Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid
[pandora-kernel.git] / kernel / trace / trace_irqsoff.c
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);