xref: /openbmc/linux/kernel/trace/trace_output.c (revision 50df3be7)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * trace_output.c
4  *
5  * Copyright (C) 2008 Red Hat Inc, Steven Rostedt <srostedt@redhat.com>
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 	const char *fmt = concatenate ? "%*phN" : "%*ph";
223 
224 	for (i = 0; i < buf_len; i += 16)
225 		trace_seq_printf(p, fmt, min(buf_len - i, 16), &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(struct trace_seq *s, unsigned long address, bool offset)
343 {
344 #ifdef CONFIG_KALLSYMS
345 	char str[KSYM_SYMBOL_LEN];
346 	const char *name;
347 
348 	if (offset)
349 		sprint_symbol(str, address);
350 	else
351 		kallsyms_lookup(address, NULL, NULL, NULL, str);
352 	name = kretprobed(str);
353 
354 	if (name && strlen(name)) {
355 		trace_seq_puts(s, name);
356 		return;
357 	}
358 #endif
359 	trace_seq_printf(s, "0x%08lx", address);
360 }
361 
362 #ifndef CONFIG_64BIT
363 # define IP_FMT "%08lx"
364 #else
365 # define IP_FMT "%016lx"
366 #endif
367 
368 static int seq_print_user_ip(struct trace_seq *s, struct mm_struct *mm,
369 			     unsigned long ip, unsigned long sym_flags)
370 {
371 	struct file *file = NULL;
372 	unsigned long vmstart = 0;
373 	int ret = 1;
374 
375 	if (s->full)
376 		return 0;
377 
378 	if (mm) {
379 		const struct vm_area_struct *vma;
380 
381 		down_read(&mm->mmap_sem);
382 		vma = find_vma(mm, ip);
383 		if (vma) {
384 			file = vma->vm_file;
385 			vmstart = vma->vm_start;
386 		}
387 		if (file) {
388 			ret = trace_seq_path(s, &file->f_path);
389 			if (ret)
390 				trace_seq_printf(s, "[+0x%lx]",
391 						 ip - vmstart);
392 		}
393 		up_read(&mm->mmap_sem);
394 	}
395 	if (ret && ((sym_flags & TRACE_ITER_SYM_ADDR) || !file))
396 		trace_seq_printf(s, " <" IP_FMT ">", ip);
397 	return !trace_seq_has_overflowed(s);
398 }
399 
400 int
401 seq_print_ip_sym(struct trace_seq *s, unsigned long ip, unsigned long sym_flags)
402 {
403 	if (!ip) {
404 		trace_seq_putc(s, '0');
405 		goto out;
406 	}
407 
408 	seq_print_sym(s, ip, sym_flags & TRACE_ITER_SYM_OFFSET);
409 
410 	if (sym_flags & TRACE_ITER_SYM_ADDR)
411 		trace_seq_printf(s, " <" IP_FMT ">", ip);
412 
413  out:
414 	return !trace_seq_has_overflowed(s);
415 }
416 
417 /**
418  * trace_print_lat_fmt - print the irq, preempt and lockdep fields
419  * @s: trace seq struct to write to
420  * @entry: The trace entry field from the ring buffer
421  *
422  * Prints the generic fields of irqs off, in hard or softirq, preempt
423  * count.
424  */
425 int trace_print_lat_fmt(struct trace_seq *s, struct trace_entry *entry)
426 {
427 	char hardsoft_irq;
428 	char need_resched;
429 	char irqs_off;
430 	int hardirq;
431 	int softirq;
432 	int nmi;
433 
434 	nmi = entry->flags & TRACE_FLAG_NMI;
435 	hardirq = entry->flags & TRACE_FLAG_HARDIRQ;
436 	softirq = entry->flags & TRACE_FLAG_SOFTIRQ;
437 
438 	irqs_off =
439 		(entry->flags & TRACE_FLAG_IRQS_OFF) ? 'd' :
440 		(entry->flags & TRACE_FLAG_IRQS_NOSUPPORT) ? 'X' :
441 		'.';
442 
443 	switch (entry->flags & (TRACE_FLAG_NEED_RESCHED |
444 				TRACE_FLAG_PREEMPT_RESCHED)) {
445 	case TRACE_FLAG_NEED_RESCHED | TRACE_FLAG_PREEMPT_RESCHED:
446 		need_resched = 'N';
447 		break;
448 	case TRACE_FLAG_NEED_RESCHED:
449 		need_resched = 'n';
450 		break;
451 	case TRACE_FLAG_PREEMPT_RESCHED:
452 		need_resched = 'p';
453 		break;
454 	default:
455 		need_resched = '.';
456 		break;
457 	}
458 
459 	hardsoft_irq =
460 		(nmi && hardirq)     ? 'Z' :
461 		nmi                  ? 'z' :
462 		(hardirq && softirq) ? 'H' :
463 		hardirq              ? 'h' :
464 		softirq              ? 's' :
465 		                       '.' ;
466 
467 	trace_seq_printf(s, "%c%c%c",
468 			 irqs_off, need_resched, hardsoft_irq);
469 
470 	if (entry->preempt_count)
471 		trace_seq_printf(s, "%x", entry->preempt_count);
472 	else
473 		trace_seq_putc(s, '.');
474 
475 	return !trace_seq_has_overflowed(s);
476 }
477 
478 static int
479 lat_print_generic(struct trace_seq *s, struct trace_entry *entry, int cpu)
480 {
481 	char comm[TASK_COMM_LEN];
482 
483 	trace_find_cmdline(entry->pid, comm);
484 
485 	trace_seq_printf(s, "%8.8s-%-5d %3d",
486 			 comm, entry->pid, cpu);
487 
488 	return trace_print_lat_fmt(s, entry);
489 }
490 
491 #undef MARK
492 #define MARK(v, s) {.val = v, .sym = s}
493 /* trace overhead mark */
494 static const struct trace_mark {
495 	unsigned long long	val; /* unit: nsec */
496 	char			sym;
497 } mark[] = {
498 	MARK(1000000000ULL	, '$'), /* 1 sec */
499 	MARK(100000000ULL	, '@'), /* 100 msec */
500 	MARK(10000000ULL	, '*'), /* 10 msec */
501 	MARK(1000000ULL		, '#'), /* 1000 usecs */
502 	MARK(100000ULL		, '!'), /* 100 usecs */
503 	MARK(10000ULL		, '+'), /* 10 usecs */
504 };
505 #undef MARK
506 
507 char trace_find_mark(unsigned long long d)
508 {
509 	int i;
510 	int size = ARRAY_SIZE(mark);
511 
512 	for (i = 0; i < size; i++) {
513 		if (d > mark[i].val)
514 			break;
515 	}
516 
517 	return (i == size) ? ' ' : mark[i].sym;
518 }
519 
520 static int
521 lat_print_timestamp(struct trace_iterator *iter, u64 next_ts)
522 {
523 	struct trace_array *tr = iter->tr;
524 	unsigned long verbose = tr->trace_flags & TRACE_ITER_VERBOSE;
525 	unsigned long in_ns = iter->iter_flags & TRACE_FILE_TIME_IN_NS;
526 	unsigned long long abs_ts = iter->ts - iter->trace_buffer->time_start;
527 	unsigned long long rel_ts = next_ts - iter->ts;
528 	struct trace_seq *s = &iter->seq;
529 
530 	if (in_ns) {
531 		abs_ts = ns2usecs(abs_ts);
532 		rel_ts = ns2usecs(rel_ts);
533 	}
534 
535 	if (verbose && in_ns) {
536 		unsigned long abs_usec = do_div(abs_ts, USEC_PER_MSEC);
537 		unsigned long abs_msec = (unsigned long)abs_ts;
538 		unsigned long rel_usec = do_div(rel_ts, USEC_PER_MSEC);
539 		unsigned long rel_msec = (unsigned long)rel_ts;
540 
541 		trace_seq_printf(
542 			s, "[%08llx] %ld.%03ldms (+%ld.%03ldms): ",
543 			ns2usecs(iter->ts),
544 			abs_msec, abs_usec,
545 			rel_msec, rel_usec);
546 
547 	} else if (verbose && !in_ns) {
548 		trace_seq_printf(
549 			s, "[%016llx] %lld (+%lld): ",
550 			iter->ts, abs_ts, rel_ts);
551 
552 	} else if (!verbose && in_ns) {
553 		trace_seq_printf(
554 			s, " %4lldus%c: ",
555 			abs_ts,
556 			trace_find_mark(rel_ts * NSEC_PER_USEC));
557 
558 	} else { /* !verbose && !in_ns */
559 		trace_seq_printf(s, " %4lld: ", abs_ts);
560 	}
561 
562 	return !trace_seq_has_overflowed(s);
563 }
564 
565 int trace_print_context(struct trace_iterator *iter)
566 {
567 	struct trace_array *tr = iter->tr;
568 	struct trace_seq *s = &iter->seq;
569 	struct trace_entry *entry = iter->ent;
570 	unsigned long long t;
571 	unsigned long secs, usec_rem;
572 	char comm[TASK_COMM_LEN];
573 
574 	trace_find_cmdline(entry->pid, comm);
575 
576 	trace_seq_printf(s, "%16s-%-5d ", comm, entry->pid);
577 
578 	if (tr->trace_flags & TRACE_ITER_RECORD_TGID) {
579 		unsigned int tgid = trace_find_tgid(entry->pid);
580 
581 		if (!tgid)
582 			trace_seq_printf(s, "(-----) ");
583 		else
584 			trace_seq_printf(s, "(%5d) ", tgid);
585 	}
586 
587 	trace_seq_printf(s, "[%03d] ", iter->cpu);
588 
589 	if (tr->trace_flags & TRACE_ITER_IRQ_INFO)
590 		trace_print_lat_fmt(s, entry);
591 
592 	if (iter->iter_flags & TRACE_FILE_TIME_IN_NS) {
593 		t = ns2usecs(iter->ts);
594 		usec_rem = do_div(t, USEC_PER_SEC);
595 		secs = (unsigned long)t;
596 		trace_seq_printf(s, " %5lu.%06lu: ", secs, usec_rem);
597 	} else
598 		trace_seq_printf(s, " %12llu: ", iter->ts);
599 
600 	return !trace_seq_has_overflowed(s);
601 }
602 
603 int trace_print_lat_context(struct trace_iterator *iter)
604 {
605 	struct trace_array *tr = iter->tr;
606 	/* trace_find_next_entry will reset ent_size */
607 	int ent_size = iter->ent_size;
608 	struct trace_seq *s = &iter->seq;
609 	u64 next_ts;
610 	struct trace_entry *entry = iter->ent,
611 			   *next_entry = trace_find_next_entry(iter, NULL,
612 							       &next_ts);
613 	unsigned long verbose = (tr->trace_flags & TRACE_ITER_VERBOSE);
614 
615 	/* Restore the original ent_size */
616 	iter->ent_size = ent_size;
617 
618 	if (!next_entry)
619 		next_ts = iter->ts;
620 
621 	if (verbose) {
622 		char comm[TASK_COMM_LEN];
623 
624 		trace_find_cmdline(entry->pid, comm);
625 
626 		trace_seq_printf(
627 			s, "%16s %5d %3d %d %08x %08lx ",
628 			comm, entry->pid, iter->cpu, entry->flags,
629 			entry->preempt_count, iter->idx);
630 	} else {
631 		lat_print_generic(s, entry, iter->cpu);
632 	}
633 
634 	lat_print_timestamp(iter, next_ts);
635 
636 	return !trace_seq_has_overflowed(s);
637 }
638 
639 /**
640  * ftrace_find_event - find a registered event
641  * @type: the type of event to look for
642  *
643  * Returns an event of type @type otherwise NULL
644  * Called with trace_event_read_lock() held.
645  */
646 struct trace_event *ftrace_find_event(int type)
647 {
648 	struct trace_event *event;
649 	unsigned key;
650 
651 	key = type & (EVENT_HASHSIZE - 1);
652 
653 	hlist_for_each_entry(event, &event_hash[key], node) {
654 		if (event->type == type)
655 			return event;
656 	}
657 
658 	return NULL;
659 }
660 
661 static LIST_HEAD(ftrace_event_list);
662 
663 static int trace_search_list(struct list_head **list)
664 {
665 	struct trace_event *e;
666 	int last = __TRACE_LAST_TYPE;
667 
668 	if (list_empty(&ftrace_event_list)) {
669 		*list = &ftrace_event_list;
670 		return last + 1;
671 	}
672 
673 	/*
674 	 * We used up all possible max events,
675 	 * lets see if somebody freed one.
676 	 */
677 	list_for_each_entry(e, &ftrace_event_list, list) {
678 		if (e->type != last + 1)
679 			break;
680 		last++;
681 	}
682 
683 	/* Did we used up all 65 thousand events??? */
684 	if ((last + 1) > TRACE_EVENT_TYPE_MAX)
685 		return 0;
686 
687 	*list = &e->list;
688 	return last + 1;
689 }
690 
691 void trace_event_read_lock(void)
692 {
693 	down_read(&trace_event_sem);
694 }
695 
696 void trace_event_read_unlock(void)
697 {
698 	up_read(&trace_event_sem);
699 }
700 
701 /**
702  * register_trace_event - register output for an event type
703  * @event: the event type to register
704  *
705  * Event types are stored in a hash and this hash is used to
706  * find a way to print an event. If the @event->type is set
707  * then it will use that type, otherwise it will assign a
708  * type to use.
709  *
710  * If you assign your own type, please make sure it is added
711  * to the trace_type enum in trace.h, to avoid collisions
712  * with the dynamic types.
713  *
714  * Returns the event type number or zero on error.
715  */
716 int register_trace_event(struct trace_event *event)
717 {
718 	unsigned key;
719 	int ret = 0;
720 
721 	down_write(&trace_event_sem);
722 
723 	if (WARN_ON(!event))
724 		goto out;
725 
726 	if (WARN_ON(!event->funcs))
727 		goto out;
728 
729 	INIT_LIST_HEAD(&event->list);
730 
731 	if (!event->type) {
732 		struct list_head *list = NULL;
733 
734 		if (next_event_type > TRACE_EVENT_TYPE_MAX) {
735 
736 			event->type = trace_search_list(&list);
737 			if (!event->type)
738 				goto out;
739 
740 		} else {
741 
742 			event->type = next_event_type++;
743 			list = &ftrace_event_list;
744 		}
745 
746 		if (WARN_ON(ftrace_find_event(event->type)))
747 			goto out;
748 
749 		list_add_tail(&event->list, list);
750 
751 	} else if (event->type > __TRACE_LAST_TYPE) {
752 		printk(KERN_WARNING "Need to add type to trace.h\n");
753 		WARN_ON(1);
754 		goto out;
755 	} else {
756 		/* Is this event already used */
757 		if (ftrace_find_event(event->type))
758 			goto out;
759 	}
760 
761 	if (event->funcs->trace == NULL)
762 		event->funcs->trace = trace_nop_print;
763 	if (event->funcs->raw == NULL)
764 		event->funcs->raw = trace_nop_print;
765 	if (event->funcs->hex == NULL)
766 		event->funcs->hex = trace_nop_print;
767 	if (event->funcs->binary == NULL)
768 		event->funcs->binary = trace_nop_print;
769 
770 	key = event->type & (EVENT_HASHSIZE - 1);
771 
772 	hlist_add_head(&event->node, &event_hash[key]);
773 
774 	ret = event->type;
775  out:
776 	up_write(&trace_event_sem);
777 
778 	return ret;
779 }
780 EXPORT_SYMBOL_GPL(register_trace_event);
781 
782 /*
783  * Used by module code with the trace_event_sem held for write.
784  */
785 int __unregister_trace_event(struct trace_event *event)
786 {
787 	hlist_del(&event->node);
788 	list_del(&event->list);
789 	return 0;
790 }
791 
792 /**
793  * unregister_trace_event - remove a no longer used event
794  * @event: the event to remove
795  */
796 int unregister_trace_event(struct trace_event *event)
797 {
798 	down_write(&trace_event_sem);
799 	__unregister_trace_event(event);
800 	up_write(&trace_event_sem);
801 
802 	return 0;
803 }
804 EXPORT_SYMBOL_GPL(unregister_trace_event);
805 
806 /*
807  * Standard events
808  */
809 
810 enum print_line_t trace_nop_print(struct trace_iterator *iter, int flags,
811 				  struct trace_event *event)
812 {
813 	trace_seq_printf(&iter->seq, "type: %d\n", iter->ent->type);
814 
815 	return trace_handle_return(&iter->seq);
816 }
817 
818 /* TRACE_FN */
819 static enum print_line_t trace_fn_trace(struct trace_iterator *iter, int flags,
820 					struct trace_event *event)
821 {
822 	struct ftrace_entry *field;
823 	struct trace_seq *s = &iter->seq;
824 
825 	trace_assign_type(field, iter->ent);
826 
827 	seq_print_ip_sym(s, field->ip, flags);
828 
829 	if ((flags & TRACE_ITER_PRINT_PARENT) && field->parent_ip) {
830 		trace_seq_puts(s, " <-");
831 		seq_print_ip_sym(s, field->parent_ip, flags);
832 	}
833 
834 	trace_seq_putc(s, '\n');
835 
836 	return trace_handle_return(s);
837 }
838 
839 static enum print_line_t trace_fn_raw(struct trace_iterator *iter, int flags,
840 				      struct trace_event *event)
841 {
842 	struct ftrace_entry *field;
843 
844 	trace_assign_type(field, iter->ent);
845 
846 	trace_seq_printf(&iter->seq, "%lx %lx\n",
847 			 field->ip,
848 			 field->parent_ip);
849 
850 	return trace_handle_return(&iter->seq);
851 }
852 
853 static enum print_line_t trace_fn_hex(struct trace_iterator *iter, int flags,
854 				      struct trace_event *event)
855 {
856 	struct ftrace_entry *field;
857 	struct trace_seq *s = &iter->seq;
858 
859 	trace_assign_type(field, iter->ent);
860 
861 	SEQ_PUT_HEX_FIELD(s, field->ip);
862 	SEQ_PUT_HEX_FIELD(s, field->parent_ip);
863 
864 	return trace_handle_return(s);
865 }
866 
867 static enum print_line_t trace_fn_bin(struct trace_iterator *iter, int flags,
868 				      struct trace_event *event)
869 {
870 	struct ftrace_entry *field;
871 	struct trace_seq *s = &iter->seq;
872 
873 	trace_assign_type(field, iter->ent);
874 
875 	SEQ_PUT_FIELD(s, field->ip);
876 	SEQ_PUT_FIELD(s, field->parent_ip);
877 
878 	return trace_handle_return(s);
879 }
880 
881 static struct trace_event_functions trace_fn_funcs = {
882 	.trace		= trace_fn_trace,
883 	.raw		= trace_fn_raw,
884 	.hex		= trace_fn_hex,
885 	.binary		= trace_fn_bin,
886 };
887 
888 static struct trace_event trace_fn_event = {
889 	.type		= TRACE_FN,
890 	.funcs		= &trace_fn_funcs,
891 };
892 
893 /* TRACE_CTX an TRACE_WAKE */
894 static enum print_line_t trace_ctxwake_print(struct trace_iterator *iter,
895 					     char *delim)
896 {
897 	struct ctx_switch_entry *field;
898 	char comm[TASK_COMM_LEN];
899 	int S, T;
900 
901 
902 	trace_assign_type(field, iter->ent);
903 
904 	T = task_index_to_char(field->next_state);
905 	S = task_index_to_char(field->prev_state);
906 	trace_find_cmdline(field->next_pid, comm);
907 	trace_seq_printf(&iter->seq,
908 			 " %5d:%3d:%c %s [%03d] %5d:%3d:%c %s\n",
909 			 field->prev_pid,
910 			 field->prev_prio,
911 			 S, delim,
912 			 field->next_cpu,
913 			 field->next_pid,
914 			 field->next_prio,
915 			 T, comm);
916 
917 	return trace_handle_return(&iter->seq);
918 }
919 
920 static enum print_line_t trace_ctx_print(struct trace_iterator *iter, int flags,
921 					 struct trace_event *event)
922 {
923 	return trace_ctxwake_print(iter, "==>");
924 }
925 
926 static enum print_line_t trace_wake_print(struct trace_iterator *iter,
927 					  int flags, struct trace_event *event)
928 {
929 	return trace_ctxwake_print(iter, "  +");
930 }
931 
932 static int trace_ctxwake_raw(struct trace_iterator *iter, char S)
933 {
934 	struct ctx_switch_entry *field;
935 	int T;
936 
937 	trace_assign_type(field, iter->ent);
938 
939 	if (!S)
940 		S = task_index_to_char(field->prev_state);
941 	T = task_index_to_char(field->next_state);
942 	trace_seq_printf(&iter->seq, "%d %d %c %d %d %d %c\n",
943 			 field->prev_pid,
944 			 field->prev_prio,
945 			 S,
946 			 field->next_cpu,
947 			 field->next_pid,
948 			 field->next_prio,
949 			 T);
950 
951 	return trace_handle_return(&iter->seq);
952 }
953 
954 static enum print_line_t trace_ctx_raw(struct trace_iterator *iter, int flags,
955 				       struct trace_event *event)
956 {
957 	return trace_ctxwake_raw(iter, 0);
958 }
959 
960 static enum print_line_t trace_wake_raw(struct trace_iterator *iter, int flags,
961 					struct trace_event *event)
962 {
963 	return trace_ctxwake_raw(iter, '+');
964 }
965 
966 
967 static int trace_ctxwake_hex(struct trace_iterator *iter, char S)
968 {
969 	struct ctx_switch_entry *field;
970 	struct trace_seq *s = &iter->seq;
971 	int T;
972 
973 	trace_assign_type(field, iter->ent);
974 
975 	if (!S)
976 		S = task_index_to_char(field->prev_state);
977 	T = task_index_to_char(field->next_state);
978 
979 	SEQ_PUT_HEX_FIELD(s, field->prev_pid);
980 	SEQ_PUT_HEX_FIELD(s, field->prev_prio);
981 	SEQ_PUT_HEX_FIELD(s, S);
982 	SEQ_PUT_HEX_FIELD(s, field->next_cpu);
983 	SEQ_PUT_HEX_FIELD(s, field->next_pid);
984 	SEQ_PUT_HEX_FIELD(s, field->next_prio);
985 	SEQ_PUT_HEX_FIELD(s, T);
986 
987 	return trace_handle_return(s);
988 }
989 
990 static enum print_line_t trace_ctx_hex(struct trace_iterator *iter, int flags,
991 				       struct trace_event *event)
992 {
993 	return trace_ctxwake_hex(iter, 0);
994 }
995 
996 static enum print_line_t trace_wake_hex(struct trace_iterator *iter, int flags,
997 					struct trace_event *event)
998 {
999 	return trace_ctxwake_hex(iter, '+');
1000 }
1001 
1002 static enum print_line_t trace_ctxwake_bin(struct trace_iterator *iter,
1003 					   int flags, struct trace_event *event)
1004 {
1005 	struct ctx_switch_entry *field;
1006 	struct trace_seq *s = &iter->seq;
1007 
1008 	trace_assign_type(field, iter->ent);
1009 
1010 	SEQ_PUT_FIELD(s, field->prev_pid);
1011 	SEQ_PUT_FIELD(s, field->prev_prio);
1012 	SEQ_PUT_FIELD(s, field->prev_state);
1013 	SEQ_PUT_FIELD(s, field->next_cpu);
1014 	SEQ_PUT_FIELD(s, field->next_pid);
1015 	SEQ_PUT_FIELD(s, field->next_prio);
1016 	SEQ_PUT_FIELD(s, field->next_state);
1017 
1018 	return trace_handle_return(s);
1019 }
1020 
1021 static struct trace_event_functions trace_ctx_funcs = {
1022 	.trace		= trace_ctx_print,
1023 	.raw		= trace_ctx_raw,
1024 	.hex		= trace_ctx_hex,
1025 	.binary		= trace_ctxwake_bin,
1026 };
1027 
1028 static struct trace_event trace_ctx_event = {
1029 	.type		= TRACE_CTX,
1030 	.funcs		= &trace_ctx_funcs,
1031 };
1032 
1033 static struct trace_event_functions trace_wake_funcs = {
1034 	.trace		= trace_wake_print,
1035 	.raw		= trace_wake_raw,
1036 	.hex		= trace_wake_hex,
1037 	.binary		= trace_ctxwake_bin,
1038 };
1039 
1040 static struct trace_event trace_wake_event = {
1041 	.type		= TRACE_WAKE,
1042 	.funcs		= &trace_wake_funcs,
1043 };
1044 
1045 /* TRACE_STACK */
1046 
1047 static enum print_line_t trace_stack_print(struct trace_iterator *iter,
1048 					   int flags, struct trace_event *event)
1049 {
1050 	struct stack_entry *field;
1051 	struct trace_seq *s = &iter->seq;
1052 	unsigned long *p;
1053 	unsigned long *end;
1054 
1055 	trace_assign_type(field, iter->ent);
1056 	end = (unsigned long *)((long)iter->ent + iter->ent_size);
1057 
1058 	trace_seq_puts(s, "<stack trace>\n");
1059 
1060 	for (p = field->caller; p && p < end && *p != ULONG_MAX; p++) {
1061 
1062 		if (trace_seq_has_overflowed(s))
1063 			break;
1064 
1065 		trace_seq_puts(s, " => ");
1066 		seq_print_ip_sym(s, *p, flags);
1067 		trace_seq_putc(s, '\n');
1068 	}
1069 
1070 	return trace_handle_return(s);
1071 }
1072 
1073 static struct trace_event_functions trace_stack_funcs = {
1074 	.trace		= trace_stack_print,
1075 };
1076 
1077 static struct trace_event trace_stack_event = {
1078 	.type		= TRACE_STACK,
1079 	.funcs		= &trace_stack_funcs,
1080 };
1081 
1082 /* TRACE_USER_STACK */
1083 static enum print_line_t trace_user_stack_print(struct trace_iterator *iter,
1084 						int flags, struct trace_event *event)
1085 {
1086 	struct trace_array *tr = iter->tr;
1087 	struct userstack_entry *field;
1088 	struct trace_seq *s = &iter->seq;
1089 	struct mm_struct *mm = NULL;
1090 	unsigned int i;
1091 
1092 	trace_assign_type(field, iter->ent);
1093 
1094 	trace_seq_puts(s, "<user stack trace>\n");
1095 
1096 	if (tr->trace_flags & TRACE_ITER_SYM_USEROBJ) {
1097 		struct task_struct *task;
1098 		/*
1099 		 * we do the lookup on the thread group leader,
1100 		 * since individual threads might have already quit!
1101 		 */
1102 		rcu_read_lock();
1103 		task = find_task_by_vpid(field->tgid);
1104 		if (task)
1105 			mm = get_task_mm(task);
1106 		rcu_read_unlock();
1107 	}
1108 
1109 	for (i = 0; i < FTRACE_STACK_ENTRIES; i++) {
1110 		unsigned long ip = field->caller[i];
1111 
1112 		if (!ip || trace_seq_has_overflowed(s))
1113 			break;
1114 
1115 		trace_seq_puts(s, " => ");
1116 		seq_print_user_ip(s, mm, ip, flags);
1117 		trace_seq_putc(s, '\n');
1118 	}
1119 
1120 	if (mm)
1121 		mmput(mm);
1122 
1123 	return trace_handle_return(s);
1124 }
1125 
1126 static struct trace_event_functions trace_user_stack_funcs = {
1127 	.trace		= trace_user_stack_print,
1128 };
1129 
1130 static struct trace_event trace_user_stack_event = {
1131 	.type		= TRACE_USER_STACK,
1132 	.funcs		= &trace_user_stack_funcs,
1133 };
1134 
1135 /* TRACE_HWLAT */
1136 static enum print_line_t
1137 trace_hwlat_print(struct trace_iterator *iter, int flags,
1138 		  struct trace_event *event)
1139 {
1140 	struct trace_entry *entry = iter->ent;
1141 	struct trace_seq *s = &iter->seq;
1142 	struct hwlat_entry *field;
1143 
1144 	trace_assign_type(field, entry);
1145 
1146 	trace_seq_printf(s, "#%-5u inner/outer(us): %4llu/%-5llu ts:%lld.%09ld",
1147 			 field->seqnum,
1148 			 field->duration,
1149 			 field->outer_duration,
1150 			 (long long)field->timestamp.tv_sec,
1151 			 field->timestamp.tv_nsec);
1152 
1153 	if (field->nmi_count) {
1154 		/*
1155 		 * The generic sched_clock() is not NMI safe, thus
1156 		 * we only record the count and not the time.
1157 		 */
1158 		if (!IS_ENABLED(CONFIG_GENERIC_SCHED_CLOCK))
1159 			trace_seq_printf(s, " nmi-total:%llu",
1160 					 field->nmi_total_ts);
1161 		trace_seq_printf(s, " nmi-count:%u",
1162 				 field->nmi_count);
1163 	}
1164 
1165 	trace_seq_putc(s, '\n');
1166 
1167 	return trace_handle_return(s);
1168 }
1169 
1170 
1171 static enum print_line_t
1172 trace_hwlat_raw(struct trace_iterator *iter, int flags,
1173 		struct trace_event *event)
1174 {
1175 	struct hwlat_entry *field;
1176 	struct trace_seq *s = &iter->seq;
1177 
1178 	trace_assign_type(field, iter->ent);
1179 
1180 	trace_seq_printf(s, "%llu %lld %lld %09ld %u\n",
1181 			 field->duration,
1182 			 field->outer_duration,
1183 			 (long long)field->timestamp.tv_sec,
1184 			 field->timestamp.tv_nsec,
1185 			 field->seqnum);
1186 
1187 	return trace_handle_return(s);
1188 }
1189 
1190 static struct trace_event_functions trace_hwlat_funcs = {
1191 	.trace		= trace_hwlat_print,
1192 	.raw		= trace_hwlat_raw,
1193 };
1194 
1195 static struct trace_event trace_hwlat_event = {
1196 	.type		= TRACE_HWLAT,
1197 	.funcs		= &trace_hwlat_funcs,
1198 };
1199 
1200 /* TRACE_BPUTS */
1201 static enum print_line_t
1202 trace_bputs_print(struct trace_iterator *iter, int flags,
1203 		   struct trace_event *event)
1204 {
1205 	struct trace_entry *entry = iter->ent;
1206 	struct trace_seq *s = &iter->seq;
1207 	struct bputs_entry *field;
1208 
1209 	trace_assign_type(field, entry);
1210 
1211 	seq_print_ip_sym(s, field->ip, flags);
1212 	trace_seq_puts(s, ": ");
1213 	trace_seq_puts(s, field->str);
1214 
1215 	return trace_handle_return(s);
1216 }
1217 
1218 
1219 static enum print_line_t
1220 trace_bputs_raw(struct trace_iterator *iter, int flags,
1221 		struct trace_event *event)
1222 {
1223 	struct bputs_entry *field;
1224 	struct trace_seq *s = &iter->seq;
1225 
1226 	trace_assign_type(field, iter->ent);
1227 
1228 	trace_seq_printf(s, ": %lx : ", field->ip);
1229 	trace_seq_puts(s, field->str);
1230 
1231 	return trace_handle_return(s);
1232 }
1233 
1234 static struct trace_event_functions trace_bputs_funcs = {
1235 	.trace		= trace_bputs_print,
1236 	.raw		= trace_bputs_raw,
1237 };
1238 
1239 static struct trace_event trace_bputs_event = {
1240 	.type		= TRACE_BPUTS,
1241 	.funcs		= &trace_bputs_funcs,
1242 };
1243 
1244 /* TRACE_BPRINT */
1245 static enum print_line_t
1246 trace_bprint_print(struct trace_iterator *iter, int flags,
1247 		   struct trace_event *event)
1248 {
1249 	struct trace_entry *entry = iter->ent;
1250 	struct trace_seq *s = &iter->seq;
1251 	struct bprint_entry *field;
1252 
1253 	trace_assign_type(field, entry);
1254 
1255 	seq_print_ip_sym(s, field->ip, flags);
1256 	trace_seq_puts(s, ": ");
1257 	trace_seq_bprintf(s, field->fmt, field->buf);
1258 
1259 	return trace_handle_return(s);
1260 }
1261 
1262 
1263 static enum print_line_t
1264 trace_bprint_raw(struct trace_iterator *iter, int flags,
1265 		 struct trace_event *event)
1266 {
1267 	struct bprint_entry *field;
1268 	struct trace_seq *s = &iter->seq;
1269 
1270 	trace_assign_type(field, iter->ent);
1271 
1272 	trace_seq_printf(s, ": %lx : ", field->ip);
1273 	trace_seq_bprintf(s, field->fmt, field->buf);
1274 
1275 	return trace_handle_return(s);
1276 }
1277 
1278 static struct trace_event_functions trace_bprint_funcs = {
1279 	.trace		= trace_bprint_print,
1280 	.raw		= trace_bprint_raw,
1281 };
1282 
1283 static struct trace_event trace_bprint_event = {
1284 	.type		= TRACE_BPRINT,
1285 	.funcs		= &trace_bprint_funcs,
1286 };
1287 
1288 /* TRACE_PRINT */
1289 static enum print_line_t trace_print_print(struct trace_iterator *iter,
1290 					   int flags, struct trace_event *event)
1291 {
1292 	struct print_entry *field;
1293 	struct trace_seq *s = &iter->seq;
1294 
1295 	trace_assign_type(field, iter->ent);
1296 
1297 	seq_print_ip_sym(s, field->ip, flags);
1298 	trace_seq_printf(s, ": %s", field->buf);
1299 
1300 	return trace_handle_return(s);
1301 }
1302 
1303 static enum print_line_t trace_print_raw(struct trace_iterator *iter, int flags,
1304 					 struct trace_event *event)
1305 {
1306 	struct print_entry *field;
1307 
1308 	trace_assign_type(field, iter->ent);
1309 
1310 	trace_seq_printf(&iter->seq, "# %lx %s", field->ip, field->buf);
1311 
1312 	return trace_handle_return(&iter->seq);
1313 }
1314 
1315 static struct trace_event_functions trace_print_funcs = {
1316 	.trace		= trace_print_print,
1317 	.raw		= trace_print_raw,
1318 };
1319 
1320 static struct trace_event trace_print_event = {
1321 	.type	 	= TRACE_PRINT,
1322 	.funcs		= &trace_print_funcs,
1323 };
1324 
1325 static enum print_line_t trace_raw_data(struct trace_iterator *iter, int flags,
1326 					 struct trace_event *event)
1327 {
1328 	struct raw_data_entry *field;
1329 	int i;
1330 
1331 	trace_assign_type(field, iter->ent);
1332 
1333 	trace_seq_printf(&iter->seq, "# %x buf:", field->id);
1334 
1335 	for (i = 0; i < iter->ent_size - offsetof(struct raw_data_entry, buf); i++)
1336 		trace_seq_printf(&iter->seq, " %02x",
1337 				 (unsigned char)field->buf[i]);
1338 
1339 	trace_seq_putc(&iter->seq, '\n');
1340 
1341 	return trace_handle_return(&iter->seq);
1342 }
1343 
1344 static struct trace_event_functions trace_raw_data_funcs = {
1345 	.trace		= trace_raw_data,
1346 	.raw		= trace_raw_data,
1347 };
1348 
1349 static struct trace_event trace_raw_data_event = {
1350 	.type	 	= TRACE_RAW_DATA,
1351 	.funcs		= &trace_raw_data_funcs,
1352 };
1353 
1354 
1355 static struct trace_event *events[] __initdata = {
1356 	&trace_fn_event,
1357 	&trace_ctx_event,
1358 	&trace_wake_event,
1359 	&trace_stack_event,
1360 	&trace_user_stack_event,
1361 	&trace_bputs_event,
1362 	&trace_bprint_event,
1363 	&trace_print_event,
1364 	&trace_hwlat_event,
1365 	&trace_raw_data_event,
1366 	NULL
1367 };
1368 
1369 __init static int init_events(void)
1370 {
1371 	struct trace_event *event;
1372 	int i, ret;
1373 
1374 	for (i = 0; events[i]; i++) {
1375 		event = events[i];
1376 
1377 		ret = register_trace_event(event);
1378 		if (!ret) {
1379 			printk(KERN_WARNING "event %d failed to register\n",
1380 			       event->type);
1381 			WARN_ON_ONCE(1);
1382 		}
1383 	}
1384 
1385 	return 0;
1386 }
1387 early_initcall(init_events);
1388