xref: /openbmc/linux/kernel/trace/trace.h (revision 239480ab)
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 #include <linux/glob.h>
19 
20 #ifdef CONFIG_FTRACE_SYSCALLS
21 #include <asm/unistd.h>		/* For NR_SYSCALLS	     */
22 #include <asm/syscall.h>	/* some archs define it here */
23 #endif
24 
25 enum trace_type {
26 	__TRACE_FIRST_TYPE = 0,
27 
28 	TRACE_FN,
29 	TRACE_CTX,
30 	TRACE_WAKE,
31 	TRACE_STACK,
32 	TRACE_PRINT,
33 	TRACE_BPRINT,
34 	TRACE_MMIO_RW,
35 	TRACE_MMIO_MAP,
36 	TRACE_BRANCH,
37 	TRACE_GRAPH_RET,
38 	TRACE_GRAPH_ENT,
39 	TRACE_USER_STACK,
40 	TRACE_BLK,
41 	TRACE_BPUTS,
42 	TRACE_HWLAT,
43 	TRACE_RAW_DATA,
44 
45 	__TRACE_LAST_TYPE,
46 };
47 
48 
49 #undef __field
50 #define __field(type, item)		type	item;
51 
52 #undef __field_struct
53 #define __field_struct(type, item)	__field(type, item)
54 
55 #undef __field_desc
56 #define __field_desc(type, container, item)
57 
58 #undef __array
59 #define __array(type, item, size)	type	item[size];
60 
61 #undef __array_desc
62 #define __array_desc(type, container, item, size)
63 
64 #undef __dynamic_array
65 #define __dynamic_array(type, item)	type	item[];
66 
67 #undef F_STRUCT
68 #define F_STRUCT(args...)		args
69 
70 #undef FTRACE_ENTRY
71 #define FTRACE_ENTRY(name, struct_name, id, tstruct, print, filter)	\
72 	struct struct_name {						\
73 		struct trace_entry	ent;				\
74 		tstruct							\
75 	}
76 
77 #undef FTRACE_ENTRY_DUP
78 #define FTRACE_ENTRY_DUP(name, name_struct, id, tstruct, printk, filter)
79 
80 #undef FTRACE_ENTRY_REG
81 #define FTRACE_ENTRY_REG(name, struct_name, id, tstruct, print,	\
82 			 filter, regfn) \
83 	FTRACE_ENTRY(name, struct_name, id, PARAMS(tstruct), PARAMS(print), \
84 		     filter)
85 
86 #undef FTRACE_ENTRY_PACKED
87 #define FTRACE_ENTRY_PACKED(name, struct_name, id, tstruct, print,	\
88 			    filter)					\
89 	FTRACE_ENTRY(name, struct_name, id, PARAMS(tstruct), PARAMS(print), \
90 		     filter) __packed
91 
92 #include "trace_entries.h"
93 
94 /*
95  * syscalls are special, and need special handling, this is why
96  * they are not included in trace_entries.h
97  */
98 struct syscall_trace_enter {
99 	struct trace_entry	ent;
100 	int			nr;
101 	unsigned long		args[];
102 };
103 
104 struct syscall_trace_exit {
105 	struct trace_entry	ent;
106 	int			nr;
107 	long			ret;
108 };
109 
110 struct kprobe_trace_entry_head {
111 	struct trace_entry	ent;
112 	unsigned long		ip;
113 };
114 
115 struct kretprobe_trace_entry_head {
116 	struct trace_entry	ent;
117 	unsigned long		func;
118 	unsigned long		ret_ip;
119 };
120 
121 /*
122  * trace_flag_type is an enumeration that holds different
123  * states when a trace occurs. These are:
124  *  IRQS_OFF		- interrupts were disabled
125  *  IRQS_NOSUPPORT	- arch does not support irqs_disabled_flags
126  *  NEED_RESCHED	- reschedule is requested
127  *  HARDIRQ		- inside an interrupt handler
128  *  SOFTIRQ		- inside a softirq handler
129  */
130 enum trace_flag_type {
131 	TRACE_FLAG_IRQS_OFF		= 0x01,
132 	TRACE_FLAG_IRQS_NOSUPPORT	= 0x02,
133 	TRACE_FLAG_NEED_RESCHED		= 0x04,
134 	TRACE_FLAG_HARDIRQ		= 0x08,
135 	TRACE_FLAG_SOFTIRQ		= 0x10,
136 	TRACE_FLAG_PREEMPT_RESCHED	= 0x20,
137 	TRACE_FLAG_NMI			= 0x40,
138 };
139 
140 #define TRACE_BUF_SIZE		1024
141 
142 struct trace_array;
143 
144 /*
145  * The CPU trace array - it consists of thousands of trace entries
146  * plus some other descriptor data: (for example which task started
147  * the trace, etc.)
148  */
149 struct trace_array_cpu {
150 	atomic_t		disabled;
151 	void			*buffer_page;	/* ring buffer spare */
152 
153 	unsigned long		entries;
154 	unsigned long		saved_latency;
155 	unsigned long		critical_start;
156 	unsigned long		critical_end;
157 	unsigned long		critical_sequence;
158 	unsigned long		nice;
159 	unsigned long		policy;
160 	unsigned long		rt_priority;
161 	unsigned long		skipped_entries;
162 	u64			preempt_timestamp;
163 	pid_t			pid;
164 	kuid_t			uid;
165 	char			comm[TASK_COMM_LEN];
166 
167 	bool			ignore_pid;
168 #ifdef CONFIG_FUNCTION_TRACER
169 	bool			ftrace_ignore_pid;
170 #endif
171 };
172 
173 struct tracer;
174 struct trace_option_dentry;
175 
176 struct trace_buffer {
177 	struct trace_array		*tr;
178 	struct ring_buffer		*buffer;
179 	struct trace_array_cpu __percpu	*data;
180 	u64				time_start;
181 	int				cpu;
182 };
183 
184 #define TRACE_FLAGS_MAX_SIZE		32
185 
186 struct trace_options {
187 	struct tracer			*tracer;
188 	struct trace_option_dentry	*topts;
189 };
190 
191 struct trace_pid_list {
192 	int				pid_max;
193 	unsigned long			*pids;
194 };
195 
196 /*
197  * The trace array - an array of per-CPU trace arrays. This is the
198  * highest level data structure that individual tracers deal with.
199  * They have on/off state as well:
200  */
201 struct trace_array {
202 	struct list_head	list;
203 	char			*name;
204 	struct trace_buffer	trace_buffer;
205 #ifdef CONFIG_TRACER_MAX_TRACE
206 	/*
207 	 * The max_buffer is used to snapshot the trace when a maximum
208 	 * latency is reached, or when the user initiates a snapshot.
209 	 * Some tracers will use this to store a maximum trace while
210 	 * it continues examining live traces.
211 	 *
212 	 * The buffers for the max_buffer are set up the same as the trace_buffer
213 	 * When a snapshot is taken, the buffer of the max_buffer is swapped
214 	 * with the buffer of the trace_buffer and the buffers are reset for
215 	 * the trace_buffer so the tracing can continue.
216 	 */
217 	struct trace_buffer	max_buffer;
218 	bool			allocated_snapshot;
219 #endif
220 #if defined(CONFIG_TRACER_MAX_TRACE) || defined(CONFIG_HWLAT_TRACER)
221 	unsigned long		max_latency;
222 #endif
223 	struct trace_pid_list	__rcu *filtered_pids;
224 	/*
225 	 * max_lock is used to protect the swapping of buffers
226 	 * when taking a max snapshot. The buffers themselves are
227 	 * protected by per_cpu spinlocks. But the action of the swap
228 	 * needs its own lock.
229 	 *
230 	 * This is defined as a arch_spinlock_t in order to help
231 	 * with performance when lockdep debugging is enabled.
232 	 *
233 	 * It is also used in other places outside the update_max_tr
234 	 * so it needs to be defined outside of the
235 	 * CONFIG_TRACER_MAX_TRACE.
236 	 */
237 	arch_spinlock_t		max_lock;
238 	int			buffer_disabled;
239 #ifdef CONFIG_FTRACE_SYSCALLS
240 	int			sys_refcount_enter;
241 	int			sys_refcount_exit;
242 	struct trace_event_file __rcu *enter_syscall_files[NR_syscalls];
243 	struct trace_event_file __rcu *exit_syscall_files[NR_syscalls];
244 #endif
245 	int			stop_count;
246 	int			clock_id;
247 	int			nr_topts;
248 	struct tracer		*current_trace;
249 	unsigned int		trace_flags;
250 	unsigned char		trace_flags_index[TRACE_FLAGS_MAX_SIZE];
251 	unsigned int		flags;
252 	raw_spinlock_t		start_lock;
253 	struct dentry		*dir;
254 	struct dentry		*options;
255 	struct dentry		*percpu_dir;
256 	struct dentry		*event_dir;
257 	struct trace_options	*topts;
258 	struct list_head	systems;
259 	struct list_head	events;
260 	cpumask_var_t		tracing_cpumask; /* only trace on set CPUs */
261 	int			ref;
262 #ifdef CONFIG_FUNCTION_TRACER
263 	struct ftrace_ops	*ops;
264 	struct trace_pid_list	__rcu *function_pids;
265 #ifdef CONFIG_DYNAMIC_FTRACE
266 	struct list_head	func_probes;
267 #endif
268 	/* function tracing enabled */
269 	int			function_enabled;
270 #endif
271 };
272 
273 enum {
274 	TRACE_ARRAY_FL_GLOBAL	= (1 << 0)
275 };
276 
277 extern struct list_head ftrace_trace_arrays;
278 
279 extern struct mutex trace_types_lock;
280 
281 extern int trace_array_get(struct trace_array *tr);
282 extern void trace_array_put(struct trace_array *tr);
283 
284 /*
285  * The global tracer (top) should be the first trace array added,
286  * but we check the flag anyway.
287  */
288 static inline struct trace_array *top_trace_array(void)
289 {
290 	struct trace_array *tr;
291 
292 	if (list_empty(&ftrace_trace_arrays))
293 		return NULL;
294 
295 	tr = list_entry(ftrace_trace_arrays.prev,
296 			typeof(*tr), list);
297 	WARN_ON(!(tr->flags & TRACE_ARRAY_FL_GLOBAL));
298 	return tr;
299 }
300 
301 #define FTRACE_CMP_TYPE(var, type) \
302 	__builtin_types_compatible_p(typeof(var), type *)
303 
304 #undef IF_ASSIGN
305 #define IF_ASSIGN(var, entry, etype, id)		\
306 	if (FTRACE_CMP_TYPE(var, etype)) {		\
307 		var = (typeof(var))(entry);		\
308 		WARN_ON(id && (entry)->type != id);	\
309 		break;					\
310 	}
311 
312 /* Will cause compile errors if type is not found. */
313 extern void __ftrace_bad_type(void);
314 
315 /*
316  * The trace_assign_type is a verifier that the entry type is
317  * the same as the type being assigned. To add new types simply
318  * add a line with the following format:
319  *
320  * IF_ASSIGN(var, ent, type, id);
321  *
322  *  Where "type" is the trace type that includes the trace_entry
323  *  as the "ent" item. And "id" is the trace identifier that is
324  *  used in the trace_type enum.
325  *
326  *  If the type can have more than one id, then use zero.
327  */
328 #define trace_assign_type(var, ent)					\
329 	do {								\
330 		IF_ASSIGN(var, ent, struct ftrace_entry, TRACE_FN);	\
331 		IF_ASSIGN(var, ent, struct ctx_switch_entry, 0);	\
332 		IF_ASSIGN(var, ent, struct stack_entry, TRACE_STACK);	\
333 		IF_ASSIGN(var, ent, struct userstack_entry, TRACE_USER_STACK);\
334 		IF_ASSIGN(var, ent, struct print_entry, TRACE_PRINT);	\
335 		IF_ASSIGN(var, ent, struct bprint_entry, TRACE_BPRINT);	\
336 		IF_ASSIGN(var, ent, struct bputs_entry, TRACE_BPUTS);	\
337 		IF_ASSIGN(var, ent, struct hwlat_entry, TRACE_HWLAT);	\
338 		IF_ASSIGN(var, ent, struct raw_data_entry, TRACE_RAW_DATA);\
339 		IF_ASSIGN(var, ent, struct trace_mmiotrace_rw,		\
340 			  TRACE_MMIO_RW);				\
341 		IF_ASSIGN(var, ent, struct trace_mmiotrace_map,		\
342 			  TRACE_MMIO_MAP);				\
343 		IF_ASSIGN(var, ent, struct trace_branch, TRACE_BRANCH); \
344 		IF_ASSIGN(var, ent, struct ftrace_graph_ent_entry,	\
345 			  TRACE_GRAPH_ENT);		\
346 		IF_ASSIGN(var, ent, struct ftrace_graph_ret_entry,	\
347 			  TRACE_GRAPH_RET);		\
348 		__ftrace_bad_type();					\
349 	} while (0)
350 
351 /*
352  * An option specific to a tracer. This is a boolean value.
353  * The bit is the bit index that sets its value on the
354  * flags value in struct tracer_flags.
355  */
356 struct tracer_opt {
357 	const char	*name; /* Will appear on the trace_options file */
358 	u32		bit; /* Mask assigned in val field in tracer_flags */
359 };
360 
361 /*
362  * The set of specific options for a tracer. Your tracer
363  * have to set the initial value of the flags val.
364  */
365 struct tracer_flags {
366 	u32			val;
367 	struct tracer_opt	*opts;
368 	struct tracer		*trace;
369 };
370 
371 /* Makes more easy to define a tracer opt */
372 #define TRACER_OPT(s, b)	.name = #s, .bit = b
373 
374 
375 struct trace_option_dentry {
376 	struct tracer_opt		*opt;
377 	struct tracer_flags		*flags;
378 	struct trace_array		*tr;
379 	struct dentry			*entry;
380 };
381 
382 /**
383  * struct tracer - a specific tracer and its callbacks to interact with tracefs
384  * @name: the name chosen to select it on the available_tracers file
385  * @init: called when one switches to this tracer (echo name > current_tracer)
386  * @reset: called when one switches to another tracer
387  * @start: called when tracing is unpaused (echo 1 > tracing_on)
388  * @stop: called when tracing is paused (echo 0 > tracing_on)
389  * @update_thresh: called when tracing_thresh is updated
390  * @open: called when the trace file is opened
391  * @pipe_open: called when the trace_pipe file is opened
392  * @close: called when the trace file is released
393  * @pipe_close: called when the trace_pipe file is released
394  * @read: override the default read callback on trace_pipe
395  * @splice_read: override the default splice_read callback on trace_pipe
396  * @selftest: selftest to run on boot (see trace_selftest.c)
397  * @print_headers: override the first lines that describe your columns
398  * @print_line: callback that prints a trace
399  * @set_flag: signals one of your private flags changed (trace_options file)
400  * @flags: your private flags
401  */
402 struct tracer {
403 	const char		*name;
404 	int			(*init)(struct trace_array *tr);
405 	void			(*reset)(struct trace_array *tr);
406 	void			(*start)(struct trace_array *tr);
407 	void			(*stop)(struct trace_array *tr);
408 	int			(*update_thresh)(struct trace_array *tr);
409 	void			(*open)(struct trace_iterator *iter);
410 	void			(*pipe_open)(struct trace_iterator *iter);
411 	void			(*close)(struct trace_iterator *iter);
412 	void			(*pipe_close)(struct trace_iterator *iter);
413 	ssize_t			(*read)(struct trace_iterator *iter,
414 					struct file *filp, char __user *ubuf,
415 					size_t cnt, loff_t *ppos);
416 	ssize_t			(*splice_read)(struct trace_iterator *iter,
417 					       struct file *filp,
418 					       loff_t *ppos,
419 					       struct pipe_inode_info *pipe,
420 					       size_t len,
421 					       unsigned int flags);
422 #ifdef CONFIG_FTRACE_STARTUP_TEST
423 	int			(*selftest)(struct tracer *trace,
424 					    struct trace_array *tr);
425 #endif
426 	void			(*print_header)(struct seq_file *m);
427 	enum print_line_t	(*print_line)(struct trace_iterator *iter);
428 	/* If you handled the flag setting, return 0 */
429 	int			(*set_flag)(struct trace_array *tr,
430 					    u32 old_flags, u32 bit, int set);
431 	/* Return 0 if OK with change, else return non-zero */
432 	int			(*flag_changed)(struct trace_array *tr,
433 						u32 mask, int set);
434 	struct tracer		*next;
435 	struct tracer_flags	*flags;
436 	int			enabled;
437 	int			ref;
438 	bool			print_max;
439 	bool			allow_instances;
440 #ifdef CONFIG_TRACER_MAX_TRACE
441 	bool			use_max_tr;
442 #endif
443 };
444 
445 
446 /* Only current can touch trace_recursion */
447 
448 /*
449  * For function tracing recursion:
450  *  The order of these bits are important.
451  *
452  *  When function tracing occurs, the following steps are made:
453  *   If arch does not support a ftrace feature:
454  *    call internal function (uses INTERNAL bits) which calls...
455  *   If callback is registered to the "global" list, the list
456  *    function is called and recursion checks the GLOBAL bits.
457  *    then this function calls...
458  *   The function callback, which can use the FTRACE bits to
459  *    check for recursion.
460  *
461  * Now if the arch does not suppport a feature, and it calls
462  * the global list function which calls the ftrace callback
463  * all three of these steps will do a recursion protection.
464  * There's no reason to do one if the previous caller already
465  * did. The recursion that we are protecting against will
466  * go through the same steps again.
467  *
468  * To prevent the multiple recursion checks, if a recursion
469  * bit is set that is higher than the MAX bit of the current
470  * check, then we know that the check was made by the previous
471  * caller, and we can skip the current check.
472  */
473 enum {
474 	TRACE_BUFFER_BIT,
475 	TRACE_BUFFER_NMI_BIT,
476 	TRACE_BUFFER_IRQ_BIT,
477 	TRACE_BUFFER_SIRQ_BIT,
478 
479 	/* Start of function recursion bits */
480 	TRACE_FTRACE_BIT,
481 	TRACE_FTRACE_NMI_BIT,
482 	TRACE_FTRACE_IRQ_BIT,
483 	TRACE_FTRACE_SIRQ_BIT,
484 
485 	/* INTERNAL_BITs must be greater than FTRACE_BITs */
486 	TRACE_INTERNAL_BIT,
487 	TRACE_INTERNAL_NMI_BIT,
488 	TRACE_INTERNAL_IRQ_BIT,
489 	TRACE_INTERNAL_SIRQ_BIT,
490 
491 	TRACE_BRANCH_BIT,
492 /*
493  * Abuse of the trace_recursion.
494  * As we need a way to maintain state if we are tracing the function
495  * graph in irq because we want to trace a particular function that
496  * was called in irq context but we have irq tracing off. Since this
497  * can only be modified by current, we can reuse trace_recursion.
498  */
499 	TRACE_IRQ_BIT,
500 };
501 
502 #define trace_recursion_set(bit)	do { (current)->trace_recursion |= (1<<(bit)); } while (0)
503 #define trace_recursion_clear(bit)	do { (current)->trace_recursion &= ~(1<<(bit)); } while (0)
504 #define trace_recursion_test(bit)	((current)->trace_recursion & (1<<(bit)))
505 
506 #define TRACE_CONTEXT_BITS	4
507 
508 #define TRACE_FTRACE_START	TRACE_FTRACE_BIT
509 #define TRACE_FTRACE_MAX	((1 << (TRACE_FTRACE_START + TRACE_CONTEXT_BITS)) - 1)
510 
511 #define TRACE_LIST_START	TRACE_INTERNAL_BIT
512 #define TRACE_LIST_MAX		((1 << (TRACE_LIST_START + TRACE_CONTEXT_BITS)) - 1)
513 
514 #define TRACE_CONTEXT_MASK	TRACE_LIST_MAX
515 
516 static __always_inline int trace_get_context_bit(void)
517 {
518 	int bit;
519 
520 	if (in_interrupt()) {
521 		if (in_nmi())
522 			bit = 0;
523 
524 		else if (in_irq())
525 			bit = 1;
526 		else
527 			bit = 2;
528 	} else
529 		bit = 3;
530 
531 	return bit;
532 }
533 
534 static __always_inline int trace_test_and_set_recursion(int start, int max)
535 {
536 	unsigned int val = current->trace_recursion;
537 	int bit;
538 
539 	/* A previous recursion check was made */
540 	if ((val & TRACE_CONTEXT_MASK) > max)
541 		return 0;
542 
543 	bit = trace_get_context_bit() + start;
544 	if (unlikely(val & (1 << bit)))
545 		return -1;
546 
547 	val |= 1 << bit;
548 	current->trace_recursion = val;
549 	barrier();
550 
551 	return bit;
552 }
553 
554 static __always_inline void trace_clear_recursion(int bit)
555 {
556 	unsigned int val = current->trace_recursion;
557 
558 	if (!bit)
559 		return;
560 
561 	bit = 1 << bit;
562 	val &= ~bit;
563 
564 	barrier();
565 	current->trace_recursion = val;
566 }
567 
568 static inline struct ring_buffer_iter *
569 trace_buffer_iter(struct trace_iterator *iter, int cpu)
570 {
571 	if (iter->buffer_iter && iter->buffer_iter[cpu])
572 		return iter->buffer_iter[cpu];
573 	return NULL;
574 }
575 
576 int tracer_init(struct tracer *t, struct trace_array *tr);
577 int tracing_is_enabled(void);
578 void tracing_reset(struct trace_buffer *buf, int cpu);
579 void tracing_reset_online_cpus(struct trace_buffer *buf);
580 void tracing_reset_current(int cpu);
581 void tracing_reset_all_online_cpus(void);
582 int tracing_open_generic(struct inode *inode, struct file *filp);
583 bool tracing_is_disabled(void);
584 int tracer_tracing_is_on(struct trace_array *tr);
585 void tracer_tracing_on(struct trace_array *tr);
586 void tracer_tracing_off(struct trace_array *tr);
587 struct dentry *trace_create_file(const char *name,
588 				 umode_t mode,
589 				 struct dentry *parent,
590 				 void *data,
591 				 const struct file_operations *fops);
592 
593 struct dentry *tracing_init_dentry(void);
594 
595 struct ring_buffer_event;
596 
597 struct ring_buffer_event *
598 trace_buffer_lock_reserve(struct ring_buffer *buffer,
599 			  int type,
600 			  unsigned long len,
601 			  unsigned long flags,
602 			  int pc);
603 
604 struct trace_entry *tracing_get_trace_entry(struct trace_array *tr,
605 						struct trace_array_cpu *data);
606 
607 struct trace_entry *trace_find_next_entry(struct trace_iterator *iter,
608 					  int *ent_cpu, u64 *ent_ts);
609 
610 void trace_buffer_unlock_commit_nostack(struct ring_buffer *buffer,
611 					struct ring_buffer_event *event);
612 
613 int trace_empty(struct trace_iterator *iter);
614 
615 void *trace_find_next_entry_inc(struct trace_iterator *iter);
616 
617 void trace_init_global_iter(struct trace_iterator *iter);
618 
619 void tracing_iter_reset(struct trace_iterator *iter, int cpu);
620 
621 void trace_function(struct trace_array *tr,
622 		    unsigned long ip,
623 		    unsigned long parent_ip,
624 		    unsigned long flags, int pc);
625 void trace_graph_function(struct trace_array *tr,
626 		    unsigned long ip,
627 		    unsigned long parent_ip,
628 		    unsigned long flags, int pc);
629 void trace_latency_header(struct seq_file *m);
630 void trace_default_header(struct seq_file *m);
631 void print_trace_header(struct seq_file *m, struct trace_iterator *iter);
632 int trace_empty(struct trace_iterator *iter);
633 
634 void trace_graph_return(struct ftrace_graph_ret *trace);
635 int trace_graph_entry(struct ftrace_graph_ent *trace);
636 void set_graph_array(struct trace_array *tr);
637 
638 void tracing_start_cmdline_record(void);
639 void tracing_stop_cmdline_record(void);
640 int register_tracer(struct tracer *type);
641 int is_tracing_stopped(void);
642 
643 loff_t tracing_lseek(struct file *file, loff_t offset, int whence);
644 
645 extern cpumask_var_t __read_mostly tracing_buffer_mask;
646 
647 #define for_each_tracing_cpu(cpu)	\
648 	for_each_cpu(cpu, tracing_buffer_mask)
649 
650 extern unsigned long nsecs_to_usecs(unsigned long nsecs);
651 
652 extern unsigned long tracing_thresh;
653 
654 /* PID filtering */
655 
656 extern int pid_max;
657 
658 bool trace_find_filtered_pid(struct trace_pid_list *filtered_pids,
659 			     pid_t search_pid);
660 bool trace_ignore_this_task(struct trace_pid_list *filtered_pids,
661 			    struct task_struct *task);
662 void trace_filter_add_remove_task(struct trace_pid_list *pid_list,
663 				  struct task_struct *self,
664 				  struct task_struct *task);
665 void *trace_pid_next(struct trace_pid_list *pid_list, void *v, loff_t *pos);
666 void *trace_pid_start(struct trace_pid_list *pid_list, loff_t *pos);
667 int trace_pid_show(struct seq_file *m, void *v);
668 void trace_free_pid_list(struct trace_pid_list *pid_list);
669 int trace_pid_write(struct trace_pid_list *filtered_pids,
670 		    struct trace_pid_list **new_pid_list,
671 		    const char __user *ubuf, size_t cnt);
672 
673 #ifdef CONFIG_TRACER_MAX_TRACE
674 void update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu);
675 void update_max_tr_single(struct trace_array *tr,
676 			  struct task_struct *tsk, int cpu);
677 #endif /* CONFIG_TRACER_MAX_TRACE */
678 
679 #ifdef CONFIG_STACKTRACE
680 void ftrace_trace_userstack(struct ring_buffer *buffer, unsigned long flags,
681 			    int pc);
682 
683 void __trace_stack(struct trace_array *tr, unsigned long flags, int skip,
684 		   int pc);
685 #else
686 static inline void ftrace_trace_userstack(struct ring_buffer *buffer,
687 					  unsigned long flags, int pc)
688 {
689 }
690 
691 static inline void __trace_stack(struct trace_array *tr, unsigned long flags,
692 				 int skip, int pc)
693 {
694 }
695 #endif /* CONFIG_STACKTRACE */
696 
697 extern u64 ftrace_now(int cpu);
698 
699 extern void trace_find_cmdline(int pid, char comm[]);
700 extern void trace_event_follow_fork(struct trace_array *tr, bool enable);
701 
702 #ifdef CONFIG_DYNAMIC_FTRACE
703 extern unsigned long ftrace_update_tot_cnt;
704 void ftrace_init_trace_array(struct trace_array *tr);
705 #else
706 static inline void ftrace_init_trace_array(struct trace_array *tr) { }
707 #endif
708 #define DYN_FTRACE_TEST_NAME trace_selftest_dynamic_test_func
709 extern int DYN_FTRACE_TEST_NAME(void);
710 #define DYN_FTRACE_TEST_NAME2 trace_selftest_dynamic_test_func2
711 extern int DYN_FTRACE_TEST_NAME2(void);
712 
713 extern bool ring_buffer_expanded;
714 extern bool tracing_selftest_disabled;
715 
716 #ifdef CONFIG_FTRACE_STARTUP_TEST
717 extern int trace_selftest_startup_function(struct tracer *trace,
718 					   struct trace_array *tr);
719 extern int trace_selftest_startup_function_graph(struct tracer *trace,
720 						 struct trace_array *tr);
721 extern int trace_selftest_startup_irqsoff(struct tracer *trace,
722 					  struct trace_array *tr);
723 extern int trace_selftest_startup_preemptoff(struct tracer *trace,
724 					     struct trace_array *tr);
725 extern int trace_selftest_startup_preemptirqsoff(struct tracer *trace,
726 						 struct trace_array *tr);
727 extern int trace_selftest_startup_wakeup(struct tracer *trace,
728 					 struct trace_array *tr);
729 extern int trace_selftest_startup_nop(struct tracer *trace,
730 					 struct trace_array *tr);
731 extern int trace_selftest_startup_sched_switch(struct tracer *trace,
732 					       struct trace_array *tr);
733 extern int trace_selftest_startup_branch(struct tracer *trace,
734 					 struct trace_array *tr);
735 /*
736  * Tracer data references selftest functions that only occur
737  * on boot up. These can be __init functions. Thus, when selftests
738  * are enabled, then the tracers need to reference __init functions.
739  */
740 #define __tracer_data		__refdata
741 #else
742 /* Tracers are seldom changed. Optimize when selftests are disabled. */
743 #define __tracer_data		__read_mostly
744 #endif /* CONFIG_FTRACE_STARTUP_TEST */
745 
746 extern void *head_page(struct trace_array_cpu *data);
747 extern unsigned long long ns2usecs(u64 nsec);
748 extern int
749 trace_vbprintk(unsigned long ip, const char *fmt, va_list args);
750 extern int
751 trace_vprintk(unsigned long ip, const char *fmt, va_list args);
752 extern int
753 trace_array_vprintk(struct trace_array *tr,
754 		    unsigned long ip, const char *fmt, va_list args);
755 int trace_array_printk(struct trace_array *tr,
756 		       unsigned long ip, const char *fmt, ...);
757 int trace_array_printk_buf(struct ring_buffer *buffer,
758 			   unsigned long ip, const char *fmt, ...);
759 void trace_printk_seq(struct trace_seq *s);
760 enum print_line_t print_trace_line(struct trace_iterator *iter);
761 
762 extern char trace_find_mark(unsigned long long duration);
763 
764 struct ftrace_hash {
765 	unsigned long		size_bits;
766 	struct hlist_head	*buckets;
767 	unsigned long		count;
768 	struct rcu_head		rcu;
769 };
770 
771 struct ftrace_func_entry *
772 ftrace_lookup_ip(struct ftrace_hash *hash, unsigned long ip);
773 
774 static __always_inline bool ftrace_hash_empty(struct ftrace_hash *hash)
775 {
776 	return !hash || !hash->count;
777 }
778 
779 /* Standard output formatting function used for function return traces */
780 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
781 
782 /* Flag options */
783 #define TRACE_GRAPH_PRINT_OVERRUN       0x1
784 #define TRACE_GRAPH_PRINT_CPU           0x2
785 #define TRACE_GRAPH_PRINT_OVERHEAD      0x4
786 #define TRACE_GRAPH_PRINT_PROC          0x8
787 #define TRACE_GRAPH_PRINT_DURATION      0x10
788 #define TRACE_GRAPH_PRINT_ABS_TIME      0x20
789 #define TRACE_GRAPH_PRINT_IRQS          0x40
790 #define TRACE_GRAPH_PRINT_TAIL          0x80
791 #define TRACE_GRAPH_SLEEP_TIME		0x100
792 #define TRACE_GRAPH_GRAPH_TIME		0x200
793 #define TRACE_GRAPH_PRINT_FILL_SHIFT	28
794 #define TRACE_GRAPH_PRINT_FILL_MASK	(0x3 << TRACE_GRAPH_PRINT_FILL_SHIFT)
795 
796 extern void ftrace_graph_sleep_time_control(bool enable);
797 extern void ftrace_graph_graph_time_control(bool enable);
798 
799 extern enum print_line_t
800 print_graph_function_flags(struct trace_iterator *iter, u32 flags);
801 extern void print_graph_headers_flags(struct seq_file *s, u32 flags);
802 extern void
803 trace_print_graph_duration(unsigned long long duration, struct trace_seq *s);
804 extern void graph_trace_open(struct trace_iterator *iter);
805 extern void graph_trace_close(struct trace_iterator *iter);
806 extern int __trace_graph_entry(struct trace_array *tr,
807 			       struct ftrace_graph_ent *trace,
808 			       unsigned long flags, int pc);
809 extern void __trace_graph_return(struct trace_array *tr,
810 				 struct ftrace_graph_ret *trace,
811 				 unsigned long flags, int pc);
812 
813 #ifdef CONFIG_DYNAMIC_FTRACE
814 extern struct ftrace_hash *ftrace_graph_hash;
815 extern struct ftrace_hash *ftrace_graph_notrace_hash;
816 
817 static inline int ftrace_graph_addr(unsigned long addr)
818 {
819 	int ret = 0;
820 
821 	preempt_disable_notrace();
822 
823 	if (ftrace_hash_empty(ftrace_graph_hash)) {
824 		ret = 1;
825 		goto out;
826 	}
827 
828 	if (ftrace_lookup_ip(ftrace_graph_hash, addr)) {
829 		/*
830 		 * If no irqs are to be traced, but a set_graph_function
831 		 * is set, and called by an interrupt handler, we still
832 		 * want to trace it.
833 		 */
834 		if (in_irq())
835 			trace_recursion_set(TRACE_IRQ_BIT);
836 		else
837 			trace_recursion_clear(TRACE_IRQ_BIT);
838 		ret = 1;
839 	}
840 
841 out:
842 	preempt_enable_notrace();
843 	return ret;
844 }
845 
846 static inline int ftrace_graph_notrace_addr(unsigned long addr)
847 {
848 	int ret = 0;
849 
850 	preempt_disable_notrace();
851 
852 	if (ftrace_lookup_ip(ftrace_graph_notrace_hash, addr))
853 		ret = 1;
854 
855 	preempt_enable_notrace();
856 	return ret;
857 }
858 #else
859 static inline int ftrace_graph_addr(unsigned long addr)
860 {
861 	return 1;
862 }
863 
864 static inline int ftrace_graph_notrace_addr(unsigned long addr)
865 {
866 	return 0;
867 }
868 #endif /* CONFIG_DYNAMIC_FTRACE */
869 
870 extern unsigned int fgraph_max_depth;
871 
872 static inline bool ftrace_graph_ignore_func(struct ftrace_graph_ent *trace)
873 {
874 	/* trace it when it is-nested-in or is a function enabled. */
875 	return !(trace->depth || ftrace_graph_addr(trace->func)) ||
876 		(trace->depth < 0) ||
877 		(fgraph_max_depth && trace->depth >= fgraph_max_depth);
878 }
879 
880 #else /* CONFIG_FUNCTION_GRAPH_TRACER */
881 static inline enum print_line_t
882 print_graph_function_flags(struct trace_iterator *iter, u32 flags)
883 {
884 	return TRACE_TYPE_UNHANDLED;
885 }
886 #endif /* CONFIG_FUNCTION_GRAPH_TRACER */
887 
888 extern struct list_head ftrace_pids;
889 
890 #ifdef CONFIG_FUNCTION_TRACER
891 struct ftrace_func_command {
892 	struct list_head	list;
893 	char			*name;
894 	int			(*func)(struct trace_array *tr,
895 					struct ftrace_hash *hash,
896 					char *func, char *cmd,
897 					char *params, int enable);
898 };
899 extern bool ftrace_filter_param __initdata;
900 static inline int ftrace_trace_task(struct trace_array *tr)
901 {
902 	return !this_cpu_read(tr->trace_buffer.data->ftrace_ignore_pid);
903 }
904 extern int ftrace_is_dead(void);
905 int ftrace_create_function_files(struct trace_array *tr,
906 				 struct dentry *parent);
907 void ftrace_destroy_function_files(struct trace_array *tr);
908 void ftrace_init_global_array_ops(struct trace_array *tr);
909 void ftrace_init_array_ops(struct trace_array *tr, ftrace_func_t func);
910 void ftrace_reset_array_ops(struct trace_array *tr);
911 int using_ftrace_ops_list_func(void);
912 void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d_tracer);
913 void ftrace_init_tracefs_toplevel(struct trace_array *tr,
914 				  struct dentry *d_tracer);
915 void ftrace_clear_pids(struct trace_array *tr);
916 int init_function_trace(void);
917 void ftrace_pid_follow_fork(struct trace_array *tr, bool enable);
918 #else
919 static inline int ftrace_trace_task(struct trace_array *tr)
920 {
921 	return 1;
922 }
923 static inline int ftrace_is_dead(void) { return 0; }
924 static inline int
925 ftrace_create_function_files(struct trace_array *tr,
926 			     struct dentry *parent)
927 {
928 	return 0;
929 }
930 static inline void ftrace_destroy_function_files(struct trace_array *tr) { }
931 static inline __init void
932 ftrace_init_global_array_ops(struct trace_array *tr) { }
933 static inline void ftrace_reset_array_ops(struct trace_array *tr) { }
934 static inline void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d) { }
935 static inline void ftrace_init_tracefs_toplevel(struct trace_array *tr, struct dentry *d) { }
936 static inline void ftrace_clear_pids(struct trace_array *tr) { }
937 static inline int init_function_trace(void) { return 0; }
938 static inline void ftrace_pid_follow_fork(struct trace_array *tr, bool enable) { }
939 /* ftace_func_t type is not defined, use macro instead of static inline */
940 #define ftrace_init_array_ops(tr, func) do { } while (0)
941 #endif /* CONFIG_FUNCTION_TRACER */
942 
943 #if defined(CONFIG_FUNCTION_TRACER) && defined(CONFIG_DYNAMIC_FTRACE)
944 
945 struct ftrace_probe_ops {
946 	void			(*func)(unsigned long ip,
947 					unsigned long parent_ip,
948 					struct trace_array *tr,
949 					struct ftrace_probe_ops *ops,
950 					void *data);
951 	int			(*init)(struct ftrace_probe_ops *ops,
952 					struct trace_array *tr,
953 					unsigned long ip, void *init_data,
954 					void **data);
955 	void			(*free)(struct ftrace_probe_ops *ops,
956 					struct trace_array *tr,
957 					unsigned long ip, void *data);
958 	int			(*print)(struct seq_file *m,
959 					 unsigned long ip,
960 					 struct ftrace_probe_ops *ops,
961 					 void *data);
962 };
963 
964 struct ftrace_func_mapper;
965 typedef int (*ftrace_mapper_func)(void *data);
966 
967 struct ftrace_func_mapper *allocate_ftrace_func_mapper(void);
968 void **ftrace_func_mapper_find_ip(struct ftrace_func_mapper *mapper,
969 					   unsigned long ip);
970 int ftrace_func_mapper_add_ip(struct ftrace_func_mapper *mapper,
971 			       unsigned long ip, void *data);
972 void *ftrace_func_mapper_remove_ip(struct ftrace_func_mapper *mapper,
973 				   unsigned long ip);
974 void free_ftrace_func_mapper(struct ftrace_func_mapper *mapper,
975 			     ftrace_mapper_func free_func);
976 
977 extern int
978 register_ftrace_function_probe(char *glob, struct trace_array *tr,
979 			       struct ftrace_probe_ops *ops, void *data);
980 extern int
981 unregister_ftrace_function_probe_func(char *glob, struct trace_array *tr,
982 				      struct ftrace_probe_ops *ops);
983 
984 int register_ftrace_command(struct ftrace_func_command *cmd);
985 int unregister_ftrace_command(struct ftrace_func_command *cmd);
986 
987 void ftrace_create_filter_files(struct ftrace_ops *ops,
988 				struct dentry *parent);
989 void ftrace_destroy_filter_files(struct ftrace_ops *ops);
990 #else
991 struct ftrace_func_command;
992 
993 static inline __init int register_ftrace_command(struct ftrace_func_command *cmd)
994 {
995 	return -EINVAL;
996 }
997 static inline __init int unregister_ftrace_command(char *cmd_name)
998 {
999 	return -EINVAL;
1000 }
1001 /*
1002  * The ops parameter passed in is usually undefined.
1003  * This must be a macro.
1004  */
1005 #define ftrace_create_filter_files(ops, parent) do { } while (0)
1006 #define ftrace_destroy_filter_files(ops) do { } while (0)
1007 #endif /* CONFIG_FUNCTION_TRACER && CONFIG_DYNAMIC_FTRACE */
1008 
1009 bool ftrace_event_is_function(struct trace_event_call *call);
1010 
1011 /*
1012  * struct trace_parser - servers for reading the user input separated by spaces
1013  * @cont: set if the input is not complete - no final space char was found
1014  * @buffer: holds the parsed user input
1015  * @idx: user input length
1016  * @size: buffer size
1017  */
1018 struct trace_parser {
1019 	bool		cont;
1020 	char		*buffer;
1021 	unsigned	idx;
1022 	unsigned	size;
1023 };
1024 
1025 static inline bool trace_parser_loaded(struct trace_parser *parser)
1026 {
1027 	return (parser->idx != 0);
1028 }
1029 
1030 static inline bool trace_parser_cont(struct trace_parser *parser)
1031 {
1032 	return parser->cont;
1033 }
1034 
1035 static inline void trace_parser_clear(struct trace_parser *parser)
1036 {
1037 	parser->cont = false;
1038 	parser->idx = 0;
1039 }
1040 
1041 extern int trace_parser_get_init(struct trace_parser *parser, int size);
1042 extern void trace_parser_put(struct trace_parser *parser);
1043 extern int trace_get_user(struct trace_parser *parser, const char __user *ubuf,
1044 	size_t cnt, loff_t *ppos);
1045 
1046 /*
1047  * Only create function graph options if function graph is configured.
1048  */
1049 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1050 # define FGRAPH_FLAGS						\
1051 		C(DISPLAY_GRAPH,	"display-graph"),
1052 #else
1053 # define FGRAPH_FLAGS
1054 #endif
1055 
1056 #ifdef CONFIG_BRANCH_TRACER
1057 # define BRANCH_FLAGS					\
1058 		C(BRANCH,		"branch"),
1059 #else
1060 # define BRANCH_FLAGS
1061 #endif
1062 
1063 #ifdef CONFIG_FUNCTION_TRACER
1064 # define FUNCTION_FLAGS						\
1065 		C(FUNCTION,		"function-trace"),	\
1066 		C(FUNC_FORK,		"function-fork"),
1067 # define FUNCTION_DEFAULT_FLAGS		TRACE_ITER_FUNCTION
1068 #else
1069 # define FUNCTION_FLAGS
1070 # define FUNCTION_DEFAULT_FLAGS		0UL
1071 # define TRACE_ITER_FUNC_FORK		0UL
1072 #endif
1073 
1074 #ifdef CONFIG_STACKTRACE
1075 # define STACK_FLAGS				\
1076 		C(STACKTRACE,		"stacktrace"),
1077 #else
1078 # define STACK_FLAGS
1079 #endif
1080 
1081 /*
1082  * trace_iterator_flags is an enumeration that defines bit
1083  * positions into trace_flags that controls the output.
1084  *
1085  * NOTE: These bits must match the trace_options array in
1086  *       trace.c (this macro guarantees it).
1087  */
1088 #define TRACE_FLAGS						\
1089 		C(PRINT_PARENT,		"print-parent"),	\
1090 		C(SYM_OFFSET,		"sym-offset"),		\
1091 		C(SYM_ADDR,		"sym-addr"),		\
1092 		C(VERBOSE,		"verbose"),		\
1093 		C(RAW,			"raw"),			\
1094 		C(HEX,			"hex"),			\
1095 		C(BIN,			"bin"),			\
1096 		C(BLOCK,		"block"),		\
1097 		C(PRINTK,		"trace_printk"),	\
1098 		C(ANNOTATE,		"annotate"),		\
1099 		C(USERSTACKTRACE,	"userstacktrace"),	\
1100 		C(SYM_USEROBJ,		"sym-userobj"),		\
1101 		C(PRINTK_MSGONLY,	"printk-msg-only"),	\
1102 		C(CONTEXT_INFO,		"context-info"),   /* Print pid/cpu/time */ \
1103 		C(LATENCY_FMT,		"latency-format"),	\
1104 		C(RECORD_CMD,		"record-cmd"),		\
1105 		C(OVERWRITE,		"overwrite"),		\
1106 		C(STOP_ON_FREE,		"disable_on_free"),	\
1107 		C(IRQ_INFO,		"irq-info"),		\
1108 		C(MARKERS,		"markers"),		\
1109 		C(EVENT_FORK,		"event-fork"),		\
1110 		FUNCTION_FLAGS					\
1111 		FGRAPH_FLAGS					\
1112 		STACK_FLAGS					\
1113 		BRANCH_FLAGS
1114 
1115 /*
1116  * By defining C, we can make TRACE_FLAGS a list of bit names
1117  * that will define the bits for the flag masks.
1118  */
1119 #undef C
1120 #define C(a, b) TRACE_ITER_##a##_BIT
1121 
1122 enum trace_iterator_bits {
1123 	TRACE_FLAGS
1124 	/* Make sure we don't go more than we have bits for */
1125 	TRACE_ITER_LAST_BIT
1126 };
1127 
1128 /*
1129  * By redefining C, we can make TRACE_FLAGS a list of masks that
1130  * use the bits as defined above.
1131  */
1132 #undef C
1133 #define C(a, b) TRACE_ITER_##a = (1 << TRACE_ITER_##a##_BIT)
1134 
1135 enum trace_iterator_flags { TRACE_FLAGS };
1136 
1137 /*
1138  * TRACE_ITER_SYM_MASK masks the options in trace_flags that
1139  * control the output of kernel symbols.
1140  */
1141 #define TRACE_ITER_SYM_MASK \
1142 	(TRACE_ITER_PRINT_PARENT|TRACE_ITER_SYM_OFFSET|TRACE_ITER_SYM_ADDR)
1143 
1144 extern struct tracer nop_trace;
1145 
1146 #ifdef CONFIG_BRANCH_TRACER
1147 extern int enable_branch_tracing(struct trace_array *tr);
1148 extern void disable_branch_tracing(void);
1149 static inline int trace_branch_enable(struct trace_array *tr)
1150 {
1151 	if (tr->trace_flags & TRACE_ITER_BRANCH)
1152 		return enable_branch_tracing(tr);
1153 	return 0;
1154 }
1155 static inline void trace_branch_disable(void)
1156 {
1157 	/* due to races, always disable */
1158 	disable_branch_tracing();
1159 }
1160 #else
1161 static inline int trace_branch_enable(struct trace_array *tr)
1162 {
1163 	return 0;
1164 }
1165 static inline void trace_branch_disable(void)
1166 {
1167 }
1168 #endif /* CONFIG_BRANCH_TRACER */
1169 
1170 /* set ring buffers to default size if not already done so */
1171 int tracing_update_buffers(void);
1172 
1173 struct ftrace_event_field {
1174 	struct list_head	link;
1175 	const char		*name;
1176 	const char		*type;
1177 	int			filter_type;
1178 	int			offset;
1179 	int			size;
1180 	int			is_signed;
1181 };
1182 
1183 struct event_filter {
1184 	int			n_preds;	/* Number assigned */
1185 	int			a_preds;	/* allocated */
1186 	struct filter_pred	*preds;
1187 	struct filter_pred	*root;
1188 	char			*filter_string;
1189 };
1190 
1191 struct event_subsystem {
1192 	struct list_head	list;
1193 	const char		*name;
1194 	struct event_filter	*filter;
1195 	int			ref_count;
1196 };
1197 
1198 struct trace_subsystem_dir {
1199 	struct list_head		list;
1200 	struct event_subsystem		*subsystem;
1201 	struct trace_array		*tr;
1202 	struct dentry			*entry;
1203 	int				ref_count;
1204 	int				nr_events;
1205 };
1206 
1207 extern int call_filter_check_discard(struct trace_event_call *call, void *rec,
1208 				     struct ring_buffer *buffer,
1209 				     struct ring_buffer_event *event);
1210 
1211 void trace_buffer_unlock_commit_regs(struct trace_array *tr,
1212 				     struct ring_buffer *buffer,
1213 				     struct ring_buffer_event *event,
1214 				     unsigned long flags, int pc,
1215 				     struct pt_regs *regs);
1216 
1217 static inline void trace_buffer_unlock_commit(struct trace_array *tr,
1218 					      struct ring_buffer *buffer,
1219 					      struct ring_buffer_event *event,
1220 					      unsigned long flags, int pc)
1221 {
1222 	trace_buffer_unlock_commit_regs(tr, buffer, event, flags, pc, NULL);
1223 }
1224 
1225 DECLARE_PER_CPU(struct ring_buffer_event *, trace_buffered_event);
1226 DECLARE_PER_CPU(int, trace_buffered_event_cnt);
1227 void trace_buffered_event_disable(void);
1228 void trace_buffered_event_enable(void);
1229 
1230 static inline void
1231 __trace_event_discard_commit(struct ring_buffer *buffer,
1232 			     struct ring_buffer_event *event)
1233 {
1234 	if (this_cpu_read(trace_buffered_event) == event) {
1235 		/* Simply release the temp buffer */
1236 		this_cpu_dec(trace_buffered_event_cnt);
1237 		return;
1238 	}
1239 	ring_buffer_discard_commit(buffer, event);
1240 }
1241 
1242 /*
1243  * Helper function for event_trigger_unlock_commit{_regs}().
1244  * If there are event triggers attached to this event that requires
1245  * filtering against its fields, then they wil be called as the
1246  * entry already holds the field information of the current event.
1247  *
1248  * It also checks if the event should be discarded or not.
1249  * It is to be discarded if the event is soft disabled and the
1250  * event was only recorded to process triggers, or if the event
1251  * filter is active and this event did not match the filters.
1252  *
1253  * Returns true if the event is discarded, false otherwise.
1254  */
1255 static inline bool
1256 __event_trigger_test_discard(struct trace_event_file *file,
1257 			     struct ring_buffer *buffer,
1258 			     struct ring_buffer_event *event,
1259 			     void *entry,
1260 			     enum event_trigger_type *tt)
1261 {
1262 	unsigned long eflags = file->flags;
1263 
1264 	if (eflags & EVENT_FILE_FL_TRIGGER_COND)
1265 		*tt = event_triggers_call(file, entry);
1266 
1267 	if (test_bit(EVENT_FILE_FL_SOFT_DISABLED_BIT, &file->flags) ||
1268 	    (unlikely(file->flags & EVENT_FILE_FL_FILTERED) &&
1269 	     !filter_match_preds(file->filter, entry))) {
1270 		__trace_event_discard_commit(buffer, event);
1271 		return true;
1272 	}
1273 
1274 	return false;
1275 }
1276 
1277 /**
1278  * event_trigger_unlock_commit - handle triggers and finish event commit
1279  * @file: The file pointer assoctiated to the event
1280  * @buffer: The ring buffer that the event is being written to
1281  * @event: The event meta data in the ring buffer
1282  * @entry: The event itself
1283  * @irq_flags: The state of the interrupts at the start of the event
1284  * @pc: The state of the preempt count at the start of the event.
1285  *
1286  * This is a helper function to handle triggers that require data
1287  * from the event itself. It also tests the event against filters and
1288  * if the event is soft disabled and should be discarded.
1289  */
1290 static inline void
1291 event_trigger_unlock_commit(struct trace_event_file *file,
1292 			    struct ring_buffer *buffer,
1293 			    struct ring_buffer_event *event,
1294 			    void *entry, unsigned long irq_flags, int pc)
1295 {
1296 	enum event_trigger_type tt = ETT_NONE;
1297 
1298 	if (!__event_trigger_test_discard(file, buffer, event, entry, &tt))
1299 		trace_buffer_unlock_commit(file->tr, buffer, event, irq_flags, pc);
1300 
1301 	if (tt)
1302 		event_triggers_post_call(file, tt, entry);
1303 }
1304 
1305 /**
1306  * event_trigger_unlock_commit_regs - handle triggers and finish event commit
1307  * @file: The file pointer assoctiated to the event
1308  * @buffer: The ring buffer that the event is being written to
1309  * @event: The event meta data in the ring buffer
1310  * @entry: The event itself
1311  * @irq_flags: The state of the interrupts at the start of the event
1312  * @pc: The state of the preempt count at the start of the event.
1313  *
1314  * This is a helper function to handle triggers that require data
1315  * from the event itself. It also tests the event against filters and
1316  * if the event is soft disabled and should be discarded.
1317  *
1318  * Same as event_trigger_unlock_commit() but calls
1319  * trace_buffer_unlock_commit_regs() instead of trace_buffer_unlock_commit().
1320  */
1321 static inline void
1322 event_trigger_unlock_commit_regs(struct trace_event_file *file,
1323 				 struct ring_buffer *buffer,
1324 				 struct ring_buffer_event *event,
1325 				 void *entry, unsigned long irq_flags, int pc,
1326 				 struct pt_regs *regs)
1327 {
1328 	enum event_trigger_type tt = ETT_NONE;
1329 
1330 	if (!__event_trigger_test_discard(file, buffer, event, entry, &tt))
1331 		trace_buffer_unlock_commit_regs(file->tr, buffer, event,
1332 						irq_flags, pc, regs);
1333 
1334 	if (tt)
1335 		event_triggers_post_call(file, tt, entry);
1336 }
1337 
1338 #define FILTER_PRED_INVALID	((unsigned short)-1)
1339 #define FILTER_PRED_IS_RIGHT	(1 << 15)
1340 #define FILTER_PRED_FOLD	(1 << 15)
1341 
1342 /*
1343  * The max preds is the size of unsigned short with
1344  * two flags at the MSBs. One bit is used for both the IS_RIGHT
1345  * and FOLD flags. The other is reserved.
1346  *
1347  * 2^14 preds is way more than enough.
1348  */
1349 #define MAX_FILTER_PRED		16384
1350 
1351 struct filter_pred;
1352 struct regex;
1353 
1354 typedef int (*filter_pred_fn_t) (struct filter_pred *pred, void *event);
1355 
1356 typedef int (*regex_match_func)(char *str, struct regex *r, int len);
1357 
1358 enum regex_type {
1359 	MATCH_FULL = 0,
1360 	MATCH_FRONT_ONLY,
1361 	MATCH_MIDDLE_ONLY,
1362 	MATCH_END_ONLY,
1363 	MATCH_GLOB,
1364 };
1365 
1366 struct regex {
1367 	char			pattern[MAX_FILTER_STR_VAL];
1368 	int			len;
1369 	int			field_len;
1370 	regex_match_func	match;
1371 };
1372 
1373 struct filter_pred {
1374 	filter_pred_fn_t 	fn;
1375 	u64 			val;
1376 	struct regex		regex;
1377 	unsigned short		*ops;
1378 	struct ftrace_event_field *field;
1379 	int 			offset;
1380 	int 			not;
1381 	int 			op;
1382 	unsigned short		index;
1383 	unsigned short		parent;
1384 	unsigned short		left;
1385 	unsigned short		right;
1386 };
1387 
1388 static inline bool is_string_field(struct ftrace_event_field *field)
1389 {
1390 	return field->filter_type == FILTER_DYN_STRING ||
1391 	       field->filter_type == FILTER_STATIC_STRING ||
1392 	       field->filter_type == FILTER_PTR_STRING ||
1393 	       field->filter_type == FILTER_COMM;
1394 }
1395 
1396 static inline bool is_function_field(struct ftrace_event_field *field)
1397 {
1398 	return field->filter_type == FILTER_TRACE_FN;
1399 }
1400 
1401 extern enum regex_type
1402 filter_parse_regex(char *buff, int len, char **search, int *not);
1403 extern void print_event_filter(struct trace_event_file *file,
1404 			       struct trace_seq *s);
1405 extern int apply_event_filter(struct trace_event_file *file,
1406 			      char *filter_string);
1407 extern int apply_subsystem_event_filter(struct trace_subsystem_dir *dir,
1408 					char *filter_string);
1409 extern void print_subsystem_event_filter(struct event_subsystem *system,
1410 					 struct trace_seq *s);
1411 extern int filter_assign_type(const char *type);
1412 extern int create_event_filter(struct trace_event_call *call,
1413 			       char *filter_str, bool set_str,
1414 			       struct event_filter **filterp);
1415 extern void free_event_filter(struct event_filter *filter);
1416 
1417 struct ftrace_event_field *
1418 trace_find_event_field(struct trace_event_call *call, char *name);
1419 
1420 extern void trace_event_enable_cmd_record(bool enable);
1421 extern int event_trace_add_tracer(struct dentry *parent, struct trace_array *tr);
1422 extern int event_trace_del_tracer(struct trace_array *tr);
1423 
1424 extern struct trace_event_file *find_event_file(struct trace_array *tr,
1425 						const char *system,
1426 						const char *event);
1427 
1428 static inline void *event_file_data(struct file *filp)
1429 {
1430 	return ACCESS_ONCE(file_inode(filp)->i_private);
1431 }
1432 
1433 extern struct mutex event_mutex;
1434 extern struct list_head ftrace_events;
1435 
1436 extern const struct file_operations event_trigger_fops;
1437 extern const struct file_operations event_hist_fops;
1438 
1439 #ifdef CONFIG_HIST_TRIGGERS
1440 extern int register_trigger_hist_cmd(void);
1441 extern int register_trigger_hist_enable_disable_cmds(void);
1442 #else
1443 static inline int register_trigger_hist_cmd(void) { return 0; }
1444 static inline int register_trigger_hist_enable_disable_cmds(void) { return 0; }
1445 #endif
1446 
1447 extern int register_trigger_cmds(void);
1448 extern void clear_event_triggers(struct trace_array *tr);
1449 
1450 struct event_trigger_data {
1451 	unsigned long			count;
1452 	int				ref;
1453 	struct event_trigger_ops	*ops;
1454 	struct event_command		*cmd_ops;
1455 	struct event_filter __rcu	*filter;
1456 	char				*filter_str;
1457 	void				*private_data;
1458 	bool				paused;
1459 	bool				paused_tmp;
1460 	struct list_head		list;
1461 	char				*name;
1462 	struct list_head		named_list;
1463 	struct event_trigger_data	*named_data;
1464 };
1465 
1466 /* Avoid typos */
1467 #define ENABLE_EVENT_STR	"enable_event"
1468 #define DISABLE_EVENT_STR	"disable_event"
1469 #define ENABLE_HIST_STR		"enable_hist"
1470 #define DISABLE_HIST_STR	"disable_hist"
1471 
1472 struct enable_trigger_data {
1473 	struct trace_event_file		*file;
1474 	bool				enable;
1475 	bool				hist;
1476 };
1477 
1478 extern int event_enable_trigger_print(struct seq_file *m,
1479 				      struct event_trigger_ops *ops,
1480 				      struct event_trigger_data *data);
1481 extern void event_enable_trigger_free(struct event_trigger_ops *ops,
1482 				      struct event_trigger_data *data);
1483 extern int event_enable_trigger_func(struct event_command *cmd_ops,
1484 				     struct trace_event_file *file,
1485 				     char *glob, char *cmd, char *param);
1486 extern int event_enable_register_trigger(char *glob,
1487 					 struct event_trigger_ops *ops,
1488 					 struct event_trigger_data *data,
1489 					 struct trace_event_file *file);
1490 extern void event_enable_unregister_trigger(char *glob,
1491 					    struct event_trigger_ops *ops,
1492 					    struct event_trigger_data *test,
1493 					    struct trace_event_file *file);
1494 extern void trigger_data_free(struct event_trigger_data *data);
1495 extern int event_trigger_init(struct event_trigger_ops *ops,
1496 			      struct event_trigger_data *data);
1497 extern int trace_event_trigger_enable_disable(struct trace_event_file *file,
1498 					      int trigger_enable);
1499 extern void update_cond_flag(struct trace_event_file *file);
1500 extern void unregister_trigger(char *glob, struct event_trigger_ops *ops,
1501 			       struct event_trigger_data *test,
1502 			       struct trace_event_file *file);
1503 extern int set_trigger_filter(char *filter_str,
1504 			      struct event_trigger_data *trigger_data,
1505 			      struct trace_event_file *file);
1506 extern struct event_trigger_data *find_named_trigger(const char *name);
1507 extern bool is_named_trigger(struct event_trigger_data *test);
1508 extern int save_named_trigger(const char *name,
1509 			      struct event_trigger_data *data);
1510 extern void del_named_trigger(struct event_trigger_data *data);
1511 extern void pause_named_trigger(struct event_trigger_data *data);
1512 extern void unpause_named_trigger(struct event_trigger_data *data);
1513 extern void set_named_trigger_data(struct event_trigger_data *data,
1514 				   struct event_trigger_data *named_data);
1515 extern int register_event_command(struct event_command *cmd);
1516 extern int unregister_event_command(struct event_command *cmd);
1517 extern int register_trigger_hist_enable_disable_cmds(void);
1518 
1519 /**
1520  * struct event_trigger_ops - callbacks for trace event triggers
1521  *
1522  * The methods in this structure provide per-event trigger hooks for
1523  * various trigger operations.
1524  *
1525  * All the methods below, except for @init() and @free(), must be
1526  * implemented.
1527  *
1528  * @func: The trigger 'probe' function called when the triggering
1529  *	event occurs.  The data passed into this callback is the data
1530  *	that was supplied to the event_command @reg() function that
1531  *	registered the trigger (see struct event_command) along with
1532  *	the trace record, rec.
1533  *
1534  * @init: An optional initialization function called for the trigger
1535  *	when the trigger is registered (via the event_command reg()
1536  *	function).  This can be used to perform per-trigger
1537  *	initialization such as incrementing a per-trigger reference
1538  *	count, for instance.  This is usually implemented by the
1539  *	generic utility function @event_trigger_init() (see
1540  *	trace_event_triggers.c).
1541  *
1542  * @free: An optional de-initialization function called for the
1543  *	trigger when the trigger is unregistered (via the
1544  *	event_command @reg() function).  This can be used to perform
1545  *	per-trigger de-initialization such as decrementing a
1546  *	per-trigger reference count and freeing corresponding trigger
1547  *	data, for instance.  This is usually implemented by the
1548  *	generic utility function @event_trigger_free() (see
1549  *	trace_event_triggers.c).
1550  *
1551  * @print: The callback function invoked to have the trigger print
1552  *	itself.  This is usually implemented by a wrapper function
1553  *	that calls the generic utility function @event_trigger_print()
1554  *	(see trace_event_triggers.c).
1555  */
1556 struct event_trigger_ops {
1557 	void			(*func)(struct event_trigger_data *data,
1558 					void *rec);
1559 	int			(*init)(struct event_trigger_ops *ops,
1560 					struct event_trigger_data *data);
1561 	void			(*free)(struct event_trigger_ops *ops,
1562 					struct event_trigger_data *data);
1563 	int			(*print)(struct seq_file *m,
1564 					 struct event_trigger_ops *ops,
1565 					 struct event_trigger_data *data);
1566 };
1567 
1568 /**
1569  * struct event_command - callbacks and data members for event commands
1570  *
1571  * Event commands are invoked by users by writing the command name
1572  * into the 'trigger' file associated with a trace event.  The
1573  * parameters associated with a specific invocation of an event
1574  * command are used to create an event trigger instance, which is
1575  * added to the list of trigger instances associated with that trace
1576  * event.  When the event is hit, the set of triggers associated with
1577  * that event is invoked.
1578  *
1579  * The data members in this structure provide per-event command data
1580  * for various event commands.
1581  *
1582  * All the data members below, except for @post_trigger, must be set
1583  * for each event command.
1584  *
1585  * @name: The unique name that identifies the event command.  This is
1586  *	the name used when setting triggers via trigger files.
1587  *
1588  * @trigger_type: A unique id that identifies the event command
1589  *	'type'.  This value has two purposes, the first to ensure that
1590  *	only one trigger of the same type can be set at a given time
1591  *	for a particular event e.g. it doesn't make sense to have both
1592  *	a traceon and traceoff trigger attached to a single event at
1593  *	the same time, so traceon and traceoff have the same type
1594  *	though they have different names.  The @trigger_type value is
1595  *	also used as a bit value for deferring the actual trigger
1596  *	action until after the current event is finished.  Some
1597  *	commands need to do this if they themselves log to the trace
1598  *	buffer (see the @post_trigger() member below).  @trigger_type
1599  *	values are defined by adding new values to the trigger_type
1600  *	enum in include/linux/trace_events.h.
1601  *
1602  * @flags: See the enum event_command_flags below.
1603  *
1604  * All the methods below, except for @set_filter() and @unreg_all(),
1605  * must be implemented.
1606  *
1607  * @func: The callback function responsible for parsing and
1608  *	registering the trigger written to the 'trigger' file by the
1609  *	user.  It allocates the trigger instance and registers it with
1610  *	the appropriate trace event.  It makes use of the other
1611  *	event_command callback functions to orchestrate this, and is
1612  *	usually implemented by the generic utility function
1613  *	@event_trigger_callback() (see trace_event_triggers.c).
1614  *
1615  * @reg: Adds the trigger to the list of triggers associated with the
1616  *	event, and enables the event trigger itself, after
1617  *	initializing it (via the event_trigger_ops @init() function).
1618  *	This is also where commands can use the @trigger_type value to
1619  *	make the decision as to whether or not multiple instances of
1620  *	the trigger should be allowed.  This is usually implemented by
1621  *	the generic utility function @register_trigger() (see
1622  *	trace_event_triggers.c).
1623  *
1624  * @unreg: Removes the trigger from the list of triggers associated
1625  *	with the event, and disables the event trigger itself, after
1626  *	initializing it (via the event_trigger_ops @free() function).
1627  *	This is usually implemented by the generic utility function
1628  *	@unregister_trigger() (see trace_event_triggers.c).
1629  *
1630  * @unreg_all: An optional function called to remove all the triggers
1631  *	from the list of triggers associated with the event.  Called
1632  *	when a trigger file is opened in truncate mode.
1633  *
1634  * @set_filter: An optional function called to parse and set a filter
1635  *	for the trigger.  If no @set_filter() method is set for the
1636  *	event command, filters set by the user for the command will be
1637  *	ignored.  This is usually implemented by the generic utility
1638  *	function @set_trigger_filter() (see trace_event_triggers.c).
1639  *
1640  * @get_trigger_ops: The callback function invoked to retrieve the
1641  *	event_trigger_ops implementation associated with the command.
1642  */
1643 struct event_command {
1644 	struct list_head	list;
1645 	char			*name;
1646 	enum event_trigger_type	trigger_type;
1647 	int			flags;
1648 	int			(*func)(struct event_command *cmd_ops,
1649 					struct trace_event_file *file,
1650 					char *glob, char *cmd, char *params);
1651 	int			(*reg)(char *glob,
1652 				       struct event_trigger_ops *ops,
1653 				       struct event_trigger_data *data,
1654 				       struct trace_event_file *file);
1655 	void			(*unreg)(char *glob,
1656 					 struct event_trigger_ops *ops,
1657 					 struct event_trigger_data *data,
1658 					 struct trace_event_file *file);
1659 	void			(*unreg_all)(struct trace_event_file *file);
1660 	int			(*set_filter)(char *filter_str,
1661 					      struct event_trigger_data *data,
1662 					      struct trace_event_file *file);
1663 	struct event_trigger_ops *(*get_trigger_ops)(char *cmd, char *param);
1664 };
1665 
1666 /**
1667  * enum event_command_flags - flags for struct event_command
1668  *
1669  * @POST_TRIGGER: A flag that says whether or not this command needs
1670  *	to have its action delayed until after the current event has
1671  *	been closed.  Some triggers need to avoid being invoked while
1672  *	an event is currently in the process of being logged, since
1673  *	the trigger may itself log data into the trace buffer.  Thus
1674  *	we make sure the current event is committed before invoking
1675  *	those triggers.  To do that, the trigger invocation is split
1676  *	in two - the first part checks the filter using the current
1677  *	trace record; if a command has the @post_trigger flag set, it
1678  *	sets a bit for itself in the return value, otherwise it
1679  *	directly invokes the trigger.  Once all commands have been
1680  *	either invoked or set their return flag, the current record is
1681  *	either committed or discarded.  At that point, if any commands
1682  *	have deferred their triggers, those commands are finally
1683  *	invoked following the close of the current event.  In other
1684  *	words, if the event_trigger_ops @func() probe implementation
1685  *	itself logs to the trace buffer, this flag should be set,
1686  *	otherwise it can be left unspecified.
1687  *
1688  * @NEEDS_REC: A flag that says whether or not this command needs
1689  *	access to the trace record in order to perform its function,
1690  *	regardless of whether or not it has a filter associated with
1691  *	it (filters make a trigger require access to the trace record
1692  *	but are not always present).
1693  */
1694 enum event_command_flags {
1695 	EVENT_CMD_FL_POST_TRIGGER	= 1,
1696 	EVENT_CMD_FL_NEEDS_REC		= 2,
1697 };
1698 
1699 static inline bool event_command_post_trigger(struct event_command *cmd_ops)
1700 {
1701 	return cmd_ops->flags & EVENT_CMD_FL_POST_TRIGGER;
1702 }
1703 
1704 static inline bool event_command_needs_rec(struct event_command *cmd_ops)
1705 {
1706 	return cmd_ops->flags & EVENT_CMD_FL_NEEDS_REC;
1707 }
1708 
1709 extern int trace_event_enable_disable(struct trace_event_file *file,
1710 				      int enable, int soft_disable);
1711 extern int tracing_alloc_snapshot(void);
1712 
1713 extern const char *__start___trace_bprintk_fmt[];
1714 extern const char *__stop___trace_bprintk_fmt[];
1715 
1716 extern const char *__start___tracepoint_str[];
1717 extern const char *__stop___tracepoint_str[];
1718 
1719 void trace_printk_control(bool enabled);
1720 void trace_printk_init_buffers(void);
1721 void trace_printk_start_comm(void);
1722 int trace_keep_overwrite(struct tracer *tracer, u32 mask, int set);
1723 int set_tracer_flag(struct trace_array *tr, unsigned int mask, int enabled);
1724 
1725 /*
1726  * Normal trace_printk() and friends allocates special buffers
1727  * to do the manipulation, as well as saves the print formats
1728  * into sections to display. But the trace infrastructure wants
1729  * to use these without the added overhead at the price of being
1730  * a bit slower (used mainly for warnings, where we don't care
1731  * about performance). The internal_trace_puts() is for such
1732  * a purpose.
1733  */
1734 #define internal_trace_puts(str) __trace_puts(_THIS_IP_, str, strlen(str))
1735 
1736 #undef FTRACE_ENTRY
1737 #define FTRACE_ENTRY(call, struct_name, id, tstruct, print, filter)	\
1738 	extern struct trace_event_call					\
1739 	__aligned(4) event_##call;
1740 #undef FTRACE_ENTRY_DUP
1741 #define FTRACE_ENTRY_DUP(call, struct_name, id, tstruct, print, filter)	\
1742 	FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print), \
1743 		     filter)
1744 #undef FTRACE_ENTRY_PACKED
1745 #define FTRACE_ENTRY_PACKED(call, struct_name, id, tstruct, print, filter) \
1746 	FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print), \
1747 		     filter)
1748 
1749 #include "trace_entries.h"
1750 
1751 #if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_FUNCTION_TRACER)
1752 int perf_ftrace_event_register(struct trace_event_call *call,
1753 			       enum trace_reg type, void *data);
1754 #else
1755 #define perf_ftrace_event_register NULL
1756 #endif
1757 
1758 #ifdef CONFIG_FTRACE_SYSCALLS
1759 void init_ftrace_syscalls(void);
1760 const char *get_syscall_name(int syscall);
1761 #else
1762 static inline void init_ftrace_syscalls(void) { }
1763 static inline const char *get_syscall_name(int syscall)
1764 {
1765 	return NULL;
1766 }
1767 #endif
1768 
1769 #ifdef CONFIG_EVENT_TRACING
1770 void trace_event_init(void);
1771 void trace_event_enum_update(struct trace_enum_map **map, int len);
1772 #else
1773 static inline void __init trace_event_init(void) { }
1774 static inline void trace_event_enum_update(struct trace_enum_map **map, int len) { }
1775 #endif
1776 
1777 extern struct trace_iterator *tracepoint_print_iter;
1778 
1779 #endif /* _LINUX_KERNEL_TRACE_H */
1780