xref: /linux/kernel/trace/trace.h (revision 2fe05e1139a555ae91f00a812cb9520e7d3022ab)
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 	/* All of these are protected by the ftrace_lock */
267 	struct list_head	func_probes;
268 	struct list_head	mod_trace;
269 	struct list_head	mod_notrace;
270 #endif
271 	/* function tracing enabled */
272 	int			function_enabled;
273 #endif
274 };
275 
276 enum {
277 	TRACE_ARRAY_FL_GLOBAL	= (1 << 0)
278 };
279 
280 extern struct list_head ftrace_trace_arrays;
281 
282 extern struct mutex trace_types_lock;
283 
284 extern int trace_array_get(struct trace_array *tr);
285 extern void trace_array_put(struct trace_array *tr);
286 
287 /*
288  * The global tracer (top) should be the first trace array added,
289  * but we check the flag anyway.
290  */
291 static inline struct trace_array *top_trace_array(void)
292 {
293 	struct trace_array *tr;
294 
295 	if (list_empty(&ftrace_trace_arrays))
296 		return NULL;
297 
298 	tr = list_entry(ftrace_trace_arrays.prev,
299 			typeof(*tr), list);
300 	WARN_ON(!(tr->flags & TRACE_ARRAY_FL_GLOBAL));
301 	return tr;
302 }
303 
304 #define FTRACE_CMP_TYPE(var, type) \
305 	__builtin_types_compatible_p(typeof(var), type *)
306 
307 #undef IF_ASSIGN
308 #define IF_ASSIGN(var, entry, etype, id)		\
309 	if (FTRACE_CMP_TYPE(var, etype)) {		\
310 		var = (typeof(var))(entry);		\
311 		WARN_ON(id && (entry)->type != id);	\
312 		break;					\
313 	}
314 
315 /* Will cause compile errors if type is not found. */
316 extern void __ftrace_bad_type(void);
317 
318 /*
319  * The trace_assign_type is a verifier that the entry type is
320  * the same as the type being assigned. To add new types simply
321  * add a line with the following format:
322  *
323  * IF_ASSIGN(var, ent, type, id);
324  *
325  *  Where "type" is the trace type that includes the trace_entry
326  *  as the "ent" item. And "id" is the trace identifier that is
327  *  used in the trace_type enum.
328  *
329  *  If the type can have more than one id, then use zero.
330  */
331 #define trace_assign_type(var, ent)					\
332 	do {								\
333 		IF_ASSIGN(var, ent, struct ftrace_entry, TRACE_FN);	\
334 		IF_ASSIGN(var, ent, struct ctx_switch_entry, 0);	\
335 		IF_ASSIGN(var, ent, struct stack_entry, TRACE_STACK);	\
336 		IF_ASSIGN(var, ent, struct userstack_entry, TRACE_USER_STACK);\
337 		IF_ASSIGN(var, ent, struct print_entry, TRACE_PRINT);	\
338 		IF_ASSIGN(var, ent, struct bprint_entry, TRACE_BPRINT);	\
339 		IF_ASSIGN(var, ent, struct bputs_entry, TRACE_BPUTS);	\
340 		IF_ASSIGN(var, ent, struct hwlat_entry, TRACE_HWLAT);	\
341 		IF_ASSIGN(var, ent, struct raw_data_entry, TRACE_RAW_DATA);\
342 		IF_ASSIGN(var, ent, struct trace_mmiotrace_rw,		\
343 			  TRACE_MMIO_RW);				\
344 		IF_ASSIGN(var, ent, struct trace_mmiotrace_map,		\
345 			  TRACE_MMIO_MAP);				\
346 		IF_ASSIGN(var, ent, struct trace_branch, TRACE_BRANCH); \
347 		IF_ASSIGN(var, ent, struct ftrace_graph_ent_entry,	\
348 			  TRACE_GRAPH_ENT);		\
349 		IF_ASSIGN(var, ent, struct ftrace_graph_ret_entry,	\
350 			  TRACE_GRAPH_RET);		\
351 		__ftrace_bad_type();					\
352 	} while (0)
353 
354 /*
355  * An option specific to a tracer. This is a boolean value.
356  * The bit is the bit index that sets its value on the
357  * flags value in struct tracer_flags.
358  */
359 struct tracer_opt {
360 	const char	*name; /* Will appear on the trace_options file */
361 	u32		bit; /* Mask assigned in val field in tracer_flags */
362 };
363 
364 /*
365  * The set of specific options for a tracer. Your tracer
366  * have to set the initial value of the flags val.
367  */
368 struct tracer_flags {
369 	u32			val;
370 	struct tracer_opt	*opts;
371 	struct tracer		*trace;
372 };
373 
374 /* Makes more easy to define a tracer opt */
375 #define TRACER_OPT(s, b)	.name = #s, .bit = b
376 
377 
378 struct trace_option_dentry {
379 	struct tracer_opt		*opt;
380 	struct tracer_flags		*flags;
381 	struct trace_array		*tr;
382 	struct dentry			*entry;
383 };
384 
385 /**
386  * struct tracer - a specific tracer and its callbacks to interact with tracefs
387  * @name: the name chosen to select it on the available_tracers file
388  * @init: called when one switches to this tracer (echo name > current_tracer)
389  * @reset: called when one switches to another tracer
390  * @start: called when tracing is unpaused (echo 1 > tracing_on)
391  * @stop: called when tracing is paused (echo 0 > tracing_on)
392  * @update_thresh: called when tracing_thresh is updated
393  * @open: called when the trace file is opened
394  * @pipe_open: called when the trace_pipe file is opened
395  * @close: called when the trace file is released
396  * @pipe_close: called when the trace_pipe file is released
397  * @read: override the default read callback on trace_pipe
398  * @splice_read: override the default splice_read callback on trace_pipe
399  * @selftest: selftest to run on boot (see trace_selftest.c)
400  * @print_headers: override the first lines that describe your columns
401  * @print_line: callback that prints a trace
402  * @set_flag: signals one of your private flags changed (trace_options file)
403  * @flags: your private flags
404  */
405 struct tracer {
406 	const char		*name;
407 	int			(*init)(struct trace_array *tr);
408 	void			(*reset)(struct trace_array *tr);
409 	void			(*start)(struct trace_array *tr);
410 	void			(*stop)(struct trace_array *tr);
411 	int			(*update_thresh)(struct trace_array *tr);
412 	void			(*open)(struct trace_iterator *iter);
413 	void			(*pipe_open)(struct trace_iterator *iter);
414 	void			(*close)(struct trace_iterator *iter);
415 	void			(*pipe_close)(struct trace_iterator *iter);
416 	ssize_t			(*read)(struct trace_iterator *iter,
417 					struct file *filp, char __user *ubuf,
418 					size_t cnt, loff_t *ppos);
419 	ssize_t			(*splice_read)(struct trace_iterator *iter,
420 					       struct file *filp,
421 					       loff_t *ppos,
422 					       struct pipe_inode_info *pipe,
423 					       size_t len,
424 					       unsigned int flags);
425 #ifdef CONFIG_FTRACE_STARTUP_TEST
426 	int			(*selftest)(struct tracer *trace,
427 					    struct trace_array *tr);
428 #endif
429 	void			(*print_header)(struct seq_file *m);
430 	enum print_line_t	(*print_line)(struct trace_iterator *iter);
431 	/* If you handled the flag setting, return 0 */
432 	int			(*set_flag)(struct trace_array *tr,
433 					    u32 old_flags, u32 bit, int set);
434 	/* Return 0 if OK with change, else return non-zero */
435 	int			(*flag_changed)(struct trace_array *tr,
436 						u32 mask, int set);
437 	struct tracer		*next;
438 	struct tracer_flags	*flags;
439 	int			enabled;
440 	int			ref;
441 	bool			print_max;
442 	bool			allow_instances;
443 #ifdef CONFIG_TRACER_MAX_TRACE
444 	bool			use_max_tr;
445 #endif
446 };
447 
448 
449 /* Only current can touch trace_recursion */
450 
451 /*
452  * For function tracing recursion:
453  *  The order of these bits are important.
454  *
455  *  When function tracing occurs, the following steps are made:
456  *   If arch does not support a ftrace feature:
457  *    call internal function (uses INTERNAL bits) which calls...
458  *   If callback is registered to the "global" list, the list
459  *    function is called and recursion checks the GLOBAL bits.
460  *    then this function calls...
461  *   The function callback, which can use the FTRACE bits to
462  *    check for recursion.
463  *
464  * Now if the arch does not suppport a feature, and it calls
465  * the global list function which calls the ftrace callback
466  * all three of these steps will do a recursion protection.
467  * There's no reason to do one if the previous caller already
468  * did. The recursion that we are protecting against will
469  * go through the same steps again.
470  *
471  * To prevent the multiple recursion checks, if a recursion
472  * bit is set that is higher than the MAX bit of the current
473  * check, then we know that the check was made by the previous
474  * caller, and we can skip the current check.
475  */
476 enum {
477 	TRACE_BUFFER_BIT,
478 	TRACE_BUFFER_NMI_BIT,
479 	TRACE_BUFFER_IRQ_BIT,
480 	TRACE_BUFFER_SIRQ_BIT,
481 
482 	/* Start of function recursion bits */
483 	TRACE_FTRACE_BIT,
484 	TRACE_FTRACE_NMI_BIT,
485 	TRACE_FTRACE_IRQ_BIT,
486 	TRACE_FTRACE_SIRQ_BIT,
487 
488 	/* INTERNAL_BITs must be greater than FTRACE_BITs */
489 	TRACE_INTERNAL_BIT,
490 	TRACE_INTERNAL_NMI_BIT,
491 	TRACE_INTERNAL_IRQ_BIT,
492 	TRACE_INTERNAL_SIRQ_BIT,
493 
494 	TRACE_BRANCH_BIT,
495 /*
496  * Abuse of the trace_recursion.
497  * As we need a way to maintain state if we are tracing the function
498  * graph in irq because we want to trace a particular function that
499  * was called in irq context but we have irq tracing off. Since this
500  * can only be modified by current, we can reuse trace_recursion.
501  */
502 	TRACE_IRQ_BIT,
503 };
504 
505 #define trace_recursion_set(bit)	do { (current)->trace_recursion |= (1<<(bit)); } while (0)
506 #define trace_recursion_clear(bit)	do { (current)->trace_recursion &= ~(1<<(bit)); } while (0)
507 #define trace_recursion_test(bit)	((current)->trace_recursion & (1<<(bit)))
508 
509 #define TRACE_CONTEXT_BITS	4
510 
511 #define TRACE_FTRACE_START	TRACE_FTRACE_BIT
512 #define TRACE_FTRACE_MAX	((1 << (TRACE_FTRACE_START + TRACE_CONTEXT_BITS)) - 1)
513 
514 #define TRACE_LIST_START	TRACE_INTERNAL_BIT
515 #define TRACE_LIST_MAX		((1 << (TRACE_LIST_START + TRACE_CONTEXT_BITS)) - 1)
516 
517 #define TRACE_CONTEXT_MASK	TRACE_LIST_MAX
518 
519 static __always_inline int trace_get_context_bit(void)
520 {
521 	int bit;
522 
523 	if (in_interrupt()) {
524 		if (in_nmi())
525 			bit = 0;
526 
527 		else if (in_irq())
528 			bit = 1;
529 		else
530 			bit = 2;
531 	} else
532 		bit = 3;
533 
534 	return bit;
535 }
536 
537 static __always_inline int trace_test_and_set_recursion(int start, int max)
538 {
539 	unsigned int val = current->trace_recursion;
540 	int bit;
541 
542 	/* A previous recursion check was made */
543 	if ((val & TRACE_CONTEXT_MASK) > max)
544 		return 0;
545 
546 	bit = trace_get_context_bit() + start;
547 	if (unlikely(val & (1 << bit)))
548 		return -1;
549 
550 	val |= 1 << bit;
551 	current->trace_recursion = val;
552 	barrier();
553 
554 	return bit;
555 }
556 
557 static __always_inline void trace_clear_recursion(int bit)
558 {
559 	unsigned int val = current->trace_recursion;
560 
561 	if (!bit)
562 		return;
563 
564 	bit = 1 << bit;
565 	val &= ~bit;
566 
567 	barrier();
568 	current->trace_recursion = val;
569 }
570 
571 static inline struct ring_buffer_iter *
572 trace_buffer_iter(struct trace_iterator *iter, int cpu)
573 {
574 	if (iter->buffer_iter && iter->buffer_iter[cpu])
575 		return iter->buffer_iter[cpu];
576 	return NULL;
577 }
578 
579 int tracer_init(struct tracer *t, struct trace_array *tr);
580 int tracing_is_enabled(void);
581 void tracing_reset(struct trace_buffer *buf, int cpu);
582 void tracing_reset_online_cpus(struct trace_buffer *buf);
583 void tracing_reset_current(int cpu);
584 void tracing_reset_all_online_cpus(void);
585 int tracing_open_generic(struct inode *inode, struct file *filp);
586 bool tracing_is_disabled(void);
587 int tracer_tracing_is_on(struct trace_array *tr);
588 void tracer_tracing_on(struct trace_array *tr);
589 void tracer_tracing_off(struct trace_array *tr);
590 struct dentry *trace_create_file(const char *name,
591 				 umode_t mode,
592 				 struct dentry *parent,
593 				 void *data,
594 				 const struct file_operations *fops);
595 
596 struct dentry *tracing_init_dentry(void);
597 
598 struct ring_buffer_event;
599 
600 struct ring_buffer_event *
601 trace_buffer_lock_reserve(struct ring_buffer *buffer,
602 			  int type,
603 			  unsigned long len,
604 			  unsigned long flags,
605 			  int pc);
606 
607 struct trace_entry *tracing_get_trace_entry(struct trace_array *tr,
608 						struct trace_array_cpu *data);
609 
610 struct trace_entry *trace_find_next_entry(struct trace_iterator *iter,
611 					  int *ent_cpu, u64 *ent_ts);
612 
613 void trace_buffer_unlock_commit_nostack(struct ring_buffer *buffer,
614 					struct ring_buffer_event *event);
615 
616 int trace_empty(struct trace_iterator *iter);
617 
618 void *trace_find_next_entry_inc(struct trace_iterator *iter);
619 
620 void trace_init_global_iter(struct trace_iterator *iter);
621 
622 void tracing_iter_reset(struct trace_iterator *iter, int cpu);
623 
624 void trace_function(struct trace_array *tr,
625 		    unsigned long ip,
626 		    unsigned long parent_ip,
627 		    unsigned long flags, int pc);
628 void trace_graph_function(struct trace_array *tr,
629 		    unsigned long ip,
630 		    unsigned long parent_ip,
631 		    unsigned long flags, int pc);
632 void trace_latency_header(struct seq_file *m);
633 void trace_default_header(struct seq_file *m);
634 void print_trace_header(struct seq_file *m, struct trace_iterator *iter);
635 int trace_empty(struct trace_iterator *iter);
636 
637 void trace_graph_return(struct ftrace_graph_ret *trace);
638 int trace_graph_entry(struct ftrace_graph_ent *trace);
639 void set_graph_array(struct trace_array *tr);
640 
641 void tracing_start_cmdline_record(void);
642 void tracing_stop_cmdline_record(void);
643 void tracing_start_tgid_record(void);
644 void tracing_stop_tgid_record(void);
645 
646 int register_tracer(struct tracer *type);
647 int is_tracing_stopped(void);
648 
649 loff_t tracing_lseek(struct file *file, loff_t offset, int whence);
650 
651 extern cpumask_var_t __read_mostly tracing_buffer_mask;
652 
653 #define for_each_tracing_cpu(cpu)	\
654 	for_each_cpu(cpu, tracing_buffer_mask)
655 
656 extern unsigned long nsecs_to_usecs(unsigned long nsecs);
657 
658 extern unsigned long tracing_thresh;
659 
660 /* PID filtering */
661 
662 extern int pid_max;
663 
664 bool trace_find_filtered_pid(struct trace_pid_list *filtered_pids,
665 			     pid_t search_pid);
666 bool trace_ignore_this_task(struct trace_pid_list *filtered_pids,
667 			    struct task_struct *task);
668 void trace_filter_add_remove_task(struct trace_pid_list *pid_list,
669 				  struct task_struct *self,
670 				  struct task_struct *task);
671 void *trace_pid_next(struct trace_pid_list *pid_list, void *v, loff_t *pos);
672 void *trace_pid_start(struct trace_pid_list *pid_list, loff_t *pos);
673 int trace_pid_show(struct seq_file *m, void *v);
674 void trace_free_pid_list(struct trace_pid_list *pid_list);
675 int trace_pid_write(struct trace_pid_list *filtered_pids,
676 		    struct trace_pid_list **new_pid_list,
677 		    const char __user *ubuf, size_t cnt);
678 
679 #ifdef CONFIG_TRACER_MAX_TRACE
680 void update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu);
681 void update_max_tr_single(struct trace_array *tr,
682 			  struct task_struct *tsk, int cpu);
683 #endif /* CONFIG_TRACER_MAX_TRACE */
684 
685 #ifdef CONFIG_STACKTRACE
686 void ftrace_trace_userstack(struct ring_buffer *buffer, unsigned long flags,
687 			    int pc);
688 
689 void __trace_stack(struct trace_array *tr, unsigned long flags, int skip,
690 		   int pc);
691 #else
692 static inline void ftrace_trace_userstack(struct ring_buffer *buffer,
693 					  unsigned long flags, int pc)
694 {
695 }
696 
697 static inline void __trace_stack(struct trace_array *tr, unsigned long flags,
698 				 int skip, int pc)
699 {
700 }
701 #endif /* CONFIG_STACKTRACE */
702 
703 extern u64 ftrace_now(int cpu);
704 
705 extern void trace_find_cmdline(int pid, char comm[]);
706 extern int trace_find_tgid(int pid);
707 extern void trace_event_follow_fork(struct trace_array *tr, bool enable);
708 
709 #ifdef CONFIG_DYNAMIC_FTRACE
710 extern unsigned long ftrace_update_tot_cnt;
711 void ftrace_init_trace_array(struct trace_array *tr);
712 #else
713 static inline void ftrace_init_trace_array(struct trace_array *tr) { }
714 #endif
715 #define DYN_FTRACE_TEST_NAME trace_selftest_dynamic_test_func
716 extern int DYN_FTRACE_TEST_NAME(void);
717 #define DYN_FTRACE_TEST_NAME2 trace_selftest_dynamic_test_func2
718 extern int DYN_FTRACE_TEST_NAME2(void);
719 
720 extern bool ring_buffer_expanded;
721 extern bool tracing_selftest_disabled;
722 
723 #ifdef CONFIG_FTRACE_STARTUP_TEST
724 extern int trace_selftest_startup_function(struct tracer *trace,
725 					   struct trace_array *tr);
726 extern int trace_selftest_startup_function_graph(struct tracer *trace,
727 						 struct trace_array *tr);
728 extern int trace_selftest_startup_irqsoff(struct tracer *trace,
729 					  struct trace_array *tr);
730 extern int trace_selftest_startup_preemptoff(struct tracer *trace,
731 					     struct trace_array *tr);
732 extern int trace_selftest_startup_preemptirqsoff(struct tracer *trace,
733 						 struct trace_array *tr);
734 extern int trace_selftest_startup_wakeup(struct tracer *trace,
735 					 struct trace_array *tr);
736 extern int trace_selftest_startup_nop(struct tracer *trace,
737 					 struct trace_array *tr);
738 extern int trace_selftest_startup_sched_switch(struct tracer *trace,
739 					       struct trace_array *tr);
740 extern int trace_selftest_startup_branch(struct tracer *trace,
741 					 struct trace_array *tr);
742 /*
743  * Tracer data references selftest functions that only occur
744  * on boot up. These can be __init functions. Thus, when selftests
745  * are enabled, then the tracers need to reference __init functions.
746  */
747 #define __tracer_data		__refdata
748 #else
749 /* Tracers are seldom changed. Optimize when selftests are disabled. */
750 #define __tracer_data		__read_mostly
751 #endif /* CONFIG_FTRACE_STARTUP_TEST */
752 
753 extern void *head_page(struct trace_array_cpu *data);
754 extern unsigned long long ns2usecs(u64 nsec);
755 extern int
756 trace_vbprintk(unsigned long ip, const char *fmt, va_list args);
757 extern int
758 trace_vprintk(unsigned long ip, const char *fmt, va_list args);
759 extern int
760 trace_array_vprintk(struct trace_array *tr,
761 		    unsigned long ip, const char *fmt, va_list args);
762 int trace_array_printk(struct trace_array *tr,
763 		       unsigned long ip, const char *fmt, ...);
764 int trace_array_printk_buf(struct ring_buffer *buffer,
765 			   unsigned long ip, const char *fmt, ...);
766 void trace_printk_seq(struct trace_seq *s);
767 enum print_line_t print_trace_line(struct trace_iterator *iter);
768 
769 extern char trace_find_mark(unsigned long long duration);
770 
771 struct ftrace_hash;
772 
773 struct ftrace_mod_load {
774 	struct list_head	list;
775 	char			*func;
776 	char			*module;
777 	int			 enable;
778 };
779 
780 enum {
781 	FTRACE_HASH_FL_MOD	= (1 << 0),
782 };
783 
784 struct ftrace_hash {
785 	unsigned long		size_bits;
786 	struct hlist_head	*buckets;
787 	unsigned long		count;
788 	unsigned long		flags;
789 	struct rcu_head		rcu;
790 };
791 
792 struct ftrace_func_entry *
793 ftrace_lookup_ip(struct ftrace_hash *hash, unsigned long ip);
794 
795 static __always_inline bool ftrace_hash_empty(struct ftrace_hash *hash)
796 {
797 	return !hash || !(hash->count || (hash->flags & FTRACE_HASH_FL_MOD));
798 }
799 
800 /* Standard output formatting function used for function return traces */
801 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
802 
803 /* Flag options */
804 #define TRACE_GRAPH_PRINT_OVERRUN       0x1
805 #define TRACE_GRAPH_PRINT_CPU           0x2
806 #define TRACE_GRAPH_PRINT_OVERHEAD      0x4
807 #define TRACE_GRAPH_PRINT_PROC          0x8
808 #define TRACE_GRAPH_PRINT_DURATION      0x10
809 #define TRACE_GRAPH_PRINT_ABS_TIME      0x20
810 #define TRACE_GRAPH_PRINT_IRQS          0x40
811 #define TRACE_GRAPH_PRINT_TAIL          0x80
812 #define TRACE_GRAPH_SLEEP_TIME		0x100
813 #define TRACE_GRAPH_GRAPH_TIME		0x200
814 #define TRACE_GRAPH_PRINT_FILL_SHIFT	28
815 #define TRACE_GRAPH_PRINT_FILL_MASK	(0x3 << TRACE_GRAPH_PRINT_FILL_SHIFT)
816 
817 extern void ftrace_graph_sleep_time_control(bool enable);
818 extern void ftrace_graph_graph_time_control(bool enable);
819 
820 extern enum print_line_t
821 print_graph_function_flags(struct trace_iterator *iter, u32 flags);
822 extern void print_graph_headers_flags(struct seq_file *s, u32 flags);
823 extern void
824 trace_print_graph_duration(unsigned long long duration, struct trace_seq *s);
825 extern void graph_trace_open(struct trace_iterator *iter);
826 extern void graph_trace_close(struct trace_iterator *iter);
827 extern int __trace_graph_entry(struct trace_array *tr,
828 			       struct ftrace_graph_ent *trace,
829 			       unsigned long flags, int pc);
830 extern void __trace_graph_return(struct trace_array *tr,
831 				 struct ftrace_graph_ret *trace,
832 				 unsigned long flags, int pc);
833 
834 #ifdef CONFIG_DYNAMIC_FTRACE
835 extern struct ftrace_hash *ftrace_graph_hash;
836 extern struct ftrace_hash *ftrace_graph_notrace_hash;
837 
838 static inline int ftrace_graph_addr(unsigned long addr)
839 {
840 	int ret = 0;
841 
842 	preempt_disable_notrace();
843 
844 	if (ftrace_hash_empty(ftrace_graph_hash)) {
845 		ret = 1;
846 		goto out;
847 	}
848 
849 	if (ftrace_lookup_ip(ftrace_graph_hash, addr)) {
850 		/*
851 		 * If no irqs are to be traced, but a set_graph_function
852 		 * is set, and called by an interrupt handler, we still
853 		 * want to trace it.
854 		 */
855 		if (in_irq())
856 			trace_recursion_set(TRACE_IRQ_BIT);
857 		else
858 			trace_recursion_clear(TRACE_IRQ_BIT);
859 		ret = 1;
860 	}
861 
862 out:
863 	preempt_enable_notrace();
864 	return ret;
865 }
866 
867 static inline int ftrace_graph_notrace_addr(unsigned long addr)
868 {
869 	int ret = 0;
870 
871 	preempt_disable_notrace();
872 
873 	if (ftrace_lookup_ip(ftrace_graph_notrace_hash, addr))
874 		ret = 1;
875 
876 	preempt_enable_notrace();
877 	return ret;
878 }
879 #else
880 static inline int ftrace_graph_addr(unsigned long addr)
881 {
882 	return 1;
883 }
884 
885 static inline int ftrace_graph_notrace_addr(unsigned long addr)
886 {
887 	return 0;
888 }
889 #endif /* CONFIG_DYNAMIC_FTRACE */
890 
891 extern unsigned int fgraph_max_depth;
892 
893 static inline bool ftrace_graph_ignore_func(struct ftrace_graph_ent *trace)
894 {
895 	/* trace it when it is-nested-in or is a function enabled. */
896 	return !(trace->depth || ftrace_graph_addr(trace->func)) ||
897 		(trace->depth < 0) ||
898 		(fgraph_max_depth && trace->depth >= fgraph_max_depth);
899 }
900 
901 #else /* CONFIG_FUNCTION_GRAPH_TRACER */
902 static inline enum print_line_t
903 print_graph_function_flags(struct trace_iterator *iter, u32 flags)
904 {
905 	return TRACE_TYPE_UNHANDLED;
906 }
907 #endif /* CONFIG_FUNCTION_GRAPH_TRACER */
908 
909 extern struct list_head ftrace_pids;
910 
911 #ifdef CONFIG_FUNCTION_TRACER
912 struct ftrace_func_command {
913 	struct list_head	list;
914 	char			*name;
915 	int			(*func)(struct trace_array *tr,
916 					struct ftrace_hash *hash,
917 					char *func, char *cmd,
918 					char *params, int enable);
919 };
920 extern bool ftrace_filter_param __initdata;
921 static inline int ftrace_trace_task(struct trace_array *tr)
922 {
923 	return !this_cpu_read(tr->trace_buffer.data->ftrace_ignore_pid);
924 }
925 extern int ftrace_is_dead(void);
926 int ftrace_create_function_files(struct trace_array *tr,
927 				 struct dentry *parent);
928 void ftrace_destroy_function_files(struct trace_array *tr);
929 void ftrace_init_global_array_ops(struct trace_array *tr);
930 void ftrace_init_array_ops(struct trace_array *tr, ftrace_func_t func);
931 void ftrace_reset_array_ops(struct trace_array *tr);
932 int using_ftrace_ops_list_func(void);
933 void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d_tracer);
934 void ftrace_init_tracefs_toplevel(struct trace_array *tr,
935 				  struct dentry *d_tracer);
936 void ftrace_clear_pids(struct trace_array *tr);
937 int init_function_trace(void);
938 void ftrace_pid_follow_fork(struct trace_array *tr, bool enable);
939 #else
940 static inline int ftrace_trace_task(struct trace_array *tr)
941 {
942 	return 1;
943 }
944 static inline int ftrace_is_dead(void) { return 0; }
945 static inline int
946 ftrace_create_function_files(struct trace_array *tr,
947 			     struct dentry *parent)
948 {
949 	return 0;
950 }
951 static inline void ftrace_destroy_function_files(struct trace_array *tr) { }
952 static inline __init void
953 ftrace_init_global_array_ops(struct trace_array *tr) { }
954 static inline void ftrace_reset_array_ops(struct trace_array *tr) { }
955 static inline void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d) { }
956 static inline void ftrace_init_tracefs_toplevel(struct trace_array *tr, struct dentry *d) { }
957 static inline void ftrace_clear_pids(struct trace_array *tr) { }
958 static inline int init_function_trace(void) { return 0; }
959 static inline void ftrace_pid_follow_fork(struct trace_array *tr, bool enable) { }
960 /* ftace_func_t type is not defined, use macro instead of static inline */
961 #define ftrace_init_array_ops(tr, func) do { } while (0)
962 #endif /* CONFIG_FUNCTION_TRACER */
963 
964 #if defined(CONFIG_FUNCTION_TRACER) && defined(CONFIG_DYNAMIC_FTRACE)
965 
966 struct ftrace_probe_ops {
967 	void			(*func)(unsigned long ip,
968 					unsigned long parent_ip,
969 					struct trace_array *tr,
970 					struct ftrace_probe_ops *ops,
971 					void *data);
972 	int			(*init)(struct ftrace_probe_ops *ops,
973 					struct trace_array *tr,
974 					unsigned long ip, void *init_data,
975 					void **data);
976 	void			(*free)(struct ftrace_probe_ops *ops,
977 					struct trace_array *tr,
978 					unsigned long ip, void *data);
979 	int			(*print)(struct seq_file *m,
980 					 unsigned long ip,
981 					 struct ftrace_probe_ops *ops,
982 					 void *data);
983 };
984 
985 struct ftrace_func_mapper;
986 typedef int (*ftrace_mapper_func)(void *data);
987 
988 struct ftrace_func_mapper *allocate_ftrace_func_mapper(void);
989 void **ftrace_func_mapper_find_ip(struct ftrace_func_mapper *mapper,
990 					   unsigned long ip);
991 int ftrace_func_mapper_add_ip(struct ftrace_func_mapper *mapper,
992 			       unsigned long ip, void *data);
993 void *ftrace_func_mapper_remove_ip(struct ftrace_func_mapper *mapper,
994 				   unsigned long ip);
995 void free_ftrace_func_mapper(struct ftrace_func_mapper *mapper,
996 			     ftrace_mapper_func free_func);
997 
998 extern int
999 register_ftrace_function_probe(char *glob, struct trace_array *tr,
1000 			       struct ftrace_probe_ops *ops, void *data);
1001 extern int
1002 unregister_ftrace_function_probe_func(char *glob, struct trace_array *tr,
1003 				      struct ftrace_probe_ops *ops);
1004 extern void clear_ftrace_function_probes(struct trace_array *tr);
1005 
1006 int register_ftrace_command(struct ftrace_func_command *cmd);
1007 int unregister_ftrace_command(struct ftrace_func_command *cmd);
1008 
1009 void ftrace_create_filter_files(struct ftrace_ops *ops,
1010 				struct dentry *parent);
1011 void ftrace_destroy_filter_files(struct ftrace_ops *ops);
1012 #else
1013 struct ftrace_func_command;
1014 
1015 static inline __init int register_ftrace_command(struct ftrace_func_command *cmd)
1016 {
1017 	return -EINVAL;
1018 }
1019 static inline __init int unregister_ftrace_command(char *cmd_name)
1020 {
1021 	return -EINVAL;
1022 }
1023 static inline void clear_ftrace_function_probes(struct trace_array *tr)
1024 {
1025 }
1026 
1027 /*
1028  * The ops parameter passed in is usually undefined.
1029  * This must be a macro.
1030  */
1031 #define ftrace_create_filter_files(ops, parent) do { } while (0)
1032 #define ftrace_destroy_filter_files(ops) do { } while (0)
1033 #endif /* CONFIG_FUNCTION_TRACER && CONFIG_DYNAMIC_FTRACE */
1034 
1035 bool ftrace_event_is_function(struct trace_event_call *call);
1036 
1037 /*
1038  * struct trace_parser - servers for reading the user input separated by spaces
1039  * @cont: set if the input is not complete - no final space char was found
1040  * @buffer: holds the parsed user input
1041  * @idx: user input length
1042  * @size: buffer size
1043  */
1044 struct trace_parser {
1045 	bool		cont;
1046 	char		*buffer;
1047 	unsigned	idx;
1048 	unsigned	size;
1049 };
1050 
1051 static inline bool trace_parser_loaded(struct trace_parser *parser)
1052 {
1053 	return (parser->idx != 0);
1054 }
1055 
1056 static inline bool trace_parser_cont(struct trace_parser *parser)
1057 {
1058 	return parser->cont;
1059 }
1060 
1061 static inline void trace_parser_clear(struct trace_parser *parser)
1062 {
1063 	parser->cont = false;
1064 	parser->idx = 0;
1065 }
1066 
1067 extern int trace_parser_get_init(struct trace_parser *parser, int size);
1068 extern void trace_parser_put(struct trace_parser *parser);
1069 extern int trace_get_user(struct trace_parser *parser, const char __user *ubuf,
1070 	size_t cnt, loff_t *ppos);
1071 
1072 /*
1073  * Only create function graph options if function graph is configured.
1074  */
1075 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1076 # define FGRAPH_FLAGS						\
1077 		C(DISPLAY_GRAPH,	"display-graph"),
1078 #else
1079 # define FGRAPH_FLAGS
1080 #endif
1081 
1082 #ifdef CONFIG_BRANCH_TRACER
1083 # define BRANCH_FLAGS					\
1084 		C(BRANCH,		"branch"),
1085 #else
1086 # define BRANCH_FLAGS
1087 #endif
1088 
1089 #ifdef CONFIG_FUNCTION_TRACER
1090 # define FUNCTION_FLAGS						\
1091 		C(FUNCTION,		"function-trace"),	\
1092 		C(FUNC_FORK,		"function-fork"),
1093 # define FUNCTION_DEFAULT_FLAGS		TRACE_ITER_FUNCTION
1094 #else
1095 # define FUNCTION_FLAGS
1096 # define FUNCTION_DEFAULT_FLAGS		0UL
1097 # define TRACE_ITER_FUNC_FORK		0UL
1098 #endif
1099 
1100 #ifdef CONFIG_STACKTRACE
1101 # define STACK_FLAGS				\
1102 		C(STACKTRACE,		"stacktrace"),
1103 #else
1104 # define STACK_FLAGS
1105 #endif
1106 
1107 /*
1108  * trace_iterator_flags is an enumeration that defines bit
1109  * positions into trace_flags that controls the output.
1110  *
1111  * NOTE: These bits must match the trace_options array in
1112  *       trace.c (this macro guarantees it).
1113  */
1114 #define TRACE_FLAGS						\
1115 		C(PRINT_PARENT,		"print-parent"),	\
1116 		C(SYM_OFFSET,		"sym-offset"),		\
1117 		C(SYM_ADDR,		"sym-addr"),		\
1118 		C(VERBOSE,		"verbose"),		\
1119 		C(RAW,			"raw"),			\
1120 		C(HEX,			"hex"),			\
1121 		C(BIN,			"bin"),			\
1122 		C(BLOCK,		"block"),		\
1123 		C(PRINTK,		"trace_printk"),	\
1124 		C(ANNOTATE,		"annotate"),		\
1125 		C(USERSTACKTRACE,	"userstacktrace"),	\
1126 		C(SYM_USEROBJ,		"sym-userobj"),		\
1127 		C(PRINTK_MSGONLY,	"printk-msg-only"),	\
1128 		C(CONTEXT_INFO,		"context-info"),   /* Print pid/cpu/time */ \
1129 		C(LATENCY_FMT,		"latency-format"),	\
1130 		C(RECORD_CMD,		"record-cmd"),		\
1131 		C(RECORD_TGID,		"record-tgid"),		\
1132 		C(OVERWRITE,		"overwrite"),		\
1133 		C(STOP_ON_FREE,		"disable_on_free"),	\
1134 		C(IRQ_INFO,		"irq-info"),		\
1135 		C(MARKERS,		"markers"),		\
1136 		C(EVENT_FORK,		"event-fork"),		\
1137 		FUNCTION_FLAGS					\
1138 		FGRAPH_FLAGS					\
1139 		STACK_FLAGS					\
1140 		BRANCH_FLAGS
1141 
1142 /*
1143  * By defining C, we can make TRACE_FLAGS a list of bit names
1144  * that will define the bits for the flag masks.
1145  */
1146 #undef C
1147 #define C(a, b) TRACE_ITER_##a##_BIT
1148 
1149 enum trace_iterator_bits {
1150 	TRACE_FLAGS
1151 	/* Make sure we don't go more than we have bits for */
1152 	TRACE_ITER_LAST_BIT
1153 };
1154 
1155 /*
1156  * By redefining C, we can make TRACE_FLAGS a list of masks that
1157  * use the bits as defined above.
1158  */
1159 #undef C
1160 #define C(a, b) TRACE_ITER_##a = (1 << TRACE_ITER_##a##_BIT)
1161 
1162 enum trace_iterator_flags { TRACE_FLAGS };
1163 
1164 /*
1165  * TRACE_ITER_SYM_MASK masks the options in trace_flags that
1166  * control the output of kernel symbols.
1167  */
1168 #define TRACE_ITER_SYM_MASK \
1169 	(TRACE_ITER_PRINT_PARENT|TRACE_ITER_SYM_OFFSET|TRACE_ITER_SYM_ADDR)
1170 
1171 extern struct tracer nop_trace;
1172 
1173 #ifdef CONFIG_BRANCH_TRACER
1174 extern int enable_branch_tracing(struct trace_array *tr);
1175 extern void disable_branch_tracing(void);
1176 static inline int trace_branch_enable(struct trace_array *tr)
1177 {
1178 	if (tr->trace_flags & TRACE_ITER_BRANCH)
1179 		return enable_branch_tracing(tr);
1180 	return 0;
1181 }
1182 static inline void trace_branch_disable(void)
1183 {
1184 	/* due to races, always disable */
1185 	disable_branch_tracing();
1186 }
1187 #else
1188 static inline int trace_branch_enable(struct trace_array *tr)
1189 {
1190 	return 0;
1191 }
1192 static inline void trace_branch_disable(void)
1193 {
1194 }
1195 #endif /* CONFIG_BRANCH_TRACER */
1196 
1197 /* set ring buffers to default size if not already done so */
1198 int tracing_update_buffers(void);
1199 
1200 struct ftrace_event_field {
1201 	struct list_head	link;
1202 	const char		*name;
1203 	const char		*type;
1204 	int			filter_type;
1205 	int			offset;
1206 	int			size;
1207 	int			is_signed;
1208 };
1209 
1210 struct event_filter {
1211 	int			n_preds;	/* Number assigned */
1212 	int			a_preds;	/* allocated */
1213 	struct filter_pred	*preds;
1214 	struct filter_pred	*root;
1215 	char			*filter_string;
1216 };
1217 
1218 struct event_subsystem {
1219 	struct list_head	list;
1220 	const char		*name;
1221 	struct event_filter	*filter;
1222 	int			ref_count;
1223 };
1224 
1225 struct trace_subsystem_dir {
1226 	struct list_head		list;
1227 	struct event_subsystem		*subsystem;
1228 	struct trace_array		*tr;
1229 	struct dentry			*entry;
1230 	int				ref_count;
1231 	int				nr_events;
1232 };
1233 
1234 extern int call_filter_check_discard(struct trace_event_call *call, void *rec,
1235 				     struct ring_buffer *buffer,
1236 				     struct ring_buffer_event *event);
1237 
1238 void trace_buffer_unlock_commit_regs(struct trace_array *tr,
1239 				     struct ring_buffer *buffer,
1240 				     struct ring_buffer_event *event,
1241 				     unsigned long flags, int pc,
1242 				     struct pt_regs *regs);
1243 
1244 static inline void trace_buffer_unlock_commit(struct trace_array *tr,
1245 					      struct ring_buffer *buffer,
1246 					      struct ring_buffer_event *event,
1247 					      unsigned long flags, int pc)
1248 {
1249 	trace_buffer_unlock_commit_regs(tr, buffer, event, flags, pc, NULL);
1250 }
1251 
1252 DECLARE_PER_CPU(struct ring_buffer_event *, trace_buffered_event);
1253 DECLARE_PER_CPU(int, trace_buffered_event_cnt);
1254 void trace_buffered_event_disable(void);
1255 void trace_buffered_event_enable(void);
1256 
1257 static inline void
1258 __trace_event_discard_commit(struct ring_buffer *buffer,
1259 			     struct ring_buffer_event *event)
1260 {
1261 	if (this_cpu_read(trace_buffered_event) == event) {
1262 		/* Simply release the temp buffer */
1263 		this_cpu_dec(trace_buffered_event_cnt);
1264 		return;
1265 	}
1266 	ring_buffer_discard_commit(buffer, event);
1267 }
1268 
1269 /*
1270  * Helper function for event_trigger_unlock_commit{_regs}().
1271  * If there are event triggers attached to this event that requires
1272  * filtering against its fields, then they wil be called as the
1273  * entry already holds the field information of the current event.
1274  *
1275  * It also checks if the event should be discarded or not.
1276  * It is to be discarded if the event is soft disabled and the
1277  * event was only recorded to process triggers, or if the event
1278  * filter is active and this event did not match the filters.
1279  *
1280  * Returns true if the event is discarded, false otherwise.
1281  */
1282 static inline bool
1283 __event_trigger_test_discard(struct trace_event_file *file,
1284 			     struct ring_buffer *buffer,
1285 			     struct ring_buffer_event *event,
1286 			     void *entry,
1287 			     enum event_trigger_type *tt)
1288 {
1289 	unsigned long eflags = file->flags;
1290 
1291 	if (eflags & EVENT_FILE_FL_TRIGGER_COND)
1292 		*tt = event_triggers_call(file, entry);
1293 
1294 	if (test_bit(EVENT_FILE_FL_SOFT_DISABLED_BIT, &file->flags) ||
1295 	    (unlikely(file->flags & EVENT_FILE_FL_FILTERED) &&
1296 	     !filter_match_preds(file->filter, entry))) {
1297 		__trace_event_discard_commit(buffer, event);
1298 		return true;
1299 	}
1300 
1301 	return false;
1302 }
1303 
1304 /**
1305  * event_trigger_unlock_commit - handle triggers and finish event commit
1306  * @file: The file pointer assoctiated to the event
1307  * @buffer: The ring buffer that the event is being written to
1308  * @event: The event meta data in the ring buffer
1309  * @entry: The event itself
1310  * @irq_flags: The state of the interrupts at the start of the event
1311  * @pc: The state of the preempt count at the start of the event.
1312  *
1313  * This is a helper function to handle triggers that require data
1314  * from the event itself. It also tests the event against filters and
1315  * if the event is soft disabled and should be discarded.
1316  */
1317 static inline void
1318 event_trigger_unlock_commit(struct trace_event_file *file,
1319 			    struct ring_buffer *buffer,
1320 			    struct ring_buffer_event *event,
1321 			    void *entry, unsigned long irq_flags, int pc)
1322 {
1323 	enum event_trigger_type tt = ETT_NONE;
1324 
1325 	if (!__event_trigger_test_discard(file, buffer, event, entry, &tt))
1326 		trace_buffer_unlock_commit(file->tr, buffer, event, irq_flags, pc);
1327 
1328 	if (tt)
1329 		event_triggers_post_call(file, tt, entry);
1330 }
1331 
1332 /**
1333  * event_trigger_unlock_commit_regs - handle triggers and finish event commit
1334  * @file: The file pointer assoctiated to the event
1335  * @buffer: The ring buffer that the event is being written to
1336  * @event: The event meta data in the ring buffer
1337  * @entry: The event itself
1338  * @irq_flags: The state of the interrupts at the start of the event
1339  * @pc: The state of the preempt count at the start of the event.
1340  *
1341  * This is a helper function to handle triggers that require data
1342  * from the event itself. It also tests the event against filters and
1343  * if the event is soft disabled and should be discarded.
1344  *
1345  * Same as event_trigger_unlock_commit() but calls
1346  * trace_buffer_unlock_commit_regs() instead of trace_buffer_unlock_commit().
1347  */
1348 static inline void
1349 event_trigger_unlock_commit_regs(struct trace_event_file *file,
1350 				 struct ring_buffer *buffer,
1351 				 struct ring_buffer_event *event,
1352 				 void *entry, unsigned long irq_flags, int pc,
1353 				 struct pt_regs *regs)
1354 {
1355 	enum event_trigger_type tt = ETT_NONE;
1356 
1357 	if (!__event_trigger_test_discard(file, buffer, event, entry, &tt))
1358 		trace_buffer_unlock_commit_regs(file->tr, buffer, event,
1359 						irq_flags, pc, regs);
1360 
1361 	if (tt)
1362 		event_triggers_post_call(file, tt, entry);
1363 }
1364 
1365 #define FILTER_PRED_INVALID	((unsigned short)-1)
1366 #define FILTER_PRED_IS_RIGHT	(1 << 15)
1367 #define FILTER_PRED_FOLD	(1 << 15)
1368 
1369 /*
1370  * The max preds is the size of unsigned short with
1371  * two flags at the MSBs. One bit is used for both the IS_RIGHT
1372  * and FOLD flags. The other is reserved.
1373  *
1374  * 2^14 preds is way more than enough.
1375  */
1376 #define MAX_FILTER_PRED		16384
1377 
1378 struct filter_pred;
1379 struct regex;
1380 
1381 typedef int (*filter_pred_fn_t) (struct filter_pred *pred, void *event);
1382 
1383 typedef int (*regex_match_func)(char *str, struct regex *r, int len);
1384 
1385 enum regex_type {
1386 	MATCH_FULL = 0,
1387 	MATCH_FRONT_ONLY,
1388 	MATCH_MIDDLE_ONLY,
1389 	MATCH_END_ONLY,
1390 	MATCH_GLOB,
1391 };
1392 
1393 struct regex {
1394 	char			pattern[MAX_FILTER_STR_VAL];
1395 	int			len;
1396 	int			field_len;
1397 	regex_match_func	match;
1398 };
1399 
1400 struct filter_pred {
1401 	filter_pred_fn_t 	fn;
1402 	u64 			val;
1403 	struct regex		regex;
1404 	unsigned short		*ops;
1405 	struct ftrace_event_field *field;
1406 	int 			offset;
1407 	int 			not;
1408 	int 			op;
1409 	unsigned short		index;
1410 	unsigned short		parent;
1411 	unsigned short		left;
1412 	unsigned short		right;
1413 };
1414 
1415 static inline bool is_string_field(struct ftrace_event_field *field)
1416 {
1417 	return field->filter_type == FILTER_DYN_STRING ||
1418 	       field->filter_type == FILTER_STATIC_STRING ||
1419 	       field->filter_type == FILTER_PTR_STRING ||
1420 	       field->filter_type == FILTER_COMM;
1421 }
1422 
1423 static inline bool is_function_field(struct ftrace_event_field *field)
1424 {
1425 	return field->filter_type == FILTER_TRACE_FN;
1426 }
1427 
1428 extern enum regex_type
1429 filter_parse_regex(char *buff, int len, char **search, int *not);
1430 extern void print_event_filter(struct trace_event_file *file,
1431 			       struct trace_seq *s);
1432 extern int apply_event_filter(struct trace_event_file *file,
1433 			      char *filter_string);
1434 extern int apply_subsystem_event_filter(struct trace_subsystem_dir *dir,
1435 					char *filter_string);
1436 extern void print_subsystem_event_filter(struct event_subsystem *system,
1437 					 struct trace_seq *s);
1438 extern int filter_assign_type(const char *type);
1439 extern int create_event_filter(struct trace_event_call *call,
1440 			       char *filter_str, bool set_str,
1441 			       struct event_filter **filterp);
1442 extern void free_event_filter(struct event_filter *filter);
1443 
1444 struct ftrace_event_field *
1445 trace_find_event_field(struct trace_event_call *call, char *name);
1446 
1447 extern void trace_event_enable_cmd_record(bool enable);
1448 extern void trace_event_enable_tgid_record(bool enable);
1449 
1450 extern int event_trace_add_tracer(struct dentry *parent, struct trace_array *tr);
1451 extern int event_trace_del_tracer(struct trace_array *tr);
1452 
1453 extern struct trace_event_file *find_event_file(struct trace_array *tr,
1454 						const char *system,
1455 						const char *event);
1456 
1457 static inline void *event_file_data(struct file *filp)
1458 {
1459 	return ACCESS_ONCE(file_inode(filp)->i_private);
1460 }
1461 
1462 extern struct mutex event_mutex;
1463 extern struct list_head ftrace_events;
1464 
1465 extern const struct file_operations event_trigger_fops;
1466 extern const struct file_operations event_hist_fops;
1467 
1468 #ifdef CONFIG_HIST_TRIGGERS
1469 extern int register_trigger_hist_cmd(void);
1470 extern int register_trigger_hist_enable_disable_cmds(void);
1471 #else
1472 static inline int register_trigger_hist_cmd(void) { return 0; }
1473 static inline int register_trigger_hist_enable_disable_cmds(void) { return 0; }
1474 #endif
1475 
1476 extern int register_trigger_cmds(void);
1477 extern void clear_event_triggers(struct trace_array *tr);
1478 
1479 struct event_trigger_data {
1480 	unsigned long			count;
1481 	int				ref;
1482 	struct event_trigger_ops	*ops;
1483 	struct event_command		*cmd_ops;
1484 	struct event_filter __rcu	*filter;
1485 	char				*filter_str;
1486 	void				*private_data;
1487 	bool				paused;
1488 	bool				paused_tmp;
1489 	struct list_head		list;
1490 	char				*name;
1491 	struct list_head		named_list;
1492 	struct event_trigger_data	*named_data;
1493 };
1494 
1495 /* Avoid typos */
1496 #define ENABLE_EVENT_STR	"enable_event"
1497 #define DISABLE_EVENT_STR	"disable_event"
1498 #define ENABLE_HIST_STR		"enable_hist"
1499 #define DISABLE_HIST_STR	"disable_hist"
1500 
1501 struct enable_trigger_data {
1502 	struct trace_event_file		*file;
1503 	bool				enable;
1504 	bool				hist;
1505 };
1506 
1507 extern int event_enable_trigger_print(struct seq_file *m,
1508 				      struct event_trigger_ops *ops,
1509 				      struct event_trigger_data *data);
1510 extern void event_enable_trigger_free(struct event_trigger_ops *ops,
1511 				      struct event_trigger_data *data);
1512 extern int event_enable_trigger_func(struct event_command *cmd_ops,
1513 				     struct trace_event_file *file,
1514 				     char *glob, char *cmd, char *param);
1515 extern int event_enable_register_trigger(char *glob,
1516 					 struct event_trigger_ops *ops,
1517 					 struct event_trigger_data *data,
1518 					 struct trace_event_file *file);
1519 extern void event_enable_unregister_trigger(char *glob,
1520 					    struct event_trigger_ops *ops,
1521 					    struct event_trigger_data *test,
1522 					    struct trace_event_file *file);
1523 extern void trigger_data_free(struct event_trigger_data *data);
1524 extern int event_trigger_init(struct event_trigger_ops *ops,
1525 			      struct event_trigger_data *data);
1526 extern int trace_event_trigger_enable_disable(struct trace_event_file *file,
1527 					      int trigger_enable);
1528 extern void update_cond_flag(struct trace_event_file *file);
1529 extern void unregister_trigger(char *glob, struct event_trigger_ops *ops,
1530 			       struct event_trigger_data *test,
1531 			       struct trace_event_file *file);
1532 extern int set_trigger_filter(char *filter_str,
1533 			      struct event_trigger_data *trigger_data,
1534 			      struct trace_event_file *file);
1535 extern struct event_trigger_data *find_named_trigger(const char *name);
1536 extern bool is_named_trigger(struct event_trigger_data *test);
1537 extern int save_named_trigger(const char *name,
1538 			      struct event_trigger_data *data);
1539 extern void del_named_trigger(struct event_trigger_data *data);
1540 extern void pause_named_trigger(struct event_trigger_data *data);
1541 extern void unpause_named_trigger(struct event_trigger_data *data);
1542 extern void set_named_trigger_data(struct event_trigger_data *data,
1543 				   struct event_trigger_data *named_data);
1544 extern int register_event_command(struct event_command *cmd);
1545 extern int unregister_event_command(struct event_command *cmd);
1546 extern int register_trigger_hist_enable_disable_cmds(void);
1547 
1548 /**
1549  * struct event_trigger_ops - callbacks for trace event triggers
1550  *
1551  * The methods in this structure provide per-event trigger hooks for
1552  * various trigger operations.
1553  *
1554  * All the methods below, except for @init() and @free(), must be
1555  * implemented.
1556  *
1557  * @func: The trigger 'probe' function called when the triggering
1558  *	event occurs.  The data passed into this callback is the data
1559  *	that was supplied to the event_command @reg() function that
1560  *	registered the trigger (see struct event_command) along with
1561  *	the trace record, rec.
1562  *
1563  * @init: An optional initialization function called for the trigger
1564  *	when the trigger is registered (via the event_command reg()
1565  *	function).  This can be used to perform per-trigger
1566  *	initialization such as incrementing a per-trigger reference
1567  *	count, for instance.  This is usually implemented by the
1568  *	generic utility function @event_trigger_init() (see
1569  *	trace_event_triggers.c).
1570  *
1571  * @free: An optional de-initialization function called for the
1572  *	trigger when the trigger is unregistered (via the
1573  *	event_command @reg() function).  This can be used to perform
1574  *	per-trigger de-initialization such as decrementing a
1575  *	per-trigger reference count and freeing corresponding trigger
1576  *	data, for instance.  This is usually implemented by the
1577  *	generic utility function @event_trigger_free() (see
1578  *	trace_event_triggers.c).
1579  *
1580  * @print: The callback function invoked to have the trigger print
1581  *	itself.  This is usually implemented by a wrapper function
1582  *	that calls the generic utility function @event_trigger_print()
1583  *	(see trace_event_triggers.c).
1584  */
1585 struct event_trigger_ops {
1586 	void			(*func)(struct event_trigger_data *data,
1587 					void *rec);
1588 	int			(*init)(struct event_trigger_ops *ops,
1589 					struct event_trigger_data *data);
1590 	void			(*free)(struct event_trigger_ops *ops,
1591 					struct event_trigger_data *data);
1592 	int			(*print)(struct seq_file *m,
1593 					 struct event_trigger_ops *ops,
1594 					 struct event_trigger_data *data);
1595 };
1596 
1597 /**
1598  * struct event_command - callbacks and data members for event commands
1599  *
1600  * Event commands are invoked by users by writing the command name
1601  * into the 'trigger' file associated with a trace event.  The
1602  * parameters associated with a specific invocation of an event
1603  * command are used to create an event trigger instance, which is
1604  * added to the list of trigger instances associated with that trace
1605  * event.  When the event is hit, the set of triggers associated with
1606  * that event is invoked.
1607  *
1608  * The data members in this structure provide per-event command data
1609  * for various event commands.
1610  *
1611  * All the data members below, except for @post_trigger, must be set
1612  * for each event command.
1613  *
1614  * @name: The unique name that identifies the event command.  This is
1615  *	the name used when setting triggers via trigger files.
1616  *
1617  * @trigger_type: A unique id that identifies the event command
1618  *	'type'.  This value has two purposes, the first to ensure that
1619  *	only one trigger of the same type can be set at a given time
1620  *	for a particular event e.g. it doesn't make sense to have both
1621  *	a traceon and traceoff trigger attached to a single event at
1622  *	the same time, so traceon and traceoff have the same type
1623  *	though they have different names.  The @trigger_type value is
1624  *	also used as a bit value for deferring the actual trigger
1625  *	action until after the current event is finished.  Some
1626  *	commands need to do this if they themselves log to the trace
1627  *	buffer (see the @post_trigger() member below).  @trigger_type
1628  *	values are defined by adding new values to the trigger_type
1629  *	enum in include/linux/trace_events.h.
1630  *
1631  * @flags: See the enum event_command_flags below.
1632  *
1633  * All the methods below, except for @set_filter() and @unreg_all(),
1634  * must be implemented.
1635  *
1636  * @func: The callback function responsible for parsing and
1637  *	registering the trigger written to the 'trigger' file by the
1638  *	user.  It allocates the trigger instance and registers it with
1639  *	the appropriate trace event.  It makes use of the other
1640  *	event_command callback functions to orchestrate this, and is
1641  *	usually implemented by the generic utility function
1642  *	@event_trigger_callback() (see trace_event_triggers.c).
1643  *
1644  * @reg: Adds the trigger to the list of triggers associated with the
1645  *	event, and enables the event trigger itself, after
1646  *	initializing it (via the event_trigger_ops @init() function).
1647  *	This is also where commands can use the @trigger_type value to
1648  *	make the decision as to whether or not multiple instances of
1649  *	the trigger should be allowed.  This is usually implemented by
1650  *	the generic utility function @register_trigger() (see
1651  *	trace_event_triggers.c).
1652  *
1653  * @unreg: Removes the trigger from the list of triggers associated
1654  *	with the event, and disables the event trigger itself, after
1655  *	initializing it (via the event_trigger_ops @free() function).
1656  *	This is usually implemented by the generic utility function
1657  *	@unregister_trigger() (see trace_event_triggers.c).
1658  *
1659  * @unreg_all: An optional function called to remove all the triggers
1660  *	from the list of triggers associated with the event.  Called
1661  *	when a trigger file is opened in truncate mode.
1662  *
1663  * @set_filter: An optional function called to parse and set a filter
1664  *	for the trigger.  If no @set_filter() method is set for the
1665  *	event command, filters set by the user for the command will be
1666  *	ignored.  This is usually implemented by the generic utility
1667  *	function @set_trigger_filter() (see trace_event_triggers.c).
1668  *
1669  * @get_trigger_ops: The callback function invoked to retrieve the
1670  *	event_trigger_ops implementation associated with the command.
1671  */
1672 struct event_command {
1673 	struct list_head	list;
1674 	char			*name;
1675 	enum event_trigger_type	trigger_type;
1676 	int			flags;
1677 	int			(*func)(struct event_command *cmd_ops,
1678 					struct trace_event_file *file,
1679 					char *glob, char *cmd, char *params);
1680 	int			(*reg)(char *glob,
1681 				       struct event_trigger_ops *ops,
1682 				       struct event_trigger_data *data,
1683 				       struct trace_event_file *file);
1684 	void			(*unreg)(char *glob,
1685 					 struct event_trigger_ops *ops,
1686 					 struct event_trigger_data *data,
1687 					 struct trace_event_file *file);
1688 	void			(*unreg_all)(struct trace_event_file *file);
1689 	int			(*set_filter)(char *filter_str,
1690 					      struct event_trigger_data *data,
1691 					      struct trace_event_file *file);
1692 	struct event_trigger_ops *(*get_trigger_ops)(char *cmd, char *param);
1693 };
1694 
1695 /**
1696  * enum event_command_flags - flags for struct event_command
1697  *
1698  * @POST_TRIGGER: A flag that says whether or not this command needs
1699  *	to have its action delayed until after the current event has
1700  *	been closed.  Some triggers need to avoid being invoked while
1701  *	an event is currently in the process of being logged, since
1702  *	the trigger may itself log data into the trace buffer.  Thus
1703  *	we make sure the current event is committed before invoking
1704  *	those triggers.  To do that, the trigger invocation is split
1705  *	in two - the first part checks the filter using the current
1706  *	trace record; if a command has the @post_trigger flag set, it
1707  *	sets a bit for itself in the return value, otherwise it
1708  *	directly invokes the trigger.  Once all commands have been
1709  *	either invoked or set their return flag, the current record is
1710  *	either committed or discarded.  At that point, if any commands
1711  *	have deferred their triggers, those commands are finally
1712  *	invoked following the close of the current event.  In other
1713  *	words, if the event_trigger_ops @func() probe implementation
1714  *	itself logs to the trace buffer, this flag should be set,
1715  *	otherwise it can be left unspecified.
1716  *
1717  * @NEEDS_REC: A flag that says whether or not this command needs
1718  *	access to the trace record in order to perform its function,
1719  *	regardless of whether or not it has a filter associated with
1720  *	it (filters make a trigger require access to the trace record
1721  *	but are not always present).
1722  */
1723 enum event_command_flags {
1724 	EVENT_CMD_FL_POST_TRIGGER	= 1,
1725 	EVENT_CMD_FL_NEEDS_REC		= 2,
1726 };
1727 
1728 static inline bool event_command_post_trigger(struct event_command *cmd_ops)
1729 {
1730 	return cmd_ops->flags & EVENT_CMD_FL_POST_TRIGGER;
1731 }
1732 
1733 static inline bool event_command_needs_rec(struct event_command *cmd_ops)
1734 {
1735 	return cmd_ops->flags & EVENT_CMD_FL_NEEDS_REC;
1736 }
1737 
1738 extern int trace_event_enable_disable(struct trace_event_file *file,
1739 				      int enable, int soft_disable);
1740 extern int tracing_alloc_snapshot(void);
1741 
1742 extern const char *__start___trace_bprintk_fmt[];
1743 extern const char *__stop___trace_bprintk_fmt[];
1744 
1745 extern const char *__start___tracepoint_str[];
1746 extern const char *__stop___tracepoint_str[];
1747 
1748 void trace_printk_control(bool enabled);
1749 void trace_printk_init_buffers(void);
1750 void trace_printk_start_comm(void);
1751 int trace_keep_overwrite(struct tracer *tracer, u32 mask, int set);
1752 int set_tracer_flag(struct trace_array *tr, unsigned int mask, int enabled);
1753 
1754 /*
1755  * Normal trace_printk() and friends allocates special buffers
1756  * to do the manipulation, as well as saves the print formats
1757  * into sections to display. But the trace infrastructure wants
1758  * to use these without the added overhead at the price of being
1759  * a bit slower (used mainly for warnings, where we don't care
1760  * about performance). The internal_trace_puts() is for such
1761  * a purpose.
1762  */
1763 #define internal_trace_puts(str) __trace_puts(_THIS_IP_, str, strlen(str))
1764 
1765 #undef FTRACE_ENTRY
1766 #define FTRACE_ENTRY(call, struct_name, id, tstruct, print, filter)	\
1767 	extern struct trace_event_call					\
1768 	__aligned(4) event_##call;
1769 #undef FTRACE_ENTRY_DUP
1770 #define FTRACE_ENTRY_DUP(call, struct_name, id, tstruct, print, filter)	\
1771 	FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print), \
1772 		     filter)
1773 #undef FTRACE_ENTRY_PACKED
1774 #define FTRACE_ENTRY_PACKED(call, struct_name, id, tstruct, print, filter) \
1775 	FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print), \
1776 		     filter)
1777 
1778 #include "trace_entries.h"
1779 
1780 #if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_FUNCTION_TRACER)
1781 int perf_ftrace_event_register(struct trace_event_call *call,
1782 			       enum trace_reg type, void *data);
1783 #else
1784 #define perf_ftrace_event_register NULL
1785 #endif
1786 
1787 #ifdef CONFIG_FTRACE_SYSCALLS
1788 void init_ftrace_syscalls(void);
1789 const char *get_syscall_name(int syscall);
1790 #else
1791 static inline void init_ftrace_syscalls(void) { }
1792 static inline const char *get_syscall_name(int syscall)
1793 {
1794 	return NULL;
1795 }
1796 #endif
1797 
1798 #ifdef CONFIG_EVENT_TRACING
1799 void trace_event_init(void);
1800 void trace_event_eval_update(struct trace_eval_map **map, int len);
1801 #else
1802 static inline void __init trace_event_init(void) { }
1803 static inline void trace_event_eval_update(struct trace_eval_map **map, int len) { }
1804 #endif
1805 
1806 extern struct trace_iterator *tracepoint_print_iter;
1807 
1808 #endif /* _LINUX_KERNEL_TRACE_H */
1809