2 #ifndef _LINUX_FTRACE_EVENT_H
3 #define _LINUX_FTRACE_EVENT_H
5 #include <linux/ring_buffer.h>
6 #include <linux/trace_seq.h>
7 #include <linux/percpu.h>
8 #include <linux/hardirq.h>
9 #include <linux/perf_event.h>
10 #include <linux/tracepoint.h>
17 struct trace_print_flags
{
22 struct trace_print_flags_u64
{
23 unsigned long long mask
;
27 const char *ftrace_print_flags_seq(struct trace_seq
*p
, const char *delim
,
29 const struct trace_print_flags
*flag_array
);
31 const char *ftrace_print_symbols_seq(struct trace_seq
*p
, unsigned long val
,
32 const struct trace_print_flags
*symbol_array
);
34 #if BITS_PER_LONG == 32
35 const char *ftrace_print_symbols_seq_u64(struct trace_seq
*p
,
36 unsigned long long val
,
37 const struct trace_print_flags_u64
41 const char *ftrace_print_bitmask_seq(struct trace_seq
*p
, void *bitmask_ptr
,
42 unsigned int bitmask_size
);
44 const char *ftrace_print_hex_seq(struct trace_seq
*p
,
45 const unsigned char *buf
, int len
);
47 struct trace_iterator
;
50 int ftrace_raw_output_prep(struct trace_iterator
*iter
,
51 struct trace_event
*event
);
54 * The trace entry - the most basic unit of tracing. This is what
55 * is printed in the end as a single line in the trace output, such as:
57 * bash-15816 [01] 235.197585: idle_cpu <- irq_enter
62 unsigned char preempt_count
;
66 #define FTRACE_MAX_EVENT \
67 ((1 << (sizeof(((struct trace_entry *)0)->type) * 8)) - 1)
70 * Trace iterator - used by printout routines who present trace
71 * results to users and which routines might sleep, etc:
73 struct trace_iterator
{
74 struct trace_array
*tr
;
76 struct trace_buffer
*trace_buffer
;
80 struct ring_buffer_iter
**buffer_iter
;
81 unsigned long iter_flags
;
83 /* trace_seq for __print_flags() and __print_symbolic() etc. */
84 struct trace_seq tmp_seq
;
86 cpumask_var_t started
;
88 /* it's true when current open file is snapshot */
91 /* The below is zeroed out in pipe_read */
93 struct trace_entry
*ent
;
94 unsigned long lost_events
;
103 /* All new field here will be zeroed out in pipe_read */
106 enum trace_iter_flags
{
107 TRACE_FILE_LAT_FMT
= 1,
108 TRACE_FILE_ANNOTATE
= 2,
109 TRACE_FILE_TIME_IN_NS
= 4,
113 typedef enum print_line_t (*trace_print_func
)(struct trace_iterator
*iter
,
114 int flags
, struct trace_event
*event
);
116 struct trace_event_functions
{
117 trace_print_func trace
;
118 trace_print_func raw
;
119 trace_print_func hex
;
120 trace_print_func binary
;
124 struct hlist_node node
;
125 struct list_head list
;
127 struct trace_event_functions
*funcs
;
130 extern int register_ftrace_event(struct trace_event
*event
);
131 extern int unregister_ftrace_event(struct trace_event
*event
);
133 /* Return values for print_line callback */
135 TRACE_TYPE_PARTIAL_LINE
= 0, /* Retry after flushing the seq */
136 TRACE_TYPE_HANDLED
= 1,
137 TRACE_TYPE_UNHANDLED
= 2, /* Relay to other output functions */
138 TRACE_TYPE_NO_CONSUME
= 3 /* Handled but ask to not consume */
141 void tracing_generic_entry_update(struct trace_entry
*entry
,
144 struct ftrace_event_file
;
146 struct ring_buffer_event
*
147 trace_event_buffer_lock_reserve(struct ring_buffer
**current_buffer
,
148 struct ftrace_event_file
*ftrace_file
,
149 int type
, unsigned long len
,
150 unsigned long flags
, int pc
);
151 struct ring_buffer_event
*
152 trace_current_buffer_lock_reserve(struct ring_buffer
**current_buffer
,
153 int type
, unsigned long len
,
154 unsigned long flags
, int pc
);
155 void trace_current_buffer_unlock_commit(struct ring_buffer
*buffer
,
156 struct ring_buffer_event
*event
,
157 unsigned long flags
, int pc
);
158 void trace_buffer_unlock_commit(struct ring_buffer
*buffer
,
159 struct ring_buffer_event
*event
,
160 unsigned long flags
, int pc
);
161 void trace_buffer_unlock_commit_regs(struct ring_buffer
*buffer
,
162 struct ring_buffer_event
*event
,
163 unsigned long flags
, int pc
,
164 struct pt_regs
*regs
);
165 void trace_current_buffer_discard_commit(struct ring_buffer
*buffer
,
166 struct ring_buffer_event
*event
);
168 void tracing_record_cmdline(struct task_struct
*tsk
);
170 int ftrace_output_call(struct trace_iterator
*iter
, char *name
, char *fmt
, ...);
176 TRACE_REG_UNREGISTER
,
177 #ifdef CONFIG_PERF_EVENTS
178 TRACE_REG_PERF_REGISTER
,
179 TRACE_REG_PERF_UNREGISTER
,
181 TRACE_REG_PERF_CLOSE
,
187 struct ftrace_event_call
;
189 struct ftrace_event_class
{
192 #ifdef CONFIG_PERF_EVENTS
195 int (*reg
)(struct ftrace_event_call
*event
,
196 enum trace_reg type
, void *data
);
197 int (*define_fields
)(struct ftrace_event_call
*);
198 struct list_head
*(*get_fields
)(struct ftrace_event_call
*);
199 struct list_head fields
;
200 int (*raw_init
)(struct ftrace_event_call
*);
203 extern int ftrace_event_reg(struct ftrace_event_call
*event
,
204 enum trace_reg type
, void *data
);
206 int ftrace_output_event(struct trace_iterator
*iter
, struct ftrace_event_call
*event
,
209 int ftrace_event_define_field(struct ftrace_event_call
*call
,
210 char *type
, int len
, char *item
, int offset
,
211 int field_size
, int sign
, int filter
);
213 struct ftrace_event_buffer
{
214 struct ring_buffer
*buffer
;
215 struct ring_buffer_event
*event
;
216 struct ftrace_event_file
*ftrace_file
;
222 void *ftrace_event_buffer_reserve(struct ftrace_event_buffer
*fbuffer
,
223 struct ftrace_event_file
*ftrace_file
,
226 void ftrace_event_buffer_commit(struct ftrace_event_buffer
*fbuffer
);
228 int ftrace_event_define_field(struct ftrace_event_call
*call
,
229 char *type
, int len
, char *item
, int offset
,
230 int field_size
, int sign
, int filter
);
233 TRACE_EVENT_FL_FILTERED_BIT
,
234 TRACE_EVENT_FL_CAP_ANY_BIT
,
235 TRACE_EVENT_FL_NO_SET_FILTER_BIT
,
236 TRACE_EVENT_FL_IGNORE_ENABLE_BIT
,
237 TRACE_EVENT_FL_WAS_ENABLED_BIT
,
238 TRACE_EVENT_FL_USE_CALL_FILTER_BIT
,
239 TRACE_EVENT_FL_TRACEPOINT_BIT
,
244 * FILTERED - The event has a filter attached
245 * CAP_ANY - Any user can enable for perf
246 * NO_SET_FILTER - Set when filter has error and is to be ignored
247 * IGNORE_ENABLE - For ftrace internal events, do not enable with debugfs file
248 * WAS_ENABLED - Set and stays set when an event was ever enabled
249 * (used for module unloading, if a module event is enabled,
250 * it is best to clear the buffers that used it).
251 * USE_CALL_FILTER - For ftrace internal events, don't use file filter
252 * TRACEPOINT - Event is a tracepoint
255 TRACE_EVENT_FL_FILTERED
= (1 << TRACE_EVENT_FL_FILTERED_BIT
),
256 TRACE_EVENT_FL_CAP_ANY
= (1 << TRACE_EVENT_FL_CAP_ANY_BIT
),
257 TRACE_EVENT_FL_NO_SET_FILTER
= (1 << TRACE_EVENT_FL_NO_SET_FILTER_BIT
),
258 TRACE_EVENT_FL_IGNORE_ENABLE
= (1 << TRACE_EVENT_FL_IGNORE_ENABLE_BIT
),
259 TRACE_EVENT_FL_WAS_ENABLED
= (1 << TRACE_EVENT_FL_WAS_ENABLED_BIT
),
260 TRACE_EVENT_FL_USE_CALL_FILTER
= (1 << TRACE_EVENT_FL_USE_CALL_FILTER_BIT
),
261 TRACE_EVENT_FL_TRACEPOINT
= (1 << TRACE_EVENT_FL_TRACEPOINT_BIT
),
264 struct ftrace_event_call
{
265 struct list_head list
;
266 struct ftrace_event_class
*class;
269 /* Set TRACE_EVENT_FL_TRACEPOINT flag when using "tp" */
270 struct tracepoint
*tp
;
272 struct trace_event event
;
273 const char *print_fmt
;
274 struct event_filter
*filter
;
275 struct list_head
*files
;
279 * bit 0: filter_active
280 * bit 1: allow trace by non root (cap any)
281 * bit 2: failed to apply filter
282 * bit 3: ftrace internal event (do not enable)
283 * bit 4: Event was enabled by module
284 * bit 5: use call filter rather than file filter
285 * bit 6: Event is a tracepoint
287 int flags
; /* static flags of different events */
289 #ifdef CONFIG_PERF_EVENTS
291 struct hlist_head __percpu
*perf_events
;
293 int (*perf_perm
)(struct ftrace_event_call
*,
294 struct perf_event
*);
298 static inline const char *
299 ftrace_event_name(struct ftrace_event_call
*call
)
301 if (call
->flags
& TRACE_EVENT_FL_TRACEPOINT
)
302 return call
->tp
? call
->tp
->name
: NULL
;
308 struct ftrace_subsystem_dir
;
311 FTRACE_EVENT_FL_ENABLED_BIT
,
312 FTRACE_EVENT_FL_RECORDED_CMD_BIT
,
313 FTRACE_EVENT_FL_FILTERED_BIT
,
314 FTRACE_EVENT_FL_NO_SET_FILTER_BIT
,
315 FTRACE_EVENT_FL_SOFT_MODE_BIT
,
316 FTRACE_EVENT_FL_SOFT_DISABLED_BIT
,
317 FTRACE_EVENT_FL_TRIGGER_MODE_BIT
,
318 FTRACE_EVENT_FL_TRIGGER_COND_BIT
,
322 * Ftrace event file flags:
323 * ENABLED - The event is enabled
324 * RECORDED_CMD - The comms should be recorded at sched_switch
325 * FILTERED - The event has a filter attached
326 * NO_SET_FILTER - Set when filter has error and is to be ignored
327 * SOFT_MODE - The event is enabled/disabled by SOFT_DISABLED
328 * SOFT_DISABLED - When set, do not trace the event (even though its
329 * tracepoint may be enabled)
330 * TRIGGER_MODE - When set, invoke the triggers associated with the event
331 * TRIGGER_COND - When set, one or more triggers has an associated filter
334 FTRACE_EVENT_FL_ENABLED
= (1 << FTRACE_EVENT_FL_ENABLED_BIT
),
335 FTRACE_EVENT_FL_RECORDED_CMD
= (1 << FTRACE_EVENT_FL_RECORDED_CMD_BIT
),
336 FTRACE_EVENT_FL_FILTERED
= (1 << FTRACE_EVENT_FL_FILTERED_BIT
),
337 FTRACE_EVENT_FL_NO_SET_FILTER
= (1 << FTRACE_EVENT_FL_NO_SET_FILTER_BIT
),
338 FTRACE_EVENT_FL_SOFT_MODE
= (1 << FTRACE_EVENT_FL_SOFT_MODE_BIT
),
339 FTRACE_EVENT_FL_SOFT_DISABLED
= (1 << FTRACE_EVENT_FL_SOFT_DISABLED_BIT
),
340 FTRACE_EVENT_FL_TRIGGER_MODE
= (1 << FTRACE_EVENT_FL_TRIGGER_MODE_BIT
),
341 FTRACE_EVENT_FL_TRIGGER_COND
= (1 << FTRACE_EVENT_FL_TRIGGER_COND_BIT
),
344 struct ftrace_event_file
{
345 struct list_head list
;
346 struct ftrace_event_call
*event_call
;
347 struct event_filter
*filter
;
349 struct trace_array
*tr
;
350 struct ftrace_subsystem_dir
*system
;
351 struct list_head triggers
;
356 * bit 1: enabled cmd record
357 * bit 2: enable/disable with the soft disable bit
358 * bit 3: soft disabled
359 * bit 4: trigger enabled
361 * Note: The bits must be set atomically to prevent races
362 * from other writers. Reads of flags do not need to be in
363 * sync as they occur in critical sections. But the way flags
364 * is currently used, these changes do not affect the code
365 * except that when a change is made, it may have a slight
366 * delay in propagating the changes to other CPUs due to
367 * caching and such. Which is mostly OK ;-)
370 atomic_t sm_ref
; /* soft-mode reference counter */
371 atomic_t tm_ref
; /* trigger-mode reference counter */
374 #define __TRACE_EVENT_FLAGS(name, value) \
375 static int __init trace_init_flags_##name(void) \
377 event_##name.flags |= value; \
380 early_initcall(trace_init_flags_##name);
382 #define __TRACE_EVENT_PERF_PERM(name, expr...) \
383 static int perf_perm_##name(struct ftrace_event_call *tp_event, \
384 struct perf_event *p_event) \
386 return ({ expr; }); \
388 static int __init trace_init_perf_perm_##name(void) \
390 event_##name.perf_perm = &perf_perm_##name; \
393 early_initcall(trace_init_perf_perm_##name);
395 #define PERF_MAX_TRACE_SIZE 2048
397 #define MAX_FILTER_STR_VAL 256 /* Should handle KSYM_SYMBOL_LEN */
399 enum event_trigger_type
{
401 ETT_TRACE_ONOFF
= (1 << 0),
402 ETT_SNAPSHOT
= (1 << 1),
403 ETT_STACKTRACE
= (1 << 2),
404 ETT_EVENT_ENABLE
= (1 << 3),
407 extern void destroy_preds(struct ftrace_event_file
*file
);
408 extern void destroy_call_preds(struct ftrace_event_call
*call
);
409 extern int filter_match_preds(struct event_filter
*filter
, void *rec
);
411 extern int filter_check_discard(struct ftrace_event_file
*file
, void *rec
,
412 struct ring_buffer
*buffer
,
413 struct ring_buffer_event
*event
);
414 extern int call_filter_check_discard(struct ftrace_event_call
*call
, void *rec
,
415 struct ring_buffer
*buffer
,
416 struct ring_buffer_event
*event
);
417 extern enum event_trigger_type
event_triggers_call(struct ftrace_event_file
*file
,
419 extern void event_triggers_post_call(struct ftrace_event_file
*file
,
420 enum event_trigger_type tt
);
423 * ftrace_trigger_soft_disabled - do triggers and test if soft disabled
424 * @file: The file pointer of the event to test
426 * If any triggers without filters are attached to this event, they
427 * will be called here. If the event is soft disabled and has no
428 * triggers that require testing the fields, it will return true,
432 ftrace_trigger_soft_disabled(struct ftrace_event_file
*file
)
434 unsigned long eflags
= file
->flags
;
436 if (!(eflags
& FTRACE_EVENT_FL_TRIGGER_COND
)) {
437 if (eflags
& FTRACE_EVENT_FL_TRIGGER_MODE
)
438 event_triggers_call(file
, NULL
);
439 if (eflags
& FTRACE_EVENT_FL_SOFT_DISABLED
)
446 * Helper function for event_trigger_unlock_commit{_regs}().
447 * If there are event triggers attached to this event that requires
448 * filtering against its fields, then they wil be called as the
449 * entry already holds the field information of the current event.
451 * It also checks if the event should be discarded or not.
452 * It is to be discarded if the event is soft disabled and the
453 * event was only recorded to process triggers, or if the event
454 * filter is active and this event did not match the filters.
456 * Returns true if the event is discarded, false otherwise.
459 __event_trigger_test_discard(struct ftrace_event_file
*file
,
460 struct ring_buffer
*buffer
,
461 struct ring_buffer_event
*event
,
463 enum event_trigger_type
*tt
)
465 unsigned long eflags
= file
->flags
;
467 if (eflags
& FTRACE_EVENT_FL_TRIGGER_COND
)
468 *tt
= event_triggers_call(file
, entry
);
470 if (test_bit(FTRACE_EVENT_FL_SOFT_DISABLED_BIT
, &file
->flags
))
471 ring_buffer_discard_commit(buffer
, event
);
472 else if (!filter_check_discard(file
, entry
, buffer
, event
))
479 * event_trigger_unlock_commit - handle triggers and finish event commit
480 * @file: The file pointer assoctiated to the event
481 * @buffer: The ring buffer that the event is being written to
482 * @event: The event meta data in the ring buffer
483 * @entry: The event itself
484 * @irq_flags: The state of the interrupts at the start of the event
485 * @pc: The state of the preempt count at the start of the event.
487 * This is a helper function to handle triggers that require data
488 * from the event itself. It also tests the event against filters and
489 * if the event is soft disabled and should be discarded.
492 event_trigger_unlock_commit(struct ftrace_event_file
*file
,
493 struct ring_buffer
*buffer
,
494 struct ring_buffer_event
*event
,
495 void *entry
, unsigned long irq_flags
, int pc
)
497 enum event_trigger_type tt
= ETT_NONE
;
499 if (!__event_trigger_test_discard(file
, buffer
, event
, entry
, &tt
))
500 trace_buffer_unlock_commit(buffer
, event
, irq_flags
, pc
);
503 event_triggers_post_call(file
, tt
);
507 * event_trigger_unlock_commit_regs - handle triggers and finish event commit
508 * @file: The file pointer assoctiated to the event
509 * @buffer: The ring buffer that the event is being written to
510 * @event: The event meta data in the ring buffer
511 * @entry: The event itself
512 * @irq_flags: The state of the interrupts at the start of the event
513 * @pc: The state of the preempt count at the start of the event.
515 * This is a helper function to handle triggers that require data
516 * from the event itself. It also tests the event against filters and
517 * if the event is soft disabled and should be discarded.
519 * Same as event_trigger_unlock_commit() but calls
520 * trace_buffer_unlock_commit_regs() instead of trace_buffer_unlock_commit().
523 event_trigger_unlock_commit_regs(struct ftrace_event_file
*file
,
524 struct ring_buffer
*buffer
,
525 struct ring_buffer_event
*event
,
526 void *entry
, unsigned long irq_flags
, int pc
,
527 struct pt_regs
*regs
)
529 enum event_trigger_type tt
= ETT_NONE
;
531 if (!__event_trigger_test_discard(file
, buffer
, event
, entry
, &tt
))
532 trace_buffer_unlock_commit_regs(buffer
, event
,
533 irq_flags
, pc
, regs
);
536 event_triggers_post_call(file
, tt
);
541 FILTER_STATIC_STRING
,
547 extern int trace_event_raw_init(struct ftrace_event_call
*call
);
548 extern int trace_define_field(struct ftrace_event_call
*call
, const char *type
,
549 const char *name
, int offset
, int size
,
550 int is_signed
, int filter_type
);
551 extern int trace_add_event_call(struct ftrace_event_call
*call
);
552 extern int trace_remove_event_call(struct ftrace_event_call
*call
);
554 #define is_signed_type(type) (((type)(-1)) < (type)1)
556 int trace_set_clr_event(const char *system
, const char *event
, int set
);
559 * The double __builtin_constant_p is because gcc will give us an error
560 * if we try to allocate the static variable to fmt if it is not a
561 * constant. Even with the outer if statement optimizing out.
563 #define event_trace_printk(ip, fmt, args...) \
565 __trace_printk_check_format(fmt, ##args); \
566 tracing_record_cmdline(current); \
567 if (__builtin_constant_p(fmt)) { \
568 static const char *trace_printk_fmt \
569 __attribute__((section("__trace_printk_fmt"))) = \
570 __builtin_constant_p(fmt) ? fmt : NULL; \
572 __trace_bprintk(ip, trace_printk_fmt, ##args); \
574 __trace_printk(ip, fmt, ##args); \
578 * tracepoint_string - register constant persistent string to trace system
579 * @str - a constant persistent string that will be referenced in tracepoints
581 * If constant strings are being used in tracepoints, it is faster and
582 * more efficient to just save the pointer to the string and reference
583 * that with a printf "%s" instead of saving the string in the ring buffer
584 * and wasting space and time.
586 * The problem with the above approach is that userspace tools that read
587 * the binary output of the trace buffers do not have access to the string.
588 * Instead they just show the address of the string which is not very
591 * With tracepoint_string(), the string will be registered to the tracing
592 * system and exported to userspace via the debugfs/tracing/printk_formats
593 * file that maps the string address to the string text. This way userspace
594 * tools that read the binary buffers have a way to map the pointers to
595 * the ASCII strings they represent.
597 * The @str used must be a constant string and persistent as it would not
598 * make sense to show a string that no longer exists. But it is still fine
599 * to be used with modules, because when modules are unloaded, if they
600 * had tracepoints, the ring buffers are cleared too. As long as the string
601 * does not change during the life of the module, it is fine to use
602 * tracepoint_string() within a module.
604 #define tracepoint_string(str) \
606 static const char *___tp_str __tracepoint_string = str; \
609 #define __tracepoint_string __attribute__((section("__tracepoint_str")))
611 #ifdef CONFIG_PERF_EVENTS
614 DECLARE_PER_CPU(struct pt_regs
, perf_trace_regs
);
616 extern int perf_trace_init(struct perf_event
*event
);
617 extern void perf_trace_destroy(struct perf_event
*event
);
618 extern int perf_trace_add(struct perf_event
*event
, int flags
);
619 extern void perf_trace_del(struct perf_event
*event
, int flags
);
620 extern int ftrace_profile_set_filter(struct perf_event
*event
, int event_id
,
622 extern void ftrace_profile_free_filter(struct perf_event
*event
);
623 extern void *perf_trace_buf_prepare(int size
, unsigned short type
,
624 struct pt_regs
*regs
, int *rctxp
);
627 perf_trace_buf_submit(void *raw_data
, int size
, int rctx
, u64 addr
,
628 u64 count
, struct pt_regs
*regs
, void *head
,
629 struct task_struct
*task
)
631 perf_tp_event(addr
, count
, raw_data
, size
, regs
, head
, rctx
, task
);
635 #endif /* _LINUX_FTRACE_EVENT_H */