[deliverable/linux.git] / tools / perf / tests / perf-record.c

/* For the CLR_() macros */
#include <pthread.h>

#include <sched.h>
#include "evlist.h"
#include "evsel.h"
#include "perf.h"
#include "debug.h"
#include "tests.h"

static int sched__get_first_possible_cpu(pid_t pid, cpu_set_t *maskp)
{
	int i, cpu = -1, nrcpus = 1024;
realloc:
	CPU_ZERO(maskp);

	if (sched_getaffinity(pid, sizeof(*maskp), maskp) == -1) {
		if (errno == EINVAL && nrcpus < (1024 << 8)) {
			nrcpus = nrcpus << 2;
			goto realloc;
		}
		perror("sched_getaffinity");
			return -1;
	}

	for (i = 0; i < nrcpus; i++) {
		if (CPU_ISSET(i, maskp)) {
			if (cpu == -1)
				cpu = i;
			else
				CPU_CLR(i, maskp);
		}
	}

	return cpu;
}

int test__PERF_RECORD(int subtest __maybe_unused)
{
	struct record_opts opts = {
		.target = {
			.uid = UINT_MAX,
			.uses_mmap = true,
		},
		.no_buffering = true,
		.mmap_pages   = 256,
	};
	cpu_set_t cpu_mask;
	size_t cpu_mask_size = sizeof(cpu_mask);
	struct perf_evlist *evlist = perf_evlist__new_dummy();
	struct perf_evsel *evsel;
	struct perf_sample sample;
	const char *cmd = "sleep";
	const char *argv[] = { cmd, "1", NULL, };
	char *bname, *mmap_filename;
	u64 prev_time = 0;
	bool found_cmd_mmap = false,
	     found_libc_mmap = false,
	     found_vdso_mmap = false,
	     found_ld_mmap = false;
	int err = -1, errs = 0, i, wakeups = 0;
	u32 cpu;
	int total_events = 0, nr_events[PERF_RECORD_MAX] = { 0, };
	char sbuf[STRERR_BUFSIZE];

	if (evlist == NULL) /* Fallback for kernels lacking PERF_COUNT_SW_DUMMY */
		evlist = perf_evlist__new_default();

	if (evlist == NULL || argv == NULL) {
		pr_debug("Not enough memory to create evlist\n");
		goto out;
	}

	/*
	 * Create maps of threads and cpus to monitor. In this case
	 * we start with all threads and cpus (-1, -1) but then in
	 * perf_evlist__prepare_workload we'll fill in the only thread
	 * we're monitoring, the one forked there.
	 */
	err = perf_evlist__create_maps(evlist, &opts.target);
	if (err < 0) {
		pr_debug("Not enough memory to create thread/cpu maps\n");
		goto out_delete_evlist;
	}

	/*
	 * Prepare the workload in argv[] to run, it'll fork it, and then wait
	 * for perf_evlist__start_workload() to exec it. This is done this way
	 * so that we have time to open the evlist (calling sys_perf_event_open
	 * on all the fds) and then mmap them.
	 */
	err = perf_evlist__prepare_workload(evlist, &opts.target, argv, false, NULL);
	if (err < 0) {
		pr_debug("Couldn't run the workload!\n");
		goto out_delete_evlist;
	}

	/*
	 * Config the evsels, setting attr->comm on the first one, etc.
	 */
	evsel = perf_evlist__first(evlist);
	perf_evsel__set_sample_bit(evsel, CPU);
	perf_evsel__set_sample_bit(evsel, TID);
	perf_evsel__set_sample_bit(evsel, TIME);
	perf_evlist__config(evlist, &opts, NULL);

	err = sched__get_first_possible_cpu(evlist->workload.pid, &cpu_mask);
	if (err < 0) {
		pr_debug("sched__get_first_possible_cpu: %s\n",
			 str_error_r(errno, sbuf, sizeof(sbuf)));
		goto out_delete_evlist;
	}

	cpu = err;

	/*
	 * So that we can check perf_sample.cpu on all the samples.
	 */
	if (sched_setaffinity(evlist->workload.pid, cpu_mask_size, &cpu_mask) < 0) {
		pr_debug("sched_setaffinity: %s\n",
			 str_error_r(errno, sbuf, sizeof(sbuf)));
		goto out_delete_evlist;
	}

	/*
	 * Call sys_perf_event_open on all the fds on all the evsels,
	 * grouping them if asked to.
	 */
	err = perf_evlist__open(evlist);
	if (err < 0) {
		pr_debug("perf_evlist__open: %s\n",
			 str_error_r(errno, sbuf, sizeof(sbuf)));
		goto out_delete_evlist;
	}

	/*
	 * mmap the first fd on a given CPU and ask for events for the other
	 * fds in the same CPU to be injected in the same mmap ring buffer
	 * (using ioctl(PERF_EVENT_IOC_SET_OUTPUT)).
	 */
	err = perf_evlist__mmap(evlist, opts.mmap_pages, false);
	if (err < 0) {
		pr_debug("perf_evlist__mmap: %s\n",
			 str_error_r(errno, sbuf, sizeof(sbuf)));
		goto out_delete_evlist;
	}

	/*
	 * Now that all is properly set up, enable the events, they will
	 * count just on workload.pid, which will start...
	 */
	perf_evlist__enable(evlist);

	/*
	 * Now!
	 */
	perf_evlist__start_workload(evlist);

	while (1) {
		int before = total_events;

		for (i = 0; i < evlist->nr_mmaps; i++) {
			union perf_event *event;

			while ((event = perf_evlist__mmap_read(evlist, i)) != NULL) {
				const u32 type = event->header.type;
				const char *name = perf_event__name(type);

				++total_events;
				if (type < PERF_RECORD_MAX)
					nr_events[type]++;

				err = perf_evlist__parse_sample(evlist, event, &sample);
				if (err < 0) {
					if (verbose)
						perf_event__fprintf(event, stderr);
					pr_debug("Couldn't parse sample\n");
					goto out_delete_evlist;
				}

				if (verbose) {
					pr_info("%" PRIu64" %d ", sample.time, sample.cpu);
					perf_event__fprintf(event, stderr);
				}

				if (prev_time > sample.time) {
					pr_debug("%s going backwards in time, prev=%" PRIu64 ", curr=%" PRIu64 "\n",
						 name, prev_time, sample.time);
					++errs;
				}

				prev_time = sample.time;

				if (sample.cpu != cpu) {
					pr_debug("%s with unexpected cpu, expected %d, got %d\n",
						 name, cpu, sample.cpu);
					++errs;
				}

				if ((pid_t)sample.pid != evlist->workload.pid) {
					pr_debug("%s with unexpected pid, expected %d, got %d\n",
						 name, evlist->workload.pid, sample.pid);
					++errs;
				}

				if ((pid_t)sample.tid != evlist->workload.pid) {
					pr_debug("%s with unexpected tid, expected %d, got %d\n",
						 name, evlist->workload.pid, sample.tid);
					++errs;
				}

				if ((type == PERF_RECORD_COMM ||
				     type == PERF_RECORD_MMAP ||
				     type == PERF_RECORD_MMAP2 ||
				     type == PERF_RECORD_FORK ||
				     type == PERF_RECORD_EXIT) &&
				     (pid_t)event->comm.pid != evlist->workload.pid) {
					pr_debug("%s with unexpected pid/tid\n", name);
					++errs;
				}

				if ((type == PERF_RECORD_COMM ||
				     type == PERF_RECORD_MMAP ||
				     type == PERF_RECORD_MMAP2) &&
				     event->comm.pid != event->comm.tid) {
					pr_debug("%s with different pid/tid!\n", name);
					++errs;
				}

				switch (type) {
				case PERF_RECORD_COMM:
					if (strcmp(event->comm.comm, cmd)) {
						pr_debug("%s with unexpected comm!\n", name);
						++errs;
					}
					break;
				case PERF_RECORD_EXIT:
					goto found_exit;
				case PERF_RECORD_MMAP:
					mmap_filename = event->mmap.filename;
					goto check_bname;
				case PERF_RECORD_MMAP2:
					mmap_filename = event->mmap2.filename;
				check_bname:
					bname = strrchr(mmap_filename, '/');
					if (bname != NULL) {
						if (!found_cmd_mmap)
							found_cmd_mmap = !strcmp(bname + 1, cmd);
						if (!found_libc_mmap)
							found_libc_mmap = !strncmp(bname + 1, "libc", 4);
						if (!found_ld_mmap)
							found_ld_mmap = !strncmp(bname + 1, "ld", 2);
					} else if (!found_vdso_mmap)
						found_vdso_mmap = !strcmp(mmap_filename, "[vdso]");
					break;

				case PERF_RECORD_SAMPLE:
					/* Just ignore samples for now */
					break;
				default:
					pr_debug("Unexpected perf_event->header.type %d!\n",
						 type);
					++errs;
				}

				perf_evlist__mmap_consume(evlist, i);
			}
		}

		/*
		 * We don't use poll here because at least at 3.1 times the
		 * PERF_RECORD_{!SAMPLE} events don't honour
		 * perf_event_attr.wakeup_events, just PERF_EVENT_SAMPLE does.
		 */
		if (total_events == before && false)
			perf_evlist__poll(evlist, -1);

		sleep(1);
		if (++wakeups > 5) {
			pr_debug("No PERF_RECORD_EXIT event!\n");
			break;
		}
	}

found_exit:
	if (nr_events[PERF_RECORD_COMM] > 1) {
		pr_debug("Excessive number of PERF_RECORD_COMM events!\n");
		++errs;
	}

	if (nr_events[PERF_RECORD_COMM] == 0) {
		pr_debug("Missing PERF_RECORD_COMM for %s!\n", cmd);
		++errs;
	}

	if (!found_cmd_mmap) {
		pr_debug("PERF_RECORD_MMAP for %s missing!\n", cmd);
		++errs;
	}

	if (!found_libc_mmap) {
		pr_debug("PERF_RECORD_MMAP for %s missing!\n", "libc");
		++errs;
	}

	if (!found_ld_mmap) {
		pr_debug("PERF_RECORD_MMAP for %s missing!\n", "ld");
		++errs;
	}

	if (!found_vdso_mmap) {
		pr_debug("PERF_RECORD_MMAP for %s missing!\n", "[vdso]");
		++errs;
	}
out_delete_evlist:
	perf_evlist__delete(evlist);
out:
	return (err < 0 || errs > 0) ? -1 : 0;
}
Commit	Line	Data
1fbe7df8 ACM	1	/* For the CLR_() macros */
	2	#include <pthread.h>
	3
16d00fee JO	4	#include <sched.h>
	5	#include "evlist.h"
	6	#include "evsel.h"
	7	#include "perf.h"
	8	#include "debug.h"
	9	#include "tests.h"
	10
	11	static int sched__get_first_possible_cpu(pid_t pid, cpu_set_t *maskp)
	12	{
	13	int i, cpu = -1, nrcpus = 1024;
	14	realloc:
	15	CPU_ZERO(maskp);
	16
	17	if (sched_getaffinity(pid, sizeof(*maskp), maskp) == -1) {
	18	if (errno == EINVAL && nrcpus < (1024 << 8)) {
	19	nrcpus = nrcpus << 2;
	20	goto realloc;
	21	}
	22	perror("sched_getaffinity");
	23	return -1;
	24	}
	25
	26	for (i = 0; i < nrcpus; i++) {
	27	if (CPU_ISSET(i, maskp)) {
	28	if (cpu == -1)
	29	cpu = i;
	30	else
	31	CPU_CLR(i, maskp);
	32	}
	33	}
	34
	35	return cpu;
	36	}
	37
721a1f53	38	int test__PERF_RECORD(int subtest __maybe_unused)
16d00fee	39	{
b4006796	40	struct record_opts opts = {
16d00fee JO	41	.target = {
	42	.uid = UINT_MAX,
	43	.uses_mmap = true,
	44	},
509051ea	45	.no_buffering = true,
509051ea	46	.mmap_pages = 256,
16d00fee JO	47	};
	48	cpu_set_t cpu_mask;
	49	size_t cpu_mask_size = sizeof(cpu_mask);
69ef8f47	50	struct perf_evlist *evlist = perf_evlist__new_dummy();
16d00fee JO	51	struct perf_evsel *evsel;
	52	struct perf_sample sample;
	53	const char *cmd = "sleep";
	54	const char *argv[] = { cmd, "1", NULL, };
5c5e854b	55	char bname, mmap_filename;
16d00fee JO	56	u64 prev_time = 0;
	57	bool found_cmd_mmap = false,
	58	found_libc_mmap = false,
	59	found_vdso_mmap = false,
	60	found_ld_mmap = false;
	61	int err = -1, errs = 0, i, wakeups = 0;
	62	u32 cpu;
	63	int total_events = 0, nr_events[PERF_RECORD_MAX] = { 0, };
ba3dfff8	64	char sbuf[STRERR_BUFSIZE];
16d00fee	65
69ef8f47 ACM	66	if (evlist == NULL) /* Fallback for kernels lacking PERF_COUNT_SW_DUMMY */
	67	evlist = perf_evlist__new_default();
	68
16d00fee JO	69	if (evlist == NULL \|\| argv == NULL) {
	70	pr_debug("Not enough memory to create evlist\n");
	71	goto out;
	72	}
	73
16d00fee JO	74	/*
	75	* Create maps of threads and cpus to monitor. In this case
	76	* we start with all threads and cpus (-1, -1) but then in
	77	* perf_evlist__prepare_workload we'll fill in the only thread
	78	* we're monitoring, the one forked there.
	79	*/
	80	err = perf_evlist__create_maps(evlist, &opts.target);
	81	if (err < 0) {
	82	pr_debug("Not enough memory to create thread/cpu maps\n");
	83	goto out_delete_evlist;
	84	}
	85
	86	/*
	87	* Prepare the workload in argv[] to run, it'll fork it, and then wait
	88	* for perf_evlist__start_workload() to exec it. This is done this way
	89	* so that we have time to open the evlist (calling sys_perf_event_open
	90	* on all the fds) and then mmap them.
	91	*/
735f7e0b	92	err = perf_evlist__prepare_workload(evlist, &opts.target, argv, false, NULL);
16d00fee JO	93	if (err < 0) {
16d00fee JO	94	pr_debug("Couldn't run the workload!\n");
03ad9747	95	goto out_delete_evlist;
16d00fee JO	96	}
	97
	98	/*
	99	* Config the evsels, setting attr->comm on the first one, etc.
	100	*/
	101	evsel = perf_evlist__first(evlist);
7be5ebe8 ACM	102	perf_evsel__set_sample_bit(evsel, CPU);
	103	perf_evsel__set_sample_bit(evsel, TID);
	104	perf_evsel__set_sample_bit(evsel, TIME);
e68ae9cf	105	perf_evlist__config(evlist, &opts, NULL);
16d00fee JO	106
	107	err = sched__get_first_possible_cpu(evlist->workload.pid, &cpu_mask);
	108	if (err < 0) {
ba3dfff8	109	pr_debug("sched__get_first_possible_cpu: %s\n",
c8b5f2c9	110	str_error_r(errno, sbuf, sizeof(sbuf)));
03ad9747	111	goto out_delete_evlist;
16d00fee JO	112	}
	113
	114	cpu = err;
	115
	116	/*
	117	* So that we can check perf_sample.cpu on all the samples.
	118	*/
	119	if (sched_setaffinity(evlist->workload.pid, cpu_mask_size, &cpu_mask) < 0) {
ba3dfff8	120	pr_debug("sched_setaffinity: %s\n",
c8b5f2c9	121	str_error_r(errno, sbuf, sizeof(sbuf)));
03ad9747	122	goto out_delete_evlist;
16d00fee JO	123	}
	124
	125	/*
	126	* Call sys_perf_event_open on all the fds on all the evsels,
	127	* grouping them if asked to.
	128	*/
	129	err = perf_evlist__open(evlist);
	130	if (err < 0) {
ba3dfff8	131	pr_debug("perf_evlist__open: %s\n",
c8b5f2c9	132	str_error_r(errno, sbuf, sizeof(sbuf)));
03ad9747	133	goto out_delete_evlist;
16d00fee JO	134	}
	135
	136	/*
	137	* mmap the first fd on a given CPU and ask for events for the other
	138	* fds in the same CPU to be injected in the same mmap ring buffer
	139	* (using ioctl(PERF_EVENT_IOC_SET_OUTPUT)).
	140	*/
	141	err = perf_evlist__mmap(evlist, opts.mmap_pages, false);
	142	if (err < 0) {
ba3dfff8	143	pr_debug("perf_evlist__mmap: %s\n",
c8b5f2c9	144	str_error_r(errno, sbuf, sizeof(sbuf)));
f26e1c7c	145	goto out_delete_evlist;
16d00fee JO	146	}
	147
	148	/*
	149	* Now that all is properly set up, enable the events, they will
	150	* count just on workload.pid, which will start...
	151	*/
	152	perf_evlist__enable(evlist);
	153
	154	/*
	155	* Now!
	156	*/
	157	perf_evlist__start_workload(evlist);
	158
	159	while (1) {
	160	int before = total_events;
	161
	162	for (i = 0; i < evlist->nr_mmaps; i++) {
	163	union perf_event *event;
	164
	165	while ((event = perf_evlist__mmap_read(evlist, i)) != NULL) {
	166	const u32 type = event->header.type;
	167	const char *name = perf_event__name(type);
	168
	169	++total_events;
	170	if (type < PERF_RECORD_MAX)
	171	nr_events[type]++;
	172
	173	err = perf_evlist__parse_sample(evlist, event, &sample);
	174	if (err < 0) {
	175	if (verbose)
	176	perf_event__fprintf(event, stderr);
	177	pr_debug("Couldn't parse sample\n");
983874d1	178	goto out_delete_evlist;
16d00fee JO	179	}
	180
	181	if (verbose) {
	182	pr_info("%" PRIu64" %d ", sample.time, sample.cpu);
	183	perf_event__fprintf(event, stderr);
	184	}
	185
	186	if (prev_time > sample.time) {
	187	pr_debug("%s going backwards in time, prev=%" PRIu64 ", curr=%" PRIu64 "\n",
	188	name, prev_time, sample.time);
	189	++errs;
	190	}
	191
	192	prev_time = sample.time;
	193
	194	if (sample.cpu != cpu) {
	195	pr_debug("%s with unexpected cpu, expected %d, got %d\n",
	196	name, cpu, sample.cpu);
	197	++errs;
	198	}
	199
	200	if ((pid_t)sample.pid != evlist->workload.pid) {
	201	pr_debug("%s with unexpected pid, expected %d, got %d\n",
	202	name, evlist->workload.pid, sample.pid);
	203	++errs;
	204	}
	205
	206	if ((pid_t)sample.tid != evlist->workload.pid) {
	207	pr_debug("%s with unexpected tid, expected %d, got %d\n",
	208	name, evlist->workload.pid, sample.tid);
	209	++errs;
	210	}
	211
	212	if ((type == PERF_RECORD_COMM \|\|
	213	type == PERF_RECORD_MMAP \|\|
5c5e854b	214	type == PERF_RECORD_MMAP2 \|\|
16d00fee JO	215	type == PERF_RECORD_FORK \|\|
	216	type == PERF_RECORD_EXIT) &&
	217	(pid_t)event->comm.pid != evlist->workload.pid) {
	218	pr_debug("%s with unexpected pid/tid\n", name);
	219	++errs;
	220	}
	221
	222	if ((type == PERF_RECORD_COMM \|\|
5c5e854b SE	223	type == PERF_RECORD_MMAP \|\|
5c5e854b SE	224	type == PERF_RECORD_MMAP2) &&
16d00fee JO	225	event->comm.pid != event->comm.tid) {
	226	pr_debug("%s with different pid/tid!\n", name);
	227	++errs;
	228	}
	229
	230	switch (type) {
	231	case PERF_RECORD_COMM:
	232	if (strcmp(event->comm.comm, cmd)) {
	233	pr_debug("%s with unexpected comm!\n", name);
	234	++errs;
	235	}
	236	break;
	237	case PERF_RECORD_EXIT:
	238	goto found_exit;
	239	case PERF_RECORD_MMAP:
5c5e854b SE	240	mmap_filename = event->mmap.filename;
	241	goto check_bname;
	242	case PERF_RECORD_MMAP2:
	243	mmap_filename = event->mmap2.filename;
	244	check_bname:
	245	bname = strrchr(mmap_filename, '/');
16d00fee JO	246	if (bname != NULL) {
	247	if (!found_cmd_mmap)
	248	found_cmd_mmap = !strcmp(bname + 1, cmd);
	249	if (!found_libc_mmap)
	250	found_libc_mmap = !strncmp(bname + 1, "libc", 4);
	251	if (!found_ld_mmap)
	252	found_ld_mmap = !strncmp(bname + 1, "ld", 2);
	253	} else if (!found_vdso_mmap)
5c5e854b	254	found_vdso_mmap = !strcmp(mmap_filename, "[vdso]");
16d00fee JO	255	break;
	256
	257	case PERF_RECORD_SAMPLE:
	258	/* Just ignore samples for now */
	259	break;
	260	default:
	261	pr_debug("Unexpected perf_event->header.type %d!\n",
	262	type);
	263	++errs;
	264	}
8e50d384 ZZ	265
8e50d384 ZZ	266	perf_evlist__mmap_consume(evlist, i);
16d00fee JO	267	}
	268	}
	269
	270	/*
	271	* We don't use poll here because at least at 3.1 times the
	272	* PERF_RECORD_{!SAMPLE} events don't honour
	273	* perf_event_attr.wakeup_events, just PERF_EVENT_SAMPLE does.
	274	*/
	275	if (total_events == before && false)
f66a889d	276	perf_evlist__poll(evlist, -1);
16d00fee JO	277
	278	sleep(1);
	279	if (++wakeups > 5) {
	280	pr_debug("No PERF_RECORD_EXIT event!\n");
	281	break;
	282	}
	283	}
	284
	285	found_exit:
	286	if (nr_events[PERF_RECORD_COMM] > 1) {
	287	pr_debug("Excessive number of PERF_RECORD_COMM events!\n");
	288	++errs;
	289	}
	290
	291	if (nr_events[PERF_RECORD_COMM] == 0) {
	292	pr_debug("Missing PERF_RECORD_COMM for %s!\n", cmd);
	293	++errs;
	294	}
	295
	296	if (!found_cmd_mmap) {
	297	pr_debug("PERF_RECORD_MMAP for %s missing!\n", cmd);
	298	++errs;
	299	}
	300
	301	if (!found_libc_mmap) {
	302	pr_debug("PERF_RECORD_MMAP for %s missing!\n", "libc");
	303	++errs;
	304	}
	305
	306	if (!found_ld_mmap) {
	307	pr_debug("PERF_RECORD_MMAP for %s missing!\n", "ld");
	308	++errs;
	309	}
	310
	311	if (!found_vdso_mmap) {
	312	pr_debug("PERF_RECORD_MMAP for %s missing!\n", "[vdso]");
	313	++errs;
	314	}
16d00fee JO	315	out_delete_evlist:
	316	perf_evlist__delete(evlist);
	317	out:
	318	return (err < 0 \|\| errs > 0) ? -1 : 0;
	319	}