trace_events.h 15.0 KB
Newer Older
1

2 3
#ifndef _LINUX_TRACE_EVENT_H
#define _LINUX_TRACE_EVENT_H
4 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>
11 12

struct trace_array;
13
struct trace_buffer;
14
struct tracer;
15
struct dentry;
16
struct bpf_prog;
17

18 19 20
const char *trace_print_flags_seq(struct trace_seq *p, const char *delim,
				  unsigned long flags,
				  const struct trace_print_flags *flag_array);
21

22 23
const char *trace_print_symbols_seq(struct trace_seq *p, unsigned long val,
				    const struct trace_print_flags *symbol_array);
24

25
#if BITS_PER_LONG == 32
26 27 28
const char *trace_print_symbols_seq_u64(struct trace_seq *p,
					unsigned long long val,
					const struct trace_print_flags_u64
29 30 31
								 *symbol_array);
#endif

32 33
const char *trace_print_bitmask_seq(struct trace_seq *p, void *bitmask_ptr,
				    unsigned int bitmask_size);
34

35
const char *trace_print_hex_seq(struct trace_seq *p,
36 37
				const unsigned char *buf, int len,
				bool spacing);
K
Kei Tokunaga 已提交
38

39
const char *trace_print_array_seq(struct trace_seq *p,
40
				   const void *buf, int count,
D
Dave Martin 已提交
41 42
				   size_t el_size);

43 44 45
struct trace_iterator;
struct trace_event;

46 47
int trace_raw_output_prep(struct trace_iterator *iter,
			  struct trace_event *event);
48

49 50 51 52 53 54 55
/*
 * The trace entry - the most basic unit of tracing. This is what
 * is printed in the end as a single line in the trace output, such as:
 *
 *     bash-15816 [01]   235.197585: idle_cpu <- irq_enter
 */
struct trace_entry {
56
	unsigned short		type;
57 58 59 60 61
	unsigned char		flags;
	unsigned char		preempt_count;
	int			pid;
};

62
#define TRACE_EVENT_TYPE_MAX						\
63 64
	((1 << (sizeof(((struct trace_entry *)0)->type) * 8)) - 1)

65 66 67 68 69 70 71
/*
 * Trace iterator - used by printout routines who present trace
 * results to users and which routines might sleep, etc:
 */
struct trace_iterator {
	struct trace_array	*tr;
	struct tracer		*trace;
72
	struct trace_buffer	*trace_buffer;
73 74 75
	void			*private;
	int			cpu_file;
	struct mutex		mutex;
76
	struct ring_buffer_iter	**buffer_iter;
77
	unsigned long		iter_flags;
78

79 80 81
	/* trace_seq for __print_flags() and __print_symbolic() etc. */
	struct trace_seq	tmp_seq;

82 83 84 85 86
	cpumask_var_t		started;

	/* it's true when current open file is snapshot */
	bool			snapshot;

87 88 89
	/* The below is zeroed out in pipe_read */
	struct trace_seq	seq;
	struct trace_entry	*ent;
90
	unsigned long		lost_events;
91
	int			leftover;
92
	int			ent_size;
93 94 95 96 97 98
	int			cpu;
	u64			ts;

	loff_t			pos;
	long			idx;

99
	/* All new field here will be zeroed out in pipe_read */
100 101
};

102 103 104 105 106 107
enum trace_iter_flags {
	TRACE_FILE_LAT_FMT	= 1,
	TRACE_FILE_ANNOTATE	= 2,
	TRACE_FILE_TIME_IN_NS	= 4,
};

108 109

typedef enum print_line_t (*trace_print_func)(struct trace_iterator *iter,
110 111 112
				      int flags, struct trace_event *event);

struct trace_event_functions {
113 114 115 116 117 118
	trace_print_func	trace;
	trace_print_func	raw;
	trace_print_func	hex;
	trace_print_func	binary;
};

119 120 121 122 123 124 125
struct trace_event {
	struct hlist_node		node;
	struct list_head		list;
	int				type;
	struct trace_event_functions	*funcs;
};

126 127
extern int register_trace_event(struct trace_event *event);
extern int unregister_trace_event(struct trace_event *event);
128 129 130 131 132 133 134 135 136

/* Return values for print_line callback */
enum print_line_t {
	TRACE_TYPE_PARTIAL_LINE	= 0,	/* Retry after flushing the seq */
	TRACE_TYPE_HANDLED	= 1,
	TRACE_TYPE_UNHANDLED	= 2,	/* Relay to other output functions */
	TRACE_TYPE_NO_CONSUME	= 3	/* Handled but ask to not consume */
};

137 138 139 140 141 142 143 144 145 146 147
/*
 * Several functions return TRACE_TYPE_PARTIAL_LINE if the trace_seq
 * overflowed, and TRACE_TYPE_HANDLED otherwise. This helper function
 * simplifies those functions and keeps them in sync.
 */
static inline enum print_line_t trace_handle_return(struct trace_seq *s)
{
	return trace_seq_has_overflowed(s) ?
		TRACE_TYPE_PARTIAL_LINE : TRACE_TYPE_HANDLED;
}

148 149 150
void tracing_generic_entry_update(struct trace_entry *entry,
				  unsigned long flags,
				  int pc);
151
struct trace_event_file;
152 153 154

struct ring_buffer_event *
trace_event_buffer_lock_reserve(struct ring_buffer **current_buffer,
155
				struct trace_event_file *trace_file,
156 157
				int type, unsigned long len,
				unsigned long flags, int pc);
158 159 160

void tracing_record_cmdline(struct task_struct *tsk);

161
int trace_output_call(struct trace_iterator *iter, char *name, char *fmt, ...);
162

163 164
struct event_filter;

165 166 167
enum trace_reg {
	TRACE_REG_REGISTER,
	TRACE_REG_UNREGISTER,
168
#ifdef CONFIG_PERF_EVENTS
169 170
	TRACE_REG_PERF_REGISTER,
	TRACE_REG_PERF_UNREGISTER,
171 172
	TRACE_REG_PERF_OPEN,
	TRACE_REG_PERF_CLOSE,
173 174
	TRACE_REG_PERF_ADD,
	TRACE_REG_PERF_DEL,
175
#endif
176 177
};

178
struct trace_event_call;
179

180
struct trace_event_class {
181
	const char		*system;
182 183 184 185
	void			*probe;
#ifdef CONFIG_PERF_EVENTS
	void			*perf_probe;
#endif
186
	int			(*reg)(struct trace_event_call *event,
187
				       enum trace_reg type, void *data);
188 189
	int			(*define_fields)(struct trace_event_call *);
	struct list_head	*(*get_fields)(struct trace_event_call *);
190
	struct list_head	fields;
191
	int			(*raw_init)(struct trace_event_call *);
192 193
};

194
extern int trace_event_reg(struct trace_event_call *event,
195
			    enum trace_reg type, void *data);
196

197
struct trace_event_buffer {
198 199
	struct ring_buffer		*buffer;
	struct ring_buffer_event	*event;
200
	struct trace_event_file		*trace_file;
201 202 203 204 205
	void				*entry;
	unsigned long			flags;
	int				pc;
};

206
void *trace_event_buffer_reserve(struct trace_event_buffer *fbuffer,
207
				  struct trace_event_file *trace_file,
208 209
				  unsigned long len);

210
void trace_event_buffer_commit(struct trace_event_buffer *fbuffer);
211

212 213
enum {
	TRACE_EVENT_FL_FILTERED_BIT,
214
	TRACE_EVENT_FL_CAP_ANY_BIT,
L
Li Zefan 已提交
215
	TRACE_EVENT_FL_NO_SET_FILTER_BIT,
216
	TRACE_EVENT_FL_IGNORE_ENABLE_BIT,
217
	TRACE_EVENT_FL_WAS_ENABLED_BIT,
218
	TRACE_EVENT_FL_TRACEPOINT_BIT,
A
Alexei Starovoitov 已提交
219
	TRACE_EVENT_FL_KPROBE_BIT,
220
	TRACE_EVENT_FL_UPROBE_BIT,
221 222
};

223 224 225 226 227
/*
 * Event flags:
 *  FILTERED	  - The event has a filter attached
 *  CAP_ANY	  - Any user can enable for perf
 *  NO_SET_FILTER - Set when filter has error and is to be ignored
228
 *  IGNORE_ENABLE - For trace internal events, do not enable with debugfs file
229 230 231
 *  WAS_ENABLED   - Set and stays set when an event was ever enabled
 *                    (used for module unloading, if a module event is enabled,
 *                     it is best to clear the buffers that used it).
232
 *  TRACEPOINT    - Event is a tracepoint
A
Alexei Starovoitov 已提交
233
 *  KPROBE        - Event is a kprobe
234
 *  UPROBE        - Event is a uprobe
235
 */
236
enum {
237
	TRACE_EVENT_FL_FILTERED		= (1 << TRACE_EVENT_FL_FILTERED_BIT),
238
	TRACE_EVENT_FL_CAP_ANY		= (1 << TRACE_EVENT_FL_CAP_ANY_BIT),
L
Li Zefan 已提交
239
	TRACE_EVENT_FL_NO_SET_FILTER	= (1 << TRACE_EVENT_FL_NO_SET_FILTER_BIT),
240
	TRACE_EVENT_FL_IGNORE_ENABLE	= (1 << TRACE_EVENT_FL_IGNORE_ENABLE_BIT),
241
	TRACE_EVENT_FL_WAS_ENABLED	= (1 << TRACE_EVENT_FL_WAS_ENABLED_BIT),
242
	TRACE_EVENT_FL_TRACEPOINT	= (1 << TRACE_EVENT_FL_TRACEPOINT_BIT),
A
Alexei Starovoitov 已提交
243
	TRACE_EVENT_FL_KPROBE		= (1 << TRACE_EVENT_FL_KPROBE_BIT),
244
	TRACE_EVENT_FL_UPROBE		= (1 << TRACE_EVENT_FL_UPROBE_BIT),
245 246
};

247 248
#define TRACE_EVENT_FL_UKPROBE (TRACE_EVENT_FL_KPROBE | TRACE_EVENT_FL_UPROBE)

249
struct trace_event_call {
250
	struct list_head	list;
251
	struct trace_event_class *class;
252 253 254 255 256
	union {
		char			*name;
		/* Set TRACE_EVENT_FL_TRACEPOINT flag when using "tp" */
		struct tracepoint	*tp;
	};
257
	struct trace_event	event;
258
	char			*print_fmt;
259
	struct event_filter	*filter;
260
	void			*mod;
261
	void			*data;
262 263 264 265
	/*
	 *   bit 0:		filter_active
	 *   bit 1:		allow trace by non root (cap any)
	 *   bit 2:		failed to apply filter
266
	 *   bit 3:		trace internal event (do not enable)
267
	 *   bit 4:		Event was enabled by module
268
	 *   bit 5:		use call filter rather than file filter
269
	 *   bit 6:		Event is a tracepoint
270
	 */
271 272 273 274 275
	int			flags; /* static flags of different events */

#ifdef CONFIG_PERF_EVENTS
	int				perf_refcount;
	struct hlist_head __percpu	*perf_events;
276
	struct bpf_prog			*prog;
277

278
	int	(*perf_perm)(struct trace_event_call *,
279
			     struct perf_event *);
280 281 282
#endif
};

283
static inline const char *
284
trace_event_name(struct trace_event_call *call)
285 286 287 288 289 290 291
{
	if (call->flags & TRACE_EVENT_FL_TRACEPOINT)
		return call->tp ? call->tp->name : NULL;
	else
		return call->name;
}

292
struct trace_array;
293
struct trace_subsystem_dir;
294 295

enum {
296 297 298 299 300 301 302 303
	EVENT_FILE_FL_ENABLED_BIT,
	EVENT_FILE_FL_RECORDED_CMD_BIT,
	EVENT_FILE_FL_FILTERED_BIT,
	EVENT_FILE_FL_NO_SET_FILTER_BIT,
	EVENT_FILE_FL_SOFT_MODE_BIT,
	EVENT_FILE_FL_SOFT_DISABLED_BIT,
	EVENT_FILE_FL_TRIGGER_MODE_BIT,
	EVENT_FILE_FL_TRIGGER_COND_BIT,
304
	EVENT_FILE_FL_PID_FILTER_BIT,
305 306 307
};

/*
308
 * Event file flags:
309
 *  ENABLED	  - The event is enabled
310
 *  RECORDED_CMD  - The comms should be recorded at sched_switch
311 312
 *  FILTERED	  - The event has a filter attached
 *  NO_SET_FILTER - Set when filter has error and is to be ignored
313 314 315
 *  SOFT_MODE     - The event is enabled/disabled by SOFT_DISABLED
 *  SOFT_DISABLED - When set, do not trace the event (even though its
 *                   tracepoint may be enabled)
316
 *  TRIGGER_MODE  - When set, invoke the triggers associated with the event
317
 *  TRIGGER_COND  - When set, one or more triggers has an associated filter
318
 *  PID_FILTER    - When set, the event is filtered based on pid
319 320
 */
enum {
321 322 323 324 325 326 327 328
	EVENT_FILE_FL_ENABLED		= (1 << EVENT_FILE_FL_ENABLED_BIT),
	EVENT_FILE_FL_RECORDED_CMD	= (1 << EVENT_FILE_FL_RECORDED_CMD_BIT),
	EVENT_FILE_FL_FILTERED		= (1 << EVENT_FILE_FL_FILTERED_BIT),
	EVENT_FILE_FL_NO_SET_FILTER	= (1 << EVENT_FILE_FL_NO_SET_FILTER_BIT),
	EVENT_FILE_FL_SOFT_MODE		= (1 << EVENT_FILE_FL_SOFT_MODE_BIT),
	EVENT_FILE_FL_SOFT_DISABLED	= (1 << EVENT_FILE_FL_SOFT_DISABLED_BIT),
	EVENT_FILE_FL_TRIGGER_MODE	= (1 << EVENT_FILE_FL_TRIGGER_MODE_BIT),
	EVENT_FILE_FL_TRIGGER_COND	= (1 << EVENT_FILE_FL_TRIGGER_COND_BIT),
329
	EVENT_FILE_FL_PID_FILTER	= (1 << EVENT_FILE_FL_PID_FILTER_BIT),
330 331
};

332
struct trace_event_file {
333
	struct list_head		list;
334
	struct trace_event_call		*event_call;
335
	struct event_filter		*filter;
336 337
	struct dentry			*dir;
	struct trace_array		*tr;
338
	struct trace_subsystem_dir	*system;
339
	struct list_head		triggers;
340

341 342
	/*
	 * 32 bit flags:
343 344
	 *   bit 0:		enabled
	 *   bit 1:		enabled cmd record
345 346
	 *   bit 2:		enable/disable with the soft disable bit
	 *   bit 3:		soft disabled
347
	 *   bit 4:		trigger enabled
348
	 *
349 350 351
	 * Note: The bits must be set atomically to prevent races
	 * from other writers. Reads of flags do not need to be in
	 * sync as they occur in critical sections. But the way flags
352
	 * is currently used, these changes do not affect the code
353 354
	 * except that when a change is made, it may have a slight
	 * delay in propagating the changes to other CPUs due to
355
	 * caching and such. Which is mostly OK ;-)
356
	 */
357
	unsigned long		flags;
358
	atomic_t		sm_ref;	/* soft-mode reference counter */
359
	atomic_t		tm_ref;	/* trigger-mode reference counter */
360 361
};

362 363 364
#define __TRACE_EVENT_FLAGS(name, value)				\
	static int __init trace_init_flags_##name(void)			\
	{								\
365
		event_##name.flags |= value;				\
366 367 368 369
		return 0;						\
	}								\
	early_initcall(trace_init_flags_##name);

370
#define __TRACE_EVENT_PERF_PERM(name, expr...)				\
371
	static int perf_perm_##name(struct trace_event_call *tp_event, \
372 373 374 375 376 377 378 379 380 381 382
				    struct perf_event *p_event)		\
	{								\
		return ({ expr; });					\
	}								\
	static int __init trace_init_perf_perm_##name(void)		\
	{								\
		event_##name.perf_perm = &perf_perm_##name;		\
		return 0;						\
	}								\
	early_initcall(trace_init_perf_perm_##name);

383
#define PERF_MAX_TRACE_SIZE	2048
384

385
#define MAX_FILTER_STR_VAL	256	/* Should handle KSYM_SYMBOL_LEN */
386

387 388
enum event_trigger_type {
	ETT_NONE		= (0),
389
	ETT_TRACE_ONOFF		= (1 << 0),
390
	ETT_SNAPSHOT		= (1 << 1),
391
	ETT_STACKTRACE		= (1 << 2),
392
	ETT_EVENT_ENABLE	= (1 << 3),
393
	ETT_EVENT_HIST		= (1 << 4),
394
	ETT_HIST_ENABLE		= (1 << 5),
395 396
};

L
Li Zefan 已提交
397
extern int filter_match_preds(struct event_filter *filter, void *rec);
398

399
extern enum event_trigger_type event_triggers_call(struct trace_event_file *file,
400
						   void *rec);
401
extern void event_triggers_post_call(struct trace_event_file *file,
402 403
				     enum event_trigger_type tt,
				     void *rec);
404

405 406
bool trace_event_ignore_this_pid(struct trace_event_file *trace_file);

407
/**
408
 * trace_trigger_soft_disabled - do triggers and test if soft disabled
409 410 411 412 413 414 415 416
 * @file: The file pointer of the event to test
 *
 * If any triggers without filters are attached to this event, they
 * will be called here. If the event is soft disabled and has no
 * triggers that require testing the fields, it will return true,
 * otherwise false.
 */
static inline bool
417
trace_trigger_soft_disabled(struct trace_event_file *file)
418 419 420
{
	unsigned long eflags = file->flags;

421 422
	if (!(eflags & EVENT_FILE_FL_TRIGGER_COND)) {
		if (eflags & EVENT_FILE_FL_TRIGGER_MODE)
423
			event_triggers_call(file, NULL);
424
		if (eflags & EVENT_FILE_FL_SOFT_DISABLED)
425
			return true;
426 427
		if (eflags & EVENT_FILE_FL_PID_FILTER)
			return trace_event_ignore_this_pid(file);
428 429 430 431
	}
	return false;
}

432
#ifdef CONFIG_BPF_EVENTS
433 434 435 436 437 438 439 440
unsigned int trace_call_bpf(struct bpf_prog *prog, void *ctx);
#else
static inline unsigned int trace_call_bpf(struct bpf_prog *prog, void *ctx)
{
	return 1;
}
#endif

441 442 443 444
enum {
	FILTER_OTHER = 0,
	FILTER_STATIC_STRING,
	FILTER_DYN_STRING,
445
	FILTER_PTR_STRING,
446
	FILTER_TRACE_FN,
447 448
	FILTER_COMM,
	FILTER_CPU,
449 450
};

451 452
extern int trace_event_raw_init(struct trace_event_call *call);
extern int trace_define_field(struct trace_event_call *call, const char *type,
453 454
			      const char *name, int offset, int size,
			      int is_signed, int filter_type);
455 456
extern int trace_add_event_call(struct trace_event_call *call);
extern int trace_remove_event_call(struct trace_event_call *call);
457
extern int trace_event_get_offsets(struct trace_event_call *call);
458

459
#define is_signed_type(type)	(((type)(-1)) < (type)1)
460

461 462
int trace_set_clr_event(const char *system, const char *event, int set);

463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481
/*
 * The double __builtin_constant_p is because gcc will give us an error
 * if we try to allocate the static variable to fmt if it is not a
 * constant. Even with the outer if statement optimizing out.
 */
#define event_trace_printk(ip, fmt, args...)				\
do {									\
	__trace_printk_check_format(fmt, ##args);			\
	tracing_record_cmdline(current);				\
	if (__builtin_constant_p(fmt)) {				\
		static const char *trace_printk_fmt			\
		  __attribute__((section("__trace_printk_fmt"))) =	\
			__builtin_constant_p(fmt) ? fmt : NULL;		\
									\
		__trace_bprintk(ip, trace_printk_fmt, ##args);		\
	} else								\
		__trace_printk(ip, fmt, ##args);			\
} while (0)

482
#ifdef CONFIG_PERF_EVENTS
L
Li Zefan 已提交
483
struct perf_event;
484 485 486

DECLARE_PER_CPU(struct pt_regs, perf_trace_regs);

487 488
extern int  perf_trace_init(struct perf_event *event);
extern void perf_trace_destroy(struct perf_event *event);
P
Peter Zijlstra 已提交
489 490
extern int  perf_trace_add(struct perf_event *event, int flags);
extern void perf_trace_del(struct perf_event *event, int flags);
491
extern int  ftrace_profile_set_filter(struct perf_event *event, int event_id,
L
Li Zefan 已提交
492 493
				     char *filter_str);
extern void ftrace_profile_free_filter(struct perf_event *event);
494 495
void perf_trace_buf_update(void *record, u16 type);
void *perf_trace_buf_alloc(int size, struct pt_regs **regs, int *rctxp);
496

497 498 499 500 501
void perf_trace_run_bpf_submit(void *raw_data, int size, int rctx,
			       struct trace_event_call *call, u64 count,
			       struct pt_regs *regs, struct hlist_head *head,
			       struct task_struct *task);

502
static inline void
503
perf_trace_buf_submit(void *raw_data, int size, int rctx, u16 type,
504 505
		       u64 count, struct pt_regs *regs, void *head,
		       struct task_struct *task)
506
{
507
	perf_tp_event(type, count, raw_data, size, regs, head, rctx, task);
508
}
L
Li Zefan 已提交
509 510
#endif

511
#endif /* _LINUX_TRACE_EVENT_H */