1 /* SPDX-License-Identifier: GPL-2.0 */
2
3 #ifndef _LINUX_TRACE_EVENT_H
4 #define _LINUX_TRACE_EVENT_H
5
6 #include <linux/ring_buffer.h>
7 #include <linux/trace_seq.h>
8 #include <linux/percpu.h>
9 #include <linux/hardirq.h>
10 #include <linux/perf_event.h>
11 #include <linux/tracepoint.h>
12
13 struct trace_array;
14 struct array_buffer;
15 struct tracer;
16 struct dentry;
17 struct bpf_prog;
18 union bpf_attr;
19
20 /* Used for event string fields when they are NULL */
21 #define EVENT_NULL_STR "(null)"
22
23 const char *trace_print_flags_seq(struct trace_seq *p, const char *delim,
24 unsigned long flags,
25 const struct trace_print_flags *flag_array);
26
27 const char *trace_print_symbols_seq(struct trace_seq *p, unsigned long val,
28 const struct trace_print_flags *symbol_array);
29
30 #if BITS_PER_LONG == 32
31 const char *trace_print_flags_seq_u64(struct trace_seq *p, const char *delim,
32 unsigned long long flags,
33 const struct trace_print_flags_u64 *flag_array);
34
35 const char *trace_print_symbols_seq_u64(struct trace_seq *p,
36 unsigned long long val,
37 const struct trace_print_flags_u64
38 *symbol_array);
39 #endif
40
41 const char *trace_print_bitmask_seq(struct trace_seq *p, void *bitmask_ptr,
42 unsigned int bitmask_size);
43
44 const char *trace_print_hex_seq(struct trace_seq *p,
45 const unsigned char *buf, int len,
46 bool concatenate);
47
48 const char *trace_print_array_seq(struct trace_seq *p,
49 const void *buf, int count,
50 size_t el_size);
51
52 const char *
53 trace_print_hex_dump_seq(struct trace_seq *p, const char *prefix_str,
54 int prefix_type, int rowsize, int groupsize,
55 const void *buf, size_t len, bool ascii);
56
57 struct trace_iterator;
58 struct trace_event;
59
60 int trace_raw_output_prep(struct trace_iterator *iter,
61 struct trace_event *event);
62 extern __printf(2, 3)
63 void trace_event_printf(struct trace_iterator *iter, const char *fmt, ...);
64
65 /* Used to find the offset and length of dynamic fields in trace events */
66 struct trace_dynamic_info {
67 #ifdef CONFIG_CPU_BIG_ENDIAN
68 u16 len;
69 u16 offset;
70 #else
71 u16 offset;
72 u16 len;
73 #endif
74 } __packed;
75
76 /*
77 * The trace entry - the most basic unit of tracing. This is what
78 * is printed in the end as a single line in the trace output, such as:
79 *
80 * bash-15816 [01] 235.197585: idle_cpu <- irq_enter
81 */
82 struct trace_entry {
83 unsigned short type;
84 unsigned char flags;
85 unsigned char preempt_count;
86 int pid;
87 };
88
89 #define TRACE_EVENT_TYPE_MAX \
90 ((1 << (sizeof(((struct trace_entry *)0)->type) * 8)) - 1)
91
92 /*
93 * Trace iterator - used by printout routines who present trace
94 * results to users and which routines might sleep, etc:
95 */
96 struct trace_iterator {
97 struct trace_array *tr;
98 struct tracer *trace;
99 struct array_buffer *array_buffer;
100 void *private;
101 int cpu_file;
102 struct mutex mutex;
103 struct ring_buffer_iter **buffer_iter;
104 unsigned long iter_flags;
105 void *temp; /* temp holder */
106 unsigned int temp_size;
107 char *fmt; /* modified format holder */
108 unsigned int fmt_size;
109 atomic_t wait_index;
110
111 /* trace_seq for __print_flags() and __print_symbolic() etc. */
112 struct trace_seq tmp_seq;
113
114 cpumask_var_t started;
115
116 /* Set when the file is closed to prevent new waiters */
117 bool closed;
118
119 /* it's true when current open file is snapshot */
120 bool snapshot;
121
122 /* The below is zeroed out in pipe_read */
123 struct trace_seq seq;
124 struct trace_entry *ent;
125 unsigned long lost_events;
126 int leftover;
127 int ent_size;
128 int cpu;
129 u64 ts;
130
131 loff_t pos;
132 long idx;
133
134 /* All new field here will be zeroed out in pipe_read */
135 };
136
137 enum trace_iter_flags {
138 TRACE_FILE_LAT_FMT = 1,
139 TRACE_FILE_ANNOTATE = 2,
140 TRACE_FILE_TIME_IN_NS = 4,
141 };
142
143
144 typedef enum print_line_t (*trace_print_func)(struct trace_iterator *iter,
145 int flags, struct trace_event *event);
146
147 struct trace_event_functions {
148 trace_print_func trace;
149 trace_print_func raw;
150 trace_print_func hex;
151 trace_print_func binary;
152 };
153
154 struct trace_event {
155 struct hlist_node node;
156 int type;
157 struct trace_event_functions *funcs;
158 };
159
160 extern int register_trace_event(struct trace_event *event);
161 extern int unregister_trace_event(struct trace_event *event);
162
163 /* Return values for print_line callback */
164 enum print_line_t {
165 TRACE_TYPE_PARTIAL_LINE = 0, /* Retry after flushing the seq */
166 TRACE_TYPE_HANDLED = 1,
167 TRACE_TYPE_UNHANDLED = 2, /* Relay to other output functions */
168 TRACE_TYPE_NO_CONSUME = 3 /* Handled but ask to not consume */
169 };
170
171 enum print_line_t trace_handle_return(struct trace_seq *s);
172
tracing_generic_entry_update(struct trace_entry * entry,unsigned short type,unsigned int trace_ctx)173 static inline void tracing_generic_entry_update(struct trace_entry *entry,
174 unsigned short type,
175 unsigned int trace_ctx)
176 {
177 entry->preempt_count = trace_ctx & 0xff;
178 entry->pid = current->pid;
179 entry->type = type;
180 entry->flags = trace_ctx >> 16;
181 }
182
183 unsigned int tracing_gen_ctx_irq_test(unsigned int irqs_status);
184
185 enum trace_flag_type {
186 TRACE_FLAG_IRQS_OFF = 0x01,
187 TRACE_FLAG_NEED_RESCHED_LAZY = 0x02,
188 TRACE_FLAG_NEED_RESCHED = 0x04,
189 TRACE_FLAG_HARDIRQ = 0x08,
190 TRACE_FLAG_SOFTIRQ = 0x10,
191 TRACE_FLAG_PREEMPT_RESCHED = 0x20,
192 TRACE_FLAG_NMI = 0x40,
193 TRACE_FLAG_BH_OFF = 0x80,
194 };
195
tracing_gen_ctx_flags(unsigned long irqflags)196 static inline unsigned int tracing_gen_ctx_flags(unsigned long irqflags)
197 {
198 unsigned int irq_status = irqs_disabled_flags(irqflags) ?
199 TRACE_FLAG_IRQS_OFF : 0;
200 return tracing_gen_ctx_irq_test(irq_status);
201 }
tracing_gen_ctx(void)202 static inline unsigned int tracing_gen_ctx(void)
203 {
204 unsigned long irqflags;
205
206 local_save_flags(irqflags);
207 return tracing_gen_ctx_flags(irqflags);
208 }
209
tracing_gen_ctx_dec(void)210 static inline unsigned int tracing_gen_ctx_dec(void)
211 {
212 unsigned int trace_ctx;
213
214 trace_ctx = tracing_gen_ctx();
215 /*
216 * Subtract one from the preemption counter if preemption is enabled,
217 * see trace_event_buffer_reserve()for details.
218 */
219 if (IS_ENABLED(CONFIG_PREEMPTION))
220 trace_ctx--;
221 return trace_ctx;
222 }
223
224 struct trace_event_file;
225
226 struct ring_buffer_event *
227 trace_event_buffer_lock_reserve(struct trace_buffer **current_buffer,
228 struct trace_event_file *trace_file,
229 int type, unsigned long len,
230 unsigned int trace_ctx);
231
232 #define TRACE_RECORD_CMDLINE BIT(0)
233 #define TRACE_RECORD_TGID BIT(1)
234
235 void tracing_record_taskinfo(struct task_struct *task, int flags);
236 void tracing_record_taskinfo_sched_switch(struct task_struct *prev,
237 struct task_struct *next, int flags);
238
239 void tracing_record_cmdline(struct task_struct *task);
240 void tracing_record_tgid(struct task_struct *task);
241
242 int trace_output_call(struct trace_iterator *iter, char *name, char *fmt, ...)
243 __printf(3, 4);
244
245 struct event_filter;
246
247 enum trace_reg {
248 TRACE_REG_REGISTER,
249 TRACE_REG_UNREGISTER,
250 #ifdef CONFIG_PERF_EVENTS
251 TRACE_REG_PERF_REGISTER,
252 TRACE_REG_PERF_UNREGISTER,
253 TRACE_REG_PERF_OPEN,
254 TRACE_REG_PERF_CLOSE,
255 /*
256 * These (ADD/DEL) use a 'boolean' return value, where 1 (true) means a
257 * custom action was taken and the default action is not to be
258 * performed.
259 */
260 TRACE_REG_PERF_ADD,
261 TRACE_REG_PERF_DEL,
262 #endif
263 };
264
265 struct trace_event_call;
266
267 #define TRACE_FUNCTION_TYPE ((const char *)~0UL)
268
269 struct trace_event_fields {
270 const char *type;
271 union {
272 struct {
273 const char *name;
274 const int size;
275 const int align;
276 const unsigned int is_signed:1;
277 unsigned int needs_test:1;
278 const int filter_type;
279 const int len;
280 };
281 int (*define_fields)(struct trace_event_call *);
282 };
283 };
284
285 struct trace_event_class {
286 const char *system;
287 void *probe;
288 #ifdef CONFIG_PERF_EVENTS
289 void *perf_probe;
290 #endif
291 int (*reg)(struct trace_event_call *event,
292 enum trace_reg type, void *data);
293 struct trace_event_fields *fields_array;
294 struct list_head *(*get_fields)(struct trace_event_call *);
295 struct list_head fields;
296 int (*raw_init)(struct trace_event_call *);
297 };
298
299 extern int trace_event_reg(struct trace_event_call *event,
300 enum trace_reg type, void *data);
301
302 struct trace_event_buffer {
303 struct trace_buffer *buffer;
304 struct ring_buffer_event *event;
305 struct trace_event_file *trace_file;
306 void *entry;
307 unsigned int trace_ctx;
308 struct pt_regs *regs;
309 };
310
311 void *trace_event_buffer_reserve(struct trace_event_buffer *fbuffer,
312 struct trace_event_file *trace_file,
313 unsigned long len);
314
315 void trace_event_buffer_commit(struct trace_event_buffer *fbuffer);
316
317 enum {
318 TRACE_EVENT_FL_CAP_ANY_BIT,
319 TRACE_EVENT_FL_NO_SET_FILTER_BIT,
320 TRACE_EVENT_FL_IGNORE_ENABLE_BIT,
321 TRACE_EVENT_FL_TRACEPOINT_BIT,
322 TRACE_EVENT_FL_DYNAMIC_BIT,
323 TRACE_EVENT_FL_KPROBE_BIT,
324 TRACE_EVENT_FL_UPROBE_BIT,
325 TRACE_EVENT_FL_EPROBE_BIT,
326 TRACE_EVENT_FL_FPROBE_BIT,
327 TRACE_EVENT_FL_CUSTOM_BIT,
328 TRACE_EVENT_FL_TEST_STR_BIT,
329 };
330
331 /*
332 * Event flags:
333 * CAP_ANY - Any user can enable for perf
334 * NO_SET_FILTER - Set when filter has error and is to be ignored
335 * IGNORE_ENABLE - For trace internal events, do not enable with debugfs file
336 * TRACEPOINT - Event is a tracepoint
337 * DYNAMIC - Event is a dynamic event (created at run time)
338 * KPROBE - Event is a kprobe
339 * UPROBE - Event is a uprobe
340 * EPROBE - Event is an event probe
341 * FPROBE - Event is an function probe
342 * CUSTOM - Event is a custom event (to be attached to an exsiting tracepoint)
343 * This is set when the custom event has not been attached
344 * to a tracepoint yet, then it is cleared when it is.
345 * TEST_STR - The event has a "%s" that points to a string outside the event
346 */
347 enum {
348 TRACE_EVENT_FL_CAP_ANY = (1 << TRACE_EVENT_FL_CAP_ANY_BIT),
349 TRACE_EVENT_FL_NO_SET_FILTER = (1 << TRACE_EVENT_FL_NO_SET_FILTER_BIT),
350 TRACE_EVENT_FL_IGNORE_ENABLE = (1 << TRACE_EVENT_FL_IGNORE_ENABLE_BIT),
351 TRACE_EVENT_FL_TRACEPOINT = (1 << TRACE_EVENT_FL_TRACEPOINT_BIT),
352 TRACE_EVENT_FL_DYNAMIC = (1 << TRACE_EVENT_FL_DYNAMIC_BIT),
353 TRACE_EVENT_FL_KPROBE = (1 << TRACE_EVENT_FL_KPROBE_BIT),
354 TRACE_EVENT_FL_UPROBE = (1 << TRACE_EVENT_FL_UPROBE_BIT),
355 TRACE_EVENT_FL_EPROBE = (1 << TRACE_EVENT_FL_EPROBE_BIT),
356 TRACE_EVENT_FL_FPROBE = (1 << TRACE_EVENT_FL_FPROBE_BIT),
357 TRACE_EVENT_FL_CUSTOM = (1 << TRACE_EVENT_FL_CUSTOM_BIT),
358 TRACE_EVENT_FL_TEST_STR = (1 << TRACE_EVENT_FL_TEST_STR_BIT),
359 };
360
361 #define TRACE_EVENT_FL_UKPROBE (TRACE_EVENT_FL_KPROBE | TRACE_EVENT_FL_UPROBE)
362
363 struct trace_event_call {
364 struct list_head list;
365 struct trace_event_class *class;
366 union {
367 const char *name;
368 /* Set TRACE_EVENT_FL_TRACEPOINT flag when using "tp" */
369 struct tracepoint *tp;
370 };
371 struct trace_event event;
372 char *print_fmt;
373 /*
374 * Static events can disappear with modules,
375 * where as dynamic ones need their own ref count.
376 */
377 union {
378 void *module;
379 atomic_t refcnt;
380 };
381 void *data;
382
383 /* See the TRACE_EVENT_FL_* flags above */
384 int flags; /* static flags of different events */
385
386 #ifdef CONFIG_PERF_EVENTS
387 int perf_refcount;
388 struct hlist_head __percpu *perf_events;
389 struct bpf_prog_array __rcu *prog_array;
390
391 int (*perf_perm)(struct trace_event_call *,
392 struct perf_event *);
393 #endif
394 };
395
396 #ifdef CONFIG_DYNAMIC_EVENTS
397 bool trace_event_dyn_try_get_ref(struct trace_event_call *call);
398 void trace_event_dyn_put_ref(struct trace_event_call *call);
399 bool trace_event_dyn_busy(struct trace_event_call *call);
400 #else
trace_event_dyn_try_get_ref(struct trace_event_call * call)401 static inline bool trace_event_dyn_try_get_ref(struct trace_event_call *call)
402 {
403 /* Without DYNAMIC_EVENTS configured, nothing should be calling this */
404 return false;
405 }
trace_event_dyn_put_ref(struct trace_event_call * call)406 static inline void trace_event_dyn_put_ref(struct trace_event_call *call)
407 {
408 }
trace_event_dyn_busy(struct trace_event_call * call)409 static inline bool trace_event_dyn_busy(struct trace_event_call *call)
410 {
411 /* Nothing should call this without DYNAIMIC_EVENTS configured. */
412 return true;
413 }
414 #endif
415
trace_event_try_get_ref(struct trace_event_call * call)416 static inline bool trace_event_try_get_ref(struct trace_event_call *call)
417 {
418 if (call->flags & TRACE_EVENT_FL_DYNAMIC)
419 return trace_event_dyn_try_get_ref(call);
420 else
421 return try_module_get(call->module);
422 }
423
trace_event_put_ref(struct trace_event_call * call)424 static inline void trace_event_put_ref(struct trace_event_call *call)
425 {
426 if (call->flags & TRACE_EVENT_FL_DYNAMIC)
427 trace_event_dyn_put_ref(call);
428 else
429 module_put(call->module);
430 }
431
432 #ifdef CONFIG_PERF_EVENTS
bpf_prog_array_valid(struct trace_event_call * call)433 static inline bool bpf_prog_array_valid(struct trace_event_call *call)
434 {
435 /*
436 * This inline function checks whether call->prog_array
437 * is valid or not. The function is called in various places,
438 * outside rcu_read_lock/unlock, as a heuristic to speed up execution.
439 *
440 * If this function returns true, and later call->prog_array
441 * becomes false inside rcu_read_lock/unlock region,
442 * we bail out then. If this function return false,
443 * there is a risk that we might miss a few events if the checking
444 * were delayed until inside rcu_read_lock/unlock region and
445 * call->prog_array happened to become non-NULL then.
446 *
447 * Here, READ_ONCE() is used instead of rcu_access_pointer().
448 * rcu_access_pointer() requires the actual definition of
449 * "struct bpf_prog_array" while READ_ONCE() only needs
450 * a declaration of the same type.
451 */
452 return !!READ_ONCE(call->prog_array);
453 }
454 #endif
455
456 static inline const char *
trace_event_name(struct trace_event_call * call)457 trace_event_name(struct trace_event_call *call)
458 {
459 if (call->flags & TRACE_EVENT_FL_CUSTOM)
460 return call->name;
461 else if (call->flags & TRACE_EVENT_FL_TRACEPOINT)
462 return call->tp ? call->tp->name : NULL;
463 else
464 return call->name;
465 }
466
467 static inline struct list_head *
trace_get_fields(struct trace_event_call * event_call)468 trace_get_fields(struct trace_event_call *event_call)
469 {
470 if (!event_call->class->get_fields)
471 return &event_call->class->fields;
472 return event_call->class->get_fields(event_call);
473 }
474
475 struct trace_subsystem_dir;
476
477 enum {
478 EVENT_FILE_FL_ENABLED_BIT,
479 EVENT_FILE_FL_RECORDED_CMD_BIT,
480 EVENT_FILE_FL_RECORDED_TGID_BIT,
481 EVENT_FILE_FL_FILTERED_BIT,
482 EVENT_FILE_FL_NO_SET_FILTER_BIT,
483 EVENT_FILE_FL_SOFT_DISABLED_BIT,
484 EVENT_FILE_FL_TRIGGER_MODE_BIT,
485 EVENT_FILE_FL_TRIGGER_COND_BIT,
486 EVENT_FILE_FL_PID_FILTER_BIT,
487 EVENT_FILE_FL_WAS_ENABLED_BIT,
488 EVENT_FILE_FL_FREED_BIT,
489 };
490
491 extern struct trace_event_file *trace_get_event_file(const char *instance,
492 const char *system,
493 const char *event);
494 extern void trace_put_event_file(struct trace_event_file *file);
495
496 #define MAX_DYNEVENT_CMD_LEN (2048)
497
498 enum dynevent_type {
499 DYNEVENT_TYPE_SYNTH = 1,
500 DYNEVENT_TYPE_KPROBE,
501 DYNEVENT_TYPE_NONE,
502 };
503
504 struct dynevent_cmd;
505
506 typedef int (*dynevent_create_fn_t)(struct dynevent_cmd *cmd);
507
508 struct dynevent_cmd {
509 struct seq_buf seq;
510 const char *event_name;
511 unsigned int n_fields;
512 enum dynevent_type type;
513 dynevent_create_fn_t run_command;
514 void *private_data;
515 };
516
517 extern int dynevent_create(struct dynevent_cmd *cmd);
518
519 extern int synth_event_delete(const char *name);
520
521 extern void synth_event_cmd_init(struct dynevent_cmd *cmd,
522 char *buf, int maxlen);
523
524 extern int __synth_event_gen_cmd_start(struct dynevent_cmd *cmd,
525 const char *name,
526 struct module *mod, ...);
527
528 #define synth_event_gen_cmd_start(cmd, name, mod, ...) \
529 __synth_event_gen_cmd_start(cmd, name, mod, ## __VA_ARGS__, NULL)
530
531 struct synth_field_desc {
532 const char *type;
533 const char *name;
534 };
535
536 extern int synth_event_gen_cmd_array_start(struct dynevent_cmd *cmd,
537 const char *name,
538 struct module *mod,
539 struct synth_field_desc *fields,
540 unsigned int n_fields);
541 extern int synth_event_create(const char *name,
542 struct synth_field_desc *fields,
543 unsigned int n_fields, struct module *mod);
544
545 extern int synth_event_add_field(struct dynevent_cmd *cmd,
546 const char *type,
547 const char *name);
548 extern int synth_event_add_field_str(struct dynevent_cmd *cmd,
549 const char *type_name);
550 extern int synth_event_add_fields(struct dynevent_cmd *cmd,
551 struct synth_field_desc *fields,
552 unsigned int n_fields);
553
554 #define synth_event_gen_cmd_end(cmd) \
555 dynevent_create(cmd)
556
557 struct synth_event;
558
559 struct synth_event_trace_state {
560 struct trace_event_buffer fbuffer;
561 struct synth_trace_event *entry;
562 struct trace_buffer *buffer;
563 struct synth_event *event;
564 unsigned int cur_field;
565 unsigned int n_u64;
566 bool disabled;
567 bool add_next;
568 bool add_name;
569 };
570
571 extern int synth_event_trace(struct trace_event_file *file,
572 unsigned int n_vals, ...);
573 extern int synth_event_trace_array(struct trace_event_file *file, u64 *vals,
574 unsigned int n_vals);
575 extern int synth_event_trace_start(struct trace_event_file *file,
576 struct synth_event_trace_state *trace_state);
577 extern int synth_event_add_next_val(u64 val,
578 struct synth_event_trace_state *trace_state);
579 extern int synth_event_add_val(const char *field_name, u64 val,
580 struct synth_event_trace_state *trace_state);
581 extern int synth_event_trace_end(struct synth_event_trace_state *trace_state);
582
583 extern int kprobe_event_delete(const char *name);
584
585 extern void kprobe_event_cmd_init(struct dynevent_cmd *cmd,
586 char *buf, int maxlen);
587
588 #define kprobe_event_gen_cmd_start(cmd, name, loc, ...) \
589 __kprobe_event_gen_cmd_start(cmd, false, name, loc, ## __VA_ARGS__, NULL)
590
591 #define kretprobe_event_gen_cmd_start(cmd, name, loc, ...) \
592 __kprobe_event_gen_cmd_start(cmd, true, name, loc, ## __VA_ARGS__, NULL)
593
594 extern int __kprobe_event_gen_cmd_start(struct dynevent_cmd *cmd,
595 bool kretprobe,
596 const char *name,
597 const char *loc, ...);
598
599 #define kprobe_event_add_fields(cmd, ...) \
600 __kprobe_event_add_fields(cmd, ## __VA_ARGS__, NULL)
601
602 #define kprobe_event_add_field(cmd, field) \
603 __kprobe_event_add_fields(cmd, field, NULL)
604
605 extern int __kprobe_event_add_fields(struct dynevent_cmd *cmd, ...);
606
607 #define kprobe_event_gen_cmd_end(cmd) \
608 dynevent_create(cmd)
609
610 #define kretprobe_event_gen_cmd_end(cmd) \
611 dynevent_create(cmd)
612
613 /*
614 * Event file flags:
615 * ENABLED - The event is enabled
616 * RECORDED_CMD - The comms should be recorded at sched_switch
617 * RECORDED_TGID - The tgids should be recorded at sched_switch
618 * FILTERED - The event has a filter attached
619 * NO_SET_FILTER - Set when filter has error and is to be ignored
620 * SOFT_DISABLED - When set, do not trace the event (even though its
621 * tracepoint may be enabled)
622 * TRIGGER_MODE - When set, invoke the triggers associated with the event
623 * TRIGGER_COND - When set, one or more triggers has an associated filter
624 * PID_FILTER - When set, the event is filtered based on pid
625 * WAS_ENABLED - Set when enabled to know to clear trace on module removal
626 * FREED - File descriptor is freed, all fields should be considered invalid
627 */
628 enum {
629 EVENT_FILE_FL_ENABLED = (1 << EVENT_FILE_FL_ENABLED_BIT),
630 EVENT_FILE_FL_RECORDED_CMD = (1 << EVENT_FILE_FL_RECORDED_CMD_BIT),
631 EVENT_FILE_FL_RECORDED_TGID = (1 << EVENT_FILE_FL_RECORDED_TGID_BIT),
632 EVENT_FILE_FL_FILTERED = (1 << EVENT_FILE_FL_FILTERED_BIT),
633 EVENT_FILE_FL_NO_SET_FILTER = (1 << EVENT_FILE_FL_NO_SET_FILTER_BIT),
634 EVENT_FILE_FL_SOFT_DISABLED = (1 << EVENT_FILE_FL_SOFT_DISABLED_BIT),
635 EVENT_FILE_FL_TRIGGER_MODE = (1 << EVENT_FILE_FL_TRIGGER_MODE_BIT),
636 EVENT_FILE_FL_TRIGGER_COND = (1 << EVENT_FILE_FL_TRIGGER_COND_BIT),
637 EVENT_FILE_FL_PID_FILTER = (1 << EVENT_FILE_FL_PID_FILTER_BIT),
638 EVENT_FILE_FL_WAS_ENABLED = (1 << EVENT_FILE_FL_WAS_ENABLED_BIT),
639 EVENT_FILE_FL_FREED = (1 << EVENT_FILE_FL_FREED_BIT),
640 };
641
642 struct trace_event_file {
643 struct list_head list;
644 struct trace_event_call *event_call;
645 struct event_filter __rcu *filter;
646 struct eventfs_inode *ei;
647 struct trace_array *tr;
648 struct trace_subsystem_dir *system;
649 struct list_head triggers;
650
651 /*
652 * 32 bit flags:
653 * bit 0: enabled
654 * bit 1: enabled cmd record
655 * bit 2: enable/disable with the soft disable bit
656 * bit 3: soft disabled
657 * bit 4: trigger enabled
658 *
659 * Note: The bits must be set atomically to prevent races
660 * from other writers. Reads of flags do not need to be in
661 * sync as they occur in critical sections. But the way flags
662 * is currently used, these changes do not affect the code
663 * except that when a change is made, it may have a slight
664 * delay in propagating the changes to other CPUs due to
665 * caching and such. Which is mostly OK ;-)
666 */
667 unsigned long flags;
668 refcount_t ref; /* ref count for opened files */
669 atomic_t sm_ref; /* soft-mode reference counter */
670 atomic_t tm_ref; /* trigger-mode reference counter */
671 };
672
673 #ifdef CONFIG_HIST_TRIGGERS
674 extern struct irq_work hist_poll_work;
675 extern wait_queue_head_t hist_poll_wq;
676
hist_poll_wakeup(void)677 static inline void hist_poll_wakeup(void)
678 {
679 if (wq_has_sleeper(&hist_poll_wq))
680 irq_work_queue(&hist_poll_work);
681 }
682
683 #define hist_poll_wait(file, wait) \
684 poll_wait(file, &hist_poll_wq, wait)
685 #endif
686
687 #define __TRACE_EVENT_FLAGS(name, value) \
688 static int __init trace_init_flags_##name(void) \
689 { \
690 event_##name.flags |= value; \
691 return 0; \
692 } \
693 early_initcall(trace_init_flags_##name);
694
695 #define __TRACE_EVENT_PERF_PERM(name, expr...) \
696 static int perf_perm_##name(struct trace_event_call *tp_event, \
697 struct perf_event *p_event) \
698 { \
699 return ({ expr; }); \
700 } \
701 static int __init trace_init_perf_perm_##name(void) \
702 { \
703 event_##name.perf_perm = &perf_perm_##name; \
704 return 0; \
705 } \
706 early_initcall(trace_init_perf_perm_##name);
707
708 #define PERF_MAX_TRACE_SIZE 8192
709
710 #define MAX_FILTER_STR_VAL 256U /* Should handle KSYM_SYMBOL_LEN */
711
712 enum event_trigger_type {
713 ETT_NONE = (0),
714 ETT_TRACE_ONOFF = (1 << 0),
715 ETT_SNAPSHOT = (1 << 1),
716 ETT_STACKTRACE = (1 << 2),
717 ETT_EVENT_ENABLE = (1 << 3),
718 ETT_EVENT_HIST = (1 << 4),
719 ETT_HIST_ENABLE = (1 << 5),
720 ETT_EVENT_EPROBE = (1 << 6),
721 };
722
723 extern int filter_match_preds(struct event_filter *filter, void *rec);
724
725 extern enum event_trigger_type
726 event_triggers_call(struct trace_event_file *file,
727 struct trace_buffer *buffer, void *rec,
728 struct ring_buffer_event *event);
729 extern void
730 event_triggers_post_call(struct trace_event_file *file,
731 enum event_trigger_type tt);
732
733 bool trace_event_ignore_this_pid(struct trace_event_file *trace_file);
734
735 bool __trace_trigger_soft_disabled(struct trace_event_file *file);
736
737 /**
738 * trace_trigger_soft_disabled - do triggers and test if soft disabled
739 * @file: The file pointer of the event to test
740 *
741 * If any triggers without filters are attached to this event, they
742 * will be called here. If the event is soft disabled and has no
743 * triggers that require testing the fields, it will return true,
744 * otherwise false.
745 */
746 static __always_inline bool
trace_trigger_soft_disabled(struct trace_event_file * file)747 trace_trigger_soft_disabled(struct trace_event_file *file)
748 {
749 unsigned long eflags = file->flags;
750
751 if (likely(!(eflags & (EVENT_FILE_FL_TRIGGER_MODE |
752 EVENT_FILE_FL_SOFT_DISABLED |
753 EVENT_FILE_FL_PID_FILTER))))
754 return false;
755
756 if (likely(eflags & EVENT_FILE_FL_TRIGGER_COND))
757 return false;
758
759 return __trace_trigger_soft_disabled(file);
760 }
761
762 #ifdef CONFIG_BPF_EVENTS
763 unsigned int trace_call_bpf(struct trace_event_call *call, void *ctx);
764 int perf_event_attach_bpf_prog(struct perf_event *event, struct bpf_prog *prog, u64 bpf_cookie);
765 void perf_event_detach_bpf_prog(struct perf_event *event);
766 int perf_event_query_prog_array(struct perf_event *event, void __user *info);
767
768 struct bpf_raw_tp_link;
769 int bpf_probe_register(struct bpf_raw_event_map *btp, struct bpf_raw_tp_link *link);
770 int bpf_probe_unregister(struct bpf_raw_event_map *btp, struct bpf_raw_tp_link *link);
771
772 struct bpf_raw_event_map *bpf_get_raw_tracepoint(const char *name);
773 void bpf_put_raw_tracepoint(struct bpf_raw_event_map *btp);
774 int bpf_get_perf_event_info(const struct perf_event *event, u32 *prog_id,
775 u32 *fd_type, const char **buf,
776 u64 *probe_offset, u64 *probe_addr,
777 unsigned long *missed);
778 int bpf_kprobe_multi_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
779 int bpf_uprobe_multi_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
780 #else
trace_call_bpf(struct trace_event_call * call,void * ctx)781 static inline unsigned int trace_call_bpf(struct trace_event_call *call, void *ctx)
782 {
783 return 1;
784 }
785
786 static inline int
perf_event_attach_bpf_prog(struct perf_event * event,struct bpf_prog * prog,u64 bpf_cookie)787 perf_event_attach_bpf_prog(struct perf_event *event, struct bpf_prog *prog, u64 bpf_cookie)
788 {
789 return -EOPNOTSUPP;
790 }
791
perf_event_detach_bpf_prog(struct perf_event * event)792 static inline void perf_event_detach_bpf_prog(struct perf_event *event) { }
793
794 static inline int
perf_event_query_prog_array(struct perf_event * event,void __user * info)795 perf_event_query_prog_array(struct perf_event *event, void __user *info)
796 {
797 return -EOPNOTSUPP;
798 }
799 struct bpf_raw_tp_link;
bpf_probe_register(struct bpf_raw_event_map * btp,struct bpf_raw_tp_link * link)800 static inline int bpf_probe_register(struct bpf_raw_event_map *btp, struct bpf_raw_tp_link *link)
801 {
802 return -EOPNOTSUPP;
803 }
bpf_probe_unregister(struct bpf_raw_event_map * btp,struct bpf_raw_tp_link * link)804 static inline int bpf_probe_unregister(struct bpf_raw_event_map *btp, struct bpf_raw_tp_link *link)
805 {
806 return -EOPNOTSUPP;
807 }
bpf_get_raw_tracepoint(const char * name)808 static inline struct bpf_raw_event_map *bpf_get_raw_tracepoint(const char *name)
809 {
810 return NULL;
811 }
bpf_put_raw_tracepoint(struct bpf_raw_event_map * btp)812 static inline void bpf_put_raw_tracepoint(struct bpf_raw_event_map *btp)
813 {
814 }
bpf_get_perf_event_info(const struct perf_event * event,u32 * prog_id,u32 * fd_type,const char ** buf,u64 * probe_offset,u64 * probe_addr,unsigned long * missed)815 static inline int bpf_get_perf_event_info(const struct perf_event *event,
816 u32 *prog_id, u32 *fd_type,
817 const char **buf, u64 *probe_offset,
818 u64 *probe_addr, unsigned long *missed)
819 {
820 return -EOPNOTSUPP;
821 }
822 static inline int
bpf_kprobe_multi_link_attach(const union bpf_attr * attr,struct bpf_prog * prog)823 bpf_kprobe_multi_link_attach(const union bpf_attr *attr, struct bpf_prog *prog)
824 {
825 return -EOPNOTSUPP;
826 }
827 static inline int
bpf_uprobe_multi_link_attach(const union bpf_attr * attr,struct bpf_prog * prog)828 bpf_uprobe_multi_link_attach(const union bpf_attr *attr, struct bpf_prog *prog)
829 {
830 return -EOPNOTSUPP;
831 }
832 #endif
833
834 enum {
835 FILTER_OTHER = 0,
836 FILTER_STATIC_STRING,
837 FILTER_DYN_STRING,
838 FILTER_RDYN_STRING,
839 FILTER_PTR_STRING,
840 FILTER_TRACE_FN,
841 FILTER_CPUMASK,
842 FILTER_COMM,
843 FILTER_CPU,
844 FILTER_STACKTRACE,
845 };
846
847 extern int trace_event_raw_init(struct trace_event_call *call);
848 extern int trace_define_field(struct trace_event_call *call, const char *type,
849 const char *name, int offset, int size,
850 int is_signed, int filter_type);
851 extern int trace_add_event_call(struct trace_event_call *call);
852 extern int trace_remove_event_call(struct trace_event_call *call);
853 extern int trace_event_get_offsets(struct trace_event_call *call);
854
855 int ftrace_set_clr_event(struct trace_array *tr, char *buf, int set);
856 int trace_set_clr_event(const char *system, const char *event, int set);
857 int trace_array_set_clr_event(struct trace_array *tr, const char *system,
858 const char *event, bool enable);
859
860 #ifdef CONFIG_PERF_EVENTS
861 struct perf_event;
862
863 DECLARE_PER_CPU(struct pt_regs, perf_trace_regs);
864
865 extern int perf_trace_init(struct perf_event *event);
866 extern void perf_trace_destroy(struct perf_event *event);
867 extern int perf_trace_add(struct perf_event *event, int flags);
868 extern void perf_trace_del(struct perf_event *event, int flags);
869 #ifdef CONFIG_KPROBE_EVENTS
870 extern int perf_kprobe_init(struct perf_event *event, bool is_retprobe);
871 extern void perf_kprobe_destroy(struct perf_event *event);
872 extern int bpf_get_kprobe_info(const struct perf_event *event,
873 u32 *fd_type, const char **symbol,
874 u64 *probe_offset, u64 *probe_addr,
875 unsigned long *missed,
876 bool perf_type_tracepoint);
877 #endif
878 #ifdef CONFIG_UPROBE_EVENTS
879 extern int perf_uprobe_init(struct perf_event *event,
880 unsigned long ref_ctr_offset, bool is_retprobe);
881 extern void perf_uprobe_destroy(struct perf_event *event);
882 extern int bpf_get_uprobe_info(const struct perf_event *event,
883 u32 *fd_type, const char **filename,
884 u64 *probe_offset, u64 *probe_addr,
885 bool perf_type_tracepoint);
886 #endif
887 extern int ftrace_profile_set_filter(struct perf_event *event, int event_id,
888 char *filter_str);
889 extern void ftrace_profile_free_filter(struct perf_event *event);
890 void perf_trace_buf_update(void *record, u16 type);
891 void *perf_trace_buf_alloc(int size, struct pt_regs **regs, int *rctxp);
892
893 int perf_event_set_bpf_prog(struct perf_event *event, struct bpf_prog *prog, u64 bpf_cookie);
894 void perf_event_free_bpf_prog(struct perf_event *event);
895
896 void bpf_trace_run1(struct bpf_raw_tp_link *link, u64 arg1);
897 void bpf_trace_run2(struct bpf_raw_tp_link *link, u64 arg1, u64 arg2);
898 void bpf_trace_run3(struct bpf_raw_tp_link *link, u64 arg1, u64 arg2,
899 u64 arg3);
900 void bpf_trace_run4(struct bpf_raw_tp_link *link, u64 arg1, u64 arg2,
901 u64 arg3, u64 arg4);
902 void bpf_trace_run5(struct bpf_raw_tp_link *link, u64 arg1, u64 arg2,
903 u64 arg3, u64 arg4, u64 arg5);
904 void bpf_trace_run6(struct bpf_raw_tp_link *link, u64 arg1, u64 arg2,
905 u64 arg3, u64 arg4, u64 arg5, u64 arg6);
906 void bpf_trace_run7(struct bpf_raw_tp_link *link, u64 arg1, u64 arg2,
907 u64 arg3, u64 arg4, u64 arg5, u64 arg6, u64 arg7);
908 void bpf_trace_run8(struct bpf_raw_tp_link *link, u64 arg1, u64 arg2,
909 u64 arg3, u64 arg4, u64 arg5, u64 arg6, u64 arg7,
910 u64 arg8);
911 void bpf_trace_run9(struct bpf_raw_tp_link *link, u64 arg1, u64 arg2,
912 u64 arg3, u64 arg4, u64 arg5, u64 arg6, u64 arg7,
913 u64 arg8, u64 arg9);
914 void bpf_trace_run10(struct bpf_raw_tp_link *link, u64 arg1, u64 arg2,
915 u64 arg3, u64 arg4, u64 arg5, u64 arg6, u64 arg7,
916 u64 arg8, u64 arg9, u64 arg10);
917 void bpf_trace_run11(struct bpf_raw_tp_link *link, u64 arg1, u64 arg2,
918 u64 arg3, u64 arg4, u64 arg5, u64 arg6, u64 arg7,
919 u64 arg8, u64 arg9, u64 arg10, u64 arg11);
920 void bpf_trace_run12(struct bpf_raw_tp_link *link, u64 arg1, u64 arg2,
921 u64 arg3, u64 arg4, u64 arg5, u64 arg6, u64 arg7,
922 u64 arg8, u64 arg9, u64 arg10, u64 arg11, u64 arg12);
923 void perf_trace_run_bpf_submit(void *raw_data, int size, int rctx,
924 struct trace_event_call *call, u64 count,
925 struct pt_regs *regs, struct hlist_head *head,
926 struct task_struct *task);
927
928 static inline void
perf_trace_buf_submit(void * raw_data,int size,int rctx,u16 type,u64 count,struct pt_regs * regs,void * head,struct task_struct * task)929 perf_trace_buf_submit(void *raw_data, int size, int rctx, u16 type,
930 u64 count, struct pt_regs *regs, void *head,
931 struct task_struct *task)
932 {
933 perf_tp_event(type, count, raw_data, size, regs, head, rctx, task);
934 }
935
936 #endif
937
938 #define TRACE_EVENT_STR_MAX 512
939
940 /*
941 * gcc warns that you can not use a va_list in an inlined
942 * function. But lets me make it into a macro :-/
943 */
944 #define __trace_event_vstr_len(fmt, va) \
945 ({ \
946 va_list __ap; \
947 int __ret; \
948 \
949 va_copy(__ap, *(va)); \
950 __ret = vsnprintf(NULL, 0, fmt, __ap) + 1; \
951 va_end(__ap); \
952 \
953 min(__ret, TRACE_EVENT_STR_MAX); \
954 })
955
956 #endif /* _LINUX_TRACE_EVENT_H */
957
958 /*
959 * Note: we keep the TRACE_CUSTOM_EVENT outside the include file ifdef protection.
960 * This is due to the way trace custom events work. If a file includes two
961 * trace event headers under one "CREATE_CUSTOM_TRACE_EVENTS" the first include
962 * will override the TRACE_CUSTOM_EVENT and break the second include.
963 */
964
965 #ifndef TRACE_CUSTOM_EVENT
966
967 #define DECLARE_CUSTOM_EVENT_CLASS(name, proto, args, tstruct, assign, print)
968 #define DEFINE_CUSTOM_EVENT(template, name, proto, args)
969 #define TRACE_CUSTOM_EVENT(name, proto, args, struct, assign, print)
970
971 #endif /* ifdef TRACE_CUSTOM_EVENT (see note above) */
972