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
  12 enum trace_type {
  13         __TRACE_FIRST_TYPE = 0,
  14
  15         TRACE_FN,
  16         TRACE_CTX,
  17         TRACE_WAKE,
  18         TRACE_CONT,
  19         TRACE_STACK,
  20         TRACE_PRINT,
  21         TRACE_SPECIAL,
  22         TRACE_MMIO_RW,
  23         TRACE_MMIO_MAP,
  24         TRACE_BOOT,
  25
  26         __TRACE_LAST_TYPE
  27 };
  28
  29 /*
  30  * The trace entry - the most basic unit of tracing. This is what
  31  * is printed in the end as a single line in the trace output, such as:
  32  *
  33  *     bash-15816 [01]   235.197585: idle_cpu <- irq_enter
  34  */
  35 struct trace_entry {
  36         unsigned char           type;
  37         unsigned char           cpu;
  38         unsigned char           flags;
  39         unsigned char           preempt_count;
  40         int                     pid;
  41 };
  42
  43 /*
  44  * Function trace entry - function address and parent function addres:
  45  */
  46 struct ftrace_entry {
  47         struct trace_entry      ent;
  48         unsigned long           ip;
  49         unsigned long           parent_ip;
  50 };
  51 extern struct tracer boot_tracer;
  52
  53 /*
  54  * Context switch trace entry - which task (and prio) we switched from/to:
  55  */
  56 struct ctx_switch_entry {
  57         struct trace_entry      ent;
  58         unsigned int            prev_pid;
  59         unsigned char           prev_prio;
  60         unsigned char           prev_state;
  61         unsigned int            next_pid;
  62         unsigned char           next_prio;
  63         unsigned char           next_state;
  64         unsigned int            next_cpu;
  65 };
  66
  67 /*
  68  * Special (free-form) trace entry:
  69  */
  70 struct special_entry {
  71         struct trace_entry      ent;
  72         unsigned long           arg1;
  73         unsigned long           arg2;
  74         unsigned long           arg3;
  75 };
  76
  77 /*
  78  * Stack-trace entry:
  79  */
  80
  81 #define FTRACE_STACK_ENTRIES    8
  82
  83 struct stack_entry {
  84         struct trace_entry      ent;
  85         unsigned long           caller[FTRACE_STACK_ENTRIES];
  86 };
  87
  88 /*
  89  * ftrace_printk entry:
  90  */
  91 struct print_entry {
  92         struct trace_entry      ent;
  93         unsigned long           ip;
  94         char                    buf[];
  95 };
  96
  97 #define TRACE_OLD_SIZE          88
  98
  99 struct trace_field_cont {
 100         unsigned char           type;
 101         /* Temporary till we get rid of this completely */
 102         char                    buf[TRACE_OLD_SIZE - 1];
 103 };
 104
 105 struct trace_mmiotrace_rw {
 106         struct trace_entry      ent;
 107         struct mmiotrace_rw     rw;
 108 };
 109
 110 struct trace_mmiotrace_map {
 111         struct trace_entry      ent;
 112         struct mmiotrace_map    map;
 113 };
 114
 115 struct trace_boot {
 116         struct trace_entry      ent;
 117         struct boot_trace       initcall;
 118 };
 119
 120 /*
 121  * trace_flag_type is an enumeration that holds different
 122  * states when a trace occurs. These are:
 123  *  IRQS_OFF            - interrupts were disabled
 124  *  IRQS_NOSUPPORT      - arch does not support irqs_disabled_flags
 125  *  NEED_RESCED         - reschedule is requested
 126  *  HARDIRQ             - inside an interrupt handler
 127  *  SOFTIRQ             - inside a softirq handler
 128  *  CONT                - multiple entries hold the trace item
 129  */
 130 enum trace_flag_type {
 131         TRACE_FLAG_IRQS_OFF             = 0x01,
 132         TRACE_FLAG_IRQS_NOSUPPORT       = 0x02,
 133         TRACE_FLAG_NEED_RESCHED         = 0x04,
 134         TRACE_FLAG_HARDIRQ              = 0x08,
 135         TRACE_FLAG_SOFTIRQ              = 0x10,
 136         TRACE_FLAG_CONT                 = 0x20,
 137 };
 138
 139 #define TRACE_BUF_SIZE          1024
 140
 141 /*
 142  * The CPU trace array - it consists of thousands of trace entries
 143  * plus some other descriptor data: (for example which task started
 144  * the trace, etc.)
 145  */
 146 struct trace_array_cpu {
 147         atomic_t                disabled;
 148
 149         /* these fields get copied into max-trace: */
 150         unsigned long           trace_idx;
 151         unsigned long           overrun;
 152         unsigned long           saved_latency;
 153         unsigned long           critical_start;
 154         unsigned long           critical_end;
 155         unsigned long           critical_sequence;
 156         unsigned long           nice;
 157         unsigned long           policy;
 158         unsigned long           rt_priority;
 159         cycle_t                 preempt_timestamp;
 160         pid_t                   pid;
 161         uid_t                   uid;
 162         char                    comm[TASK_COMM_LEN];
 163 };
 164
 165 struct trace_iterator;
 166
 167 /*
 168  * The trace array - an array of per-CPU trace arrays. This is the
 169  * highest level data structure that individual tracers deal with.
 170  * They have on/off state as well:
 171  */
 172 struct trace_array {
 173         struct ring_buffer      *buffer;
 174         unsigned long           entries;
 175         long                    ctrl;
 176         int                     cpu;
 177         cycle_t                 time_start;
 178         struct task_struct      *waiter;
 179         struct trace_array_cpu  *data[NR_CPUS];
 180 };
 181
 182 #define FTRACE_CMP_TYPE(var, type) \
 183         __builtin_types_compatible_p(typeof(var), type *)
 184
 185 #undef IF_ASSIGN
 186 #define IF_ASSIGN(var, entry, etype, id)                \
 187         if (FTRACE_CMP_TYPE(var, etype)) {              \
 188                 var = (typeof(var))(entry);             \
 189                 WARN_ON(id && (entry)->type != id);     \
 190                 break;                                  \
 191         }
 192
 193 /* Will cause compile errors if type is not found. */
 194 extern void __ftrace_bad_type(void);
 195
 196 /*
 197  * The trace_assign_type is a verifier that the entry type is
 198  * the same as the type being assigned. To add new types simply
 199  * add a line with the following format:
 200  *
 201  * IF_ASSIGN(var, ent, type, id);
 202  *
 203  *  Where "type" is the trace type that includes the trace_entry
 204  *  as the "ent" item. And "id" is the trace identifier that is
 205  *  used in the trace_type enum.
 206  *
 207  *  If the type can have more than one id, then use zero.
 208  */
 209 #define trace_assign_type(var, ent)                                     \
 210         do {                                                            \
 211                 IF_ASSIGN(var, ent, struct ftrace_entry, TRACE_FN);     \
 212                 IF_ASSIGN(var, ent, struct ctx_switch_entry, 0);        \
 213                 IF_ASSIGN(var, ent, struct trace_field_cont, TRACE_CONT); \
 214                 IF_ASSIGN(var, ent, struct stack_entry, TRACE_STACK);   \
 215                 IF_ASSIGN(var, ent, struct print_entry, TRACE_PRINT);   \
 216                 IF_ASSIGN(var, ent, struct special_entry, 0);           \
 217                 IF_ASSIGN(var, ent, struct trace_mmiotrace_rw,          \
 218                           TRACE_MMIO_RW);                               \
 219                 IF_ASSIGN(var, ent, struct trace_mmiotrace_map,         \
 220                           TRACE_MMIO_MAP);                              \
 221                 IF_ASSIGN(var, ent, struct trace_boot, TRACE_BOOT);     \
 222                 __ftrace_bad_type();                                    \
 223         } while (0)
 224
 225 /* Return values for print_line callback */
 226 enum print_line_t {
 227         TRACE_TYPE_PARTIAL_LINE = 0,    /* Retry after flushing the seq */
 228         TRACE_TYPE_HANDLED      = 1,
 229         TRACE_TYPE_UNHANDLED    = 2     /* Relay to other output functions */
 230 };
 231
 232 /*
 233  * A specific tracer, represented by methods that operate on a trace array:
 234  */
 235 struct tracer {
 236         const char              *name;
 237         void                    (*init)(struct trace_array *tr);
 238         void                    (*reset)(struct trace_array *tr);
 239         void                    (*start)(struct trace_array *tr);
 240         void                    (*stop)(struct trace_array *tr);
 241         void                    (*open)(struct trace_iterator *iter);
 242         void                    (*pipe_open)(struct trace_iterator *iter);
 243         void                    (*close)(struct trace_iterator *iter);
 244         ssize_t                 (*read)(struct trace_iterator *iter,
 245                                         struct file *filp, char __user *ubuf,
 246                                         size_t cnt, loff_t *ppos);
 247 #ifdef CONFIG_FTRACE_STARTUP_TEST
 248         int                     (*selftest)(struct tracer *trace,
 249                                             struct trace_array *tr);
 250 #endif
 251         enum print_line_t       (*print_line)(struct trace_iterator *iter);
 252         struct tracer           *next;
 253         int                     print_max;
 254 };
 255
 256 struct trace_seq {
 257         unsigned char           buffer[PAGE_SIZE];
 258         unsigned int            len;
 259         unsigned int            readpos;
 260 };
 261
 262 /*
 263  * Trace iterator - used by printout routines who present trace
 264  * results to users and which routines might sleep, etc:
 265  */
 266 struct trace_iterator {
 267         struct trace_array      *tr;
 268         struct tracer           *trace;
 269         void                    *private;
 270         struct ring_buffer_iter *buffer_iter[NR_CPUS];
 271
 272         /* The below is zeroed out in pipe_read */
 273         struct trace_seq        seq;
 274         struct trace_entry      *ent;
 275         int                     cpu;
 276         u64                     ts;
 277
 278         unsigned long           iter_flags;
 279         loff_t                  pos;
 280         long                    idx;
 281 };
 282
 283 int tracing_is_enabled(void);
 284 void trace_wake_up(void);
 285 void tracing_reset(struct trace_array *tr, int cpu);
 286 int tracing_open_generic(struct inode *inode, struct file *filp);
 287 struct dentry *tracing_init_dentry(void);
 288 void init_tracer_sysprof_debugfs(struct dentry *d_tracer);
 289
 290 struct trace_entry *tracing_get_trace_entry(struct trace_array *tr,
 291                                                 struct trace_array_cpu *data);
 292 void tracing_generic_entry_update(struct trace_entry *entry,
 293                                   unsigned long flags,
 294                                   int pc);
 295
 296 void ftrace(struct trace_array *tr,
 297                             struct trace_array_cpu *data,
 298                             unsigned long ip,
 299                             unsigned long parent_ip,
 300                             unsigned long flags, int pc);
 301 void tracing_sched_switch_trace(struct trace_array *tr,
 302                                 struct trace_array_cpu *data,
 303                                 struct task_struct *prev,
 304                                 struct task_struct *next,
 305                                 unsigned long flags, int pc);
 306 void tracing_record_cmdline(struct task_struct *tsk);
 307
 308 void tracing_sched_wakeup_trace(struct trace_array *tr,
 309                                 struct trace_array_cpu *data,
 310                                 struct task_struct *wakee,
 311                                 struct task_struct *cur,
 312                                 unsigned long flags, int pc);
 313 void trace_special(struct trace_array *tr,
 314                    struct trace_array_cpu *data,
 315                    unsigned long arg1,
 316                    unsigned long arg2,
 317                    unsigned long arg3, int pc);
 318 void trace_function(struct trace_array *tr,
 319                     struct trace_array_cpu *data,
 320                     unsigned long ip,
 321                     unsigned long parent_ip,
 322                     unsigned long flags, int pc);
 323
 324 void tracing_start_cmdline_record(void);
 325 void tracing_stop_cmdline_record(void);
 326 void tracing_sched_switch_assign_trace(struct trace_array *tr);
 327 void tracing_stop_sched_switch_record(void);
 328 void tracing_start_sched_switch_record(void);
 329 int register_tracer(struct tracer *type);
 330 void unregister_tracer(struct tracer *type);
 331
 332 extern unsigned long nsecs_to_usecs(unsigned long nsecs);
 333
 334 extern unsigned long tracing_max_latency;
 335 extern unsigned long tracing_thresh;
 336
 337 void update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu);
 338 void update_max_tr_single(struct trace_array *tr,
 339                           struct task_struct *tsk, int cpu);
 340
 341 extern cycle_t ftrace_now(int cpu);
 342
 343 #ifdef CONFIG_FUNCTION_TRACER
 344 void tracing_start_function_trace(void);
 345 void tracing_stop_function_trace(void);
 346 #else
 347 # define tracing_start_function_trace()         do { } while (0)
 348 # define tracing_stop_function_trace()          do { } while (0)
 349 #endif
 350
 351 #ifdef CONFIG_CONTEXT_SWITCH_TRACER
 352 typedef void
 353 (*tracer_switch_func_t)(void *private,
 354                         void *__rq,
 355                         struct task_struct *prev,
 356                         struct task_struct *next);
 357
 358 struct tracer_switch_ops {
 359         tracer_switch_func_t            func;
 360         void                            *private;
 361         struct tracer_switch_ops        *next;
 362 };
 363
 364 #endif /* CONFIG_CONTEXT_SWITCH_TRACER */
 365
 366 #ifdef CONFIG_DYNAMIC_FTRACE
 367 extern unsigned long ftrace_update_tot_cnt;
 368 #define DYN_FTRACE_TEST_NAME trace_selftest_dynamic_test_func
 369 extern int DYN_FTRACE_TEST_NAME(void);
 370 #endif
 371
 372 #ifdef CONFIG_FTRACE_STARTUP_TEST
 373 extern int trace_selftest_startup_function(struct tracer *trace,
 374                                            struct trace_array *tr);
 375 extern int trace_selftest_startup_irqsoff(struct tracer *trace,
 376                                           struct trace_array *tr);
 377 extern int trace_selftest_startup_preemptoff(struct tracer *trace,
 378                                              struct trace_array *tr);
 379 extern int trace_selftest_startup_preemptirqsoff(struct tracer *trace,
 380                                                  struct trace_array *tr);
 381 extern int trace_selftest_startup_wakeup(struct tracer *trace,
 382                                          struct trace_array *tr);
 383 extern int trace_selftest_startup_nop(struct tracer *trace,
 384                                          struct trace_array *tr);
 385 extern int trace_selftest_startup_sched_switch(struct tracer *trace,
 386                                                struct trace_array *tr);
 387 extern int trace_selftest_startup_sysprof(struct tracer *trace,
 388                                                struct trace_array *tr);
 389 #endif /* CONFIG_FTRACE_STARTUP_TEST */
 390
 391 extern void *head_page(struct trace_array_cpu *data);
 392 extern int trace_seq_printf(struct trace_seq *s, const char *fmt, ...);
 393 extern void trace_seq_print_cont(struct trace_seq *s,
 394                                  struct trace_iterator *iter);
 395 extern ssize_t trace_seq_to_user(struct trace_seq *s, char __user *ubuf,
 396                                  size_t cnt);
 397 extern long ns2usecs(cycle_t nsec);
 398 extern int trace_vprintk(unsigned long ip, const char *fmt, va_list args);
 399
 400 extern unsigned long trace_flags;
 401
 402 /*
 403  * trace_iterator_flags is an enumeration that defines bit
 404  * positions into trace_flags that controls the output.
 405  *
 406  * NOTE: These bits must match the trace_options array in
 407  *       trace.c.
 408  */
 409 enum trace_iterator_flags {
 410         TRACE_ITER_PRINT_PARENT         = 0x01,
 411         TRACE_ITER_SYM_OFFSET           = 0x02,
 412         TRACE_ITER_SYM_ADDR             = 0x04,
 413         TRACE_ITER_VERBOSE              = 0x08,
 414         TRACE_ITER_RAW                  = 0x10,
 415         TRACE_ITER_HEX                  = 0x20,
 416         TRACE_ITER_BIN                  = 0x40,
 417         TRACE_ITER_BLOCK                = 0x80,
 418         TRACE_ITER_STACKTRACE           = 0x100,
 419         TRACE_ITER_SCHED_TREE           = 0x200,
 420         TRACE_ITER_PRINTK               = 0x400,
 421         TRACE_ITER_PREEMPTONLY          = 0x800,
 422 };
 423
 424 extern struct tracer nop_trace;
 425
 426 /**
 427  * ftrace_preempt_disable - disable preemption scheduler safe
 428  *
 429  * When tracing can happen inside the scheduler, there exists
 430  * cases that the tracing might happen before the need_resched
 431  * flag is checked. If this happens and the tracer calls
 432  * preempt_enable (after a disable), a schedule might take place
 433  * causing an infinite recursion.
 434  *
 435  * To prevent this, we read the need_recshed flag before
 436  * disabling preemption. When we want to enable preemption we
 437  * check the flag, if it is set, then we call preempt_enable_no_resched.
 438  * Otherwise, we call preempt_enable.
 439  *
 440  * The rational for doing the above is that if need resched is set
 441  * and we have yet to reschedule, we are either in an atomic location
 442  * (where we do not need to check for scheduling) or we are inside
 443  * the scheduler and do not want to resched.
 444  */
 445 static inline int ftrace_preempt_disable(void)
 446 {
 447         int resched;
 448
 449         resched = need_resched();
 450         preempt_disable_notrace();
 451
 452         return resched;
 453 }
 454
 455 /**
 456  * ftrace_preempt_enable - enable preemption scheduler safe
 457  * @resched: the return value from ftrace_preempt_disable
 458  *
 459  * This is a scheduler safe way to enable preemption and not miss
 460  * any preemption checks. The disabled saved the state of preemption.
 461  * If resched is set, then we were either inside an atomic or
 462  * are inside the scheduler (we would have already scheduled
 463  * otherwise). In this case, we do not want to call normal
 464  * preempt_enable, but preempt_enable_no_resched instead.
 465  */
 466 static inline void ftrace_preempt_enable(int resched)
 467 {
 468         if (resched)
 469                 preempt_enable_no_resched_notrace();
 470         else
 471                 preempt_enable_notrace();
 472 }
 473
 474 #endif /* _LINUX_KERNEL_TRACE_H */