xref: /openbmc/linux/kernel/trace/trace.h (revision fa7f32422ea1ac276b45b96a540ed5981caaa61f)
1 
2 #ifndef _LINUX_KERNEL_TRACE_H
3 #define _LINUX_KERNEL_TRACE_H
4 
5 #include <linux/fs.h>
6 #include <linux/atomic.h>
7 #include <linux/sched.h>
8 #include <linux/clocksource.h>
9 #include <linux/ring_buffer.h>
10 #include <linux/mmiotrace.h>
11 #include <linux/tracepoint.h>
12 #include <linux/ftrace.h>
13 #include <linux/hw_breakpoint.h>
14 #include <linux/trace_seq.h>
15 #include <linux/trace_events.h>
16 #include <linux/compiler.h>
17 #include <linux/trace_seq.h>
18 
19 #ifdef CONFIG_FTRACE_SYSCALLS
20 #include <asm/unistd.h>		/* For NR_SYSCALLS	     */
21 #include <asm/syscall.h>	/* some archs define it here */
22 #endif
23 
24 enum trace_type {
25 	__TRACE_FIRST_TYPE = 0,
26 
27 	TRACE_FN,
28 	TRACE_CTX,
29 	TRACE_WAKE,
30 	TRACE_STACK,
31 	TRACE_PRINT,
32 	TRACE_BPRINT,
33 	TRACE_MMIO_RW,
34 	TRACE_MMIO_MAP,
35 	TRACE_BRANCH,
36 	TRACE_GRAPH_RET,
37 	TRACE_GRAPH_ENT,
38 	TRACE_USER_STACK,
39 	TRACE_BLK,
40 	TRACE_BPUTS,
41 	TRACE_HWLAT,
42 
43 	__TRACE_LAST_TYPE,
44 };
45 
46 
47 #undef __field
48 #define __field(type, item)		type	item;
49 
50 #undef __field_struct
51 #define __field_struct(type, item)	__field(type, item)
52 
53 #undef __field_desc
54 #define __field_desc(type, container, item)
55 
56 #undef __array
57 #define __array(type, item, size)	type	item[size];
58 
59 #undef __array_desc
60 #define __array_desc(type, container, item, size)
61 
62 #undef __dynamic_array
63 #define __dynamic_array(type, item)	type	item[];
64 
65 #undef F_STRUCT
66 #define F_STRUCT(args...)		args
67 
68 #undef FTRACE_ENTRY
69 #define FTRACE_ENTRY(name, struct_name, id, tstruct, print, filter)	\
70 	struct struct_name {						\
71 		struct trace_entry	ent;				\
72 		tstruct							\
73 	}
74 
75 #undef FTRACE_ENTRY_DUP
76 #define FTRACE_ENTRY_DUP(name, name_struct, id, tstruct, printk, filter)
77 
78 #undef FTRACE_ENTRY_REG
79 #define FTRACE_ENTRY_REG(name, struct_name, id, tstruct, print,	\
80 			 filter, regfn) \
81 	FTRACE_ENTRY(name, struct_name, id, PARAMS(tstruct), PARAMS(print), \
82 		     filter)
83 
84 #undef FTRACE_ENTRY_PACKED
85 #define FTRACE_ENTRY_PACKED(name, struct_name, id, tstruct, print,	\
86 			    filter)					\
87 	FTRACE_ENTRY(name, struct_name, id, PARAMS(tstruct), PARAMS(print), \
88 		     filter) __packed
89 
90 #include "trace_entries.h"
91 
92 /*
93  * syscalls are special, and need special handling, this is why
94  * they are not included in trace_entries.h
95  */
96 struct syscall_trace_enter {
97 	struct trace_entry	ent;
98 	int			nr;
99 	unsigned long		args[];
100 };
101 
102 struct syscall_trace_exit {
103 	struct trace_entry	ent;
104 	int			nr;
105 	long			ret;
106 };
107 
108 struct kprobe_trace_entry_head {
109 	struct trace_entry	ent;
110 	unsigned long		ip;
111 };
112 
113 struct kretprobe_trace_entry_head {
114 	struct trace_entry	ent;
115 	unsigned long		func;
116 	unsigned long		ret_ip;
117 };
118 
119 /*
120  * trace_flag_type is an enumeration that holds different
121  * states when a trace occurs. These are:
122  *  IRQS_OFF		- interrupts were disabled
123  *  IRQS_NOSUPPORT	- arch does not support irqs_disabled_flags
124  *  NEED_RESCHED	- reschedule is requested
125  *  HARDIRQ		- inside an interrupt handler
126  *  SOFTIRQ		- inside a softirq handler
127  */
128 enum trace_flag_type {
129 	TRACE_FLAG_IRQS_OFF		= 0x01,
130 	TRACE_FLAG_IRQS_NOSUPPORT	= 0x02,
131 	TRACE_FLAG_NEED_RESCHED		= 0x04,
132 	TRACE_FLAG_HARDIRQ		= 0x08,
133 	TRACE_FLAG_SOFTIRQ		= 0x10,
134 	TRACE_FLAG_PREEMPT_RESCHED	= 0x20,
135 	TRACE_FLAG_NMI			= 0x40,
136 };
137 
138 #define TRACE_BUF_SIZE		1024
139 
140 struct trace_array;
141 
142 /*
143  * The CPU trace array - it consists of thousands of trace entries
144  * plus some other descriptor data: (for example which task started
145  * the trace, etc.)
146  */
147 struct trace_array_cpu {
148 	atomic_t		disabled;
149 	void			*buffer_page;	/* ring buffer spare */
150 
151 	unsigned long		entries;
152 	unsigned long		saved_latency;
153 	unsigned long		critical_start;
154 	unsigned long		critical_end;
155 	unsigned long		critical_sequence;
156 	unsigned long		nice;
157 	unsigned long		policy;
158 	unsigned long		rt_priority;
159 	unsigned long		skipped_entries;
160 	cycle_t			preempt_timestamp;
161 	pid_t			pid;
162 	kuid_t			uid;
163 	char			comm[TASK_COMM_LEN];
164 
165 	bool			ignore_pid;
166 #ifdef CONFIG_FUNCTION_TRACER
167 	bool			ftrace_ignore_pid;
168 #endif
169 };
170 
171 struct tracer;
172 struct trace_option_dentry;
173 
174 struct trace_buffer {
175 	struct trace_array		*tr;
176 	struct ring_buffer		*buffer;
177 	struct trace_array_cpu __percpu	*data;
178 	cycle_t				time_start;
179 	int				cpu;
180 };
181 
182 #define TRACE_FLAGS_MAX_SIZE		32
183 
184 struct trace_options {
185 	struct tracer			*tracer;
186 	struct trace_option_dentry	*topts;
187 };
188 
189 struct trace_pid_list {
190 	int				pid_max;
191 	unsigned long			*pids;
192 };
193 
194 /*
195  * The trace array - an array of per-CPU trace arrays. This is the
196  * highest level data structure that individual tracers deal with.
197  * They have on/off state as well:
198  */
199 struct trace_array {
200 	struct list_head	list;
201 	char			*name;
202 	struct trace_buffer	trace_buffer;
203 #ifdef CONFIG_TRACER_MAX_TRACE
204 	/*
205 	 * The max_buffer is used to snapshot the trace when a maximum
206 	 * latency is reached, or when the user initiates a snapshot.
207 	 * Some tracers will use this to store a maximum trace while
208 	 * it continues examining live traces.
209 	 *
210 	 * The buffers for the max_buffer are set up the same as the trace_buffer
211 	 * When a snapshot is taken, the buffer of the max_buffer is swapped
212 	 * with the buffer of the trace_buffer and the buffers are reset for
213 	 * the trace_buffer so the tracing can continue.
214 	 */
215 	struct trace_buffer	max_buffer;
216 	bool			allocated_snapshot;
217 #endif
218 #if defined(CONFIG_TRACER_MAX_TRACE) || defined(CONFIG_HWLAT_TRACER)
219 	unsigned long		max_latency;
220 #endif
221 	struct trace_pid_list	__rcu *filtered_pids;
222 	/*
223 	 * max_lock is used to protect the swapping of buffers
224 	 * when taking a max snapshot. The buffers themselves are
225 	 * protected by per_cpu spinlocks. But the action of the swap
226 	 * needs its own lock.
227 	 *
228 	 * This is defined as a arch_spinlock_t in order to help
229 	 * with performance when lockdep debugging is enabled.
230 	 *
231 	 * It is also used in other places outside the update_max_tr
232 	 * so it needs to be defined outside of the
233 	 * CONFIG_TRACER_MAX_TRACE.
234 	 */
235 	arch_spinlock_t		max_lock;
236 	int			buffer_disabled;
237 #ifdef CONFIG_FTRACE_SYSCALLS
238 	int			sys_refcount_enter;
239 	int			sys_refcount_exit;
240 	struct trace_event_file __rcu *enter_syscall_files[NR_syscalls];
241 	struct trace_event_file __rcu *exit_syscall_files[NR_syscalls];
242 #endif
243 	int			stop_count;
244 	int			clock_id;
245 	int			nr_topts;
246 	struct tracer		*current_trace;
247 	unsigned int		trace_flags;
248 	unsigned char		trace_flags_index[TRACE_FLAGS_MAX_SIZE];
249 	unsigned int		flags;
250 	raw_spinlock_t		start_lock;
251 	struct dentry		*dir;
252 	struct dentry		*options;
253 	struct dentry		*percpu_dir;
254 	struct dentry		*event_dir;
255 	struct trace_options	*topts;
256 	struct list_head	systems;
257 	struct list_head	events;
258 	cpumask_var_t		tracing_cpumask; /* only trace on set CPUs */
259 	int			ref;
260 #ifdef CONFIG_FUNCTION_TRACER
261 	struct ftrace_ops	*ops;
262 	struct trace_pid_list	__rcu *function_pids;
263 	/* function tracing enabled */
264 	int			function_enabled;
265 #endif
266 };
267 
268 enum {
269 	TRACE_ARRAY_FL_GLOBAL	= (1 << 0)
270 };
271 
272 extern struct list_head ftrace_trace_arrays;
273 
274 extern struct mutex trace_types_lock;
275 
276 extern int trace_array_get(struct trace_array *tr);
277 extern void trace_array_put(struct trace_array *tr);
278 
279 /*
280  * The global tracer (top) should be the first trace array added,
281  * but we check the flag anyway.
282  */
283 static inline struct trace_array *top_trace_array(void)
284 {
285 	struct trace_array *tr;
286 
287 	if (list_empty(&ftrace_trace_arrays))
288 		return NULL;
289 
290 	tr = list_entry(ftrace_trace_arrays.prev,
291 			typeof(*tr), list);
292 	WARN_ON(!(tr->flags & TRACE_ARRAY_FL_GLOBAL));
293 	return tr;
294 }
295 
296 #define FTRACE_CMP_TYPE(var, type) \
297 	__builtin_types_compatible_p(typeof(var), type *)
298 
299 #undef IF_ASSIGN
300 #define IF_ASSIGN(var, entry, etype, id)		\
301 	if (FTRACE_CMP_TYPE(var, etype)) {		\
302 		var = (typeof(var))(entry);		\
303 		WARN_ON(id && (entry)->type != id);	\
304 		break;					\
305 	}
306 
307 /* Will cause compile errors if type is not found. */
308 extern void __ftrace_bad_type(void);
309 
310 /*
311  * The trace_assign_type is a verifier that the entry type is
312  * the same as the type being assigned. To add new types simply
313  * add a line with the following format:
314  *
315  * IF_ASSIGN(var, ent, type, id);
316  *
317  *  Where "type" is the trace type that includes the trace_entry
318  *  as the "ent" item. And "id" is the trace identifier that is
319  *  used in the trace_type enum.
320  *
321  *  If the type can have more than one id, then use zero.
322  */
323 #define trace_assign_type(var, ent)					\
324 	do {								\
325 		IF_ASSIGN(var, ent, struct ftrace_entry, TRACE_FN);	\
326 		IF_ASSIGN(var, ent, struct ctx_switch_entry, 0);	\
327 		IF_ASSIGN(var, ent, struct stack_entry, TRACE_STACK);	\
328 		IF_ASSIGN(var, ent, struct userstack_entry, TRACE_USER_STACK);\
329 		IF_ASSIGN(var, ent, struct print_entry, TRACE_PRINT);	\
330 		IF_ASSIGN(var, ent, struct bprint_entry, TRACE_BPRINT);	\
331 		IF_ASSIGN(var, ent, struct bputs_entry, TRACE_BPUTS);	\
332 		IF_ASSIGN(var, ent, struct hwlat_entry, TRACE_HWLAT);	\
333 		IF_ASSIGN(var, ent, struct trace_mmiotrace_rw,		\
334 			  TRACE_MMIO_RW);				\
335 		IF_ASSIGN(var, ent, struct trace_mmiotrace_map,		\
336 			  TRACE_MMIO_MAP);				\
337 		IF_ASSIGN(var, ent, struct trace_branch, TRACE_BRANCH); \
338 		IF_ASSIGN(var, ent, struct ftrace_graph_ent_entry,	\
339 			  TRACE_GRAPH_ENT);		\
340 		IF_ASSIGN(var, ent, struct ftrace_graph_ret_entry,	\
341 			  TRACE_GRAPH_RET);		\
342 		__ftrace_bad_type();					\
343 	} while (0)
344 
345 /*
346  * An option specific to a tracer. This is a boolean value.
347  * The bit is the bit index that sets its value on the
348  * flags value in struct tracer_flags.
349  */
350 struct tracer_opt {
351 	const char	*name; /* Will appear on the trace_options file */
352 	u32		bit; /* Mask assigned in val field in tracer_flags */
353 };
354 
355 /*
356  * The set of specific options for a tracer. Your tracer
357  * have to set the initial value of the flags val.
358  */
359 struct tracer_flags {
360 	u32			val;
361 	struct tracer_opt	*opts;
362 	struct tracer		*trace;
363 };
364 
365 /* Makes more easy to define a tracer opt */
366 #define TRACER_OPT(s, b)	.name = #s, .bit = b
367 
368 
369 struct trace_option_dentry {
370 	struct tracer_opt		*opt;
371 	struct tracer_flags		*flags;
372 	struct trace_array		*tr;
373 	struct dentry			*entry;
374 };
375 
376 /**
377  * struct tracer - a specific tracer and its callbacks to interact with tracefs
378  * @name: the name chosen to select it on the available_tracers file
379  * @init: called when one switches to this tracer (echo name > current_tracer)
380  * @reset: called when one switches to another tracer
381  * @start: called when tracing is unpaused (echo 1 > tracing_on)
382  * @stop: called when tracing is paused (echo 0 > tracing_on)
383  * @update_thresh: called when tracing_thresh is updated
384  * @open: called when the trace file is opened
385  * @pipe_open: called when the trace_pipe file is opened
386  * @close: called when the trace file is released
387  * @pipe_close: called when the trace_pipe file is released
388  * @read: override the default read callback on trace_pipe
389  * @splice_read: override the default splice_read callback on trace_pipe
390  * @selftest: selftest to run on boot (see trace_selftest.c)
391  * @print_headers: override the first lines that describe your columns
392  * @print_line: callback that prints a trace
393  * @set_flag: signals one of your private flags changed (trace_options file)
394  * @flags: your private flags
395  */
396 struct tracer {
397 	const char		*name;
398 	int			(*init)(struct trace_array *tr);
399 	void			(*reset)(struct trace_array *tr);
400 	void			(*start)(struct trace_array *tr);
401 	void			(*stop)(struct trace_array *tr);
402 	int			(*update_thresh)(struct trace_array *tr);
403 	void			(*open)(struct trace_iterator *iter);
404 	void			(*pipe_open)(struct trace_iterator *iter);
405 	void			(*close)(struct trace_iterator *iter);
406 	void			(*pipe_close)(struct trace_iterator *iter);
407 	ssize_t			(*read)(struct trace_iterator *iter,
408 					struct file *filp, char __user *ubuf,
409 					size_t cnt, loff_t *ppos);
410 	ssize_t			(*splice_read)(struct trace_iterator *iter,
411 					       struct file *filp,
412 					       loff_t *ppos,
413 					       struct pipe_inode_info *pipe,
414 					       size_t len,
415 					       unsigned int flags);
416 #ifdef CONFIG_FTRACE_STARTUP_TEST
417 	int			(*selftest)(struct tracer *trace,
418 					    struct trace_array *tr);
419 #endif
420 	void			(*print_header)(struct seq_file *m);
421 	enum print_line_t	(*print_line)(struct trace_iterator *iter);
422 	/* If you handled the flag setting, return 0 */
423 	int			(*set_flag)(struct trace_array *tr,
424 					    u32 old_flags, u32 bit, int set);
425 	/* Return 0 if OK with change, else return non-zero */
426 	int			(*flag_changed)(struct trace_array *tr,
427 						u32 mask, int set);
428 	struct tracer		*next;
429 	struct tracer_flags	*flags;
430 	int			enabled;
431 	int			ref;
432 	bool			print_max;
433 	bool			allow_instances;
434 #ifdef CONFIG_TRACER_MAX_TRACE
435 	bool			use_max_tr;
436 #endif
437 };
438 
439 
440 /* Only current can touch trace_recursion */
441 
442 /*
443  * For function tracing recursion:
444  *  The order of these bits are important.
445  *
446  *  When function tracing occurs, the following steps are made:
447  *   If arch does not support a ftrace feature:
448  *    call internal function (uses INTERNAL bits) which calls...
449  *   If callback is registered to the "global" list, the list
450  *    function is called and recursion checks the GLOBAL bits.
451  *    then this function calls...
452  *   The function callback, which can use the FTRACE bits to
453  *    check for recursion.
454  *
455  * Now if the arch does not suppport a feature, and it calls
456  * the global list function which calls the ftrace callback
457  * all three of these steps will do a recursion protection.
458  * There's no reason to do one if the previous caller already
459  * did. The recursion that we are protecting against will
460  * go through the same steps again.
461  *
462  * To prevent the multiple recursion checks, if a recursion
463  * bit is set that is higher than the MAX bit of the current
464  * check, then we know that the check was made by the previous
465  * caller, and we can skip the current check.
466  */
467 enum {
468 	TRACE_BUFFER_BIT,
469 	TRACE_BUFFER_NMI_BIT,
470 	TRACE_BUFFER_IRQ_BIT,
471 	TRACE_BUFFER_SIRQ_BIT,
472 
473 	/* Start of function recursion bits */
474 	TRACE_FTRACE_BIT,
475 	TRACE_FTRACE_NMI_BIT,
476 	TRACE_FTRACE_IRQ_BIT,
477 	TRACE_FTRACE_SIRQ_BIT,
478 
479 	/* INTERNAL_BITs must be greater than FTRACE_BITs */
480 	TRACE_INTERNAL_BIT,
481 	TRACE_INTERNAL_NMI_BIT,
482 	TRACE_INTERNAL_IRQ_BIT,
483 	TRACE_INTERNAL_SIRQ_BIT,
484 
485 	TRACE_BRANCH_BIT,
486 /*
487  * Abuse of the trace_recursion.
488  * As we need a way to maintain state if we are tracing the function
489  * graph in irq because we want to trace a particular function that
490  * was called in irq context but we have irq tracing off. Since this
491  * can only be modified by current, we can reuse trace_recursion.
492  */
493 	TRACE_IRQ_BIT,
494 };
495 
496 #define trace_recursion_set(bit)	do { (current)->trace_recursion |= (1<<(bit)); } while (0)
497 #define trace_recursion_clear(bit)	do { (current)->trace_recursion &= ~(1<<(bit)); } while (0)
498 #define trace_recursion_test(bit)	((current)->trace_recursion & (1<<(bit)))
499 
500 #define TRACE_CONTEXT_BITS	4
501 
502 #define TRACE_FTRACE_START	TRACE_FTRACE_BIT
503 #define TRACE_FTRACE_MAX	((1 << (TRACE_FTRACE_START + TRACE_CONTEXT_BITS)) - 1)
504 
505 #define TRACE_LIST_START	TRACE_INTERNAL_BIT
506 #define TRACE_LIST_MAX		((1 << (TRACE_LIST_START + TRACE_CONTEXT_BITS)) - 1)
507 
508 #define TRACE_CONTEXT_MASK	TRACE_LIST_MAX
509 
510 static __always_inline int trace_get_context_bit(void)
511 {
512 	int bit;
513 
514 	if (in_interrupt()) {
515 		if (in_nmi())
516 			bit = 0;
517 
518 		else if (in_irq())
519 			bit = 1;
520 		else
521 			bit = 2;
522 	} else
523 		bit = 3;
524 
525 	return bit;
526 }
527 
528 static __always_inline int trace_test_and_set_recursion(int start, int max)
529 {
530 	unsigned int val = current->trace_recursion;
531 	int bit;
532 
533 	/* A previous recursion check was made */
534 	if ((val & TRACE_CONTEXT_MASK) > max)
535 		return 0;
536 
537 	bit = trace_get_context_bit() + start;
538 	if (unlikely(val & (1 << bit)))
539 		return -1;
540 
541 	val |= 1 << bit;
542 	current->trace_recursion = val;
543 	barrier();
544 
545 	return bit;
546 }
547 
548 static __always_inline void trace_clear_recursion(int bit)
549 {
550 	unsigned int val = current->trace_recursion;
551 
552 	if (!bit)
553 		return;
554 
555 	bit = 1 << bit;
556 	val &= ~bit;
557 
558 	barrier();
559 	current->trace_recursion = val;
560 }
561 
562 static inline struct ring_buffer_iter *
563 trace_buffer_iter(struct trace_iterator *iter, int cpu)
564 {
565 	if (iter->buffer_iter && iter->buffer_iter[cpu])
566 		return iter->buffer_iter[cpu];
567 	return NULL;
568 }
569 
570 int tracer_init(struct tracer *t, struct trace_array *tr);
571 int tracing_is_enabled(void);
572 void tracing_reset(struct trace_buffer *buf, int cpu);
573 void tracing_reset_online_cpus(struct trace_buffer *buf);
574 void tracing_reset_current(int cpu);
575 void tracing_reset_all_online_cpus(void);
576 int tracing_open_generic(struct inode *inode, struct file *filp);
577 bool tracing_is_disabled(void);
578 int tracer_tracing_is_on(struct trace_array *tr);
579 struct dentry *trace_create_file(const char *name,
580 				 umode_t mode,
581 				 struct dentry *parent,
582 				 void *data,
583 				 const struct file_operations *fops);
584 
585 struct dentry *tracing_init_dentry(void);
586 
587 struct ring_buffer_event;
588 
589 struct ring_buffer_event *
590 trace_buffer_lock_reserve(struct ring_buffer *buffer,
591 			  int type,
592 			  unsigned long len,
593 			  unsigned long flags,
594 			  int pc);
595 
596 struct trace_entry *tracing_get_trace_entry(struct trace_array *tr,
597 						struct trace_array_cpu *data);
598 
599 struct trace_entry *trace_find_next_entry(struct trace_iterator *iter,
600 					  int *ent_cpu, u64 *ent_ts);
601 
602 void __buffer_unlock_commit(struct ring_buffer *buffer,
603 			    struct ring_buffer_event *event);
604 
605 int trace_empty(struct trace_iterator *iter);
606 
607 void *trace_find_next_entry_inc(struct trace_iterator *iter);
608 
609 void trace_init_global_iter(struct trace_iterator *iter);
610 
611 void tracing_iter_reset(struct trace_iterator *iter, int cpu);
612 
613 void trace_function(struct trace_array *tr,
614 		    unsigned long ip,
615 		    unsigned long parent_ip,
616 		    unsigned long flags, int pc);
617 void trace_graph_function(struct trace_array *tr,
618 		    unsigned long ip,
619 		    unsigned long parent_ip,
620 		    unsigned long flags, int pc);
621 void trace_latency_header(struct seq_file *m);
622 void trace_default_header(struct seq_file *m);
623 void print_trace_header(struct seq_file *m, struct trace_iterator *iter);
624 int trace_empty(struct trace_iterator *iter);
625 
626 void trace_graph_return(struct ftrace_graph_ret *trace);
627 int trace_graph_entry(struct ftrace_graph_ent *trace);
628 void set_graph_array(struct trace_array *tr);
629 
630 void tracing_start_cmdline_record(void);
631 void tracing_stop_cmdline_record(void);
632 int register_tracer(struct tracer *type);
633 int is_tracing_stopped(void);
634 
635 loff_t tracing_lseek(struct file *file, loff_t offset, int whence);
636 
637 extern cpumask_var_t __read_mostly tracing_buffer_mask;
638 
639 #define for_each_tracing_cpu(cpu)	\
640 	for_each_cpu(cpu, tracing_buffer_mask)
641 
642 extern unsigned long nsecs_to_usecs(unsigned long nsecs);
643 
644 extern unsigned long tracing_thresh;
645 
646 /* PID filtering */
647 
648 extern int pid_max;
649 
650 bool trace_find_filtered_pid(struct trace_pid_list *filtered_pids,
651 			     pid_t search_pid);
652 bool trace_ignore_this_task(struct trace_pid_list *filtered_pids,
653 			    struct task_struct *task);
654 void trace_filter_add_remove_task(struct trace_pid_list *pid_list,
655 				  struct task_struct *self,
656 				  struct task_struct *task);
657 void *trace_pid_next(struct trace_pid_list *pid_list, void *v, loff_t *pos);
658 void *trace_pid_start(struct trace_pid_list *pid_list, loff_t *pos);
659 int trace_pid_show(struct seq_file *m, void *v);
660 void trace_free_pid_list(struct trace_pid_list *pid_list);
661 int trace_pid_write(struct trace_pid_list *filtered_pids,
662 		    struct trace_pid_list **new_pid_list,
663 		    const char __user *ubuf, size_t cnt);
664 
665 #ifdef CONFIG_TRACER_MAX_TRACE
666 void update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu);
667 void update_max_tr_single(struct trace_array *tr,
668 			  struct task_struct *tsk, int cpu);
669 #endif /* CONFIG_TRACER_MAX_TRACE */
670 
671 #ifdef CONFIG_STACKTRACE
672 void ftrace_trace_userstack(struct ring_buffer *buffer, unsigned long flags,
673 			    int pc);
674 
675 void __trace_stack(struct trace_array *tr, unsigned long flags, int skip,
676 		   int pc);
677 #else
678 static inline void ftrace_trace_userstack(struct ring_buffer *buffer,
679 					  unsigned long flags, int pc)
680 {
681 }
682 
683 static inline void __trace_stack(struct trace_array *tr, unsigned long flags,
684 				 int skip, int pc)
685 {
686 }
687 #endif /* CONFIG_STACKTRACE */
688 
689 extern cycle_t ftrace_now(int cpu);
690 
691 extern void trace_find_cmdline(int pid, char comm[]);
692 extern void trace_event_follow_fork(struct trace_array *tr, bool enable);
693 
694 #ifdef CONFIG_DYNAMIC_FTRACE
695 extern unsigned long ftrace_update_tot_cnt;
696 #endif
697 #define DYN_FTRACE_TEST_NAME trace_selftest_dynamic_test_func
698 extern int DYN_FTRACE_TEST_NAME(void);
699 #define DYN_FTRACE_TEST_NAME2 trace_selftest_dynamic_test_func2
700 extern int DYN_FTRACE_TEST_NAME2(void);
701 
702 extern bool ring_buffer_expanded;
703 extern bool tracing_selftest_disabled;
704 
705 #ifdef CONFIG_FTRACE_STARTUP_TEST
706 extern int trace_selftest_startup_function(struct tracer *trace,
707 					   struct trace_array *tr);
708 extern int trace_selftest_startup_function_graph(struct tracer *trace,
709 						 struct trace_array *tr);
710 extern int trace_selftest_startup_irqsoff(struct tracer *trace,
711 					  struct trace_array *tr);
712 extern int trace_selftest_startup_preemptoff(struct tracer *trace,
713 					     struct trace_array *tr);
714 extern int trace_selftest_startup_preemptirqsoff(struct tracer *trace,
715 						 struct trace_array *tr);
716 extern int trace_selftest_startup_wakeup(struct tracer *trace,
717 					 struct trace_array *tr);
718 extern int trace_selftest_startup_nop(struct tracer *trace,
719 					 struct trace_array *tr);
720 extern int trace_selftest_startup_sched_switch(struct tracer *trace,
721 					       struct trace_array *tr);
722 extern int trace_selftest_startup_branch(struct tracer *trace,
723 					 struct trace_array *tr);
724 /*
725  * Tracer data references selftest functions that only occur
726  * on boot up. These can be __init functions. Thus, when selftests
727  * are enabled, then the tracers need to reference __init functions.
728  */
729 #define __tracer_data		__refdata
730 #else
731 /* Tracers are seldom changed. Optimize when selftests are disabled. */
732 #define __tracer_data		__read_mostly
733 #endif /* CONFIG_FTRACE_STARTUP_TEST */
734 
735 extern void *head_page(struct trace_array_cpu *data);
736 extern unsigned long long ns2usecs(cycle_t nsec);
737 extern int
738 trace_vbprintk(unsigned long ip, const char *fmt, va_list args);
739 extern int
740 trace_vprintk(unsigned long ip, const char *fmt, va_list args);
741 extern int
742 trace_array_vprintk(struct trace_array *tr,
743 		    unsigned long ip, const char *fmt, va_list args);
744 int trace_array_printk(struct trace_array *tr,
745 		       unsigned long ip, const char *fmt, ...);
746 int trace_array_printk_buf(struct ring_buffer *buffer,
747 			   unsigned long ip, const char *fmt, ...);
748 void trace_printk_seq(struct trace_seq *s);
749 enum print_line_t print_trace_line(struct trace_iterator *iter);
750 
751 extern char trace_find_mark(unsigned long long duration);
752 
753 /* Standard output formatting function used for function return traces */
754 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
755 
756 /* Flag options */
757 #define TRACE_GRAPH_PRINT_OVERRUN       0x1
758 #define TRACE_GRAPH_PRINT_CPU           0x2
759 #define TRACE_GRAPH_PRINT_OVERHEAD      0x4
760 #define TRACE_GRAPH_PRINT_PROC          0x8
761 #define TRACE_GRAPH_PRINT_DURATION      0x10
762 #define TRACE_GRAPH_PRINT_ABS_TIME      0x20
763 #define TRACE_GRAPH_PRINT_IRQS          0x40
764 #define TRACE_GRAPH_PRINT_TAIL          0x80
765 #define TRACE_GRAPH_SLEEP_TIME		0x100
766 #define TRACE_GRAPH_GRAPH_TIME		0x200
767 #define TRACE_GRAPH_PRINT_FILL_SHIFT	28
768 #define TRACE_GRAPH_PRINT_FILL_MASK	(0x3 << TRACE_GRAPH_PRINT_FILL_SHIFT)
769 
770 extern void ftrace_graph_sleep_time_control(bool enable);
771 extern void ftrace_graph_graph_time_control(bool enable);
772 
773 extern enum print_line_t
774 print_graph_function_flags(struct trace_iterator *iter, u32 flags);
775 extern void print_graph_headers_flags(struct seq_file *s, u32 flags);
776 extern void
777 trace_print_graph_duration(unsigned long long duration, struct trace_seq *s);
778 extern void graph_trace_open(struct trace_iterator *iter);
779 extern void graph_trace_close(struct trace_iterator *iter);
780 extern int __trace_graph_entry(struct trace_array *tr,
781 			       struct ftrace_graph_ent *trace,
782 			       unsigned long flags, int pc);
783 extern void __trace_graph_return(struct trace_array *tr,
784 				 struct ftrace_graph_ret *trace,
785 				 unsigned long flags, int pc);
786 
787 
788 #ifdef CONFIG_DYNAMIC_FTRACE
789 /* TODO: make this variable */
790 #define FTRACE_GRAPH_MAX_FUNCS		32
791 extern int ftrace_graph_count;
792 extern unsigned long ftrace_graph_funcs[FTRACE_GRAPH_MAX_FUNCS];
793 extern int ftrace_graph_notrace_count;
794 extern unsigned long ftrace_graph_notrace_funcs[FTRACE_GRAPH_MAX_FUNCS];
795 
796 static inline int ftrace_graph_addr(unsigned long addr)
797 {
798 	int i;
799 
800 	if (!ftrace_graph_count)
801 		return 1;
802 
803 	for (i = 0; i < ftrace_graph_count; i++) {
804 		if (addr == ftrace_graph_funcs[i]) {
805 			/*
806 			 * If no irqs are to be traced, but a set_graph_function
807 			 * is set, and called by an interrupt handler, we still
808 			 * want to trace it.
809 			 */
810 			if (in_irq())
811 				trace_recursion_set(TRACE_IRQ_BIT);
812 			else
813 				trace_recursion_clear(TRACE_IRQ_BIT);
814 			return 1;
815 		}
816 	}
817 
818 	return 0;
819 }
820 
821 static inline int ftrace_graph_notrace_addr(unsigned long addr)
822 {
823 	int i;
824 
825 	if (!ftrace_graph_notrace_count)
826 		return 0;
827 
828 	for (i = 0; i < ftrace_graph_notrace_count; i++) {
829 		if (addr == ftrace_graph_notrace_funcs[i])
830 			return 1;
831 	}
832 
833 	return 0;
834 }
835 #else
836 static inline int ftrace_graph_addr(unsigned long addr)
837 {
838 	return 1;
839 }
840 
841 static inline int ftrace_graph_notrace_addr(unsigned long addr)
842 {
843 	return 0;
844 }
845 #endif /* CONFIG_DYNAMIC_FTRACE */
846 #else /* CONFIG_FUNCTION_GRAPH_TRACER */
847 static inline enum print_line_t
848 print_graph_function_flags(struct trace_iterator *iter, u32 flags)
849 {
850 	return TRACE_TYPE_UNHANDLED;
851 }
852 #endif /* CONFIG_FUNCTION_GRAPH_TRACER */
853 
854 extern struct list_head ftrace_pids;
855 
856 #ifdef CONFIG_FUNCTION_TRACER
857 extern bool ftrace_filter_param __initdata;
858 static inline int ftrace_trace_task(struct trace_array *tr)
859 {
860 	return !this_cpu_read(tr->trace_buffer.data->ftrace_ignore_pid);
861 }
862 extern int ftrace_is_dead(void);
863 int ftrace_create_function_files(struct trace_array *tr,
864 				 struct dentry *parent);
865 void ftrace_destroy_function_files(struct trace_array *tr);
866 void ftrace_init_global_array_ops(struct trace_array *tr);
867 void ftrace_init_array_ops(struct trace_array *tr, ftrace_func_t func);
868 void ftrace_reset_array_ops(struct trace_array *tr);
869 int using_ftrace_ops_list_func(void);
870 void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d_tracer);
871 void ftrace_init_tracefs_toplevel(struct trace_array *tr,
872 				  struct dentry *d_tracer);
873 #else
874 static inline int ftrace_trace_task(struct trace_array *tr)
875 {
876 	return 1;
877 }
878 static inline int ftrace_is_dead(void) { return 0; }
879 static inline int
880 ftrace_create_function_files(struct trace_array *tr,
881 			     struct dentry *parent)
882 {
883 	return 0;
884 }
885 static inline void ftrace_destroy_function_files(struct trace_array *tr) { }
886 static inline __init void
887 ftrace_init_global_array_ops(struct trace_array *tr) { }
888 static inline void ftrace_reset_array_ops(struct trace_array *tr) { }
889 static inline void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d) { }
890 static inline void ftrace_init_tracefs_toplevel(struct trace_array *tr, struct dentry *d) { }
891 /* ftace_func_t type is not defined, use macro instead of static inline */
892 #define ftrace_init_array_ops(tr, func) do { } while (0)
893 #endif /* CONFIG_FUNCTION_TRACER */
894 
895 #if defined(CONFIG_FUNCTION_TRACER) && defined(CONFIG_DYNAMIC_FTRACE)
896 void ftrace_create_filter_files(struct ftrace_ops *ops,
897 				struct dentry *parent);
898 void ftrace_destroy_filter_files(struct ftrace_ops *ops);
899 #else
900 /*
901  * The ops parameter passed in is usually undefined.
902  * This must be a macro.
903  */
904 #define ftrace_create_filter_files(ops, parent) do { } while (0)
905 #define ftrace_destroy_filter_files(ops) do { } while (0)
906 #endif /* CONFIG_FUNCTION_TRACER && CONFIG_DYNAMIC_FTRACE */
907 
908 bool ftrace_event_is_function(struct trace_event_call *call);
909 
910 /*
911  * struct trace_parser - servers for reading the user input separated by spaces
912  * @cont: set if the input is not complete - no final space char was found
913  * @buffer: holds the parsed user input
914  * @idx: user input length
915  * @size: buffer size
916  */
917 struct trace_parser {
918 	bool		cont;
919 	char		*buffer;
920 	unsigned	idx;
921 	unsigned	size;
922 };
923 
924 static inline bool trace_parser_loaded(struct trace_parser *parser)
925 {
926 	return (parser->idx != 0);
927 }
928 
929 static inline bool trace_parser_cont(struct trace_parser *parser)
930 {
931 	return parser->cont;
932 }
933 
934 static inline void trace_parser_clear(struct trace_parser *parser)
935 {
936 	parser->cont = false;
937 	parser->idx = 0;
938 }
939 
940 extern int trace_parser_get_init(struct trace_parser *parser, int size);
941 extern void trace_parser_put(struct trace_parser *parser);
942 extern int trace_get_user(struct trace_parser *parser, const char __user *ubuf,
943 	size_t cnt, loff_t *ppos);
944 
945 /*
946  * Only create function graph options if function graph is configured.
947  */
948 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
949 # define FGRAPH_FLAGS						\
950 		C(DISPLAY_GRAPH,	"display-graph"),
951 #else
952 # define FGRAPH_FLAGS
953 #endif
954 
955 #ifdef CONFIG_BRANCH_TRACER
956 # define BRANCH_FLAGS					\
957 		C(BRANCH,		"branch"),
958 #else
959 # define BRANCH_FLAGS
960 #endif
961 
962 #ifdef CONFIG_FUNCTION_TRACER
963 # define FUNCTION_FLAGS						\
964 		C(FUNCTION,		"function-trace"),
965 # define FUNCTION_DEFAULT_FLAGS		TRACE_ITER_FUNCTION
966 #else
967 # define FUNCTION_FLAGS
968 # define FUNCTION_DEFAULT_FLAGS		0UL
969 #endif
970 
971 #ifdef CONFIG_STACKTRACE
972 # define STACK_FLAGS				\
973 		C(STACKTRACE,		"stacktrace"),
974 #else
975 # define STACK_FLAGS
976 #endif
977 
978 /*
979  * trace_iterator_flags is an enumeration that defines bit
980  * positions into trace_flags that controls the output.
981  *
982  * NOTE: These bits must match the trace_options array in
983  *       trace.c (this macro guarantees it).
984  */
985 #define TRACE_FLAGS						\
986 		C(PRINT_PARENT,		"print-parent"),	\
987 		C(SYM_OFFSET,		"sym-offset"),		\
988 		C(SYM_ADDR,		"sym-addr"),		\
989 		C(VERBOSE,		"verbose"),		\
990 		C(RAW,			"raw"),			\
991 		C(HEX,			"hex"),			\
992 		C(BIN,			"bin"),			\
993 		C(BLOCK,		"block"),		\
994 		C(PRINTK,		"trace_printk"),	\
995 		C(ANNOTATE,		"annotate"),		\
996 		C(USERSTACKTRACE,	"userstacktrace"),	\
997 		C(SYM_USEROBJ,		"sym-userobj"),		\
998 		C(PRINTK_MSGONLY,	"printk-msg-only"),	\
999 		C(CONTEXT_INFO,		"context-info"),   /* Print pid/cpu/time */ \
1000 		C(LATENCY_FMT,		"latency-format"),	\
1001 		C(RECORD_CMD,		"record-cmd"),		\
1002 		C(OVERWRITE,		"overwrite"),		\
1003 		C(STOP_ON_FREE,		"disable_on_free"),	\
1004 		C(IRQ_INFO,		"irq-info"),		\
1005 		C(MARKERS,		"markers"),		\
1006 		C(EVENT_FORK,		"event-fork"),		\
1007 		FUNCTION_FLAGS					\
1008 		FGRAPH_FLAGS					\
1009 		STACK_FLAGS					\
1010 		BRANCH_FLAGS
1011 
1012 /*
1013  * By defining C, we can make TRACE_FLAGS a list of bit names
1014  * that will define the bits for the flag masks.
1015  */
1016 #undef C
1017 #define C(a, b) TRACE_ITER_##a##_BIT
1018 
1019 enum trace_iterator_bits {
1020 	TRACE_FLAGS
1021 	/* Make sure we don't go more than we have bits for */
1022 	TRACE_ITER_LAST_BIT
1023 };
1024 
1025 /*
1026  * By redefining C, we can make TRACE_FLAGS a list of masks that
1027  * use the bits as defined above.
1028  */
1029 #undef C
1030 #define C(a, b) TRACE_ITER_##a = (1 << TRACE_ITER_##a##_BIT)
1031 
1032 enum trace_iterator_flags { TRACE_FLAGS };
1033 
1034 /*
1035  * TRACE_ITER_SYM_MASK masks the options in trace_flags that
1036  * control the output of kernel symbols.
1037  */
1038 #define TRACE_ITER_SYM_MASK \
1039 	(TRACE_ITER_PRINT_PARENT|TRACE_ITER_SYM_OFFSET|TRACE_ITER_SYM_ADDR)
1040 
1041 extern struct tracer nop_trace;
1042 
1043 #ifdef CONFIG_BRANCH_TRACER
1044 extern int enable_branch_tracing(struct trace_array *tr);
1045 extern void disable_branch_tracing(void);
1046 static inline int trace_branch_enable(struct trace_array *tr)
1047 {
1048 	if (tr->trace_flags & TRACE_ITER_BRANCH)
1049 		return enable_branch_tracing(tr);
1050 	return 0;
1051 }
1052 static inline void trace_branch_disable(void)
1053 {
1054 	/* due to races, always disable */
1055 	disable_branch_tracing();
1056 }
1057 #else
1058 static inline int trace_branch_enable(struct trace_array *tr)
1059 {
1060 	return 0;
1061 }
1062 static inline void trace_branch_disable(void)
1063 {
1064 }
1065 #endif /* CONFIG_BRANCH_TRACER */
1066 
1067 /* set ring buffers to default size if not already done so */
1068 int tracing_update_buffers(void);
1069 
1070 struct ftrace_event_field {
1071 	struct list_head	link;
1072 	const char		*name;
1073 	const char		*type;
1074 	int			filter_type;
1075 	int			offset;
1076 	int			size;
1077 	int			is_signed;
1078 };
1079 
1080 struct event_filter {
1081 	int			n_preds;	/* Number assigned */
1082 	int			a_preds;	/* allocated */
1083 	struct filter_pred	*preds;
1084 	struct filter_pred	*root;
1085 	char			*filter_string;
1086 };
1087 
1088 struct event_subsystem {
1089 	struct list_head	list;
1090 	const char		*name;
1091 	struct event_filter	*filter;
1092 	int			ref_count;
1093 };
1094 
1095 struct trace_subsystem_dir {
1096 	struct list_head		list;
1097 	struct event_subsystem		*subsystem;
1098 	struct trace_array		*tr;
1099 	struct dentry			*entry;
1100 	int				ref_count;
1101 	int				nr_events;
1102 };
1103 
1104 extern int call_filter_check_discard(struct trace_event_call *call, void *rec,
1105 				     struct ring_buffer *buffer,
1106 				     struct ring_buffer_event *event);
1107 
1108 void trace_buffer_unlock_commit_regs(struct trace_array *tr,
1109 				     struct ring_buffer *buffer,
1110 				     struct ring_buffer_event *event,
1111 				     unsigned long flags, int pc,
1112 				     struct pt_regs *regs);
1113 
1114 static inline void trace_buffer_unlock_commit(struct trace_array *tr,
1115 					      struct ring_buffer *buffer,
1116 					      struct ring_buffer_event *event,
1117 					      unsigned long flags, int pc)
1118 {
1119 	trace_buffer_unlock_commit_regs(tr, buffer, event, flags, pc, NULL);
1120 }
1121 
1122 DECLARE_PER_CPU(struct ring_buffer_event *, trace_buffered_event);
1123 DECLARE_PER_CPU(int, trace_buffered_event_cnt);
1124 void trace_buffered_event_disable(void);
1125 void trace_buffered_event_enable(void);
1126 
1127 static inline void
1128 __trace_event_discard_commit(struct ring_buffer *buffer,
1129 			     struct ring_buffer_event *event)
1130 {
1131 	if (this_cpu_read(trace_buffered_event) == event) {
1132 		/* Simply release the temp buffer */
1133 		this_cpu_dec(trace_buffered_event_cnt);
1134 		return;
1135 	}
1136 	ring_buffer_discard_commit(buffer, event);
1137 }
1138 
1139 /*
1140  * Helper function for event_trigger_unlock_commit{_regs}().
1141  * If there are event triggers attached to this event that requires
1142  * filtering against its fields, then they wil be called as the
1143  * entry already holds the field information of the current event.
1144  *
1145  * It also checks if the event should be discarded or not.
1146  * It is to be discarded if the event is soft disabled and the
1147  * event was only recorded to process triggers, or if the event
1148  * filter is active and this event did not match the filters.
1149  *
1150  * Returns true if the event is discarded, false otherwise.
1151  */
1152 static inline bool
1153 __event_trigger_test_discard(struct trace_event_file *file,
1154 			     struct ring_buffer *buffer,
1155 			     struct ring_buffer_event *event,
1156 			     void *entry,
1157 			     enum event_trigger_type *tt)
1158 {
1159 	unsigned long eflags = file->flags;
1160 
1161 	if (eflags & EVENT_FILE_FL_TRIGGER_COND)
1162 		*tt = event_triggers_call(file, entry);
1163 
1164 	if (test_bit(EVENT_FILE_FL_SOFT_DISABLED_BIT, &file->flags) ||
1165 	    (unlikely(file->flags & EVENT_FILE_FL_FILTERED) &&
1166 	     !filter_match_preds(file->filter, entry))) {
1167 		__trace_event_discard_commit(buffer, event);
1168 		return true;
1169 	}
1170 
1171 	return false;
1172 }
1173 
1174 /**
1175  * event_trigger_unlock_commit - handle triggers and finish event commit
1176  * @file: The file pointer assoctiated to the event
1177  * @buffer: The ring buffer that the event is being written to
1178  * @event: The event meta data in the ring buffer
1179  * @entry: The event itself
1180  * @irq_flags: The state of the interrupts at the start of the event
1181  * @pc: The state of the preempt count at the start of the event.
1182  *
1183  * This is a helper function to handle triggers that require data
1184  * from the event itself. It also tests the event against filters and
1185  * if the event is soft disabled and should be discarded.
1186  */
1187 static inline void
1188 event_trigger_unlock_commit(struct trace_event_file *file,
1189 			    struct ring_buffer *buffer,
1190 			    struct ring_buffer_event *event,
1191 			    void *entry, unsigned long irq_flags, int pc)
1192 {
1193 	enum event_trigger_type tt = ETT_NONE;
1194 
1195 	if (!__event_trigger_test_discard(file, buffer, event, entry, &tt))
1196 		trace_buffer_unlock_commit(file->tr, buffer, event, irq_flags, pc);
1197 
1198 	if (tt)
1199 		event_triggers_post_call(file, tt, entry);
1200 }
1201 
1202 /**
1203  * event_trigger_unlock_commit_regs - handle triggers and finish event commit
1204  * @file: The file pointer assoctiated to the event
1205  * @buffer: The ring buffer that the event is being written to
1206  * @event: The event meta data in the ring buffer
1207  * @entry: The event itself
1208  * @irq_flags: The state of the interrupts at the start of the event
1209  * @pc: The state of the preempt count at the start of the event.
1210  *
1211  * This is a helper function to handle triggers that require data
1212  * from the event itself. It also tests the event against filters and
1213  * if the event is soft disabled and should be discarded.
1214  *
1215  * Same as event_trigger_unlock_commit() but calls
1216  * trace_buffer_unlock_commit_regs() instead of trace_buffer_unlock_commit().
1217  */
1218 static inline void
1219 event_trigger_unlock_commit_regs(struct trace_event_file *file,
1220 				 struct ring_buffer *buffer,
1221 				 struct ring_buffer_event *event,
1222 				 void *entry, unsigned long irq_flags, int pc,
1223 				 struct pt_regs *regs)
1224 {
1225 	enum event_trigger_type tt = ETT_NONE;
1226 
1227 	if (!__event_trigger_test_discard(file, buffer, event, entry, &tt))
1228 		trace_buffer_unlock_commit_regs(file->tr, buffer, event,
1229 						irq_flags, pc, regs);
1230 
1231 	if (tt)
1232 		event_triggers_post_call(file, tt, entry);
1233 }
1234 
1235 #define FILTER_PRED_INVALID	((unsigned short)-1)
1236 #define FILTER_PRED_IS_RIGHT	(1 << 15)
1237 #define FILTER_PRED_FOLD	(1 << 15)
1238 
1239 /*
1240  * The max preds is the size of unsigned short with
1241  * two flags at the MSBs. One bit is used for both the IS_RIGHT
1242  * and FOLD flags. The other is reserved.
1243  *
1244  * 2^14 preds is way more than enough.
1245  */
1246 #define MAX_FILTER_PRED		16384
1247 
1248 struct filter_pred;
1249 struct regex;
1250 
1251 typedef int (*filter_pred_fn_t) (struct filter_pred *pred, void *event);
1252 
1253 typedef int (*regex_match_func)(char *str, struct regex *r, int len);
1254 
1255 enum regex_type {
1256 	MATCH_FULL = 0,
1257 	MATCH_FRONT_ONLY,
1258 	MATCH_MIDDLE_ONLY,
1259 	MATCH_END_ONLY,
1260 };
1261 
1262 struct regex {
1263 	char			pattern[MAX_FILTER_STR_VAL];
1264 	int			len;
1265 	int			field_len;
1266 	regex_match_func	match;
1267 };
1268 
1269 struct filter_pred {
1270 	filter_pred_fn_t 	fn;
1271 	u64 			val;
1272 	struct regex		regex;
1273 	unsigned short		*ops;
1274 	struct ftrace_event_field *field;
1275 	int 			offset;
1276 	int 			not;
1277 	int 			op;
1278 	unsigned short		index;
1279 	unsigned short		parent;
1280 	unsigned short		left;
1281 	unsigned short		right;
1282 };
1283 
1284 static inline bool is_string_field(struct ftrace_event_field *field)
1285 {
1286 	return field->filter_type == FILTER_DYN_STRING ||
1287 	       field->filter_type == FILTER_STATIC_STRING ||
1288 	       field->filter_type == FILTER_PTR_STRING;
1289 }
1290 
1291 static inline bool is_function_field(struct ftrace_event_field *field)
1292 {
1293 	return field->filter_type == FILTER_TRACE_FN;
1294 }
1295 
1296 extern enum regex_type
1297 filter_parse_regex(char *buff, int len, char **search, int *not);
1298 extern void print_event_filter(struct trace_event_file *file,
1299 			       struct trace_seq *s);
1300 extern int apply_event_filter(struct trace_event_file *file,
1301 			      char *filter_string);
1302 extern int apply_subsystem_event_filter(struct trace_subsystem_dir *dir,
1303 					char *filter_string);
1304 extern void print_subsystem_event_filter(struct event_subsystem *system,
1305 					 struct trace_seq *s);
1306 extern int filter_assign_type(const char *type);
1307 extern int create_event_filter(struct trace_event_call *call,
1308 			       char *filter_str, bool set_str,
1309 			       struct event_filter **filterp);
1310 extern void free_event_filter(struct event_filter *filter);
1311 
1312 struct ftrace_event_field *
1313 trace_find_event_field(struct trace_event_call *call, char *name);
1314 
1315 extern void trace_event_enable_cmd_record(bool enable);
1316 extern int event_trace_add_tracer(struct dentry *parent, struct trace_array *tr);
1317 extern int event_trace_del_tracer(struct trace_array *tr);
1318 
1319 extern struct trace_event_file *find_event_file(struct trace_array *tr,
1320 						const char *system,
1321 						const char *event);
1322 
1323 static inline void *event_file_data(struct file *filp)
1324 {
1325 	return ACCESS_ONCE(file_inode(filp)->i_private);
1326 }
1327 
1328 extern struct mutex event_mutex;
1329 extern struct list_head ftrace_events;
1330 
1331 extern const struct file_operations event_trigger_fops;
1332 extern const struct file_operations event_hist_fops;
1333 
1334 #ifdef CONFIG_HIST_TRIGGERS
1335 extern int register_trigger_hist_cmd(void);
1336 extern int register_trigger_hist_enable_disable_cmds(void);
1337 #else
1338 static inline int register_trigger_hist_cmd(void) { return 0; }
1339 static inline int register_trigger_hist_enable_disable_cmds(void) { return 0; }
1340 #endif
1341 
1342 extern int register_trigger_cmds(void);
1343 extern void clear_event_triggers(struct trace_array *tr);
1344 
1345 struct event_trigger_data {
1346 	unsigned long			count;
1347 	int				ref;
1348 	struct event_trigger_ops	*ops;
1349 	struct event_command		*cmd_ops;
1350 	struct event_filter __rcu	*filter;
1351 	char				*filter_str;
1352 	void				*private_data;
1353 	bool				paused;
1354 	bool				paused_tmp;
1355 	struct list_head		list;
1356 	char				*name;
1357 	struct list_head		named_list;
1358 	struct event_trigger_data	*named_data;
1359 };
1360 
1361 /* Avoid typos */
1362 #define ENABLE_EVENT_STR	"enable_event"
1363 #define DISABLE_EVENT_STR	"disable_event"
1364 #define ENABLE_HIST_STR		"enable_hist"
1365 #define DISABLE_HIST_STR	"disable_hist"
1366 
1367 struct enable_trigger_data {
1368 	struct trace_event_file		*file;
1369 	bool				enable;
1370 	bool				hist;
1371 };
1372 
1373 extern int event_enable_trigger_print(struct seq_file *m,
1374 				      struct event_trigger_ops *ops,
1375 				      struct event_trigger_data *data);
1376 extern void event_enable_trigger_free(struct event_trigger_ops *ops,
1377 				      struct event_trigger_data *data);
1378 extern int event_enable_trigger_func(struct event_command *cmd_ops,
1379 				     struct trace_event_file *file,
1380 				     char *glob, char *cmd, char *param);
1381 extern int event_enable_register_trigger(char *glob,
1382 					 struct event_trigger_ops *ops,
1383 					 struct event_trigger_data *data,
1384 					 struct trace_event_file *file);
1385 extern void event_enable_unregister_trigger(char *glob,
1386 					    struct event_trigger_ops *ops,
1387 					    struct event_trigger_data *test,
1388 					    struct trace_event_file *file);
1389 extern void trigger_data_free(struct event_trigger_data *data);
1390 extern int event_trigger_init(struct event_trigger_ops *ops,
1391 			      struct event_trigger_data *data);
1392 extern int trace_event_trigger_enable_disable(struct trace_event_file *file,
1393 					      int trigger_enable);
1394 extern void update_cond_flag(struct trace_event_file *file);
1395 extern void unregister_trigger(char *glob, struct event_trigger_ops *ops,
1396 			       struct event_trigger_data *test,
1397 			       struct trace_event_file *file);
1398 extern int set_trigger_filter(char *filter_str,
1399 			      struct event_trigger_data *trigger_data,
1400 			      struct trace_event_file *file);
1401 extern struct event_trigger_data *find_named_trigger(const char *name);
1402 extern bool is_named_trigger(struct event_trigger_data *test);
1403 extern int save_named_trigger(const char *name,
1404 			      struct event_trigger_data *data);
1405 extern void del_named_trigger(struct event_trigger_data *data);
1406 extern void pause_named_trigger(struct event_trigger_data *data);
1407 extern void unpause_named_trigger(struct event_trigger_data *data);
1408 extern void set_named_trigger_data(struct event_trigger_data *data,
1409 				   struct event_trigger_data *named_data);
1410 extern int register_event_command(struct event_command *cmd);
1411 extern int unregister_event_command(struct event_command *cmd);
1412 extern int register_trigger_hist_enable_disable_cmds(void);
1413 
1414 /**
1415  * struct event_trigger_ops - callbacks for trace event triggers
1416  *
1417  * The methods in this structure provide per-event trigger hooks for
1418  * various trigger operations.
1419  *
1420  * All the methods below, except for @init() and @free(), must be
1421  * implemented.
1422  *
1423  * @func: The trigger 'probe' function called when the triggering
1424  *	event occurs.  The data passed into this callback is the data
1425  *	that was supplied to the event_command @reg() function that
1426  *	registered the trigger (see struct event_command) along with
1427  *	the trace record, rec.
1428  *
1429  * @init: An optional initialization function called for the trigger
1430  *	when the trigger is registered (via the event_command reg()
1431  *	function).  This can be used to perform per-trigger
1432  *	initialization such as incrementing a per-trigger reference
1433  *	count, for instance.  This is usually implemented by the
1434  *	generic utility function @event_trigger_init() (see
1435  *	trace_event_triggers.c).
1436  *
1437  * @free: An optional de-initialization function called for the
1438  *	trigger when the trigger is unregistered (via the
1439  *	event_command @reg() function).  This can be used to perform
1440  *	per-trigger de-initialization such as decrementing a
1441  *	per-trigger reference count and freeing corresponding trigger
1442  *	data, for instance.  This is usually implemented by the
1443  *	generic utility function @event_trigger_free() (see
1444  *	trace_event_triggers.c).
1445  *
1446  * @print: The callback function invoked to have the trigger print
1447  *	itself.  This is usually implemented by a wrapper function
1448  *	that calls the generic utility function @event_trigger_print()
1449  *	(see trace_event_triggers.c).
1450  */
1451 struct event_trigger_ops {
1452 	void			(*func)(struct event_trigger_data *data,
1453 					void *rec);
1454 	int			(*init)(struct event_trigger_ops *ops,
1455 					struct event_trigger_data *data);
1456 	void			(*free)(struct event_trigger_ops *ops,
1457 					struct event_trigger_data *data);
1458 	int			(*print)(struct seq_file *m,
1459 					 struct event_trigger_ops *ops,
1460 					 struct event_trigger_data *data);
1461 };
1462 
1463 /**
1464  * struct event_command - callbacks and data members for event commands
1465  *
1466  * Event commands are invoked by users by writing the command name
1467  * into the 'trigger' file associated with a trace event.  The
1468  * parameters associated with a specific invocation of an event
1469  * command are used to create an event trigger instance, which is
1470  * added to the list of trigger instances associated with that trace
1471  * event.  When the event is hit, the set of triggers associated with
1472  * that event is invoked.
1473  *
1474  * The data members in this structure provide per-event command data
1475  * for various event commands.
1476  *
1477  * All the data members below, except for @post_trigger, must be set
1478  * for each event command.
1479  *
1480  * @name: The unique name that identifies the event command.  This is
1481  *	the name used when setting triggers via trigger files.
1482  *
1483  * @trigger_type: A unique id that identifies the event command
1484  *	'type'.  This value has two purposes, the first to ensure that
1485  *	only one trigger of the same type can be set at a given time
1486  *	for a particular event e.g. it doesn't make sense to have both
1487  *	a traceon and traceoff trigger attached to a single event at
1488  *	the same time, so traceon and traceoff have the same type
1489  *	though they have different names.  The @trigger_type value is
1490  *	also used as a bit value for deferring the actual trigger
1491  *	action until after the current event is finished.  Some
1492  *	commands need to do this if they themselves log to the trace
1493  *	buffer (see the @post_trigger() member below).  @trigger_type
1494  *	values are defined by adding new values to the trigger_type
1495  *	enum in include/linux/trace_events.h.
1496  *
1497  * @flags: See the enum event_command_flags below.
1498  *
1499  * All the methods below, except for @set_filter() and @unreg_all(),
1500  * must be implemented.
1501  *
1502  * @func: The callback function responsible for parsing and
1503  *	registering the trigger written to the 'trigger' file by the
1504  *	user.  It allocates the trigger instance and registers it with
1505  *	the appropriate trace event.  It makes use of the other
1506  *	event_command callback functions to orchestrate this, and is
1507  *	usually implemented by the generic utility function
1508  *	@event_trigger_callback() (see trace_event_triggers.c).
1509  *
1510  * @reg: Adds the trigger to the list of triggers associated with the
1511  *	event, and enables the event trigger itself, after
1512  *	initializing it (via the event_trigger_ops @init() function).
1513  *	This is also where commands can use the @trigger_type value to
1514  *	make the decision as to whether or not multiple instances of
1515  *	the trigger should be allowed.  This is usually implemented by
1516  *	the generic utility function @register_trigger() (see
1517  *	trace_event_triggers.c).
1518  *
1519  * @unreg: Removes the trigger from the list of triggers associated
1520  *	with the event, and disables the event trigger itself, after
1521  *	initializing it (via the event_trigger_ops @free() function).
1522  *	This is usually implemented by the generic utility function
1523  *	@unregister_trigger() (see trace_event_triggers.c).
1524  *
1525  * @unreg_all: An optional function called to remove all the triggers
1526  *	from the list of triggers associated with the event.  Called
1527  *	when a trigger file is opened in truncate mode.
1528  *
1529  * @set_filter: An optional function called to parse and set a filter
1530  *	for the trigger.  If no @set_filter() method is set for the
1531  *	event command, filters set by the user for the command will be
1532  *	ignored.  This is usually implemented by the generic utility
1533  *	function @set_trigger_filter() (see trace_event_triggers.c).
1534  *
1535  * @get_trigger_ops: The callback function invoked to retrieve the
1536  *	event_trigger_ops implementation associated with the command.
1537  */
1538 struct event_command {
1539 	struct list_head	list;
1540 	char			*name;
1541 	enum event_trigger_type	trigger_type;
1542 	int			flags;
1543 	int			(*func)(struct event_command *cmd_ops,
1544 					struct trace_event_file *file,
1545 					char *glob, char *cmd, char *params);
1546 	int			(*reg)(char *glob,
1547 				       struct event_trigger_ops *ops,
1548 				       struct event_trigger_data *data,
1549 				       struct trace_event_file *file);
1550 	void			(*unreg)(char *glob,
1551 					 struct event_trigger_ops *ops,
1552 					 struct event_trigger_data *data,
1553 					 struct trace_event_file *file);
1554 	void			(*unreg_all)(struct trace_event_file *file);
1555 	int			(*set_filter)(char *filter_str,
1556 					      struct event_trigger_data *data,
1557 					      struct trace_event_file *file);
1558 	struct event_trigger_ops *(*get_trigger_ops)(char *cmd, char *param);
1559 };
1560 
1561 /**
1562  * enum event_command_flags - flags for struct event_command
1563  *
1564  * @POST_TRIGGER: A flag that says whether or not this command needs
1565  *	to have its action delayed until after the current event has
1566  *	been closed.  Some triggers need to avoid being invoked while
1567  *	an event is currently in the process of being logged, since
1568  *	the trigger may itself log data into the trace buffer.  Thus
1569  *	we make sure the current event is committed before invoking
1570  *	those triggers.  To do that, the trigger invocation is split
1571  *	in two - the first part checks the filter using the current
1572  *	trace record; if a command has the @post_trigger flag set, it
1573  *	sets a bit for itself in the return value, otherwise it
1574  *	directly invokes the trigger.  Once all commands have been
1575  *	either invoked or set their return flag, the current record is
1576  *	either committed or discarded.  At that point, if any commands
1577  *	have deferred their triggers, those commands are finally
1578  *	invoked following the close of the current event.  In other
1579  *	words, if the event_trigger_ops @func() probe implementation
1580  *	itself logs to the trace buffer, this flag should be set,
1581  *	otherwise it can be left unspecified.
1582  *
1583  * @NEEDS_REC: A flag that says whether or not this command needs
1584  *	access to the trace record in order to perform its function,
1585  *	regardless of whether or not it has a filter associated with
1586  *	it (filters make a trigger require access to the trace record
1587  *	but are not always present).
1588  */
1589 enum event_command_flags {
1590 	EVENT_CMD_FL_POST_TRIGGER	= 1,
1591 	EVENT_CMD_FL_NEEDS_REC		= 2,
1592 };
1593 
1594 static inline bool event_command_post_trigger(struct event_command *cmd_ops)
1595 {
1596 	return cmd_ops->flags & EVENT_CMD_FL_POST_TRIGGER;
1597 }
1598 
1599 static inline bool event_command_needs_rec(struct event_command *cmd_ops)
1600 {
1601 	return cmd_ops->flags & EVENT_CMD_FL_NEEDS_REC;
1602 }
1603 
1604 extern int trace_event_enable_disable(struct trace_event_file *file,
1605 				      int enable, int soft_disable);
1606 extern int tracing_alloc_snapshot(void);
1607 
1608 extern const char *__start___trace_bprintk_fmt[];
1609 extern const char *__stop___trace_bprintk_fmt[];
1610 
1611 extern const char *__start___tracepoint_str[];
1612 extern const char *__stop___tracepoint_str[];
1613 
1614 void trace_printk_control(bool enabled);
1615 void trace_printk_init_buffers(void);
1616 void trace_printk_start_comm(void);
1617 int trace_keep_overwrite(struct tracer *tracer, u32 mask, int set);
1618 int set_tracer_flag(struct trace_array *tr, unsigned int mask, int enabled);
1619 
1620 /*
1621  * Normal trace_printk() and friends allocates special buffers
1622  * to do the manipulation, as well as saves the print formats
1623  * into sections to display. But the trace infrastructure wants
1624  * to use these without the added overhead at the price of being
1625  * a bit slower (used mainly for warnings, where we don't care
1626  * about performance). The internal_trace_puts() is for such
1627  * a purpose.
1628  */
1629 #define internal_trace_puts(str) __trace_puts(_THIS_IP_, str, strlen(str))
1630 
1631 #undef FTRACE_ENTRY
1632 #define FTRACE_ENTRY(call, struct_name, id, tstruct, print, filter)	\
1633 	extern struct trace_event_call					\
1634 	__aligned(4) event_##call;
1635 #undef FTRACE_ENTRY_DUP
1636 #define FTRACE_ENTRY_DUP(call, struct_name, id, tstruct, print, filter)	\
1637 	FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print), \
1638 		     filter)
1639 #undef FTRACE_ENTRY_PACKED
1640 #define FTRACE_ENTRY_PACKED(call, struct_name, id, tstruct, print, filter) \
1641 	FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print), \
1642 		     filter)
1643 
1644 #include "trace_entries.h"
1645 
1646 #if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_FUNCTION_TRACER)
1647 int perf_ftrace_event_register(struct trace_event_call *call,
1648 			       enum trace_reg type, void *data);
1649 #else
1650 #define perf_ftrace_event_register NULL
1651 #endif
1652 
1653 #ifdef CONFIG_FTRACE_SYSCALLS
1654 void init_ftrace_syscalls(void);
1655 const char *get_syscall_name(int syscall);
1656 #else
1657 static inline void init_ftrace_syscalls(void) { }
1658 static inline const char *get_syscall_name(int syscall)
1659 {
1660 	return NULL;
1661 }
1662 #endif
1663 
1664 #ifdef CONFIG_EVENT_TRACING
1665 void trace_event_init(void);
1666 void trace_event_enum_update(struct trace_enum_map **map, int len);
1667 #else
1668 static inline void __init trace_event_init(void) { }
1669 static inline void trace_event_enum_update(struct trace_enum_map **map, int len) { }
1670 #endif
1671 
1672 extern struct trace_iterator *tracepoint_print_iter;
1673 
1674 #endif /* _LINUX_KERNEL_TRACE_H */
1675