trace_functions_graph.c 28.7 KB
Newer Older
1
2
3
/*
 *
 * Function graph tracer.
4
 * Copyright (c) 2008-2009 Frederic Weisbecker <fweisbec@gmail.com>
5
6
7
8
9
10
11
12
13
14
 * Mostly borrowed from function tracer which
 * is Copyright (c) Steven Rostedt <srostedt@redhat.com>
 *
 */
#include <linux/debugfs.h>
#include <linux/uaccess.h>
#include <linux/ftrace.h>
#include <linux/fs.h>

#include "trace.h"
15
#include "trace_output.h"
16

17
struct fgraph_cpu_data {
18
19
	pid_t		last_pid;
	int		depth;
20
	int		ignore;
21
	unsigned long	enter_funcs[FTRACE_RETFUNC_DEPTH];
22
23
24
25
26
27
28
29
30
31
};

struct fgraph_data {
	struct fgraph_cpu_data		*cpu_data;

	/* Place to preserve last processed entry. */
	struct ftrace_graph_ent_entry	ent;
	struct ftrace_graph_ret_entry	ret;
	int				failed;
	int				cpu;
32
33
};

34
#define TRACE_GRAPH_INDENT	2
35

36
/* Flag options */
37
#define TRACE_GRAPH_PRINT_OVERRUN	0x1
38
39
#define TRACE_GRAPH_PRINT_CPU		0x2
#define TRACE_GRAPH_PRINT_OVERHEAD	0x4
40
#define TRACE_GRAPH_PRINT_PROC		0x8
41
42
#define TRACE_GRAPH_PRINT_DURATION	0x10
#define TRACE_GRAPH_PRINT_ABS_TIME	0X20
43

44
static struct tracer_opt trace_opts[] = {
45
	/* Display overruns? (for self-debug purpose) */
46
47
48
49
50
	{ TRACER_OPT(funcgraph-overrun, TRACE_GRAPH_PRINT_OVERRUN) },
	/* Display CPU ? */
	{ TRACER_OPT(funcgraph-cpu, TRACE_GRAPH_PRINT_CPU) },
	/* Display Overhead ? */
	{ TRACER_OPT(funcgraph-overhead, TRACE_GRAPH_PRINT_OVERHEAD) },
51
52
	/* Display proc name/pid */
	{ TRACER_OPT(funcgraph-proc, TRACE_GRAPH_PRINT_PROC) },
53
54
55
56
	/* Display duration of execution */
	{ TRACER_OPT(funcgraph-duration, TRACE_GRAPH_PRINT_DURATION) },
	/* Display absolute time of an entry */
	{ TRACER_OPT(funcgraph-abstime, TRACE_GRAPH_PRINT_ABS_TIME) },
57
58
59
60
	{ } /* Empty entry */
};

static struct tracer_flags tracer_flags = {
61
	/* Don't display overruns and proc by default */
62
63
	.val = TRACE_GRAPH_PRINT_CPU | TRACE_GRAPH_PRINT_OVERHEAD |
	       TRACE_GRAPH_PRINT_DURATION,
64
65
66
	.opts = trace_opts
};

67
static struct trace_array *graph_array;
68

69

70
71
/* Add a function return address to the trace stack on thread info.*/
int
72
73
ftrace_push_return_trace(unsigned long ret, unsigned long func, int *depth,
			 unsigned long frame_pointer)
74
{
75
	unsigned long long calltime;
76
77
78
79
80
	int index;

	if (!current->ret_stack)
		return -EBUSY;

81
82
83
84
85
86
	/*
	 * We must make sure the ret_stack is tested before we read
	 * anything else.
	 */
	smp_rmb();

87
88
89
90
91
92
	/* The return trace stack is full */
	if (current->curr_ret_stack == FTRACE_RETFUNC_DEPTH - 1) {
		atomic_inc(&current->trace_overrun);
		return -EBUSY;
	}

93
94
	calltime = trace_clock_local();

95
96
97
98
	index = ++current->curr_ret_stack;
	barrier();
	current->ret_stack[index].ret = ret;
	current->ret_stack[index].func = func;
99
	current->ret_stack[index].calltime = calltime;
100
	current->ret_stack[index].subtime = 0;
101
	current->ret_stack[index].fp = frame_pointer;
102
103
104
105
106
107
	*depth = index;

	return 0;
}

/* Retrieve a function return address to the trace stack on thread info.*/
108
static void
109
110
ftrace_pop_return_trace(struct ftrace_graph_ret *trace, unsigned long *ret,
			unsigned long frame_pointer)
111
112
113
114
115
116
117
118
119
120
121
122
123
{
	int index;

	index = current->curr_ret_stack;

	if (unlikely(index < 0)) {
		ftrace_graph_stop();
		WARN_ON(1);
		/* Might as well panic, otherwise we have no where to go */
		*ret = (unsigned long)panic;
		return;
	}

124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
#ifdef CONFIG_HAVE_FUNCTION_GRAPH_FP_TEST
	/*
	 * The arch may choose to record the frame pointer used
	 * and check it here to make sure that it is what we expect it
	 * to be. If gcc does not set the place holder of the return
	 * address in the frame pointer, and does a copy instead, then
	 * the function graph trace will fail. This test detects this
	 * case.
	 *
	 * Currently, x86_32 with optimize for size (-Os) makes the latest
	 * gcc do the above.
	 */
	if (unlikely(current->ret_stack[index].fp != frame_pointer)) {
		ftrace_graph_stop();
		WARN(1, "Bad frame pointer: expected %lx, received %lx\n"
139
		     "  from func %ps return to %lx\n",
140
141
142
143
144
145
146
147
148
		     current->ret_stack[index].fp,
		     frame_pointer,
		     (void *)current->ret_stack[index].func,
		     current->ret_stack[index].ret);
		*ret = (unsigned long)panic;
		return;
	}
#endif

149
150
151
152
153
154
155
156
157
158
159
	*ret = current->ret_stack[index].ret;
	trace->func = current->ret_stack[index].func;
	trace->calltime = current->ret_stack[index].calltime;
	trace->overrun = atomic_read(&current->trace_overrun);
	trace->depth = index;
}

/*
 * Send the trace to the ring-buffer.
 * @return the original return address.
 */
160
unsigned long ftrace_return_to_handler(unsigned long frame_pointer)
161
162
163
164
{
	struct ftrace_graph_ret trace;
	unsigned long ret;

165
	ftrace_pop_return_trace(&trace, &ret, frame_pointer);
166
	trace.rettime = trace_clock_local();
167
	ftrace_graph_return(&trace);
168
169
	barrier();
	current->curr_ret_stack--;
170
171
172
173
174
175
176
177
178
179
180

	if (unlikely(!ret)) {
		ftrace_graph_stop();
		WARN_ON(1);
		/* Might as well panic. What else to do? */
		ret = (unsigned long)panic;
	}

	return ret;
}

181
182
183
184
185
186
187
static int __trace_graph_entry(struct trace_array *tr,
				struct ftrace_graph_ent *trace,
				unsigned long flags,
				int pc)
{
	struct ftrace_event_call *call = &event_funcgraph_entry;
	struct ring_buffer_event *event;
188
	struct ring_buffer *buffer = tr->buffer;
189
190
	struct ftrace_graph_ent_entry *entry;

191
	if (unlikely(__this_cpu_read(per_cpu_var(ftrace_cpu_disabled))))
192
193
		return 0;

194
	event = trace_buffer_lock_reserve(buffer, TRACE_GRAPH_ENT,
195
196
197
198
199
					  sizeof(*entry), flags, pc);
	if (!event)
		return 0;
	entry	= ring_buffer_event_data(event);
	entry->graph_ent			= *trace;
200
201
	if (!filter_current_check_discard(buffer, call, entry, event))
		ring_buffer_unlock_commit(buffer, event);
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218

	return 1;
}

int trace_graph_entry(struct ftrace_graph_ent *trace)
{
	struct trace_array *tr = graph_array;
	struct trace_array_cpu *data;
	unsigned long flags;
	long disabled;
	int ret;
	int cpu;
	int pc;

	if (!ftrace_trace_task(current))
		return 0;

219
220
	/* trace it when it is-nested-in or is a function enabled. */
	if (!(trace->depth || ftrace_graph_addr(trace->func)))
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
		return 0;

	local_irq_save(flags);
	cpu = raw_smp_processor_id();
	data = tr->data[cpu];
	disabled = atomic_inc_return(&data->disabled);
	if (likely(disabled == 1)) {
		pc = preempt_count();
		ret = __trace_graph_entry(tr, trace, flags, pc);
	} else {
		ret = 0;
	}

	atomic_dec(&data->disabled);
	local_irq_restore(flags);

	return ret;
}

static void __trace_graph_return(struct trace_array *tr,
				struct ftrace_graph_ret *trace,
				unsigned long flags,
				int pc)
{
	struct ftrace_event_call *call = &event_funcgraph_exit;
	struct ring_buffer_event *event;
247
	struct ring_buffer *buffer = tr->buffer;
248
249
	struct ftrace_graph_ret_entry *entry;

250
	if (unlikely(__this_cpu_read(per_cpu_var(ftrace_cpu_disabled))))
251
252
		return;

253
	event = trace_buffer_lock_reserve(buffer, TRACE_GRAPH_RET,
254
255
256
257
258
					  sizeof(*entry), flags, pc);
	if (!event)
		return;
	entry	= ring_buffer_event_data(event);
	entry->ret				= *trace;
259
260
	if (!filter_current_check_discard(buffer, call, entry, event))
		ring_buffer_unlock_commit(buffer, event);
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
}

void trace_graph_return(struct ftrace_graph_ret *trace)
{
	struct trace_array *tr = graph_array;
	struct trace_array_cpu *data;
	unsigned long flags;
	long disabled;
	int cpu;
	int pc;

	local_irq_save(flags);
	cpu = raw_smp_processor_id();
	data = tr->data[cpu];
	disabled = atomic_inc_return(&data->disabled);
	if (likely(disabled == 1)) {
		pc = preempt_count();
		__trace_graph_return(tr, trace, flags, pc);
	}
	atomic_dec(&data->disabled);
	local_irq_restore(flags);
}

284
285
286
287
288
289
290
291
292
void set_graph_array(struct trace_array *tr)
{
	graph_array = tr;

	/* Make graph_array visible before we start tracing */

	smp_mb();
}

293
294
static int graph_trace_init(struct trace_array *tr)
{
295
296
	int ret;

297
	set_graph_array(tr);
298
299
	ret = register_ftrace_graph(&trace_graph_return,
				    &trace_graph_entry);
300
301
302
303
304
	if (ret)
		return ret;
	tracing_start_cmdline_record();

	return 0;
305
306
307
308
}

static void graph_trace_reset(struct trace_array *tr)
{
309
310
	tracing_stop_cmdline_record();
	unregister_ftrace_graph();
311
312
}

313
static int max_bytes_for_cpu;
314
315
316
317
318
319

static enum print_line_t
print_graph_cpu(struct trace_seq *s, int cpu)
{
	int ret;

320
321
322
323
324
	/*
	 * Start with a space character - to make it stand out
	 * to the right a bit when trace output is pasted into
	 * email:
	 */
325
	ret = trace_seq_printf(s, " %*d) ", max_bytes_for_cpu, cpu);
326
	if (!ret)
327
328
		return TRACE_TYPE_PARTIAL_LINE;

329
330
331
	return TRACE_TYPE_HANDLED;
}

332
333
334
335
336
#define TRACE_GRAPH_PROCINFO_LENGTH	14

static enum print_line_t
print_graph_proc(struct trace_seq *s, pid_t pid)
{
337
	char comm[TASK_COMM_LEN];
338
339
	/* sign + log10(MAX_INT) + '\0' */
	char pid_str[11];
340
341
342
343
	int spaces = 0;
	int ret;
	int len;
	int i;
344

345
	trace_find_cmdline(pid, comm);
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
	comm[7] = '\0';
	sprintf(pid_str, "%d", pid);

	/* 1 stands for the "-" character */
	len = strlen(comm) + strlen(pid_str) + 1;

	if (len < TRACE_GRAPH_PROCINFO_LENGTH)
		spaces = TRACE_GRAPH_PROCINFO_LENGTH - len;

	/* First spaces to align center */
	for (i = 0; i < spaces / 2; i++) {
		ret = trace_seq_printf(s, " ");
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
	}

	ret = trace_seq_printf(s, "%s-%s", comm, pid_str);
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;

	/* Last spaces to align center */
	for (i = 0; i < spaces - (spaces / 2); i++) {
		ret = trace_seq_printf(s, " ");
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
	}
	return TRACE_TYPE_HANDLED;
}

375

376
377
378
static enum print_line_t
print_graph_lat_fmt(struct trace_seq *s, struct trace_entry *entry)
{
379
	if (!trace_seq_putc(s, ' '))
380
381
		return 0;

382
	return trace_print_lat_fmt(s, entry);
383
384
}

385
/* If the pid changed since the last trace, output this event */
386
static enum print_line_t
387
verif_pid(struct trace_seq *s, pid_t pid, int cpu, struct fgraph_data *data)
388
{
389
	pid_t prev_pid;
390
	pid_t *last_pid;
391
	int ret;
392

393
	if (!data)
394
395
		return TRACE_TYPE_HANDLED;

396
	last_pid = &(per_cpu_ptr(data->cpu_data, cpu)->last_pid);
397
398

	if (*last_pid == pid)
399
		return TRACE_TYPE_HANDLED;
400

401
402
	prev_pid = *last_pid;
	*last_pid = pid;
403

404
405
	if (prev_pid == -1)
		return TRACE_TYPE_HANDLED;
406
407
408
409
410
411
412
413
414
/*
 * Context-switch trace line:

 ------------------------------------------
 | 1)  migration/0--1  =>  sshd-1755
 ------------------------------------------

 */
	ret = trace_seq_printf(s,
415
		" ------------------------------------------\n");
416
	if (!ret)
417
		return TRACE_TYPE_PARTIAL_LINE;
418
419
420

	ret = print_graph_cpu(s, cpu);
	if (ret == TRACE_TYPE_PARTIAL_LINE)
421
		return TRACE_TYPE_PARTIAL_LINE;
422
423
424

	ret = print_graph_proc(s, prev_pid);
	if (ret == TRACE_TYPE_PARTIAL_LINE)
425
		return TRACE_TYPE_PARTIAL_LINE;
426
427
428

	ret = trace_seq_printf(s, " => ");
	if (!ret)
429
		return TRACE_TYPE_PARTIAL_LINE;
430
431
432

	ret = print_graph_proc(s, pid);
	if (ret == TRACE_TYPE_PARTIAL_LINE)
433
		return TRACE_TYPE_PARTIAL_LINE;
434
435
436
437

	ret = trace_seq_printf(s,
		"\n ------------------------------------------\n\n");
	if (!ret)
438
		return TRACE_TYPE_PARTIAL_LINE;
439

440
	return TRACE_TYPE_HANDLED;
441
442
}

443
444
static struct ftrace_graph_ret_entry *
get_return_for_leaf(struct trace_iterator *iter,
445
446
		struct ftrace_graph_ent_entry *curr)
{
447
448
	struct fgraph_data *data = iter->private;
	struct ring_buffer_iter *ring_iter = NULL;
449
450
451
	struct ring_buffer_event *event;
	struct ftrace_graph_ret_entry *next;

452
453
454
455
456
457
458
459
	/*
	 * If the previous output failed to write to the seq buffer,
	 * then we just reuse the data from before.
	 */
	if (data && data->failed) {
		curr = &data->ent;
		next = &data->ret;
	} else {
460

461
462
463
464
465
466
467
468
469
470
471
472
473
474
		ring_iter = iter->buffer_iter[iter->cpu];

		/* First peek to compare current entry and the next one */
		if (ring_iter)
			event = ring_buffer_iter_peek(ring_iter, NULL);
		else {
			/*
			 * We need to consume the current entry to see
			 * the next one.
			 */
			ring_buffer_consume(iter->tr->buffer, iter->cpu, NULL);
			event = ring_buffer_peek(iter->tr->buffer, iter->cpu,
						 NULL);
		}
475

476
477
478
479
		if (!event)
			return NULL;

		next = ring_buffer_event_data(event);
480

481
482
483
484
485
486
487
488
489
		if (data) {
			/*
			 * Save current and next entries for later reference
			 * if the output fails.
			 */
			data->ent = *curr;
			data->ret = *next;
		}
	}
490
491

	if (next->ent.type != TRACE_GRAPH_RET)
492
		return NULL;
493
494
495

	if (curr->ent.pid != next->ent.pid ||
			curr->graph_ent.func != next->ret.func)
496
		return NULL;
497

498
499
500
501
502
	/* this is a leaf, now advance the iterator */
	if (ring_iter)
		ring_buffer_read(ring_iter, NULL);

	return next;
503
504
}

505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
/* Signal a overhead of time execution to the output */
static int
print_graph_overhead(unsigned long long duration, struct trace_seq *s)
{
	/* If duration disappear, we don't need anything */
	if (!(tracer_flags.val & TRACE_GRAPH_PRINT_DURATION))
		return 1;

	/* Non nested entry or return */
	if (duration == -1)
		return trace_seq_printf(s, "  ");

	if (tracer_flags.val & TRACE_GRAPH_PRINT_OVERHEAD) {
		/* Duration exceeded 100 msecs */
		if (duration > 100000ULL)
			return trace_seq_printf(s, "! ");

		/* Duration exceeded 10 msecs */
		if (duration > 10000ULL)
			return trace_seq_printf(s, "+ ");
	}

	return trace_seq_printf(s, "  ");
}

530
531
532
533
534
535
536
537
538
539
540
static int print_graph_abs_time(u64 t, struct trace_seq *s)
{
	unsigned long usecs_rem;

	usecs_rem = do_div(t, NSEC_PER_SEC);
	usecs_rem /= 1000;

	return trace_seq_printf(s, "%5lu.%06lu |  ",
			(unsigned long)t, usecs_rem);
}

541
static enum print_line_t
542
print_graph_irq(struct trace_iterator *iter, unsigned long addr,
543
		enum trace_type type, int cpu, pid_t pid)
544
545
{
	int ret;
546
	struct trace_seq *s = &iter->seq;
547
548
549
550
551

	if (addr < (unsigned long)__irqentry_text_start ||
		addr >= (unsigned long)__irqentry_text_end)
		return TRACE_TYPE_UNHANDLED;

552
553
554
555
556
557
558
	/* Absolute time */
	if (tracer_flags.val & TRACE_GRAPH_PRINT_ABS_TIME) {
		ret = print_graph_abs_time(iter->ts, s);
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
	}

559
560
561
562
563
564
	/* Cpu */
	if (tracer_flags.val & TRACE_GRAPH_PRINT_CPU) {
		ret = print_graph_cpu(s, cpu);
		if (ret == TRACE_TYPE_PARTIAL_LINE)
			return TRACE_TYPE_PARTIAL_LINE;
	}
565

566
567
568
569
570
571
572
573
574
	/* Proc */
	if (tracer_flags.val & TRACE_GRAPH_PRINT_PROC) {
		ret = print_graph_proc(s, pid);
		if (ret == TRACE_TYPE_PARTIAL_LINE)
			return TRACE_TYPE_PARTIAL_LINE;
		ret = trace_seq_printf(s, " | ");
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
	}
575

576
577
578
579
	/* No overhead */
	ret = print_graph_overhead(-1, s);
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;
580

581
582
583
584
585
586
587
588
589
590
591
592
	if (type == TRACE_GRAPH_ENT)
		ret = trace_seq_printf(s, "==========>");
	else
		ret = trace_seq_printf(s, "<==========");

	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;

	/* Don't close the duration column if haven't one */
	if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION)
		trace_seq_printf(s, " |");
	ret = trace_seq_printf(s, "\n");
593
594
595
596
597

	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;
	return TRACE_TYPE_HANDLED;
}
598

599
600
enum print_line_t
trace_print_graph_duration(unsigned long long duration, struct trace_seq *s)
601
602
{
	unsigned long nsecs_rem = do_div(duration, 1000);
603
604
605
606
607
608
609
610
611
	/* log10(ULONG_MAX) + '\0' */
	char msecs_str[21];
	char nsecs_str[5];
	int ret, len;
	int i;

	sprintf(msecs_str, "%lu", (unsigned long) duration);

	/* Print msecs */
612
	ret = trace_seq_printf(s, "%s", msecs_str);
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;

	len = strlen(msecs_str);

	/* Print nsecs (we don't want to exceed 7 numbers) */
	if (len < 7) {
		snprintf(nsecs_str, 8 - len, "%03lu", nsecs_rem);
		ret = trace_seq_printf(s, ".%s", nsecs_str);
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
		len += strlen(nsecs_str);
	}

	ret = trace_seq_printf(s, " us ");
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;

	/* Print remaining spaces to fit the row's width */
	for (i = len; i < 7; i++) {
		ret = trace_seq_printf(s, " ");
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
	}
637
638
639
640
641
642
643
644
645
646
647
	return TRACE_TYPE_HANDLED;
}

static enum print_line_t
print_graph_duration(unsigned long long duration, struct trace_seq *s)
{
	int ret;

	ret = trace_print_graph_duration(duration, s);
	if (ret != TRACE_TYPE_HANDLED)
		return ret;
648
649
650
651
652

	ret = trace_seq_printf(s, "|  ");
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;

653
	return TRACE_TYPE_HANDLED;
654
655
656
}

/* Case of a leaf function on its call entry */
657
static enum print_line_t
658
print_graph_entry_leaf(struct trace_iterator *iter,
659
660
		struct ftrace_graph_ent_entry *entry,
		struct ftrace_graph_ret_entry *ret_entry, struct trace_seq *s)
661
{
662
	struct fgraph_data *data = iter->private;
663
664
665
	struct ftrace_graph_ret *graph_ret;
	struct ftrace_graph_ent *call;
	unsigned long long duration;
666
	int ret;
667
	int i;
668

669
670
671
672
	graph_ret = &ret_entry->ret;
	call = &entry->graph_ent;
	duration = graph_ret->rettime - graph_ret->calltime;

673
	if (data) {
674
		struct fgraph_cpu_data *cpu_data;
675
		int cpu = iter->cpu;
676
677

		cpu_data = per_cpu_ptr(data->cpu_data, cpu);
678
679
680
681
682
683

		/*
		 * Comments display at + 1 to depth. Since
		 * this is a leaf function, keep the comments
		 * equal to this depth.
		 */
684
685
686
687
688
		cpu_data->depth = call->depth - 1;

		/* No need to keep this function around for this depth */
		if (call->depth < FTRACE_RETFUNC_DEPTH)
			cpu_data->enter_funcs[call->depth] = 0;
689
690
	}

691
	/* Overhead */
692
693
694
	ret = print_graph_overhead(duration, s);
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;
695
696

	/* Duration */
697
698
699
700
701
	if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION) {
		ret = print_graph_duration(duration, s);
		if (ret == TRACE_TYPE_PARTIAL_LINE)
			return TRACE_TYPE_PARTIAL_LINE;
	}
702

703
704
705
706
707
708
709
	/* Function */
	for (i = 0; i < call->depth * TRACE_GRAPH_INDENT; i++) {
		ret = trace_seq_printf(s, " ");
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
	}

710
	ret = trace_seq_printf(s, "%ps();\n", (void *)call->func);
711
712
713
714
715
716
717
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;

	return TRACE_TYPE_HANDLED;
}

static enum print_line_t
718
719
720
print_graph_entry_nested(struct trace_iterator *iter,
			 struct ftrace_graph_ent_entry *entry,
			 struct trace_seq *s, int cpu)
721
722
{
	struct ftrace_graph_ent *call = &entry->graph_ent;
723
724
725
726
727
	struct fgraph_data *data = iter->private;
	int ret;
	int i;

	if (data) {
728
		struct fgraph_cpu_data *cpu_data;
729
730
		int cpu = iter->cpu;

731
732
733
734
735
736
		cpu_data = per_cpu_ptr(data->cpu_data, cpu);
		cpu_data->depth = call->depth;

		/* Save this function pointer to see if the exit matches */
		if (call->depth < FTRACE_RETFUNC_DEPTH)
			cpu_data->enter_funcs[call->depth] = call->func;
737
	}
738
739

	/* No overhead */
740
741
742
	ret = print_graph_overhead(-1, s);
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;
743

744
745
	/* No time */
	if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION) {
746
747
748
749
750
		ret = trace_seq_printf(s, "            |  ");
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
	}

751
	/* Function */
752
753
	for (i = 0; i < call->depth * TRACE_GRAPH_INDENT; i++) {
		ret = trace_seq_printf(s, " ");
754
755
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
756
757
	}

758
	ret = trace_seq_printf(s, "%ps() {\n", (void *)call->func);
759
760
761
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;

762
763
764
765
766
	/*
	 * we already consumed the current entry to check the next one
	 * and see if this is a leaf.
	 */
	return TRACE_TYPE_NO_CONSUME;
767
768
}

769
static enum print_line_t
770
771
print_graph_prologue(struct trace_iterator *iter, struct trace_seq *s,
		     int type, unsigned long addr)
772
{
773
	struct fgraph_data *data = iter->private;
774
	struct trace_entry *ent = iter->ent;
775
776
	int cpu = iter->cpu;
	int ret;
777

778
	/* Pid */
779
	if (verif_pid(s, ent->pid, cpu, data) == TRACE_TYPE_PARTIAL_LINE)
780
781
		return TRACE_TYPE_PARTIAL_LINE;

782
783
784
785
786
787
	if (type) {
		/* Interrupt */
		ret = print_graph_irq(iter, addr, type, cpu, ent->pid);
		if (ret == TRACE_TYPE_PARTIAL_LINE)
			return TRACE_TYPE_PARTIAL_LINE;
	}
788

789
790
791
792
793
794
795
	/* Absolute time */
	if (tracer_flags.val & TRACE_GRAPH_PRINT_ABS_TIME) {
		ret = print_graph_abs_time(iter->ts, s);
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
	}

796
797
798
	/* Cpu */
	if (tracer_flags.val & TRACE_GRAPH_PRINT_CPU) {
		ret = print_graph_cpu(s, cpu);
799
800
801
802
803
804
		if (ret == TRACE_TYPE_PARTIAL_LINE)
			return TRACE_TYPE_PARTIAL_LINE;
	}

	/* Proc */
	if (tracer_flags.val & TRACE_GRAPH_PRINT_PROC) {
805
		ret = print_graph_proc(s, ent->pid);
806
807
808
809
		if (ret == TRACE_TYPE_PARTIAL_LINE)
			return TRACE_TYPE_PARTIAL_LINE;

		ret = trace_seq_printf(s, " | ");
810
811
812
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
	}
813

814
815
816
817
818
819
820
	/* Latency format */
	if (trace_flags & TRACE_ITER_LATENCY_FMT) {
		ret = print_graph_lat_fmt(s, ent);
		if (ret == TRACE_TYPE_PARTIAL_LINE)
			return TRACE_TYPE_PARTIAL_LINE;
	}

821
822
823
824
825
826
827
	return 0;
}

static enum print_line_t
print_graph_entry(struct ftrace_graph_ent_entry *field, struct trace_seq *s,
			struct trace_iterator *iter)
{
828
	struct fgraph_data *data = iter->private;
829
830
	struct ftrace_graph_ent *call = &field->graph_ent;
	struct ftrace_graph_ret_entry *leaf_ret;
831
832
	static enum print_line_t ret;
	int cpu = iter->cpu;
833
834
835
836

	if (print_graph_prologue(iter, s, TRACE_GRAPH_ENT, call->func))
		return TRACE_TYPE_PARTIAL_LINE;

837
838
	leaf_ret = get_return_for_leaf(iter, field);
	if (leaf_ret)
839
		ret = print_graph_entry_leaf(iter, field, leaf_ret, s);
840
	else
841
		ret = print_graph_entry_nested(iter, field, s, cpu);
842

843
844
845
846
847
848
849
850
851
852
853
854
855
	if (data) {
		/*
		 * If we failed to write our output, then we need to make
		 * note of it. Because we already consumed our entry.
		 */
		if (s->full) {
			data->failed = 1;
			data->cpu = cpu;
		} else
			data->failed = 0;
	}

	return ret;
856
857
}

858
859
static enum print_line_t
print_graph_return(struct ftrace_graph_ret *trace, struct trace_seq *s,
860
		   struct trace_entry *ent, struct trace_iterator *iter)
861
{
862
	unsigned long long duration = trace->rettime - trace->calltime;
863
864
865
	struct fgraph_data *data = iter->private;
	pid_t pid = ent->pid;
	int cpu = iter->cpu;
866
	int func_match = 1;
867
868
869
870
	int ret;
	int i;

	if (data) {
871
872
873
874
		struct fgraph_cpu_data *cpu_data;
		int cpu = iter->cpu;

		cpu_data = per_cpu_ptr(data->cpu_data, cpu);
875
876
877
878
879
880

		/*
		 * Comments display at + 1 to depth. This is the
		 * return from a function, we now want the comments
		 * to display at the same level of the bracket.
		 */
881
882
883
884
885
886
887
		cpu_data->depth = trace->depth - 1;

		if (trace->depth < FTRACE_RETFUNC_DEPTH) {
			if (cpu_data->enter_funcs[trace->depth] != trace->func)
				func_match = 0;
			cpu_data->enter_funcs[trace->depth] = 0;
		}
888
	}
889

890
	if (print_graph_prologue(iter, s, 0, 0))
891
892
		return TRACE_TYPE_PARTIAL_LINE;

893
	/* Overhead */
894
895
896
	ret = print_graph_overhead(duration, s);
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;
897
898

	/* Duration */
899
900
901
902
903
	if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION) {
		ret = print_graph_duration(duration, s);
		if (ret == TRACE_TYPE_PARTIAL_LINE)
			return TRACE_TYPE_PARTIAL_LINE;
	}
904
905

	/* Closing brace */
906
907
	for (i = 0; i < trace->depth * TRACE_GRAPH_INDENT; i++) {
		ret = trace_seq_printf(s, " ");
908
909
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
910
911
	}

912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
	/*
	 * If the return function does not have a matching entry,
	 * then the entry was lost. Instead of just printing
	 * the '}' and letting the user guess what function this
	 * belongs to, write out the function name.
	 */
	if (func_match) {
		ret = trace_seq_printf(s, "}\n");
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
	} else {
		ret = trace_seq_printf(s, "} (%ps)\n", (void *)trace->func);
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
	}
927

928
	/* Overrun */
929
930
931
	if (tracer_flags.val & TRACE_GRAPH_PRINT_OVERRUN) {
		ret = trace_seq_printf(s, " (Overruns: %lu)\n",
					trace->overrun);
932
933
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
934
	}
935

936
	ret = print_graph_irq(iter, trace->func, TRACE_GRAPH_RET, cpu, pid);
937
938
939
	if (ret == TRACE_TYPE_PARTIAL_LINE)
		return TRACE_TYPE_PARTIAL_LINE;

940
941
942
	return TRACE_TYPE_HANDLED;
}

943
static enum print_line_t
944
945
print_graph_comment(struct trace_seq *s,  struct trace_entry *ent,
		    struct trace_iterator *iter)
946
{
947
	unsigned long sym_flags = (trace_flags & TRACE_ITER_SYM_MASK);
948
	struct fgraph_data *data = iter->private;
949
	struct trace_event *event;
950
	int depth = 0;
951
	int ret;
952
953
954
	int i;

	if (data)
955
		depth = per_cpu_ptr(data->cpu_data, iter->cpu)->depth;
956

957
	if (print_graph_prologue(iter, s, 0, 0))
958
959
		return TRACE_TYPE_PARTIAL_LINE;

960
	/* No overhead */
961
962
963
964
965
966
967
	ret = print_graph_overhead(-1, s);
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;

	/* No time */
	if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION) {
		ret = trace_seq_printf(s, "            |  ");
968
969
970
971
972
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
	}

	/* Indentation */
973
974
	if (depth > 0)
		for (i = 0; i < (depth + 1) * TRACE_GRAPH_INDENT; i++) {
975
976
977
978
979
980
			ret = trace_seq_printf(s, " ");
			if (!ret)
				return TRACE_TYPE_PARTIAL_LINE;
		}

	/* The comment */
981
982
983
984
	ret = trace_seq_printf(s, "/* ");
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;

985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
	switch (iter->ent->type) {
	case TRACE_BPRINT:
		ret = trace_print_bprintk_msg_only(iter);
		if (ret != TRACE_TYPE_HANDLED)
			return ret;
		break;
	case TRACE_PRINT:
		ret = trace_print_printk_msg_only(iter);
		if (ret != TRACE_TYPE_HANDLED)
			return ret;
		break;
	default:
		event = ftrace_find_event(ent->type);
		if (!event)
			return TRACE_TYPE_UNHANDLED;

		ret = event->trace(iter, sym_flags);
		if (ret != TRACE_TYPE_HANDLED)
			return ret;
	}
1005

1006
1007
1008
1009
1010
1011
	/* Strip ending newline */
	if (s->buffer[s->len - 1] == '\n') {
		s->buffer[s->len - 1] = '\0';
		s->len--;
	}

1012
1013
1014
1015
1016
1017
1018
1019
	ret = trace_seq_printf(s, " */\n");
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;

	return TRACE_TYPE_HANDLED;
}


1020
1021
1022
enum print_line_t
print_graph_function(struct trace_iterator *iter)
{
1023
1024
	struct ftrace_graph_ent_entry *field;
	struct fgraph_data *data = iter->private;
1025
	struct trace_entry *entry = iter->ent;
1026
	struct trace_seq *s = &iter->seq;
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
	int cpu = iter->cpu;
	int ret;

	if (data && per_cpu_ptr(data->cpu_data, cpu)->ignore) {
		per_cpu_ptr(data->cpu_data, cpu)->ignore = 0;
		return TRACE_TYPE_HANDLED;
	}

	/*
	 * If the last output failed, there's a possibility we need
	 * to print out the missing entry which would never go out.
	 */
	if (data && data->failed) {
		field = &data->ent;
		iter->cpu = data->cpu;
		ret = print_graph_entry(field, s, iter);
		if (ret == TRACE_TYPE_HANDLED && iter->cpu != cpu) {
			per_cpu_ptr(data->cpu_data, iter->cpu)->ignore = 1;
			ret = TRACE_TYPE_NO_CONSUME;
		}
		iter->cpu = cpu;
		return ret;
	}
1050

1051
1052
	switch (entry->type) {
	case TRACE_GRAPH_ENT: {
1053
1054
1055
1056
1057
1058
		/*
		 * print_graph_entry() may consume the current event,
		 * thus @field may become invalid, so we need to save it.
		 * sizeof(struct ftrace_graph_ent_entry) is very small,
		 * it can be safely saved at the stack.
		 */
1059
		struct ftrace_graph_ent_entry saved;
1060
		trace_assign_type(field, entry);
1061
1062
		saved = *field;
		return print_graph_entry(&saved, s, iter);
1063
1064
1065
1066
	}
	case TRACE_GRAPH_RET: {
		struct ftrace_graph_ret_entry *field;
		trace_assign_type(field, entry);
1067
		return print_graph_return(&field->ret, s, entry, iter);
1068
1069
	}
	default:
1070
		return print_graph_comment(s, entry, iter);
1071
	}
1072
1073

	return TRACE_TYPE_HANDLED;
1074
1075
}

1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
static void print_lat_header(struct seq_file *s)
{
	static const char spaces[] = "                "	/* 16 spaces */
		"    "					/* 4 spaces */
		"                 ";			/* 17 spaces */
	int size = 0;

	if (tracer_flags.val & TRACE_GRAPH_PRINT_ABS_TIME)
		size += 16;
	if (tracer_flags.val & TRACE_GRAPH_PRINT_CPU)
		size += 4;
	if (tracer_flags.val & TRACE_GRAPH_PRINT_PROC)
		size += 17;

	seq_printf(s, "#%.*s  _-----=> irqs-off        \n", size, spaces);
	seq_printf(s, "#%.*s / _----=> need-resched    \n", size, spaces);
	seq_printf(s, "#%.*s| / _---=> hardirq/softirq \n", size, spaces);
	seq_printf(s, "#%.*s|| / _--=> preempt-depth   \n", size, spaces);
1094
1095
	seq_printf(s, "#%.*s||| / _-=> lock-depth      \n", size, spaces);
	seq_printf(s, "#%.*s|||| /                     \n", size, spaces);
1096
1097
}

1098
1099
static void print_graph_headers(struct seq_file *s)
{
1100
1101
1102
1103
1104
	int lat = trace_flags & TRACE_ITER_LATENCY_FMT;

	if (lat)
		print_lat_header(s);

1105
	/* 1st line */
1106
	seq_printf(s, "#");
1107
1108
	if (tracer_flags.val & TRACE_GRAPH_PRINT_ABS_TIME)
		seq_printf(s, "     TIME       ");
1109
	if (tracer_flags.val & TRACE_GRAPH_PRINT_CPU)
1110
		seq_printf(s, " CPU");
1111
	if (tracer_flags.val & TRACE_GRAPH_PRINT_PROC)
1112
1113
		seq_printf(s, "  TASK/PID       ");
	if (lat)
1114
		seq_printf(s, "|||||");
1115
1116
1117
	if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION)
		seq_printf(s, "  DURATION   ");
	seq_printf(s, "               FUNCTION CALLS\n");
1118
1119

	/* 2nd line */
1120
	seq_printf(s, "#");
1121
1122
	if (tracer_flags.val & TRACE_GRAPH_PRINT_ABS_TIME)
		seq_printf(s, "      |         ");
1123
	if (tracer_flags.val & TRACE_GRAPH_PRINT_CPU)
1124
		seq_printf(s, " |  ");
1125
	if (tracer_flags.val & TRACE_GRAPH_PRINT_PROC)
1126
1127
		seq_printf(s, "   |    |        ");
	if (lat)
1128
		seq_printf(s, "|||||");
1129
1130
1131
	if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION)
		seq_printf(s, "   |   |      ");
	seq_printf(s, "               |   |   |   |\n");
1132
}
1133
1134
1135

static void graph_trace_open(struct trace_iterator *iter)
{
1136
	/* pid and depth on the last trace processed */
1137
	struct fgraph_data *data;
1138
1139
	int cpu;

1140
1141
1142
	iter->private = NULL;

	data = kzalloc(sizeof(*data), GFP_KERNEL);
1143
	if (!data)
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
		goto out_err;

	data->cpu_data = alloc_percpu(struct fgraph_cpu_data);
	if (!data->cpu_data)
		goto out_err_free;

	for_each_possible_cpu(cpu) {
		pid_t *pid = &(per_cpu_ptr(data->cpu_data, cpu)->last_pid);
		int *depth = &(per_cpu_ptr(data->cpu_data, cpu)->depth);
		int *ignore = &(per_cpu_ptr(data->cpu_data, cpu)->ignore);
		*pid = -1;
		*depth = 0;
		*ignore = 0;
	}
1158

1159
	iter->private = data;
1160
1161
1162
1163
1164
1165
1166

	return;

 out_err_free:
	kfree(data);
 out_err:
	pr_warning("function graph tracer: not enough memory\n");
1167
1168
1169
1170
}

static void graph_trace_close(struct trace_iterator *iter)
{
1171
1172
1173
1174
1175
1176
	struct fgraph_data *data = iter->private;

	if (data) {
		free_percpu(data->cpu_data);
		kfree(data);
	}
1177
1178
}

1179
static struct tracer graph_trace __read_mostly = {
1180
	.name		= "function_graph",
1181
	.open		= graph_trace_open,
1182
	.pipe_open	= graph_trace_open,
1183
	.close		= graph_trace_close,
1184
	.pipe_close	= graph_trace_close,
1185
	.wait_pipe	= poll_wait_pipe,
1186
1187
	.init		= graph_trace_init,
	.reset		= graph_trace_reset,
1188
1189
	.print_line	= print_graph_function,
	.print_header	= print_graph_headers,
1190
	.flags		= &tracer_flags,
1191
1192
1193
#ifdef CONFIG_FTRACE_SELFTEST
	.selftest	= trace_selftest_startup_function_graph,
#endif
1194
1195
1196
1197
};

static __init int init_graph_trace(void)
{
1198
1199
	max_bytes_for_cpu = snprintf(NULL, 0, "%d", nr_cpu_ids - 1);

1200
1201
1202
1203
	return register_tracer(&graph_trace);
}

device_initcall(init_graph_trace);