1 // SPDX-License-Identifier: GPL-2.0
3 #ifndef _LINUX_KERNEL_TRACE_H
4 #define _LINUX_KERNEL_TRACE_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 #include <linux/ctype.h>
24 #ifdef CONFIG_FTRACE_SYSCALLS
25 #include <asm/unistd.h> /* For NR_SYSCALLS */
26 #include <asm/syscall.h> /* some archs define it here */
30 __TRACE_FIRST_TYPE = 0,
54 #define __field(type, item) type item;
57 #define __field_fn(type, item) type item;
60 #define __field_struct(type, item) __field(type, item)
63 #define __field_desc(type, container, item)
66 #define __field_packed(type, container, item)
69 #define __array(type, item, size) type item[size];
72 #define __array_desc(type, container, item, size)
74 #undef __dynamic_array
75 #define __dynamic_array(type, item) type item[];
78 #define F_STRUCT(args...) args
81 #define FTRACE_ENTRY(name, struct_name, id, tstruct, print) \
82 struct struct_name { \
83 struct trace_entry ent; \
87 #undef FTRACE_ENTRY_DUP
88 #define FTRACE_ENTRY_DUP(name, name_struct, id, tstruct, printk)
90 #undef FTRACE_ENTRY_REG
91 #define FTRACE_ENTRY_REG(name, struct_name, id, tstruct, print, regfn) \
92 FTRACE_ENTRY(name, struct_name, id, PARAMS(tstruct), PARAMS(print))
94 #undef FTRACE_ENTRY_PACKED
95 #define FTRACE_ENTRY_PACKED(name, struct_name, id, tstruct, print) \
96 FTRACE_ENTRY(name, struct_name, id, PARAMS(tstruct), PARAMS(print)) __packed
98 #include "trace_entries.h"
100 /* Use this for memory failure errors */
101 #define MEM_FAIL(condition, fmt, ...) ({ \
102 static bool __section(.data.once) __warned; \
103 int __ret_warn_once = !!(condition); \
105 if (unlikely(__ret_warn_once && !__warned)) { \
107 pr_err("ERROR: " fmt, ##__VA_ARGS__); \
109 unlikely(__ret_warn_once); \
113 * syscalls are special, and need special handling, this is why
114 * they are not included in trace_entries.h
116 struct syscall_trace_enter {
117 struct trace_entry ent;
119 unsigned long args[];
122 struct syscall_trace_exit {
123 struct trace_entry ent;
128 struct kprobe_trace_entry_head {
129 struct trace_entry ent;
133 struct kretprobe_trace_entry_head {
134 struct trace_entry ent;
136 unsigned long ret_ip;
140 * trace_flag_type is an enumeration that holds different
141 * states when a trace occurs. These are:
142 * IRQS_OFF - interrupts were disabled
143 * IRQS_NOSUPPORT - arch does not support irqs_disabled_flags
144 * NEED_RESCHED - reschedule is requested
145 * HARDIRQ - inside an interrupt handler
146 * SOFTIRQ - inside a softirq handler
148 enum trace_flag_type {
149 TRACE_FLAG_IRQS_OFF = 0x01,
150 TRACE_FLAG_IRQS_NOSUPPORT = 0x02,
151 TRACE_FLAG_NEED_RESCHED = 0x04,
152 TRACE_FLAG_HARDIRQ = 0x08,
153 TRACE_FLAG_SOFTIRQ = 0x10,
154 TRACE_FLAG_PREEMPT_RESCHED = 0x20,
155 TRACE_FLAG_NMI = 0x40,
158 #define TRACE_BUF_SIZE 1024
163 * The CPU trace array - it consists of thousands of trace entries
164 * plus some other descriptor data: (for example which task started
167 struct trace_array_cpu {
169 void *buffer_page; /* ring buffer spare */
171 unsigned long entries;
172 unsigned long saved_latency;
173 unsigned long critical_start;
174 unsigned long critical_end;
175 unsigned long critical_sequence;
177 unsigned long policy;
178 unsigned long rt_priority;
179 unsigned long skipped_entries;
180 u64 preempt_timestamp;
183 char comm[TASK_COMM_LEN];
185 #ifdef CONFIG_FUNCTION_TRACER
186 int ftrace_ignore_pid;
192 struct trace_option_dentry;
194 struct array_buffer {
195 struct trace_array *tr;
196 struct trace_buffer *buffer;
197 struct trace_array_cpu __percpu *data;
202 #define TRACE_FLAGS_MAX_SIZE 32
204 struct trace_options {
205 struct tracer *tracer;
206 struct trace_option_dentry *topts;
209 struct trace_pid_list {
216 TRACE_NO_PIDS = BIT(1),
219 static inline bool pid_type_enabled(int type, struct trace_pid_list *pid_list,
220 struct trace_pid_list *no_pid_list)
222 /* Return true if the pid list in type has pids */
223 return ((type & TRACE_PIDS) && pid_list) ||
224 ((type & TRACE_NO_PIDS) && no_pid_list);
227 static inline bool still_need_pid_events(int type, struct trace_pid_list *pid_list,
228 struct trace_pid_list *no_pid_list)
231 * Turning off what is in @type, return true if the "other"
232 * pid list, still has pids in it.
234 return (!(type & TRACE_PIDS) && pid_list) ||
235 (!(type & TRACE_NO_PIDS) && no_pid_list);
238 typedef bool (*cond_update_fn_t)(struct trace_array *tr, void *cond_data);
241 * struct cond_snapshot - conditional snapshot data and callback
243 * The cond_snapshot structure encapsulates a callback function and
244 * data associated with the snapshot for a given tracing instance.
246 * When a snapshot is taken conditionally, by invoking
247 * tracing_snapshot_cond(tr, cond_data), the cond_data passed in is
248 * passed in turn to the cond_snapshot.update() function. That data
249 * can be compared by the update() implementation with the cond_data
250 * contained within the struct cond_snapshot instance associated with
251 * the trace_array. Because the tr->max_lock is held throughout the
252 * update() call, the update() function can directly retrieve the
253 * cond_snapshot and cond_data associated with the per-instance
254 * snapshot associated with the trace_array.
256 * The cond_snapshot.update() implementation can save data to be
257 * associated with the snapshot if it decides to, and returns 'true'
258 * in that case, or it returns 'false' if the conditional snapshot
259 * shouldn't be taken.
261 * The cond_snapshot instance is created and associated with the
262 * user-defined cond_data by tracing_cond_snapshot_enable().
263 * Likewise, the cond_snapshot instance is destroyed and is no longer
264 * associated with the trace instance by
265 * tracing_cond_snapshot_disable().
267 * The method below is required.
269 * @update: When a conditional snapshot is invoked, the update()
270 * callback function is invoked with the tr->max_lock held. The
271 * update() implementation signals whether or not to actually
272 * take the snapshot, by returning 'true' if so, 'false' if no
273 * snapshot should be taken. Because the max_lock is held for
274 * the duration of update(), the implementation is safe to
275 * directly retrieved and save any implementation data it needs
276 * to in association with the snapshot.
278 struct cond_snapshot {
280 cond_update_fn_t update;
284 * The trace array - an array of per-CPU trace arrays. This is the
285 * highest level data structure that individual tracers deal with.
286 * They have on/off state as well:
289 struct list_head list;
291 struct array_buffer array_buffer;
292 #ifdef CONFIG_TRACER_MAX_TRACE
294 * The max_buffer is used to snapshot the trace when a maximum
295 * latency is reached, or when the user initiates a snapshot.
296 * Some tracers will use this to store a maximum trace while
297 * it continues examining live traces.
299 * The buffers for the max_buffer are set up the same as the array_buffer
300 * When a snapshot is taken, the buffer of the max_buffer is swapped
301 * with the buffer of the array_buffer and the buffers are reset for
302 * the array_buffer so the tracing can continue.
304 struct array_buffer max_buffer;
305 bool allocated_snapshot;
307 #if defined(CONFIG_TRACER_MAX_TRACE) || defined(CONFIG_HWLAT_TRACER)
308 unsigned long max_latency;
309 #ifdef CONFIG_FSNOTIFY
310 struct dentry *d_max_latency;
311 struct work_struct fsnotify_work;
312 struct irq_work fsnotify_irqwork;
315 struct trace_pid_list __rcu *filtered_pids;
316 struct trace_pid_list __rcu *filtered_no_pids;
318 * max_lock is used to protect the swapping of buffers
319 * when taking a max snapshot. The buffers themselves are
320 * protected by per_cpu spinlocks. But the action of the swap
321 * needs its own lock.
323 * This is defined as a arch_spinlock_t in order to help
324 * with performance when lockdep debugging is enabled.
326 * It is also used in other places outside the update_max_tr
327 * so it needs to be defined outside of the
328 * CONFIG_TRACER_MAX_TRACE.
330 arch_spinlock_t max_lock;
332 #ifdef CONFIG_FTRACE_SYSCALLS
333 int sys_refcount_enter;
334 int sys_refcount_exit;
335 struct trace_event_file __rcu *enter_syscall_files[NR_syscalls];
336 struct trace_event_file __rcu *exit_syscall_files[NR_syscalls];
343 unsigned int n_err_log_entries;
344 struct tracer *current_trace;
345 unsigned int trace_flags;
346 unsigned char trace_flags_index[TRACE_FLAGS_MAX_SIZE];
348 raw_spinlock_t start_lock;
349 struct list_head err_log;
351 struct dentry *options;
352 struct dentry *percpu_dir;
353 struct dentry *event_dir;
354 struct trace_options *topts;
355 struct list_head systems;
356 struct list_head events;
357 struct trace_event_file *trace_marker_file;
358 cpumask_var_t tracing_cpumask; /* only trace on set CPUs */
361 #ifdef CONFIG_FUNCTION_TRACER
362 struct ftrace_ops *ops;
363 struct trace_pid_list __rcu *function_pids;
364 struct trace_pid_list __rcu *function_no_pids;
365 #ifdef CONFIG_DYNAMIC_FTRACE
366 /* All of these are protected by the ftrace_lock */
367 struct list_head func_probes;
368 struct list_head mod_trace;
369 struct list_head mod_notrace;
371 /* function tracing enabled */
372 int function_enabled;
374 int time_stamp_abs_ref;
375 struct list_head hist_vars;
376 #ifdef CONFIG_TRACER_SNAPSHOT
377 struct cond_snapshot *cond_snapshot;
382 TRACE_ARRAY_FL_GLOBAL = (1 << 0)
385 extern struct list_head ftrace_trace_arrays;
387 extern struct mutex trace_types_lock;
389 extern int trace_array_get(struct trace_array *tr);
390 extern int tracing_check_open_get_tr(struct trace_array *tr);
391 extern struct trace_array *trace_array_find(const char *instance);
392 extern struct trace_array *trace_array_find_get(const char *instance);
394 extern int tracing_set_time_stamp_abs(struct trace_array *tr, bool abs);
395 extern int tracing_set_clock(struct trace_array *tr, const char *clockstr);
397 extern bool trace_clock_in_ns(struct trace_array *tr);
400 * The global tracer (top) should be the first trace array added,
401 * but we check the flag anyway.
403 static inline struct trace_array *top_trace_array(void)
405 struct trace_array *tr;
407 if (list_empty(&ftrace_trace_arrays))
410 tr = list_entry(ftrace_trace_arrays.prev,
412 WARN_ON(!(tr->flags & TRACE_ARRAY_FL_GLOBAL));
416 #define FTRACE_CMP_TYPE(var, type) \
417 __builtin_types_compatible_p(typeof(var), type *)
420 #define IF_ASSIGN(var, entry, etype, id) \
421 if (FTRACE_CMP_TYPE(var, etype)) { \
422 var = (typeof(var))(entry); \
423 WARN_ON(id != 0 && (entry)->type != id); \
427 /* Will cause compile errors if type is not found. */
428 extern void __ftrace_bad_type(void);
431 * The trace_assign_type is a verifier that the entry type is
432 * the same as the type being assigned. To add new types simply
433 * add a line with the following format:
435 * IF_ASSIGN(var, ent, type, id);
437 * Where "type" is the trace type that includes the trace_entry
438 * as the "ent" item. And "id" is the trace identifier that is
439 * used in the trace_type enum.
441 * If the type can have more than one id, then use zero.
443 #define trace_assign_type(var, ent) \
445 IF_ASSIGN(var, ent, struct ftrace_entry, TRACE_FN); \
446 IF_ASSIGN(var, ent, struct ctx_switch_entry, 0); \
447 IF_ASSIGN(var, ent, struct stack_entry, TRACE_STACK); \
448 IF_ASSIGN(var, ent, struct userstack_entry, TRACE_USER_STACK);\
449 IF_ASSIGN(var, ent, struct print_entry, TRACE_PRINT); \
450 IF_ASSIGN(var, ent, struct bprint_entry, TRACE_BPRINT); \
451 IF_ASSIGN(var, ent, struct bputs_entry, TRACE_BPUTS); \
452 IF_ASSIGN(var, ent, struct hwlat_entry, TRACE_HWLAT); \
453 IF_ASSIGN(var, ent, struct raw_data_entry, TRACE_RAW_DATA);\
454 IF_ASSIGN(var, ent, struct trace_mmiotrace_rw, \
456 IF_ASSIGN(var, ent, struct trace_mmiotrace_map, \
458 IF_ASSIGN(var, ent, struct trace_branch, TRACE_BRANCH); \
459 IF_ASSIGN(var, ent, struct ftrace_graph_ent_entry, \
461 IF_ASSIGN(var, ent, struct ftrace_graph_ret_entry, \
463 __ftrace_bad_type(); \
467 * An option specific to a tracer. This is a boolean value.
468 * The bit is the bit index that sets its value on the
469 * flags value in struct tracer_flags.
472 const char *name; /* Will appear on the trace_options file */
473 u32 bit; /* Mask assigned in val field in tracer_flags */
477 * The set of specific options for a tracer. Your tracer
478 * have to set the initial value of the flags val.
480 struct tracer_flags {
482 struct tracer_opt *opts;
483 struct tracer *trace;
486 /* Makes more easy to define a tracer opt */
487 #define TRACER_OPT(s, b) .name = #s, .bit = b
490 struct trace_option_dentry {
491 struct tracer_opt *opt;
492 struct tracer_flags *flags;
493 struct trace_array *tr;
494 struct dentry *entry;
498 * struct tracer - a specific tracer and its callbacks to interact with tracefs
499 * @name: the name chosen to select it on the available_tracers file
500 * @init: called when one switches to this tracer (echo name > current_tracer)
501 * @reset: called when one switches to another tracer
502 * @start: called when tracing is unpaused (echo 1 > tracing_on)
503 * @stop: called when tracing is paused (echo 0 > tracing_on)
504 * @update_thresh: called when tracing_thresh is updated
505 * @open: called when the trace file is opened
506 * @pipe_open: called when the trace_pipe file is opened
507 * @close: called when the trace file is released
508 * @pipe_close: called when the trace_pipe file is released
509 * @read: override the default read callback on trace_pipe
510 * @splice_read: override the default splice_read callback on trace_pipe
511 * @selftest: selftest to run on boot (see trace_selftest.c)
512 * @print_headers: override the first lines that describe your columns
513 * @print_line: callback that prints a trace
514 * @set_flag: signals one of your private flags changed (trace_options file)
515 * @flags: your private flags
519 int (*init)(struct trace_array *tr);
520 void (*reset)(struct trace_array *tr);
521 void (*start)(struct trace_array *tr);
522 void (*stop)(struct trace_array *tr);
523 int (*update_thresh)(struct trace_array *tr);
524 void (*open)(struct trace_iterator *iter);
525 void (*pipe_open)(struct trace_iterator *iter);
526 void (*close)(struct trace_iterator *iter);
527 void (*pipe_close)(struct trace_iterator *iter);
528 ssize_t (*read)(struct trace_iterator *iter,
529 struct file *filp, char __user *ubuf,
530 size_t cnt, loff_t *ppos);
531 ssize_t (*splice_read)(struct trace_iterator *iter,
534 struct pipe_inode_info *pipe,
537 #ifdef CONFIG_FTRACE_STARTUP_TEST
538 int (*selftest)(struct tracer *trace,
539 struct trace_array *tr);
541 void (*print_header)(struct seq_file *m);
542 enum print_line_t (*print_line)(struct trace_iterator *iter);
543 /* If you handled the flag setting, return 0 */
544 int (*set_flag)(struct trace_array *tr,
545 u32 old_flags, u32 bit, int set);
546 /* Return 0 if OK with change, else return non-zero */
547 int (*flag_changed)(struct trace_array *tr,
550 struct tracer_flags *flags;
553 bool allow_instances;
554 #ifdef CONFIG_TRACER_MAX_TRACE
557 /* True if tracer cannot be enabled in kernel param */
562 /* Only current can touch trace_recursion */
565 * For function tracing recursion:
566 * The order of these bits are important.
568 * When function tracing occurs, the following steps are made:
569 * If arch does not support a ftrace feature:
570 * call internal function (uses INTERNAL bits) which calls...
571 * If callback is registered to the "global" list, the list
572 * function is called and recursion checks the GLOBAL bits.
573 * then this function calls...
574 * The function callback, which can use the FTRACE bits to
575 * check for recursion.
577 * Now if the arch does not support a feature, and it calls
578 * the global list function which calls the ftrace callback
579 * all three of these steps will do a recursion protection.
580 * There's no reason to do one if the previous caller already
581 * did. The recursion that we are protecting against will
582 * go through the same steps again.
584 * To prevent the multiple recursion checks, if a recursion
585 * bit is set that is higher than the MAX bit of the current
586 * check, then we know that the check was made by the previous
587 * caller, and we can skip the current check.
590 /* Function recursion bits */
592 TRACE_FTRACE_NMI_BIT,
593 TRACE_FTRACE_IRQ_BIT,
594 TRACE_FTRACE_SIRQ_BIT,
596 /* INTERNAL_BITs must be greater than FTRACE_BITs */
598 TRACE_INTERNAL_NMI_BIT,
599 TRACE_INTERNAL_IRQ_BIT,
600 TRACE_INTERNAL_SIRQ_BIT,
604 * Abuse of the trace_recursion.
605 * As we need a way to maintain state if we are tracing the function
606 * graph in irq because we want to trace a particular function that
607 * was called in irq context but we have irq tracing off. Since this
608 * can only be modified by current, we can reuse trace_recursion.
612 /* Set if the function is in the set_graph_function file */
616 * In the very unlikely case that an interrupt came in
617 * at a start of graph tracing, and we want to trace
618 * the function in that interrupt, the depth can be greater
619 * than zero, because of the preempted start of a previous
620 * trace. In an even more unlikely case, depth could be 2
621 * if a softirq interrupted the start of graph tracing,
622 * followed by an interrupt preempting a start of graph
623 * tracing in the softirq, and depth can even be 3
624 * if an NMI came in at the start of an interrupt function
625 * that preempted a softirq start of a function that
626 * preempted normal context!!!! Luckily, it can't be
627 * greater than 3, so the next two bits are a mask
628 * of what the depth is when we set TRACE_GRAPH_BIT
631 TRACE_GRAPH_DEPTH_START_BIT,
632 TRACE_GRAPH_DEPTH_END_BIT,
635 * To implement set_graph_notrace, if this bit is set, we ignore
636 * function graph tracing of called functions, until the return
637 * function is called to clear it.
639 TRACE_GRAPH_NOTRACE_BIT,
642 #define trace_recursion_set(bit) do { (current)->trace_recursion |= (1<<(bit)); } while (0)
643 #define trace_recursion_clear(bit) do { (current)->trace_recursion &= ~(1<<(bit)); } while (0)
644 #define trace_recursion_test(bit) ((current)->trace_recursion & (1<<(bit)))
646 #define trace_recursion_depth() \
647 (((current)->trace_recursion >> TRACE_GRAPH_DEPTH_START_BIT) & 3)
648 #define trace_recursion_set_depth(depth) \
650 current->trace_recursion &= \
651 ~(3 << TRACE_GRAPH_DEPTH_START_BIT); \
652 current->trace_recursion |= \
653 ((depth) & 3) << TRACE_GRAPH_DEPTH_START_BIT; \
656 #define TRACE_CONTEXT_BITS 4
658 #define TRACE_FTRACE_START TRACE_FTRACE_BIT
659 #define TRACE_FTRACE_MAX ((1 << (TRACE_FTRACE_START + TRACE_CONTEXT_BITS)) - 1)
661 #define TRACE_LIST_START TRACE_INTERNAL_BIT
662 #define TRACE_LIST_MAX ((1 << (TRACE_LIST_START + TRACE_CONTEXT_BITS)) - 1)
664 #define TRACE_CONTEXT_MASK TRACE_LIST_MAX
666 static __always_inline int trace_get_context_bit(void)
670 if (in_interrupt()) {
684 static __always_inline int trace_test_and_set_recursion(int start, int max)
686 unsigned int val = current->trace_recursion;
689 /* A previous recursion check was made */
690 if ((val & TRACE_CONTEXT_MASK) > max)
693 bit = trace_get_context_bit() + start;
694 if (unlikely(val & (1 << bit)))
698 current->trace_recursion = val;
704 static __always_inline void trace_clear_recursion(int bit)
706 unsigned int val = current->trace_recursion;
715 current->trace_recursion = val;
718 static inline struct ring_buffer_iter *
719 trace_buffer_iter(struct trace_iterator *iter, int cpu)
721 return iter->buffer_iter ? iter->buffer_iter[cpu] : NULL;
724 int tracer_init(struct tracer *t, struct trace_array *tr);
725 int tracing_is_enabled(void);
726 void tracing_reset_online_cpus(struct array_buffer *buf);
727 void tracing_reset_current(int cpu);
728 void tracing_reset_all_online_cpus(void);
729 int tracing_open_generic(struct inode *inode, struct file *filp);
730 int tracing_open_generic_tr(struct inode *inode, struct file *filp);
731 bool tracing_is_disabled(void);
732 bool tracer_tracing_is_on(struct trace_array *tr);
733 void tracer_tracing_on(struct trace_array *tr);
734 void tracer_tracing_off(struct trace_array *tr);
735 struct dentry *trace_create_file(const char *name,
737 struct dentry *parent,
739 const struct file_operations *fops);
741 int tracing_init_dentry(void);
743 struct ring_buffer_event;
745 struct ring_buffer_event *
746 trace_buffer_lock_reserve(struct trace_buffer *buffer,
752 struct trace_entry *tracing_get_trace_entry(struct trace_array *tr,
753 struct trace_array_cpu *data);
755 struct trace_entry *trace_find_next_entry(struct trace_iterator *iter,
756 int *ent_cpu, u64 *ent_ts);
758 void trace_buffer_unlock_commit_nostack(struct trace_buffer *buffer,
759 struct ring_buffer_event *event);
761 int trace_empty(struct trace_iterator *iter);
763 void *trace_find_next_entry_inc(struct trace_iterator *iter);
765 void trace_init_global_iter(struct trace_iterator *iter);
767 void tracing_iter_reset(struct trace_iterator *iter, int cpu);
769 unsigned long trace_total_entries_cpu(struct trace_array *tr, int cpu);
770 unsigned long trace_total_entries(struct trace_array *tr);
772 void trace_function(struct trace_array *tr,
774 unsigned long parent_ip,
775 unsigned long flags, int pc);
776 void trace_graph_function(struct trace_array *tr,
778 unsigned long parent_ip,
779 unsigned long flags, int pc);
780 void trace_latency_header(struct seq_file *m);
781 void trace_default_header(struct seq_file *m);
782 void print_trace_header(struct seq_file *m, struct trace_iterator *iter);
783 int trace_empty(struct trace_iterator *iter);
785 void trace_graph_return(struct ftrace_graph_ret *trace);
786 int trace_graph_entry(struct ftrace_graph_ent *trace);
787 void set_graph_array(struct trace_array *tr);
789 void tracing_start_cmdline_record(void);
790 void tracing_stop_cmdline_record(void);
791 void tracing_start_tgid_record(void);
792 void tracing_stop_tgid_record(void);
794 int register_tracer(struct tracer *type);
795 int is_tracing_stopped(void);
797 loff_t tracing_lseek(struct file *file, loff_t offset, int whence);
799 extern cpumask_var_t __read_mostly tracing_buffer_mask;
801 #define for_each_tracing_cpu(cpu) \
802 for_each_cpu(cpu, tracing_buffer_mask)
804 extern unsigned long nsecs_to_usecs(unsigned long nsecs);
806 extern unsigned long tracing_thresh;
812 bool trace_find_filtered_pid(struct trace_pid_list *filtered_pids,
814 bool trace_ignore_this_task(struct trace_pid_list *filtered_pids,
815 struct trace_pid_list *filtered_no_pids,
816 struct task_struct *task);
817 void trace_filter_add_remove_task(struct trace_pid_list *pid_list,
818 struct task_struct *self,
819 struct task_struct *task);
820 void *trace_pid_next(struct trace_pid_list *pid_list, void *v, loff_t *pos);
821 void *trace_pid_start(struct trace_pid_list *pid_list, loff_t *pos);
822 int trace_pid_show(struct seq_file *m, void *v);
823 void trace_free_pid_list(struct trace_pid_list *pid_list);
824 int trace_pid_write(struct trace_pid_list *filtered_pids,
825 struct trace_pid_list **new_pid_list,
826 const char __user *ubuf, size_t cnt);
828 #ifdef CONFIG_TRACER_MAX_TRACE
829 void update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu,
831 void update_max_tr_single(struct trace_array *tr,
832 struct task_struct *tsk, int cpu);
833 #endif /* CONFIG_TRACER_MAX_TRACE */
835 #if (defined(CONFIG_TRACER_MAX_TRACE) || defined(CONFIG_HWLAT_TRACER)) && \
836 defined(CONFIG_FSNOTIFY)
838 void latency_fsnotify(struct trace_array *tr);
842 static inline void latency_fsnotify(struct trace_array *tr) { }
846 #ifdef CONFIG_STACKTRACE
847 void __trace_stack(struct trace_array *tr, unsigned long flags, int skip,
850 static inline void __trace_stack(struct trace_array *tr, unsigned long flags,
854 #endif /* CONFIG_STACKTRACE */
856 extern u64 ftrace_now(int cpu);
858 extern void trace_find_cmdline(int pid, char comm[]);
859 extern int trace_find_tgid(int pid);
860 extern void trace_event_follow_fork(struct trace_array *tr, bool enable);
862 #ifdef CONFIG_DYNAMIC_FTRACE
863 extern unsigned long ftrace_update_tot_cnt;
864 extern unsigned long ftrace_number_of_pages;
865 extern unsigned long ftrace_number_of_groups;
866 void ftrace_init_trace_array(struct trace_array *tr);
868 static inline void ftrace_init_trace_array(struct trace_array *tr) { }
870 #define DYN_FTRACE_TEST_NAME trace_selftest_dynamic_test_func
871 extern int DYN_FTRACE_TEST_NAME(void);
872 #define DYN_FTRACE_TEST_NAME2 trace_selftest_dynamic_test_func2
873 extern int DYN_FTRACE_TEST_NAME2(void);
875 extern bool ring_buffer_expanded;
876 extern bool tracing_selftest_disabled;
878 #ifdef CONFIG_FTRACE_STARTUP_TEST
879 extern int trace_selftest_startup_function(struct tracer *trace,
880 struct trace_array *tr);
881 extern int trace_selftest_startup_function_graph(struct tracer *trace,
882 struct trace_array *tr);
883 extern int trace_selftest_startup_irqsoff(struct tracer *trace,
884 struct trace_array *tr);
885 extern int trace_selftest_startup_preemptoff(struct tracer *trace,
886 struct trace_array *tr);
887 extern int trace_selftest_startup_preemptirqsoff(struct tracer *trace,
888 struct trace_array *tr);
889 extern int trace_selftest_startup_wakeup(struct tracer *trace,
890 struct trace_array *tr);
891 extern int trace_selftest_startup_nop(struct tracer *trace,
892 struct trace_array *tr);
893 extern int trace_selftest_startup_branch(struct tracer *trace,
894 struct trace_array *tr);
896 * Tracer data references selftest functions that only occur
897 * on boot up. These can be __init functions. Thus, when selftests
898 * are enabled, then the tracers need to reference __init functions.
900 #define __tracer_data __refdata
902 /* Tracers are seldom changed. Optimize when selftests are disabled. */
903 #define __tracer_data __read_mostly
904 #endif /* CONFIG_FTRACE_STARTUP_TEST */
906 extern void *head_page(struct trace_array_cpu *data);
907 extern unsigned long long ns2usecs(u64 nsec);
909 trace_vbprintk(unsigned long ip, const char *fmt, va_list args);
911 trace_vprintk(unsigned long ip, const char *fmt, va_list args);
913 trace_array_vprintk(struct trace_array *tr,
914 unsigned long ip, const char *fmt, va_list args);
915 int trace_array_printk_buf(struct trace_buffer *buffer,
916 unsigned long ip, const char *fmt, ...);
917 void trace_printk_seq(struct trace_seq *s);
918 enum print_line_t print_trace_line(struct trace_iterator *iter);
920 extern char trace_find_mark(unsigned long long duration);
924 struct ftrace_mod_load {
925 struct list_head list;
932 FTRACE_HASH_FL_MOD = (1 << 0),
936 unsigned long size_bits;
937 struct hlist_head *buckets;
943 struct ftrace_func_entry *
944 ftrace_lookup_ip(struct ftrace_hash *hash, unsigned long ip);
946 static __always_inline bool ftrace_hash_empty(struct ftrace_hash *hash)
948 return !hash || !(hash->count || (hash->flags & FTRACE_HASH_FL_MOD));
951 /* Standard output formatting function used for function return traces */
952 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
955 #define TRACE_GRAPH_PRINT_OVERRUN 0x1
956 #define TRACE_GRAPH_PRINT_CPU 0x2
957 #define TRACE_GRAPH_PRINT_OVERHEAD 0x4
958 #define TRACE_GRAPH_PRINT_PROC 0x8
959 #define TRACE_GRAPH_PRINT_DURATION 0x10
960 #define TRACE_GRAPH_PRINT_ABS_TIME 0x20
961 #define TRACE_GRAPH_PRINT_REL_TIME 0x40
962 #define TRACE_GRAPH_PRINT_IRQS 0x80
963 #define TRACE_GRAPH_PRINT_TAIL 0x100
964 #define TRACE_GRAPH_SLEEP_TIME 0x200
965 #define TRACE_GRAPH_GRAPH_TIME 0x400
966 #define TRACE_GRAPH_PRINT_FILL_SHIFT 28
967 #define TRACE_GRAPH_PRINT_FILL_MASK (0x3 << TRACE_GRAPH_PRINT_FILL_SHIFT)
969 extern void ftrace_graph_sleep_time_control(bool enable);
971 #ifdef CONFIG_FUNCTION_PROFILER
972 extern void ftrace_graph_graph_time_control(bool enable);
974 static inline void ftrace_graph_graph_time_control(bool enable) { }
977 extern enum print_line_t
978 print_graph_function_flags(struct trace_iterator *iter, u32 flags);
979 extern void print_graph_headers_flags(struct seq_file *s, u32 flags);
981 trace_print_graph_duration(unsigned long long duration, struct trace_seq *s);
982 extern void graph_trace_open(struct trace_iterator *iter);
983 extern void graph_trace_close(struct trace_iterator *iter);
984 extern int __trace_graph_entry(struct trace_array *tr,
985 struct ftrace_graph_ent *trace,
986 unsigned long flags, int pc);
987 extern void __trace_graph_return(struct trace_array *tr,
988 struct ftrace_graph_ret *trace,
989 unsigned long flags, int pc);
991 #ifdef CONFIG_DYNAMIC_FTRACE
992 extern struct ftrace_hash __rcu *ftrace_graph_hash;
993 extern struct ftrace_hash __rcu *ftrace_graph_notrace_hash;
995 static inline int ftrace_graph_addr(struct ftrace_graph_ent *trace)
997 unsigned long addr = trace->func;
999 struct ftrace_hash *hash;
1001 preempt_disable_notrace();
1004 * Have to open code "rcu_dereference_sched()" because the
1005 * function graph tracer can be called when RCU is not
1007 * Protected with schedule_on_each_cpu(ftrace_sync)
1009 hash = rcu_dereference_protected(ftrace_graph_hash, !preemptible());
1011 if (ftrace_hash_empty(hash)) {
1016 if (ftrace_lookup_ip(hash, addr)) {
1019 * This needs to be cleared on the return functions
1020 * when the depth is zero.
1022 trace_recursion_set(TRACE_GRAPH_BIT);
1023 trace_recursion_set_depth(trace->depth);
1026 * If no irqs are to be traced, but a set_graph_function
1027 * is set, and called by an interrupt handler, we still
1031 trace_recursion_set(TRACE_IRQ_BIT);
1033 trace_recursion_clear(TRACE_IRQ_BIT);
1038 preempt_enable_notrace();
1042 static inline void ftrace_graph_addr_finish(struct ftrace_graph_ret *trace)
1044 if (trace_recursion_test(TRACE_GRAPH_BIT) &&
1045 trace->depth == trace_recursion_depth())
1046 trace_recursion_clear(TRACE_GRAPH_BIT);
1049 static inline int ftrace_graph_notrace_addr(unsigned long addr)
1052 struct ftrace_hash *notrace_hash;
1054 preempt_disable_notrace();
1057 * Have to open code "rcu_dereference_sched()" because the
1058 * function graph tracer can be called when RCU is not
1060 * Protected with schedule_on_each_cpu(ftrace_sync)
1062 notrace_hash = rcu_dereference_protected(ftrace_graph_notrace_hash,
1065 if (ftrace_lookup_ip(notrace_hash, addr))
1068 preempt_enable_notrace();
1072 static inline int ftrace_graph_addr(struct ftrace_graph_ent *trace)
1077 static inline int ftrace_graph_notrace_addr(unsigned long addr)
1081 static inline void ftrace_graph_addr_finish(struct ftrace_graph_ret *trace)
1083 #endif /* CONFIG_DYNAMIC_FTRACE */
1085 extern unsigned int fgraph_max_depth;
1087 static inline bool ftrace_graph_ignore_func(struct ftrace_graph_ent *trace)
1089 /* trace it when it is-nested-in or is a function enabled. */
1090 return !(trace_recursion_test(TRACE_GRAPH_BIT) ||
1091 ftrace_graph_addr(trace)) ||
1092 (trace->depth < 0) ||
1093 (fgraph_max_depth && trace->depth >= fgraph_max_depth);
1096 #else /* CONFIG_FUNCTION_GRAPH_TRACER */
1097 static inline enum print_line_t
1098 print_graph_function_flags(struct trace_iterator *iter, u32 flags)
1100 return TRACE_TYPE_UNHANDLED;
1102 #endif /* CONFIG_FUNCTION_GRAPH_TRACER */
1104 extern struct list_head ftrace_pids;
1106 #ifdef CONFIG_FUNCTION_TRACER
1108 #define FTRACE_PID_IGNORE -1
1109 #define FTRACE_PID_TRACE -2
1111 struct ftrace_func_command {
1112 struct list_head list;
1114 int (*func)(struct trace_array *tr,
1115 struct ftrace_hash *hash,
1116 char *func, char *cmd,
1117 char *params, int enable);
1119 extern bool ftrace_filter_param __initdata;
1120 static inline int ftrace_trace_task(struct trace_array *tr)
1122 return this_cpu_read(tr->array_buffer.data->ftrace_ignore_pid) !=
1125 extern int ftrace_is_dead(void);
1126 int ftrace_create_function_files(struct trace_array *tr,
1127 struct dentry *parent);
1128 void ftrace_destroy_function_files(struct trace_array *tr);
1129 int ftrace_allocate_ftrace_ops(struct trace_array *tr);
1130 void ftrace_free_ftrace_ops(struct trace_array *tr);
1131 void ftrace_init_global_array_ops(struct trace_array *tr);
1132 void ftrace_init_array_ops(struct trace_array *tr, ftrace_func_t func);
1133 void ftrace_reset_array_ops(struct trace_array *tr);
1134 void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d_tracer);
1135 void ftrace_init_tracefs_toplevel(struct trace_array *tr,
1136 struct dentry *d_tracer);
1137 void ftrace_clear_pids(struct trace_array *tr);
1138 int init_function_trace(void);
1139 void ftrace_pid_follow_fork(struct trace_array *tr, bool enable);
1141 static inline int ftrace_trace_task(struct trace_array *tr)
1145 static inline int ftrace_is_dead(void) { return 0; }
1147 ftrace_create_function_files(struct trace_array *tr,
1148 struct dentry *parent)
1152 static inline int ftrace_allocate_ftrace_ops(struct trace_array *tr)
1156 static inline void ftrace_free_ftrace_ops(struct trace_array *tr) { }
1157 static inline void ftrace_destroy_function_files(struct trace_array *tr) { }
1158 static inline __init void
1159 ftrace_init_global_array_ops(struct trace_array *tr) { }
1160 static inline void ftrace_reset_array_ops(struct trace_array *tr) { }
1161 static inline void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d) { }
1162 static inline void ftrace_init_tracefs_toplevel(struct trace_array *tr, struct dentry *d) { }
1163 static inline void ftrace_clear_pids(struct trace_array *tr) { }
1164 static inline int init_function_trace(void) { return 0; }
1165 static inline void ftrace_pid_follow_fork(struct trace_array *tr, bool enable) { }
1166 /* ftace_func_t type is not defined, use macro instead of static inline */
1167 #define ftrace_init_array_ops(tr, func) do { } while (0)
1168 #endif /* CONFIG_FUNCTION_TRACER */
1170 #if defined(CONFIG_FUNCTION_TRACER) && defined(CONFIG_DYNAMIC_FTRACE)
1172 struct ftrace_probe_ops {
1173 void (*func)(unsigned long ip,
1174 unsigned long parent_ip,
1175 struct trace_array *tr,
1176 struct ftrace_probe_ops *ops,
1178 int (*init)(struct ftrace_probe_ops *ops,
1179 struct trace_array *tr,
1180 unsigned long ip, void *init_data,
1182 void (*free)(struct ftrace_probe_ops *ops,
1183 struct trace_array *tr,
1184 unsigned long ip, void *data);
1185 int (*print)(struct seq_file *m,
1187 struct ftrace_probe_ops *ops,
1191 struct ftrace_func_mapper;
1192 typedef int (*ftrace_mapper_func)(void *data);
1194 struct ftrace_func_mapper *allocate_ftrace_func_mapper(void);
1195 void **ftrace_func_mapper_find_ip(struct ftrace_func_mapper *mapper,
1197 int ftrace_func_mapper_add_ip(struct ftrace_func_mapper *mapper,
1198 unsigned long ip, void *data);
1199 void *ftrace_func_mapper_remove_ip(struct ftrace_func_mapper *mapper,
1201 void free_ftrace_func_mapper(struct ftrace_func_mapper *mapper,
1202 ftrace_mapper_func free_func);
1205 register_ftrace_function_probe(char *glob, struct trace_array *tr,
1206 struct ftrace_probe_ops *ops, void *data);
1208 unregister_ftrace_function_probe_func(char *glob, struct trace_array *tr,
1209 struct ftrace_probe_ops *ops);
1210 extern void clear_ftrace_function_probes(struct trace_array *tr);
1212 int register_ftrace_command(struct ftrace_func_command *cmd);
1213 int unregister_ftrace_command(struct ftrace_func_command *cmd);
1215 void ftrace_create_filter_files(struct ftrace_ops *ops,
1216 struct dentry *parent);
1217 void ftrace_destroy_filter_files(struct ftrace_ops *ops);
1219 extern int ftrace_set_filter(struct ftrace_ops *ops, unsigned char *buf,
1220 int len, int reset);
1221 extern int ftrace_set_notrace(struct ftrace_ops *ops, unsigned char *buf,
1222 int len, int reset);
1224 struct ftrace_func_command;
1226 static inline __init int register_ftrace_command(struct ftrace_func_command *cmd)
1230 static inline __init int unregister_ftrace_command(char *cmd_name)
1234 static inline void clear_ftrace_function_probes(struct trace_array *tr)
1239 * The ops parameter passed in is usually undefined.
1240 * This must be a macro.
1242 #define ftrace_create_filter_files(ops, parent) do { } while (0)
1243 #define ftrace_destroy_filter_files(ops) do { } while (0)
1244 #endif /* CONFIG_FUNCTION_TRACER && CONFIG_DYNAMIC_FTRACE */
1246 bool ftrace_event_is_function(struct trace_event_call *call);
1249 * struct trace_parser - servers for reading the user input separated by spaces
1250 * @cont: set if the input is not complete - no final space char was found
1251 * @buffer: holds the parsed user input
1252 * @idx: user input length
1253 * @size: buffer size
1255 struct trace_parser {
1262 static inline bool trace_parser_loaded(struct trace_parser *parser)
1264 return (parser->idx != 0);
1267 static inline bool trace_parser_cont(struct trace_parser *parser)
1269 return parser->cont;
1272 static inline void trace_parser_clear(struct trace_parser *parser)
1274 parser->cont = false;
1278 extern int trace_parser_get_init(struct trace_parser *parser, int size);
1279 extern void trace_parser_put(struct trace_parser *parser);
1280 extern int trace_get_user(struct trace_parser *parser, const char __user *ubuf,
1281 size_t cnt, loff_t *ppos);
1284 * Only create function graph options if function graph is configured.
1286 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1287 # define FGRAPH_FLAGS \
1288 C(DISPLAY_GRAPH, "display-graph"),
1290 # define FGRAPH_FLAGS
1293 #ifdef CONFIG_BRANCH_TRACER
1294 # define BRANCH_FLAGS \
1295 C(BRANCH, "branch"),
1297 # define BRANCH_FLAGS
1300 #ifdef CONFIG_FUNCTION_TRACER
1301 # define FUNCTION_FLAGS \
1302 C(FUNCTION, "function-trace"), \
1303 C(FUNC_FORK, "function-fork"),
1304 # define FUNCTION_DEFAULT_FLAGS TRACE_ITER_FUNCTION
1306 # define FUNCTION_FLAGS
1307 # define FUNCTION_DEFAULT_FLAGS 0UL
1308 # define TRACE_ITER_FUNC_FORK 0UL
1311 #ifdef CONFIG_STACKTRACE
1312 # define STACK_FLAGS \
1313 C(STACKTRACE, "stacktrace"),
1315 # define STACK_FLAGS
1319 * trace_iterator_flags is an enumeration that defines bit
1320 * positions into trace_flags that controls the output.
1322 * NOTE: These bits must match the trace_options array in
1323 * trace.c (this macro guarantees it).
1325 #define TRACE_FLAGS \
1326 C(PRINT_PARENT, "print-parent"), \
1327 C(SYM_OFFSET, "sym-offset"), \
1328 C(SYM_ADDR, "sym-addr"), \
1329 C(VERBOSE, "verbose"), \
1333 C(BLOCK, "block"), \
1334 C(PRINTK, "trace_printk"), \
1335 C(ANNOTATE, "annotate"), \
1336 C(USERSTACKTRACE, "userstacktrace"), \
1337 C(SYM_USEROBJ, "sym-userobj"), \
1338 C(PRINTK_MSGONLY, "printk-msg-only"), \
1339 C(CONTEXT_INFO, "context-info"), /* Print pid/cpu/time */ \
1340 C(LATENCY_FMT, "latency-format"), \
1341 C(RECORD_CMD, "record-cmd"), \
1342 C(RECORD_TGID, "record-tgid"), \
1343 C(OVERWRITE, "overwrite"), \
1344 C(STOP_ON_FREE, "disable_on_free"), \
1345 C(IRQ_INFO, "irq-info"), \
1346 C(MARKERS, "markers"), \
1347 C(EVENT_FORK, "event-fork"), \
1348 C(PAUSE_ON_TRACE, "pause-on-trace"), \
1355 * By defining C, we can make TRACE_FLAGS a list of bit names
1356 * that will define the bits for the flag masks.
1359 #define C(a, b) TRACE_ITER_##a##_BIT
1361 enum trace_iterator_bits {
1363 /* Make sure we don't go more than we have bits for */
1368 * By redefining C, we can make TRACE_FLAGS a list of masks that
1369 * use the bits as defined above.
1372 #define C(a, b) TRACE_ITER_##a = (1 << TRACE_ITER_##a##_BIT)
1374 enum trace_iterator_flags { TRACE_FLAGS };
1377 * TRACE_ITER_SYM_MASK masks the options in trace_flags that
1378 * control the output of kernel symbols.
1380 #define TRACE_ITER_SYM_MASK \
1381 (TRACE_ITER_PRINT_PARENT|TRACE_ITER_SYM_OFFSET|TRACE_ITER_SYM_ADDR)
1383 extern struct tracer nop_trace;
1385 #ifdef CONFIG_BRANCH_TRACER
1386 extern int enable_branch_tracing(struct trace_array *tr);
1387 extern void disable_branch_tracing(void);
1388 static inline int trace_branch_enable(struct trace_array *tr)
1390 if (tr->trace_flags & TRACE_ITER_BRANCH)
1391 return enable_branch_tracing(tr);
1394 static inline void trace_branch_disable(void)
1396 /* due to races, always disable */
1397 disable_branch_tracing();
1400 static inline int trace_branch_enable(struct trace_array *tr)
1404 static inline void trace_branch_disable(void)
1407 #endif /* CONFIG_BRANCH_TRACER */
1409 /* set ring buffers to default size if not already done so */
1410 int tracing_update_buffers(void);
1412 struct ftrace_event_field {
1413 struct list_head link;
1424 struct event_filter {
1425 struct prog_entry __rcu *prog;
1426 char *filter_string;
1429 struct event_subsystem {
1430 struct list_head list;
1432 struct event_filter *filter;
1436 struct trace_subsystem_dir {
1437 struct list_head list;
1438 struct event_subsystem *subsystem;
1439 struct trace_array *tr;
1440 struct dentry *entry;
1445 extern int call_filter_check_discard(struct trace_event_call *call, void *rec,
1446 struct trace_buffer *buffer,
1447 struct ring_buffer_event *event);
1449 void trace_buffer_unlock_commit_regs(struct trace_array *tr,
1450 struct trace_buffer *buffer,
1451 struct ring_buffer_event *event,
1452 unsigned long flags, int pc,
1453 struct pt_regs *regs);
1455 static inline void trace_buffer_unlock_commit(struct trace_array *tr,
1456 struct trace_buffer *buffer,
1457 struct ring_buffer_event *event,
1458 unsigned long flags, int pc)
1460 trace_buffer_unlock_commit_regs(tr, buffer, event, flags, pc, NULL);
1463 DECLARE_PER_CPU(struct ring_buffer_event *, trace_buffered_event);
1464 DECLARE_PER_CPU(int, trace_buffered_event_cnt);
1465 void trace_buffered_event_disable(void);
1466 void trace_buffered_event_enable(void);
1469 __trace_event_discard_commit(struct trace_buffer *buffer,
1470 struct ring_buffer_event *event)
1472 if (this_cpu_read(trace_buffered_event) == event) {
1473 /* Simply release the temp buffer */
1474 this_cpu_dec(trace_buffered_event_cnt);
1477 ring_buffer_discard_commit(buffer, event);
1481 * Helper function for event_trigger_unlock_commit{_regs}().
1482 * If there are event triggers attached to this event that requires
1483 * filtering against its fields, then they will be called as the
1484 * entry already holds the field information of the current event.
1486 * It also checks if the event should be discarded or not.
1487 * It is to be discarded if the event is soft disabled and the
1488 * event was only recorded to process triggers, or if the event
1489 * filter is active and this event did not match the filters.
1491 * Returns true if the event is discarded, false otherwise.
1494 __event_trigger_test_discard(struct trace_event_file *file,
1495 struct trace_buffer *buffer,
1496 struct ring_buffer_event *event,
1498 enum event_trigger_type *tt)
1500 unsigned long eflags = file->flags;
1502 if (eflags & EVENT_FILE_FL_TRIGGER_COND)
1503 *tt = event_triggers_call(file, entry, event);
1505 if (test_bit(EVENT_FILE_FL_SOFT_DISABLED_BIT, &file->flags) ||
1506 (unlikely(file->flags & EVENT_FILE_FL_FILTERED) &&
1507 !filter_match_preds(file->filter, entry))) {
1508 __trace_event_discard_commit(buffer, event);
1516 * event_trigger_unlock_commit - handle triggers and finish event commit
1517 * @file: The file pointer assoctiated to the event
1518 * @buffer: The ring buffer that the event is being written to
1519 * @event: The event meta data in the ring buffer
1520 * @entry: The event itself
1521 * @irq_flags: The state of the interrupts at the start of the event
1522 * @pc: The state of the preempt count at the start of the event.
1524 * This is a helper function to handle triggers that require data
1525 * from the event itself. It also tests the event against filters and
1526 * if the event is soft disabled and should be discarded.
1529 event_trigger_unlock_commit(struct trace_event_file *file,
1530 struct trace_buffer *buffer,
1531 struct ring_buffer_event *event,
1532 void *entry, unsigned long irq_flags, int pc)
1534 enum event_trigger_type tt = ETT_NONE;
1536 if (!__event_trigger_test_discard(file, buffer, event, entry, &tt))
1537 trace_buffer_unlock_commit(file->tr, buffer, event, irq_flags, pc);
1540 event_triggers_post_call(file, tt);
1544 * event_trigger_unlock_commit_regs - handle triggers and finish event commit
1545 * @file: The file pointer assoctiated to the event
1546 * @buffer: The ring buffer that the event is being written to
1547 * @event: The event meta data in the ring buffer
1548 * @entry: The event itself
1549 * @irq_flags: The state of the interrupts at the start of the event
1550 * @pc: The state of the preempt count at the start of the event.
1552 * This is a helper function to handle triggers that require data
1553 * from the event itself. It also tests the event against filters and
1554 * if the event is soft disabled and should be discarded.
1556 * Same as event_trigger_unlock_commit() but calls
1557 * trace_buffer_unlock_commit_regs() instead of trace_buffer_unlock_commit().
1560 event_trigger_unlock_commit_regs(struct trace_event_file *file,
1561 struct trace_buffer *buffer,
1562 struct ring_buffer_event *event,
1563 void *entry, unsigned long irq_flags, int pc,
1564 struct pt_regs *regs)
1566 enum event_trigger_type tt = ETT_NONE;
1568 if (!__event_trigger_test_discard(file, buffer, event, entry, &tt))
1569 trace_buffer_unlock_commit_regs(file->tr, buffer, event,
1570 irq_flags, pc, regs);
1573 event_triggers_post_call(file, tt);
1576 #define FILTER_PRED_INVALID ((unsigned short)-1)
1577 #define FILTER_PRED_IS_RIGHT (1 << 15)
1578 #define FILTER_PRED_FOLD (1 << 15)
1581 * The max preds is the size of unsigned short with
1582 * two flags at the MSBs. One bit is used for both the IS_RIGHT
1583 * and FOLD flags. The other is reserved.
1585 * 2^14 preds is way more than enough.
1587 #define MAX_FILTER_PRED 16384
1592 typedef int (*filter_pred_fn_t) (struct filter_pred *pred, void *event);
1594 typedef int (*regex_match_func)(char *str, struct regex *r, int len);
1606 char pattern[MAX_FILTER_STR_VAL];
1609 regex_match_func match;
1612 struct filter_pred {
1613 filter_pred_fn_t fn;
1616 unsigned short *ops;
1617 struct ftrace_event_field *field;
1623 static inline bool is_string_field(struct ftrace_event_field *field)
1625 return field->filter_type == FILTER_DYN_STRING ||
1626 field->filter_type == FILTER_STATIC_STRING ||
1627 field->filter_type == FILTER_PTR_STRING ||
1628 field->filter_type == FILTER_COMM;
1631 static inline bool is_function_field(struct ftrace_event_field *field)
1633 return field->filter_type == FILTER_TRACE_FN;
1636 extern enum regex_type
1637 filter_parse_regex(char *buff, int len, char **search, int *not);
1638 extern void print_event_filter(struct trace_event_file *file,
1639 struct trace_seq *s);
1640 extern int apply_event_filter(struct trace_event_file *file,
1641 char *filter_string);
1642 extern int apply_subsystem_event_filter(struct trace_subsystem_dir *dir,
1643 char *filter_string);
1644 extern void print_subsystem_event_filter(struct event_subsystem *system,
1645 struct trace_seq *s);
1646 extern int filter_assign_type(const char *type);
1647 extern int create_event_filter(struct trace_array *tr,
1648 struct trace_event_call *call,
1649 char *filter_str, bool set_str,
1650 struct event_filter **filterp);
1651 extern void free_event_filter(struct event_filter *filter);
1653 struct ftrace_event_field *
1654 trace_find_event_field(struct trace_event_call *call, char *name);
1656 extern void trace_event_enable_cmd_record(bool enable);
1657 extern void trace_event_enable_tgid_record(bool enable);
1659 extern int event_trace_init(void);
1660 extern int event_trace_add_tracer(struct dentry *parent, struct trace_array *tr);
1661 extern int event_trace_del_tracer(struct trace_array *tr);
1662 extern void __trace_early_add_events(struct trace_array *tr);
1664 extern struct trace_event_file *__find_event_file(struct trace_array *tr,
1667 extern struct trace_event_file *find_event_file(struct trace_array *tr,
1671 static inline void *event_file_data(struct file *filp)
1673 return READ_ONCE(file_inode(filp)->i_private);
1676 extern struct mutex event_mutex;
1677 extern struct list_head ftrace_events;
1679 extern const struct file_operations event_trigger_fops;
1680 extern const struct file_operations event_hist_fops;
1681 extern const struct file_operations event_hist_debug_fops;
1682 extern const struct file_operations event_inject_fops;
1684 #ifdef CONFIG_HIST_TRIGGERS
1685 extern int register_trigger_hist_cmd(void);
1686 extern int register_trigger_hist_enable_disable_cmds(void);
1688 static inline int register_trigger_hist_cmd(void) { return 0; }
1689 static inline int register_trigger_hist_enable_disable_cmds(void) { return 0; }
1692 extern int register_trigger_cmds(void);
1693 extern void clear_event_triggers(struct trace_array *tr);
1695 struct event_trigger_data {
1696 unsigned long count;
1698 struct event_trigger_ops *ops;
1699 struct event_command *cmd_ops;
1700 struct event_filter __rcu *filter;
1705 struct list_head list;
1707 struct list_head named_list;
1708 struct event_trigger_data *named_data;
1712 #define ENABLE_EVENT_STR "enable_event"
1713 #define DISABLE_EVENT_STR "disable_event"
1714 #define ENABLE_HIST_STR "enable_hist"
1715 #define DISABLE_HIST_STR "disable_hist"
1717 struct enable_trigger_data {
1718 struct trace_event_file *file;
1723 extern int event_enable_trigger_print(struct seq_file *m,
1724 struct event_trigger_ops *ops,
1725 struct event_trigger_data *data);
1726 extern void event_enable_trigger_free(struct event_trigger_ops *ops,
1727 struct event_trigger_data *data);
1728 extern int event_enable_trigger_func(struct event_command *cmd_ops,
1729 struct trace_event_file *file,
1730 char *glob, char *cmd, char *param);
1731 extern int event_enable_register_trigger(char *glob,
1732 struct event_trigger_ops *ops,
1733 struct event_trigger_data *data,
1734 struct trace_event_file *file);
1735 extern void event_enable_unregister_trigger(char *glob,
1736 struct event_trigger_ops *ops,
1737 struct event_trigger_data *test,
1738 struct trace_event_file *file);
1739 extern void trigger_data_free(struct event_trigger_data *data);
1740 extern int event_trigger_init(struct event_trigger_ops *ops,
1741 struct event_trigger_data *data);
1742 extern int trace_event_trigger_enable_disable(struct trace_event_file *file,
1743 int trigger_enable);
1744 extern void update_cond_flag(struct trace_event_file *file);
1745 extern int set_trigger_filter(char *filter_str,
1746 struct event_trigger_data *trigger_data,
1747 struct trace_event_file *file);
1748 extern struct event_trigger_data *find_named_trigger(const char *name);
1749 extern bool is_named_trigger(struct event_trigger_data *test);
1750 extern int save_named_trigger(const char *name,
1751 struct event_trigger_data *data);
1752 extern void del_named_trigger(struct event_trigger_data *data);
1753 extern void pause_named_trigger(struct event_trigger_data *data);
1754 extern void unpause_named_trigger(struct event_trigger_data *data);
1755 extern void set_named_trigger_data(struct event_trigger_data *data,
1756 struct event_trigger_data *named_data);
1757 extern struct event_trigger_data *
1758 get_named_trigger_data(struct event_trigger_data *data);
1759 extern int register_event_command(struct event_command *cmd);
1760 extern int unregister_event_command(struct event_command *cmd);
1761 extern int register_trigger_hist_enable_disable_cmds(void);
1764 * struct event_trigger_ops - callbacks for trace event triggers
1766 * The methods in this structure provide per-event trigger hooks for
1767 * various trigger operations.
1769 * All the methods below, except for @init() and @free(), must be
1772 * @func: The trigger 'probe' function called when the triggering
1773 * event occurs. The data passed into this callback is the data
1774 * that was supplied to the event_command @reg() function that
1775 * registered the trigger (see struct event_command) along with
1776 * the trace record, rec.
1778 * @init: An optional initialization function called for the trigger
1779 * when the trigger is registered (via the event_command reg()
1780 * function). This can be used to perform per-trigger
1781 * initialization such as incrementing a per-trigger reference
1782 * count, for instance. This is usually implemented by the
1783 * generic utility function @event_trigger_init() (see
1784 * trace_event_triggers.c).
1786 * @free: An optional de-initialization function called for the
1787 * trigger when the trigger is unregistered (via the
1788 * event_command @reg() function). This can be used to perform
1789 * per-trigger de-initialization such as decrementing a
1790 * per-trigger reference count and freeing corresponding trigger
1791 * data, for instance. This is usually implemented by the
1792 * generic utility function @event_trigger_free() (see
1793 * trace_event_triggers.c).
1795 * @print: The callback function invoked to have the trigger print
1796 * itself. This is usually implemented by a wrapper function
1797 * that calls the generic utility function @event_trigger_print()
1798 * (see trace_event_triggers.c).
1800 struct event_trigger_ops {
1801 void (*func)(struct event_trigger_data *data,
1803 struct ring_buffer_event *rbe);
1804 int (*init)(struct event_trigger_ops *ops,
1805 struct event_trigger_data *data);
1806 void (*free)(struct event_trigger_ops *ops,
1807 struct event_trigger_data *data);
1808 int (*print)(struct seq_file *m,
1809 struct event_trigger_ops *ops,
1810 struct event_trigger_data *data);
1814 * struct event_command - callbacks and data members for event commands
1816 * Event commands are invoked by users by writing the command name
1817 * into the 'trigger' file associated with a trace event. The
1818 * parameters associated with a specific invocation of an event
1819 * command are used to create an event trigger instance, which is
1820 * added to the list of trigger instances associated with that trace
1821 * event. When the event is hit, the set of triggers associated with
1822 * that event is invoked.
1824 * The data members in this structure provide per-event command data
1825 * for various event commands.
1827 * All the data members below, except for @post_trigger, must be set
1828 * for each event command.
1830 * @name: The unique name that identifies the event command. This is
1831 * the name used when setting triggers via trigger files.
1833 * @trigger_type: A unique id that identifies the event command
1834 * 'type'. This value has two purposes, the first to ensure that
1835 * only one trigger of the same type can be set at a given time
1836 * for a particular event e.g. it doesn't make sense to have both
1837 * a traceon and traceoff trigger attached to a single event at
1838 * the same time, so traceon and traceoff have the same type
1839 * though they have different names. The @trigger_type value is
1840 * also used as a bit value for deferring the actual trigger
1841 * action until after the current event is finished. Some
1842 * commands need to do this if they themselves log to the trace
1843 * buffer (see the @post_trigger() member below). @trigger_type
1844 * values are defined by adding new values to the trigger_type
1845 * enum in include/linux/trace_events.h.
1847 * @flags: See the enum event_command_flags below.
1849 * All the methods below, except for @set_filter() and @unreg_all(),
1850 * must be implemented.
1852 * @func: The callback function responsible for parsing and
1853 * registering the trigger written to the 'trigger' file by the
1854 * user. It allocates the trigger instance and registers it with
1855 * the appropriate trace event. It makes use of the other
1856 * event_command callback functions to orchestrate this, and is
1857 * usually implemented by the generic utility function
1858 * @event_trigger_callback() (see trace_event_triggers.c).
1860 * @reg: Adds the trigger to the list of triggers associated with the
1861 * event, and enables the event trigger itself, after
1862 * initializing it (via the event_trigger_ops @init() function).
1863 * This is also where commands can use the @trigger_type value to
1864 * make the decision as to whether or not multiple instances of
1865 * the trigger should be allowed. This is usually implemented by
1866 * the generic utility function @register_trigger() (see
1867 * trace_event_triggers.c).
1869 * @unreg: Removes the trigger from the list of triggers associated
1870 * with the event, and disables the event trigger itself, after
1871 * initializing it (via the event_trigger_ops @free() function).
1872 * This is usually implemented by the generic utility function
1873 * @unregister_trigger() (see trace_event_triggers.c).
1875 * @unreg_all: An optional function called to remove all the triggers
1876 * from the list of triggers associated with the event. Called
1877 * when a trigger file is opened in truncate mode.
1879 * @set_filter: An optional function called to parse and set a filter
1880 * for the trigger. If no @set_filter() method is set for the
1881 * event command, filters set by the user for the command will be
1882 * ignored. This is usually implemented by the generic utility
1883 * function @set_trigger_filter() (see trace_event_triggers.c).
1885 * @get_trigger_ops: The callback function invoked to retrieve the
1886 * event_trigger_ops implementation associated with the command.
1888 struct event_command {
1889 struct list_head list;
1891 enum event_trigger_type trigger_type;
1893 int (*func)(struct event_command *cmd_ops,
1894 struct trace_event_file *file,
1895 char *glob, char *cmd, char *params);
1896 int (*reg)(char *glob,
1897 struct event_trigger_ops *ops,
1898 struct event_trigger_data *data,
1899 struct trace_event_file *file);
1900 void (*unreg)(char *glob,
1901 struct event_trigger_ops *ops,
1902 struct event_trigger_data *data,
1903 struct trace_event_file *file);
1904 void (*unreg_all)(struct trace_event_file *file);
1905 int (*set_filter)(char *filter_str,
1906 struct event_trigger_data *data,
1907 struct trace_event_file *file);
1908 struct event_trigger_ops *(*get_trigger_ops)(char *cmd, char *param);
1912 * enum event_command_flags - flags for struct event_command
1914 * @POST_TRIGGER: A flag that says whether or not this command needs
1915 * to have its action delayed until after the current event has
1916 * been closed. Some triggers need to avoid being invoked while
1917 * an event is currently in the process of being logged, since
1918 * the trigger may itself log data into the trace buffer. Thus
1919 * we make sure the current event is committed before invoking
1920 * those triggers. To do that, the trigger invocation is split
1921 * in two - the first part checks the filter using the current
1922 * trace record; if a command has the @post_trigger flag set, it
1923 * sets a bit for itself in the return value, otherwise it
1924 * directly invokes the trigger. Once all commands have been
1925 * either invoked or set their return flag, the current record is
1926 * either committed or discarded. At that point, if any commands
1927 * have deferred their triggers, those commands are finally
1928 * invoked following the close of the current event. In other
1929 * words, if the event_trigger_ops @func() probe implementation
1930 * itself logs to the trace buffer, this flag should be set,
1931 * otherwise it can be left unspecified.
1933 * @NEEDS_REC: A flag that says whether or not this command needs
1934 * access to the trace record in order to perform its function,
1935 * regardless of whether or not it has a filter associated with
1936 * it (filters make a trigger require access to the trace record
1937 * but are not always present).
1939 enum event_command_flags {
1940 EVENT_CMD_FL_POST_TRIGGER = 1,
1941 EVENT_CMD_FL_NEEDS_REC = 2,
1944 static inline bool event_command_post_trigger(struct event_command *cmd_ops)
1946 return cmd_ops->flags & EVENT_CMD_FL_POST_TRIGGER;
1949 static inline bool event_command_needs_rec(struct event_command *cmd_ops)
1951 return cmd_ops->flags & EVENT_CMD_FL_NEEDS_REC;
1954 extern int trace_event_enable_disable(struct trace_event_file *file,
1955 int enable, int soft_disable);
1956 extern int tracing_alloc_snapshot(void);
1957 extern void tracing_snapshot_cond(struct trace_array *tr, void *cond_data);
1958 extern int tracing_snapshot_cond_enable(struct trace_array *tr, void *cond_data, cond_update_fn_t update);
1960 extern int tracing_snapshot_cond_disable(struct trace_array *tr);
1961 extern void *tracing_cond_snapshot_data(struct trace_array *tr);
1963 extern const char *__start___trace_bprintk_fmt[];
1964 extern const char *__stop___trace_bprintk_fmt[];
1966 extern const char *__start___tracepoint_str[];
1967 extern const char *__stop___tracepoint_str[];
1969 void trace_printk_control(bool enabled);
1970 void trace_printk_start_comm(void);
1971 int trace_keep_overwrite(struct tracer *tracer, u32 mask, int set);
1972 int set_tracer_flag(struct trace_array *tr, unsigned int mask, int enabled);
1974 /* Used from boot time tracer */
1975 extern int trace_set_options(struct trace_array *tr, char *option);
1976 extern int tracing_set_tracer(struct trace_array *tr, const char *buf);
1977 extern ssize_t tracing_resize_ring_buffer(struct trace_array *tr,
1978 unsigned long size, int cpu_id);
1979 extern int tracing_set_cpumask(struct trace_array *tr,
1980 cpumask_var_t tracing_cpumask_new);
1983 #define MAX_EVENT_NAME_LEN 64
1985 extern int trace_run_command(const char *buf, int (*createfn)(int, char**));
1986 extern ssize_t trace_parse_run_command(struct file *file,
1987 const char __user *buffer, size_t count, loff_t *ppos,
1988 int (*createfn)(int, char**));
1990 extern unsigned int err_pos(char *cmd, const char *str);
1991 extern void tracing_log_err(struct trace_array *tr,
1992 const char *loc, const char *cmd,
1993 const char **errs, u8 type, u8 pos);
1996 * Normal trace_printk() and friends allocates special buffers
1997 * to do the manipulation, as well as saves the print formats
1998 * into sections to display. But the trace infrastructure wants
1999 * to use these without the added overhead at the price of being
2000 * a bit slower (used mainly for warnings, where we don't care
2001 * about performance). The internal_trace_puts() is for such
2004 #define internal_trace_puts(str) __trace_puts(_THIS_IP_, str, strlen(str))
2007 #define FTRACE_ENTRY(call, struct_name, id, tstruct, print) \
2008 extern struct trace_event_call \
2009 __aligned(4) event_##call;
2010 #undef FTRACE_ENTRY_DUP
2011 #define FTRACE_ENTRY_DUP(call, struct_name, id, tstruct, print) \
2012 FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print))
2013 #undef FTRACE_ENTRY_PACKED
2014 #define FTRACE_ENTRY_PACKED(call, struct_name, id, tstruct, print) \
2015 FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print))
2017 #include "trace_entries.h"
2019 #if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_FUNCTION_TRACER)
2020 int perf_ftrace_event_register(struct trace_event_call *call,
2021 enum trace_reg type, void *data);
2023 #define perf_ftrace_event_register NULL
2026 #ifdef CONFIG_FTRACE_SYSCALLS
2027 void init_ftrace_syscalls(void);
2028 const char *get_syscall_name(int syscall);
2030 static inline void init_ftrace_syscalls(void) { }
2031 static inline const char *get_syscall_name(int syscall)
2037 #ifdef CONFIG_EVENT_TRACING
2038 void trace_event_init(void);
2039 void trace_event_eval_update(struct trace_eval_map **map, int len);
2040 /* Used from boot time tracer */
2041 extern int ftrace_set_clr_event(struct trace_array *tr, char *buf, int set);
2042 extern int trigger_process_regex(struct trace_event_file *file, char *buff);
2044 static inline void __init trace_event_init(void) { }
2045 static inline void trace_event_eval_update(struct trace_eval_map **map, int len) { }
2048 #ifdef CONFIG_TRACER_SNAPSHOT
2049 void tracing_snapshot_instance(struct trace_array *tr);
2050 int tracing_alloc_snapshot_instance(struct trace_array *tr);
2052 static inline void tracing_snapshot_instance(struct trace_array *tr) { }
2053 static inline int tracing_alloc_snapshot_instance(struct trace_array *tr)
2059 #ifdef CONFIG_PREEMPT_TRACER
2060 void tracer_preempt_on(unsigned long a0, unsigned long a1);
2061 void tracer_preempt_off(unsigned long a0, unsigned long a1);
2063 static inline void tracer_preempt_on(unsigned long a0, unsigned long a1) { }
2064 static inline void tracer_preempt_off(unsigned long a0, unsigned long a1) { }
2066 #ifdef CONFIG_IRQSOFF_TRACER
2067 void tracer_hardirqs_on(unsigned long a0, unsigned long a1);
2068 void tracer_hardirqs_off(unsigned long a0, unsigned long a1);
2070 static inline void tracer_hardirqs_on(unsigned long a0, unsigned long a1) { }
2071 static inline void tracer_hardirqs_off(unsigned long a0, unsigned long a1) { }
2074 extern struct trace_iterator *tracepoint_print_iter;
2077 * Reset the state of the trace_iterator so that it can read consumed data.
2078 * Normally, the trace_iterator is used for reading the data when it is not
2079 * consumed, and must retain state.
2081 static __always_inline void trace_iterator_reset(struct trace_iterator *iter)
2083 const size_t offset = offsetof(struct trace_iterator, seq);
2086 * Keep gcc from complaining about overwriting more than just one
2087 * member in the structure.
2089 memset((char *)iter + offset, 0, sizeof(struct trace_iterator) - offset);
2094 /* Check the name is good for event/group/fields */
2095 static inline bool is_good_name(const char *name)
2097 if (!isalpha(*name) && *name != '_')
2099 while (*++name != '\0') {
2100 if (!isalpha(*name) && !isdigit(*name) && *name != '_')
2106 #endif /* _LINUX_KERNEL_TRACE_H */