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