xref: /linux/kernel/trace/trace.h (revision 7895d662bab8827176d44326d0a7423221287ca9)
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/trace.h>
15 #include <linux/hw_breakpoint.h>
16 #include <linux/trace_seq.h>
17 #include <linux/trace_events.h>
18 #include <linux/compiler.h>
19 #include <linux/glob.h>
20 #include <linux/irq_work.h>
21 #include <linux/workqueue.h>
22 
23 #ifdef CONFIG_FTRACE_SYSCALLS
24 #include <asm/unistd.h>		/* For NR_SYSCALLS	     */
25 #include <asm/syscall.h>	/* some archs define it here */
26 #endif
27 
28 enum trace_type {
29 	__TRACE_FIRST_TYPE = 0,
30 
31 	TRACE_FN,
32 	TRACE_CTX,
33 	TRACE_WAKE,
34 	TRACE_STACK,
35 	TRACE_PRINT,
36 	TRACE_BPRINT,
37 	TRACE_MMIO_RW,
38 	TRACE_MMIO_MAP,
39 	TRACE_BRANCH,
40 	TRACE_GRAPH_RET,
41 	TRACE_GRAPH_ENT,
42 	TRACE_USER_STACK,
43 	TRACE_BLK,
44 	TRACE_BPUTS,
45 	TRACE_HWLAT,
46 	TRACE_RAW_DATA,
47 
48 	__TRACE_LAST_TYPE,
49 };
50 
51 
52 #undef __field
53 #define __field(type, item)		type	item;
54 
55 #undef __field_fn
56 #define __field_fn(type, item)		type	item;
57 
58 #undef __field_struct
59 #define __field_struct(type, item)	__field(type, item)
60 
61 #undef __field_desc
62 #define __field_desc(type, container, item)
63 
64 #undef __field_packed
65 #define __field_packed(type, container, item)
66 
67 #undef __array
68 #define __array(type, item, size)	type	item[size];
69 
70 #undef __array_desc
71 #define __array_desc(type, container, item, size)
72 
73 #undef __dynamic_array
74 #define __dynamic_array(type, item)	type	item[];
75 
76 #undef F_STRUCT
77 #define F_STRUCT(args...)		args
78 
79 #undef FTRACE_ENTRY
80 #define FTRACE_ENTRY(name, struct_name, id, tstruct, print)		\
81 	struct struct_name {						\
82 		struct trace_entry	ent;				\
83 		tstruct							\
84 	}
85 
86 #undef FTRACE_ENTRY_DUP
87 #define FTRACE_ENTRY_DUP(name, name_struct, id, tstruct, printk)
88 
89 #undef FTRACE_ENTRY_REG
90 #define FTRACE_ENTRY_REG(name, struct_name, id, tstruct, print,	regfn)	\
91 	FTRACE_ENTRY(name, struct_name, id, PARAMS(tstruct), PARAMS(print))
92 
93 #undef FTRACE_ENTRY_PACKED
94 #define FTRACE_ENTRY_PACKED(name, struct_name, id, tstruct, print)	\
95 	FTRACE_ENTRY(name, struct_name, id, PARAMS(tstruct), PARAMS(print)) __packed
96 
97 #include "trace_entries.h"
98 
99 /* Use this for memory failure errors */
100 #define MEM_FAIL(condition, fmt, ...) ({			\
101 	static bool __section(.data.once) __warned;		\
102 	int __ret_warn_once = !!(condition);			\
103 								\
104 	if (unlikely(__ret_warn_once && !__warned)) {		\
105 		__warned = true;				\
106 		pr_err("ERROR: " fmt, ##__VA_ARGS__);		\
107 	}							\
108 	unlikely(__ret_warn_once);				\
109 })
110 
111 /*
112  * syscalls are special, and need special handling, this is why
113  * they are not included in trace_entries.h
114  */
115 struct syscall_trace_enter {
116 	struct trace_entry	ent;
117 	int			nr;
118 	unsigned long		args[];
119 };
120 
121 struct syscall_trace_exit {
122 	struct trace_entry	ent;
123 	int			nr;
124 	long			ret;
125 };
126 
127 struct kprobe_trace_entry_head {
128 	struct trace_entry	ent;
129 	unsigned long		ip;
130 };
131 
132 struct kretprobe_trace_entry_head {
133 	struct trace_entry	ent;
134 	unsigned long		func;
135 	unsigned long		ret_ip;
136 };
137 
138 /*
139  * trace_flag_type is an enumeration that holds different
140  * states when a trace occurs. These are:
141  *  IRQS_OFF		- interrupts were disabled
142  *  IRQS_NOSUPPORT	- arch does not support irqs_disabled_flags
143  *  NEED_RESCHED	- reschedule is requested
144  *  HARDIRQ		- inside an interrupt handler
145  *  SOFTIRQ		- inside a softirq handler
146  */
147 enum trace_flag_type {
148 	TRACE_FLAG_IRQS_OFF		= 0x01,
149 	TRACE_FLAG_IRQS_NOSUPPORT	= 0x02,
150 	TRACE_FLAG_NEED_RESCHED		= 0x04,
151 	TRACE_FLAG_HARDIRQ		= 0x08,
152 	TRACE_FLAG_SOFTIRQ		= 0x10,
153 	TRACE_FLAG_PREEMPT_RESCHED	= 0x20,
154 	TRACE_FLAG_NMI			= 0x40,
155 };
156 
157 #define TRACE_BUF_SIZE		1024
158 
159 struct trace_array;
160 
161 /*
162  * The CPU trace array - it consists of thousands of trace entries
163  * plus some other descriptor data: (for example which task started
164  * the trace, etc.)
165  */
166 struct trace_array_cpu {
167 	atomic_t		disabled;
168 	void			*buffer_page;	/* ring buffer spare */
169 
170 	unsigned long		entries;
171 	unsigned long		saved_latency;
172 	unsigned long		critical_start;
173 	unsigned long		critical_end;
174 	unsigned long		critical_sequence;
175 	unsigned long		nice;
176 	unsigned long		policy;
177 	unsigned long		rt_priority;
178 	unsigned long		skipped_entries;
179 	u64			preempt_timestamp;
180 	pid_t			pid;
181 	kuid_t			uid;
182 	char			comm[TASK_COMM_LEN];
183 
184 #ifdef CONFIG_FUNCTION_TRACER
185 	int			ftrace_ignore_pid;
186 #endif
187 	bool			ignore_pid;
188 };
189 
190 struct tracer;
191 struct trace_option_dentry;
192 
193 struct array_buffer {
194 	struct trace_array		*tr;
195 	struct trace_buffer		*buffer;
196 	struct trace_array_cpu __percpu	*data;
197 	u64				time_start;
198 	int				cpu;
199 };
200 
201 #define TRACE_FLAGS_MAX_SIZE		32
202 
203 struct trace_options {
204 	struct tracer			*tracer;
205 	struct trace_option_dentry	*topts;
206 };
207 
208 struct trace_pid_list {
209 	int				pid_max;
210 	unsigned long			*pids;
211 };
212 
213 enum {
214 	TRACE_PIDS		= BIT(0),
215 	TRACE_NO_PIDS		= BIT(1),
216 };
217 
218 static inline bool pid_type_enabled(int type, struct trace_pid_list *pid_list,
219 				    struct trace_pid_list *no_pid_list)
220 {
221 	/* Return true if the pid list in type has pids */
222 	return ((type & TRACE_PIDS) && pid_list) ||
223 		((type & TRACE_NO_PIDS) && no_pid_list);
224 }
225 
226 static inline bool still_need_pid_events(int type, struct trace_pid_list *pid_list,
227 					 struct trace_pid_list *no_pid_list)
228 {
229 	/*
230 	 * Turning off what is in @type, return true if the "other"
231 	 * pid list, still has pids in it.
232 	 */
233 	return (!(type & TRACE_PIDS) && pid_list) ||
234 		(!(type & TRACE_NO_PIDS) && no_pid_list);
235 }
236 
237 typedef bool (*cond_update_fn_t)(struct trace_array *tr, void *cond_data);
238 
239 /**
240  * struct cond_snapshot - conditional snapshot data and callback
241  *
242  * The cond_snapshot structure encapsulates a callback function and
243  * data associated with the snapshot for a given tracing instance.
244  *
245  * When a snapshot is taken conditionally, by invoking
246  * tracing_snapshot_cond(tr, cond_data), the cond_data passed in is
247  * passed in turn to the cond_snapshot.update() function.  That data
248  * can be compared by the update() implementation with the cond_data
249  * contained wihin the struct cond_snapshot instance associated with
250  * the trace_array.  Because the tr->max_lock is held throughout the
251  * update() call, the update() function can directly retrieve the
252  * cond_snapshot and cond_data associated with the per-instance
253  * snapshot associated with the trace_array.
254  *
255  * The cond_snapshot.update() implementation can save data to be
256  * associated with the snapshot if it decides to, and returns 'true'
257  * in that case, or it returns 'false' if the conditional snapshot
258  * shouldn't be taken.
259  *
260  * The cond_snapshot instance is created and associated with the
261  * user-defined cond_data by tracing_cond_snapshot_enable().
262  * Likewise, the cond_snapshot instance is destroyed and is no longer
263  * associated with the trace instance by
264  * tracing_cond_snapshot_disable().
265  *
266  * The method below is required.
267  *
268  * @update: When a conditional snapshot is invoked, the update()
269  *	callback function is invoked with the tr->max_lock held.  The
270  *	update() implementation signals whether or not to actually
271  *	take the snapshot, by returning 'true' if so, 'false' if no
272  *	snapshot should be taken.  Because the max_lock is held for
273  *	the duration of update(), the implementation is safe to
274  *	directly retrieven and save any implementation data it needs
275  *	to in association with the snapshot.
276  */
277 struct cond_snapshot {
278 	void				*cond_data;
279 	cond_update_fn_t		update;
280 };
281 
282 /*
283  * The trace array - an array of per-CPU trace arrays. This is the
284  * highest level data structure that individual tracers deal with.
285  * They have on/off state as well:
286  */
287 struct trace_array {
288 	struct list_head	list;
289 	char			*name;
290 	struct array_buffer	array_buffer;
291 #ifdef CONFIG_TRACER_MAX_TRACE
292 	/*
293 	 * The max_buffer is used to snapshot the trace when a maximum
294 	 * latency is reached, or when the user initiates a snapshot.
295 	 * Some tracers will use this to store a maximum trace while
296 	 * it continues examining live traces.
297 	 *
298 	 * The buffers for the max_buffer are set up the same as the array_buffer
299 	 * When a snapshot is taken, the buffer of the max_buffer is swapped
300 	 * with the buffer of the array_buffer and the buffers are reset for
301 	 * the array_buffer so the tracing can continue.
302 	 */
303 	struct array_buffer	max_buffer;
304 	bool			allocated_snapshot;
305 #endif
306 #if defined(CONFIG_TRACER_MAX_TRACE) || defined(CONFIG_HWLAT_TRACER)
307 	unsigned long		max_latency;
308 #ifdef CONFIG_FSNOTIFY
309 	struct dentry		*d_max_latency;
310 	struct work_struct	fsnotify_work;
311 	struct irq_work		fsnotify_irqwork;
312 #endif
313 #endif
314 	struct trace_pid_list	__rcu *filtered_pids;
315 	struct trace_pid_list	__rcu *filtered_no_pids;
316 	/*
317 	 * max_lock is used to protect the swapping of buffers
318 	 * when taking a max snapshot. The buffers themselves are
319 	 * protected by per_cpu spinlocks. But the action of the swap
320 	 * needs its own lock.
321 	 *
322 	 * This is defined as a arch_spinlock_t in order to help
323 	 * with performance when lockdep debugging is enabled.
324 	 *
325 	 * It is also used in other places outside the update_max_tr
326 	 * so it needs to be defined outside of the
327 	 * CONFIG_TRACER_MAX_TRACE.
328 	 */
329 	arch_spinlock_t		max_lock;
330 	int			buffer_disabled;
331 #ifdef CONFIG_FTRACE_SYSCALLS
332 	int			sys_refcount_enter;
333 	int			sys_refcount_exit;
334 	struct trace_event_file __rcu *enter_syscall_files[NR_syscalls];
335 	struct trace_event_file __rcu *exit_syscall_files[NR_syscalls];
336 #endif
337 	int			stop_count;
338 	int			clock_id;
339 	int			nr_topts;
340 	bool			clear_trace;
341 	int			buffer_percent;
342 	unsigned int		n_err_log_entries;
343 	struct tracer		*current_trace;
344 	unsigned int		trace_flags;
345 	unsigned char		trace_flags_index[TRACE_FLAGS_MAX_SIZE];
346 	unsigned int		flags;
347 	raw_spinlock_t		start_lock;
348 	struct list_head	err_log;
349 	struct dentry		*dir;
350 	struct dentry		*options;
351 	struct dentry		*percpu_dir;
352 	struct dentry		*event_dir;
353 	struct trace_options	*topts;
354 	struct list_head	systems;
355 	struct list_head	events;
356 	struct trace_event_file *trace_marker_file;
357 	cpumask_var_t		tracing_cpumask; /* only trace on set CPUs */
358 	int			ref;
359 	int			trace_ref;
360 #ifdef CONFIG_FUNCTION_TRACER
361 	struct ftrace_ops	*ops;
362 	struct trace_pid_list	__rcu *function_pids;
363 	struct trace_pid_list	__rcu *function_no_pids;
364 #ifdef CONFIG_DYNAMIC_FTRACE
365 	/* All of these are protected by the ftrace_lock */
366 	struct list_head	func_probes;
367 	struct list_head	mod_trace;
368 	struct list_head	mod_notrace;
369 #endif
370 	/* function tracing enabled */
371 	int			function_enabled;
372 #endif
373 	int			time_stamp_abs_ref;
374 	struct list_head	hist_vars;
375 #ifdef CONFIG_TRACER_SNAPSHOT
376 	struct cond_snapshot	*cond_snapshot;
377 #endif
378 };
379 
380 enum {
381 	TRACE_ARRAY_FL_GLOBAL	= (1 << 0)
382 };
383 
384 extern struct list_head ftrace_trace_arrays;
385 
386 extern struct mutex trace_types_lock;
387 
388 extern int trace_array_get(struct trace_array *tr);
389 extern int tracing_check_open_get_tr(struct trace_array *tr);
390 extern struct trace_array *trace_array_find(const char *instance);
391 extern struct trace_array *trace_array_find_get(const char *instance);
392 
393 extern int tracing_set_time_stamp_abs(struct trace_array *tr, bool abs);
394 extern int tracing_set_clock(struct trace_array *tr, const char *clockstr);
395 
396 extern bool trace_clock_in_ns(struct trace_array *tr);
397 
398 /*
399  * The global tracer (top) should be the first trace array added,
400  * but we check the flag anyway.
401  */
402 static inline struct trace_array *top_trace_array(void)
403 {
404 	struct trace_array *tr;
405 
406 	if (list_empty(&ftrace_trace_arrays))
407 		return NULL;
408 
409 	tr = list_entry(ftrace_trace_arrays.prev,
410 			typeof(*tr), list);
411 	WARN_ON(!(tr->flags & TRACE_ARRAY_FL_GLOBAL));
412 	return tr;
413 }
414 
415 #define FTRACE_CMP_TYPE(var, type) \
416 	__builtin_types_compatible_p(typeof(var), type *)
417 
418 #undef IF_ASSIGN
419 #define IF_ASSIGN(var, entry, etype, id)			\
420 	if (FTRACE_CMP_TYPE(var, etype)) {			\
421 		var = (typeof(var))(entry);			\
422 		WARN_ON(id != 0 && (entry)->type != id);	\
423 		break;						\
424 	}
425 
426 /* Will cause compile errors if type is not found. */
427 extern void __ftrace_bad_type(void);
428 
429 /*
430  * The trace_assign_type is a verifier that the entry type is
431  * the same as the type being assigned. To add new types simply
432  * add a line with the following format:
433  *
434  * IF_ASSIGN(var, ent, type, id);
435  *
436  *  Where "type" is the trace type that includes the trace_entry
437  *  as the "ent" item. And "id" is the trace identifier that is
438  *  used in the trace_type enum.
439  *
440  *  If the type can have more than one id, then use zero.
441  */
442 #define trace_assign_type(var, ent)					\
443 	do {								\
444 		IF_ASSIGN(var, ent, struct ftrace_entry, TRACE_FN);	\
445 		IF_ASSIGN(var, ent, struct ctx_switch_entry, 0);	\
446 		IF_ASSIGN(var, ent, struct stack_entry, TRACE_STACK);	\
447 		IF_ASSIGN(var, ent, struct userstack_entry, TRACE_USER_STACK);\
448 		IF_ASSIGN(var, ent, struct print_entry, TRACE_PRINT);	\
449 		IF_ASSIGN(var, ent, struct bprint_entry, TRACE_BPRINT);	\
450 		IF_ASSIGN(var, ent, struct bputs_entry, TRACE_BPUTS);	\
451 		IF_ASSIGN(var, ent, struct hwlat_entry, TRACE_HWLAT);	\
452 		IF_ASSIGN(var, ent, struct raw_data_entry, TRACE_RAW_DATA);\
453 		IF_ASSIGN(var, ent, struct trace_mmiotrace_rw,		\
454 			  TRACE_MMIO_RW);				\
455 		IF_ASSIGN(var, ent, struct trace_mmiotrace_map,		\
456 			  TRACE_MMIO_MAP);				\
457 		IF_ASSIGN(var, ent, struct trace_branch, TRACE_BRANCH); \
458 		IF_ASSIGN(var, ent, struct ftrace_graph_ent_entry,	\
459 			  TRACE_GRAPH_ENT);		\
460 		IF_ASSIGN(var, ent, struct ftrace_graph_ret_entry,	\
461 			  TRACE_GRAPH_RET);		\
462 		__ftrace_bad_type();					\
463 	} while (0)
464 
465 /*
466  * An option specific to a tracer. This is a boolean value.
467  * The bit is the bit index that sets its value on the
468  * flags value in struct tracer_flags.
469  */
470 struct tracer_opt {
471 	const char	*name; /* Will appear on the trace_options file */
472 	u32		bit; /* Mask assigned in val field in tracer_flags */
473 };
474 
475 /*
476  * The set of specific options for a tracer. Your tracer
477  * have to set the initial value of the flags val.
478  */
479 struct tracer_flags {
480 	u32			val;
481 	struct tracer_opt	*opts;
482 	struct tracer		*trace;
483 };
484 
485 /* Makes more easy to define a tracer opt */
486 #define TRACER_OPT(s, b)	.name = #s, .bit = b
487 
488 
489 struct trace_option_dentry {
490 	struct tracer_opt		*opt;
491 	struct tracer_flags		*flags;
492 	struct trace_array		*tr;
493 	struct dentry			*entry;
494 };
495 
496 /**
497  * struct tracer - a specific tracer and its callbacks to interact with tracefs
498  * @name: the name chosen to select it on the available_tracers file
499  * @init: called when one switches to this tracer (echo name > current_tracer)
500  * @reset: called when one switches to another tracer
501  * @start: called when tracing is unpaused (echo 1 > tracing_on)
502  * @stop: called when tracing is paused (echo 0 > tracing_on)
503  * @update_thresh: called when tracing_thresh is updated
504  * @open: called when the trace file is opened
505  * @pipe_open: called when the trace_pipe file is opened
506  * @close: called when the trace file is released
507  * @pipe_close: called when the trace_pipe file is released
508  * @read: override the default read callback on trace_pipe
509  * @splice_read: override the default splice_read callback on trace_pipe
510  * @selftest: selftest to run on boot (see trace_selftest.c)
511  * @print_headers: override the first lines that describe your columns
512  * @print_line: callback that prints a trace
513  * @set_flag: signals one of your private flags changed (trace_options file)
514  * @flags: your private flags
515  */
516 struct tracer {
517 	const char		*name;
518 	int			(*init)(struct trace_array *tr);
519 	void			(*reset)(struct trace_array *tr);
520 	void			(*start)(struct trace_array *tr);
521 	void			(*stop)(struct trace_array *tr);
522 	int			(*update_thresh)(struct trace_array *tr);
523 	void			(*open)(struct trace_iterator *iter);
524 	void			(*pipe_open)(struct trace_iterator *iter);
525 	void			(*close)(struct trace_iterator *iter);
526 	void			(*pipe_close)(struct trace_iterator *iter);
527 	ssize_t			(*read)(struct trace_iterator *iter,
528 					struct file *filp, char __user *ubuf,
529 					size_t cnt, loff_t *ppos);
530 	ssize_t			(*splice_read)(struct trace_iterator *iter,
531 					       struct file *filp,
532 					       loff_t *ppos,
533 					       struct pipe_inode_info *pipe,
534 					       size_t len,
535 					       unsigned int flags);
536 #ifdef CONFIG_FTRACE_STARTUP_TEST
537 	int			(*selftest)(struct tracer *trace,
538 					    struct trace_array *tr);
539 #endif
540 	void			(*print_header)(struct seq_file *m);
541 	enum print_line_t	(*print_line)(struct trace_iterator *iter);
542 	/* If you handled the flag setting, return 0 */
543 	int			(*set_flag)(struct trace_array *tr,
544 					    u32 old_flags, u32 bit, int set);
545 	/* Return 0 if OK with change, else return non-zero */
546 	int			(*flag_changed)(struct trace_array *tr,
547 						u32 mask, int set);
548 	struct tracer		*next;
549 	struct tracer_flags	*flags;
550 	int			enabled;
551 	bool			print_max;
552 	bool			allow_instances;
553 #ifdef CONFIG_TRACER_MAX_TRACE
554 	bool			use_max_tr;
555 #endif
556 	/* True if tracer cannot be enabled in kernel param */
557 	bool			noboot;
558 };
559 
560 
561 /* Only current can touch trace_recursion */
562 
563 /*
564  * For function tracing recursion:
565  *  The order of these bits are important.
566  *
567  *  When function tracing occurs, the following steps are made:
568  *   If arch does not support a ftrace feature:
569  *    call internal function (uses INTERNAL bits) which calls...
570  *   If callback is registered to the "global" list, the list
571  *    function is called and recursion checks the GLOBAL bits.
572  *    then this function calls...
573  *   The function callback, which can use the FTRACE bits to
574  *    check for recursion.
575  *
576  * Now if the arch does not suppport a feature, and it calls
577  * the global list function which calls the ftrace callback
578  * all three of these steps will do a recursion protection.
579  * There's no reason to do one if the previous caller already
580  * did. The recursion that we are protecting against will
581  * go through the same steps again.
582  *
583  * To prevent the multiple recursion checks, if a recursion
584  * bit is set that is higher than the MAX bit of the current
585  * check, then we know that the check was made by the previous
586  * caller, and we can skip the current check.
587  */
588 enum {
589 	/* Function recursion bits */
590 	TRACE_FTRACE_BIT,
591 	TRACE_FTRACE_NMI_BIT,
592 	TRACE_FTRACE_IRQ_BIT,
593 	TRACE_FTRACE_SIRQ_BIT,
594 
595 	/* INTERNAL_BITs must be greater than FTRACE_BITs */
596 	TRACE_INTERNAL_BIT,
597 	TRACE_INTERNAL_NMI_BIT,
598 	TRACE_INTERNAL_IRQ_BIT,
599 	TRACE_INTERNAL_SIRQ_BIT,
600 
601 	TRACE_BRANCH_BIT,
602 /*
603  * Abuse of the trace_recursion.
604  * As we need a way to maintain state if we are tracing the function
605  * graph in irq because we want to trace a particular function that
606  * was called in irq context but we have irq tracing off. Since this
607  * can only be modified by current, we can reuse trace_recursion.
608  */
609 	TRACE_IRQ_BIT,
610 
611 	/* Set if the function is in the set_graph_function file */
612 	TRACE_GRAPH_BIT,
613 
614 	/*
615 	 * In the very unlikely case that an interrupt came in
616 	 * at a start of graph tracing, and we want to trace
617 	 * the function in that interrupt, the depth can be greater
618 	 * than zero, because of the preempted start of a previous
619 	 * trace. In an even more unlikely case, depth could be 2
620 	 * if a softirq interrupted the start of graph tracing,
621 	 * followed by an interrupt preempting a start of graph
622 	 * tracing in the softirq, and depth can even be 3
623 	 * if an NMI came in at the start of an interrupt function
624 	 * that preempted a softirq start of a function that
625 	 * preempted normal context!!!! Luckily, it can't be
626 	 * greater than 3, so the next two bits are a mask
627 	 * of what the depth is when we set TRACE_GRAPH_BIT
628 	 */
629 
630 	TRACE_GRAPH_DEPTH_START_BIT,
631 	TRACE_GRAPH_DEPTH_END_BIT,
632 
633 	/*
634 	 * To implement set_graph_notrace, if this bit is set, we ignore
635 	 * function graph tracing of called functions, until the return
636 	 * function is called to clear it.
637 	 */
638 	TRACE_GRAPH_NOTRACE_BIT,
639 };
640 
641 #define trace_recursion_set(bit)	do { (current)->trace_recursion |= (1<<(bit)); } while (0)
642 #define trace_recursion_clear(bit)	do { (current)->trace_recursion &= ~(1<<(bit)); } while (0)
643 #define trace_recursion_test(bit)	((current)->trace_recursion & (1<<(bit)))
644 
645 #define trace_recursion_depth() \
646 	(((current)->trace_recursion >> TRACE_GRAPH_DEPTH_START_BIT) & 3)
647 #define trace_recursion_set_depth(depth) \
648 	do {								\
649 		current->trace_recursion &=				\
650 			~(3 << TRACE_GRAPH_DEPTH_START_BIT);		\
651 		current->trace_recursion |=				\
652 			((depth) & 3) << TRACE_GRAPH_DEPTH_START_BIT;	\
653 	} while (0)
654 
655 #define TRACE_CONTEXT_BITS	4
656 
657 #define TRACE_FTRACE_START	TRACE_FTRACE_BIT
658 #define TRACE_FTRACE_MAX	((1 << (TRACE_FTRACE_START + TRACE_CONTEXT_BITS)) - 1)
659 
660 #define TRACE_LIST_START	TRACE_INTERNAL_BIT
661 #define TRACE_LIST_MAX		((1 << (TRACE_LIST_START + TRACE_CONTEXT_BITS)) - 1)
662 
663 #define TRACE_CONTEXT_MASK	TRACE_LIST_MAX
664 
665 static __always_inline int trace_get_context_bit(void)
666 {
667 	int bit;
668 
669 	if (in_interrupt()) {
670 		if (in_nmi())
671 			bit = 0;
672 
673 		else if (in_irq())
674 			bit = 1;
675 		else
676 			bit = 2;
677 	} else
678 		bit = 3;
679 
680 	return bit;
681 }
682 
683 static __always_inline int trace_test_and_set_recursion(int start, int max)
684 {
685 	unsigned int val = current->trace_recursion;
686 	int bit;
687 
688 	/* A previous recursion check was made */
689 	if ((val & TRACE_CONTEXT_MASK) > max)
690 		return 0;
691 
692 	bit = trace_get_context_bit() + start;
693 	if (unlikely(val & (1 << bit)))
694 		return -1;
695 
696 	val |= 1 << bit;
697 	current->trace_recursion = val;
698 	barrier();
699 
700 	return bit;
701 }
702 
703 static __always_inline void trace_clear_recursion(int bit)
704 {
705 	unsigned int val = current->trace_recursion;
706 
707 	if (!bit)
708 		return;
709 
710 	bit = 1 << bit;
711 	val &= ~bit;
712 
713 	barrier();
714 	current->trace_recursion = val;
715 }
716 
717 static inline struct ring_buffer_iter *
718 trace_buffer_iter(struct trace_iterator *iter, int cpu)
719 {
720 	return iter->buffer_iter ? iter->buffer_iter[cpu] : NULL;
721 }
722 
723 int tracer_init(struct tracer *t, struct trace_array *tr);
724 int tracing_is_enabled(void);
725 void tracing_reset_online_cpus(struct array_buffer *buf);
726 void tracing_reset_current(int cpu);
727 void tracing_reset_all_online_cpus(void);
728 int tracing_open_generic(struct inode *inode, struct file *filp);
729 int tracing_open_generic_tr(struct inode *inode, struct file *filp);
730 bool tracing_is_disabled(void);
731 bool tracer_tracing_is_on(struct trace_array *tr);
732 void tracer_tracing_on(struct trace_array *tr);
733 void tracer_tracing_off(struct trace_array *tr);
734 struct dentry *trace_create_file(const char *name,
735 				 umode_t mode,
736 				 struct dentry *parent,
737 				 void *data,
738 				 const struct file_operations *fops);
739 
740 struct dentry *tracing_init_dentry(void);
741 
742 struct ring_buffer_event;
743 
744 struct ring_buffer_event *
745 trace_buffer_lock_reserve(struct trace_buffer *buffer,
746 			  int type,
747 			  unsigned long len,
748 			  unsigned long flags,
749 			  int pc);
750 
751 struct trace_entry *tracing_get_trace_entry(struct trace_array *tr,
752 						struct trace_array_cpu *data);
753 
754 struct trace_entry *trace_find_next_entry(struct trace_iterator *iter,
755 					  int *ent_cpu, u64 *ent_ts);
756 
757 void trace_buffer_unlock_commit_nostack(struct trace_buffer *buffer,
758 					struct ring_buffer_event *event);
759 
760 int trace_empty(struct trace_iterator *iter);
761 
762 void *trace_find_next_entry_inc(struct trace_iterator *iter);
763 
764 void trace_init_global_iter(struct trace_iterator *iter);
765 
766 void tracing_iter_reset(struct trace_iterator *iter, int cpu);
767 
768 unsigned long trace_total_entries_cpu(struct trace_array *tr, int cpu);
769 unsigned long trace_total_entries(struct trace_array *tr);
770 
771 void trace_function(struct trace_array *tr,
772 		    unsigned long ip,
773 		    unsigned long parent_ip,
774 		    unsigned long flags, int pc);
775 void trace_graph_function(struct trace_array *tr,
776 		    unsigned long ip,
777 		    unsigned long parent_ip,
778 		    unsigned long flags, int pc);
779 void trace_latency_header(struct seq_file *m);
780 void trace_default_header(struct seq_file *m);
781 void print_trace_header(struct seq_file *m, struct trace_iterator *iter);
782 int trace_empty(struct trace_iterator *iter);
783 
784 void trace_graph_return(struct ftrace_graph_ret *trace);
785 int trace_graph_entry(struct ftrace_graph_ent *trace);
786 void set_graph_array(struct trace_array *tr);
787 
788 void tracing_start_cmdline_record(void);
789 void tracing_stop_cmdline_record(void);
790 void tracing_start_tgid_record(void);
791 void tracing_stop_tgid_record(void);
792 
793 int register_tracer(struct tracer *type);
794 int is_tracing_stopped(void);
795 
796 loff_t tracing_lseek(struct file *file, loff_t offset, int whence);
797 
798 extern cpumask_var_t __read_mostly tracing_buffer_mask;
799 
800 #define for_each_tracing_cpu(cpu)	\
801 	for_each_cpu(cpu, tracing_buffer_mask)
802 
803 extern unsigned long nsecs_to_usecs(unsigned long nsecs);
804 
805 extern unsigned long tracing_thresh;
806 
807 /* PID filtering */
808 
809 extern int pid_max;
810 
811 bool trace_find_filtered_pid(struct trace_pid_list *filtered_pids,
812 			     pid_t search_pid);
813 bool trace_ignore_this_task(struct trace_pid_list *filtered_pids,
814 			    struct trace_pid_list *filtered_no_pids,
815 			    struct task_struct *task);
816 void trace_filter_add_remove_task(struct trace_pid_list *pid_list,
817 				  struct task_struct *self,
818 				  struct task_struct *task);
819 void *trace_pid_next(struct trace_pid_list *pid_list, void *v, loff_t *pos);
820 void *trace_pid_start(struct trace_pid_list *pid_list, loff_t *pos);
821 int trace_pid_show(struct seq_file *m, void *v);
822 void trace_free_pid_list(struct trace_pid_list *pid_list);
823 int trace_pid_write(struct trace_pid_list *filtered_pids,
824 		    struct trace_pid_list **new_pid_list,
825 		    const char __user *ubuf, size_t cnt);
826 
827 #ifdef CONFIG_TRACER_MAX_TRACE
828 void update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu,
829 		   void *cond_data);
830 void update_max_tr_single(struct trace_array *tr,
831 			  struct task_struct *tsk, int cpu);
832 #endif /* CONFIG_TRACER_MAX_TRACE */
833 
834 #if (defined(CONFIG_TRACER_MAX_TRACE) || defined(CONFIG_HWLAT_TRACER)) && \
835 	defined(CONFIG_FSNOTIFY)
836 
837 void latency_fsnotify(struct trace_array *tr);
838 
839 #else
840 
841 static inline void latency_fsnotify(struct trace_array *tr) { }
842 
843 #endif
844 
845 #ifdef CONFIG_STACKTRACE
846 void __trace_stack(struct trace_array *tr, unsigned long flags, int skip,
847 		   int pc);
848 #else
849 static inline void __trace_stack(struct trace_array *tr, unsigned long flags,
850 				 int skip, int pc)
851 {
852 }
853 #endif /* CONFIG_STACKTRACE */
854 
855 extern u64 ftrace_now(int cpu);
856 
857 extern void trace_find_cmdline(int pid, char comm[]);
858 extern int trace_find_tgid(int pid);
859 extern void trace_event_follow_fork(struct trace_array *tr, bool enable);
860 
861 #ifdef CONFIG_DYNAMIC_FTRACE
862 extern unsigned long ftrace_update_tot_cnt;
863 extern unsigned long ftrace_number_of_pages;
864 extern unsigned long ftrace_number_of_groups;
865 void ftrace_init_trace_array(struct trace_array *tr);
866 #else
867 static inline void ftrace_init_trace_array(struct trace_array *tr) { }
868 #endif
869 #define DYN_FTRACE_TEST_NAME trace_selftest_dynamic_test_func
870 extern int DYN_FTRACE_TEST_NAME(void);
871 #define DYN_FTRACE_TEST_NAME2 trace_selftest_dynamic_test_func2
872 extern int DYN_FTRACE_TEST_NAME2(void);
873 
874 extern bool ring_buffer_expanded;
875 extern bool tracing_selftest_disabled;
876 
877 #ifdef CONFIG_FTRACE_STARTUP_TEST
878 extern int trace_selftest_startup_function(struct tracer *trace,
879 					   struct trace_array *tr);
880 extern int trace_selftest_startup_function_graph(struct tracer *trace,
881 						 struct trace_array *tr);
882 extern int trace_selftest_startup_irqsoff(struct tracer *trace,
883 					  struct trace_array *tr);
884 extern int trace_selftest_startup_preemptoff(struct tracer *trace,
885 					     struct trace_array *tr);
886 extern int trace_selftest_startup_preemptirqsoff(struct tracer *trace,
887 						 struct trace_array *tr);
888 extern int trace_selftest_startup_wakeup(struct tracer *trace,
889 					 struct trace_array *tr);
890 extern int trace_selftest_startup_nop(struct tracer *trace,
891 					 struct trace_array *tr);
892 extern int trace_selftest_startup_branch(struct tracer *trace,
893 					 struct trace_array *tr);
894 /*
895  * Tracer data references selftest functions that only occur
896  * on boot up. These can be __init functions. Thus, when selftests
897  * are enabled, then the tracers need to reference __init functions.
898  */
899 #define __tracer_data		__refdata
900 #else
901 /* Tracers are seldom changed. Optimize when selftests are disabled. */
902 #define __tracer_data		__read_mostly
903 #endif /* CONFIG_FTRACE_STARTUP_TEST */
904 
905 extern void *head_page(struct trace_array_cpu *data);
906 extern unsigned long long ns2usecs(u64 nsec);
907 extern int
908 trace_vbprintk(unsigned long ip, const char *fmt, va_list args);
909 extern int
910 trace_vprintk(unsigned long ip, const char *fmt, va_list args);
911 extern int
912 trace_array_vprintk(struct trace_array *tr,
913 		    unsigned long ip, const char *fmt, va_list args);
914 int trace_array_printk_buf(struct trace_buffer *buffer,
915 			   unsigned long ip, const char *fmt, ...);
916 void trace_printk_seq(struct trace_seq *s);
917 enum print_line_t print_trace_line(struct trace_iterator *iter);
918 
919 extern char trace_find_mark(unsigned long long duration);
920 
921 struct ftrace_hash;
922 
923 struct ftrace_mod_load {
924 	struct list_head	list;
925 	char			*func;
926 	char			*module;
927 	int			 enable;
928 };
929 
930 enum {
931 	FTRACE_HASH_FL_MOD	= (1 << 0),
932 };
933 
934 struct ftrace_hash {
935 	unsigned long		size_bits;
936 	struct hlist_head	*buckets;
937 	unsigned long		count;
938 	unsigned long		flags;
939 	struct rcu_head		rcu;
940 };
941 
942 struct ftrace_func_entry *
943 ftrace_lookup_ip(struct ftrace_hash *hash, unsigned long ip);
944 
945 static __always_inline bool ftrace_hash_empty(struct ftrace_hash *hash)
946 {
947 	return !hash || !(hash->count || (hash->flags & FTRACE_HASH_FL_MOD));
948 }
949 
950 /* Standard output formatting function used for function return traces */
951 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
952 
953 /* Flag options */
954 #define TRACE_GRAPH_PRINT_OVERRUN       0x1
955 #define TRACE_GRAPH_PRINT_CPU           0x2
956 #define TRACE_GRAPH_PRINT_OVERHEAD      0x4
957 #define TRACE_GRAPH_PRINT_PROC          0x8
958 #define TRACE_GRAPH_PRINT_DURATION      0x10
959 #define TRACE_GRAPH_PRINT_ABS_TIME      0x20
960 #define TRACE_GRAPH_PRINT_REL_TIME      0x40
961 #define TRACE_GRAPH_PRINT_IRQS          0x80
962 #define TRACE_GRAPH_PRINT_TAIL          0x100
963 #define TRACE_GRAPH_SLEEP_TIME          0x200
964 #define TRACE_GRAPH_GRAPH_TIME          0x400
965 #define TRACE_GRAPH_PRINT_FILL_SHIFT	28
966 #define TRACE_GRAPH_PRINT_FILL_MASK	(0x3 << TRACE_GRAPH_PRINT_FILL_SHIFT)
967 
968 extern void ftrace_graph_sleep_time_control(bool enable);
969 
970 #ifdef CONFIG_FUNCTION_PROFILER
971 extern void ftrace_graph_graph_time_control(bool enable);
972 #else
973 static inline void ftrace_graph_graph_time_control(bool enable) { }
974 #endif
975 
976 extern enum print_line_t
977 print_graph_function_flags(struct trace_iterator *iter, u32 flags);
978 extern void print_graph_headers_flags(struct seq_file *s, u32 flags);
979 extern void
980 trace_print_graph_duration(unsigned long long duration, struct trace_seq *s);
981 extern void graph_trace_open(struct trace_iterator *iter);
982 extern void graph_trace_close(struct trace_iterator *iter);
983 extern int __trace_graph_entry(struct trace_array *tr,
984 			       struct ftrace_graph_ent *trace,
985 			       unsigned long flags, int pc);
986 extern void __trace_graph_return(struct trace_array *tr,
987 				 struct ftrace_graph_ret *trace,
988 				 unsigned long flags, int pc);
989 
990 #ifdef CONFIG_DYNAMIC_FTRACE
991 extern struct ftrace_hash __rcu *ftrace_graph_hash;
992 extern struct ftrace_hash __rcu *ftrace_graph_notrace_hash;
993 
994 static inline int ftrace_graph_addr(struct ftrace_graph_ent *trace)
995 {
996 	unsigned long addr = trace->func;
997 	int ret = 0;
998 	struct ftrace_hash *hash;
999 
1000 	preempt_disable_notrace();
1001 
1002 	/*
1003 	 * Have to open code "rcu_dereference_sched()" because the
1004 	 * function graph tracer can be called when RCU is not
1005 	 * "watching".
1006 	 * Protected with schedule_on_each_cpu(ftrace_sync)
1007 	 */
1008 	hash = rcu_dereference_protected(ftrace_graph_hash, !preemptible());
1009 
1010 	if (ftrace_hash_empty(hash)) {
1011 		ret = 1;
1012 		goto out;
1013 	}
1014 
1015 	if (ftrace_lookup_ip(hash, addr)) {
1016 
1017 		/*
1018 		 * This needs to be cleared on the return functions
1019 		 * when the depth is zero.
1020 		 */
1021 		trace_recursion_set(TRACE_GRAPH_BIT);
1022 		trace_recursion_set_depth(trace->depth);
1023 
1024 		/*
1025 		 * If no irqs are to be traced, but a set_graph_function
1026 		 * is set, and called by an interrupt handler, we still
1027 		 * want to trace it.
1028 		 */
1029 		if (in_irq())
1030 			trace_recursion_set(TRACE_IRQ_BIT);
1031 		else
1032 			trace_recursion_clear(TRACE_IRQ_BIT);
1033 		ret = 1;
1034 	}
1035 
1036 out:
1037 	preempt_enable_notrace();
1038 	return ret;
1039 }
1040 
1041 static inline void ftrace_graph_addr_finish(struct ftrace_graph_ret *trace)
1042 {
1043 	if (trace_recursion_test(TRACE_GRAPH_BIT) &&
1044 	    trace->depth == trace_recursion_depth())
1045 		trace_recursion_clear(TRACE_GRAPH_BIT);
1046 }
1047 
1048 static inline int ftrace_graph_notrace_addr(unsigned long addr)
1049 {
1050 	int ret = 0;
1051 	struct ftrace_hash *notrace_hash;
1052 
1053 	preempt_disable_notrace();
1054 
1055 	/*
1056 	 * Have to open code "rcu_dereference_sched()" because the
1057 	 * function graph tracer can be called when RCU is not
1058 	 * "watching".
1059 	 * Protected with schedule_on_each_cpu(ftrace_sync)
1060 	 */
1061 	notrace_hash = rcu_dereference_protected(ftrace_graph_notrace_hash,
1062 						 !preemptible());
1063 
1064 	if (ftrace_lookup_ip(notrace_hash, addr))
1065 		ret = 1;
1066 
1067 	preempt_enable_notrace();
1068 	return ret;
1069 }
1070 #else
1071 static inline int ftrace_graph_addr(struct ftrace_graph_ent *trace)
1072 {
1073 	return 1;
1074 }
1075 
1076 static inline int ftrace_graph_notrace_addr(unsigned long addr)
1077 {
1078 	return 0;
1079 }
1080 static inline void ftrace_graph_addr_finish(struct ftrace_graph_ret *trace)
1081 { }
1082 #endif /* CONFIG_DYNAMIC_FTRACE */
1083 
1084 extern unsigned int fgraph_max_depth;
1085 
1086 static inline bool ftrace_graph_ignore_func(struct ftrace_graph_ent *trace)
1087 {
1088 	/* trace it when it is-nested-in or is a function enabled. */
1089 	return !(trace_recursion_test(TRACE_GRAPH_BIT) ||
1090 		 ftrace_graph_addr(trace)) ||
1091 		(trace->depth < 0) ||
1092 		(fgraph_max_depth && trace->depth >= fgraph_max_depth);
1093 }
1094 
1095 #else /* CONFIG_FUNCTION_GRAPH_TRACER */
1096 static inline enum print_line_t
1097 print_graph_function_flags(struct trace_iterator *iter, u32 flags)
1098 {
1099 	return TRACE_TYPE_UNHANDLED;
1100 }
1101 #endif /* CONFIG_FUNCTION_GRAPH_TRACER */
1102 
1103 extern struct list_head ftrace_pids;
1104 
1105 #ifdef CONFIG_FUNCTION_TRACER
1106 
1107 #define FTRACE_PID_IGNORE	-1
1108 #define FTRACE_PID_TRACE	-2
1109 
1110 struct ftrace_func_command {
1111 	struct list_head	list;
1112 	char			*name;
1113 	int			(*func)(struct trace_array *tr,
1114 					struct ftrace_hash *hash,
1115 					char *func, char *cmd,
1116 					char *params, int enable);
1117 };
1118 extern bool ftrace_filter_param __initdata;
1119 static inline int ftrace_trace_task(struct trace_array *tr)
1120 {
1121 	return this_cpu_read(tr->array_buffer.data->ftrace_ignore_pid) !=
1122 		FTRACE_PID_IGNORE;
1123 }
1124 extern int ftrace_is_dead(void);
1125 int ftrace_create_function_files(struct trace_array *tr,
1126 				 struct dentry *parent);
1127 void ftrace_destroy_function_files(struct trace_array *tr);
1128 void ftrace_init_global_array_ops(struct trace_array *tr);
1129 void ftrace_init_array_ops(struct trace_array *tr, ftrace_func_t func);
1130 void ftrace_reset_array_ops(struct trace_array *tr);
1131 void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d_tracer);
1132 void ftrace_init_tracefs_toplevel(struct trace_array *tr,
1133 				  struct dentry *d_tracer);
1134 void ftrace_clear_pids(struct trace_array *tr);
1135 int init_function_trace(void);
1136 void ftrace_pid_follow_fork(struct trace_array *tr, bool enable);
1137 #else
1138 static inline int ftrace_trace_task(struct trace_array *tr)
1139 {
1140 	return 1;
1141 }
1142 static inline int ftrace_is_dead(void) { return 0; }
1143 static inline int
1144 ftrace_create_function_files(struct trace_array *tr,
1145 			     struct dentry *parent)
1146 {
1147 	return 0;
1148 }
1149 static inline void ftrace_destroy_function_files(struct trace_array *tr) { }
1150 static inline __init void
1151 ftrace_init_global_array_ops(struct trace_array *tr) { }
1152 static inline void ftrace_reset_array_ops(struct trace_array *tr) { }
1153 static inline void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d) { }
1154 static inline void ftrace_init_tracefs_toplevel(struct trace_array *tr, struct dentry *d) { }
1155 static inline void ftrace_clear_pids(struct trace_array *tr) { }
1156 static inline int init_function_trace(void) { return 0; }
1157 static inline void ftrace_pid_follow_fork(struct trace_array *tr, bool enable) { }
1158 /* ftace_func_t type is not defined, use macro instead of static inline */
1159 #define ftrace_init_array_ops(tr, func) do { } while (0)
1160 #endif /* CONFIG_FUNCTION_TRACER */
1161 
1162 #if defined(CONFIG_FUNCTION_TRACER) && defined(CONFIG_DYNAMIC_FTRACE)
1163 
1164 struct ftrace_probe_ops {
1165 	void			(*func)(unsigned long ip,
1166 					unsigned long parent_ip,
1167 					struct trace_array *tr,
1168 					struct ftrace_probe_ops *ops,
1169 					void *data);
1170 	int			(*init)(struct ftrace_probe_ops *ops,
1171 					struct trace_array *tr,
1172 					unsigned long ip, void *init_data,
1173 					void **data);
1174 	void			(*free)(struct ftrace_probe_ops *ops,
1175 					struct trace_array *tr,
1176 					unsigned long ip, void *data);
1177 	int			(*print)(struct seq_file *m,
1178 					 unsigned long ip,
1179 					 struct ftrace_probe_ops *ops,
1180 					 void *data);
1181 };
1182 
1183 struct ftrace_func_mapper;
1184 typedef int (*ftrace_mapper_func)(void *data);
1185 
1186 struct ftrace_func_mapper *allocate_ftrace_func_mapper(void);
1187 void **ftrace_func_mapper_find_ip(struct ftrace_func_mapper *mapper,
1188 					   unsigned long ip);
1189 int ftrace_func_mapper_add_ip(struct ftrace_func_mapper *mapper,
1190 			       unsigned long ip, void *data);
1191 void *ftrace_func_mapper_remove_ip(struct ftrace_func_mapper *mapper,
1192 				   unsigned long ip);
1193 void free_ftrace_func_mapper(struct ftrace_func_mapper *mapper,
1194 			     ftrace_mapper_func free_func);
1195 
1196 extern int
1197 register_ftrace_function_probe(char *glob, struct trace_array *tr,
1198 			       struct ftrace_probe_ops *ops, void *data);
1199 extern int
1200 unregister_ftrace_function_probe_func(char *glob, struct trace_array *tr,
1201 				      struct ftrace_probe_ops *ops);
1202 extern void clear_ftrace_function_probes(struct trace_array *tr);
1203 
1204 int register_ftrace_command(struct ftrace_func_command *cmd);
1205 int unregister_ftrace_command(struct ftrace_func_command *cmd);
1206 
1207 void ftrace_create_filter_files(struct ftrace_ops *ops,
1208 				struct dentry *parent);
1209 void ftrace_destroy_filter_files(struct ftrace_ops *ops);
1210 
1211 extern int ftrace_set_filter(struct ftrace_ops *ops, unsigned char *buf,
1212 			     int len, int reset);
1213 extern int ftrace_set_notrace(struct ftrace_ops *ops, unsigned char *buf,
1214 			      int len, int reset);
1215 #else
1216 struct ftrace_func_command;
1217 
1218 static inline __init int register_ftrace_command(struct ftrace_func_command *cmd)
1219 {
1220 	return -EINVAL;
1221 }
1222 static inline __init int unregister_ftrace_command(char *cmd_name)
1223 {
1224 	return -EINVAL;
1225 }
1226 static inline void clear_ftrace_function_probes(struct trace_array *tr)
1227 {
1228 }
1229 
1230 /*
1231  * The ops parameter passed in is usually undefined.
1232  * This must be a macro.
1233  */
1234 #define ftrace_create_filter_files(ops, parent) do { } while (0)
1235 #define ftrace_destroy_filter_files(ops) do { } while (0)
1236 #endif /* CONFIG_FUNCTION_TRACER && CONFIG_DYNAMIC_FTRACE */
1237 
1238 bool ftrace_event_is_function(struct trace_event_call *call);
1239 
1240 /*
1241  * struct trace_parser - servers for reading the user input separated by spaces
1242  * @cont: set if the input is not complete - no final space char was found
1243  * @buffer: holds the parsed user input
1244  * @idx: user input length
1245  * @size: buffer size
1246  */
1247 struct trace_parser {
1248 	bool		cont;
1249 	char		*buffer;
1250 	unsigned	idx;
1251 	unsigned	size;
1252 };
1253 
1254 static inline bool trace_parser_loaded(struct trace_parser *parser)
1255 {
1256 	return (parser->idx != 0);
1257 }
1258 
1259 static inline bool trace_parser_cont(struct trace_parser *parser)
1260 {
1261 	return parser->cont;
1262 }
1263 
1264 static inline void trace_parser_clear(struct trace_parser *parser)
1265 {
1266 	parser->cont = false;
1267 	parser->idx = 0;
1268 }
1269 
1270 extern int trace_parser_get_init(struct trace_parser *parser, int size);
1271 extern void trace_parser_put(struct trace_parser *parser);
1272 extern int trace_get_user(struct trace_parser *parser, const char __user *ubuf,
1273 	size_t cnt, loff_t *ppos);
1274 
1275 /*
1276  * Only create function graph options if function graph is configured.
1277  */
1278 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1279 # define FGRAPH_FLAGS						\
1280 		C(DISPLAY_GRAPH,	"display-graph"),
1281 #else
1282 # define FGRAPH_FLAGS
1283 #endif
1284 
1285 #ifdef CONFIG_BRANCH_TRACER
1286 # define BRANCH_FLAGS					\
1287 		C(BRANCH,		"branch"),
1288 #else
1289 # define BRANCH_FLAGS
1290 #endif
1291 
1292 #ifdef CONFIG_FUNCTION_TRACER
1293 # define FUNCTION_FLAGS						\
1294 		C(FUNCTION,		"function-trace"),	\
1295 		C(FUNC_FORK,		"function-fork"),
1296 # define FUNCTION_DEFAULT_FLAGS		TRACE_ITER_FUNCTION
1297 #else
1298 # define FUNCTION_FLAGS
1299 # define FUNCTION_DEFAULT_FLAGS		0UL
1300 # define TRACE_ITER_FUNC_FORK		0UL
1301 #endif
1302 
1303 #ifdef CONFIG_STACKTRACE
1304 # define STACK_FLAGS				\
1305 		C(STACKTRACE,		"stacktrace"),
1306 #else
1307 # define STACK_FLAGS
1308 #endif
1309 
1310 /*
1311  * trace_iterator_flags is an enumeration that defines bit
1312  * positions into trace_flags that controls the output.
1313  *
1314  * NOTE: These bits must match the trace_options array in
1315  *       trace.c (this macro guarantees it).
1316  */
1317 #define TRACE_FLAGS						\
1318 		C(PRINT_PARENT,		"print-parent"),	\
1319 		C(SYM_OFFSET,		"sym-offset"),		\
1320 		C(SYM_ADDR,		"sym-addr"),		\
1321 		C(VERBOSE,		"verbose"),		\
1322 		C(RAW,			"raw"),			\
1323 		C(HEX,			"hex"),			\
1324 		C(BIN,			"bin"),			\
1325 		C(BLOCK,		"block"),		\
1326 		C(PRINTK,		"trace_printk"),	\
1327 		C(ANNOTATE,		"annotate"),		\
1328 		C(USERSTACKTRACE,	"userstacktrace"),	\
1329 		C(SYM_USEROBJ,		"sym-userobj"),		\
1330 		C(PRINTK_MSGONLY,	"printk-msg-only"),	\
1331 		C(CONTEXT_INFO,		"context-info"),   /* Print pid/cpu/time */ \
1332 		C(LATENCY_FMT,		"latency-format"),	\
1333 		C(RECORD_CMD,		"record-cmd"),		\
1334 		C(RECORD_TGID,		"record-tgid"),		\
1335 		C(OVERWRITE,		"overwrite"),		\
1336 		C(STOP_ON_FREE,		"disable_on_free"),	\
1337 		C(IRQ_INFO,		"irq-info"),		\
1338 		C(MARKERS,		"markers"),		\
1339 		C(EVENT_FORK,		"event-fork"),		\
1340 		C(PAUSE_ON_TRACE,	"pause-on-trace"),	\
1341 		FUNCTION_FLAGS					\
1342 		FGRAPH_FLAGS					\
1343 		STACK_FLAGS					\
1344 		BRANCH_FLAGS
1345 
1346 /*
1347  * By defining C, we can make TRACE_FLAGS a list of bit names
1348  * that will define the bits for the flag masks.
1349  */
1350 #undef C
1351 #define C(a, b) TRACE_ITER_##a##_BIT
1352 
1353 enum trace_iterator_bits {
1354 	TRACE_FLAGS
1355 	/* Make sure we don't go more than we have bits for */
1356 	TRACE_ITER_LAST_BIT
1357 };
1358 
1359 /*
1360  * By redefining C, we can make TRACE_FLAGS a list of masks that
1361  * use the bits as defined above.
1362  */
1363 #undef C
1364 #define C(a, b) TRACE_ITER_##a = (1 << TRACE_ITER_##a##_BIT)
1365 
1366 enum trace_iterator_flags { TRACE_FLAGS };
1367 
1368 /*
1369  * TRACE_ITER_SYM_MASK masks the options in trace_flags that
1370  * control the output of kernel symbols.
1371  */
1372 #define TRACE_ITER_SYM_MASK \
1373 	(TRACE_ITER_PRINT_PARENT|TRACE_ITER_SYM_OFFSET|TRACE_ITER_SYM_ADDR)
1374 
1375 extern struct tracer nop_trace;
1376 
1377 #ifdef CONFIG_BRANCH_TRACER
1378 extern int enable_branch_tracing(struct trace_array *tr);
1379 extern void disable_branch_tracing(void);
1380 static inline int trace_branch_enable(struct trace_array *tr)
1381 {
1382 	if (tr->trace_flags & TRACE_ITER_BRANCH)
1383 		return enable_branch_tracing(tr);
1384 	return 0;
1385 }
1386 static inline void trace_branch_disable(void)
1387 {
1388 	/* due to races, always disable */
1389 	disable_branch_tracing();
1390 }
1391 #else
1392 static inline int trace_branch_enable(struct trace_array *tr)
1393 {
1394 	return 0;
1395 }
1396 static inline void trace_branch_disable(void)
1397 {
1398 }
1399 #endif /* CONFIG_BRANCH_TRACER */
1400 
1401 /* set ring buffers to default size if not already done so */
1402 int tracing_update_buffers(void);
1403 
1404 struct ftrace_event_field {
1405 	struct list_head	link;
1406 	const char		*name;
1407 	const char		*type;
1408 	int			filter_type;
1409 	int			offset;
1410 	int			size;
1411 	int			is_signed;
1412 };
1413 
1414 struct prog_entry;
1415 
1416 struct event_filter {
1417 	struct prog_entry __rcu	*prog;
1418 	char			*filter_string;
1419 };
1420 
1421 struct event_subsystem {
1422 	struct list_head	list;
1423 	const char		*name;
1424 	struct event_filter	*filter;
1425 	int			ref_count;
1426 };
1427 
1428 struct trace_subsystem_dir {
1429 	struct list_head		list;
1430 	struct event_subsystem		*subsystem;
1431 	struct trace_array		*tr;
1432 	struct dentry			*entry;
1433 	int				ref_count;
1434 	int				nr_events;
1435 };
1436 
1437 extern int call_filter_check_discard(struct trace_event_call *call, void *rec,
1438 				     struct trace_buffer *buffer,
1439 				     struct ring_buffer_event *event);
1440 
1441 void trace_buffer_unlock_commit_regs(struct trace_array *tr,
1442 				     struct trace_buffer *buffer,
1443 				     struct ring_buffer_event *event,
1444 				     unsigned long flags, int pc,
1445 				     struct pt_regs *regs);
1446 
1447 static inline void trace_buffer_unlock_commit(struct trace_array *tr,
1448 					      struct trace_buffer *buffer,
1449 					      struct ring_buffer_event *event,
1450 					      unsigned long flags, int pc)
1451 {
1452 	trace_buffer_unlock_commit_regs(tr, buffer, event, flags, pc, NULL);
1453 }
1454 
1455 DECLARE_PER_CPU(struct ring_buffer_event *, trace_buffered_event);
1456 DECLARE_PER_CPU(int, trace_buffered_event_cnt);
1457 void trace_buffered_event_disable(void);
1458 void trace_buffered_event_enable(void);
1459 
1460 static inline void
1461 __trace_event_discard_commit(struct trace_buffer *buffer,
1462 			     struct ring_buffer_event *event)
1463 {
1464 	if (this_cpu_read(trace_buffered_event) == event) {
1465 		/* Simply release the temp buffer */
1466 		this_cpu_dec(trace_buffered_event_cnt);
1467 		return;
1468 	}
1469 	ring_buffer_discard_commit(buffer, event);
1470 }
1471 
1472 /*
1473  * Helper function for event_trigger_unlock_commit{_regs}().
1474  * If there are event triggers attached to this event that requires
1475  * filtering against its fields, then they wil be called as the
1476  * entry already holds the field information of the current event.
1477  *
1478  * It also checks if the event should be discarded or not.
1479  * It is to be discarded if the event is soft disabled and the
1480  * event was only recorded to process triggers, or if the event
1481  * filter is active and this event did not match the filters.
1482  *
1483  * Returns true if the event is discarded, false otherwise.
1484  */
1485 static inline bool
1486 __event_trigger_test_discard(struct trace_event_file *file,
1487 			     struct trace_buffer *buffer,
1488 			     struct ring_buffer_event *event,
1489 			     void *entry,
1490 			     enum event_trigger_type *tt)
1491 {
1492 	unsigned long eflags = file->flags;
1493 
1494 	if (eflags & EVENT_FILE_FL_TRIGGER_COND)
1495 		*tt = event_triggers_call(file, entry, event);
1496 
1497 	if (test_bit(EVENT_FILE_FL_SOFT_DISABLED_BIT, &file->flags) ||
1498 	    (unlikely(file->flags & EVENT_FILE_FL_FILTERED) &&
1499 	     !filter_match_preds(file->filter, entry))) {
1500 		__trace_event_discard_commit(buffer, event);
1501 		return true;
1502 	}
1503 
1504 	return false;
1505 }
1506 
1507 /**
1508  * event_trigger_unlock_commit - handle triggers and finish event commit
1509  * @file: The file pointer assoctiated to the event
1510  * @buffer: The ring buffer that the event is being written to
1511  * @event: The event meta data in the ring buffer
1512  * @entry: The event itself
1513  * @irq_flags: The state of the interrupts at the start of the event
1514  * @pc: The state of the preempt count at the start of the event.
1515  *
1516  * This is a helper function to handle triggers that require data
1517  * from the event itself. It also tests the event against filters and
1518  * if the event is soft disabled and should be discarded.
1519  */
1520 static inline void
1521 event_trigger_unlock_commit(struct trace_event_file *file,
1522 			    struct trace_buffer *buffer,
1523 			    struct ring_buffer_event *event,
1524 			    void *entry, unsigned long irq_flags, int pc)
1525 {
1526 	enum event_trigger_type tt = ETT_NONE;
1527 
1528 	if (!__event_trigger_test_discard(file, buffer, event, entry, &tt))
1529 		trace_buffer_unlock_commit(file->tr, buffer, event, irq_flags, pc);
1530 
1531 	if (tt)
1532 		event_triggers_post_call(file, tt);
1533 }
1534 
1535 /**
1536  * event_trigger_unlock_commit_regs - handle triggers and finish event commit
1537  * @file: The file pointer assoctiated to the event
1538  * @buffer: The ring buffer that the event is being written to
1539  * @event: The event meta data in the ring buffer
1540  * @entry: The event itself
1541  * @irq_flags: The state of the interrupts at the start of the event
1542  * @pc: The state of the preempt count at the start of the event.
1543  *
1544  * This is a helper function to handle triggers that require data
1545  * from the event itself. It also tests the event against filters and
1546  * if the event is soft disabled and should be discarded.
1547  *
1548  * Same as event_trigger_unlock_commit() but calls
1549  * trace_buffer_unlock_commit_regs() instead of trace_buffer_unlock_commit().
1550  */
1551 static inline void
1552 event_trigger_unlock_commit_regs(struct trace_event_file *file,
1553 				 struct trace_buffer *buffer,
1554 				 struct ring_buffer_event *event,
1555 				 void *entry, unsigned long irq_flags, int pc,
1556 				 struct pt_regs *regs)
1557 {
1558 	enum event_trigger_type tt = ETT_NONE;
1559 
1560 	if (!__event_trigger_test_discard(file, buffer, event, entry, &tt))
1561 		trace_buffer_unlock_commit_regs(file->tr, buffer, event,
1562 						irq_flags, pc, regs);
1563 
1564 	if (tt)
1565 		event_triggers_post_call(file, tt);
1566 }
1567 
1568 #define FILTER_PRED_INVALID	((unsigned short)-1)
1569 #define FILTER_PRED_IS_RIGHT	(1 << 15)
1570 #define FILTER_PRED_FOLD	(1 << 15)
1571 
1572 /*
1573  * The max preds is the size of unsigned short with
1574  * two flags at the MSBs. One bit is used for both the IS_RIGHT
1575  * and FOLD flags. The other is reserved.
1576  *
1577  * 2^14 preds is way more than enough.
1578  */
1579 #define MAX_FILTER_PRED		16384
1580 
1581 struct filter_pred;
1582 struct regex;
1583 
1584 typedef int (*filter_pred_fn_t) (struct filter_pred *pred, void *event);
1585 
1586 typedef int (*regex_match_func)(char *str, struct regex *r, int len);
1587 
1588 enum regex_type {
1589 	MATCH_FULL = 0,
1590 	MATCH_FRONT_ONLY,
1591 	MATCH_MIDDLE_ONLY,
1592 	MATCH_END_ONLY,
1593 	MATCH_GLOB,
1594 	MATCH_INDEX,
1595 };
1596 
1597 struct regex {
1598 	char			pattern[MAX_FILTER_STR_VAL];
1599 	int			len;
1600 	int			field_len;
1601 	regex_match_func	match;
1602 };
1603 
1604 struct filter_pred {
1605 	filter_pred_fn_t 	fn;
1606 	u64 			val;
1607 	struct regex		regex;
1608 	unsigned short		*ops;
1609 	struct ftrace_event_field *field;
1610 	int 			offset;
1611 	int			not;
1612 	int 			op;
1613 };
1614 
1615 static inline bool is_string_field(struct ftrace_event_field *field)
1616 {
1617 	return field->filter_type == FILTER_DYN_STRING ||
1618 	       field->filter_type == FILTER_STATIC_STRING ||
1619 	       field->filter_type == FILTER_PTR_STRING ||
1620 	       field->filter_type == FILTER_COMM;
1621 }
1622 
1623 static inline bool is_function_field(struct ftrace_event_field *field)
1624 {
1625 	return field->filter_type == FILTER_TRACE_FN;
1626 }
1627 
1628 extern enum regex_type
1629 filter_parse_regex(char *buff, int len, char **search, int *not);
1630 extern void print_event_filter(struct trace_event_file *file,
1631 			       struct trace_seq *s);
1632 extern int apply_event_filter(struct trace_event_file *file,
1633 			      char *filter_string);
1634 extern int apply_subsystem_event_filter(struct trace_subsystem_dir *dir,
1635 					char *filter_string);
1636 extern void print_subsystem_event_filter(struct event_subsystem *system,
1637 					 struct trace_seq *s);
1638 extern int filter_assign_type(const char *type);
1639 extern int create_event_filter(struct trace_array *tr,
1640 			       struct trace_event_call *call,
1641 			       char *filter_str, bool set_str,
1642 			       struct event_filter **filterp);
1643 extern void free_event_filter(struct event_filter *filter);
1644 
1645 struct ftrace_event_field *
1646 trace_find_event_field(struct trace_event_call *call, char *name);
1647 
1648 extern void trace_event_enable_cmd_record(bool enable);
1649 extern void trace_event_enable_tgid_record(bool enable);
1650 
1651 extern int event_trace_init(void);
1652 extern int event_trace_add_tracer(struct dentry *parent, struct trace_array *tr);
1653 extern int event_trace_del_tracer(struct trace_array *tr);
1654 
1655 extern struct trace_event_file *__find_event_file(struct trace_array *tr,
1656 						  const char *system,
1657 						  const char *event);
1658 extern struct trace_event_file *find_event_file(struct trace_array *tr,
1659 						const char *system,
1660 						const char *event);
1661 
1662 static inline void *event_file_data(struct file *filp)
1663 {
1664 	return READ_ONCE(file_inode(filp)->i_private);
1665 }
1666 
1667 extern struct mutex event_mutex;
1668 extern struct list_head ftrace_events;
1669 
1670 extern const struct file_operations event_trigger_fops;
1671 extern const struct file_operations event_hist_fops;
1672 extern const struct file_operations event_hist_debug_fops;
1673 extern const struct file_operations event_inject_fops;
1674 
1675 #ifdef CONFIG_HIST_TRIGGERS
1676 extern int register_trigger_hist_cmd(void);
1677 extern int register_trigger_hist_enable_disable_cmds(void);
1678 #else
1679 static inline int register_trigger_hist_cmd(void) { return 0; }
1680 static inline int register_trigger_hist_enable_disable_cmds(void) { return 0; }
1681 #endif
1682 
1683 extern int register_trigger_cmds(void);
1684 extern void clear_event_triggers(struct trace_array *tr);
1685 
1686 struct event_trigger_data {
1687 	unsigned long			count;
1688 	int				ref;
1689 	struct event_trigger_ops	*ops;
1690 	struct event_command		*cmd_ops;
1691 	struct event_filter __rcu	*filter;
1692 	char				*filter_str;
1693 	void				*private_data;
1694 	bool				paused;
1695 	bool				paused_tmp;
1696 	struct list_head		list;
1697 	char				*name;
1698 	struct list_head		named_list;
1699 	struct event_trigger_data	*named_data;
1700 };
1701 
1702 /* Avoid typos */
1703 #define ENABLE_EVENT_STR	"enable_event"
1704 #define DISABLE_EVENT_STR	"disable_event"
1705 #define ENABLE_HIST_STR		"enable_hist"
1706 #define DISABLE_HIST_STR	"disable_hist"
1707 
1708 struct enable_trigger_data {
1709 	struct trace_event_file		*file;
1710 	bool				enable;
1711 	bool				hist;
1712 };
1713 
1714 extern int event_enable_trigger_print(struct seq_file *m,
1715 				      struct event_trigger_ops *ops,
1716 				      struct event_trigger_data *data);
1717 extern void event_enable_trigger_free(struct event_trigger_ops *ops,
1718 				      struct event_trigger_data *data);
1719 extern int event_enable_trigger_func(struct event_command *cmd_ops,
1720 				     struct trace_event_file *file,
1721 				     char *glob, char *cmd, char *param);
1722 extern int event_enable_register_trigger(char *glob,
1723 					 struct event_trigger_ops *ops,
1724 					 struct event_trigger_data *data,
1725 					 struct trace_event_file *file);
1726 extern void event_enable_unregister_trigger(char *glob,
1727 					    struct event_trigger_ops *ops,
1728 					    struct event_trigger_data *test,
1729 					    struct trace_event_file *file);
1730 extern void trigger_data_free(struct event_trigger_data *data);
1731 extern int event_trigger_init(struct event_trigger_ops *ops,
1732 			      struct event_trigger_data *data);
1733 extern int trace_event_trigger_enable_disable(struct trace_event_file *file,
1734 					      int trigger_enable);
1735 extern void update_cond_flag(struct trace_event_file *file);
1736 extern int set_trigger_filter(char *filter_str,
1737 			      struct event_trigger_data *trigger_data,
1738 			      struct trace_event_file *file);
1739 extern struct event_trigger_data *find_named_trigger(const char *name);
1740 extern bool is_named_trigger(struct event_trigger_data *test);
1741 extern int save_named_trigger(const char *name,
1742 			      struct event_trigger_data *data);
1743 extern void del_named_trigger(struct event_trigger_data *data);
1744 extern void pause_named_trigger(struct event_trigger_data *data);
1745 extern void unpause_named_trigger(struct event_trigger_data *data);
1746 extern void set_named_trigger_data(struct event_trigger_data *data,
1747 				   struct event_trigger_data *named_data);
1748 extern struct event_trigger_data *
1749 get_named_trigger_data(struct event_trigger_data *data);
1750 extern int register_event_command(struct event_command *cmd);
1751 extern int unregister_event_command(struct event_command *cmd);
1752 extern int register_trigger_hist_enable_disable_cmds(void);
1753 
1754 /**
1755  * struct event_trigger_ops - callbacks for trace event triggers
1756  *
1757  * The methods in this structure provide per-event trigger hooks for
1758  * various trigger operations.
1759  *
1760  * All the methods below, except for @init() and @free(), must be
1761  * implemented.
1762  *
1763  * @func: The trigger 'probe' function called when the triggering
1764  *	event occurs.  The data passed into this callback is the data
1765  *	that was supplied to the event_command @reg() function that
1766  *	registered the trigger (see struct event_command) along with
1767  *	the trace record, rec.
1768  *
1769  * @init: An optional initialization function called for the trigger
1770  *	when the trigger is registered (via the event_command reg()
1771  *	function).  This can be used to perform per-trigger
1772  *	initialization such as incrementing a per-trigger reference
1773  *	count, for instance.  This is usually implemented by the
1774  *	generic utility function @event_trigger_init() (see
1775  *	trace_event_triggers.c).
1776  *
1777  * @free: An optional de-initialization function called for the
1778  *	trigger when the trigger is unregistered (via the
1779  *	event_command @reg() function).  This can be used to perform
1780  *	per-trigger de-initialization such as decrementing a
1781  *	per-trigger reference count and freeing corresponding trigger
1782  *	data, for instance.  This is usually implemented by the
1783  *	generic utility function @event_trigger_free() (see
1784  *	trace_event_triggers.c).
1785  *
1786  * @print: The callback function invoked to have the trigger print
1787  *	itself.  This is usually implemented by a wrapper function
1788  *	that calls the generic utility function @event_trigger_print()
1789  *	(see trace_event_triggers.c).
1790  */
1791 struct event_trigger_ops {
1792 	void			(*func)(struct event_trigger_data *data,
1793 					void *rec,
1794 					struct ring_buffer_event *rbe);
1795 	int			(*init)(struct event_trigger_ops *ops,
1796 					struct event_trigger_data *data);
1797 	void			(*free)(struct event_trigger_ops *ops,
1798 					struct event_trigger_data *data);
1799 	int			(*print)(struct seq_file *m,
1800 					 struct event_trigger_ops *ops,
1801 					 struct event_trigger_data *data);
1802 };
1803 
1804 /**
1805  * struct event_command - callbacks and data members for event commands
1806  *
1807  * Event commands are invoked by users by writing the command name
1808  * into the 'trigger' file associated with a trace event.  The
1809  * parameters associated with a specific invocation of an event
1810  * command are used to create an event trigger instance, which is
1811  * added to the list of trigger instances associated with that trace
1812  * event.  When the event is hit, the set of triggers associated with
1813  * that event is invoked.
1814  *
1815  * The data members in this structure provide per-event command data
1816  * for various event commands.
1817  *
1818  * All the data members below, except for @post_trigger, must be set
1819  * for each event command.
1820  *
1821  * @name: The unique name that identifies the event command.  This is
1822  *	the name used when setting triggers via trigger files.
1823  *
1824  * @trigger_type: A unique id that identifies the event command
1825  *	'type'.  This value has two purposes, the first to ensure that
1826  *	only one trigger of the same type can be set at a given time
1827  *	for a particular event e.g. it doesn't make sense to have both
1828  *	a traceon and traceoff trigger attached to a single event at
1829  *	the same time, so traceon and traceoff have the same type
1830  *	though they have different names.  The @trigger_type value is
1831  *	also used as a bit value for deferring the actual trigger
1832  *	action until after the current event is finished.  Some
1833  *	commands need to do this if they themselves log to the trace
1834  *	buffer (see the @post_trigger() member below).  @trigger_type
1835  *	values are defined by adding new values to the trigger_type
1836  *	enum in include/linux/trace_events.h.
1837  *
1838  * @flags: See the enum event_command_flags below.
1839  *
1840  * All the methods below, except for @set_filter() and @unreg_all(),
1841  * must be implemented.
1842  *
1843  * @func: The callback function responsible for parsing and
1844  *	registering the trigger written to the 'trigger' file by the
1845  *	user.  It allocates the trigger instance and registers it with
1846  *	the appropriate trace event.  It makes use of the other
1847  *	event_command callback functions to orchestrate this, and is
1848  *	usually implemented by the generic utility function
1849  *	@event_trigger_callback() (see trace_event_triggers.c).
1850  *
1851  * @reg: Adds the trigger to the list of triggers associated with the
1852  *	event, and enables the event trigger itself, after
1853  *	initializing it (via the event_trigger_ops @init() function).
1854  *	This is also where commands can use the @trigger_type value to
1855  *	make the decision as to whether or not multiple instances of
1856  *	the trigger should be allowed.  This is usually implemented by
1857  *	the generic utility function @register_trigger() (see
1858  *	trace_event_triggers.c).
1859  *
1860  * @unreg: Removes the trigger from the list of triggers associated
1861  *	with the event, and disables the event trigger itself, after
1862  *	initializing it (via the event_trigger_ops @free() function).
1863  *	This is usually implemented by the generic utility function
1864  *	@unregister_trigger() (see trace_event_triggers.c).
1865  *
1866  * @unreg_all: An optional function called to remove all the triggers
1867  *	from the list of triggers associated with the event.  Called
1868  *	when a trigger file is opened in truncate mode.
1869  *
1870  * @set_filter: An optional function called to parse and set a filter
1871  *	for the trigger.  If no @set_filter() method is set for the
1872  *	event command, filters set by the user for the command will be
1873  *	ignored.  This is usually implemented by the generic utility
1874  *	function @set_trigger_filter() (see trace_event_triggers.c).
1875  *
1876  * @get_trigger_ops: The callback function invoked to retrieve the
1877  *	event_trigger_ops implementation associated with the command.
1878  */
1879 struct event_command {
1880 	struct list_head	list;
1881 	char			*name;
1882 	enum event_trigger_type	trigger_type;
1883 	int			flags;
1884 	int			(*func)(struct event_command *cmd_ops,
1885 					struct trace_event_file *file,
1886 					char *glob, char *cmd, char *params);
1887 	int			(*reg)(char *glob,
1888 				       struct event_trigger_ops *ops,
1889 				       struct event_trigger_data *data,
1890 				       struct trace_event_file *file);
1891 	void			(*unreg)(char *glob,
1892 					 struct event_trigger_ops *ops,
1893 					 struct event_trigger_data *data,
1894 					 struct trace_event_file *file);
1895 	void			(*unreg_all)(struct trace_event_file *file);
1896 	int			(*set_filter)(char *filter_str,
1897 					      struct event_trigger_data *data,
1898 					      struct trace_event_file *file);
1899 	struct event_trigger_ops *(*get_trigger_ops)(char *cmd, char *param);
1900 };
1901 
1902 /**
1903  * enum event_command_flags - flags for struct event_command
1904  *
1905  * @POST_TRIGGER: A flag that says whether or not this command needs
1906  *	to have its action delayed until after the current event has
1907  *	been closed.  Some triggers need to avoid being invoked while
1908  *	an event is currently in the process of being logged, since
1909  *	the trigger may itself log data into the trace buffer.  Thus
1910  *	we make sure the current event is committed before invoking
1911  *	those triggers.  To do that, the trigger invocation is split
1912  *	in two - the first part checks the filter using the current
1913  *	trace record; if a command has the @post_trigger flag set, it
1914  *	sets a bit for itself in the return value, otherwise it
1915  *	directly invokes the trigger.  Once all commands have been
1916  *	either invoked or set their return flag, the current record is
1917  *	either committed or discarded.  At that point, if any commands
1918  *	have deferred their triggers, those commands are finally
1919  *	invoked following the close of the current event.  In other
1920  *	words, if the event_trigger_ops @func() probe implementation
1921  *	itself logs to the trace buffer, this flag should be set,
1922  *	otherwise it can be left unspecified.
1923  *
1924  * @NEEDS_REC: A flag that says whether or not this command needs
1925  *	access to the trace record in order to perform its function,
1926  *	regardless of whether or not it has a filter associated with
1927  *	it (filters make a trigger require access to the trace record
1928  *	but are not always present).
1929  */
1930 enum event_command_flags {
1931 	EVENT_CMD_FL_POST_TRIGGER	= 1,
1932 	EVENT_CMD_FL_NEEDS_REC		= 2,
1933 };
1934 
1935 static inline bool event_command_post_trigger(struct event_command *cmd_ops)
1936 {
1937 	return cmd_ops->flags & EVENT_CMD_FL_POST_TRIGGER;
1938 }
1939 
1940 static inline bool event_command_needs_rec(struct event_command *cmd_ops)
1941 {
1942 	return cmd_ops->flags & EVENT_CMD_FL_NEEDS_REC;
1943 }
1944 
1945 extern int trace_event_enable_disable(struct trace_event_file *file,
1946 				      int enable, int soft_disable);
1947 extern int tracing_alloc_snapshot(void);
1948 extern void tracing_snapshot_cond(struct trace_array *tr, void *cond_data);
1949 extern int tracing_snapshot_cond_enable(struct trace_array *tr, void *cond_data, cond_update_fn_t update);
1950 
1951 extern int tracing_snapshot_cond_disable(struct trace_array *tr);
1952 extern void *tracing_cond_snapshot_data(struct trace_array *tr);
1953 
1954 extern const char *__start___trace_bprintk_fmt[];
1955 extern const char *__stop___trace_bprintk_fmt[];
1956 
1957 extern const char *__start___tracepoint_str[];
1958 extern const char *__stop___tracepoint_str[];
1959 
1960 void trace_printk_control(bool enabled);
1961 void trace_printk_start_comm(void);
1962 int trace_keep_overwrite(struct tracer *tracer, u32 mask, int set);
1963 int set_tracer_flag(struct trace_array *tr, unsigned int mask, int enabled);
1964 
1965 /* Used from boot time tracer */
1966 extern int trace_set_options(struct trace_array *tr, char *option);
1967 extern int tracing_set_tracer(struct trace_array *tr, const char *buf);
1968 extern ssize_t tracing_resize_ring_buffer(struct trace_array *tr,
1969 					  unsigned long size, int cpu_id);
1970 extern int tracing_set_cpumask(struct trace_array *tr,
1971 				cpumask_var_t tracing_cpumask_new);
1972 
1973 
1974 #define MAX_EVENT_NAME_LEN	64
1975 
1976 extern int trace_run_command(const char *buf, int (*createfn)(int, char**));
1977 extern ssize_t trace_parse_run_command(struct file *file,
1978 		const char __user *buffer, size_t count, loff_t *ppos,
1979 		int (*createfn)(int, char**));
1980 
1981 extern unsigned int err_pos(char *cmd, const char *str);
1982 extern void tracing_log_err(struct trace_array *tr,
1983 			    const char *loc, const char *cmd,
1984 			    const char **errs, u8 type, u8 pos);
1985 
1986 /*
1987  * Normal trace_printk() and friends allocates special buffers
1988  * to do the manipulation, as well as saves the print formats
1989  * into sections to display. But the trace infrastructure wants
1990  * to use these without the added overhead at the price of being
1991  * a bit slower (used mainly for warnings, where we don't care
1992  * about performance). The internal_trace_puts() is for such
1993  * a purpose.
1994  */
1995 #define internal_trace_puts(str) __trace_puts(_THIS_IP_, str, strlen(str))
1996 
1997 #undef FTRACE_ENTRY
1998 #define FTRACE_ENTRY(call, struct_name, id, tstruct, print)	\
1999 	extern struct trace_event_call					\
2000 	__aligned(4) event_##call;
2001 #undef FTRACE_ENTRY_DUP
2002 #define FTRACE_ENTRY_DUP(call, struct_name, id, tstruct, print)	\
2003 	FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print))
2004 #undef FTRACE_ENTRY_PACKED
2005 #define FTRACE_ENTRY_PACKED(call, struct_name, id, tstruct, print) \
2006 	FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print))
2007 
2008 #include "trace_entries.h"
2009 
2010 #if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_FUNCTION_TRACER)
2011 int perf_ftrace_event_register(struct trace_event_call *call,
2012 			       enum trace_reg type, void *data);
2013 #else
2014 #define perf_ftrace_event_register NULL
2015 #endif
2016 
2017 #ifdef CONFIG_FTRACE_SYSCALLS
2018 void init_ftrace_syscalls(void);
2019 const char *get_syscall_name(int syscall);
2020 #else
2021 static inline void init_ftrace_syscalls(void) { }
2022 static inline const char *get_syscall_name(int syscall)
2023 {
2024 	return NULL;
2025 }
2026 #endif
2027 
2028 #ifdef CONFIG_EVENT_TRACING
2029 void trace_event_init(void);
2030 void trace_event_eval_update(struct trace_eval_map **map, int len);
2031 /* Used from boot time tracer */
2032 extern int ftrace_set_clr_event(struct trace_array *tr, char *buf, int set);
2033 extern int trigger_process_regex(struct trace_event_file *file, char *buff);
2034 #else
2035 static inline void __init trace_event_init(void) { }
2036 static inline void trace_event_eval_update(struct trace_eval_map **map, int len) { }
2037 #endif
2038 
2039 #ifdef CONFIG_TRACER_SNAPSHOT
2040 void tracing_snapshot_instance(struct trace_array *tr);
2041 int tracing_alloc_snapshot_instance(struct trace_array *tr);
2042 #else
2043 static inline void tracing_snapshot_instance(struct trace_array *tr) { }
2044 static inline int tracing_alloc_snapshot_instance(struct trace_array *tr)
2045 {
2046 	return 0;
2047 }
2048 #endif
2049 
2050 #ifdef CONFIG_PREEMPT_TRACER
2051 void tracer_preempt_on(unsigned long a0, unsigned long a1);
2052 void tracer_preempt_off(unsigned long a0, unsigned long a1);
2053 #else
2054 static inline void tracer_preempt_on(unsigned long a0, unsigned long a1) { }
2055 static inline void tracer_preempt_off(unsigned long a0, unsigned long a1) { }
2056 #endif
2057 #ifdef CONFIG_IRQSOFF_TRACER
2058 void tracer_hardirqs_on(unsigned long a0, unsigned long a1);
2059 void tracer_hardirqs_off(unsigned long a0, unsigned long a1);
2060 #else
2061 static inline void tracer_hardirqs_on(unsigned long a0, unsigned long a1) { }
2062 static inline void tracer_hardirqs_off(unsigned long a0, unsigned long a1) { }
2063 #endif
2064 
2065 extern struct trace_iterator *tracepoint_print_iter;
2066 
2067 /*
2068  * Reset the state of the trace_iterator so that it can read consumed data.
2069  * Normally, the trace_iterator is used for reading the data when it is not
2070  * consumed, and must retain state.
2071  */
2072 static __always_inline void trace_iterator_reset(struct trace_iterator *iter)
2073 {
2074 	const size_t offset = offsetof(struct trace_iterator, seq);
2075 
2076 	/*
2077 	 * Keep gcc from complaining about overwriting more than just one
2078 	 * member in the structure.
2079 	 */
2080 	memset((char *)iter + offset, 0, sizeof(struct trace_iterator) - offset);
2081 
2082 	iter->pos = -1;
2083 }
2084 
2085 #endif /* _LINUX_KERNEL_TRACE_H */
2086