builtin-trace.c 95.0 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
/*
 * builtin-trace.c
 *
 * Builtin 'trace' command:
 *
 * Display a continuously updated trace of any workload, CPU, specific PID,
 * system wide, etc.  Default format is loosely strace like, but any other
 * event may be specified using --event.
 *
 * Copyright (C) 2012, 2013, 2014, 2015 Red Hat Inc, Arnaldo Carvalho de Melo <acme@redhat.com>
 *
 * Initially based on the 'trace' prototype by Thomas Gleixner:
 *
 * http://lwn.net/Articles/415728/ ("Announcing a new utility: 'trace'")
 *
 * Released under the GPL v2. (and only v2, not any later version)
 */

19
#include <traceevent/event-parse.h>
20
#include <api/fs/tracing_path.h>
A
Arnaldo Carvalho de Melo 已提交
21
#include "builtin.h"
22
#include "util/cgroup.h"
23
#include "util/color.h"
24
#include "util/debug.h"
25
#include "util/env.h"
26
#include "util/event.h"
A
Arnaldo Carvalho de Melo 已提交
27
#include "util/evlist.h"
28
#include <subcmd/exec-cmd.h>
29
#include "util/machine.h"
30
#include "util/path.h"
31
#include "util/session.h"
32
#include "util/thread.h"
33
#include <subcmd/parse-options.h>
34
#include "util/strlist.h"
35
#include "util/intlist.h"
A
Arnaldo Carvalho de Melo 已提交
36
#include "util/thread_map.h"
37
#include "util/stat.h"
38
#include "trace/beauty/beauty.h"
39
#include "trace-event.h"
40
#include "util/parse-events.h"
41
#include "util/bpf-loader.h"
42
#include "callchain.h"
43
#include "print_binary.h"
44
#include "string2.h"
45
#include "syscalltbl.h"
46
#include "rb_resort.h"
A
Arnaldo Carvalho de Melo 已提交
47

48
#include <errno.h>
49
#include <inttypes.h>
50
#include <poll.h>
51
#include <signal.h>
A
Arnaldo Carvalho de Melo 已提交
52
#include <stdlib.h>
53
#include <string.h>
54
#include <linux/err.h>
55
#include <linux/filter.h>
56
#include <linux/kernel.h>
57
#include <linux/random.h>
58
#include <linux/stringify.h>
59
#include <linux/time64.h>
60
#include <fcntl.h>
A
Arnaldo Carvalho de Melo 已提交
61

62 63
#include "sane_ctype.h"

64 65 66 67
#ifndef O_CLOEXEC
# define O_CLOEXEC		02000000
#endif

68 69 70 71
#ifndef F_LINUX_SPECIFIC_BASE
# define F_LINUX_SPECIFIC_BASE	1024
#endif

72 73
struct trace {
	struct perf_tool	tool;
74
	struct syscalltbl	*sctbl;
75 76 77 78 79
	struct {
		int		max;
		struct syscall  *table;
		struct {
			struct perf_evsel *sys_enter,
80 81
					  *sys_exit,
					  *augmented;
82 83 84 85 86 87
		}		events;
	} syscalls;
	struct record_opts	opts;
	struct perf_evlist	*evlist;
	struct machine		*host;
	struct thread		*current;
88
	struct cgroup		*cgroup;
89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106
	u64			base_time;
	FILE			*output;
	unsigned long		nr_events;
	struct strlist		*ev_qualifier;
	struct {
		size_t		nr;
		int		*entries;
	}			ev_qualifier_ids;
	struct {
		size_t		nr;
		pid_t		*entries;
	}			filter_pids;
	double			duration_filter;
	double			runtime_ms;
	struct {
		u64		vfs_getname,
				proc_getname;
	} stats;
107
	unsigned int		max_stack;
108
	unsigned int		min_stack;
109 110 111 112 113 114 115
	bool			not_ev_qualifier;
	bool			live;
	bool			full_time;
	bool			sched;
	bool			multiple_threads;
	bool			summary;
	bool			summary_only;
116
	bool			failure_only;
117
	bool			show_comm;
118
	bool			print_sample;
119 120
	bool			show_tool_stats;
	bool			trace_syscalls;
121
	bool			kernel_syscallchains;
122 123 124 125
	bool			force;
	bool			vfs_getname;
	int			trace_pgfaults;
};
126

127 128 129 130 131 132 133 134 135 136 137
struct tp_field {
	int offset;
	union {
		u64 (*integer)(struct tp_field *field, struct perf_sample *sample);
		void *(*pointer)(struct tp_field *field, struct perf_sample *sample);
	};
};

#define TP_UINT_FIELD(bits) \
static u64 tp_field__u##bits(struct tp_field *field, struct perf_sample *sample) \
{ \
138 139 140
	u##bits value; \
	memcpy(&value, sample->raw_data + field->offset, sizeof(value)); \
	return value;  \
141 142 143 144 145 146 147 148 149 150
}

TP_UINT_FIELD(8);
TP_UINT_FIELD(16);
TP_UINT_FIELD(32);
TP_UINT_FIELD(64);

#define TP_UINT_FIELD__SWAPPED(bits) \
static u64 tp_field__swapped_u##bits(struct tp_field *field, struct perf_sample *sample) \
{ \
151 152
	u##bits value; \
	memcpy(&value, sample->raw_data + field->offset, sizeof(value)); \
153 154 155 156 157 158 159
	return bswap_##bits(value);\
}

TP_UINT_FIELD__SWAPPED(16);
TP_UINT_FIELD__SWAPPED(32);
TP_UINT_FIELD__SWAPPED(64);

160
static int __tp_field__init_uint(struct tp_field *field, int size, int offset, bool needs_swap)
161
{
162
	field->offset = offset;
163

164
	switch (size) {
165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183
	case 1:
		field->integer = tp_field__u8;
		break;
	case 2:
		field->integer = needs_swap ? tp_field__swapped_u16 : tp_field__u16;
		break;
	case 4:
		field->integer = needs_swap ? tp_field__swapped_u32 : tp_field__u32;
		break;
	case 8:
		field->integer = needs_swap ? tp_field__swapped_u64 : tp_field__u64;
		break;
	default:
		return -1;
	}

	return 0;
}

184 185 186 187 188
static int tp_field__init_uint(struct tp_field *field, struct format_field *format_field, bool needs_swap)
{
	return __tp_field__init_uint(field, format_field->size, format_field->offset, needs_swap);
}

189 190 191 192 193
static void *tp_field__ptr(struct tp_field *field, struct perf_sample *sample)
{
	return sample->raw_data + field->offset;
}

194
static int __tp_field__init_ptr(struct tp_field *field, int offset)
195
{
196
	field->offset = offset;
197 198 199 200
	field->pointer = tp_field__ptr;
	return 0;
}

201 202 203 204 205
static int tp_field__init_ptr(struct tp_field *field, struct format_field *format_field)
{
	return __tp_field__init_ptr(field, format_field->offset);
}

206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246
struct syscall_tp {
	struct tp_field id;
	union {
		struct tp_field args, ret;
	};
};

static int perf_evsel__init_tp_uint_field(struct perf_evsel *evsel,
					  struct tp_field *field,
					  const char *name)
{
	struct format_field *format_field = perf_evsel__field(evsel, name);

	if (format_field == NULL)
		return -1;

	return tp_field__init_uint(field, format_field, evsel->needs_swap);
}

#define perf_evsel__init_sc_tp_uint_field(evsel, name) \
	({ struct syscall_tp *sc = evsel->priv;\
	   perf_evsel__init_tp_uint_field(evsel, &sc->name, #name); })

static int perf_evsel__init_tp_ptr_field(struct perf_evsel *evsel,
					 struct tp_field *field,
					 const char *name)
{
	struct format_field *format_field = perf_evsel__field(evsel, name);

	if (format_field == NULL)
		return -1;

	return tp_field__init_ptr(field, format_field);
}

#define perf_evsel__init_sc_tp_ptr_field(evsel, name) \
	({ struct syscall_tp *sc = evsel->priv;\
	   perf_evsel__init_tp_ptr_field(evsel, &sc->name, #name); })

static void perf_evsel__delete_priv(struct perf_evsel *evsel)
{
247
	zfree(&evsel->priv);
248 249 250
	perf_evsel__delete(evsel);
}

251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266
static int perf_evsel__init_syscall_tp(struct perf_evsel *evsel)
{
	struct syscall_tp *sc = evsel->priv = malloc(sizeof(struct syscall_tp));

	if (evsel->priv != NULL) {
		if (perf_evsel__init_tp_uint_field(evsel, &sc->id, "__syscall_nr"))
			goto out_delete;
		return 0;
	}

	return -ENOMEM;
out_delete:
	zfree(&evsel->priv);
	return -ENOENT;
}

267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290
static int perf_evsel__init_augmented_syscall_tp(struct perf_evsel *evsel)
{
	struct syscall_tp *sc = evsel->priv = malloc(sizeof(struct syscall_tp));

	if (evsel->priv != NULL) {       /* field, sizeof_field, offsetof_field */
		if (__tp_field__init_uint(&sc->id, sizeof(long), sizeof(long long), evsel->needs_swap))
			goto out_delete;

		return 0;
	}

	return -ENOMEM;
out_delete:
	zfree(&evsel->priv);
	return -EINVAL;
}

static int perf_evsel__init_augmented_syscall_tp_args(struct perf_evsel *evsel)
{
	struct syscall_tp *sc = evsel->priv;

	return __tp_field__init_ptr(&sc->args, sc->id.offset + sizeof(u64));
}

291
static int perf_evsel__init_raw_syscall_tp(struct perf_evsel *evsel, void *handler)
292 293 294 295 296 297 298 299 300 301 302 303 304
{
	evsel->priv = malloc(sizeof(struct syscall_tp));
	if (evsel->priv != NULL) {
		if (perf_evsel__init_sc_tp_uint_field(evsel, id))
			goto out_delete;

		evsel->handler = handler;
		return 0;
	}

	return -ENOMEM;

out_delete:
305
	zfree(&evsel->priv);
306 307 308
	return -ENOENT;
}

309
static struct perf_evsel *perf_evsel__raw_syscall_newtp(const char *direction, void *handler)
310
{
311
	struct perf_evsel *evsel = perf_evsel__newtp("raw_syscalls", direction);
312

313
	/* older kernel (e.g., RHEL6) use syscalls:{enter,exit} */
314
	if (IS_ERR(evsel))
315 316
		evsel = perf_evsel__newtp("syscalls", direction);

317 318 319
	if (IS_ERR(evsel))
		return NULL;

320
	if (perf_evsel__init_raw_syscall_tp(evsel, handler))
321
		goto out_delete;
322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337

	return evsel;

out_delete:
	perf_evsel__delete_priv(evsel);
	return NULL;
}

#define perf_evsel__sc_tp_uint(evsel, name, sample) \
	({ struct syscall_tp *fields = evsel->priv; \
	   fields->name.integer(&fields->name, sample); })

#define perf_evsel__sc_tp_ptr(evsel, name, sample) \
	({ struct syscall_tp *fields = evsel->priv; \
	   fields->name.pointer(&fields->name, sample); })

338 339 340
size_t strarray__scnprintf(struct strarray *sa, char *bf, size_t size, const char *intfmt, int val)
{
	int idx = val - sa->offset;
341

342
	if (idx < 0 || idx >= sa->nr_entries || sa->entries[idx] == NULL)
343
		return scnprintf(bf, size, intfmt, val);
344

345
	return scnprintf(bf, size, "%s", sa->entries[idx]);
346 347
}

348 349 350
static size_t __syscall_arg__scnprintf_strarray(char *bf, size_t size,
						const char *intfmt,
					        struct syscall_arg *arg)
351
{
352
	return strarray__scnprintf(arg->parm, bf, size, intfmt, arg->val);
353 354
}

355 356 357 358 359 360
static size_t syscall_arg__scnprintf_strarray(char *bf, size_t size,
					      struct syscall_arg *arg)
{
	return __syscall_arg__scnprintf_strarray(bf, size, "%d", arg);
}

361 362
#define SCA_STRARRAY syscall_arg__scnprintf_strarray

363 364 365 366 367 368 369 370 371 372
struct strarrays {
	int		nr_entries;
	struct strarray **entries;
};

#define DEFINE_STRARRAYS(array) struct strarrays strarrays__##array = { \
	.nr_entries = ARRAY_SIZE(array), \
	.entries = array, \
}

373 374
size_t syscall_arg__scnprintf_strarrays(char *bf, size_t size,
					struct syscall_arg *arg)
375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392
{
	struct strarrays *sas = arg->parm;
	int i;

	for (i = 0; i < sas->nr_entries; ++i) {
		struct strarray *sa = sas->entries[i];
		int idx = arg->val - sa->offset;

		if (idx >= 0 && idx < sa->nr_entries) {
			if (sa->entries[idx] == NULL)
				break;
			return scnprintf(bf, size, "%s", sa->entries[idx]);
		}
	}

	return scnprintf(bf, size, "%d", arg->val);
}

393 394 395 396
#ifndef AT_FDCWD
#define AT_FDCWD	-100
#endif

397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414
static size_t syscall_arg__scnprintf_fd_at(char *bf, size_t size,
					   struct syscall_arg *arg)
{
	int fd = arg->val;

	if (fd == AT_FDCWD)
		return scnprintf(bf, size, "CWD");

	return syscall_arg__scnprintf_fd(bf, size, arg);
}

#define SCA_FDAT syscall_arg__scnprintf_fd_at

static size_t syscall_arg__scnprintf_close_fd(char *bf, size_t size,
					      struct syscall_arg *arg);

#define SCA_CLOSE_FD syscall_arg__scnprintf_close_fd

415
size_t syscall_arg__scnprintf_hex(char *bf, size_t size, struct syscall_arg *arg)
416
{
417
	return scnprintf(bf, size, "%#lx", arg->val);
418 419
}

420
size_t syscall_arg__scnprintf_int(char *bf, size_t size, struct syscall_arg *arg)
421 422 423 424
{
	return scnprintf(bf, size, "%d", arg->val);
}

425 426 427 428 429
size_t syscall_arg__scnprintf_long(char *bf, size_t size, struct syscall_arg *arg)
{
	return scnprintf(bf, size, "%ld", arg->val);
}

430 431 432 433 434 435
static const char *bpf_cmd[] = {
	"MAP_CREATE", "MAP_LOOKUP_ELEM", "MAP_UPDATE_ELEM", "MAP_DELETE_ELEM",
	"MAP_GET_NEXT_KEY", "PROG_LOAD",
};
static DEFINE_STRARRAY(bpf_cmd);

436 437
static const char *epoll_ctl_ops[] = { "ADD", "DEL", "MOD", };
static DEFINE_STRARRAY_OFFSET(epoll_ctl_ops, 1);
438

439 440 441
static const char *itimers[] = { "REAL", "VIRTUAL", "PROF", };
static DEFINE_STRARRAY(itimers);

442 443 444 445 446 447 448 449 450
static const char *keyctl_options[] = {
	"GET_KEYRING_ID", "JOIN_SESSION_KEYRING", "UPDATE", "REVOKE", "CHOWN",
	"SETPERM", "DESCRIBE", "CLEAR", "LINK", "UNLINK", "SEARCH", "READ",
	"INSTANTIATE", "NEGATE", "SET_REQKEY_KEYRING", "SET_TIMEOUT",
	"ASSUME_AUTHORITY", "GET_SECURITY", "SESSION_TO_PARENT", "REJECT",
	"INSTANTIATE_IOV", "INVALIDATE", "GET_PERSISTENT",
};
static DEFINE_STRARRAY(keyctl_options);

451 452 453 454 455 456 457 458 459
static const char *whences[] = { "SET", "CUR", "END",
#ifdef SEEK_DATA
"DATA",
#endif
#ifdef SEEK_HOLE
"HOLE",
#endif
};
static DEFINE_STRARRAY(whences);
460

461 462
static const char *fcntl_cmds[] = {
	"DUPFD", "GETFD", "SETFD", "GETFL", "SETFL", "GETLK", "SETLK",
463 464 465
	"SETLKW", "SETOWN", "GETOWN", "SETSIG", "GETSIG", "GETLK64",
	"SETLK64", "SETLKW64", "SETOWN_EX", "GETOWN_EX",
	"GETOWNER_UIDS",
466 467 468
};
static DEFINE_STRARRAY(fcntl_cmds);

469 470 471
static const char *fcntl_linux_specific_cmds[] = {
	"SETLEASE", "GETLEASE", "NOTIFY", [5] =	"CANCELLK", "DUPFD_CLOEXEC",
	"SETPIPE_SZ", "GETPIPE_SZ", "ADD_SEALS", "GET_SEALS",
472
	"GET_RW_HINT", "SET_RW_HINT", "GET_FILE_RW_HINT", "SET_FILE_RW_HINT",
473 474 475 476 477 478 479 480 481 482 483
};

static DEFINE_STRARRAY_OFFSET(fcntl_linux_specific_cmds, F_LINUX_SPECIFIC_BASE);

static struct strarray *fcntl_cmds_arrays[] = {
	&strarray__fcntl_cmds,
	&strarray__fcntl_linux_specific_cmds,
};

static DEFINE_STRARRAYS(fcntl_cmds_arrays);

484 485 486 487 488 489 490
static const char *rlimit_resources[] = {
	"CPU", "FSIZE", "DATA", "STACK", "CORE", "RSS", "NPROC", "NOFILE",
	"MEMLOCK", "AS", "LOCKS", "SIGPENDING", "MSGQUEUE", "NICE", "RTPRIO",
	"RTTIME",
};
static DEFINE_STRARRAY(rlimit_resources);

491 492 493
static const char *sighow[] = { "BLOCK", "UNBLOCK", "SETMASK", };
static DEFINE_STRARRAY(sighow);

494 495
static const char *clockid[] = {
	"REALTIME", "MONOTONIC", "PROCESS_CPUTIME_ID", "THREAD_CPUTIME_ID",
496 497
	"MONOTONIC_RAW", "REALTIME_COARSE", "MONOTONIC_COARSE", "BOOTTIME",
	"REALTIME_ALARM", "BOOTTIME_ALARM", "SGI_CYCLE", "TAI"
498 499 500
};
static DEFINE_STRARRAY(clockid);

501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527
static size_t syscall_arg__scnprintf_access_mode(char *bf, size_t size,
						 struct syscall_arg *arg)
{
	size_t printed = 0;
	int mode = arg->val;

	if (mode == F_OK) /* 0 */
		return scnprintf(bf, size, "F");
#define	P_MODE(n) \
	if (mode & n##_OK) { \
		printed += scnprintf(bf + printed, size - printed, "%s", #n); \
		mode &= ~n##_OK; \
	}

	P_MODE(R);
	P_MODE(W);
	P_MODE(X);
#undef P_MODE

	if (mode)
		printed += scnprintf(bf + printed, size - printed, "|%#x", mode);

	return printed;
}

#define SCA_ACCMODE syscall_arg__scnprintf_access_mode

528 529 530 531 532
static size_t syscall_arg__scnprintf_filename(char *bf, size_t size,
					      struct syscall_arg *arg);

#define SCA_FILENAME syscall_arg__scnprintf_filename

533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555
static size_t syscall_arg__scnprintf_pipe_flags(char *bf, size_t size,
						struct syscall_arg *arg)
{
	int printed = 0, flags = arg->val;

#define	P_FLAG(n) \
	if (flags & O_##n) { \
		printed += scnprintf(bf + printed, size - printed, "%s%s", printed ? "|" : "", #n); \
		flags &= ~O_##n; \
	}

	P_FLAG(CLOEXEC);
	P_FLAG(NONBLOCK);
#undef P_FLAG

	if (flags)
		printed += scnprintf(bf + printed, size - printed, "%s%#x", printed ? "|" : "", flags);

	return printed;
}

#define SCA_PIPE_FLAGS syscall_arg__scnprintf_pipe_flags

556 557 558 559 560 561 562
#ifndef GRND_NONBLOCK
#define GRND_NONBLOCK	0x0001
#endif
#ifndef GRND_RANDOM
#define GRND_RANDOM	0x0002
#endif

563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585
static size_t syscall_arg__scnprintf_getrandom_flags(char *bf, size_t size,
						   struct syscall_arg *arg)
{
	int printed = 0, flags = arg->val;

#define	P_FLAG(n) \
	if (flags & GRND_##n) { \
		printed += scnprintf(bf + printed, size - printed, "%s%s", printed ? "|" : "", #n); \
		flags &= ~GRND_##n; \
	}

	P_FLAG(RANDOM);
	P_FLAG(NONBLOCK);
#undef P_FLAG

	if (flags)
		printed += scnprintf(bf + printed, size - printed, "%s%#x", printed ? "|" : "", flags);

	return printed;
}

#define SCA_GETRANDOM_FLAGS syscall_arg__scnprintf_getrandom_flags

586 587 588
#define STRARRAY(name, array) \
	  { .scnprintf	= SCA_STRARRAY, \
	    .parm	= &strarray__##array, }
589

590
#include "trace/beauty/arch_errno_names.c"
591
#include "trace/beauty/eventfd.c"
592
#include "trace/beauty/futex_op.c"
593
#include "trace/beauty/futex_val3.c"
594
#include "trace/beauty/mmap.c"
595
#include "trace/beauty/mode_t.c"
596
#include "trace/beauty/msg_flags.c"
597
#include "trace/beauty/open_flags.c"
598
#include "trace/beauty/perf_event_open.c"
599
#include "trace/beauty/pid.c"
600
#include "trace/beauty/sched_policy.c"
601
#include "trace/beauty/seccomp.c"
602
#include "trace/beauty/signum.c"
603
#include "trace/beauty/socket_type.c"
604
#include "trace/beauty/waitid_options.c"
605

606 607 608
struct syscall_arg_fmt {
	size_t	   (*scnprintf)(char *bf, size_t size, struct syscall_arg *arg);
	void	   *parm;
609
	const char *name;
610
	bool	   show_zero;
611 612
};

A
Arnaldo Carvalho de Melo 已提交
613 614
static struct syscall_fmt {
	const char *name;
615
	const char *alias;
616
	struct syscall_arg_fmt arg[6];
617
	u8	   nr_args;
618
	bool	   errpid;
A
Arnaldo Carvalho de Melo 已提交
619
	bool	   timeout;
620
	bool	   hexret;
A
Arnaldo Carvalho de Melo 已提交
621
} syscall_fmts[] = {
622
	{ .name	    = "access",
623
	  .arg = { [1] = { .scnprintf = SCA_ACCMODE,  /* mode */ }, }, },
624 625
	{ .name	    = "bind",
	  .arg = { [1] = { .scnprintf = SCA_SOCKADDR, /* umyaddr */ }, }, },
626
	{ .name	    = "bpf",
627
	  .arg = { [0] = STRARRAY(cmd, bpf_cmd), }, },
628
	{ .name	    = "brk",	    .hexret = true,
629
	  .arg = { [0] = { .scnprintf = SCA_HEX, /* brk */ }, }, },
630
	{ .name     = "clock_gettime",
631
	  .arg = { [0] = STRARRAY(clk_id, clockid), }, },
632 633 634 635 636 637
	{ .name	    = "clone",	    .errpid = true, .nr_args = 5,
	  .arg = { [0] = { .name = "flags",	    .scnprintf = SCA_CLONE_FLAGS, },
		   [1] = { .name = "child_stack",   .scnprintf = SCA_HEX, },
		   [2] = { .name = "parent_tidptr", .scnprintf = SCA_HEX, },
		   [3] = { .name = "child_tidptr",  .scnprintf = SCA_HEX, },
		   [4] = { .name = "tls",	    .scnprintf = SCA_HEX, }, }, },
638
	{ .name	    = "close",
639
	  .arg = { [0] = { .scnprintf = SCA_CLOSE_FD, /* fd */ }, }, },
640 641
	{ .name	    = "connect",
	  .arg = { [1] = { .scnprintf = SCA_SOCKADDR, /* servaddr */ }, }, },
642
	{ .name	    = "epoll_ctl",
643
	  .arg = { [1] = STRARRAY(op, epoll_ctl_ops), }, },
644
	{ .name	    = "eventfd2",
645
	  .arg = { [1] = { .scnprintf = SCA_EFD_FLAGS, /* flags */ }, }, },
646
	{ .name	    = "fchmodat",
647
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* fd */ }, }, },
648
	{ .name	    = "fchownat",
649
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* fd */ }, }, },
650
	{ .name	    = "fcntl",
651
	  .arg = { [1] = { .scnprintf = SCA_FCNTL_CMD, /* cmd */
652 653
			   .parm      = &strarrays__fcntl_cmds_arrays,
			   .show_zero = true, },
654
		   [2] = { .scnprintf =  SCA_FCNTL_ARG, /* arg */ }, }, },
655
	{ .name	    = "flock",
656
	  .arg = { [1] = { .scnprintf = SCA_FLOCK, /* cmd */ }, }, },
657 658 659
	{ .name	    = "fstat", .alias = "newfstat", },
	{ .name	    = "fstatat", .alias = "newfstatat", },
	{ .name	    = "futex",
660 661
	  .arg = { [1] = { .scnprintf = SCA_FUTEX_OP, /* op */ },
		   [5] = { .scnprintf = SCA_FUTEX_VAL3, /* val3 */ }, }, },
662
	{ .name	    = "futimesat",
663
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* fd */ }, }, },
664
	{ .name	    = "getitimer",
665
	  .arg = { [0] = STRARRAY(which, itimers), }, },
666
	{ .name	    = "getpid",	    .errpid = true, },
667
	{ .name	    = "getpgid",    .errpid = true, },
668
	{ .name	    = "getppid",    .errpid = true, },
669
	{ .name	    = "getrandom",
670
	  .arg = { [2] = { .scnprintf = SCA_GETRANDOM_FLAGS, /* flags */ }, }, },
671
	{ .name	    = "getrlimit",
672
	  .arg = { [0] = STRARRAY(resource, rlimit_resources), }, },
673
	{ .name	    = "gettid",	    .errpid = true, },
674
	{ .name	    = "ioctl",
675
	  .arg = {
676 677 678 679
#if defined(__i386__) || defined(__x86_64__)
/*
 * FIXME: Make this available to all arches.
 */
680
		   [1] = { .scnprintf = SCA_IOCTL_CMD, /* cmd */ },
681
		   [2] = { .scnprintf = SCA_HEX, /* arg */ }, }, },
682
#else
683
		   [2] = { .scnprintf = SCA_HEX, /* arg */ }, }, },
684
#endif
685 686 687 688 689 690
	{ .name	    = "kcmp",	    .nr_args = 5,
	  .arg = { [0] = { .name = "pid1",	.scnprintf = SCA_PID, },
		   [1] = { .name = "pid2",	.scnprintf = SCA_PID, },
		   [2] = { .name = "type",	.scnprintf = SCA_KCMP_TYPE, },
		   [3] = { .name = "idx1",	.scnprintf = SCA_KCMP_IDX, },
		   [4] = { .name = "idx2",	.scnprintf = SCA_KCMP_IDX, }, }, },
691
	{ .name	    = "keyctl",
692
	  .arg = { [0] = STRARRAY(option, keyctl_options), }, },
693
	{ .name	    = "kill",
694
	  .arg = { [1] = { .scnprintf = SCA_SIGNUM, /* sig */ }, }, },
695
	{ .name	    = "linkat",
696
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* fd */ }, }, },
697
	{ .name	    = "lseek",
698
	  .arg = { [2] = STRARRAY(whence, whences), }, },
699 700
	{ .name	    = "lstat", .alias = "newlstat", },
	{ .name     = "madvise",
701 702
	  .arg = { [0] = { .scnprintf = SCA_HEX,      /* start */ },
		   [2] = { .scnprintf = SCA_MADV_BHV, /* behavior */ }, }, },
703
	{ .name	    = "mkdirat",
704
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* fd */ }, }, },
705
	{ .name	    = "mknodat",
706
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* fd */ }, }, },
707
	{ .name	    = "mlock",
708
	  .arg = { [0] = { .scnprintf = SCA_HEX, /* addr */ }, }, },
709
	{ .name	    = "mlockall",
710
	  .arg = { [0] = { .scnprintf = SCA_HEX, /* addr */ }, }, },
711
	{ .name	    = "mmap",	    .hexret = true,
J
Jiri Olsa 已提交
712 713 714 715
/* The standard mmap maps to old_mmap on s390x */
#if defined(__s390x__)
	.alias = "old_mmap",
#endif
716 717 718
	  .arg = { [0] = { .scnprintf = SCA_HEX,	/* addr */ },
		   [2] = { .scnprintf = SCA_MMAP_PROT,	/* prot */ },
		   [3] = { .scnprintf = SCA_MMAP_FLAGS,	/* flags */ }, }, },
719
	{ .name	    = "mprotect",
720 721
	  .arg = { [0] = { .scnprintf = SCA_HEX,	/* start */ },
		   [2] = { .scnprintf = SCA_MMAP_PROT,	/* prot */ }, }, },
722
	{ .name	    = "mq_unlink",
723
	  .arg = { [0] = { .scnprintf = SCA_FILENAME, /* u_name */ }, }, },
724
	{ .name	    = "mremap",	    .hexret = true,
725 726 727
	  .arg = { [0] = { .scnprintf = SCA_HEX,	  /* addr */ },
		   [3] = { .scnprintf = SCA_MREMAP_FLAGS, /* flags */ },
		   [4] = { .scnprintf = SCA_HEX,	  /* new_addr */ }, }, },
728
	{ .name	    = "munlock",
729
	  .arg = { [0] = { .scnprintf = SCA_HEX, /* addr */ }, }, },
730
	{ .name	    = "munmap",
731
	  .arg = { [0] = { .scnprintf = SCA_HEX, /* addr */ }, }, },
732
	{ .name	    = "name_to_handle_at",
733
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* dfd */ }, }, },
734
	{ .name	    = "newfstatat",
735
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* dfd */ }, }, },
736
	{ .name	    = "open",
737
	  .arg = { [1] = { .scnprintf = SCA_OPEN_FLAGS, /* flags */ }, }, },
738
	{ .name	    = "open_by_handle_at",
739 740
	  .arg = { [0] = { .scnprintf = SCA_FDAT,	/* dfd */ },
		   [2] = { .scnprintf = SCA_OPEN_FLAGS, /* flags */ }, }, },
741
	{ .name	    = "openat",
742 743
	  .arg = { [0] = { .scnprintf = SCA_FDAT,	/* dfd */ },
		   [2] = { .scnprintf = SCA_OPEN_FLAGS, /* flags */ }, }, },
744
	{ .name	    = "perf_event_open",
745 746 747
	  .arg = { [2] = { .scnprintf = SCA_INT,	/* cpu */ },
		   [3] = { .scnprintf = SCA_FD,		/* group_fd */ },
		   [4] = { .scnprintf = SCA_PERF_FLAGS, /* flags */ }, }, },
748
	{ .name	    = "pipe2",
749
	  .arg = { [1] = { .scnprintf = SCA_PIPE_FLAGS, /* flags */ }, }, },
750 751 752 753 754 755 756 757
	{ .name	    = "pkey_alloc",
	  .arg = { [1] = { .scnprintf = SCA_PKEY_ALLOC_ACCESS_RIGHTS,	/* access_rights */ }, }, },
	{ .name	    = "pkey_free",
	  .arg = { [0] = { .scnprintf = SCA_INT,	/* key */ }, }, },
	{ .name	    = "pkey_mprotect",
	  .arg = { [0] = { .scnprintf = SCA_HEX,	/* start */ },
		   [2] = { .scnprintf = SCA_MMAP_PROT,	/* prot */ },
		   [3] = { .scnprintf = SCA_INT,	/* pkey */ }, }, },
758 759
	{ .name	    = "poll", .timeout = true, },
	{ .name	    = "ppoll", .timeout = true, },
760 761 762 763
	{ .name	    = "prctl", .alias = "arch_prctl",
	  .arg = { [0] = { .scnprintf = SCA_PRCTL_OPTION, /* option */ },
		   [1] = { .scnprintf = SCA_PRCTL_ARG2, /* arg2 */ },
		   [2] = { .scnprintf = SCA_PRCTL_ARG3, /* arg3 */ }, }, },
764 765 766
	{ .name	    = "pread", .alias = "pread64", },
	{ .name	    = "preadv", .alias = "pread", },
	{ .name	    = "prlimit64",
767
	  .arg = { [1] = STRARRAY(resource, rlimit_resources), }, },
768 769
	{ .name	    = "pwrite", .alias = "pwrite64", },
	{ .name	    = "readlinkat",
770
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* dfd */ }, }, },
771
	{ .name	    = "recvfrom",
772
	  .arg = { [3] = { .scnprintf = SCA_MSG_FLAGS, /* flags */ }, }, },
773
	{ .name	    = "recvmmsg",
774
	  .arg = { [3] = { .scnprintf = SCA_MSG_FLAGS, /* flags */ }, }, },
775
	{ .name	    = "recvmsg",
776
	  .arg = { [2] = { .scnprintf = SCA_MSG_FLAGS, /* flags */ }, }, },
777
	{ .name	    = "renameat",
778
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* dfd */ }, }, },
779
	{ .name	    = "rt_sigaction",
780
	  .arg = { [0] = { .scnprintf = SCA_SIGNUM, /* sig */ }, }, },
781
	{ .name	    = "rt_sigprocmask",
782
	  .arg = { [0] = STRARRAY(how, sighow), }, },
783
	{ .name	    = "rt_sigqueueinfo",
784
	  .arg = { [1] = { .scnprintf = SCA_SIGNUM, /* sig */ }, }, },
785
	{ .name	    = "rt_tgsigqueueinfo",
786
	  .arg = { [2] = { .scnprintf = SCA_SIGNUM, /* sig */ }, }, },
787
	{ .name	    = "sched_setscheduler",
788
	  .arg = { [1] = { .scnprintf = SCA_SCHED_POLICY, /* policy */ }, }, },
789
	{ .name	    = "seccomp",
790 791
	  .arg = { [0] = { .scnprintf = SCA_SECCOMP_OP,	   /* op */ },
		   [1] = { .scnprintf = SCA_SECCOMP_FLAGS, /* flags */ }, }, },
792 793
	{ .name	    = "select", .timeout = true, },
	{ .name	    = "sendmmsg",
794
	  .arg = { [3] = { .scnprintf = SCA_MSG_FLAGS, /* flags */ }, }, },
795
	{ .name	    = "sendmsg",
796
	  .arg = { [2] = { .scnprintf = SCA_MSG_FLAGS, /* flags */ }, }, },
797
	{ .name	    = "sendto",
798 799
	  .arg = { [3] = { .scnprintf = SCA_MSG_FLAGS, /* flags */ },
		   [4] = { .scnprintf = SCA_SOCKADDR, /* addr */ }, }, },
800
	{ .name	    = "set_tid_address", .errpid = true, },
801
	{ .name	    = "setitimer",
802
	  .arg = { [0] = STRARRAY(which, itimers), }, },
803
	{ .name	    = "setrlimit",
804
	  .arg = { [0] = STRARRAY(resource, rlimit_resources), }, },
805
	{ .name	    = "socket",
806
	  .arg = { [0] = STRARRAY(family, socket_families),
807 808
		   [1] = { .scnprintf = SCA_SK_TYPE, /* type */ },
		   [2] = { .scnprintf = SCA_SK_PROTO, /* protocol */ }, }, },
809
	{ .name	    = "socketpair",
810
	  .arg = { [0] = STRARRAY(family, socket_families),
811 812
		   [1] = { .scnprintf = SCA_SK_TYPE, /* type */ },
		   [2] = { .scnprintf = SCA_SK_PROTO, /* protocol */ }, }, },
813 814
	{ .name	    = "stat", .alias = "newstat", },
	{ .name	    = "statx",
815 816 817
	  .arg = { [0] = { .scnprintf = SCA_FDAT,	 /* fdat */ },
		   [2] = { .scnprintf = SCA_STATX_FLAGS, /* flags */ } ,
		   [3] = { .scnprintf = SCA_STATX_MASK,	 /* mask */ }, }, },
818
	{ .name	    = "swapoff",
819
	  .arg = { [0] = { .scnprintf = SCA_FILENAME, /* specialfile */ }, }, },
820
	{ .name	    = "swapon",
821
	  .arg = { [0] = { .scnprintf = SCA_FILENAME, /* specialfile */ }, }, },
822
	{ .name	    = "symlinkat",
823
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* dfd */ }, }, },
824
	{ .name	    = "tgkill",
825
	  .arg = { [2] = { .scnprintf = SCA_SIGNUM, /* sig */ }, }, },
826
	{ .name	    = "tkill",
827
	  .arg = { [1] = { .scnprintf = SCA_SIGNUM, /* sig */ }, }, },
828 829
	{ .name	    = "uname", .alias = "newuname", },
	{ .name	    = "unlinkat",
830
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* dfd */ }, }, },
831
	{ .name	    = "utimensat",
832
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* dirfd */ }, }, },
833
	{ .name	    = "wait4",	    .errpid = true,
834
	  .arg = { [2] = { .scnprintf = SCA_WAITID_OPTIONS, /* options */ }, }, },
835
	{ .name	    = "waitid",	    .errpid = true,
836
	  .arg = { [3] = { .scnprintf = SCA_WAITID_OPTIONS, /* options */ }, }, },
A
Arnaldo Carvalho de Melo 已提交
837 838 839 840 841 842 843 844 845 846 847 848 849 850
};

static int syscall_fmt__cmp(const void *name, const void *fmtp)
{
	const struct syscall_fmt *fmt = fmtp;
	return strcmp(name, fmt->name);
}

static struct syscall_fmt *syscall_fmt__find(const char *name)
{
	const int nmemb = ARRAY_SIZE(syscall_fmts);
	return bsearch(name, syscall_fmts, nmemb, sizeof(struct syscall_fmt), syscall_fmt__cmp);
}

851 852 853
/*
 * is_exit: is this "exit" or "exit_group"?
 * is_open: is this "open" or "openat"? To associate the fd returned in sys_exit with the pathname in sys_enter.
854
 * args_size: sum of the sizes of the syscall arguments, anything after that is augmented stuff: pathname for openat, etc.
855
 */
A
Arnaldo Carvalho de Melo 已提交
856 857
struct syscall {
	struct event_format *tp_format;
858
	int		    nr_args;
859
	int		    args_size;
860 861
	bool		    is_exit;
	bool		    is_open;
862
	struct format_field *args;
A
Arnaldo Carvalho de Melo 已提交
863 864
	const char	    *name;
	struct syscall_fmt  *fmt;
865
	struct syscall_arg_fmt *arg_fmt;
A
Arnaldo Carvalho de Melo 已提交
866 867
};

868 869 870 871 872 873 874 875
/*
 * We need to have this 'calculated' boolean because in some cases we really
 * don't know what is the duration of a syscall, for instance, when we start
 * a session and some threads are waiting for a syscall to finish, say 'poll',
 * in which case all we can do is to print "( ? ) for duration and for the
 * start timestamp.
 */
static size_t fprintf_duration(unsigned long t, bool calculated, FILE *fp)
876 877 878 879
{
	double duration = (double)t / NSEC_PER_MSEC;
	size_t printed = fprintf(fp, "(");

880
	if (!calculated)
881
		printed += fprintf(fp, "         ");
882
	else if (duration >= 1.0)
883 884 885 886 887
		printed += color_fprintf(fp, PERF_COLOR_RED, "%6.3f ms", duration);
	else if (duration >= 0.01)
		printed += color_fprintf(fp, PERF_COLOR_YELLOW, "%6.3f ms", duration);
	else
		printed += color_fprintf(fp, PERF_COLOR_NORMAL, "%6.3f ms", duration);
888
	return printed + fprintf(fp, "): ");
889 890
}

891 892 893 894
/**
 * filename.ptr: The filename char pointer that will be vfs_getname'd
 * filename.entry_str_pos: Where to insert the string translated from
 *                         filename.ptr by the vfs_getname tracepoint/kprobe.
895 896
 * ret_scnprintf: syscall args may set this to a different syscall return
 *                formatter, for instance, fcntl may return fds, file flags, etc.
897
 */
898 899 900
struct thread_trace {
	u64		  entry_time;
	bool		  entry_pending;
901
	unsigned long	  nr_events;
902
	unsigned long	  pfmaj, pfmin;
903
	char		  *entry_str;
904
	double		  runtime_ms;
905
	size_t		  (*ret_scnprintf)(char *bf, size_t size, struct syscall_arg *arg);
906 907
        struct {
		unsigned long ptr;
908 909 910 911
		short int     entry_str_pos;
		bool	      pending_open;
		unsigned int  namelen;
		char	      *name;
912
	} filename;
913 914 915 916
	struct {
		int	  max;
		char	  **table;
	} paths;
917 918

	struct intlist *syscall_stats;
919 920 921 922
};

static struct thread_trace *thread_trace__new(void)
{
923 924 925 926 927
	struct thread_trace *ttrace =  zalloc(sizeof(struct thread_trace));

	if (ttrace)
		ttrace->paths.max = -1;

928 929
	ttrace->syscall_stats = intlist__new(NULL);

930
	return ttrace;
931 932
}

933
static struct thread_trace *thread__trace(struct thread *thread, FILE *fp)
934
{
935 936
	struct thread_trace *ttrace;

937 938 939
	if (thread == NULL)
		goto fail;

940 941
	if (thread__priv(thread) == NULL)
		thread__set_priv(thread, thread_trace__new());
942

943
	if (thread__priv(thread) == NULL)
944 945
		goto fail;

946
	ttrace = thread__priv(thread);
947 948 949
	++ttrace->nr_events;

	return ttrace;
950
fail:
951
	color_fprintf(fp, PERF_COLOR_RED,
952 953 954 955
		      "WARNING: not enough memory, dropping samples!\n");
	return NULL;
}

956 957

void syscall_arg__set_ret_scnprintf(struct syscall_arg *arg,
958
				    size_t (*ret_scnprintf)(char *bf, size_t size, struct syscall_arg *arg))
959 960 961 962 963 964
{
	struct thread_trace *ttrace = thread__priv(arg->thread);

	ttrace->ret_scnprintf = ret_scnprintf;
}

965 966 967
#define TRACE_PFMAJ		(1 << 0)
#define TRACE_PFMIN		(1 << 1)

968 969
static const size_t trace__entry_str_size = 2048;

970
static int trace__set_fd_pathname(struct thread *thread, int fd, const char *pathname)
971
{
972
	struct thread_trace *ttrace = thread__priv(thread);
973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995

	if (fd > ttrace->paths.max) {
		char **npath = realloc(ttrace->paths.table, (fd + 1) * sizeof(char *));

		if (npath == NULL)
			return -1;

		if (ttrace->paths.max != -1) {
			memset(npath + ttrace->paths.max + 1, 0,
			       (fd - ttrace->paths.max) * sizeof(char *));
		} else {
			memset(npath, 0, (fd + 1) * sizeof(char *));
		}

		ttrace->paths.table = npath;
		ttrace->paths.max   = fd;
	}

	ttrace->paths.table[fd] = strdup(pathname);

	return ttrace->paths.table[fd] != NULL ? 0 : -1;
}

996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021
static int thread__read_fd_path(struct thread *thread, int fd)
{
	char linkname[PATH_MAX], pathname[PATH_MAX];
	struct stat st;
	int ret;

	if (thread->pid_ == thread->tid) {
		scnprintf(linkname, sizeof(linkname),
			  "/proc/%d/fd/%d", thread->pid_, fd);
	} else {
		scnprintf(linkname, sizeof(linkname),
			  "/proc/%d/task/%d/fd/%d", thread->pid_, thread->tid, fd);
	}

	if (lstat(linkname, &st) < 0 || st.st_size + 1 > (off_t)sizeof(pathname))
		return -1;

	ret = readlink(linkname, pathname, sizeof(pathname));

	if (ret < 0 || ret > st.st_size)
		return -1;

	pathname[ret] = '\0';
	return trace__set_fd_pathname(thread, fd, pathname);
}

1022 1023
static const char *thread__fd_path(struct thread *thread, int fd,
				   struct trace *trace)
1024
{
1025
	struct thread_trace *ttrace = thread__priv(thread);
1026 1027 1028 1029 1030 1031 1032

	if (ttrace == NULL)
		return NULL;

	if (fd < 0)
		return NULL;

1033
	if ((fd > ttrace->paths.max || ttrace->paths.table[fd] == NULL)) {
1034 1035 1036
		if (!trace->live)
			return NULL;
		++trace->stats.proc_getname;
1037
		if (thread__read_fd_path(thread, fd))
1038 1039
			return NULL;
	}
1040 1041 1042 1043

	return ttrace->paths.table[fd];
}

1044
size_t syscall_arg__scnprintf_fd(char *bf, size_t size, struct syscall_arg *arg)
1045 1046 1047
{
	int fd = arg->val;
	size_t printed = scnprintf(bf, size, "%d", fd);
1048
	const char *path = thread__fd_path(arg->thread, fd, arg->trace);
1049 1050 1051 1052 1053 1054 1055

	if (path)
		printed += scnprintf(bf + printed, size - printed, "<%s>", path);

	return printed;
}

1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072
size_t pid__scnprintf_fd(struct trace *trace, pid_t pid, int fd, char *bf, size_t size)
{
        size_t printed = scnprintf(bf, size, "%d", fd);
	struct thread *thread = machine__find_thread(trace->host, pid, pid);

	if (thread) {
		const char *path = thread__fd_path(thread, fd, trace);

		if (path)
			printed += scnprintf(bf + printed, size - printed, "<%s>", path);

		thread__put(thread);
	}

        return printed;
}

1073 1074 1075 1076 1077
static size_t syscall_arg__scnprintf_close_fd(char *bf, size_t size,
					      struct syscall_arg *arg)
{
	int fd = arg->val;
	size_t printed = syscall_arg__scnprintf_fd(bf, size, arg);
1078
	struct thread_trace *ttrace = thread__priv(arg->thread);
1079

1080 1081
	if (ttrace && fd >= 0 && fd <= ttrace->paths.max)
		zfree(&ttrace->paths.table[fd]);
1082 1083 1084 1085

	return printed;
}

1086 1087 1088 1089 1090 1091 1092 1093 1094
static void thread__set_filename_pos(struct thread *thread, const char *bf,
				     unsigned long ptr)
{
	struct thread_trace *ttrace = thread__priv(thread);

	ttrace->filename.ptr = ptr;
	ttrace->filename.entry_str_pos = bf - ttrace->entry_str;
}

1095 1096 1097 1098 1099 1100 1101
static size_t syscall_arg__scnprintf_augmented_string(struct syscall_arg *arg, char *bf, size_t size)
{
	struct augmented_arg *augmented_arg = arg->augmented.args;

	return scnprintf(bf, size, "%.*s", augmented_arg->size, augmented_arg->value);
}

1102 1103 1104 1105 1106
static size_t syscall_arg__scnprintf_filename(char *bf, size_t size,
					      struct syscall_arg *arg)
{
	unsigned long ptr = arg->val;

1107 1108 1109
	if (arg->augmented.args)
		return syscall_arg__scnprintf_augmented_string(arg, bf, size);

1110 1111 1112 1113 1114 1115 1116
	if (!arg->trace->vfs_getname)
		return scnprintf(bf, size, "%#x", ptr);

	thread__set_filename_pos(arg->thread, bf, ptr);
	return 0;
}

1117 1118 1119 1120 1121
static bool trace__filter_duration(struct trace *trace, double t)
{
	return t < (trace->duration_filter * NSEC_PER_MSEC);
}

1122
static size_t __trace__fprintf_tstamp(struct trace *trace, u64 tstamp, FILE *fp)
1123 1124 1125
{
	double ts = (double)(tstamp - trace->base_time) / NSEC_PER_MSEC;

1126
	return fprintf(fp, "%10.3f ", ts);
1127 1128
}

1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142
/*
 * We're handling tstamp=0 as an undefined tstamp, i.e. like when we are
 * using ttrace->entry_time for a thread that receives a sys_exit without
 * first having received a sys_enter ("poll" issued before tracing session
 * starts, lost sys_enter exit due to ring buffer overflow).
 */
static size_t trace__fprintf_tstamp(struct trace *trace, u64 tstamp, FILE *fp)
{
	if (tstamp > 0)
		return __trace__fprintf_tstamp(trace, tstamp, fp);

	return fprintf(fp, "         ? ");
}

1143
static bool done = false;
1144
static bool interrupted = false;
1145

1146
static void sig_handler(int sig)
1147 1148
{
	done = true;
1149
	interrupted = sig == SIGINT;
1150 1151
}

1152
static size_t trace__fprintf_comm_tid(struct trace *trace, struct thread *thread, FILE *fp)
1153
{
1154
	size_t printed = 0;
1155

1156 1157
	if (trace->multiple_threads) {
		if (trace->show_comm)
1158
			printed += fprintf(fp, "%.14s/", thread__comm_str(thread));
1159
		printed += fprintf(fp, "%d ", thread->tid);
1160
	}
1161 1162 1163 1164

	return printed;
}

1165 1166 1167 1168 1169 1170 1171 1172
static size_t trace__fprintf_entry_head(struct trace *trace, struct thread *thread,
					u64 duration, bool duration_calculated, u64 tstamp, FILE *fp)
{
	size_t printed = trace__fprintf_tstamp(trace, tstamp, fp);
	printed += fprintf_duration(duration, duration_calculated, fp);
	return printed + trace__fprintf_comm_tid(trace, thread, fp);
}

1173
static int trace__process_event(struct trace *trace, struct machine *machine,
1174
				union perf_event *event, struct perf_sample *sample)
1175 1176 1177 1178 1179
{
	int ret = 0;

	switch (event->header.type) {
	case PERF_RECORD_LOST:
1180
		color_fprintf(trace->output, PERF_COLOR_RED,
1181
			      "LOST %" PRIu64 " events!\n", event->lost.lost);
1182
		ret = machine__process_lost_event(machine, event, sample);
1183
		break;
1184
	default:
1185
		ret = machine__process_event(machine, event, sample);
1186 1187 1188 1189 1190 1191
		break;
	}

	return ret;
}

1192
static int trace__tool_process(struct perf_tool *tool,
1193
			       union perf_event *event,
1194
			       struct perf_sample *sample,
1195 1196
			       struct machine *machine)
{
1197
	struct trace *trace = container_of(tool, struct trace, tool);
1198
	return trace__process_event(trace, machine, event, sample);
1199 1200
}

1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218
static char *trace__machine__resolve_kernel_addr(void *vmachine, unsigned long long *addrp, char **modp)
{
	struct machine *machine = vmachine;

	if (machine->kptr_restrict_warned)
		return NULL;

	if (symbol_conf.kptr_restrict) {
		pr_warning("Kernel address maps (/proc/{kallsyms,modules}) are restricted.\n\n"
			   "Check /proc/sys/kernel/kptr_restrict.\n\n"
			   "Kernel samples will not be resolved.\n");
		machine->kptr_restrict_warned = true;
		return NULL;
	}

	return machine__resolve_kernel_addr(vmachine, addrp, modp);
}

1219 1220
static int trace__symbols_init(struct trace *trace, struct perf_evlist *evlist)
{
1221
	int err = symbol__init(NULL);
1222 1223 1224 1225

	if (err)
		return err;

1226 1227 1228
	trace->host = machine__new_host();
	if (trace->host == NULL)
		return -ENOMEM;
1229

1230 1231 1232
	err = trace_event__register_resolver(trace->host, trace__machine__resolve_kernel_addr);
	if (err < 0)
		goto out;
1233

1234
	err = __machine__synthesize_threads(trace->host, &trace->tool, &trace->opts.target,
1235
					    evlist->threads, trace__tool_process, false,
1236
					    trace->opts.proc_map_timeout, 1);
1237
out:
1238 1239 1240 1241 1242 1243
	if (err)
		symbol__exit();

	return err;
}

1244 1245 1246 1247 1248 1249 1250 1251
static void trace__symbols__exit(struct trace *trace)
{
	machine__exit(trace->host);
	trace->host = NULL;

	symbol__exit();
}

1252
static int syscall__alloc_arg_fmts(struct syscall *sc, int nr_args)
1253
{
1254
	int idx;
1255

1256 1257 1258
	if (nr_args == 6 && sc->fmt && sc->fmt->nr_args != 0)
		nr_args = sc->fmt->nr_args;

1259
	sc->arg_fmt = calloc(nr_args, sizeof(*sc->arg_fmt));
1260
	if (sc->arg_fmt == NULL)
1261 1262
		return -1;

1263 1264
	for (idx = 0; idx < nr_args; ++idx) {
		if (sc->fmt)
1265
			sc->arg_fmt[idx] = sc->fmt->arg[idx];
1266
	}
1267

1268 1269 1270 1271 1272 1273
	sc->nr_args = nr_args;
	return 0;
}

static int syscall__set_arg_fmts(struct syscall *sc)
{
1274
	struct format_field *field, *last_field = NULL;
1275 1276 1277
	int idx = 0, len;

	for (field = sc->args; field; field = field->next, ++idx) {
1278 1279
		last_field = field;

1280 1281
		if (sc->fmt && sc->fmt->arg[idx].scnprintf)
			continue;
1282

1283
		if (strcmp(field->type, "const char *") == 0 &&
1284 1285 1286
			 (strcmp(field->name, "filename") == 0 ||
			  strcmp(field->name, "path") == 0 ||
			  strcmp(field->name, "pathname") == 0))
1287
			sc->arg_fmt[idx].scnprintf = SCA_FILENAME;
1288
		else if (field->flags & FIELD_IS_POINTER)
1289
			sc->arg_fmt[idx].scnprintf = syscall_arg__scnprintf_hex;
1290
		else if (strcmp(field->type, "pid_t") == 0)
1291
			sc->arg_fmt[idx].scnprintf = SCA_PID;
1292
		else if (strcmp(field->type, "umode_t") == 0)
1293
			sc->arg_fmt[idx].scnprintf = SCA_MODE_T;
1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305
		else if ((strcmp(field->type, "int") == 0 ||
			  strcmp(field->type, "unsigned int") == 0 ||
			  strcmp(field->type, "long") == 0) &&
			 (len = strlen(field->name)) >= 2 &&
			 strcmp(field->name + len - 2, "fd") == 0) {
			/*
			 * /sys/kernel/tracing/events/syscalls/sys_enter*
			 * egrep 'field:.*fd;' .../format|sed -r 's/.*field:([a-z ]+) [a-z_]*fd.+/\1/g'|sort|uniq -c
			 * 65 int
			 * 23 unsigned int
			 * 7 unsigned long
			 */
1306
			sc->arg_fmt[idx].scnprintf = SCA_FD;
1307
		}
1308 1309
	}

1310 1311 1312
	if (last_field)
		sc->args_size = last_field->offset + last_field->size;

1313 1314 1315
	return 0;
}

A
Arnaldo Carvalho de Melo 已提交
1316 1317 1318 1319
static int trace__read_syscall_info(struct trace *trace, int id)
{
	char tp_name[128];
	struct syscall *sc;
1320
	const char *name = syscalltbl__name(trace->sctbl, id);
1321 1322 1323

	if (name == NULL)
		return -1;
A
Arnaldo Carvalho de Melo 已提交
1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342

	if (id > trace->syscalls.max) {
		struct syscall *nsyscalls = realloc(trace->syscalls.table, (id + 1) * sizeof(*sc));

		if (nsyscalls == NULL)
			return -1;

		if (trace->syscalls.max != -1) {
			memset(nsyscalls + trace->syscalls.max + 1, 0,
			       (id - trace->syscalls.max) * sizeof(*sc));
		} else {
			memset(nsyscalls, 0, (id + 1) * sizeof(*sc));
		}

		trace->syscalls.table = nsyscalls;
		trace->syscalls.max   = id;
	}

	sc = trace->syscalls.table + id;
1343
	sc->name = name;
1344

1345
	sc->fmt  = syscall_fmt__find(sc->name);
A
Arnaldo Carvalho de Melo 已提交
1346

1347
	snprintf(tp_name, sizeof(tp_name), "sys_enter_%s", sc->name);
1348
	sc->tp_format = trace_event__tp_format("syscalls", tp_name);
1349

1350
	if (IS_ERR(sc->tp_format) && sc->fmt && sc->fmt->alias) {
1351
		snprintf(tp_name, sizeof(tp_name), "sys_enter_%s", sc->fmt->alias);
1352
		sc->tp_format = trace_event__tp_format("syscalls", tp_name);
1353
	}
A
Arnaldo Carvalho de Melo 已提交
1354

1355 1356 1357
	if (syscall__alloc_arg_fmts(sc, IS_ERR(sc->tp_format) ? 6 : sc->tp_format->format.nr_fields))
		return -1;

1358
	if (IS_ERR(sc->tp_format))
1359 1360
		return -1;

1361
	sc->args = sc->tp_format->format.fields;
1362 1363 1364 1365 1366 1367
	/*
	 * We need to check and discard the first variable '__syscall_nr'
	 * or 'nr' that mean the syscall number. It is needless here.
	 * So drop '__syscall_nr' or 'nr' field but does not exist on older kernels.
	 */
	if (sc->args && (!strcmp(sc->args->name, "__syscall_nr") || !strcmp(sc->args->name, "nr"))) {
1368 1369 1370 1371
		sc->args = sc->args->next;
		--sc->nr_args;
	}

1372
	sc->is_exit = !strcmp(name, "exit_group") || !strcmp(name, "exit");
1373
	sc->is_open = !strcmp(name, "open") || !strcmp(name, "openat");
1374

1375
	return syscall__set_arg_fmts(sc);
A
Arnaldo Carvalho de Melo 已提交
1376 1377
}

1378 1379
static int trace__validate_ev_qualifier(struct trace *trace)
{
1380
	int err = 0, i;
1381
	size_t nr_allocated;
1382 1383
	struct str_node *pos;

1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394
	trace->ev_qualifier_ids.nr = strlist__nr_entries(trace->ev_qualifier);
	trace->ev_qualifier_ids.entries = malloc(trace->ev_qualifier_ids.nr *
						 sizeof(trace->ev_qualifier_ids.entries[0]));

	if (trace->ev_qualifier_ids.entries == NULL) {
		fputs("Error:\tNot enough memory for allocating events qualifier ids\n",
		       trace->output);
		err = -EINVAL;
		goto out;
	}

1395
	nr_allocated = trace->ev_qualifier_ids.nr;
1396 1397
	i = 0;

1398
	strlist__for_each_entry(pos, trace->ev_qualifier) {
1399
		const char *sc = pos->s;
1400
		int id = syscalltbl__id(trace->sctbl, sc), match_next = -1;
1401

1402
		if (id < 0) {
1403 1404 1405 1406
			id = syscalltbl__strglobmatch_first(trace->sctbl, sc, &match_next);
			if (id >= 0)
				goto matches;

1407 1408 1409 1410 1411 1412 1413 1414 1415
			if (err == 0) {
				fputs("Error:\tInvalid syscall ", trace->output);
				err = -EINVAL;
			} else {
				fputs(", ", trace->output);
			}

			fputs(sc, trace->output);
		}
1416
matches:
1417
		trace->ev_qualifier_ids.entries[i++] = id;
1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440
		if (match_next == -1)
			continue;

		while (1) {
			id = syscalltbl__strglobmatch_next(trace->sctbl, sc, &match_next);
			if (id < 0)
				break;
			if (nr_allocated == trace->ev_qualifier_ids.nr) {
				void *entries;

				nr_allocated += 8;
				entries = realloc(trace->ev_qualifier_ids.entries,
						  nr_allocated * sizeof(trace->ev_qualifier_ids.entries[0]));
				if (entries == NULL) {
					err = -ENOMEM;
					fputs("\nError:\t Not enough memory for parsing\n", trace->output);
					goto out_free;
				}
				trace->ev_qualifier_ids.entries = entries;
			}
			trace->ev_qualifier_ids.nr++;
			trace->ev_qualifier_ids.entries[i++] = id;
		}
1441 1442 1443 1444 1445
	}

	if (err < 0) {
		fputs("\nHint:\ttry 'perf list syscalls:sys_enter_*'"
		      "\nHint:\tand: 'man syscalls'\n", trace->output);
1446
out_free:
1447 1448
		zfree(&trace->ev_qualifier_ids.entries);
		trace->ev_qualifier_ids.nr = 0;
1449
	}
1450
out:
1451 1452 1453
	return err;
}

1454 1455 1456 1457 1458 1459 1460 1461
/*
 * args is to be interpreted as a series of longs but we need to handle
 * 8-byte unaligned accesses. args points to raw_data within the event
 * and raw_data is guaranteed to be 8-byte unaligned because it is
 * preceded by raw_size which is a u32. So we need to copy args to a temp
 * variable to read it. Most notably this avoids extended load instructions
 * on unaligned addresses
 */
1462
unsigned long syscall_arg__val(struct syscall_arg *arg, u8 idx)
1463 1464
{
	unsigned long val;
1465
	unsigned char *p = arg->args + sizeof(unsigned long) * idx;
1466 1467 1468 1469 1470

	memcpy(&val, p, sizeof(val));
	return val;
}

1471 1472 1473 1474 1475 1476 1477 1478 1479
static size_t syscall__scnprintf_name(struct syscall *sc, char *bf, size_t size,
				      struct syscall_arg *arg)
{
	if (sc->arg_fmt && sc->arg_fmt[arg->idx].name)
		return scnprintf(bf, size, "%s: ", sc->arg_fmt[arg->idx].name);

	return scnprintf(bf, size, "arg%d: ", arg->idx);
}

1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491
static size_t syscall__scnprintf_val(struct syscall *sc, char *bf, size_t size,
				     struct syscall_arg *arg, unsigned long val)
{
	if (sc->arg_fmt && sc->arg_fmt[arg->idx].scnprintf) {
		arg->val = val;
		if (sc->arg_fmt[arg->idx].parm)
			arg->parm = sc->arg_fmt[arg->idx].parm;
		return sc->arg_fmt[arg->idx].scnprintf(bf, size, arg);
	}
	return scnprintf(bf, size, "%ld", val);
}

1492
static size_t syscall__scnprintf_args(struct syscall *sc, char *bf, size_t size,
1493 1494
				      unsigned char *args, void *augmented_args, int augmented_args_size,
				      struct trace *trace, struct thread *thread)
A
Arnaldo Carvalho de Melo 已提交
1495 1496
{
	size_t printed = 0;
1497
	unsigned long val;
1498 1499 1500
	u8 bit = 1;
	struct syscall_arg arg = {
		.args	= args,
1501 1502 1503 1504
		.augmented = {
			.size = augmented_args_size,
			.args = augmented_args,
		},
1505 1506 1507 1508 1509
		.idx	= 0,
		.mask	= 0,
		.trace  = trace,
		.thread = thread,
	};
1510 1511 1512 1513 1514 1515 1516 1517
	struct thread_trace *ttrace = thread__priv(thread);

	/*
	 * Things like fcntl will set this in its 'cmd' formatter to pick the
	 * right formatter for the return value (an fd? file flags?), which is
	 * not needed for syscalls that always return a given type, say an fd.
	 */
	ttrace->ret_scnprintf = NULL;
A
Arnaldo Carvalho de Melo 已提交
1518

1519
	if (sc->args != NULL) {
A
Arnaldo Carvalho de Melo 已提交
1520
		struct format_field *field;
1521

1522
		for (field = sc->args; field;
1523 1524
		     field = field->next, ++arg.idx, bit <<= 1) {
			if (arg.mask & bit)
1525
				continue;
1526

1527
			val = syscall_arg__val(&arg, arg.idx);
1528

1529 1530 1531 1532 1533
			/*
 			 * Suppress this argument if its value is zero and
 			 * and we don't have a string associated in an
 			 * strarray for it.
 			 */
1534
			if (val == 0 &&
1535
			    !(sc->arg_fmt &&
1536 1537
			      (sc->arg_fmt[arg.idx].show_zero ||
			       sc->arg_fmt[arg.idx].scnprintf == SCA_STRARRAY ||
1538 1539
			       sc->arg_fmt[arg.idx].scnprintf == SCA_STRARRAYS) &&
			      sc->arg_fmt[arg.idx].parm))
1540 1541
				continue;

1542
			printed += scnprintf(bf + printed, size - printed,
1543
					     "%s%s: ", printed ? ", " : "", field->name);
1544
			printed += syscall__scnprintf_val(sc, bf + printed, size - printed, &arg, val);
A
Arnaldo Carvalho de Melo 已提交
1545
		}
1546 1547 1548 1549 1550 1551
	} else if (IS_ERR(sc->tp_format)) {
		/*
		 * If we managed to read the tracepoint /format file, then we
		 * may end up not having any args, like with gettid(), so only
		 * print the raw args when we didn't manage to read it.
		 */
1552
		while (arg.idx < sc->nr_args) {
1553 1554 1555
			if (arg.mask & bit)
				goto next_arg;
			val = syscall_arg__val(&arg, arg.idx);
1556 1557 1558
			if (printed)
				printed += scnprintf(bf + printed, size - printed, ", ");
			printed += syscall__scnprintf_name(sc, bf + printed, size - printed, &arg);
1559 1560 1561 1562
			printed += syscall__scnprintf_val(sc, bf + printed, size - printed, &arg, val);
next_arg:
			++arg.idx;
			bit <<= 1;
A
Arnaldo Carvalho de Melo 已提交
1563 1564 1565 1566 1567 1568
		}
	}

	return printed;
}

1569
typedef int (*tracepoint_handler)(struct trace *trace, struct perf_evsel *evsel,
1570
				  union perf_event *event,
1571 1572 1573
				  struct perf_sample *sample);

static struct syscall *trace__syscall_info(struct trace *trace,
1574
					   struct perf_evsel *evsel, int id)
1575 1576 1577
{

	if (id < 0) {
1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593

		/*
		 * XXX: Noticed on x86_64, reproduced as far back as 3.0.36, haven't tried
		 * before that, leaving at a higher verbosity level till that is
		 * explained. Reproduced with plain ftrace with:
		 *
		 * echo 1 > /t/events/raw_syscalls/sys_exit/enable
		 * grep "NR -1 " /t/trace_pipe
		 *
		 * After generating some load on the machine.
 		 */
		if (verbose > 1) {
			static u64 n;
			fprintf(trace->output, "Invalid syscall %d id, skipping (%s, %" PRIu64 ") ...\n",
				id, perf_evsel__name(evsel), ++n);
		}
1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606
		return NULL;
	}

	if ((id > trace->syscalls.max || trace->syscalls.table[id].name == NULL) &&
	    trace__read_syscall_info(trace, id))
		goto out_cant_read;

	if ((id > trace->syscalls.max || trace->syscalls.table[id].name == NULL))
		goto out_cant_read;

	return &trace->syscalls.table[id];

out_cant_read:
1607
	if (verbose > 0) {
1608 1609 1610 1611 1612
		fprintf(trace->output, "Problems reading syscall %d", id);
		if (id <= trace->syscalls.max && trace->syscalls.table[id].name != NULL)
			fprintf(trace->output, "(%s)", trace->syscalls.table[id].name);
		fputs(" information\n", trace->output);
	}
1613 1614 1615
	return NULL;
}

1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641
static void thread__update_stats(struct thread_trace *ttrace,
				 int id, struct perf_sample *sample)
{
	struct int_node *inode;
	struct stats *stats;
	u64 duration = 0;

	inode = intlist__findnew(ttrace->syscall_stats, id);
	if (inode == NULL)
		return;

	stats = inode->priv;
	if (stats == NULL) {
		stats = malloc(sizeof(struct stats));
		if (stats == NULL)
			return;
		init_stats(stats);
		inode->priv = stats;
	}

	if (ttrace->entry_time && sample->time > ttrace->entry_time)
		duration = sample->time - ttrace->entry_time;

	update_stats(stats, duration);
}

1642
static int trace__printf_interrupted_entry(struct trace *trace)
1643 1644 1645 1646
{
	struct thread_trace *ttrace;
	size_t printed;

1647
	if (trace->failure_only || trace->current == NULL)
1648 1649 1650 1651 1652 1653 1654
		return 0;

	ttrace = thread__priv(trace->current);

	if (!ttrace->entry_pending)
		return 0;

1655
	printed  = trace__fprintf_entry_head(trace, trace->current, 0, false, ttrace->entry_time, trace->output);
1656 1657 1658 1659 1660 1661
	printed += fprintf(trace->output, "%-70s) ...\n", ttrace->entry_str);
	ttrace->entry_pending = false;

	return printed;
}

1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678
static int trace__fprintf_sample(struct trace *trace, struct perf_evsel *evsel,
				 struct perf_sample *sample, struct thread *thread)
{
	int printed = 0;

	if (trace->print_sample) {
		double ts = (double)sample->time / NSEC_PER_MSEC;

		printed += fprintf(trace->output, "%22s %10.3f %s %d/%d [%d]\n",
				   perf_evsel__name(evsel), ts,
				   thread__comm_str(thread),
				   sample->pid, sample->tid, sample->cpu);
	}

	return printed;
}

1679
static int trace__sys_enter(struct trace *trace, struct perf_evsel *evsel,
1680
			    union perf_event *event __maybe_unused,
1681 1682
			    struct perf_sample *sample)
{
1683
	char *msg;
1684
	void *args;
1685
	size_t printed = 0;
1686
	struct thread *thread;
1687
	int id = perf_evsel__sc_tp_uint(evsel, id, sample), err = -1;
1688
	struct syscall *sc = trace__syscall_info(trace, evsel, id);
1689 1690 1691 1692
	struct thread_trace *ttrace;

	if (sc == NULL)
		return -1;
1693

1694
	thread = machine__findnew_thread(trace->host, sample->pid, sample->tid);
1695
	ttrace = thread__trace(thread, trace->output);
1696
	if (ttrace == NULL)
1697
		goto out_put;
1698

1699 1700
	trace__fprintf_sample(trace, evsel, sample, thread);

1701
	args = perf_evsel__sc_tp_ptr(evsel, args, sample);
1702 1703

	if (ttrace->entry_str == NULL) {
1704
		ttrace->entry_str = malloc(trace__entry_str_size);
1705
		if (!ttrace->entry_str)
1706
			goto out_put;
1707 1708
	}

1709
	if (!(trace->duration_filter || trace->summary_only || trace->min_stack))
1710
		trace__printf_interrupted_entry(trace);
1711

1712 1713
	ttrace->entry_time = sample->time;
	msg = ttrace->entry_str;
1714
	printed += scnprintf(msg + printed, trace__entry_str_size - printed, "%s(", sc->name);
1715

1716
	printed += syscall__scnprintf_args(sc, msg + printed, trace__entry_str_size - printed,
1717
					   args, NULL, 0, trace, thread);
1718

1719
	if (sc->is_exit) {
1720
		if (!(trace->duration_filter || trace->summary_only || trace->failure_only || trace->min_stack)) {
1721
			trace__fprintf_entry_head(trace, thread, 0, false, ttrace->entry_time, trace->output);
1722
			fprintf(trace->output, "%-70s)\n", ttrace->entry_str);
1723
		}
1724
	} else {
1725
		ttrace->entry_pending = true;
1726 1727 1728
		/* See trace__vfs_getname & trace__sys_exit */
		ttrace->filename.pending_open = false;
	}
1729

1730 1731 1732 1733
	if (trace->current != thread) {
		thread__put(trace->current);
		trace->current = thread__get(thread);
	}
1734 1735 1736 1737
	err = 0;
out_put:
	thread__put(thread);
	return err;
1738 1739
}

1740 1741 1742 1743 1744
static int trace__fprintf_sys_enter(struct trace *trace, struct perf_evsel *evsel,
				    struct perf_sample *sample)
{
	struct thread_trace *ttrace;
	struct thread *thread;
1745 1746
	int id = perf_evsel__sc_tp_uint(evsel, id, sample), err = -1;
	struct syscall *sc = trace__syscall_info(trace, evsel, id);
1747
	char msg[1024];
1748 1749
	void *args, *augmented_args = NULL;
	int augmented_args_size;
1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762

	if (sc == NULL)
		return -1;

	thread = machine__findnew_thread(trace->host, sample->pid, sample->tid);
	ttrace = thread__trace(thread, trace->output);
	/*
	 * We need to get ttrace just to make sure it is there when syscall__scnprintf_args()
	 * and the rest of the beautifiers accessing it via struct syscall_arg touches it.
	 */
	if (ttrace == NULL)
		goto out_put;

1763
	args = perf_evsel__sc_tp_ptr(evsel, args, sample);
1764 1765 1766 1767 1768
	augmented_args_size = sample->raw_size - sc->args_size;
	if (augmented_args_size > 0)
		augmented_args = sample->raw_data + sc->args_size;

	syscall__scnprintf_args(sc, msg, sizeof(msg), args, augmented_args, augmented_args_size, trace, thread);
1769 1770 1771 1772 1773 1774 1775
	fprintf(trace->output, "%s", msg);
	err = 0;
out_put:
	thread__put(thread);
	return err;
}

1776 1777 1778
static int trace__resolve_callchain(struct trace *trace, struct perf_evsel *evsel,
				    struct perf_sample *sample,
				    struct callchain_cursor *cursor)
1779 1780
{
	struct addr_location al;
1781 1782 1783
	int max_stack = evsel->attr.sample_max_stack ?
			evsel->attr.sample_max_stack :
			trace->max_stack;
1784 1785

	if (machine__resolve(trace->host, &al, sample) < 0 ||
1786
	    thread__resolve_callchain(al.thread, cursor, evsel, sample, NULL, NULL, max_stack))
1787 1788 1789 1790 1791 1792 1793
		return -1;

	return 0;
}

static int trace__fprintf_callchain(struct trace *trace, struct perf_sample *sample)
{
1794
	/* TODO: user-configurable print_opts */
1795 1796 1797
	const unsigned int print_opts = EVSEL__PRINT_SYM |
				        EVSEL__PRINT_DSO |
				        EVSEL__PRINT_UNKNOWN_AS_ADDR;
1798

1799
	return sample__fprintf_callchain(sample, 38, print_opts, &callchain_cursor, trace->output);
1800 1801
}

1802 1803 1804 1805 1806 1807 1808 1809
static const char *errno_to_name(struct perf_evsel *evsel, int err)
{
	struct perf_env *env = perf_evsel__env(evsel);
	const char *arch_name = perf_env__arch(env);

	return arch_syscalls__strerrno(arch_name, err);
}

1810
static int trace__sys_exit(struct trace *trace, struct perf_evsel *evsel,
1811
			   union perf_event *event __maybe_unused,
1812 1813
			   struct perf_sample *sample)
{
1814
	long ret;
1815
	u64 duration = 0;
1816
	bool duration_calculated = false;
1817
	struct thread *thread;
1818
	int id = perf_evsel__sc_tp_uint(evsel, id, sample), err = -1, callchain_ret = 0;
1819
	struct syscall *sc = trace__syscall_info(trace, evsel, id);
1820 1821 1822 1823
	struct thread_trace *ttrace;

	if (sc == NULL)
		return -1;
1824

1825
	thread = machine__findnew_thread(trace->host, sample->pid, sample->tid);
1826
	ttrace = thread__trace(thread, trace->output);
1827
	if (ttrace == NULL)
1828
		goto out_put;
1829

1830 1831
	trace__fprintf_sample(trace, evsel, sample, thread);

1832 1833 1834
	if (trace->summary)
		thread__update_stats(ttrace, id, sample);

1835
	ret = perf_evsel__sc_tp_uint(evsel, ret, sample);
1836

1837
	if (sc->is_open && ret >= 0 && ttrace->filename.pending_open) {
1838 1839
		trace__set_fd_pathname(thread, ret, ttrace->filename.name);
		ttrace->filename.pending_open = false;
1840 1841 1842
		++trace->stats.vfs_getname;
	}

1843
	if (ttrace->entry_time) {
1844
		duration = sample->time - ttrace->entry_time;
1845 1846
		if (trace__filter_duration(trace, duration))
			goto out;
1847
		duration_calculated = true;
1848 1849
	} else if (trace->duration_filter)
		goto out;
1850

1851 1852 1853 1854 1855 1856 1857 1858 1859
	if (sample->callchain) {
		callchain_ret = trace__resolve_callchain(trace, evsel, sample, &callchain_cursor);
		if (callchain_ret == 0) {
			if (callchain_cursor.nr < trace->min_stack)
				goto out;
			callchain_ret = 1;
		}
	}

1860
	if (trace->summary_only || (ret >= 0 && trace->failure_only))
D
David Ahern 已提交
1861 1862
		goto out;

1863
	trace__fprintf_entry_head(trace, thread, duration, duration_calculated, ttrace->entry_time, trace->output);
1864 1865

	if (ttrace->entry_pending) {
1866
		fprintf(trace->output, "%-70s", ttrace->entry_str);
1867
	} else {
1868 1869 1870
		fprintf(trace->output, " ... [");
		color_fprintf(trace->output, PERF_COLOR_YELLOW, "continued");
		fprintf(trace->output, "]: %s()", sc->name);
1871 1872
	}

1873
	if (sc->fmt == NULL) {
1874 1875
		if (ret < 0)
			goto errno_print;
1876
signed_print:
1877
		fprintf(trace->output, ") = %ld", ret);
1878 1879
	} else if (ret < 0) {
errno_print: {
1880
		char bf[STRERR_BUFSIZE];
1881
		const char *emsg = str_error_r(-ret, bf, sizeof(bf)),
1882
			   *e = errno_to_name(evsel, -ret);
1883

1884
		fprintf(trace->output, ") = -1 %s %s", e, emsg);
1885
	}
1886
	} else if (ret == 0 && sc->fmt->timeout)
1887
		fprintf(trace->output, ") = 0 Timeout");
1888 1889
	else if (ttrace->ret_scnprintf) {
		char bf[1024];
1890 1891 1892 1893 1894 1895
		struct syscall_arg arg = {
			.val	= ret,
			.thread	= thread,
			.trace	= trace,
		};
		ttrace->ret_scnprintf(bf, sizeof(bf), &arg);
1896 1897 1898
		ttrace->ret_scnprintf = NULL;
		fprintf(trace->output, ") = %s", bf);
	} else if (sc->fmt->hexret)
1899
		fprintf(trace->output, ") = %#lx", ret);
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909
	else if (sc->fmt->errpid) {
		struct thread *child = machine__find_thread(trace->host, ret, ret);

		if (child != NULL) {
			fprintf(trace->output, ") = %ld", ret);
			if (child->comm_set)
				fprintf(trace->output, " (%s)", thread__comm_str(child));
			thread__put(child);
		}
	} else
1910
		goto signed_print;
1911

1912
	fputc('\n', trace->output);
1913

1914 1915 1916 1917
	if (callchain_ret > 0)
		trace__fprintf_callchain(trace, sample);
	else if (callchain_ret < 0)
		pr_err("Problem processing %s callchain, skipping...\n", perf_evsel__name(evsel));
1918
out:
1919
	ttrace->entry_pending = false;
1920 1921 1922 1923
	err = 0;
out_put:
	thread__put(thread);
	return err;
1924 1925
}

1926
static int trace__vfs_getname(struct trace *trace, struct perf_evsel *evsel,
1927
			      union perf_event *event __maybe_unused,
1928 1929
			      struct perf_sample *sample)
{
1930 1931 1932 1933 1934
	struct thread *thread = machine__findnew_thread(trace->host, sample->pid, sample->tid);
	struct thread_trace *ttrace;
	size_t filename_len, entry_str_len, to_move;
	ssize_t remaining_space;
	char *pos;
1935
	const char *filename = perf_evsel__rawptr(evsel, sample, "pathname");
1936 1937 1938 1939 1940 1941

	if (!thread)
		goto out;

	ttrace = thread__priv(thread);
	if (!ttrace)
1942
		goto out_put;
1943

1944
	filename_len = strlen(filename);
1945
	if (filename_len == 0)
1946
		goto out_put;
1947 1948 1949 1950 1951

	if (ttrace->filename.namelen < filename_len) {
		char *f = realloc(ttrace->filename.name, filename_len + 1);

		if (f == NULL)
1952
			goto out_put;
1953 1954 1955 1956 1957 1958 1959 1960

		ttrace->filename.namelen = filename_len;
		ttrace->filename.name = f;
	}

	strcpy(ttrace->filename.name, filename);
	ttrace->filename.pending_open = true;

1961
	if (!ttrace->filename.ptr)
1962
		goto out_put;
1963 1964 1965 1966

	entry_str_len = strlen(ttrace->entry_str);
	remaining_space = trace__entry_str_size - entry_str_len - 1; /* \0 */
	if (remaining_space <= 0)
1967
		goto out_put;
1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980

	if (filename_len > (size_t)remaining_space) {
		filename += filename_len - remaining_space;
		filename_len = remaining_space;
	}

	to_move = entry_str_len - ttrace->filename.entry_str_pos + 1; /* \0 */
	pos = ttrace->entry_str + ttrace->filename.entry_str_pos;
	memmove(pos + filename_len, pos, to_move);
	memcpy(pos, filename, filename_len);

	ttrace->filename.ptr = 0;
	ttrace->filename.entry_str_pos = 0;
1981 1982
out_put:
	thread__put(thread);
1983
out:
1984 1985 1986
	return 0;
}

1987
static int trace__sched_stat_runtime(struct trace *trace, struct perf_evsel *evsel,
1988
				     union perf_event *event __maybe_unused,
1989 1990 1991 1992
				     struct perf_sample *sample)
{
        u64 runtime = perf_evsel__intval(evsel, sample, "runtime");
	double runtime_ms = (double)runtime / NSEC_PER_MSEC;
1993
	struct thread *thread = machine__findnew_thread(trace->host,
1994 1995
							sample->pid,
							sample->tid);
1996
	struct thread_trace *ttrace = thread__trace(thread, trace->output);
1997 1998 1999 2000 2001 2002

	if (ttrace == NULL)
		goto out_dump;

	ttrace->runtime_ms += runtime_ms;
	trace->runtime_ms += runtime_ms;
2003
out_put:
2004
	thread__put(thread);
2005 2006 2007
	return 0;

out_dump:
2008
	fprintf(trace->output, "%s: comm=%s,pid=%u,runtime=%" PRIu64 ",vruntime=%" PRIu64 ")\n",
2009 2010 2011 2012 2013
	       evsel->name,
	       perf_evsel__strval(evsel, sample, "comm"),
	       (pid_t)perf_evsel__intval(evsel, sample, "pid"),
	       runtime,
	       perf_evsel__intval(evsel, sample, "vruntime"));
2014
	goto out_put;
2015 2016
}

2017 2018
static int bpf_output__printer(enum binary_printer_ops op,
			       unsigned int val, void *extra __maybe_unused, FILE *fp)
2019 2020 2021 2022 2023
{
	unsigned char ch = (unsigned char)val;

	switch (op) {
	case BINARY_PRINT_CHAR_DATA:
2024
		return fprintf(fp, "%c", isprint(ch) ? ch : '.');
2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036
	case BINARY_PRINT_DATA_BEGIN:
	case BINARY_PRINT_LINE_BEGIN:
	case BINARY_PRINT_ADDR:
	case BINARY_PRINT_NUM_DATA:
	case BINARY_PRINT_NUM_PAD:
	case BINARY_PRINT_SEP:
	case BINARY_PRINT_CHAR_PAD:
	case BINARY_PRINT_LINE_END:
	case BINARY_PRINT_DATA_END:
	default:
		break;
	}
2037 2038

	return 0;
2039 2040 2041 2042 2043
}

static void bpf_output__fprintf(struct trace *trace,
				struct perf_sample *sample)
{
2044 2045
	binary__fprintf(sample->raw_data, sample->raw_size, 8,
			bpf_output__printer, NULL, trace->output);
2046 2047
}

2048 2049 2050 2051
static int trace__event_handler(struct trace *trace, struct perf_evsel *evsel,
				union perf_event *event __maybe_unused,
				struct perf_sample *sample)
{
2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062
	int callchain_ret = 0;

	if (sample->callchain) {
		callchain_ret = trace__resolve_callchain(trace, evsel, sample, &callchain_cursor);
		if (callchain_ret == 0) {
			if (callchain_cursor.nr < trace->min_stack)
				goto out;
			callchain_ret = 1;
		}
	}

2063
	trace__printf_interrupted_entry(trace);
2064
	trace__fprintf_tstamp(trace, sample->time, trace->output);
2065 2066 2067 2068

	if (trace->trace_syscalls)
		fprintf(trace->output, "(         ): ");

2069 2070 2071 2072 2073 2074 2075 2076
	if (evsel == trace->syscalls.events.augmented) {
		int id = perf_evsel__sc_tp_uint(evsel, id, sample);
		struct syscall *sc = trace__syscall_info(trace, evsel, id);

		if (sc) {
			struct thread *thread = machine__findnew_thread(trace->host, sample->pid, sample->tid);

			if (thread) {
2077
				trace__fprintf_comm_tid(trace, thread, trace->output);
2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092
				fprintf(trace->output, "%s(", sc->name);
				trace__fprintf_sys_enter(trace, evsel, sample);
				fputc(')', trace->output);
				thread__put(thread);
				goto newline;
			}
		}

		/*
		 * XXX: Not having the associated syscall info or not finding/adding
		 * 	the thread should never happen, but if it does...
		 * 	fall thru and print it as a bpf_output event.
		 */
	}

2093
	fprintf(trace->output, "%s:", evsel->name);
2094

2095
	if (perf_evsel__is_bpf_output(evsel)) {
2096
		bpf_output__fprintf(trace, sample);
2097
	} else if (evsel->tp_format) {
2098 2099 2100 2101 2102 2103
		if (strncmp(evsel->tp_format->name, "sys_enter_", 10) ||
		    trace__fprintf_sys_enter(trace, evsel, sample)) {
			event_format__fprintf(evsel->tp_format, sample->cpu,
					      sample->raw_data, sample->raw_size,
					      trace->output);
		}
2104 2105
	}

2106
newline:
C
Changbin Du 已提交
2107
	fprintf(trace->output, "\n");
2108

2109 2110 2111 2112 2113
	if (callchain_ret > 0)
		trace__fprintf_callchain(trace, sample);
	else if (callchain_ret < 0)
		pr_err("Problem processing %s callchain, skipping...\n", perf_evsel__name(evsel));
out:
2114 2115 2116
	return 0;
}

2117 2118 2119 2120 2121
static void print_location(FILE *f, struct perf_sample *sample,
			   struct addr_location *al,
			   bool print_dso, bool print_sym)
{

2122
	if ((verbose > 0 || print_dso) && al->map)
2123 2124
		fprintf(f, "%s@", al->map->dso->long_name);

2125
	if ((verbose > 0 || print_sym) && al->sym)
2126
		fprintf(f, "%s+0x%" PRIx64, al->sym->name,
2127 2128
			al->addr - al->sym->start);
	else if (al->map)
2129
		fprintf(f, "0x%" PRIx64, al->addr);
2130
	else
2131
		fprintf(f, "0x%" PRIx64, sample->addr);
2132 2133 2134 2135
}

static int trace__pgfault(struct trace *trace,
			  struct perf_evsel *evsel,
2136
			  union perf_event *event __maybe_unused,
2137 2138 2139 2140 2141
			  struct perf_sample *sample)
{
	struct thread *thread;
	struct addr_location al;
	char map_type = 'd';
2142
	struct thread_trace *ttrace;
2143
	int err = -1;
2144
	int callchain_ret = 0;
2145 2146

	thread = machine__findnew_thread(trace->host, sample->pid, sample->tid);
2147 2148 2149 2150 2151 2152 2153 2154 2155 2156

	if (sample->callchain) {
		callchain_ret = trace__resolve_callchain(trace, evsel, sample, &callchain_cursor);
		if (callchain_ret == 0) {
			if (callchain_cursor.nr < trace->min_stack)
				goto out_put;
			callchain_ret = 1;
		}
	}

2157 2158
	ttrace = thread__trace(thread, trace->output);
	if (ttrace == NULL)
2159
		goto out_put;
2160 2161 2162 2163 2164 2165 2166

	if (evsel->attr.config == PERF_COUNT_SW_PAGE_FAULTS_MAJ)
		ttrace->pfmaj++;
	else
		ttrace->pfmin++;

	if (trace->summary_only)
2167
		goto out;
2168

2169
	thread__find_symbol(thread, sample->cpumode, sample->ip, &al);
2170

2171
	trace__fprintf_entry_head(trace, thread, 0, true, sample->time, trace->output);
2172 2173 2174 2175 2176 2177 2178 2179 2180

	fprintf(trace->output, "%sfault [",
		evsel->attr.config == PERF_COUNT_SW_PAGE_FAULTS_MAJ ?
		"maj" : "min");

	print_location(trace->output, sample, &al, false, true);

	fprintf(trace->output, "] => ");

2181
	thread__find_symbol(thread, sample->cpumode, sample->addr, &al);
2182 2183

	if (!al.map) {
2184
		thread__find_symbol(thread, sample->cpumode, sample->addr, &al);
2185 2186 2187 2188 2189 2190 2191 2192 2193 2194

		if (al.map)
			map_type = 'x';
		else
			map_type = '?';
	}

	print_location(trace->output, sample, &al, true, false);

	fprintf(trace->output, " (%c%c)\n", map_type, al.level);
2195

2196 2197 2198 2199
	if (callchain_ret > 0)
		trace__fprintf_callchain(trace, sample);
	else if (callchain_ret < 0)
		pr_err("Problem processing %s callchain, skipping...\n", perf_evsel__name(evsel));
2200 2201 2202 2203 2204
out:
	err = 0;
out_put:
	thread__put(thread);
	return err;
2205 2206
}

2207
static void trace__set_base_time(struct trace *trace,
2208
				 struct perf_evsel *evsel,
2209 2210
				 struct perf_sample *sample)
{
2211 2212 2213 2214 2215 2216 2217 2218 2219 2220
	/*
	 * BPF events were not setting PERF_SAMPLE_TIME, so be more robust
	 * and don't use sample->time unconditionally, we may end up having
	 * some other event in the future without PERF_SAMPLE_TIME for good
	 * reason, i.e. we may not be interested in its timestamps, just in
	 * it taking place, picking some piece of information when it
	 * appears in our event stream (vfs_getname comes to mind).
	 */
	if (trace->base_time == 0 && !trace->full_time &&
	    (evsel->attr.sample_type & PERF_SAMPLE_TIME))
2221 2222 2223
		trace->base_time = sample->time;
}

2224
static int trace__process_sample(struct perf_tool *tool,
2225
				 union perf_event *event,
2226 2227 2228 2229 2230
				 struct perf_sample *sample,
				 struct perf_evsel *evsel,
				 struct machine *machine __maybe_unused)
{
	struct trace *trace = container_of(tool, struct trace, tool);
2231
	struct thread *thread;
2232 2233
	int err = 0;

2234
	tracepoint_handler handler = evsel->handler;
2235

2236 2237
	thread = machine__findnew_thread(trace->host, sample->pid, sample->tid);
	if (thread && thread__is_filtered(thread))
2238
		goto out;
2239

2240
	trace__set_base_time(trace, evsel, sample);
2241

2242 2243
	if (handler) {
		++trace->nr_events;
2244
		handler(trace, evsel, event, sample);
2245
	}
2246 2247
out:
	thread__put(thread);
2248 2249 2250
	return err;
}

2251
static int trace__record(struct trace *trace, int argc, const char **argv)
D
David Ahern 已提交
2252 2253 2254 2255 2256 2257 2258 2259 2260 2261
{
	unsigned int rec_argc, i, j;
	const char **rec_argv;
	const char * const record_args[] = {
		"record",
		"-R",
		"-m", "1024",
		"-c", "1",
	};

2262 2263 2264 2265 2266 2267 2268
	const char * const sc_args[] = { "-e", };
	unsigned int sc_args_nr = ARRAY_SIZE(sc_args);
	const char * const majpf_args[] = { "-e", "major-faults" };
	unsigned int majpf_args_nr = ARRAY_SIZE(majpf_args);
	const char * const minpf_args[] = { "-e", "minor-faults" };
	unsigned int minpf_args_nr = ARRAY_SIZE(minpf_args);

2269
	/* +1 is for the event string below */
2270 2271
	rec_argc = ARRAY_SIZE(record_args) + sc_args_nr + 1 +
		majpf_args_nr + minpf_args_nr + argc;
D
David Ahern 已提交
2272 2273 2274 2275 2276
	rec_argv = calloc(rec_argc + 1, sizeof(char *));

	if (rec_argv == NULL)
		return -ENOMEM;

2277
	j = 0;
D
David Ahern 已提交
2278
	for (i = 0; i < ARRAY_SIZE(record_args); i++)
2279 2280
		rec_argv[j++] = record_args[i];

2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291
	if (trace->trace_syscalls) {
		for (i = 0; i < sc_args_nr; i++)
			rec_argv[j++] = sc_args[i];

		/* event string may be different for older kernels - e.g., RHEL6 */
		if (is_valid_tracepoint("raw_syscalls:sys_enter"))
			rec_argv[j++] = "raw_syscalls:sys_enter,raw_syscalls:sys_exit";
		else if (is_valid_tracepoint("syscalls:sys_enter"))
			rec_argv[j++] = "syscalls:sys_enter,syscalls:sys_exit";
		else {
			pr_err("Neither raw_syscalls nor syscalls events exist.\n");
2292
			free(rec_argv);
2293 2294
			return -1;
		}
2295 2296
	}

2297 2298 2299 2300 2301 2302 2303 2304 2305 2306
	if (trace->trace_pgfaults & TRACE_PFMAJ)
		for (i = 0; i < majpf_args_nr; i++)
			rec_argv[j++] = majpf_args[i];

	if (trace->trace_pgfaults & TRACE_PFMIN)
		for (i = 0; i < minpf_args_nr; i++)
			rec_argv[j++] = minpf_args[i];

	for (i = 0; i < (unsigned int)argc; i++)
		rec_argv[j++] = argv[i];
D
David Ahern 已提交
2307

2308
	return cmd_record(j, rec_argv);
D
David Ahern 已提交
2309 2310
}

2311 2312
static size_t trace__fprintf_thread_summary(struct trace *trace, FILE *fp);

2313
static bool perf_evlist__add_vfs_getname(struct perf_evlist *evlist)
2314
{
2315
	struct perf_evsel *evsel = perf_evsel__newtp("probe", "vfs_getname");
2316 2317

	if (IS_ERR(evsel))
2318
		return false;
2319 2320 2321

	if (perf_evsel__field(evsel, "pathname") == NULL) {
		perf_evsel__delete(evsel);
2322
		return false;
2323 2324
	}

2325
	evsel->handler = trace__vfs_getname;
2326
	perf_evlist__add(evlist, evsel);
2327
	return true;
2328 2329
}

2330
static struct perf_evsel *perf_evsel__new_pgfault(u64 config)
2331 2332 2333 2334 2335 2336 2337 2338
{
	struct perf_evsel *evsel;
	struct perf_event_attr attr = {
		.type = PERF_TYPE_SOFTWARE,
		.mmap_data = 1,
	};

	attr.config = config;
2339
	attr.sample_period = 1;
2340 2341 2342 2343

	event_attr_init(&attr);

	evsel = perf_evsel__new(&attr);
2344 2345
	if (evsel)
		evsel->handler = trace__pgfault;
2346

2347
	return evsel;
2348 2349
}

2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365
static void trace__handle_event(struct trace *trace, union perf_event *event, struct perf_sample *sample)
{
	const u32 type = event->header.type;
	struct perf_evsel *evsel;

	if (type != PERF_RECORD_SAMPLE) {
		trace__process_event(trace, trace->host, event, sample);
		return;
	}

	evsel = perf_evlist__id2evsel(trace->evlist, sample->id);
	if (evsel == NULL) {
		fprintf(trace->output, "Unknown tp ID %" PRIu64 ", skipping...\n", sample->id);
		return;
	}

2366 2367
	trace__set_base_time(trace, evsel, sample);

2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378
	if (evsel->attr.type == PERF_TYPE_TRACEPOINT &&
	    sample->raw_data == NULL) {
		fprintf(trace->output, "%s sample with no payload for tid: %d, cpu %d, raw_size=%d, skipping...\n",
		       perf_evsel__name(evsel), sample->tid,
		       sample->cpu, sample->raw_size);
	} else {
		tracepoint_handler handler = evsel->handler;
		handler(trace, evsel, event, sample);
	}
}

2379 2380 2381 2382 2383 2384
static int trace__add_syscall_newtp(struct trace *trace)
{
	int ret = -1;
	struct perf_evlist *evlist = trace->evlist;
	struct perf_evsel *sys_enter, *sys_exit;

2385
	sys_enter = perf_evsel__raw_syscall_newtp("sys_enter", trace__sys_enter);
2386 2387 2388 2389 2390 2391
	if (sys_enter == NULL)
		goto out;

	if (perf_evsel__init_sc_tp_ptr_field(sys_enter, args))
		goto out_delete_sys_enter;

2392
	sys_exit = perf_evsel__raw_syscall_newtp("sys_exit", trace__sys_exit);
2393 2394 2395 2396 2397 2398
	if (sys_exit == NULL)
		goto out_delete_sys_enter;

	if (perf_evsel__init_sc_tp_uint_field(sys_exit, ret))
		goto out_delete_sys_exit;

2399 2400 2401
	perf_evsel__config_callchain(sys_enter, &trace->opts, &callchain_param);
	perf_evsel__config_callchain(sys_exit, &trace->opts, &callchain_param);

2402 2403 2404
	perf_evlist__add(evlist, sys_enter);
	perf_evlist__add(evlist, sys_exit);

2405
	if (callchain_param.enabled && !trace->kernel_syscallchains) {
2406 2407 2408 2409 2410 2411 2412 2413
		/*
		 * We're interested only in the user space callchain
		 * leading to the syscall, allow overriding that for
		 * debugging reasons using --kernel_syscall_callchains
		 */
		sys_exit->attr.exclude_callchain_kernel = 1;
	}

2414 2415
	trace->syscalls.events.sys_enter = sys_enter;
	trace->syscalls.events.sys_exit  = sys_exit;
2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427

	ret = 0;
out:
	return ret;

out_delete_sys_exit:
	perf_evsel__delete_priv(sys_exit);
out_delete_sys_enter:
	perf_evsel__delete_priv(sys_enter);
	goto out;
}

2428 2429 2430
static int trace__set_ev_qualifier_filter(struct trace *trace)
{
	int err = -1;
2431
	struct perf_evsel *sys_exit;
2432 2433 2434 2435 2436 2437 2438
	char *filter = asprintf_expr_inout_ints("id", !trace->not_ev_qualifier,
						trace->ev_qualifier_ids.nr,
						trace->ev_qualifier_ids.entries);

	if (filter == NULL)
		goto out_enomem;

2439 2440
	if (!perf_evsel__append_tp_filter(trace->syscalls.events.sys_enter,
					  filter)) {
2441
		sys_exit = trace->syscalls.events.sys_exit;
2442
		err = perf_evsel__append_tp_filter(sys_exit, filter);
2443
	}
2444 2445 2446 2447 2448 2449 2450 2451

	free(filter);
out:
	return err;
out_enomem:
	errno = ENOMEM;
	goto out;
}
2452

2453 2454
static int trace__set_filter_loop_pids(struct trace *trace)
{
2455
	unsigned int nr = 1;
2456 2457 2458
	pid_t pids[32] = {
		getpid(),
	};
2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472
	struct thread *thread = machine__find_thread(trace->host, pids[0], pids[0]);

	while (thread && nr < ARRAY_SIZE(pids)) {
		struct thread *parent = machine__find_thread(trace->host, thread->ppid, thread->ppid);

		if (parent == NULL)
			break;

		if (!strcmp(thread__comm_str(parent), "sshd")) {
			pids[nr++] = parent->tid;
			break;
		}
		thread = parent;
	}
2473 2474 2475 2476

	return perf_evlist__set_filter_pids(trace->evlist, nr, pids);
}

2477
static int trace__run(struct trace *trace, int argc, const char **argv)
A
Arnaldo Carvalho de Melo 已提交
2478
{
2479
	struct perf_evlist *evlist = trace->evlist;
2480
	struct perf_evsel *evsel, *pgfault_maj = NULL, *pgfault_min = NULL;
2481 2482
	int err = -1, i;
	unsigned long before;
2483
	const bool forks = argc > 0;
2484
	bool draining = false;
A
Arnaldo Carvalho de Melo 已提交
2485

2486 2487
	trace->live = true;

2488
	if (trace->trace_syscalls && trace__add_syscall_newtp(trace))
2489
		goto out_error_raw_syscalls;
A
Arnaldo Carvalho de Melo 已提交
2490

2491
	if (trace->trace_syscalls)
2492
		trace->vfs_getname = perf_evlist__add_vfs_getname(evlist);
2493

2494 2495 2496 2497
	if ((trace->trace_pgfaults & TRACE_PFMAJ)) {
		pgfault_maj = perf_evsel__new_pgfault(PERF_COUNT_SW_PAGE_FAULTS_MAJ);
		if (pgfault_maj == NULL)
			goto out_error_mem;
2498
		perf_evsel__config_callchain(pgfault_maj, &trace->opts, &callchain_param);
2499
		perf_evlist__add(evlist, pgfault_maj);
2500
	}
2501

2502 2503 2504 2505
	if ((trace->trace_pgfaults & TRACE_PFMIN)) {
		pgfault_min = perf_evsel__new_pgfault(PERF_COUNT_SW_PAGE_FAULTS_MIN);
		if (pgfault_min == NULL)
			goto out_error_mem;
2506
		perf_evsel__config_callchain(pgfault_min, &trace->opts, &callchain_param);
2507 2508
		perf_evlist__add(evlist, pgfault_min);
	}
2509

2510
	if (trace->sched &&
2511 2512 2513
	    perf_evlist__add_newtp(evlist, "sched", "sched_stat_runtime",
				   trace__sched_stat_runtime))
		goto out_error_sched_stat_runtime;
2514

2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542
	/*
	 * If a global cgroup was set, apply it to all the events without an
	 * explicit cgroup. I.e.:
	 *
	 * 	trace -G A -e sched:*switch
	 *
	 * Will set all raw_syscalls:sys_{enter,exit}, pgfault, vfs_getname, etc
	 * _and_ sched:sched_switch to the 'A' cgroup, while:
	 *
	 * trace -e sched:*switch -G A
	 *
	 * will only set the sched:sched_switch event to the 'A' cgroup, all the
	 * other events (raw_syscalls:sys_{enter,exit}, etc are left "without"
	 * a cgroup (on the root cgroup, sys wide, etc).
	 *
	 * Multiple cgroups:
	 *
	 * trace -G A -e sched:*switch -G B
	 *
	 * the syscall ones go to the 'A' cgroup, the sched:sched_switch goes
	 * to the 'B' cgroup.
	 *
	 * evlist__set_default_cgroup() grabs a reference of the passed cgroup
	 * only for the evsels still without a cgroup, i.e. evsel->cgroup == NULL.
	 */
	if (trace->cgroup)
		evlist__set_default_cgroup(trace->evlist, trace->cgroup);

A
Arnaldo Carvalho de Melo 已提交
2543 2544
	err = perf_evlist__create_maps(evlist, &trace->opts.target);
	if (err < 0) {
2545
		fprintf(trace->output, "Problems parsing the target to trace, check your options!\n");
A
Arnaldo Carvalho de Melo 已提交
2546 2547 2548
		goto out_delete_evlist;
	}

2549 2550
	err = trace__symbols_init(trace, evlist);
	if (err < 0) {
2551
		fprintf(trace->output, "Problems initializing symbol libraries!\n");
2552
		goto out_delete_evlist;
2553 2554
	}

2555
	perf_evlist__config(evlist, &trace->opts, &callchain_param);
2556

2557 2558 2559 2560
	signal(SIGCHLD, sig_handler);
	signal(SIGINT, sig_handler);

	if (forks) {
2561
		err = perf_evlist__prepare_workload(evlist, &trace->opts.target,
2562
						    argv, false, NULL);
2563
		if (err < 0) {
2564
			fprintf(trace->output, "Couldn't run the workload!\n");
2565
			goto out_delete_evlist;
2566 2567 2568
		}
	}

A
Arnaldo Carvalho de Melo 已提交
2569
	err = perf_evlist__open(evlist);
2570 2571
	if (err < 0)
		goto out_error_open;
A
Arnaldo Carvalho de Melo 已提交
2572

2573 2574 2575 2576 2577 2578 2579 2580 2581 2582
	err = bpf__apply_obj_config();
	if (err) {
		char errbuf[BUFSIZ];

		bpf__strerror_apply_obj_config(err, errbuf, sizeof(errbuf));
		pr_err("ERROR: Apply config to BPF failed: %s\n",
			 errbuf);
		goto out_error_open;
	}

2583 2584 2585 2586 2587 2588
	/*
	 * Better not use !target__has_task() here because we need to cover the
	 * case where no threads were specified in the command line, but a
	 * workload was, and in that case we will fill in the thread_map when
	 * we fork the workload in perf_evlist__prepare_workload.
	 */
2589 2590
	if (trace->filter_pids.nr > 0)
		err = perf_evlist__set_filter_pids(evlist, trace->filter_pids.nr, trace->filter_pids.entries);
2591
	else if (thread_map__pid(evlist->threads, 0) == -1)
2592
		err = trace__set_filter_loop_pids(trace);
2593

2594 2595 2596
	if (err < 0)
		goto out_error_mem;

2597 2598 2599 2600 2601
	if (trace->ev_qualifier_ids.nr > 0) {
		err = trace__set_ev_qualifier_filter(trace);
		if (err < 0)
			goto out_errno;

2602 2603 2604
		pr_debug("event qualifier tracepoint filter: %s\n",
			 trace->syscalls.events.sys_exit->filter);
	}
2605

2606 2607 2608
	err = perf_evlist__apply_filters(evlist, &evsel);
	if (err < 0)
		goto out_error_apply_filters;
2609

2610
	err = perf_evlist__mmap(evlist, trace->opts.mmap_pages);
2611 2612
	if (err < 0)
		goto out_error_mmap;
A
Arnaldo Carvalho de Melo 已提交
2613

A
Alexis Berlemont 已提交
2614
	if (!target__none(&trace->opts.target) && !trace->opts.initial_delay)
2615 2616
		perf_evlist__enable(evlist);

2617 2618 2619
	if (forks)
		perf_evlist__start_workload(evlist);

A
Alexis Berlemont 已提交
2620 2621 2622 2623 2624
	if (trace->opts.initial_delay) {
		usleep(trace->opts.initial_delay * 1000);
		perf_evlist__enable(evlist);
	}

2625
	trace->multiple_threads = thread_map__pid(evlist->threads, 0) == -1 ||
2626 2627
				  evlist->threads->nr > 1 ||
				  perf_evlist__first(evlist)->attr.inherit;
2628 2629 2630 2631 2632 2633 2634 2635

	/*
	 * Now that we already used evsel->attr to ask the kernel to setup the
	 * events, lets reuse evsel->attr.sample_max_stack as the limit in
	 * trace__resolve_callchain(), allowing per-event max-stack settings
	 * to override an explicitely set --max-stack global setting.
	 */
	evlist__for_each_entry(evlist, evsel) {
2636
		if (evsel__has_callchain(evsel) &&
2637 2638 2639
		    evsel->attr.sample_max_stack == 0)
			evsel->attr.sample_max_stack = trace->max_stack;
	}
A
Arnaldo Carvalho de Melo 已提交
2640
again:
2641
	before = trace->nr_events;
A
Arnaldo Carvalho de Melo 已提交
2642 2643 2644

	for (i = 0; i < evlist->nr_mmaps; i++) {
		union perf_event *event;
2645
		struct perf_mmap *md;
A
Arnaldo Carvalho de Melo 已提交
2646

2647
		md = &evlist->mmap[i];
2648
		if (perf_mmap__read_init(md) < 0)
2649 2650
			continue;

2651
		while ((event = perf_mmap__read_event(md)) != NULL) {
A
Arnaldo Carvalho de Melo 已提交
2652 2653
			struct perf_sample sample;

2654
			++trace->nr_events;
A
Arnaldo Carvalho de Melo 已提交
2655 2656 2657

			err = perf_evlist__parse_sample(evlist, event, &sample);
			if (err) {
2658
				fprintf(trace->output, "Can't parse sample, err = %d, skipping...\n", err);
2659
				goto next_event;
A
Arnaldo Carvalho de Melo 已提交
2660 2661
			}

2662
			trace__handle_event(trace, event, &sample);
2663
next_event:
2664
			perf_mmap__consume(md);
2665

2666 2667
			if (interrupted)
				goto out_disable;
2668 2669 2670 2671 2672

			if (done && !draining) {
				perf_evlist__disable(evlist);
				draining = true;
			}
A
Arnaldo Carvalho de Melo 已提交
2673
		}
2674
		perf_mmap__read_done(md);
A
Arnaldo Carvalho de Melo 已提交
2675 2676
	}

2677
	if (trace->nr_events == before) {
2678
		int timeout = done ? 100 : -1;
2679

2680 2681 2682 2683
		if (!draining && perf_evlist__poll(evlist, timeout) > 0) {
			if (perf_evlist__filter_pollfd(evlist, POLLERR | POLLHUP) == 0)
				draining = true;

2684
			goto again;
2685
		}
2686 2687
	} else {
		goto again;
2688 2689
	}

2690
out_disable:
2691 2692
	thread__zput(trace->current);

2693
	perf_evlist__disable(evlist);
A
Arnaldo Carvalho de Melo 已提交
2694

2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706
	if (!err) {
		if (trace->summary)
			trace__fprintf_thread_summary(trace, trace->output);

		if (trace->show_tool_stats) {
			fprintf(trace->output, "Stats:\n "
					       " vfs_getname : %" PRIu64 "\n"
					       " proc_getname: %" PRIu64 "\n",
				trace->stats.vfs_getname,
				trace->stats.proc_getname);
		}
	}
2707

A
Arnaldo Carvalho de Melo 已提交
2708
out_delete_evlist:
2709 2710
	trace__symbols__exit(trace);

A
Arnaldo Carvalho de Melo 已提交
2711
	perf_evlist__delete(evlist);
2712
	cgroup__put(trace->cgroup);
2713
	trace->evlist = NULL;
2714
	trace->live = false;
A
Arnaldo Carvalho de Melo 已提交
2715
	return err;
2716 2717
{
	char errbuf[BUFSIZ];
2718

2719
out_error_sched_stat_runtime:
2720
	tracing_path__strerror_open_tp(errno, errbuf, sizeof(errbuf), "sched", "sched_stat_runtime");
2721 2722
	goto out_error;

2723
out_error_raw_syscalls:
2724
	tracing_path__strerror_open_tp(errno, errbuf, sizeof(errbuf), "raw_syscalls", "sys_(enter|exit)");
2725 2726
	goto out_error;

2727 2728 2729 2730
out_error_mmap:
	perf_evlist__strerror_mmap(evlist, errno, errbuf, sizeof(errbuf));
	goto out_error;

2731 2732 2733 2734
out_error_open:
	perf_evlist__strerror_open(evlist, errno, errbuf, sizeof(errbuf));

out_error:
2735
	fprintf(trace->output, "%s\n", errbuf);
2736
	goto out_delete_evlist;
2737 2738 2739 2740 2741

out_error_apply_filters:
	fprintf(trace->output,
		"Failed to set filter \"%s\" on event %s with %d (%s)\n",
		evsel->filter, perf_evsel__name(evsel), errno,
2742
		str_error_r(errno, errbuf, sizeof(errbuf)));
2743
	goto out_delete_evlist;
A
Arnaldo Carvalho de Melo 已提交
2744
}
2745 2746 2747
out_error_mem:
	fprintf(trace->output, "Not enough memory to run!\n");
	goto out_delete_evlist;
2748 2749 2750 2751

out_errno:
	fprintf(trace->output, "errno=%d,%s\n", errno, strerror(errno));
	goto out_delete_evlist;
2752
}
A
Arnaldo Carvalho de Melo 已提交
2753

2754 2755 2756
static int trace__replay(struct trace *trace)
{
	const struct perf_evsel_str_handler handlers[] = {
2757
		{ "probe:vfs_getname",	     trace__vfs_getname, },
2758
	};
2759
	struct perf_data data = {
J
Jiri Olsa 已提交
2760 2761 2762 2763 2764
		.file      = {
			.path = input_name,
		},
		.mode      = PERF_DATA_MODE_READ,
		.force     = trace->force,
2765
	};
2766
	struct perf_session *session;
2767
	struct perf_evsel *evsel;
2768 2769 2770 2771
	int err = -1;

	trace->tool.sample	  = trace__process_sample;
	trace->tool.mmap	  = perf_event__process_mmap;
D
David Ahern 已提交
2772
	trace->tool.mmap2	  = perf_event__process_mmap2;
2773 2774 2775 2776
	trace->tool.comm	  = perf_event__process_comm;
	trace->tool.exit	  = perf_event__process_exit;
	trace->tool.fork	  = perf_event__process_fork;
	trace->tool.attr	  = perf_event__process_attr;
2777
	trace->tool.tracing_data  = perf_event__process_tracing_data;
2778
	trace->tool.build_id	  = perf_event__process_build_id;
2779
	trace->tool.namespaces	  = perf_event__process_namespaces;
2780

2781
	trace->tool.ordered_events = true;
2782 2783 2784 2785 2786
	trace->tool.ordering_requires_timestamps = true;

	/* add tid to output */
	trace->multiple_threads = true;

2787
	session = perf_session__new(&data, false, &trace->tool);
2788
	if (session == NULL)
2789
		return -1;
2790

2791 2792 2793 2794 2795 2796
	if (trace->opts.target.pid)
		symbol_conf.pid_list_str = strdup(trace->opts.target.pid);

	if (trace->opts.target.tid)
		symbol_conf.tid_list_str = strdup(trace->opts.target.tid);

2797
	if (symbol__init(&session->header.env) < 0)
2798 2799
		goto out;

2800 2801
	trace->host = &session->machines.host;

2802 2803 2804 2805
	err = perf_session__set_tracepoints_handlers(session, handlers);
	if (err)
		goto out;

2806 2807
	evsel = perf_evlist__find_tracepoint_by_name(session->evlist,
						     "raw_syscalls:sys_enter");
2808 2809 2810 2811
	/* older kernels have syscalls tp versus raw_syscalls */
	if (evsel == NULL)
		evsel = perf_evlist__find_tracepoint_by_name(session->evlist,
							     "syscalls:sys_enter");
2812

2813
	if (evsel &&
2814
	    (perf_evsel__init_raw_syscall_tp(evsel, trace__sys_enter) < 0 ||
2815
	    perf_evsel__init_sc_tp_ptr_field(evsel, args))) {
2816 2817 2818 2819 2820 2821
		pr_err("Error during initialize raw_syscalls:sys_enter event\n");
		goto out;
	}

	evsel = perf_evlist__find_tracepoint_by_name(session->evlist,
						     "raw_syscalls:sys_exit");
2822 2823 2824
	if (evsel == NULL)
		evsel = perf_evlist__find_tracepoint_by_name(session->evlist,
							     "syscalls:sys_exit");
2825
	if (evsel &&
2826
	    (perf_evsel__init_raw_syscall_tp(evsel, trace__sys_exit) < 0 ||
2827
	    perf_evsel__init_sc_tp_uint_field(evsel, ret))) {
2828
		pr_err("Error during initialize raw_syscalls:sys_exit event\n");
2829 2830 2831
		goto out;
	}

2832
	evlist__for_each_entry(session->evlist, evsel) {
2833 2834 2835 2836 2837 2838 2839
		if (evsel->attr.type == PERF_TYPE_SOFTWARE &&
		    (evsel->attr.config == PERF_COUNT_SW_PAGE_FAULTS_MAJ ||
		     evsel->attr.config == PERF_COUNT_SW_PAGE_FAULTS_MIN ||
		     evsel->attr.config == PERF_COUNT_SW_PAGE_FAULTS))
			evsel->handler = trace__pgfault;
	}

2840 2841
	setup_pager();

2842
	err = perf_session__process_events(session);
2843 2844 2845
	if (err)
		pr_err("Failed to process events, error %d", err);

2846 2847 2848
	else if (trace->summary)
		trace__fprintf_thread_summary(trace, trace->output);

2849 2850 2851 2852 2853 2854
out:
	perf_session__delete(session);

	return err;
}

2855 2856 2857 2858
static size_t trace__fprintf_threads_header(FILE *fp)
{
	size_t printed;

2859
	printed  = fprintf(fp, "\n Summary of events:\n\n");
2860 2861 2862 2863

	return printed;
}

2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877
DEFINE_RESORT_RB(syscall_stats, a->msecs > b->msecs,
	struct stats 	*stats;
	double		msecs;
	int		syscall;
)
{
	struct int_node *source = rb_entry(nd, struct int_node, rb_node);
	struct stats *stats = source->priv;

	entry->syscall = source->i;
	entry->stats   = stats;
	entry->msecs   = stats ? (u64)stats->n * (avg_stats(stats) / NSEC_PER_MSEC) : 0;
}

2878 2879 2880 2881 2882
static size_t thread__dump_stats(struct thread_trace *ttrace,
				 struct trace *trace, FILE *fp)
{
	size_t printed = 0;
	struct syscall *sc;
2883 2884
	struct rb_node *nd;
	DECLARE_RESORT_RB_INTLIST(syscall_stats, ttrace->syscall_stats);
2885

2886
	if (syscall_stats == NULL)
2887 2888 2889 2890
		return 0;

	printed += fprintf(fp, "\n");

2891 2892 2893
	printed += fprintf(fp, "   syscall            calls    total       min       avg       max      stddev\n");
	printed += fprintf(fp, "                               (msec)    (msec)    (msec)    (msec)        (%%)\n");
	printed += fprintf(fp, "   --------------- -------- --------- --------- --------- ---------     ------\n");
2894

2895
	resort_rb__for_each_entry(nd, syscall_stats) {
2896
		struct stats *stats = syscall_stats_entry->stats;
2897 2898 2899 2900 2901 2902 2903 2904 2905 2906
		if (stats) {
			double min = (double)(stats->min) / NSEC_PER_MSEC;
			double max = (double)(stats->max) / NSEC_PER_MSEC;
			double avg = avg_stats(stats);
			double pct;
			u64 n = (u64) stats->n;

			pct = avg ? 100.0 * stddev_stats(stats)/avg : 0.0;
			avg /= NSEC_PER_MSEC;

2907
			sc = &trace->syscalls.table[syscall_stats_entry->syscall];
2908
			printed += fprintf(fp, "   %-15s", sc->name);
2909
			printed += fprintf(fp, " %8" PRIu64 " %9.3f %9.3f %9.3f",
2910
					   n, syscall_stats_entry->msecs, min, avg);
P
Pekka Enberg 已提交
2911
			printed += fprintf(fp, " %9.3f %9.2f%%\n", max, pct);
2912 2913 2914
		}
	}

2915
	resort_rb__delete(syscall_stats);
2916
	printed += fprintf(fp, "\n\n");
2917 2918 2919 2920

	return printed;
}

2921
static size_t trace__fprintf_thread(FILE *fp, struct thread *thread, struct trace *trace)
2922
{
2923
	size_t printed = 0;
2924
	struct thread_trace *ttrace = thread__priv(thread);
2925 2926 2927 2928 2929 2930 2931
	double ratio;

	if (ttrace == NULL)
		return 0;

	ratio = (double)ttrace->nr_events / trace->nr_events * 100.0;

2932
	printed += fprintf(fp, " %s (%d), ", thread__comm_str(thread), thread->tid);
2933
	printed += fprintf(fp, "%lu events, ", ttrace->nr_events);
2934
	printed += fprintf(fp, "%.1f%%", ratio);
2935 2936 2937 2938
	if (ttrace->pfmaj)
		printed += fprintf(fp, ", %lu majfaults", ttrace->pfmaj);
	if (ttrace->pfmin)
		printed += fprintf(fp, ", %lu minfaults", ttrace->pfmin);
2939 2940 2941 2942 2943
	if (trace->sched)
		printed += fprintf(fp, ", %.3f msec\n", ttrace->runtime_ms);
	else if (fputc('\n', fp) != EOF)
		++printed;

2944
	printed += thread__dump_stats(ttrace, trace, fp);
2945

2946 2947
	return printed;
}
2948

2949 2950 2951 2952 2953 2954 2955 2956 2957 2958
static unsigned long thread__nr_events(struct thread_trace *ttrace)
{
	return ttrace ? ttrace->nr_events : 0;
}

DEFINE_RESORT_RB(threads, (thread__nr_events(a->thread->priv) < thread__nr_events(b->thread->priv)),
	struct thread *thread;
)
{
	entry->thread = rb_entry(nd, struct thread, rb_node);
2959 2960
}

2961 2962
static size_t trace__fprintf_thread_summary(struct trace *trace, FILE *fp)
{
2963 2964
	size_t printed = trace__fprintf_threads_header(fp);
	struct rb_node *nd;
2965
	int i;
2966

2967 2968
	for (i = 0; i < THREADS__TABLE_SIZE; i++) {
		DECLARE_RESORT_RB_MACHINE_THREADS(threads, trace->host, i);
2969

2970 2971 2972 2973
		if (threads == NULL) {
			fprintf(fp, "%s", "Error sorting output by nr_events!\n");
			return 0;
		}
2974

2975 2976
		resort_rb__for_each_entry(nd, threads)
			printed += trace__fprintf_thread(fp, threads_entry->thread, trace);
2977

2978 2979
		resort_rb__delete(threads);
	}
2980
	return printed;
2981 2982
}

2983 2984 2985 2986 2987 2988 2989 2990 2991
static int trace__set_duration(const struct option *opt, const char *str,
			       int unset __maybe_unused)
{
	struct trace *trace = opt->value;

	trace->duration_filter = atof(str);
	return 0;
}

2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023
static int trace__set_filter_pids(const struct option *opt, const char *str,
				  int unset __maybe_unused)
{
	int ret = -1;
	size_t i;
	struct trace *trace = opt->value;
	/*
	 * FIXME: introduce a intarray class, plain parse csv and create a
	 * { int nr, int entries[] } struct...
	 */
	struct intlist *list = intlist__new(str);

	if (list == NULL)
		return -1;

	i = trace->filter_pids.nr = intlist__nr_entries(list) + 1;
	trace->filter_pids.entries = calloc(i, sizeof(pid_t));

	if (trace->filter_pids.entries == NULL)
		goto out;

	trace->filter_pids.entries[0] = getpid();

	for (i = 1; i < trace->filter_pids.nr; ++i)
		trace->filter_pids.entries[i] = intlist__entry(list, i - 1)->i;

	intlist__delete(list);
	ret = 0;
out:
	return ret;
}

3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040
static int trace__open_output(struct trace *trace, const char *filename)
{
	struct stat st;

	if (!stat(filename, &st) && st.st_size) {
		char oldname[PATH_MAX];

		scnprintf(oldname, sizeof(oldname), "%s.old", filename);
		unlink(oldname);
		rename(filename, oldname);
	}

	trace->output = fopen(filename, "w");

	return trace->output == NULL ? -errno : 0;
}

3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057
static int parse_pagefaults(const struct option *opt, const char *str,
			    int unset __maybe_unused)
{
	int *trace_pgfaults = opt->value;

	if (strcmp(str, "all") == 0)
		*trace_pgfaults |= TRACE_PFMAJ | TRACE_PFMIN;
	else if (strcmp(str, "maj") == 0)
		*trace_pgfaults |= TRACE_PFMAJ;
	else if (strcmp(str, "min") == 0)
		*trace_pgfaults |= TRACE_PFMIN;
	else
		return -1;

	return 0;
}

3058 3059 3060 3061
static void evlist__set_evsel_handler(struct perf_evlist *evlist, void *handler)
{
	struct perf_evsel *evsel;

3062
	evlist__for_each_entry(evlist, evsel)
3063 3064 3065
		evsel->handler = handler;
}

3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095
static int evlist__set_syscall_tp_fields(struct perf_evlist *evlist)
{
	struct perf_evsel *evsel;

	evlist__for_each_entry(evlist, evsel) {
		if (evsel->priv || !evsel->tp_format)
			continue;

		if (strcmp(evsel->tp_format->system, "syscalls"))
			continue;

		if (perf_evsel__init_syscall_tp(evsel))
			return -1;

		if (!strncmp(evsel->tp_format->name, "sys_enter_", 10)) {
			struct syscall_tp *sc = evsel->priv;

			if (__tp_field__init_ptr(&sc->args, sc->id.offset + sizeof(u64)))
				return -1;
		} else if (!strncmp(evsel->tp_format->name, "sys_exit_", 9)) {
			struct syscall_tp *sc = evsel->priv;

			if (__tp_field__init_uint(&sc->ret, sizeof(u64), sc->id.offset + sizeof(u64), evsel->needs_swap))
				return -1;
		}
	}

	return 0;
}

3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109
/*
 * XXX: Hackish, just splitting the combined -e+--event (syscalls
 * (raw_syscalls:{sys_{enter,exit}} + events (tracepoints, HW, SW, etc) to use
 * existing facilities unchanged (trace->ev_qualifier + parse_options()).
 *
 * It'd be better to introduce a parse_options() variant that would return a
 * list with the terms it didn't match to an event...
 */
static int trace__parse_events_option(const struct option *opt, const char *str,
				      int unset __maybe_unused)
{
	struct trace *trace = (struct trace *)opt->value;
	const char *s = str;
	char *sep = NULL, *lists[2] = { NULL, NULL, };
3110
	int len = strlen(str) + 1, err = -1, list, idx;
3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126
	char *strace_groups_dir = system_path(STRACE_GROUPS_DIR);
	char group_name[PATH_MAX];

	if (strace_groups_dir == NULL)
		return -1;

	if (*s == '!') {
		++s;
		trace->not_ev_qualifier = true;
	}

	while (1) {
		if ((sep = strchr(s, ',')) != NULL)
			*sep = '\0';

		list = 0;
3127 3128
		if (syscalltbl__id(trace->sctbl, s) >= 0 ||
		    syscalltbl__strglobmatch_first(trace->sctbl, s, &idx) >= 0) {
3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164
			list = 1;
		} else {
			path__join(group_name, sizeof(group_name), strace_groups_dir, s);
			if (access(group_name, R_OK) == 0)
				list = 1;
		}

		if (lists[list]) {
			sprintf(lists[list] + strlen(lists[list]), ",%s", s);
		} else {
			lists[list] = malloc(len);
			if (lists[list] == NULL)
				goto out;
			strcpy(lists[list], s);
		}

		if (!sep)
			break;

		*sep = ',';
		s = sep + 1;
	}

	if (lists[1] != NULL) {
		struct strlist_config slist_config = {
			.dirname = strace_groups_dir,
		};

		trace->ev_qualifier = strlist__new(lists[1], &slist_config);
		if (trace->ev_qualifier == NULL) {
			fputs("Not enough memory to parse event qualifier", trace->output);
			goto out;
		}

		if (trace__validate_ev_qualifier(trace))
			goto out;
3165
		trace->trace_syscalls = true;
3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182
	}

	err = 0;

	if (lists[0]) {
		struct option o = OPT_CALLBACK('e', "event", &trace->evlist, "event",
					       "event selector. use 'perf list' to list available events",
					       parse_events_option);
		err = parse_events_option(&o, lists[0], 0);
	}
out:
	if (sep)
		*sep = ',';

	return err;
}

3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194
static int trace__parse_cgroups(const struct option *opt, const char *str, int unset)
{
	struct trace *trace = opt->value;

	if (!list_empty(&trace->evlist->entries))
		return parse_cgroups(opt, str, unset);

	trace->cgroup = evlist__findnew_cgroup(trace->evlist, str);

	return 0;
}

3195
int cmd_trace(int argc, const char **argv)
A
Arnaldo Carvalho de Melo 已提交
3196
{
3197
	const char *trace_usage[] = {
3198 3199
		"perf trace [<options>] [<command>]",
		"perf trace [<options>] -- <command> [<options>]",
D
David Ahern 已提交
3200 3201
		"perf trace record [<options>] [<command>]",
		"perf trace record [<options>] -- <command> [<options>]",
A
Arnaldo Carvalho de Melo 已提交
3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214
		NULL
	};
	struct trace trace = {
		.syscalls = {
			. max = -1,
		},
		.opts = {
			.target = {
				.uid	   = UINT_MAX,
				.uses_mmap = true,
			},
			.user_freq     = UINT_MAX,
			.user_interval = ULLONG_MAX,
3215
			.no_buffering  = true,
3216
			.mmap_pages    = UINT_MAX,
3217
			.proc_map_timeout  = 500,
A
Arnaldo Carvalho de Melo 已提交
3218
		},
3219
		.output = stderr,
3220
		.show_comm = true,
3221
		.trace_syscalls = false,
3222
		.kernel_syscallchains = false,
3223
		.max_stack = UINT_MAX,
A
Arnaldo Carvalho de Melo 已提交
3224
	};
3225
	const char *output_name = NULL;
A
Arnaldo Carvalho de Melo 已提交
3226
	const struct option trace_options[] = {
3227 3228 3229
	OPT_CALLBACK('e', "event", &trace, "event",
		     "event/syscall selector. use 'perf list' to list available events",
		     trace__parse_events_option),
3230 3231
	OPT_BOOLEAN(0, "comm", &trace.show_comm,
		    "show the thread COMM next to its id"),
3232
	OPT_BOOLEAN(0, "tool_stats", &trace.show_tool_stats, "show tool stats"),
3233 3234
	OPT_CALLBACK(0, "expr", &trace, "expr", "list of syscalls/events to trace",
		     trace__parse_events_option),
3235
	OPT_STRING('o', "output", &output_name, "file", "output file name"),
3236
	OPT_STRING('i', "input", &input_name, "file", "Analyze events in file"),
A
Arnaldo Carvalho de Melo 已提交
3237 3238
	OPT_STRING('p', "pid", &trace.opts.target.pid, "pid",
		    "trace events on existing process id"),
3239
	OPT_STRING('t', "tid", &trace.opts.target.tid, "tid",
A
Arnaldo Carvalho de Melo 已提交
3240
		    "trace events on existing thread id"),
3241 3242
	OPT_CALLBACK(0, "filter-pids", &trace, "CSV list of pids",
		     "pids to filter (by the kernel)", trace__set_filter_pids),
3243
	OPT_BOOLEAN('a', "all-cpus", &trace.opts.target.system_wide,
A
Arnaldo Carvalho de Melo 已提交
3244
		    "system-wide collection from all CPUs"),
3245
	OPT_STRING('C', "cpu", &trace.opts.target.cpu_list, "cpu",
A
Arnaldo Carvalho de Melo 已提交
3246
		    "list of cpus to monitor"),
3247
	OPT_BOOLEAN(0, "no-inherit", &trace.opts.no_inherit,
A
Arnaldo Carvalho de Melo 已提交
3248
		    "child tasks do not inherit counters"),
3249 3250 3251
	OPT_CALLBACK('m', "mmap-pages", &trace.opts.mmap_pages, "pages",
		     "number of mmap data pages",
		     perf_evlist__parse_mmap_pages),
3252
	OPT_STRING('u', "uid", &trace.opts.target.uid_str, "user",
A
Arnaldo Carvalho de Melo 已提交
3253
		   "user to profile"),
3254 3255 3256
	OPT_CALLBACK(0, "duration", &trace, "float",
		     "show only events with duration > N.M ms",
		     trace__set_duration),
3257
	OPT_BOOLEAN(0, "sched", &trace.sched, "show blocking scheduler events"),
3258
	OPT_INCR('v', "verbose", &verbose, "be more verbose"),
3259 3260
	OPT_BOOLEAN('T', "time", &trace.full_time,
		    "Show full timestamp, not time relative to first start"),
3261 3262
	OPT_BOOLEAN(0, "failure", &trace.failure_only,
		    "Show only syscalls that failed"),
D
David Ahern 已提交
3263 3264 3265 3266
	OPT_BOOLEAN('s', "summary", &trace.summary_only,
		    "Show only syscall summary with statistics"),
	OPT_BOOLEAN('S', "with-summary", &trace.summary,
		    "Show all syscalls and summary with statistics"),
3267 3268
	OPT_CALLBACK_DEFAULT('F', "pf", &trace.trace_pgfaults, "all|maj|min",
		     "Trace pagefaults", parse_pagefaults, "maj"),
3269
	OPT_BOOLEAN(0, "syscalls", &trace.trace_syscalls, "Trace syscalls"),
3270
	OPT_BOOLEAN('f', "force", &trace.force, "don't complain, do it"),
3271 3272 3273
	OPT_CALLBACK(0, "call-graph", &trace.opts,
		     "record_mode[,record_size]", record_callchain_help,
		     &record_parse_callchain_opt),
3274 3275
	OPT_BOOLEAN(0, "kernel-syscall-graph", &trace.kernel_syscallchains,
		    "Show the kernel callchains on the syscall exit path"),
3276 3277 3278
	OPT_UINTEGER(0, "min-stack", &trace.min_stack,
		     "Set the minimum stack depth when parsing the callchain, "
		     "anything below the specified depth will be ignored."),
3279 3280 3281
	OPT_UINTEGER(0, "max-stack", &trace.max_stack,
		     "Set the maximum stack depth when parsing the callchain, "
		     "anything beyond the specified depth will be ignored. "
3282
		     "Default: kernel.perf_event_max_stack or " __stringify(PERF_MAX_STACK_DEPTH)),
3283 3284
	OPT_BOOLEAN(0, "print-sample", &trace.print_sample,
			"print the PERF_RECORD_SAMPLE PERF_SAMPLE_ info, for debugging"),
3285 3286
	OPT_UINTEGER(0, "proc-map-timeout", &trace.opts.proc_map_timeout,
			"per thread proc mmap processing timeout in ms"),
3287 3288
	OPT_CALLBACK('G', "cgroup", &trace, "name", "monitor event in cgroup name only",
		     trace__parse_cgroups),
A
Alexis Berlemont 已提交
3289 3290 3291
	OPT_UINTEGER('D', "delay", &trace.opts.initial_delay,
		     "ms to wait before starting measurement after program "
		     "start"),
A
Arnaldo Carvalho de Melo 已提交
3292 3293
	OPT_END()
	};
3294
	bool __maybe_unused max_stack_user_set = true;
3295
	bool mmap_pages_user_set = true;
3296
	struct perf_evsel *evsel;
3297
	const char * const trace_subcommands[] = { "record", NULL };
3298
	int err = -1;
3299
	char bf[BUFSIZ];
A
Arnaldo Carvalho de Melo 已提交
3300

3301 3302 3303
	signal(SIGSEGV, sighandler_dump_stack);
	signal(SIGFPE, sighandler_dump_stack);

3304
	trace.evlist = perf_evlist__new();
3305
	trace.sctbl = syscalltbl__new();
3306

3307
	if (trace.evlist == NULL || trace.sctbl == NULL) {
3308
		pr_err("Not enough memory to run!\n");
3309
		err = -ENOMEM;
3310 3311 3312
		goto out;
	}

3313 3314
	argc = parse_options_subcommand(argc, argv, trace_options, trace_subcommands,
				 trace_usage, PARSE_OPT_STOP_AT_NON_OPTION);
D
David Ahern 已提交
3315

3316 3317 3318 3319 3320
	if ((nr_cgroups || trace.cgroup) && !trace.opts.target.system_wide) {
		usage_with_options_msg(trace_usage, trace_options,
				       "cgroup monitoring only available in system-wide mode");
	}

3321 3322 3323
	evsel = bpf__setup_output_event(trace.evlist, "__augmented_syscalls__");
	if (IS_ERR(evsel)) {
		bpf__strerror_setup_output_event(trace.evlist, PTR_ERR(evsel), bf, sizeof(bf));
3324 3325 3326 3327
		pr_err("ERROR: Setup trace syscalls enter failed: %s\n", bf);
		goto out;
	}

3328 3329 3330 3331 3332 3333 3334
	if (evsel) {
		if (perf_evsel__init_augmented_syscall_tp(evsel) ||
		    perf_evsel__init_augmented_syscall_tp_args(evsel))
			goto out;
		trace.syscalls.events.augmented = evsel;
	}

3335 3336 3337 3338 3339 3340 3341
	err = bpf__setup_stdout(trace.evlist);
	if (err) {
		bpf__strerror_setup_stdout(trace.evlist, err, bf, sizeof(bf));
		pr_err("ERROR: Setup BPF stdout failed: %s\n", bf);
		goto out;
	}

3342 3343
	err = -1;

3344 3345 3346 3347 3348
	if (trace.trace_pgfaults) {
		trace.opts.sample_address = true;
		trace.opts.sample_time = true;
	}

3349 3350 3351
	if (trace.opts.mmap_pages == UINT_MAX)
		mmap_pages_user_set = false;

3352
	if (trace.max_stack == UINT_MAX) {
3353
		trace.max_stack = input_name ? PERF_MAX_STACK_DEPTH : sysctl__max_stack();
3354 3355 3356 3357
		max_stack_user_set = false;
	}

#ifdef HAVE_DWARF_UNWIND_SUPPORT
3358
	if ((trace.min_stack || max_stack_user_set) && !callchain_param.enabled) {
3359
		record_opts__parse_callchain(&trace.opts, &callchain_param, "dwarf", false);
3360
	}
3361 3362
#endif

3363
	if (callchain_param.enabled) {
3364 3365 3366
		if (!mmap_pages_user_set && geteuid() == 0)
			trace.opts.mmap_pages = perf_event_mlock_kb_in_pages() * 4;

3367
		symbol_conf.use_callchain = true;
3368
	}
3369

3370
	if (trace.evlist->nr_entries > 0) {
3371
		evlist__set_evsel_handler(trace.evlist, trace__event_handler);
3372 3373 3374 3375 3376
		if (evlist__set_syscall_tp_fields(trace.evlist)) {
			perror("failed to set syscalls:* tracepoint fields");
			goto out;
		}
	}
3377

3378 3379 3380 3381 3382 3383 3384
	if ((argc >= 1) && (strcmp(argv[0], "record") == 0))
		return trace__record(&trace, argc-1, &argv[1]);

	/* summary_only implies summary option, but don't overwrite summary if set */
	if (trace.summary_only)
		trace.summary = trace.summary_only;

3385 3386
	if (!trace.trace_syscalls && !trace.trace_pgfaults &&
	    trace.evlist->nr_entries == 0 /* Was --events used? */) {
3387
		trace.trace_syscalls = true;
3388 3389
	}

3390 3391 3392 3393 3394 3395 3396 3397
	if (output_name != NULL) {
		err = trace__open_output(&trace, output_name);
		if (err < 0) {
			perror("failed to create output file");
			goto out;
		}
	}

3398
	err = target__validate(&trace.opts.target);
3399
	if (err) {
3400
		target__strerror(&trace.opts.target, err, bf, sizeof(bf));
3401 3402
		fprintf(trace.output, "%s", bf);
		goto out_close;
3403 3404
	}

3405
	err = target__parse_uid(&trace.opts.target);
A
Arnaldo Carvalho de Melo 已提交
3406
	if (err) {
3407
		target__strerror(&trace.opts.target, err, bf, sizeof(bf));
3408 3409
		fprintf(trace.output, "%s", bf);
		goto out_close;
A
Arnaldo Carvalho de Melo 已提交
3410 3411
	}

3412
	if (!argc && target__none(&trace.opts.target))
3413 3414
		trace.opts.target.system_wide = true;

3415 3416 3417 3418
	if (input_name)
		err = trace__replay(&trace);
	else
		err = trace__run(&trace, argc, argv);
3419

3420 3421 3422 3423
out_close:
	if (output_name != NULL)
		fclose(trace.output);
out:
3424
	return err;
A
Arnaldo Carvalho de Melo 已提交
3425
}