trace_events.h 15.3 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

154 155 156 157 158 159 160 161 162
#define TRACE_RECORD_CMDLINE	BIT(0)
#define TRACE_RECORD_TGID	BIT(1)

void tracing_record_taskinfo(struct task_struct *task, int flags);
void tracing_record_taskinfo_sched_switch(struct task_struct *prev,
					  struct task_struct *next, int flags);

void tracing_record_cmdline(struct task_struct *task);
void tracing_record_tgid(struct task_struct *task);
163

164
int trace_output_call(struct trace_iterator *iter, char *name, char *fmt, ...);
165

166 167
struct event_filter;

168 169 170
enum trace_reg {
	TRACE_REG_REGISTER,
	TRACE_REG_UNREGISTER,
171
#ifdef CONFIG_PERF_EVENTS
172 173
	TRACE_REG_PERF_REGISTER,
	TRACE_REG_PERF_UNREGISTER,
174 175
	TRACE_REG_PERF_OPEN,
	TRACE_REG_PERF_CLOSE,
176 177
	TRACE_REG_PERF_ADD,
	TRACE_REG_PERF_DEL,
178
#endif
179 180
};

181
struct trace_event_call;
182

183
struct trace_event_class {
184
	const char		*system;
185 186 187 188
	void			*probe;
#ifdef CONFIG_PERF_EVENTS
	void			*perf_probe;
#endif
189
	int			(*reg)(struct trace_event_call *event,
190
				       enum trace_reg type, void *data);
191 192
	int			(*define_fields)(struct trace_event_call *);
	struct list_head	*(*get_fields)(struct trace_event_call *);
193
	struct list_head	fields;
194
	int			(*raw_init)(struct trace_event_call *);
195 196
};

197
extern int trace_event_reg(struct trace_event_call *event,
198
			    enum trace_reg type, void *data);
199

200
struct trace_event_buffer {
201 202
	struct ring_buffer		*buffer;
	struct ring_buffer_event	*event;
203
	struct trace_event_file		*trace_file;
204 205 206 207 208
	void				*entry;
	unsigned long			flags;
	int				pc;
};

209
void *trace_event_buffer_reserve(struct trace_event_buffer *fbuffer,
210
				  struct trace_event_file *trace_file,
211 212
				  unsigned long len);

213
void trace_event_buffer_commit(struct trace_event_buffer *fbuffer);
214

215 216
enum {
	TRACE_EVENT_FL_FILTERED_BIT,
217
	TRACE_EVENT_FL_CAP_ANY_BIT,
L
Li Zefan 已提交
218
	TRACE_EVENT_FL_NO_SET_FILTER_BIT,
219
	TRACE_EVENT_FL_IGNORE_ENABLE_BIT,
220
	TRACE_EVENT_FL_TRACEPOINT_BIT,
A
Alexei Starovoitov 已提交
221
	TRACE_EVENT_FL_KPROBE_BIT,
222
	TRACE_EVENT_FL_UPROBE_BIT,
223 224
};

225 226 227 228 229
/*
 * 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
230
 *  IGNORE_ENABLE - For trace internal events, do not enable with debugfs file
231
 *  TRACEPOINT    - Event is a tracepoint
A
Alexei Starovoitov 已提交
232
 *  KPROBE        - Event is a kprobe
233
 *  UPROBE        - Event is a uprobe
234
 */
235
enum {
236
	TRACE_EVENT_FL_FILTERED		= (1 << TRACE_EVENT_FL_FILTERED_BIT),
237
	TRACE_EVENT_FL_CAP_ANY		= (1 << TRACE_EVENT_FL_CAP_ANY_BIT),
L
Li Zefan 已提交
238
	TRACE_EVENT_FL_NO_SET_FILTER	= (1 << TRACE_EVENT_FL_NO_SET_FILTER_BIT),
239
	TRACE_EVENT_FL_IGNORE_ENABLE	= (1 << TRACE_EVENT_FL_IGNORE_ENABLE_BIT),
240
	TRACE_EVENT_FL_TRACEPOINT	= (1 << TRACE_EVENT_FL_TRACEPOINT_BIT),
A
Alexei Starovoitov 已提交
241
	TRACE_EVENT_FL_KPROBE		= (1 << TRACE_EVENT_FL_KPROBE_BIT),
242
	TRACE_EVENT_FL_UPROBE		= (1 << TRACE_EVENT_FL_UPROBE_BIT),
243 244
};

245 246
#define TRACE_EVENT_FL_UKPROBE (TRACE_EVENT_FL_KPROBE | TRACE_EVENT_FL_UPROBE)

247
struct trace_event_call {
248
	struct list_head	list;
249
	struct trace_event_class *class;
250 251 252 253 254
	union {
		char			*name;
		/* Set TRACE_EVENT_FL_TRACEPOINT flag when using "tp" */
		struct tracepoint	*tp;
	};
255
	struct trace_event	event;
256
	char			*print_fmt;
257
	struct event_filter	*filter;
258
	void			*mod;
259
	void			*data;
260 261 262 263
	/*
	 *   bit 0:		filter_active
	 *   bit 1:		allow trace by non root (cap any)
	 *   bit 2:		failed to apply filter
264
	 *   bit 3:		trace internal event (do not enable)
265
	 *   bit 4:		Event was enabled by module
266
	 *   bit 5:		use call filter rather than file filter
267
	 *   bit 6:		Event is a tracepoint
268
	 */
269 270 271 272 273
	int			flags; /* static flags of different events */

#ifdef CONFIG_PERF_EVENTS
	int				perf_refcount;
	struct hlist_head __percpu	*perf_events;
274
	struct bpf_prog			*prog;
275
	struct perf_event		*bpf_prog_owner;
276

277
	int	(*perf_perm)(struct trace_event_call *,
278
			     struct perf_event *);
279 280 281
#endif
};

282
static inline const char *
283
trace_event_name(struct trace_event_call *call)
284 285 286 287 288 289 290
{
	if (call->flags & TRACE_EVENT_FL_TRACEPOINT)
		return call->tp ? call->tp->name : NULL;
	else
		return call->name;
}

291
struct trace_array;
292
struct trace_subsystem_dir;
293 294

enum {
295 296
	EVENT_FILE_FL_ENABLED_BIT,
	EVENT_FILE_FL_RECORDED_CMD_BIT,
297
	EVENT_FILE_FL_RECORDED_TGID_BIT,
298 299 300 301 302 303
	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
	EVENT_FILE_FL_WAS_ENABLED_BIT,
306 307 308
};

/*
309
 * Event file flags:
310
 *  ENABLED	  - The event is enabled
311
 *  RECORDED_CMD  - The comms should be recorded at sched_switch
312
 *  RECORDED_TGID - The tgids should be recorded at sched_switch
313 314
 *  FILTERED	  - The event has a filter attached
 *  NO_SET_FILTER - Set when filter has error and is to be ignored
315 316 317
 *  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)
318
 *  TRIGGER_MODE  - When set, invoke the triggers associated with the event
319
 *  TRIGGER_COND  - When set, one or more triggers has an associated filter
320
 *  PID_FILTER    - When set, the event is filtered based on pid
321
 *  WAS_ENABLED   - Set when enabled to know to clear trace on module removal
322 323
 */
enum {
324 325
	EVENT_FILE_FL_ENABLED		= (1 << EVENT_FILE_FL_ENABLED_BIT),
	EVENT_FILE_FL_RECORDED_CMD	= (1 << EVENT_FILE_FL_RECORDED_CMD_BIT),
326
	EVENT_FILE_FL_RECORDED_TGID	= (1 << EVENT_FILE_FL_RECORDED_TGID_BIT),
327 328 329 330 331 332
	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),
333
	EVENT_FILE_FL_PID_FILTER	= (1 << EVENT_FILE_FL_PID_FILTER_BIT),
334
	EVENT_FILE_FL_WAS_ENABLED	= (1 << EVENT_FILE_FL_WAS_ENABLED_BIT),
335 336
};

337
struct trace_event_file {
338
	struct list_head		list;
339
	struct trace_event_call		*event_call;
340
	struct event_filter __rcu	*filter;
341 342
	struct dentry			*dir;
	struct trace_array		*tr;
343
	struct trace_subsystem_dir	*system;
344
	struct list_head		triggers;
345

346 347
	/*
	 * 32 bit flags:
348 349
	 *   bit 0:		enabled
	 *   bit 1:		enabled cmd record
350 351
	 *   bit 2:		enable/disable with the soft disable bit
	 *   bit 3:		soft disabled
352
	 *   bit 4:		trigger enabled
353
	 *
354 355 356
	 * 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
357
	 * is currently used, these changes do not affect the code
358 359
	 * except that when a change is made, it may have a slight
	 * delay in propagating the changes to other CPUs due to
360
	 * caching and such. Which is mostly OK ;-)
361
	 */
362
	unsigned long		flags;
363
	atomic_t		sm_ref;	/* soft-mode reference counter */
364
	atomic_t		tm_ref;	/* trigger-mode reference counter */
365 366
};

367 368 369
#define __TRACE_EVENT_FLAGS(name, value)				\
	static int __init trace_init_flags_##name(void)			\
	{								\
370
		event_##name.flags |= value;				\
371 372 373 374
		return 0;						\
	}								\
	early_initcall(trace_init_flags_##name);

375
#define __TRACE_EVENT_PERF_PERM(name, expr...)				\
376
	static int perf_perm_##name(struct trace_event_call *tp_event, \
377 378 379 380 381 382 383 384 385 386 387
				    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);

388
#define PERF_MAX_TRACE_SIZE	2048
389

390
#define MAX_FILTER_STR_VAL	256	/* Should handle KSYM_SYMBOL_LEN */
391

392 393
enum event_trigger_type {
	ETT_NONE		= (0),
394
	ETT_TRACE_ONOFF		= (1 << 0),
395
	ETT_SNAPSHOT		= (1 << 1),
396
	ETT_STACKTRACE		= (1 << 2),
397
	ETT_EVENT_ENABLE	= (1 << 3),
398
	ETT_EVENT_HIST		= (1 << 4),
399
	ETT_HIST_ENABLE		= (1 << 5),
400 401
};

L
Li Zefan 已提交
402
extern int filter_match_preds(struct event_filter *filter, void *rec);
403

404
extern enum event_trigger_type event_triggers_call(struct trace_event_file *file,
405
						   void *rec);
406
extern void event_triggers_post_call(struct trace_event_file *file,
407 408
				     enum event_trigger_type tt,
				     void *rec);
409

410 411
bool trace_event_ignore_this_pid(struct trace_event_file *trace_file);

412
/**
413
 * trace_trigger_soft_disabled - do triggers and test if soft disabled
414 415 416 417 418 419 420 421
 * @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
422
trace_trigger_soft_disabled(struct trace_event_file *file)
423 424 425
{
	unsigned long eflags = file->flags;

426 427
	if (!(eflags & EVENT_FILE_FL_TRIGGER_COND)) {
		if (eflags & EVENT_FILE_FL_TRIGGER_MODE)
428
			event_triggers_call(file, NULL);
429
		if (eflags & EVENT_FILE_FL_SOFT_DISABLED)
430
			return true;
431 432
		if (eflags & EVENT_FILE_FL_PID_FILTER)
			return trace_event_ignore_this_pid(file);
433 434 435 436
	}
	return false;
}

437
#ifdef CONFIG_BPF_EVENTS
438 439 440 441 442 443 444 445
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

446 447 448 449
enum {
	FILTER_OTHER = 0,
	FILTER_STATIC_STRING,
	FILTER_DYN_STRING,
450
	FILTER_PTR_STRING,
451
	FILTER_TRACE_FN,
452 453
	FILTER_COMM,
	FILTER_CPU,
454 455
};

456 457
extern int trace_event_raw_init(struct trace_event_call *call);
extern int trace_define_field(struct trace_event_call *call, const char *type,
458 459
			      const char *name, int offset, int size,
			      int is_signed, int filter_type);
460 461
extern int trace_add_event_call(struct trace_event_call *call);
extern int trace_remove_event_call(struct trace_event_call *call);
462
extern int trace_event_get_offsets(struct trace_event_call *call);
463

464
#define is_signed_type(type)	(((type)(-1)) < (type)1)
465

466 467
int trace_set_clr_event(const char *system, const char *event, int set);

468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486
/*
 * 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)

487
#ifdef CONFIG_PERF_EVENTS
L
Li Zefan 已提交
488
struct perf_event;
489 490 491

DECLARE_PER_CPU(struct pt_regs, perf_trace_regs);

492 493
extern int  perf_trace_init(struct perf_event *event);
extern void perf_trace_destroy(struct perf_event *event);
P
Peter Zijlstra 已提交
494 495
extern int  perf_trace_add(struct perf_event *event, int flags);
extern void perf_trace_del(struct perf_event *event, int flags);
496
extern int  ftrace_profile_set_filter(struct perf_event *event, int event_id,
L
Li Zefan 已提交
497 498
				     char *filter_str);
extern void ftrace_profile_free_filter(struct perf_event *event);
499 500
void perf_trace_buf_update(void *record, u16 type);
void *perf_trace_buf_alloc(int size, struct pt_regs **regs, int *rctxp);
501

502 503 504 505 506
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);

507
static inline void
508
perf_trace_buf_submit(void *raw_data, int size, int rctx, u16 type,
509
		       u64 count, struct pt_regs *regs, void *head,
510
		       struct task_struct *task, struct perf_event *event)
511
{
512
	perf_tp_event(type, count, raw_data, size, regs, head, rctx, task, event);
513
}
L
Li Zefan 已提交
514 515
#endif

516
#endif /* _LINUX_TRACE_EVENT_H */