xref: /linux/kernel/trace/trace_output.c (revision e5c86679d5e864947a52fb31e45a425dea3e7fa9)
1 /*
2  * trace_output.c
3  *
4  * Copyright (C) 2008 Red Hat Inc, Steven Rostedt <srostedt@redhat.com>
5  *
6  */
7 
8 #include <linux/module.h>
9 #include <linux/mutex.h>
10 #include <linux/ftrace.h>
11 #include <linux/sched/clock.h>
12 #include <linux/sched/mm.h>
13 
14 #include "trace_output.h"
15 
16 /* must be a power of 2 */
17 #define EVENT_HASHSIZE	128
18 
19 DECLARE_RWSEM(trace_event_sem);
20 
21 static struct hlist_head event_hash[EVENT_HASHSIZE] __read_mostly;
22 
23 static int next_event_type = __TRACE_LAST_TYPE + 1;
24 
25 enum print_line_t trace_print_bputs_msg_only(struct trace_iterator *iter)
26 {
27 	struct trace_seq *s = &iter->seq;
28 	struct trace_entry *entry = iter->ent;
29 	struct bputs_entry *field;
30 
31 	trace_assign_type(field, entry);
32 
33 	trace_seq_puts(s, field->str);
34 
35 	return trace_handle_return(s);
36 }
37 
38 enum print_line_t trace_print_bprintk_msg_only(struct trace_iterator *iter)
39 {
40 	struct trace_seq *s = &iter->seq;
41 	struct trace_entry *entry = iter->ent;
42 	struct bprint_entry *field;
43 
44 	trace_assign_type(field, entry);
45 
46 	trace_seq_bprintf(s, field->fmt, field->buf);
47 
48 	return trace_handle_return(s);
49 }
50 
51 enum print_line_t trace_print_printk_msg_only(struct trace_iterator *iter)
52 {
53 	struct trace_seq *s = &iter->seq;
54 	struct trace_entry *entry = iter->ent;
55 	struct print_entry *field;
56 
57 	trace_assign_type(field, entry);
58 
59 	trace_seq_puts(s, field->buf);
60 
61 	return trace_handle_return(s);
62 }
63 
64 const char *
65 trace_print_flags_seq(struct trace_seq *p, const char *delim,
66 		      unsigned long flags,
67 		      const struct trace_print_flags *flag_array)
68 {
69 	unsigned long mask;
70 	const char *str;
71 	const char *ret = trace_seq_buffer_ptr(p);
72 	int i, first = 1;
73 
74 	for (i = 0;  flag_array[i].name && flags; i++) {
75 
76 		mask = flag_array[i].mask;
77 		if ((flags & mask) != mask)
78 			continue;
79 
80 		str = flag_array[i].name;
81 		flags &= ~mask;
82 		if (!first && delim)
83 			trace_seq_puts(p, delim);
84 		else
85 			first = 0;
86 		trace_seq_puts(p, str);
87 	}
88 
89 	/* check for left over flags */
90 	if (flags) {
91 		if (!first && delim)
92 			trace_seq_puts(p, delim);
93 		trace_seq_printf(p, "0x%lx", flags);
94 	}
95 
96 	trace_seq_putc(p, 0);
97 
98 	return ret;
99 }
100 EXPORT_SYMBOL(trace_print_flags_seq);
101 
102 const char *
103 trace_print_symbols_seq(struct trace_seq *p, unsigned long val,
104 			const struct trace_print_flags *symbol_array)
105 {
106 	int i;
107 	const char *ret = trace_seq_buffer_ptr(p);
108 
109 	for (i = 0;  symbol_array[i].name; i++) {
110 
111 		if (val != symbol_array[i].mask)
112 			continue;
113 
114 		trace_seq_puts(p, symbol_array[i].name);
115 		break;
116 	}
117 
118 	if (ret == (const char *)(trace_seq_buffer_ptr(p)))
119 		trace_seq_printf(p, "0x%lx", val);
120 
121 	trace_seq_putc(p, 0);
122 
123 	return ret;
124 }
125 EXPORT_SYMBOL(trace_print_symbols_seq);
126 
127 #if BITS_PER_LONG == 32
128 const char *
129 trace_print_flags_seq_u64(struct trace_seq *p, const char *delim,
130 		      unsigned long long flags,
131 		      const struct trace_print_flags_u64 *flag_array)
132 {
133 	unsigned long long mask;
134 	const char *str;
135 	const char *ret = trace_seq_buffer_ptr(p);
136 	int i, first = 1;
137 
138 	for (i = 0;  flag_array[i].name && flags; i++) {
139 
140 		mask = flag_array[i].mask;
141 		if ((flags & mask) != mask)
142 			continue;
143 
144 		str = flag_array[i].name;
145 		flags &= ~mask;
146 		if (!first && delim)
147 			trace_seq_puts(p, delim);
148 		else
149 			first = 0;
150 		trace_seq_puts(p, str);
151 	}
152 
153 	/* check for left over flags */
154 	if (flags) {
155 		if (!first && delim)
156 			trace_seq_puts(p, delim);
157 		trace_seq_printf(p, "0x%llx", flags);
158 	}
159 
160 	trace_seq_putc(p, 0);
161 
162 	return ret;
163 }
164 EXPORT_SYMBOL(trace_print_flags_seq_u64);
165 
166 const char *
167 trace_print_symbols_seq_u64(struct trace_seq *p, unsigned long long val,
168 			 const struct trace_print_flags_u64 *symbol_array)
169 {
170 	int i;
171 	const char *ret = trace_seq_buffer_ptr(p);
172 
173 	for (i = 0;  symbol_array[i].name; i++) {
174 
175 		if (val != symbol_array[i].mask)
176 			continue;
177 
178 		trace_seq_puts(p, symbol_array[i].name);
179 		break;
180 	}
181 
182 	if (ret == (const char *)(trace_seq_buffer_ptr(p)))
183 		trace_seq_printf(p, "0x%llx", val);
184 
185 	trace_seq_putc(p, 0);
186 
187 	return ret;
188 }
189 EXPORT_SYMBOL(trace_print_symbols_seq_u64);
190 #endif
191 
192 const char *
193 trace_print_bitmask_seq(struct trace_seq *p, void *bitmask_ptr,
194 			unsigned int bitmask_size)
195 {
196 	const char *ret = trace_seq_buffer_ptr(p);
197 
198 	trace_seq_bitmask(p, bitmask_ptr, bitmask_size * 8);
199 	trace_seq_putc(p, 0);
200 
201 	return ret;
202 }
203 EXPORT_SYMBOL_GPL(trace_print_bitmask_seq);
204 
205 /**
206  * trace_print_hex_seq - print buffer as hex sequence
207  * @p: trace seq struct to write to
208  * @buf: The buffer to print
209  * @buf_len: Length of @buf in bytes
210  * @concatenate: Print @buf as single hex string or with spacing
211  *
212  * Prints the passed buffer as a hex sequence either as a whole,
213  * single hex string if @concatenate is true or with spacing after
214  * each byte in case @concatenate is false.
215  */
216 const char *
217 trace_print_hex_seq(struct trace_seq *p, const unsigned char *buf, int buf_len,
218 		    bool concatenate)
219 {
220 	int i;
221 	const char *ret = trace_seq_buffer_ptr(p);
222 
223 	for (i = 0; i < buf_len; i++)
224 		trace_seq_printf(p, "%s%2.2x", concatenate || i == 0 ? "" : " ",
225 				 buf[i]);
226 	trace_seq_putc(p, 0);
227 
228 	return ret;
229 }
230 EXPORT_SYMBOL(trace_print_hex_seq);
231 
232 const char *
233 trace_print_array_seq(struct trace_seq *p, const void *buf, int count,
234 		      size_t el_size)
235 {
236 	const char *ret = trace_seq_buffer_ptr(p);
237 	const char *prefix = "";
238 	void *ptr = (void *)buf;
239 	size_t buf_len = count * el_size;
240 
241 	trace_seq_putc(p, '{');
242 
243 	while (ptr < buf + buf_len) {
244 		switch (el_size) {
245 		case 1:
246 			trace_seq_printf(p, "%s0x%x", prefix,
247 					 *(u8 *)ptr);
248 			break;
249 		case 2:
250 			trace_seq_printf(p, "%s0x%x", prefix,
251 					 *(u16 *)ptr);
252 			break;
253 		case 4:
254 			trace_seq_printf(p, "%s0x%x", prefix,
255 					 *(u32 *)ptr);
256 			break;
257 		case 8:
258 			trace_seq_printf(p, "%s0x%llx", prefix,
259 					 *(u64 *)ptr);
260 			break;
261 		default:
262 			trace_seq_printf(p, "BAD SIZE:%zu 0x%x", el_size,
263 					 *(u8 *)ptr);
264 			el_size = 1;
265 		}
266 		prefix = ",";
267 		ptr += el_size;
268 	}
269 
270 	trace_seq_putc(p, '}');
271 	trace_seq_putc(p, 0);
272 
273 	return ret;
274 }
275 EXPORT_SYMBOL(trace_print_array_seq);
276 
277 int trace_raw_output_prep(struct trace_iterator *iter,
278 			  struct trace_event *trace_event)
279 {
280 	struct trace_event_call *event;
281 	struct trace_seq *s = &iter->seq;
282 	struct trace_seq *p = &iter->tmp_seq;
283 	struct trace_entry *entry;
284 
285 	event = container_of(trace_event, struct trace_event_call, event);
286 	entry = iter->ent;
287 
288 	if (entry->type != event->event.type) {
289 		WARN_ON_ONCE(1);
290 		return TRACE_TYPE_UNHANDLED;
291 	}
292 
293 	trace_seq_init(p);
294 	trace_seq_printf(s, "%s: ", trace_event_name(event));
295 
296 	return trace_handle_return(s);
297 }
298 EXPORT_SYMBOL(trace_raw_output_prep);
299 
300 static int trace_output_raw(struct trace_iterator *iter, char *name,
301 			    char *fmt, va_list ap)
302 {
303 	struct trace_seq *s = &iter->seq;
304 
305 	trace_seq_printf(s, "%s: ", name);
306 	trace_seq_vprintf(s, fmt, ap);
307 
308 	return trace_handle_return(s);
309 }
310 
311 int trace_output_call(struct trace_iterator *iter, char *name, char *fmt, ...)
312 {
313 	va_list ap;
314 	int ret;
315 
316 	va_start(ap, fmt);
317 	ret = trace_output_raw(iter, name, fmt, ap);
318 	va_end(ap);
319 
320 	return ret;
321 }
322 EXPORT_SYMBOL_GPL(trace_output_call);
323 
324 #ifdef CONFIG_KRETPROBES
325 static inline const char *kretprobed(const char *name)
326 {
327 	static const char tramp_name[] = "kretprobe_trampoline";
328 	int size = sizeof(tramp_name);
329 
330 	if (strncmp(tramp_name, name, size) == 0)
331 		return "[unknown/kretprobe'd]";
332 	return name;
333 }
334 #else
335 static inline const char *kretprobed(const char *name)
336 {
337 	return name;
338 }
339 #endif /* CONFIG_KRETPROBES */
340 
341 static void
342 seq_print_sym_short(struct trace_seq *s, const char *fmt, unsigned long address)
343 {
344 #ifdef CONFIG_KALLSYMS
345 	char str[KSYM_SYMBOL_LEN];
346 	const char *name;
347 
348 	kallsyms_lookup(address, NULL, NULL, NULL, str);
349 
350 	name = kretprobed(str);
351 
352 	trace_seq_printf(s, fmt, name);
353 #endif
354 }
355 
356 static void
357 seq_print_sym_offset(struct trace_seq *s, const char *fmt,
358 		     unsigned long address)
359 {
360 #ifdef CONFIG_KALLSYMS
361 	char str[KSYM_SYMBOL_LEN];
362 	const char *name;
363 
364 	sprint_symbol(str, address);
365 	name = kretprobed(str);
366 
367 	trace_seq_printf(s, fmt, name);
368 #endif
369 }
370 
371 #ifndef CONFIG_64BIT
372 # define IP_FMT "%08lx"
373 #else
374 # define IP_FMT "%016lx"
375 #endif
376 
377 static int seq_print_user_ip(struct trace_seq *s, struct mm_struct *mm,
378 			     unsigned long ip, unsigned long sym_flags)
379 {
380 	struct file *file = NULL;
381 	unsigned long vmstart = 0;
382 	int ret = 1;
383 
384 	if (s->full)
385 		return 0;
386 
387 	if (mm) {
388 		const struct vm_area_struct *vma;
389 
390 		down_read(&mm->mmap_sem);
391 		vma = find_vma(mm, ip);
392 		if (vma) {
393 			file = vma->vm_file;
394 			vmstart = vma->vm_start;
395 		}
396 		if (file) {
397 			ret = trace_seq_path(s, &file->f_path);
398 			if (ret)
399 				trace_seq_printf(s, "[+0x%lx]",
400 						 ip - vmstart);
401 		}
402 		up_read(&mm->mmap_sem);
403 	}
404 	if (ret && ((sym_flags & TRACE_ITER_SYM_ADDR) || !file))
405 		trace_seq_printf(s, " <" IP_FMT ">", ip);
406 	return !trace_seq_has_overflowed(s);
407 }
408 
409 int
410 seq_print_ip_sym(struct trace_seq *s, unsigned long ip, unsigned long sym_flags)
411 {
412 	if (!ip) {
413 		trace_seq_putc(s, '0');
414 		goto out;
415 	}
416 
417 	if (sym_flags & TRACE_ITER_SYM_OFFSET)
418 		seq_print_sym_offset(s, "%s", ip);
419 	else
420 		seq_print_sym_short(s, "%s", ip);
421 
422 	if (sym_flags & TRACE_ITER_SYM_ADDR)
423 		trace_seq_printf(s, " <" IP_FMT ">", ip);
424 
425  out:
426 	return !trace_seq_has_overflowed(s);
427 }
428 
429 /**
430  * trace_print_lat_fmt - print the irq, preempt and lockdep fields
431  * @s: trace seq struct to write to
432  * @entry: The trace entry field from the ring buffer
433  *
434  * Prints the generic fields of irqs off, in hard or softirq, preempt
435  * count.
436  */
437 int trace_print_lat_fmt(struct trace_seq *s, struct trace_entry *entry)
438 {
439 	char hardsoft_irq;
440 	char need_resched;
441 	char irqs_off;
442 	int hardirq;
443 	int softirq;
444 	int nmi;
445 
446 	nmi = entry->flags & TRACE_FLAG_NMI;
447 	hardirq = entry->flags & TRACE_FLAG_HARDIRQ;
448 	softirq = entry->flags & TRACE_FLAG_SOFTIRQ;
449 
450 	irqs_off =
451 		(entry->flags & TRACE_FLAG_IRQS_OFF) ? 'd' :
452 		(entry->flags & TRACE_FLAG_IRQS_NOSUPPORT) ? 'X' :
453 		'.';
454 
455 	switch (entry->flags & (TRACE_FLAG_NEED_RESCHED |
456 				TRACE_FLAG_PREEMPT_RESCHED)) {
457 	case TRACE_FLAG_NEED_RESCHED | TRACE_FLAG_PREEMPT_RESCHED:
458 		need_resched = 'N';
459 		break;
460 	case TRACE_FLAG_NEED_RESCHED:
461 		need_resched = 'n';
462 		break;
463 	case TRACE_FLAG_PREEMPT_RESCHED:
464 		need_resched = 'p';
465 		break;
466 	default:
467 		need_resched = '.';
468 		break;
469 	}
470 
471 	hardsoft_irq =
472 		(nmi && hardirq)     ? 'Z' :
473 		nmi                  ? 'z' :
474 		(hardirq && softirq) ? 'H' :
475 		hardirq              ? 'h' :
476 		softirq              ? 's' :
477 		                       '.' ;
478 
479 	trace_seq_printf(s, "%c%c%c",
480 			 irqs_off, need_resched, hardsoft_irq);
481 
482 	if (entry->preempt_count)
483 		trace_seq_printf(s, "%x", entry->preempt_count);
484 	else
485 		trace_seq_putc(s, '.');
486 
487 	return !trace_seq_has_overflowed(s);
488 }
489 
490 static int
491 lat_print_generic(struct trace_seq *s, struct trace_entry *entry, int cpu)
492 {
493 	char comm[TASK_COMM_LEN];
494 
495 	trace_find_cmdline(entry->pid, comm);
496 
497 	trace_seq_printf(s, "%8.8s-%-5d %3d",
498 			 comm, entry->pid, cpu);
499 
500 	return trace_print_lat_fmt(s, entry);
501 }
502 
503 #undef MARK
504 #define MARK(v, s) {.val = v, .sym = s}
505 /* trace overhead mark */
506 static const struct trace_mark {
507 	unsigned long long	val; /* unit: nsec */
508 	char			sym;
509 } mark[] = {
510 	MARK(1000000000ULL	, '$'), /* 1 sec */
511 	MARK(100000000ULL	, '@'), /* 100 msec */
512 	MARK(10000000ULL	, '*'), /* 10 msec */
513 	MARK(1000000ULL		, '#'), /* 1000 usecs */
514 	MARK(100000ULL		, '!'), /* 100 usecs */
515 	MARK(10000ULL		, '+'), /* 10 usecs */
516 };
517 #undef MARK
518 
519 char trace_find_mark(unsigned long long d)
520 {
521 	int i;
522 	int size = ARRAY_SIZE(mark);
523 
524 	for (i = 0; i < size; i++) {
525 		if (d > mark[i].val)
526 			break;
527 	}
528 
529 	return (i == size) ? ' ' : mark[i].sym;
530 }
531 
532 static int
533 lat_print_timestamp(struct trace_iterator *iter, u64 next_ts)
534 {
535 	struct trace_array *tr = iter->tr;
536 	unsigned long verbose = tr->trace_flags & TRACE_ITER_VERBOSE;
537 	unsigned long in_ns = iter->iter_flags & TRACE_FILE_TIME_IN_NS;
538 	unsigned long long abs_ts = iter->ts - iter->trace_buffer->time_start;
539 	unsigned long long rel_ts = next_ts - iter->ts;
540 	struct trace_seq *s = &iter->seq;
541 
542 	if (in_ns) {
543 		abs_ts = ns2usecs(abs_ts);
544 		rel_ts = ns2usecs(rel_ts);
545 	}
546 
547 	if (verbose && in_ns) {
548 		unsigned long abs_usec = do_div(abs_ts, USEC_PER_MSEC);
549 		unsigned long abs_msec = (unsigned long)abs_ts;
550 		unsigned long rel_usec = do_div(rel_ts, USEC_PER_MSEC);
551 		unsigned long rel_msec = (unsigned long)rel_ts;
552 
553 		trace_seq_printf(
554 			s, "[%08llx] %ld.%03ldms (+%ld.%03ldms): ",
555 			ns2usecs(iter->ts),
556 			abs_msec, abs_usec,
557 			rel_msec, rel_usec);
558 
559 	} else if (verbose && !in_ns) {
560 		trace_seq_printf(
561 			s, "[%016llx] %lld (+%lld): ",
562 			iter->ts, abs_ts, rel_ts);
563 
564 	} else if (!verbose && in_ns) {
565 		trace_seq_printf(
566 			s, " %4lldus%c: ",
567 			abs_ts,
568 			trace_find_mark(rel_ts * NSEC_PER_USEC));
569 
570 	} else { /* !verbose && !in_ns */
571 		trace_seq_printf(s, " %4lld: ", abs_ts);
572 	}
573 
574 	return !trace_seq_has_overflowed(s);
575 }
576 
577 int trace_print_context(struct trace_iterator *iter)
578 {
579 	struct trace_array *tr = iter->tr;
580 	struct trace_seq *s = &iter->seq;
581 	struct trace_entry *entry = iter->ent;
582 	unsigned long long t;
583 	unsigned long secs, usec_rem;
584 	char comm[TASK_COMM_LEN];
585 
586 	trace_find_cmdline(entry->pid, comm);
587 
588 	trace_seq_printf(s, "%16s-%-5d [%03d] ",
589 			       comm, entry->pid, iter->cpu);
590 
591 	if (tr->trace_flags & TRACE_ITER_IRQ_INFO)
592 		trace_print_lat_fmt(s, entry);
593 
594 	if (iter->iter_flags & TRACE_FILE_TIME_IN_NS) {
595 		t = ns2usecs(iter->ts);
596 		usec_rem = do_div(t, USEC_PER_SEC);
597 		secs = (unsigned long)t;
598 		trace_seq_printf(s, " %5lu.%06lu: ", secs, usec_rem);
599 	} else
600 		trace_seq_printf(s, " %12llu: ", iter->ts);
601 
602 	return !trace_seq_has_overflowed(s);
603 }
604 
605 int trace_print_lat_context(struct trace_iterator *iter)
606 {
607 	struct trace_array *tr = iter->tr;
608 	/* trace_find_next_entry will reset ent_size */
609 	int ent_size = iter->ent_size;
610 	struct trace_seq *s = &iter->seq;
611 	u64 next_ts;
612 	struct trace_entry *entry = iter->ent,
613 			   *next_entry = trace_find_next_entry(iter, NULL,
614 							       &next_ts);
615 	unsigned long verbose = (tr->trace_flags & TRACE_ITER_VERBOSE);
616 
617 	/* Restore the original ent_size */
618 	iter->ent_size = ent_size;
619 
620 	if (!next_entry)
621 		next_ts = iter->ts;
622 
623 	if (verbose) {
624 		char comm[TASK_COMM_LEN];
625 
626 		trace_find_cmdline(entry->pid, comm);
627 
628 		trace_seq_printf(
629 			s, "%16s %5d %3d %d %08x %08lx ",
630 			comm, entry->pid, iter->cpu, entry->flags,
631 			entry->preempt_count, iter->idx);
632 	} else {
633 		lat_print_generic(s, entry, iter->cpu);
634 	}
635 
636 	lat_print_timestamp(iter, next_ts);
637 
638 	return !trace_seq_has_overflowed(s);
639 }
640 
641 static const char state_to_char[] = TASK_STATE_TO_CHAR_STR;
642 
643 static int task_state_char(unsigned long state)
644 {
645 	int bit = state ? __ffs(state) + 1 : 0;
646 
647 	return bit < sizeof(state_to_char) - 1 ? state_to_char[bit] : '?';
648 }
649 
650 /**
651  * ftrace_find_event - find a registered event
652  * @type: the type of event to look for
653  *
654  * Returns an event of type @type otherwise NULL
655  * Called with trace_event_read_lock() held.
656  */
657 struct trace_event *ftrace_find_event(int type)
658 {
659 	struct trace_event *event;
660 	unsigned key;
661 
662 	key = type & (EVENT_HASHSIZE - 1);
663 
664 	hlist_for_each_entry(event, &event_hash[key], node) {
665 		if (event->type == type)
666 			return event;
667 	}
668 
669 	return NULL;
670 }
671 
672 static LIST_HEAD(ftrace_event_list);
673 
674 static int trace_search_list(struct list_head **list)
675 {
676 	struct trace_event *e;
677 	int last = __TRACE_LAST_TYPE;
678 
679 	if (list_empty(&ftrace_event_list)) {
680 		*list = &ftrace_event_list;
681 		return last + 1;
682 	}
683 
684 	/*
685 	 * We used up all possible max events,
686 	 * lets see if somebody freed one.
687 	 */
688 	list_for_each_entry(e, &ftrace_event_list, list) {
689 		if (e->type != last + 1)
690 			break;
691 		last++;
692 	}
693 
694 	/* Did we used up all 65 thousand events??? */
695 	if ((last + 1) > TRACE_EVENT_TYPE_MAX)
696 		return 0;
697 
698 	*list = &e->list;
699 	return last + 1;
700 }
701 
702 void trace_event_read_lock(void)
703 {
704 	down_read(&trace_event_sem);
705 }
706 
707 void trace_event_read_unlock(void)
708 {
709 	up_read(&trace_event_sem);
710 }
711 
712 /**
713  * register_trace_event - register output for an event type
714  * @event: the event type to register
715  *
716  * Event types are stored in a hash and this hash is used to
717  * find a way to print an event. If the @event->type is set
718  * then it will use that type, otherwise it will assign a
719  * type to use.
720  *
721  * If you assign your own type, please make sure it is added
722  * to the trace_type enum in trace.h, to avoid collisions
723  * with the dynamic types.
724  *
725  * Returns the event type number or zero on error.
726  */
727 int register_trace_event(struct trace_event *event)
728 {
729 	unsigned key;
730 	int ret = 0;
731 
732 	down_write(&trace_event_sem);
733 
734 	if (WARN_ON(!event))
735 		goto out;
736 
737 	if (WARN_ON(!event->funcs))
738 		goto out;
739 
740 	INIT_LIST_HEAD(&event->list);
741 
742 	if (!event->type) {
743 		struct list_head *list = NULL;
744 
745 		if (next_event_type > TRACE_EVENT_TYPE_MAX) {
746 
747 			event->type = trace_search_list(&list);
748 			if (!event->type)
749 				goto out;
750 
751 		} else {
752 
753 			event->type = next_event_type++;
754 			list = &ftrace_event_list;
755 		}
756 
757 		if (WARN_ON(ftrace_find_event(event->type)))
758 			goto out;
759 
760 		list_add_tail(&event->list, list);
761 
762 	} else if (event->type > __TRACE_LAST_TYPE) {
763 		printk(KERN_WARNING "Need to add type to trace.h\n");
764 		WARN_ON(1);
765 		goto out;
766 	} else {
767 		/* Is this event already used */
768 		if (ftrace_find_event(event->type))
769 			goto out;
770 	}
771 
772 	if (event->funcs->trace == NULL)
773 		event->funcs->trace = trace_nop_print;
774 	if (event->funcs->raw == NULL)
775 		event->funcs->raw = trace_nop_print;
776 	if (event->funcs->hex == NULL)
777 		event->funcs->hex = trace_nop_print;
778 	if (event->funcs->binary == NULL)
779 		event->funcs->binary = trace_nop_print;
780 
781 	key = event->type & (EVENT_HASHSIZE - 1);
782 
783 	hlist_add_head(&event->node, &event_hash[key]);
784 
785 	ret = event->type;
786  out:
787 	up_write(&trace_event_sem);
788 
789 	return ret;
790 }
791 EXPORT_SYMBOL_GPL(register_trace_event);
792 
793 /*
794  * Used by module code with the trace_event_sem held for write.
795  */
796 int __unregister_trace_event(struct trace_event *event)
797 {
798 	hlist_del(&event->node);
799 	list_del(&event->list);
800 	return 0;
801 }
802 
803 /**
804  * unregister_trace_event - remove a no longer used event
805  * @event: the event to remove
806  */
807 int unregister_trace_event(struct trace_event *event)
808 {
809 	down_write(&trace_event_sem);
810 	__unregister_trace_event(event);
811 	up_write(&trace_event_sem);
812 
813 	return 0;
814 }
815 EXPORT_SYMBOL_GPL(unregister_trace_event);
816 
817 /*
818  * Standard events
819  */
820 
821 enum print_line_t trace_nop_print(struct trace_iterator *iter, int flags,
822 				  struct trace_event *event)
823 {
824 	trace_seq_printf(&iter->seq, "type: %d\n", iter->ent->type);
825 
826 	return trace_handle_return(&iter->seq);
827 }
828 
829 /* TRACE_FN */
830 static enum print_line_t trace_fn_trace(struct trace_iterator *iter, int flags,
831 					struct trace_event *event)
832 {
833 	struct ftrace_entry *field;
834 	struct trace_seq *s = &iter->seq;
835 
836 	trace_assign_type(field, iter->ent);
837 
838 	seq_print_ip_sym(s, field->ip, flags);
839 
840 	if ((flags & TRACE_ITER_PRINT_PARENT) && field->parent_ip) {
841 		trace_seq_puts(s, " <-");
842 		seq_print_ip_sym(s, field->parent_ip, flags);
843 	}
844 
845 	trace_seq_putc(s, '\n');
846 
847 	return trace_handle_return(s);
848 }
849 
850 static enum print_line_t trace_fn_raw(struct trace_iterator *iter, int flags,
851 				      struct trace_event *event)
852 {
853 	struct ftrace_entry *field;
854 
855 	trace_assign_type(field, iter->ent);
856 
857 	trace_seq_printf(&iter->seq, "%lx %lx\n",
858 			 field->ip,
859 			 field->parent_ip);
860 
861 	return trace_handle_return(&iter->seq);
862 }
863 
864 static enum print_line_t trace_fn_hex(struct trace_iterator *iter, int flags,
865 				      struct trace_event *event)
866 {
867 	struct ftrace_entry *field;
868 	struct trace_seq *s = &iter->seq;
869 
870 	trace_assign_type(field, iter->ent);
871 
872 	SEQ_PUT_HEX_FIELD(s, field->ip);
873 	SEQ_PUT_HEX_FIELD(s, field->parent_ip);
874 
875 	return trace_handle_return(s);
876 }
877 
878 static enum print_line_t trace_fn_bin(struct trace_iterator *iter, int flags,
879 				      struct trace_event *event)
880 {
881 	struct ftrace_entry *field;
882 	struct trace_seq *s = &iter->seq;
883 
884 	trace_assign_type(field, iter->ent);
885 
886 	SEQ_PUT_FIELD(s, field->ip);
887 	SEQ_PUT_FIELD(s, field->parent_ip);
888 
889 	return trace_handle_return(s);
890 }
891 
892 static struct trace_event_functions trace_fn_funcs = {
893 	.trace		= trace_fn_trace,
894 	.raw		= trace_fn_raw,
895 	.hex		= trace_fn_hex,
896 	.binary		= trace_fn_bin,
897 };
898 
899 static struct trace_event trace_fn_event = {
900 	.type		= TRACE_FN,
901 	.funcs		= &trace_fn_funcs,
902 };
903 
904 /* TRACE_CTX an TRACE_WAKE */
905 static enum print_line_t trace_ctxwake_print(struct trace_iterator *iter,
906 					     char *delim)
907 {
908 	struct ctx_switch_entry *field;
909 	char comm[TASK_COMM_LEN];
910 	int S, T;
911 
912 
913 	trace_assign_type(field, iter->ent);
914 
915 	T = task_state_char(field->next_state);
916 	S = task_state_char(field->prev_state);
917 	trace_find_cmdline(field->next_pid, comm);
918 	trace_seq_printf(&iter->seq,
919 			 " %5d:%3d:%c %s [%03d] %5d:%3d:%c %s\n",
920 			 field->prev_pid,
921 			 field->prev_prio,
922 			 S, delim,
923 			 field->next_cpu,
924 			 field->next_pid,
925 			 field->next_prio,
926 			 T, comm);
927 
928 	return trace_handle_return(&iter->seq);
929 }
930 
931 static enum print_line_t trace_ctx_print(struct trace_iterator *iter, int flags,
932 					 struct trace_event *event)
933 {
934 	return trace_ctxwake_print(iter, "==>");
935 }
936 
937 static enum print_line_t trace_wake_print(struct trace_iterator *iter,
938 					  int flags, struct trace_event *event)
939 {
940 	return trace_ctxwake_print(iter, "  +");
941 }
942 
943 static int trace_ctxwake_raw(struct trace_iterator *iter, char S)
944 {
945 	struct ctx_switch_entry *field;
946 	int T;
947 
948 	trace_assign_type(field, iter->ent);
949 
950 	if (!S)
951 		S = task_state_char(field->prev_state);
952 	T = task_state_char(field->next_state);
953 	trace_seq_printf(&iter->seq, "%d %d %c %d %d %d %c\n",
954 			 field->prev_pid,
955 			 field->prev_prio,
956 			 S,
957 			 field->next_cpu,
958 			 field->next_pid,
959 			 field->next_prio,
960 			 T);
961 
962 	return trace_handle_return(&iter->seq);
963 }
964 
965 static enum print_line_t trace_ctx_raw(struct trace_iterator *iter, int flags,
966 				       struct trace_event *event)
967 {
968 	return trace_ctxwake_raw(iter, 0);
969 }
970 
971 static enum print_line_t trace_wake_raw(struct trace_iterator *iter, int flags,
972 					struct trace_event *event)
973 {
974 	return trace_ctxwake_raw(iter, '+');
975 }
976 
977 
978 static int trace_ctxwake_hex(struct trace_iterator *iter, char S)
979 {
980 	struct ctx_switch_entry *field;
981 	struct trace_seq *s = &iter->seq;
982 	int T;
983 
984 	trace_assign_type(field, iter->ent);
985 
986 	if (!S)
987 		S = task_state_char(field->prev_state);
988 	T = task_state_char(field->next_state);
989 
990 	SEQ_PUT_HEX_FIELD(s, field->prev_pid);
991 	SEQ_PUT_HEX_FIELD(s, field->prev_prio);
992 	SEQ_PUT_HEX_FIELD(s, S);
993 	SEQ_PUT_HEX_FIELD(s, field->next_cpu);
994 	SEQ_PUT_HEX_FIELD(s, field->next_pid);
995 	SEQ_PUT_HEX_FIELD(s, field->next_prio);
996 	SEQ_PUT_HEX_FIELD(s, T);
997 
998 	return trace_handle_return(s);
999 }
1000 
1001 static enum print_line_t trace_ctx_hex(struct trace_iterator *iter, int flags,
1002 				       struct trace_event *event)
1003 {
1004 	return trace_ctxwake_hex(iter, 0);
1005 }
1006 
1007 static enum print_line_t trace_wake_hex(struct trace_iterator *iter, int flags,
1008 					struct trace_event *event)
1009 {
1010 	return trace_ctxwake_hex(iter, '+');
1011 }
1012 
1013 static enum print_line_t trace_ctxwake_bin(struct trace_iterator *iter,
1014 					   int flags, struct trace_event *event)
1015 {
1016 	struct ctx_switch_entry *field;
1017 	struct trace_seq *s = &iter->seq;
1018 
1019 	trace_assign_type(field, iter->ent);
1020 
1021 	SEQ_PUT_FIELD(s, field->prev_pid);
1022 	SEQ_PUT_FIELD(s, field->prev_prio);
1023 	SEQ_PUT_FIELD(s, field->prev_state);
1024 	SEQ_PUT_FIELD(s, field->next_cpu);
1025 	SEQ_PUT_FIELD(s, field->next_pid);
1026 	SEQ_PUT_FIELD(s, field->next_prio);
1027 	SEQ_PUT_FIELD(s, field->next_state);
1028 
1029 	return trace_handle_return(s);
1030 }
1031 
1032 static struct trace_event_functions trace_ctx_funcs = {
1033 	.trace		= trace_ctx_print,
1034 	.raw		= trace_ctx_raw,
1035 	.hex		= trace_ctx_hex,
1036 	.binary		= trace_ctxwake_bin,
1037 };
1038 
1039 static struct trace_event trace_ctx_event = {
1040 	.type		= TRACE_CTX,
1041 	.funcs		= &trace_ctx_funcs,
1042 };
1043 
1044 static struct trace_event_functions trace_wake_funcs = {
1045 	.trace		= trace_wake_print,
1046 	.raw		= trace_wake_raw,
1047 	.hex		= trace_wake_hex,
1048 	.binary		= trace_ctxwake_bin,
1049 };
1050 
1051 static struct trace_event trace_wake_event = {
1052 	.type		= TRACE_WAKE,
1053 	.funcs		= &trace_wake_funcs,
1054 };
1055 
1056 /* TRACE_STACK */
1057 
1058 static enum print_line_t trace_stack_print(struct trace_iterator *iter,
1059 					   int flags, struct trace_event *event)
1060 {
1061 	struct stack_entry *field;
1062 	struct trace_seq *s = &iter->seq;
1063 	unsigned long *p;
1064 	unsigned long *end;
1065 
1066 	trace_assign_type(field, iter->ent);
1067 	end = (unsigned long *)((long)iter->ent + iter->ent_size);
1068 
1069 	trace_seq_puts(s, "<stack trace>\n");
1070 
1071 	for (p = field->caller; p && *p != ULONG_MAX && p < end; p++) {
1072 
1073 		if (trace_seq_has_overflowed(s))
1074 			break;
1075 
1076 		trace_seq_puts(s, " => ");
1077 		seq_print_ip_sym(s, *p, flags);
1078 		trace_seq_putc(s, '\n');
1079 	}
1080 
1081 	return trace_handle_return(s);
1082 }
1083 
1084 static struct trace_event_functions trace_stack_funcs = {
1085 	.trace		= trace_stack_print,
1086 };
1087 
1088 static struct trace_event trace_stack_event = {
1089 	.type		= TRACE_STACK,
1090 	.funcs		= &trace_stack_funcs,
1091 };
1092 
1093 /* TRACE_USER_STACK */
1094 static enum print_line_t trace_user_stack_print(struct trace_iterator *iter,
1095 						int flags, struct trace_event *event)
1096 {
1097 	struct trace_array *tr = iter->tr;
1098 	struct userstack_entry *field;
1099 	struct trace_seq *s = &iter->seq;
1100 	struct mm_struct *mm = NULL;
1101 	unsigned int i;
1102 
1103 	trace_assign_type(field, iter->ent);
1104 
1105 	trace_seq_puts(s, "<user stack trace>\n");
1106 
1107 	if (tr->trace_flags & TRACE_ITER_SYM_USEROBJ) {
1108 		struct task_struct *task;
1109 		/*
1110 		 * we do the lookup on the thread group leader,
1111 		 * since individual threads might have already quit!
1112 		 */
1113 		rcu_read_lock();
1114 		task = find_task_by_vpid(field->tgid);
1115 		if (task)
1116 			mm = get_task_mm(task);
1117 		rcu_read_unlock();
1118 	}
1119 
1120 	for (i = 0; i < FTRACE_STACK_ENTRIES; i++) {
1121 		unsigned long ip = field->caller[i];
1122 
1123 		if (ip == ULONG_MAX || trace_seq_has_overflowed(s))
1124 			break;
1125 
1126 		trace_seq_puts(s, " => ");
1127 
1128 		if (!ip) {
1129 			trace_seq_puts(s, "??");
1130 			trace_seq_putc(s, '\n');
1131 			continue;
1132 		}
1133 
1134 		seq_print_user_ip(s, mm, ip, flags);
1135 		trace_seq_putc(s, '\n');
1136 	}
1137 
1138 	if (mm)
1139 		mmput(mm);
1140 
1141 	return trace_handle_return(s);
1142 }
1143 
1144 static struct trace_event_functions trace_user_stack_funcs = {
1145 	.trace		= trace_user_stack_print,
1146 };
1147 
1148 static struct trace_event trace_user_stack_event = {
1149 	.type		= TRACE_USER_STACK,
1150 	.funcs		= &trace_user_stack_funcs,
1151 };
1152 
1153 /* TRACE_HWLAT */
1154 static enum print_line_t
1155 trace_hwlat_print(struct trace_iterator *iter, int flags,
1156 		  struct trace_event *event)
1157 {
1158 	struct trace_entry *entry = iter->ent;
1159 	struct trace_seq *s = &iter->seq;
1160 	struct hwlat_entry *field;
1161 
1162 	trace_assign_type(field, entry);
1163 
1164 	trace_seq_printf(s, "#%-5u inner/outer(us): %4llu/%-5llu ts:%ld.%09ld",
1165 			 field->seqnum,
1166 			 field->duration,
1167 			 field->outer_duration,
1168 			 field->timestamp.tv_sec,
1169 			 field->timestamp.tv_nsec);
1170 
1171 	if (field->nmi_count) {
1172 		/*
1173 		 * The generic sched_clock() is not NMI safe, thus
1174 		 * we only record the count and not the time.
1175 		 */
1176 		if (!IS_ENABLED(CONFIG_GENERIC_SCHED_CLOCK))
1177 			trace_seq_printf(s, " nmi-total:%llu",
1178 					 field->nmi_total_ts);
1179 		trace_seq_printf(s, " nmi-count:%u",
1180 				 field->nmi_count);
1181 	}
1182 
1183 	trace_seq_putc(s, '\n');
1184 
1185 	return trace_handle_return(s);
1186 }
1187 
1188 
1189 static enum print_line_t
1190 trace_hwlat_raw(struct trace_iterator *iter, int flags,
1191 		struct trace_event *event)
1192 {
1193 	struct hwlat_entry *field;
1194 	struct trace_seq *s = &iter->seq;
1195 
1196 	trace_assign_type(field, iter->ent);
1197 
1198 	trace_seq_printf(s, "%llu %lld %ld %09ld %u\n",
1199 			 field->duration,
1200 			 field->outer_duration,
1201 			 field->timestamp.tv_sec,
1202 			 field->timestamp.tv_nsec,
1203 			 field->seqnum);
1204 
1205 	return trace_handle_return(s);
1206 }
1207 
1208 static struct trace_event_functions trace_hwlat_funcs = {
1209 	.trace		= trace_hwlat_print,
1210 	.raw		= trace_hwlat_raw,
1211 };
1212 
1213 static struct trace_event trace_hwlat_event = {
1214 	.type		= TRACE_HWLAT,
1215 	.funcs		= &trace_hwlat_funcs,
1216 };
1217 
1218 /* TRACE_BPUTS */
1219 static enum print_line_t
1220 trace_bputs_print(struct trace_iterator *iter, int flags,
1221 		   struct trace_event *event)
1222 {
1223 	struct trace_entry *entry = iter->ent;
1224 	struct trace_seq *s = &iter->seq;
1225 	struct bputs_entry *field;
1226 
1227 	trace_assign_type(field, entry);
1228 
1229 	seq_print_ip_sym(s, field->ip, flags);
1230 	trace_seq_puts(s, ": ");
1231 	trace_seq_puts(s, field->str);
1232 
1233 	return trace_handle_return(s);
1234 }
1235 
1236 
1237 static enum print_line_t
1238 trace_bputs_raw(struct trace_iterator *iter, int flags,
1239 		struct trace_event *event)
1240 {
1241 	struct bputs_entry *field;
1242 	struct trace_seq *s = &iter->seq;
1243 
1244 	trace_assign_type(field, iter->ent);
1245 
1246 	trace_seq_printf(s, ": %lx : ", field->ip);
1247 	trace_seq_puts(s, field->str);
1248 
1249 	return trace_handle_return(s);
1250 }
1251 
1252 static struct trace_event_functions trace_bputs_funcs = {
1253 	.trace		= trace_bputs_print,
1254 	.raw		= trace_bputs_raw,
1255 };
1256 
1257 static struct trace_event trace_bputs_event = {
1258 	.type		= TRACE_BPUTS,
1259 	.funcs		= &trace_bputs_funcs,
1260 };
1261 
1262 /* TRACE_BPRINT */
1263 static enum print_line_t
1264 trace_bprint_print(struct trace_iterator *iter, int flags,
1265 		   struct trace_event *event)
1266 {
1267 	struct trace_entry *entry = iter->ent;
1268 	struct trace_seq *s = &iter->seq;
1269 	struct bprint_entry *field;
1270 
1271 	trace_assign_type(field, entry);
1272 
1273 	seq_print_ip_sym(s, field->ip, flags);
1274 	trace_seq_puts(s, ": ");
1275 	trace_seq_bprintf(s, field->fmt, field->buf);
1276 
1277 	return trace_handle_return(s);
1278 }
1279 
1280 
1281 static enum print_line_t
1282 trace_bprint_raw(struct trace_iterator *iter, int flags,
1283 		 struct trace_event *event)
1284 {
1285 	struct bprint_entry *field;
1286 	struct trace_seq *s = &iter->seq;
1287 
1288 	trace_assign_type(field, iter->ent);
1289 
1290 	trace_seq_printf(s, ": %lx : ", field->ip);
1291 	trace_seq_bprintf(s, field->fmt, field->buf);
1292 
1293 	return trace_handle_return(s);
1294 }
1295 
1296 static struct trace_event_functions trace_bprint_funcs = {
1297 	.trace		= trace_bprint_print,
1298 	.raw		= trace_bprint_raw,
1299 };
1300 
1301 static struct trace_event trace_bprint_event = {
1302 	.type		= TRACE_BPRINT,
1303 	.funcs		= &trace_bprint_funcs,
1304 };
1305 
1306 /* TRACE_PRINT */
1307 static enum print_line_t trace_print_print(struct trace_iterator *iter,
1308 					   int flags, struct trace_event *event)
1309 {
1310 	struct print_entry *field;
1311 	struct trace_seq *s = &iter->seq;
1312 
1313 	trace_assign_type(field, iter->ent);
1314 
1315 	seq_print_ip_sym(s, field->ip, flags);
1316 	trace_seq_printf(s, ": %s", field->buf);
1317 
1318 	return trace_handle_return(s);
1319 }
1320 
1321 static enum print_line_t trace_print_raw(struct trace_iterator *iter, int flags,
1322 					 struct trace_event *event)
1323 {
1324 	struct print_entry *field;
1325 
1326 	trace_assign_type(field, iter->ent);
1327 
1328 	trace_seq_printf(&iter->seq, "# %lx %s", field->ip, field->buf);
1329 
1330 	return trace_handle_return(&iter->seq);
1331 }
1332 
1333 static struct trace_event_functions trace_print_funcs = {
1334 	.trace		= trace_print_print,
1335 	.raw		= trace_print_raw,
1336 };
1337 
1338 static struct trace_event trace_print_event = {
1339 	.type	 	= TRACE_PRINT,
1340 	.funcs		= &trace_print_funcs,
1341 };
1342 
1343 static enum print_line_t trace_raw_data(struct trace_iterator *iter, int flags,
1344 					 struct trace_event *event)
1345 {
1346 	struct raw_data_entry *field;
1347 	int i;
1348 
1349 	trace_assign_type(field, iter->ent);
1350 
1351 	trace_seq_printf(&iter->seq, "# %x buf:", field->id);
1352 
1353 	for (i = 0; i < iter->ent_size - offsetof(struct raw_data_entry, buf); i++)
1354 		trace_seq_printf(&iter->seq, " %02x",
1355 				 (unsigned char)field->buf[i]);
1356 
1357 	trace_seq_putc(&iter->seq, '\n');
1358 
1359 	return trace_handle_return(&iter->seq);
1360 }
1361 
1362 static struct trace_event_functions trace_raw_data_funcs = {
1363 	.trace		= trace_raw_data,
1364 	.raw		= trace_raw_data,
1365 };
1366 
1367 static struct trace_event trace_raw_data_event = {
1368 	.type	 	= TRACE_RAW_DATA,
1369 	.funcs		= &trace_raw_data_funcs,
1370 };
1371 
1372 
1373 static struct trace_event *events[] __initdata = {
1374 	&trace_fn_event,
1375 	&trace_ctx_event,
1376 	&trace_wake_event,
1377 	&trace_stack_event,
1378 	&trace_user_stack_event,
1379 	&trace_bputs_event,
1380 	&trace_bprint_event,
1381 	&trace_print_event,
1382 	&trace_hwlat_event,
1383 	&trace_raw_data_event,
1384 	NULL
1385 };
1386 
1387 __init static int init_events(void)
1388 {
1389 	struct trace_event *event;
1390 	int i, ret;
1391 
1392 	for (i = 0; events[i]; i++) {
1393 		event = events[i];
1394 
1395 		ret = register_trace_event(event);
1396 		if (!ret) {
1397 			printk(KERN_WARNING "event %d failed to register\n",
1398 			       event->type);
1399 			WARN_ON_ONCE(1);
1400 		}
1401 	}
1402 
1403 	return 0;
1404 }
1405 early_initcall(init_events);
1406