1/* SPDX-License-Identifier: GPL-2.0 */
   2
   3#ifndef _LINUX_KERNEL_TRACE_H
   4#define _LINUX_KERNEL_TRACE_H
   5
   6#include <linux/fs.h>
   7#include <linux/atomic.h>
   8#include <linux/sched.h>
   9#include <linux/clocksource.h>
  10#include <linux/ring_buffer.h>
  11#include <linux/mmiotrace.h>
  12#include <linux/tracepoint.h>
  13#include <linux/ftrace.h>
  14#include <linux/hw_breakpoint.h>
  15#include <linux/trace_seq.h>
  16#include <linux/trace_events.h>
  17#include <linux/compiler.h>
  18#include <linux/trace_seq.h>
  19#include <linux/glob.h>
  20
  21#ifdef CONFIG_FTRACE_SYSCALLS
  22#include <asm/unistd.h>		/* For NR_SYSCALLS	     */
  23#include <asm/syscall.h>	/* some archs define it here */
  24#endif
  25
  26enum trace_type {
  27	__TRACE_FIRST_TYPE = 0,
  28
  29	TRACE_FN,
  30	TRACE_CTX,
  31	TRACE_WAKE,
  32	TRACE_STACK,
  33	TRACE_PRINT,
  34	TRACE_BPRINT,
  35	TRACE_MMIO_RW,
  36	TRACE_MMIO_MAP,
  37	TRACE_BRANCH,
  38	TRACE_GRAPH_RET,
  39	TRACE_GRAPH_ENT,
  40	TRACE_USER_STACK,
  41	TRACE_BLK,
  42	TRACE_BPUTS,
  43	TRACE_HWLAT,
  44	TRACE_RAW_DATA,
  45
  46	__TRACE_LAST_TYPE,
  47};
  48
  49
  50#undef __field
  51#define __field(type, item)		type	item;
  52
  53#undef __field_struct
  54#define __field_struct(type, item)	__field(type, item)
  55
  56#undef __field_desc
  57#define __field_desc(type, container, item)
  58
  59#undef __array
  60#define __array(type, item, size)	type	item[size];
  61
  62#undef __array_desc
  63#define __array_desc(type, container, item, size)
  64
  65#undef __dynamic_array
  66#define __dynamic_array(type, item)	type	item[];
  67
  68#undef F_STRUCT
  69#define F_STRUCT(args...)		args
  70
  71#undef FTRACE_ENTRY
  72#define FTRACE_ENTRY(name, struct_name, id, tstruct, print, filter)	\
  73	struct struct_name {						\
  74		struct trace_entry	ent;				\
  75		tstruct							\
  76	}
  77
  78#undef FTRACE_ENTRY_DUP
  79#define FTRACE_ENTRY_DUP(name, name_struct, id, tstruct, printk, filter)
  80
  81#undef FTRACE_ENTRY_REG
  82#define FTRACE_ENTRY_REG(name, struct_name, id, tstruct, print,	\
  83			 filter, regfn) \
  84	FTRACE_ENTRY(name, struct_name, id, PARAMS(tstruct), PARAMS(print), \
  85		     filter)
  86
  87#undef FTRACE_ENTRY_PACKED
  88#define FTRACE_ENTRY_PACKED(name, struct_name, id, tstruct, print,	\
  89			    filter)					\
  90	FTRACE_ENTRY(name, struct_name, id, PARAMS(tstruct), PARAMS(print), \
  91		     filter) __packed
  92
  93#include "trace_entries.h"
  94
  95/*
  96 * syscalls are special, and need special handling, this is why
  97 * they are not included in trace_entries.h
  98 */
  99struct syscall_trace_enter {
 100	struct trace_entry	ent;
 101	int			nr;
 102	unsigned long		args[];
 103};
 104
 105struct syscall_trace_exit {
 106	struct trace_entry	ent;
 107	int			nr;
 108	long			ret;
 109};
 110
 111struct kprobe_trace_entry_head {
 112	struct trace_entry	ent;
 113	unsigned long		ip;
 114};
 115
 116struct kretprobe_trace_entry_head {
 117	struct trace_entry	ent;
 118	unsigned long		func;
 119	unsigned long		ret_ip;
 120};
 121
 122/*
 123 * trace_flag_type is an enumeration that holds different
 124 * states when a trace occurs. These are:
 125 *  IRQS_OFF		- interrupts were disabled
 126 *  IRQS_NOSUPPORT	- arch does not support irqs_disabled_flags
 127 *  NEED_RESCHED	- reschedule is requested
 128 *  HARDIRQ		- inside an interrupt handler
 129 *  SOFTIRQ		- inside a softirq handler
 130 */
 131enum trace_flag_type {
 132	TRACE_FLAG_IRQS_OFF		= 0x01,
 133	TRACE_FLAG_IRQS_NOSUPPORT	= 0x02,
 134	TRACE_FLAG_NEED_RESCHED		= 0x04,
 135	TRACE_FLAG_HARDIRQ		= 0x08,
 136	TRACE_FLAG_SOFTIRQ		= 0x10,
 137	TRACE_FLAG_PREEMPT_RESCHED	= 0x20,
 138	TRACE_FLAG_NMI			= 0x40,
 139};
 140
 141#define TRACE_BUF_SIZE		1024
 142
 143struct trace_array;
 144
 145/*
 146 * The CPU trace array - it consists of thousands of trace entries
 147 * plus some other descriptor data: (for example which task started
 148 * the trace, etc.)
 149 */
 150struct trace_array_cpu {
 151	atomic_t		disabled;
 152	void			*buffer_page;	/* ring buffer spare */
 153
 154	unsigned long		entries;
 155	unsigned long		saved_latency;
 156	unsigned long		critical_start;
 157	unsigned long		critical_end;
 158	unsigned long		critical_sequence;
 159	unsigned long		nice;
 160	unsigned long		policy;
 161	unsigned long		rt_priority;
 162	unsigned long		skipped_entries;
 163	u64			preempt_timestamp;
 164	pid_t			pid;
 165	kuid_t			uid;
 166	char			comm[TASK_COMM_LEN];
 167
 168	bool			ignore_pid;
 169#ifdef CONFIG_FUNCTION_TRACER
 170	bool			ftrace_ignore_pid;
 171#endif
 172};
 173
 174struct tracer;
 175struct trace_option_dentry;
 176
 177struct trace_buffer {
 178	struct trace_array		*tr;
 179	struct ring_buffer		*buffer;
 180	struct trace_array_cpu __percpu	*data;
 181	u64				time_start;
 182	int				cpu;
 183};
 184
 185#define TRACE_FLAGS_MAX_SIZE		32
 186
 187struct trace_options {
 188	struct tracer			*tracer;
 189	struct trace_option_dentry	*topts;
 190};
 191
 192struct trace_pid_list {
 193	int				pid_max;
 194	unsigned long			*pids;
 
 195};
 196
 197/*
 198 * The trace array - an array of per-CPU trace arrays. This is the
 199 * highest level data structure that individual tracers deal with.
 200 * They have on/off state as well:
 201 */
 202struct trace_array {
 203	struct list_head	list;
 204	char			*name;
 205	struct trace_buffer	trace_buffer;
 206#ifdef CONFIG_TRACER_MAX_TRACE
 207	/*
 208	 * The max_buffer is used to snapshot the trace when a maximum
 209	 * latency is reached, or when the user initiates a snapshot.
 210	 * Some tracers will use this to store a maximum trace while
 211	 * it continues examining live traces.
 212	 *
 213	 * The buffers for the max_buffer are set up the same as the trace_buffer
 214	 * When a snapshot is taken, the buffer of the max_buffer is swapped
 215	 * with the buffer of the trace_buffer and the buffers are reset for
 216	 * the trace_buffer so the tracing can continue.
 217	 */
 218	struct trace_buffer	max_buffer;
 219	bool			allocated_snapshot;
 220#endif
 221#if defined(CONFIG_TRACER_MAX_TRACE) || defined(CONFIG_HWLAT_TRACER)
 222	unsigned long		max_latency;
 223#endif
 224	struct trace_pid_list	__rcu *filtered_pids;
 225	/*
 226	 * max_lock is used to protect the swapping of buffers
 227	 * when taking a max snapshot. The buffers themselves are
 228	 * protected by per_cpu spinlocks. But the action of the swap
 229	 * needs its own lock.
 230	 *
 231	 * This is defined as a arch_spinlock_t in order to help
 232	 * with performance when lockdep debugging is enabled.
 233	 *
 234	 * It is also used in other places outside the update_max_tr
 235	 * so it needs to be defined outside of the
 236	 * CONFIG_TRACER_MAX_TRACE.
 237	 */
 238	arch_spinlock_t		max_lock;
 239	int			buffer_disabled;
 240#ifdef CONFIG_FTRACE_SYSCALLS
 241	int			sys_refcount_enter;
 242	int			sys_refcount_exit;
 243	struct trace_event_file __rcu *enter_syscall_files[NR_syscalls];
 244	struct trace_event_file __rcu *exit_syscall_files[NR_syscalls];
 245#endif
 246	int			stop_count;
 247	int			clock_id;
 248	int			nr_topts;
 249	bool			clear_trace;
 250	struct tracer		*current_trace;
 251	unsigned int		trace_flags;
 252	unsigned char		trace_flags_index[TRACE_FLAGS_MAX_SIZE];
 253	unsigned int		flags;
 254	raw_spinlock_t		start_lock;
 255	struct dentry		*dir;
 256	struct dentry		*options;
 257	struct dentry		*percpu_dir;
 258	struct dentry		*event_dir;
 259	struct trace_options	*topts;
 260	struct list_head	systems;
 261	struct list_head	events;
 262	cpumask_var_t		tracing_cpumask; /* only trace on set CPUs */
 263	int			ref;
 264#ifdef CONFIG_FUNCTION_TRACER
 265	struct ftrace_ops	*ops;
 266	struct trace_pid_list	__rcu *function_pids;
 267#ifdef CONFIG_DYNAMIC_FTRACE
 268	/* All of these are protected by the ftrace_lock */
 269	struct list_head	func_probes;
 270	struct list_head	mod_trace;
 271	struct list_head	mod_notrace;
 272#endif
 273	/* function tracing enabled */
 274	int			function_enabled;
 275#endif
 276	int			time_stamp_abs_ref;
 277	struct list_head	hist_vars;
 278};
 279
 280enum {
 281	TRACE_ARRAY_FL_GLOBAL	= (1 << 0)
 282};
 283
 284extern struct list_head ftrace_trace_arrays;
 285
 286extern struct mutex trace_types_lock;
 287
 288extern int trace_array_get(struct trace_array *tr);
 289extern void trace_array_put(struct trace_array *tr);
 290
 291extern int tracing_set_time_stamp_abs(struct trace_array *tr, bool abs);
 292extern int tracing_set_clock(struct trace_array *tr, const char *clockstr);
 293
 294extern bool trace_clock_in_ns(struct trace_array *tr);
 295
 296/*
 297 * The global tracer (top) should be the first trace array added,
 298 * but we check the flag anyway.
 299 */
 300static inline struct trace_array *top_trace_array(void)
 301{
 302	struct trace_array *tr;
 303
 304	if (list_empty(&ftrace_trace_arrays))
 305		return NULL;
 306
 307	tr = list_entry(ftrace_trace_arrays.prev,
 308			typeof(*tr), list);
 309	WARN_ON(!(tr->flags & TRACE_ARRAY_FL_GLOBAL));
 310	return tr;
 311}
 312
 313#define FTRACE_CMP_TYPE(var, type) \
 314	__builtin_types_compatible_p(typeof(var), type *)
 315
 316#undef IF_ASSIGN
 317#define IF_ASSIGN(var, entry, etype, id)		\
 318	if (FTRACE_CMP_TYPE(var, etype)) {		\
 319		var = (typeof(var))(entry);		\
 320		WARN_ON(id && (entry)->type != id);	\
 321		break;					\
 322	}
 323
 324/* Will cause compile errors if type is not found. */
 325extern void __ftrace_bad_type(void);
 326
 327/*
 328 * The trace_assign_type is a verifier that the entry type is
 329 * the same as the type being assigned. To add new types simply
 330 * add a line with the following format:
 331 *
 332 * IF_ASSIGN(var, ent, type, id);
 333 *
 334 *  Where "type" is the trace type that includes the trace_entry
 335 *  as the "ent" item. And "id" is the trace identifier that is
 336 *  used in the trace_type enum.
 337 *
 338 *  If the type can have more than one id, then use zero.
 339 */
 340#define trace_assign_type(var, ent)					\
 341	do {								\
 342		IF_ASSIGN(var, ent, struct ftrace_entry, TRACE_FN);	\
 343		IF_ASSIGN(var, ent, struct ctx_switch_entry, 0);	\
 344		IF_ASSIGN(var, ent, struct stack_entry, TRACE_STACK);	\
 345		IF_ASSIGN(var, ent, struct userstack_entry, TRACE_USER_STACK);\
 346		IF_ASSIGN(var, ent, struct print_entry, TRACE_PRINT);	\
 347		IF_ASSIGN(var, ent, struct bprint_entry, TRACE_BPRINT);	\
 348		IF_ASSIGN(var, ent, struct bputs_entry, TRACE_BPUTS);	\
 349		IF_ASSIGN(var, ent, struct hwlat_entry, TRACE_HWLAT);	\
 350		IF_ASSIGN(var, ent, struct raw_data_entry, TRACE_RAW_DATA);\
 351		IF_ASSIGN(var, ent, struct trace_mmiotrace_rw,		\
 352			  TRACE_MMIO_RW);				\
 353		IF_ASSIGN(var, ent, struct trace_mmiotrace_map,		\
 354			  TRACE_MMIO_MAP);				\
 355		IF_ASSIGN(var, ent, struct trace_branch, TRACE_BRANCH); \
 356		IF_ASSIGN(var, ent, struct ftrace_graph_ent_entry,	\
 357			  TRACE_GRAPH_ENT);		\
 358		IF_ASSIGN(var, ent, struct ftrace_graph_ret_entry,	\
 359			  TRACE_GRAPH_RET);		\
 360		__ftrace_bad_type();					\
 361	} while (0)
 362
 363/*
 364 * An option specific to a tracer. This is a boolean value.
 365 * The bit is the bit index that sets its value on the
 366 * flags value in struct tracer_flags.
 367 */
 368struct tracer_opt {
 369	const char	*name; /* Will appear on the trace_options file */
 370	u32		bit; /* Mask assigned in val field in tracer_flags */
 371};
 372
 373/*
 374 * The set of specific options for a tracer. Your tracer
 375 * have to set the initial value of the flags val.
 376 */
 377struct tracer_flags {
 378	u32			val;
 379	struct tracer_opt	*opts;
 380	struct tracer		*trace;
 381};
 382
 383/* Makes more easy to define a tracer opt */
 384#define TRACER_OPT(s, b)	.name = #s, .bit = b
 385
 386
 387struct trace_option_dentry {
 388	struct tracer_opt		*opt;
 389	struct tracer_flags		*flags;
 390	struct trace_array		*tr;
 391	struct dentry			*entry;
 392};
 393
 394/**
 395 * struct tracer - a specific tracer and its callbacks to interact with tracefs
 396 * @name: the name chosen to select it on the available_tracers file
 397 * @init: called when one switches to this tracer (echo name > current_tracer)
 398 * @reset: called when one switches to another tracer
 399 * @start: called when tracing is unpaused (echo 1 > tracing_on)
 400 * @stop: called when tracing is paused (echo 0 > tracing_on)
 401 * @update_thresh: called when tracing_thresh is updated
 402 * @open: called when the trace file is opened
 403 * @pipe_open: called when the trace_pipe file is opened
 404 * @close: called when the trace file is released
 405 * @pipe_close: called when the trace_pipe file is released
 406 * @read: override the default read callback on trace_pipe
 407 * @splice_read: override the default splice_read callback on trace_pipe
 408 * @selftest: selftest to run on boot (see trace_selftest.c)
 409 * @print_headers: override the first lines that describe your columns
 410 * @print_line: callback that prints a trace
 411 * @set_flag: signals one of your private flags changed (trace_options file)
 412 * @flags: your private flags
 413 */
 414struct tracer {
 415	const char		*name;
 416	int			(*init)(struct trace_array *tr);
 417	void			(*reset)(struct trace_array *tr);
 418	void			(*start)(struct trace_array *tr);
 419	void			(*stop)(struct trace_array *tr);
 420	int			(*update_thresh)(struct trace_array *tr);
 421	void			(*open)(struct trace_iterator *iter);
 422	void			(*pipe_open)(struct trace_iterator *iter);
 423	void			(*close)(struct trace_iterator *iter);
 424	void			(*pipe_close)(struct trace_iterator *iter);
 425	ssize_t			(*read)(struct trace_iterator *iter,
 426					struct file *filp, char __user *ubuf,
 427					size_t cnt, loff_t *ppos);
 428	ssize_t			(*splice_read)(struct trace_iterator *iter,
 429					       struct file *filp,
 430					       loff_t *ppos,
 431					       struct pipe_inode_info *pipe,
 432					       size_t len,
 433					       unsigned int flags);
 434#ifdef CONFIG_FTRACE_STARTUP_TEST
 435	int			(*selftest)(struct tracer *trace,
 436					    struct trace_array *tr);
 437#endif
 438	void			(*print_header)(struct seq_file *m);
 439	enum print_line_t	(*print_line)(struct trace_iterator *iter);
 440	/* If you handled the flag setting, return 0 */
 441	int			(*set_flag)(struct trace_array *tr,
 442					    u32 old_flags, u32 bit, int set);
 443	/* Return 0 if OK with change, else return non-zero */
 444	int			(*flag_changed)(struct trace_array *tr,
 445						u32 mask, int set);
 446	struct tracer		*next;
 447	struct tracer_flags	*flags;
 448	int			enabled;
 449	int			ref;
 450	bool			print_max;
 451	bool			allow_instances;
 452#ifdef CONFIG_TRACER_MAX_TRACE
 453	bool			use_max_tr;
 454#endif
 455	/* True if tracer cannot be enabled in kernel param */
 456	bool			noboot;
 457};
 458
 459
 460/* Only current can touch trace_recursion */
 461
 462/*
 463 * For function tracing recursion:
 464 *  The order of these bits are important.
 465 *
 466 *  When function tracing occurs, the following steps are made:
 467 *   If arch does not support a ftrace feature:
 468 *    call internal function (uses INTERNAL bits) which calls...
 469 *   If callback is registered to the "global" list, the list
 470 *    function is called and recursion checks the GLOBAL bits.
 471 *    then this function calls...
 472 *   The function callback, which can use the FTRACE bits to
 473 *    check for recursion.
 474 *
 475 * Now if the arch does not suppport a feature, and it calls
 476 * the global list function which calls the ftrace callback
 477 * all three of these steps will do a recursion protection.
 478 * There's no reason to do one if the previous caller already
 479 * did. The recursion that we are protecting against will
 480 * go through the same steps again.
 481 *
 482 * To prevent the multiple recursion checks, if a recursion
 483 * bit is set that is higher than the MAX bit of the current
 484 * check, then we know that the check was made by the previous
 485 * caller, and we can skip the current check.
 486 */
 487enum {
 488	TRACE_BUFFER_BIT,
 489	TRACE_BUFFER_NMI_BIT,
 490	TRACE_BUFFER_IRQ_BIT,
 491	TRACE_BUFFER_SIRQ_BIT,
 492
 493	/* Start of function recursion bits */
 494	TRACE_FTRACE_BIT,
 495	TRACE_FTRACE_NMI_BIT,
 496	TRACE_FTRACE_IRQ_BIT,
 497	TRACE_FTRACE_SIRQ_BIT,
 498
 499	/* INTERNAL_BITs must be greater than FTRACE_BITs */
 500	TRACE_INTERNAL_BIT,
 501	TRACE_INTERNAL_NMI_BIT,
 502	TRACE_INTERNAL_IRQ_BIT,
 503	TRACE_INTERNAL_SIRQ_BIT,
 504
 505	TRACE_BRANCH_BIT,
 506/*
 507 * Abuse of the trace_recursion.
 508 * As we need a way to maintain state if we are tracing the function
 509 * graph in irq because we want to trace a particular function that
 510 * was called in irq context but we have irq tracing off. Since this
 511 * can only be modified by current, we can reuse trace_recursion.
 512 */
 513	TRACE_IRQ_BIT,
 514};
 515
 516#define trace_recursion_set(bit)	do { (current)->trace_recursion |= (1<<(bit)); } while (0)
 517#define trace_recursion_clear(bit)	do { (current)->trace_recursion &= ~(1<<(bit)); } while (0)
 518#define trace_recursion_test(bit)	((current)->trace_recursion & (1<<(bit)))
 519
 520#define TRACE_CONTEXT_BITS	4
 521
 522#define TRACE_FTRACE_START	TRACE_FTRACE_BIT
 523#define TRACE_FTRACE_MAX	((1 << (TRACE_FTRACE_START + TRACE_CONTEXT_BITS)) - 1)
 524
 525#define TRACE_LIST_START	TRACE_INTERNAL_BIT
 526#define TRACE_LIST_MAX		((1 << (TRACE_LIST_START + TRACE_CONTEXT_BITS)) - 1)
 527
 528#define TRACE_CONTEXT_MASK	TRACE_LIST_MAX
 529
 530static __always_inline int trace_get_context_bit(void)
 531{
 532	int bit;
 533
 534	if (in_interrupt()) {
 535		if (in_nmi())
 536			bit = 0;
 537
 538		else if (in_irq())
 539			bit = 1;
 540		else
 541			bit = 2;
 542	} else
 543		bit = 3;
 544
 545	return bit;
 546}
 547
 548static __always_inline int trace_test_and_set_recursion(int start, int max)
 549{
 550	unsigned int val = current->trace_recursion;
 551	int bit;
 552
 553	/* A previous recursion check was made */
 554	if ((val & TRACE_CONTEXT_MASK) > max)
 555		return 0;
 556
 557	bit = trace_get_context_bit() + start;
 558	if (unlikely(val & (1 << bit)))
 559		return -1;
 560
 561	val |= 1 << bit;
 562	current->trace_recursion = val;
 563	barrier();
 564
 565	return bit;
 566}
 567
 568static __always_inline void trace_clear_recursion(int bit)
 569{
 570	unsigned int val = current->trace_recursion;
 571
 572	if (!bit)
 573		return;
 574
 575	bit = 1 << bit;
 576	val &= ~bit;
 577
 578	barrier();
 579	current->trace_recursion = val;
 580}
 581
 582static inline struct ring_buffer_iter *
 583trace_buffer_iter(struct trace_iterator *iter, int cpu)
 584{
 585	if (iter->buffer_iter && iter->buffer_iter[cpu])
 586		return iter->buffer_iter[cpu];
 587	return NULL;
 588}
 589
 590int tracer_init(struct tracer *t, struct trace_array *tr);
 591int tracing_is_enabled(void);
 592void tracing_reset(struct trace_buffer *buf, int cpu);
 593void tracing_reset_online_cpus(struct trace_buffer *buf);
 594void tracing_reset_current(int cpu);
 595void tracing_reset_all_online_cpus(void);
 596int tracing_open_generic(struct inode *inode, struct file *filp);
 597bool tracing_is_disabled(void);
 598int tracer_tracing_is_on(struct trace_array *tr);
 599void tracer_tracing_on(struct trace_array *tr);
 600void tracer_tracing_off(struct trace_array *tr);
 601struct dentry *trace_create_file(const char *name,
 602				 umode_t mode,
 603				 struct dentry *parent,
 604				 void *data,
 605				 const struct file_operations *fops);
 606
 607struct dentry *tracing_init_dentry(void);
 608
 609struct ring_buffer_event;
 610
 611struct ring_buffer_event *
 612trace_buffer_lock_reserve(struct ring_buffer *buffer,
 613			  int type,
 614			  unsigned long len,
 615			  unsigned long flags,
 616			  int pc);
 617
 618struct trace_entry *tracing_get_trace_entry(struct trace_array *tr,
 619						struct trace_array_cpu *data);
 620
 621struct trace_entry *trace_find_next_entry(struct trace_iterator *iter,
 622					  int *ent_cpu, u64 *ent_ts);
 623
 624void trace_buffer_unlock_commit_nostack(struct ring_buffer *buffer,
 625					struct ring_buffer_event *event);
 626
 627int trace_empty(struct trace_iterator *iter);
 628
 629void *trace_find_next_entry_inc(struct trace_iterator *iter);
 630
 631void trace_init_global_iter(struct trace_iterator *iter);
 632
 633void tracing_iter_reset(struct trace_iterator *iter, int cpu);
 634
 635void trace_function(struct trace_array *tr,
 636		    unsigned long ip,
 637		    unsigned long parent_ip,
 638		    unsigned long flags, int pc);
 639void trace_graph_function(struct trace_array *tr,
 640		    unsigned long ip,
 641		    unsigned long parent_ip,
 642		    unsigned long flags, int pc);
 643void trace_latency_header(struct seq_file *m);
 644void trace_default_header(struct seq_file *m);
 645void print_trace_header(struct seq_file *m, struct trace_iterator *iter);
 646int trace_empty(struct trace_iterator *iter);
 647
 648void trace_graph_return(struct ftrace_graph_ret *trace);
 649int trace_graph_entry(struct ftrace_graph_ent *trace);
 650void set_graph_array(struct trace_array *tr);
 651
 652void tracing_start_cmdline_record(void);
 653void tracing_stop_cmdline_record(void);
 654void tracing_start_tgid_record(void);
 655void tracing_stop_tgid_record(void);
 656
 657int register_tracer(struct tracer *type);
 658int is_tracing_stopped(void);
 659
 660loff_t tracing_lseek(struct file *file, loff_t offset, int whence);
 661
 662extern cpumask_var_t __read_mostly tracing_buffer_mask;
 663
 664#define for_each_tracing_cpu(cpu)	\
 665	for_each_cpu(cpu, tracing_buffer_mask)
 666
 667extern unsigned long nsecs_to_usecs(unsigned long nsecs);
 668
 669extern unsigned long tracing_thresh;
 670
 671/* PID filtering */
 672
 673extern int pid_max;
 674
 675bool trace_find_filtered_pid(struct trace_pid_list *filtered_pids,
 676			     pid_t search_pid);
 677bool trace_ignore_this_task(struct trace_pid_list *filtered_pids,
 678			    struct task_struct *task);
 679void trace_filter_add_remove_task(struct trace_pid_list *pid_list,
 680				  struct task_struct *self,
 681				  struct task_struct *task);
 682void *trace_pid_next(struct trace_pid_list *pid_list, void *v, loff_t *pos);
 683void *trace_pid_start(struct trace_pid_list *pid_list, loff_t *pos);
 684int trace_pid_show(struct seq_file *m, void *v);
 685void trace_free_pid_list(struct trace_pid_list *pid_list);
 686int trace_pid_write(struct trace_pid_list *filtered_pids,
 687		    struct trace_pid_list **new_pid_list,
 688		    const char __user *ubuf, size_t cnt);
 689
 690#ifdef CONFIG_TRACER_MAX_TRACE
 691void update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu);
 692void update_max_tr_single(struct trace_array *tr,
 693			  struct task_struct *tsk, int cpu);
 694#endif /* CONFIG_TRACER_MAX_TRACE */
 695
 696#ifdef CONFIG_STACKTRACE
 697void ftrace_trace_userstack(struct ring_buffer *buffer, unsigned long flags,
 698			    int pc);
 699
 700void __trace_stack(struct trace_array *tr, unsigned long flags, int skip,
 701		   int pc);
 702#else
 703static inline void ftrace_trace_userstack(struct ring_buffer *buffer,
 704					  unsigned long flags, int pc)
 705{
 706}
 707
 708static inline void __trace_stack(struct trace_array *tr, unsigned long flags,
 709				 int skip, int pc)
 710{
 711}
 712#endif /* CONFIG_STACKTRACE */
 713
 714extern u64 ftrace_now(int cpu);
 715
 716extern void trace_find_cmdline(int pid, char comm[]);
 717extern int trace_find_tgid(int pid);
 718extern void trace_event_follow_fork(struct trace_array *tr, bool enable);
 719
 720#ifdef CONFIG_DYNAMIC_FTRACE
 721extern unsigned long ftrace_update_tot_cnt;
 722void ftrace_init_trace_array(struct trace_array *tr);
 723#else
 724static inline void ftrace_init_trace_array(struct trace_array *tr) { }
 725#endif
 726#define DYN_FTRACE_TEST_NAME trace_selftest_dynamic_test_func
 727extern int DYN_FTRACE_TEST_NAME(void);
 728#define DYN_FTRACE_TEST_NAME2 trace_selftest_dynamic_test_func2
 729extern int DYN_FTRACE_TEST_NAME2(void);
 730
 731extern bool ring_buffer_expanded;
 732extern bool tracing_selftest_disabled;
 733
 734#ifdef CONFIG_FTRACE_STARTUP_TEST
 735extern int trace_selftest_startup_function(struct tracer *trace,
 736					   struct trace_array *tr);
 737extern int trace_selftest_startup_function_graph(struct tracer *trace,
 738						 struct trace_array *tr);
 739extern int trace_selftest_startup_irqsoff(struct tracer *trace,
 740					  struct trace_array *tr);
 741extern int trace_selftest_startup_preemptoff(struct tracer *trace,
 742					     struct trace_array *tr);
 743extern int trace_selftest_startup_preemptirqsoff(struct tracer *trace,
 744						 struct trace_array *tr);
 745extern int trace_selftest_startup_wakeup(struct tracer *trace,
 746					 struct trace_array *tr);
 747extern int trace_selftest_startup_nop(struct tracer *trace,
 748					 struct trace_array *tr);
 
 
 749extern int trace_selftest_startup_branch(struct tracer *trace,
 750					 struct trace_array *tr);
 751/*
 752 * Tracer data references selftest functions that only occur
 753 * on boot up. These can be __init functions. Thus, when selftests
 754 * are enabled, then the tracers need to reference __init functions.
 755 */
 756#define __tracer_data		__refdata
 757#else
 758/* Tracers are seldom changed. Optimize when selftests are disabled. */
 759#define __tracer_data		__read_mostly
 760#endif /* CONFIG_FTRACE_STARTUP_TEST */
 761
 762extern void *head_page(struct trace_array_cpu *data);
 763extern unsigned long long ns2usecs(u64 nsec);
 764extern int
 765trace_vbprintk(unsigned long ip, const char *fmt, va_list args);
 766extern int
 767trace_vprintk(unsigned long ip, const char *fmt, va_list args);
 768extern int
 769trace_array_vprintk(struct trace_array *tr,
 770		    unsigned long ip, const char *fmt, va_list args);
 771int trace_array_printk(struct trace_array *tr,
 772		       unsigned long ip, const char *fmt, ...);
 773int trace_array_printk_buf(struct ring_buffer *buffer,
 774			   unsigned long ip, const char *fmt, ...);
 775void trace_printk_seq(struct trace_seq *s);
 776enum print_line_t print_trace_line(struct trace_iterator *iter);
 777
 778extern char trace_find_mark(unsigned long long duration);
 779
 780struct ftrace_hash;
 781
 782struct ftrace_mod_load {
 783	struct list_head	list;
 784	char			*func;
 785	char			*module;
 786	int			 enable;
 787};
 788
 789enum {
 790	FTRACE_HASH_FL_MOD	= (1 << 0),
 791};
 792
 793struct ftrace_hash {
 794	unsigned long		size_bits;
 795	struct hlist_head	*buckets;
 796	unsigned long		count;
 797	unsigned long		flags;
 798	struct rcu_head		rcu;
 799};
 800
 801struct ftrace_func_entry *
 802ftrace_lookup_ip(struct ftrace_hash *hash, unsigned long ip);
 803
 804static __always_inline bool ftrace_hash_empty(struct ftrace_hash *hash)
 805{
 806	return !hash || !(hash->count || (hash->flags & FTRACE_HASH_FL_MOD));
 807}
 808
 809/* Standard output formatting function used for function return traces */
 810#ifdef CONFIG_FUNCTION_GRAPH_TRACER
 811
 812/* Flag options */
 813#define TRACE_GRAPH_PRINT_OVERRUN       0x1
 814#define TRACE_GRAPH_PRINT_CPU           0x2
 815#define TRACE_GRAPH_PRINT_OVERHEAD      0x4
 816#define TRACE_GRAPH_PRINT_PROC          0x8
 817#define TRACE_GRAPH_PRINT_DURATION      0x10
 818#define TRACE_GRAPH_PRINT_ABS_TIME      0x20
 819#define TRACE_GRAPH_PRINT_IRQS          0x40
 820#define TRACE_GRAPH_PRINT_TAIL          0x80
 821#define TRACE_GRAPH_SLEEP_TIME		0x100
 822#define TRACE_GRAPH_GRAPH_TIME		0x200
 823#define TRACE_GRAPH_PRINT_FILL_SHIFT	28
 824#define TRACE_GRAPH_PRINT_FILL_MASK	(0x3 << TRACE_GRAPH_PRINT_FILL_SHIFT)
 825
 826extern void ftrace_graph_sleep_time_control(bool enable);
 827extern void ftrace_graph_graph_time_control(bool enable);
 828
 829extern enum print_line_t
 830print_graph_function_flags(struct trace_iterator *iter, u32 flags);
 831extern void print_graph_headers_flags(struct seq_file *s, u32 flags);
 832extern void
 833trace_print_graph_duration(unsigned long long duration, struct trace_seq *s);
 834extern void graph_trace_open(struct trace_iterator *iter);
 835extern void graph_trace_close(struct trace_iterator *iter);
 836extern int __trace_graph_entry(struct trace_array *tr,
 837			       struct ftrace_graph_ent *trace,
 838			       unsigned long flags, int pc);
 839extern void __trace_graph_return(struct trace_array *tr,
 840				 struct ftrace_graph_ret *trace,
 841				 unsigned long flags, int pc);
 842
 
 843#ifdef CONFIG_DYNAMIC_FTRACE
 844extern struct ftrace_hash *ftrace_graph_hash;
 845extern struct ftrace_hash *ftrace_graph_notrace_hash;
 
 
 
 
 846
 847static inline int ftrace_graph_addr(unsigned long addr)
 848{
 849	int ret = 0;
 850
 851	preempt_disable_notrace();
 
 852
 853	if (ftrace_hash_empty(ftrace_graph_hash)) {
 854		ret = 1;
 855		goto out;
 
 
 
 
 
 
 
 
 
 
 856	}
 857
 858	if (ftrace_lookup_ip(ftrace_graph_hash, addr)) {
 859		/*
 860		 * If no irqs are to be traced, but a set_graph_function
 861		 * is set, and called by an interrupt handler, we still
 862		 * want to trace it.
 863		 */
 864		if (in_irq())
 865			trace_recursion_set(TRACE_IRQ_BIT);
 866		else
 867			trace_recursion_clear(TRACE_IRQ_BIT);
 868		ret = 1;
 869	}
 870
 871out:
 872	preempt_enable_notrace();
 873	return ret;
 874}
 875
 876static inline int ftrace_graph_notrace_addr(unsigned long addr)
 877{
 878	int ret = 0;
 879
 880	preempt_disable_notrace();
 
 881
 882	if (ftrace_lookup_ip(ftrace_graph_notrace_hash, addr))
 883		ret = 1;
 
 
 884
 885	preempt_enable_notrace();
 886	return ret;
 887}
 888#else
 889static inline int ftrace_graph_addr(unsigned long addr)
 890{
 891	return 1;
 892}
 893
 894static inline int ftrace_graph_notrace_addr(unsigned long addr)
 895{
 896	return 0;
 897}
 898#endif /* CONFIG_DYNAMIC_FTRACE */
 899
 900extern unsigned int fgraph_max_depth;
 901
 902static inline bool ftrace_graph_ignore_func(struct ftrace_graph_ent *trace)
 903{
 904	/* trace it when it is-nested-in or is a function enabled. */
 905	return !(trace->depth || ftrace_graph_addr(trace->func)) ||
 906		(trace->depth < 0) ||
 907		(fgraph_max_depth && trace->depth >= fgraph_max_depth);
 908}
 909
 910#else /* CONFIG_FUNCTION_GRAPH_TRACER */
 911static inline enum print_line_t
 912print_graph_function_flags(struct trace_iterator *iter, u32 flags)
 913{
 914	return TRACE_TYPE_UNHANDLED;
 915}
 916#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
 917
 918extern struct list_head ftrace_pids;
 919
 920#ifdef CONFIG_FUNCTION_TRACER
 921struct ftrace_func_command {
 922	struct list_head	list;
 923	char			*name;
 924	int			(*func)(struct trace_array *tr,
 925					struct ftrace_hash *hash,
 926					char *func, char *cmd,
 927					char *params, int enable);
 928};
 929extern bool ftrace_filter_param __initdata;
 930static inline int ftrace_trace_task(struct trace_array *tr)
 931{
 932	return !this_cpu_read(tr->trace_buffer.data->ftrace_ignore_pid);
 
 
 
 933}
 934extern int ftrace_is_dead(void);
 935int ftrace_create_function_files(struct trace_array *tr,
 936				 struct dentry *parent);
 937void ftrace_destroy_function_files(struct trace_array *tr);
 938void ftrace_init_global_array_ops(struct trace_array *tr);
 939void ftrace_init_array_ops(struct trace_array *tr, ftrace_func_t func);
 940void ftrace_reset_array_ops(struct trace_array *tr);
 941int using_ftrace_ops_list_func(void);
 942void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d_tracer);
 943void ftrace_init_tracefs_toplevel(struct trace_array *tr,
 944				  struct dentry *d_tracer);
 945void ftrace_clear_pids(struct trace_array *tr);
 946int init_function_trace(void);
 947void ftrace_pid_follow_fork(struct trace_array *tr, bool enable);
 948#else
 949static inline int ftrace_trace_task(struct trace_array *tr)
 950{
 951	return 1;
 952}
 953static inline int ftrace_is_dead(void) { return 0; }
 954static inline int
 955ftrace_create_function_files(struct trace_array *tr,
 956			     struct dentry *parent)
 957{
 958	return 0;
 959}
 960static inline void ftrace_destroy_function_files(struct trace_array *tr) { }
 961static inline __init void
 962ftrace_init_global_array_ops(struct trace_array *tr) { }
 963static inline void ftrace_reset_array_ops(struct trace_array *tr) { }
 964static inline void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d) { }
 965static inline void ftrace_init_tracefs_toplevel(struct trace_array *tr, struct dentry *d) { }
 966static inline void ftrace_clear_pids(struct trace_array *tr) { }
 967static inline int init_function_trace(void) { return 0; }
 968static inline void ftrace_pid_follow_fork(struct trace_array *tr, bool enable) { }
 969/* ftace_func_t type is not defined, use macro instead of static inline */
 970#define ftrace_init_array_ops(tr, func) do { } while (0)
 971#endif /* CONFIG_FUNCTION_TRACER */
 972
 973#if defined(CONFIG_FUNCTION_TRACER) && defined(CONFIG_DYNAMIC_FTRACE)
 974
 975struct ftrace_probe_ops {
 976	void			(*func)(unsigned long ip,
 977					unsigned long parent_ip,
 978					struct trace_array *tr,
 979					struct ftrace_probe_ops *ops,
 980					void *data);
 981	int			(*init)(struct ftrace_probe_ops *ops,
 982					struct trace_array *tr,
 983					unsigned long ip, void *init_data,
 984					void **data);
 985	void			(*free)(struct ftrace_probe_ops *ops,
 986					struct trace_array *tr,
 987					unsigned long ip, void *data);
 988	int			(*print)(struct seq_file *m,
 989					 unsigned long ip,
 990					 struct ftrace_probe_ops *ops,
 991					 void *data);
 992};
 993
 994struct ftrace_func_mapper;
 995typedef int (*ftrace_mapper_func)(void *data);
 996
 997struct ftrace_func_mapper *allocate_ftrace_func_mapper(void);
 998void **ftrace_func_mapper_find_ip(struct ftrace_func_mapper *mapper,
 999					   unsigned long ip);
1000int ftrace_func_mapper_add_ip(struct ftrace_func_mapper *mapper,
1001			       unsigned long ip, void *data);
1002void *ftrace_func_mapper_remove_ip(struct ftrace_func_mapper *mapper,
1003				   unsigned long ip);
1004void free_ftrace_func_mapper(struct ftrace_func_mapper *mapper,
1005			     ftrace_mapper_func free_func);
1006
1007extern int
1008register_ftrace_function_probe(char *glob, struct trace_array *tr,
1009			       struct ftrace_probe_ops *ops, void *data);
1010extern int
1011unregister_ftrace_function_probe_func(char *glob, struct trace_array *tr,
1012				      struct ftrace_probe_ops *ops);
1013extern void clear_ftrace_function_probes(struct trace_array *tr);
1014
1015int register_ftrace_command(struct ftrace_func_command *cmd);
1016int unregister_ftrace_command(struct ftrace_func_command *cmd);
1017
1018void ftrace_create_filter_files(struct ftrace_ops *ops,
1019				struct dentry *parent);
1020void ftrace_destroy_filter_files(struct ftrace_ops *ops);
1021#else
1022struct ftrace_func_command;
1023
1024static inline __init int register_ftrace_command(struct ftrace_func_command *cmd)
1025{
1026	return -EINVAL;
1027}
1028static inline __init int unregister_ftrace_command(char *cmd_name)
1029{
1030	return -EINVAL;
1031}
1032static inline void clear_ftrace_function_probes(struct trace_array *tr)
1033{
1034}
1035
1036/*
1037 * The ops parameter passed in is usually undefined.
1038 * This must be a macro.
1039 */
1040#define ftrace_create_filter_files(ops, parent) do { } while (0)
1041#define ftrace_destroy_filter_files(ops) do { } while (0)
1042#endif /* CONFIG_FUNCTION_TRACER && CONFIG_DYNAMIC_FTRACE */
1043
1044bool ftrace_event_is_function(struct trace_event_call *call);
1045
1046/*
1047 * struct trace_parser - servers for reading the user input separated by spaces
1048 * @cont: set if the input is not complete - no final space char was found
1049 * @buffer: holds the parsed user input
1050 * @idx: user input length
1051 * @size: buffer size
1052 */
1053struct trace_parser {
1054	bool		cont;
1055	char		*buffer;
1056	unsigned	idx;
1057	unsigned	size;
1058};
1059
1060static inline bool trace_parser_loaded(struct trace_parser *parser)
1061{
1062	return (parser->idx != 0);
1063}
1064
1065static inline bool trace_parser_cont(struct trace_parser *parser)
1066{
1067	return parser->cont;
1068}
1069
1070static inline void trace_parser_clear(struct trace_parser *parser)
1071{
1072	parser->cont = false;
1073	parser->idx = 0;
1074}
1075
1076extern int trace_parser_get_init(struct trace_parser *parser, int size);
1077extern void trace_parser_put(struct trace_parser *parser);
1078extern int trace_get_user(struct trace_parser *parser, const char __user *ubuf,
1079	size_t cnt, loff_t *ppos);
1080
1081/*
1082 * Only create function graph options if function graph is configured.
1083 */
1084#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1085# define FGRAPH_FLAGS						\
1086		C(DISPLAY_GRAPH,	"display-graph"),
1087#else
1088# define FGRAPH_FLAGS
1089#endif
1090
1091#ifdef CONFIG_BRANCH_TRACER
1092# define BRANCH_FLAGS					\
1093		C(BRANCH,		"branch"),
1094#else
1095# define BRANCH_FLAGS
1096#endif
1097
1098#ifdef CONFIG_FUNCTION_TRACER
1099# define FUNCTION_FLAGS						\
1100		C(FUNCTION,		"function-trace"),	\
1101		C(FUNC_FORK,		"function-fork"),
1102# define FUNCTION_DEFAULT_FLAGS		TRACE_ITER_FUNCTION
1103#else
1104# define FUNCTION_FLAGS
1105# define FUNCTION_DEFAULT_FLAGS		0UL
1106# define TRACE_ITER_FUNC_FORK		0UL
1107#endif
1108
1109#ifdef CONFIG_STACKTRACE
1110# define STACK_FLAGS				\
1111		C(STACKTRACE,		"stacktrace"),
1112#else
1113# define STACK_FLAGS
1114#endif
1115
1116/*
1117 * trace_iterator_flags is an enumeration that defines bit
1118 * positions into trace_flags that controls the output.
1119 *
1120 * NOTE: These bits must match the trace_options array in
1121 *       trace.c (this macro guarantees it).
1122 */
1123#define TRACE_FLAGS						\
1124		C(PRINT_PARENT,		"print-parent"),	\
1125		C(SYM_OFFSET,		"sym-offset"),		\
1126		C(SYM_ADDR,		"sym-addr"),		\
1127		C(VERBOSE,		"verbose"),		\
1128		C(RAW,			"raw"),			\
1129		C(HEX,			"hex"),			\
1130		C(BIN,			"bin"),			\
1131		C(BLOCK,		"block"),		\
1132		C(PRINTK,		"trace_printk"),	\
1133		C(ANNOTATE,		"annotate"),		\
1134		C(USERSTACKTRACE,	"userstacktrace"),	\
1135		C(SYM_USEROBJ,		"sym-userobj"),		\
1136		C(PRINTK_MSGONLY,	"printk-msg-only"),	\
1137		C(CONTEXT_INFO,		"context-info"),   /* Print pid/cpu/time */ \
1138		C(LATENCY_FMT,		"latency-format"),	\
1139		C(RECORD_CMD,		"record-cmd"),		\
1140		C(RECORD_TGID,		"record-tgid"),		\
1141		C(OVERWRITE,		"overwrite"),		\
1142		C(STOP_ON_FREE,		"disable_on_free"),	\
1143		C(IRQ_INFO,		"irq-info"),		\
1144		C(MARKERS,		"markers"),		\
1145		C(EVENT_FORK,		"event-fork"),		\
1146		FUNCTION_FLAGS					\
1147		FGRAPH_FLAGS					\
1148		STACK_FLAGS					\
1149		BRANCH_FLAGS
1150
1151/*
1152 * By defining C, we can make TRACE_FLAGS a list of bit names
1153 * that will define the bits for the flag masks.
1154 */
1155#undef C
1156#define C(a, b) TRACE_ITER_##a##_BIT
1157
1158enum trace_iterator_bits {
1159	TRACE_FLAGS
1160	/* Make sure we don't go more than we have bits for */
1161	TRACE_ITER_LAST_BIT
1162};
1163
1164/*
1165 * By redefining C, we can make TRACE_FLAGS a list of masks that
1166 * use the bits as defined above.
1167 */
1168#undef C
1169#define C(a, b) TRACE_ITER_##a = (1 << TRACE_ITER_##a##_BIT)
1170
1171enum trace_iterator_flags { TRACE_FLAGS };
1172
1173/*
1174 * TRACE_ITER_SYM_MASK masks the options in trace_flags that
1175 * control the output of kernel symbols.
1176 */
1177#define TRACE_ITER_SYM_MASK \
1178	(TRACE_ITER_PRINT_PARENT|TRACE_ITER_SYM_OFFSET|TRACE_ITER_SYM_ADDR)
1179
1180extern struct tracer nop_trace;
1181
1182#ifdef CONFIG_BRANCH_TRACER
1183extern int enable_branch_tracing(struct trace_array *tr);
1184extern void disable_branch_tracing(void);
1185static inline int trace_branch_enable(struct trace_array *tr)
1186{
1187	if (tr->trace_flags & TRACE_ITER_BRANCH)
1188		return enable_branch_tracing(tr);
1189	return 0;
1190}
1191static inline void trace_branch_disable(void)
1192{
1193	/* due to races, always disable */
1194	disable_branch_tracing();
1195}
1196#else
1197static inline int trace_branch_enable(struct trace_array *tr)
1198{
1199	return 0;
1200}
1201static inline void trace_branch_disable(void)
1202{
1203}
1204#endif /* CONFIG_BRANCH_TRACER */
1205
1206/* set ring buffers to default size if not already done so */
1207int tracing_update_buffers(void);
1208
1209struct ftrace_event_field {
1210	struct list_head	link;
1211	const char		*name;
1212	const char		*type;
1213	int			filter_type;
1214	int			offset;
1215	int			size;
1216	int			is_signed;
1217};
1218
1219struct prog_entry;
1220
1221struct event_filter {
1222	struct prog_entry __rcu	*prog;
 
 
 
1223	char			*filter_string;
1224};
1225
1226struct event_subsystem {
1227	struct list_head	list;
1228	const char		*name;
1229	struct event_filter	*filter;
1230	int			ref_count;
1231};
1232
1233struct trace_subsystem_dir {
1234	struct list_head		list;
1235	struct event_subsystem		*subsystem;
1236	struct trace_array		*tr;
1237	struct dentry			*entry;
1238	int				ref_count;
1239	int				nr_events;
1240};
1241
1242extern int call_filter_check_discard(struct trace_event_call *call, void *rec,
1243				     struct ring_buffer *buffer,
1244				     struct ring_buffer_event *event);
1245
1246void trace_buffer_unlock_commit_regs(struct trace_array *tr,
1247				     struct ring_buffer *buffer,
1248				     struct ring_buffer_event *event,
1249				     unsigned long flags, int pc,
1250				     struct pt_regs *regs);
1251
1252static inline void trace_buffer_unlock_commit(struct trace_array *tr,
1253					      struct ring_buffer *buffer,
1254					      struct ring_buffer_event *event,
1255					      unsigned long flags, int pc)
1256{
1257	trace_buffer_unlock_commit_regs(tr, buffer, event, flags, pc, NULL);
1258}
1259
1260DECLARE_PER_CPU(struct ring_buffer_event *, trace_buffered_event);
1261DECLARE_PER_CPU(int, trace_buffered_event_cnt);
1262void trace_buffered_event_disable(void);
1263void trace_buffered_event_enable(void);
1264
1265static inline void
1266__trace_event_discard_commit(struct ring_buffer *buffer,
1267			     struct ring_buffer_event *event)
1268{
1269	if (this_cpu_read(trace_buffered_event) == event) {
1270		/* Simply release the temp buffer */
1271		this_cpu_dec(trace_buffered_event_cnt);
1272		return;
1273	}
1274	ring_buffer_discard_commit(buffer, event);
1275}
1276
1277/*
1278 * Helper function for event_trigger_unlock_commit{_regs}().
1279 * If there are event triggers attached to this event that requires
1280 * filtering against its fields, then they wil be called as the
1281 * entry already holds the field information of the current event.
1282 *
1283 * It also checks if the event should be discarded or not.
1284 * It is to be discarded if the event is soft disabled and the
1285 * event was only recorded to process triggers, or if the event
1286 * filter is active and this event did not match the filters.
1287 *
1288 * Returns true if the event is discarded, false otherwise.
1289 */
1290static inline bool
1291__event_trigger_test_discard(struct trace_event_file *file,
1292			     struct ring_buffer *buffer,
1293			     struct ring_buffer_event *event,
1294			     void *entry,
1295			     enum event_trigger_type *tt)
1296{
1297	unsigned long eflags = file->flags;
1298
1299	if (eflags & EVENT_FILE_FL_TRIGGER_COND)
1300		*tt = event_triggers_call(file, entry, event);
1301
1302	if (test_bit(EVENT_FILE_FL_SOFT_DISABLED_BIT, &file->flags) ||
1303	    (unlikely(file->flags & EVENT_FILE_FL_FILTERED) &&
1304	     !filter_match_preds(file->filter, entry))) {
1305		__trace_event_discard_commit(buffer, event);
1306		return true;
1307	}
1308
1309	return false;
1310}
1311
1312/**
1313 * event_trigger_unlock_commit - handle triggers and finish event commit
1314 * @file: The file pointer assoctiated to the event
1315 * @buffer: The ring buffer that the event is being written to
1316 * @event: The event meta data in the ring buffer
1317 * @entry: The event itself
1318 * @irq_flags: The state of the interrupts at the start of the event
1319 * @pc: The state of the preempt count at the start of the event.
1320 *
1321 * This is a helper function to handle triggers that require data
1322 * from the event itself. It also tests the event against filters and
1323 * if the event is soft disabled and should be discarded.
1324 */
1325static inline void
1326event_trigger_unlock_commit(struct trace_event_file *file,
1327			    struct ring_buffer *buffer,
1328			    struct ring_buffer_event *event,
1329			    void *entry, unsigned long irq_flags, int pc)
1330{
1331	enum event_trigger_type tt = ETT_NONE;
1332
1333	if (!__event_trigger_test_discard(file, buffer, event, entry, &tt))
1334		trace_buffer_unlock_commit(file->tr, buffer, event, irq_flags, pc);
1335
1336	if (tt)
1337		event_triggers_post_call(file, tt, entry, event);
1338}
1339
1340/**
1341 * event_trigger_unlock_commit_regs - handle triggers and finish event commit
1342 * @file: The file pointer assoctiated to the event
1343 * @buffer: The ring buffer that the event is being written to
1344 * @event: The event meta data in the ring buffer
1345 * @entry: The event itself
1346 * @irq_flags: The state of the interrupts at the start of the event
1347 * @pc: The state of the preempt count at the start of the event.
1348 *
1349 * This is a helper function to handle triggers that require data
1350 * from the event itself. It also tests the event against filters and
1351 * if the event is soft disabled and should be discarded.
1352 *
1353 * Same as event_trigger_unlock_commit() but calls
1354 * trace_buffer_unlock_commit_regs() instead of trace_buffer_unlock_commit().
1355 */
1356static inline void
1357event_trigger_unlock_commit_regs(struct trace_event_file *file,
1358				 struct ring_buffer *buffer,
1359				 struct ring_buffer_event *event,
1360				 void *entry, unsigned long irq_flags, int pc,
1361				 struct pt_regs *regs)
1362{
1363	enum event_trigger_type tt = ETT_NONE;
1364
1365	if (!__event_trigger_test_discard(file, buffer, event, entry, &tt))
1366		trace_buffer_unlock_commit_regs(file->tr, buffer, event,
1367						irq_flags, pc, regs);
1368
1369	if (tt)
1370		event_triggers_post_call(file, tt, entry, event);
1371}
1372
1373#define FILTER_PRED_INVALID	((unsigned short)-1)
1374#define FILTER_PRED_IS_RIGHT	(1 << 15)
1375#define FILTER_PRED_FOLD	(1 << 15)
1376
1377/*
1378 * The max preds is the size of unsigned short with
1379 * two flags at the MSBs. One bit is used for both the IS_RIGHT
1380 * and FOLD flags. The other is reserved.
1381 *
1382 * 2^14 preds is way more than enough.
1383 */
1384#define MAX_FILTER_PRED		16384
1385
1386struct filter_pred;
1387struct regex;
1388
1389typedef int (*filter_pred_fn_t) (struct filter_pred *pred, void *event);
1390
1391typedef int (*regex_match_func)(char *str, struct regex *r, int len);
1392
1393enum regex_type {
1394	MATCH_FULL = 0,
1395	MATCH_FRONT_ONLY,
1396	MATCH_MIDDLE_ONLY,
1397	MATCH_END_ONLY,
1398	MATCH_GLOB,
1399};
1400
1401struct regex {
1402	char			pattern[MAX_FILTER_STR_VAL];
1403	int			len;
1404	int			field_len;
1405	regex_match_func	match;
1406};
1407
1408struct filter_pred {
1409	filter_pred_fn_t 	fn;
1410	u64 			val;
1411	struct regex		regex;
1412	unsigned short		*ops;
1413	struct ftrace_event_field *field;
1414	int 			offset;
1415	int			not;
1416	int 			op;
 
 
 
 
1417};
1418
1419static inline bool is_string_field(struct ftrace_event_field *field)
1420{
1421	return field->filter_type == FILTER_DYN_STRING ||
1422	       field->filter_type == FILTER_STATIC_STRING ||
1423	       field->filter_type == FILTER_PTR_STRING ||
1424	       field->filter_type == FILTER_COMM;
1425}
1426
1427static inline bool is_function_field(struct ftrace_event_field *field)
1428{
1429	return field->filter_type == FILTER_TRACE_FN;
1430}
1431
1432extern enum regex_type
1433filter_parse_regex(char *buff, int len, char **search, int *not);
1434extern void print_event_filter(struct trace_event_file *file,
1435			       struct trace_seq *s);
1436extern int apply_event_filter(struct trace_event_file *file,
1437			      char *filter_string);
1438extern int apply_subsystem_event_filter(struct trace_subsystem_dir *dir,
1439					char *filter_string);
1440extern void print_subsystem_event_filter(struct event_subsystem *system,
1441					 struct trace_seq *s);
1442extern int filter_assign_type(const char *type);
1443extern int create_event_filter(struct trace_event_call *call,
1444			       char *filter_str, bool set_str,
1445			       struct event_filter **filterp);
1446extern void free_event_filter(struct event_filter *filter);
1447
1448struct ftrace_event_field *
1449trace_find_event_field(struct trace_event_call *call, char *name);
1450
1451extern void trace_event_enable_cmd_record(bool enable);
1452extern void trace_event_enable_tgid_record(bool enable);
1453
1454extern int event_trace_add_tracer(struct dentry *parent, struct trace_array *tr);
1455extern int event_trace_del_tracer(struct trace_array *tr);
1456
1457extern struct trace_event_file *find_event_file(struct trace_array *tr,
1458						const char *system,
1459						const char *event);
1460
1461static inline void *event_file_data(struct file *filp)
1462{
1463	return READ_ONCE(file_inode(filp)->i_private);
1464}
1465
1466extern struct mutex event_mutex;
1467extern struct list_head ftrace_events;
1468
1469extern const struct file_operations event_trigger_fops;
1470extern const struct file_operations event_hist_fops;
1471
1472#ifdef CONFIG_HIST_TRIGGERS
1473extern int register_trigger_hist_cmd(void);
1474extern int register_trigger_hist_enable_disable_cmds(void);
1475#else
1476static inline int register_trigger_hist_cmd(void) { return 0; }
1477static inline int register_trigger_hist_enable_disable_cmds(void) { return 0; }
1478#endif
1479
1480extern int register_trigger_cmds(void);
1481extern void clear_event_triggers(struct trace_array *tr);
1482
1483struct event_trigger_data {
1484	unsigned long			count;
1485	int				ref;
1486	struct event_trigger_ops	*ops;
1487	struct event_command		*cmd_ops;
1488	struct event_filter __rcu	*filter;
1489	char				*filter_str;
1490	void				*private_data;
1491	bool				paused;
1492	bool				paused_tmp;
1493	struct list_head		list;
1494	char				*name;
1495	struct list_head		named_list;
1496	struct event_trigger_data	*named_data;
1497};
1498
1499/* Avoid typos */
1500#define ENABLE_EVENT_STR	"enable_event"
1501#define DISABLE_EVENT_STR	"disable_event"
1502#define ENABLE_HIST_STR		"enable_hist"
1503#define DISABLE_HIST_STR	"disable_hist"
1504
1505struct enable_trigger_data {
1506	struct trace_event_file		*file;
1507	bool				enable;
1508	bool				hist;
1509};
1510
1511extern int event_enable_trigger_print(struct seq_file *m,
1512				      struct event_trigger_ops *ops,
1513				      struct event_trigger_data *data);
1514extern void event_enable_trigger_free(struct event_trigger_ops *ops,
1515				      struct event_trigger_data *data);
1516extern int event_enable_trigger_func(struct event_command *cmd_ops,
1517				     struct trace_event_file *file,
1518				     char *glob, char *cmd, char *param);
1519extern int event_enable_register_trigger(char *glob,
1520					 struct event_trigger_ops *ops,
1521					 struct event_trigger_data *data,
1522					 struct trace_event_file *file);
1523extern void event_enable_unregister_trigger(char *glob,
1524					    struct event_trigger_ops *ops,
1525					    struct event_trigger_data *test,
1526					    struct trace_event_file *file);
1527extern void trigger_data_free(struct event_trigger_data *data);
1528extern int event_trigger_init(struct event_trigger_ops *ops,
1529			      struct event_trigger_data *data);
1530extern int trace_event_trigger_enable_disable(struct trace_event_file *file,
1531					      int trigger_enable);
1532extern void update_cond_flag(struct trace_event_file *file);
1533extern void unregister_trigger(char *glob, struct event_trigger_ops *ops,
1534			       struct event_trigger_data *test,
1535			       struct trace_event_file *file);
1536extern int set_trigger_filter(char *filter_str,
1537			      struct event_trigger_data *trigger_data,
1538			      struct trace_event_file *file);
1539extern struct event_trigger_data *find_named_trigger(const char *name);
1540extern bool is_named_trigger(struct event_trigger_data *test);
1541extern int save_named_trigger(const char *name,
1542			      struct event_trigger_data *data);
1543extern void del_named_trigger(struct event_trigger_data *data);
1544extern void pause_named_trigger(struct event_trigger_data *data);
1545extern void unpause_named_trigger(struct event_trigger_data *data);
1546extern void set_named_trigger_data(struct event_trigger_data *data,
1547				   struct event_trigger_data *named_data);
1548extern struct event_trigger_data *
1549get_named_trigger_data(struct event_trigger_data *data);
1550extern int register_event_command(struct event_command *cmd);
1551extern int unregister_event_command(struct event_command *cmd);
1552extern int register_trigger_hist_enable_disable_cmds(void);
1553
1554/**
1555 * struct event_trigger_ops - callbacks for trace event triggers
1556 *
1557 * The methods in this structure provide per-event trigger hooks for
1558 * various trigger operations.
1559 *
1560 * All the methods below, except for @init() and @free(), must be
1561 * implemented.
1562 *
1563 * @func: The trigger 'probe' function called when the triggering
1564 *	event occurs.  The data passed into this callback is the data
1565 *	that was supplied to the event_command @reg() function that
1566 *	registered the trigger (see struct event_command) along with
1567 *	the trace record, rec.
1568 *
1569 * @init: An optional initialization function called for the trigger
1570 *	when the trigger is registered (via the event_command reg()
1571 *	function).  This can be used to perform per-trigger
1572 *	initialization such as incrementing a per-trigger reference
1573 *	count, for instance.  This is usually implemented by the
1574 *	generic utility function @event_trigger_init() (see
1575 *	trace_event_triggers.c).
1576 *
1577 * @free: An optional de-initialization function called for the
1578 *	trigger when the trigger is unregistered (via the
1579 *	event_command @reg() function).  This can be used to perform
1580 *	per-trigger de-initialization such as decrementing a
1581 *	per-trigger reference count and freeing corresponding trigger
1582 *	data, for instance.  This is usually implemented by the
1583 *	generic utility function @event_trigger_free() (see
1584 *	trace_event_triggers.c).
1585 *
1586 * @print: The callback function invoked to have the trigger print
1587 *	itself.  This is usually implemented by a wrapper function
1588 *	that calls the generic utility function @event_trigger_print()
1589 *	(see trace_event_triggers.c).
1590 */
1591struct event_trigger_ops {
1592	void			(*func)(struct event_trigger_data *data,
1593					void *rec,
1594					struct ring_buffer_event *rbe);
1595	int			(*init)(struct event_trigger_ops *ops,
1596					struct event_trigger_data *data);
1597	void			(*free)(struct event_trigger_ops *ops,
1598					struct event_trigger_data *data);
1599	int			(*print)(struct seq_file *m,
1600					 struct event_trigger_ops *ops,
1601					 struct event_trigger_data *data);
1602};
1603
1604/**
1605 * struct event_command - callbacks and data members for event commands
1606 *
1607 * Event commands are invoked by users by writing the command name
1608 * into the 'trigger' file associated with a trace event.  The
1609 * parameters associated with a specific invocation of an event
1610 * command are used to create an event trigger instance, which is
1611 * added to the list of trigger instances associated with that trace
1612 * event.  When the event is hit, the set of triggers associated with
1613 * that event is invoked.
1614 *
1615 * The data members in this structure provide per-event command data
1616 * for various event commands.
1617 *
1618 * All the data members below, except for @post_trigger, must be set
1619 * for each event command.
1620 *
1621 * @name: The unique name that identifies the event command.  This is
1622 *	the name used when setting triggers via trigger files.
1623 *
1624 * @trigger_type: A unique id that identifies the event command
1625 *	'type'.  This value has two purposes, the first to ensure that
1626 *	only one trigger of the same type can be set at a given time
1627 *	for a particular event e.g. it doesn't make sense to have both
1628 *	a traceon and traceoff trigger attached to a single event at
1629 *	the same time, so traceon and traceoff have the same type
1630 *	though they have different names.  The @trigger_type value is
1631 *	also used as a bit value for deferring the actual trigger
1632 *	action until after the current event is finished.  Some
1633 *	commands need to do this if they themselves log to the trace
1634 *	buffer (see the @post_trigger() member below).  @trigger_type
1635 *	values are defined by adding new values to the trigger_type
1636 *	enum in include/linux/trace_events.h.
1637 *
1638 * @flags: See the enum event_command_flags below.
1639 *
1640 * All the methods below, except for @set_filter() and @unreg_all(),
1641 * must be implemented.
1642 *
1643 * @func: The callback function responsible for parsing and
1644 *	registering the trigger written to the 'trigger' file by the
1645 *	user.  It allocates the trigger instance and registers it with
1646 *	the appropriate trace event.  It makes use of the other
1647 *	event_command callback functions to orchestrate this, and is
1648 *	usually implemented by the generic utility function
1649 *	@event_trigger_callback() (see trace_event_triggers.c).
1650 *
1651 * @reg: Adds the trigger to the list of triggers associated with the
1652 *	event, and enables the event trigger itself, after
1653 *	initializing it (via the event_trigger_ops @init() function).
1654 *	This is also where commands can use the @trigger_type value to
1655 *	make the decision as to whether or not multiple instances of
1656 *	the trigger should be allowed.  This is usually implemented by
1657 *	the generic utility function @register_trigger() (see
1658 *	trace_event_triggers.c).
1659 *
1660 * @unreg: Removes the trigger from the list of triggers associated
1661 *	with the event, and disables the event trigger itself, after
1662 *	initializing it (via the event_trigger_ops @free() function).
1663 *	This is usually implemented by the generic utility function
1664 *	@unregister_trigger() (see trace_event_triggers.c).
1665 *
1666 * @unreg_all: An optional function called to remove all the triggers
1667 *	from the list of triggers associated with the event.  Called
1668 *	when a trigger file is opened in truncate mode.
1669 *
1670 * @set_filter: An optional function called to parse and set a filter
1671 *	for the trigger.  If no @set_filter() method is set for the
1672 *	event command, filters set by the user for the command will be
1673 *	ignored.  This is usually implemented by the generic utility
1674 *	function @set_trigger_filter() (see trace_event_triggers.c).
1675 *
1676 * @get_trigger_ops: The callback function invoked to retrieve the
1677 *	event_trigger_ops implementation associated with the command.
1678 */
1679struct event_command {
1680	struct list_head	list;
1681	char			*name;
1682	enum event_trigger_type	trigger_type;
1683	int			flags;
1684	int			(*func)(struct event_command *cmd_ops,
1685					struct trace_event_file *file,
1686					char *glob, char *cmd, char *params);
1687	int			(*reg)(char *glob,
1688				       struct event_trigger_ops *ops,
1689				       struct event_trigger_data *data,
1690				       struct trace_event_file *file);
1691	void			(*unreg)(char *glob,
1692					 struct event_trigger_ops *ops,
1693					 struct event_trigger_data *data,
1694					 struct trace_event_file *file);
1695	void			(*unreg_all)(struct trace_event_file *file);
1696	int			(*set_filter)(char *filter_str,
1697					      struct event_trigger_data *data,
1698					      struct trace_event_file *file);
1699	struct event_trigger_ops *(*get_trigger_ops)(char *cmd, char *param);
1700};
1701
1702/**
1703 * enum event_command_flags - flags for struct event_command
1704 *
1705 * @POST_TRIGGER: A flag that says whether or not this command needs
1706 *	to have its action delayed until after the current event has
1707 *	been closed.  Some triggers need to avoid being invoked while
1708 *	an event is currently in the process of being logged, since
1709 *	the trigger may itself log data into the trace buffer.  Thus
1710 *	we make sure the current event is committed before invoking
1711 *	those triggers.  To do that, the trigger invocation is split
1712 *	in two - the first part checks the filter using the current
1713 *	trace record; if a command has the @post_trigger flag set, it
1714 *	sets a bit for itself in the return value, otherwise it
1715 *	directly invokes the trigger.  Once all commands have been
1716 *	either invoked or set their return flag, the current record is
1717 *	either committed or discarded.  At that point, if any commands
1718 *	have deferred their triggers, those commands are finally
1719 *	invoked following the close of the current event.  In other
1720 *	words, if the event_trigger_ops @func() probe implementation
1721 *	itself logs to the trace buffer, this flag should be set,
1722 *	otherwise it can be left unspecified.
1723 *
1724 * @NEEDS_REC: A flag that says whether or not this command needs
1725 *	access to the trace record in order to perform its function,
1726 *	regardless of whether or not it has a filter associated with
1727 *	it (filters make a trigger require access to the trace record
1728 *	but are not always present).
1729 */
1730enum event_command_flags {
1731	EVENT_CMD_FL_POST_TRIGGER	= 1,
1732	EVENT_CMD_FL_NEEDS_REC		= 2,
1733};
1734
1735static inline bool event_command_post_trigger(struct event_command *cmd_ops)
1736{
1737	return cmd_ops->flags & EVENT_CMD_FL_POST_TRIGGER;
1738}
1739
1740static inline bool event_command_needs_rec(struct event_command *cmd_ops)
1741{
1742	return cmd_ops->flags & EVENT_CMD_FL_NEEDS_REC;
1743}
1744
1745extern int trace_event_enable_disable(struct trace_event_file *file,
1746				      int enable, int soft_disable);
1747extern int tracing_alloc_snapshot(void);
1748
1749extern const char *__start___trace_bprintk_fmt[];
1750extern const char *__stop___trace_bprintk_fmt[];
1751
1752extern const char *__start___tracepoint_str[];
1753extern const char *__stop___tracepoint_str[];
1754
1755void trace_printk_control(bool enabled);
1756void trace_printk_init_buffers(void);
1757void trace_printk_start_comm(void);
1758int trace_keep_overwrite(struct tracer *tracer, u32 mask, int set);
1759int set_tracer_flag(struct trace_array *tr, unsigned int mask, int enabled);
1760
1761#define MAX_EVENT_NAME_LEN	64
1762
1763extern int trace_run_command(const char *buf, int (*createfn)(int, char**));
1764extern ssize_t trace_parse_run_command(struct file *file,
1765		const char __user *buffer, size_t count, loff_t *ppos,
1766		int (*createfn)(int, char**));
1767
1768/*
1769 * Normal trace_printk() and friends allocates special buffers
1770 * to do the manipulation, as well as saves the print formats
1771 * into sections to display. But the trace infrastructure wants
1772 * to use these without the added overhead at the price of being
1773 * a bit slower (used mainly for warnings, where we don't care
1774 * about performance). The internal_trace_puts() is for such
1775 * a purpose.
1776 */
1777#define internal_trace_puts(str) __trace_puts(_THIS_IP_, str, strlen(str))
1778
1779#undef FTRACE_ENTRY
1780#define FTRACE_ENTRY(call, struct_name, id, tstruct, print, filter)	\
1781	extern struct trace_event_call					\
1782	__aligned(4) event_##call;
1783#undef FTRACE_ENTRY_DUP
1784#define FTRACE_ENTRY_DUP(call, struct_name, id, tstruct, print, filter)	\
1785	FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print), \
1786		     filter)
1787#undef FTRACE_ENTRY_PACKED
1788#define FTRACE_ENTRY_PACKED(call, struct_name, id, tstruct, print, filter) \
1789	FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print), \
1790		     filter)
1791
1792#include "trace_entries.h"
1793
1794#if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_FUNCTION_TRACER)
1795int perf_ftrace_event_register(struct trace_event_call *call,
1796			       enum trace_reg type, void *data);
1797#else
1798#define perf_ftrace_event_register NULL
1799#endif
1800
1801#ifdef CONFIG_FTRACE_SYSCALLS
1802void init_ftrace_syscalls(void);
1803const char *get_syscall_name(int syscall);
1804#else
1805static inline void init_ftrace_syscalls(void) { }
1806static inline const char *get_syscall_name(int syscall)
1807{
1808	return NULL;
1809}
1810#endif
1811
1812#ifdef CONFIG_EVENT_TRACING
1813void trace_event_init(void);
1814void trace_event_eval_update(struct trace_eval_map **map, int len);
1815#else
1816static inline void __init trace_event_init(void) { }
1817static inline void trace_event_eval_update(struct trace_eval_map **map, int len) { }
1818#endif
1819
1820#ifdef CONFIG_TRACER_SNAPSHOT
1821void tracing_snapshot_instance(struct trace_array *tr);
1822int tracing_alloc_snapshot_instance(struct trace_array *tr);
1823#else
1824static inline void tracing_snapshot_instance(struct trace_array *tr) { }
1825static inline int tracing_alloc_snapshot_instance(struct trace_array *tr)
1826{
1827	return 0;
1828}
1829#endif
1830
1831extern struct trace_iterator *tracepoint_print_iter;
1832
1833#endif /* _LINUX_KERNEL_TRACE_H */