[linux-2.6-block.git] / Documentation / perf_counter / builtin-record.c



#include "perf.h"
#include "builtin.h"
#include "util/util.h"
#include "util/parse-options.h"
#include "util/parse-events.h"
#include "util/string.h"

#include <unistd.h>
#include <sched.h>

#define ALIGN(x, a)		__ALIGN_MASK(x, (typeof(x))(a)-1)
#define __ALIGN_MASK(x, mask)	(((x)+(mask))&~(mask))

static int			default_interval = 100000;
static int			event_count[MAX_COUNTERS];

static int			fd[MAX_NR_CPUS][MAX_COUNTERS];
static int			nr_cpus				=  0;
static unsigned int		page_size;
static unsigned int		mmap_pages			= 16;
static int			output;
static const char		*output_name			= "perf.data";
static int			group				= 0;
static unsigned int		realtime_prio			= 0;
static int			system_wide			= 0;
static pid_t			target_pid			= -1;
static int			inherit				= 1;
static int			force				= 0;

const unsigned int default_count[] = {
	1000000,
	1000000,
	  10000,
	  10000,
	1000000,
	  10000,
};

struct mmap_data {
	int counter;
	void *base;
	unsigned int mask;
	unsigned int prev;
};

static unsigned int mmap_read_head(struct mmap_data *md)
{
	struct perf_counter_mmap_page *pc = md->base;
	int head;

	head = pc->data_head;
	rmb();

	return head;
}

static long events;
static struct timeval last_read, this_read;

static void mmap_read(struct mmap_data *md)
{
	unsigned int head = mmap_read_head(md);
	unsigned int old = md->prev;
	unsigned char *data = md->base + page_size;
	unsigned long size;
	void *buf;
	int diff;

	gettimeofday(&this_read, NULL);

	/*
	 * If we're further behind than half the buffer, there's a chance
	 * the writer will bite our tail and screw up the events under us.
	 *
	 * If we somehow ended up ahead of the head, we got messed up.
	 *
	 * In either case, truncate and restart at head.
	 */
	diff = head - old;
	if (diff > md->mask / 2 || diff < 0) {
		struct timeval iv;
		unsigned long msecs;

		timersub(&this_read, &last_read, &iv);
		msecs = iv.tv_sec*1000 + iv.tv_usec/1000;

		fprintf(stderr, "WARNING: failed to keep up with mmap data."
				"  Last read %lu msecs ago.\n", msecs);

		/*
		 * head points to a known good entry, start there.
		 */
		old = head;
	}

	last_read = this_read;

	if (old != head)
		events++;

	size = head - old;

	if ((old & md->mask) + size != (head & md->mask)) {
		buf = &data[old & md->mask];
		size = md->mask + 1 - (old & md->mask);
		old += size;
		while (size) {
			int ret = write(output, buf, size);
			if (ret < 0) {
				perror("failed to write");
				exit(-1);
			}
			size -= ret;
			buf += ret;
		}
	}

	buf = &data[old & md->mask];
	size = head - old;
	old += size;
	while (size) {
		int ret = write(output, buf, size);
		if (ret < 0) {
			perror("failed to write");
			exit(-1);
		}
		size -= ret;
		buf += ret;
	}

	md->prev = old;
}

static volatile int done = 0;

static void sig_handler(int sig)
{
	done = 1;
}

static struct pollfd event_array[MAX_NR_CPUS * MAX_COUNTERS];
static struct mmap_data mmap_array[MAX_NR_CPUS][MAX_COUNTERS];

static int nr_poll;
static int nr_cpu;

struct mmap_event {
	struct perf_event_header	header;
	__u32				pid;
	__u32				tid;
	__u64				start;
	__u64				len;
	__u64				pgoff;
	char				filename[PATH_MAX];
};

struct comm_event {
	struct perf_event_header	header;
	__u32				pid;
	__u32				tid;
	char				comm[16];
};

static void pid_synthesize_comm_event(pid_t pid, int full)
{
	struct comm_event comm_ev;
	char filename[PATH_MAX];
	char bf[BUFSIZ];
	int fd, ret;
	size_t size;
	char *field, *sep;
	DIR *tasks;
	struct dirent dirent, *next;

	snprintf(filename, sizeof(filename), "/proc/%d/stat", pid);

	fd = open(filename, O_RDONLY);
	if (fd < 0) {
		fprintf(stderr, "couldn't open %s\n", filename);
		exit(EXIT_FAILURE);
	}
	if (read(fd, bf, sizeof(bf)) < 0) {
		fprintf(stderr, "couldn't read %s\n", filename);
		exit(EXIT_FAILURE);
	}
	close(fd);

	/* 9027 (cat) R 6747 9027 6747 34816 9027 ... */
	memset(&comm_ev, 0, sizeof(comm_ev));
	field = strchr(bf, '(');
	if (field == NULL)
		goto out_failure;
	sep = strchr(++field, ')');
	if (sep == NULL)
		goto out_failure;
	size = sep - field;
	memcpy(comm_ev.comm, field, size++);

	comm_ev.pid = pid;
	comm_ev.header.type = PERF_EVENT_COMM;
	size = ALIGN(size, sizeof(uint64_t));
	comm_ev.header.size = sizeof(comm_ev) - (sizeof(comm_ev.comm) - size);

	if (!full) {
		comm_ev.tid = pid;

		ret = write(output, &comm_ev, comm_ev.header.size);
		if (ret < 0) {
			perror("failed to write");
			exit(-1);
		}
		return;
	}

	snprintf(filename, sizeof(filename), "/proc/%d/task", pid);

	tasks = opendir(filename);
	while (!readdir_r(tasks, &dirent, &next) && next) {
		char *end;
		pid = strtol(dirent.d_name, &end, 10);
		if (*end)
			continue;

		comm_ev.tid = pid;

		ret = write(output, &comm_ev, comm_ev.header.size);
		if (ret < 0) {
			perror("failed to write");
			exit(-1);
		}
	}
	closedir(tasks);
	return;

out_failure:
	fprintf(stderr, "couldn't get COMM and pgid, malformed %s\n",
		filename);
	exit(EXIT_FAILURE);
}

static void pid_synthesize_mmap_events(pid_t pid)
{
	char filename[PATH_MAX];
	FILE *fp;

	snprintf(filename, sizeof(filename), "/proc/%d/maps", pid);

	fp = fopen(filename, "r");
	if (fp == NULL) {
		fprintf(stderr, "couldn't open %s\n", filename);
		exit(EXIT_FAILURE);
	}
	while (1) {
		char bf[BUFSIZ], *pbf = bf;
		struct mmap_event mmap_ev = {
			.header.type = PERF_EVENT_MMAP,
		};
		int n;
		size_t size;
		if (fgets(bf, sizeof(bf), fp) == NULL)
			break;

		/* 00400000-0040c000 r-xp 00000000 fd:01 41038  /bin/cat */
		n = hex2u64(pbf, &mmap_ev.start);
		if (n < 0)
			continue;
		pbf += n + 1;
		n = hex2u64(pbf, &mmap_ev.len);
		if (n < 0)
			continue;
		pbf += n + 3;
		if (*pbf == 'x') { /* vm_exec */
			char *execname = strrchr(bf, ' ');

			if (execname == NULL || execname[1] != '/')
				continue;

			execname += 1;
			size = strlen(execname);
			execname[size - 1] = '\0'; /* Remove \n */
			memcpy(mmap_ev.filename, execname, size);
			size = ALIGN(size, sizeof(uint64_t));
			mmap_ev.len -= mmap_ev.start;
			mmap_ev.header.size = (sizeof(mmap_ev) -
					       (sizeof(mmap_ev.filename) - size));
			mmap_ev.pid = pid;
			mmap_ev.tid = pid;

			if (write(output, &mmap_ev, mmap_ev.header.size) < 0) {
				perror("failed to write");
				exit(-1);
			}
		}
	}

	fclose(fp);
}

static void synthesize_events(void)
{
	DIR *proc;
	struct dirent dirent, *next;

	proc = opendir("/proc");

	while (!readdir_r(proc, &dirent, &next) && next) {
		char *end;
		pid_t pid;

		pid = strtol(dirent.d_name, &end, 10);
		if (*end) /* only interested in proper numerical dirents */
			continue;

		pid_synthesize_comm_event(pid, 1);
		pid_synthesize_mmap_events(pid);
	}

	closedir(proc);
}

static void open_counters(int cpu, pid_t pid)
{
	struct perf_counter_attr attr;
	int counter, group_fd;
	int track = 1;

	if (pid > 0) {
		pid_synthesize_comm_event(pid, 0);
		pid_synthesize_mmap_events(pid);
	}

	group_fd = -1;
	for (counter = 0; counter < nr_counters; counter++) {

		memset(&attr, 0, sizeof(attr));
		attr.config		= event_id[counter];
		attr.sample_period	= event_count[counter];
		attr.sample_type	= PERF_SAMPLE_IP | PERF_SAMPLE_TID;
		attr.mmap		= track;
		attr.comm		= track;
		attr.inherit	= (cpu < 0) && inherit;

		track = 0; // only the first counter needs these

		fd[nr_cpu][counter] =
			sys_perf_counter_open(&attr, pid, cpu, group_fd, 0);

		if (fd[nr_cpu][counter] < 0) {
			int err = errno;
			printf("kerneltop error: syscall returned with %d (%s)\n",
					fd[nr_cpu][counter], strerror(err));
			if (err == EPERM)
				printf("Are you root?\n");
			exit(-1);
		}
		assert(fd[nr_cpu][counter] >= 0);
		fcntl(fd[nr_cpu][counter], F_SETFL, O_NONBLOCK);

		/*
		 * First counter acts as the group leader:
		 */
		if (group && group_fd == -1)
			group_fd = fd[nr_cpu][counter];

		event_array[nr_poll].fd = fd[nr_cpu][counter];
		event_array[nr_poll].events = POLLIN;
		nr_poll++;

		mmap_array[nr_cpu][counter].counter = counter;
		mmap_array[nr_cpu][counter].prev = 0;
		mmap_array[nr_cpu][counter].mask = mmap_pages*page_size - 1;
		mmap_array[nr_cpu][counter].base = mmap(NULL, (mmap_pages+1)*page_size,
				PROT_READ, MAP_SHARED, fd[nr_cpu][counter], 0);
		if (mmap_array[nr_cpu][counter].base == MAP_FAILED) {
			printf("kerneltop error: failed to mmap with %d (%s)\n",
					errno, strerror(errno));
			exit(-1);
		}
	}
	nr_cpu++;
}

static int __cmd_record(int argc, const char **argv)
{
	int i, counter;
	pid_t pid;
	int ret;
	struct stat st;

	page_size = sysconf(_SC_PAGE_SIZE);
	nr_cpus = sysconf(_SC_NPROCESSORS_ONLN);
	assert(nr_cpus <= MAX_NR_CPUS);
	assert(nr_cpus >= 0);

	if (!stat(output_name, &st) && !force) {
		fprintf(stderr, "Error, output file: %s exists, use -f to overwrite.\n",
				output_name);
		exit(-1);
	}

	output = open(output_name, O_CREAT|O_TRUNC|O_RDWR, S_IRUSR|S_IWUSR);
	if (output < 0) {
		perror("failed to create output file");
		exit(-1);
	}

	if (!system_wide) {
		open_counters(-1, target_pid != -1 ? target_pid : 0);
	} else for (i = 0; i < nr_cpus; i++)
		open_counters(i, target_pid);

	signal(SIGCHLD, sig_handler);
	signal(SIGINT, sig_handler);

	if (target_pid == -1 && argc) {
		pid = fork();
		if (pid < 0)
			perror("failed to fork");

		if (!pid) {
			if (execvp(argv[0], (char **)argv)) {
				perror(argv[0]);
				exit(-1);
			}
		}
	}

	if (realtime_prio) {
		struct sched_param param;

		param.sched_priority = realtime_prio;
		if (sched_setscheduler(0, SCHED_FIFO, &param)) {
			printf("Could not set realtime priority.\n");
			exit(-1);
		}
	}

	if (system_wide)
		synthesize_events();

	while (!done) {
		int hits = events;

		for (i = 0; i < nr_cpu; i++) {
			for (counter = 0; counter < nr_counters; counter++)
				mmap_read(&mmap_array[i][counter]);
		}

		if (hits == events)
			ret = poll(event_array, nr_poll, 100);
	}

	return 0;
}

static const char * const record_usage[] = {
	"perf record [<options>] [<command>]",
	"perf record [<options>] -- <command> [<options>]",
	NULL
};

static char events_help_msg[EVENTS_HELP_MAX];

static const struct option options[] = {
	OPT_CALLBACK('e', "event", NULL, "event",
		     events_help_msg, parse_events),
	OPT_INTEGER('c', "count", &default_interval,
		    "event period to sample"),
	OPT_INTEGER('m', "mmap-pages", &mmap_pages,
		    "number of mmap data pages"),
	OPT_STRING('o', "output", &output_name, "file",
		    "output file name"),
	OPT_BOOLEAN('i', "inherit", &inherit,
		    "child tasks inherit counters"),
	OPT_INTEGER('p', "pid", &target_pid,
		    "record events on existing pid"),
	OPT_INTEGER('r', "realtime", &realtime_prio,
		    "collect data with this RT SCHED_FIFO priority"),
	OPT_BOOLEAN('a', "all-cpus", &system_wide,
			    "system-wide collection from all CPUs"),
	OPT_BOOLEAN('f', "force", &force,
			"overwrite existing data file"),
	OPT_END()
};

int cmd_record(int argc, const char **argv, const char *prefix)
{
	int counter;

	create_events_help(events_help_msg);

	argc = parse_options(argc, argv, options, record_usage, 0);
	if (!argc && target_pid == -1 && !system_wide)
		usage_with_options(record_usage, options);

	if (!nr_counters) {
		nr_counters = 1;
		event_id[0] = 0;
	}

	for (counter = 0; counter < nr_counters; counter++) {
		if (event_count[counter])
			continue;

		event_count[counter] = default_interval;
	}

	return __cmd_record(argc, argv);
}
Commit	Line	Data
de9ac07b PZ	1
de9ac07b PZ	2
1a482f38	3	#include "perf.h"
16f762a2	4	#include "builtin.h"
6eda5838	5	#include "util/util.h"
0e9b20b8	6	#include "util/parse-options.h"
8ad8db37	7	#include "util/parse-events.h"
a0055ae2	8	#include "util/string.h"
6eda5838	9
97124d5e	10	#include <unistd.h>
de9ac07b	11	#include <sched.h>
de9ac07b	12
0e9b20b8 IM	13	#define ALIGN(x, a) __ALIGN_MASK(x, (typeof(x))(a)-1)
0e9b20b8 IM	14	#define __ALIGN_MASK(x, mask) (((x)+(mask))&~(mask))
1a853e36	15
de9ac07b PZ	16	static int default_interval = 100000;
de9ac07b PZ	17	static int event_count[MAX_COUNTERS];
8ad8db37	18
de9ac07b PZ	19	static int fd[MAX_NR_CPUS][MAX_COUNTERS];
	20	static int nr_cpus = 0;
	21	static unsigned int page_size;
	22	static unsigned int mmap_pages = 16;
	23	static int output;
23ac9cbe	24	static const char *output_name = "perf.data";
de9ac07b	25	static int group = 0;
16c8a109 PZ	26	static unsigned int realtime_prio = 0;
16c8a109 PZ	27	static int system_wide = 0;
1a853e36	28	static pid_t target_pid = -1;
16c8a109	29	static int inherit = 1;
97124d5e	30	static int force = 0;
de9ac07b PZ	31
	32	const unsigned int default_count[] = {
	33	1000000,
	34	1000000,
	35	10000,
	36	10000,
	37	1000000,
	38	10000,
	39	};
	40
de9ac07b PZ	41	struct mmap_data {
	42	int counter;
	43	void *base;
	44	unsigned int mask;
	45	unsigned int prev;
	46	};
	47
	48	static unsigned int mmap_read_head(struct mmap_data *md)
	49	{
	50	struct perf_counter_mmap_page *pc = md->base;
	51	int head;
	52
	53	head = pc->data_head;
	54	rmb();
	55
	56	return head;
	57	}
	58
	59	static long events;
	60	static struct timeval last_read, this_read;
	61
	62	static void mmap_read(struct mmap_data *md)
	63	{
	64	unsigned int head = mmap_read_head(md);
	65	unsigned int old = md->prev;
	66	unsigned char *data = md->base + page_size;
	67	unsigned long size;
	68	void *buf;
	69	int diff;
	70
	71	gettimeofday(&this_read, NULL);
	72
	73	/*
	74	* If we're further behind than half the buffer, there's a chance
	75	* the writer will bite our tail and screw up the events under us.
	76	*
	77	* If we somehow ended up ahead of the head, we got messed up.
	78	*
	79	* In either case, truncate and restart at head.
	80	*/
	81	diff = head - old;
	82	if (diff > md->mask / 2 \|\| diff < 0) {
	83	struct timeval iv;
	84	unsigned long msecs;
	85
	86	timersub(&this_read, &last_read, &iv);
	87	msecs = iv.tv_sec*1000 + iv.tv_usec/1000;
	88
	89	fprintf(stderr, "WARNING: failed to keep up with mmap data."
	90	" Last read %lu msecs ago.\n", msecs);
	91
	92	/*
	93	* head points to a known good entry, start there.
	94	*/
	95	old = head;
	96	}
	97
	98	last_read = this_read;
	99
	100	if (old != head)
	101	events++;
	102
	103	size = head - old;
	104
105	if ((old & md->mask) + size != (head & md->mask)) {
106	buf = &data[old & md->mask];
107	size = md->mask + 1 - (old & md->mask);
108	old += size;
109	while (size) {
110	int ret = write(output, buf, size);
111	if (ret < 0) {
112	perror("failed to write");
113	exit(-1);
114	}
115	size -= ret;
116	buf += ret;
117	}
118	}
119
120	buf = &data[old & md->mask];
121	size = head - old;
122	old += size;
123	while (size) {
124	int ret = write(output, buf, size);
125	if (ret < 0) {
126	perror("failed to write");
127	exit(-1);
128	}
129	size -= ret;
130	buf += ret;
131	}
132
133	md->prev = old;
134	}
135
136	static volatile int done = 0;
137
16c8a109	138	static void sig_handler(int sig)
de9ac07b	139	{
16c8a109	140	done = 1;
de9ac07b PZ	141	}
de9ac07b PZ	142
16c8a109 PZ	143	static struct pollfd event_array[MAX_NR_CPUS * MAX_COUNTERS];
	144	static struct mmap_data mmap_array[MAX_NR_CPUS][MAX_COUNTERS];
	145
	146	static int nr_poll;
	147	static int nr_cpu;
	148
1a853e36	149	struct mmap_event {
16f762a2 IM	150	struct perf_event_header header;
	151	__u32 pid;
	152	__u32 tid;
	153	__u64 start;
	154	__u64 len;
	155	__u64 pgoff;
	156	char filename[PATH_MAX];
1a853e36	157	};
16f762a2	158
1a853e36	159	struct comm_event {
16f762a2 IM	160	struct perf_event_header header;
	161	__u32 pid;
	162	__u32 tid;
	163	char comm[16];
1a853e36 ACM	164	};
1a853e36 ACM	165
f70e87d7	166	static void pid_synthesize_comm_event(pid_t pid, int full)
1a853e36	167	{
16f762a2	168	struct comm_event comm_ev;
1a853e36 ACM	169	char filename[PATH_MAX];
1a853e36 ACM	170	char bf[BUFSIZ];
a0055ae2	171	int fd, ret;
1a853e36	172	size_t size;
a0055ae2	173	char field, sep;
f70e87d7 PZ	174	DIR *tasks;
f70e87d7 PZ	175	struct dirent dirent, *next;
1a853e36 ACM	176
	177	snprintf(filename, sizeof(filename), "/proc/%d/stat", pid);
	178
	179	fd = open(filename, O_RDONLY);
	180	if (fd < 0) {
	181	fprintf(stderr, "couldn't open %s\n", filename);
	182	exit(EXIT_FAILURE);
	183	}
	184	if (read(fd, bf, sizeof(bf)) < 0) {
	185	fprintf(stderr, "couldn't read %s\n", filename);
	186	exit(EXIT_FAILURE);
	187	}
	188	close(fd);
	189
a0055ae2	190	/* 9027 (cat) R 6747 9027 6747 34816 9027 ... */
1a853e36	191	memset(&comm_ev, 0, sizeof(comm_ev));
a0055ae2 ACM	192	field = strchr(bf, '(');
	193	if (field == NULL)
	194	goto out_failure;
	195	sep = strchr(++field, ')');
	196	if (sep == NULL)
	197	goto out_failure;
	198	size = sep - field;
	199	memcpy(comm_ev.comm, field, size++);
f70e87d7 PZ	200
f70e87d7 PZ	201	comm_ev.pid = pid;
1a853e36	202	comm_ev.header.type = PERF_EVENT_COMM;
1a853e36 ACM	203	size = ALIGN(size, sizeof(uint64_t));
1a853e36 ACM	204	comm_ev.header.size = sizeof(comm_ev) - (sizeof(comm_ev.comm) - size);
16f762a2	205
f70e87d7 PZ	206	if (!full) {
	207	comm_ev.tid = pid;
	208
	209	ret = write(output, &comm_ev, comm_ev.header.size);
	210	if (ret < 0) {
	211	perror("failed to write");
	212	exit(-1);
	213	}
	214	return;
	215	}
	216
	217	snprintf(filename, sizeof(filename), "/proc/%d/task", pid);
	218
	219	tasks = opendir(filename);
	220	while (!readdir_r(tasks, &dirent, &next) && next) {
	221	char *end;
	222	pid = strtol(dirent.d_name, &end, 10);
	223	if (*end)
	224	continue;
	225
	226	comm_ev.tid = pid;
	227
	228	ret = write(output, &comm_ev, comm_ev.header.size);
	229	if (ret < 0) {
	230	perror("failed to write");
	231	exit(-1);
	232	}
1a853e36	233	}
f70e87d7 PZ	234	closedir(tasks);
	235	return;
	236
a0055ae2 ACM	237	out_failure:
	238	fprintf(stderr, "couldn't get COMM and pgid, malformed %s\n",
	239	filename);
	240	exit(EXIT_FAILURE);
1a853e36 ACM	241	}
1a853e36 ACM	242
f70e87d7	243	static void pid_synthesize_mmap_events(pid_t pid)
1a853e36 ACM	244	{
	245	char filename[PATH_MAX];
	246	FILE *fp;
	247
	248	snprintf(filename, sizeof(filename), "/proc/%d/maps", pid);
	249
	250	fp = fopen(filename, "r");
	251	if (fp == NULL) {
	252	fprintf(stderr, "couldn't open %s\n", filename);
	253	exit(EXIT_FAILURE);
	254	}
	255	while (1) {
a0055ae2	256	char bf[BUFSIZ], *pbf = bf;
1a853e36 ACM	257	struct mmap_event mmap_ev = {
	258	.header.type = PERF_EVENT_MMAP,
	259	};
a0055ae2	260	int n;
1a853e36 ACM	261	size_t size;
	262	if (fgets(bf, sizeof(bf), fp) == NULL)
	263	break;
	264
	265	/* 00400000-0040c000 r-xp 00000000 fd:01 41038 /bin/cat */
a0055ae2 ACM	266	n = hex2u64(pbf, &mmap_ev.start);
	267	if (n < 0)
	268	continue;
	269	pbf += n + 1;
	270	n = hex2u64(pbf, &mmap_ev.len);
	271	if (n < 0)
	272	continue;
	273	pbf += n + 3;
	274	if (pbf == 'x') { / vm_exec */
1a853e36 ACM	275	char *execname = strrchr(bf, ' ');
	276
	277	if (execname == NULL \|\| execname[1] != '/')
	278	continue;
	279
	280	execname += 1;
	281	size = strlen(execname);
	282	execname[size - 1] = '\0'; /* Remove \n */
	283	memcpy(mmap_ev.filename, execname, size);
	284	size = ALIGN(size, sizeof(uint64_t));
	285	mmap_ev.len -= mmap_ev.start;
	286	mmap_ev.header.size = (sizeof(mmap_ev) -
	287	(sizeof(mmap_ev.filename) - size));
f70e87d7	288	mmap_ev.pid = pid;
1a853e36 ACM	289	mmap_ev.tid = pid;
	290
	291	if (write(output, &mmap_ev, mmap_ev.header.size) < 0) {
	292	perror("failed to write");
	293	exit(-1);
	294	}
	295	}
	296	}
	297
	298	fclose(fp);
	299	}
	300
f70e87d7 PZ	301	static void synthesize_events(void)
	302	{
	303	DIR *proc;
	304	struct dirent dirent, *next;
	305
	306	proc = opendir("/proc");
	307
	308	while (!readdir_r(proc, &dirent, &next) && next) {
	309	char *end;
	310	pid_t pid;
	311
	312	pid = strtol(dirent.d_name, &end, 10);
	313	if (end) / only interested in proper numerical dirents */
	314	continue;
	315
	316	pid_synthesize_comm_event(pid, 1);
	317	pid_synthesize_mmap_events(pid);
	318	}
	319
	320	closedir(proc);
	321	}
	322
1a853e36	323	static void open_counters(int cpu, pid_t pid)
de9ac07b	324	{
c70975bc	325	struct perf_counter_attr attr;
16c8a109 PZ	326	int counter, group_fd;
16c8a109 PZ	327	int track = 1;
16c8a109	328
1a853e36	329	if (pid > 0) {
f70e87d7 PZ	330	pid_synthesize_comm_event(pid, 0);
f70e87d7 PZ	331	pid_synthesize_mmap_events(pid);
1a853e36	332	}
16c8a109 PZ	333
	334	group_fd = -1;
	335	for (counter = 0; counter < nr_counters; counter++) {
	336
c70975bc PZ	337	memset(&attr, 0, sizeof(attr));
	338	attr.config = event_id[counter];
	339	attr.sample_period = event_count[counter];
	340	attr.sample_type = PERF_SAMPLE_IP \| PERF_SAMPLE_TID;
	341	attr.mmap = track;
	342	attr.comm = track;
	343	attr.inherit = (cpu < 0) && inherit;
16c8a109 PZ	344
	345	track = 0; // only the first counter needs these
	346
	347	fd[nr_cpu][counter] =
c70975bc	348	sys_perf_counter_open(&attr, pid, cpu, group_fd, 0);
16c8a109 PZ	349
	350	if (fd[nr_cpu][counter] < 0) {
	351	int err = errno;
	352	printf("kerneltop error: syscall returned with %d (%s)\n",
	353	fd[nr_cpu][counter], strerror(err));
	354	if (err == EPERM)
	355	printf("Are you root?\n");
	356	exit(-1);
	357	}
	358	assert(fd[nr_cpu][counter] >= 0);
	359	fcntl(fd[nr_cpu][counter], F_SETFL, O_NONBLOCK);
	360
	361	/*
	362	* First counter acts as the group leader:
	363	*/
	364	if (group && group_fd == -1)
	365	group_fd = fd[nr_cpu][counter];
	366
	367	event_array[nr_poll].fd = fd[nr_cpu][counter];
	368	event_array[nr_poll].events = POLLIN;
	369	nr_poll++;
	370
	371	mmap_array[nr_cpu][counter].counter = counter;
	372	mmap_array[nr_cpu][counter].prev = 0;
	373	mmap_array[nr_cpu][counter].mask = mmap_pages*page_size - 1;
	374	mmap_array[nr_cpu][counter].base = mmap(NULL, (mmap_pages+1)*page_size,
	375	PROT_READ, MAP_SHARED, fd[nr_cpu][counter], 0);
	376	if (mmap_array[nr_cpu][counter].base == MAP_FAILED) {
	377	printf("kerneltop error: failed to mmap with %d (%s)\n",
	378	errno, strerror(errno));
	379	exit(-1);
	380	}
	381	}
	382	nr_cpu++;
	383	}
	384
0e9b20b8	385	static int __cmd_record(int argc, const char **argv)
16c8a109 PZ	386	{
16c8a109 PZ	387	int i, counter;
de9ac07b PZ	388	pid_t pid;
de9ac07b PZ	389	int ret;
97124d5e	390	struct stat st;
de9ac07b PZ	391
de9ac07b PZ	392	page_size = sysconf(_SC_PAGE_SIZE);
de9ac07b PZ	393	nr_cpus = sysconf(_SC_NPROCESSORS_ONLN);
	394	assert(nr_cpus <= MAX_NR_CPUS);
	395	assert(nr_cpus >= 0);
	396
97124d5e PZ	397	if (!stat(output_name, &st) && !force) {
	398	fprintf(stderr, "Error, output file: %s exists, use -f to overwrite.\n",
	399	output_name);
	400	exit(-1);
	401	}
	402
	403	output = open(output_name, O_CREAT\|O_TRUNC\|O_RDWR, S_IRUSR\|S_IWUSR);
de9ac07b PZ	404	if (output < 0) {
	405	perror("failed to create output file");
	406	exit(-1);
	407	}
	408
1a853e36 ACM	409	if (!system_wide) {
	410	open_counters(-1, target_pid != -1 ? target_pid : 0);
	411	} else for (i = 0; i < nr_cpus; i++)
	412	open_counters(i, target_pid);
de9ac07b	413
16c8a109 PZ	414	signal(SIGCHLD, sig_handler);
16c8a109 PZ	415	signal(SIGINT, sig_handler);
de9ac07b	416
ef65b2a0	417	if (target_pid == -1 && argc) {
1a853e36 ACM	418	pid = fork();
	419	if (pid < 0)
	420	perror("failed to fork");
de9ac07b	421
1a853e36	422	if (!pid) {
0e9b20b8	423	if (execvp(argv[0], (char **)argv)) {
1a853e36 ACM	424	perror(argv[0]);
	425	exit(-1);
	426	}
de9ac07b PZ	427	}
	428	}
	429
	430	if (realtime_prio) {
	431	struct sched_param param;
	432
	433	param.sched_priority = realtime_prio;
	434	if (sched_setscheduler(0, SCHED_FIFO, &param)) {
	435	printf("Could not set realtime priority.\n");
	436	exit(-1);
	437	}
	438	}
	439
f70e87d7 PZ	440	if (system_wide)
f70e87d7 PZ	441	synthesize_events();
de9ac07b PZ	442
	443	while (!done) {
	444	int hits = events;
	445
16c8a109	446	for (i = 0; i < nr_cpu; i++) {
de9ac07b PZ	447	for (counter = 0; counter < nr_counters; counter++)
	448	mmap_read(&mmap_array[i][counter]);
	449	}
	450
	451	if (hits == events)
	452	ret = poll(event_array, nr_poll, 100);
	453	}
	454
	455	return 0;
	456	}
0e9b20b8	457
0e9b20b8	458	static const char * const record_usage[] = {
9e096753 MG	459	"perf record [<options>] [<command>]",
9e096753 MG	460	"perf record [<options>] -- <command> [<options>]",
0e9b20b8 IM	461	NULL
	462	};
	463
8ad8db37 IM	464	static char events_help_msg[EVENTS_HELP_MAX];
8ad8db37 IM	465
5242519b	466	static const struct option options[] = {
0e9b20b8	467	OPT_CALLBACK('e', "event", NULL, "event",
8ad8db37	468	events_help_msg, parse_events),
0e9b20b8 IM	469	OPT_INTEGER('c', "count", &default_interval,
	470	"event period to sample"),
	471	OPT_INTEGER('m', "mmap-pages", &mmap_pages,
	472	"number of mmap data pages"),
	473	OPT_STRING('o', "output", &output_name, "file",
	474	"output file name"),
	475	OPT_BOOLEAN('i', "inherit", &inherit,
	476	"child tasks inherit counters"),
	477	OPT_INTEGER('p', "pid", &target_pid,
	478	"record events on existing pid"),
	479	OPT_INTEGER('r', "realtime", &realtime_prio,
	480	"collect data with this RT SCHED_FIFO priority"),
	481	OPT_BOOLEAN('a', "all-cpus", &system_wide,
	482	"system-wide collection from all CPUs"),
97124d5e PZ	483	OPT_BOOLEAN('f', "force", &force,
97124d5e PZ	484	"overwrite existing data file"),
0e9b20b8 IM	485	OPT_END()
	486	};
	487
	488	int cmd_record(int argc, const char *argv, const char prefix)
	489	{
	490	int counter;
	491
8ad8db37	492	create_events_help(events_help_msg);
0e9b20b8 IM	493
0e9b20b8 IM	494	argc = parse_options(argc, argv, options, record_usage, 0);
ef65b2a0	495	if (!argc && target_pid == -1 && !system_wide)
0e9b20b8 IM	496	usage_with_options(record_usage, options);
	497
	498	if (!nr_counters) {
	499	nr_counters = 1;
	500	event_id[0] = 0;
	501	}
	502
	503	for (counter = 0; counter < nr_counters; counter++) {
	504	if (event_count[counter])
	505	continue;
	506
	507	event_count[counter] = default_interval;
	508	}
	509
	510	return __cmd_record(argc, argv);
	511	}