xref: /openbmc/linux/kernel/trace/trace_irqsoff.c (revision 9b799b78)
1 /*
2  * trace irqs off critical timings
3  *
4  * Copyright (C) 2007-2008 Steven Rostedt <srostedt@redhat.com>
5  * Copyright (C) 2008 Ingo Molnar <mingo@redhat.com>
6  *
7  * From code in the latency_tracer, that is:
8  *
9  *  Copyright (C) 2004-2006 Ingo Molnar
10  *  Copyright (C) 2004 Nadia Yvette Chambers
11  */
12 #include <linux/kallsyms.h>
13 #include <linux/uaccess.h>
14 #include <linux/module.h>
15 #include <linux/ftrace.h>
16 
17 #include "trace.h"
18 
19 static struct trace_array		*irqsoff_trace __read_mostly;
20 static int				tracer_enabled __read_mostly;
21 
22 static DEFINE_PER_CPU(int, tracing_cpu);
23 
24 static DEFINE_RAW_SPINLOCK(max_trace_lock);
25 
26 enum {
27 	TRACER_IRQS_OFF		= (1 << 1),
28 	TRACER_PREEMPT_OFF	= (1 << 2),
29 };
30 
31 static int trace_type __read_mostly;
32 
33 static int save_flags;
34 static bool function_enabled;
35 
36 static void stop_irqsoff_tracer(struct trace_array *tr, int graph);
37 static int start_irqsoff_tracer(struct trace_array *tr, int graph);
38 
39 #ifdef CONFIG_PREEMPT_TRACER
40 static inline int
41 preempt_trace(void)
42 {
43 	return ((trace_type & TRACER_PREEMPT_OFF) && preempt_count());
44 }
45 #else
46 # define preempt_trace() (0)
47 #endif
48 
49 #ifdef CONFIG_IRQSOFF_TRACER
50 static inline int
51 irq_trace(void)
52 {
53 	return ((trace_type & TRACER_IRQS_OFF) &&
54 		irqs_disabled());
55 }
56 #else
57 # define irq_trace() (0)
58 #endif
59 
60 #define TRACE_DISPLAY_GRAPH	1
61 
62 static struct tracer_opt trace_opts[] = {
63 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
64 	/* display latency trace as call graph */
65 	{ TRACER_OPT(display-graph, TRACE_DISPLAY_GRAPH) },
66 #endif
67 	{ } /* Empty entry */
68 };
69 
70 static struct tracer_flags tracer_flags = {
71 	.val  = 0,
72 	.opts = trace_opts,
73 };
74 
75 #define is_graph() (tracer_flags.val & TRACE_DISPLAY_GRAPH)
76 
77 /*
78  * Sequence count - we record it when starting a measurement and
79  * skip the latency if the sequence has changed - some other section
80  * did a maximum and could disturb our measurement with serial console
81  * printouts, etc. Truly coinciding maximum latencies should be rare
82  * and what happens together happens separately as well, so this doesn't
83  * decrease the validity of the maximum found:
84  */
85 static __cacheline_aligned_in_smp	unsigned long max_sequence;
86 
87 #ifdef CONFIG_FUNCTION_TRACER
88 /*
89  * Prologue for the preempt and irqs off function tracers.
90  *
91  * Returns 1 if it is OK to continue, and data->disabled is
92  *            incremented.
93  *         0 if the trace is to be ignored, and data->disabled
94  *            is kept the same.
95  *
96  * Note, this function is also used outside this ifdef but
97  *  inside the #ifdef of the function graph tracer below.
98  *  This is OK, since the function graph tracer is
99  *  dependent on the function tracer.
100  */
101 static int func_prolog_dec(struct trace_array *tr,
102 			   struct trace_array_cpu **data,
103 			   unsigned long *flags)
104 {
105 	long disabled;
106 	int cpu;
107 
108 	/*
109 	 * Does not matter if we preempt. We test the flags
110 	 * afterward, to see if irqs are disabled or not.
111 	 * If we preempt and get a false positive, the flags
112 	 * test will fail.
113 	 */
114 	cpu = raw_smp_processor_id();
115 	if (likely(!per_cpu(tracing_cpu, cpu)))
116 		return 0;
117 
118 	local_save_flags(*flags);
119 	/* slight chance to get a false positive on tracing_cpu */
120 	if (!irqs_disabled_flags(*flags))
121 		return 0;
122 
123 	*data = per_cpu_ptr(tr->trace_buffer.data, cpu);
124 	disabled = atomic_inc_return(&(*data)->disabled);
125 
126 	if (likely(disabled == 1))
127 		return 1;
128 
129 	atomic_dec(&(*data)->disabled);
130 
131 	return 0;
132 }
133 
134 /*
135  * irqsoff uses its own tracer function to keep the overhead down:
136  */
137 static void
138 irqsoff_tracer_call(unsigned long ip, unsigned long parent_ip,
139 		    struct ftrace_ops *op, struct pt_regs *pt_regs)
140 {
141 	struct trace_array *tr = irqsoff_trace;
142 	struct trace_array_cpu *data;
143 	unsigned long flags;
144 
145 	if (!func_prolog_dec(tr, &data, &flags))
146 		return;
147 
148 	trace_function(tr, ip, parent_ip, flags, preempt_count());
149 
150 	atomic_dec(&data->disabled);
151 }
152 #endif /* CONFIG_FUNCTION_TRACER */
153 
154 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
155 static int
156 irqsoff_set_flag(struct trace_array *tr, u32 old_flags, u32 bit, int set)
157 {
158 	int cpu;
159 
160 	if (!(bit & TRACE_DISPLAY_GRAPH))
161 		return -EINVAL;
162 
163 	if (!(is_graph() ^ set))
164 		return 0;
165 
166 	stop_irqsoff_tracer(irqsoff_trace, !set);
167 
168 	for_each_possible_cpu(cpu)
169 		per_cpu(tracing_cpu, cpu) = 0;
170 
171 	tr->max_latency = 0;
172 	tracing_reset_online_cpus(&irqsoff_trace->trace_buffer);
173 
174 	return start_irqsoff_tracer(irqsoff_trace, set);
175 }
176 
177 static int irqsoff_graph_entry(struct ftrace_graph_ent *trace)
178 {
179 	struct trace_array *tr = irqsoff_trace;
180 	struct trace_array_cpu *data;
181 	unsigned long flags;
182 	int ret;
183 	int pc;
184 
185 	if (!func_prolog_dec(tr, &data, &flags))
186 		return 0;
187 
188 	pc = preempt_count();
189 	ret = __trace_graph_entry(tr, trace, flags, pc);
190 	atomic_dec(&data->disabled);
191 
192 	return ret;
193 }
194 
195 static void irqsoff_graph_return(struct ftrace_graph_ret *trace)
196 {
197 	struct trace_array *tr = irqsoff_trace;
198 	struct trace_array_cpu *data;
199 	unsigned long flags;
200 	int pc;
201 
202 	if (!func_prolog_dec(tr, &data, &flags))
203 		return;
204 
205 	pc = preempt_count();
206 	__trace_graph_return(tr, trace, flags, pc);
207 	atomic_dec(&data->disabled);
208 }
209 
210 static void irqsoff_trace_open(struct trace_iterator *iter)
211 {
212 	if (is_graph())
213 		graph_trace_open(iter);
214 
215 }
216 
217 static void irqsoff_trace_close(struct trace_iterator *iter)
218 {
219 	if (iter->private)
220 		graph_trace_close(iter);
221 }
222 
223 #define GRAPH_TRACER_FLAGS (TRACE_GRAPH_PRINT_CPU | \
224 			    TRACE_GRAPH_PRINT_PROC | \
225 			    TRACE_GRAPH_PRINT_ABS_TIME | \
226 			    TRACE_GRAPH_PRINT_DURATION)
227 
228 static enum print_line_t irqsoff_print_line(struct trace_iterator *iter)
229 {
230 	/*
231 	 * In graph mode call the graph tracer output function,
232 	 * otherwise go with the TRACE_FN event handler
233 	 */
234 	if (is_graph())
235 		return print_graph_function_flags(iter, GRAPH_TRACER_FLAGS);
236 
237 	return TRACE_TYPE_UNHANDLED;
238 }
239 
240 static void irqsoff_print_header(struct seq_file *s)
241 {
242 	if (is_graph())
243 		print_graph_headers_flags(s, GRAPH_TRACER_FLAGS);
244 	else
245 		trace_default_header(s);
246 }
247 
248 static void
249 __trace_function(struct trace_array *tr,
250 		 unsigned long ip, unsigned long parent_ip,
251 		 unsigned long flags, int pc)
252 {
253 	if (is_graph())
254 		trace_graph_function(tr, ip, parent_ip, flags, pc);
255 	else
256 		trace_function(tr, ip, parent_ip, flags, pc);
257 }
258 
259 #else
260 #define __trace_function trace_function
261 
262 static int
263 irqsoff_set_flag(struct trace_array *tr, u32 old_flags, u32 bit, int set)
264 {
265 	return -EINVAL;
266 }
267 
268 static int irqsoff_graph_entry(struct ftrace_graph_ent *trace)
269 {
270 	return -1;
271 }
272 
273 static enum print_line_t irqsoff_print_line(struct trace_iterator *iter)
274 {
275 	return TRACE_TYPE_UNHANDLED;
276 }
277 
278 static void irqsoff_graph_return(struct ftrace_graph_ret *trace) { }
279 static void irqsoff_trace_open(struct trace_iterator *iter) { }
280 static void irqsoff_trace_close(struct trace_iterator *iter) { }
281 
282 #ifdef CONFIG_FUNCTION_TRACER
283 static void irqsoff_print_header(struct seq_file *s)
284 {
285 	trace_default_header(s);
286 }
287 #else
288 static void irqsoff_print_header(struct seq_file *s)
289 {
290 	trace_latency_header(s);
291 }
292 #endif /* CONFIG_FUNCTION_TRACER */
293 #endif /* CONFIG_FUNCTION_GRAPH_TRACER */
294 
295 /*
296  * Should this new latency be reported/recorded?
297  */
298 static int report_latency(struct trace_array *tr, cycle_t delta)
299 {
300 	if (tracing_thresh) {
301 		if (delta < tracing_thresh)
302 			return 0;
303 	} else {
304 		if (delta <= tr->max_latency)
305 			return 0;
306 	}
307 	return 1;
308 }
309 
310 static void
311 check_critical_timing(struct trace_array *tr,
312 		      struct trace_array_cpu *data,
313 		      unsigned long parent_ip,
314 		      int cpu)
315 {
316 	cycle_t T0, T1, delta;
317 	unsigned long flags;
318 	int pc;
319 
320 	T0 = data->preempt_timestamp;
321 	T1 = ftrace_now(cpu);
322 	delta = T1-T0;
323 
324 	local_save_flags(flags);
325 
326 	pc = preempt_count();
327 
328 	if (!report_latency(tr, delta))
329 		goto out;
330 
331 	raw_spin_lock_irqsave(&max_trace_lock, flags);
332 
333 	/* check if we are still the max latency */
334 	if (!report_latency(tr, delta))
335 		goto out_unlock;
336 
337 	__trace_function(tr, CALLER_ADDR0, parent_ip, flags, pc);
338 	/* Skip 5 functions to get to the irq/preempt enable function */
339 	__trace_stack(tr, flags, 5, pc);
340 
341 	if (data->critical_sequence != max_sequence)
342 		goto out_unlock;
343 
344 	data->critical_end = parent_ip;
345 
346 	if (likely(!is_tracing_stopped())) {
347 		tr->max_latency = delta;
348 		update_max_tr_single(tr, current, cpu);
349 	}
350 
351 	max_sequence++;
352 
353 out_unlock:
354 	raw_spin_unlock_irqrestore(&max_trace_lock, flags);
355 
356 out:
357 	data->critical_sequence = max_sequence;
358 	data->preempt_timestamp = ftrace_now(cpu);
359 	__trace_function(tr, CALLER_ADDR0, parent_ip, flags, pc);
360 }
361 
362 static inline void
363 start_critical_timing(unsigned long ip, unsigned long parent_ip)
364 {
365 	int cpu;
366 	struct trace_array *tr = irqsoff_trace;
367 	struct trace_array_cpu *data;
368 	unsigned long flags;
369 
370 	if (!tracer_enabled || !tracing_is_enabled())
371 		return;
372 
373 	cpu = raw_smp_processor_id();
374 
375 	if (per_cpu(tracing_cpu, cpu))
376 		return;
377 
378 	data = per_cpu_ptr(tr->trace_buffer.data, cpu);
379 
380 	if (unlikely(!data) || atomic_read(&data->disabled))
381 		return;
382 
383 	atomic_inc(&data->disabled);
384 
385 	data->critical_sequence = max_sequence;
386 	data->preempt_timestamp = ftrace_now(cpu);
387 	data->critical_start = parent_ip ? : ip;
388 
389 	local_save_flags(flags);
390 
391 	__trace_function(tr, ip, parent_ip, flags, preempt_count());
392 
393 	per_cpu(tracing_cpu, cpu) = 1;
394 
395 	atomic_dec(&data->disabled);
396 }
397 
398 static inline void
399 stop_critical_timing(unsigned long ip, unsigned long parent_ip)
400 {
401 	int cpu;
402 	struct trace_array *tr = irqsoff_trace;
403 	struct trace_array_cpu *data;
404 	unsigned long flags;
405 
406 	cpu = raw_smp_processor_id();
407 	/* Always clear the tracing cpu on stopping the trace */
408 	if (unlikely(per_cpu(tracing_cpu, cpu)))
409 		per_cpu(tracing_cpu, cpu) = 0;
410 	else
411 		return;
412 
413 	if (!tracer_enabled || !tracing_is_enabled())
414 		return;
415 
416 	data = per_cpu_ptr(tr->trace_buffer.data, cpu);
417 
418 	if (unlikely(!data) ||
419 	    !data->critical_start || atomic_read(&data->disabled))
420 		return;
421 
422 	atomic_inc(&data->disabled);
423 
424 	local_save_flags(flags);
425 	__trace_function(tr, ip, parent_ip, flags, preempt_count());
426 	check_critical_timing(tr, data, parent_ip ? : ip, cpu);
427 	data->critical_start = 0;
428 	atomic_dec(&data->disabled);
429 }
430 
431 /* start and stop critical timings used to for stoppage (in idle) */
432 void start_critical_timings(void)
433 {
434 	if (preempt_trace() || irq_trace())
435 		start_critical_timing(CALLER_ADDR0, CALLER_ADDR1);
436 }
437 EXPORT_SYMBOL_GPL(start_critical_timings);
438 
439 void stop_critical_timings(void)
440 {
441 	if (preempt_trace() || irq_trace())
442 		stop_critical_timing(CALLER_ADDR0, CALLER_ADDR1);
443 }
444 EXPORT_SYMBOL_GPL(stop_critical_timings);
445 
446 #ifdef CONFIG_IRQSOFF_TRACER
447 #ifdef CONFIG_PROVE_LOCKING
448 void time_hardirqs_on(unsigned long a0, unsigned long a1)
449 {
450 	if (!preempt_trace() && irq_trace())
451 		stop_critical_timing(a0, a1);
452 }
453 
454 void time_hardirqs_off(unsigned long a0, unsigned long a1)
455 {
456 	if (!preempt_trace() && irq_trace())
457 		start_critical_timing(a0, a1);
458 }
459 
460 #else /* !CONFIG_PROVE_LOCKING */
461 
462 /*
463  * Stubs:
464  */
465 
466 void trace_softirqs_on(unsigned long ip)
467 {
468 }
469 
470 void trace_softirqs_off(unsigned long ip)
471 {
472 }
473 
474 inline void print_irqtrace_events(struct task_struct *curr)
475 {
476 }
477 
478 /*
479  * We are only interested in hardirq on/off events:
480  */
481 void trace_hardirqs_on(void)
482 {
483 	if (!preempt_trace() && irq_trace())
484 		stop_critical_timing(CALLER_ADDR0, CALLER_ADDR1);
485 }
486 EXPORT_SYMBOL(trace_hardirqs_on);
487 
488 void trace_hardirqs_off(void)
489 {
490 	if (!preempt_trace() && irq_trace())
491 		start_critical_timing(CALLER_ADDR0, CALLER_ADDR1);
492 }
493 EXPORT_SYMBOL(trace_hardirqs_off);
494 
495 __visible void trace_hardirqs_on_caller(unsigned long caller_addr)
496 {
497 	if (!preempt_trace() && irq_trace())
498 		stop_critical_timing(CALLER_ADDR0, caller_addr);
499 }
500 EXPORT_SYMBOL(trace_hardirqs_on_caller);
501 
502 __visible void trace_hardirqs_off_caller(unsigned long caller_addr)
503 {
504 	if (!preempt_trace() && irq_trace())
505 		start_critical_timing(CALLER_ADDR0, caller_addr);
506 }
507 EXPORT_SYMBOL(trace_hardirqs_off_caller);
508 
509 #endif /* CONFIG_PROVE_LOCKING */
510 #endif /*  CONFIG_IRQSOFF_TRACER */
511 
512 #ifdef CONFIG_PREEMPT_TRACER
513 void trace_preempt_on(unsigned long a0, unsigned long a1)
514 {
515 	if (preempt_trace() && !irq_trace())
516 		stop_critical_timing(a0, a1);
517 }
518 
519 void trace_preempt_off(unsigned long a0, unsigned long a1)
520 {
521 	if (preempt_trace() && !irq_trace())
522 		start_critical_timing(a0, a1);
523 }
524 #endif /* CONFIG_PREEMPT_TRACER */
525 
526 static int register_irqsoff_function(struct trace_array *tr, int graph, int set)
527 {
528 	int ret;
529 
530 	/* 'set' is set if TRACE_ITER_FUNCTION is about to be set */
531 	if (function_enabled || (!set && !(trace_flags & TRACE_ITER_FUNCTION)))
532 		return 0;
533 
534 	if (graph)
535 		ret = register_ftrace_graph(&irqsoff_graph_return,
536 					    &irqsoff_graph_entry);
537 	else
538 		ret = register_ftrace_function(tr->ops);
539 
540 	if (!ret)
541 		function_enabled = true;
542 
543 	return ret;
544 }
545 
546 static void unregister_irqsoff_function(struct trace_array *tr, int graph)
547 {
548 	if (!function_enabled)
549 		return;
550 
551 	if (graph)
552 		unregister_ftrace_graph();
553 	else
554 		unregister_ftrace_function(tr->ops);
555 
556 	function_enabled = false;
557 }
558 
559 static void irqsoff_function_set(struct trace_array *tr, int set)
560 {
561 	if (set)
562 		register_irqsoff_function(tr, is_graph(), 1);
563 	else
564 		unregister_irqsoff_function(tr, is_graph());
565 }
566 
567 static int irqsoff_flag_changed(struct trace_array *tr, u32 mask, int set)
568 {
569 	struct tracer *tracer = tr->current_trace;
570 
571 	if (mask & TRACE_ITER_FUNCTION)
572 		irqsoff_function_set(tr, set);
573 
574 	return trace_keep_overwrite(tracer, mask, set);
575 }
576 
577 static int start_irqsoff_tracer(struct trace_array *tr, int graph)
578 {
579 	int ret;
580 
581 	ret = register_irqsoff_function(tr, graph, 0);
582 
583 	if (!ret && tracing_is_enabled())
584 		tracer_enabled = 1;
585 	else
586 		tracer_enabled = 0;
587 
588 	return ret;
589 }
590 
591 static void stop_irqsoff_tracer(struct trace_array *tr, int graph)
592 {
593 	tracer_enabled = 0;
594 
595 	unregister_irqsoff_function(tr, graph);
596 }
597 
598 static bool irqsoff_busy;
599 
600 static int __irqsoff_tracer_init(struct trace_array *tr)
601 {
602 	if (irqsoff_busy)
603 		return -EBUSY;
604 
605 	save_flags = trace_flags;
606 
607 	/* non overwrite screws up the latency tracers */
608 	set_tracer_flag(tr, TRACE_ITER_OVERWRITE, 1);
609 	set_tracer_flag(tr, TRACE_ITER_LATENCY_FMT, 1);
610 
611 	tr->max_latency = 0;
612 	irqsoff_trace = tr;
613 	/* make sure that the tracer is visible */
614 	smp_wmb();
615 	tracing_reset_online_cpus(&tr->trace_buffer);
616 
617 	ftrace_init_array_ops(tr, irqsoff_tracer_call);
618 
619 	/* Only toplevel instance supports graph tracing */
620 	if (start_irqsoff_tracer(tr, (tr->flags & TRACE_ARRAY_FL_GLOBAL &&
621 				      is_graph())))
622 		printk(KERN_ERR "failed to start irqsoff tracer\n");
623 
624 	irqsoff_busy = true;
625 	return 0;
626 }
627 
628 static void irqsoff_tracer_reset(struct trace_array *tr)
629 {
630 	int lat_flag = save_flags & TRACE_ITER_LATENCY_FMT;
631 	int overwrite_flag = save_flags & TRACE_ITER_OVERWRITE;
632 
633 	stop_irqsoff_tracer(tr, is_graph());
634 
635 	set_tracer_flag(tr, TRACE_ITER_LATENCY_FMT, lat_flag);
636 	set_tracer_flag(tr, TRACE_ITER_OVERWRITE, overwrite_flag);
637 	ftrace_reset_array_ops(tr);
638 
639 	irqsoff_busy = false;
640 }
641 
642 static void irqsoff_tracer_start(struct trace_array *tr)
643 {
644 	tracer_enabled = 1;
645 }
646 
647 static void irqsoff_tracer_stop(struct trace_array *tr)
648 {
649 	tracer_enabled = 0;
650 }
651 
652 #ifdef CONFIG_IRQSOFF_TRACER
653 static int irqsoff_tracer_init(struct trace_array *tr)
654 {
655 	trace_type = TRACER_IRQS_OFF;
656 
657 	return __irqsoff_tracer_init(tr);
658 }
659 static struct tracer irqsoff_tracer __read_mostly =
660 {
661 	.name		= "irqsoff",
662 	.init		= irqsoff_tracer_init,
663 	.reset		= irqsoff_tracer_reset,
664 	.start		= irqsoff_tracer_start,
665 	.stop		= irqsoff_tracer_stop,
666 	.print_max	= true,
667 	.print_header   = irqsoff_print_header,
668 	.print_line     = irqsoff_print_line,
669 	.flags		= &tracer_flags,
670 	.set_flag	= irqsoff_set_flag,
671 	.flag_changed	= irqsoff_flag_changed,
672 #ifdef CONFIG_FTRACE_SELFTEST
673 	.selftest    = trace_selftest_startup_irqsoff,
674 #endif
675 	.open           = irqsoff_trace_open,
676 	.close          = irqsoff_trace_close,
677 	.allow_instances = true,
678 	.use_max_tr	= true,
679 };
680 # define register_irqsoff(trace) register_tracer(&trace)
681 #else
682 # define register_irqsoff(trace) do { } while (0)
683 #endif
684 
685 #ifdef CONFIG_PREEMPT_TRACER
686 static int preemptoff_tracer_init(struct trace_array *tr)
687 {
688 	trace_type = TRACER_PREEMPT_OFF;
689 
690 	return __irqsoff_tracer_init(tr);
691 }
692 
693 static struct tracer preemptoff_tracer __read_mostly =
694 {
695 	.name		= "preemptoff",
696 	.init		= preemptoff_tracer_init,
697 	.reset		= irqsoff_tracer_reset,
698 	.start		= irqsoff_tracer_start,
699 	.stop		= irqsoff_tracer_stop,
700 	.print_max	= true,
701 	.print_header   = irqsoff_print_header,
702 	.print_line     = irqsoff_print_line,
703 	.flags		= &tracer_flags,
704 	.set_flag	= irqsoff_set_flag,
705 	.flag_changed	= irqsoff_flag_changed,
706 #ifdef CONFIG_FTRACE_SELFTEST
707 	.selftest    = trace_selftest_startup_preemptoff,
708 #endif
709 	.open		= irqsoff_trace_open,
710 	.close		= irqsoff_trace_close,
711 	.allow_instances = true,
712 	.use_max_tr	= true,
713 };
714 # define register_preemptoff(trace) register_tracer(&trace)
715 #else
716 # define register_preemptoff(trace) do { } while (0)
717 #endif
718 
719 #if defined(CONFIG_IRQSOFF_TRACER) && \
720 	defined(CONFIG_PREEMPT_TRACER)
721 
722 static int preemptirqsoff_tracer_init(struct trace_array *tr)
723 {
724 	trace_type = TRACER_IRQS_OFF | TRACER_PREEMPT_OFF;
725 
726 	return __irqsoff_tracer_init(tr);
727 }
728 
729 static struct tracer preemptirqsoff_tracer __read_mostly =
730 {
731 	.name		= "preemptirqsoff",
732 	.init		= preemptirqsoff_tracer_init,
733 	.reset		= irqsoff_tracer_reset,
734 	.start		= irqsoff_tracer_start,
735 	.stop		= irqsoff_tracer_stop,
736 	.print_max	= true,
737 	.print_header   = irqsoff_print_header,
738 	.print_line     = irqsoff_print_line,
739 	.flags		= &tracer_flags,
740 	.set_flag	= irqsoff_set_flag,
741 	.flag_changed	= irqsoff_flag_changed,
742 #ifdef CONFIG_FTRACE_SELFTEST
743 	.selftest    = trace_selftest_startup_preemptirqsoff,
744 #endif
745 	.open		= irqsoff_trace_open,
746 	.close		= irqsoff_trace_close,
747 	.allow_instances = true,
748 	.use_max_tr	= true,
749 };
750 
751 # define register_preemptirqsoff(trace) register_tracer(&trace)
752 #else
753 # define register_preemptirqsoff(trace) do { } while (0)
754 #endif
755 
756 __init static int init_irqsoff_tracer(void)
757 {
758 	register_irqsoff(irqsoff_tracer);
759 	register_preemptoff(preemptoff_tracer);
760 	register_preemptirqsoff(preemptirqsoff_tracer);
761 
762 	return 0;
763 }
764 core_initcall(init_irqsoff_tracer);
765