xref: /openbmc/linux/kernel/trace/trace_output.c (revision da2ef666)
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 
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 	char str[KSYM_SYMBOL_LEN];
345 #ifdef CONFIG_KALLSYMS
346 	const char *name;
347 
348 	kallsyms_lookup(address, NULL, NULL, NULL, str);
349 
350 	name = kretprobed(str);
351 
352 	if (name && strlen(name)) {
353 		trace_seq_printf(s, fmt, name);
354 		return;
355 	}
356 #endif
357 	snprintf(str, KSYM_SYMBOL_LEN, "0x%08lx", address);
358 	trace_seq_printf(s, fmt, str);
359 }
360 
361 static void
362 seq_print_sym_offset(struct trace_seq *s, const char *fmt,
363 		     unsigned long address)
364 {
365 	char str[KSYM_SYMBOL_LEN];
366 #ifdef CONFIG_KALLSYMS
367 	const char *name;
368 
369 	sprint_symbol(str, address);
370 	name = kretprobed(str);
371 
372 	if (name && strlen(name)) {
373 		trace_seq_printf(s, fmt, name);
374 		return;
375 	}
376 #endif
377 	snprintf(str, KSYM_SYMBOL_LEN, "0x%08lx", address);
378 	trace_seq_printf(s, fmt, str);
379 }
380 
381 #ifndef CONFIG_64BIT
382 # define IP_FMT "%08lx"
383 #else
384 # define IP_FMT "%016lx"
385 #endif
386 
387 static int seq_print_user_ip(struct trace_seq *s, struct mm_struct *mm,
388 			     unsigned long ip, unsigned long sym_flags)
389 {
390 	struct file *file = NULL;
391 	unsigned long vmstart = 0;
392 	int ret = 1;
393 
394 	if (s->full)
395 		return 0;
396 
397 	if (mm) {
398 		const struct vm_area_struct *vma;
399 
400 		down_read(&mm->mmap_sem);
401 		vma = find_vma(mm, ip);
402 		if (vma) {
403 			file = vma->vm_file;
404 			vmstart = vma->vm_start;
405 		}
406 		if (file) {
407 			ret = trace_seq_path(s, &file->f_path);
408 			if (ret)
409 				trace_seq_printf(s, "[+0x%lx]",
410 						 ip - vmstart);
411 		}
412 		up_read(&mm->mmap_sem);
413 	}
414 	if (ret && ((sym_flags & TRACE_ITER_SYM_ADDR) || !file))
415 		trace_seq_printf(s, " <" IP_FMT ">", ip);
416 	return !trace_seq_has_overflowed(s);
417 }
418 
419 int
420 seq_print_ip_sym(struct trace_seq *s, unsigned long ip, unsigned long sym_flags)
421 {
422 	if (!ip) {
423 		trace_seq_putc(s, '0');
424 		goto out;
425 	}
426 
427 	if (sym_flags & TRACE_ITER_SYM_OFFSET)
428 		seq_print_sym_offset(s, "%s", ip);
429 	else
430 		seq_print_sym_short(s, "%s", ip);
431 
432 	if (sym_flags & TRACE_ITER_SYM_ADDR)
433 		trace_seq_printf(s, " <" IP_FMT ">", ip);
434 
435  out:
436 	return !trace_seq_has_overflowed(s);
437 }
438 
439 /**
440  * trace_print_lat_fmt - print the irq, preempt and lockdep fields
441  * @s: trace seq struct to write to
442  * @entry: The trace entry field from the ring buffer
443  *
444  * Prints the generic fields of irqs off, in hard or softirq, preempt
445  * count.
446  */
447 int trace_print_lat_fmt(struct trace_seq *s, struct trace_entry *entry)
448 {
449 	char hardsoft_irq;
450 	char need_resched;
451 	char irqs_off;
452 	int hardirq;
453 	int softirq;
454 	int nmi;
455 
456 	nmi = entry->flags & TRACE_FLAG_NMI;
457 	hardirq = entry->flags & TRACE_FLAG_HARDIRQ;
458 	softirq = entry->flags & TRACE_FLAG_SOFTIRQ;
459 
460 	irqs_off =
461 		(entry->flags & TRACE_FLAG_IRQS_OFF) ? 'd' :
462 		(entry->flags & TRACE_FLAG_IRQS_NOSUPPORT) ? 'X' :
463 		'.';
464 
465 	switch (entry->flags & (TRACE_FLAG_NEED_RESCHED |
466 				TRACE_FLAG_PREEMPT_RESCHED)) {
467 	case TRACE_FLAG_NEED_RESCHED | TRACE_FLAG_PREEMPT_RESCHED:
468 		need_resched = 'N';
469 		break;
470 	case TRACE_FLAG_NEED_RESCHED:
471 		need_resched = 'n';
472 		break;
473 	case TRACE_FLAG_PREEMPT_RESCHED:
474 		need_resched = 'p';
475 		break;
476 	default:
477 		need_resched = '.';
478 		break;
479 	}
480 
481 	hardsoft_irq =
482 		(nmi && hardirq)     ? 'Z' :
483 		nmi                  ? 'z' :
484 		(hardirq && softirq) ? 'H' :
485 		hardirq              ? 'h' :
486 		softirq              ? 's' :
487 		                       '.' ;
488 
489 	trace_seq_printf(s, "%c%c%c",
490 			 irqs_off, need_resched, hardsoft_irq);
491 
492 	if (entry->preempt_count)
493 		trace_seq_printf(s, "%x", entry->preempt_count);
494 	else
495 		trace_seq_putc(s, '.');
496 
497 	return !trace_seq_has_overflowed(s);
498 }
499 
500 static int
501 lat_print_generic(struct trace_seq *s, struct trace_entry *entry, int cpu)
502 {
503 	char comm[TASK_COMM_LEN];
504 
505 	trace_find_cmdline(entry->pid, comm);
506 
507 	trace_seq_printf(s, "%8.8s-%-5d %3d",
508 			 comm, entry->pid, cpu);
509 
510 	return trace_print_lat_fmt(s, entry);
511 }
512 
513 #undef MARK
514 #define MARK(v, s) {.val = v, .sym = s}
515 /* trace overhead mark */
516 static const struct trace_mark {
517 	unsigned long long	val; /* unit: nsec */
518 	char			sym;
519 } mark[] = {
520 	MARK(1000000000ULL	, '$'), /* 1 sec */
521 	MARK(100000000ULL	, '@'), /* 100 msec */
522 	MARK(10000000ULL	, '*'), /* 10 msec */
523 	MARK(1000000ULL		, '#'), /* 1000 usecs */
524 	MARK(100000ULL		, '!'), /* 100 usecs */
525 	MARK(10000ULL		, '+'), /* 10 usecs */
526 };
527 #undef MARK
528 
529 char trace_find_mark(unsigned long long d)
530 {
531 	int i;
532 	int size = ARRAY_SIZE(mark);
533 
534 	for (i = 0; i < size; i++) {
535 		if (d > mark[i].val)
536 			break;
537 	}
538 
539 	return (i == size) ? ' ' : mark[i].sym;
540 }
541 
542 static int
543 lat_print_timestamp(struct trace_iterator *iter, u64 next_ts)
544 {
545 	struct trace_array *tr = iter->tr;
546 	unsigned long verbose = tr->trace_flags & TRACE_ITER_VERBOSE;
547 	unsigned long in_ns = iter->iter_flags & TRACE_FILE_TIME_IN_NS;
548 	unsigned long long abs_ts = iter->ts - iter->trace_buffer->time_start;
549 	unsigned long long rel_ts = next_ts - iter->ts;
550 	struct trace_seq *s = &iter->seq;
551 
552 	if (in_ns) {
553 		abs_ts = ns2usecs(abs_ts);
554 		rel_ts = ns2usecs(rel_ts);
555 	}
556 
557 	if (verbose && in_ns) {
558 		unsigned long abs_usec = do_div(abs_ts, USEC_PER_MSEC);
559 		unsigned long abs_msec = (unsigned long)abs_ts;
560 		unsigned long rel_usec = do_div(rel_ts, USEC_PER_MSEC);
561 		unsigned long rel_msec = (unsigned long)rel_ts;
562 
563 		trace_seq_printf(
564 			s, "[%08llx] %ld.%03ldms (+%ld.%03ldms): ",
565 			ns2usecs(iter->ts),
566 			abs_msec, abs_usec,
567 			rel_msec, rel_usec);
568 
569 	} else if (verbose && !in_ns) {
570 		trace_seq_printf(
571 			s, "[%016llx] %lld (+%lld): ",
572 			iter->ts, abs_ts, rel_ts);
573 
574 	} else if (!verbose && in_ns) {
575 		trace_seq_printf(
576 			s, " %4lldus%c: ",
577 			abs_ts,
578 			trace_find_mark(rel_ts * NSEC_PER_USEC));
579 
580 	} else { /* !verbose && !in_ns */
581 		trace_seq_printf(s, " %4lld: ", abs_ts);
582 	}
583 
584 	return !trace_seq_has_overflowed(s);
585 }
586 
587 int trace_print_context(struct trace_iterator *iter)
588 {
589 	struct trace_array *tr = iter->tr;
590 	struct trace_seq *s = &iter->seq;
591 	struct trace_entry *entry = iter->ent;
592 	unsigned long long t;
593 	unsigned long secs, usec_rem;
594 	char comm[TASK_COMM_LEN];
595 
596 	trace_find_cmdline(entry->pid, comm);
597 
598 	trace_seq_printf(s, "%16s-%-5d ", comm, entry->pid);
599 
600 	if (tr->trace_flags & TRACE_ITER_RECORD_TGID) {
601 		unsigned int tgid = trace_find_tgid(entry->pid);
602 
603 		if (!tgid)
604 			trace_seq_printf(s, "(-----) ");
605 		else
606 			trace_seq_printf(s, "(%5d) ", tgid);
607 	}
608 
609 	trace_seq_printf(s, "[%03d] ", iter->cpu);
610 
611 	if (tr->trace_flags & TRACE_ITER_IRQ_INFO)
612 		trace_print_lat_fmt(s, entry);
613 
614 	if (iter->iter_flags & TRACE_FILE_TIME_IN_NS) {
615 		t = ns2usecs(iter->ts);
616 		usec_rem = do_div(t, USEC_PER_SEC);
617 		secs = (unsigned long)t;
618 		trace_seq_printf(s, " %5lu.%06lu: ", secs, usec_rem);
619 	} else
620 		trace_seq_printf(s, " %12llu: ", iter->ts);
621 
622 	return !trace_seq_has_overflowed(s);
623 }
624 
625 int trace_print_lat_context(struct trace_iterator *iter)
626 {
627 	struct trace_array *tr = iter->tr;
628 	/* trace_find_next_entry will reset ent_size */
629 	int ent_size = iter->ent_size;
630 	struct trace_seq *s = &iter->seq;
631 	u64 next_ts;
632 	struct trace_entry *entry = iter->ent,
633 			   *next_entry = trace_find_next_entry(iter, NULL,
634 							       &next_ts);
635 	unsigned long verbose = (tr->trace_flags & TRACE_ITER_VERBOSE);
636 
637 	/* Restore the original ent_size */
638 	iter->ent_size = ent_size;
639 
640 	if (!next_entry)
641 		next_ts = iter->ts;
642 
643 	if (verbose) {
644 		char comm[TASK_COMM_LEN];
645 
646 		trace_find_cmdline(entry->pid, comm);
647 
648 		trace_seq_printf(
649 			s, "%16s %5d %3d %d %08x %08lx ",
650 			comm, entry->pid, iter->cpu, entry->flags,
651 			entry->preempt_count, iter->idx);
652 	} else {
653 		lat_print_generic(s, entry, iter->cpu);
654 	}
655 
656 	lat_print_timestamp(iter, next_ts);
657 
658 	return !trace_seq_has_overflowed(s);
659 }
660 
661 /**
662  * ftrace_find_event - find a registered event
663  * @type: the type of event to look for
664  *
665  * Returns an event of type @type otherwise NULL
666  * Called with trace_event_read_lock() held.
667  */
668 struct trace_event *ftrace_find_event(int type)
669 {
670 	struct trace_event *event;
671 	unsigned key;
672 
673 	key = type & (EVENT_HASHSIZE - 1);
674 
675 	hlist_for_each_entry(event, &event_hash[key], node) {
676 		if (event->type == type)
677 			return event;
678 	}
679 
680 	return NULL;
681 }
682 
683 static LIST_HEAD(ftrace_event_list);
684 
685 static int trace_search_list(struct list_head **list)
686 {
687 	struct trace_event *e;
688 	int last = __TRACE_LAST_TYPE;
689 
690 	if (list_empty(&ftrace_event_list)) {
691 		*list = &ftrace_event_list;
692 		return last + 1;
693 	}
694 
695 	/*
696 	 * We used up all possible max events,
697 	 * lets see if somebody freed one.
698 	 */
699 	list_for_each_entry(e, &ftrace_event_list, list) {
700 		if (e->type != last + 1)
701 			break;
702 		last++;
703 	}
704 
705 	/* Did we used up all 65 thousand events??? */
706 	if ((last + 1) > TRACE_EVENT_TYPE_MAX)
707 		return 0;
708 
709 	*list = &e->list;
710 	return last + 1;
711 }
712 
713 void trace_event_read_lock(void)
714 {
715 	down_read(&trace_event_sem);
716 }
717 
718 void trace_event_read_unlock(void)
719 {
720 	up_read(&trace_event_sem);
721 }
722 
723 /**
724  * register_trace_event - register output for an event type
725  * @event: the event type to register
726  *
727  * Event types are stored in a hash and this hash is used to
728  * find a way to print an event. If the @event->type is set
729  * then it will use that type, otherwise it will assign a
730  * type to use.
731  *
732  * If you assign your own type, please make sure it is added
733  * to the trace_type enum in trace.h, to avoid collisions
734  * with the dynamic types.
735  *
736  * Returns the event type number or zero on error.
737  */
738 int register_trace_event(struct trace_event *event)
739 {
740 	unsigned key;
741 	int ret = 0;
742 
743 	down_write(&trace_event_sem);
744 
745 	if (WARN_ON(!event))
746 		goto out;
747 
748 	if (WARN_ON(!event->funcs))
749 		goto out;
750 
751 	INIT_LIST_HEAD(&event->list);
752 
753 	if (!event->type) {
754 		struct list_head *list = NULL;
755 
756 		if (next_event_type > TRACE_EVENT_TYPE_MAX) {
757 
758 			event->type = trace_search_list(&list);
759 			if (!event->type)
760 				goto out;
761 
762 		} else {
763 
764 			event->type = next_event_type++;
765 			list = &ftrace_event_list;
766 		}
767 
768 		if (WARN_ON(ftrace_find_event(event->type)))
769 			goto out;
770 
771 		list_add_tail(&event->list, list);
772 
773 	} else if (event->type > __TRACE_LAST_TYPE) {
774 		printk(KERN_WARNING "Need to add type to trace.h\n");
775 		WARN_ON(1);
776 		goto out;
777 	} else {
778 		/* Is this event already used */
779 		if (ftrace_find_event(event->type))
780 			goto out;
781 	}
782 
783 	if (event->funcs->trace == NULL)
784 		event->funcs->trace = trace_nop_print;
785 	if (event->funcs->raw == NULL)
786 		event->funcs->raw = trace_nop_print;
787 	if (event->funcs->hex == NULL)
788 		event->funcs->hex = trace_nop_print;
789 	if (event->funcs->binary == NULL)
790 		event->funcs->binary = trace_nop_print;
791 
792 	key = event->type & (EVENT_HASHSIZE - 1);
793 
794 	hlist_add_head(&event->node, &event_hash[key]);
795 
796 	ret = event->type;
797  out:
798 	up_write(&trace_event_sem);
799 
800 	return ret;
801 }
802 EXPORT_SYMBOL_GPL(register_trace_event);
803 
804 /*
805  * Used by module code with the trace_event_sem held for write.
806  */
807 int __unregister_trace_event(struct trace_event *event)
808 {
809 	hlist_del(&event->node);
810 	list_del(&event->list);
811 	return 0;
812 }
813 
814 /**
815  * unregister_trace_event - remove a no longer used event
816  * @event: the event to remove
817  */
818 int unregister_trace_event(struct trace_event *event)
819 {
820 	down_write(&trace_event_sem);
821 	__unregister_trace_event(event);
822 	up_write(&trace_event_sem);
823 
824 	return 0;
825 }
826 EXPORT_SYMBOL_GPL(unregister_trace_event);
827 
828 /*
829  * Standard events
830  */
831 
832 enum print_line_t trace_nop_print(struct trace_iterator *iter, int flags,
833 				  struct trace_event *event)
834 {
835 	trace_seq_printf(&iter->seq, "type: %d\n", iter->ent->type);
836 
837 	return trace_handle_return(&iter->seq);
838 }
839 
840 /* TRACE_FN */
841 static enum print_line_t trace_fn_trace(struct trace_iterator *iter, int flags,
842 					struct trace_event *event)
843 {
844 	struct ftrace_entry *field;
845 	struct trace_seq *s = &iter->seq;
846 
847 	trace_assign_type(field, iter->ent);
848 
849 	seq_print_ip_sym(s, field->ip, flags);
850 
851 	if ((flags & TRACE_ITER_PRINT_PARENT) && field->parent_ip) {
852 		trace_seq_puts(s, " <-");
853 		seq_print_ip_sym(s, field->parent_ip, flags);
854 	}
855 
856 	trace_seq_putc(s, '\n');
857 
858 	return trace_handle_return(s);
859 }
860 
861 static enum print_line_t trace_fn_raw(struct trace_iterator *iter, int flags,
862 				      struct trace_event *event)
863 {
864 	struct ftrace_entry *field;
865 
866 	trace_assign_type(field, iter->ent);
867 
868 	trace_seq_printf(&iter->seq, "%lx %lx\n",
869 			 field->ip,
870 			 field->parent_ip);
871 
872 	return trace_handle_return(&iter->seq);
873 }
874 
875 static enum print_line_t trace_fn_hex(struct trace_iterator *iter, int flags,
876 				      struct trace_event *event)
877 {
878 	struct ftrace_entry *field;
879 	struct trace_seq *s = &iter->seq;
880 
881 	trace_assign_type(field, iter->ent);
882 
883 	SEQ_PUT_HEX_FIELD(s, field->ip);
884 	SEQ_PUT_HEX_FIELD(s, field->parent_ip);
885 
886 	return trace_handle_return(s);
887 }
888 
889 static enum print_line_t trace_fn_bin(struct trace_iterator *iter, int flags,
890 				      struct trace_event *event)
891 {
892 	struct ftrace_entry *field;
893 	struct trace_seq *s = &iter->seq;
894 
895 	trace_assign_type(field, iter->ent);
896 
897 	SEQ_PUT_FIELD(s, field->ip);
898 	SEQ_PUT_FIELD(s, field->parent_ip);
899 
900 	return trace_handle_return(s);
901 }
902 
903 static struct trace_event_functions trace_fn_funcs = {
904 	.trace		= trace_fn_trace,
905 	.raw		= trace_fn_raw,
906 	.hex		= trace_fn_hex,
907 	.binary		= trace_fn_bin,
908 };
909 
910 static struct trace_event trace_fn_event = {
911 	.type		= TRACE_FN,
912 	.funcs		= &trace_fn_funcs,
913 };
914 
915 /* TRACE_CTX an TRACE_WAKE */
916 static enum print_line_t trace_ctxwake_print(struct trace_iterator *iter,
917 					     char *delim)
918 {
919 	struct ctx_switch_entry *field;
920 	char comm[TASK_COMM_LEN];
921 	int S, T;
922 
923 
924 	trace_assign_type(field, iter->ent);
925 
926 	T = task_index_to_char(field->next_state);
927 	S = task_index_to_char(field->prev_state);
928 	trace_find_cmdline(field->next_pid, comm);
929 	trace_seq_printf(&iter->seq,
930 			 " %5d:%3d:%c %s [%03d] %5d:%3d:%c %s\n",
931 			 field->prev_pid,
932 			 field->prev_prio,
933 			 S, delim,
934 			 field->next_cpu,
935 			 field->next_pid,
936 			 field->next_prio,
937 			 T, comm);
938 
939 	return trace_handle_return(&iter->seq);
940 }
941 
942 static enum print_line_t trace_ctx_print(struct trace_iterator *iter, int flags,
943 					 struct trace_event *event)
944 {
945 	return trace_ctxwake_print(iter, "==>");
946 }
947 
948 static enum print_line_t trace_wake_print(struct trace_iterator *iter,
949 					  int flags, struct trace_event *event)
950 {
951 	return trace_ctxwake_print(iter, "  +");
952 }
953 
954 static int trace_ctxwake_raw(struct trace_iterator *iter, char S)
955 {
956 	struct ctx_switch_entry *field;
957 	int T;
958 
959 	trace_assign_type(field, iter->ent);
960 
961 	if (!S)
962 		S = task_index_to_char(field->prev_state);
963 	T = task_index_to_char(field->next_state);
964 	trace_seq_printf(&iter->seq, "%d %d %c %d %d %d %c\n",
965 			 field->prev_pid,
966 			 field->prev_prio,
967 			 S,
968 			 field->next_cpu,
969 			 field->next_pid,
970 			 field->next_prio,
971 			 T);
972 
973 	return trace_handle_return(&iter->seq);
974 }
975 
976 static enum print_line_t trace_ctx_raw(struct trace_iterator *iter, int flags,
977 				       struct trace_event *event)
978 {
979 	return trace_ctxwake_raw(iter, 0);
980 }
981 
982 static enum print_line_t trace_wake_raw(struct trace_iterator *iter, int flags,
983 					struct trace_event *event)
984 {
985 	return trace_ctxwake_raw(iter, '+');
986 }
987 
988 
989 static int trace_ctxwake_hex(struct trace_iterator *iter, char S)
990 {
991 	struct ctx_switch_entry *field;
992 	struct trace_seq *s = &iter->seq;
993 	int T;
994 
995 	trace_assign_type(field, iter->ent);
996 
997 	if (!S)
998 		S = task_index_to_char(field->prev_state);
999 	T = task_index_to_char(field->next_state);
1000 
1001 	SEQ_PUT_HEX_FIELD(s, field->prev_pid);
1002 	SEQ_PUT_HEX_FIELD(s, field->prev_prio);
1003 	SEQ_PUT_HEX_FIELD(s, S);
1004 	SEQ_PUT_HEX_FIELD(s, field->next_cpu);
1005 	SEQ_PUT_HEX_FIELD(s, field->next_pid);
1006 	SEQ_PUT_HEX_FIELD(s, field->next_prio);
1007 	SEQ_PUT_HEX_FIELD(s, T);
1008 
1009 	return trace_handle_return(s);
1010 }
1011 
1012 static enum print_line_t trace_ctx_hex(struct trace_iterator *iter, int flags,
1013 				       struct trace_event *event)
1014 {
1015 	return trace_ctxwake_hex(iter, 0);
1016 }
1017 
1018 static enum print_line_t trace_wake_hex(struct trace_iterator *iter, int flags,
1019 					struct trace_event *event)
1020 {
1021 	return trace_ctxwake_hex(iter, '+');
1022 }
1023 
1024 static enum print_line_t trace_ctxwake_bin(struct trace_iterator *iter,
1025 					   int flags, struct trace_event *event)
1026 {
1027 	struct ctx_switch_entry *field;
1028 	struct trace_seq *s = &iter->seq;
1029 
1030 	trace_assign_type(field, iter->ent);
1031 
1032 	SEQ_PUT_FIELD(s, field->prev_pid);
1033 	SEQ_PUT_FIELD(s, field->prev_prio);
1034 	SEQ_PUT_FIELD(s, field->prev_state);
1035 	SEQ_PUT_FIELD(s, field->next_cpu);
1036 	SEQ_PUT_FIELD(s, field->next_pid);
1037 	SEQ_PUT_FIELD(s, field->next_prio);
1038 	SEQ_PUT_FIELD(s, field->next_state);
1039 
1040 	return trace_handle_return(s);
1041 }
1042 
1043 static struct trace_event_functions trace_ctx_funcs = {
1044 	.trace		= trace_ctx_print,
1045 	.raw		= trace_ctx_raw,
1046 	.hex		= trace_ctx_hex,
1047 	.binary		= trace_ctxwake_bin,
1048 };
1049 
1050 static struct trace_event trace_ctx_event = {
1051 	.type		= TRACE_CTX,
1052 	.funcs		= &trace_ctx_funcs,
1053 };
1054 
1055 static struct trace_event_functions trace_wake_funcs = {
1056 	.trace		= trace_wake_print,
1057 	.raw		= trace_wake_raw,
1058 	.hex		= trace_wake_hex,
1059 	.binary		= trace_ctxwake_bin,
1060 };
1061 
1062 static struct trace_event trace_wake_event = {
1063 	.type		= TRACE_WAKE,
1064 	.funcs		= &trace_wake_funcs,
1065 };
1066 
1067 /* TRACE_STACK */
1068 
1069 static enum print_line_t trace_stack_print(struct trace_iterator *iter,
1070 					   int flags, struct trace_event *event)
1071 {
1072 	struct stack_entry *field;
1073 	struct trace_seq *s = &iter->seq;
1074 	unsigned long *p;
1075 	unsigned long *end;
1076 
1077 	trace_assign_type(field, iter->ent);
1078 	end = (unsigned long *)((long)iter->ent + iter->ent_size);
1079 
1080 	trace_seq_puts(s, "<stack trace>\n");
1081 
1082 	for (p = field->caller; p && *p != ULONG_MAX && p < end; p++) {
1083 
1084 		if (trace_seq_has_overflowed(s))
1085 			break;
1086 
1087 		trace_seq_puts(s, " => ");
1088 		seq_print_ip_sym(s, *p, flags);
1089 		trace_seq_putc(s, '\n');
1090 	}
1091 
1092 	return trace_handle_return(s);
1093 }
1094 
1095 static struct trace_event_functions trace_stack_funcs = {
1096 	.trace		= trace_stack_print,
1097 };
1098 
1099 static struct trace_event trace_stack_event = {
1100 	.type		= TRACE_STACK,
1101 	.funcs		= &trace_stack_funcs,
1102 };
1103 
1104 /* TRACE_USER_STACK */
1105 static enum print_line_t trace_user_stack_print(struct trace_iterator *iter,
1106 						int flags, struct trace_event *event)
1107 {
1108 	struct trace_array *tr = iter->tr;
1109 	struct userstack_entry *field;
1110 	struct trace_seq *s = &iter->seq;
1111 	struct mm_struct *mm = NULL;
1112 	unsigned int i;
1113 
1114 	trace_assign_type(field, iter->ent);
1115 
1116 	trace_seq_puts(s, "<user stack trace>\n");
1117 
1118 	if (tr->trace_flags & TRACE_ITER_SYM_USEROBJ) {
1119 		struct task_struct *task;
1120 		/*
1121 		 * we do the lookup on the thread group leader,
1122 		 * since individual threads might have already quit!
1123 		 */
1124 		rcu_read_lock();
1125 		task = find_task_by_vpid(field->tgid);
1126 		if (task)
1127 			mm = get_task_mm(task);
1128 		rcu_read_unlock();
1129 	}
1130 
1131 	for (i = 0; i < FTRACE_STACK_ENTRIES; i++) {
1132 		unsigned long ip = field->caller[i];
1133 
1134 		if (ip == ULONG_MAX || trace_seq_has_overflowed(s))
1135 			break;
1136 
1137 		trace_seq_puts(s, " => ");
1138 
1139 		if (!ip) {
1140 			trace_seq_puts(s, "??");
1141 			trace_seq_putc(s, '\n');
1142 			continue;
1143 		}
1144 
1145 		seq_print_user_ip(s, mm, ip, flags);
1146 		trace_seq_putc(s, '\n');
1147 	}
1148 
1149 	if (mm)
1150 		mmput(mm);
1151 
1152 	return trace_handle_return(s);
1153 }
1154 
1155 static struct trace_event_functions trace_user_stack_funcs = {
1156 	.trace		= trace_user_stack_print,
1157 };
1158 
1159 static struct trace_event trace_user_stack_event = {
1160 	.type		= TRACE_USER_STACK,
1161 	.funcs		= &trace_user_stack_funcs,
1162 };
1163 
1164 /* TRACE_HWLAT */
1165 static enum print_line_t
1166 trace_hwlat_print(struct trace_iterator *iter, int flags,
1167 		  struct trace_event *event)
1168 {
1169 	struct trace_entry *entry = iter->ent;
1170 	struct trace_seq *s = &iter->seq;
1171 	struct hwlat_entry *field;
1172 
1173 	trace_assign_type(field, entry);
1174 
1175 	trace_seq_printf(s, "#%-5u inner/outer(us): %4llu/%-5llu ts:%lld.%09ld",
1176 			 field->seqnum,
1177 			 field->duration,
1178 			 field->outer_duration,
1179 			 (long long)field->timestamp.tv_sec,
1180 			 field->timestamp.tv_nsec);
1181 
1182 	if (field->nmi_count) {
1183 		/*
1184 		 * The generic sched_clock() is not NMI safe, thus
1185 		 * we only record the count and not the time.
1186 		 */
1187 		if (!IS_ENABLED(CONFIG_GENERIC_SCHED_CLOCK))
1188 			trace_seq_printf(s, " nmi-total:%llu",
1189 					 field->nmi_total_ts);
1190 		trace_seq_printf(s, " nmi-count:%u",
1191 				 field->nmi_count);
1192 	}
1193 
1194 	trace_seq_putc(s, '\n');
1195 
1196 	return trace_handle_return(s);
1197 }
1198 
1199 
1200 static enum print_line_t
1201 trace_hwlat_raw(struct trace_iterator *iter, int flags,
1202 		struct trace_event *event)
1203 {
1204 	struct hwlat_entry *field;
1205 	struct trace_seq *s = &iter->seq;
1206 
1207 	trace_assign_type(field, iter->ent);
1208 
1209 	trace_seq_printf(s, "%llu %lld %lld %09ld %u\n",
1210 			 field->duration,
1211 			 field->outer_duration,
1212 			 (long long)field->timestamp.tv_sec,
1213 			 field->timestamp.tv_nsec,
1214 			 field->seqnum);
1215 
1216 	return trace_handle_return(s);
1217 }
1218 
1219 static struct trace_event_functions trace_hwlat_funcs = {
1220 	.trace		= trace_hwlat_print,
1221 	.raw		= trace_hwlat_raw,
1222 };
1223 
1224 static struct trace_event trace_hwlat_event = {
1225 	.type		= TRACE_HWLAT,
1226 	.funcs		= &trace_hwlat_funcs,
1227 };
1228 
1229 /* TRACE_BPUTS */
1230 static enum print_line_t
1231 trace_bputs_print(struct trace_iterator *iter, int flags,
1232 		   struct trace_event *event)
1233 {
1234 	struct trace_entry *entry = iter->ent;
1235 	struct trace_seq *s = &iter->seq;
1236 	struct bputs_entry *field;
1237 
1238 	trace_assign_type(field, entry);
1239 
1240 	seq_print_ip_sym(s, field->ip, flags);
1241 	trace_seq_puts(s, ": ");
1242 	trace_seq_puts(s, field->str);
1243 
1244 	return trace_handle_return(s);
1245 }
1246 
1247 
1248 static enum print_line_t
1249 trace_bputs_raw(struct trace_iterator *iter, int flags,
1250 		struct trace_event *event)
1251 {
1252 	struct bputs_entry *field;
1253 	struct trace_seq *s = &iter->seq;
1254 
1255 	trace_assign_type(field, iter->ent);
1256 
1257 	trace_seq_printf(s, ": %lx : ", field->ip);
1258 	trace_seq_puts(s, field->str);
1259 
1260 	return trace_handle_return(s);
1261 }
1262 
1263 static struct trace_event_functions trace_bputs_funcs = {
1264 	.trace		= trace_bputs_print,
1265 	.raw		= trace_bputs_raw,
1266 };
1267 
1268 static struct trace_event trace_bputs_event = {
1269 	.type		= TRACE_BPUTS,
1270 	.funcs		= &trace_bputs_funcs,
1271 };
1272 
1273 /* TRACE_BPRINT */
1274 static enum print_line_t
1275 trace_bprint_print(struct trace_iterator *iter, int flags,
1276 		   struct trace_event *event)
1277 {
1278 	struct trace_entry *entry = iter->ent;
1279 	struct trace_seq *s = &iter->seq;
1280 	struct bprint_entry *field;
1281 
1282 	trace_assign_type(field, entry);
1283 
1284 	seq_print_ip_sym(s, field->ip, flags);
1285 	trace_seq_puts(s, ": ");
1286 	trace_seq_bprintf(s, field->fmt, field->buf);
1287 
1288 	return trace_handle_return(s);
1289 }
1290 
1291 
1292 static enum print_line_t
1293 trace_bprint_raw(struct trace_iterator *iter, int flags,
1294 		 struct trace_event *event)
1295 {
1296 	struct bprint_entry *field;
1297 	struct trace_seq *s = &iter->seq;
1298 
1299 	trace_assign_type(field, iter->ent);
1300 
1301 	trace_seq_printf(s, ": %lx : ", field->ip);
1302 	trace_seq_bprintf(s, field->fmt, field->buf);
1303 
1304 	return trace_handle_return(s);
1305 }
1306 
1307 static struct trace_event_functions trace_bprint_funcs = {
1308 	.trace		= trace_bprint_print,
1309 	.raw		= trace_bprint_raw,
1310 };
1311 
1312 static struct trace_event trace_bprint_event = {
1313 	.type		= TRACE_BPRINT,
1314 	.funcs		= &trace_bprint_funcs,
1315 };
1316 
1317 /* TRACE_PRINT */
1318 static enum print_line_t trace_print_print(struct trace_iterator *iter,
1319 					   int flags, struct trace_event *event)
1320 {
1321 	struct print_entry *field;
1322 	struct trace_seq *s = &iter->seq;
1323 
1324 	trace_assign_type(field, iter->ent);
1325 
1326 	seq_print_ip_sym(s, field->ip, flags);
1327 	trace_seq_printf(s, ": %s", field->buf);
1328 
1329 	return trace_handle_return(s);
1330 }
1331 
1332 static enum print_line_t trace_print_raw(struct trace_iterator *iter, int flags,
1333 					 struct trace_event *event)
1334 {
1335 	struct print_entry *field;
1336 
1337 	trace_assign_type(field, iter->ent);
1338 
1339 	trace_seq_printf(&iter->seq, "# %lx %s", field->ip, field->buf);
1340 
1341 	return trace_handle_return(&iter->seq);
1342 }
1343 
1344 static struct trace_event_functions trace_print_funcs = {
1345 	.trace		= trace_print_print,
1346 	.raw		= trace_print_raw,
1347 };
1348 
1349 static struct trace_event trace_print_event = {
1350 	.type	 	= TRACE_PRINT,
1351 	.funcs		= &trace_print_funcs,
1352 };
1353 
1354 static enum print_line_t trace_raw_data(struct trace_iterator *iter, int flags,
1355 					 struct trace_event *event)
1356 {
1357 	struct raw_data_entry *field;
1358 	int i;
1359 
1360 	trace_assign_type(field, iter->ent);
1361 
1362 	trace_seq_printf(&iter->seq, "# %x buf:", field->id);
1363 
1364 	for (i = 0; i < iter->ent_size - offsetof(struct raw_data_entry, buf); i++)
1365 		trace_seq_printf(&iter->seq, " %02x",
1366 				 (unsigned char)field->buf[i]);
1367 
1368 	trace_seq_putc(&iter->seq, '\n');
1369 
1370 	return trace_handle_return(&iter->seq);
1371 }
1372 
1373 static struct trace_event_functions trace_raw_data_funcs = {
1374 	.trace		= trace_raw_data,
1375 	.raw		= trace_raw_data,
1376 };
1377 
1378 static struct trace_event trace_raw_data_event = {
1379 	.type	 	= TRACE_RAW_DATA,
1380 	.funcs		= &trace_raw_data_funcs,
1381 };
1382 
1383 
1384 static struct trace_event *events[] __initdata = {
1385 	&trace_fn_event,
1386 	&trace_ctx_event,
1387 	&trace_wake_event,
1388 	&trace_stack_event,
1389 	&trace_user_stack_event,
1390 	&trace_bputs_event,
1391 	&trace_bprint_event,
1392 	&trace_print_event,
1393 	&trace_hwlat_event,
1394 	&trace_raw_data_event,
1395 	NULL
1396 };
1397 
1398 __init static int init_events(void)
1399 {
1400 	struct trace_event *event;
1401 	int i, ret;
1402 
1403 	for (i = 0; events[i]; i++) {
1404 		event = events[i];
1405 
1406 		ret = register_trace_event(event);
1407 		if (!ret) {
1408 			printk(KERN_WARNING "event %d failed to register\n",
1409 			       event->type);
1410 			WARN_ON_ONCE(1);
1411 		}
1412 	}
1413 
1414 	return 0;
1415 }
1416 early_initcall(init_events);
1417