &stackid, &next_id) == 0) {
		bpf_map_delete_elem(stack_map, &next_id);
		stackid = next_id;
	}
}

static inline int generate_load(void)
{
	if (system("dd if=/dev/zero of=/dev/null count=5000k status=none") < 0) {
		printf("failed to generate some load with dd: %s\n", strerror(errno));
		return -1;
	}

	return 0;
}

static void test_perf_event_all_cpu(struct perf_event_attr *attr)
{
	int nr_cpus = sysconf(_SC_NPROCESSORS_CONF);
	int *pmu_fd = malloc(nr_cpus * sizeof(int));
	int i, error = 0;

	/* system wide perf event, no need to inherit */
	attr->inherit = 0;

	/* open perf_event on all cpus */
	for (i = 0; i < nr_cpus; i++) {
		pmu_fd[i] = sys_perf_event_open(attr, -1, i, -1, 0);
		if (pmu_fd[i] < 0) {
			printf("sys_perf_event_open failed\n");
			error = 1;
			goto all_cpu_err;
		}
		assert(ioctl(pmu_fd[i], PERF_EVENT_IOC_SET_BPF, prog_fd[0]) == 0);
		assert(ioctl(pmu_fd[i], PERF_EVENT_IOC_ENABLE) == 0);
	}

	if (generate_load() < 0) {
		error = 1;
		goto all_cpu_err;
	}
	print_stacks();
all_cpu_err:
	for (i--; i >= 0; i--) {
		ioctl(pmu_fd[i], PERF_EVENT_IOC_DISABLE);
		close(pmu_fd[i]);
	}
	free(pmu_fd);
	if (error)
		int_exit(0);
}

static void test_perf_event_task(struct perf_event_attr *attr)
{
	int pmu_fd, error = 0;

	/* per task perf event, enable inherit so the "dd ..." command can be traced properly.
	 * Enabling inherit will cause bpf_perf_prog_read_time helper failure.
	 */
	attr->inherit = 1;

	/* open task bound event */
	pmu_fd = sys_perf_event_open(attr, 0, -1, -1, 0);
	if (pmu_fd < 0) {
		printf("sys_perf_event_open failed\n");
		int_exit(0);
	}
	assert(ioctl(pmu_fd, PERF_EVENT_IOC_SET_BPF, prog_fd[0]) == 0);
	assert(ioctl(pmu_fd, PERF_EVENT_IOC_ENABLE) == 0);

	if (generate_load() < 0) {
		error = 1;
		goto err;
	}
	print_stacks();
err:
	ioctl(pmu_fd, PERF_EVENT_IOC_DISABLE);
	close(pmu_fd);
	if (error)
		int_exit(0);
}

static void test_bpf_perf_event(void)
{
	struct perf_event_attr attr_type_hw = {
		.sample_freq = SAMPLE_FREQ,
		.freq = 1,
		.type = PERF_TYPE_HARDWARE,
		.config = PERF_COUNT_HW_CPU_CYCLES,
	};
	struct perf_event_attr attr_type_sw = {
		.sample_freq = SAMPLE_FREQ,
		.freq = 1,
		.type = PERF_TYPE_SOFTWARE,
		.config = PERF_COUNT_SW_CPU_CLOCK,
	};
	struct perf_event_attr attr_hw_cache_l1d = {
		.sample_freq = SAMPLE_FREQ,
		.freq = 1,
		.type = PERF_TYPE_HW_CACHE,
		.config =
			PERF_COUNT_HW_CACHE_L1D |
			(PERF_COUNT_HW_CACHE_OP_READ << 8) |
			(PERF_COUNT_HW_CACHE_RESULT_ACCESS << 16),
	};
	struct perf_event_attr attr_hw_cache_branch_miss = {
		.sample_freq = SAMPLE_FREQ,
		.freq = 1,
		.type = PERF_TYPE_HW_CACHE,
		.config =
			PERF_COUNT_HW_CACHE_BPU |
			(PERF_COUNT_HW_CACHE_OP_READ << 8) |
			(PERF_COUNT_HW_CACHE_RESULT_MISS << 16),
	};
	struct perf_event_attr attr_type_raw = {
		.sample_freq = SAMPLE_FREQ,
		.freq = 1,
		.type = PERF_TYPE_RAW,
		/* Intel Instruction Retired */
		.config = 0xc0,
	};
	struct perf_event_attr attr_type_raw_lock_load = {
		.sample_freq = SAMPLE_FREQ,
		.freq = 1,
		.type = PERF_TYPE_RAW,
		/* Intel MEM_UOPS_RETIRED.LOCK_LOADS */
		.config = 0x21d0,
		/* Request to record lock address from PEBS */
		.sample_type = PERF_SAMPLE_ADDR,
		/* Record address value requires precise event */
		.precise_ip = 2,
	};

	printf("Test HW_CPU_CYCLES\n");
	test_perf_event_all_cpu(&attr_type_hw);
	test_perf_event_task(&attr_type_hw);

	printf("Test SW_CPU_CLOCK\n");
	test_perf_event_all_cpu(&attr_type_sw);
	test_perf_event_task(&attr_type_sw);

	printf("Test HW_CACHE_L1D\n");
	test_perf_event_all_cpu(&attr_hw_cache_l1d);
	test_perf_event_task(&attr_hw_cache_l1d);

	printf("Test HW_CACHE_BPU\n");
	test_perf_event_all_cpu(&attr_hw_cache_branch_miss);
	test_perf_event_task(&attr_hw_cache_branch_miss);

	printf("Test Instruction Retired\n");
	test_perf_event_all_cpu(&attr_type_raw);
	test_perf_event_task(&attr_type_raw);

	printf("Test Lock Load\n");
	test_perf_event_all_cpu(&attr_type_raw_lock_load);
	test_perf_event_task(&attr_type_raw_lock_load);

	printf("*** PASS ***\n");
}


int main(int argc, char **argv)
{
	struct rlimit r = {RLIM_INFINITY, RLIM_INFINITY};
	char filename[256];

	snprintf(filename, sizeof(filename), "%s_kern.o", argv[0]);
	setrlimit(RLIMIT_MEMLOCK, &r);

	signal(SIGINT, int_exit);
	signal(SIGTERM, int_exit);

	if (load_kallsyms()) {
		printf("failed to process /proc/kallsyms\n");
		return 1;
	}

	if (load_bpf_file(filename)) {
		printf("%s", bpf_log_buf);
		return 2;
	}

	if (fork() == 0) {
		read_trace_pipe();
		return 0;
	}
	test_bpf_perf_event();
	int_exit(0);
	return 0;
}
Commit	Line	Data
25763b3c	1	// SPDX-License-Identifier: GPL-2.0-only
1c47910e	2	/* Copyright (c) 2016 Facebook
1c47910e AS	3	*/
	4	#include <stdio.h>
	5	#include <unistd.h>
	6	#include <stdlib.h>
	7	#include <stdbool.h>
	8	#include <string.h>
	9	#include <fcntl.h>
	10	#include <poll.h>
	11	#include <sys/ioctl.h>
	12	#include <linux/perf_event.h>
	13	#include <linux/bpf.h>
	14	#include <signal.h>
	15	#include <assert.h>
	16	#include <errno.h>
	17	#include <sys/resource.h>
	18	#include "libbpf.h"
	19	#include "bpf_load.h"
205c8ada	20	#include "perf-sys.h"
28dbf861	21	#include "trace_helpers.h"
1c47910e AS	22
	23	#define SAMPLE_FREQ 50
	24
	25	static bool sys_read_seen, sys_write_seen;
	26
	27	static void print_ksym(__u64 addr)
	28	{
	29	struct ksym *sym;
	30
	31	if (!addr)
	32	return;
	33	sym = ksym_search(addr);
e67b2c71 DL	34	if (!sym) {
	35	printf("ksym not found. Is kallsyms loaded?\n");
	36	return;
	37	}
	38
1c47910e AS	39	printf("%s;", sym->name);
	40	if (!strcmp(sym->name, "sys_read"))
	41	sys_read_seen = true;
	42	else if (!strcmp(sym->name, "sys_write"))
	43	sys_write_seen = true;
	44	}
	45
	46	static void print_addr(__u64 addr)
	47	{
	48	if (!addr)
	49	return;
	50	printf("%llx;", addr);
	51	}
	52
	53	#define TASK_COMM_LEN 16
	54
	55	struct key_t {
	56	char comm[TASK_COMM_LEN];
	57	__u32 kernstack;
	58	__u32 userstack;
	59	};
	60
	61	static void print_stack(struct key_t *key, __u64 count)
	62	{
	63	__u64 ip[PERF_MAX_STACK_DEPTH] = {};
	64	static bool warned;
	65	int i;
	66
	67	printf("%3lld %s;", count, key->comm);
d40fc181	68	if (bpf_map_lookup_elem(map_fd[1], &key->kernstack, ip) != 0) {
1c47910e AS	69	printf("---;");
	70	} else {
	71	for (i = PERF_MAX_STACK_DEPTH - 1; i >= 0; i--)
	72	print_ksym(ip[i]);
	73	}
	74	printf("-;");
d40fc181	75	if (bpf_map_lookup_elem(map_fd[1], &key->userstack, ip) != 0) {
1c47910e AS	76	printf("---;");
	77	} else {
	78	for (i = PERF_MAX_STACK_DEPTH - 1; i >= 0; i--)
	79	print_addr(ip[i]);
	80	}
41e9a804 TQ	81	if (count < 6)
	82	printf("\r");
	83	else
	84	printf("\n");
1c47910e AS	85
	86	if (key->kernstack == -EEXIST && !warned) {
	87	printf("stackmap collisions seen. Consider increasing size\n");
	88	warned = true;
	89	} else if ((int)key->kernstack < 0 && (int)key->userstack < 0) {
	90	printf("err stackid %d %d\n", key->kernstack, key->userstack);
	91	}
	92	}
	93
	94	static void int_exit(int sig)
	95	{
	96	kill(0, SIGKILL);
	97	exit(0);
	98	}
	99
	100	static void print_stacks(void)
	101	{
	102	struct key_t key = {}, next_key;
	103	__u64 value;
	104	__u32 stackid = 0, next_id;
	105	int fd = map_fd[0], stack_map = map_fd[1];
	106
	107	sys_read_seen = sys_write_seen = false;
d40fc181 JS	108	while (bpf_map_get_next_key(fd, &key, &next_key) == 0) {
d40fc181 JS	109	bpf_map_lookup_elem(fd, &next_key, &value);
1c47910e	110	print_stack(&next_key, value);
d40fc181	111	bpf_map_delete_elem(fd, &next_key);
1c47910e AS	112	key = next_key;
1c47910e AS	113	}
41e9a804	114	printf("\n");
1c47910e AS	115	if (!sys_read_seen \|\| !sys_write_seen) {
	116	printf("BUG kernel stack doesn't contain sys_read() and sys_write()\n");
	117	int_exit(0);
	118	}
	119
	120	/* clear stack map */
d40fc181 JS	121	while (bpf_map_get_next_key(stack_map, &stackid, &next_id) == 0) {
d40fc181 JS	122	bpf_map_delete_elem(stack_map, &next_id);
1c47910e AS	123	stackid = next_id;
	124	}
	125	}
	126
492b7e89 TS	127	static inline int generate_load(void)
	128	{
	129	if (system("dd if=/dev/zero of=/dev/null count=5000k status=none") < 0) {
	130	printf("failed to generate some load with dd: %s\n", strerror(errno));
	131	return -1;
	132	}
	133
	134	return 0;
	135	}
	136
1c47910e AS	137	static void test_perf_event_all_cpu(struct perf_event_attr *attr)
	138	{
	139	int nr_cpus = sysconf(_SC_NPROCESSORS_CONF);
	140	int pmu_fd = malloc(nr_cpus sizeof(int));
41e9a804	141	int i, error = 0;
1c47910e	142
81b9cf80 YS	143	/* system wide perf event, no need to inherit */
	144	attr->inherit = 0;
	145
1c47910e AS	146	/* open perf_event on all cpus */
1c47910e AS	147	for (i = 0; i < nr_cpus; i++) {
205c8ada	148	pmu_fd[i] = sys_perf_event_open(attr, -1, i, -1, 0);
1c47910e	149	if (pmu_fd[i] < 0) {
205c8ada	150	printf("sys_perf_event_open failed\n");
41e9a804	151	error = 1;
1c47910e AS	152	goto all_cpu_err;
	153	}
	154	assert(ioctl(pmu_fd[i], PERF_EVENT_IOC_SET_BPF, prog_fd[0]) == 0);
41e9a804	155	assert(ioctl(pmu_fd[i], PERF_EVENT_IOC_ENABLE) == 0);
1c47910e	156	}
492b7e89 TS	157
	158	if (generate_load() < 0) {
	159	error = 1;
	160	goto all_cpu_err;
	161	}
1c47910e AS	162	print_stacks();
1c47910e AS	163	all_cpu_err:
41e9a804 TQ	164	for (i--; i >= 0; i--) {
41e9a804 TQ	165	ioctl(pmu_fd[i], PERF_EVENT_IOC_DISABLE);
1c47910e	166	close(pmu_fd[i]);
41e9a804	167	}
1c47910e	168	free(pmu_fd);
41e9a804 TQ	169	if (error)
41e9a804 TQ	170	int_exit(0);
1c47910e AS	171	}
	172
	173	static void test_perf_event_task(struct perf_event_attr *attr)
	174	{
492b7e89	175	int pmu_fd, error = 0;
1c47910e	176
81b9cf80 YS	177	/* per task perf event, enable inherit so the "dd ..." command can be traced properly.
	178	* Enabling inherit will cause bpf_perf_prog_read_time helper failure.
	179	*/
	180	attr->inherit = 1;
	181
1c47910e	182	/* open task bound event */
205c8ada	183	pmu_fd = sys_perf_event_open(attr, 0, -1, -1, 0);
1c47910e	184	if (pmu_fd < 0) {
205c8ada	185	printf("sys_perf_event_open failed\n");
41e9a804	186	int_exit(0);
1c47910e AS	187	}
1c47910e AS	188	assert(ioctl(pmu_fd, PERF_EVENT_IOC_SET_BPF, prog_fd[0]) == 0);
41e9a804	189	assert(ioctl(pmu_fd, PERF_EVENT_IOC_ENABLE) == 0);
492b7e89 TS	190
	191	if (generate_load() < 0) {
	192	error = 1;
	193	goto err;
	194	}
1c47910e	195	print_stacks();
492b7e89	196	err:
41e9a804	197	ioctl(pmu_fd, PERF_EVENT_IOC_DISABLE);
1c47910e	198	close(pmu_fd);
492b7e89 TS	199	if (error)
492b7e89 TS	200	int_exit(0);
1c47910e AS	201	}
	202
	203	static void test_bpf_perf_event(void)
	204	{
	205	struct perf_event_attr attr_type_hw = {
	206	.sample_freq = SAMPLE_FREQ,
	207	.freq = 1,
	208	.type = PERF_TYPE_HARDWARE,
	209	.config = PERF_COUNT_HW_CPU_CYCLES,
1c47910e AS	210	};
	211	struct perf_event_attr attr_type_sw = {
	212	.sample_freq = SAMPLE_FREQ,
	213	.freq = 1,
	214	.type = PERF_TYPE_SOFTWARE,
	215	.config = PERF_COUNT_SW_CPU_CLOCK,
1c47910e	216	};
41e9a804 TQ	217	struct perf_event_attr attr_hw_cache_l1d = {
	218	.sample_freq = SAMPLE_FREQ,
	219	.freq = 1,
	220	.type = PERF_TYPE_HW_CACHE,
	221	.config =
	222	PERF_COUNT_HW_CACHE_L1D \|
	223	(PERF_COUNT_HW_CACHE_OP_READ << 8) \|
	224	(PERF_COUNT_HW_CACHE_RESULT_ACCESS << 16),
41e9a804 TQ	225	};
	226	struct perf_event_attr attr_hw_cache_branch_miss = {
	227	.sample_freq = SAMPLE_FREQ,
	228	.freq = 1,
	229	.type = PERF_TYPE_HW_CACHE,
	230	.config =
	231	PERF_COUNT_HW_CACHE_BPU \|
	232	(PERF_COUNT_HW_CACHE_OP_READ << 8) \|
	233	(PERF_COUNT_HW_CACHE_RESULT_MISS << 16),
41e9a804 TQ	234	};
	235	struct perf_event_attr attr_type_raw = {
	236	.sample_freq = SAMPLE_FREQ,
	237	.freq = 1,
	238	.type = PERF_TYPE_RAW,
	239	/* Intel Instruction Retired */
	240	.config = 0xc0,
41e9a804	241	};
12fe1225 TQ	242	struct perf_event_attr attr_type_raw_lock_load = {
	243	.sample_freq = SAMPLE_FREQ,
	244	.freq = 1,
	245	.type = PERF_TYPE_RAW,
	246	/* Intel MEM_UOPS_RETIRED.LOCK_LOADS */
	247	.config = 0x21d0,
	248	/* Request to record lock address from PEBS */
	249	.sample_type = PERF_SAMPLE_ADDR,
	250	/* Record address value requires precise event */
	251	.precise_ip = 2,
	252	};
1c47910e	253
41e9a804	254	printf("Test HW_CPU_CYCLES\n");
1c47910e AS	255	test_perf_event_all_cpu(&attr_type_hw);
1c47910e AS	256	test_perf_event_task(&attr_type_hw);
41e9a804 TQ	257
41e9a804 TQ	258	printf("Test SW_CPU_CLOCK\n");
1c47910e AS	259	test_perf_event_all_cpu(&attr_type_sw);
1c47910e AS	260	test_perf_event_task(&attr_type_sw);
41e9a804 TQ	261
	262	printf("Test HW_CACHE_L1D\n");
	263	test_perf_event_all_cpu(&attr_hw_cache_l1d);
	264	test_perf_event_task(&attr_hw_cache_l1d);
	265
	266	printf("Test HW_CACHE_BPU\n");
	267	test_perf_event_all_cpu(&attr_hw_cache_branch_miss);
	268	test_perf_event_task(&attr_hw_cache_branch_miss);
	269
	270	printf("Test Instruction Retired\n");
	271	test_perf_event_all_cpu(&attr_type_raw);
	272	test_perf_event_task(&attr_type_raw);
	273
12fe1225 TQ	274	printf("Test Lock Load\n");
	275	test_perf_event_all_cpu(&attr_type_raw_lock_load);
	276	test_perf_event_task(&attr_type_raw_lock_load);
	277
41e9a804	278	printf("* PASS *\n");
1c47910e AS	279	}
	280
	281
	282	int main(int argc, char **argv)
	283	{
	284	struct rlimit r = {RLIM_INFINITY, RLIM_INFINITY};
	285	char filename[256];
	286
	287	snprintf(filename, sizeof(filename), "%s_kern.o", argv[0]);
	288	setrlimit(RLIMIT_MEMLOCK, &r);
	289
	290	signal(SIGINT, int_exit);
ad990dbe	291	signal(SIGTERM, int_exit);
1c47910e AS	292
	293	if (load_kallsyms()) {
	294	printf("failed to process /proc/kallsyms\n");
	295	return 1;
	296	}
	297
	298	if (load_bpf_file(filename)) {
	299	printf("%s", bpf_log_buf);
	300	return 2;
	301	}
	302
	303	if (fork() == 0) {
	304	read_trace_pipe();
	305	return 0;
	306	}
	307	test_bpf_perf_event();
1c47910e AS	308	int_exit(0);
	309	return 0;
	310	}