builtin-kvm.c 34.3 KB
Newer Older
1 2 3
#include "builtin.h"
#include "perf.h"

X
Xiao Guangrong 已提交
4
#include "util/evsel.h"
D
David Ahern 已提交
5
#include "util/evlist.h"
6 7 8 9 10 11
#include "util/util.h"
#include "util/cache.h"
#include "util/symbol.h"
#include "util/thread.h"
#include "util/header.h"
#include "util/session.h"
12
#include "util/intlist.h"
13 14 15
#include "util/parse-options.h"
#include "util/trace-event.h"
#include "util/debug.h"
16
#include <api/fs/debugfs.h>
X
Xiao Guangrong 已提交
17 18
#include "util/tool.h"
#include "util/stat.h"
D
David Ahern 已提交
19
#include "util/top.h"
20
#include "util/data.h"
21 22

#include <sys/prctl.h>
23
#ifdef HAVE_TIMERFD_SUPPORT
D
David Ahern 已提交
24
#include <sys/timerfd.h>
25
#endif
26

D
David Ahern 已提交
27
#include <termios.h>
28 29 30 31
#include <semaphore.h>
#include <pthread.h>
#include <math.h>

32
#ifdef HAVE_KVM_STAT_SUPPORT
33
#include <asm/kvm_perf.h>
34
#include "util/kvm-stat.h"
X
Xiao Guangrong 已提交
35

36 37 38
void exit_event_get_key(struct perf_evsel *evsel,
			struct perf_sample *sample,
			struct event_key *key)
X
Xiao Guangrong 已提交
39 40
{
	key->info = 0;
41
	key->key = perf_evsel__intval(evsel, sample, KVM_EXIT_REASON);
X
Xiao Guangrong 已提交
42 43
}

44
bool kvm_exit_event(struct perf_evsel *evsel)
X
Xiao Guangrong 已提交
45
{
46
	return !strcmp(evsel->name, KVM_EXIT_TRACE);
X
Xiao Guangrong 已提交
47 48
}

49 50
bool exit_event_begin(struct perf_evsel *evsel,
		      struct perf_sample *sample, struct event_key *key)
X
Xiao Guangrong 已提交
51
{
52 53
	if (kvm_exit_event(evsel)) {
		exit_event_get_key(evsel, sample, key);
X
Xiao Guangrong 已提交
54 55 56 57 58 59
		return true;
	}

	return false;
}

60
bool kvm_entry_event(struct perf_evsel *evsel)
X
Xiao Guangrong 已提交
61
{
62
	return !strcmp(evsel->name, KVM_ENTRY_TRACE);
X
Xiao Guangrong 已提交
63 64
}

65 66 67
bool exit_event_end(struct perf_evsel *evsel,
		    struct perf_sample *sample __maybe_unused,
		    struct event_key *key __maybe_unused)
X
Xiao Guangrong 已提交
68
{
69
	return kvm_entry_event(evsel);
X
Xiao Guangrong 已提交
70 71
}

72 73 74
static const char *get_exit_reason(struct perf_kvm_stat *kvm,
				   struct exit_reasons_table *tbl,
				   u64 exit_code)
X
Xiao Guangrong 已提交
75
{
76
	while (tbl->reason != NULL) {
77 78 79
		if (tbl->exit_code == exit_code)
			return tbl->reason;
		tbl++;
X
Xiao Guangrong 已提交
80 81 82
	}

	pr_err("unknown kvm exit code:%lld on %s\n",
83
		(unsigned long long)exit_code, kvm->exit_reasons_isa);
X
Xiao Guangrong 已提交
84 85 86
	return "UNKNOWN";
}

87 88 89
void exit_event_decode_key(struct perf_kvm_stat *kvm,
			   struct event_key *key,
			   char *decode)
X
Xiao Guangrong 已提交
90
{
91 92
	const char *exit_reason = get_exit_reason(kvm, kvm->exit_reasons,
						  key->key);
X
Xiao Guangrong 已提交
93

94
	scnprintf(decode, DECODE_STR_LEN, "%s", exit_reason);
X
Xiao Guangrong 已提交
95 96
}

97
static bool register_kvm_events_ops(struct perf_kvm_stat *kvm)
X
Xiao Guangrong 已提交
98
{
99
	struct kvm_reg_events_ops *events_ops = kvm_reg_events_ops;
X
Xiao Guangrong 已提交
100

101 102 103 104 105
	for (events_ops = kvm_reg_events_ops; events_ops->name; events_ops++) {
		if (!strcmp(events_ops->name, kvm->report_event)) {
			kvm->events_ops = events_ops->ops;
			return true;
		}
X
Xiao Guangrong 已提交
106 107 108 109 110 111 112 113 114 115 116 117
	}

	return false;
}

struct vcpu_event_record {
	int vcpu_id;
	u64 start_time;
	struct kvm_event *last_event;
};


118
static void init_kvm_event_record(struct perf_kvm_stat *kvm)
X
Xiao Guangrong 已提交
119
{
120
	unsigned int i;
X
Xiao Guangrong 已提交
121

122
	for (i = 0; i < EVENTS_CACHE_SIZE; i++)
123
		INIT_LIST_HEAD(&kvm->kvm_events_cache[i]);
X
Xiao Guangrong 已提交
124 125
}

126
#ifdef HAVE_TIMERFD_SUPPORT
D
David Ahern 已提交
127 128 129 130 131
static void clear_events_cache_stats(struct list_head *kvm_events_cache)
{
	struct list_head *head;
	struct kvm_event *event;
	unsigned int i;
132
	int j;
D
David Ahern 已提交
133 134 135 136 137

	for (i = 0; i < EVENTS_CACHE_SIZE; i++) {
		head = &kvm_events_cache[i];
		list_for_each_entry(event, head, hash_entry) {
			/* reset stats for event */
138 139 140 141 142 143 144
			event->total.time = 0;
			init_stats(&event->total.stats);

			for (j = 0; j < event->max_vcpu; ++j) {
				event->vcpu[j].time = 0;
				init_stats(&event->vcpu[j].stats);
			}
D
David Ahern 已提交
145 146 147
		}
	}
}
148
#endif
D
David Ahern 已提交
149

X
Xiao Guangrong 已提交
150 151 152 153 154 155 156 157
static int kvm_events_hash_fn(u64 key)
{
	return key & (EVENTS_CACHE_SIZE - 1);
}

static bool kvm_event_expand(struct kvm_event *event, int vcpu_id)
{
	int old_max_vcpu = event->max_vcpu;
D
David Ahern 已提交
158
	void *prev;
X
Xiao Guangrong 已提交
159 160 161 162 163 164 165

	if (vcpu_id < event->max_vcpu)
		return true;

	while (event->max_vcpu <= vcpu_id)
		event->max_vcpu += DEFAULT_VCPU_NUM;

D
David Ahern 已提交
166
	prev = event->vcpu;
X
Xiao Guangrong 已提交
167 168 169
	event->vcpu = realloc(event->vcpu,
			      event->max_vcpu * sizeof(*event->vcpu));
	if (!event->vcpu) {
D
David Ahern 已提交
170
		free(prev);
X
Xiao Guangrong 已提交
171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190
		pr_err("Not enough memory\n");
		return false;
	}

	memset(event->vcpu + old_max_vcpu, 0,
	       (event->max_vcpu - old_max_vcpu) * sizeof(*event->vcpu));
	return true;
}

static struct kvm_event *kvm_alloc_init_event(struct event_key *key)
{
	struct kvm_event *event;

	event = zalloc(sizeof(*event));
	if (!event) {
		pr_err("Not enough memory\n");
		return NULL;
	}

	event->key = *key;
191
	init_stats(&event->total.stats);
X
Xiao Guangrong 已提交
192 193 194
	return event;
}

195
static struct kvm_event *find_create_kvm_event(struct perf_kvm_stat *kvm,
196
					       struct event_key *key)
X
Xiao Guangrong 已提交
197 198 199 200 201 202
{
	struct kvm_event *event;
	struct list_head *head;

	BUG_ON(key->key == INVALID_KEY);

203
	head = &kvm->kvm_events_cache[kvm_events_hash_fn(key->key)];
204
	list_for_each_entry(event, head, hash_entry) {
X
Xiao Guangrong 已提交
205 206
		if (event->key.key == key->key && event->key.info == key->info)
			return event;
207
	}
X
Xiao Guangrong 已提交
208 209 210 211 212 213 214 215 216

	event = kvm_alloc_init_event(key);
	if (!event)
		return NULL;

	list_add(&event->hash_entry, head);
	return event;
}

217
static bool handle_begin_event(struct perf_kvm_stat *kvm,
218
			       struct vcpu_event_record *vcpu_record,
X
Xiao Guangrong 已提交
219 220 221 222 223
			       struct event_key *key, u64 timestamp)
{
	struct kvm_event *event = NULL;

	if (key->key != INVALID_KEY)
224
		event = find_create_kvm_event(kvm, key);
X
Xiao Guangrong 已提交
225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251

	vcpu_record->last_event = event;
	vcpu_record->start_time = timestamp;
	return true;
}

static void
kvm_update_event_stats(struct kvm_event_stats *kvm_stats, u64 time_diff)
{
	kvm_stats->time += time_diff;
	update_stats(&kvm_stats->stats, time_diff);
}

static double kvm_event_rel_stddev(int vcpu_id, struct kvm_event *event)
{
	struct kvm_event_stats *kvm_stats = &event->total;

	if (vcpu_id != -1)
		kvm_stats = &event->vcpu[vcpu_id];

	return rel_stddev_stats(stddev_stats(&kvm_stats->stats),
				avg_stats(&kvm_stats->stats));
}

static bool update_kvm_event(struct kvm_event *event, int vcpu_id,
			     u64 time_diff)
{
252 253 254 255
	if (vcpu_id == -1) {
		kvm_update_event_stats(&event->total, time_diff);
		return true;
	}
X
Xiao Guangrong 已提交
256 257 258 259 260 261 262 263

	if (!kvm_event_expand(event, vcpu_id))
		return false;

	kvm_update_event_stats(&event->vcpu[vcpu_id], time_diff);
	return true;
}

264 265 266 267 268 269 270 271 272 273 274
static bool skip_event(const char *event)
{
	const char * const *skip_events;

	for (skip_events = kvm_skip_events; *skip_events; skip_events++)
		if (!strcmp(event, *skip_events))
			return true;

	return false;
}

275
static bool handle_end_event(struct perf_kvm_stat *kvm,
276 277
			     struct vcpu_event_record *vcpu_record,
			     struct event_key *key,
278
			     struct perf_sample *sample)
X
Xiao Guangrong 已提交
279 280 281
{
	struct kvm_event *event;
	u64 time_begin, time_diff;
282 283 284 285 286 287
	int vcpu;

	if (kvm->trace_vcpu == -1)
		vcpu = -1;
	else
		vcpu = vcpu_record->vcpu_id;
X
Xiao Guangrong 已提交
288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305

	event = vcpu_record->last_event;
	time_begin = vcpu_record->start_time;

	/* The begin event is not caught. */
	if (!time_begin)
		return true;

	/*
	 * In some case, the 'begin event' only records the start timestamp,
	 * the actual event is recognized in the 'end event' (e.g. mmio-event).
	 */

	/* Both begin and end events did not get the key. */
	if (!event && key->key == INVALID_KEY)
		return true;

	if (!event)
306
		event = find_create_kvm_event(kvm, key);
X
Xiao Guangrong 已提交
307 308 309 310 311 312 313

	if (!event)
		return false;

	vcpu_record->last_event = NULL;
	vcpu_record->start_time = 0;

D
David Ahern 已提交
314
	/* seems to happen once in a while during live mode */
315
	if (sample->time < time_begin) {
D
David Ahern 已提交
316 317 318
		pr_debug("End time before begin time; skipping event.\n");
		return true;
	}
X
Xiao Guangrong 已提交
319

320 321 322
	time_diff = sample->time - time_begin;

	if (kvm->duration && time_diff > kvm->duration) {
323
		char decode[DECODE_STR_LEN];
324 325

		kvm->events_ops->decode_key(kvm, &event->key, decode);
326
		if (!skip_event(decode)) {
327 328 329 330 331 332
			pr_info("%" PRIu64 " VM %d, vcpu %d: %s event took %" PRIu64 "usec\n",
				 sample->time, sample->pid, vcpu_record->vcpu_id,
				 decode, time_diff/1000);
		}
	}

333
	return update_kvm_event(event, vcpu, time_diff);
X
Xiao Guangrong 已提交
334 335
}

336 337 338 339
static
struct vcpu_event_record *per_vcpu_record(struct thread *thread,
					  struct perf_evsel *evsel,
					  struct perf_sample *sample)
X
Xiao Guangrong 已提交
340 341
{
	/* Only kvm_entry records vcpu id. */
342
	if (!thread->priv && kvm_entry_event(evsel)) {
X
Xiao Guangrong 已提交
343 344
		struct vcpu_event_record *vcpu_record;

345
		vcpu_record = zalloc(sizeof(*vcpu_record));
X
Xiao Guangrong 已提交
346
		if (!vcpu_record) {
347
			pr_err("%s: Not enough memory\n", __func__);
X
Xiao Guangrong 已提交
348 349 350
			return NULL;
		}

351
		vcpu_record->vcpu_id = perf_evsel__intval(evsel, sample, VCPU_ID);
X
Xiao Guangrong 已提交
352 353 354
		thread->priv = vcpu_record;
	}

355
	return thread->priv;
X
Xiao Guangrong 已提交
356 357
}

358
static bool handle_kvm_event(struct perf_kvm_stat *kvm,
359 360
			     struct thread *thread,
			     struct perf_evsel *evsel,
361
			     struct perf_sample *sample)
X
Xiao Guangrong 已提交
362 363 364 365
{
	struct vcpu_event_record *vcpu_record;
	struct event_key key = {.key = INVALID_KEY};

366
	vcpu_record = per_vcpu_record(thread, evsel, sample);
X
Xiao Guangrong 已提交
367 368 369
	if (!vcpu_record)
		return true;

370 371 372 373 374
	/* only process events for vcpus user cares about */
	if ((kvm->trace_vcpu != -1) &&
	    (kvm->trace_vcpu != vcpu_record->vcpu_id))
		return true;

375 376
	if (kvm->events_ops->is_begin_event(evsel, sample, &key))
		return handle_begin_event(kvm, vcpu_record, &key, sample->time);
X
Xiao Guangrong 已提交
377

378
	if (kvm->events_ops->is_end_event(evsel, sample, &key))
379
		return handle_end_event(kvm, vcpu_record, &key, sample);
X
Xiao Guangrong 已提交
380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407

	return true;
}

#define GET_EVENT_KEY(func, field)					\
static u64 get_event_ ##func(struct kvm_event *event, int vcpu)		\
{									\
	if (vcpu == -1)							\
		return event->total.field;				\
									\
	if (vcpu >= event->max_vcpu)					\
		return 0;						\
									\
	return event->vcpu[vcpu].field;					\
}

#define COMPARE_EVENT_KEY(func, field)					\
GET_EVENT_KEY(func, field)						\
static int compare_kvm_event_ ## func(struct kvm_event *one,		\
					struct kvm_event *two, int vcpu)\
{									\
	return get_event_ ##func(one, vcpu) >				\
				get_event_ ##func(two, vcpu);		\
}

GET_EVENT_KEY(time, time);
COMPARE_EVENT_KEY(count, stats.n);
COMPARE_EVENT_KEY(mean, stats.mean);
408 409
GET_EVENT_KEY(max, stats.max);
GET_EVENT_KEY(min, stats.min);
X
Xiao Guangrong 已提交
410 411 412 413 414 415 416 417 418 419

#define DEF_SORT_NAME_KEY(name, compare_key)				\
	{ #name, compare_kvm_event_ ## compare_key }

static struct kvm_event_key keys[] = {
	DEF_SORT_NAME_KEY(sample, count),
	DEF_SORT_NAME_KEY(time, mean),
	{ NULL, NULL }
};

420
static bool select_key(struct perf_kvm_stat *kvm)
X
Xiao Guangrong 已提交
421 422 423 424
{
	int i;

	for (i = 0; keys[i].name; i++) {
425 426
		if (!strcmp(keys[i].name, kvm->sort_key)) {
			kvm->compare = keys[i].key;
X
Xiao Guangrong 已提交
427 428 429 430
			return true;
		}
	}

431
	pr_err("Unknown compare key:%s\n", kvm->sort_key);
X
Xiao Guangrong 已提交
432 433 434
	return false;
}

435 436
static void insert_to_result(struct rb_root *result, struct kvm_event *event,
			     key_cmp_fun bigger, int vcpu)
X
Xiao Guangrong 已提交
437
{
438
	struct rb_node **rb = &result->rb_node;
X
Xiao Guangrong 已提交
439 440 441 442 443 444 445 446 447 448 449 450 451 452
	struct rb_node *parent = NULL;
	struct kvm_event *p;

	while (*rb) {
		p = container_of(*rb, struct kvm_event, rb);
		parent = *rb;

		if (bigger(event, p, vcpu))
			rb = &(*rb)->rb_left;
		else
			rb = &(*rb)->rb_right;
	}

	rb_link_node(&event->rb, parent, rb);
453
	rb_insert_color(&event->rb, result);
X
Xiao Guangrong 已提交
454 455
}

456 457
static void
update_total_count(struct perf_kvm_stat *kvm, struct kvm_event *event)
X
Xiao Guangrong 已提交
458
{
459 460 461 462
	int vcpu = kvm->trace_vcpu;

	kvm->total_count += get_event_count(event, vcpu);
	kvm->total_time += get_event_time(event, vcpu);
X
Xiao Guangrong 已提交
463 464 465 466 467 468 469
}

static bool event_is_valid(struct kvm_event *event, int vcpu)
{
	return !!get_event_count(event, vcpu);
}

470
static void sort_result(struct perf_kvm_stat *kvm)
X
Xiao Guangrong 已提交
471 472
{
	unsigned int i;
473
	int vcpu = kvm->trace_vcpu;
X
Xiao Guangrong 已提交
474 475
	struct kvm_event *event;

476 477
	for (i = 0; i < EVENTS_CACHE_SIZE; i++) {
		list_for_each_entry(event, &kvm->kvm_events_cache[i], hash_entry) {
X
Xiao Guangrong 已提交
478
			if (event_is_valid(event, vcpu)) {
479 480 481
				update_total_count(kvm, event);
				insert_to_result(&kvm->result, event,
						 kvm->compare, vcpu);
X
Xiao Guangrong 已提交
482
			}
483 484
		}
	}
X
Xiao Guangrong 已提交
485 486 487
}

/* returns left most element of result, and erase it */
488
static struct kvm_event *pop_from_result(struct rb_root *result)
X
Xiao Guangrong 已提交
489
{
490
	struct rb_node *node = rb_first(result);
X
Xiao Guangrong 已提交
491 492 493 494

	if (!node)
		return NULL;

495
	rb_erase(node, result);
X
Xiao Guangrong 已提交
496 497 498
	return container_of(node, struct kvm_event, rb);
}

D
David Ahern 已提交
499
static void print_vcpu_info(struct perf_kvm_stat *kvm)
X
Xiao Guangrong 已提交
500
{
D
David Ahern 已提交
501 502
	int vcpu = kvm->trace_vcpu;

X
Xiao Guangrong 已提交
503 504
	pr_info("Analyze events for ");

D
David Ahern 已提交
505 506 507 508 509 510 511 512 513
	if (kvm->live) {
		if (kvm->opts.target.system_wide)
			pr_info("all VMs, ");
		else if (kvm->opts.target.pid)
			pr_info("pid(s) %s, ", kvm->opts.target.pid);
		else
			pr_info("dazed and confused on what is monitored, ");
	}

X
Xiao Guangrong 已提交
514 515 516 517 518 519
	if (vcpu == -1)
		pr_info("all VCPUs:\n\n");
	else
		pr_info("VCPU %d:\n\n", vcpu);
}

D
David Ahern 已提交
520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535
static void show_timeofday(void)
{
	char date[64];
	struct timeval tv;
	struct tm ltime;

	gettimeofday(&tv, NULL);
	if (localtime_r(&tv.tv_sec, &ltime)) {
		strftime(date, sizeof(date), "%H:%M:%S", &ltime);
		pr_info("%s.%06ld", date, tv.tv_usec);
	} else
		pr_info("00:00:00.000000");

	return;
}

536
static void print_result(struct perf_kvm_stat *kvm)
X
Xiao Guangrong 已提交
537
{
538
	char decode[DECODE_STR_LEN];
X
Xiao Guangrong 已提交
539
	struct kvm_event *event;
540
	int vcpu = kvm->trace_vcpu;
X
Xiao Guangrong 已提交
541

D
David Ahern 已提交
542 543 544 545 546
	if (kvm->live) {
		puts(CONSOLE_CLEAR);
		show_timeofday();
	}

X
Xiao Guangrong 已提交
547
	pr_info("\n\n");
D
David Ahern 已提交
548
	print_vcpu_info(kvm);
549
	pr_info("%*s ", DECODE_STR_LEN, kvm->events_ops->name);
X
Xiao Guangrong 已提交
550 551 552 553
	pr_info("%10s ", "Samples");
	pr_info("%9s ", "Samples%");

	pr_info("%9s ", "Time%");
554 555
	pr_info("%10s ", "Min Time");
	pr_info("%10s ", "Max Time");
X
Xiao Guangrong 已提交
556 557 558
	pr_info("%16s ", "Avg time");
	pr_info("\n\n");

559
	while ((event = pop_from_result(&kvm->result))) {
560
		u64 ecount, etime, max, min;
X
Xiao Guangrong 已提交
561 562 563

		ecount = get_event_count(event, vcpu);
		etime = get_event_time(event, vcpu);
564 565
		max = get_event_max(event, vcpu);
		min = get_event_min(event, vcpu);
X
Xiao Guangrong 已提交
566

567
		kvm->events_ops->decode_key(kvm, &event->key, decode);
568
		pr_info("%*s ", DECODE_STR_LEN, decode);
X
Xiao Guangrong 已提交
569
		pr_info("%10llu ", (unsigned long long)ecount);
570 571
		pr_info("%8.2f%% ", (double)ecount / kvm->total_count * 100);
		pr_info("%8.2f%% ", (double)etime / kvm->total_time * 100);
572 573
		pr_info("%8" PRIu64 "us ", min / 1000);
		pr_info("%8" PRIu64 "us ", max / 1000);
X
Xiao Guangrong 已提交
574 575 576 577 578
		pr_info("%9.2fus ( +-%7.2f%% )", (double)etime / ecount/1e3,
			kvm_event_rel_stddev(vcpu, event));
		pr_info("\n");
	}

579 580
	pr_info("\nTotal Samples:%" PRIu64 ", Total events handled time:%.2fus.\n\n",
		kvm->total_count, kvm->total_time / 1e3);
D
David Ahern 已提交
581 582 583 584 585

	if (kvm->lost_events)
		pr_info("\nLost events: %" PRIu64 "\n\n", kvm->lost_events);
}

586
#ifdef HAVE_TIMERFD_SUPPORT
D
David Ahern 已提交
587 588 589 590 591 592 593 594 595
static int process_lost_event(struct perf_tool *tool,
			      union perf_event *event __maybe_unused,
			      struct perf_sample *sample __maybe_unused,
			      struct machine *machine __maybe_unused)
{
	struct perf_kvm_stat *kvm = container_of(tool, struct perf_kvm_stat, tool);

	kvm->lost_events++;
	return 0;
X
Xiao Guangrong 已提交
596
}
597
#endif
X
Xiao Guangrong 已提交
598

599 600 601 602 603 604 605 606 607
static bool skip_sample(struct perf_kvm_stat *kvm,
			struct perf_sample *sample)
{
	if (kvm->pid_list && intlist__find(kvm->pid_list, sample->pid) == NULL)
		return true;

	return false;
}

608
static int process_sample_event(struct perf_tool *tool,
X
Xiao Guangrong 已提交
609 610 611 612 613
				union perf_event *event,
				struct perf_sample *sample,
				struct perf_evsel *evsel,
				struct machine *machine)
{
614
	struct thread *thread;
615 616
	struct perf_kvm_stat *kvm = container_of(tool, struct perf_kvm_stat,
						 tool);
X
Xiao Guangrong 已提交
617

618 619 620
	if (skip_sample(kvm, sample))
		return 0;

621
	thread = machine__findnew_thread(machine, sample->pid, sample->tid);
X
Xiao Guangrong 已提交
622 623 624 625 626 627
	if (thread == NULL) {
		pr_debug("problem processing %d event, skipping it.\n",
			event->header.type);
		return -1;
	}

628
	if (!handle_kvm_event(kvm, thread, evsel, sample))
X
Xiao Guangrong 已提交
629 630 631 632 633
		return -1;

	return 0;
}

D
David Ahern 已提交
634
static int cpu_isa_config(struct perf_kvm_stat *kvm)
X
Xiao Guangrong 已提交
635
{
D
David Ahern 已提交
636
	char buf[64], *cpuid;
637
	int err;
D
David Ahern 已提交
638 639 640 641

	if (kvm->live) {
		err = get_cpuid(buf, sizeof(buf));
		if (err != 0) {
642
			pr_err("Failed to look up CPU type\n");
D
David Ahern 已提交
643 644 645 646 647
			return err;
		}
		cpuid = buf;
	} else
		cpuid = kvm->session->header.env.cpuid;
X
Xiao Guangrong 已提交
648

649 650 651
	if (!cpuid) {
		pr_err("Failed to look up CPU type\n");
		return -EINVAL;
D
David Ahern 已提交
652 653
	}

654 655 656
	err = cpu_isa_init(kvm, cpuid);
	if (err == -ENOTSUP)
		pr_err("CPU %s is not supported.\n", cpuid);
D
David Ahern 已提交
657

658
	return err;
D
David Ahern 已提交
659 660 661 662 663 664 665 666 667 668 669 670
}

static bool verify_vcpu(int vcpu)
{
	if (vcpu != -1 && vcpu < 0) {
		pr_err("Invalid vcpu:%d.\n", vcpu);
		return false;
	}

	return true;
}

671
#ifdef HAVE_TIMERFD_SUPPORT
D
David Ahern 已提交
672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688
/* keeping the max events to a modest level to keep
 * the processing of samples per mmap smooth.
 */
#define PERF_KVM__MAX_EVENTS_PER_MMAP  25

static s64 perf_kvm__mmap_read_idx(struct perf_kvm_stat *kvm, int idx,
				   u64 *mmap_time)
{
	union perf_event *event;
	struct perf_sample sample;
	s64 n = 0;
	int err;

	*mmap_time = ULLONG_MAX;
	while ((event = perf_evlist__mmap_read(kvm->evlist, idx)) != NULL) {
		err = perf_evlist__parse_sample(kvm->evlist, event, &sample);
		if (err) {
689
			perf_evlist__mmap_consume(kvm->evlist, idx);
D
David Ahern 已提交
690 691 692 693 694
			pr_err("Failed to parse sample\n");
			return -1;
		}

		err = perf_session_queue_event(kvm->session, event, &sample, 0);
695 696 697 698 699 700
		/*
		 * FIXME: Here we can't consume the event, as perf_session_queue_event will
		 *        point to it, and it'll get possibly overwritten by the kernel.
		 */
		perf_evlist__mmap_consume(kvm->evlist, idx);

D
David Ahern 已提交
701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950
		if (err) {
			pr_err("Failed to enqueue sample: %d\n", err);
			return -1;
		}

		/* save time stamp of our first sample for this mmap */
		if (n == 0)
			*mmap_time = sample.time;

		/* limit events per mmap handled all at once */
		n++;
		if (n == PERF_KVM__MAX_EVENTS_PER_MMAP)
			break;
	}

	return n;
}

static int perf_kvm__mmap_read(struct perf_kvm_stat *kvm)
{
	int i, err, throttled = 0;
	s64 n, ntotal = 0;
	u64 flush_time = ULLONG_MAX, mmap_time;

	for (i = 0; i < kvm->evlist->nr_mmaps; i++) {
		n = perf_kvm__mmap_read_idx(kvm, i, &mmap_time);
		if (n < 0)
			return -1;

		/* flush time is going to be the minimum of all the individual
		 * mmap times. Essentially, we flush all the samples queued up
		 * from the last pass under our minimal start time -- that leaves
		 * a very small race for samples to come in with a lower timestamp.
		 * The ioctl to return the perf_clock timestamp should close the
		 * race entirely.
		 */
		if (mmap_time < flush_time)
			flush_time = mmap_time;

		ntotal += n;
		if (n == PERF_KVM__MAX_EVENTS_PER_MMAP)
			throttled = 1;
	}

	/* flush queue after each round in which we processed events */
	if (ntotal) {
		kvm->session->ordered_samples.next_flush = flush_time;
		err = kvm->tool.finished_round(&kvm->tool, NULL, kvm->session);
		if (err) {
			if (kvm->lost_events)
				pr_info("\nLost events: %" PRIu64 "\n\n",
					kvm->lost_events);
			return err;
		}
	}

	return throttled;
}

static volatile int done;

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

static int perf_kvm__timerfd_create(struct perf_kvm_stat *kvm)
{
	struct itimerspec new_value;
	int rc = -1;

	kvm->timerfd = timerfd_create(CLOCK_MONOTONIC, TFD_NONBLOCK);
	if (kvm->timerfd < 0) {
		pr_err("timerfd_create failed\n");
		goto out;
	}

	new_value.it_value.tv_sec = kvm->display_time;
	new_value.it_value.tv_nsec = 0;
	new_value.it_interval.tv_sec = kvm->display_time;
	new_value.it_interval.tv_nsec = 0;

	if (timerfd_settime(kvm->timerfd, 0, &new_value, NULL) != 0) {
		pr_err("timerfd_settime failed: %d\n", errno);
		close(kvm->timerfd);
		goto out;
	}

	rc = 0;
out:
	return rc;
}

static int perf_kvm__handle_timerfd(struct perf_kvm_stat *kvm)
{
	uint64_t c;
	int rc;

	rc = read(kvm->timerfd, &c, sizeof(uint64_t));
	if (rc < 0) {
		if (errno == EAGAIN)
			return 0;

		pr_err("Failed to read timer fd: %d\n", errno);
		return -1;
	}

	if (rc != sizeof(uint64_t)) {
		pr_err("Error reading timer fd - invalid size returned\n");
		return -1;
	}

	if (c != 1)
		pr_debug("Missed timer beats: %" PRIu64 "\n", c-1);

	/* update display */
	sort_result(kvm);
	print_result(kvm);

	/* reset counts */
	clear_events_cache_stats(kvm->kvm_events_cache);
	kvm->total_count = 0;
	kvm->total_time = 0;
	kvm->lost_events = 0;

	return 0;
}

static int fd_set_nonblock(int fd)
{
	long arg = 0;

	arg = fcntl(fd, F_GETFL);
	if (arg < 0) {
		pr_err("Failed to get current flags for fd %d\n", fd);
		return -1;
	}

	if (fcntl(fd, F_SETFL, arg | O_NONBLOCK) < 0) {
		pr_err("Failed to set non-block option on fd %d\n", fd);
		return -1;
	}

	return 0;
}

static
int perf_kvm__handle_stdin(struct termios *tc_now, struct termios *tc_save)
{
	int c;

	tcsetattr(0, TCSANOW, tc_now);
	c = getc(stdin);
	tcsetattr(0, TCSAFLUSH, tc_save);

	if (c == 'q')
		return 1;

	return 0;
}

static int kvm_events_live_report(struct perf_kvm_stat *kvm)
{
	struct pollfd *pollfds = NULL;
	int nr_fds, nr_stdin, ret, err = -EINVAL;
	struct termios tc, save;

	/* live flag must be set first */
	kvm->live = true;

	ret = cpu_isa_config(kvm);
	if (ret < 0)
		return ret;

	if (!verify_vcpu(kvm->trace_vcpu) ||
	    !select_key(kvm) ||
	    !register_kvm_events_ops(kvm)) {
		goto out;
	}

	init_kvm_event_record(kvm);

	tcgetattr(0, &save);
	tc = save;
	tc.c_lflag &= ~(ICANON | ECHO);
	tc.c_cc[VMIN] = 0;
	tc.c_cc[VTIME] = 0;

	signal(SIGINT, sig_handler);
	signal(SIGTERM, sig_handler);

	/* copy pollfds -- need to add timerfd and stdin */
	nr_fds = kvm->evlist->nr_fds;
	pollfds = zalloc(sizeof(struct pollfd) * (nr_fds + 2));
	if (!pollfds) {
		err = -ENOMEM;
		goto out;
	}
	memcpy(pollfds, kvm->evlist->pollfd,
		sizeof(struct pollfd) * kvm->evlist->nr_fds);

	/* add timer fd */
	if (perf_kvm__timerfd_create(kvm) < 0) {
		err = -1;
		goto out;
	}

	pollfds[nr_fds].fd = kvm->timerfd;
	pollfds[nr_fds].events = POLLIN;
	nr_fds++;

	pollfds[nr_fds].fd = fileno(stdin);
	pollfds[nr_fds].events = POLLIN;
	nr_stdin = nr_fds;
	nr_fds++;
	if (fd_set_nonblock(fileno(stdin)) != 0)
		goto out;

	/* everything is good - enable the events and process */
	perf_evlist__enable(kvm->evlist);

	while (!done) {
		int rc;

		rc = perf_kvm__mmap_read(kvm);
		if (rc < 0)
			break;

		err = perf_kvm__handle_timerfd(kvm);
		if (err)
			goto out;

		if (pollfds[nr_stdin].revents & POLLIN)
			done = perf_kvm__handle_stdin(&tc, &save);

		if (!rc && !done)
			err = poll(pollfds, nr_fds, 100);
	}

	perf_evlist__disable(kvm->evlist);

	if (err == 0) {
		sort_result(kvm);
		print_result(kvm);
	}

out:
	if (kvm->timerfd >= 0)
		close(kvm->timerfd);

951
	free(pollfds);
D
David Ahern 已提交
952 953 954 955 956 957 958 959 960 961 962 963 964 965 966
	return err;
}

static int kvm_live_open_events(struct perf_kvm_stat *kvm)
{
	int err, rc = -1;
	struct perf_evsel *pos;
	struct perf_evlist *evlist = kvm->evlist;

	perf_evlist__config(evlist, &kvm->opts);

	/*
	 * Note: exclude_{guest,host} do not apply here.
	 *       This command processes KVM tracepoints from host only
	 */
967
	evlist__for_each(evlist, pos) {
D
David Ahern 已提交
968 969 970
		struct perf_event_attr *attr = &pos->attr;

		/* make sure these *are* set */
971 972 973 974
		perf_evsel__set_sample_bit(pos, TID);
		perf_evsel__set_sample_bit(pos, TIME);
		perf_evsel__set_sample_bit(pos, CPU);
		perf_evsel__set_sample_bit(pos, RAW);
D
David Ahern 已提交
975
		/* make sure these are *not*; want as small a sample as possible */
976 977 978 979 980
		perf_evsel__reset_sample_bit(pos, PERIOD);
		perf_evsel__reset_sample_bit(pos, IP);
		perf_evsel__reset_sample_bit(pos, CALLCHAIN);
		perf_evsel__reset_sample_bit(pos, ADDR);
		perf_evsel__reset_sample_bit(pos, READ);
D
David Ahern 已提交
981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997
		attr->mmap = 0;
		attr->comm = 0;
		attr->task = 0;

		attr->sample_period = 1;

		attr->watermark = 0;
		attr->wakeup_events = 1000;

		/* will enable all once we are ready */
		attr->disabled = 1;
	}

	err = perf_evlist__open(evlist);
	if (err < 0) {
		printf("Couldn't create the events: %s\n", strerror(errno));
		goto out;
X
Xiao Guangrong 已提交
998 999
	}

D
David Ahern 已提交
1000 1001 1002 1003 1004 1005 1006 1007 1008 1009
	if (perf_evlist__mmap(evlist, kvm->opts.mmap_pages, false) < 0) {
		ui__error("Failed to mmap the events: %s\n", strerror(errno));
		perf_evlist__close(evlist);
		goto out;
	}

	rc = 0;

out:
	return rc;
X
Xiao Guangrong 已提交
1010
}
1011
#endif
X
Xiao Guangrong 已提交
1012

1013
static int read_events(struct perf_kvm_stat *kvm)
X
Xiao Guangrong 已提交
1014 1015 1016
{
	int ret;

1017 1018 1019 1020 1021
	struct perf_tool eops = {
		.sample			= process_sample_event,
		.comm			= perf_event__process_comm,
		.ordered_samples	= true,
	};
1022
	struct perf_data_file file = {
1023
		.path = kvm->file_name,
1024 1025
		.mode = PERF_DATA_MODE_READ,
	};
1026 1027

	kvm->tool = eops;
1028
	kvm->session = perf_session__new(&file, false, &kvm->tool);
1029
	if (!kvm->session) {
X
Xiao Guangrong 已提交
1030 1031 1032 1033
		pr_err("Initializing perf session failed\n");
		return -EINVAL;
	}

1034
	if (!perf_session__has_traces(kvm->session, "kvm record"))
X
Xiao Guangrong 已提交
1035 1036 1037 1038 1039 1040
		return -EINVAL;

	/*
	 * Do not use 'isa' recorded in kvm_exit tracepoint since it is not
	 * traced in the old kernel.
	 */
D
David Ahern 已提交
1041
	ret = cpu_isa_config(kvm);
X
Xiao Guangrong 已提交
1042 1043 1044
	if (ret < 0)
		return ret;

1045
	return perf_session__process_events(kvm->session, &kvm->tool);
X
Xiao Guangrong 已提交
1046 1047
}

1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060
static int parse_target_str(struct perf_kvm_stat *kvm)
{
	if (kvm->pid_str) {
		kvm->pid_list = intlist__new(kvm->pid_str);
		if (kvm->pid_list == NULL) {
			pr_err("Error parsing process id string\n");
			return -EINVAL;
		}
	}

	return 0;
}

1061
static int kvm_events_report_vcpu(struct perf_kvm_stat *kvm)
X
Xiao Guangrong 已提交
1062 1063
{
	int ret = -EINVAL;
1064
	int vcpu = kvm->trace_vcpu;
X
Xiao Guangrong 已提交
1065

1066 1067 1068
	if (parse_target_str(kvm) != 0)
		goto exit;

X
Xiao Guangrong 已提交
1069 1070 1071
	if (!verify_vcpu(vcpu))
		goto exit;

1072
	if (!select_key(kvm))
X
Xiao Guangrong 已提交
1073 1074
		goto exit;

1075
	if (!register_kvm_events_ops(kvm))
X
Xiao Guangrong 已提交
1076 1077
		goto exit;

1078
	init_kvm_event_record(kvm);
X
Xiao Guangrong 已提交
1079 1080
	setup_pager();

1081
	ret = read_events(kvm);
X
Xiao Guangrong 已提交
1082 1083 1084
	if (ret)
		goto exit;

1085 1086 1087
	sort_result(kvm);
	print_result(kvm);

X
Xiao Guangrong 已提交
1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099
exit:
	return ret;
}

#define STRDUP_FAIL_EXIT(s)		\
	({	char *_p;		\
	_p = strdup(s);		\
		if (!_p)		\
			return -ENOMEM;	\
		_p;			\
	})

1100 1101
static int
kvm_events_record(struct perf_kvm_stat *kvm, int argc, const char **argv)
X
Xiao Guangrong 已提交
1102
{
1103
	unsigned int rec_argc, i, j, events_tp_size;
X
Xiao Guangrong 已提交
1104
	const char **rec_argv;
1105 1106 1107 1108 1109 1110
	const char * const record_args[] = {
		"record",
		"-R",
		"-m", "1024",
		"-c", "1",
	};
1111 1112 1113 1114 1115
	const char * const *events_tp;
	events_tp_size = 0;

	for (events_tp = kvm_events_tp; *events_tp; events_tp++)
		events_tp_size++;
X
Xiao Guangrong 已提交
1116

1117
	rec_argc = ARRAY_SIZE(record_args) + argc + 2 +
1118
		   2 * events_tp_size;
X
Xiao Guangrong 已提交
1119 1120 1121 1122 1123 1124 1125 1126
	rec_argv = calloc(rec_argc + 1, sizeof(char *));

	if (rec_argv == NULL)
		return -ENOMEM;

	for (i = 0; i < ARRAY_SIZE(record_args); i++)
		rec_argv[i] = STRDUP_FAIL_EXIT(record_args[i]);

1127
	for (j = 0; j < events_tp_size; j++) {
1128 1129 1130 1131
		rec_argv[i++] = "-e";
		rec_argv[i++] = STRDUP_FAIL_EXIT(kvm_events_tp[j]);
	}

X
Xiao Guangrong 已提交
1132
	rec_argv[i++] = STRDUP_FAIL_EXIT("-o");
1133
	rec_argv[i++] = STRDUP_FAIL_EXIT(kvm->file_name);
X
Xiao Guangrong 已提交
1134 1135 1136 1137 1138 1139 1140

	for (j = 1; j < (unsigned int)argc; j++, i++)
		rec_argv[i] = argv[j];

	return cmd_record(i, rec_argv, NULL);
}

1141 1142
static int
kvm_events_report(struct perf_kvm_stat *kvm, int argc, const char **argv)
1143 1144 1145 1146 1147 1148 1149 1150 1151
{
	const struct option kvm_events_report_options[] = {
		OPT_STRING(0, "event", &kvm->report_event, "report event",
			    "event for reporting: vmexit, mmio, ioport"),
		OPT_INTEGER(0, "vcpu", &kvm->trace_vcpu,
			    "vcpu id to report"),
		OPT_STRING('k', "key", &kvm->sort_key, "sort-key",
			    "key for sorting: sample(sort by samples number)"
			    " time (sort by avg time)"),
1152 1153
		OPT_STRING('p', "pid", &kvm->pid_str, "pid",
			   "analyze events only for given process id(s)"),
1154 1155
		OPT_END()
	};
X
Xiao Guangrong 已提交
1156

1157 1158 1159 1160
	const char * const kvm_events_report_usage[] = {
		"perf kvm stat report [<options>]",
		NULL
	};
X
Xiao Guangrong 已提交
1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172

	symbol__init();

	if (argc) {
		argc = parse_options(argc, argv,
				     kvm_events_report_options,
				     kvm_events_report_usage, 0);
		if (argc)
			usage_with_options(kvm_events_report_usage,
					   kvm_events_report_options);
	}

1173
	return kvm_events_report_vcpu(kvm);
X
Xiao Guangrong 已提交
1174 1175
}

1176
#ifdef HAVE_TIMERFD_SUPPORT
D
David Ahern 已提交
1177 1178 1179 1180 1181
static struct perf_evlist *kvm_live_event_list(void)
{
	struct perf_evlist *evlist;
	char *tp, *name, *sys;
	int err = -1;
1182
	const char * const *events_tp;
D
David Ahern 已提交
1183 1184 1185 1186 1187

	evlist = perf_evlist__new();
	if (evlist == NULL)
		return NULL;

1188
	for (events_tp = kvm_events_tp; *events_tp; events_tp++) {
D
David Ahern 已提交
1189

1190
		tp = strdup(*events_tp);
D
David Ahern 已提交
1191 1192 1193 1194 1195 1196 1197 1198
		if (tp == NULL)
			goto out;

		/* split tracepoint into subsystem and name */
		sys = tp;
		name = strchr(tp, ':');
		if (name == NULL) {
			pr_err("Error parsing %s tracepoint: subsystem delimiter not found\n",
1199
			       *events_tp);
D
David Ahern 已提交
1200 1201 1202 1203 1204 1205 1206
			free(tp);
			goto out;
		}
		*name = '\0';
		name++;

		if (perf_evlist__add_newtp(evlist, sys, name, NULL)) {
1207
			pr_err("Failed to add %s tracepoint to the list\n", *events_tp);
D
David Ahern 已提交
1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234
			free(tp);
			goto out;
		}

		free(tp);
	}

	err = 0;

out:
	if (err) {
		perf_evlist__delete(evlist);
		evlist = NULL;
	}

	return evlist;
}

static int kvm_events_live(struct perf_kvm_stat *kvm,
			   int argc, const char **argv)
{
	char errbuf[BUFSIZ];
	int err;

	const struct option live_options[] = {
		OPT_STRING('p', "pid", &kvm->opts.target.pid, "pid",
			"record events on existing process id"),
1235 1236 1237
		OPT_CALLBACK('m', "mmap-pages", &kvm->opts.mmap_pages, "pages",
			"number of mmap data pages",
			perf_evlist__parse_mmap_pages),
D
David Ahern 已提交
1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250
		OPT_INCR('v', "verbose", &verbose,
			"be more verbose (show counter open errors, etc)"),
		OPT_BOOLEAN('a', "all-cpus", &kvm->opts.target.system_wide,
			"system-wide collection from all CPUs"),
		OPT_UINTEGER('d', "display", &kvm->display_time,
			"time in seconds between display updates"),
		OPT_STRING(0, "event", &kvm->report_event, "report event",
			"event for reporting: vmexit, mmio, ioport"),
		OPT_INTEGER(0, "vcpu", &kvm->trace_vcpu,
			"vcpu id to report"),
		OPT_STRING('k', "key", &kvm->sort_key, "sort-key",
			"key for sorting: sample(sort by samples number)"
			" time (sort by avg time)"),
1251 1252
		OPT_U64(0, "duration", &kvm->duration,
		    "show events other than HALT that take longer than duration usecs"),
D
David Ahern 已提交
1253 1254 1255 1256 1257 1258
		OPT_END()
	};
	const char * const live_usage[] = {
		"perf kvm stat live [<options>]",
		NULL
	};
1259 1260 1261
	struct perf_data_file file = {
		.mode = PERF_DATA_MODE_WRITE,
	};
D
David Ahern 已提交
1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293


	/* event handling */
	kvm->tool.sample = process_sample_event;
	kvm->tool.comm   = perf_event__process_comm;
	kvm->tool.exit   = perf_event__process_exit;
	kvm->tool.fork   = perf_event__process_fork;
	kvm->tool.lost   = process_lost_event;
	kvm->tool.ordered_samples = true;
	perf_tool__fill_defaults(&kvm->tool);

	/* set defaults */
	kvm->display_time = 1;
	kvm->opts.user_interval = 1;
	kvm->opts.mmap_pages = 512;
	kvm->opts.target.uses_mmap = false;
	kvm->opts.target.uid_str = NULL;
	kvm->opts.target.uid = UINT_MAX;

	symbol__init();
	disable_buildid_cache();

	use_browser = 0;
	setup_browser(false);

	if (argc) {
		argc = parse_options(argc, argv, live_options,
				     live_usage, 0);
		if (argc)
			usage_with_options(live_usage, live_options);
	}

1294 1295
	kvm->duration *= NSEC_PER_USEC;   /* convert usec to nsec */

D
David Ahern 已提交
1296 1297 1298
	/*
	 * target related setups
	 */
1299
	err = target__validate(&kvm->opts.target);
D
David Ahern 已提交
1300
	if (err) {
1301
		target__strerror(&kvm->opts.target, err, errbuf, BUFSIZ);
D
David Ahern 已提交
1302 1303 1304
		ui__warning("%s", errbuf);
	}

1305
	if (target__none(&kvm->opts.target))
D
David Ahern 已提交
1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325
		kvm->opts.target.system_wide = true;


	/*
	 * generate the event list
	 */
	kvm->evlist = kvm_live_event_list();
	if (kvm->evlist == NULL) {
		err = -1;
		goto out;
	}

	symbol_conf.nr_events = kvm->evlist->nr_entries;

	if (perf_evlist__create_maps(kvm->evlist, &kvm->opts.target) < 0)
		usage_with_options(live_usage, live_options);

	/*
	 * perf session
	 */
1326
	kvm->session = perf_session__new(&file, false, &kvm->tool);
D
David Ahern 已提交
1327 1328 1329 1330 1331 1332
	if (kvm->session == NULL) {
		err = -ENOMEM;
		goto out;
	}
	kvm->session->evlist = kvm->evlist;
	perf_session__set_id_hdr_size(kvm->session);
1333 1334
	machine__synthesize_threads(&kvm->session->machines.host, &kvm->opts.target,
				    kvm->evlist->threads, false);
D
David Ahern 已提交
1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346
	err = kvm_live_open_events(kvm);
	if (err)
		goto out;

	err = kvm_events_live_report(kvm);

out:
	exit_browser(0);

	if (kvm->session)
		perf_session__delete(kvm->session);
	kvm->session = NULL;
1347
	if (kvm->evlist)
D
David Ahern 已提交
1348 1349 1350 1351
		perf_evlist__delete(kvm->evlist);

	return err;
}
1352
#endif
D
David Ahern 已提交
1353

X
Xiao Guangrong 已提交
1354 1355 1356 1357 1358 1359 1360
static void print_kvm_stat_usage(void)
{
	printf("Usage: perf kvm stat <command>\n\n");

	printf("# Available commands:\n");
	printf("\trecord: record kvm events\n");
	printf("\treport: report statistical data of kvm events\n");
D
David Ahern 已提交
1361
	printf("\tlive:   live reporting of statistical data of kvm events\n");
X
Xiao Guangrong 已提交
1362 1363 1364 1365

	printf("\nOtherwise, it is the alias of 'perf stat':\n");
}

1366
static int kvm_cmd_stat(const char *file_name, int argc, const char **argv)
X
Xiao Guangrong 已提交
1367
{
1368 1369 1370 1371 1372 1373 1374 1375 1376
	struct perf_kvm_stat kvm = {
		.file_name = file_name,

		.trace_vcpu	= -1,
		.report_event	= "vmexit",
		.sort_key	= "sample",

	};

X
Xiao Guangrong 已提交
1377 1378 1379 1380 1381 1382
	if (argc == 1) {
		print_kvm_stat_usage();
		goto perf_stat;
	}

	if (!strncmp(argv[1], "rec", 3))
1383
		return kvm_events_record(&kvm, argc - 1, argv + 1);
X
Xiao Guangrong 已提交
1384 1385

	if (!strncmp(argv[1], "rep", 3))
1386
		return kvm_events_report(&kvm, argc - 1 , argv + 1);
X
Xiao Guangrong 已提交
1387

1388
#ifdef HAVE_TIMERFD_SUPPORT
D
David Ahern 已提交
1389 1390
	if (!strncmp(argv[1], "live", 4))
		return kvm_events_live(&kvm, argc - 1 , argv + 1);
1391
#endif
D
David Ahern 已提交
1392

X
Xiao Guangrong 已提交
1393 1394 1395
perf_stat:
	return cmd_stat(argc, argv, NULL);
}
1396
#endif /* HAVE_KVM_STAT_SUPPORT */
X
Xiao Guangrong 已提交
1397

1398
static int __cmd_record(const char *file_name, int argc, const char **argv)
1399 1400 1401 1402 1403 1404 1405 1406
{
	int rec_argc, i = 0, j;
	const char **rec_argv;

	rec_argc = argc + 2;
	rec_argv = calloc(rec_argc + 1, sizeof(char *));
	rec_argv[i++] = strdup("record");
	rec_argv[i++] = strdup("-o");
1407
	rec_argv[i++] = strdup(file_name);
1408 1409 1410 1411 1412 1413 1414 1415
	for (j = 1; j < argc; j++, i++)
		rec_argv[i] = argv[j];

	BUG_ON(i != rec_argc);

	return cmd_record(i, rec_argv, NULL);
}

1416
static int __cmd_report(const char *file_name, int argc, const char **argv)
1417 1418 1419 1420 1421 1422 1423 1424
{
	int rec_argc, i = 0, j;
	const char **rec_argv;

	rec_argc = argc + 2;
	rec_argv = calloc(rec_argc + 1, sizeof(char *));
	rec_argv[i++] = strdup("report");
	rec_argv[i++] = strdup("-i");
1425
	rec_argv[i++] = strdup(file_name);
1426 1427 1428 1429 1430 1431 1432 1433
	for (j = 1; j < argc; j++, i++)
		rec_argv[i] = argv[j];

	BUG_ON(i != rec_argc);

	return cmd_report(i, rec_argv, NULL);
}

1434 1435
static int
__cmd_buildid_list(const char *file_name, int argc, const char **argv)
1436 1437 1438 1439 1440 1441 1442 1443
{
	int rec_argc, i = 0, j;
	const char **rec_argv;

	rec_argc = argc + 2;
	rec_argv = calloc(rec_argc + 1, sizeof(char *));
	rec_argv[i++] = strdup("buildid-list");
	rec_argv[i++] = strdup("-i");
1444
	rec_argv[i++] = strdup(file_name);
1445 1446 1447 1448 1449 1450 1451 1452
	for (j = 1; j < argc; j++, i++)
		rec_argv[i] = argv[j];

	BUG_ON(i != rec_argc);

	return cmd_buildid_list(i, rec_argv, NULL);
}

1453
int cmd_kvm(int argc, const char **argv, const char *prefix __maybe_unused)
1454
{
1455
	const char *file_name = NULL;
1456
	const struct option kvm_options[] = {
1457
		OPT_STRING('i', "input", &file_name, "file",
1458
			   "Input file name"),
1459
		OPT_STRING('o', "output", &file_name, "file",
1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473
			   "Output file name"),
		OPT_BOOLEAN(0, "guest", &perf_guest,
			    "Collect guest os data"),
		OPT_BOOLEAN(0, "host", &perf_host,
			    "Collect host os data"),
		OPT_STRING(0, "guestmount", &symbol_conf.guestmount, "directory",
			   "guest mount directory under which every guest os"
			   " instance has a subdir"),
		OPT_STRING(0, "guestvmlinux", &symbol_conf.default_guest_vmlinux_name,
			   "file", "file saving guest os vmlinux"),
		OPT_STRING(0, "guestkallsyms", &symbol_conf.default_guest_kallsyms,
			   "file", "file saving guest os /proc/kallsyms"),
		OPT_STRING(0, "guestmodules", &symbol_conf.default_guest_modules,
			   "file", "file saving guest os /proc/modules"),
1474 1475
		OPT_INCR('v', "verbose", &verbose,
			    "be more verbose (show counter open errors, etc)"),
1476 1477 1478
		OPT_END()
	};

1479 1480 1481
	const char *const kvm_subcommands[] = { "top", "record", "report", "diff",
						"buildid-list", "stat", NULL };
	const char *kvm_usage[] = { NULL, NULL };
1482

1483 1484
	perf_host  = 0;
	perf_guest = 1;
1485

1486 1487
	argc = parse_options_subcommand(argc, argv, kvm_options, kvm_subcommands, kvm_usage,
					PARSE_OPT_STOP_AT_NON_OPTION);
1488 1489 1490 1491 1492 1493
	if (!argc)
		usage_with_options(kvm_usage, kvm_options);

	if (!perf_host)
		perf_guest = 1;

1494
	if (!file_name) {
1495
		file_name = get_filename_for_perf_kvm();
1496

1497
		if (!file_name) {
1498 1499 1500
			pr_err("Failed to allocate memory for filename\n");
			return -ENOMEM;
		}
1501 1502 1503
	}

	if (!strncmp(argv[0], "rec", 3))
1504
		return __cmd_record(file_name, argc, argv);
1505
	else if (!strncmp(argv[0], "rep", 3))
1506
		return __cmd_report(file_name, argc, argv);
1507 1508 1509 1510 1511
	else if (!strncmp(argv[0], "diff", 4))
		return cmd_diff(argc, argv, NULL);
	else if (!strncmp(argv[0], "top", 3))
		return cmd_top(argc, argv, NULL);
	else if (!strncmp(argv[0], "buildid-list", 12))
1512
		return __cmd_buildid_list(file_name, argc, argv);
1513
#ifdef HAVE_KVM_STAT_SUPPORT
X
Xiao Guangrong 已提交
1514
	else if (!strncmp(argv[0], "stat", 4))
1515
		return kvm_cmd_stat(file_name, argc, argv);
1516
#endif
1517 1518 1519 1520 1521
	else
		usage_with_options(kvm_usage, kvm_options);

	return 0;
}