linux/kernel/trace/trace.h
<<
>>
Prefs
   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        struct trace_event_file *trace_marker_file;
 263        cpumask_var_t           tracing_cpumask; /* only trace on set CPUs */
 264        int                     ref;
 265#ifdef CONFIG_FUNCTION_TRACER
 266        struct ftrace_ops       *ops;
 267        struct trace_pid_list   __rcu *function_pids;
 268#ifdef CONFIG_DYNAMIC_FTRACE
 269        /* All of these are protected by the ftrace_lock */
 270        struct list_head        func_probes;
 271        struct list_head        mod_trace;
 272        struct list_head        mod_notrace;
 273#endif
 274        /* function tracing enabled */
 275        int                     function_enabled;
 276#endif
 277        int                     time_stamp_abs_ref;
 278        struct list_head        hist_vars;
 279};
 280
 281enum {
 282        TRACE_ARRAY_FL_GLOBAL   = (1 << 0)
 283};
 284
 285extern struct list_head ftrace_trace_arrays;
 286
 287extern struct mutex trace_types_lock;
 288
 289extern int trace_array_get(struct trace_array *tr);
 290extern void trace_array_put(struct trace_array *tr);
 291
 292extern int tracing_set_time_stamp_abs(struct trace_array *tr, bool abs);
 293extern int tracing_set_clock(struct trace_array *tr, const char *clockstr);
 294
 295extern bool trace_clock_in_ns(struct trace_array *tr);
 296
 297/*
 298 * The global tracer (top) should be the first trace array added,
 299 * but we check the flag anyway.
 300 */
 301static inline struct trace_array *top_trace_array(void)
 302{
 303        struct trace_array *tr;
 304
 305        if (list_empty(&ftrace_trace_arrays))
 306                return NULL;
 307
 308        tr = list_entry(ftrace_trace_arrays.prev,
 309                        typeof(*tr), list);
 310        WARN_ON(!(tr->flags & TRACE_ARRAY_FL_GLOBAL));
 311        return tr;
 312}
 313
 314#define FTRACE_CMP_TYPE(var, type) \
 315        __builtin_types_compatible_p(typeof(var), type *)
 316
 317#undef IF_ASSIGN
 318#define IF_ASSIGN(var, entry, etype, id)                \
 319        if (FTRACE_CMP_TYPE(var, etype)) {              \
 320                var = (typeof(var))(entry);             \
 321                WARN_ON(id && (entry)->type != id);     \
 322                break;                                  \
 323        }
 324
 325/* Will cause compile errors if type is not found. */
 326extern void __ftrace_bad_type(void);
 327
 328/*
 329 * The trace_assign_type is a verifier that the entry type is
 330 * the same as the type being assigned. To add new types simply
 331 * add a line with the following format:
 332 *
 333 * IF_ASSIGN(var, ent, type, id);
 334 *
 335 *  Where "type" is the trace type that includes the trace_entry
 336 *  as the "ent" item. And "id" is the trace identifier that is
 337 *  used in the trace_type enum.
 338 *
 339 *  If the type can have more than one id, then use zero.
 340 */
 341#define trace_assign_type(var, ent)                                     \
 342        do {                                                            \
 343                IF_ASSIGN(var, ent, struct ftrace_entry, TRACE_FN);     \
 344                IF_ASSIGN(var, ent, struct ctx_switch_entry, 0);        \
 345                IF_ASSIGN(var, ent, struct stack_entry, TRACE_STACK);   \
 346                IF_ASSIGN(var, ent, struct userstack_entry, TRACE_USER_STACK);\
 347                IF_ASSIGN(var, ent, struct print_entry, TRACE_PRINT);   \
 348                IF_ASSIGN(var, ent, struct bprint_entry, TRACE_BPRINT); \
 349                IF_ASSIGN(var, ent, struct bputs_entry, TRACE_BPUTS);   \
 350                IF_ASSIGN(var, ent, struct hwlat_entry, TRACE_HWLAT);   \
 351                IF_ASSIGN(var, ent, struct raw_data_entry, TRACE_RAW_DATA);\
 352                IF_ASSIGN(var, ent, struct trace_mmiotrace_rw,          \
 353                          TRACE_MMIO_RW);                               \
 354                IF_ASSIGN(var, ent, struct trace_mmiotrace_map,         \
 355                          TRACE_MMIO_MAP);                              \
 356                IF_ASSIGN(var, ent, struct trace_branch, TRACE_BRANCH); \
 357                IF_ASSIGN(var, ent, struct ftrace_graph_ent_entry,      \
 358                          TRACE_GRAPH_ENT);             \
 359                IF_ASSIGN(var, ent, struct ftrace_graph_ret_entry,      \
 360                          TRACE_GRAPH_RET);             \
 361                __ftrace_bad_type();                                    \
 362        } while (0)
 363
 364/*
 365 * An option specific to a tracer. This is a boolean value.
 366 * The bit is the bit index that sets its value on the
 367 * flags value in struct tracer_flags.
 368 */
 369struct tracer_opt {
 370        const char      *name; /* Will appear on the trace_options file */
 371        u32             bit; /* Mask assigned in val field in tracer_flags */
 372};
 373
 374/*
 375 * The set of specific options for a tracer. Your tracer
 376 * have to set the initial value of the flags val.
 377 */
 378struct tracer_flags {
 379        u32                     val;
 380        struct tracer_opt       *opts;
 381        struct tracer           *trace;
 382};
 383
 384/* Makes more easy to define a tracer opt */
 385#define TRACER_OPT(s, b)        .name = #s, .bit = b
 386
 387
 388struct trace_option_dentry {
 389        struct tracer_opt               *opt;
 390        struct tracer_flags             *flags;
 391        struct trace_array              *tr;
 392        struct dentry                   *entry;
 393};
 394
 395/**
 396 * struct tracer - a specific tracer and its callbacks to interact with tracefs
 397 * @name: the name chosen to select it on the available_tracers file
 398 * @init: called when one switches to this tracer (echo name > current_tracer)
 399 * @reset: called when one switches to another tracer
 400 * @start: called when tracing is unpaused (echo 1 > tracing_on)
 401 * @stop: called when tracing is paused (echo 0 > tracing_on)
 402 * @update_thresh: called when tracing_thresh is updated
 403 * @open: called when the trace file is opened
 404 * @pipe_open: called when the trace_pipe file is opened
 405 * @close: called when the trace file is released
 406 * @pipe_close: called when the trace_pipe file is released
 407 * @read: override the default read callback on trace_pipe
 408 * @splice_read: override the default splice_read callback on trace_pipe
 409 * @selftest: selftest to run on boot (see trace_selftest.c)
 410 * @print_headers: override the first lines that describe your columns
 411 * @print_line: callback that prints a trace
 412 * @set_flag: signals one of your private flags changed (trace_options file)
 413 * @flags: your private flags
 414 */
 415struct tracer {
 416        const char              *name;
 417        int                     (*init)(struct trace_array *tr);
 418        void                    (*reset)(struct trace_array *tr);
 419        void                    (*start)(struct trace_array *tr);
 420        void                    (*stop)(struct trace_array *tr);
 421        int                     (*update_thresh)(struct trace_array *tr);
 422        void                    (*open)(struct trace_iterator *iter);
 423        void                    (*pipe_open)(struct trace_iterator *iter);
 424        void                    (*close)(struct trace_iterator *iter);
 425        void                    (*pipe_close)(struct trace_iterator *iter);
 426        ssize_t                 (*read)(struct trace_iterator *iter,
 427                                        struct file *filp, char __user *ubuf,
 428                                        size_t cnt, loff_t *ppos);
 429        ssize_t                 (*splice_read)(struct trace_iterator *iter,
 430                                               struct file *filp,
 431                                               loff_t *ppos,
 432                                               struct pipe_inode_info *pipe,
 433                                               size_t len,
 434                                               unsigned int flags);
 435#ifdef CONFIG_FTRACE_STARTUP_TEST
 436        int                     (*selftest)(struct tracer *trace,
 437                                            struct trace_array *tr);
 438#endif
 439        void                    (*print_header)(struct seq_file *m);
 440        enum print_line_t       (*print_line)(struct trace_iterator *iter);
 441        /* If you handled the flag setting, return 0 */
 442        int                     (*set_flag)(struct trace_array *tr,
 443                                            u32 old_flags, u32 bit, int set);
 444        /* Return 0 if OK with change, else return non-zero */
 445        int                     (*flag_changed)(struct trace_array *tr,
 446                                                u32 mask, int set);
 447        struct tracer           *next;
 448        struct tracer_flags     *flags;
 449        int                     enabled;
 450        int                     ref;
 451        bool                    print_max;
 452        bool                    allow_instances;
 453#ifdef CONFIG_TRACER_MAX_TRACE
 454        bool                    use_max_tr;
 455#endif
 456        /* True if tracer cannot be enabled in kernel param */
 457        bool                    noboot;
 458};
 459
 460
 461/* Only current can touch trace_recursion */
 462
 463/*
 464 * For function tracing recursion:
 465 *  The order of these bits are important.
 466 *
 467 *  When function tracing occurs, the following steps are made:
 468 *   If arch does not support a ftrace feature:
 469 *    call internal function (uses INTERNAL bits) which calls...
 470 *   If callback is registered to the "global" list, the list
 471 *    function is called and recursion checks the GLOBAL bits.
 472 *    then this function calls...
 473 *   The function callback, which can use the FTRACE bits to
 474 *    check for recursion.
 475 *
 476 * Now if the arch does not suppport a feature, and it calls
 477 * the global list function which calls the ftrace callback
 478 * all three of these steps will do a recursion protection.
 479 * There's no reason to do one if the previous caller already
 480 * did. The recursion that we are protecting against will
 481 * go through the same steps again.
 482 *
 483 * To prevent the multiple recursion checks, if a recursion
 484 * bit is set that is higher than the MAX bit of the current
 485 * check, then we know that the check was made by the previous
 486 * caller, and we can skip the current check.
 487 */
 488enum {
 489        TRACE_BUFFER_BIT,
 490        TRACE_BUFFER_NMI_BIT,
 491        TRACE_BUFFER_IRQ_BIT,
 492        TRACE_BUFFER_SIRQ_BIT,
 493
 494        /* Start of function recursion bits */
 495        TRACE_FTRACE_BIT,
 496        TRACE_FTRACE_NMI_BIT,
 497        TRACE_FTRACE_IRQ_BIT,
 498        TRACE_FTRACE_SIRQ_BIT,
 499
 500        /* INTERNAL_BITs must be greater than FTRACE_BITs */
 501        TRACE_INTERNAL_BIT,
 502        TRACE_INTERNAL_NMI_BIT,
 503        TRACE_INTERNAL_IRQ_BIT,
 504        TRACE_INTERNAL_SIRQ_BIT,
 505
 506        TRACE_BRANCH_BIT,
 507/*
 508 * Abuse of the trace_recursion.
 509 * As we need a way to maintain state if we are tracing the function
 510 * graph in irq because we want to trace a particular function that
 511 * was called in irq context but we have irq tracing off. Since this
 512 * can only be modified by current, we can reuse trace_recursion.
 513 */
 514        TRACE_IRQ_BIT,
 515};
 516
 517#define trace_recursion_set(bit)        do { (current)->trace_recursion |= (1<<(bit)); } while (0)
 518#define trace_recursion_clear(bit)      do { (current)->trace_recursion &= ~(1<<(bit)); } while (0)
 519#define trace_recursion_test(bit)       ((current)->trace_recursion & (1<<(bit)))
 520
 521#define TRACE_CONTEXT_BITS      4
 522
 523#define TRACE_FTRACE_START      TRACE_FTRACE_BIT
 524#define TRACE_FTRACE_MAX        ((1 << (TRACE_FTRACE_START + TRACE_CONTEXT_BITS)) - 1)
 525
 526#define TRACE_LIST_START        TRACE_INTERNAL_BIT
 527#define TRACE_LIST_MAX          ((1 << (TRACE_LIST_START + TRACE_CONTEXT_BITS)) - 1)
 528
 529#define TRACE_CONTEXT_MASK      TRACE_LIST_MAX
 530
 531static __always_inline int trace_get_context_bit(void)
 532{
 533        int bit;
 534
 535        if (in_interrupt()) {
 536                if (in_nmi())
 537                        bit = 0;
 538
 539                else if (in_irq())
 540                        bit = 1;
 541                else
 542                        bit = 2;
 543        } else
 544                bit = 3;
 545
 546        return bit;
 547}
 548
 549static __always_inline int trace_test_and_set_recursion(int start, int max)
 550{
 551        unsigned int val = current->trace_recursion;
 552        int bit;
 553
 554        /* A previous recursion check was made */
 555        if ((val & TRACE_CONTEXT_MASK) > max)
 556                return 0;
 557
 558        bit = trace_get_context_bit() + start;
 559        if (unlikely(val & (1 << bit)))
 560                return -1;
 561
 562        val |= 1 << bit;
 563        current->trace_recursion = val;
 564        barrier();
 565
 566        return bit;
 567}
 568
 569static __always_inline void trace_clear_recursion(int bit)
 570{
 571        unsigned int val = current->trace_recursion;
 572
 573        if (!bit)
 574                return;
 575
 576        bit = 1 << bit;
 577        val &= ~bit;
 578
 579        barrier();
 580        current->trace_recursion = val;
 581}
 582
 583static inline struct ring_buffer_iter *
 584trace_buffer_iter(struct trace_iterator *iter, int cpu)
 585{
 586        return iter->buffer_iter ? iter->buffer_iter[cpu] : NULL;
 587}
 588
 589int tracer_init(struct tracer *t, struct trace_array *tr);
 590int tracing_is_enabled(void);
 591void tracing_reset(struct trace_buffer *buf, int cpu);
 592void tracing_reset_online_cpus(struct trace_buffer *buf);
 593void tracing_reset_current(int cpu);
 594void tracing_reset_all_online_cpus(void);
 595int tracing_open_generic(struct inode *inode, struct file *filp);
 596bool tracing_is_disabled(void);
 597int tracer_tracing_is_on(struct trace_array *tr);
 598void tracer_tracing_on(struct trace_array *tr);
 599void tracer_tracing_off(struct trace_array *tr);
 600struct dentry *trace_create_file(const char *name,
 601                                 umode_t mode,
 602                                 struct dentry *parent,
 603                                 void *data,
 604                                 const struct file_operations *fops);
 605
 606struct dentry *tracing_init_dentry(void);
 607
 608struct ring_buffer_event;
 609
 610struct ring_buffer_event *
 611trace_buffer_lock_reserve(struct ring_buffer *buffer,
 612                          int type,
 613                          unsigned long len,
 614                          unsigned long flags,
 615                          int pc);
 616
 617struct trace_entry *tracing_get_trace_entry(struct trace_array *tr,
 618                                                struct trace_array_cpu *data);
 619
 620struct trace_entry *trace_find_next_entry(struct trace_iterator *iter,
 621                                          int *ent_cpu, u64 *ent_ts);
 622
 623void trace_buffer_unlock_commit_nostack(struct ring_buffer *buffer,
 624                                        struct ring_buffer_event *event);
 625
 626int trace_empty(struct trace_iterator *iter);
 627
 628void *trace_find_next_entry_inc(struct trace_iterator *iter);
 629
 630void trace_init_global_iter(struct trace_iterator *iter);
 631
 632void tracing_iter_reset(struct trace_iterator *iter, int cpu);
 633
 634void trace_function(struct trace_array *tr,
 635                    unsigned long ip,
 636                    unsigned long parent_ip,
 637                    unsigned long flags, int pc);
 638void trace_graph_function(struct trace_array *tr,
 639                    unsigned long ip,
 640                    unsigned long parent_ip,
 641                    unsigned long flags, int pc);
 642void trace_latency_header(struct seq_file *m);
 643void trace_default_header(struct seq_file *m);
 644void print_trace_header(struct seq_file *m, struct trace_iterator *iter);
 645int trace_empty(struct trace_iterator *iter);
 646
 647void trace_graph_return(struct ftrace_graph_ret *trace);
 648int trace_graph_entry(struct ftrace_graph_ent *trace);
 649void set_graph_array(struct trace_array *tr);
 650
 651void tracing_start_cmdline_record(void);
 652void tracing_stop_cmdline_record(void);
 653void tracing_start_tgid_record(void);
 654void tracing_stop_tgid_record(void);
 655
 656int register_tracer(struct tracer *type);
 657int is_tracing_stopped(void);
 658
 659loff_t tracing_lseek(struct file *file, loff_t offset, int whence);
 660
 661extern cpumask_var_t __read_mostly tracing_buffer_mask;
 662
 663#define for_each_tracing_cpu(cpu)       \
 664        for_each_cpu(cpu, tracing_buffer_mask)
 665
 666extern unsigned long nsecs_to_usecs(unsigned long nsecs);
 667
 668extern unsigned long tracing_thresh;
 669
 670/* PID filtering */
 671
 672extern int pid_max;
 673
 674bool trace_find_filtered_pid(struct trace_pid_list *filtered_pids,
 675                             pid_t search_pid);
 676bool trace_ignore_this_task(struct trace_pid_list *filtered_pids,
 677                            struct task_struct *task);
 678void trace_filter_add_remove_task(struct trace_pid_list *pid_list,
 679                                  struct task_struct *self,
 680                                  struct task_struct *task);
 681void *trace_pid_next(struct trace_pid_list *pid_list, void *v, loff_t *pos);
 682void *trace_pid_start(struct trace_pid_list *pid_list, loff_t *pos);
 683int trace_pid_show(struct seq_file *m, void *v);
 684void trace_free_pid_list(struct trace_pid_list *pid_list);
 685int trace_pid_write(struct trace_pid_list *filtered_pids,
 686                    struct trace_pid_list **new_pid_list,
 687                    const char __user *ubuf, size_t cnt);
 688
 689#ifdef CONFIG_TRACER_MAX_TRACE
 690void update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu);
 691void update_max_tr_single(struct trace_array *tr,
 692                          struct task_struct *tsk, int cpu);
 693#endif /* CONFIG_TRACER_MAX_TRACE */
 694
 695#ifdef CONFIG_STACKTRACE
 696void ftrace_trace_userstack(struct ring_buffer *buffer, unsigned long flags,
 697                            int pc);
 698
 699void __trace_stack(struct trace_array *tr, unsigned long flags, int skip,
 700                   int pc);
 701#else
 702static inline void ftrace_trace_userstack(struct ring_buffer *buffer,
 703                                          unsigned long flags, int pc)
 704{
 705}
 706
 707static inline void __trace_stack(struct trace_array *tr, unsigned long flags,
 708                                 int skip, int pc)
 709{
 710}
 711#endif /* CONFIG_STACKTRACE */
 712
 713extern u64 ftrace_now(int cpu);
 714
 715extern void trace_find_cmdline(int pid, char comm[]);
 716extern int trace_find_tgid(int pid);
 717extern void trace_event_follow_fork(struct trace_array *tr, bool enable);
 718
 719#ifdef CONFIG_DYNAMIC_FTRACE
 720extern unsigned long ftrace_update_tot_cnt;
 721void ftrace_init_trace_array(struct trace_array *tr);
 722#else
 723static inline void ftrace_init_trace_array(struct trace_array *tr) { }
 724#endif
 725#define DYN_FTRACE_TEST_NAME trace_selftest_dynamic_test_func
 726extern int DYN_FTRACE_TEST_NAME(void);
 727#define DYN_FTRACE_TEST_NAME2 trace_selftest_dynamic_test_func2
 728extern int DYN_FTRACE_TEST_NAME2(void);
 729
 730extern bool ring_buffer_expanded;
 731extern bool tracing_selftest_disabled;
 732
 733#ifdef CONFIG_FTRACE_STARTUP_TEST
 734extern int trace_selftest_startup_function(struct tracer *trace,
 735                                           struct trace_array *tr);
 736extern int trace_selftest_startup_function_graph(struct tracer *trace,
 737                                                 struct trace_array *tr);
 738extern int trace_selftest_startup_irqsoff(struct tracer *trace,
 739                                          struct trace_array *tr);
 740extern int trace_selftest_startup_preemptoff(struct tracer *trace,
 741                                             struct trace_array *tr);
 742extern int trace_selftest_startup_preemptirqsoff(struct tracer *trace,
 743                                                 struct trace_array *tr);
 744extern int trace_selftest_startup_wakeup(struct tracer *trace,
 745                                         struct trace_array *tr);
 746extern int trace_selftest_startup_nop(struct tracer *trace,
 747                                         struct trace_array *tr);
 748extern int trace_selftest_startup_branch(struct tracer *trace,
 749                                         struct trace_array *tr);
 750/*
 751 * Tracer data references selftest functions that only occur
 752 * on boot up. These can be __init functions. Thus, when selftests
 753 * are enabled, then the tracers need to reference __init functions.
 754 */
 755#define __tracer_data           __refdata
 756#else
 757/* Tracers are seldom changed. Optimize when selftests are disabled. */
 758#define __tracer_data           __read_mostly
 759#endif /* CONFIG_FTRACE_STARTUP_TEST */
 760
 761extern void *head_page(struct trace_array_cpu *data);
 762extern unsigned long long ns2usecs(u64 nsec);
 763extern int
 764trace_vbprintk(unsigned long ip, const char *fmt, va_list args);
 765extern int
 766trace_vprintk(unsigned long ip, const char *fmt, va_list args);
 767extern int
 768trace_array_vprintk(struct trace_array *tr,
 769                    unsigned long ip, const char *fmt, va_list args);
 770int trace_array_printk(struct trace_array *tr,
 771                       unsigned long ip, const char *fmt, ...);
 772int trace_array_printk_buf(struct ring_buffer *buffer,
 773                           unsigned long ip, const char *fmt, ...);
 774void trace_printk_seq(struct trace_seq *s);
 775enum print_line_t print_trace_line(struct trace_iterator *iter);
 776
 777extern char trace_find_mark(unsigned long long duration);
 778
 779struct ftrace_hash;
 780
 781struct ftrace_mod_load {
 782        struct list_head        list;
 783        char                    *func;
 784        char                    *module;
 785        int                      enable;
 786};
 787
 788enum {
 789        FTRACE_HASH_FL_MOD      = (1 << 0),
 790};
 791
 792struct ftrace_hash {
 793        unsigned long           size_bits;
 794        struct hlist_head       *buckets;
 795        unsigned long           count;
 796        unsigned long           flags;
 797        struct rcu_head         rcu;
 798};
 799
 800struct ftrace_func_entry *
 801ftrace_lookup_ip(struct ftrace_hash *hash, unsigned long ip);
 802
 803static __always_inline bool ftrace_hash_empty(struct ftrace_hash *hash)
 804{
 805        return !hash || !(hash->count || (hash->flags & FTRACE_HASH_FL_MOD));
 806}
 807
 808/* Standard output formatting function used for function return traces */
 809#ifdef CONFIG_FUNCTION_GRAPH_TRACER
 810
 811/* Flag options */
 812#define TRACE_GRAPH_PRINT_OVERRUN       0x1
 813#define TRACE_GRAPH_PRINT_CPU           0x2
 814#define TRACE_GRAPH_PRINT_OVERHEAD      0x4
 815#define TRACE_GRAPH_PRINT_PROC          0x8
 816#define TRACE_GRAPH_PRINT_DURATION      0x10
 817#define TRACE_GRAPH_PRINT_ABS_TIME      0x20
 818#define TRACE_GRAPH_PRINT_IRQS          0x40
 819#define TRACE_GRAPH_PRINT_TAIL          0x80
 820#define TRACE_GRAPH_SLEEP_TIME          0x100
 821#define TRACE_GRAPH_GRAPH_TIME          0x200
 822#define TRACE_GRAPH_PRINT_FILL_SHIFT    28
 823#define TRACE_GRAPH_PRINT_FILL_MASK     (0x3 << TRACE_GRAPH_PRINT_FILL_SHIFT)
 824
 825extern void ftrace_graph_sleep_time_control(bool enable);
 826extern void ftrace_graph_graph_time_control(bool enable);
 827
 828extern enum print_line_t
 829print_graph_function_flags(struct trace_iterator *iter, u32 flags);
 830extern void print_graph_headers_flags(struct seq_file *s, u32 flags);
 831extern void
 832trace_print_graph_duration(unsigned long long duration, struct trace_seq *s);
 833extern void graph_trace_open(struct trace_iterator *iter);
 834extern void graph_trace_close(struct trace_iterator *iter);
 835extern int __trace_graph_entry(struct trace_array *tr,
 836                               struct ftrace_graph_ent *trace,
 837                               unsigned long flags, int pc);
 838extern void __trace_graph_return(struct trace_array *tr,
 839                                 struct ftrace_graph_ret *trace,
 840                                 unsigned long flags, int pc);
 841
 842#ifdef CONFIG_DYNAMIC_FTRACE
 843extern struct ftrace_hash *ftrace_graph_hash;
 844extern struct ftrace_hash *ftrace_graph_notrace_hash;
 845
 846static inline int ftrace_graph_addr(unsigned long addr)
 847{
 848        int ret = 0;
 849
 850        preempt_disable_notrace();
 851
 852        if (ftrace_hash_empty(ftrace_graph_hash)) {
 853                ret = 1;
 854                goto out;
 855        }
 856
 857        if (ftrace_lookup_ip(ftrace_graph_hash, addr)) {
 858                /*
 859                 * If no irqs are to be traced, but a set_graph_function
 860                 * is set, and called by an interrupt handler, we still
 861                 * want to trace it.
 862                 */
 863                if (in_irq())
 864                        trace_recursion_set(TRACE_IRQ_BIT);
 865                else
 866                        trace_recursion_clear(TRACE_IRQ_BIT);
 867                ret = 1;
 868        }
 869
 870out:
 871        preempt_enable_notrace();
 872        return ret;
 873}
 874
 875static inline int ftrace_graph_notrace_addr(unsigned long addr)
 876{
 877        int ret = 0;
 878
 879        preempt_disable_notrace();
 880
 881        if (ftrace_lookup_ip(ftrace_graph_notrace_hash, addr))
 882                ret = 1;
 883
 884        preempt_enable_notrace();
 885        return ret;
 886}
 887#else
 888static inline int ftrace_graph_addr(unsigned long addr)
 889{
 890        return 1;
 891}
 892
 893static inline int ftrace_graph_notrace_addr(unsigned long addr)
 894{
 895        return 0;
 896}
 897#endif /* CONFIG_DYNAMIC_FTRACE */
 898
 899extern unsigned int fgraph_max_depth;
 900
 901static inline bool ftrace_graph_ignore_func(struct ftrace_graph_ent *trace)
 902{
 903        /* trace it when it is-nested-in or is a function enabled. */
 904        return !(trace->depth || ftrace_graph_addr(trace->func)) ||
 905                (trace->depth < 0) ||
 906                (fgraph_max_depth && trace->depth >= fgraph_max_depth);
 907}
 908
 909#else /* CONFIG_FUNCTION_GRAPH_TRACER */
 910static inline enum print_line_t
 911print_graph_function_flags(struct trace_iterator *iter, u32 flags)
 912{
 913        return TRACE_TYPE_UNHANDLED;
 914}
 915#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
 916
 917extern struct list_head ftrace_pids;
 918
 919#ifdef CONFIG_FUNCTION_TRACER
 920struct ftrace_func_command {
 921        struct list_head        list;
 922        char                    *name;
 923        int                     (*func)(struct trace_array *tr,
 924                                        struct ftrace_hash *hash,
 925                                        char *func, char *cmd,
 926                                        char *params, int enable);
 927};
 928extern bool ftrace_filter_param __initdata;
 929static inline int ftrace_trace_task(struct trace_array *tr)
 930{
 931        return !this_cpu_read(tr->trace_buffer.data->ftrace_ignore_pid);
 932}
 933extern int ftrace_is_dead(void);
 934int ftrace_create_function_files(struct trace_array *tr,
 935                                 struct dentry *parent);
 936void ftrace_destroy_function_files(struct trace_array *tr);
 937void ftrace_init_global_array_ops(struct trace_array *tr);
 938void ftrace_init_array_ops(struct trace_array *tr, ftrace_func_t func);
 939void ftrace_reset_array_ops(struct trace_array *tr);
 940int using_ftrace_ops_list_func(void);
 941void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d_tracer);
 942void ftrace_init_tracefs_toplevel(struct trace_array *tr,
 943                                  struct dentry *d_tracer);
 944void ftrace_clear_pids(struct trace_array *tr);
 945int init_function_trace(void);
 946void ftrace_pid_follow_fork(struct trace_array *tr, bool enable);
 947#else
 948static inline int ftrace_trace_task(struct trace_array *tr)
 949{
 950        return 1;
 951}
 952static inline int ftrace_is_dead(void) { return 0; }
 953static inline int
 954ftrace_create_function_files(struct trace_array *tr,
 955                             struct dentry *parent)
 956{
 957        return 0;
 958}
 959static inline void ftrace_destroy_function_files(struct trace_array *tr) { }
 960static inline __init void
 961ftrace_init_global_array_ops(struct trace_array *tr) { }
 962static inline void ftrace_reset_array_ops(struct trace_array *tr) { }
 963static inline void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d) { }
 964static inline void ftrace_init_tracefs_toplevel(struct trace_array *tr, struct dentry *d) { }
 965static inline void ftrace_clear_pids(struct trace_array *tr) { }
 966static inline int init_function_trace(void) { return 0; }
 967static inline void ftrace_pid_follow_fork(struct trace_array *tr, bool enable) { }
 968/* ftace_func_t type is not defined, use macro instead of static inline */
 969#define ftrace_init_array_ops(tr, func) do { } while (0)
 970#endif /* CONFIG_FUNCTION_TRACER */
 971
 972#if defined(CONFIG_FUNCTION_TRACER) && defined(CONFIG_DYNAMIC_FTRACE)
 973
 974struct ftrace_probe_ops {
 975        void                    (*func)(unsigned long ip,
 976                                        unsigned long parent_ip,
 977                                        struct trace_array *tr,
 978                                        struct ftrace_probe_ops *ops,
 979                                        void *data);
 980        int                     (*init)(struct ftrace_probe_ops *ops,
 981                                        struct trace_array *tr,
 982                                        unsigned long ip, void *init_data,
 983                                        void **data);
 984        void                    (*free)(struct ftrace_probe_ops *ops,
 985                                        struct trace_array *tr,
 986                                        unsigned long ip, void *data);
 987        int                     (*print)(struct seq_file *m,
 988                                         unsigned long ip,
 989                                         struct ftrace_probe_ops *ops,
 990                                         void *data);
 991};
 992
 993struct ftrace_func_mapper;
 994typedef int (*ftrace_mapper_func)(void *data);
 995
 996struct ftrace_func_mapper *allocate_ftrace_func_mapper(void);
 997void **ftrace_func_mapper_find_ip(struct ftrace_func_mapper *mapper,
 998                                           unsigned long ip);
 999int ftrace_func_mapper_add_ip(struct ftrace_func_mapper *mapper,
1000                               unsigned long ip, void *data);
1001void *ftrace_func_mapper_remove_ip(struct ftrace_func_mapper *mapper,
1002                                   unsigned long ip);
1003void free_ftrace_func_mapper(struct ftrace_func_mapper *mapper,
1004                             ftrace_mapper_func free_func);
1005
1006extern int
1007register_ftrace_function_probe(char *glob, struct trace_array *tr,
1008                               struct ftrace_probe_ops *ops, void *data);
1009extern int
1010unregister_ftrace_function_probe_func(char *glob, struct trace_array *tr,
1011                                      struct ftrace_probe_ops *ops);
1012extern void clear_ftrace_function_probes(struct trace_array *tr);
1013
1014int register_ftrace_command(struct ftrace_func_command *cmd);
1015int unregister_ftrace_command(struct ftrace_func_command *cmd);
1016
1017void ftrace_create_filter_files(struct ftrace_ops *ops,
1018                                struct dentry *parent);
1019void ftrace_destroy_filter_files(struct ftrace_ops *ops);
1020#else
1021struct ftrace_func_command;
1022
1023static inline __init int register_ftrace_command(struct ftrace_func_command *cmd)
1024{
1025        return -EINVAL;
1026}
1027static inline __init int unregister_ftrace_command(char *cmd_name)
1028{
1029        return -EINVAL;
1030}
1031static inline void clear_ftrace_function_probes(struct trace_array *tr)
1032{
1033}
1034
1035/*
1036 * The ops parameter passed in is usually undefined.
1037 * This must be a macro.
1038 */
1039#define ftrace_create_filter_files(ops, parent) do { } while (0)
1040#define ftrace_destroy_filter_files(ops) do { } while (0)
1041#endif /* CONFIG_FUNCTION_TRACER && CONFIG_DYNAMIC_FTRACE */
1042
1043bool ftrace_event_is_function(struct trace_event_call *call);
1044
1045/*
1046 * struct trace_parser - servers for reading the user input separated by spaces
1047 * @cont: set if the input is not complete - no final space char was found
1048 * @buffer: holds the parsed user input
1049 * @idx: user input length
1050 * @size: buffer size
1051 */
1052struct trace_parser {
1053        bool            cont;
1054        char            *buffer;
1055        unsigned        idx;
1056        unsigned        size;
1057};
1058
1059static inline bool trace_parser_loaded(struct trace_parser *parser)
1060{
1061        return (parser->idx != 0);
1062}
1063
1064static inline bool trace_parser_cont(struct trace_parser *parser)
1065{
1066        return parser->cont;
1067}
1068
1069static inline void trace_parser_clear(struct trace_parser *parser)
1070{
1071        parser->cont = false;
1072        parser->idx = 0;
1073}
1074
1075extern int trace_parser_get_init(struct trace_parser *parser, int size);
1076extern void trace_parser_put(struct trace_parser *parser);
1077extern int trace_get_user(struct trace_parser *parser, const char __user *ubuf,
1078        size_t cnt, loff_t *ppos);
1079
1080/*
1081 * Only create function graph options if function graph is configured.
1082 */
1083#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1084# define FGRAPH_FLAGS                                           \
1085                C(DISPLAY_GRAPH,        "display-graph"),
1086#else
1087# define FGRAPH_FLAGS
1088#endif
1089
1090#ifdef CONFIG_BRANCH_TRACER
1091# define BRANCH_FLAGS                                   \
1092                C(BRANCH,               "branch"),
1093#else
1094# define BRANCH_FLAGS
1095#endif
1096
1097#ifdef CONFIG_FUNCTION_TRACER
1098# define FUNCTION_FLAGS                                         \
1099                C(FUNCTION,             "function-trace"),      \
1100                C(FUNC_FORK,            "function-fork"),
1101# define FUNCTION_DEFAULT_FLAGS         TRACE_ITER_FUNCTION
1102#else
1103# define FUNCTION_FLAGS
1104# define FUNCTION_DEFAULT_FLAGS         0UL
1105# define TRACE_ITER_FUNC_FORK           0UL
1106#endif
1107
1108#ifdef CONFIG_STACKTRACE
1109# define STACK_FLAGS                            \
1110                C(STACKTRACE,           "stacktrace"),
1111#else
1112# define STACK_FLAGS
1113#endif
1114
1115/*
1116 * trace_iterator_flags is an enumeration that defines bit
1117 * positions into trace_flags that controls the output.
1118 *
1119 * NOTE: These bits must match the trace_options array in
1120 *       trace.c (this macro guarantees it).
1121 */
1122#define TRACE_FLAGS                                             \
1123                C(PRINT_PARENT,         "print-parent"),        \
1124                C(SYM_OFFSET,           "sym-offset"),          \
1125                C(SYM_ADDR,             "sym-addr"),            \
1126                C(VERBOSE,              "verbose"),             \
1127                C(RAW,                  "raw"),                 \
1128                C(HEX,                  "hex"),                 \
1129                C(BIN,                  "bin"),                 \
1130                C(BLOCK,                "block"),               \
1131                C(PRINTK,               "trace_printk"),        \
1132                C(ANNOTATE,             "annotate"),            \
1133                C(USERSTACKTRACE,       "userstacktrace"),      \
1134                C(SYM_USEROBJ,          "sym-userobj"),         \
1135                C(PRINTK_MSGONLY,       "printk-msg-only"),     \
1136                C(CONTEXT_INFO,         "context-info"),   /* Print pid/cpu/time */ \
1137                C(LATENCY_FMT,          "latency-format"),      \
1138                C(RECORD_CMD,           "record-cmd"),          \
1139                C(RECORD_TGID,          "record-tgid"),         \
1140                C(OVERWRITE,            "overwrite"),           \
1141                C(STOP_ON_FREE,         "disable_on_free"),     \
1142                C(IRQ_INFO,             "irq-info"),            \
1143                C(MARKERS,              "markers"),             \
1144                C(EVENT_FORK,           "event-fork"),          \
1145                FUNCTION_FLAGS                                  \
1146                FGRAPH_FLAGS                                    \
1147                STACK_FLAGS                                     \
1148                BRANCH_FLAGS
1149
1150/*
1151 * By defining C, we can make TRACE_FLAGS a list of bit names
1152 * that will define the bits for the flag masks.
1153 */
1154#undef C
1155#define C(a, b) TRACE_ITER_##a##_BIT
1156
1157enum trace_iterator_bits {
1158        TRACE_FLAGS
1159        /* Make sure we don't go more than we have bits for */
1160        TRACE_ITER_LAST_BIT
1161};
1162
1163/*
1164 * By redefining C, we can make TRACE_FLAGS a list of masks that
1165 * use the bits as defined above.
1166 */
1167#undef C
1168#define C(a, b) TRACE_ITER_##a = (1 << TRACE_ITER_##a##_BIT)
1169
1170enum trace_iterator_flags { TRACE_FLAGS };
1171
1172/*
1173 * TRACE_ITER_SYM_MASK masks the options in trace_flags that
1174 * control the output of kernel symbols.
1175 */
1176#define TRACE_ITER_SYM_MASK \
1177        (TRACE_ITER_PRINT_PARENT|TRACE_ITER_SYM_OFFSET|TRACE_ITER_SYM_ADDR)
1178
1179extern struct tracer nop_trace;
1180
1181#ifdef CONFIG_BRANCH_TRACER
1182extern int enable_branch_tracing(struct trace_array *tr);
1183extern void disable_branch_tracing(void);
1184static inline int trace_branch_enable(struct trace_array *tr)
1185{
1186        if (tr->trace_flags & TRACE_ITER_BRANCH)
1187                return enable_branch_tracing(tr);
1188        return 0;
1189}
1190static inline void trace_branch_disable(void)
1191{
1192        /* due to races, always disable */
1193        disable_branch_tracing();
1194}
1195#else
1196static inline int trace_branch_enable(struct trace_array *tr)
1197{
1198        return 0;
1199}
1200static inline void trace_branch_disable(void)
1201{
1202}
1203#endif /* CONFIG_BRANCH_TRACER */
1204
1205/* set ring buffers to default size if not already done so */
1206int tracing_update_buffers(void);
1207
1208struct ftrace_event_field {
1209        struct list_head        link;
1210        const char              *name;
1211        const char              *type;
1212        int                     filter_type;
1213        int                     offset;
1214        int                     size;
1215        int                     is_signed;
1216};
1217
1218struct prog_entry;
1219
1220struct event_filter {
1221        struct prog_entry __rcu *prog;
1222        char                    *filter_string;
1223};
1224
1225struct event_subsystem {
1226        struct list_head        list;
1227        const char              *name;
1228        struct event_filter     *filter;
1229        int                     ref_count;
1230};
1231
1232struct trace_subsystem_dir {
1233        struct list_head                list;
1234        struct event_subsystem          *subsystem;
1235        struct trace_array              *tr;
1236        struct dentry                   *entry;
1237        int                             ref_count;
1238        int                             nr_events;
1239};
1240
1241extern int call_filter_check_discard(struct trace_event_call *call, void *rec,
1242                                     struct ring_buffer *buffer,
1243                                     struct ring_buffer_event *event);
1244
1245void trace_buffer_unlock_commit_regs(struct trace_array *tr,
1246                                     struct ring_buffer *buffer,
1247                                     struct ring_buffer_event *event,
1248                                     unsigned long flags, int pc,
1249                                     struct pt_regs *regs);
1250
1251static inline void trace_buffer_unlock_commit(struct trace_array *tr,
1252                                              struct ring_buffer *buffer,
1253                                              struct ring_buffer_event *event,
1254                                              unsigned long flags, int pc)
1255{
1256        trace_buffer_unlock_commit_regs(tr, buffer, event, flags, pc, NULL);
1257}
1258
1259DECLARE_PER_CPU(struct ring_buffer_event *, trace_buffered_event);
1260DECLARE_PER_CPU(int, trace_buffered_event_cnt);
1261void trace_buffered_event_disable(void);
1262void trace_buffered_event_enable(void);
1263
1264static inline void
1265__trace_event_discard_commit(struct ring_buffer *buffer,
1266                             struct ring_buffer_event *event)
1267{
1268        if (this_cpu_read(trace_buffered_event) == event) {
1269                /* Simply release the temp buffer */
1270                this_cpu_dec(trace_buffered_event_cnt);
1271                return;
1272        }
1273        ring_buffer_discard_commit(buffer, event);
1274}
1275
1276/*
1277 * Helper function for event_trigger_unlock_commit{_regs}().
1278 * If there are event triggers attached to this event that requires
1279 * filtering against its fields, then they wil be called as the
1280 * entry already holds the field information of the current event.
1281 *
1282 * It also checks if the event should be discarded or not.
1283 * It is to be discarded if the event is soft disabled and the
1284 * event was only recorded to process triggers, or if the event
1285 * filter is active and this event did not match the filters.
1286 *
1287 * Returns true if the event is discarded, false otherwise.
1288 */
1289static inline bool
1290__event_trigger_test_discard(struct trace_event_file *file,
1291                             struct ring_buffer *buffer,
1292                             struct ring_buffer_event *event,
1293                             void *entry,
1294                             enum event_trigger_type *tt)
1295{
1296        unsigned long eflags = file->flags;
1297
1298        if (eflags & EVENT_FILE_FL_TRIGGER_COND)
1299                *tt = event_triggers_call(file, entry, event);
1300
1301        if (test_bit(EVENT_FILE_FL_SOFT_DISABLED_BIT, &file->flags) ||
1302            (unlikely(file->flags & EVENT_FILE_FL_FILTERED) &&
1303             !filter_match_preds(file->filter, entry))) {
1304                __trace_event_discard_commit(buffer, event);
1305                return true;
1306        }
1307
1308        return false;
1309}
1310
1311/**
1312 * event_trigger_unlock_commit - handle triggers and finish event commit
1313 * @file: The file pointer assoctiated to the event
1314 * @buffer: The ring buffer that the event is being written to
1315 * @event: The event meta data in the ring buffer
1316 * @entry: The event itself
1317 * @irq_flags: The state of the interrupts at the start of the event
1318 * @pc: The state of the preempt count at the start of the event.
1319 *
1320 * This is a helper function to handle triggers that require data
1321 * from the event itself. It also tests the event against filters and
1322 * if the event is soft disabled and should be discarded.
1323 */
1324static inline void
1325event_trigger_unlock_commit(struct trace_event_file *file,
1326                            struct ring_buffer *buffer,
1327                            struct ring_buffer_event *event,
1328                            void *entry, unsigned long irq_flags, int pc)
1329{
1330        enum event_trigger_type tt = ETT_NONE;
1331
1332        if (!__event_trigger_test_discard(file, buffer, event, entry, &tt))
1333                trace_buffer_unlock_commit(file->tr, buffer, event, irq_flags, pc);
1334
1335        if (tt)
1336                event_triggers_post_call(file, tt);
1337}
1338
1339/**
1340 * event_trigger_unlock_commit_regs - handle triggers and finish event commit
1341 * @file: The file pointer assoctiated to the event
1342 * @buffer: The ring buffer that the event is being written to
1343 * @event: The event meta data in the ring buffer
1344 * @entry: The event itself
1345 * @irq_flags: The state of the interrupts at the start of the event
1346 * @pc: The state of the preempt count at the start of the event.
1347 *
1348 * This is a helper function to handle triggers that require data
1349 * from the event itself. It also tests the event against filters and
1350 * if the event is soft disabled and should be discarded.
1351 *
1352 * Same as event_trigger_unlock_commit() but calls
1353 * trace_buffer_unlock_commit_regs() instead of trace_buffer_unlock_commit().
1354 */
1355static inline void
1356event_trigger_unlock_commit_regs(struct trace_event_file *file,
1357                                 struct ring_buffer *buffer,
1358                                 struct ring_buffer_event *event,
1359                                 void *entry, unsigned long irq_flags, int pc,
1360                                 struct pt_regs *regs)
1361{
1362        enum event_trigger_type tt = ETT_NONE;
1363
1364        if (!__event_trigger_test_discard(file, buffer, event, entry, &tt))
1365                trace_buffer_unlock_commit_regs(file->tr, buffer, event,
1366                                                irq_flags, pc, regs);
1367
1368        if (tt)
1369                event_triggers_post_call(file, tt);
1370}
1371
1372#define FILTER_PRED_INVALID     ((unsigned short)-1)
1373#define FILTER_PRED_IS_RIGHT    (1 << 15)
1374#define FILTER_PRED_FOLD        (1 << 15)
1375
1376/*
1377 * The max preds is the size of unsigned short with
1378 * two flags at the MSBs. One bit is used for both the IS_RIGHT
1379 * and FOLD flags. The other is reserved.
1380 *
1381 * 2^14 preds is way more than enough.
1382 */
1383#define MAX_FILTER_PRED         16384
1384
1385struct filter_pred;
1386struct regex;
1387
1388typedef int (*filter_pred_fn_t) (struct filter_pred *pred, void *event);
1389
1390typedef int (*regex_match_func)(char *str, struct regex *r, int len);
1391
1392enum regex_type {
1393        MATCH_FULL = 0,
1394        MATCH_FRONT_ONLY,
1395        MATCH_MIDDLE_ONLY,
1396        MATCH_END_ONLY,
1397        MATCH_GLOB,
1398};
1399
1400struct regex {
1401        char                    pattern[MAX_FILTER_STR_VAL];
1402        int                     len;
1403        int                     field_len;
1404        regex_match_func        match;
1405};
1406
1407struct filter_pred {
1408        filter_pred_fn_t        fn;
1409        u64                     val;
1410        struct regex            regex;
1411        unsigned short          *ops;
1412        struct ftrace_event_field *field;
1413        int                     offset;
1414        int                     not;
1415        int                     op;
1416};
1417
1418static inline bool is_string_field(struct ftrace_event_field *field)
1419{
1420        return field->filter_type == FILTER_DYN_STRING ||
1421               field->filter_type == FILTER_STATIC_STRING ||
1422               field->filter_type == FILTER_PTR_STRING ||
1423               field->filter_type == FILTER_COMM;
1424}
1425
1426static inline bool is_function_field(struct ftrace_event_field *field)
1427{
1428        return field->filter_type == FILTER_TRACE_FN;
1429}
1430
1431extern enum regex_type
1432filter_parse_regex(char *buff, int len, char **search, int *not);
1433extern void print_event_filter(struct trace_event_file *file,
1434                               struct trace_seq *s);
1435extern int apply_event_filter(struct trace_event_file *file,
1436                              char *filter_string);
1437extern int apply_subsystem_event_filter(struct trace_subsystem_dir *dir,
1438                                        char *filter_string);
1439extern void print_subsystem_event_filter(struct event_subsystem *system,
1440                                         struct trace_seq *s);
1441extern int filter_assign_type(const char *type);
1442extern int create_event_filter(struct trace_event_call *call,
1443                               char *filter_str, bool set_str,
1444                               struct event_filter **filterp);
1445extern void free_event_filter(struct event_filter *filter);
1446
1447struct ftrace_event_field *
1448trace_find_event_field(struct trace_event_call *call, char *name);
1449
1450extern void trace_event_enable_cmd_record(bool enable);
1451extern void trace_event_enable_tgid_record(bool enable);
1452
1453extern int event_trace_init(void);
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);
1460extern struct trace_event_file *find_event_file(struct trace_array *tr,
1461                                                const char *system,
1462                                                const char *event);
1463
1464static inline void *event_file_data(struct file *filp)
1465{
1466        return READ_ONCE(file_inode(filp)->i_private);
1467}
1468
1469extern struct mutex event_mutex;
1470extern struct list_head ftrace_events;
1471
1472extern const struct file_operations event_trigger_fops;
1473extern const struct file_operations event_hist_fops;
1474
1475#ifdef CONFIG_HIST_TRIGGERS
1476extern int register_trigger_hist_cmd(void);
1477extern int register_trigger_hist_enable_disable_cmds(void);
1478#else
1479static inline int register_trigger_hist_cmd(void) { return 0; }
1480static inline int register_trigger_hist_enable_disable_cmds(void) { return 0; }
1481#endif
1482
1483extern int register_trigger_cmds(void);
1484extern void clear_event_triggers(struct trace_array *tr);
1485
1486struct event_trigger_data {
1487        unsigned long                   count;
1488        int                             ref;
1489        struct event_trigger_ops        *ops;
1490        struct event_command            *cmd_ops;
1491        struct event_filter __rcu       *filter;
1492        char                            *filter_str;
1493        void                            *private_data;
1494        bool                            paused;
1495        bool                            paused_tmp;
1496        struct list_head                list;
1497        char                            *name;
1498        struct list_head                named_list;
1499        struct event_trigger_data       *named_data;
1500};
1501
1502/* Avoid typos */
1503#define ENABLE_EVENT_STR        "enable_event"
1504#define DISABLE_EVENT_STR       "disable_event"
1505#define ENABLE_HIST_STR         "enable_hist"
1506#define DISABLE_HIST_STR        "disable_hist"
1507
1508struct enable_trigger_data {
1509        struct trace_event_file         *file;
1510        bool                            enable;
1511        bool                            hist;
1512};
1513
1514extern int event_enable_trigger_print(struct seq_file *m,
1515                                      struct event_trigger_ops *ops,
1516                                      struct event_trigger_data *data);
1517extern void event_enable_trigger_free(struct event_trigger_ops *ops,
1518                                      struct event_trigger_data *data);
1519extern int event_enable_trigger_func(struct event_command *cmd_ops,
1520                                     struct trace_event_file *file,
1521                                     char *glob, char *cmd, char *param);
1522extern int event_enable_register_trigger(char *glob,
1523                                         struct event_trigger_ops *ops,
1524                                         struct event_trigger_data *data,
1525                                         struct trace_event_file *file);
1526extern void event_enable_unregister_trigger(char *glob,
1527                                            struct event_trigger_ops *ops,
1528                                            struct event_trigger_data *test,
1529                                            struct trace_event_file *file);
1530extern void trigger_data_free(struct event_trigger_data *data);
1531extern int event_trigger_init(struct event_trigger_ops *ops,
1532                              struct event_trigger_data *data);
1533extern int trace_event_trigger_enable_disable(struct trace_event_file *file,
1534                                              int trigger_enable);
1535extern void update_cond_flag(struct trace_event_file *file);
1536extern void unregister_trigger(char *glob, struct event_trigger_ops *ops,
1537                               struct event_trigger_data *test,
1538                               struct trace_event_file *file);
1539extern int set_trigger_filter(char *filter_str,
1540                              struct event_trigger_data *trigger_data,
1541                              struct trace_event_file *file);
1542extern struct event_trigger_data *find_named_trigger(const char *name);
1543extern bool is_named_trigger(struct event_trigger_data *test);
1544extern int save_named_trigger(const char *name,
1545                              struct event_trigger_data *data);
1546extern void del_named_trigger(struct event_trigger_data *data);
1547extern void pause_named_trigger(struct event_trigger_data *data);
1548extern void unpause_named_trigger(struct event_trigger_data *data);
1549extern void set_named_trigger_data(struct event_trigger_data *data,
1550                                   struct event_trigger_data *named_data);
1551extern struct event_trigger_data *
1552get_named_trigger_data(struct event_trigger_data *data);
1553extern int register_event_command(struct event_command *cmd);
1554extern int unregister_event_command(struct event_command *cmd);
1555extern int register_trigger_hist_enable_disable_cmds(void);
1556
1557/**
1558 * struct event_trigger_ops - callbacks for trace event triggers
1559 *
1560 * The methods in this structure provide per-event trigger hooks for
1561 * various trigger operations.
1562 *
1563 * All the methods below, except for @init() and @free(), must be
1564 * implemented.
1565 *
1566 * @func: The trigger 'probe' function called when the triggering
1567 *      event occurs.  The data passed into this callback is the data
1568 *      that was supplied to the event_command @reg() function that
1569 *      registered the trigger (see struct event_command) along with
1570 *      the trace record, rec.
1571 *
1572 * @init: An optional initialization function called for the trigger
1573 *      when the trigger is registered (via the event_command reg()
1574 *      function).  This can be used to perform per-trigger
1575 *      initialization such as incrementing a per-trigger reference
1576 *      count, for instance.  This is usually implemented by the
1577 *      generic utility function @event_trigger_init() (see
1578 *      trace_event_triggers.c).
1579 *
1580 * @free: An optional de-initialization function called for the
1581 *      trigger when the trigger is unregistered (via the
1582 *      event_command @reg() function).  This can be used to perform
1583 *      per-trigger de-initialization such as decrementing a
1584 *      per-trigger reference count and freeing corresponding trigger
1585 *      data, for instance.  This is usually implemented by the
1586 *      generic utility function @event_trigger_free() (see
1587 *      trace_event_triggers.c).
1588 *
1589 * @print: The callback function invoked to have the trigger print
1590 *      itself.  This is usually implemented by a wrapper function
1591 *      that calls the generic utility function @event_trigger_print()
1592 *      (see trace_event_triggers.c).
1593 */
1594struct event_trigger_ops {
1595        void                    (*func)(struct event_trigger_data *data,
1596                                        void *rec,
1597                                        struct ring_buffer_event *rbe);
1598        int                     (*init)(struct event_trigger_ops *ops,
1599                                        struct event_trigger_data *data);
1600        void                    (*free)(struct event_trigger_ops *ops,
1601                                        struct event_trigger_data *data);
1602        int                     (*print)(struct seq_file *m,
1603                                         struct event_trigger_ops *ops,
1604                                         struct event_trigger_data *data);
1605};
1606
1607/**
1608 * struct event_command - callbacks and data members for event commands
1609 *
1610 * Event commands are invoked by users by writing the command name
1611 * into the 'trigger' file associated with a trace event.  The
1612 * parameters associated with a specific invocation of an event
1613 * command are used to create an event trigger instance, which is
1614 * added to the list of trigger instances associated with that trace
1615 * event.  When the event is hit, the set of triggers associated with
1616 * that event is invoked.
1617 *
1618 * The data members in this structure provide per-event command data
1619 * for various event commands.
1620 *
1621 * All the data members below, except for @post_trigger, must be set
1622 * for each event command.
1623 *
1624 * @name: The unique name that identifies the event command.  This is
1625 *      the name used when setting triggers via trigger files.
1626 *
1627 * @trigger_type: A unique id that identifies the event command
1628 *      'type'.  This value has two purposes, the first to ensure that
1629 *      only one trigger of the same type can be set at a given time
1630 *      for a particular event e.g. it doesn't make sense to have both
1631 *      a traceon and traceoff trigger attached to a single event at
1632 *      the same time, so traceon and traceoff have the same type
1633 *      though they have different names.  The @trigger_type value is
1634 *      also used as a bit value for deferring the actual trigger
1635 *      action until after the current event is finished.  Some
1636 *      commands need to do this if they themselves log to the trace
1637 *      buffer (see the @post_trigger() member below).  @trigger_type
1638 *      values are defined by adding new values to the trigger_type
1639 *      enum in include/linux/trace_events.h.
1640 *
1641 * @flags: See the enum event_command_flags below.
1642 *
1643 * All the methods below, except for @set_filter() and @unreg_all(),
1644 * must be implemented.
1645 *
1646 * @func: The callback function responsible for parsing and
1647 *      registering the trigger written to the 'trigger' file by the
1648 *      user.  It allocates the trigger instance and registers it with
1649 *      the appropriate trace event.  It makes use of the other
1650 *      event_command callback functions to orchestrate this, and is
1651 *      usually implemented by the generic utility function
1652 *      @event_trigger_callback() (see trace_event_triggers.c).
1653 *
1654 * @reg: Adds the trigger to the list of triggers associated with the
1655 *      event, and enables the event trigger itself, after
1656 *      initializing it (via the event_trigger_ops @init() function).
1657 *      This is also where commands can use the @trigger_type value to
1658 *      make the decision as to whether or not multiple instances of
1659 *      the trigger should be allowed.  This is usually implemented by
1660 *      the generic utility function @register_trigger() (see
1661 *      trace_event_triggers.c).
1662 *
1663 * @unreg: Removes the trigger from the list of triggers associated
1664 *      with the event, and disables the event trigger itself, after
1665 *      initializing it (via the event_trigger_ops @free() function).
1666 *      This is usually implemented by the generic utility function
1667 *      @unregister_trigger() (see trace_event_triggers.c).
1668 *
1669 * @unreg_all: An optional function called to remove all the triggers
1670 *      from the list of triggers associated with the event.  Called
1671 *      when a trigger file is opened in truncate mode.
1672 *
1673 * @set_filter: An optional function called to parse and set a filter
1674 *      for the trigger.  If no @set_filter() method is set for the
1675 *      event command, filters set by the user for the command will be
1676 *      ignored.  This is usually implemented by the generic utility
1677 *      function @set_trigger_filter() (see trace_event_triggers.c).
1678 *
1679 * @get_trigger_ops: The callback function invoked to retrieve the
1680 *      event_trigger_ops implementation associated with the command.
1681 */
1682struct event_command {
1683        struct list_head        list;
1684        char                    *name;
1685        enum event_trigger_type trigger_type;
1686        int                     flags;
1687        int                     (*func)(struct event_command *cmd_ops,
1688                                        struct trace_event_file *file,
1689                                        char *glob, char *cmd, char *params);
1690        int                     (*reg)(char *glob,
1691                                       struct event_trigger_ops *ops,
1692                                       struct event_trigger_data *data,
1693                                       struct trace_event_file *file);
1694        void                    (*unreg)(char *glob,
1695                                         struct event_trigger_ops *ops,
1696                                         struct event_trigger_data *data,
1697                                         struct trace_event_file *file);
1698        void                    (*unreg_all)(struct trace_event_file *file);
1699        int                     (*set_filter)(char *filter_str,
1700                                              struct event_trigger_data *data,
1701                                              struct trace_event_file *file);
1702        struct event_trigger_ops *(*get_trigger_ops)(char *cmd, char *param);
1703};
1704
1705/**
1706 * enum event_command_flags - flags for struct event_command
1707 *
1708 * @POST_TRIGGER: A flag that says whether or not this command needs
1709 *      to have its action delayed until after the current event has
1710 *      been closed.  Some triggers need to avoid being invoked while
1711 *      an event is currently in the process of being logged, since
1712 *      the trigger may itself log data into the trace buffer.  Thus
1713 *      we make sure the current event is committed before invoking
1714 *      those triggers.  To do that, the trigger invocation is split
1715 *      in two - the first part checks the filter using the current
1716 *      trace record; if a command has the @post_trigger flag set, it
1717 *      sets a bit for itself in the return value, otherwise it
1718 *      directly invokes the trigger.  Once all commands have been
1719 *      either invoked or set their return flag, the current record is
1720 *      either committed or discarded.  At that point, if any commands
1721 *      have deferred their triggers, those commands are finally
1722 *      invoked following the close of the current event.  In other
1723 *      words, if the event_trigger_ops @func() probe implementation
1724 *      itself logs to the trace buffer, this flag should be set,
1725 *      otherwise it can be left unspecified.
1726 *
1727 * @NEEDS_REC: A flag that says whether or not this command needs
1728 *      access to the trace record in order to perform its function,
1729 *      regardless of whether or not it has a filter associated with
1730 *      it (filters make a trigger require access to the trace record
1731 *      but are not always present).
1732 */
1733enum event_command_flags {
1734        EVENT_CMD_FL_POST_TRIGGER       = 1,
1735        EVENT_CMD_FL_NEEDS_REC          = 2,
1736};
1737
1738static inline bool event_command_post_trigger(struct event_command *cmd_ops)
1739{
1740        return cmd_ops->flags & EVENT_CMD_FL_POST_TRIGGER;
1741}
1742
1743static inline bool event_command_needs_rec(struct event_command *cmd_ops)
1744{
1745        return cmd_ops->flags & EVENT_CMD_FL_NEEDS_REC;
1746}
1747
1748extern int trace_event_enable_disable(struct trace_event_file *file,
1749                                      int enable, int soft_disable);
1750extern int tracing_alloc_snapshot(void);
1751
1752extern const char *__start___trace_bprintk_fmt[];
1753extern const char *__stop___trace_bprintk_fmt[];
1754
1755extern const char *__start___tracepoint_str[];
1756extern const char *__stop___tracepoint_str[];
1757
1758void trace_printk_control(bool enabled);
1759void trace_printk_init_buffers(void);
1760void trace_printk_start_comm(void);
1761int trace_keep_overwrite(struct tracer *tracer, u32 mask, int set);
1762int set_tracer_flag(struct trace_array *tr, unsigned int mask, int enabled);
1763
1764#define MAX_EVENT_NAME_LEN      64
1765
1766extern int trace_run_command(const char *buf, int (*createfn)(int, char**));
1767extern ssize_t trace_parse_run_command(struct file *file,
1768                const char __user *buffer, size_t count, loff_t *ppos,
1769                int (*createfn)(int, char**));
1770
1771/*
1772 * Normal trace_printk() and friends allocates special buffers
1773 * to do the manipulation, as well as saves the print formats
1774 * into sections to display. But the trace infrastructure wants
1775 * to use these without the added overhead at the price of being
1776 * a bit slower (used mainly for warnings, where we don't care
1777 * about performance). The internal_trace_puts() is for such
1778 * a purpose.
1779 */
1780#define internal_trace_puts(str) __trace_puts(_THIS_IP_, str, strlen(str))
1781
1782#undef FTRACE_ENTRY
1783#define FTRACE_ENTRY(call, struct_name, id, tstruct, print, filter)     \
1784        extern struct trace_event_call                                  \
1785        __aligned(4) event_##call;
1786#undef FTRACE_ENTRY_DUP
1787#define FTRACE_ENTRY_DUP(call, struct_name, id, tstruct, print, filter) \
1788        FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print), \
1789                     filter)
1790#undef FTRACE_ENTRY_PACKED
1791#define FTRACE_ENTRY_PACKED(call, struct_name, id, tstruct, print, filter) \
1792        FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print), \
1793                     filter)
1794
1795#include "trace_entries.h"
1796
1797#if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_FUNCTION_TRACER)
1798int perf_ftrace_event_register(struct trace_event_call *call,
1799                               enum trace_reg type, void *data);
1800#else
1801#define perf_ftrace_event_register NULL
1802#endif
1803
1804#ifdef CONFIG_FTRACE_SYSCALLS
1805void init_ftrace_syscalls(void);
1806const char *get_syscall_name(int syscall);
1807#else
1808static inline void init_ftrace_syscalls(void) { }
1809static inline const char *get_syscall_name(int syscall)
1810{
1811        return NULL;
1812}
1813#endif
1814
1815#ifdef CONFIG_EVENT_TRACING
1816void trace_event_init(void);
1817void trace_event_eval_update(struct trace_eval_map **map, int len);
1818#else
1819static inline void __init trace_event_init(void) { }
1820static inline void trace_event_eval_update(struct trace_eval_map **map, int len) { }
1821#endif
1822
1823#ifdef CONFIG_TRACER_SNAPSHOT
1824void tracing_snapshot_instance(struct trace_array *tr);
1825int tracing_alloc_snapshot_instance(struct trace_array *tr);
1826#else
1827static inline void tracing_snapshot_instance(struct trace_array *tr) { }
1828static inline int tracing_alloc_snapshot_instance(struct trace_array *tr)
1829{
1830        return 0;
1831}
1832#endif
1833
1834extern struct trace_iterator *tracepoint_print_iter;
1835
1836#endif /* _LINUX_KERNEL_TRACE_H */
1837