builtin-top.c

	switch (origin) {
	case PERF_RECORD_MISC_USER:
		++us_samples;
		if (hide_user_symbols)
			return;
		machine = perf_session__find_host_machine(session);
		break;
	case PERF_RECORD_MISC_KERNEL:
		++kernel_samples;
		if (hide_kernel_symbols)
			return;
		machine = perf_session__find_host_machine(session);
		break;
	case PERF_RECORD_MISC_GUEST_KERNEL:
		++guest_kernel_samples;
		machine = perf_session__find_machine(session, self->ip.pid);
		break;
	case PERF_RECORD_MISC_GUEST_USER:
		++guest_us_samples;
		/*
		 * TODO: we don't process guest user from host side
		 * except simple counting.
		 */
		return;
	default:
		return;
	}

	if (!machine && perf_guest) {
		pr_err("Can't find guest [%d]'s kernel information\n",
			self->ip.pid);
		return;
	}

	if (self->header.misc & PERF_RECORD_MISC_EXACT_IP)
		exact_samples++;

	if (event__preprocess_sample(self, session, &al, sample,
				     symbol_filter) < 0 ||
	    al.filtered)
		return;

	if (al.sym == NULL) {
		/*
		 * As we do lazy loading of symtabs we only will know if the
		 * specified vmlinux file is invalid when we actually have a
		 * hit in kernel space and then try to load it. So if we get
		 * here and there are _no_ symbols in the DSO backing the
		 * kernel map, bail out.
		 *
		 * We may never get here, for instance, if we use -K/
		 * --hide-kernel-symbols, even if the user specifies an
		 * invalid --vmlinux ;-)
		 */
		if (al.map == machine->vmlinux_maps[MAP__FUNCTION] &&
		    RB_EMPTY_ROOT(&al.map->dso->symbols[MAP__FUNCTION])) {
			pr_err("The %s file can't be used\n",
			       symbol_conf.vmlinux_name);
			exit(1);
		}

		return;
	}

	/* let's see, whether we need to install initial sym_filter_entry */
	if (sym_filter_entry_sched) {
		sym_filter_entry = sym_filter_entry_sched;
		sym_filter_entry_sched = NULL;
		if (parse_source(sym_filter_entry) < 0) {
			struct symbol *sym = sym_entry__symbol(sym_filter_entry);

			pr_err("Can't annotate %s", sym->name);
			if (sym_filter_entry->map->dso->origin == DSO__ORIG_KERNEL) {
				pr_err(": No vmlinux file was found in the path:\n");
				machine__fprintf_vmlinux_path(machine, stderr);
			} else
				pr_err(".\n");
			exit(1);
		}
	}

	syme = symbol__priv(al.sym);
	if (!syme->skip) {
		syme->count[evsel->idx]++;
		syme->origin = origin;
		record_precise_ip(syme, evsel->idx, ip);
		pthread_mutex_lock(&active_symbols_lock);
		if (list_empty(&syme->node) || !syme->node.next)
			__list_insert_active_sym(syme);
		pthread_mutex_unlock(&active_symbols_lock);
	}
}

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

static int perf_evsel__alloc_mmap_per_thread(struct perf_evsel *evsel,
					     int ncpus, int nthreads)
{
	evsel->priv = xyarray__new(ncpus, nthreads, sizeof(struct mmap_data));
	return evsel->priv != NULL ? 0 : -ENOMEM;
}

static void perf_evsel__free_mmap(struct perf_evsel *evsel)
{
	xyarray__delete(evsel->priv);
	evsel->priv = NULL;
}

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

	head = pc->data_head;
	rmb();

	return head;
}

static void perf_session__mmap_read_counter(struct perf_session *self,
					    struct perf_evsel *evsel,
					    int cpu, int thread_idx)
{
	struct xyarray *mmap_array = evsel->priv;
	struct mmap_data *md = xyarray__entry(mmap_array, cpu, thread_idx);
	unsigned int head = mmap_read_head(md);
	unsigned int old = md->prev;
	unsigned char *data = md->base + page_size;
	struct sample_data sample;
	int diff;

	/*
	 * If we're further behind than half the buffer, there's a chance
	 * the writer will bite our tail and mess up the samples 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) {
		fprintf(stderr, "WARNING: failed to keep up with mmap data.\n");

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

	for (; old != head;) {
		event_t *event = (event_t *)&data[old & md->mask];

		event_t event_copy;

		size_t size = event->header.size;

		/*
		 * Event straddles the mmap boundary -- header should always
		 * be inside due to u64 alignment of output.
		 */
		if ((old & md->mask) + size != ((old + size) & md->mask)) {
			unsigned int offset = old;
			unsigned int len = min(sizeof(*event), size), cpy;
			void *dst = &event_copy;

			do {
				cpy = min(md->mask + 1 - (offset & md->mask), len);
				memcpy(dst, &data[offset & md->mask], cpy);
				offset += cpy;
				dst += cpy;
				len -= cpy;
			} while (len);

			event = &event_copy;
		}

		event__parse_sample(event, self, &sample);
		if (event->header.type == PERF_RECORD_SAMPLE)
			event__process_sample(event, &sample, self, evsel);
		else
			event__process(event, &sample, self);
		old += size;
	}

	md->prev = old;
}

static struct pollfd *event_array;

static void perf_session__mmap_read(struct perf_session *self)
{
	struct perf_evsel *counter;
	int i, thread_index;

	for (i = 0; i < cpus->nr; i++) {
		list_for_each_entry(counter, &evsel_list, node) {
			for (thread_index = 0;
				thread_index < threads->nr;
				thread_index++) {
				perf_session__mmap_read_counter(self,
					counter, i, thread_index);
			}
		}
	}
}

int nr_poll;
int group_fd;

static void start_counter(int i, struct perf_evsel *evsel)
{
	struct xyarray *mmap_array = evsel->priv;
	struct mmap_data *mm;
	struct perf_event_attr *attr;
	int cpu = -1;
	int thread_index;

	if (target_tid == -1)
		cpu = cpus->map[i];

	attr = &evsel->attr;

	attr->sample_type	= PERF_SAMPLE_IP | PERF_SAMPLE_TID;

	if (freq) {
		attr->sample_type	|= PERF_SAMPLE_PERIOD;
		attr->freq		= 1;
		attr->sample_freq	= freq;
	}

	attr->inherit		= (cpu < 0) && inherit;
	attr->mmap		= 1;

	for (thread_index = 0; thread_index < threads->nr; thread_index++) {
try_again:
		FD(evsel, i, thread_index) = sys_perf_event_open(attr,
				threads->map[thread_index], cpu, group_fd, 0);

		if (FD(evsel, i, thread_index) < 0) {
			int err = errno;

			if (err == EPERM || err == EACCES)
				die("Permission error - are you root?\n"
					"\t Consider tweaking"
					" /proc/sys/kernel/perf_event_paranoid.\n");
			/*
			 * If it's cycles then fall back to hrtimer
			 * based cpu-clock-tick sw counter, which
			 * is always available even if no PMU support:
			 */
			if (attr->type == PERF_TYPE_HARDWARE
					&& attr->config == PERF_COUNT_HW_CPU_CYCLES) {

				if (verbose)
					warning(" ... trying to fall back to cpu-clock-ticks\n");

				attr->type = PERF_TYPE_SOFTWARE;
				attr->config = PERF_COUNT_SW_CPU_CLOCK;
				goto try_again;
			}
			printf("\n");
			error("sys_perf_event_open() syscall returned with %d (%s).  /bin/dmesg may provide additional information.\n",
					FD(evsel, i, thread_index), strerror(err));
			die("No CONFIG_PERF_EVENTS=y kernel support configured?\n");
			exit(-1);
		}
		assert(FD(evsel, i, thread_index) >= 0);
		fcntl(FD(evsel, i, thread_index), F_SETFL, O_NONBLOCK);

		/*
		 * First counter acts as the group leader:
		 */
		if (group && group_fd == -1)
			group_fd = FD(evsel, i, thread_index);

		event_array[nr_poll].fd = FD(evsel, i, thread_index);
		event_array[nr_poll].events = POLLIN;
		nr_poll++;

		mm = xyarray__entry(mmap_array, i, thread_index);
		mm->prev = 0;
		mm->mask = mmap_pages*page_size - 1;
		mm->base = mmap(NULL, (mmap_pages+1)*page_size,
				PROT_READ, MAP_SHARED, FD(evsel, i, thread_index), 0);
		if (mm->base == MAP_FAILED)
			die("failed to mmap with %d (%s)\n", errno, strerror(errno));
	}
}

static int __cmd_top(void)
{
	pthread_t thread;
	struct perf_evsel *counter;
	int i, ret;
	/*
	 * FIXME: perf_session__new should allow passing a O_MMAP, so that all this
	 * mmap reading, etc is encapsulated in it. Use O_WRONLY for now.
	 */
	struct perf_session *session = perf_session__new(NULL, O_WRONLY, false, false, NULL);
	if (session == NULL)
		return -ENOMEM;

	if (target_tid != -1)
		event__synthesize_thread(target_tid, event__process, session);
	else
		event__synthesize_threads(event__process, session);

	for (i = 0; i < cpus->nr; i++) {
		group_fd = -1;
		list_for_each_entry(counter, &evsel_list, node)
			start_counter(i, counter);
	}

	/* Wait for a minimal set of events before starting the snapshot */
	poll(&event_array[0], nr_poll, 100);

	perf_session__mmap_read(session);

	if (pthread_create(&thread, NULL, display_thread, session)) {
		printf("Could not create display thread.\n");
		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);
		}
	}

	while (1) {
		int hits = samples;

		perf_session__mmap_read(session);

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

	return 0;
}

static const char * const top_usage[] = {
	"perf top [<options>]",
	NULL
};

static const struct option options[] = {
	OPT_CALLBACK('e', "event", NULL, "event",
		     "event selector. use 'perf list' to list available events",
		     parse_events),
	OPT_INTEGER('c', "count", &default_interval,
		    "event period to sample"),
	OPT_INTEGER('p', "pid", &target_pid,
		    "profile events on existing process id"),
	OPT_INTEGER('t', "tid", &target_tid,
		    "profile events on existing thread id"),
	OPT_BOOLEAN('a', "all-cpus", &system_wide,
			    "system-wide collection from all CPUs"),
	OPT_STRING('C', "cpu", &cpu_list, "cpu",
		    "list of cpus to monitor"),
	OPT_STRING('k', "vmlinux", &symbol_conf.vmlinux_name,
		   "file", "vmlinux pathname"),
	OPT_BOOLEAN('K', "hide_kernel_symbols", &hide_kernel_symbols,
		    "hide kernel symbols"),
	OPT_UINTEGER('m', "mmap-pages", &mmap_pages, "number of mmap data pages"),
	OPT_INTEGER('r', "realtime", &realtime_prio,
		    "collect data with this RT SCHED_FIFO priority"),
	OPT_INTEGER('d', "delay", &delay_secs,
		    "number of seconds to delay between refreshes"),
	OPT_BOOLEAN('D', "dump-symtab", &dump_symtab,
			    "dump the symbol table used for profiling"),
	OPT_INTEGER('f', "count-filter", &count_filter,
		    "only display functions with more events than this"),
	OPT_BOOLEAN('g', "group", &group,
			    "put the counters into a counter group"),
	OPT_BOOLEAN('i', "inherit", &inherit,
		    "child tasks inherit counters"),
	OPT_STRING('s', "sym-annotate", &sym_filter, "symbol name",
		    "symbol to annotate"),
	OPT_BOOLEAN('z', "zero", &zero,
		    "zero history across updates"),
	OPT_INTEGER('F', "freq", &freq,
		    "profile at this frequency"),
	OPT_INTEGER('E', "entries", &print_entries,
		    "display this many functions"),
	OPT_BOOLEAN('U', "hide_user_symbols", &hide_user_symbols,
		    "hide user symbols"),
	OPT_INCR('v', "verbose", &verbose,
		    "be more verbose (show counter open errors, etc)"),
	OPT_END()
};

int cmd_top(int argc, const char **argv, const char *prefix __used)
{
	struct perf_evsel *pos;
	int status = -ENOMEM;

	page_size = sysconf(_SC_PAGE_SIZE);

	argc = parse_options(argc, argv, options, top_usage, 0);
	if (argc)
		usage_with_options(top_usage, options);

	if (target_pid != -1)
		target_tid = target_pid;

	threads = thread_map__new(target_pid, target_tid);
	if (threads == NULL) {
		pr_err("Problems finding threads of monitor\n");
		usage_with_options(top_usage, options);
	}

	event_array = malloc((sizeof(struct pollfd) *
			      MAX_NR_CPUS * MAX_COUNTERS * threads->nr));
	if (!event_array)
		return -ENOMEM;

	/* CPU and PID are mutually exclusive */
	if (target_tid > 0 && cpu_list) {
		printf("WARNING: PID switch overriding CPU\n");
		sleep(1);
		cpu_list = NULL;
	}

	if (!nr_counters && perf_evsel_list__create_default() < 0) {
		pr_err("Not enough memory for event selector list\n");
		return -ENOMEM;
	}

	if (delay_secs < 1)
		delay_secs = 1;

	/*
	 * User specified count overrides default frequency.
	 */
	if (default_interval)
		freq = 0;
	else if (freq) {
		default_interval = freq;
	} else {
		fprintf(stderr, "frequency and count are zero, aborting\n");
		exit(EXIT_FAILURE);
	}

	if (target_tid != -1)
		cpus = cpu_map__dummy_new();
	else
		cpus = cpu_map__new(cpu_list);

	if (cpus == NULL)
		usage_with_options(top_usage, options);

	list_for_each_entry(pos, &evsel_list, node) {
		if (perf_evsel__alloc_mmap_per_thread(pos, cpus->nr, threads->nr) < 0 ||
		    perf_evsel__alloc_fd(pos, cpus->nr, threads->nr) < 0)
			goto out_free_fd;
		/*
		 * Fill in the ones not specifically initialized via -c:
		 */
		if (pos->attr.sample_period)
			continue;

		pos->attr.sample_period = default_interval;
	}

	sym_evsel = list_entry(evsel_list.next, struct perf_evsel, node);

	symbol_conf.priv_size = (sizeof(struct sym_entry) +
				 (nr_counters + 1) * sizeof(unsigned long));

	symbol_conf.try_vmlinux_path = (symbol_conf.vmlinux_name == NULL);
	if (symbol__init() < 0)
		return -1;

	get_term_dimensions(&winsize);
	if (print_entries == 0) {
		update_print_entries(&winsize);
		signal(SIGWINCH, sig_winch_handler);
	}

	status = __cmd_top();
out_free_fd:
	list_for_each_entry(pos, &evsel_list, node)
		perf_evsel__free_mmap(pos);
	perf_evsel_list__delete();

	return status;
}