kernel/trace/trace.h

   1 #ifndef _LINUX_KERNEL_TRACE_H
   2 #define _LINUX_KERNEL_TRACE_H
   3
   4 #include <linux/fs.h>
   5 #include <asm/atomic.h>
   6 #include <linux/sched.h>
   7 #include <linux/clocksource.h>
   8 #include <linux/ring_buffer.h>
   9 #include <linux/mmiotrace.h>
  10 #include <linux/ftrace.h>
  11 #include <trace/boot.h>
  12
  13 enum trace_type {
  14         __TRACE_FIRST_TYPE = 0,
  15
  16         TRACE_FN,
  17         TRACE_CTX,
  18         TRACE_WAKE,
  19         TRACE_CONT,
  20         TRACE_STACK,
  21         TRACE_PRINT,
  22         TRACE_SPECIAL,
  23         TRACE_MMIO_RW,
  24         TRACE_MMIO_MAP,
  25         TRACE_BRANCH,
  26         TRACE_BOOT_CALL,
  27         TRACE_BOOT_RET,
  28         TRACE_FN_RET,
  29         TRACE_USER_STACK,
  30         TRACE_BTS,
  31         TRACE_POWER,
  32
  33         __TRACE_LAST_TYPE
  34 };
  35
  36 /*
  37  * The trace entry - the most basic unit of tracing. This is what
  38  * is printed in the end as a single line in the trace output, such as:
  39  *
  40  *     bash-15816 [01]   235.197585: idle_cpu <- irq_enter
  41  */
  42 struct trace_entry {
  43         unsigned char           type;
  44         unsigned char           cpu;
  45         unsigned char           flags;
  46         unsigned char           preempt_count;
  47         int                     pid;
  48         int                     tgid;
  49 };
  50
  51 /*
  52  * Function trace entry - function address and parent function addres:
  53  */
  54 struct ftrace_entry {
  55         struct trace_entry      ent;
  56         unsigned long           ip;
  57         unsigned long           parent_ip;
  58 };
  59
  60 /* Function return entry */
  61 struct ftrace_ret_entry {
  62         struct trace_entry      ent;
  63         unsigned long           ip;
  64         unsigned long           parent_ip;
  65         unsigned long long      calltime;
  66         unsigned long long      rettime;
  67         unsigned long           overrun;
  68 };
  69 extern struct tracer boot_tracer;
  70
  71 /*
  72  * Context switch trace entry - which task (and prio) we switched from/to:
  73  */
  74 struct ctx_switch_entry {
  75         struct trace_entry      ent;
  76         unsigned int            prev_pid;
  77         unsigned char           prev_prio;
  78         unsigned char           prev_state;
  79         unsigned int            next_pid;
  80         unsigned char           next_prio;
  81         unsigned char           next_state;
  82         unsigned int            next_cpu;
  83 };
  84
  85 /*
  86  * Special (free-form) trace entry:
  87  */
  88 struct special_entry {
  89         struct trace_entry      ent;
  90         unsigned long           arg1;
  91         unsigned long           arg2;
  92         unsigned long           arg3;
  93 };
  94
  95 /*
  96  * Stack-trace entry:
  97  */
  98
  99 #define FTRACE_STACK_ENTRIES    8
 100
 101 struct stack_entry {
 102         struct trace_entry      ent;
 103         unsigned long           caller[FTRACE_STACK_ENTRIES];
 104 };
 105
 106 struct userstack_entry {
 107         struct trace_entry      ent;
 108         unsigned long           caller[FTRACE_STACK_ENTRIES];
 109 };
 110
 111 /*
 112  * ftrace_printk entry:
 113  */
 114 struct print_entry {
 115         struct trace_entry      ent;
 116         unsigned long           ip;
 117         char                    buf[];
 118 };
 119
 120 #define TRACE_OLD_SIZE          88
 121
 122 struct trace_field_cont {
 123         unsigned char           type;
 124         /* Temporary till we get rid of this completely */
 125         char                    buf[TRACE_OLD_SIZE - 1];
 126 };
 127
 128 struct trace_mmiotrace_rw {
 129         struct trace_entry      ent;
 130         struct mmiotrace_rw     rw;
 131 };
 132
 133 struct trace_mmiotrace_map {
 134         struct trace_entry      ent;
 135         struct mmiotrace_map    map;
 136 };
 137
 138 struct trace_boot_call {
 139         struct trace_entry      ent;
 140         struct boot_trace_call boot_call;
 141 };
 142
 143 struct trace_boot_ret {
 144         struct trace_entry      ent;
 145         struct boot_trace_ret boot_ret;
 146 };
 147
 148 #define TRACE_FUNC_SIZE 30
 149 #define TRACE_FILE_SIZE 20
 150 struct trace_branch {
 151         struct trace_entry      ent;
 152         unsigned                line;
 153         char                    func[TRACE_FUNC_SIZE+1];
 154         char                    file[TRACE_FILE_SIZE+1];
 155         char                    correct;
 156 };
 157
 158 struct bts_entry {
 159         struct trace_entry      ent;
 160         unsigned long           from;
 161         unsigned long           to;
 162 };
 163
 164 struct trace_power {
 165         struct trace_entry      ent;
 166         struct power_trace      state_data;
 167 };
 168
 169 /*
 170  * trace_flag_type is an enumeration that holds different
 171  * states when a trace occurs. These are:
 172  *  IRQS_OFF            - interrupts were disabled
 173  *  IRQS_NOSUPPORT      - arch does not support irqs_disabled_flags
 174  *  NEED_RESCED         - reschedule is requested
 175  *  HARDIRQ             - inside an interrupt handler
 176  *  SOFTIRQ             - inside a softirq handler
 177  *  CONT                - multiple entries hold the trace item
 178  */
 179 enum trace_flag_type {
 180         TRACE_FLAG_IRQS_OFF             = 0x01,
 181         TRACE_FLAG_IRQS_NOSUPPORT       = 0x02,
 182         TRACE_FLAG_NEED_RESCHED         = 0x04,
 183         TRACE_FLAG_HARDIRQ              = 0x08,
 184         TRACE_FLAG_SOFTIRQ              = 0x10,
 185         TRACE_FLAG_CONT                 = 0x20,
 186 };
 187
 188 #define TRACE_BUF_SIZE          1024
 189
 190 /*
 191  * The CPU trace array - it consists of thousands of trace entries
 192  * plus some other descriptor data: (for example which task started
 193  * the trace, etc.)
 194  */
 195 struct trace_array_cpu {
 196         atomic_t                disabled;
 197
 198         /* these fields get copied into max-trace: */
 199         unsigned long           trace_idx;
 200         unsigned long           overrun;
 201         unsigned long           saved_latency;
 202         unsigned long           critical_start;
 203         unsigned long           critical_end;
 204         unsigned long           critical_sequence;
 205         unsigned long           nice;
 206         unsigned long           policy;
 207         unsigned long           rt_priority;
 208         cycle_t                 preempt_timestamp;
 209         pid_t                   pid;
 210         uid_t                   uid;
 211         char                    comm[TASK_COMM_LEN];
 212 };
 213
 214 struct trace_iterator;
 215
 216 /*
 217  * The trace array - an array of per-CPU trace arrays. This is the
 218  * highest level data structure that individual tracers deal with.
 219  * They have on/off state as well:
 220  */
 221 struct trace_array {
 222         struct ring_buffer      *buffer;
 223         unsigned long           entries;
 224         int                     cpu;
 225         cycle_t                 time_start;
 226         struct task_struct      *waiter;
 227         struct trace_array_cpu  *data[NR_CPUS];
 228 };
 229
 230 #define FTRACE_CMP_TYPE(var, type) \
 231         __builtin_types_compatible_p(typeof(var), type *)
 232
 233 #undef IF_ASSIGN
 234 #define IF_ASSIGN(var, entry, etype, id)                \
 235         if (FTRACE_CMP_TYPE(var, etype)) {              \
 236                 var = (typeof(var))(entry);             \
 237                 WARN_ON(id && (entry)->type != id);     \
 238                 break;                                  \
 239         }
 240
 241 /* Will cause compile errors if type is not found. */
 242 extern void __ftrace_bad_type(void);
 243
 244 /*
 245  * The trace_assign_type is a verifier that the entry type is
 246  * the same as the type being assigned. To add new types simply
 247  * add a line with the following format:
 248  *
 249  * IF_ASSIGN(var, ent, type, id);
 250  *
 251  *  Where "type" is the trace type that includes the trace_entry
 252  *  as the "ent" item. And "id" is the trace identifier that is
 253  *  used in the trace_type enum.
 254  *
 255  *  If the type can have more than one id, then use zero.
 256  */
 257 #define trace_assign_type(var, ent)                                     \
 258         do {                                                            \
 259                 IF_ASSIGN(var, ent, struct ftrace_entry, TRACE_FN);     \
 260                 IF_ASSIGN(var, ent, struct ctx_switch_entry, 0);        \
 261                 IF_ASSIGN(var, ent, struct trace_field_cont, TRACE_CONT); \
 262                 IF_ASSIGN(var, ent, struct stack_entry, TRACE_STACK);   \
 263                 IF_ASSIGN(var, ent, struct userstack_entry, TRACE_USER_STACK);\
 264                 IF_ASSIGN(var, ent, struct print_entry, TRACE_PRINT);   \
 265                 IF_ASSIGN(var, ent, struct special_entry, 0);           \
 266                 IF_ASSIGN(var, ent, struct trace_mmiotrace_rw,          \
 267                           TRACE_MMIO_RW);                               \
 268                 IF_ASSIGN(var, ent, struct trace_mmiotrace_map,         \
 269                           TRACE_MMIO_MAP);                              \
 270                 IF_ASSIGN(var, ent, struct trace_boot_call, TRACE_BOOT_CALL);\
 271                 IF_ASSIGN(var, ent, struct trace_boot_ret, TRACE_BOOT_RET);\
 272                 IF_ASSIGN(var, ent, struct trace_branch, TRACE_BRANCH); \
 273                 IF_ASSIGN(var, ent, struct ftrace_ret_entry, TRACE_FN_RET);\
 274                 IF_ASSIGN(var, ent, struct bts_entry, TRACE_BTS);\
 275                 IF_ASSIGN(var, ent, struct trace_power, TRACE_POWER); \
 276                 __ftrace_bad_type();                                    \
 277         } while (0)
 278
 279 /* Return values for print_line callback */
 280 enum print_line_t {
 281         TRACE_TYPE_PARTIAL_LINE = 0,    /* Retry after flushing the seq */
 282         TRACE_TYPE_HANDLED      = 1,
 283         TRACE_TYPE_UNHANDLED    = 2     /* Relay to other output functions */
 284 };
 285
 286
 287 /*
 288  * An option specific to a tracer. This is a boolean value.
 289  * The bit is the bit index that sets its value on the
 290  * flags value in struct tracer_flags.
 291  */
 292 struct tracer_opt {
 293         const char      *name; /* Will appear on the trace_options file */
 294         u32             bit; /* Mask assigned in val field in tracer_flags */
 295 };
 296
 297 /*
 298  * The set of specific options for a tracer. Your tracer
 299  * have to set the initial value of the flags val.
 300  */
 301 struct tracer_flags {
 302         u32                     val;
 303         struct tracer_opt       *opts;
 304 };
 305
 306 /* Makes more easy to define a tracer opt */
 307 #define TRACER_OPT(s, b)        .name = #s, .bit = b
 308
 309 /*
 310  * A specific tracer, represented by methods that operate on a trace array:
 311  */
 312 struct tracer {
 313         const char              *name;
 314         /* Your tracer should raise a warning if init fails */
 315         int                     (*init)(struct trace_array *tr);
 316         void                    (*reset)(struct trace_array *tr);
 317         void                    (*start)(struct trace_array *tr);
 318         void                    (*stop)(struct trace_array *tr);
 319         void                    (*open)(struct trace_iterator *iter);
 320         void                    (*pipe_open)(struct trace_iterator *iter);
 321         void                    (*close)(struct trace_iterator *iter);
 322         ssize_t                 (*read)(struct trace_iterator *iter,
 323                                         struct file *filp, char __user *ubuf,
 324                                         size_t cnt, loff_t *ppos);
 325 #ifdef CONFIG_FTRACE_STARTUP_TEST
 326         int                     (*selftest)(struct tracer *trace,
 327                                             struct trace_array *tr);
 328 #endif
 329         void                    (*print_header)(struct seq_file *m);
 330         enum print_line_t       (*print_line)(struct trace_iterator *iter);
 331         /* If you handled the flag setting, return 0 */
 332         int                     (*set_flag)(u32 old_flags, u32 bit, int set);
 333         struct tracer           *next;
 334         int                     print_max;
 335         struct tracer_flags     *flags;
 336 };
 337
 338 struct trace_seq {
 339         unsigned char           buffer[PAGE_SIZE];
 340         unsigned int            len;
 341         unsigned int            readpos;
 342 };
 343
 344 /*
 345  * Trace iterator - used by printout routines who present trace
 346  * results to users and which routines might sleep, etc:
 347  */
 348 struct trace_iterator {
 349         struct trace_array      *tr;
 350         struct tracer           *trace;
 351         void                    *private;
 352         struct ring_buffer_iter *buffer_iter[NR_CPUS];
 353
 354         /* The below is zeroed out in pipe_read */
 355         struct trace_seq        seq;
 356         struct trace_entry      *ent;
 357         int                     cpu;
 358         u64                     ts;
 359
 360         unsigned long           iter_flags;
 361         loff_t                  pos;
 362         long                    idx;
 363
 364         cpumask_t               started;
 365 };
 366
 367 int tracing_is_enabled(void);
 368 void trace_wake_up(void);
 369 void tracing_reset(struct trace_array *tr, int cpu);
 370 int tracing_open_generic(struct inode *inode, struct file *filp);
 371 struct dentry *tracing_init_dentry(void);
 372 void init_tracer_sysprof_debugfs(struct dentry *d_tracer);
 373
 374 struct trace_entry *tracing_get_trace_entry(struct trace_array *tr,
 375                                                 struct trace_array_cpu *data);
 376 void tracing_generic_entry_update(struct trace_entry *entry,
 377                                   unsigned long flags,
 378                                   int pc);
 379
 380 void ftrace(struct trace_array *tr,
 381                             struct trace_array_cpu *data,
 382                             unsigned long ip,
 383                             unsigned long parent_ip,
 384                             unsigned long flags, int pc);
 385 void tracing_sched_switch_trace(struct trace_array *tr,
 386                                 struct trace_array_cpu *data,
 387                                 struct task_struct *prev,
 388                                 struct task_struct *next,
 389                                 unsigned long flags, int pc);
 390 void tracing_record_cmdline(struct task_struct *tsk);
 391
 392 void tracing_sched_wakeup_trace(struct trace_array *tr,
 393                                 struct trace_array_cpu *data,
 394                                 struct task_struct *wakee,
 395                                 struct task_struct *cur,
 396                                 unsigned long flags, int pc);
 397 void trace_special(struct trace_array *tr,
 398                    struct trace_array_cpu *data,
 399                    unsigned long arg1,
 400                    unsigned long arg2,
 401                    unsigned long arg3, int pc);
 402 void trace_function(struct trace_array *tr,
 403                     struct trace_array_cpu *data,
 404                     unsigned long ip,
 405                     unsigned long parent_ip,
 406                     unsigned long flags, int pc);
 407 void
 408 trace_function_return(struct ftrace_retfunc *trace);
 409
 410 void trace_bts(struct trace_array *tr,
 411                unsigned long from,
 412                unsigned long to);
 413
 414 void tracing_start_cmdline_record(void);
 415 void tracing_stop_cmdline_record(void);
 416 void tracing_sched_switch_assign_trace(struct trace_array *tr);
 417 void tracing_stop_sched_switch_record(void);
 418 void tracing_start_sched_switch_record(void);
 419 int register_tracer(struct tracer *type);
 420 void unregister_tracer(struct tracer *type);
 421
 422 extern unsigned long nsecs_to_usecs(unsigned long nsecs);
 423
 424 extern unsigned long tracing_max_latency;
 425 extern unsigned long tracing_thresh;
 426
 427 void update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu);
 428 void update_max_tr_single(struct trace_array *tr,
 429                           struct task_struct *tsk, int cpu);
 430
 431 extern cycle_t ftrace_now(int cpu);
 432
 433 #ifdef CONFIG_FUNCTION_TRACER
 434 void tracing_start_function_trace(void);
 435 void tracing_stop_function_trace(void);
 436 #else
 437 # define tracing_start_function_trace()         do { } while (0)
 438 # define tracing_stop_function_trace()          do { } while (0)
 439 #endif
 440
 441 #ifdef CONFIG_CONTEXT_SWITCH_TRACER
 442 typedef void
 443 (*tracer_switch_func_t)(void *private,
 444                         void *__rq,
 445                         struct task_struct *prev,
 446                         struct task_struct *next);
 447
 448 struct tracer_switch_ops {
 449         tracer_switch_func_t            func;
 450         void                            *private;
 451         struct tracer_switch_ops        *next;
 452 };
 453
 454 #endif /* CONFIG_CONTEXT_SWITCH_TRACER */
 455
 456 #ifdef CONFIG_DYNAMIC_FTRACE
 457 extern unsigned long ftrace_update_tot_cnt;
 458 #define DYN_FTRACE_TEST_NAME trace_selftest_dynamic_test_func
 459 extern int DYN_FTRACE_TEST_NAME(void);
 460 #endif
 461
 462 #ifdef CONFIG_FTRACE_STARTUP_TEST
 463 extern int trace_selftest_startup_function(struct tracer *trace,
 464                                            struct trace_array *tr);
 465 extern int trace_selftest_startup_irqsoff(struct tracer *trace,
 466                                           struct trace_array *tr);
 467 extern int trace_selftest_startup_preemptoff(struct tracer *trace,
 468                                              struct trace_array *tr);
 469 extern int trace_selftest_startup_preemptirqsoff(struct tracer *trace,
 470                                                  struct trace_array *tr);
 471 extern int trace_selftest_startup_wakeup(struct tracer *trace,
 472                                          struct trace_array *tr);
 473 extern int trace_selftest_startup_nop(struct tracer *trace,
 474                                          struct trace_array *tr);
 475 extern int trace_selftest_startup_sched_switch(struct tracer *trace,
 476                                                struct trace_array *tr);
 477 extern int trace_selftest_startup_sysprof(struct tracer *trace,
 478                                                struct trace_array *tr);
 479 extern int trace_selftest_startup_branch(struct tracer *trace,
 480                                          struct trace_array *tr);
 481 #endif /* CONFIG_FTRACE_STARTUP_TEST */
 482
 483 extern void *head_page(struct trace_array_cpu *data);
 484 extern int trace_seq_printf(struct trace_seq *s, const char *fmt, ...);
 485 extern void trace_seq_print_cont(struct trace_seq *s,
 486                                  struct trace_iterator *iter);
 487
 488 extern int
 489 seq_print_ip_sym(struct trace_seq *s, unsigned long ip,
 490                 unsigned long sym_flags);
 491 extern ssize_t trace_seq_to_user(struct trace_seq *s, char __user *ubuf,
 492                                  size_t cnt);
 493 extern long ns2usecs(cycle_t nsec);
 494 extern int trace_vprintk(unsigned long ip, const char *fmt, va_list args);
 495
 496 extern unsigned long trace_flags;
 497
 498 /* Standard output formatting function used for function return traces */
 499 #ifdef CONFIG_FUNCTION_RET_TRACER
 500 extern enum print_line_t print_return_function(struct trace_iterator *iter);
 501 #else
 502 static inline enum print_line_t
 503 print_return_function(struct trace_iterator *iter)
 504 {
 505         return TRACE_TYPE_UNHANDLED;
 506 }
 507 #endif
 508
 509 /*
 510  * trace_iterator_flags is an enumeration that defines bit
 511  * positions into trace_flags that controls the output.
 512  *
 513  * NOTE: These bits must match the trace_options array in
 514  *       trace.c.
 515  */
 516 enum trace_iterator_flags {
 517         TRACE_ITER_PRINT_PARENT         = 0x01,
 518         TRACE_ITER_SYM_OFFSET           = 0x02,
 519         TRACE_ITER_SYM_ADDR             = 0x04,
 520         TRACE_ITER_VERBOSE              = 0x08,
 521         TRACE_ITER_RAW                  = 0x10,
 522         TRACE_ITER_HEX                  = 0x20,
 523         TRACE_ITER_BIN                  = 0x40,
 524         TRACE_ITER_BLOCK                = 0x80,
 525         TRACE_ITER_STACKTRACE           = 0x100,
 526         TRACE_ITER_SCHED_TREE           = 0x200,
 527         TRACE_ITER_PRINTK               = 0x400,
 528         TRACE_ITER_PREEMPTONLY          = 0x800,
 529         TRACE_ITER_BRANCH               = 0x1000,
 530         TRACE_ITER_ANNOTATE             = 0x2000,
 531         TRACE_ITER_USERSTACKTRACE       = 0x4000,
 532         TRACE_ITER_SYM_USEROBJ          = 0x8000
 533 };
 534
 535 /*
 536  * TRACE_ITER_SYM_MASK masks the options in trace_flags that
 537  * control the output of kernel symbols.
 538  */
 539 #define TRACE_ITER_SYM_MASK \
 540         (TRACE_ITER_PRINT_PARENT|TRACE_ITER_SYM_OFFSET|TRACE_ITER_SYM_ADDR)
 541
 542 extern struct tracer nop_trace;
 543
 544 /**
 545  * ftrace_preempt_disable - disable preemption scheduler safe
 546  *
 547  * When tracing can happen inside the scheduler, there exists
 548  * cases that the tracing might happen before the need_resched
 549  * flag is checked. If this happens and the tracer calls
 550  * preempt_enable (after a disable), a schedule might take place
 551  * causing an infinite recursion.
 552  *
 553  * To prevent this, we read the need_recshed flag before
 554  * disabling preemption. When we want to enable preemption we
 555  * check the flag, if it is set, then we call preempt_enable_no_resched.
 556  * Otherwise, we call preempt_enable.
 557  *
 558  * The rational for doing the above is that if need resched is set
 559  * and we have yet to reschedule, we are either in an atomic location
 560  * (where we do not need to check for scheduling) or we are inside
 561  * the scheduler and do not want to resched.
 562  */
 563 static inline int ftrace_preempt_disable(void)
 564 {
 565         int resched;
 566
 567         resched = need_resched();
 568         preempt_disable_notrace();
 569
 570         return resched;
 571 }
 572
 573 /**
 574  * ftrace_preempt_enable - enable preemption scheduler safe
 575  * @resched: the return value from ftrace_preempt_disable
 576  *
 577  * This is a scheduler safe way to enable preemption and not miss
 578  * any preemption checks. The disabled saved the state of preemption.
 579  * If resched is set, then we were either inside an atomic or
 580  * are inside the scheduler (we would have already scheduled
 581  * otherwise). In this case, we do not want to call normal
 582  * preempt_enable, but preempt_enable_no_resched instead.
 583  */
 584 static inline void ftrace_preempt_enable(int resched)
 585 {
 586         if (resched)
 587                 preempt_enable_no_resched_notrace();
 588         else
 589                 preempt_enable_notrace();
 590 }
 591
 592 #ifdef CONFIG_BRANCH_TRACER
 593 extern int enable_branch_tracing(struct trace_array *tr);
 594 extern void disable_branch_tracing(void);
 595 static inline int trace_branch_enable(struct trace_array *tr)
 596 {
 597         if (trace_flags & TRACE_ITER_BRANCH)
 598                 return enable_branch_tracing(tr);
 599         return 0;
 600 }
 601 static inline void trace_branch_disable(void)
 602 {
 603         /* due to races, always disable */
 604         disable_branch_tracing();
 605 }
 606 #else
 607 static inline int trace_branch_enable(struct trace_array *tr)
 608 {
 609         return 0;
 610 }
 611 static inline void trace_branch_disable(void)
 612 {
 613 }
 614 #endif /* CONFIG_BRANCH_TRACER */
 615
 616 #endif /* _LINUX_KERNEL_TRACE_H */