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

#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);

	err = sched__get_first_possible_cpu(evlist->workload.pid, &cpu_mask);
	if (err < 0) {
		pr_debug("sched__get_first_possible_cpu: %s\n",
			 strerror_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",
			 strerror_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",
			 strerror_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",
			 strerror_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
16d00fee JO	1	#include <sched.h>
	2	#include "evlist.h"
	3	#include "evsel.h"
	4	#include "perf.h"
	5	#include "debug.h"
	6	#include "tests.h"
	7
	8	static int sched__get_first_possible_cpu(pid_t pid, cpu_set_t *maskp)
	9	{
	10	int i, cpu = -1, nrcpus = 1024;
	11	realloc:
	12	CPU_ZERO(maskp);
	13
	14	if (sched_getaffinity(pid, sizeof(*maskp), maskp) == -1) {
	15	if (errno == EINVAL && nrcpus < (1024 << 8)) {
	16	nrcpus = nrcpus << 2;
	17	goto realloc;
	18	}
	19	perror("sched_getaffinity");
	20	return -1;
	21	}
	22
	23	for (i = 0; i < nrcpus; i++) {
	24	if (CPU_ISSET(i, maskp)) {
	25	if (cpu == -1)
	26	cpu = i;
	27	else
	28	CPU_CLR(i, maskp);
	29	}
	30	}
	31
	32	return cpu;
	33	}
	34
721a1f53	35	int test__PERF_RECORD(int subtest __maybe_unused)
16d00fee	36	{
b4006796	37	struct record_opts opts = {
16d00fee JO	38	.target = {
	39	.uid = UINT_MAX,
	40	.uses_mmap = true,
	41	},
509051ea	42	.no_buffering = true,
509051ea	43	.mmap_pages = 256,
16d00fee JO	44	};
	45	cpu_set_t cpu_mask;
	46	size_t cpu_mask_size = sizeof(cpu_mask);
69ef8f47	47	struct perf_evlist *evlist = perf_evlist__new_dummy();
16d00fee JO	48	struct perf_evsel *evsel;
	49	struct perf_sample sample;
	50	const char *cmd = "sleep";
	51	const char *argv[] = { cmd, "1", NULL, };
5c5e854b	52	char bname, mmap_filename;
16d00fee JO	53	u64 prev_time = 0;
	54	bool found_cmd_mmap = false,
	55	found_libc_mmap = false,
	56	found_vdso_mmap = false,
	57	found_ld_mmap = false;
	58	int err = -1, errs = 0, i, wakeups = 0;
	59	u32 cpu;
	60	int total_events = 0, nr_events[PERF_RECORD_MAX] = { 0, };
ba3dfff8	61	char sbuf[STRERR_BUFSIZE];
16d00fee	62
69ef8f47 ACM	63	if (evlist == NULL) /* Fallback for kernels lacking PERF_COUNT_SW_DUMMY */
	64	evlist = perf_evlist__new_default();
	65
16d00fee JO	66	if (evlist == NULL \|\| argv == NULL) {
	67	pr_debug("Not enough memory to create evlist\n");
	68	goto out;
	69	}
	70
16d00fee JO	71	/*
	72	* Create maps of threads and cpus to monitor. In this case
	73	* we start with all threads and cpus (-1, -1) but then in
	74	* perf_evlist__prepare_workload we'll fill in the only thread
	75	* we're monitoring, the one forked there.
	76	*/
	77	err = perf_evlist__create_maps(evlist, &opts.target);
	78	if (err < 0) {
	79	pr_debug("Not enough memory to create thread/cpu maps\n");
	80	goto out_delete_evlist;
	81	}
	82
	83	/*
	84	* Prepare the workload in argv[] to run, it'll fork it, and then wait
	85	* for perf_evlist__start_workload() to exec it. This is done this way
	86	* so that we have time to open the evlist (calling sys_perf_event_open
	87	* on all the fds) and then mmap them.
	88	*/
735f7e0b	89	err = perf_evlist__prepare_workload(evlist, &opts.target, argv, false, NULL);
16d00fee JO	90	if (err < 0) {
16d00fee JO	91	pr_debug("Couldn't run the workload!\n");
03ad9747	92	goto out_delete_evlist;
16d00fee JO	93	}
	94
	95	/*
	96	* Config the evsels, setting attr->comm on the first one, etc.
	97	*/
	98	evsel = perf_evlist__first(evlist);
7be5ebe8 ACM	99	perf_evsel__set_sample_bit(evsel, CPU);
	100	perf_evsel__set_sample_bit(evsel, TID);
	101	perf_evsel__set_sample_bit(evsel, TIME);
f77a9518	102	perf_evlist__config(evlist, &opts);
16d00fee JO	103
	104	err = sched__get_first_possible_cpu(evlist->workload.pid, &cpu_mask);
	105	if (err < 0) {
ba3dfff8 MH	106	pr_debug("sched__get_first_possible_cpu: %s\n",
ba3dfff8 MH	107	strerror_r(errno, sbuf, sizeof(sbuf)));
03ad9747	108	goto out_delete_evlist;
16d00fee JO	109	}
	110
	111	cpu = err;
	112
	113	/*
	114	* So that we can check perf_sample.cpu on all the samples.
	115	*/
	116	if (sched_setaffinity(evlist->workload.pid, cpu_mask_size, &cpu_mask) < 0) {
ba3dfff8 MH	117	pr_debug("sched_setaffinity: %s\n",
ba3dfff8 MH	118	strerror_r(errno, sbuf, sizeof(sbuf)));
03ad9747	119	goto out_delete_evlist;
16d00fee JO	120	}
	121
	122	/*
	123	* Call sys_perf_event_open on all the fds on all the evsels,
	124	* grouping them if asked to.
	125	*/
	126	err = perf_evlist__open(evlist);
	127	if (err < 0) {
ba3dfff8 MH	128	pr_debug("perf_evlist__open: %s\n",
ba3dfff8 MH	129	strerror_r(errno, sbuf, sizeof(sbuf)));
03ad9747	130	goto out_delete_evlist;
16d00fee JO	131	}
	132
	133	/*
	134	* mmap the first fd on a given CPU and ask for events for the other
	135	* fds in the same CPU to be injected in the same mmap ring buffer
	136	* (using ioctl(PERF_EVENT_IOC_SET_OUTPUT)).
	137	*/
	138	err = perf_evlist__mmap(evlist, opts.mmap_pages, false);
	139	if (err < 0) {
ba3dfff8 MH	140	pr_debug("perf_evlist__mmap: %s\n",
ba3dfff8 MH	141	strerror_r(errno, sbuf, sizeof(sbuf)));
f26e1c7c	142	goto out_delete_evlist;
16d00fee JO	143	}
	144
	145	/*
	146	* Now that all is properly set up, enable the events, they will
	147	* count just on workload.pid, which will start...
	148	*/
	149	perf_evlist__enable(evlist);
	150
	151	/*
	152	* Now!
	153	*/
	154	perf_evlist__start_workload(evlist);
	155
	156	while (1) {
	157	int before = total_events;
	158
	159	for (i = 0; i < evlist->nr_mmaps; i++) {
	160	union perf_event *event;
	161
	162	while ((event = perf_evlist__mmap_read(evlist, i)) != NULL) {
	163	const u32 type = event->header.type;
	164	const char *name = perf_event__name(type);
	165
	166	++total_events;
	167	if (type < PERF_RECORD_MAX)
	168	nr_events[type]++;
	169
	170	err = perf_evlist__parse_sample(evlist, event, &sample);
	171	if (err < 0) {
	172	if (verbose)
	173	perf_event__fprintf(event, stderr);
	174	pr_debug("Couldn't parse sample\n");
983874d1	175	goto out_delete_evlist;
16d00fee JO	176	}
	177
	178	if (verbose) {
	179	pr_info("%" PRIu64" %d ", sample.time, sample.cpu);
	180	perf_event__fprintf(event, stderr);
	181	}
	182
	183	if (prev_time > sample.time) {
	184	pr_debug("%s going backwards in time, prev=%" PRIu64 ", curr=%" PRIu64 "\n",
	185	name, prev_time, sample.time);
	186	++errs;
	187	}
	188
	189	prev_time = sample.time;
	190
	191	if (sample.cpu != cpu) {
	192	pr_debug("%s with unexpected cpu, expected %d, got %d\n",
	193	name, cpu, sample.cpu);
	194	++errs;
	195	}
	196
	197	if ((pid_t)sample.pid != evlist->workload.pid) {
	198	pr_debug("%s with unexpected pid, expected %d, got %d\n",
	199	name, evlist->workload.pid, sample.pid);
	200	++errs;
	201	}
	202
	203	if ((pid_t)sample.tid != evlist->workload.pid) {
	204	pr_debug("%s with unexpected tid, expected %d, got %d\n",
	205	name, evlist->workload.pid, sample.tid);
	206	++errs;
	207	}
	208
	209	if ((type == PERF_RECORD_COMM \|\|
	210	type == PERF_RECORD_MMAP \|\|
5c5e854b	211	type == PERF_RECORD_MMAP2 \|\|
16d00fee JO	212	type == PERF_RECORD_FORK \|\|
	213	type == PERF_RECORD_EXIT) &&
	214	(pid_t)event->comm.pid != evlist->workload.pid) {
	215	pr_debug("%s with unexpected pid/tid\n", name);
	216	++errs;
	217	}
	218
	219	if ((type == PERF_RECORD_COMM \|\|
5c5e854b SE	220	type == PERF_RECORD_MMAP \|\|
5c5e854b SE	221	type == PERF_RECORD_MMAP2) &&
16d00fee JO	222	event->comm.pid != event->comm.tid) {
	223	pr_debug("%s with different pid/tid!\n", name);
	224	++errs;
	225	}
	226
	227	switch (type) {
	228	case PERF_RECORD_COMM:
	229	if (strcmp(event->comm.comm, cmd)) {
	230	pr_debug("%s with unexpected comm!\n", name);
	231	++errs;
	232	}
	233	break;
	234	case PERF_RECORD_EXIT:
	235	goto found_exit;
	236	case PERF_RECORD_MMAP:
5c5e854b SE	237	mmap_filename = event->mmap.filename;
	238	goto check_bname;
	239	case PERF_RECORD_MMAP2:
	240	mmap_filename = event->mmap2.filename;
	241	check_bname:
	242	bname = strrchr(mmap_filename, '/');
16d00fee JO	243	if (bname != NULL) {
	244	if (!found_cmd_mmap)
	245	found_cmd_mmap = !strcmp(bname + 1, cmd);
	246	if (!found_libc_mmap)
	247	found_libc_mmap = !strncmp(bname + 1, "libc", 4);
	248	if (!found_ld_mmap)
	249	found_ld_mmap = !strncmp(bname + 1, "ld", 2);
	250	} else if (!found_vdso_mmap)
5c5e854b	251	found_vdso_mmap = !strcmp(mmap_filename, "[vdso]");
16d00fee JO	252	break;
	253
	254	case PERF_RECORD_SAMPLE:
	255	/* Just ignore samples for now */
	256	break;
	257	default:
	258	pr_debug("Unexpected perf_event->header.type %d!\n",
	259	type);
	260	++errs;
	261	}
8e50d384 ZZ	262
8e50d384 ZZ	263	perf_evlist__mmap_consume(evlist, i);
16d00fee JO	264	}
	265	}
	266
	267	/*
	268	* We don't use poll here because at least at 3.1 times the
	269	* PERF_RECORD_{!SAMPLE} events don't honour
	270	* perf_event_attr.wakeup_events, just PERF_EVENT_SAMPLE does.
	271	*/
	272	if (total_events == before && false)
f66a889d	273	perf_evlist__poll(evlist, -1);
16d00fee JO	274
	275	sleep(1);
	276	if (++wakeups > 5) {
	277	pr_debug("No PERF_RECORD_EXIT event!\n");
	278	break;
	279	}
	280	}
	281
	282	found_exit:
	283	if (nr_events[PERF_RECORD_COMM] > 1) {
	284	pr_debug("Excessive number of PERF_RECORD_COMM events!\n");
	285	++errs;
	286	}
	287
	288	if (nr_events[PERF_RECORD_COMM] == 0) {
	289	pr_debug("Missing PERF_RECORD_COMM for %s!\n", cmd);
	290	++errs;
	291	}
	292
	293	if (!found_cmd_mmap) {
	294	pr_debug("PERF_RECORD_MMAP for %s missing!\n", cmd);
	295	++errs;
	296	}
	297
	298	if (!found_libc_mmap) {
	299	pr_debug("PERF_RECORD_MMAP for %s missing!\n", "libc");
	300	++errs;
	301	}
	302
	303	if (!found_ld_mmap) {
	304	pr_debug("PERF_RECORD_MMAP for %s missing!\n", "ld");
	305	++errs;
	306	}
	307
	308	if (!found_vdso_mmap) {
	309	pr_debug("PERF_RECORD_MMAP for %s missing!\n", "[vdso]");
	310	++errs;
	311	}
16d00fee JO	312	out_delete_evlist:
	313	perf_evlist__delete(evlist);
	314	out:
	315	return (err < 0 \|\| errs > 0) ? -1 : 0;
	316	}