xref: /linux/kernel/trace/trace.h (revision fd639726bf15fca8ee1a00dce8e0096d0ad9bd18)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 
3 #ifndef _LINUX_KERNEL_TRACE_H
4 #define _LINUX_KERNEL_TRACE_H
5 
6 #include <linux/fs.h>
7 #include <linux/atomic.h>
8 #include <linux/sched.h>
9 #include <linux/clocksource.h>
10 #include <linux/ring_buffer.h>
11 #include <linux/mmiotrace.h>
12 #include <linux/tracepoint.h>
13 #include <linux/ftrace.h>
14 #include <linux/hw_breakpoint.h>
15 #include <linux/trace_seq.h>
16 #include <linux/trace_events.h>
17 #include <linux/compiler.h>
18 #include <linux/trace_seq.h>
19 #include <linux/glob.h>
20 
21 #ifdef CONFIG_FTRACE_SYSCALLS
22 #include <asm/unistd.h>		/* For NR_SYSCALLS	     */
23 #include <asm/syscall.h>	/* some archs define it here */
24 #endif
25 
26 enum trace_type {
27 	__TRACE_FIRST_TYPE = 0,
28 
29 	TRACE_FN,
30 	TRACE_CTX,
31 	TRACE_WAKE,
32 	TRACE_STACK,
33 	TRACE_PRINT,
34 	TRACE_BPRINT,
35 	TRACE_MMIO_RW,
36 	TRACE_MMIO_MAP,
37 	TRACE_BRANCH,
38 	TRACE_GRAPH_RET,
39 	TRACE_GRAPH_ENT,
40 	TRACE_USER_STACK,
41 	TRACE_BLK,
42 	TRACE_BPUTS,
43 	TRACE_HWLAT,
44 	TRACE_RAW_DATA,
45 
46 	__TRACE_LAST_TYPE,
47 };
48 
49 
50 #undef __field
51 #define __field(type, item)		type	item;
52 
53 #undef __field_struct
54 #define __field_struct(type, item)	__field(type, item)
55 
56 #undef __field_desc
57 #define __field_desc(type, container, item)
58 
59 #undef __array
60 #define __array(type, item, size)	type	item[size];
61 
62 #undef __array_desc
63 #define __array_desc(type, container, item, size)
64 
65 #undef __dynamic_array
66 #define __dynamic_array(type, item)	type	item[];
67 
68 #undef F_STRUCT
69 #define F_STRUCT(args...)		args
70 
71 #undef FTRACE_ENTRY
72 #define FTRACE_ENTRY(name, struct_name, id, tstruct, print, filter)	\
73 	struct struct_name {						\
74 		struct trace_entry	ent;				\
75 		tstruct							\
76 	}
77 
78 #undef FTRACE_ENTRY_DUP
79 #define FTRACE_ENTRY_DUP(name, name_struct, id, tstruct, printk, filter)
80 
81 #undef FTRACE_ENTRY_REG
82 #define FTRACE_ENTRY_REG(name, struct_name, id, tstruct, print,	\
83 			 filter, regfn) \
84 	FTRACE_ENTRY(name, struct_name, id, PARAMS(tstruct), PARAMS(print), \
85 		     filter)
86 
87 #undef FTRACE_ENTRY_PACKED
88 #define FTRACE_ENTRY_PACKED(name, struct_name, id, tstruct, print,	\
89 			    filter)					\
90 	FTRACE_ENTRY(name, struct_name, id, PARAMS(tstruct), PARAMS(print), \
91 		     filter) __packed
92 
93 #include "trace_entries.h"
94 
95 /*
96  * syscalls are special, and need special handling, this is why
97  * they are not included in trace_entries.h
98  */
99 struct syscall_trace_enter {
100 	struct trace_entry	ent;
101 	int			nr;
102 	unsigned long		args[];
103 };
104 
105 struct syscall_trace_exit {
106 	struct trace_entry	ent;
107 	int			nr;
108 	long			ret;
109 };
110 
111 struct kprobe_trace_entry_head {
112 	struct trace_entry	ent;
113 	unsigned long		ip;
114 };
115 
116 struct kretprobe_trace_entry_head {
117 	struct trace_entry	ent;
118 	unsigned long		func;
119 	unsigned long		ret_ip;
120 };
121 
122 /*
123  * trace_flag_type is an enumeration that holds different
124  * states when a trace occurs. These are:
125  *  IRQS_OFF		- interrupts were disabled
126  *  IRQS_NOSUPPORT	- arch does not support irqs_disabled_flags
127  *  NEED_RESCHED	- reschedule is requested
128  *  HARDIRQ		- inside an interrupt handler
129  *  SOFTIRQ		- inside a softirq handler
130  */
131 enum trace_flag_type {
132 	TRACE_FLAG_IRQS_OFF		= 0x01,
133 	TRACE_FLAG_IRQS_NOSUPPORT	= 0x02,
134 	TRACE_FLAG_NEED_RESCHED		= 0x04,
135 	TRACE_FLAG_HARDIRQ		= 0x08,
136 	TRACE_FLAG_SOFTIRQ		= 0x10,
137 	TRACE_FLAG_PREEMPT_RESCHED	= 0x20,
138 	TRACE_FLAG_NMI			= 0x40,
139 };
140 
141 #define TRACE_BUF_SIZE		1024
142 
143 struct trace_array;
144 
145 /*
146  * The CPU trace array - it consists of thousands of trace entries
147  * plus some other descriptor data: (for example which task started
148  * the trace, etc.)
149  */
150 struct trace_array_cpu {
151 	atomic_t		disabled;
152 	void			*buffer_page;	/* ring buffer spare */
153 
154 	unsigned long		entries;
155 	unsigned long		saved_latency;
156 	unsigned long		critical_start;
157 	unsigned long		critical_end;
158 	unsigned long		critical_sequence;
159 	unsigned long		nice;
160 	unsigned long		policy;
161 	unsigned long		rt_priority;
162 	unsigned long		skipped_entries;
163 	u64			preempt_timestamp;
164 	pid_t			pid;
165 	kuid_t			uid;
166 	char			comm[TASK_COMM_LEN];
167 
168 	bool			ignore_pid;
169 #ifdef CONFIG_FUNCTION_TRACER
170 	bool			ftrace_ignore_pid;
171 #endif
172 };
173 
174 struct tracer;
175 struct trace_option_dentry;
176 
177 struct trace_buffer {
178 	struct trace_array		*tr;
179 	struct ring_buffer		*buffer;
180 	struct trace_array_cpu __percpu	*data;
181 	u64				time_start;
182 	int				cpu;
183 };
184 
185 #define TRACE_FLAGS_MAX_SIZE		32
186 
187 struct trace_options {
188 	struct tracer			*tracer;
189 	struct trace_option_dentry	*topts;
190 };
191 
192 struct trace_pid_list {
193 	int				pid_max;
194 	unsigned long			*pids;
195 };
196 
197 /*
198  * The trace array - an array of per-CPU trace arrays. This is the
199  * highest level data structure that individual tracers deal with.
200  * They have on/off state as well:
201  */
202 struct trace_array {
203 	struct list_head	list;
204 	char			*name;
205 	struct trace_buffer	trace_buffer;
206 #ifdef CONFIG_TRACER_MAX_TRACE
207 	/*
208 	 * The max_buffer is used to snapshot the trace when a maximum
209 	 * latency is reached, or when the user initiates a snapshot.
210 	 * Some tracers will use this to store a maximum trace while
211 	 * it continues examining live traces.
212 	 *
213 	 * The buffers for the max_buffer are set up the same as the trace_buffer
214 	 * When a snapshot is taken, the buffer of the max_buffer is swapped
215 	 * with the buffer of the trace_buffer and the buffers are reset for
216 	 * the trace_buffer so the tracing can continue.
217 	 */
218 	struct trace_buffer	max_buffer;
219 	bool			allocated_snapshot;
220 #endif
221 #if defined(CONFIG_TRACER_MAX_TRACE) || defined(CONFIG_HWLAT_TRACER)
222 	unsigned long		max_latency;
223 #endif
224 	struct trace_pid_list	__rcu *filtered_pids;
225 	/*
226 	 * max_lock is used to protect the swapping of buffers
227 	 * when taking a max snapshot. The buffers themselves are
228 	 * protected by per_cpu spinlocks. But the action of the swap
229 	 * needs its own lock.
230 	 *
231 	 * This is defined as a arch_spinlock_t in order to help
232 	 * with performance when lockdep debugging is enabled.
233 	 *
234 	 * It is also used in other places outside the update_max_tr
235 	 * so it needs to be defined outside of the
236 	 * CONFIG_TRACER_MAX_TRACE.
237 	 */
238 	arch_spinlock_t		max_lock;
239 	int			buffer_disabled;
240 #ifdef CONFIG_FTRACE_SYSCALLS
241 	int			sys_refcount_enter;
242 	int			sys_refcount_exit;
243 	struct trace_event_file __rcu *enter_syscall_files[NR_syscalls];
244 	struct trace_event_file __rcu *exit_syscall_files[NR_syscalls];
245 #endif
246 	int			stop_count;
247 	int			clock_id;
248 	int			nr_topts;
249 	bool			clear_trace;
250 	struct tracer		*current_trace;
251 	unsigned int		trace_flags;
252 	unsigned char		trace_flags_index[TRACE_FLAGS_MAX_SIZE];
253 	unsigned int		flags;
254 	raw_spinlock_t		start_lock;
255 	struct dentry		*dir;
256 	struct dentry		*options;
257 	struct dentry		*percpu_dir;
258 	struct dentry		*event_dir;
259 	struct trace_options	*topts;
260 	struct list_head	systems;
261 	struct list_head	events;
262 	cpumask_var_t		tracing_cpumask; /* only trace on set CPUs */
263 	int			ref;
264 #ifdef CONFIG_FUNCTION_TRACER
265 	struct ftrace_ops	*ops;
266 	struct trace_pid_list	__rcu *function_pids;
267 #ifdef CONFIG_DYNAMIC_FTRACE
268 	/* All of these are protected by the ftrace_lock */
269 	struct list_head	func_probes;
270 	struct list_head	mod_trace;
271 	struct list_head	mod_notrace;
272 #endif
273 	/* function tracing enabled */
274 	int			function_enabled;
275 #endif
276 };
277 
278 enum {
279 	TRACE_ARRAY_FL_GLOBAL	= (1 << 0)
280 };
281 
282 extern struct list_head ftrace_trace_arrays;
283 
284 extern struct mutex trace_types_lock;
285 
286 extern int trace_array_get(struct trace_array *tr);
287 extern void trace_array_put(struct trace_array *tr);
288 
289 /*
290  * The global tracer (top) should be the first trace array added,
291  * but we check the flag anyway.
292  */
293 static inline struct trace_array *top_trace_array(void)
294 {
295 	struct trace_array *tr;
296 
297 	if (list_empty(&ftrace_trace_arrays))
298 		return NULL;
299 
300 	tr = list_entry(ftrace_trace_arrays.prev,
301 			typeof(*tr), list);
302 	WARN_ON(!(tr->flags & TRACE_ARRAY_FL_GLOBAL));
303 	return tr;
304 }
305 
306 #define FTRACE_CMP_TYPE(var, type) \
307 	__builtin_types_compatible_p(typeof(var), type *)
308 
309 #undef IF_ASSIGN
310 #define IF_ASSIGN(var, entry, etype, id)		\
311 	if (FTRACE_CMP_TYPE(var, etype)) {		\
312 		var = (typeof(var))(entry);		\
313 		WARN_ON(id && (entry)->type != id);	\
314 		break;					\
315 	}
316 
317 /* Will cause compile errors if type is not found. */
318 extern void __ftrace_bad_type(void);
319 
320 /*
321  * The trace_assign_type is a verifier that the entry type is
322  * the same as the type being assigned. To add new types simply
323  * add a line with the following format:
324  *
325  * IF_ASSIGN(var, ent, type, id);
326  *
327  *  Where "type" is the trace type that includes the trace_entry
328  *  as the "ent" item. And "id" is the trace identifier that is
329  *  used in the trace_type enum.
330  *
331  *  If the type can have more than one id, then use zero.
332  */
333 #define trace_assign_type(var, ent)					\
334 	do {								\
335 		IF_ASSIGN(var, ent, struct ftrace_entry, TRACE_FN);	\
336 		IF_ASSIGN(var, ent, struct ctx_switch_entry, 0);	\
337 		IF_ASSIGN(var, ent, struct stack_entry, TRACE_STACK);	\
338 		IF_ASSIGN(var, ent, struct userstack_entry, TRACE_USER_STACK);\
339 		IF_ASSIGN(var, ent, struct print_entry, TRACE_PRINT);	\
340 		IF_ASSIGN(var, ent, struct bprint_entry, TRACE_BPRINT);	\
341 		IF_ASSIGN(var, ent, struct bputs_entry, TRACE_BPUTS);	\
342 		IF_ASSIGN(var, ent, struct hwlat_entry, TRACE_HWLAT);	\
343 		IF_ASSIGN(var, ent, struct raw_data_entry, TRACE_RAW_DATA);\
344 		IF_ASSIGN(var, ent, struct trace_mmiotrace_rw,		\
345 			  TRACE_MMIO_RW);				\
346 		IF_ASSIGN(var, ent, struct trace_mmiotrace_map,		\
347 			  TRACE_MMIO_MAP);				\
348 		IF_ASSIGN(var, ent, struct trace_branch, TRACE_BRANCH); \
349 		IF_ASSIGN(var, ent, struct ftrace_graph_ent_entry,	\
350 			  TRACE_GRAPH_ENT);		\
351 		IF_ASSIGN(var, ent, struct ftrace_graph_ret_entry,	\
352 			  TRACE_GRAPH_RET);		\
353 		__ftrace_bad_type();					\
354 	} while (0)
355 
356 /*
357  * An option specific to a tracer. This is a boolean value.
358  * The bit is the bit index that sets its value on the
359  * flags value in struct tracer_flags.
360  */
361 struct tracer_opt {
362 	const char	*name; /* Will appear on the trace_options file */
363 	u32		bit; /* Mask assigned in val field in tracer_flags */
364 };
365 
366 /*
367  * The set of specific options for a tracer. Your tracer
368  * have to set the initial value of the flags val.
369  */
370 struct tracer_flags {
371 	u32			val;
372 	struct tracer_opt	*opts;
373 	struct tracer		*trace;
374 };
375 
376 /* Makes more easy to define a tracer opt */
377 #define TRACER_OPT(s, b)	.name = #s, .bit = b
378 
379 
380 struct trace_option_dentry {
381 	struct tracer_opt		*opt;
382 	struct tracer_flags		*flags;
383 	struct trace_array		*tr;
384 	struct dentry			*entry;
385 };
386 
387 /**
388  * struct tracer - a specific tracer and its callbacks to interact with tracefs
389  * @name: the name chosen to select it on the available_tracers file
390  * @init: called when one switches to this tracer (echo name > current_tracer)
391  * @reset: called when one switches to another tracer
392  * @start: called when tracing is unpaused (echo 1 > tracing_on)
393  * @stop: called when tracing is paused (echo 0 > tracing_on)
394  * @update_thresh: called when tracing_thresh is updated
395  * @open: called when the trace file is opened
396  * @pipe_open: called when the trace_pipe file is opened
397  * @close: called when the trace file is released
398  * @pipe_close: called when the trace_pipe file is released
399  * @read: override the default read callback on trace_pipe
400  * @splice_read: override the default splice_read callback on trace_pipe
401  * @selftest: selftest to run on boot (see trace_selftest.c)
402  * @print_headers: override the first lines that describe your columns
403  * @print_line: callback that prints a trace
404  * @set_flag: signals one of your private flags changed (trace_options file)
405  * @flags: your private flags
406  */
407 struct tracer {
408 	const char		*name;
409 	int			(*init)(struct trace_array *tr);
410 	void			(*reset)(struct trace_array *tr);
411 	void			(*start)(struct trace_array *tr);
412 	void			(*stop)(struct trace_array *tr);
413 	int			(*update_thresh)(struct trace_array *tr);
414 	void			(*open)(struct trace_iterator *iter);
415 	void			(*pipe_open)(struct trace_iterator *iter);
416 	void			(*close)(struct trace_iterator *iter);
417 	void			(*pipe_close)(struct trace_iterator *iter);
418 	ssize_t			(*read)(struct trace_iterator *iter,
419 					struct file *filp, char __user *ubuf,
420 					size_t cnt, loff_t *ppos);
421 	ssize_t			(*splice_read)(struct trace_iterator *iter,
422 					       struct file *filp,
423 					       loff_t *ppos,
424 					       struct pipe_inode_info *pipe,
425 					       size_t len,
426 					       unsigned int flags);
427 #ifdef CONFIG_FTRACE_STARTUP_TEST
428 	int			(*selftest)(struct tracer *trace,
429 					    struct trace_array *tr);
430 #endif
431 	void			(*print_header)(struct seq_file *m);
432 	enum print_line_t	(*print_line)(struct trace_iterator *iter);
433 	/* If you handled the flag setting, return 0 */
434 	int			(*set_flag)(struct trace_array *tr,
435 					    u32 old_flags, u32 bit, int set);
436 	/* Return 0 if OK with change, else return non-zero */
437 	int			(*flag_changed)(struct trace_array *tr,
438 						u32 mask, int set);
439 	struct tracer		*next;
440 	struct tracer_flags	*flags;
441 	int			enabled;
442 	int			ref;
443 	bool			print_max;
444 	bool			allow_instances;
445 #ifdef CONFIG_TRACER_MAX_TRACE
446 	bool			use_max_tr;
447 #endif
448 	/* True if tracer cannot be enabled in kernel param */
449 	bool			noboot;
450 };
451 
452 
453 /* Only current can touch trace_recursion */
454 
455 /*
456  * For function tracing recursion:
457  *  The order of these bits are important.
458  *
459  *  When function tracing occurs, the following steps are made:
460  *   If arch does not support a ftrace feature:
461  *    call internal function (uses INTERNAL bits) which calls...
462  *   If callback is registered to the "global" list, the list
463  *    function is called and recursion checks the GLOBAL bits.
464  *    then this function calls...
465  *   The function callback, which can use the FTRACE bits to
466  *    check for recursion.
467  *
468  * Now if the arch does not suppport a feature, and it calls
469  * the global list function which calls the ftrace callback
470  * all three of these steps will do a recursion protection.
471  * There's no reason to do one if the previous caller already
472  * did. The recursion that we are protecting against will
473  * go through the same steps again.
474  *
475  * To prevent the multiple recursion checks, if a recursion
476  * bit is set that is higher than the MAX bit of the current
477  * check, then we know that the check was made by the previous
478  * caller, and we can skip the current check.
479  */
480 enum {
481 	TRACE_BUFFER_BIT,
482 	TRACE_BUFFER_NMI_BIT,
483 	TRACE_BUFFER_IRQ_BIT,
484 	TRACE_BUFFER_SIRQ_BIT,
485 
486 	/* Start of function recursion bits */
487 	TRACE_FTRACE_BIT,
488 	TRACE_FTRACE_NMI_BIT,
489 	TRACE_FTRACE_IRQ_BIT,
490 	TRACE_FTRACE_SIRQ_BIT,
491 
492 	/* INTERNAL_BITs must be greater than FTRACE_BITs */
493 	TRACE_INTERNAL_BIT,
494 	TRACE_INTERNAL_NMI_BIT,
495 	TRACE_INTERNAL_IRQ_BIT,
496 	TRACE_INTERNAL_SIRQ_BIT,
497 
498 	TRACE_BRANCH_BIT,
499 /*
500  * Abuse of the trace_recursion.
501  * As we need a way to maintain state if we are tracing the function
502  * graph in irq because we want to trace a particular function that
503  * was called in irq context but we have irq tracing off. Since this
504  * can only be modified by current, we can reuse trace_recursion.
505  */
506 	TRACE_IRQ_BIT,
507 };
508 
509 #define trace_recursion_set(bit)	do { (current)->trace_recursion |= (1<<(bit)); } while (0)
510 #define trace_recursion_clear(bit)	do { (current)->trace_recursion &= ~(1<<(bit)); } while (0)
511 #define trace_recursion_test(bit)	((current)->trace_recursion & (1<<(bit)))
512 
513 #define TRACE_CONTEXT_BITS	4
514 
515 #define TRACE_FTRACE_START	TRACE_FTRACE_BIT
516 #define TRACE_FTRACE_MAX	((1 << (TRACE_FTRACE_START + TRACE_CONTEXT_BITS)) - 1)
517 
518 #define TRACE_LIST_START	TRACE_INTERNAL_BIT
519 #define TRACE_LIST_MAX		((1 << (TRACE_LIST_START + TRACE_CONTEXT_BITS)) - 1)
520 
521 #define TRACE_CONTEXT_MASK	TRACE_LIST_MAX
522 
523 static __always_inline int trace_get_context_bit(void)
524 {
525 	int bit;
526 
527 	if (in_interrupt()) {
528 		if (in_nmi())
529 			bit = 0;
530 
531 		else if (in_irq())
532 			bit = 1;
533 		else
534 			bit = 2;
535 	} else
536 		bit = 3;
537 
538 	return bit;
539 }
540 
541 static __always_inline int trace_test_and_set_recursion(int start, int max)
542 {
543 	unsigned int val = current->trace_recursion;
544 	int bit;
545 
546 	/* A previous recursion check was made */
547 	if ((val & TRACE_CONTEXT_MASK) > max)
548 		return 0;
549 
550 	bit = trace_get_context_bit() + start;
551 	if (unlikely(val & (1 << bit)))
552 		return -1;
553 
554 	val |= 1 << bit;
555 	current->trace_recursion = val;
556 	barrier();
557 
558 	return bit;
559 }
560 
561 static __always_inline void trace_clear_recursion(int bit)
562 {
563 	unsigned int val = current->trace_recursion;
564 
565 	if (!bit)
566 		return;
567 
568 	bit = 1 << bit;
569 	val &= ~bit;
570 
571 	barrier();
572 	current->trace_recursion = val;
573 }
574 
575 static inline struct ring_buffer_iter *
576 trace_buffer_iter(struct trace_iterator *iter, int cpu)
577 {
578 	if (iter->buffer_iter && iter->buffer_iter[cpu])
579 		return iter->buffer_iter[cpu];
580 	return NULL;
581 }
582 
583 int tracer_init(struct tracer *t, struct trace_array *tr);
584 int tracing_is_enabled(void);
585 void tracing_reset(struct trace_buffer *buf, int cpu);
586 void tracing_reset_online_cpus(struct trace_buffer *buf);
587 void tracing_reset_current(int cpu);
588 void tracing_reset_all_online_cpus(void);
589 int tracing_open_generic(struct inode *inode, struct file *filp);
590 bool tracing_is_disabled(void);
591 int tracer_tracing_is_on(struct trace_array *tr);
592 void tracer_tracing_on(struct trace_array *tr);
593 void tracer_tracing_off(struct trace_array *tr);
594 struct dentry *trace_create_file(const char *name,
595 				 umode_t mode,
596 				 struct dentry *parent,
597 				 void *data,
598 				 const struct file_operations *fops);
599 
600 struct dentry *tracing_init_dentry(void);
601 
602 struct ring_buffer_event;
603 
604 struct ring_buffer_event *
605 trace_buffer_lock_reserve(struct ring_buffer *buffer,
606 			  int type,
607 			  unsigned long len,
608 			  unsigned long flags,
609 			  int pc);
610 
611 struct trace_entry *tracing_get_trace_entry(struct trace_array *tr,
612 						struct trace_array_cpu *data);
613 
614 struct trace_entry *trace_find_next_entry(struct trace_iterator *iter,
615 					  int *ent_cpu, u64 *ent_ts);
616 
617 void trace_buffer_unlock_commit_nostack(struct ring_buffer *buffer,
618 					struct ring_buffer_event *event);
619 
620 int trace_empty(struct trace_iterator *iter);
621 
622 void *trace_find_next_entry_inc(struct trace_iterator *iter);
623 
624 void trace_init_global_iter(struct trace_iterator *iter);
625 
626 void tracing_iter_reset(struct trace_iterator *iter, int cpu);
627 
628 void trace_function(struct trace_array *tr,
629 		    unsigned long ip,
630 		    unsigned long parent_ip,
631 		    unsigned long flags, int pc);
632 void trace_graph_function(struct trace_array *tr,
633 		    unsigned long ip,
634 		    unsigned long parent_ip,
635 		    unsigned long flags, int pc);
636 void trace_latency_header(struct seq_file *m);
637 void trace_default_header(struct seq_file *m);
638 void print_trace_header(struct seq_file *m, struct trace_iterator *iter);
639 int trace_empty(struct trace_iterator *iter);
640 
641 void trace_graph_return(struct ftrace_graph_ret *trace);
642 int trace_graph_entry(struct ftrace_graph_ent *trace);
643 void set_graph_array(struct trace_array *tr);
644 
645 void tracing_start_cmdline_record(void);
646 void tracing_stop_cmdline_record(void);
647 void tracing_start_tgid_record(void);
648 void tracing_stop_tgid_record(void);
649 
650 int register_tracer(struct tracer *type);
651 int is_tracing_stopped(void);
652 
653 loff_t tracing_lseek(struct file *file, loff_t offset, int whence);
654 
655 extern cpumask_var_t __read_mostly tracing_buffer_mask;
656 
657 #define for_each_tracing_cpu(cpu)	\
658 	for_each_cpu(cpu, tracing_buffer_mask)
659 
660 extern unsigned long nsecs_to_usecs(unsigned long nsecs);
661 
662 extern unsigned long tracing_thresh;
663 
664 /* PID filtering */
665 
666 extern int pid_max;
667 
668 bool trace_find_filtered_pid(struct trace_pid_list *filtered_pids,
669 			     pid_t search_pid);
670 bool trace_ignore_this_task(struct trace_pid_list *filtered_pids,
671 			    struct task_struct *task);
672 void trace_filter_add_remove_task(struct trace_pid_list *pid_list,
673 				  struct task_struct *self,
674 				  struct task_struct *task);
675 void *trace_pid_next(struct trace_pid_list *pid_list, void *v, loff_t *pos);
676 void *trace_pid_start(struct trace_pid_list *pid_list, loff_t *pos);
677 int trace_pid_show(struct seq_file *m, void *v);
678 void trace_free_pid_list(struct trace_pid_list *pid_list);
679 int trace_pid_write(struct trace_pid_list *filtered_pids,
680 		    struct trace_pid_list **new_pid_list,
681 		    const char __user *ubuf, size_t cnt);
682 
683 #ifdef CONFIG_TRACER_MAX_TRACE
684 void update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu);
685 void update_max_tr_single(struct trace_array *tr,
686 			  struct task_struct *tsk, int cpu);
687 #endif /* CONFIG_TRACER_MAX_TRACE */
688 
689 #ifdef CONFIG_STACKTRACE
690 void ftrace_trace_userstack(struct ring_buffer *buffer, unsigned long flags,
691 			    int pc);
692 
693 void __trace_stack(struct trace_array *tr, unsigned long flags, int skip,
694 		   int pc);
695 #else
696 static inline void ftrace_trace_userstack(struct ring_buffer *buffer,
697 					  unsigned long flags, int pc)
698 {
699 }
700 
701 static inline void __trace_stack(struct trace_array *tr, unsigned long flags,
702 				 int skip, int pc)
703 {
704 }
705 #endif /* CONFIG_STACKTRACE */
706 
707 extern u64 ftrace_now(int cpu);
708 
709 extern void trace_find_cmdline(int pid, char comm[]);
710 extern int trace_find_tgid(int pid);
711 extern void trace_event_follow_fork(struct trace_array *tr, bool enable);
712 
713 #ifdef CONFIG_DYNAMIC_FTRACE
714 extern unsigned long ftrace_update_tot_cnt;
715 void ftrace_init_trace_array(struct trace_array *tr);
716 #else
717 static inline void ftrace_init_trace_array(struct trace_array *tr) { }
718 #endif
719 #define DYN_FTRACE_TEST_NAME trace_selftest_dynamic_test_func
720 extern int DYN_FTRACE_TEST_NAME(void);
721 #define DYN_FTRACE_TEST_NAME2 trace_selftest_dynamic_test_func2
722 extern int DYN_FTRACE_TEST_NAME2(void);
723 
724 extern bool ring_buffer_expanded;
725 extern bool tracing_selftest_disabled;
726 
727 #ifdef CONFIG_FTRACE_STARTUP_TEST
728 extern int trace_selftest_startup_function(struct tracer *trace,
729 					   struct trace_array *tr);
730 extern int trace_selftest_startup_function_graph(struct tracer *trace,
731 						 struct trace_array *tr);
732 extern int trace_selftest_startup_irqsoff(struct tracer *trace,
733 					  struct trace_array *tr);
734 extern int trace_selftest_startup_preemptoff(struct tracer *trace,
735 					     struct trace_array *tr);
736 extern int trace_selftest_startup_preemptirqsoff(struct tracer *trace,
737 						 struct trace_array *tr);
738 extern int trace_selftest_startup_wakeup(struct tracer *trace,
739 					 struct trace_array *tr);
740 extern int trace_selftest_startup_nop(struct tracer *trace,
741 					 struct trace_array *tr);
742 extern int trace_selftest_startup_branch(struct tracer *trace,
743 					 struct trace_array *tr);
744 /*
745  * Tracer data references selftest functions that only occur
746  * on boot up. These can be __init functions. Thus, when selftests
747  * are enabled, then the tracers need to reference __init functions.
748  */
749 #define __tracer_data		__refdata
750 #else
751 /* Tracers are seldom changed. Optimize when selftests are disabled. */
752 #define __tracer_data		__read_mostly
753 #endif /* CONFIG_FTRACE_STARTUP_TEST */
754 
755 extern void *head_page(struct trace_array_cpu *data);
756 extern unsigned long long ns2usecs(u64 nsec);
757 extern int
758 trace_vbprintk(unsigned long ip, const char *fmt, va_list args);
759 extern int
760 trace_vprintk(unsigned long ip, const char *fmt, va_list args);
761 extern int
762 trace_array_vprintk(struct trace_array *tr,
763 		    unsigned long ip, const char *fmt, va_list args);
764 int trace_array_printk(struct trace_array *tr,
765 		       unsigned long ip, const char *fmt, ...);
766 int trace_array_printk_buf(struct ring_buffer *buffer,
767 			   unsigned long ip, const char *fmt, ...);
768 void trace_printk_seq(struct trace_seq *s);
769 enum print_line_t print_trace_line(struct trace_iterator *iter);
770 
771 extern char trace_find_mark(unsigned long long duration);
772 
773 struct ftrace_hash;
774 
775 struct ftrace_mod_load {
776 	struct list_head	list;
777 	char			*func;
778 	char			*module;
779 	int			 enable;
780 };
781 
782 enum {
783 	FTRACE_HASH_FL_MOD	= (1 << 0),
784 };
785 
786 struct ftrace_hash {
787 	unsigned long		size_bits;
788 	struct hlist_head	*buckets;
789 	unsigned long		count;
790 	unsigned long		flags;
791 	struct rcu_head		rcu;
792 };
793 
794 struct ftrace_func_entry *
795 ftrace_lookup_ip(struct ftrace_hash *hash, unsigned long ip);
796 
797 static __always_inline bool ftrace_hash_empty(struct ftrace_hash *hash)
798 {
799 	return !hash || !(hash->count || (hash->flags & FTRACE_HASH_FL_MOD));
800 }
801 
802 /* Standard output formatting function used for function return traces */
803 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
804 
805 /* Flag options */
806 #define TRACE_GRAPH_PRINT_OVERRUN       0x1
807 #define TRACE_GRAPH_PRINT_CPU           0x2
808 #define TRACE_GRAPH_PRINT_OVERHEAD      0x4
809 #define TRACE_GRAPH_PRINT_PROC          0x8
810 #define TRACE_GRAPH_PRINT_DURATION      0x10
811 #define TRACE_GRAPH_PRINT_ABS_TIME      0x20
812 #define TRACE_GRAPH_PRINT_IRQS          0x40
813 #define TRACE_GRAPH_PRINT_TAIL          0x80
814 #define TRACE_GRAPH_SLEEP_TIME		0x100
815 #define TRACE_GRAPH_GRAPH_TIME		0x200
816 #define TRACE_GRAPH_PRINT_FILL_SHIFT	28
817 #define TRACE_GRAPH_PRINT_FILL_MASK	(0x3 << TRACE_GRAPH_PRINT_FILL_SHIFT)
818 
819 extern void ftrace_graph_sleep_time_control(bool enable);
820 extern void ftrace_graph_graph_time_control(bool enable);
821 
822 extern enum print_line_t
823 print_graph_function_flags(struct trace_iterator *iter, u32 flags);
824 extern void print_graph_headers_flags(struct seq_file *s, u32 flags);
825 extern void
826 trace_print_graph_duration(unsigned long long duration, struct trace_seq *s);
827 extern void graph_trace_open(struct trace_iterator *iter);
828 extern void graph_trace_close(struct trace_iterator *iter);
829 extern int __trace_graph_entry(struct trace_array *tr,
830 			       struct ftrace_graph_ent *trace,
831 			       unsigned long flags, int pc);
832 extern void __trace_graph_return(struct trace_array *tr,
833 				 struct ftrace_graph_ret *trace,
834 				 unsigned long flags, int pc);
835 
836 #ifdef CONFIG_DYNAMIC_FTRACE
837 extern struct ftrace_hash *ftrace_graph_hash;
838 extern struct ftrace_hash *ftrace_graph_notrace_hash;
839 
840 static inline int ftrace_graph_addr(unsigned long addr)
841 {
842 	int ret = 0;
843 
844 	preempt_disable_notrace();
845 
846 	if (ftrace_hash_empty(ftrace_graph_hash)) {
847 		ret = 1;
848 		goto out;
849 	}
850 
851 	if (ftrace_lookup_ip(ftrace_graph_hash, addr)) {
852 		/*
853 		 * If no irqs are to be traced, but a set_graph_function
854 		 * is set, and called by an interrupt handler, we still
855 		 * want to trace it.
856 		 */
857 		if (in_irq())
858 			trace_recursion_set(TRACE_IRQ_BIT);
859 		else
860 			trace_recursion_clear(TRACE_IRQ_BIT);
861 		ret = 1;
862 	}
863 
864 out:
865 	preempt_enable_notrace();
866 	return ret;
867 }
868 
869 static inline int ftrace_graph_notrace_addr(unsigned long addr)
870 {
871 	int ret = 0;
872 
873 	preempt_disable_notrace();
874 
875 	if (ftrace_lookup_ip(ftrace_graph_notrace_hash, addr))
876 		ret = 1;
877 
878 	preempt_enable_notrace();
879 	return ret;
880 }
881 #else
882 static inline int ftrace_graph_addr(unsigned long addr)
883 {
884 	return 1;
885 }
886 
887 static inline int ftrace_graph_notrace_addr(unsigned long addr)
888 {
889 	return 0;
890 }
891 #endif /* CONFIG_DYNAMIC_FTRACE */
892 
893 extern unsigned int fgraph_max_depth;
894 
895 static inline bool ftrace_graph_ignore_func(struct ftrace_graph_ent *trace)
896 {
897 	/* trace it when it is-nested-in or is a function enabled. */
898 	return !(trace->depth || ftrace_graph_addr(trace->func)) ||
899 		(trace->depth < 0) ||
900 		(fgraph_max_depth && trace->depth >= fgraph_max_depth);
901 }
902 
903 #else /* CONFIG_FUNCTION_GRAPH_TRACER */
904 static inline enum print_line_t
905 print_graph_function_flags(struct trace_iterator *iter, u32 flags)
906 {
907 	return TRACE_TYPE_UNHANDLED;
908 }
909 #endif /* CONFIG_FUNCTION_GRAPH_TRACER */
910 
911 extern struct list_head ftrace_pids;
912 
913 #ifdef CONFIG_FUNCTION_TRACER
914 struct ftrace_func_command {
915 	struct list_head	list;
916 	char			*name;
917 	int			(*func)(struct trace_array *tr,
918 					struct ftrace_hash *hash,
919 					char *func, char *cmd,
920 					char *params, int enable);
921 };
922 extern bool ftrace_filter_param __initdata;
923 static inline int ftrace_trace_task(struct trace_array *tr)
924 {
925 	return !this_cpu_read(tr->trace_buffer.data->ftrace_ignore_pid);
926 }
927 extern int ftrace_is_dead(void);
928 int ftrace_create_function_files(struct trace_array *tr,
929 				 struct dentry *parent);
930 void ftrace_destroy_function_files(struct trace_array *tr);
931 void ftrace_init_global_array_ops(struct trace_array *tr);
932 void ftrace_init_array_ops(struct trace_array *tr, ftrace_func_t func);
933 void ftrace_reset_array_ops(struct trace_array *tr);
934 int using_ftrace_ops_list_func(void);
935 void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d_tracer);
936 void ftrace_init_tracefs_toplevel(struct trace_array *tr,
937 				  struct dentry *d_tracer);
938 void ftrace_clear_pids(struct trace_array *tr);
939 int init_function_trace(void);
940 void ftrace_pid_follow_fork(struct trace_array *tr, bool enable);
941 #else
942 static inline int ftrace_trace_task(struct trace_array *tr)
943 {
944 	return 1;
945 }
946 static inline int ftrace_is_dead(void) { return 0; }
947 static inline int
948 ftrace_create_function_files(struct trace_array *tr,
949 			     struct dentry *parent)
950 {
951 	return 0;
952 }
953 static inline void ftrace_destroy_function_files(struct trace_array *tr) { }
954 static inline __init void
955 ftrace_init_global_array_ops(struct trace_array *tr) { }
956 static inline void ftrace_reset_array_ops(struct trace_array *tr) { }
957 static inline void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d) { }
958 static inline void ftrace_init_tracefs_toplevel(struct trace_array *tr, struct dentry *d) { }
959 static inline void ftrace_clear_pids(struct trace_array *tr) { }
960 static inline int init_function_trace(void) { return 0; }
961 static inline void ftrace_pid_follow_fork(struct trace_array *tr, bool enable) { }
962 /* ftace_func_t type is not defined, use macro instead of static inline */
963 #define ftrace_init_array_ops(tr, func) do { } while (0)
964 #endif /* CONFIG_FUNCTION_TRACER */
965 
966 #if defined(CONFIG_FUNCTION_TRACER) && defined(CONFIG_DYNAMIC_FTRACE)
967 
968 struct ftrace_probe_ops {
969 	void			(*func)(unsigned long ip,
970 					unsigned long parent_ip,
971 					struct trace_array *tr,
972 					struct ftrace_probe_ops *ops,
973 					void *data);
974 	int			(*init)(struct ftrace_probe_ops *ops,
975 					struct trace_array *tr,
976 					unsigned long ip, void *init_data,
977 					void **data);
978 	void			(*free)(struct ftrace_probe_ops *ops,
979 					struct trace_array *tr,
980 					unsigned long ip, void *data);
981 	int			(*print)(struct seq_file *m,
982 					 unsigned long ip,
983 					 struct ftrace_probe_ops *ops,
984 					 void *data);
985 };
986 
987 struct ftrace_func_mapper;
988 typedef int (*ftrace_mapper_func)(void *data);
989 
990 struct ftrace_func_mapper *allocate_ftrace_func_mapper(void);
991 void **ftrace_func_mapper_find_ip(struct ftrace_func_mapper *mapper,
992 					   unsigned long ip);
993 int ftrace_func_mapper_add_ip(struct ftrace_func_mapper *mapper,
994 			       unsigned long ip, void *data);
995 void *ftrace_func_mapper_remove_ip(struct ftrace_func_mapper *mapper,
996 				   unsigned long ip);
997 void free_ftrace_func_mapper(struct ftrace_func_mapper *mapper,
998 			     ftrace_mapper_func free_func);
999 
1000 extern int
1001 register_ftrace_function_probe(char *glob, struct trace_array *tr,
1002 			       struct ftrace_probe_ops *ops, void *data);
1003 extern int
1004 unregister_ftrace_function_probe_func(char *glob, struct trace_array *tr,
1005 				      struct ftrace_probe_ops *ops);
1006 extern void clear_ftrace_function_probes(struct trace_array *tr);
1007 
1008 int register_ftrace_command(struct ftrace_func_command *cmd);
1009 int unregister_ftrace_command(struct ftrace_func_command *cmd);
1010 
1011 void ftrace_create_filter_files(struct ftrace_ops *ops,
1012 				struct dentry *parent);
1013 void ftrace_destroy_filter_files(struct ftrace_ops *ops);
1014 #else
1015 struct ftrace_func_command;
1016 
1017 static inline __init int register_ftrace_command(struct ftrace_func_command *cmd)
1018 {
1019 	return -EINVAL;
1020 }
1021 static inline __init int unregister_ftrace_command(char *cmd_name)
1022 {
1023 	return -EINVAL;
1024 }
1025 static inline void clear_ftrace_function_probes(struct trace_array *tr)
1026 {
1027 }
1028 
1029 /*
1030  * The ops parameter passed in is usually undefined.
1031  * This must be a macro.
1032  */
1033 #define ftrace_create_filter_files(ops, parent) do { } while (0)
1034 #define ftrace_destroy_filter_files(ops) do { } while (0)
1035 #endif /* CONFIG_FUNCTION_TRACER && CONFIG_DYNAMIC_FTRACE */
1036 
1037 bool ftrace_event_is_function(struct trace_event_call *call);
1038 
1039 /*
1040  * struct trace_parser - servers for reading the user input separated by spaces
1041  * @cont: set if the input is not complete - no final space char was found
1042  * @buffer: holds the parsed user input
1043  * @idx: user input length
1044  * @size: buffer size
1045  */
1046 struct trace_parser {
1047 	bool		cont;
1048 	char		*buffer;
1049 	unsigned	idx;
1050 	unsigned	size;
1051 };
1052 
1053 static inline bool trace_parser_loaded(struct trace_parser *parser)
1054 {
1055 	return (parser->idx != 0);
1056 }
1057 
1058 static inline bool trace_parser_cont(struct trace_parser *parser)
1059 {
1060 	return parser->cont;
1061 }
1062 
1063 static inline void trace_parser_clear(struct trace_parser *parser)
1064 {
1065 	parser->cont = false;
1066 	parser->idx = 0;
1067 }
1068 
1069 extern int trace_parser_get_init(struct trace_parser *parser, int size);
1070 extern void trace_parser_put(struct trace_parser *parser);
1071 extern int trace_get_user(struct trace_parser *parser, const char __user *ubuf,
1072 	size_t cnt, loff_t *ppos);
1073 
1074 /*
1075  * Only create function graph options if function graph is configured.
1076  */
1077 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1078 # define FGRAPH_FLAGS						\
1079 		C(DISPLAY_GRAPH,	"display-graph"),
1080 #else
1081 # define FGRAPH_FLAGS
1082 #endif
1083 
1084 #ifdef CONFIG_BRANCH_TRACER
1085 # define BRANCH_FLAGS					\
1086 		C(BRANCH,		"branch"),
1087 #else
1088 # define BRANCH_FLAGS
1089 #endif
1090 
1091 #ifdef CONFIG_FUNCTION_TRACER
1092 # define FUNCTION_FLAGS						\
1093 		C(FUNCTION,		"function-trace"),	\
1094 		C(FUNC_FORK,		"function-fork"),
1095 # define FUNCTION_DEFAULT_FLAGS		TRACE_ITER_FUNCTION
1096 #else
1097 # define FUNCTION_FLAGS
1098 # define FUNCTION_DEFAULT_FLAGS		0UL
1099 # define TRACE_ITER_FUNC_FORK		0UL
1100 #endif
1101 
1102 #ifdef CONFIG_STACKTRACE
1103 # define STACK_FLAGS				\
1104 		C(STACKTRACE,		"stacktrace"),
1105 #else
1106 # define STACK_FLAGS
1107 #endif
1108 
1109 /*
1110  * trace_iterator_flags is an enumeration that defines bit
1111  * positions into trace_flags that controls the output.
1112  *
1113  * NOTE: These bits must match the trace_options array in
1114  *       trace.c (this macro guarantees it).
1115  */
1116 #define TRACE_FLAGS						\
1117 		C(PRINT_PARENT,		"print-parent"),	\
1118 		C(SYM_OFFSET,		"sym-offset"),		\
1119 		C(SYM_ADDR,		"sym-addr"),		\
1120 		C(VERBOSE,		"verbose"),		\
1121 		C(RAW,			"raw"),			\
1122 		C(HEX,			"hex"),			\
1123 		C(BIN,			"bin"),			\
1124 		C(BLOCK,		"block"),		\
1125 		C(PRINTK,		"trace_printk"),	\
1126 		C(ANNOTATE,		"annotate"),		\
1127 		C(USERSTACKTRACE,	"userstacktrace"),	\
1128 		C(SYM_USEROBJ,		"sym-userobj"),		\
1129 		C(PRINTK_MSGONLY,	"printk-msg-only"),	\
1130 		C(CONTEXT_INFO,		"context-info"),   /* Print pid/cpu/time */ \
1131 		C(LATENCY_FMT,		"latency-format"),	\
1132 		C(RECORD_CMD,		"record-cmd"),		\
1133 		C(RECORD_TGID,		"record-tgid"),		\
1134 		C(OVERWRITE,		"overwrite"),		\
1135 		C(STOP_ON_FREE,		"disable_on_free"),	\
1136 		C(IRQ_INFO,		"irq-info"),		\
1137 		C(MARKERS,		"markers"),		\
1138 		C(EVENT_FORK,		"event-fork"),		\
1139 		FUNCTION_FLAGS					\
1140 		FGRAPH_FLAGS					\
1141 		STACK_FLAGS					\
1142 		BRANCH_FLAGS
1143 
1144 /*
1145  * By defining C, we can make TRACE_FLAGS a list of bit names
1146  * that will define the bits for the flag masks.
1147  */
1148 #undef C
1149 #define C(a, b) TRACE_ITER_##a##_BIT
1150 
1151 enum trace_iterator_bits {
1152 	TRACE_FLAGS
1153 	/* Make sure we don't go more than we have bits for */
1154 	TRACE_ITER_LAST_BIT
1155 };
1156 
1157 /*
1158  * By redefining C, we can make TRACE_FLAGS a list of masks that
1159  * use the bits as defined above.
1160  */
1161 #undef C
1162 #define C(a, b) TRACE_ITER_##a = (1 << TRACE_ITER_##a##_BIT)
1163 
1164 enum trace_iterator_flags { TRACE_FLAGS };
1165 
1166 /*
1167  * TRACE_ITER_SYM_MASK masks the options in trace_flags that
1168  * control the output of kernel symbols.
1169  */
1170 #define TRACE_ITER_SYM_MASK \
1171 	(TRACE_ITER_PRINT_PARENT|TRACE_ITER_SYM_OFFSET|TRACE_ITER_SYM_ADDR)
1172 
1173 extern struct tracer nop_trace;
1174 
1175 #ifdef CONFIG_BRANCH_TRACER
1176 extern int enable_branch_tracing(struct trace_array *tr);
1177 extern void disable_branch_tracing(void);
1178 static inline int trace_branch_enable(struct trace_array *tr)
1179 {
1180 	if (tr->trace_flags & TRACE_ITER_BRANCH)
1181 		return enable_branch_tracing(tr);
1182 	return 0;
1183 }
1184 static inline void trace_branch_disable(void)
1185 {
1186 	/* due to races, always disable */
1187 	disable_branch_tracing();
1188 }
1189 #else
1190 static inline int trace_branch_enable(struct trace_array *tr)
1191 {
1192 	return 0;
1193 }
1194 static inline void trace_branch_disable(void)
1195 {
1196 }
1197 #endif /* CONFIG_BRANCH_TRACER */
1198 
1199 /* set ring buffers to default size if not already done so */
1200 int tracing_update_buffers(void);
1201 
1202 struct ftrace_event_field {
1203 	struct list_head	link;
1204 	const char		*name;
1205 	const char		*type;
1206 	int			filter_type;
1207 	int			offset;
1208 	int			size;
1209 	int			is_signed;
1210 };
1211 
1212 struct event_filter {
1213 	int			n_preds;	/* Number assigned */
1214 	int			a_preds;	/* allocated */
1215 	struct filter_pred __rcu	*preds;
1216 	struct filter_pred __rcu	*root;
1217 	char				*filter_string;
1218 };
1219 
1220 struct event_subsystem {
1221 	struct list_head	list;
1222 	const char		*name;
1223 	struct event_filter	*filter;
1224 	int			ref_count;
1225 };
1226 
1227 struct trace_subsystem_dir {
1228 	struct list_head		list;
1229 	struct event_subsystem		*subsystem;
1230 	struct trace_array		*tr;
1231 	struct dentry			*entry;
1232 	int				ref_count;
1233 	int				nr_events;
1234 };
1235 
1236 extern int call_filter_check_discard(struct trace_event_call *call, void *rec,
1237 				     struct ring_buffer *buffer,
1238 				     struct ring_buffer_event *event);
1239 
1240 void trace_buffer_unlock_commit_regs(struct trace_array *tr,
1241 				     struct ring_buffer *buffer,
1242 				     struct ring_buffer_event *event,
1243 				     unsigned long flags, int pc,
1244 				     struct pt_regs *regs);
1245 
1246 static inline void trace_buffer_unlock_commit(struct trace_array *tr,
1247 					      struct ring_buffer *buffer,
1248 					      struct ring_buffer_event *event,
1249 					      unsigned long flags, int pc)
1250 {
1251 	trace_buffer_unlock_commit_regs(tr, buffer, event, flags, pc, NULL);
1252 }
1253 
1254 DECLARE_PER_CPU(struct ring_buffer_event *, trace_buffered_event);
1255 DECLARE_PER_CPU(int, trace_buffered_event_cnt);
1256 void trace_buffered_event_disable(void);
1257 void trace_buffered_event_enable(void);
1258 
1259 static inline void
1260 __trace_event_discard_commit(struct ring_buffer *buffer,
1261 			     struct ring_buffer_event *event)
1262 {
1263 	if (this_cpu_read(trace_buffered_event) == event) {
1264 		/* Simply release the temp buffer */
1265 		this_cpu_dec(trace_buffered_event_cnt);
1266 		return;
1267 	}
1268 	ring_buffer_discard_commit(buffer, event);
1269 }
1270 
1271 /*
1272  * Helper function for event_trigger_unlock_commit{_regs}().
1273  * If there are event triggers attached to this event that requires
1274  * filtering against its fields, then they wil be called as the
1275  * entry already holds the field information of the current event.
1276  *
1277  * It also checks if the event should be discarded or not.
1278  * It is to be discarded if the event is soft disabled and the
1279  * event was only recorded to process triggers, or if the event
1280  * filter is active and this event did not match the filters.
1281  *
1282  * Returns true if the event is discarded, false otherwise.
1283  */
1284 static inline bool
1285 __event_trigger_test_discard(struct trace_event_file *file,
1286 			     struct ring_buffer *buffer,
1287 			     struct ring_buffer_event *event,
1288 			     void *entry,
1289 			     enum event_trigger_type *tt)
1290 {
1291 	unsigned long eflags = file->flags;
1292 
1293 	if (eflags & EVENT_FILE_FL_TRIGGER_COND)
1294 		*tt = event_triggers_call(file, entry);
1295 
1296 	if (test_bit(EVENT_FILE_FL_SOFT_DISABLED_BIT, &file->flags) ||
1297 	    (unlikely(file->flags & EVENT_FILE_FL_FILTERED) &&
1298 	     !filter_match_preds(file->filter, entry))) {
1299 		__trace_event_discard_commit(buffer, event);
1300 		return true;
1301 	}
1302 
1303 	return false;
1304 }
1305 
1306 /**
1307  * event_trigger_unlock_commit - handle triggers and finish event commit
1308  * @file: The file pointer assoctiated to the event
1309  * @buffer: The ring buffer that the event is being written to
1310  * @event: The event meta data in the ring buffer
1311  * @entry: The event itself
1312  * @irq_flags: The state of the interrupts at the start of the event
1313  * @pc: The state of the preempt count at the start of the event.
1314  *
1315  * This is a helper function to handle triggers that require data
1316  * from the event itself. It also tests the event against filters and
1317  * if the event is soft disabled and should be discarded.
1318  */
1319 static inline void
1320 event_trigger_unlock_commit(struct trace_event_file *file,
1321 			    struct ring_buffer *buffer,
1322 			    struct ring_buffer_event *event,
1323 			    void *entry, unsigned long irq_flags, int pc)
1324 {
1325 	enum event_trigger_type tt = ETT_NONE;
1326 
1327 	if (!__event_trigger_test_discard(file, buffer, event, entry, &tt))
1328 		trace_buffer_unlock_commit(file->tr, buffer, event, irq_flags, pc);
1329 
1330 	if (tt)
1331 		event_triggers_post_call(file, tt, entry);
1332 }
1333 
1334 /**
1335  * event_trigger_unlock_commit_regs - handle triggers and finish event commit
1336  * @file: The file pointer assoctiated to the event
1337  * @buffer: The ring buffer that the event is being written to
1338  * @event: The event meta data in the ring buffer
1339  * @entry: The event itself
1340  * @irq_flags: The state of the interrupts at the start of the event
1341  * @pc: The state of the preempt count at the start of the event.
1342  *
1343  * This is a helper function to handle triggers that require data
1344  * from the event itself. It also tests the event against filters and
1345  * if the event is soft disabled and should be discarded.
1346  *
1347  * Same as event_trigger_unlock_commit() but calls
1348  * trace_buffer_unlock_commit_regs() instead of trace_buffer_unlock_commit().
1349  */
1350 static inline void
1351 event_trigger_unlock_commit_regs(struct trace_event_file *file,
1352 				 struct ring_buffer *buffer,
1353 				 struct ring_buffer_event *event,
1354 				 void *entry, unsigned long irq_flags, int pc,
1355 				 struct pt_regs *regs)
1356 {
1357 	enum event_trigger_type tt = ETT_NONE;
1358 
1359 	if (!__event_trigger_test_discard(file, buffer, event, entry, &tt))
1360 		trace_buffer_unlock_commit_regs(file->tr, buffer, event,
1361 						irq_flags, pc, regs);
1362 
1363 	if (tt)
1364 		event_triggers_post_call(file, tt, entry);
1365 }
1366 
1367 #define FILTER_PRED_INVALID	((unsigned short)-1)
1368 #define FILTER_PRED_IS_RIGHT	(1 << 15)
1369 #define FILTER_PRED_FOLD	(1 << 15)
1370 
1371 /*
1372  * The max preds is the size of unsigned short with
1373  * two flags at the MSBs. One bit is used for both the IS_RIGHT
1374  * and FOLD flags. The other is reserved.
1375  *
1376  * 2^14 preds is way more than enough.
1377  */
1378 #define MAX_FILTER_PRED		16384
1379 
1380 struct filter_pred;
1381 struct regex;
1382 
1383 typedef int (*filter_pred_fn_t) (struct filter_pred *pred, void *event);
1384 
1385 typedef int (*regex_match_func)(char *str, struct regex *r, int len);
1386 
1387 enum regex_type {
1388 	MATCH_FULL = 0,
1389 	MATCH_FRONT_ONLY,
1390 	MATCH_MIDDLE_ONLY,
1391 	MATCH_END_ONLY,
1392 	MATCH_GLOB,
1393 };
1394 
1395 struct regex {
1396 	char			pattern[MAX_FILTER_STR_VAL];
1397 	int			len;
1398 	int			field_len;
1399 	regex_match_func	match;
1400 };
1401 
1402 struct filter_pred {
1403 	filter_pred_fn_t 	fn;
1404 	u64 			val;
1405 	struct regex		regex;
1406 	unsigned short		*ops;
1407 	struct ftrace_event_field *field;
1408 	int 			offset;
1409 	int 			not;
1410 	int 			op;
1411 	unsigned short		index;
1412 	unsigned short		parent;
1413 	unsigned short		left;
1414 	unsigned short		right;
1415 };
1416 
1417 static inline bool is_string_field(struct ftrace_event_field *field)
1418 {
1419 	return field->filter_type == FILTER_DYN_STRING ||
1420 	       field->filter_type == FILTER_STATIC_STRING ||
1421 	       field->filter_type == FILTER_PTR_STRING ||
1422 	       field->filter_type == FILTER_COMM;
1423 }
1424 
1425 static inline bool is_function_field(struct ftrace_event_field *field)
1426 {
1427 	return field->filter_type == FILTER_TRACE_FN;
1428 }
1429 
1430 extern enum regex_type
1431 filter_parse_regex(char *buff, int len, char **search, int *not);
1432 extern void print_event_filter(struct trace_event_file *file,
1433 			       struct trace_seq *s);
1434 extern int apply_event_filter(struct trace_event_file *file,
1435 			      char *filter_string);
1436 extern int apply_subsystem_event_filter(struct trace_subsystem_dir *dir,
1437 					char *filter_string);
1438 extern void print_subsystem_event_filter(struct event_subsystem *system,
1439 					 struct trace_seq *s);
1440 extern int filter_assign_type(const char *type);
1441 extern int create_event_filter(struct trace_event_call *call,
1442 			       char *filter_str, bool set_str,
1443 			       struct event_filter **filterp);
1444 extern void free_event_filter(struct event_filter *filter);
1445 
1446 struct ftrace_event_field *
1447 trace_find_event_field(struct trace_event_call *call, char *name);
1448 
1449 extern void trace_event_enable_cmd_record(bool enable);
1450 extern void trace_event_enable_tgid_record(bool enable);
1451 
1452 extern int event_trace_add_tracer(struct dentry *parent, struct trace_array *tr);
1453 extern int event_trace_del_tracer(struct trace_array *tr);
1454 
1455 extern struct trace_event_file *find_event_file(struct trace_array *tr,
1456 						const char *system,
1457 						const char *event);
1458 
1459 static inline void *event_file_data(struct file *filp)
1460 {
1461 	return READ_ONCE(file_inode(filp)->i_private);
1462 }
1463 
1464 extern struct mutex event_mutex;
1465 extern struct list_head ftrace_events;
1466 
1467 extern const struct file_operations event_trigger_fops;
1468 extern const struct file_operations event_hist_fops;
1469 
1470 #ifdef CONFIG_HIST_TRIGGERS
1471 extern int register_trigger_hist_cmd(void);
1472 extern int register_trigger_hist_enable_disable_cmds(void);
1473 #else
1474 static inline int register_trigger_hist_cmd(void) { return 0; }
1475 static inline int register_trigger_hist_enable_disable_cmds(void) { return 0; }
1476 #endif
1477 
1478 extern int register_trigger_cmds(void);
1479 extern void clear_event_triggers(struct trace_array *tr);
1480 
1481 struct event_trigger_data {
1482 	unsigned long			count;
1483 	int				ref;
1484 	struct event_trigger_ops	*ops;
1485 	struct event_command		*cmd_ops;
1486 	struct event_filter __rcu	*filter;
1487 	char				*filter_str;
1488 	void				*private_data;
1489 	bool				paused;
1490 	bool				paused_tmp;
1491 	struct list_head		list;
1492 	char				*name;
1493 	struct list_head		named_list;
1494 	struct event_trigger_data	*named_data;
1495 };
1496 
1497 /* Avoid typos */
1498 #define ENABLE_EVENT_STR	"enable_event"
1499 #define DISABLE_EVENT_STR	"disable_event"
1500 #define ENABLE_HIST_STR		"enable_hist"
1501 #define DISABLE_HIST_STR	"disable_hist"
1502 
1503 struct enable_trigger_data {
1504 	struct trace_event_file		*file;
1505 	bool				enable;
1506 	bool				hist;
1507 };
1508 
1509 extern int event_enable_trigger_print(struct seq_file *m,
1510 				      struct event_trigger_ops *ops,
1511 				      struct event_trigger_data *data);
1512 extern void event_enable_trigger_free(struct event_trigger_ops *ops,
1513 				      struct event_trigger_data *data);
1514 extern int event_enable_trigger_func(struct event_command *cmd_ops,
1515 				     struct trace_event_file *file,
1516 				     char *glob, char *cmd, char *param);
1517 extern int event_enable_register_trigger(char *glob,
1518 					 struct event_trigger_ops *ops,
1519 					 struct event_trigger_data *data,
1520 					 struct trace_event_file *file);
1521 extern void event_enable_unregister_trigger(char *glob,
1522 					    struct event_trigger_ops *ops,
1523 					    struct event_trigger_data *test,
1524 					    struct trace_event_file *file);
1525 extern void trigger_data_free(struct event_trigger_data *data);
1526 extern int event_trigger_init(struct event_trigger_ops *ops,
1527 			      struct event_trigger_data *data);
1528 extern int trace_event_trigger_enable_disable(struct trace_event_file *file,
1529 					      int trigger_enable);
1530 extern void update_cond_flag(struct trace_event_file *file);
1531 extern void unregister_trigger(char *glob, struct event_trigger_ops *ops,
1532 			       struct event_trigger_data *test,
1533 			       struct trace_event_file *file);
1534 extern int set_trigger_filter(char *filter_str,
1535 			      struct event_trigger_data *trigger_data,
1536 			      struct trace_event_file *file);
1537 extern struct event_trigger_data *find_named_trigger(const char *name);
1538 extern bool is_named_trigger(struct event_trigger_data *test);
1539 extern int save_named_trigger(const char *name,
1540 			      struct event_trigger_data *data);
1541 extern void del_named_trigger(struct event_trigger_data *data);
1542 extern void pause_named_trigger(struct event_trigger_data *data);
1543 extern void unpause_named_trigger(struct event_trigger_data *data);
1544 extern void set_named_trigger_data(struct event_trigger_data *data,
1545 				   struct event_trigger_data *named_data);
1546 extern int register_event_command(struct event_command *cmd);
1547 extern int unregister_event_command(struct event_command *cmd);
1548 extern int register_trigger_hist_enable_disable_cmds(void);
1549 
1550 /**
1551  * struct event_trigger_ops - callbacks for trace event triggers
1552  *
1553  * The methods in this structure provide per-event trigger hooks for
1554  * various trigger operations.
1555  *
1556  * All the methods below, except for @init() and @free(), must be
1557  * implemented.
1558  *
1559  * @func: The trigger 'probe' function called when the triggering
1560  *	event occurs.  The data passed into this callback is the data
1561  *	that was supplied to the event_command @reg() function that
1562  *	registered the trigger (see struct event_command) along with
1563  *	the trace record, rec.
1564  *
1565  * @init: An optional initialization function called for the trigger
1566  *	when the trigger is registered (via the event_command reg()
1567  *	function).  This can be used to perform per-trigger
1568  *	initialization such as incrementing a per-trigger reference
1569  *	count, for instance.  This is usually implemented by the
1570  *	generic utility function @event_trigger_init() (see
1571  *	trace_event_triggers.c).
1572  *
1573  * @free: An optional de-initialization function called for the
1574  *	trigger when the trigger is unregistered (via the
1575  *	event_command @reg() function).  This can be used to perform
1576  *	per-trigger de-initialization such as decrementing a
1577  *	per-trigger reference count and freeing corresponding trigger
1578  *	data, for instance.  This is usually implemented by the
1579  *	generic utility function @event_trigger_free() (see
1580  *	trace_event_triggers.c).
1581  *
1582  * @print: The callback function invoked to have the trigger print
1583  *	itself.  This is usually implemented by a wrapper function
1584  *	that calls the generic utility function @event_trigger_print()
1585  *	(see trace_event_triggers.c).
1586  */
1587 struct event_trigger_ops {
1588 	void			(*func)(struct event_trigger_data *data,
1589 					void *rec);
1590 	int			(*init)(struct event_trigger_ops *ops,
1591 					struct event_trigger_data *data);
1592 	void			(*free)(struct event_trigger_ops *ops,
1593 					struct event_trigger_data *data);
1594 	int			(*print)(struct seq_file *m,
1595 					 struct event_trigger_ops *ops,
1596 					 struct event_trigger_data *data);
1597 };
1598 
1599 /**
1600  * struct event_command - callbacks and data members for event commands
1601  *
1602  * Event commands are invoked by users by writing the command name
1603  * into the 'trigger' file associated with a trace event.  The
1604  * parameters associated with a specific invocation of an event
1605  * command are used to create an event trigger instance, which is
1606  * added to the list of trigger instances associated with that trace
1607  * event.  When the event is hit, the set of triggers associated with
1608  * that event is invoked.
1609  *
1610  * The data members in this structure provide per-event command data
1611  * for various event commands.
1612  *
1613  * All the data members below, except for @post_trigger, must be set
1614  * for each event command.
1615  *
1616  * @name: The unique name that identifies the event command.  This is
1617  *	the name used when setting triggers via trigger files.
1618  *
1619  * @trigger_type: A unique id that identifies the event command
1620  *	'type'.  This value has two purposes, the first to ensure that
1621  *	only one trigger of the same type can be set at a given time
1622  *	for a particular event e.g. it doesn't make sense to have both
1623  *	a traceon and traceoff trigger attached to a single event at
1624  *	the same time, so traceon and traceoff have the same type
1625  *	though they have different names.  The @trigger_type value is
1626  *	also used as a bit value for deferring the actual trigger
1627  *	action until after the current event is finished.  Some
1628  *	commands need to do this if they themselves log to the trace
1629  *	buffer (see the @post_trigger() member below).  @trigger_type
1630  *	values are defined by adding new values to the trigger_type
1631  *	enum in include/linux/trace_events.h.
1632  *
1633  * @flags: See the enum event_command_flags below.
1634  *
1635  * All the methods below, except for @set_filter() and @unreg_all(),
1636  * must be implemented.
1637  *
1638  * @func: The callback function responsible for parsing and
1639  *	registering the trigger written to the 'trigger' file by the
1640  *	user.  It allocates the trigger instance and registers it with
1641  *	the appropriate trace event.  It makes use of the other
1642  *	event_command callback functions to orchestrate this, and is
1643  *	usually implemented by the generic utility function
1644  *	@event_trigger_callback() (see trace_event_triggers.c).
1645  *
1646  * @reg: Adds the trigger to the list of triggers associated with the
1647  *	event, and enables the event trigger itself, after
1648  *	initializing it (via the event_trigger_ops @init() function).
1649  *	This is also where commands can use the @trigger_type value to
1650  *	make the decision as to whether or not multiple instances of
1651  *	the trigger should be allowed.  This is usually implemented by
1652  *	the generic utility function @register_trigger() (see
1653  *	trace_event_triggers.c).
1654  *
1655  * @unreg: Removes the trigger from the list of triggers associated
1656  *	with the event, and disables the event trigger itself, after
1657  *	initializing it (via the event_trigger_ops @free() function).
1658  *	This is usually implemented by the generic utility function
1659  *	@unregister_trigger() (see trace_event_triggers.c).
1660  *
1661  * @unreg_all: An optional function called to remove all the triggers
1662  *	from the list of triggers associated with the event.  Called
1663  *	when a trigger file is opened in truncate mode.
1664  *
1665  * @set_filter: An optional function called to parse and set a filter
1666  *	for the trigger.  If no @set_filter() method is set for the
1667  *	event command, filters set by the user for the command will be
1668  *	ignored.  This is usually implemented by the generic utility
1669  *	function @set_trigger_filter() (see trace_event_triggers.c).
1670  *
1671  * @get_trigger_ops: The callback function invoked to retrieve the
1672  *	event_trigger_ops implementation associated with the command.
1673  */
1674 struct event_command {
1675 	struct list_head	list;
1676 	char			*name;
1677 	enum event_trigger_type	trigger_type;
1678 	int			flags;
1679 	int			(*func)(struct event_command *cmd_ops,
1680 					struct trace_event_file *file,
1681 					char *glob, char *cmd, char *params);
1682 	int			(*reg)(char *glob,
1683 				       struct event_trigger_ops *ops,
1684 				       struct event_trigger_data *data,
1685 				       struct trace_event_file *file);
1686 	void			(*unreg)(char *glob,
1687 					 struct event_trigger_ops *ops,
1688 					 struct event_trigger_data *data,
1689 					 struct trace_event_file *file);
1690 	void			(*unreg_all)(struct trace_event_file *file);
1691 	int			(*set_filter)(char *filter_str,
1692 					      struct event_trigger_data *data,
1693 					      struct trace_event_file *file);
1694 	struct event_trigger_ops *(*get_trigger_ops)(char *cmd, char *param);
1695 };
1696 
1697 /**
1698  * enum event_command_flags - flags for struct event_command
1699  *
1700  * @POST_TRIGGER: A flag that says whether or not this command needs
1701  *	to have its action delayed until after the current event has
1702  *	been closed.  Some triggers need to avoid being invoked while
1703  *	an event is currently in the process of being logged, since
1704  *	the trigger may itself log data into the trace buffer.  Thus
1705  *	we make sure the current event is committed before invoking
1706  *	those triggers.  To do that, the trigger invocation is split
1707  *	in two - the first part checks the filter using the current
1708  *	trace record; if a command has the @post_trigger flag set, it
1709  *	sets a bit for itself in the return value, otherwise it
1710  *	directly invokes the trigger.  Once all commands have been
1711  *	either invoked or set their return flag, the current record is
1712  *	either committed or discarded.  At that point, if any commands
1713  *	have deferred their triggers, those commands are finally
1714  *	invoked following the close of the current event.  In other
1715  *	words, if the event_trigger_ops @func() probe implementation
1716  *	itself logs to the trace buffer, this flag should be set,
1717  *	otherwise it can be left unspecified.
1718  *
1719  * @NEEDS_REC: A flag that says whether or not this command needs
1720  *	access to the trace record in order to perform its function,
1721  *	regardless of whether or not it has a filter associated with
1722  *	it (filters make a trigger require access to the trace record
1723  *	but are not always present).
1724  */
1725 enum event_command_flags {
1726 	EVENT_CMD_FL_POST_TRIGGER	= 1,
1727 	EVENT_CMD_FL_NEEDS_REC		= 2,
1728 };
1729 
1730 static inline bool event_command_post_trigger(struct event_command *cmd_ops)
1731 {
1732 	return cmd_ops->flags & EVENT_CMD_FL_POST_TRIGGER;
1733 }
1734 
1735 static inline bool event_command_needs_rec(struct event_command *cmd_ops)
1736 {
1737 	return cmd_ops->flags & EVENT_CMD_FL_NEEDS_REC;
1738 }
1739 
1740 extern int trace_event_enable_disable(struct trace_event_file *file,
1741 				      int enable, int soft_disable);
1742 extern int tracing_alloc_snapshot(void);
1743 
1744 extern const char *__start___trace_bprintk_fmt[];
1745 extern const char *__stop___trace_bprintk_fmt[];
1746 
1747 extern const char *__start___tracepoint_str[];
1748 extern const char *__stop___tracepoint_str[];
1749 
1750 void trace_printk_control(bool enabled);
1751 void trace_printk_init_buffers(void);
1752 void trace_printk_start_comm(void);
1753 int trace_keep_overwrite(struct tracer *tracer, u32 mask, int set);
1754 int set_tracer_flag(struct trace_array *tr, unsigned int mask, int enabled);
1755 
1756 #define MAX_EVENT_NAME_LEN	64
1757 
1758 extern int trace_run_command(const char *buf, int (*createfn)(int, char**));
1759 extern ssize_t trace_parse_run_command(struct file *file,
1760 		const char __user *buffer, size_t count, loff_t *ppos,
1761 		int (*createfn)(int, char**));
1762 
1763 /*
1764  * Normal trace_printk() and friends allocates special buffers
1765  * to do the manipulation, as well as saves the print formats
1766  * into sections to display. But the trace infrastructure wants
1767  * to use these without the added overhead at the price of being
1768  * a bit slower (used mainly for warnings, where we don't care
1769  * about performance). The internal_trace_puts() is for such
1770  * a purpose.
1771  */
1772 #define internal_trace_puts(str) __trace_puts(_THIS_IP_, str, strlen(str))
1773 
1774 #undef FTRACE_ENTRY
1775 #define FTRACE_ENTRY(call, struct_name, id, tstruct, print, filter)	\
1776 	extern struct trace_event_call					\
1777 	__aligned(4) event_##call;
1778 #undef FTRACE_ENTRY_DUP
1779 #define FTRACE_ENTRY_DUP(call, struct_name, id, tstruct, print, filter)	\
1780 	FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print), \
1781 		     filter)
1782 #undef FTRACE_ENTRY_PACKED
1783 #define FTRACE_ENTRY_PACKED(call, struct_name, id, tstruct, print, filter) \
1784 	FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print), \
1785 		     filter)
1786 
1787 #include "trace_entries.h"
1788 
1789 #if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_FUNCTION_TRACER)
1790 int perf_ftrace_event_register(struct trace_event_call *call,
1791 			       enum trace_reg type, void *data);
1792 #else
1793 #define perf_ftrace_event_register NULL
1794 #endif
1795 
1796 #ifdef CONFIG_FTRACE_SYSCALLS
1797 void init_ftrace_syscalls(void);
1798 const char *get_syscall_name(int syscall);
1799 #else
1800 static inline void init_ftrace_syscalls(void) { }
1801 static inline const char *get_syscall_name(int syscall)
1802 {
1803 	return NULL;
1804 }
1805 #endif
1806 
1807 #ifdef CONFIG_EVENT_TRACING
1808 void trace_event_init(void);
1809 void trace_event_eval_update(struct trace_eval_map **map, int len);
1810 #else
1811 static inline void __init trace_event_init(void) { }
1812 static inline void trace_event_eval_update(struct trace_eval_map **map, int len) { }
1813 #endif
1814 
1815 extern struct trace_iterator *tracepoint_print_iter;
1816 
1817 #endif /* _LINUX_KERNEL_TRACE_H */
1818