trace_events.h 14.9 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
R
Ross Zwisler 已提交
26 27 28 29
const char *trace_print_flags_seq_u64(struct trace_seq *p, const char *delim,
		      unsigned long long flags,
		      const struct trace_print_flags_u64 *flag_array);

30 31 32
const char *trace_print_symbols_seq_u64(struct trace_seq *p,
					unsigned long long val,
					const struct trace_print_flags_u64
33 34 35
								 *symbol_array);
#endif

36 37
const char *trace_print_bitmask_seq(struct trace_seq *p, void *bitmask_ptr,
				    unsigned int bitmask_size);
38

39
const char *trace_print_hex_seq(struct trace_seq *p,
40
				const unsigned char *buf, int len,
41
				bool concatenate);
K
Kei Tokunaga 已提交
42

43
const char *trace_print_array_seq(struct trace_seq *p,
44
				   const void *buf, int count,
D
Dave Martin 已提交
45 46
				   size_t el_size);

47 48 49
struct trace_iterator;
struct trace_event;

50 51
int trace_raw_output_prep(struct trace_iterator *iter,
			  struct trace_event *event);
52

53 54 55 56 57 58 59
/*
 * 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 {
60
	unsigned short		type;
61 62 63 64 65
	unsigned char		flags;
	unsigned char		preempt_count;
	int			pid;
};

66
#define TRACE_EVENT_TYPE_MAX						\
67 68
	((1 << (sizeof(((struct trace_entry *)0)->type) * 8)) - 1)

69 70 71 72 73 74 75
/*
 * 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;
76
	struct trace_buffer	*trace_buffer;
77 78 79
	void			*private;
	int			cpu_file;
	struct mutex		mutex;
80
	struct ring_buffer_iter	**buffer_iter;
81
	unsigned long		iter_flags;
82

83 84 85
	/* trace_seq for __print_flags() and __print_symbolic() etc. */
	struct trace_seq	tmp_seq;

86 87 88 89 90
	cpumask_var_t		started;

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

91 92 93
	/* The below is zeroed out in pipe_read */
	struct trace_seq	seq;
	struct trace_entry	*ent;
94
	unsigned long		lost_events;
95
	int			leftover;
96
	int			ent_size;
97 98 99 100 101 102
	int			cpu;
	u64			ts;

	loff_t			pos;
	long			idx;

103
	/* All new field here will be zeroed out in pipe_read */
104 105
};

106 107 108 109 110 111
enum trace_iter_flags {
	TRACE_FILE_LAT_FMT	= 1,
	TRACE_FILE_ANNOTATE	= 2,
	TRACE_FILE_TIME_IN_NS	= 4,
};

112 113

typedef enum print_line_t (*trace_print_func)(struct trace_iterator *iter,
114 115 116
				      int flags, struct trace_event *event);

struct trace_event_functions {
117 118 119 120 121 122
	trace_print_func	trace;
	trace_print_func	raw;
	trace_print_func	hex;
	trace_print_func	binary;
};

123 124 125 126 127 128 129
struct trace_event {
	struct hlist_node		node;
	struct list_head		list;
	int				type;
	struct trace_event_functions	*funcs;
};

130 131
extern int register_trace_event(struct trace_event *event);
extern int unregister_trace_event(struct trace_event *event);
132 133 134 135 136 137 138 139 140

/* 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 */
};

141
enum print_line_t trace_handle_return(struct trace_seq *s);
142

143 144 145
void tracing_generic_entry_update(struct trace_entry *entry,
				  unsigned long flags,
				  int pc);
146
struct trace_event_file;
147 148 149

struct ring_buffer_event *
trace_event_buffer_lock_reserve(struct ring_buffer **current_buffer,
150
				struct trace_event_file *trace_file,
151 152
				int type, unsigned long len,
				unsigned long flags, int pc);
153 154 155

void tracing_record_cmdline(struct task_struct *tsk);

156
int trace_output_call(struct trace_iterator *iter, char *name, char *fmt, ...);
157

158 159
struct event_filter;

160 161 162
enum trace_reg {
	TRACE_REG_REGISTER,
	TRACE_REG_UNREGISTER,
163
#ifdef CONFIG_PERF_EVENTS
164 165
	TRACE_REG_PERF_REGISTER,
	TRACE_REG_PERF_UNREGISTER,
166 167
	TRACE_REG_PERF_OPEN,
	TRACE_REG_PERF_CLOSE,
168 169
	TRACE_REG_PERF_ADD,
	TRACE_REG_PERF_DEL,
170
#endif
171 172
};

173
struct trace_event_call;
174

175
struct trace_event_class {
176
	const char		*system;
177 178 179 180
	void			*probe;
#ifdef CONFIG_PERF_EVENTS
	void			*perf_probe;
#endif
181
	int			(*reg)(struct trace_event_call *event,
182
				       enum trace_reg type, void *data);
183 184
	int			(*define_fields)(struct trace_event_call *);
	struct list_head	*(*get_fields)(struct trace_event_call *);
185
	struct list_head	fields;
186
	int			(*raw_init)(struct trace_event_call *);
187 188
};

189
extern int trace_event_reg(struct trace_event_call *event,
190
			    enum trace_reg type, void *data);
191

192
struct trace_event_buffer {
193 194
	struct ring_buffer		*buffer;
	struct ring_buffer_event	*event;
195
	struct trace_event_file		*trace_file;
196 197 198 199 200
	void				*entry;
	unsigned long			flags;
	int				pc;
};

201
void *trace_event_buffer_reserve(struct trace_event_buffer *fbuffer,
202
				  struct trace_event_file *trace_file,
203 204
				  unsigned long len);

205
void trace_event_buffer_commit(struct trace_event_buffer *fbuffer);
206

207 208
enum {
	TRACE_EVENT_FL_FILTERED_BIT,
209
	TRACE_EVENT_FL_CAP_ANY_BIT,
L
Li Zefan 已提交
210
	TRACE_EVENT_FL_NO_SET_FILTER_BIT,
211
	TRACE_EVENT_FL_IGNORE_ENABLE_BIT,
212
	TRACE_EVENT_FL_WAS_ENABLED_BIT,
213
	TRACE_EVENT_FL_TRACEPOINT_BIT,
A
Alexei Starovoitov 已提交
214
	TRACE_EVENT_FL_KPROBE_BIT,
215
	TRACE_EVENT_FL_UPROBE_BIT,
216 217
};

218 219 220 221 222
/*
 * 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
223
 *  IGNORE_ENABLE - For trace internal events, do not enable with debugfs file
224 225 226
 *  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).
227
 *  TRACEPOINT    - Event is a tracepoint
A
Alexei Starovoitov 已提交
228
 *  KPROBE        - Event is a kprobe
229
 *  UPROBE        - Event is a uprobe
230
 */
231
enum {
232
	TRACE_EVENT_FL_FILTERED		= (1 << TRACE_EVENT_FL_FILTERED_BIT),
233
	TRACE_EVENT_FL_CAP_ANY		= (1 << TRACE_EVENT_FL_CAP_ANY_BIT),
L
Li Zefan 已提交
234
	TRACE_EVENT_FL_NO_SET_FILTER	= (1 << TRACE_EVENT_FL_NO_SET_FILTER_BIT),
235
	TRACE_EVENT_FL_IGNORE_ENABLE	= (1 << TRACE_EVENT_FL_IGNORE_ENABLE_BIT),
236
	TRACE_EVENT_FL_WAS_ENABLED	= (1 << TRACE_EVENT_FL_WAS_ENABLED_BIT),
237
	TRACE_EVENT_FL_TRACEPOINT	= (1 << TRACE_EVENT_FL_TRACEPOINT_BIT),
A
Alexei Starovoitov 已提交
238
	TRACE_EVENT_FL_KPROBE		= (1 << TRACE_EVENT_FL_KPROBE_BIT),
239
	TRACE_EVENT_FL_UPROBE		= (1 << TRACE_EVENT_FL_UPROBE_BIT),
240 241
};

242 243
#define TRACE_EVENT_FL_UKPROBE (TRACE_EVENT_FL_KPROBE | TRACE_EVENT_FL_UPROBE)

244
struct trace_event_call {
245
	struct list_head	list;
246
	struct trace_event_class *class;
247 248 249 250 251
	union {
		char			*name;
		/* Set TRACE_EVENT_FL_TRACEPOINT flag when using "tp" */
		struct tracepoint	*tp;
	};
252
	struct trace_event	event;
253
	char			*print_fmt;
254
	struct event_filter	*filter;
255
	void			*mod;
256
	void			*data;
257 258 259 260
	/*
	 *   bit 0:		filter_active
	 *   bit 1:		allow trace by non root (cap any)
	 *   bit 2:		failed to apply filter
261
	 *   bit 3:		trace internal event (do not enable)
262
	 *   bit 4:		Event was enabled by module
263
	 *   bit 5:		use call filter rather than file filter
264
	 *   bit 6:		Event is a tracepoint
265
	 */
266 267 268 269 270
	int			flags; /* static flags of different events */

#ifdef CONFIG_PERF_EVENTS
	int				perf_refcount;
	struct hlist_head __percpu	*perf_events;
271
	struct bpf_prog			*prog;
272

273
	int	(*perf_perm)(struct trace_event_call *,
274
			     struct perf_event *);
275 276 277
#endif
};

278
static inline const char *
279
trace_event_name(struct trace_event_call *call)
280 281 282 283 284 285 286
{
	if (call->flags & TRACE_EVENT_FL_TRACEPOINT)
		return call->tp ? call->tp->name : NULL;
	else
		return call->name;
}

287
struct trace_array;
288
struct trace_subsystem_dir;
289 290

enum {
291 292 293 294 295 296 297 298
	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,
299
	EVENT_FILE_FL_PID_FILTER_BIT,
300 301 302
};

/*
303
 * Event file flags:
304
 *  ENABLED	  - The event is enabled
305
 *  RECORDED_CMD  - The comms should be recorded at sched_switch
306 307
 *  FILTERED	  - The event has a filter attached
 *  NO_SET_FILTER - Set when filter has error and is to be ignored
308 309 310
 *  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)
311
 *  TRIGGER_MODE  - When set, invoke the triggers associated with the event
312
 *  TRIGGER_COND  - When set, one or more triggers has an associated filter
313
 *  PID_FILTER    - When set, the event is filtered based on pid
314 315
 */
enum {
316 317 318 319 320 321 322 323
	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),
324
	EVENT_FILE_FL_PID_FILTER	= (1 << EVENT_FILE_FL_PID_FILTER_BIT),
325 326
};

327
struct trace_event_file {
328
	struct list_head		list;
329
	struct trace_event_call		*event_call;
330
	struct event_filter		*filter;
331 332
	struct dentry			*dir;
	struct trace_array		*tr;
333
	struct trace_subsystem_dir	*system;
334
	struct list_head		triggers;
335

336 337
	/*
	 * 32 bit flags:
338 339
	 *   bit 0:		enabled
	 *   bit 1:		enabled cmd record
340 341
	 *   bit 2:		enable/disable with the soft disable bit
	 *   bit 3:		soft disabled
342
	 *   bit 4:		trigger enabled
343
	 *
344 345 346
	 * 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
347
	 * is currently used, these changes do not affect the code
348 349
	 * except that when a change is made, it may have a slight
	 * delay in propagating the changes to other CPUs due to
350
	 * caching and such. Which is mostly OK ;-)
351
	 */
352
	unsigned long		flags;
353
	atomic_t		sm_ref;	/* soft-mode reference counter */
354
	atomic_t		tm_ref;	/* trigger-mode reference counter */
355 356
};

357 358 359
#define __TRACE_EVENT_FLAGS(name, value)				\
	static int __init trace_init_flags_##name(void)			\
	{								\
360
		event_##name.flags |= value;				\
361 362 363 364
		return 0;						\
	}								\
	early_initcall(trace_init_flags_##name);

365
#define __TRACE_EVENT_PERF_PERM(name, expr...)				\
366
	static int perf_perm_##name(struct trace_event_call *tp_event, \
367 368 369 370 371 372 373 374 375 376 377
				    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);

378
#define PERF_MAX_TRACE_SIZE	2048
379

380
#define MAX_FILTER_STR_VAL	256	/* Should handle KSYM_SYMBOL_LEN */
381

382 383
enum event_trigger_type {
	ETT_NONE		= (0),
384
	ETT_TRACE_ONOFF		= (1 << 0),
385
	ETT_SNAPSHOT		= (1 << 1),
386
	ETT_STACKTRACE		= (1 << 2),
387
	ETT_EVENT_ENABLE	= (1 << 3),
388
	ETT_EVENT_HIST		= (1 << 4),
389
	ETT_HIST_ENABLE		= (1 << 5),
390 391
};

L
Li Zefan 已提交
392
extern int filter_match_preds(struct event_filter *filter, void *rec);
393

394
extern enum event_trigger_type event_triggers_call(struct trace_event_file *file,
395
						   void *rec);
396
extern void event_triggers_post_call(struct trace_event_file *file,
397 398
				     enum event_trigger_type tt,
				     void *rec);
399

400 401
bool trace_event_ignore_this_pid(struct trace_event_file *trace_file);

402
/**
403
 * trace_trigger_soft_disabled - do triggers and test if soft disabled
404 405 406 407 408 409 410 411
 * @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
412
trace_trigger_soft_disabled(struct trace_event_file *file)
413 414 415
{
	unsigned long eflags = file->flags;

416 417
	if (!(eflags & EVENT_FILE_FL_TRIGGER_COND)) {
		if (eflags & EVENT_FILE_FL_TRIGGER_MODE)
418
			event_triggers_call(file, NULL);
419
		if (eflags & EVENT_FILE_FL_SOFT_DISABLED)
420
			return true;
421 422
		if (eflags & EVENT_FILE_FL_PID_FILTER)
			return trace_event_ignore_this_pid(file);
423 424 425 426
	}
	return false;
}

427
#ifdef CONFIG_BPF_EVENTS
428 429 430 431 432 433 434 435
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

436 437 438 439
enum {
	FILTER_OTHER = 0,
	FILTER_STATIC_STRING,
	FILTER_DYN_STRING,
440
	FILTER_PTR_STRING,
441
	FILTER_TRACE_FN,
442 443
	FILTER_COMM,
	FILTER_CPU,
444 445
};

446 447
extern int trace_event_raw_init(struct trace_event_call *call);
extern int trace_define_field(struct trace_event_call *call, const char *type,
448 449
			      const char *name, int offset, int size,
			      int is_signed, int filter_type);
450 451
extern int trace_add_event_call(struct trace_event_call *call);
extern int trace_remove_event_call(struct trace_event_call *call);
452
extern int trace_event_get_offsets(struct trace_event_call *call);
453

454
#define is_signed_type(type)	(((type)(-1)) < (type)1)
455

456 457
int trace_set_clr_event(const char *system, const char *event, int set);

458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476
/*
 * 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)

477
#ifdef CONFIG_PERF_EVENTS
L
Li Zefan 已提交
478
struct perf_event;
479 480 481

DECLARE_PER_CPU(struct pt_regs, perf_trace_regs);

482 483
extern int  perf_trace_init(struct perf_event *event);
extern void perf_trace_destroy(struct perf_event *event);
P
Peter Zijlstra 已提交
484 485
extern int  perf_trace_add(struct perf_event *event, int flags);
extern void perf_trace_del(struct perf_event *event, int flags);
486
extern int  ftrace_profile_set_filter(struct perf_event *event, int event_id,
L
Li Zefan 已提交
487 488
				     char *filter_str);
extern void ftrace_profile_free_filter(struct perf_event *event);
489 490
void perf_trace_buf_update(void *record, u16 type);
void *perf_trace_buf_alloc(int size, struct pt_regs **regs, int *rctxp);
491

492 493 494 495 496
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);

497
static inline void
498
perf_trace_buf_submit(void *raw_data, int size, int rctx, u16 type,
499 500
		       u64 count, struct pt_regs *regs, void *head,
		       struct task_struct *task)
501
{
502
	perf_tp_event(type, count, raw_data, size, regs, head, rctx, task);
503
}
L
Li Zefan 已提交
504 505
#endif

506
#endif /* _LINUX_TRACE_EVENT_H */