1 #ifndef _LINUX_KERNEL_TRACE_H
2 #define _LINUX_KERNEL_TRACE_H
5 #include <asm/atomic.h>
6 #include <linux/sched.h>
7 #include <linux/clocksource.h>
8 #include <linux/ring_buffer.h>
9 #include <linux/mmiotrace.h>
10 #include <linux/ftrace.h>
11 #include <trace/boot.h>
12 #include <trace/kmemtrace.h>
13 #include <trace/power.h>
16 __TRACE_FIRST_TYPE = 0,
45 * The trace entry - the most basic unit of tracing. This is what
46 * is printed in the end as a single line in the trace output, such as:
48 * bash-15816 [01] 235.197585: idle_cpu <- irq_enter
53 unsigned char preempt_count;
59 * Function trace entry - function address and parent function addres:
62 struct trace_entry ent;
64 unsigned long parent_ip;
67 /* Function call entry */
68 struct ftrace_graph_ent_entry {
69 struct trace_entry ent;
70 struct ftrace_graph_ent graph_ent;
73 /* Function return entry */
74 struct ftrace_graph_ret_entry {
75 struct trace_entry ent;
76 struct ftrace_graph_ret ret;
78 extern struct tracer boot_tracer;
81 * Context switch trace entry - which task (and prio) we switched from/to:
83 struct ctx_switch_entry {
84 struct trace_entry ent;
85 unsigned int prev_pid;
86 unsigned char prev_prio;
87 unsigned char prev_state;
88 unsigned int next_pid;
89 unsigned char next_prio;
90 unsigned char next_state;
91 unsigned int next_cpu;
95 * Special (free-form) trace entry:
97 struct special_entry {
98 struct trace_entry ent;
108 #define FTRACE_STACK_ENTRIES 8
111 struct trace_entry ent;
112 unsigned long caller[FTRACE_STACK_ENTRIES];
115 struct userstack_entry {
116 struct trace_entry ent;
117 unsigned long caller[FTRACE_STACK_ENTRIES];
121 * trace_printk entry:
123 struct bprint_entry {
124 struct trace_entry ent;
131 struct trace_entry ent;
136 #define TRACE_OLD_SIZE 88
138 struct trace_field_cont {
140 /* Temporary till we get rid of this completely */
141 char buf[TRACE_OLD_SIZE - 1];
144 struct trace_mmiotrace_rw {
145 struct trace_entry ent;
146 struct mmiotrace_rw rw;
149 struct trace_mmiotrace_map {
150 struct trace_entry ent;
151 struct mmiotrace_map map;
154 struct trace_boot_call {
155 struct trace_entry ent;
156 struct boot_trace_call boot_call;
159 struct trace_boot_ret {
160 struct trace_entry ent;
161 struct boot_trace_ret boot_ret;
164 #define TRACE_FUNC_SIZE 30
165 #define TRACE_FILE_SIZE 20
166 struct trace_branch {
167 struct trace_entry ent;
169 char func[TRACE_FUNC_SIZE+1];
170 char file[TRACE_FILE_SIZE+1];
174 struct hw_branch_entry {
175 struct trace_entry ent;
181 struct trace_entry ent;
182 struct power_trace state_data;
185 struct kmemtrace_alloc_entry {
186 struct trace_entry ent;
187 enum kmemtrace_type_id type_id;
188 unsigned long call_site;
196 struct kmemtrace_free_entry {
197 struct trace_entry ent;
198 enum kmemtrace_type_id type_id;
199 unsigned long call_site;
203 struct syscall_trace_enter {
204 struct trace_entry ent;
206 unsigned long args[];
209 struct syscall_trace_exit {
210 struct trace_entry ent;
217 * trace_flag_type is an enumeration that holds different
218 * states when a trace occurs. These are:
219 * IRQS_OFF - interrupts were disabled
220 * IRQS_NOSUPPORT - arch does not support irqs_disabled_flags
221 * NEED_RESCED - reschedule is requested
222 * HARDIRQ - inside an interrupt handler
223 * SOFTIRQ - inside a softirq handler
225 enum trace_flag_type {
226 TRACE_FLAG_IRQS_OFF = 0x01,
227 TRACE_FLAG_IRQS_NOSUPPORT = 0x02,
228 TRACE_FLAG_NEED_RESCHED = 0x04,
229 TRACE_FLAG_HARDIRQ = 0x08,
230 TRACE_FLAG_SOFTIRQ = 0x10,
233 #define TRACE_BUF_SIZE 1024
236 * The CPU trace array - it consists of thousands of trace entries
237 * plus some other descriptor data: (for example which task started
240 struct trace_array_cpu {
242 void *buffer_page; /* ring buffer spare */
244 /* these fields get copied into max-trace: */
245 unsigned long trace_idx;
246 unsigned long overrun;
247 unsigned long saved_latency;
248 unsigned long critical_start;
249 unsigned long critical_end;
250 unsigned long critical_sequence;
252 unsigned long policy;
253 unsigned long rt_priority;
254 cycle_t preempt_timestamp;
257 char comm[TASK_COMM_LEN];
260 struct trace_iterator;
263 * The trace array - an array of per-CPU trace arrays. This is the
264 * highest level data structure that individual tracers deal with.
265 * They have on/off state as well:
268 struct ring_buffer *buffer;
269 unsigned long entries;
272 struct task_struct *waiter;
273 struct trace_array_cpu *data[NR_CPUS];
276 #define FTRACE_CMP_TYPE(var, type) \
277 __builtin_types_compatible_p(typeof(var), type *)
280 #define IF_ASSIGN(var, entry, etype, id) \
281 if (FTRACE_CMP_TYPE(var, etype)) { \
282 var = (typeof(var))(entry); \
283 WARN_ON(id && (entry)->type != id); \
287 /* Will cause compile errors if type is not found. */
288 extern void __ftrace_bad_type(void);
291 * The trace_assign_type is a verifier that the entry type is
292 * the same as the type being assigned. To add new types simply
293 * add a line with the following format:
295 * IF_ASSIGN(var, ent, type, id);
297 * Where "type" is the trace type that includes the trace_entry
298 * as the "ent" item. And "id" is the trace identifier that is
299 * used in the trace_type enum.
301 * If the type can have more than one id, then use zero.
303 #define trace_assign_type(var, ent) \
305 IF_ASSIGN(var, ent, struct ftrace_entry, TRACE_FN); \
306 IF_ASSIGN(var, ent, struct ctx_switch_entry, 0); \
307 IF_ASSIGN(var, ent, struct stack_entry, TRACE_STACK); \
308 IF_ASSIGN(var, ent, struct userstack_entry, TRACE_USER_STACK);\
309 IF_ASSIGN(var, ent, struct print_entry, TRACE_PRINT); \
310 IF_ASSIGN(var, ent, struct bprint_entry, TRACE_BPRINT); \
311 IF_ASSIGN(var, ent, struct special_entry, 0); \
312 IF_ASSIGN(var, ent, struct trace_mmiotrace_rw, \
314 IF_ASSIGN(var, ent, struct trace_mmiotrace_map, \
316 IF_ASSIGN(var, ent, struct trace_boot_call, TRACE_BOOT_CALL);\
317 IF_ASSIGN(var, ent, struct trace_boot_ret, TRACE_BOOT_RET);\
318 IF_ASSIGN(var, ent, struct trace_branch, TRACE_BRANCH); \
319 IF_ASSIGN(var, ent, struct ftrace_graph_ent_entry, \
321 IF_ASSIGN(var, ent, struct ftrace_graph_ret_entry, \
323 IF_ASSIGN(var, ent, struct hw_branch_entry, TRACE_HW_BRANCHES);\
324 IF_ASSIGN(var, ent, struct trace_power, TRACE_POWER); \
325 IF_ASSIGN(var, ent, struct kmemtrace_alloc_entry, \
327 IF_ASSIGN(var, ent, struct kmemtrace_free_entry, \
329 IF_ASSIGN(var, ent, struct syscall_trace_enter, \
330 TRACE_SYSCALL_ENTER); \
331 IF_ASSIGN(var, ent, struct syscall_trace_exit, \
332 TRACE_SYSCALL_EXIT); \
333 __ftrace_bad_type(); \
336 /* Return values for print_line callback */
338 TRACE_TYPE_PARTIAL_LINE = 0, /* Retry after flushing the seq */
339 TRACE_TYPE_HANDLED = 1,
340 TRACE_TYPE_UNHANDLED = 2, /* Relay to other output functions */
341 TRACE_TYPE_NO_CONSUME = 3 /* Handled but ask to not consume */
346 * An option specific to a tracer. This is a boolean value.
347 * The bit is the bit index that sets its value on the
348 * flags value in struct tracer_flags.
351 const char *name; /* Will appear on the trace_options file */
352 u32 bit; /* Mask assigned in val field in tracer_flags */
356 * The set of specific options for a tracer. Your tracer
357 * have to set the initial value of the flags val.
359 struct tracer_flags {
361 struct tracer_opt *opts;
364 /* Makes more easy to define a tracer opt */
365 #define TRACER_OPT(s, b) .name = #s, .bit = b
369 * struct tracer - a specific tracer and its callbacks to interact with debugfs
370 * @name: the name chosen to select it on the available_tracers file
371 * @init: called when one switches to this tracer (echo name > current_tracer)
372 * @reset: called when one switches to another tracer
373 * @start: called when tracing is unpaused (echo 1 > tracing_enabled)
374 * @stop: called when tracing is paused (echo 0 > tracing_enabled)
375 * @open: called when the trace file is opened
376 * @pipe_open: called when the trace_pipe file is opened
377 * @wait_pipe: override how the user waits for traces on trace_pipe
378 * @close: called when the trace file is released
379 * @read: override the default read callback on trace_pipe
380 * @splice_read: override the default splice_read callback on trace_pipe
381 * @selftest: selftest to run on boot (see trace_selftest.c)
382 * @print_headers: override the first lines that describe your columns
383 * @print_line: callback that prints a trace
384 * @set_flag: signals one of your private flags changed (trace_options file)
385 * @flags: your private flags
389 int (*init)(struct trace_array *tr);
390 void (*reset)(struct trace_array *tr);
391 void (*start)(struct trace_array *tr);
392 void (*stop)(struct trace_array *tr);
393 void (*open)(struct trace_iterator *iter);
394 void (*pipe_open)(struct trace_iterator *iter);
395 void (*wait_pipe)(struct trace_iterator *iter);
396 void (*close)(struct trace_iterator *iter);
397 ssize_t (*read)(struct trace_iterator *iter,
398 struct file *filp, char __user *ubuf,
399 size_t cnt, loff_t *ppos);
400 ssize_t (*splice_read)(struct trace_iterator *iter,
403 struct pipe_inode_info *pipe,
406 #ifdef CONFIG_FTRACE_STARTUP_TEST
407 int (*selftest)(struct tracer *trace,
408 struct trace_array *tr);
410 void (*print_header)(struct seq_file *m);
411 enum print_line_t (*print_line)(struct trace_iterator *iter);
412 /* If you handled the flag setting, return 0 */
413 int (*set_flag)(u32 old_flags, u32 bit, int set);
416 struct tracer_flags *flags;
417 struct tracer_stat *stats;
421 unsigned char buffer[PAGE_SIZE];
423 unsigned int readpos;
427 trace_seq_init(struct trace_seq *s)
434 #define TRACE_PIPE_ALL_CPU -1
437 * Trace iterator - used by printout routines who present trace
438 * results to users and which routines might sleep, etc:
440 struct trace_iterator {
441 struct trace_array *tr;
442 struct tracer *trace;
446 struct ring_buffer_iter *buffer_iter[NR_CPUS];
448 /* The below is zeroed out in pipe_read */
449 struct trace_seq seq;
450 struct trace_entry *ent;
454 unsigned long iter_flags;
458 cpumask_var_t started;
461 int tracer_init(struct tracer *t, struct trace_array *tr);
462 int tracing_is_enabled(void);
463 void trace_wake_up(void);
464 void tracing_reset(struct trace_array *tr, int cpu);
465 void tracing_reset_online_cpus(struct trace_array *tr);
466 int tracing_open_generic(struct inode *inode, struct file *filp);
467 struct dentry *tracing_init_dentry(void);
468 void init_tracer_sysprof_debugfs(struct dentry *d_tracer);
470 struct ring_buffer_event;
472 struct ring_buffer_event *trace_buffer_lock_reserve(struct trace_array *tr,
477 void trace_buffer_unlock_commit(struct trace_array *tr,
478 struct ring_buffer_event *event,
479 unsigned long flags, int pc);
481 struct ring_buffer_event *
482 trace_current_buffer_lock_reserve(unsigned char type, unsigned long len,
483 unsigned long flags, int pc);
484 void trace_current_buffer_unlock_commit(struct ring_buffer_event *event,
485 unsigned long flags, int pc);
487 struct trace_entry *tracing_get_trace_entry(struct trace_array *tr,
488 struct trace_array_cpu *data);
490 struct trace_entry *trace_find_next_entry(struct trace_iterator *iter,
491 int *ent_cpu, u64 *ent_ts);
493 void tracing_generic_entry_update(struct trace_entry *entry,
497 void default_wait_pipe(struct trace_iterator *iter);
498 void poll_wait_pipe(struct trace_iterator *iter);
500 void ftrace(struct trace_array *tr,
501 struct trace_array_cpu *data,
503 unsigned long parent_ip,
504 unsigned long flags, int pc);
505 void tracing_sched_switch_trace(struct trace_array *tr,
506 struct task_struct *prev,
507 struct task_struct *next,
508 unsigned long flags, int pc);
509 void tracing_record_cmdline(struct task_struct *tsk);
511 void tracing_sched_wakeup_trace(struct trace_array *tr,
512 struct task_struct *wakee,
513 struct task_struct *cur,
514 unsigned long flags, int pc);
515 void trace_special(struct trace_array *tr,
516 struct trace_array_cpu *data,
519 unsigned long arg3, int pc);
520 void trace_function(struct trace_array *tr,
522 unsigned long parent_ip,
523 unsigned long flags, int pc);
525 void trace_graph_return(struct ftrace_graph_ret *trace);
526 int trace_graph_entry(struct ftrace_graph_ent *trace);
528 void tracing_start_cmdline_record(void);
529 void tracing_stop_cmdline_record(void);
530 void tracing_sched_switch_assign_trace(struct trace_array *tr);
531 void tracing_stop_sched_switch_record(void);
532 void tracing_start_sched_switch_record(void);
533 int register_tracer(struct tracer *type);
534 void unregister_tracer(struct tracer *type);
536 extern unsigned long nsecs_to_usecs(unsigned long nsecs);
538 extern unsigned long tracing_max_latency;
539 extern unsigned long tracing_thresh;
541 void update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu);
542 void update_max_tr_single(struct trace_array *tr,
543 struct task_struct *tsk, int cpu);
545 void __trace_stack(struct trace_array *tr,
549 extern cycle_t ftrace_now(int cpu);
551 #ifdef CONFIG_CONTEXT_SWITCH_TRACER
553 (*tracer_switch_func_t)(void *private,
555 struct task_struct *prev,
556 struct task_struct *next);
558 struct tracer_switch_ops {
559 tracer_switch_func_t func;
561 struct tracer_switch_ops *next;
563 #endif /* CONFIG_CONTEXT_SWITCH_TRACER */
565 extern void trace_find_cmdline(int pid, char comm[]);
567 #ifdef CONFIG_DYNAMIC_FTRACE
568 extern unsigned long ftrace_update_tot_cnt;
569 #define DYN_FTRACE_TEST_NAME trace_selftest_dynamic_test_func
570 extern int DYN_FTRACE_TEST_NAME(void);
573 #ifdef CONFIG_FTRACE_STARTUP_TEST
574 extern int trace_selftest_startup_function(struct tracer *trace,
575 struct trace_array *tr);
576 extern int trace_selftest_startup_function_graph(struct tracer *trace,
577 struct trace_array *tr);
578 extern int trace_selftest_startup_irqsoff(struct tracer *trace,
579 struct trace_array *tr);
580 extern int trace_selftest_startup_preemptoff(struct tracer *trace,
581 struct trace_array *tr);
582 extern int trace_selftest_startup_preemptirqsoff(struct tracer *trace,
583 struct trace_array *tr);
584 extern int trace_selftest_startup_wakeup(struct tracer *trace,
585 struct trace_array *tr);
586 extern int trace_selftest_startup_nop(struct tracer *trace,
587 struct trace_array *tr);
588 extern int trace_selftest_startup_sched_switch(struct tracer *trace,
589 struct trace_array *tr);
590 extern int trace_selftest_startup_sysprof(struct tracer *trace,
591 struct trace_array *tr);
592 extern int trace_selftest_startup_branch(struct tracer *trace,
593 struct trace_array *tr);
594 #endif /* CONFIG_FTRACE_STARTUP_TEST */
596 extern void *head_page(struct trace_array_cpu *data);
597 extern long ns2usecs(cycle_t nsec);
599 trace_vbprintk(unsigned long ip, const char *fmt, va_list args);
601 trace_vprintk(unsigned long ip, const char *fmt, va_list args);
603 extern unsigned long trace_flags;
605 /* Standard output formatting function used for function return traces */
606 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
607 extern enum print_line_t print_graph_function(struct trace_iterator *iter);
609 #ifdef CONFIG_DYNAMIC_FTRACE
610 /* TODO: make this variable */
611 #define FTRACE_GRAPH_MAX_FUNCS 32
612 extern int ftrace_graph_count;
613 extern unsigned long ftrace_graph_funcs[FTRACE_GRAPH_MAX_FUNCS];
615 static inline int ftrace_graph_addr(unsigned long addr)
619 if (!ftrace_graph_count || test_tsk_trace_graph(current))
622 for (i = 0; i < ftrace_graph_count; i++) {
623 if (addr == ftrace_graph_funcs[i])
630 static inline int ftrace_trace_addr(unsigned long addr)
634 static inline int ftrace_graph_addr(unsigned long addr)
638 #endif /* CONFIG_DYNAMIC_FTRACE */
640 #else /* CONFIG_FUNCTION_GRAPH_TRACER */
641 static inline enum print_line_t
642 print_graph_function(struct trace_iterator *iter)
644 return TRACE_TYPE_UNHANDLED;
646 #endif /* CONFIG_FUNCTION_GRAPH_TRACER */
648 extern struct pid *ftrace_pid_trace;
650 static inline int ftrace_trace_task(struct task_struct *task)
652 if (!ftrace_pid_trace)
655 return test_tsk_trace_trace(task);
659 * trace_iterator_flags is an enumeration that defines bit
660 * positions into trace_flags that controls the output.
662 * NOTE: These bits must match the trace_options array in
665 enum trace_iterator_flags {
666 TRACE_ITER_PRINT_PARENT = 0x01,
667 TRACE_ITER_SYM_OFFSET = 0x02,
668 TRACE_ITER_SYM_ADDR = 0x04,
669 TRACE_ITER_VERBOSE = 0x08,
670 TRACE_ITER_RAW = 0x10,
671 TRACE_ITER_HEX = 0x20,
672 TRACE_ITER_BIN = 0x40,
673 TRACE_ITER_BLOCK = 0x80,
674 TRACE_ITER_STACKTRACE = 0x100,
675 TRACE_ITER_SCHED_TREE = 0x200,
676 TRACE_ITER_PRINTK = 0x400,
677 TRACE_ITER_PREEMPTONLY = 0x800,
678 TRACE_ITER_BRANCH = 0x1000,
679 TRACE_ITER_ANNOTATE = 0x2000,
680 TRACE_ITER_USERSTACKTRACE = 0x4000,
681 TRACE_ITER_SYM_USEROBJ = 0x8000,
682 TRACE_ITER_PRINTK_MSGONLY = 0x10000,
683 TRACE_ITER_CONTEXT_INFO = 0x20000, /* Print pid/cpu/time */
684 TRACE_ITER_LATENCY_FMT = 0x40000,
685 TRACE_ITER_GLOBAL_CLK = 0x80000,
686 TRACE_ITER_SLEEP_TIME = 0x100000,
690 * TRACE_ITER_SYM_MASK masks the options in trace_flags that
691 * control the output of kernel symbols.
693 #define TRACE_ITER_SYM_MASK \
694 (TRACE_ITER_PRINT_PARENT|TRACE_ITER_SYM_OFFSET|TRACE_ITER_SYM_ADDR)
696 extern struct tracer nop_trace;
699 * ftrace_preempt_disable - disable preemption scheduler safe
701 * When tracing can happen inside the scheduler, there exists
702 * cases that the tracing might happen before the need_resched
703 * flag is checked. If this happens and the tracer calls
704 * preempt_enable (after a disable), a schedule might take place
705 * causing an infinite recursion.
707 * To prevent this, we read the need_resched flag before
708 * disabling preemption. When we want to enable preemption we
709 * check the flag, if it is set, then we call preempt_enable_no_resched.
710 * Otherwise, we call preempt_enable.
712 * The rational for doing the above is that if need_resched is set
713 * and we have yet to reschedule, we are either in an atomic location
714 * (where we do not need to check for scheduling) or we are inside
715 * the scheduler and do not want to resched.
717 static inline int ftrace_preempt_disable(void)
721 resched = need_resched();
722 preempt_disable_notrace();
728 * ftrace_preempt_enable - enable preemption scheduler safe
729 * @resched: the return value from ftrace_preempt_disable
731 * This is a scheduler safe way to enable preemption and not miss
732 * any preemption checks. The disabled saved the state of preemption.
733 * If resched is set, then we are either inside an atomic or
734 * are inside the scheduler (we would have already scheduled
735 * otherwise). In this case, we do not want to call normal
736 * preempt_enable, but preempt_enable_no_resched instead.
738 static inline void ftrace_preempt_enable(int resched)
741 preempt_enable_no_resched_notrace();
743 preempt_enable_notrace();
746 #ifdef CONFIG_BRANCH_TRACER
747 extern int enable_branch_tracing(struct trace_array *tr);
748 extern void disable_branch_tracing(void);
749 static inline int trace_branch_enable(struct trace_array *tr)
751 if (trace_flags & TRACE_ITER_BRANCH)
752 return enable_branch_tracing(tr);
755 static inline void trace_branch_disable(void)
757 /* due to races, always disable */
758 disable_branch_tracing();
761 static inline int trace_branch_enable(struct trace_array *tr)
765 static inline void trace_branch_disable(void)
768 #endif /* CONFIG_BRANCH_TRACER */
770 /* set ring buffers to default size if not already done so */
771 int tracing_update_buffers(void);
773 /* trace event type bit fields, not numeric */
775 TRACE_EVENT_TYPE_PRINTF = 1,
776 TRACE_EVENT_TYPE_RAW = 2,
779 struct ftrace_event_call {
784 int (*regfunc)(void);
785 void (*unregfunc)(void);
787 int (*raw_init)(void);
788 int (*show_format)(struct trace_seq *s);
790 #ifdef CONFIG_EVENT_PROFILE
791 atomic_t profile_count;
792 int (*profile_enable)(struct ftrace_event_call *);
793 void (*profile_disable)(struct ftrace_event_call *);
797 void event_trace_printk(unsigned long ip, const char *fmt, ...);
798 extern struct ftrace_event_call __start_ftrace_events[];
799 extern struct ftrace_event_call __stop_ftrace_events[];
801 #define for_each_event(event) \
802 for (event = __start_ftrace_events; \
803 (unsigned long)event < (unsigned long)__stop_ftrace_events; \
806 extern const char *__start___trace_bprintk_fmt[];
807 extern const char *__stop___trace_bprintk_fmt[];
810 * The double __builtin_constant_p is because gcc will give us an error
811 * if we try to allocate the static variable to fmt if it is not a
812 * constant. Even with the outer if statement optimizing out.
814 #define event_trace_printk(ip, fmt, args...) \
816 __trace_printk_check_format(fmt, ##args); \
817 tracing_record_cmdline(current); \
818 if (__builtin_constant_p(fmt)) { \
819 static const char *trace_printk_fmt \
820 __attribute__((section("__trace_printk_fmt"))) = \
821 __builtin_constant_p(fmt) ? fmt : NULL; \
823 __trace_bprintk(ip, trace_printk_fmt, ##args); \
825 __trace_printk(ip, fmt, ##args); \
828 #endif /* _LINUX_KERNEL_TRACE_H */