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