Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
[linux-2.6] / kernel / trace / trace_irqsoff.c
1 /*
2  * trace irqs off criticall 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 William Lee Irwin III
11  */
12 #include <linux/kallsyms.h>
13 #include <linux/debugfs.h>
14 #include <linux/uaccess.h>
15 #include <linux/module.h>
16 #include <linux/ftrace.h>
17 #include <linux/fs.h>
18
19 #include "trace.h"
20
21 static struct trace_array               *irqsoff_trace __read_mostly;
22 static int                              tracer_enabled __read_mostly;
23
24 static DEFINE_PER_CPU(int, tracing_cpu);
25
26 static DEFINE_SPINLOCK(max_trace_lock);
27
28 enum {
29         TRACER_IRQS_OFF         = (1 << 1),
30         TRACER_PREEMPT_OFF      = (1 << 2),
31 };
32
33 static int trace_type __read_mostly;
34
35 #ifdef CONFIG_PREEMPT_TRACER
36 static inline int
37 preempt_trace(void)
38 {
39         return ((trace_type & TRACER_PREEMPT_OFF) && preempt_count());
40 }
41 #else
42 # define preempt_trace() (0)
43 #endif
44
45 #ifdef CONFIG_IRQSOFF_TRACER
46 static inline int
47 irq_trace(void)
48 {
49         return ((trace_type & TRACER_IRQS_OFF) &&
50                 irqs_disabled());
51 }
52 #else
53 # define irq_trace() (0)
54 #endif
55
56 /*
57  * Sequence count - we record it when starting a measurement and
58  * skip the latency if the sequence has changed - some other section
59  * did a maximum and could disturb our measurement with serial console
60  * printouts, etc. Truly coinciding maximum latencies should be rare
61  * and what happens together happens separately as well, so this doesnt
62  * decrease the validity of the maximum found:
63  */
64 static __cacheline_aligned_in_smp       unsigned long max_sequence;
65
66 #ifdef CONFIG_FUNCTION_TRACER
67 /*
68  * irqsoff uses its own tracer function to keep the overhead down:
69  */
70 static void
71 irqsoff_tracer_call(unsigned long ip, unsigned long parent_ip)
72 {
73         struct trace_array *tr = irqsoff_trace;
74         struct trace_array_cpu *data;
75         unsigned long flags;
76         long disabled;
77         int cpu;
78
79         /*
80          * Does not matter if we preempt. We test the flags
81          * afterward, to see if irqs are disabled or not.
82          * If we preempt and get a false positive, the flags
83          * test will fail.
84          */
85         cpu = raw_smp_processor_id();
86         if (likely(!per_cpu(tracing_cpu, cpu)))
87                 return;
88
89         local_save_flags(flags);
90         /* slight chance to get a false positive on tracing_cpu */
91         if (!irqs_disabled_flags(flags))
92                 return;
93
94         data = tr->data[cpu];
95         disabled = atomic_inc_return(&data->disabled);
96
97         if (likely(disabled == 1))
98                 trace_function(tr, data, ip, parent_ip, flags, preempt_count());
99
100         atomic_dec(&data->disabled);
101 }
102
103 static struct ftrace_ops trace_ops __read_mostly =
104 {
105         .func = irqsoff_tracer_call,
106 };
107 #endif /* CONFIG_FUNCTION_TRACER */
108
109 /*
110  * Should this new latency be reported/recorded?
111  */
112 static int report_latency(cycle_t delta)
113 {
114         if (tracing_thresh) {
115                 if (delta < tracing_thresh)
116                         return 0;
117         } else {
118                 if (delta <= tracing_max_latency)
119                         return 0;
120         }
121         return 1;
122 }
123
124 static void
125 check_critical_timing(struct trace_array *tr,
126                       struct trace_array_cpu *data,
127                       unsigned long parent_ip,
128                       int cpu)
129 {
130         unsigned long latency, t0, t1;
131         cycle_t T0, T1, delta;
132         unsigned long flags;
133         int pc;
134
135         /*
136          * usecs conversion is slow so we try to delay the conversion
137          * as long as possible:
138          */
139         T0 = data->preempt_timestamp;
140         T1 = ftrace_now(cpu);
141         delta = T1-T0;
142
143         local_save_flags(flags);
144
145         pc = preempt_count();
146
147         if (!report_latency(delta))
148                 goto out;
149
150         spin_lock_irqsave(&max_trace_lock, flags);
151
152         /* check if we are still the max latency */
153         if (!report_latency(delta))
154                 goto out_unlock;
155
156         trace_function(tr, data, CALLER_ADDR0, parent_ip, flags, pc);
157
158         latency = nsecs_to_usecs(delta);
159
160         if (data->critical_sequence != max_sequence)
161                 goto out_unlock;
162
163         tracing_max_latency = delta;
164         t0 = nsecs_to_usecs(T0);
165         t1 = nsecs_to_usecs(T1);
166
167         data->critical_end = parent_ip;
168
169         update_max_tr_single(tr, current, cpu);
170
171         max_sequence++;
172
173 out_unlock:
174         spin_unlock_irqrestore(&max_trace_lock, flags);
175
176 out:
177         data->critical_sequence = max_sequence;
178         data->preempt_timestamp = ftrace_now(cpu);
179         tracing_reset(tr, cpu);
180         trace_function(tr, data, CALLER_ADDR0, parent_ip, flags, pc);
181 }
182
183 static inline void
184 start_critical_timing(unsigned long ip, unsigned long parent_ip)
185 {
186         int cpu;
187         struct trace_array *tr = irqsoff_trace;
188         struct trace_array_cpu *data;
189         unsigned long flags;
190
191         if (likely(!tracer_enabled))
192                 return;
193
194         cpu = raw_smp_processor_id();
195
196         if (per_cpu(tracing_cpu, cpu))
197                 return;
198
199         data = tr->data[cpu];
200
201         if (unlikely(!data) || atomic_read(&data->disabled))
202                 return;
203
204         atomic_inc(&data->disabled);
205
206         data->critical_sequence = max_sequence;
207         data->preempt_timestamp = ftrace_now(cpu);
208         data->critical_start = parent_ip ? : ip;
209         tracing_reset(tr, cpu);
210
211         local_save_flags(flags);
212
213         trace_function(tr, data, ip, parent_ip, flags, preempt_count());
214
215         per_cpu(tracing_cpu, cpu) = 1;
216
217         atomic_dec(&data->disabled);
218 }
219
220 static inline void
221 stop_critical_timing(unsigned long ip, unsigned long parent_ip)
222 {
223         int cpu;
224         struct trace_array *tr = irqsoff_trace;
225         struct trace_array_cpu *data;
226         unsigned long flags;
227
228         cpu = raw_smp_processor_id();
229         /* Always clear the tracing cpu on stopping the trace */
230         if (unlikely(per_cpu(tracing_cpu, cpu)))
231                 per_cpu(tracing_cpu, cpu) = 0;
232         else
233                 return;
234
235         if (!tracer_enabled)
236                 return;
237
238         data = tr->data[cpu];
239
240         if (unlikely(!data) ||
241             !data->critical_start || atomic_read(&data->disabled))
242                 return;
243
244         atomic_inc(&data->disabled);
245
246         local_save_flags(flags);
247         trace_function(tr, data, ip, parent_ip, flags, preempt_count());
248         check_critical_timing(tr, data, parent_ip ? : ip, cpu);
249         data->critical_start = 0;
250         atomic_dec(&data->disabled);
251 }
252
253 /* start and stop critical timings used to for stoppage (in idle) */
254 void start_critical_timings(void)
255 {
256         if (preempt_trace() || irq_trace())
257                 start_critical_timing(CALLER_ADDR0, CALLER_ADDR1);
258 }
259 EXPORT_SYMBOL_GPL(start_critical_timings);
260
261 void stop_critical_timings(void)
262 {
263         if (preempt_trace() || irq_trace())
264                 stop_critical_timing(CALLER_ADDR0, CALLER_ADDR1);
265 }
266 EXPORT_SYMBOL_GPL(stop_critical_timings);
267
268 #ifdef CONFIG_IRQSOFF_TRACER
269 #ifdef CONFIG_PROVE_LOCKING
270 void time_hardirqs_on(unsigned long a0, unsigned long a1)
271 {
272         if (!preempt_trace() && irq_trace())
273                 stop_critical_timing(a0, a1);
274 }
275
276 void time_hardirqs_off(unsigned long a0, unsigned long a1)
277 {
278         if (!preempt_trace() && irq_trace())
279                 start_critical_timing(a0, a1);
280 }
281
282 #else /* !CONFIG_PROVE_LOCKING */
283
284 /*
285  * Stubs:
286  */
287
288 void early_boot_irqs_off(void)
289 {
290 }
291
292 void early_boot_irqs_on(void)
293 {
294 }
295
296 void trace_softirqs_on(unsigned long ip)
297 {
298 }
299
300 void trace_softirqs_off(unsigned long ip)
301 {
302 }
303
304 inline void print_irqtrace_events(struct task_struct *curr)
305 {
306 }
307
308 /*
309  * We are only interested in hardirq on/off events:
310  */
311 void trace_hardirqs_on(void)
312 {
313         if (!preempt_trace() && irq_trace())
314                 stop_critical_timing(CALLER_ADDR0, CALLER_ADDR1);
315 }
316 EXPORT_SYMBOL(trace_hardirqs_on);
317
318 void trace_hardirqs_off(void)
319 {
320         if (!preempt_trace() && irq_trace())
321                 start_critical_timing(CALLER_ADDR0, CALLER_ADDR1);
322 }
323 EXPORT_SYMBOL(trace_hardirqs_off);
324
325 void trace_hardirqs_on_caller(unsigned long caller_addr)
326 {
327         if (!preempt_trace() && irq_trace())
328                 stop_critical_timing(CALLER_ADDR0, caller_addr);
329 }
330 EXPORT_SYMBOL(trace_hardirqs_on_caller);
331
332 void trace_hardirqs_off_caller(unsigned long caller_addr)
333 {
334         if (!preempt_trace() && irq_trace())
335                 start_critical_timing(CALLER_ADDR0, caller_addr);
336 }
337 EXPORT_SYMBOL(trace_hardirqs_off_caller);
338
339 #endif /* CONFIG_PROVE_LOCKING */
340 #endif /*  CONFIG_IRQSOFF_TRACER */
341
342 #ifdef CONFIG_PREEMPT_TRACER
343 void trace_preempt_on(unsigned long a0, unsigned long a1)
344 {
345         if (preempt_trace())
346                 stop_critical_timing(a0, a1);
347 }
348
349 void trace_preempt_off(unsigned long a0, unsigned long a1)
350 {
351         if (preempt_trace())
352                 start_critical_timing(a0, a1);
353 }
354 #endif /* CONFIG_PREEMPT_TRACER */
355
356 /*
357  * save_tracer_enabled is used to save the state of the tracer_enabled
358  * variable when we disable it when we open a trace output file.
359  */
360 static int save_tracer_enabled;
361
362 static void start_irqsoff_tracer(struct trace_array *tr)
363 {
364         register_ftrace_function(&trace_ops);
365         if (tracing_is_enabled()) {
366                 tracer_enabled = 1;
367                 save_tracer_enabled = 1;
368         } else {
369                 tracer_enabled = 0;
370                 save_tracer_enabled = 0;
371         }
372 }
373
374 static void stop_irqsoff_tracer(struct trace_array *tr)
375 {
376         tracer_enabled = 0;
377         save_tracer_enabled = 0;
378         unregister_ftrace_function(&trace_ops);
379 }
380
381 static void __irqsoff_tracer_init(struct trace_array *tr)
382 {
383         tracing_max_latency = 0;
384         irqsoff_trace = tr;
385         /* make sure that the tracer is visible */
386         smp_wmb();
387         start_irqsoff_tracer(tr);
388 }
389
390 static void irqsoff_tracer_reset(struct trace_array *tr)
391 {
392         stop_irqsoff_tracer(tr);
393 }
394
395 static void irqsoff_tracer_start(struct trace_array *tr)
396 {
397         tracer_enabled = 1;
398         save_tracer_enabled = 1;
399 }
400
401 static void irqsoff_tracer_stop(struct trace_array *tr)
402 {
403         tracer_enabled = 0;
404         save_tracer_enabled = 0;
405 }
406
407 static void irqsoff_tracer_open(struct trace_iterator *iter)
408 {
409         /* stop the trace while dumping */
410         tracer_enabled = 0;
411 }
412
413 static void irqsoff_tracer_close(struct trace_iterator *iter)
414 {
415         /* restart tracing */
416         tracer_enabled = save_tracer_enabled;
417 }
418
419 #ifdef CONFIG_IRQSOFF_TRACER
420 static int irqsoff_tracer_init(struct trace_array *tr)
421 {
422         trace_type = TRACER_IRQS_OFF;
423
424         __irqsoff_tracer_init(tr);
425         return 0;
426 }
427 static struct tracer irqsoff_tracer __read_mostly =
428 {
429         .name           = "irqsoff",
430         .init           = irqsoff_tracer_init,
431         .reset          = irqsoff_tracer_reset,
432         .start          = irqsoff_tracer_start,
433         .stop           = irqsoff_tracer_stop,
434         .open           = irqsoff_tracer_open,
435         .close          = irqsoff_tracer_close,
436         .print_max      = 1,
437 #ifdef CONFIG_FTRACE_SELFTEST
438         .selftest    = trace_selftest_startup_irqsoff,
439 #endif
440 };
441 # define register_irqsoff(trace) register_tracer(&trace)
442 #else
443 # define register_irqsoff(trace) do { } while (0)
444 #endif
445
446 #ifdef CONFIG_PREEMPT_TRACER
447 static int preemptoff_tracer_init(struct trace_array *tr)
448 {
449         trace_type = TRACER_PREEMPT_OFF;
450
451         __irqsoff_tracer_init(tr);
452         return 0;
453 }
454
455 static struct tracer preemptoff_tracer __read_mostly =
456 {
457         .name           = "preemptoff",
458         .init           = preemptoff_tracer_init,
459         .reset          = irqsoff_tracer_reset,
460         .start          = irqsoff_tracer_start,
461         .stop           = irqsoff_tracer_stop,
462         .open           = irqsoff_tracer_open,
463         .close          = irqsoff_tracer_close,
464         .print_max      = 1,
465 #ifdef CONFIG_FTRACE_SELFTEST
466         .selftest    = trace_selftest_startup_preemptoff,
467 #endif
468 };
469 # define register_preemptoff(trace) register_tracer(&trace)
470 #else
471 # define register_preemptoff(trace) do { } while (0)
472 #endif
473
474 #if defined(CONFIG_IRQSOFF_TRACER) && \
475         defined(CONFIG_PREEMPT_TRACER)
476
477 static int preemptirqsoff_tracer_init(struct trace_array *tr)
478 {
479         trace_type = TRACER_IRQS_OFF | TRACER_PREEMPT_OFF;
480
481         __irqsoff_tracer_init(tr);
482         return 0;
483 }
484
485 static struct tracer preemptirqsoff_tracer __read_mostly =
486 {
487         .name           = "preemptirqsoff",
488         .init           = preemptirqsoff_tracer_init,
489         .reset          = irqsoff_tracer_reset,
490         .start          = irqsoff_tracer_start,
491         .stop           = irqsoff_tracer_stop,
492         .open           = irqsoff_tracer_open,
493         .close          = irqsoff_tracer_close,
494         .print_max      = 1,
495 #ifdef CONFIG_FTRACE_SELFTEST
496         .selftest    = trace_selftest_startup_preemptirqsoff,
497 #endif
498 };
499
500 # define register_preemptirqsoff(trace) register_tracer(&trace)
501 #else
502 # define register_preemptirqsoff(trace) do { } while (0)
503 #endif
504
505 __init static int init_irqsoff_tracer(void)
506 {
507         register_irqsoff(irqsoff_tracer);
508         register_preemptoff(preemptoff_tracer);
509         register_preemptirqsoff(preemptirqsoff_tracer);
510
511         return 0;
512 }
513 device_initcall(init_irqsoff_tracer);