trace_functions_graph.c 38.2 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
 * 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>
12
#include <linux/slab.h>
13 14 15
#include <linux/fs.h>

#include "trace.h"
16
#include "trace_output.h"
17

18 19 20
/* When set, irq functions will be ignored */
static int ftrace_graph_skip_irqs;

21
struct fgraph_cpu_data {
22 23
	pid_t		last_pid;
	int		depth;
24
	int		depth_irq;
25
	int		ignore;
26
	unsigned long	enter_funcs[FTRACE_RETFUNC_DEPTH];
27 28 29
};

struct fgraph_data {
30
	struct fgraph_cpu_data __percpu *cpu_data;
31 32 33 34 35 36

	/* Place to preserve last processed entry. */
	struct ftrace_graph_ent_entry	ent;
	struct ftrace_graph_ret_entry	ret;
	int				failed;
	int				cpu;
37 38
};

39
#define TRACE_GRAPH_INDENT	2
40

41 42
static unsigned int max_depth;

43
static struct tracer_opt trace_opts[] = {
44
	/* Display overruns? (for self-debug purpose) */
45 46 47 48 49
	{ 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) },
50 51
	/* Display proc name/pid */
	{ TRACER_OPT(funcgraph-proc, TRACE_GRAPH_PRINT_PROC) },
52 53 54 55
	/* 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) },
56 57
	/* Display interrupts */
	{ TRACER_OPT(funcgraph-irqs, TRACE_GRAPH_PRINT_IRQS) },
58 59
	/* Display function name after trailing } */
	{ TRACER_OPT(funcgraph-tail, TRACE_GRAPH_PRINT_TAIL) },
60 61 62 63
	{ } /* Empty entry */
};

static struct tracer_flags tracer_flags = {
64
	/* Don't display overruns, proc, or tail by default */
65
	.val = TRACE_GRAPH_PRINT_CPU | TRACE_GRAPH_PRINT_OVERHEAD |
66
	       TRACE_GRAPH_PRINT_DURATION | TRACE_GRAPH_PRINT_IRQS,
67 68 69
	.opts = trace_opts
};

70
static struct trace_array *graph_array;
71

72 73 74 75 76 77
/*
 * DURATION column is being also used to display IRQ signs,
 * following values are used by print_graph_irq and others
 * to fill in space into DURATION column.
 */
enum {
78 79 80
	FLAGS_FILL_FULL  = 1 << TRACE_GRAPH_PRINT_FILL_SHIFT,
	FLAGS_FILL_START = 2 << TRACE_GRAPH_PRINT_FILL_SHIFT,
	FLAGS_FILL_END   = 3 << TRACE_GRAPH_PRINT_FILL_SHIFT,
81 82 83 84 85
};

static enum print_line_t
print_graph_duration(unsigned long long duration, struct trace_seq *s,
		     u32 flags);
86

87 88
/* Add a function return address to the trace stack on thread info.*/
int
89 90
ftrace_push_return_trace(unsigned long ret, unsigned long func, int *depth,
			 unsigned long frame_pointer)
91
{
92
	unsigned long long calltime;
93 94 95 96 97
	int index;

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

98 99 100 101 102 103
	/*
	 * We must make sure the ret_stack is tested before we read
	 * anything else.
	 */
	smp_rmb();

104 105 106 107 108 109
	/* The return trace stack is full */
	if (current->curr_ret_stack == FTRACE_RETFUNC_DEPTH - 1) {
		atomic_inc(&current->trace_overrun);
		return -EBUSY;
	}

110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128
	/*
	 * The curr_ret_stack is an index to ftrace return stack of
	 * current task.  Its value should be in [0, FTRACE_RETFUNC_
	 * DEPTH) when the function graph tracer is used.  To support
	 * filtering out specific functions, it makes the index
	 * negative by subtracting huge value (FTRACE_NOTRACE_DEPTH)
	 * so when it sees a negative index the ftrace will ignore
	 * the record.  And the index gets recovered when returning
	 * from the filtered function by adding the FTRACE_NOTRACE_
	 * DEPTH and then it'll continue to record functions normally.
	 *
	 * The curr_ret_stack is initialized to -1 and get increased
	 * in this function.  So it can be less than -1 only if it was
	 * filtered out via ftrace_graph_notrace_addr() which can be
	 * set from set_graph_notrace file in debugfs by user.
	 */
	if (current->curr_ret_stack < -1)
		return -EBUSY;

129 130
	calltime = trace_clock_local();

131
	index = ++current->curr_ret_stack;
132 133
	if (ftrace_graph_notrace_addr(func))
		current->curr_ret_stack -= FTRACE_NOTRACE_DEPTH;
134 135 136
	barrier();
	current->ret_stack[index].ret = ret;
	current->ret_stack[index].func = func;
137
	current->ret_stack[index].calltime = calltime;
138
	current->ret_stack[index].subtime = 0;
139
	current->ret_stack[index].fp = frame_pointer;
140
	*depth = current->curr_ret_stack;
141 142 143 144 145

	return 0;
}

/* Retrieve a function return address to the trace stack on thread info.*/
146
static void
147 148
ftrace_pop_return_trace(struct ftrace_graph_ret *trace, unsigned long *ret,
			unsigned long frame_pointer)
149 150 151 152 153
{
	int index;

	index = current->curr_ret_stack;

154 155 156 157 158 159 160 161 162 163 164
	/*
	 * A negative index here means that it's just returned from a
	 * notrace'd function.  Recover index to get an original
	 * return address.  See ftrace_push_return_trace().
	 *
	 * TODO: Need to check whether the stack gets corrupted.
	 */
	if (index < 0)
		index += FTRACE_NOTRACE_DEPTH;

	if (unlikely(index < 0 || index >= FTRACE_RETFUNC_DEPTH)) {
165 166 167 168 169 170 171
		ftrace_graph_stop();
		WARN_ON(1);
		/* Might as well panic, otherwise we have no where to go */
		*ret = (unsigned long)panic;
		return;
	}

172
#if defined(CONFIG_HAVE_FUNCTION_GRAPH_FP_TEST) && !defined(CC_USING_FENTRY)
173 174 175 176 177 178 179 180 181 182
	/*
	 * 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.
183 184 185
	 *
	 * Note, -mfentry does not use frame pointers, and this test
	 *  is not needed if CC_USING_FENTRY is set.
186 187 188 189
	 */
	if (unlikely(current->ret_stack[index].fp != frame_pointer)) {
		ftrace_graph_stop();
		WARN(1, "Bad frame pointer: expected %lx, received %lx\n"
190
		     "  from func %ps return to %lx\n",
191 192 193 194 195 196 197 198 199
		     current->ret_stack[index].fp,
		     frame_pointer,
		     (void *)current->ret_stack[index].func,
		     current->ret_stack[index].ret);
		*ret = (unsigned long)panic;
		return;
	}
#endif

200 201 202 203 204 205 206 207 208 209 210
	*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.
 */
211
unsigned long ftrace_return_to_handler(unsigned long frame_pointer)
212 213 214 215
{
	struct ftrace_graph_ret trace;
	unsigned long ret;

216
	ftrace_pop_return_trace(&trace, &ret, frame_pointer);
217
	trace.rettime = trace_clock_local();
218 219
	barrier();
	current->curr_ret_stack--;
220 221 222 223 224 225 226 227 228
	/*
	 * The curr_ret_stack can be less than -1 only if it was
	 * filtered out and it's about to return from the function.
	 * Recover the index and continue to trace normal functions.
	 */
	if (current->curr_ret_stack < -1) {
		current->curr_ret_stack += FTRACE_NOTRACE_DEPTH;
		return ret;
	}
229

230 231 232 233 234 235 236
	/*
	 * The trace should run after decrementing the ret counter
	 * in case an interrupt were to come in. We don't want to
	 * lose the interrupt if max_depth is set.
	 */
	ftrace_graph_return(&trace);

237 238 239 240 241 242 243 244 245 246
	if (unlikely(!ret)) {
		ftrace_graph_stop();
		WARN_ON(1);
		/* Might as well panic. What else to do? */
		ret = (unsigned long)panic;
	}

	return ret;
}

247
int __trace_graph_entry(struct trace_array *tr,
248 249 250 251 252 253
				struct ftrace_graph_ent *trace,
				unsigned long flags,
				int pc)
{
	struct ftrace_event_call *call = &event_funcgraph_entry;
	struct ring_buffer_event *event;
254
	struct ring_buffer *buffer = tr->trace_buffer.buffer;
255 256
	struct ftrace_graph_ent_entry *entry;

257
	if (unlikely(__this_cpu_read(ftrace_cpu_disabled)))
258 259
		return 0;

260
	event = trace_buffer_lock_reserve(buffer, TRACE_GRAPH_ENT,
261 262 263 264 265
					  sizeof(*entry), flags, pc);
	if (!event)
		return 0;
	entry	= ring_buffer_event_data(event);
	entry->graph_ent			= *trace;
266
	if (!call_filter_check_discard(call, entry, buffer, event))
267
		__buffer_unlock_commit(buffer, event);
268 269 270 271

	return 1;
}

272 273
static inline int ftrace_graph_ignore_irqs(void)
{
274
	if (!ftrace_graph_skip_irqs || trace_recursion_test(TRACE_IRQ_BIT))
275 276 277 278 279
		return 0;

	return in_irq();
}

280 281 282 283 284 285 286 287 288 289 290 291 292
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;

293
	/* trace it when it is-nested-in or is a function enabled. */
294
	if ((!(trace->depth || ftrace_graph_addr(trace->func)) ||
295
	     ftrace_graph_ignore_irqs()) || (trace->depth < 0) ||
296
	    (max_depth && trace->depth >= max_depth))
297 298
		return 0;

299 300 301 302 303 304 305 306 307 308
	/*
	 * Do not trace a function if it's filtered by set_graph_notrace.
	 * Make the index of ret stack negative to indicate that it should
	 * ignore further functions.  But it needs its own ret stack entry
	 * to recover the original index in order to continue tracing after
	 * returning from the function.
	 */
	if (ftrace_graph_notrace_addr(trace->func))
		return 1;

309 310
	local_irq_save(flags);
	cpu = raw_smp_processor_id();
311
	data = per_cpu_ptr(tr->trace_buffer.data, cpu);
312 313 314 315 316 317 318 319 320 321 322 323 324 325
	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;
}

326 327 328 329 330 331 332 333
int trace_graph_thresh_entry(struct ftrace_graph_ent *trace)
{
	if (tracing_thresh)
		return 1;
	else
		return trace_graph_entry(trace);
}

334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361
static void
__trace_graph_function(struct trace_array *tr,
		unsigned long ip, unsigned long flags, int pc)
{
	u64 time = trace_clock_local();
	struct ftrace_graph_ent ent = {
		.func  = ip,
		.depth = 0,
	};
	struct ftrace_graph_ret ret = {
		.func     = ip,
		.depth    = 0,
		.calltime = time,
		.rettime  = time,
	};

	__trace_graph_entry(tr, &ent, flags, pc);
	__trace_graph_return(tr, &ret, flags, pc);
}

void
trace_graph_function(struct trace_array *tr,
		unsigned long ip, unsigned long parent_ip,
		unsigned long flags, int pc)
{
	__trace_graph_function(tr, ip, flags, pc);
}

362
void __trace_graph_return(struct trace_array *tr,
363 364 365 366 367 368
				struct ftrace_graph_ret *trace,
				unsigned long flags,
				int pc)
{
	struct ftrace_event_call *call = &event_funcgraph_exit;
	struct ring_buffer_event *event;
369
	struct ring_buffer *buffer = tr->trace_buffer.buffer;
370 371
	struct ftrace_graph_ret_entry *entry;

372
	if (unlikely(__this_cpu_read(ftrace_cpu_disabled)))
373 374
		return;

375
	event = trace_buffer_lock_reserve(buffer, TRACE_GRAPH_RET,
376 377 378 379 380
					  sizeof(*entry), flags, pc);
	if (!event)
		return;
	entry	= ring_buffer_event_data(event);
	entry->ret				= *trace;
381
	if (!call_filter_check_discard(call, entry, buffer, event))
382
		__buffer_unlock_commit(buffer, event);
383 384 385 386 387 388 389 390 391 392 393 394 395
}

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();
396
	data = per_cpu_ptr(tr->trace_buffer.data, cpu);
397 398 399 400 401 402 403 404 405
	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);
}

406 407 408 409 410 411 412 413 414
void set_graph_array(struct trace_array *tr)
{
	graph_array = tr;

	/* Make graph_array visible before we start tracing */

	smp_mb();
}

415 416 417 418 419 420 421 422 423
void trace_graph_thresh_return(struct ftrace_graph_ret *trace)
{
	if (tracing_thresh &&
	    (trace->rettime - trace->calltime < tracing_thresh))
		return;
	else
		trace_graph_return(trace);
}

424 425
static int graph_trace_init(struct trace_array *tr)
{
426 427
	int ret;

428
	set_graph_array(tr);
429 430 431 432 433 434
	if (tracing_thresh)
		ret = register_ftrace_graph(&trace_graph_thresh_return,
					    &trace_graph_thresh_entry);
	else
		ret = register_ftrace_graph(&trace_graph_return,
					    &trace_graph_entry);
435 436 437 438 439
	if (ret)
		return ret;
	tracing_start_cmdline_record();

	return 0;
440 441 442 443
}

static void graph_trace_reset(struct trace_array *tr)
{
444 445
	tracing_stop_cmdline_record();
	unregister_ftrace_graph();
446 447
}

448
static int max_bytes_for_cpu;
449 450 451 452 453 454

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

455 456 457 458 459
	/*
	 * Start with a space character - to make it stand out
	 * to the right a bit when trace output is pasted into
	 * email:
	 */
460
	ret = trace_seq_printf(s, " %*d) ", max_bytes_for_cpu, cpu);
461
	if (!ret)
462 463
		return TRACE_TYPE_PARTIAL_LINE;

464 465 466
	return TRACE_TYPE_HANDLED;
}

467 468 469 470 471
#define TRACE_GRAPH_PROCINFO_LENGTH	14

static enum print_line_t
print_graph_proc(struct trace_seq *s, pid_t pid)
{
472
	char comm[TASK_COMM_LEN];
473 474
	/* sign + log10(MAX_INT) + '\0' */
	char pid_str[11];
475 476 477 478
	int spaces = 0;
	int ret;
	int len;
	int i;
479

480
	trace_find_cmdline(pid, comm);
481 482 483 484 485 486 487 488 489 490 491
	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++) {
492
		ret = trace_seq_putc(s, ' ');
493 494 495 496 497 498 499 500 501 502
		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++) {
503
		ret = trace_seq_putc(s, ' ');
504 505 506 507 508 509
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
	}
	return TRACE_TYPE_HANDLED;
}

510

511 512 513
static enum print_line_t
print_graph_lat_fmt(struct trace_seq *s, struct trace_entry *entry)
{
514
	if (!trace_seq_putc(s, ' '))
515 516
		return 0;

517
	return trace_print_lat_fmt(s, entry);
518 519
}

520
/* If the pid changed since the last trace, output this event */
521
static enum print_line_t
522
verif_pid(struct trace_seq *s, pid_t pid, int cpu, struct fgraph_data *data)
523
{
524
	pid_t prev_pid;
525
	pid_t *last_pid;
526
	int ret;
527

528
	if (!data)
529 530
		return TRACE_TYPE_HANDLED;

531
	last_pid = &(per_cpu_ptr(data->cpu_data, cpu)->last_pid);
532 533

	if (*last_pid == pid)
534
		return TRACE_TYPE_HANDLED;
535

536 537
	prev_pid = *last_pid;
	*last_pid = pid;
538

539 540
	if (prev_pid == -1)
		return TRACE_TYPE_HANDLED;
541 542 543 544 545 546 547 548
/*
 * Context-switch trace line:

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

 */
549
	ret = trace_seq_puts(s,
550
		" ------------------------------------------\n");
551
	if (!ret)
552
		return TRACE_TYPE_PARTIAL_LINE;
553 554 555

	ret = print_graph_cpu(s, cpu);
	if (ret == TRACE_TYPE_PARTIAL_LINE)
556
		return TRACE_TYPE_PARTIAL_LINE;
557 558 559

	ret = print_graph_proc(s, prev_pid);
	if (ret == TRACE_TYPE_PARTIAL_LINE)
560
		return TRACE_TYPE_PARTIAL_LINE;
561

562
	ret = trace_seq_puts(s, " => ");
563
	if (!ret)
564
		return TRACE_TYPE_PARTIAL_LINE;
565 566 567

	ret = print_graph_proc(s, pid);
	if (ret == TRACE_TYPE_PARTIAL_LINE)
568
		return TRACE_TYPE_PARTIAL_LINE;
569

570
	ret = trace_seq_puts(s,
571 572
		"\n ------------------------------------------\n\n");
	if (!ret)
573
		return TRACE_TYPE_PARTIAL_LINE;
574

575
	return TRACE_TYPE_HANDLED;
576 577
}

578 579
static struct ftrace_graph_ret_entry *
get_return_for_leaf(struct trace_iterator *iter,
580 581
		struct ftrace_graph_ent_entry *curr)
{
582 583
	struct fgraph_data *data = iter->private;
	struct ring_buffer_iter *ring_iter = NULL;
584 585 586
	struct ring_buffer_event *event;
	struct ftrace_graph_ret_entry *next;

587 588 589 590 591 592 593 594
	/*
	 * 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 {
595

596
		ring_iter = trace_buffer_iter(iter, iter->cpu);
597 598 599 600 601 602 603 604 605

		/* 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.
			 */
606
			ring_buffer_consume(iter->trace_buffer->buffer, iter->cpu,
607
					    NULL, NULL);
608
			event = ring_buffer_peek(iter->trace_buffer->buffer, iter->cpu,
609
						 NULL, NULL);
610
		}
611

612 613 614 615
		if (!event)
			return NULL;

		next = ring_buffer_event_data(event);
616

617 618 619 620 621 622
		if (data) {
			/*
			 * Save current and next entries for later reference
			 * if the output fails.
			 */
			data->ent = *curr;
623 624 625 626 627 628 629 630 631
			/*
			 * If the next event is not a return type, then
			 * we only care about what type it is. Otherwise we can
			 * safely copy the entire event.
			 */
			if (next->ent.type == TRACE_GRAPH_RET)
				data->ret = *next;
			else
				data->ret.ent.type = next->ent.type;
632 633
		}
	}
634 635

	if (next->ent.type != TRACE_GRAPH_RET)
636
		return NULL;
637 638 639

	if (curr->ent.pid != next->ent.pid ||
			curr->graph_ent.func != next->ret.func)
640
		return NULL;
641

642 643 644 645 646
	/* this is a leaf, now advance the iterator */
	if (ring_iter)
		ring_buffer_read(ring_iter, NULL);

	return next;
647 648
}

649 650 651 652 653 654 655 656 657 658 659
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);
}

660
static enum print_line_t
661
print_graph_irq(struct trace_iterator *iter, unsigned long addr,
662
		enum trace_type type, int cpu, pid_t pid, u32 flags)
663 664
{
	int ret;
665
	struct trace_seq *s = &iter->seq;
666 667 668 669 670

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

671 672 673 674 675 676 677
	if (trace_flags & TRACE_ITER_CONTEXT_INFO) {
		/* Absolute time */
		if (flags & TRACE_GRAPH_PRINT_ABS_TIME) {
			ret = print_graph_abs_time(iter->ts, s);
			if (!ret)
				return TRACE_TYPE_PARTIAL_LINE;
		}
678

679 680 681 682 683 684
		/* Cpu */
		if (flags & TRACE_GRAPH_PRINT_CPU) {
			ret = print_graph_cpu(s, cpu);
			if (ret == TRACE_TYPE_PARTIAL_LINE)
				return TRACE_TYPE_PARTIAL_LINE;
		}
685

686 687 688 689 690
		/* Proc */
		if (flags & TRACE_GRAPH_PRINT_PROC) {
			ret = print_graph_proc(s, pid);
			if (ret == TRACE_TYPE_PARTIAL_LINE)
				return TRACE_TYPE_PARTIAL_LINE;
691
			ret = trace_seq_puts(s, " | ");
692 693 694
			if (!ret)
				return TRACE_TYPE_PARTIAL_LINE;
		}
695
	}
696

697
	/* No overhead */
698
	ret = print_graph_duration(0, s, flags | FLAGS_FILL_START);
699 700
	if (ret != TRACE_TYPE_HANDLED)
		return ret;
701

702
	if (type == TRACE_GRAPH_ENT)
703
		ret = trace_seq_puts(s, "==========>");
704
	else
705
		ret = trace_seq_puts(s, "<==========");
706 707 708 709

	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;

710
	ret = print_graph_duration(0, s, flags | FLAGS_FILL_END);
711 712 713
	if (ret != TRACE_TYPE_HANDLED)
		return ret;

714
	ret = trace_seq_putc(s, '\n');
715 716 717 718 719

	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;
	return TRACE_TYPE_HANDLED;
}
720

721 722
enum print_line_t
trace_print_graph_duration(unsigned long long duration, struct trace_seq *s)
723 724
{
	unsigned long nsecs_rem = do_div(duration, 1000);
725 726 727 728 729 730 731 732 733
	/* 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 */
734
	ret = trace_seq_printf(s, "%s", msecs_str);
735 736 737 738 739 740 741
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;

	len = strlen(msecs_str);

	/* Print nsecs (we don't want to exceed 7 numbers) */
	if (len < 7) {
742 743 744
		size_t slen = min_t(size_t, sizeof(nsecs_str), 8UL - len);

		snprintf(nsecs_str, slen, "%03lu", nsecs_rem);
745 746 747 748 749 750
		ret = trace_seq_printf(s, ".%s", nsecs_str);
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
		len += strlen(nsecs_str);
	}

751
	ret = trace_seq_puts(s, " us ");
752 753 754 755 756
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;

	/* Print remaining spaces to fit the row's width */
	for (i = len; i < 7; i++) {
757
		ret = trace_seq_putc(s, ' ');
758 759 760
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
	}
761 762 763 764
	return TRACE_TYPE_HANDLED;
}

static enum print_line_t
765 766
print_graph_duration(unsigned long long duration, struct trace_seq *s,
		     u32 flags)
767
{
768 769
	int ret = -1;

770 771 772
	if (!(flags & TRACE_GRAPH_PRINT_DURATION) ||
	    !(trace_flags & TRACE_ITER_CONTEXT_INFO))
			return TRACE_TYPE_HANDLED;
773 774

	/* No real adata, just filling the column with spaces */
775 776
	switch (flags & TRACE_GRAPH_PRINT_FILL_MASK) {
	case FLAGS_FILL_FULL:
777
		ret = trace_seq_puts(s, "              |  ");
778
		return ret ? TRACE_TYPE_HANDLED : TRACE_TYPE_PARTIAL_LINE;
779
	case FLAGS_FILL_START:
780
		ret = trace_seq_puts(s, "  ");
781
		return ret ? TRACE_TYPE_HANDLED : TRACE_TYPE_PARTIAL_LINE;
782
	case FLAGS_FILL_END:
783
		ret = trace_seq_puts(s, " |");
784 785 786 787 788 789 790
		return ret ? TRACE_TYPE_HANDLED : TRACE_TYPE_PARTIAL_LINE;
	}

	/* Signal a overhead of time execution to the output */
	if (flags & TRACE_GRAPH_PRINT_OVERHEAD) {
		/* Duration exceeded 100 msecs */
		if (duration > 100000ULL)
791
			ret = trace_seq_puts(s, "! ");
792 793
		/* Duration exceeded 10 msecs */
		else if (duration > 10000ULL)
794
			ret = trace_seq_puts(s, "+ ");
795 796 797 798 799 800 801 802
	}

	/*
	 * The -1 means we either did not exceed the duration tresholds
	 * or we dont want to print out the overhead. Either way we need
	 * to fill out the space.
	 */
	if (ret == -1)
803
		ret = trace_seq_puts(s, "  ");
804 805 806 807

	/* Catching here any failure happenned above */
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;
808 809 810 811

	ret = trace_print_graph_duration(duration, s);
	if (ret != TRACE_TYPE_HANDLED)
		return ret;
812

813
	ret = trace_seq_puts(s, "|  ");
814 815 816
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;

817
	return TRACE_TYPE_HANDLED;
818 819 820
}

/* Case of a leaf function on its call entry */
821
static enum print_line_t
822
print_graph_entry_leaf(struct trace_iterator *iter,
823
		struct ftrace_graph_ent_entry *entry,
824 825
		struct ftrace_graph_ret_entry *ret_entry,
		struct trace_seq *s, u32 flags)
826
{
827
	struct fgraph_data *data = iter->private;
828 829 830
	struct ftrace_graph_ret *graph_ret;
	struct ftrace_graph_ent *call;
	unsigned long long duration;
831
	int ret;
832
	int i;
833

834 835 836 837
	graph_ret = &ret_entry->ret;
	call = &entry->graph_ent;
	duration = graph_ret->rettime - graph_ret->calltime;

838
	if (data) {
839
		struct fgraph_cpu_data *cpu_data;
840
		int cpu = iter->cpu;
841 842

		cpu_data = per_cpu_ptr(data->cpu_data, cpu);
843 844 845 846 847 848

		/*
		 * Comments display at + 1 to depth. Since
		 * this is a leaf function, keep the comments
		 * equal to this depth.
		 */
849 850 851 852 853
		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;
854 855
	}

856 857 858
	/* Overhead and duration */
	ret = print_graph_duration(duration, s, flags);
	if (ret == TRACE_TYPE_PARTIAL_LINE)
859
		return TRACE_TYPE_PARTIAL_LINE;
860

861 862
	/* Function */
	for (i = 0; i < call->depth * TRACE_GRAPH_INDENT; i++) {
863
		ret = trace_seq_putc(s, ' ');
864 865 866 867
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
	}

868
	ret = trace_seq_printf(s, "%ps();\n", (void *)call->func);
869 870 871 872 873 874 875
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;

	return TRACE_TYPE_HANDLED;
}

static enum print_line_t
876 877
print_graph_entry_nested(struct trace_iterator *iter,
			 struct ftrace_graph_ent_entry *entry,
878
			 struct trace_seq *s, int cpu, u32 flags)
879 880
{
	struct ftrace_graph_ent *call = &entry->graph_ent;
881 882 883 884 885
	struct fgraph_data *data = iter->private;
	int ret;
	int i;

	if (data) {
886
		struct fgraph_cpu_data *cpu_data;
887 888
		int cpu = iter->cpu;

889 890 891 892 893 894
		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;
895
	}
896

897
	/* No time */
898
	ret = print_graph_duration(0, s, flags | FLAGS_FILL_FULL);
899 900
	if (ret != TRACE_TYPE_HANDLED)
		return ret;
901

902
	/* Function */
903
	for (i = 0; i < call->depth * TRACE_GRAPH_INDENT; i++) {
904
		ret = trace_seq_putc(s, ' ');
905 906
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
907 908
	}

909
	ret = trace_seq_printf(s, "%ps() {\n", (void *)call->func);
910 911 912
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;

913 914 915 916 917
	/*
	 * we already consumed the current entry to check the next one
	 * and see if this is a leaf.
	 */
	return TRACE_TYPE_NO_CONSUME;
918 919
}

920
static enum print_line_t
921
print_graph_prologue(struct trace_iterator *iter, struct trace_seq *s,
922
		     int type, unsigned long addr, u32 flags)
923
{
924
	struct fgraph_data *data = iter->private;
925
	struct trace_entry *ent = iter->ent;
926 927
	int cpu = iter->cpu;
	int ret;
928

929
	/* Pid */
930
	if (verif_pid(s, ent->pid, cpu, data) == TRACE_TYPE_PARTIAL_LINE)
931 932
		return TRACE_TYPE_PARTIAL_LINE;

933 934
	if (type) {
		/* Interrupt */
935
		ret = print_graph_irq(iter, addr, type, cpu, ent->pid, flags);
936 937 938
		if (ret == TRACE_TYPE_PARTIAL_LINE)
			return TRACE_TYPE_PARTIAL_LINE;
	}
939

940 941 942
	if (!(trace_flags & TRACE_ITER_CONTEXT_INFO))
		return 0;

943
	/* Absolute time */
944
	if (flags & TRACE_GRAPH_PRINT_ABS_TIME) {
945 946 947 948 949
		ret = print_graph_abs_time(iter->ts, s);
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
	}

950
	/* Cpu */
951
	if (flags & TRACE_GRAPH_PRINT_CPU) {
952
		ret = print_graph_cpu(s, cpu);
953 954 955 956 957
		if (ret == TRACE_TYPE_PARTIAL_LINE)
			return TRACE_TYPE_PARTIAL_LINE;
	}

	/* Proc */
958
	if (flags & TRACE_GRAPH_PRINT_PROC) {
959
		ret = print_graph_proc(s, ent->pid);
960 961 962
		if (ret == TRACE_TYPE_PARTIAL_LINE)
			return TRACE_TYPE_PARTIAL_LINE;

963
		ret = trace_seq_puts(s, " | ");
964 965 966
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
	}
967

968 969 970 971 972 973 974
	/* 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;
	}

975 976 977
	return 0;
}

978 979 980 981 982
/*
 * Entry check for irq code
 *
 * returns 1 if
 *  - we are inside irq code
Lucas De Marchi's avatar
Lucas De Marchi committed
983
 *  - we just entered irq code
984 985 986 987 988 989 990 991 992 993
 *
 * retunns 0 if
 *  - funcgraph-interrupts option is set
 *  - we are not inside irq code
 */
static int
check_irq_entry(struct trace_iterator *iter, u32 flags,
		unsigned long addr, int depth)
{
	int cpu = iter->cpu;
994
	int *depth_irq;
995 996
	struct fgraph_data *data = iter->private;

997 998 999 1000 1001 1002 1003
	/*
	 * If we are either displaying irqs, or we got called as
	 * a graph event and private data does not exist,
	 * then we bypass the irq check.
	 */
	if ((flags & TRACE_GRAPH_PRINT_IRQS) ||
	    (!data))
1004 1005
		return 0;

1006 1007
	depth_irq = &(per_cpu_ptr(data->cpu_data, cpu)->depth_irq);

1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039
	/*
	 * We are inside the irq code
	 */
	if (*depth_irq >= 0)
		return 1;

	if ((addr < (unsigned long)__irqentry_text_start) ||
	    (addr >= (unsigned long)__irqentry_text_end))
		return 0;

	/*
	 * We are entering irq code.
	 */
	*depth_irq = depth;
	return 1;
}

/*
 * Return check for irq code
 *
 * returns 1 if
 *  - we are inside irq code
 *  - we just left irq code
 *
 * returns 0 if
 *  - funcgraph-interrupts option is set
 *  - we are not inside irq code
 */
static int
check_irq_return(struct trace_iterator *iter, u32 flags, int depth)
{
	int cpu = iter->cpu;
1040
	int *depth_irq;
1041 1042
	struct fgraph_data *data = iter->private;

1043 1044 1045 1046 1047 1048 1049
	/*
	 * If we are either displaying irqs, or we got called as
	 * a graph event and private data does not exist,
	 * then we bypass the irq check.
	 */
	if ((flags & TRACE_GRAPH_PRINT_IRQS) ||
	    (!data))
1050 1051
		return 0;

1052 1053
	depth_irq = &(per_cpu_ptr(data->cpu_data, cpu)->depth_irq);

1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079
	/*
	 * We are not inside the irq code.
	 */
	if (*depth_irq == -1)
		return 0;

	/*
	 * We are inside the irq code, and this is returning entry.
	 * Let's not trace it and clear the entry depth, since
	 * we are out of irq code.
	 *
	 * This condition ensures that we 'leave the irq code' once
	 * we are out of the entry depth. Thus protecting us from
	 * the RETURN entry loss.
	 */
	if (*depth_irq >= depth) {
		*depth_irq = -1;
		return 1;
	}

	/*
	 * We are inside the irq code, and this is not the entry.
	 */
	return 1;
}

1080 1081
static enum print_line_t
print_graph_entry(struct ftrace_graph_ent_entry *field, struct trace_seq *s,
1082
			struct trace_iterator *iter, u32 flags)
1083
{
1084
	struct fgraph_data *data = iter->private;
1085 1086
	struct ftrace_graph_ent *call = &field->graph_ent;
	struct ftrace_graph_ret_entry *leaf_ret;
1087 1088
	static enum print_line_t ret;
	int cpu = iter->cpu;
1089

1090 1091 1092
	if (check_irq_entry(iter, flags, call->func, call->depth))
		return TRACE_TYPE_HANDLED;

1093
	if (print_graph_prologue(iter, s, TRACE_GRAPH_ENT, call->func, flags))
1094 1095
		return TRACE_TYPE_PARTIAL_LINE;

1096 1097
	leaf_ret = get_return_for_leaf(iter, field);
	if (leaf_ret)
1098
		ret = print_graph_entry_leaf(iter, field, leaf_ret, s, flags);
1099
	else
1100
		ret = print_graph_entry_nested(iter, field, s, cpu, flags);
1101

1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114
	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;
1115 1116
}

1117 1118
static enum print_line_t
print_graph_return(struct ftrace_graph_ret *trace, struct trace_seq *s,
1119 1120
		   struct trace_entry *ent, struct trace_iterator *iter,
		   u32 flags)
1121
{
1122
	unsigned long long duration = trace->rettime - trace->calltime;
1123 1124 1125
	struct fgraph_data *data = iter->private;
	pid_t pid = ent->pid;
	int cpu = iter->cpu;
1126
	int func_match = 1;
1127 1128 1129
	int ret;
	int i;

1130 1131 1132
	if (check_irq_return(iter, flags, trace->depth))
		return TRACE_TYPE_HANDLED;

1133
	if (data) {
1134 1135 1136 1137
		struct fgraph_cpu_data *cpu_data;
		int cpu = iter->cpu;

		cpu_data = per_cpu_ptr(data->cpu_data, cpu);
1138 1139 1140 1141 1142 1143

		/*
		 * 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.
		 */
1144 1145 1146 1147 1148 1149 1150
		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;
		}
1151
	}
1152

1153
	if (print_graph_prologue(iter, s, 0, 0, flags))
1154 1155
		return TRACE_TYPE_PARTIAL_LINE;

1156 1157 1158
	/* Overhead and duration */
	ret = print_graph_duration(duration, s, flags);
	if (ret == TRACE_TYPE_PARTIAL_LINE)
1159
		return TRACE_TYPE_PARTIAL_LINE;
1160

1161
	/* Closing brace */
1162
	for (i = 0; i < trace->depth * TRACE_GRAPH_INDENT; i++) {
1163
		ret = trace_seq_putc(s, ' ');
1164 1165
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
1166 1167
	}

1168 1169 1170 1171
	/*
	 * 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
1172 1173
	 * belongs to, write out the function name. Always do
	 * that if the funcgraph-tail option is enabled.
1174
	 */
1175
	if (func_match && !(flags & TRACE_GRAPH_PRINT_TAIL)) {
1176
		ret = trace_seq_puts(s, "}\n");
1177 1178 1179
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
	} else {
1180
		ret = trace_seq_printf(s, "} /* %ps */\n", (void *)trace->func);
1181 1182 1183
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
	}
1184

1185
	/* Overrun */
1186
	if (flags & TRACE_GRAPH_PRINT_OVERRUN) {
1187 1188
		ret = trace_seq_printf(s, " (Overruns: %lu)\n",
					trace->overrun);
1189 1190
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;