1 /*
2  *
3  * Function graph tracer.
4  * Copyright (c) 2008-2009 Frederic Weisbecker <[email protected]>
5  * Mostly borrowed from function tracer which
6  * is Copyright (c) Steven Rostedt <[email protected]>
7  *
8  */
9 #include <linux/debugfs.h>
10 #include <linux/uaccess.h>
11 #include <linux/ftrace.h>
12 #include <linux/fs.h>
13 
14 #include "trace.h"
15 #include "trace_output.h"
16 
17 struct fgraph_cpu_data {
18 	pid_t		last_pid;
19 	int		depth;
20 	int		ignore;
21 	unsigned long	enter_funcs[FTRACE_RETFUNC_DEPTH];
22 };
23 
24 struct fgraph_data {
25 	struct fgraph_cpu_data		*cpu_data;
26 
27 	/* Place to preserve last processed entry. */
28 	struct ftrace_graph_ent_entry	ent;
29 	struct ftrace_graph_ret_entry	ret;
30 	int				failed;
31 	int				cpu;
32 };
33 
34 #define TRACE_GRAPH_INDENT	2
35 
36 /* Flag options */
37 #define TRACE_GRAPH_PRINT_OVERRUN	0x1
38 #define TRACE_GRAPH_PRINT_CPU		0x2
39 #define TRACE_GRAPH_PRINT_OVERHEAD	0x4
40 #define TRACE_GRAPH_PRINT_PROC		0x8
41 #define TRACE_GRAPH_PRINT_DURATION	0x10
42 #define TRACE_GRAPH_PRINT_ABS_TIME	0X20
43 
44 static struct tracer_opt trace_opts[] = {
45 	/* Display overruns? (for self-debug purpose) */
46 	{ TRACER_OPT(funcgraph-overrun, TRACE_GRAPH_PRINT_OVERRUN) },
47 	/* Display CPU ? */
48 	{ TRACER_OPT(funcgraph-cpu, TRACE_GRAPH_PRINT_CPU) },
49 	/* Display Overhead ? */
50 	{ TRACER_OPT(funcgraph-overhead, TRACE_GRAPH_PRINT_OVERHEAD) },
51 	/* Display proc name/pid */
52 	{ TRACER_OPT(funcgraph-proc, TRACE_GRAPH_PRINT_PROC) },
53 	/* Display duration of execution */
54 	{ TRACER_OPT(funcgraph-duration, TRACE_GRAPH_PRINT_DURATION) },
55 	/* Display absolute time of an entry */
56 	{ TRACER_OPT(funcgraph-abstime, TRACE_GRAPH_PRINT_ABS_TIME) },
57 	{ } /* Empty entry */
58 };
59 
60 static struct tracer_flags tracer_flags = {
61 	/* Don't display overruns and proc by default */
62 	.val = TRACE_GRAPH_PRINT_CPU | TRACE_GRAPH_PRINT_OVERHEAD |
63 	       TRACE_GRAPH_PRINT_DURATION,
64 	.opts = trace_opts
65 };
66 
67 static struct trace_array *graph_array;
68 
69 
70 /* Add a function return address to the trace stack on thread info.*/
71 int
72 ftrace_push_return_trace(unsigned long ret, unsigned long func, int *depth,
73 			 unsigned long frame_pointer)
74 {
75 	unsigned long long calltime;
76 	int index;
77 
78 	if (!current->ret_stack)
79 		return -EBUSY;
80 
81 	/*
82 	 * We must make sure the ret_stack is tested before we read
83 	 * anything else.
84 	 */
85 	smp_rmb();
86 
87 	/* The return trace stack is full */
88 	if (current->curr_ret_stack == FTRACE_RETFUNC_DEPTH - 1) {
89 		atomic_inc(&current->trace_overrun);
90 		return -EBUSY;
91 	}
92 
93 	calltime = trace_clock_local();
94 
95 	index = ++current->curr_ret_stack;
96 	barrier();
97 	current->ret_stack[index].ret = ret;
98 	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 	*depth = index;
103 
104 	return 0;
105 }
106 
107 /* Retrieve a function return address to the trace stack on thread info.*/
108 static void
109 ftrace_pop_return_trace(struct ftrace_graph_ret *trace, unsigned long *ret,
110 			unsigned long frame_pointer)
111 {
112 	int index;
113 
114 	index = current->curr_ret_stack;
115 
116 	if (unlikely(index < 0)) {
117 		ftrace_graph_stop();
118 		WARN_ON(1);
119 		/* Might as well panic, otherwise we have no where to go */
120 		*ret = (unsigned long)panic;
121 		return;
122 	}
123 
124 #ifdef CONFIG_HAVE_FUNCTION_GRAPH_FP_TEST
125 	/*
126 	 * The arch may choose to record the frame pointer used
127 	 * and check it here to make sure that it is what we expect it
128 	 * to be. If gcc does not set the place holder of the return
129 	 * address in the frame pointer, and does a copy instead, then
130 	 * the function graph trace will fail. This test detects this
131 	 * case.
132 	 *
133 	 * Currently, x86_32 with optimize for size (-Os) makes the latest
134 	 * gcc do the above.
135 	 */
136 	if (unlikely(current->ret_stack[index].fp != frame_pointer)) {
137 		ftrace_graph_stop();
138 		WARN(1, "Bad frame pointer: expected %lx, received %lx\n"
139 		     "  from func %ps return to %lx\n",
140 		     current->ret_stack[index].fp,
141 		     frame_pointer,
142 		     (void *)current->ret_stack[index].func,
143 		     current->ret_stack[index].ret);
144 		*ret = (unsigned long)panic;
145 		return;
146 	}
147 #endif
148 
149 	*ret = current->ret_stack[index].ret;
150 	trace->func = current->ret_stack[index].func;
151 	trace->calltime = current->ret_stack[index].calltime;
152 	trace->overrun = atomic_read(&current->trace_overrun);
153 	trace->depth = index;
154 }
155 
156 /*
157  * Send the trace to the ring-buffer.
158  * @return the original return address.
159  */
160 unsigned long ftrace_return_to_handler(unsigned long frame_pointer)
161 {
162 	struct ftrace_graph_ret trace;
163 	unsigned long ret;
164 
165 	ftrace_pop_return_trace(&trace, &ret, frame_pointer);
166 	trace.rettime = trace_clock_local();
167 	ftrace_graph_return(&trace);
168 	barrier();
169 	current->curr_ret_stack--;
170 
171 	if (unlikely(!ret)) {
172 		ftrace_graph_stop();
173 		WARN_ON(1);
174 		/* Might as well panic. What else to do? */
175 		ret = (unsigned long)panic;
176 	}
177 
178 	return ret;
179 }
180 
181 static int __trace_graph_entry(struct trace_array *tr,
182 				struct ftrace_graph_ent *trace,
183 				unsigned long flags,
184 				int pc)
185 {
186 	struct ftrace_event_call *call = &event_funcgraph_entry;
187 	struct ring_buffer_event *event;
188 	struct ring_buffer *buffer = tr->buffer;
189 	struct ftrace_graph_ent_entry *entry;
190 
191 	if (unlikely(__this_cpu_read(ftrace_cpu_disabled)))
192 		return 0;
193 
194 	event = trace_buffer_lock_reserve(buffer, TRACE_GRAPH_ENT,
195 					  sizeof(*entry), flags, pc);
196 	if (!event)
197 		return 0;
198 	entry	= ring_buffer_event_data(event);
199 	entry->graph_ent			= *trace;
200 	if (!filter_current_check_discard(buffer, call, entry, event))
201 		ring_buffer_unlock_commit(buffer, event);
202 
203 	return 1;
204 }
205 
206 int trace_graph_entry(struct ftrace_graph_ent *trace)
207 {
208 	struct trace_array *tr = graph_array;
209 	struct trace_array_cpu *data;
210 	unsigned long flags;
211 	long disabled;
212 	int ret;
213 	int cpu;
214 	int pc;
215 
216 	if (!ftrace_trace_task(current))
217 		return 0;
218 
219 	/* trace it when it is-nested-in or is a function enabled. */
220 	if (!(trace->depth || ftrace_graph_addr(trace->func)))
221 		return 0;
222 
223 	local_irq_save(flags);
224 	cpu = raw_smp_processor_id();
225 	data = tr->data[cpu];
226 	disabled = atomic_inc_return(&data->disabled);
227 	if (likely(disabled == 1)) {
228 		pc = preempt_count();
229 		ret = __trace_graph_entry(tr, trace, flags, pc);
230 	} else {
231 		ret = 0;
232 	}
233 
234 	atomic_dec(&data->disabled);
235 	local_irq_restore(flags);
236 
237 	return ret;
238 }
239 
240 int trace_graph_thresh_entry(struct ftrace_graph_ent *trace)
241 {
242 	if (tracing_thresh)
243 		return 1;
244 	else
245 		return trace_graph_entry(trace);
246 }
247 
248 static void __trace_graph_return(struct trace_array *tr,
249 				struct ftrace_graph_ret *trace,
250 				unsigned long flags,
251 				int pc)
252 {
253 	struct ftrace_event_call *call = &event_funcgraph_exit;
254 	struct ring_buffer_event *event;
255 	struct ring_buffer *buffer = tr->buffer;
256 	struct ftrace_graph_ret_entry *entry;
257 
258 	if (unlikely(__this_cpu_read(ftrace_cpu_disabled)))
259 		return;
260 
261 	event = trace_buffer_lock_reserve(buffer, TRACE_GRAPH_RET,
262 					  sizeof(*entry), flags, pc);
263 	if (!event)
264 		return;
265 	entry	= ring_buffer_event_data(event);
266 	entry->ret				= *trace;
267 	if (!filter_current_check_discard(buffer, call, entry, event))
268 		ring_buffer_unlock_commit(buffer, event);
269 }
270 
271 void trace_graph_return(struct ftrace_graph_ret *trace)
272 {
273 	struct trace_array *tr = graph_array;
274 	struct trace_array_cpu *data;
275 	unsigned long flags;
276 	long disabled;
277 	int cpu;
278 	int pc;
279 
280 	local_irq_save(flags);
281 	cpu = raw_smp_processor_id();
282 	data = tr->data[cpu];
283 	disabled = atomic_inc_return(&data->disabled);
284 	if (likely(disabled == 1)) {
285 		pc = preempt_count();
286 		__trace_graph_return(tr, trace, flags, pc);
287 	}
288 	atomic_dec(&data->disabled);
289 	local_irq_restore(flags);
290 }
291 
292 void set_graph_array(struct trace_array *tr)
293 {
294 	graph_array = tr;
295 
296 	/* Make graph_array visible before we start tracing */
297 
298 	smp_mb();
299 }
300 
301 void trace_graph_thresh_return(struct ftrace_graph_ret *trace)
302 {
303 	if (tracing_thresh &&
304 	    (trace->rettime - trace->calltime < tracing_thresh))
305 		return;
306 	else
307 		trace_graph_return(trace);
308 }
309 
310 static int graph_trace_init(struct trace_array *tr)
311 {
312 	int ret;
313 
314 	set_graph_array(tr);
315 	if (tracing_thresh)
316 		ret = register_ftrace_graph(&trace_graph_thresh_return,
317 					    &trace_graph_thresh_entry);
318 	else
319 		ret = register_ftrace_graph(&trace_graph_return,
320 					    &trace_graph_entry);
321 	if (ret)
322 		return ret;
323 	tracing_start_cmdline_record();
324 
325 	return 0;
326 }
327 
328 static void graph_trace_reset(struct trace_array *tr)
329 {
330 	tracing_stop_cmdline_record();
331 	unregister_ftrace_graph();
332 }
333 
334 static int max_bytes_for_cpu;
335 
336 static enum print_line_t
337 print_graph_cpu(struct trace_seq *s, int cpu)
338 {
339 	int ret;
340 
341 	/*
342 	 * Start with a space character - to make it stand out
343 	 * to the right a bit when trace output is pasted into
344 	 * email:
345 	 */
346 	ret = trace_seq_printf(s, " %*d) ", max_bytes_for_cpu, cpu);
347 	if (!ret)
348 		return TRACE_TYPE_PARTIAL_LINE;
349 
350 	return TRACE_TYPE_HANDLED;
351 }
352 
353 #define TRACE_GRAPH_PROCINFO_LENGTH	14
354 
355 static enum print_line_t
356 print_graph_proc(struct trace_seq *s, pid_t pid)
357 {
358 	char comm[TASK_COMM_LEN];
359 	/* sign + log10(MAX_INT) + '\0' */
360 	char pid_str[11];
361 	int spaces = 0;
362 	int ret;
363 	int len;
364 	int i;
365 
366 	trace_find_cmdline(pid, comm);
367 	comm[7] = '\0';
368 	sprintf(pid_str, "%d", pid);
369 
370 	/* 1 stands for the "-" character */
371 	len = strlen(comm) + strlen(pid_str) + 1;
372 
373 	if (len < TRACE_GRAPH_PROCINFO_LENGTH)
374 		spaces = TRACE_GRAPH_PROCINFO_LENGTH - len;
375 
376 	/* First spaces to align center */
377 	for (i = 0; i < spaces / 2; i++) {
378 		ret = trace_seq_printf(s, " ");
379 		if (!ret)
380 			return TRACE_TYPE_PARTIAL_LINE;
381 	}
382 
383 	ret = trace_seq_printf(s, "%s-%s", comm, pid_str);
384 	if (!ret)
385 		return TRACE_TYPE_PARTIAL_LINE;
386 
387 	/* Last spaces to align center */
388 	for (i = 0; i < spaces - (spaces / 2); i++) {
389 		ret = trace_seq_printf(s, " ");
390 		if (!ret)
391 			return TRACE_TYPE_PARTIAL_LINE;
392 	}
393 	return TRACE_TYPE_HANDLED;
394 }
395 
396 
397 static enum print_line_t
398 print_graph_lat_fmt(struct trace_seq *s, struct trace_entry *entry)
399 {
400 	if (!trace_seq_putc(s, ' '))
401 		return 0;
402 
403 	return trace_print_lat_fmt(s, entry);
404 }
405 
406 /* If the pid changed since the last trace, output this event */
407 static enum print_line_t
408 verif_pid(struct trace_seq *s, pid_t pid, int cpu, struct fgraph_data *data)
409 {
410 	pid_t prev_pid;
411 	pid_t *last_pid;
412 	int ret;
413 
414 	if (!data)
415 		return TRACE_TYPE_HANDLED;
416 
417 	last_pid = &(per_cpu_ptr(data->cpu_data, cpu)->last_pid);
418 
419 	if (*last_pid == pid)
420 		return TRACE_TYPE_HANDLED;
421 
422 	prev_pid = *last_pid;
423 	*last_pid = pid;
424 
425 	if (prev_pid == -1)
426 		return TRACE_TYPE_HANDLED;
427 /*
428  * Context-switch trace line:
429 
430  ------------------------------------------
431  | 1)  migration/0--1  =>  sshd-1755
432  ------------------------------------------
433 
434  */
435 	ret = trace_seq_printf(s,
436 		" ------------------------------------------\n");
437 	if (!ret)
438 		return TRACE_TYPE_PARTIAL_LINE;
439 
440 	ret = print_graph_cpu(s, cpu);
441 	if (ret == TRACE_TYPE_PARTIAL_LINE)
442 		return TRACE_TYPE_PARTIAL_LINE;
443 
444 	ret = print_graph_proc(s, prev_pid);
445 	if (ret == TRACE_TYPE_PARTIAL_LINE)
446 		return TRACE_TYPE_PARTIAL_LINE;
447 
448 	ret = trace_seq_printf(s, " => ");
449 	if (!ret)
450 		return TRACE_TYPE_PARTIAL_LINE;
451 
452 	ret = print_graph_proc(s, pid);
453 	if (ret == TRACE_TYPE_PARTIAL_LINE)
454 		return TRACE_TYPE_PARTIAL_LINE;
455 
456 	ret = trace_seq_printf(s,
457 		"\n ------------------------------------------\n\n");
458 	if (!ret)
459 		return TRACE_TYPE_PARTIAL_LINE;
460 
461 	return TRACE_TYPE_HANDLED;
462 }
463 
464 static struct ftrace_graph_ret_entry *
465 get_return_for_leaf(struct trace_iterator *iter,
466 		struct ftrace_graph_ent_entry *curr)
467 {
468 	struct fgraph_data *data = iter->private;
469 	struct ring_buffer_iter *ring_iter = NULL;
470 	struct ring_buffer_event *event;
471 	struct ftrace_graph_ret_entry *next;
472 
473 	/*
474 	 * If the previous output failed to write to the seq buffer,
475 	 * then we just reuse the data from before.
476 	 */
477 	if (data && data->failed) {
478 		curr = &data->ent;
479 		next = &data->ret;
480 	} else {
481 
482 		ring_iter = iter->buffer_iter[iter->cpu];
483 
484 		/* First peek to compare current entry and the next one */
485 		if (ring_iter)
486 			event = ring_buffer_iter_peek(ring_iter, NULL);
487 		else {
488 			/*
489 			 * We need to consume the current entry to see
490 			 * the next one.
491 			 */
492 			ring_buffer_consume(iter->tr->buffer, iter->cpu, NULL);
493 			event = ring_buffer_peek(iter->tr->buffer, iter->cpu,
494 						 NULL);
495 		}
496 
497 		if (!event)
498 			return NULL;
499 
500 		next = ring_buffer_event_data(event);
501 
502 		if (data) {
503 			/*
504 			 * Save current and next entries for later reference
505 			 * if the output fails.
506 			 */
507 			data->ent = *curr;
508 			data->ret = *next;
509 		}
510 	}
511 
512 	if (next->ent.type != TRACE_GRAPH_RET)
513 		return NULL;
514 
515 	if (curr->ent.pid != next->ent.pid ||
516 			curr->graph_ent.func != next->ret.func)
517 		return NULL;
518 
519 	/* this is a leaf, now advance the iterator */
520 	if (ring_iter)
521 		ring_buffer_read(ring_iter, NULL);
522 
523 	return next;
524 }
525 
526 /* Signal a overhead of time execution to the output */
527 static int
528 print_graph_overhead(unsigned long long duration, struct trace_seq *s)
529 {
530 	/* If duration disappear, we don't need anything */
531 	if (!(tracer_flags.val & TRACE_GRAPH_PRINT_DURATION))
532 		return 1;
533 
534 	/* Non nested entry or return */
535 	if (duration == -1)
536 		return trace_seq_printf(s, "  ");
537 
538 	if (tracer_flags.val & TRACE_GRAPH_PRINT_OVERHEAD) {
539 		/* Duration exceeded 100 msecs */
540 		if (duration > 100000ULL)
541 			return trace_seq_printf(s, "! ");
542 
543 		/* Duration exceeded 10 msecs */
544 		if (duration > 10000ULL)
545 			return trace_seq_printf(s, "+ ");
546 	}
547 
548 	return trace_seq_printf(s, "  ");
549 }
550 
551 static int print_graph_abs_time(u64 t, struct trace_seq *s)
552 {
553 	unsigned long usecs_rem;
554 
555 	usecs_rem = do_div(t, NSEC_PER_SEC);
556 	usecs_rem /= 1000;
557 
558 	return trace_seq_printf(s, "%5lu.%06lu |  ",
559 			(unsigned long)t, usecs_rem);
560 }
561 
562 static enum print_line_t
563 print_graph_irq(struct trace_iterator *iter, unsigned long addr,
564 		enum trace_type type, int cpu, pid_t pid)
565 {
566 	int ret;
567 	struct trace_seq *s = &iter->seq;
568 
569 	if (addr < (unsigned long)__irqentry_text_start ||
570 		addr >= (unsigned long)__irqentry_text_end)
571 		return TRACE_TYPE_UNHANDLED;
572 
573 	/* Absolute time */
574 	if (tracer_flags.val & TRACE_GRAPH_PRINT_ABS_TIME) {
575 		ret = print_graph_abs_time(iter->ts, s);
576 		if (!ret)
577 			return TRACE_TYPE_PARTIAL_LINE;
578 	}
579 
580 	/* Cpu */
581 	if (tracer_flags.val & TRACE_GRAPH_PRINT_CPU) {
582 		ret = print_graph_cpu(s, cpu);
583 		if (ret == TRACE_TYPE_PARTIAL_LINE)
584 			return TRACE_TYPE_PARTIAL_LINE;
585 	}
586 
587 	/* Proc */
588 	if (tracer_flags.val & TRACE_GRAPH_PRINT_PROC) {
589 		ret = print_graph_proc(s, pid);
590 		if (ret == TRACE_TYPE_PARTIAL_LINE)
591 			return TRACE_TYPE_PARTIAL_LINE;
592 		ret = trace_seq_printf(s, " | ");
593 		if (!ret)
594 			return TRACE_TYPE_PARTIAL_LINE;
595 	}
596 
597 	/* No overhead */
598 	ret = print_graph_overhead(-1, s);
599 	if (!ret)
600 		return TRACE_TYPE_PARTIAL_LINE;
601 
602 	if (type == TRACE_GRAPH_ENT)
603 		ret = trace_seq_printf(s, "==========>");
604 	else
605 		ret = trace_seq_printf(s, "<==========");
606 
607 	if (!ret)
608 		return TRACE_TYPE_PARTIAL_LINE;
609 
610 	/* Don't close the duration column if haven't one */
611 	if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION)
612 		trace_seq_printf(s, " |");
613 	ret = trace_seq_printf(s, "\n");
614 
615 	if (!ret)
616 		return TRACE_TYPE_PARTIAL_LINE;
617 	return TRACE_TYPE_HANDLED;
618 }
619 
620 enum print_line_t
621 trace_print_graph_duration(unsigned long long duration, struct trace_seq *s)
622 {
623 	unsigned long nsecs_rem = do_div(duration, 1000);
624 	/* log10(ULONG_MAX) + '\0' */
625 	char msecs_str[21];
626 	char nsecs_str[5];
627 	int ret, len;
628 	int i;
629 
630 	sprintf(msecs_str, "%lu", (unsigned long) duration);
631 
632 	/* Print msecs */
633 	ret = trace_seq_printf(s, "%s", msecs_str);
634 	if (!ret)
635 		return TRACE_TYPE_PARTIAL_LINE;
636 
637 	len = strlen(msecs_str);
638 
639 	/* Print nsecs (we don't want to exceed 7 numbers) */
640 	if (len < 7) {
641 		snprintf(nsecs_str, 8 - len, "%03lu", nsecs_rem);
642 		ret = trace_seq_printf(s, ".%s", nsecs_str);
643 		if (!ret)
644 			return TRACE_TYPE_PARTIAL_LINE;
645 		len += strlen(nsecs_str);
646 	}
647 
648 	ret = trace_seq_printf(s, " us ");
649 	if (!ret)
650 		return TRACE_TYPE_PARTIAL_LINE;
651 
652 	/* Print remaining spaces to fit the row's width */
653 	for (i = len; i < 7; i++) {
654 		ret = trace_seq_printf(s, " ");
655 		if (!ret)
656 			return TRACE_TYPE_PARTIAL_LINE;
657 	}
658 	return TRACE_TYPE_HANDLED;
659 }
660 
661 static enum print_line_t
662 print_graph_duration(unsigned long long duration, struct trace_seq *s)
663 {
664 	int ret;
665 
666 	ret = trace_print_graph_duration(duration, s);
667 	if (ret != TRACE_TYPE_HANDLED)
668 		return ret;
669 
670 	ret = trace_seq_printf(s, "|  ");
671 	if (!ret)
672 		return TRACE_TYPE_PARTIAL_LINE;
673 
674 	return TRACE_TYPE_HANDLED;
675 }
676 
677 /* Case of a leaf function on its call entry */
678 static enum print_line_t
679 print_graph_entry_leaf(struct trace_iterator *iter,
680 		struct ftrace_graph_ent_entry *entry,
681 		struct ftrace_graph_ret_entry *ret_entry, struct trace_seq *s)
682 {
683 	struct fgraph_data *data = iter->private;
684 	struct ftrace_graph_ret *graph_ret;
685 	struct ftrace_graph_ent *call;
686 	unsigned long long duration;
687 	int ret;
688 	int i;
689 
690 	graph_ret = &ret_entry->ret;
691 	call = &entry->graph_ent;
692 	duration = graph_ret->rettime - graph_ret->calltime;
693 
694 	if (data) {
695 		struct fgraph_cpu_data *cpu_data;
696 		int cpu = iter->cpu;
697 
698 		cpu_data = per_cpu_ptr(data->cpu_data, cpu);
699 
700 		/*
701 		 * Comments display at + 1 to depth. Since
702 		 * this is a leaf function, keep the comments
703 		 * equal to this depth.
704 		 */
705 		cpu_data->depth = call->depth - 1;
706 
707 		/* No need to keep this function around for this depth */
708 		if (call->depth < FTRACE_RETFUNC_DEPTH)
709 			cpu_data->enter_funcs[call->depth] = 0;
710 	}
711 
712 	/* Overhead */
713 	ret = print_graph_overhead(duration, s);
714 	if (!ret)
715 		return TRACE_TYPE_PARTIAL_LINE;
716 
717 	/* Duration */
718 	if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION) {
719 		ret = print_graph_duration(duration, s);
720 		if (ret == TRACE_TYPE_PARTIAL_LINE)
721 			return TRACE_TYPE_PARTIAL_LINE;
722 	}
723 
724 	/* Function */
725 	for (i = 0; i < call->depth * TRACE_GRAPH_INDENT; i++) {
726 		ret = trace_seq_printf(s, " ");
727 		if (!ret)
728 			return TRACE_TYPE_PARTIAL_LINE;
729 	}
730 
731 	ret = trace_seq_printf(s, "%ps();\n", (void *)call->func);
732 	if (!ret)
733 		return TRACE_TYPE_PARTIAL_LINE;
734 
735 	return TRACE_TYPE_HANDLED;
736 }
737 
738 static enum print_line_t
739 print_graph_entry_nested(struct trace_iterator *iter,
740 			 struct ftrace_graph_ent_entry *entry,
741 			 struct trace_seq *s, int cpu)
742 {
743 	struct ftrace_graph_ent *call = &entry->graph_ent;
744 	struct fgraph_data *data = iter->private;
745 	int ret;
746 	int i;
747 
748 	if (data) {
749 		struct fgraph_cpu_data *cpu_data;
750 		int cpu = iter->cpu;
751 
752 		cpu_data = per_cpu_ptr(data->cpu_data, cpu);
753 		cpu_data->depth = call->depth;
754 
755 		/* Save this function pointer to see if the exit matches */
756 		if (call->depth < FTRACE_RETFUNC_DEPTH)
757 			cpu_data->enter_funcs[call->depth] = call->func;
758 	}
759 
760 	/* No overhead */
761 	ret = print_graph_overhead(-1, s);
762 	if (!ret)
763 		return TRACE_TYPE_PARTIAL_LINE;
764 
765 	/* No time */
766 	if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION) {
767 		ret = trace_seq_printf(s, "            |  ");
768 		if (!ret)
769 			return TRACE_TYPE_PARTIAL_LINE;
770 	}
771 
772 	/* Function */
773 	for (i = 0; i < call->depth * TRACE_GRAPH_INDENT; i++) {
774 		ret = trace_seq_printf(s, " ");
775 		if (!ret)
776 			return TRACE_TYPE_PARTIAL_LINE;
777 	}
778 
779 	ret = trace_seq_printf(s, "%ps() {\n", (void *)call->func);
780 	if (!ret)
781 		return TRACE_TYPE_PARTIAL_LINE;
782 
783 	/*
784 	 * we already consumed the current entry to check the next one
785 	 * and see if this is a leaf.
786 	 */
787 	return TRACE_TYPE_NO_CONSUME;
788 }
789 
790 static enum print_line_t
791 print_graph_prologue(struct trace_iterator *iter, struct trace_seq *s,
792 		     int type, unsigned long addr)
793 {
794 	struct fgraph_data *data = iter->private;
795 	struct trace_entry *ent = iter->ent;
796 	int cpu = iter->cpu;
797 	int ret;
798 
799 	/* Pid */
800 	if (verif_pid(s, ent->pid, cpu, data) == TRACE_TYPE_PARTIAL_LINE)
801 		return TRACE_TYPE_PARTIAL_LINE;
802 
803 	if (type) {
804 		/* Interrupt */
805 		ret = print_graph_irq(iter, addr, type, cpu, ent->pid);
806 		if (ret == TRACE_TYPE_PARTIAL_LINE)
807 			return TRACE_TYPE_PARTIAL_LINE;
808 	}
809 
810 	/* Absolute time */
811 	if (tracer_flags.val & TRACE_GRAPH_PRINT_ABS_TIME) {
812 		ret = print_graph_abs_time(iter->ts, s);
813 		if (!ret)
814 			return TRACE_TYPE_PARTIAL_LINE;
815 	}
816 
817 	/* Cpu */
818 	if (tracer_flags.val & TRACE_GRAPH_PRINT_CPU) {
819 		ret = print_graph_cpu(s, cpu);
820 		if (ret == TRACE_TYPE_PARTIAL_LINE)
821 			return TRACE_TYPE_PARTIAL_LINE;
822 	}
823 
824 	/* Proc */
825 	if (tracer_flags.val & TRACE_GRAPH_PRINT_PROC) {
826 		ret = print_graph_proc(s, ent->pid);
827 		if (ret == TRACE_TYPE_PARTIAL_LINE)
828 			return TRACE_TYPE_PARTIAL_LINE;
829 
830 		ret = trace_seq_printf(s, " | ");
831 		if (!ret)
832 			return TRACE_TYPE_PARTIAL_LINE;
833 	}
834 
835 	/* Latency format */
836 	if (trace_flags & TRACE_ITER_LATENCY_FMT) {
837 		ret = print_graph_lat_fmt(s, ent);
838 		if (ret == TRACE_TYPE_PARTIAL_LINE)
839 			return TRACE_TYPE_PARTIAL_LINE;
840 	}
841 
842 	return 0;
843 }
844 
845 static enum print_line_t
846 print_graph_entry(struct ftrace_graph_ent_entry *field, struct trace_seq *s,
847 			struct trace_iterator *iter)
848 {
849 	struct fgraph_data *data = iter->private;
850 	struct ftrace_graph_ent *call = &field->graph_ent;
851 	struct ftrace_graph_ret_entry *leaf_ret;
852 	static enum print_line_t ret;
853 	int cpu = iter->cpu;
854 
855 	if (print_graph_prologue(iter, s, TRACE_GRAPH_ENT, call->func))
856 		return TRACE_TYPE_PARTIAL_LINE;
857 
858 	leaf_ret = get_return_for_leaf(iter, field);
859 	if (leaf_ret)
860 		ret = print_graph_entry_leaf(iter, field, leaf_ret, s);
861 	else
862 		ret = print_graph_entry_nested(iter, field, s, cpu);
863 
864 	if (data) {
865 		/*
866 		 * If we failed to write our output, then we need to make
867 		 * note of it. Because we already consumed our entry.
868 		 */
869 		if (s->full) {
870 			data->failed = 1;
871 			data->cpu = cpu;
872 		} else
873 			data->failed = 0;
874 	}
875 
876 	return ret;
877 }
878 
879 static enum print_line_t
880 print_graph_return(struct ftrace_graph_ret *trace, struct trace_seq *s,
881 		   struct trace_entry *ent, struct trace_iterator *iter)
882 {
883 	unsigned long long duration = trace->rettime - trace->calltime;
884 	struct fgraph_data *data = iter->private;
885 	pid_t pid = ent->pid;
886 	int cpu = iter->cpu;
887 	int func_match = 1;
888 	int ret;
889 	int i;
890 
891 	if (data) {
892 		struct fgraph_cpu_data *cpu_data;
893 		int cpu = iter->cpu;
894 
895 		cpu_data = per_cpu_ptr(data->cpu_data, cpu);
896 
897 		/*
898 		 * Comments display at + 1 to depth. This is the
899 		 * return from a function, we now want the comments
900 		 * to display at the same level of the bracket.
901 		 */
902 		cpu_data->depth = trace->depth - 1;
903 
904 		if (trace->depth < FTRACE_RETFUNC_DEPTH) {
905 			if (cpu_data->enter_funcs[trace->depth] != trace->func)
906 				func_match = 0;
907 			cpu_data->enter_funcs[trace->depth] = 0;
908 		}
909 	}
910 
911 	if (print_graph_prologue(iter, s, 0, 0))
912 		return TRACE_TYPE_PARTIAL_LINE;
913 
914 	/* Overhead */
915 	ret = print_graph_overhead(duration, s);
916 	if (!ret)
917 		return TRACE_TYPE_PARTIAL_LINE;
918 
919 	/* Duration */
920 	if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION) {
921 		ret = print_graph_duration(duration, s);
922 		if (ret == TRACE_TYPE_PARTIAL_LINE)
923 			return TRACE_TYPE_PARTIAL_LINE;
924 	}
925 
926 	/* Closing brace */
927 	for (i = 0; i < trace->depth * TRACE_GRAPH_INDENT; i++) {
928 		ret = trace_seq_printf(s, " ");
929 		if (!ret)
930 			return TRACE_TYPE_PARTIAL_LINE;
931 	}
932 
933 	/*
934 	 * If the return function does not have a matching entry,
935 	 * then the entry was lost. Instead of just printing
936 	 * the '}' and letting the user guess what function this
937 	 * belongs to, write out the function name.
938 	 */
939 	if (func_match) {
940 		ret = trace_seq_printf(s, "}\n");
941 		if (!ret)
942 			return TRACE_TYPE_PARTIAL_LINE;
943 	} else {
944 		ret = trace_seq_printf(s, "} /* %ps */\n", (void *)trace->func);
945 		if (!ret)
946 			return TRACE_TYPE_PARTIAL_LINE;
947 	}
948 
949 	/* Overrun */
950 	if (tracer_flags.val & TRACE_GRAPH_PRINT_OVERRUN) {
951 		ret = trace_seq_printf(s, " (Overruns: %lu)\n",
952 					trace->overrun);
953 		if (!ret)
954 			return TRACE_TYPE_PARTIAL_LINE;
955 	}
956 
957 	ret = print_graph_irq(iter, trace->func, TRACE_GRAPH_RET, cpu, pid);
958 	if (ret == TRACE_TYPE_PARTIAL_LINE)
959 		return TRACE_TYPE_PARTIAL_LINE;
960 
961 	return TRACE_TYPE_HANDLED;
962 }
963 
964 static enum print_line_t
965 print_graph_comment(struct trace_seq *s,  struct trace_entry *ent,
966 		    struct trace_iterator *iter)
967 {
968 	unsigned long sym_flags = (trace_flags & TRACE_ITER_SYM_MASK);
969 	struct fgraph_data *data = iter->private;
970 	struct trace_event *event;
971 	int depth = 0;
972 	int ret;
973 	int i;
974 
975 	if (data)
976 		depth = per_cpu_ptr(data->cpu_data, iter->cpu)->depth;
977 
978 	if (print_graph_prologue(iter, s, 0, 0))
979 		return TRACE_TYPE_PARTIAL_LINE;
980 
981 	/* No overhead */
982 	ret = print_graph_overhead(-1, s);
983 	if (!ret)
984 		return TRACE_TYPE_PARTIAL_LINE;
985 
986 	/* No time */
987 	if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION) {
988 		ret = trace_seq_printf(s, "            |  ");
989 		if (!ret)
990 			return TRACE_TYPE_PARTIAL_LINE;
991 	}
992 
993 	/* Indentation */
994 	if (depth > 0)
995 		for (i = 0; i < (depth + 1) * TRACE_GRAPH_INDENT; i++) {
996 			ret = trace_seq_printf(s, " ");
997 			if (!ret)
998 				return TRACE_TYPE_PARTIAL_LINE;
999 		}
1000 
1001 	/* The comment */
1002 	ret = trace_seq_printf(s, "/* ");
1003 	if (!ret)
1004 		return TRACE_TYPE_PARTIAL_LINE;
1005 
1006 	switch (iter->ent->type) {
1007 	case TRACE_BPRINT:
1008 		ret = trace_print_bprintk_msg_only(iter);
1009 		if (ret != TRACE_TYPE_HANDLED)
1010 			return ret;
1011 		break;
1012 	case TRACE_PRINT:
1013 		ret = trace_print_printk_msg_only(iter);
1014 		if (ret != TRACE_TYPE_HANDLED)
1015 			return ret;
1016 		break;
1017 	default:
1018 		event = ftrace_find_event(ent->type);
1019 		if (!event)
1020 			return TRACE_TYPE_UNHANDLED;
1021 
1022 		ret = event->trace(iter, sym_flags);
1023 		if (ret != TRACE_TYPE_HANDLED)
1024 			return ret;
1025 	}
1026 
1027 	/* Strip ending newline */
1028 	if (s->buffer[s->len - 1] == '\n') {
1029 		s->buffer[s->len - 1] = '\0';
1030 		s->len--;
1031 	}
1032 
1033 	ret = trace_seq_printf(s, " */\n");
1034 	if (!ret)
1035 		return TRACE_TYPE_PARTIAL_LINE;
1036 
1037 	return TRACE_TYPE_HANDLED;
1038 }
1039 
1040 
1041 enum print_line_t
1042 print_graph_function(struct trace_iterator *iter)
1043 {
1044 	struct ftrace_graph_ent_entry *field;
1045 	struct fgraph_data *data = iter->private;
1046 	struct trace_entry *entry = iter->ent;
1047 	struct trace_seq *s = &iter->seq;
1048 	int cpu = iter->cpu;
1049 	int ret;
1050 
1051 	if (data && per_cpu_ptr(data->cpu_data, cpu)->ignore) {
1052 		per_cpu_ptr(data->cpu_data, cpu)->ignore = 0;
1053 		return TRACE_TYPE_HANDLED;
1054 	}
1055 
1056 	/*
1057 	 * If the last output failed, there's a possibility we need
1058 	 * to print out the missing entry which would never go out.
1059 	 */
1060 	if (data && data->failed) {
1061 		field = &data->ent;
1062 		iter->cpu = data->cpu;
1063 		ret = print_graph_entry(field, s, iter);
1064 		if (ret == TRACE_TYPE_HANDLED && iter->cpu != cpu) {
1065 			per_cpu_ptr(data->cpu_data, iter->cpu)->ignore = 1;
1066 			ret = TRACE_TYPE_NO_CONSUME;
1067 		}
1068 		iter->cpu = cpu;
1069 		return ret;
1070 	}
1071 
1072 	switch (entry->type) {
1073 	case TRACE_GRAPH_ENT: {
1074 		/*
1075 		 * print_graph_entry() may consume the current event,
1076 		 * thus @field may become invalid, so we need to save it.
1077 		 * sizeof(struct ftrace_graph_ent_entry) is very small,
1078 		 * it can be safely saved at the stack.
1079 		 */
1080 		struct ftrace_graph_ent_entry saved;
1081 		trace_assign_type(field, entry);
1082 		saved = *field;
1083 		return print_graph_entry(&saved, s, iter);
1084 	}
1085 	case TRACE_GRAPH_RET: {
1086 		struct ftrace_graph_ret_entry *field;
1087 		trace_assign_type(field, entry);
1088 		return print_graph_return(&field->ret, s, entry, iter);
1089 	}
1090 	default:
1091 		return print_graph_comment(s, entry, iter);
1092 	}
1093 
1094 	return TRACE_TYPE_HANDLED;
1095 }
1096 
1097 static void print_lat_header(struct seq_file *s)
1098 {
1099 	static const char spaces[] = "                "	/* 16 spaces */
1100 		"    "					/* 4 spaces */
1101 		"                 ";			/* 17 spaces */
1102 	int size = 0;
1103 
1104 	if (tracer_flags.val & TRACE_GRAPH_PRINT_ABS_TIME)
1105 		size += 16;
1106 	if (tracer_flags.val & TRACE_GRAPH_PRINT_CPU)
1107 		size += 4;
1108 	if (tracer_flags.val & TRACE_GRAPH_PRINT_PROC)
1109 		size += 17;
1110 
1111 	seq_printf(s, "#%.*s  _-----=> irqs-off        \n", size, spaces);
1112 	seq_printf(s, "#%.*s / _----=> need-resched    \n", size, spaces);
1113 	seq_printf(s, "#%.*s| / _---=> hardirq/softirq \n", size, spaces);
1114 	seq_printf(s, "#%.*s|| / _--=> preempt-depth   \n", size, spaces);
1115 	seq_printf(s, "#%.*s||| / _-=> lock-depth      \n", size, spaces);
1116 	seq_printf(s, "#%.*s|||| /                     \n", size, spaces);
1117 }
1118 
1119 static void print_graph_headers(struct seq_file *s)
1120 {
1121 	int lat = trace_flags & TRACE_ITER_LATENCY_FMT;
1122 
1123 	if (lat)
1124 		print_lat_header(s);
1125 
1126 	/* 1st line */
1127 	seq_printf(s, "#");
1128 	if (tracer_flags.val & TRACE_GRAPH_PRINT_ABS_TIME)
1129 		seq_printf(s, "     TIME       ");
1130 	if (tracer_flags.val & TRACE_GRAPH_PRINT_CPU)
1131 		seq_printf(s, " CPU");
1132 	if (tracer_flags.val & TRACE_GRAPH_PRINT_PROC)
1133 		seq_printf(s, "  TASK/PID       ");
1134 	if (lat)
1135 		seq_printf(s, "|||||");
1136 	if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION)
1137 		seq_printf(s, "  DURATION   ");
1138 	seq_printf(s, "               FUNCTION CALLS\n");
1139 
1140 	/* 2nd line */
1141 	seq_printf(s, "#");
1142 	if (tracer_flags.val & TRACE_GRAPH_PRINT_ABS_TIME)
1143 		seq_printf(s, "      |         ");
1144 	if (tracer_flags.val & TRACE_GRAPH_PRINT_CPU)
1145 		seq_printf(s, " |  ");
1146 	if (tracer_flags.val & TRACE_GRAPH_PRINT_PROC)
1147 		seq_printf(s, "   |    |        ");
1148 	if (lat)
1149 		seq_printf(s, "|||||");
1150 	if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION)
1151 		seq_printf(s, "   |   |      ");
1152 	seq_printf(s, "               |   |   |   |\n");
1153 }
1154 
1155 static void graph_trace_open(struct trace_iterator *iter)
1156 {
1157 	/* pid and depth on the last trace processed */
1158 	struct fgraph_data *data;
1159 	int cpu;
1160 
1161 	iter->private = NULL;
1162 
1163 	data = kzalloc(sizeof(*data), GFP_KERNEL);
1164 	if (!data)
1165 		goto out_err;
1166 
1167 	data->cpu_data = alloc_percpu(struct fgraph_cpu_data);
1168 	if (!data->cpu_data)
1169 		goto out_err_free;
1170 
1171 	for_each_possible_cpu(cpu) {
1172 		pid_t *pid = &(per_cpu_ptr(data->cpu_data, cpu)->last_pid);
1173 		int *depth = &(per_cpu_ptr(data->cpu_data, cpu)->depth);
1174 		int *ignore = &(per_cpu_ptr(data->cpu_data, cpu)->ignore);
1175 		*pid = -1;
1176 		*depth = 0;
1177 		*ignore = 0;
1178 	}
1179 
1180 	iter->private = data;
1181 
1182 	return;
1183 
1184  out_err_free:
1185 	kfree(data);
1186  out_err:
1187 	pr_warning("function graph tracer: not enough memory\n");
1188 }
1189 
1190 static void graph_trace_close(struct trace_iterator *iter)
1191 {
1192 	struct fgraph_data *data = iter->private;
1193 
1194 	if (data) {
1195 		free_percpu(data->cpu_data);
1196 		kfree(data);
1197 	}
1198 }
1199 
1200 static struct tracer graph_trace __read_mostly = {
1201 	.name		= "function_graph",
1202 	.open		= graph_trace_open,
1203 	.pipe_open	= graph_trace_open,
1204 	.close		= graph_trace_close,
1205 	.pipe_close	= graph_trace_close,
1206 	.wait_pipe	= poll_wait_pipe,
1207 	.init		= graph_trace_init,
1208 	.reset		= graph_trace_reset,
1209 	.print_line	= print_graph_function,
1210 	.print_header	= print_graph_headers,
1211 	.flags		= &tracer_flags,
1212 #ifdef CONFIG_FTRACE_SELFTEST
1213 	.selftest	= trace_selftest_startup_function_graph,
1214 #endif
1215 };
1216 
1217 static __init int init_graph_trace(void)
1218 {
1219 	max_bytes_for_cpu = snprintf(NULL, 0, "%d", nr_cpu_ids - 1);
1220 
1221 	return register_tracer(&graph_trace);
1222 }
1223 
1224 device_initcall(init_graph_trace);
1225