builtin-trace.c 130 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
/*
 * 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'")
 */

17
#include "util/record.h"
18
#include <traceevent/event-parse.h>
19
#include <api/fs/tracing_path.h>
20
#include <bpf/bpf.h>
21
#include "util/bpf_map.h"
22
#include "util/rlimit.h"
23
#include "builtin.h"
24
#include "util/cgroup.h"
25
#include "util/color.h"
26
#include "util/config.h"
27
#include "util/debug.h"
28
#include "util/dso.h"
29
#include "util/env.h"
30
#include "util/event.h"
31 32
#include "util/evsel.h"
#include "util/evsel_fprintf.h"
33
#include "util/synthetic-events.h"
34
#include "util/evlist.h"
35
#include "util/evswitch.h"
36
#include "util/mmap.h"
37
#include <subcmd/pager.h>
38
#include <subcmd/exec-cmd.h>
39
#include "util/machine.h"
40
#include "util/map.h"
41
#include "util/symbol.h"
42
#include "util/path.h"
43
#include "util/session.h"
44
#include "util/thread.h"
45
#include <subcmd/parse-options.h>
46
#include "util/strlist.h"
47
#include "util/intlist.h"
48
#include "util/thread_map.h"
49
#include "util/stat.h"
50
#include "util/tool.h"
51
#include "util/util.h"
52
#include "trace/beauty/beauty.h"
53
#include "trace-event.h"
54
#include "util/parse-events.h"
55
#include "util/bpf-loader.h"
56
#include "callchain.h"
57
#include "print_binary.h"
58
#include "string2.h"
59
#include "syscalltbl.h"
60
#include "rb_resort.h"
61
#include "../perf.h"
62

63
#include <errno.h>
64
#include <inttypes.h>
65
#include <poll.h>
66
#include <signal.h>
67
#include <stdlib.h>
68
#include <string.h>
69
#include <linux/err.h>
70
#include <linux/filter.h>
71
#include <linux/kernel.h>
72
#include <linux/random.h>
73
#include <linux/stringify.h>
74
#include <linux/time64.h>
75
#include <linux/zalloc.h>
76
#include <fcntl.h>
77
#include <sys/sysmacros.h>
78

79
#include <linux/ctype.h>
80

81 82 83 84
#ifndef O_CLOEXEC
# define O_CLOEXEC		02000000
#endif

85 86 87 88
#ifndef F_LINUX_SPECIFIC_BASE
# define F_LINUX_SPECIFIC_BASE	1024
#endif

89 90
struct trace {
	struct perf_tool	tool;
91
	struct syscalltbl	*sctbl;
92 93
	struct {
		struct syscall  *table;
94
		struct bpf_map  *map;
95 96 97 98
		struct { // per syscall BPF_MAP_TYPE_PROG_ARRAY
			struct bpf_map  *sys_enter,
					*sys_exit;
		}		prog_array;
99
		struct {
100
			struct evsel *sys_enter,
101 102
					  *sys_exit,
					  *augmented;
103
		}		events;
104
		struct bpf_program *unaugmented_prog;
105
	} syscalls;
106 107 108
	struct {
		struct bpf_map *map;
	} dump;
109
	struct record_opts	opts;
110
	struct evlist	*evlist;
111 112
	struct machine		*host;
	struct thread		*current;
113
	struct bpf_object	*bpf_obj;
114
	struct cgroup		*cgroup;
115 116 117
	u64			base_time;
	FILE			*output;
	unsigned long		nr_events;
118 119
	unsigned long		nr_events_printed;
	unsigned long		max_events;
120
	struct evswitch		evswitch;
121 122 123 124 125 126 127 128
	struct strlist		*ev_qualifier;
	struct {
		size_t		nr;
		int		*entries;
	}			ev_qualifier_ids;
	struct {
		size_t		nr;
		pid_t		*entries;
129
		struct bpf_map  *map;
130 131 132 133 134 135 136
	}			filter_pids;
	double			duration_filter;
	double			runtime_ms;
	struct {
		u64		vfs_getname,
				proc_getname;
	} stats;
137
	unsigned int		max_stack;
138
	unsigned int		min_stack;
139
	int			raw_augmented_syscalls_args_size;
140
	bool			raw_augmented_syscalls;
141
	bool			fd_path_disabled;
142
	bool			sort_events;
143 144 145 146 147 148 149
	bool			not_ev_qualifier;
	bool			live;
	bool			full_time;
	bool			sched;
	bool			multiple_threads;
	bool			summary;
	bool			summary_only;
150
	bool			failure_only;
151
	bool			show_comm;
152
	bool			print_sample;
153 154
	bool			show_tool_stats;
	bool			trace_syscalls;
155
	bool			kernel_syscallchains;
156
	s16			args_alignment;
157
	bool			show_tstamp;
158
	bool			show_duration;
159
	bool			show_zeros;
160
	bool			show_arg_names;
161
	bool			show_string_prefix;
162 163 164
	bool			force;
	bool			vfs_getname;
	int			trace_pgfaults;
165
	char			*perfconfig_events;
166 167 168 169
	struct {
		struct ordered_events	data;
		u64			last;
	} oe;
170
};
171

172 173 174 175 176 177 178 179 180 181 182
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) \
{ \
183 184 185
	u##bits value; \
	memcpy(&value, sample->raw_data + field->offset, sizeof(value)); \
	return value;  \
186 187 188 189 190 191 192 193 194 195
}

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) \
{ \
196 197
	u##bits value; \
	memcpy(&value, sample->raw_data + field->offset, sizeof(value)); \
198 199 200 201 202 203 204
	return bswap_##bits(value);\
}

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

205
static int __tp_field__init_uint(struct tp_field *field, int size, int offset, bool needs_swap)
206
{
207
	field->offset = offset;
208

209
	switch (size) {
210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228
	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;
}

229
static int tp_field__init_uint(struct tp_field *field, struct tep_format_field *format_field, bool needs_swap)
230 231 232 233
{
	return __tp_field__init_uint(field, format_field->size, format_field->offset, needs_swap);
}

234 235 236 237 238
static void *tp_field__ptr(struct tp_field *field, struct perf_sample *sample)
{
	return sample->raw_data + field->offset;
}

239
static int __tp_field__init_ptr(struct tp_field *field, int offset)
240
{
241
	field->offset = offset;
242 243 244 245
	field->pointer = tp_field__ptr;
	return 0;
}

246
static int tp_field__init_ptr(struct tp_field *field, struct tep_format_field *format_field)
247 248 249 250
{
	return __tp_field__init_ptr(field, format_field->offset);
}

251 252 253 254 255 256 257
struct syscall_tp {
	struct tp_field id;
	union {
		struct tp_field args, ret;
	};
};

258
static int perf_evsel__init_tp_uint_field(struct evsel *evsel,
259 260 261
					  struct tp_field *field,
					  const char *name)
{
262
	struct tep_format_field *format_field = perf_evsel__field(evsel, name);
263 264 265 266 267 268 269 270 271 272 273

	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); })

274
static int perf_evsel__init_tp_ptr_field(struct evsel *evsel,
275 276 277
					 struct tp_field *field,
					 const char *name)
{
278
	struct tep_format_field *format_field = perf_evsel__field(evsel, name);
279 280 281 282 283 284 285 286 287 288 289

	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); })

290
static void evsel__delete_priv(struct evsel *evsel)
291
{
292
	zfree(&evsel->priv);
293
	evsel__delete(evsel);
294 295
}

296
static int perf_evsel__init_syscall_tp(struct evsel *evsel)
297 298 299 300
{
	struct syscall_tp *sc = evsel->priv = malloc(sizeof(struct syscall_tp));

	if (evsel->priv != NULL) {
301 302
		if (perf_evsel__init_tp_uint_field(evsel, &sc->id, "__syscall_nr") &&
		    perf_evsel__init_tp_uint_field(evsel, &sc->id, "nr"))
303 304 305 306 307 308 309 310 311 312
			goto out_delete;
		return 0;
	}

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

313
static int perf_evsel__init_augmented_syscall_tp(struct evsel *evsel, struct evsel *tp)
314 315 316
{
	struct syscall_tp *sc = evsel->priv = malloc(sizeof(struct syscall_tp));

317 318 319 320 321 322 323
	if (evsel->priv != NULL) {
		struct tep_format_field *syscall_id = perf_evsel__field(tp, "id");
		if (syscall_id == NULL)
			syscall_id = perf_evsel__field(tp, "__syscall_nr");
		if (syscall_id == NULL)
			goto out_delete;
		if (__tp_field__init_uint(&sc->id, syscall_id->size, syscall_id->offset, evsel->needs_swap))
324 325 326 327 328 329 330 331 332 333 334
			goto out_delete;

		return 0;
	}

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

335
static int perf_evsel__init_augmented_syscall_tp_args(struct evsel *evsel)
336 337 338 339 340 341
{
	struct syscall_tp *sc = evsel->priv;

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

342
static int perf_evsel__init_augmented_syscall_tp_ret(struct evsel *evsel)
343 344 345 346 347 348
{
	struct syscall_tp *sc = evsel->priv;

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

349
static int perf_evsel__init_raw_syscall_tp(struct evsel *evsel, void *handler)
350 351 352 353 354 355 356 357 358 359 360 361 362
{
	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:
363
	zfree(&evsel->priv);
364 365 366
	return -ENOENT;
}

367
static struct evsel *perf_evsel__raw_syscall_newtp(const char *direction, void *handler)
368
{
369
	struct evsel *evsel = perf_evsel__newtp("raw_syscalls", direction);
370

371
	/* older kernel (e.g., RHEL6) use syscalls:{enter,exit} */
372
	if (IS_ERR(evsel))
373 374
		evsel = perf_evsel__newtp("syscalls", direction);

375 376 377
	if (IS_ERR(evsel))
		return NULL;

378
	if (perf_evsel__init_raw_syscall_tp(evsel, handler))
379
		goto out_delete;
380 381 382 383

	return evsel;

out_delete:
384
	evsel__delete_priv(evsel);
385 386 387 388 389 390 391 392 393 394 395
	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); })

396
size_t strarray__scnprintf(struct strarray *sa, char *bf, size_t size, const char *intfmt, bool show_prefix, int val)
397 398
{
	int idx = val - sa->offset;
399

400 401 402 403 404 405
	if (idx < 0 || idx >= sa->nr_entries || sa->entries[idx] == NULL) {
		size_t printed = scnprintf(bf, size, intfmt, val);
		if (show_prefix)
			printed += scnprintf(bf + printed, size - printed, " /* %s??? */", sa->prefix);
		return printed;
	}
406

407
	return scnprintf(bf, size, "%s%s", show_prefix ? sa->prefix : "", sa->entries[idx]);
408 409
}

410 411 412
static size_t __syscall_arg__scnprintf_strarray(char *bf, size_t size,
						const char *intfmt,
					        struct syscall_arg *arg)
413
{
414
	return strarray__scnprintf(arg->parm, bf, size, intfmt, arg->show_string_prefix, arg->val);
415 416
}

417 418 419 420 421 422
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);
}

423 424
#define SCA_STRARRAY syscall_arg__scnprintf_strarray

425 426 427 428 429
size_t syscall_arg__scnprintf_strarray_flags(char *bf, size_t size, struct syscall_arg *arg)
{
	return strarray__scnprintf_flags(arg->parm, bf, size, arg->show_string_prefix, arg->val);
}

430
size_t strarrays__scnprintf(struct strarrays *sas, char *bf, size_t size, const char *intfmt, bool show_prefix, int val)
431
{
432
	size_t printed;
433 434 435 436
	int i;

	for (i = 0; i < sas->nr_entries; ++i) {
		struct strarray *sa = sas->entries[i];
437
		int idx = val - sa->offset;
438 439 440 441

		if (idx >= 0 && idx < sa->nr_entries) {
			if (sa->entries[idx] == NULL)
				break;
442
			return scnprintf(bf, size, "%s%s", show_prefix ? sa->prefix : "", sa->entries[idx]);
443 444 445
		}
	}

446 447 448 449
	printed = scnprintf(bf, size, intfmt, val);
	if (show_prefix)
		printed += scnprintf(bf + printed, size - printed, " /* %s??? */", sas->entries[0]->prefix);
	return printed;
450 451 452 453 454 455
}

size_t syscall_arg__scnprintf_strarrays(char *bf, size_t size,
					struct syscall_arg *arg)
{
	return strarrays__scnprintf(arg->parm, bf, size, "%d", arg->show_string_prefix, arg->val);
456 457
}

458 459 460 461
#ifndef AT_FDCWD
#define AT_FDCWD	-100
#endif

462 463 464 465
static size_t syscall_arg__scnprintf_fd_at(char *bf, size_t size,
					   struct syscall_arg *arg)
{
	int fd = arg->val;
466
	const char *prefix = "AT_FD";
467 468

	if (fd == AT_FDCWD)
469
		return scnprintf(bf, size, "%s%s", arg->show_string_prefix ? prefix : "", "CWD");
470 471 472 473 474 475 476 477 478 479 480

	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

481
size_t syscall_arg__scnprintf_hex(char *bf, size_t size, struct syscall_arg *arg)
482
{
483
	return scnprintf(bf, size, "%#lx", arg->val);
484 485
}

486 487 488 489 490 491 492
size_t syscall_arg__scnprintf_ptr(char *bf, size_t size, struct syscall_arg *arg)
{
	if (arg->val == 0)
		return scnprintf(bf, size, "NULL");
	return syscall_arg__scnprintf_hex(bf, size, arg);
}

493
size_t syscall_arg__scnprintf_int(char *bf, size_t size, struct syscall_arg *arg)
494 495 496 497
{
	return scnprintf(bf, size, "%d", arg->val);
}

498 499 500 501 502
size_t syscall_arg__scnprintf_long(char *bf, size_t size, struct syscall_arg *arg)
{
	return scnprintf(bf, size, "%ld", arg->val);
}

503 504 505 506
static const char *bpf_cmd[] = {
	"MAP_CREATE", "MAP_LOOKUP_ELEM", "MAP_UPDATE_ELEM", "MAP_DELETE_ELEM",
	"MAP_GET_NEXT_KEY", "PROG_LOAD",
};
507
static DEFINE_STRARRAY(bpf_cmd, "BPF_");
508

509 510 511 512 513
static const char *fsmount_flags[] = {
	[1] = "CLOEXEC",
};
static DEFINE_STRARRAY(fsmount_flags, "FSMOUNT_");

514 515 516 517
#include "trace/beauty/generated/fsconfig_arrays.c"

static DEFINE_STRARRAY(fsconfig_cmds, "FSCONFIG_");

518
static const char *epoll_ctl_ops[] = { "ADD", "DEL", "MOD", };
519
static DEFINE_STRARRAY_OFFSET(epoll_ctl_ops, "EPOLL_CTL_", 1);
520

521
static const char *itimers[] = { "REAL", "VIRTUAL", "PROF", };
522
static DEFINE_STRARRAY(itimers, "ITIMER_");
523

524 525 526 527 528 529 530
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",
};
531
static DEFINE_STRARRAY(keyctl_options, "KEYCTL_");
532

533 534 535 536 537 538 539 540
static const char *whences[] = { "SET", "CUR", "END",
#ifdef SEEK_DATA
"DATA",
#endif
#ifdef SEEK_HOLE
"HOLE",
#endif
};
541
static DEFINE_STRARRAY(whences, "SEEK_");
542

543 544
static const char *fcntl_cmds[] = {
	"DUPFD", "GETFD", "SETFD", "GETFL", "SETFL", "GETLK", "SETLK",
545 546 547
	"SETLKW", "SETOWN", "GETOWN", "SETSIG", "GETSIG", "GETLK64",
	"SETLK64", "SETLKW64", "SETOWN_EX", "GETOWN_EX",
	"GETOWNER_UIDS",
548
};
549
static DEFINE_STRARRAY(fcntl_cmds, "F_");
550

551 552 553
static const char *fcntl_linux_specific_cmds[] = {
	"SETLEASE", "GETLEASE", "NOTIFY", [5] =	"CANCELLK", "DUPFD_CLOEXEC",
	"SETPIPE_SZ", "GETPIPE_SZ", "ADD_SEALS", "GET_SEALS",
554
	"GET_RW_HINT", "SET_RW_HINT", "GET_FILE_RW_HINT", "SET_FILE_RW_HINT",
555 556
};

557
static DEFINE_STRARRAY_OFFSET(fcntl_linux_specific_cmds, "F_", F_LINUX_SPECIFIC_BASE);
558 559 560 561 562 563 564 565

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

static DEFINE_STRARRAYS(fcntl_cmds_arrays);

566 567 568 569 570
static const char *rlimit_resources[] = {
	"CPU", "FSIZE", "DATA", "STACK", "CORE", "RSS", "NPROC", "NOFILE",
	"MEMLOCK", "AS", "LOCKS", "SIGPENDING", "MSGQUEUE", "NICE", "RTPRIO",
	"RTTIME",
};
571
static DEFINE_STRARRAY(rlimit_resources, "RLIMIT_");
572

573
static const char *sighow[] = { "BLOCK", "UNBLOCK", "SETMASK", };
574
static DEFINE_STRARRAY(sighow, "SIG_");
575

576 577
static const char *clockid[] = {
	"REALTIME", "MONOTONIC", "PROCESS_CPUTIME_ID", "THREAD_CPUTIME_ID",
578 579
	"MONOTONIC_RAW", "REALTIME_COARSE", "MONOTONIC_COARSE", "BOOTTIME",
	"REALTIME_ALARM", "BOOTTIME_ALARM", "SGI_CYCLE", "TAI"
580
};
581
static DEFINE_STRARRAY(clockid, "CLOCK_");
582

583 584 585
static size_t syscall_arg__scnprintf_access_mode(char *bf, size_t size,
						 struct syscall_arg *arg)
{
586 587
	bool show_prefix = arg->show_string_prefix;
	const char *suffix = "_OK";
588 589 590 591
	size_t printed = 0;
	int mode = arg->val;

	if (mode == F_OK) /* 0 */
592
		return scnprintf(bf, size, "F%s", show_prefix ? suffix : "");
593 594
#define	P_MODE(n) \
	if (mode & n##_OK) { \
595
		printed += scnprintf(bf + printed, size - printed, "%s%s", #n, show_prefix ? suffix : ""); \
596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611
		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

612 613 614 615 616
static size_t syscall_arg__scnprintf_filename(char *bf, size_t size,
					      struct syscall_arg *arg);

#define SCA_FILENAME syscall_arg__scnprintf_filename

617 618 619
static size_t syscall_arg__scnprintf_pipe_flags(char *bf, size_t size,
						struct syscall_arg *arg)
{
620 621
	bool show_prefix = arg->show_string_prefix;
	const char *prefix = "O_";
622 623 624 625
	int printed = 0, flags = arg->val;

#define	P_FLAG(n) \
	if (flags & O_##n) { \
626
		printed += scnprintf(bf + printed, size - printed, "%s%s%s", printed ? "|" : "", show_prefix ? prefix : "", #n); \
627 628 629 630 631 632 633 634 635 636 637 638 639 640 641
		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

642 643 644 645 646 647 648
#ifndef GRND_NONBLOCK
#define GRND_NONBLOCK	0x0001
#endif
#ifndef GRND_RANDOM
#define GRND_RANDOM	0x0002
#endif

649 650 651
static size_t syscall_arg__scnprintf_getrandom_flags(char *bf, size_t size,
						   struct syscall_arg *arg)
{
652 653
	bool show_prefix = arg->show_string_prefix;
	const char *prefix = "GRND_";
654 655 656 657
	int printed = 0, flags = arg->val;

#define	P_FLAG(n) \
	if (flags & GRND_##n) { \
658
		printed += scnprintf(bf + printed, size - printed, "%s%s%s", printed ? "|" : "", show_prefix ? prefix : "", #n); \
659 660 661 662 663 664 665 666 667 668 669 670 671 672 673
		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

674 675 676
#define STRARRAY(name, array) \
	  { .scnprintf	= SCA_STRARRAY, \
	    .parm	= &strarray__##array, }
677

678 679 680 681
#define STRARRAY_FLAGS(name, array) \
	  { .scnprintf	= SCA_STRARRAY_FLAGS, \
	    .parm	= &strarray__##array, }

682
#include "trace/beauty/arch_errno_names.c"
683
#include "trace/beauty/eventfd.c"
684
#include "trace/beauty/futex_op.c"
685
#include "trace/beauty/futex_val3.c"
686
#include "trace/beauty/mmap.c"
687
#include "trace/beauty/mode_t.c"
688
#include "trace/beauty/msg_flags.c"
689
#include "trace/beauty/open_flags.c"
690
#include "trace/beauty/perf_event_open.c"
691
#include "trace/beauty/pid.c"
692
#include "trace/beauty/sched_policy.c"
693
#include "trace/beauty/seccomp.c"
694
#include "trace/beauty/signum.c"
695
#include "trace/beauty/socket_type.c"
696
#include "trace/beauty/waitid_options.c"
697

698 699
struct syscall_arg_fmt {
	size_t	   (*scnprintf)(char *bf, size_t size, struct syscall_arg *arg);
700
	unsigned long (*mask_val)(struct syscall_arg *arg, unsigned long val);
701
	void	   *parm;
702
	const char *name;
703
	bool	   show_zero;
704 705
};

706
struct syscall_fmt {
707
	const char *name;
708
	const char *alias;
709 710 711 712
	struct {
		const char *sys_enter,
			   *sys_exit;
	}	   bpf_prog_name;
713
	struct syscall_arg_fmt arg[6];
714
	u8	   nr_args;
715
	bool	   errpid;
716
	bool	   timeout;
717
	bool	   hexret;
718 719 720
};

static struct syscall_fmt syscall_fmts[] = {
721
	{ .name	    = "access",
722
	  .arg = { [1] = { .scnprintf = SCA_ACCMODE,  /* mode */ }, }, },
723 724 725
	{ .name	    = "arch_prctl",
	  .arg = { [0] = { .scnprintf = SCA_X86_ARCH_PRCTL_CODE, /* code */ },
		   [1] = { .scnprintf = SCA_PTR, /* arg2 */ }, }, },
726
	{ .name	    = "bind",
727
	  .arg = { [0] = { .scnprintf = SCA_INT, /* fd */ },
728 729
		   [1] = { .scnprintf = SCA_SOCKADDR, /* umyaddr */ },
		   [2] = { .scnprintf = SCA_INT, /* addrlen */ }, }, },
730
	{ .name	    = "bpf",
731
	  .arg = { [0] = STRARRAY(cmd, bpf_cmd), }, },
732
	{ .name	    = "brk",	    .hexret = true,
733
	  .arg = { [0] = { .scnprintf = SCA_PTR, /* brk */ }, }, },
734
	{ .name     = "clock_gettime",
735
	  .arg = { [0] = STRARRAY(clk_id, clockid), }, },
736 737 738 739 740 741
	{ .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, }, }, },
742
	{ .name	    = "close",
743
	  .arg = { [0] = { .scnprintf = SCA_CLOSE_FD, /* fd */ }, }, },
744
	{ .name	    = "connect",
745 746
	  .arg = { [0] = { .scnprintf = SCA_INT, /* fd */ },
		   [1] = { .scnprintf = SCA_SOCKADDR, /* servaddr */ },
747
		   [2] = { .scnprintf = SCA_INT, /* addrlen */ }, }, },
748
	{ .name	    = "epoll_ctl",
749
	  .arg = { [1] = STRARRAY(op, epoll_ctl_ops), }, },
750
	{ .name	    = "eventfd2",
751
	  .arg = { [1] = { .scnprintf = SCA_EFD_FLAGS, /* flags */ }, }, },
752
	{ .name	    = "fchmodat",
753
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* fd */ }, }, },
754
	{ .name	    = "fchownat",
755
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* fd */ }, }, },
756
	{ .name	    = "fcntl",
757
	  .arg = { [1] = { .scnprintf = SCA_FCNTL_CMD, /* cmd */
758 759
			   .parm      = &strarrays__fcntl_cmds_arrays,
			   .show_zero = true, },
760
		   [2] = { .scnprintf =  SCA_FCNTL_ARG, /* arg */ }, }, },
761
	{ .name	    = "flock",
762
	  .arg = { [1] = { .scnprintf = SCA_FLOCK, /* cmd */ }, }, },
763 764
	{ .name     = "fsconfig",
	  .arg = { [1] = STRARRAY(cmd, fsconfig_cmds), }, },
765 766 767
	{ .name     = "fsmount",
	  .arg = { [1] = STRARRAY_FLAGS(flags, fsmount_flags),
		   [2] = { .scnprintf = SCA_FSMOUNT_ATTR_FLAGS, /* attr_flags */ }, }, },
768 769 770 771
	{ .name     = "fspick",
	  .arg = { [0] = { .scnprintf = SCA_FDAT,	  /* dfd */ },
		   [1] = { .scnprintf = SCA_FILENAME,	  /* path */ },
		   [2] = { .scnprintf = SCA_FSPICK_FLAGS, /* flags */ }, }, },
772 773 774
	{ .name	    = "fstat", .alias = "newfstat", },
	{ .name	    = "fstatat", .alias = "newfstatat", },
	{ .name	    = "futex",
775 776
	  .arg = { [1] = { .scnprintf = SCA_FUTEX_OP, /* op */ },
		   [5] = { .scnprintf = SCA_FUTEX_VAL3, /* val3 */ }, }, },
777
	{ .name	    = "futimesat",
778
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* fd */ }, }, },
779
	{ .name	    = "getitimer",
780
	  .arg = { [0] = STRARRAY(which, itimers), }, },
781
	{ .name	    = "getpid",	    .errpid = true, },
782
	{ .name	    = "getpgid",    .errpid = true, },
783
	{ .name	    = "getppid",    .errpid = true, },
784
	{ .name	    = "getrandom",
785
	  .arg = { [2] = { .scnprintf = SCA_GETRANDOM_FLAGS, /* flags */ }, }, },
786
	{ .name	    = "getrlimit",
787
	  .arg = { [0] = STRARRAY(resource, rlimit_resources), }, },
788
	{ .name	    = "gettid",	    .errpid = true, },
789
	{ .name	    = "ioctl",
790
	  .arg = {
791 792 793 794
#if defined(__i386__) || defined(__x86_64__)
/*
 * FIXME: Make this available to all arches.
 */
795
		   [1] = { .scnprintf = SCA_IOCTL_CMD, /* cmd */ },
796
		   [2] = { .scnprintf = SCA_HEX, /* arg */ }, }, },
797
#else
798
		   [2] = { .scnprintf = SCA_HEX, /* arg */ }, }, },
799
#endif
800 801 802 803 804 805
	{ .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, }, }, },
806
	{ .name	    = "keyctl",
807
	  .arg = { [0] = STRARRAY(option, keyctl_options), }, },
808
	{ .name	    = "kill",
809
	  .arg = { [1] = { .scnprintf = SCA_SIGNUM, /* sig */ }, }, },
810
	{ .name	    = "linkat",
811
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* fd */ }, }, },
812
	{ .name	    = "lseek",
813
	  .arg = { [2] = STRARRAY(whence, whences), }, },
814 815
	{ .name	    = "lstat", .alias = "newlstat", },
	{ .name     = "madvise",
816 817
	  .arg = { [0] = { .scnprintf = SCA_HEX,      /* start */ },
		   [2] = { .scnprintf = SCA_MADV_BHV, /* behavior */ }, }, },
818
	{ .name	    = "mkdirat",
819
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* fd */ }, }, },
820
	{ .name	    = "mknodat",
821
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* fd */ }, }, },
822
	{ .name	    = "mmap",	    .hexret = true,
823 824 825 826
/* The standard mmap maps to old_mmap on s390x */
#if defined(__s390x__)
	.alias = "old_mmap",
#endif
827
	  .arg = { [2] = { .scnprintf = SCA_MMAP_PROT,	/* prot */ },
828 829
		   [3] = { .scnprintf = SCA_MMAP_FLAGS,	/* flags */ },
		   [5] = { .scnprintf = SCA_HEX,	/* offset */ }, }, },
830
	{ .name	    = "mount",
831 832
	  .arg = { [0] = { .scnprintf = SCA_FILENAME, /* dev_name */ },
		   [3] = { .scnprintf = SCA_MOUNT_FLAGS, /* flags */
833
			   .mask_val  = SCAMV_MOUNT_FLAGS, /* flags */ }, }, },
834 835 836 837 838 839
	{ .name	    = "move_mount",
	  .arg = { [0] = { .scnprintf = SCA_FDAT,	/* from_dfd */ },
		   [1] = { .scnprintf = SCA_FILENAME, /* from_pathname */ },
		   [2] = { .scnprintf = SCA_FDAT,	/* to_dfd */ },
		   [3] = { .scnprintf = SCA_FILENAME, /* to_pathname */ },
		   [4] = { .scnprintf = SCA_MOVE_MOUNT_FLAGS, /* flags */ }, }, },
840
	{ .name	    = "mprotect",
841 842
	  .arg = { [0] = { .scnprintf = SCA_HEX,	/* start */ },
		   [2] = { .scnprintf = SCA_MMAP_PROT,	/* prot */ }, }, },
843
	{ .name	    = "mq_unlink",
844
	  .arg = { [0] = { .scnprintf = SCA_FILENAME, /* u_name */ }, }, },
845
	{ .name	    = "mremap",	    .hexret = true,
846
	  .arg = { [3] = { .scnprintf = SCA_MREMAP_FLAGS, /* flags */ }, }, },
847
	{ .name	    = "name_to_handle_at",
848
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* dfd */ }, }, },
849
	{ .name	    = "newfstatat",
850
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* dfd */ }, }, },
851
	{ .name	    = "open",
852
	  .arg = { [1] = { .scnprintf = SCA_OPEN_FLAGS, /* flags */ }, }, },
853
	{ .name	    = "open_by_handle_at",
854 855
	  .arg = { [0] = { .scnprintf = SCA_FDAT,	/* dfd */ },
		   [2] = { .scnprintf = SCA_OPEN_FLAGS, /* flags */ }, }, },
856
	{ .name	    = "openat",
857 858
	  .arg = { [0] = { .scnprintf = SCA_FDAT,	/* dfd */ },
		   [2] = { .scnprintf = SCA_OPEN_FLAGS, /* flags */ }, }, },
859
	{ .name	    = "perf_event_open",
860 861 862
	  .arg = { [2] = { .scnprintf = SCA_INT,	/* cpu */ },
		   [3] = { .scnprintf = SCA_FD,		/* group_fd */ },
		   [4] = { .scnprintf = SCA_PERF_FLAGS, /* flags */ }, }, },
863
	{ .name	    = "pipe2",
864
	  .arg = { [1] = { .scnprintf = SCA_PIPE_FLAGS, /* flags */ }, }, },
865 866 867 868 869 870 871 872
	{ .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 */ }, }, },
873 874
	{ .name	    = "poll", .timeout = true, },
	{ .name	    = "ppoll", .timeout = true, },
875
	{ .name	    = "prctl",
876 877 878
	  .arg = { [0] = { .scnprintf = SCA_PRCTL_OPTION, /* option */ },
		   [1] = { .scnprintf = SCA_PRCTL_ARG2, /* arg2 */ },
		   [2] = { .scnprintf = SCA_PRCTL_ARG3, /* arg3 */ }, }, },
879 880 881
	{ .name	    = "pread", .alias = "pread64", },
	{ .name	    = "preadv", .alias = "pread", },
	{ .name	    = "prlimit64",
882
	  .arg = { [1] = STRARRAY(resource, rlimit_resources), }, },
883 884
	{ .name	    = "pwrite", .alias = "pwrite64", },
	{ .name	    = "readlinkat",
885
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* dfd */ }, }, },
886
	{ .name	    = "recvfrom",
887
	  .arg = { [3] = { .scnprintf = SCA_MSG_FLAGS, /* flags */ }, }, },
888
	{ .name	    = "recvmmsg",
889
	  .arg = { [3] = { .scnprintf = SCA_MSG_FLAGS, /* flags */ }, }, },
890
	{ .name	    = "recvmsg",
891
	  .arg = { [2] = { .scnprintf = SCA_MSG_FLAGS, /* flags */ }, }, },
892
	{ .name	    = "renameat",
893 894
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* olddirfd */ },
		   [2] = { .scnprintf = SCA_FDAT, /* newdirfd */ }, }, },
895 896
	{ .name	    = "renameat2",
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* olddirfd */ },
897 898
		   [2] = { .scnprintf = SCA_FDAT, /* newdirfd */ },
		   [4] = { .scnprintf = SCA_RENAMEAT2_FLAGS, /* flags */ }, }, },
899
	{ .name	    = "rt_sigaction",
900
	  .arg = { [0] = { .scnprintf = SCA_SIGNUM, /* sig */ }, }, },
901
	{ .name	    = "rt_sigprocmask",
902
	  .arg = { [0] = STRARRAY(how, sighow), }, },
903
	{ .name	    = "rt_sigqueueinfo",
904
	  .arg = { [1] = { .scnprintf = SCA_SIGNUM, /* sig */ }, }, },
905
	{ .name	    = "rt_tgsigqueueinfo",
906
	  .arg = { [2] = { .scnprintf = SCA_SIGNUM, /* sig */ }, }, },
907
	{ .name	    = "sched_setscheduler",
908
	  .arg = { [1] = { .scnprintf = SCA_SCHED_POLICY, /* policy */ }, }, },
909
	{ .name	    = "seccomp",
910 911
	  .arg = { [0] = { .scnprintf = SCA_SECCOMP_OP,	   /* op */ },
		   [1] = { .scnprintf = SCA_SECCOMP_FLAGS, /* flags */ }, }, },
912
	{ .name	    = "select", .timeout = true, },
913
	{ .name	    = "sendfile", .alias = "sendfile64", },
914
	{ .name	    = "sendmmsg",
915
	  .arg = { [3] = { .scnprintf = SCA_MSG_FLAGS, /* flags */ }, }, },
916
	{ .name	    = "sendmsg",
917
	  .arg = { [2] = { .scnprintf = SCA_MSG_FLAGS, /* flags */ }, }, },
918
	{ .name	    = "sendto",
919 920
	  .arg = { [3] = { .scnprintf = SCA_MSG_FLAGS, /* flags */ },
		   [4] = { .scnprintf = SCA_SOCKADDR, /* addr */ }, }, },
921
	{ .name	    = "set_tid_address", .errpid = true, },
922
	{ .name	    = "setitimer",
923
	  .arg = { [0] = STRARRAY(which, itimers), }, },
924
	{ .name	    = "setrlimit",
925
	  .arg = { [0] = STRARRAY(resource, rlimit_resources), }, },
926
	{ .name	    = "socket",
927
	  .arg = { [0] = STRARRAY(family, socket_families),
928 929
		   [1] = { .scnprintf = SCA_SK_TYPE, /* type */ },
		   [2] = { .scnprintf = SCA_SK_PROTO, /* protocol */ }, }, },
930
	{ .name	    = "socketpair",
931
	  .arg = { [0] = STRARRAY(family, socket_families),
932 933
		   [1] = { .scnprintf = SCA_SK_TYPE, /* type */ },
		   [2] = { .scnprintf = SCA_SK_PROTO, /* protocol */ }, }, },
934 935
	{ .name	    = "stat", .alias = "newstat", },
	{ .name	    = "statx",
936 937 938
	  .arg = { [0] = { .scnprintf = SCA_FDAT,	 /* fdat */ },
		   [2] = { .scnprintf = SCA_STATX_FLAGS, /* flags */ } ,
		   [3] = { .scnprintf = SCA_STATX_MASK,	 /* mask */ }, }, },
939
	{ .name	    = "swapoff",
940
	  .arg = { [0] = { .scnprintf = SCA_FILENAME, /* specialfile */ }, }, },
941
	{ .name	    = "swapon",
942
	  .arg = { [0] = { .scnprintf = SCA_FILENAME, /* specialfile */ }, }, },
943
	{ .name	    = "symlinkat",
944
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* dfd */ }, }, },
945 946
	{ .name	    = "sync_file_range",
	  .arg = { [3] = { .scnprintf = SCA_SYNC_FILE_RANGE_FLAGS, /* flags */ }, }, },
947
	{ .name	    = "tgkill",
948
	  .arg = { [2] = { .scnprintf = SCA_SIGNUM, /* sig */ }, }, },
949
	{ .name	    = "tkill",
950
	  .arg = { [1] = { .scnprintf = SCA_SIGNUM, /* sig */ }, }, },
951 952
	{ .name     = "umount2", .alias = "umount",
	  .arg = { [0] = { .scnprintf = SCA_FILENAME, /* name */ }, }, },
953 954
	{ .name	    = "uname", .alias = "newuname", },
	{ .name	    = "unlinkat",
955
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* dfd */ }, }, },
956
	{ .name	    = "utimensat",
957
	  .arg = { [0] = { .scnprintf = SCA_FDAT, /* dirfd */ }, }, },
958
	{ .name	    = "wait4",	    .errpid = true,
959
	  .arg = { [2] = { .scnprintf = SCA_WAITID_OPTIONS, /* options */ }, }, },
960
	{ .name	    = "waitid",	    .errpid = true,
961
	  .arg = { [3] = { .scnprintf = SCA_WAITID_OPTIONS, /* options */ }, }, },
962 963 964 965 966 967 968 969
};

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

970 971 972 973 974
static struct syscall_fmt *__syscall_fmt__find(struct syscall_fmt *fmts, const int nmemb, const char *name)
{
	return bsearch(name, fmts, nmemb, sizeof(struct syscall_fmt), syscall_fmt__cmp);
}

975 976 977
static struct syscall_fmt *syscall_fmt__find(const char *name)
{
	const int nmemb = ARRAY_SIZE(syscall_fmts);
978
	return __syscall_fmt__find(syscall_fmts, nmemb, name);
979 980
}

981
static struct syscall_fmt *__syscall_fmt__find_by_alias(struct syscall_fmt *fmts, const int nmemb, const char *alias)
982
{
983
	int i;
984 985

	for (i = 0; i < nmemb; ++i) {
986 987
		if (fmts[i].alias && strcmp(fmts[i].alias, alias) == 0)
			return &fmts[i];
988 989 990 991 992
	}

	return NULL;
}

993 994 995 996 997 998
static struct syscall_fmt *syscall_fmt__find_by_alias(const char *alias)
{
	const int nmemb = ARRAY_SIZE(syscall_fmts);
	return __syscall_fmt__find_by_alias(syscall_fmts, nmemb, alias);
}

999 1000 1001
/*
 * 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.
1002
 * args_size: sum of the sizes of the syscall arguments, anything after that is augmented stuff: pathname for openat, etc.
1003
 * nonexistent: Just a hole in the syscall table, syscall id not allocated
1004
 */
1005
struct syscall {
1006
	struct tep_event    *tp_format;
1007
	int		    nr_args;
1008
	int		    args_size;
1009 1010 1011 1012
	struct {
		struct bpf_program *sys_enter,
				   *sys_exit;
	}		    bpf_prog;
1013 1014
	bool		    is_exit;
	bool		    is_open;
1015
	bool		    nonexistent;
1016
	struct tep_format_field *args;
1017 1018
	const char	    *name;
	struct syscall_fmt  *fmt;
1019
	struct syscall_arg_fmt *arg_fmt;
1020 1021
};

1022 1023 1024 1025 1026
/*
 * Must match what is in the BPF program:
 *
 * tools/perf/examples/bpf/augmented_raw_syscalls.c
 */
1027 1028
struct bpf_map_syscall_entry {
	bool	enabled;
1029
	u16	string_args_len[6];
1030 1031
};

1032 1033 1034 1035 1036 1037 1038 1039
/*
 * 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)
1040 1041 1042 1043
{
	double duration = (double)t / NSEC_PER_MSEC;
	size_t printed = fprintf(fp, "(");

1044
	if (!calculated)
1045
		printed += fprintf(fp, "         ");
1046
	else if (duration >= 1.0)
1047 1048 1049 1050 1051
		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);
1052
	return printed + fprintf(fp, "): ");
1053 1054
}

1055 1056 1057 1058
/**
 * 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.
1059 1060
 * ret_scnprintf: syscall args may set this to a different syscall return
 *                formatter, for instance, fcntl may return fds, file flags, etc.
1061
 */
1062 1063 1064
struct thread_trace {
	u64		  entry_time;
	bool		  entry_pending;
1065
	unsigned long	  nr_events;
1066
	unsigned long	  pfmaj, pfmin;
1067
	char		  *entry_str;
1068
	double		  runtime_ms;
1069
	size_t		  (*ret_scnprintf)(char *bf, size_t size, struct syscall_arg *arg);
1070 1071
        struct {
		unsigned long ptr;
1072 1073 1074 1075
		short int     entry_str_pos;
		bool	      pending_open;
		unsigned int  namelen;
		char	      *name;
1076
	} filename;
1077
	struct {
1078 1079 1080
		int	      max;
		struct file   *table;
	} files;
1081 1082

	struct intlist *syscall_stats;
1083 1084 1085 1086
};

static struct thread_trace *thread_trace__new(void)
{
1087 1088
	struct thread_trace *ttrace =  zalloc(sizeof(struct thread_trace));

1089
	if (ttrace) {
1090
		ttrace->files.max = -1;
1091 1092
		ttrace->syscall_stats = intlist__new(NULL);
	}
1093

1094
	return ttrace;
1095 1096
}

1097
static struct thread_trace *thread__trace(struct thread *thread, FILE *fp)
1098
{
1099 1100
	struct thread_trace *ttrace;

1101 1102 1103
	if (thread == NULL)
		goto fail;

1104 1105
	if (thread__priv(thread) == NULL)
		thread__set_priv(thread, thread_trace__new());
1106

1107
	if (thread__priv(thread) == NULL)
1108 1109
		goto fail;

1110
	ttrace = thread__priv(thread);
1111 1112 1113
	++ttrace->nr_events;

	return ttrace;
1114
fail:
1115
	color_fprintf(fp, PERF_COLOR_RED,
1116 1117 1118 1119
		      "WARNING: not enough memory, dropping samples!\n");
	return NULL;
}

1120 1121

void syscall_arg__set_ret_scnprintf(struct syscall_arg *arg,
1122
				    size_t (*ret_scnprintf)(char *bf, size_t size, struct syscall_arg *arg))
1123 1124 1125 1126 1127 1128
{
	struct thread_trace *ttrace = thread__priv(arg->thread);

	ttrace->ret_scnprintf = ret_scnprintf;
}

1129 1130 1131
#define TRACE_PFMAJ		(1 << 0)
#define TRACE_PFMIN		(1 << 1)

1132 1133
static const size_t trace__entry_str_size = 2048;

1134
static struct file *thread_trace__files_entry(struct thread_trace *ttrace, int fd)
1135
{
1136 1137 1138
	if (fd < 0)
		return NULL;

1139 1140
	if (fd > ttrace->files.max) {
		struct file *nfiles = realloc(ttrace->files.table, (fd + 1) * sizeof(struct file));
1141

1142
		if (nfiles == NULL)
1143
			return NULL;
1144

1145 1146 1147
		if (ttrace->files.max != -1) {
			memset(nfiles + ttrace->files.max + 1, 0,
			       (fd - ttrace->files.max) * sizeof(struct file));
1148
		} else {
1149
			memset(nfiles, 0, (fd + 1) * sizeof(struct file));
1150 1151
		}

1152 1153
		ttrace->files.table = nfiles;
		ttrace->files.max   = fd;
1154 1155
	}

1156 1157 1158
	return ttrace->files.table + fd;
}

1159 1160 1161 1162 1163
struct file *thread__files_entry(struct thread *thread, int fd)
{
	return thread_trace__files_entry(thread__priv(thread), fd);
}

1164 1165 1166 1167 1168 1169
static int trace__set_fd_pathname(struct thread *thread, int fd, const char *pathname)
{
	struct thread_trace *ttrace = thread__priv(thread);
	struct file *file = thread_trace__files_entry(ttrace, fd);

	if (file != NULL) {
1170 1171 1172
		struct stat st;
		if (stat(pathname, &st) == 0)
			file->dev_maj = major(st.st_rdev);
1173 1174 1175 1176
		file->pathname = strdup(pathname);
		if (file->pathname)
			return 0;
	}
1177

1178
	return -1;
1179 1180
}

1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206
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);
}

1207 1208
static const char *thread__fd_path(struct thread *thread, int fd,
				   struct trace *trace)
1209
{
1210
	struct thread_trace *ttrace = thread__priv(thread);
1211

1212
	if (ttrace == NULL || trace->fd_path_disabled)
1213 1214 1215 1216 1217
		return NULL;

	if (fd < 0)
		return NULL;

1218
	if ((fd > ttrace->files.max || ttrace->files.table[fd].pathname == NULL)) {
1219 1220 1221
		if (!trace->live)
			return NULL;
		++trace->stats.proc_getname;
1222
		if (thread__read_fd_path(thread, fd))
1223 1224
			return NULL;
	}
1225

1226
	return ttrace->files.table[fd].pathname;
1227 1228
}

1229
size_t syscall_arg__scnprintf_fd(char *bf, size_t size, struct syscall_arg *arg)
1230 1231 1232
{
	int fd = arg->val;
	size_t printed = scnprintf(bf, size, "%d", fd);
1233
	const char *path = thread__fd_path(arg->thread, fd, arg->trace);
1234 1235 1236 1237 1238 1239 1240

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

	return printed;
}

1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257
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;
}

1258 1259 1260 1261 1262
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);
1263
	struct thread_trace *ttrace = thread__priv(arg->thread);
1264

1265 1266
	if (ttrace && fd >= 0 && fd <= ttrace->files.max)
		zfree(&ttrace->files.table[fd].pathname);
1267 1268 1269 1270

	return printed;
}

1271 1272 1273 1274 1275 1276 1277 1278 1279
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;
}

1280 1281 1282
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;
1283 1284 1285 1286 1287 1288
	size_t printed = scnprintf(bf, size, "\"%.*s\"", augmented_arg->size, augmented_arg->value);
	/*
	 * So that the next arg with a payload can consume its augmented arg, i.e. for rename* syscalls
	 * we would have two strings, each prefixed by its size.
	 */
	int consumed = sizeof(*augmented_arg) + augmented_arg->size;
1289

1290
	arg->augmented.args = ((void *)arg->augmented.args) + consumed;
1291
	arg->augmented.size -= consumed;
1292

1293
	return printed;
1294 1295
}

1296 1297 1298 1299 1300
static size_t syscall_arg__scnprintf_filename(char *bf, size_t size,
					      struct syscall_arg *arg)
{
	unsigned long ptr = arg->val;

1301 1302 1303
	if (arg->augmented.args)
		return syscall_arg__scnprintf_augmented_string(arg, bf, size);

1304 1305 1306 1307 1308 1309 1310
	if (!arg->trace->vfs_getname)
		return scnprintf(bf, size, "%#x", ptr);

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

1311 1312 1313 1314 1315
static bool trace__filter_duration(struct trace *trace, double t)
{
	return t < (trace->duration_filter * NSEC_PER_MSEC);
}

1316
static size_t __trace__fprintf_tstamp(struct trace *trace, u64 tstamp, FILE *fp)
1317 1318 1319
{
	double ts = (double)(tstamp - trace->base_time) / NSEC_PER_MSEC;

1320
	return fprintf(fp, "%10.3f ", ts);
1321 1322
}

1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336
/*
 * 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, "         ? ");
}

1337
static bool done = false;
1338
static bool interrupted = false;
1339

1340
static void sig_handler(int sig)
1341 1342
{
	done = true;
1343
	interrupted = sig == SIGINT;
1344 1345
}

1346
static size_t trace__fprintf_comm_tid(struct trace *trace, struct thread *thread, FILE *fp)
1347
{
1348
	size_t printed = 0;
1349

1350 1351
	if (trace->multiple_threads) {
		if (trace->show_comm)
1352
			printed += fprintf(fp, "%.14s/", thread__comm_str(thread));
1353
		printed += fprintf(fp, "%d ", thread->tid);
1354
	}
1355 1356 1357 1358

	return printed;
}

1359 1360 1361
static size_t trace__fprintf_entry_head(struct trace *trace, struct thread *thread,
					u64 duration, bool duration_calculated, u64 tstamp, FILE *fp)
{
1362 1363 1364 1365
	size_t printed = 0;

	if (trace->show_tstamp)
		printed = trace__fprintf_tstamp(trace, tstamp, fp);
1366 1367
	if (trace->show_duration)
		printed += fprintf_duration(duration, duration_calculated, fp);
1368 1369 1370
	return printed + trace__fprintf_comm_tid(trace, thread, fp);
}

1371
static int trace__process_event(struct trace *trace, struct machine *machine,
1372
				union perf_event *event, struct perf_sample *sample)
1373 1374 1375 1376 1377
{
	int ret = 0;

	switch (event->header.type) {
	case PERF_RECORD_LOST:
1378
		color_fprintf(trace->output, PERF_COLOR_RED,
1379
			      "LOST %" PRIu64 " events!\n", event->lost.lost);
1380
		ret = machine__process_lost_event(machine, event, sample);
1381
		break;
1382
	default:
1383
		ret = machine__process_event(machine, event, sample);
1384 1385 1386 1387 1388 1389
		break;
	}

	return ret;
}

1390
static int trace__tool_process(struct perf_tool *tool,
1391
			       union perf_event *event,
1392
			       struct perf_sample *sample,
1393 1394
			       struct machine *machine)
{
1395
	struct trace *trace = container_of(tool, struct trace, tool);
1396
	return trace__process_event(trace, machine, event, sample);
1397 1398
}

1399 1400 1401 1402 1403 1404 1405 1406 1407
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"
1408
			   "Check /proc/sys/kernel/kptr_restrict and /proc/sys/kernel/perf_event_paranoid.\n\n"
1409 1410 1411 1412 1413 1414 1415 1416
			   "Kernel samples will not be resolved.\n");
		machine->kptr_restrict_warned = true;
		return NULL;
	}

	return machine__resolve_kernel_addr(vmachine, addrp, modp);
}

1417
static int trace__symbols_init(struct trace *trace, struct evlist *evlist)
1418
{
1419
	int err = symbol__init(NULL);
1420 1421 1422 1423

	if (err)
		return err;

1424 1425 1426
	trace->host = machine__new_host();
	if (trace->host == NULL)
		return -ENOMEM;
1427

1428 1429 1430
	err = trace_event__register_resolver(trace->host, trace__machine__resolve_kernel_addr);
	if (err < 0)
		goto out;
1431

1432
	err = __machine__synthesize_threads(trace->host, &trace->tool, &trace->opts.target,
1433
					    evlist->core.threads, trace__tool_process, false,
1434
					    1);
1435
out:
1436 1437 1438 1439 1440 1441
	if (err)
		symbol__exit();

	return err;
}

1442 1443 1444 1445 1446 1447 1448 1449
static void trace__symbols__exit(struct trace *trace)
{
	machine__exit(trace->host);
	trace->host = NULL;

	symbol__exit();
}

1450
static int syscall__alloc_arg_fmts(struct syscall *sc, int nr_args)
1451
{
1452
	int idx;
1453

1454 1455 1456
	if (nr_args == 6 && sc->fmt && sc->fmt->nr_args != 0)
		nr_args = sc->fmt->nr_args;

1457
	sc->arg_fmt = calloc(nr_args, sizeof(*sc->arg_fmt));
1458
	if (sc->arg_fmt == NULL)
1459 1460
		return -1;

1461 1462
	for (idx = 0; idx < nr_args; ++idx) {
		if (sc->fmt)
1463
			sc->arg_fmt[idx] = sc->fmt->arg[idx];
1464
	}
1465

1466 1467 1468 1469 1470 1471
	sc->nr_args = nr_args;
	return 0;
}

static int syscall__set_arg_fmts(struct syscall *sc)
{
1472
	struct tep_format_field *field, *last_field = NULL;
1473 1474 1475
	int idx = 0, len;

	for (field = sc->args; field; field = field->next, ++idx) {
1476 1477
		last_field = field;

1478 1479
		if (sc->fmt && sc->fmt->arg[idx].scnprintf)
			continue;
1480

1481 1482
		len = strlen(field->name);

1483
		if (strcmp(field->type, "const char *") == 0 &&
1484 1485
		    ((len >= 4 && strcmp(field->name + len - 4, "name") == 0) ||
		     strstr(field->name, "path") != NULL))
1486
			sc->arg_fmt[idx].scnprintf = SCA_FILENAME;
1487 1488
		else if ((field->flags & TEP_FIELD_IS_POINTER) || strstr(field->name, "addr"))
			sc->arg_fmt[idx].scnprintf = SCA_PTR;
1489
		else if (strcmp(field->type, "pid_t") == 0)
1490
			sc->arg_fmt[idx].scnprintf = SCA_PID;
1491
		else if (strcmp(field->type, "umode_t") == 0)
1492
			sc->arg_fmt[idx].scnprintf = SCA_MODE_T;
1493 1494 1495
		else if ((strcmp(field->type, "int") == 0 ||
			  strcmp(field->type, "unsigned int") == 0 ||
			  strcmp(field->type, "long") == 0) &&
1496
			 len >= 2 && strcmp(field->name + len - 2, "fd") == 0) {
1497 1498 1499 1500 1501 1502 1503
			/*
			 * /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
			 */
1504
			sc->arg_fmt[idx].scnprintf = SCA_FD;
1505
		}
1506 1507
	}

1508 1509 1510
	if (last_field)
		sc->args_size = last_field->offset + last_field->size;

1511 1512 1513
	return 0;
}

1514 1515 1516 1517
static int trace__read_syscall_info(struct trace *trace, int id)
{
	char tp_name[128];
	struct syscall *sc;
1518
	const char *name = syscalltbl__name(trace->sctbl, id);
1519

1520
	if (trace->syscalls.table == NULL) {
1521
		trace->syscalls.table = calloc(trace->sctbl->syscalls.max_id + 1, sizeof(*sc));
1522
		if (trace->syscalls.table == NULL)
1523
			return -ENOMEM;
1524 1525 1526
	}

	sc = trace->syscalls.table + id;
1527 1528
	if (sc->nonexistent)
		return 0;
1529

1530 1531 1532 1533 1534 1535
	if (name == NULL) {
		sc->nonexistent = true;
		return 0;
	}

	sc->name = name;
1536
	sc->fmt  = syscall_fmt__find(sc->name);
1537

1538
	snprintf(tp_name, sizeof(tp_name), "sys_enter_%s", sc->name);
1539
	sc->tp_format = trace_event__tp_format("syscalls", tp_name);
1540

1541
	if (IS_ERR(sc->tp_format) && sc->fmt && sc->fmt->alias) {
1542
		snprintf(tp_name, sizeof(tp_name), "sys_enter_%s", sc->fmt->alias);
1543
		sc->tp_format = trace_event__tp_format("syscalls", tp_name);
1544
	}
1545

1546
	if (syscall__alloc_arg_fmts(sc, IS_ERR(sc->tp_format) ? 6 : sc->tp_format->format.nr_fields))
1547
		return -ENOMEM;
1548

1549
	if (IS_ERR(sc->tp_format))
1550
		return PTR_ERR(sc->tp_format);
1551

1552
	sc->args = sc->tp_format->format.fields;
1553 1554 1555 1556 1557 1558
	/*
	 * 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"))) {
1559 1560 1561 1562
		sc->args = sc->args->next;
		--sc->nr_args;
	}

1563
	sc->is_exit = !strcmp(name, "exit_group") || !strcmp(name, "exit");
1564
	sc->is_open = !strcmp(name, "open") || !strcmp(name, "openat");
1565

1566
	return syscall__set_arg_fmts(sc);
1567 1568
}

1569 1570 1571 1572 1573 1574 1575
static int intcmp(const void *a, const void *b)
{
	const int *one = a, *another = b;

	return *one - *another;
}

1576 1577
static int trace__validate_ev_qualifier(struct trace *trace)
{
1578
	int err = 0;
1579
	bool printed_invalid_prefix = false;
1580
	struct str_node *pos;
1581
	size_t nr_used = 0, nr_allocated = strlist__nr_entries(trace->ev_qualifier);
1582

1583
	trace->ev_qualifier_ids.entries = malloc(nr_allocated *
1584 1585 1586 1587 1588 1589 1590 1591 1592
						 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;
	}

1593
	strlist__for_each_entry(pos, trace->ev_qualifier) {
1594
		const char *sc = pos->s;
1595
		int id = syscalltbl__id(trace->sctbl, sc), match_next = -1;
1596

1597
		if (id < 0) {
1598 1599 1600 1601
			id = syscalltbl__strglobmatch_first(trace->sctbl, sc, &match_next);
			if (id >= 0)
				goto matches;

1602 1603 1604
			if (!printed_invalid_prefix) {
				pr_debug("Skipping unknown syscalls: ");
				printed_invalid_prefix = true;
1605
			} else {
1606
				pr_debug(", ");
1607 1608
			}

1609 1610
			pr_debug("%s", sc);
			continue;
1611
		}
1612
matches:
1613
		trace->ev_qualifier_ids.entries[nr_used++] = id;
1614 1615 1616 1617 1618 1619 1620
		if (match_next == -1)
			continue;

		while (1) {
			id = syscalltbl__strglobmatch_next(trace->sctbl, sc, &match_next);
			if (id < 0)
				break;
1621
			if (nr_allocated == nr_used) {
1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633
				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;
			}
1634
			trace->ev_qualifier_ids.entries[nr_used++] = id;
1635
		}
1636 1637
	}

1638
	trace->ev_qualifier_ids.nr = nr_used;
1639
	qsort(trace->ev_qualifier_ids.entries, nr_used, sizeof(int), intcmp);
1640
out:
1641 1642
	if (printed_invalid_prefix)
		pr_debug("\n");
1643
	return err;
1644 1645 1646 1647
out_free:
	zfree(&trace->ev_qualifier_ids.entries);
	trace->ev_qualifier_ids.nr = 0;
	goto out;
1648 1649
}

1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665
static __maybe_unused bool trace__syscall_enabled(struct trace *trace, int id)
{
	bool in_ev_qualifier;

	if (trace->ev_qualifier_ids.nr == 0)
		return true;

	in_ev_qualifier = bsearch(&id, trace->ev_qualifier_ids.entries,
				  trace->ev_qualifier_ids.nr, sizeof(int), intcmp) != NULL;

	if (in_ev_qualifier)
	       return !trace->not_ev_qualifier;

	return trace->not_ev_qualifier;
}

1666 1667 1668 1669 1670 1671 1672 1673
/*
 * 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
 */
1674
unsigned long syscall_arg__val(struct syscall_arg *arg, u8 idx)
1675 1676
{
	unsigned long val;
1677
	unsigned char *p = arg->args + sizeof(unsigned long) * idx;
1678 1679 1680 1681 1682

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

1683 1684 1685 1686 1687 1688 1689 1690 1691
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);
}

1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704
/*
 * Check if the value is in fact zero, i.e. mask whatever needs masking, such
 * as mount 'flags' argument that needs ignoring some magic flag, see comment
 * in tools/perf/trace/beauty/mount_flags.c
 */
static unsigned long syscall__mask_val(struct syscall *sc, struct syscall_arg *arg, unsigned long val)
{
	if (sc->arg_fmt && sc->arg_fmt[arg->idx].mask_val)
		return sc->arg_fmt[arg->idx].mask_val(arg, val);

	return val;
}

1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716
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);
}

1717
static size_t syscall__scnprintf_args(struct syscall *sc, char *bf, size_t size,
1718 1719
				      unsigned char *args, void *augmented_args, int augmented_args_size,
				      struct trace *trace, struct thread *thread)
1720 1721
{
	size_t printed = 0;
1722
	unsigned long val;
1723 1724 1725
	u8 bit = 1;
	struct syscall_arg arg = {
		.args	= args,
1726 1727 1728 1729
		.augmented = {
			.size = augmented_args_size,
			.args = augmented_args,
		},
1730 1731 1732 1733
		.idx	= 0,
		.mask	= 0,
		.trace  = trace,
		.thread = thread,
1734
		.show_string_prefix = trace->show_string_prefix,
1735
	};
1736 1737 1738 1739 1740 1741 1742 1743
	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;
1744

1745
	if (sc->args != NULL) {
1746
		struct tep_format_field *field;
1747

1748
		for (field = sc->args; field;
1749 1750
		     field = field->next, ++arg.idx, bit <<= 1) {
			if (arg.mask & bit)
1751
				continue;
1752

1753
			val = syscall_arg__val(&arg, arg.idx);
1754 1755 1756 1757 1758
			/*
			 * Some syscall args need some mask, most don't and
			 * return val untouched.
			 */
			val = syscall__mask_val(sc, &arg, val);
1759

1760 1761 1762 1763 1764
			/*
 			 * Suppress this argument if its value is zero and
 			 * and we don't have a string associated in an
 			 * strarray for it.
 			 */
1765
			if (val == 0 &&
1766
			    !trace->show_zeros &&
1767
			    !(sc->arg_fmt &&
1768 1769
			      (sc->arg_fmt[arg.idx].show_zero ||
			       sc->arg_fmt[arg.idx].scnprintf == SCA_STRARRAY ||
1770 1771
			       sc->arg_fmt[arg.idx].scnprintf == SCA_STRARRAYS) &&
			      sc->arg_fmt[arg.idx].parm))
1772 1773
				continue;

1774 1775 1776 1777 1778
			printed += scnprintf(bf + printed, size - printed, "%s", printed ? ", " : "");

			if (trace->show_arg_names)
				printed += scnprintf(bf + printed, size - printed, "%s: ", field->name);

1779
			printed += syscall__scnprintf_val(sc, bf + printed, size - printed, &arg, val);
1780
		}
1781 1782 1783 1784 1785 1786
	} 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.
		 */
1787
		while (arg.idx < sc->nr_args) {
1788 1789 1790
			if (arg.mask & bit)
				goto next_arg;
			val = syscall_arg__val(&arg, arg.idx);
1791 1792 1793
			if (printed)
				printed += scnprintf(bf + printed, size - printed, ", ");
			printed += syscall__scnprintf_name(sc, bf + printed, size - printed, &arg);
1794 1795 1796 1797
			printed += syscall__scnprintf_val(sc, bf + printed, size - printed, &arg, val);
next_arg:
			++arg.idx;
			bit <<= 1;
1798 1799 1800 1801 1802 1803
		}
	}

	return printed;
}

1804
typedef int (*tracepoint_handler)(struct trace *trace, struct evsel *evsel,
1805
				  union perf_event *event,
1806 1807 1808
				  struct perf_sample *sample);

static struct syscall *trace__syscall_info(struct trace *trace,
1809
					   struct evsel *evsel, int id)
1810
{
1811
	int err = 0;
1812 1813

	if (id < 0) {
1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829

		/*
		 * 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);
		}
1830 1831 1832
		return NULL;
	}

1833 1834
	err = -EINVAL;

1835
	if (id > trace->sctbl->syscalls.max_id)
1836 1837
		goto out_cant_read;

1838 1839
	if ((trace->syscalls.table == NULL || trace->syscalls.table[id].name == NULL) &&
	    (err = trace__read_syscall_info(trace, id)) != 0)
1840 1841
		goto out_cant_read;

1842 1843 1844 1845 1846 1847
	if (trace->syscalls.table[id].name == NULL) {
		if (trace->syscalls.table[id].nonexistent)
			return NULL;
		goto out_cant_read;
	}

1848 1849 1850
	return &trace->syscalls.table[id];

out_cant_read:
1851
	if (verbose > 0) {
1852 1853
		char sbuf[STRERR_BUFSIZE];
		fprintf(trace->output, "Problems reading syscall %d: %d (%s)", id, -err, str_error_r(-err, sbuf, sizeof(sbuf)));
1854
		if (id <= trace->sctbl->syscalls.max_id && trace->syscalls.table[id].name != NULL)
1855 1856 1857
			fprintf(trace->output, "(%s)", trace->syscalls.table[id].name);
		fputs(" information\n", trace->output);
	}
1858 1859 1860
	return NULL;
}

1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886
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);
}

1887
static int trace__printf_interrupted_entry(struct trace *trace)
1888 1889 1890
{
	struct thread_trace *ttrace;
	size_t printed;
1891
	int len;
1892

1893
	if (trace->failure_only || trace->current == NULL)
1894 1895 1896 1897 1898 1899 1900
		return 0;

	ttrace = thread__priv(trace->current);

	if (!ttrace->entry_pending)
		return 0;

1901
	printed  = trace__fprintf_entry_head(trace, trace->current, 0, false, ttrace->entry_time, trace->output);
1902 1903 1904 1905
	printed += len = fprintf(trace->output, "%s)", ttrace->entry_str);

	if (len < trace->args_alignment - 4)
		printed += fprintf(trace->output, "%-*s", trace->args_alignment - 4 - len, " ");
1906

1907 1908 1909
	printed += fprintf(trace->output, " ...\n");

	ttrace->entry_pending = false;
1910 1911
	++trace->nr_events_printed;

1912 1913 1914
	return printed;
}

1915
static int trace__fprintf_sample(struct trace *trace, struct evsel *evsel,
1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931
				 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;
}

1932
static void *syscall__augmented_args(struct syscall *sc, struct perf_sample *sample, int *augmented_args_size, int raw_augmented_args_size)
1933 1934
{
	void *augmented_args = NULL;
1935 1936
	/*
	 * For now with BPF raw_augmented we hook into raw_syscalls:sys_enter
1937 1938 1939 1940 1941
	 * and there we get all 6 syscall args plus the tracepoint common fields
	 * that gets calculated at the start and the syscall_nr (another long).
	 * So we check if that is the case and if so don't look after the
	 * sc->args_size but always after the full raw_syscalls:sys_enter payload,
	 * which is fixed.
1942 1943 1944 1945 1946 1947 1948
	 *
	 * We'll revisit this later to pass s->args_size to the BPF augmenter
	 * (now tools/perf/examples/bpf/augmented_raw_syscalls.c, so that it
	 * copies only what we need for each syscall, like what happens when we
	 * use syscalls:sys_enter_NAME, so that we reduce the kernel/userspace
	 * traffic to just what is needed for each syscall.
	 */
1949
	int args_size = raw_augmented_args_size ?: sc->args_size;
1950

1951
	*augmented_args_size = sample->raw_size - args_size;
1952
	if (*augmented_args_size > 0)
1953
		augmented_args = sample->raw_data + args_size;
1954 1955 1956 1957

	return augmented_args;
}

1958
static int trace__sys_enter(struct trace *trace, struct evsel *evsel,
1959
			    union perf_event *event __maybe_unused,
1960 1961
			    struct perf_sample *sample)
{
1962
	char *msg;
1963
	void *args;
1964
	int printed = 0;
1965
	struct thread *thread;
1966
	int id = perf_evsel__sc_tp_uint(evsel, id, sample), err = -1;
1967 1968
	int augmented_args_size = 0;
	void *augmented_args = NULL;
1969
	struct syscall *sc = trace__syscall_info(trace, evsel, id);
1970 1971 1972 1973
	struct thread_trace *ttrace;

	if (sc == NULL)
		return -1;
1974

1975
	thread = machine__findnew_thread(trace->host, sample->pid, sample->tid);
1976
	ttrace = thread__trace(thread, trace->output);
1977
	if (ttrace == NULL)
1978
		goto out_put;
1979

1980 1981
	trace__fprintf_sample(trace, evsel, sample, thread);

1982
	args = perf_evsel__sc_tp_ptr(evsel, args, sample);
1983 1984

	if (ttrace->entry_str == NULL) {
1985
		ttrace->entry_str = malloc(trace__entry_str_size);
1986
		if (!ttrace->entry_str)
1987
			goto out_put;
1988 1989
	}

1990
	if (!(trace->duration_filter || trace->summary_only || trace->min_stack))
1991
		trace__printf_interrupted_entry(trace);
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002
	/*
	 * If this is raw_syscalls.sys_enter, then it always comes with the 6 possible
	 * arguments, even if the syscall being handled, say "openat", uses only 4 arguments
	 * this breaks syscall__augmented_args() check for augmented args, as we calculate
	 * syscall->args_size using each syscalls:sys_enter_NAME tracefs format file,
	 * so when handling, say the openat syscall, we end up getting 6 args for the
	 * raw_syscalls:sys_enter event, when we expected just 4, we end up mistakenly
	 * thinking that the extra 2 u64 args are the augmented filename, so just check
	 * here and avoid using augmented syscalls when the evsel is the raw_syscalls one.
	 */
	if (evsel != trace->syscalls.events.sys_enter)
2003
		augmented_args = syscall__augmented_args(sc, sample, &augmented_args_size, trace->raw_augmented_syscalls_args_size);
2004 2005
	ttrace->entry_time = sample->time;
	msg = ttrace->entry_str;
2006
	printed += scnprintf(msg + printed, trace__entry_str_size - printed, "%s(", sc->name);
2007

2008
	printed += syscall__scnprintf_args(sc, msg + printed, trace__entry_str_size - printed,
2009
					   args, augmented_args, augmented_args_size, trace, thread);
2010

2011
	if (sc->is_exit) {
2012
		if (!(trace->duration_filter || trace->summary_only || trace->failure_only || trace->min_stack)) {
2013 2014
			int alignment = 0;

2015
			trace__fprintf_entry_head(trace, thread, 0, false, ttrace->entry_time, trace->output);
2016 2017 2018 2019
			printed = fprintf(trace->output, "%s)", ttrace->entry_str);
			if (trace->args_alignment > printed)
				alignment = trace->args_alignment - printed;
			fprintf(trace->output, "%*s= ?\n", alignment, " ");
2020
		}
2021
	} else {
2022
		ttrace->entry_pending = true;
2023 2024 2025
		/* See trace__vfs_getname & trace__sys_exit */
		ttrace->filename.pending_open = false;
	}
2026

2027 2028 2029 2030
	if (trace->current != thread) {
		thread__put(trace->current);
		trace->current = thread__get(thread);
	}
2031 2032 2033 2034
	err = 0;
out_put:
	thread__put(thread);
	return err;
2035 2036
}

2037
static int trace__fprintf_sys_enter(struct trace *trace, struct evsel *evsel,
2038 2039 2040 2041
				    struct perf_sample *sample)
{
	struct thread_trace *ttrace;
	struct thread *thread;
2042 2043
	int id = perf_evsel__sc_tp_uint(evsel, id, sample), err = -1;
	struct syscall *sc = trace__syscall_info(trace, evsel, id);
2044
	char msg[1024];
2045 2046
	void *args, *augmented_args = NULL;
	int augmented_args_size;
2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059

	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;

2060
	args = perf_evsel__sc_tp_ptr(evsel, args, sample);
2061
	augmented_args = syscall__augmented_args(sc, sample, &augmented_args_size, trace->raw_augmented_syscalls_args_size);
2062
	syscall__scnprintf_args(sc, msg, sizeof(msg), args, augmented_args, augmented_args_size, trace, thread);
2063 2064 2065 2066 2067 2068 2069
	fprintf(trace->output, "%s", msg);
	err = 0;
out_put:
	thread__put(thread);
	return err;
}

2070
static int trace__resolve_callchain(struct trace *trace, struct evsel *evsel,
2071 2072
				    struct perf_sample *sample,
				    struct callchain_cursor *cursor)
2073 2074
{
	struct addr_location al;
2075 2076
	int max_stack = evsel->core.attr.sample_max_stack ?
			evsel->core.attr.sample_max_stack :
2077
			trace->max_stack;
2078
	int err;
2079

2080
	if (machine__resolve(trace->host, &al, sample) < 0)
2081 2082
		return -1;

2083 2084 2085
	err = thread__resolve_callchain(al.thread, cursor, evsel, sample, NULL, NULL, max_stack);
	addr_location__put(&al);
	return err;
2086 2087 2088 2089
}

static int trace__fprintf_callchain(struct trace *trace, struct perf_sample *sample)
{
2090
	/* TODO: user-configurable print_opts */
2091 2092 2093
	const unsigned int print_opts = EVSEL__PRINT_SYM |
				        EVSEL__PRINT_DSO |
				        EVSEL__PRINT_UNKNOWN_AS_ADDR;
2094

2095
	return sample__fprintf_callchain(sample, 38, print_opts, &callchain_cursor, symbol_conf.bt_stop_list, trace->output);
2096 2097
}

2098
static const char *errno_to_name(struct evsel *evsel, int err)
2099 2100 2101 2102 2103 2104 2105
{
	struct perf_env *env = perf_evsel__env(evsel);
	const char *arch_name = perf_env__arch(env);

	return arch_syscalls__strerrno(arch_name, err);
}

2106
static int trace__sys_exit(struct trace *trace, struct evsel *evsel,
2107
			   union perf_event *event __maybe_unused,
2108 2109
			   struct perf_sample *sample)
{
2110
	long ret;
2111
	u64 duration = 0;
2112
	bool duration_calculated = false;
2113
	struct thread *thread;
2114 2115
	int id = perf_evsel__sc_tp_uint(evsel, id, sample), err = -1, callchain_ret = 0, printed = 0;
	int alignment = trace->args_alignment;
2116
	struct syscall *sc = trace__syscall_info(trace, evsel, id);
2117 2118 2119 2120
	struct thread_trace *ttrace;

	if (sc == NULL)
		return -1;
2121

2122
	thread = machine__findnew_thread(trace->host, sample->pid, sample->tid);
2123
	ttrace = thread__trace(thread, trace->output);
2124
	if (ttrace == NULL)
2125
		goto out_put;
2126

2127 2128
	trace__fprintf_sample(trace, evsel, sample, thread);

2129 2130 2131
	if (trace->summary)
		thread__update_stats(ttrace, id, sample);

2132
	ret = perf_evsel__sc_tp_uint(evsel, ret, sample);
2133

2134
	if (!trace->fd_path_disabled && sc->is_open && ret >= 0 && ttrace->filename.pending_open) {
2135 2136
		trace__set_fd_pathname(thread, ret, ttrace->filename.name);
		ttrace->filename.pending_open = false;
2137 2138 2139
		++trace->stats.vfs_getname;
	}

2140
	if (ttrace->entry_time) {
2141
		duration = sample->time - ttrace->entry_time;
2142 2143
		if (trace__filter_duration(trace, duration))
			goto out;
2144
		duration_calculated = true;
2145 2146
	} else if (trace->duration_filter)
		goto out;
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;
			callchain_ret = 1;
		}
	}

2157
	if (trace->summary_only || (ret >= 0 && trace->failure_only))
2158 2159
		goto out;

2160
	trace__fprintf_entry_head(trace, thread, duration, duration_calculated, ttrace->entry_time, trace->output);
2161 2162

	if (ttrace->entry_pending) {
2163
		printed = fprintf(trace->output, "%s", ttrace->entry_str);
2164
	} else {
2165
		printed += fprintf(trace->output, " ... [");
2166
		color_fprintf(trace->output, PERF_COLOR_YELLOW, "continued");
2167 2168
		printed += 9;
		printed += fprintf(trace->output, "]: %s()", sc->name);
2169 2170
	}

2171 2172 2173 2174 2175 2176 2177 2178 2179
	printed++; /* the closing ')' */

	if (alignment > printed)
		alignment -= printed;
	else
		alignment = 0;

	fprintf(trace->output, ")%*s= ", alignment, " ");

2180
	if (sc->fmt == NULL) {
2181 2182
		if (ret < 0)
			goto errno_print;
2183
signed_print:
2184
		fprintf(trace->output, "%ld", ret);
2185 2186
	} else if (ret < 0) {
errno_print: {
2187
		char bf[STRERR_BUFSIZE];
2188
		const char *emsg = str_error_r(-ret, bf, sizeof(bf)),
2189
			   *e = errno_to_name(evsel, -ret);
2190

2191
		fprintf(trace->output, "-1 %s (%s)", e, emsg);
2192
	}
2193
	} else if (ret == 0 && sc->fmt->timeout)
2194
		fprintf(trace->output, "0 (Timeout)");
2195 2196
	else if (ttrace->ret_scnprintf) {
		char bf[1024];
2197 2198 2199 2200 2201 2202
		struct syscall_arg arg = {
			.val	= ret,
			.thread	= thread,
			.trace	= trace,
		};
		ttrace->ret_scnprintf(bf, sizeof(bf), &arg);
2203
		ttrace->ret_scnprintf = NULL;
2204
		fprintf(trace->output, "%s", bf);
2205
	} else if (sc->fmt->hexret)
2206
		fprintf(trace->output, "%#lx", ret);
2207 2208 2209 2210
	else if (sc->fmt->errpid) {
		struct thread *child = machine__find_thread(trace->host, ret, ret);

		if (child != NULL) {
2211
			fprintf(trace->output, "%ld", ret);
2212 2213 2214 2215 2216
			if (child->comm_set)
				fprintf(trace->output, " (%s)", thread__comm_str(child));
			thread__put(child);
		}
	} else
2217
		goto signed_print;
2218

2219
	fputc('\n', trace->output);
2220

2221 2222 2223 2224 2225 2226 2227
	/*
	 * We only consider an 'event' for the sake of --max-events a non-filtered
	 * sys_enter + sys_exit and other tracepoint events.
	 */
	if (++trace->nr_events_printed == trace->max_events && trace->max_events != ULONG_MAX)
		interrupted = true;

2228 2229 2230 2231
	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));
2232
out:
2233
	ttrace->entry_pending = false;
2234 2235 2236 2237
	err = 0;
out_put:
	thread__put(thread);
	return err;
2238 2239
}

2240
static int trace__vfs_getname(struct trace *trace, struct evsel *evsel,
2241
			      union perf_event *event __maybe_unused,
2242 2243
			      struct perf_sample *sample)
{
2244 2245 2246 2247 2248
	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;
2249
	const char *filename = perf_evsel__rawptr(evsel, sample, "pathname");
2250 2251 2252 2253 2254 2255

	if (!thread)
		goto out;

	ttrace = thread__priv(thread);
	if (!ttrace)
2256
		goto out_put;
2257

2258
	filename_len = strlen(filename);
2259
	if (filename_len == 0)
2260
		goto out_put;
2261 2262 2263 2264 2265

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

		if (f == NULL)
2266
			goto out_put;
2267 2268 2269 2270 2271 2272 2273 2274

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

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

2275
	if (!ttrace->filename.ptr)
2276
		goto out_put;
2277 2278 2279 2280

	entry_str_len = strlen(ttrace->entry_str);
	remaining_space = trace__entry_str_size - entry_str_len - 1; /* \0 */
	if (remaining_space <= 0)
2281
		goto out_put;
2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294

	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;
2295 2296
out_put:
	thread__put(thread);
2297
out:
2298 2299 2300
	return 0;
}

2301
static int trace__sched_stat_runtime(struct trace *trace, struct evsel *evsel,
2302
				     union perf_event *event __maybe_unused,
2303 2304 2305 2306
				     struct perf_sample *sample)
{
        u64 runtime = perf_evsel__intval(evsel, sample, "runtime");
	double runtime_ms = (double)runtime / NSEC_PER_MSEC;
2307
	struct thread *thread = machine__findnew_thread(trace->host,
2308 2309
							sample->pid,
							sample->tid);
2310
	struct thread_trace *ttrace = thread__trace(thread, trace->output);
2311 2312 2313 2314 2315 2316

	if (ttrace == NULL)
		goto out_dump;

	ttrace->runtime_ms += runtime_ms;
	trace->runtime_ms += runtime_ms;
2317
out_put:
2318
	thread__put(thread);
2319 2320 2321
	return 0;

out_dump:
2322
	fprintf(trace->output, "%s: comm=%s,pid=%u,runtime=%" PRIu64 ",vruntime=%" PRIu64 ")\n",
2323 2324 2325 2326 2327
	       evsel->name,
	       perf_evsel__strval(evsel, sample, "comm"),
	       (pid_t)perf_evsel__intval(evsel, sample, "pid"),
	       runtime,
	       perf_evsel__intval(evsel, sample, "vruntime"));
2328
	goto out_put;
2329 2330
}

2331 2332
static int bpf_output__printer(enum binary_printer_ops op,
			       unsigned int val, void *extra __maybe_unused, FILE *fp)
2333 2334 2335 2336 2337
{
	unsigned char ch = (unsigned char)val;

	switch (op) {
	case BINARY_PRINT_CHAR_DATA:
2338
		return fprintf(fp, "%c", isprint(ch) ? ch : '.');
2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350
	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;
	}
2351 2352

	return 0;
2353 2354 2355 2356 2357
}

static void bpf_output__fprintf(struct trace *trace,
				struct perf_sample *sample)
{
2358 2359
	binary__fprintf(sample->raw_data, sample->raw_size, 8,
			bpf_output__printer, NULL, trace->output);
2360
	++trace->nr_events_printed;
2361 2362
}

2363
static int trace__event_handler(struct trace *trace, struct evsel *evsel,
2364 2365 2366
				union perf_event *event __maybe_unused,
				struct perf_sample *sample)
{
2367
	struct thread *thread;
2368
	int callchain_ret = 0;
2369 2370 2371 2372 2373 2374 2375 2376 2377 2378
	/*
	 * Check if we called perf_evsel__disable(evsel) due to, for instance,
	 * this event's max_events having been hit and this is an entry coming
	 * from the ring buffer that we should discard, since the max events
	 * have already been considered/printed.
	 */
	if (evsel->disabled)
		return 0;

	thread = machine__findnew_thread(trace->host, sample->pid, sample->tid);
2379 2380 2381 2382 2383 2384 2385 2386 2387 2388

	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;
		}
	}

2389
	trace__printf_interrupted_entry(trace);
2390
	trace__fprintf_tstamp(trace, sample->time, trace->output);
2391

2392
	if (trace->trace_syscalls && trace->show_duration)
2393 2394
		fprintf(trace->output, "(         ): ");

2395 2396 2397
	if (thread)
		trace__fprintf_comm_tid(trace, thread, trace->output);

2398 2399 2400 2401 2402
	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) {
2403 2404 2405 2406
			fprintf(trace->output, "%s(", sc->name);
			trace__fprintf_sys_enter(trace, evsel, sample);
			fputc(')', trace->output);
			goto newline;
2407 2408 2409 2410 2411 2412 2413 2414 2415
		}

		/*
		 * 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.
		 */
	}

2416
	fprintf(trace->output, "%s:", evsel->name);
2417

2418
	if (perf_evsel__is_bpf_output(evsel)) {
2419
		bpf_output__fprintf(trace, sample);
2420
	} else if (evsel->tp_format) {
2421 2422 2423 2424 2425
		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);
2426
			++trace->nr_events_printed;
2427 2428

			if (evsel->max_events != ULONG_MAX && ++evsel->nr_events_printed == evsel->max_events) {
2429
				evsel__disable(evsel);
2430
				evsel__close(evsel);
2431
			}
2432
		}
2433 2434
	}

2435
newline:
2436
	fprintf(trace->output, "\n");
2437

2438 2439 2440 2441 2442
	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:
2443
	thread__put(thread);
2444 2445 2446
	return 0;
}

2447 2448 2449 2450 2451
static void print_location(FILE *f, struct perf_sample *sample,
			   struct addr_location *al,
			   bool print_dso, bool print_sym)
{

2452
	if ((verbose > 0 || print_dso) && al->map)
2453 2454
		fprintf(f, "%s@", al->map->dso->long_name);

2455
	if ((verbose > 0 || print_sym) && al->sym)
2456
		fprintf(f, "%s+0x%" PRIx64, al->sym->name,
2457 2458
			al->addr - al->sym->start);
	else if (al->map)
2459
		fprintf(f, "0x%" PRIx64, al->addr);
2460
	else
2461
		fprintf(f, "0x%" PRIx64, sample->addr);
2462 2463 2464
}

static int trace__pgfault(struct trace *trace,
2465
			  struct evsel *evsel,
2466
			  union perf_event *event __maybe_unused,
2467 2468 2469 2470 2471
			  struct perf_sample *sample)
{
	struct thread *thread;
	struct addr_location al;
	char map_type = 'd';
2472
	struct thread_trace *ttrace;
2473
	int err = -1;
2474
	int callchain_ret = 0;
2475 2476

	thread = machine__findnew_thread(trace->host, sample->pid, sample->tid);
2477 2478 2479 2480 2481 2482 2483 2484 2485 2486

	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;
		}
	}

2487 2488
	ttrace = thread__trace(thread, trace->output);
	if (ttrace == NULL)
2489
		goto out_put;
2490

2491
	if (evsel->core.attr.config == PERF_COUNT_SW_PAGE_FAULTS_MAJ)
2492 2493 2494 2495 2496
		ttrace->pfmaj++;
	else
		ttrace->pfmin++;

	if (trace->summary_only)
2497
		goto out;
2498

2499
	thread__find_symbol(thread, sample->cpumode, sample->ip, &al);
2500

2501
	trace__fprintf_entry_head(trace, thread, 0, true, sample->time, trace->output);
2502 2503

	fprintf(trace->output, "%sfault [",
2504
		evsel->core.attr.config == PERF_COUNT_SW_PAGE_FAULTS_MAJ ?
2505 2506 2507 2508 2509 2510
		"maj" : "min");

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

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

2511
	thread__find_symbol(thread, sample->cpumode, sample->addr, &al);
2512 2513

	if (!al.map) {
2514
		thread__find_symbol(thread, sample->cpumode, sample->addr, &al);
2515 2516 2517 2518 2519 2520 2521 2522 2523 2524

		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);
2525

2526 2527 2528 2529
	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));
2530 2531

	++trace->nr_events_printed;
2532 2533 2534 2535 2536
out:
	err = 0;
out_put:
	thread__put(thread);
	return err;
2537 2538
}

2539
static void trace__set_base_time(struct trace *trace,
2540
				 struct evsel *evsel,
2541 2542
				 struct perf_sample *sample)
{
2543 2544 2545 2546 2547 2548 2549 2550 2551
	/*
	 * 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 &&
2552
	    (evsel->core.attr.sample_type & PERF_SAMPLE_TIME))
2553 2554 2555
		trace->base_time = sample->time;
}

2556
static int trace__process_sample(struct perf_tool *tool,
2557
				 union perf_event *event,
2558
				 struct perf_sample *sample,
2559
				 struct evsel *evsel,
2560 2561 2562
				 struct machine *machine __maybe_unused)
{
	struct trace *trace = container_of(tool, struct trace, tool);
2563
	struct thread *thread;
2564 2565
	int err = 0;

2566
	tracepoint_handler handler = evsel->handler;
2567

2568 2569
	thread = machine__findnew_thread(trace->host, sample->pid, sample->tid);
	if (thread && thread__is_filtered(thread))
2570
		goto out;
2571

2572
	trace__set_base_time(trace, evsel, sample);
2573

2574 2575
	if (handler) {
		++trace->nr_events;
2576
		handler(trace, evsel, event, sample);
2577
	}
2578 2579
out:
	thread__put(thread);
2580 2581 2582
	return err;
}

2583
static int trace__record(struct trace *trace, int argc, const char **argv)
David Ahern's avatar
David Ahern committed
2584 2585 2586 2587 2588 2589 2590 2591 2592 2593
{
	unsigned int rec_argc, i, j;
	const char **rec_argv;
	const char * const record_args[] = {
		"record",
		"-R",
		"-m", "1024",
		"-c", "1",
	};

2594 2595 2596 2597 2598 2599 2600
	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);

2601
	/* +1 is for the event string below */
2602 2603
	rec_argc = ARRAY_SIZE(record_args) + sc_args_nr + 1 +
		majpf_args_nr + minpf_args_nr + argc;
David Ahern's avatar
David Ahern committed
2604 2605 2606 2607 2608
	rec_argv = calloc(rec_argc + 1, sizeof(char *));

	if (rec_argv == NULL)
		return -ENOMEM;

2609
	j = 0;
David Ahern's avatar
David Ahern committed
2610
	for (i = 0; i < ARRAY_SIZE(record_args); i++)
2611 2612
		rec_argv[j++] = record_args[i];

2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623
	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");
2624
			free(rec_argv);
2625 2626
			return -1;
		}
2627 2628
	}

2629 2630 2631 2632 2633 2634 2635 2636 2637 2638
	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];
David Ahern's avatar
David Ahern committed
2639

2640
	return cmd_record(j, rec_argv);
David Ahern's avatar
David Ahern committed
2641 2642
}

2643 2644
static size_t trace__fprintf_thread_summary(struct trace *trace, FILE *fp);

2645
static bool evlist__add_vfs_getname(struct evlist *evlist)
2646
{
2647
	bool found = false;
2648
	struct evsel *evsel, *tmp;
2649 2650
	struct parse_events_error err = { .idx = 0, };
	int ret = parse_events(evlist, "probe:vfs_getname*", &err);
2651

2652
	if (ret)
2653
		return false;
2654

2655 2656 2657 2658 2659 2660 2661 2662 2663 2664
	evlist__for_each_entry_safe(evlist, evsel, tmp) {
		if (!strstarts(perf_evsel__name(evsel), "probe:vfs_getname"))
			continue;

		if (perf_evsel__field(evsel, "pathname")) {
			evsel->handler = trace__vfs_getname;
			found = true;
			continue;
		}

2665
		list_del_init(&evsel->core.node);
2666
		evsel->evlist = NULL;
2667
		evsel__delete(evsel);
2668 2669
	}

2670
	return found;
2671 2672
}

2673
static struct evsel *perf_evsel__new_pgfault(u64 config)
2674
{
2675
	struct evsel *evsel;
2676 2677 2678 2679 2680 2681
	struct perf_event_attr attr = {
		.type = PERF_TYPE_SOFTWARE,
		.mmap_data = 1,
	};

	attr.config = config;
2682
	attr.sample_period = 1;
2683 2684 2685

	event_attr_init(&attr);

2686
	evsel = evsel__new(&attr);
2687 2688
	if (evsel)
		evsel->handler = trace__pgfault;
2689

2690
	return evsel;
2691 2692
}

2693 2694 2695
static void trace__handle_event(struct trace *trace, union perf_event *event, struct perf_sample *sample)
{
	const u32 type = event->header.type;
2696
	struct evsel *evsel;
2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708

	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;
	}

2709 2710 2711
	if (evswitch__discard(&trace->evswitch, evsel))
		return;

2712 2713
	trace__set_base_time(trace, evsel, sample);

2714
	if (evsel->core.attr.type == PERF_TYPE_TRACEPOINT &&
2715 2716 2717 2718 2719 2720 2721 2722
	    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);
	}
2723 2724 2725

	if (trace->nr_events_printed >= trace->max_events && trace->max_events != ULONG_MAX)
		interrupted = true;
2726 2727
}

2728 2729 2730
static int trace__add_syscall_newtp(struct trace *trace)
{
	int ret = -1;
2731
	struct evlist *evlist = trace->evlist;
2732
	struct evsel *sys_enter, *sys_exit;
2733

2734
	sys_enter = perf_evsel__raw_syscall_newtp("sys_enter", trace__sys_enter);
2735 2736 2737 2738 2739 2740
	if (sys_enter == NULL)
		goto out;

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

2741
	sys_exit = perf_evsel__raw_syscall_newtp("sys_exit", trace__sys_exit);
2742 2743 2744 2745 2746 2747
	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;

2748 2749 2750
	perf_evsel__config_callchain(sys_enter, &trace->opts, &callchain_param);
	perf_evsel__config_callchain(sys_exit, &trace->opts, &callchain_param);

2751 2752
	evlist__add(evlist, sys_enter);
	evlist__add(evlist, sys_exit);
2753

2754
	if (callchain_param.enabled && !trace->kernel_syscallchains) {
2755 2756 2757 2758 2759
		/*
		 * We're interested only in the user space callchain
		 * leading to the syscall, allow overriding that for
		 * debugging reasons using --kernel_syscall_callchains
		 */
2760
		sys_exit->core.attr.exclude_callchain_kernel = 1;
2761 2762
	}

2763 2764
	trace->syscalls.events.sys_enter = sys_enter;
	trace->syscalls.events.sys_exit  = sys_exit;
2765 2766 2767 2768 2769 2770

	ret = 0;
out:
	return ret;

out_delete_sys_exit:
2771
	evsel__delete_priv(sys_exit);
2772
out_delete_sys_enter:
2773
	evsel__delete_priv(sys_enter);
2774 2775 2776
	goto out;
}

2777
static int trace__set_ev_qualifier_tp_filter(struct trace *trace)
2778 2779
{
	int err = -1;
2780
	struct evsel *sys_exit;
2781 2782 2783 2784 2785 2786 2787
	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;

2788 2789
	if (!perf_evsel__append_tp_filter(trace->syscalls.events.sys_enter,
					  filter)) {
2790
		sys_exit = trace->syscalls.events.sys_exit;
2791
		err = perf_evsel__append_tp_filter(sys_exit, filter);
2792
	}
2793 2794 2795 2796 2797 2798 2799 2800

	free(filter);
out:
	return err;
out_enomem:
	errno = ENOMEM;
	goto out;
}
2801

2802
#ifdef HAVE_LIBBPF_SUPPORT
2803 2804 2805 2806 2807 2808 2809 2810
static struct bpf_program *trace__find_bpf_program_by_title(struct trace *trace, const char *name)
{
	if (trace->bpf_obj == NULL)
		return NULL;

	return bpf_object__find_program_by_title(trace->bpf_obj, name);
}

2811 2812 2813 2814 2815
static struct bpf_program *trace__find_syscall_bpf_prog(struct trace *trace, struct syscall *sc,
							const char *prog_name, const char *type)
{
	struct bpf_program *prog;

2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827
	if (prog_name == NULL) {
		char default_prog_name[256];
		scnprintf(default_prog_name, sizeof(default_prog_name), "!syscalls:sys_%s_%s", type, sc->name);
		prog = trace__find_bpf_program_by_title(trace, default_prog_name);
		if (prog != NULL)
			goto out_found;
		if (sc->fmt && sc->fmt->alias) {
			scnprintf(default_prog_name, sizeof(default_prog_name), "!syscalls:sys_%s_%s", type, sc->fmt->alias);
			prog = trace__find_bpf_program_by_title(trace, default_prog_name);
			if (prog != NULL)
				goto out_found;
		}
2828
		goto out_unaugmented;
2829
	}
2830 2831

	prog = trace__find_bpf_program_by_title(trace, prog_name);
2832 2833 2834

	if (prog != NULL) {
out_found:
2835
		return prog;
2836
	}
2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850

	pr_debug("Couldn't find BPF prog \"%s\" to associate with syscalls:sys_%s_%s, not augmenting it\n",
		 prog_name, type, sc->name);
out_unaugmented:
	return trace->syscalls.unaugmented_prog;
}

static void trace__init_syscall_bpf_progs(struct trace *trace, int id)
{
	struct syscall *sc = trace__syscall_info(trace, NULL, id);

	if (sc == NULL)
		return;

2851 2852
	sc->bpf_prog.sys_enter = trace__find_syscall_bpf_prog(trace, sc, sc->fmt ? sc->fmt->bpf_prog_name.sys_enter : NULL, "enter");
	sc->bpf_prog.sys_exit  = trace__find_syscall_bpf_prog(trace, sc, sc->fmt ? sc->fmt->bpf_prog_name.sys_exit  : NULL,  "exit");
2853 2854
}

2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866
static int trace__bpf_prog_sys_enter_fd(struct trace *trace, int id)
{
	struct syscall *sc = trace__syscall_info(trace, NULL, id);
	return sc ? bpf_program__fd(sc->bpf_prog.sys_enter) : bpf_program__fd(trace->syscalls.unaugmented_prog);
}

static int trace__bpf_prog_sys_exit_fd(struct trace *trace, int id)
{
	struct syscall *sc = trace__syscall_info(trace, NULL, id);
	return sc ? bpf_program__fd(sc->bpf_prog.sys_exit) : bpf_program__fd(trace->syscalls.unaugmented_prog);
}

2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885
static void trace__init_bpf_map_syscall_args(struct trace *trace, int id, struct bpf_map_syscall_entry *entry)
{
	struct syscall *sc = trace__syscall_info(trace, NULL, id);
	int arg = 0;

	if (sc == NULL)
		goto out;

	for (; arg < sc->nr_args; ++arg) {
		entry->string_args_len[arg] = 0;
		if (sc->arg_fmt[arg].scnprintf == SCA_FILENAME) {
			/* Should be set like strace -s strsize */
			entry->string_args_len[arg] = PATH_MAX;
		}
	}
out:
	for (; arg < 6; ++arg)
		entry->string_args_len[arg] = 0;
}
2886 2887 2888
static int trace__set_ev_qualifier_bpf_filter(struct trace *trace)
{
	int fd = bpf_map__fd(trace->syscalls.map);
2889 2890 2891
	struct bpf_map_syscall_entry value = {
		.enabled = !trace->not_ev_qualifier,
	};
2892 2893 2894 2895 2896 2897
	int err = 0;
	size_t i;

	for (i = 0; i < trace->ev_qualifier_ids.nr; ++i) {
		int key = trace->ev_qualifier_ids.entries[i];

2898
		if (value.enabled) {
2899
			trace__init_bpf_map_syscall_args(trace, key, &value);
2900 2901
			trace__init_syscall_bpf_progs(trace, key);
		}
2902

2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913
		err = bpf_map_update_elem(fd, &key, &value, BPF_EXIST);
		if (err)
			break;
	}

	return err;
}

static int __trace__init_syscalls_bpf_map(struct trace *trace, bool enabled)
{
	int fd = bpf_map__fd(trace->syscalls.map);
2914 2915 2916
	struct bpf_map_syscall_entry value = {
		.enabled = enabled,
	};
2917 2918 2919
	int err = 0, key;

	for (key = 0; key < trace->sctbl->syscalls.nr_entries; ++key) {
2920
		if (enabled)
2921 2922
			trace__init_bpf_map_syscall_args(trace, key, &value);

2923
		err = bpf_map_update_elem(fd, &key, &value, BPF_ANY);
2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939
		if (err)
			break;
	}

	return err;
}

static int trace__init_syscalls_bpf_map(struct trace *trace)
{
	bool enabled = true;

	if (trace->ev_qualifier_ids.nr)
		enabled = trace->not_ev_qualifier;

	return __trace__init_syscalls_bpf_map(trace, enabled);
}
2940

2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 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 3024 3025 3026 3027 3028
static struct bpf_program *trace__find_usable_bpf_prog_entry(struct trace *trace, struct syscall *sc)
{
	struct tep_format_field *field, *candidate_field;
	int id;

	/*
	 * We're only interested in syscalls that have a pointer:
	 */
	for (field = sc->args; field; field = field->next) {
		if (field->flags & TEP_FIELD_IS_POINTER)
			goto try_to_find_pair;
	}

	return NULL;

try_to_find_pair:
	for (id = 0; id < trace->sctbl->syscalls.nr_entries; ++id) {
		struct syscall *pair = trace__syscall_info(trace, NULL, id);
		struct bpf_program *pair_prog;
		bool is_candidate = false;

		if (pair == NULL || pair == sc ||
		    pair->bpf_prog.sys_enter == trace->syscalls.unaugmented_prog)
			continue;

		for (field = sc->args, candidate_field = pair->args;
		     field && candidate_field; field = field->next, candidate_field = candidate_field->next) {
			bool is_pointer = field->flags & TEP_FIELD_IS_POINTER,
			     candidate_is_pointer = candidate_field->flags & TEP_FIELD_IS_POINTER;

			if (is_pointer) {
			       if (!candidate_is_pointer) {
					// The candidate just doesn't copies our pointer arg, might copy other pointers we want.
					continue;
			       }
			} else {
				if (candidate_is_pointer) {
					// The candidate might copy a pointer we don't have, skip it.
					goto next_candidate;
				}
				continue;
			}

			if (strcmp(field->type, candidate_field->type))
				goto next_candidate;

			is_candidate = true;
		}

		if (!is_candidate)
			goto next_candidate;

		/*
		 * Check if the tentative pair syscall augmenter has more pointers, if it has,
		 * then it may be collecting that and we then can't use it, as it would collect
		 * more than what is common to the two syscalls.
		 */
		if (candidate_field) {
			for (candidate_field = candidate_field->next; candidate_field; candidate_field = candidate_field->next)
				if (candidate_field->flags & TEP_FIELD_IS_POINTER)
					goto next_candidate;
		}

		pair_prog = pair->bpf_prog.sys_enter;
		/*
		 * If the pair isn't enabled, then its bpf_prog.sys_enter will not
		 * have been searched for, so search it here and if it returns the
		 * unaugmented one, then ignore it, otherwise we'll reuse that BPF
		 * program for a filtered syscall on a non-filtered one.
		 *
		 * For instance, we have "!syscalls:sys_enter_renameat" and that is
		 * useful for "renameat2".
		 */
		if (pair_prog == NULL) {
			pair_prog = trace__find_syscall_bpf_prog(trace, pair, pair->fmt ? pair->fmt->bpf_prog_name.sys_enter : NULL, "enter");
			if (pair_prog == trace->syscalls.unaugmented_prog)
				goto next_candidate;
		}

		pr_debug("Reusing \"%s\" BPF sys_enter augmenter for \"%s\"\n", pair->name, sc->name);
		return pair_prog;
	next_candidate:
		continue;
	}

	return NULL;
}

3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053
static int trace__init_syscalls_bpf_prog_array_maps(struct trace *trace)
{
	int map_enter_fd = bpf_map__fd(trace->syscalls.prog_array.sys_enter),
	    map_exit_fd  = bpf_map__fd(trace->syscalls.prog_array.sys_exit);
	int err = 0, key;

	for (key = 0; key < trace->sctbl->syscalls.nr_entries; ++key) {
		int prog_fd;

		if (!trace__syscall_enabled(trace, key))
			continue;

		trace__init_syscall_bpf_progs(trace, key);

		// It'll get at least the "!raw_syscalls:unaugmented"
		prog_fd = trace__bpf_prog_sys_enter_fd(trace, key);
		err = bpf_map_update_elem(map_enter_fd, &key, &prog_fd, BPF_ANY);
		if (err)
			break;
		prog_fd = trace__bpf_prog_sys_exit_fd(trace, key);
		err = bpf_map_update_elem(map_exit_fd, &key, &prog_fd, BPF_ANY);
		if (err)
			break;
	}

3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 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 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117
	/*
	 * Now lets do a second pass looking for enabled syscalls without
	 * an augmenter that have a signature that is a superset of another
	 * syscall with an augmenter so that we can auto-reuse it.
	 *
	 * I.e. if we have an augmenter for the "open" syscall that has
	 * this signature:
	 *
	 *   int open(const char *pathname, int flags, mode_t mode);
	 *
	 * I.e. that will collect just the first string argument, then we
	 * can reuse it for the 'creat' syscall, that has this signature:
	 *
	 *   int creat(const char *pathname, mode_t mode);
	 *
	 * and for:
	 *
	 *   int stat(const char *pathname, struct stat *statbuf);
	 *   int lstat(const char *pathname, struct stat *statbuf);
	 *
	 * Because the 'open' augmenter will collect the first arg as a string,
	 * and leave alone all the other args, which already helps with
	 * beautifying 'stat' and 'lstat''s pathname arg.
	 *
	 * Then, in time, when 'stat' gets an augmenter that collects both
	 * first and second arg (this one on the raw_syscalls:sys_exit prog
	 * array tail call, then that one will be used.
	 */
	for (key = 0; key < trace->sctbl->syscalls.nr_entries; ++key) {
		struct syscall *sc = trace__syscall_info(trace, NULL, key);
		struct bpf_program *pair_prog;
		int prog_fd;

		if (sc == NULL || sc->bpf_prog.sys_enter == NULL)
			continue;

		/*
		 * For now we're just reusing the sys_enter prog, and if it
		 * already has an augmenter, we don't need to find one.
		 */
		if (sc->bpf_prog.sys_enter != trace->syscalls.unaugmented_prog)
			continue;

		/*
		 * Look at all the other syscalls for one that has a signature
		 * that is close enough that we can share:
		 */
		pair_prog = trace__find_usable_bpf_prog_entry(trace, sc);
		if (pair_prog == NULL)
			continue;

		sc->bpf_prog.sys_enter = pair_prog;

		/*
		 * Update the BPF_MAP_TYPE_PROG_SHARED for raw_syscalls:sys_enter
		 * with the fd for the program we're reusing:
		 */
		prog_fd = bpf_program__fd(sc->bpf_prog.sys_enter);
		err = bpf_map_update_elem(map_enter_fd, &key, &prog_fd, BPF_ANY);
		if (err)
			break;
	}


3118 3119
	return err;
}
3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140

static void trace__delete_augmented_syscalls(struct trace *trace)
{
	struct evsel *evsel, *tmp;

	evlist__remove(trace->evlist, trace->syscalls.events.augmented);
	evsel__delete(trace->syscalls.events.augmented);
	trace->syscalls.events.augmented = NULL;

	evlist__for_each_entry_safe(trace->evlist, tmp, evsel) {
		if (evsel->bpf_obj == trace->bpf_obj) {
			evlist__remove(trace->evlist, evsel);
			evsel__delete(evsel);
		}

	}

	bpf_object__close(trace->bpf_obj);
	trace->bpf_obj = NULL;
}
#else // HAVE_LIBBPF_SUPPORT
3141 3142 3143 3144 3145 3146 3147 3148 3149
static int trace__set_ev_qualifier_bpf_filter(struct trace *trace __maybe_unused)
{
	return 0;
}

static int trace__init_syscalls_bpf_map(struct trace *trace __maybe_unused)
{
	return 0;
}
3150 3151 3152 3153 3154 3155

static struct bpf_program *trace__find_bpf_program_by_title(struct trace *trace __maybe_unused,
							    const char *name __maybe_unused)
{
	return NULL;
}
3156 3157 3158 3159 3160

static int trace__init_syscalls_bpf_prog_array_maps(struct trace *trace __maybe_unused)
{
	return 0;
}
3161 3162 3163 3164

static void trace__delete_augmented_syscalls(struct trace *trace __maybe_unused)
{
}
3165 3166
#endif // HAVE_LIBBPF_SUPPORT

3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181
static bool trace__only_augmented_syscalls_evsels(struct trace *trace)
{
	struct evsel *evsel;

	evlist__for_each_entry(trace->evlist, evsel) {
		if (evsel == trace->syscalls.events.augmented ||
		    evsel->bpf_obj == trace->bpf_obj)
			continue;

		return false;
	}

	return true;
}

3182 3183
static int trace__set_ev_qualifier_filter(struct trace *trace)
{
3184 3185
	if (trace->syscalls.map)
		return trace__set_ev_qualifier_bpf_filter(trace);
3186 3187 3188
	if (trace->syscalls.events.sys_enter)
		return trace__set_ev_qualifier_tp_filter(trace);
	return 0;
3189 3190
}

3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208
static int bpf_map__set_filter_pids(struct bpf_map *map __maybe_unused,
				    size_t npids __maybe_unused, pid_t *pids __maybe_unused)
{
	int err = 0;
#ifdef HAVE_LIBBPF_SUPPORT
	bool value = true;
	int map_fd = bpf_map__fd(map);
	size_t i;

	for (i = 0; i < npids; ++i) {
		err = bpf_map_update_elem(map_fd, &pids[i], &value, BPF_ANY);
		if (err)
			break;
	}
#endif
	return err;
}

3209 3210
static int trace__set_filter_loop_pids(struct trace *trace)
{
3211
	unsigned int nr = 1, err;
3212 3213 3214
	pid_t pids[32] = {
		getpid(),
	};
3215 3216 3217 3218 3219 3220 3221 3222
	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;

3223 3224
		if (!strcmp(thread__comm_str(parent), "sshd") ||
		    strstarts(thread__comm_str(parent), "gnome-terminal")) {
3225 3226 3227 3228 3229
			pids[nr++] = parent->tid;
			break;
		}
		thread = parent;
	}
3230

3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253
	err = perf_evlist__set_tp_filter_pids(trace->evlist, nr, pids);
	if (!err && trace->filter_pids.map)
		err = bpf_map__set_filter_pids(trace->filter_pids.map, nr, pids);

	return err;
}

static int trace__set_filter_pids(struct trace *trace)
{
	int err = 0;
	/*
	 * 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.
	 */
	if (trace->filter_pids.nr > 0) {
		err = perf_evlist__set_tp_filter_pids(trace->evlist, trace->filter_pids.nr,
						      trace->filter_pids.entries);
		if (!err && trace->filter_pids.map) {
			err = bpf_map__set_filter_pids(trace->filter_pids.map, trace->filter_pids.nr,
						       trace->filter_pids.entries);
		}
3254
	} else if (perf_thread_map__pid(trace->evlist->core.threads, 0) == -1) {
3255 3256 3257 3258
		err = trace__set_filter_loop_pids(trace);
	}

	return err;
3259 3260
}

3261
static int __trace__deliver_event(struct trace *trace, union perf_event *event)
3262
{
3263
	struct evlist *evlist = trace->evlist;
3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275
	struct perf_sample sample;
	int err;

	err = perf_evlist__parse_sample(evlist, event, &sample);
	if (err)
		fprintf(trace->output, "Can't parse sample, err = %d, skipping...\n", err);
	else
		trace__handle_event(trace, event, &sample);

	return 0;
}

3276
static int __trace__flush_events(struct trace *trace)
3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287
{
	u64 first = ordered_events__first_time(&trace->oe.data);
	u64 flush = trace->oe.last - NSEC_PER_SEC;

	/* Is there some thing to flush.. */
	if (first && first < flush)
		return ordered_events__flush_time(&trace->oe.data, flush);

	return 0;
}

3288 3289 3290 3291 3292
static int trace__flush_events(struct trace *trace)
{
	return !trace->sort_events ? 0 : __trace__flush_events(trace);
}

3293
static int trace__deliver_event(struct trace *trace, union perf_event *event)
3294 3295 3296
{
	int err;

3297 3298 3299 3300
	if (!trace->sort_events)
		return __trace__deliver_event(trace, event);

	err = perf_evlist__parse_sample_timestamp(trace->evlist, event, &trace->oe.last);
3301 3302 3303 3304 3305 3306 3307
	if (err && err != -1)
		return err;

	err = ordered_events__queue(&trace->oe.data, event, trace->oe.last, 0);
	if (err)
		return err;

3308
	return trace__flush_events(trace);
3309 3310 3311 3312 3313 3314 3315
}

static int ordered_events__deliver_event(struct ordered_events *oe,
					 struct ordered_event *event)
{
	struct trace *trace = container_of(oe, struct trace, oe.data);

3316
	return __trace__deliver_event(trace, event->event);
3317 3318
}

3319
static int trace__run(struct trace *trace, int argc, const char **argv)
3320
{
3321
	struct evlist *evlist = trace->evlist;
3322
	struct evsel *evsel, *pgfault_maj = NULL, *pgfault_min = NULL;
3323 3324
	int err = -1, i;
	unsigned long before;
3325
	const bool forks = argc > 0;
3326
	bool draining = false;
3327

3328 3329
	trace->live = true;

3330 3331 3332
	if (!trace->raw_augmented_syscalls) {
		if (trace->trace_syscalls && trace__add_syscall_newtp(trace))
			goto out_error_raw_syscalls;
3333

3334
		if (trace->trace_syscalls)
3335
			trace->vfs_getname = evlist__add_vfs_getname(evlist);
3336
	}
3337

3338 3339 3340 3341
	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;
3342
		perf_evsel__config_callchain(pgfault_maj, &trace->opts, &callchain_param);
3343
		evlist__add(evlist, pgfault_maj);
3344
	}
3345

3346 3347 3348 3349
	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;
3350
		perf_evsel__config_callchain(pgfault_min, &trace->opts, &callchain_param);
3351
		evlist__add(evlist, pgfault_min);
3352
	}
3353

3354
	if (trace->sched &&
3355 3356 3357
	    perf_evlist__add_newtp(evlist, "sched", "sched_stat_runtime",
				   trace__sched_stat_runtime))
		goto out_error_sched_stat_runtime;
3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385
	/*
	 * 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);

3386 3387
	err = perf_evlist__create_maps(evlist, &trace->opts.target);
	if (err < 0) {
3388
		fprintf(trace->output, "Problems parsing the target to trace, check your options!\n");
3389 3390 3391
		goto out_delete_evlist;
	}

3392 3393
	err = trace__symbols_init(trace, evlist);
	if (err < 0) {
3394
		fprintf(trace->output, "Problems initializing symbol libraries!\n");
3395
		goto out_delete_evlist;
3396 3397
	}

3398
	perf_evlist__config(evlist, &trace->opts, &callchain_param);
3399

3400 3401 3402 3403
	signal(SIGCHLD, sig_handler);
	signal(SIGINT, sig_handler);

	if (forks) {
3404
		err = perf_evlist__prepare_workload(evlist, &trace->opts.target,
3405
						    argv, false, NULL);
3406
		if (err < 0) {
3407
			fprintf(trace->output, "Couldn't run the workload!\n");
3408
			goto out_delete_evlist;
3409 3410 3411
		}
	}

3412
	err = evlist__open(evlist);
3413 3414
	if (err < 0)
		goto out_error_open;
3415

3416 3417 3418 3419 3420 3421 3422 3423 3424 3425
	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;
	}

3426
	err = trace__set_filter_pids(trace);
3427 3428 3429
	if (err < 0)
		goto out_error_mem;

3430 3431 3432
	if (trace->syscalls.map)
		trace__init_syscalls_bpf_map(trace);

3433 3434 3435
	if (trace->syscalls.prog_array.sys_enter)
		trace__init_syscalls_bpf_prog_array_maps(trace);

3436 3437 3438 3439 3440
	if (trace->ev_qualifier_ids.nr > 0) {
		err = trace__set_ev_qualifier_filter(trace);
		if (err < 0)
			goto out_errno;

3441 3442 3443 3444
		if (trace->syscalls.events.sys_exit) {
			pr_debug("event qualifier tracepoint filter: %s\n",
				 trace->syscalls.events.sys_exit->filter);
		}
3445
	}
3446

3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459
	/*
	 * If the "close" syscall is not traced, then we will not have the
	 * opportunity to, in syscall_arg__scnprintf_close_fd() invalidate the
	 * fd->pathname table and were ending up showing the last value set by
	 * syscalls opening a pathname and associating it with a descriptor or
	 * reading it from /proc/pid/fd/ in cases where that doesn't make
	 * sense.
	 *
	 *  So just disable this beautifier (SCA_FD, SCA_FDAT) when 'close' is
	 *  not in use.
	 */
	trace->fd_path_disabled = !trace__syscall_enabled(trace, syscalltbl__id(trace->sctbl, "close"));

3460 3461 3462
	err = perf_evlist__apply_filters(evlist, &evsel);
	if (err < 0)
		goto out_error_apply_filters;
3463

3464 3465 3466
	if (trace->dump.map)
		bpf_map__fprintf(trace->dump.map, trace->output);

3467
	err = evlist__mmap(evlist, trace->opts.mmap_pages);
3468 3469
	if (err < 0)
		goto out_error_mmap;
3470

3471
	if (!target__none(&trace->opts.target) && !trace->opts.initial_delay)
3472
		evlist__enable(evlist);
3473

3474 3475 3476
	if (forks)
		perf_evlist__start_workload(evlist);

3477 3478
	if (trace->opts.initial_delay) {
		usleep(trace->opts.initial_delay * 1000);
3479
		evlist__enable(evlist);
3480 3481
	}

3482
	trace->multiple_threads = perf_thread_map__pid(evlist->core.threads, 0) == -1 ||
3483
				  evlist->core.threads->nr > 1 ||
3484
				  evlist__first(evlist)->core.attr.inherit;
3485 3486

	/*
3487 3488
	 * Now that we already used evsel->core.attr to ask the kernel to setup the
	 * events, lets reuse evsel->core.attr.sample_max_stack as the limit in
3489
	 * trace__resolve_callchain(), allowing per-event max-stack settings
3490
	 * to override an explicitly set --max-stack global setting.
3491 3492
	 */
	evlist__for_each_entry(evlist, evsel) {
3493
		if (evsel__has_callchain(evsel) &&
3494 3495
		    evsel->core.attr.sample_max_stack == 0)
			evsel->core.attr.sample_max_stack = trace->max_stack;
3496
	}
3497
again:
3498
	before = trace->nr_events;
3499

3500
	for (i = 0; i < evlist->core.nr_mmaps; i++) {
3501
		union perf_event *event;
3502
		struct mmap *md;
3503

3504
		md = &evlist->mmap[i];
3505
		if (perf_mmap__read_init(md) < 0)
3506 3507
			continue;

3508
		while ((event = perf_mmap__read_event(md)) != NULL) {
3509
			++trace->nr_events;
3510

3511
			err = trace__deliver_event(trace, event);
3512 3513
			if (err)
				goto out_disable;
3514

3515
			perf_mmap__consume(md);
3516

3517 3518
			if (interrupted)
				goto out_disable;
3519 3520

			if (done && !draining) {
3521
				evlist__disable(evlist);
3522 3523
				draining = true;
			}
3524
		}
3525
		perf_mmap__read_done(md);
3526 3527
	}

3528
	if (trace->nr_events == before) {
3529
		int timeout = done ? 100 : -1;
3530

3531
		if (!draining && evlist__poll(evlist, timeout) > 0) {
3532
			if (evlist__filter_pollfd(evlist, POLLERR | POLLHUP | POLLNVAL) == 0)
3533 3534
				draining = true;

3535
			goto again;
3536
		} else {
3537
			if (trace__flush_events(trace))
3538
				goto out_disable;
3539
		}
3540 3541
	} else {
		goto again;
3542 3543
	}

3544
out_disable:
3545 3546
	thread__zput(trace->current);

3547
	evlist__disable(evlist);
3548

3549 3550
	if (trace->sort_events)
		ordered_events__flush(&trace->oe.data, OE_FLUSH__FINAL);
3551

3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563
	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);
		}
	}
3564

3565
out_delete_evlist:
3566 3567
	trace__symbols__exit(trace);

3568
	evlist__delete(evlist);
3569
	cgroup__put(trace->cgroup);
3570
	trace->evlist = NULL;
3571
	trace->live = false;
3572
	return err;
3573 3574
{
	char errbuf[BUFSIZ];
3575

3576
out_error_sched_stat_runtime:
3577
	tracing_path__strerror_open_tp(errno, errbuf, sizeof(errbuf), "sched", "sched_stat_runtime");
3578 3579
	goto out_error;

3580
out_error_raw_syscalls:
3581
	tracing_path__strerror_open_tp(errno, errbuf, sizeof(errbuf), "raw_syscalls", "sys_(enter|exit)");
3582 3583
	goto out_error;

3584 3585 3586 3587
out_error_mmap:
	perf_evlist__strerror_mmap(evlist, errno, errbuf, sizeof(errbuf));
	goto out_error;

3588 3589 3590 3591
out_error_open:
	perf_evlist__strerror_open(evlist, errno, errbuf, sizeof(errbuf));

out_error:
3592
	fprintf(trace->output, "%s\n", errbuf);
3593
	goto out_delete_evlist;
3594 3595 3596 3597 3598

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,
3599
		str_error_r(errno, errbuf, sizeof(errbuf)));
3600
	goto out_delete_evlist;
3601
}
3602 3603 3604
out_error_mem:
	fprintf(trace->output, "Not enough memory to run!\n");
	goto out_delete_evlist;
3605 3606 3607 3608

out_errno:
	fprintf(trace->output, "errno=%d,%s\n", errno, strerror(errno));
	goto out_delete_evlist;
3609
}
3610

3611 3612
static int trace__replay(struct trace *trace)
{
3613
	const struct evsel_str_handler handlers[] = {
3614
		{ "probe:vfs_getname",	     trace__vfs_getname, },
3615
	};
3616
	struct perf_data data = {
3617 3618 3619
		.path  = input_name,
		.mode  = PERF_DATA_MODE_READ,
		.force = trace->force,
3620
	};
3621
	struct perf_session *session;
3622
	struct evsel *evsel;
3623 3624 3625 3626
	int err = -1;

	trace->tool.sample	  = trace__process_sample;
	trace->tool.mmap	  = perf_event__process_mmap;
David Ahern's avatar
David Ahern committed
3627
	trace->tool.mmap2	  = perf_event__process_mmap2;
3628 3629 3630 3631
	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;
3632
	trace->tool.tracing_data  = perf_event__process_tracing_data;
3633
	trace->tool.build_id	  = perf_event__process_build_id;
3634
	trace->tool.namespaces	  = perf_event__process_namespaces;
3635

3636
	trace->tool.ordered_events = true;
3637 3638 3639 3640 3641
	trace->tool.ordering_requires_timestamps = true;

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

3642
	session = perf_session__new(&data, false, &trace->tool);
3643 3644
	if (IS_ERR(session))
		return PTR_ERR(session);
3645

3646 3647 3648 3649 3650 3651
	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);

3652
	if (symbol__init(&session->header.env) < 0)
3653 3654
		goto out;

3655 3656
	trace->host = &session->machines.host;

3657 3658 3659 3660
	err = perf_session__set_tracepoints_handlers(session, handlers);
	if (err)
		goto out;

3661 3662
	evsel = perf_evlist__find_tracepoint_by_name(session->evlist,
						     "raw_syscalls:sys_enter");
3663 3664 3665 3666
	/* older kernels have syscalls tp versus raw_syscalls */
	if (evsel == NULL)
		evsel = perf_evlist__find_tracepoint_by_name(session->evlist,
							     "syscalls:sys_enter");
3667

3668
	if (evsel &&
3669
	    (perf_evsel__init_raw_syscall_tp(evsel, trace__sys_enter) < 0 ||
3670
	    perf_evsel__init_sc_tp_ptr_field(evsel, args))) {
3671 3672 3673 3674 3675 3676
		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");
3677 3678 3679
	if (evsel == NULL)
		evsel = perf_evlist__find_tracepoint_by_name(session->evlist,
							     "syscalls:sys_exit");
3680
	if (evsel &&
3681
	    (perf_evsel__init_raw_syscall_tp(evsel, trace__sys_exit) < 0 ||
3682
	    perf_evsel__init_sc_tp_uint_field(evsel, ret))) {
3683
		pr_err("Error during initialize raw_syscalls:sys_exit event\n");
3684 3685 3686
		goto out;
	}

3687
	evlist__for_each_entry(session->evlist, evsel) {
3688 3689 3690 3691
		if (evsel->core.attr.type == PERF_TYPE_SOFTWARE &&
		    (evsel->core.attr.config == PERF_COUNT_SW_PAGE_FAULTS_MAJ ||
		     evsel->core.attr.config == PERF_COUNT_SW_PAGE_FAULTS_MIN ||
		     evsel->core.attr.config == PERF_COUNT_SW_PAGE_FAULTS))
3692 3693 3694
			evsel->handler = trace__pgfault;
	}

3695 3696
	setup_pager();

3697
	err = perf_session__process_events(session);
3698 3699 3700
	if (err)
		pr_err("Failed to process events, error %d", err);

3701 3702 3703
	else if (trace->summary)
		trace__fprintf_thread_summary(trace, trace->output);

3704 3705 3706 3707 3708 3709
out:
	perf_session__delete(session);

	return err;
}

3710 3711 3712 3713
static size_t trace__fprintf_threads_header(FILE *fp)
{
	size_t printed;

3714
	printed  = fprintf(fp, "\n Summary of events:\n\n");
3715 3716 3717 3718

	return printed;
}

3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732
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;
}

3733 3734 3735 3736 3737
static size_t thread__dump_stats(struct thread_trace *ttrace,
				 struct trace *trace, FILE *fp)
{
	size_t printed = 0;
	struct syscall *sc;
3738 3739
	struct rb_node *nd;
	DECLARE_RESORT_RB_INTLIST(syscall_stats, ttrace->syscall_stats);
3740

3741
	if (syscall_stats == NULL)
3742 3743 3744 3745
		return 0;

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

3746 3747 3748
	printed += fprintf(fp, "   syscall            calls    total       min       avg       max      stddev\n");
	printed += fprintf(fp, "                               (msec)    (msec)    (msec)    (msec)        (%%)\n");
	printed += fprintf(fp, "   --------------- -------- --------- --------- --------- ---------     ------\n");
3749

3750
	resort_rb__for_each_entry(nd, syscall_stats) {
3751
		struct stats *stats = syscall_stats_entry->stats;
3752 3753 3754 3755 3756 3757 3758 3759 3760 3761
		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;

3762
			sc = &trace->syscalls.table[syscall_stats_entry->syscall];
3763
			printed += fprintf(fp, "   %-15s", sc->name);
3764
			printed += fprintf(fp, " %8" PRIu64 " %9.3f %9.3f %9.3f",
3765
					   n, syscall_stats_entry->msecs, min, avg);
3766
			printed += fprintf(fp, " %9.3f %9.2f%%\n", max, pct);
3767 3768 3769
		}
	}

3770
	resort_rb__delete(syscall_stats);
3771
	printed += fprintf(fp, "\n\n");
3772 3773 3774 3775

	return printed;
}

3776
static size_t trace__fprintf_thread(FILE *fp, struct thread *thread, struct trace *trace)
3777
{
3778
	size_t printed = 0;
3779
	struct thread_trace *ttrace = thread__priv(thread);
3780 3781 3782 3783 3784 3785 3786
	double ratio;

	if (ttrace == NULL)
		return 0;

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

3787
	printed += fprintf(fp, " %s (%d), ", thread__comm_str(thread), thread->tid);
3788
	printed += fprintf(fp, "%lu events, ", ttrace->nr_events);
3789
	printed += fprintf(fp, "%.1f%%", ratio);
3790 3791 3792 3793
	if (ttrace->pfmaj)
		printed += fprintf(fp, ", %lu majfaults", ttrace->pfmaj);
	if (ttrace->pfmin)
		printed += fprintf(fp, ", %lu minfaults", ttrace->pfmin);
3794 3795 3796 3797 3798
	if (trace->sched)
		printed += fprintf(fp, ", %.3f msec\n", ttrace->runtime_ms);
	else if (fputc('\n', fp) != EOF)
		++printed;

3799
	printed += thread__dump_stats(ttrace, trace, fp);
3800

3801 3802
	return printed;
}
3803

3804 3805 3806 3807 3808 3809 3810 3811 3812 3813
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);
3814 3815
}

3816 3817
static size_t trace__fprintf_thread_summary(struct trace *trace, FILE *fp)
{
3818 3819
	size_t printed = trace__fprintf_threads_header(fp);
	struct rb_node *nd;
3820
	int i;
3821

3822 3823
	for (i = 0; i < THREADS__TABLE_SIZE; i++) {
		DECLARE_RESORT_RB_MACHINE_THREADS(threads, trace->host, i);
3824

3825 3826 3827 3828
		if (threads == NULL) {
			fprintf(fp, "%s", "Error sorting output by nr_events!\n");
			return 0;
		}
3829

3830 3831
		resort_rb__for_each_entry(nd, threads)
			printed += trace__fprintf_thread(fp, threads_entry->thread, trace);
3832

3833 3834
		resort_rb__delete(threads);
	}
3835
	return printed;
3836 3837
}

3838 3839 3840 3841 3842 3843 3844 3845 3846
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;
}

3847 3848
static int trace__set_filter_pids_from_option(const struct option *opt, const char *str,
					      int unset __maybe_unused)
3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878
{
	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;
}

3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895
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;
}

3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912
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;
}

3913
static void evlist__set_default_evsel_handler(struct evlist *evlist, void *handler)
3914
{
3915
	struct evsel *evsel;
3916

3917 3918 3919 3920
	evlist__for_each_entry(evlist, evsel) {
		if (evsel->handler == NULL)
			evsel->handler = handler;
	}
3921 3922
}

3923
static int evlist__set_syscall_tp_fields(struct evlist *evlist)
3924
{
3925
	struct evsel *evsel;
3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952

	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;
}

3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966
/*
 * 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, };
3967
	int len = strlen(str) + 1, err = -1, list, idx;
3968 3969
	char *strace_groups_dir = system_path(STRACE_GROUPS_DIR);
	char group_name[PATH_MAX];
3970
	struct syscall_fmt *fmt;
3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984

	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;
3985 3986
		if (syscalltbl__id(trace->sctbl, s) >= 0 ||
		    syscalltbl__strglobmatch_first(trace->sctbl, s, &idx) >= 0) {
3987
			list = 1;
3988 3989 3990 3991 3992 3993 3994
			goto do_concat;
		}

		fmt = syscall_fmt__find_by_alias(s);
		if (fmt != NULL) {
			list = 1;
			s = fmt->name;
3995 3996 3997 3998 3999
		} else {
			path__join(group_name, sizeof(group_name), strace_groups_dir, s);
			if (access(group_name, R_OK) == 0)
				list = 1;
		}
4000
do_concat:
4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029
		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;
4030
		trace->trace_syscalls = true;
4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047
	}

	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;
}

4048 4049 4050 4051
static int trace__parse_cgroups(const struct option *opt, const char *str, int unset)
{
	struct trace *trace = opt->value;

4052
	if (!list_empty(&trace->evlist->core.entries))
4053 4054 4055 4056 4057 4058 4059
		return parse_cgroups(opt, str, unset);

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

	return 0;
}

4060
static struct bpf_map *trace__find_bpf_map_by_name(struct trace *trace, const char *name)
4061
{
4062 4063
	if (trace->bpf_obj == NULL)
		return NULL;
4064

4065
	return bpf_object__find_map_by_name(trace->bpf_obj, name);
4066 4067 4068 4069
}

static void trace__set_bpf_map_filtered_pids(struct trace *trace)
{
4070
	trace->filter_pids.map = trace__find_bpf_map_by_name(trace, "pids_filtered");
4071 4072
}

4073 4074
static void trace__set_bpf_map_syscalls(struct trace *trace)
{
4075
	trace->syscalls.map = trace__find_bpf_map_by_name(trace, "syscalls");
4076 4077
	trace->syscalls.prog_array.sys_enter = trace__find_bpf_map_by_name(trace, "syscalls_sys_enter");
	trace->syscalls.prog_array.sys_exit  = trace__find_bpf_map_by_name(trace, "syscalls_sys_exit");
4078 4079
}

4080 4081
static int trace__config(const char *var, const char *value, void *arg)
{
4082
	struct trace *trace = arg;
4083 4084 4085
	int err = 0;

	if (!strcmp(var, "trace.add_events")) {
4086 4087 4088 4089 4090
		trace->perfconfig_events = strdup(value);
		if (trace->perfconfig_events == NULL) {
			pr_err("Not enough memory for %s\n", "trace.add_events");
			return -1;
		}
4091 4092
	} else if (!strcmp(var, "trace.show_timestamp")) {
		trace->show_tstamp = perf_config_bool(var, value);
4093 4094
	} else if (!strcmp(var, "trace.show_duration")) {
		trace->show_duration = perf_config_bool(var, value);
4095 4096 4097 4098
	} else if (!strcmp(var, "trace.show_arg_names")) {
		trace->show_arg_names = perf_config_bool(var, value);
		if (!trace->show_arg_names)
			trace->show_zeros = true;
4099
	} else if (!strcmp(var, "trace.show_zeros")) {
4100 4101 4102 4103 4104 4105
		bool new_show_zeros = perf_config_bool(var, value);
		if (!trace->show_arg_names && !new_show_zeros) {
			pr_warning("trace.show_zeros has to be set when trace.show_arg_names=no\n");
			goto out;
		}
		trace->show_zeros = new_show_zeros;
4106 4107
	} else if (!strcmp(var, "trace.show_prefix")) {
		trace->show_string_prefix = perf_config_bool(var, value);
4108 4109
	} else if (!strcmp(var, "trace.no_inherit")) {
		trace->opts.no_inherit = perf_config_bool(var, value);
4110 4111 4112 4113
	} else if (!strcmp(var, "trace.args_alignment")) {
		int args_alignment = 0;
		if (perf_config_int(&args_alignment, var, value) == 0)
			trace->args_alignment = args_alignment;
4114
	}
4115
out:
4116 4117 4118
	return err;
}

4119
int cmd_trace(int argc, const char **argv)
4120
{
4121
	const char *trace_usage[] = {
4122 4123
		"perf trace [<options>] [<command>]",
		"perf trace [<options>] -- <command> [<options>]",
David Ahern's avatar
David Ahern committed
4124 4125
		"perf trace record [<options>] [<command>]",
		"perf trace record [<options>] -- <command> [<options>]",
4126 4127 4128 4129 4130 4131 4132 4133 4134 4135
		NULL
	};
	struct trace trace = {
		.opts = {
			.target = {
				.uid	   = UINT_MAX,
				.uses_mmap = true,
			},
			.user_freq     = UINT_MAX,
			.user_interval = ULLONG_MAX,
4136
			.no_buffering  = true,
4137
			.mmap_pages    = UINT_MAX,
4138
		},
4139
		.output = stderr,
4140
		.show_comm = true,
4141
		.show_tstamp = true,
4142
		.show_duration = true,
4143
		.show_arg_names = true,
4144
		.args_alignment = 70,
4145
		.trace_syscalls = false,
4146
		.kernel_syscallchains = false,
4147
		.max_stack = UINT_MAX,
4148
		.max_events = ULONG_MAX,
4149
	};
4150
	const char *map_dump_str = NULL;
4151
	const char *output_name = NULL;
4152
	const struct option trace_options[] = {
4153 4154 4155
	OPT_CALLBACK('e', "event", &trace, "event",
		     "event/syscall selector. use 'perf list' to list available events",
		     trace__parse_events_option),
4156 4157
	OPT_BOOLEAN(0, "comm", &trace.show_comm,
		    "show the thread COMM next to its id"),
4158
	OPT_BOOLEAN(0, "tool_stats", &trace.show_tool_stats, "show tool stats"),
4159 4160
	OPT_CALLBACK(0, "expr", &trace, "expr", "list of syscalls/events to trace",
		     trace__parse_events_option),
4161
	OPT_STRING('o', "output", &output_name, "file", "output file name"),
4162
	OPT_STRING('i', "input", &input_name, "file", "Analyze events in file"),
4163 4164
	OPT_STRING('p', "pid", &trace.opts.target.pid, "pid",
		    "trace events on existing process id"),
4165
	OPT_STRING('t', "tid", &trace.opts.target.tid, "tid",
4166
		    "trace events on existing thread id"),
4167
	OPT_CALLBACK(0, "filter-pids", &trace, "CSV list of pids",
4168
		     "pids to filter (by the kernel)", trace__set_filter_pids_from_option),
4169
	OPT_BOOLEAN('a', "all-cpus", &trace.opts.target.system_wide,
4170
		    "system-wide collection from all CPUs"),
4171
	OPT_STRING('C', "cpu", &trace.opts.target.cpu_list, "cpu",
4172
		    "list of cpus to monitor"),
4173
	OPT_BOOLEAN(0, "no-inherit", &trace.opts.no_inherit,
4174
		    "child tasks do not inherit counters"),
4175 4176 4177
	OPT_CALLBACK('m', "mmap-pages", &trace.opts.mmap_pages, "pages",
		     "number of mmap data pages",
		     perf_evlist__parse_mmap_pages),
4178
	OPT_STRING('u', "uid", &trace.opts.target.uid_str, "user",
4179
		   "user to profile"),
4180 4181 4182
	OPT_CALLBACK(0, "duration", &trace, "float",
		     "show only events with duration > N.M ms",
		     trace__set_duration),
4183 4184 4185
#ifdef HAVE_LIBBPF_SUPPORT
	OPT_STRING(0, "map-dump", &map_dump_str, "BPF map", "BPF map to periodically dump"),
#endif
4186
	OPT_BOOLEAN(0, "sched", &trace.sched, "show blocking scheduler events"),
4187
	OPT_INCR('v', "verbose", &verbose, "be more verbose"),
4188 4189
	OPT_BOOLEAN('T', "time", &trace.full_time,
		    "Show full timestamp, not time relative to first start"),
4190 4191
	OPT_BOOLEAN(0, "failure", &trace.failure_only,
		    "Show only syscalls that failed"),
4192 4193 4194 4195
	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"),
4196 4197
	OPT_CALLBACK_DEFAULT('F', "pf", &trace.trace_pgfaults, "all|maj|min",
		     "Trace pagefaults", parse_pagefaults, "maj"),
4198
	OPT_BOOLEAN(0, "syscalls", &trace.trace_syscalls, "Trace syscalls"),
4199
	OPT_BOOLEAN('f', "force", &trace.force, "don't complain, do it"),
4200 4201 4202
	OPT_CALLBACK(0, "call-graph", &trace.opts,
		     "record_mode[,record_size]", record_callchain_help,
		     &record_parse_callchain_opt),
4203 4204
	OPT_BOOLEAN(0, "kernel-syscall-graph", &trace.kernel_syscallchains,
		    "Show the kernel callchains on the syscall exit path"),
4205 4206
	OPT_ULONG(0, "max-events", &trace.max_events,
		"Set the maximum number of events to print, exit after that is reached. "),
4207 4208 4209
	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."),
4210 4211 4212
	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. "
4213
		     "Default: kernel.perf_event_max_stack or " __stringify(PERF_MAX_STACK_DEPTH)),
4214 4215
	OPT_BOOLEAN(0, "sort-events", &trace.sort_events,
			"Sort batch of events before processing, use if getting out of order events"),
4216 4217
	OPT_BOOLEAN(0, "print-sample", &trace.print_sample,
			"print the PERF_RECORD_SAMPLE PERF_SAMPLE_ info, for debugging"),
4218
	OPT_UINTEGER(0, "proc-map-timeout", &proc_map_timeout,
4219
			"per thread proc mmap processing timeout in ms"),
4220 4221
	OPT_CALLBACK('G', "cgroup", &trace, "name", "monitor event in cgroup name only",
		     trace__parse_cgroups),
4222 4223 4224
	OPT_UINTEGER('D', "delay", &trace.opts.initial_delay,
		     "ms to wait before starting measurement after program "
		     "start"),
4225
	OPTS_EVSWITCH(&trace.evswitch),
4226 4227
	OPT_END()
	};
4228
	bool __maybe_unused max_stack_user_set = true;
4229
	bool mmap_pages_user_set = true;
4230
	struct evsel *evsel;
4231
	const char * const trace_subcommands[] = { "record", NULL };
4232
	int err = -1;
4233
	char bf[BUFSIZ];
4234

4235 4236 4237
	signal(SIGSEGV, sighandler_dump_stack);
	signal(SIGFPE, sighandler_dump_stack);

4238
	trace.evlist = evlist__new();
4239
	trace.sctbl = syscalltbl__new();
4240

4241
	if (trace.evlist == NULL || trace.sctbl == NULL) {
4242
		pr_err("Not enough memory to run!\n");
4243
		err = -ENOMEM;
4244 4245 4246
		goto out;
	}

4247 4248 4249 4250 4251 4252 4253 4254 4255
	/*
	 * Parsing .perfconfig may entail creating a BPF event, that may need
	 * to create BPF maps, so bump RLIM_MEMLOCK as the default 64K setting
	 * is too small. This affects just this process, not touching the
	 * global setting. If it fails we'll get something in 'perf trace -v'
	 * to help diagnose the problem.
	 */
	rlimit__bump_memlock();

4256 4257 4258 4259
	err = perf_config(trace__config, &trace);
	if (err)
		goto out;

4260 4261
	argc = parse_options_subcommand(argc, argv, trace_options, trace_subcommands,
				 trace_usage, PARSE_OPT_STOP_AT_NON_OPTION);
4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277

	/*
	 * Here we already passed thru trace__parse_events_option() and it has
	 * already figured out if -e syscall_name, if not but if --event
	 * foo:bar was used, the user is interested _just_ in those, say,
	 * tracepoint events, not in the strace-like syscall-name-based mode.
	 *
	 * This is important because we need to check if strace-like mode is
	 * needed to decided if we should filter out the eBPF
	 * __augmented_syscalls__ code, if it is in the mix, say, via
	 * .perfconfig trace.add_events, and filter those out.
	 */
	if (!trace.trace_syscalls && !trace.trace_pgfaults &&
	    trace.evlist->core.nr_entries == 0 /* Was --events used? */) {
		trace.trace_syscalls = true;
	}
4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292
	/*
	 * Now that we have --verbose figured out, lets see if we need to parse
	 * events from .perfconfig, so that if those events fail parsing, say some
	 * BPF program fails, then we'll be able to use --verbose to see what went
	 * wrong in more detail.
	 */
	if (trace.perfconfig_events != NULL) {
		struct parse_events_error parse_err = { .idx = 0, };

		err = parse_events(trace.evlist, trace.perfconfig_events, &parse_err);
		if (err) {
			parse_events_print_error(&parse_err, trace.perfconfig_events);
			goto out;
		}
	}
4293

4294 4295 4296 4297 4298
	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");
	}

4299 4300 4301
	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));
4302 4303 4304 4305
		pr_err("ERROR: Setup trace syscalls enter failed: %s\n", bf);
		goto out;
	}

4306
	if (evsel) {
4307
		trace.syscalls.events.augmented = evsel;
4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321

		evsel = perf_evlist__find_tracepoint_by_name(trace.evlist, "raw_syscalls:sys_enter");
		if (evsel == NULL) {
			pr_err("ERROR: raw_syscalls:sys_enter not found in the augmented BPF object\n");
			goto out;
		}

		if (evsel->bpf_obj == NULL) {
			pr_err("ERROR: raw_syscalls:sys_enter not associated to a BPF object\n");
			goto out;
		}

		trace.bpf_obj = evsel->bpf_obj;

4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360
		/*
		 * If we have _just_ the augmenter event but don't have a
		 * explicit --syscalls, then assume we want all strace-like
		 * syscalls:
		 */
		if (!trace.trace_syscalls && trace__only_augmented_syscalls_evsels(&trace))
			trace.trace_syscalls = true;
		/*
		 * So, if we have a syscall augmenter, but trace_syscalls, aka
		 * strace-like syscall tracing is not set, then we need to trow
		 * away the augmenter, i.e. all the events that were created
		 * from that BPF object file.
		 *
		 * This is more to fix the current .perfconfig trace.add_events
		 * style of setting up the strace-like eBPF based syscall point
		 * payload augmenter.
		 *
		 * All this complexity will be avoided by adding an alternative
		 * to trace.add_events in the form of
		 * trace.bpf_augmented_syscalls, that will be only parsed if we
		 * need it.
		 *
		 * .perfconfig trace.add_events is still useful if we want, for
		 * instance, have msr_write.msr in some .perfconfig profile based
		 * 'perf trace --config determinism.profile' mode, where for some
		 * particular goal/workload type we want a set of events and
		 * output mode (with timings, etc) instead of having to add
		 * all via the command line.
		 *
		 * Also --config to specify an alternate .perfconfig file needs
		 * to be implemented.
		 */
		if (!trace.trace_syscalls) {
			trace__delete_augmented_syscalls(&trace);
		} else {
			trace__set_bpf_map_filtered_pids(&trace);
			trace__set_bpf_map_syscalls(&trace);
			trace.syscalls.unaugmented_prog = trace__find_bpf_program_by_title(&trace, "!raw_syscalls:unaugmented");
		}
4361
	}
4362

4363 4364 4365 4366 4367 4368 4369
	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;
	}

4370 4371
	err = -1;

4372
	if (map_dump_str) {
4373
		trace.dump.map = trace__find_bpf_map_by_name(&trace, map_dump_str);
4374 4375 4376 4377 4378 4379
		if (trace.dump.map == NULL) {
			pr_err("ERROR: BPF map \"%s\" not found\n", map_dump_str);
			goto out;
		}
	}

4380 4381 4382 4383 4384
	if (trace.trace_pgfaults) {
		trace.opts.sample_address = true;
		trace.opts.sample_time = true;
	}

4385 4386 4387
	if (trace.opts.mmap_pages == UINT_MAX)
		mmap_pages_user_set = false;

4388
	if (trace.max_stack == UINT_MAX) {
4389
		trace.max_stack = input_name ? PERF_MAX_STACK_DEPTH : sysctl__max_stack();
4390 4391 4392 4393
		max_stack_user_set = false;
	}

#ifdef HAVE_DWARF_UNWIND_SUPPORT
4394
	if ((trace.min_stack || max_stack_user_set) && !callchain_param.enabled) {
4395
		record_opts__parse_callchain(&trace.opts, &callchain_param, "dwarf", false);
4396
	}
4397 4398
#endif

4399
	if (callchain_param.enabled) {
4400 4401 4402
		if (!mmap_pages_user_set && geteuid() == 0)
			trace.opts.mmap_pages = perf_event_mlock_kb_in_pages() * 4;

4403
		symbol_conf.use_callchain = true;
4404
	}
4405

4406
	if (trace.evlist->core.nr_entries > 0) {
4407
		evlist__set_default_evsel_handler(trace.evlist, trace__event_handler);
4408 4409 4410 4411 4412
		if (evlist__set_syscall_tp_fields(trace.evlist)) {
			perror("failed to set syscalls:* tracepoint fields");
			goto out;
		}
	}
4413

4414 4415 4416 4417
	if (trace.sort_events) {
		ordered_events__init(&trace.oe.data, ordered_events__deliver_event, &trace);
		ordered_events__set_copy_on_queue(&trace.oe.data, true);
	}
4418

4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431
	/*
	 * If we are augmenting syscalls, then combine what we put in the
	 * __augmented_syscalls__ BPF map with what is in the
	 * syscalls:sys_exit_FOO tracepoints, i.e. just like we do without BPF,
	 * combining raw_syscalls:sys_enter with raw_syscalls:sys_exit.
	 *
	 * We'll switch to look at two BPF maps, one for sys_enter and the
	 * other for sys_exit when we start augmenting the sys_exit paths with
	 * buffers that are being copied from kernel to userspace, think 'read'
	 * syscall.
	 */
	if (trace.syscalls.events.augmented) {
		evlist__for_each_entry(trace.evlist, evsel) {
4432 4433 4434 4435 4436 4437 4438
			bool raw_syscalls_sys_exit = strcmp(perf_evsel__name(evsel), "raw_syscalls:sys_exit") == 0;

			if (raw_syscalls_sys_exit) {
				trace.raw_augmented_syscalls = true;
				goto init_augmented_syscall_tp;
			}

4439 4440
			if (trace.syscalls.events.augmented->priv == NULL &&
			    strstr(perf_evsel__name(evsel), "syscalls:sys_enter")) {
4441
				struct evsel *augmented = trace.syscalls.events.augmented;
4442 4443 4444
				if (perf_evsel__init_augmented_syscall_tp(augmented, evsel) ||
				    perf_evsel__init_augmented_syscall_tp_args(augmented))
					goto out;
4445 4446 4447 4448 4449
				/*
				 * Augmented is __augmented_syscalls__ BPF_OUTPUT event
				 * Above we made sure we can get from the payload the tp fields
				 * that we get from syscalls:sys_enter tracefs format file.
				 */
4450
				augmented->handler = trace__sys_enter;
4451 4452 4453 4454 4455 4456 4457 4458 4459 4460
				/*
				 * Now we do the same for the *syscalls:sys_enter event so that
				 * if we handle it directly, i.e. if the BPF prog returns 0 so
				 * as not to filter it, then we'll handle it just like we would
				 * for the BPF_OUTPUT one:
				 */
				if (perf_evsel__init_augmented_syscall_tp(evsel, evsel) ||
				    perf_evsel__init_augmented_syscall_tp_args(evsel))
					goto out;
				evsel->handler = trace__sys_enter;
4461 4462
			}

4463
			if (strstarts(perf_evsel__name(evsel), "syscalls:sys_exit_")) {
4464
				struct syscall_tp *sc;
4465
init_augmented_syscall_tp:
4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489
				if (perf_evsel__init_augmented_syscall_tp(evsel, evsel))
					goto out;
				sc = evsel->priv;
				/*
				 * For now with BPF raw_augmented we hook into
				 * raw_syscalls:sys_enter and there we get all
				 * 6 syscall args plus the tracepoint common
				 * fields and the syscall_nr (another long).
				 * So we check if that is the case and if so
				 * don't look after the sc->args_size but
				 * always after the full raw_syscalls:sys_enter
				 * payload, which is fixed.
				 *
				 * We'll revisit this later to pass
				 * s->args_size to the BPF augmenter (now
				 * tools/perf/examples/bpf/augmented_raw_syscalls.c,
				 * so that it copies only what we need for each
				 * syscall, like what happens when we use
				 * syscalls:sys_enter_NAME, so that we reduce
				 * the kernel/userspace traffic to just what is
				 * needed for each syscall.
				 */
				if (trace.raw_augmented_syscalls)
					trace.raw_augmented_syscalls_args_size = (6 + 1) * sizeof(long) + sc->id.offset;
4490 4491 4492 4493 4494 4495
				perf_evsel__init_augmented_syscall_tp_ret(evsel);
				evsel->handler = trace__sys_exit;
			}
		}
	}

4496 4497 4498 4499 4500 4501 4502
	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;

4503 4504 4505 4506 4507 4508 4509 4510
	if (output_name != NULL) {
		err = trace__open_output(&trace, output_name);
		if (err < 0) {
			perror("failed to create output file");
			goto out;
		}
	}

4511 4512 4513 4514
	err = evswitch__init(&trace.evswitch, trace.evlist, stderr);
	if (err)
		goto out_close;

4515
	err = target__validate(&trace.opts.target);
4516
	if (err) {
4517
		target__strerror(&trace.opts.target, err, bf, sizeof(bf));
4518 4519
		fprintf(trace.output, "%s", bf);
		goto out_close;
4520 4521
	}

4522
	err = target__parse_uid(&trace.opts.target);
4523
	if (err) {
4524
		target__strerror(&trace.opts.target, err, bf, sizeof(bf));
4525 4526
		fprintf(trace.output, "%s", bf);
		goto out_close;
4527 4528
	}

4529
	if (!argc && target__none(&trace.opts.target))
4530 4531
		trace.opts.target.system_wide = true;

4532 4533 4534 4535
	if (input_name)
		err = trace__replay(&trace);
	else
		err = trace__run(&trace, argc, argv);
4536

4537 4538 4539 4540
out_close:
	if (output_name != NULL)
		fclose(trace.output);
out:
4541
	zfree(&trace.perfconfig_events);
4542
	return err;
4543
}