1 /* SPDX-License-Identifier: GPL-2.0 */ 2 /* 3 * If TRACE_SYSTEM is defined, that will be the directory created 4 * in the ftrace directory under /sys/kernel/tracing/events/<system> 5 * 6 * The define_trace.h below will also look for a file name of 7 * TRACE_SYSTEM.h where TRACE_SYSTEM is what is defined here. 8 * In this case, it would look for sample-trace.h 9 * 10 * If the header name will be different than the system name 11 * (as in this case), then you can override the header name that 12 * define_trace.h will look up by defining TRACE_INCLUDE_FILE 13 * 14 * This file is called trace-events-sample.h but we want the system 15 * to be called "sample-trace". Therefore we must define the name of this 16 * file: 17 * 18 * #define TRACE_INCLUDE_FILE trace-events-sample 19 * 20 * As we do an the bottom of this file. 21 * 22 * Notice that TRACE_SYSTEM should be defined outside of #if 23 * protection, just like TRACE_INCLUDE_FILE. 24 */ 25 #undef TRACE_SYSTEM 26 #define TRACE_SYSTEM sample-trace 27 28 /* 29 * TRACE_SYSTEM is expected to be a C valid variable (alpha-numeric 30 * and underscore), although it may start with numbers. If for some 31 * reason it is not, you need to add the following lines: 32 */ 33 #undef TRACE_SYSTEM_VAR 34 #define TRACE_SYSTEM_VAR sample_trace 35 /* 36 * But the above is only needed if TRACE_SYSTEM is not alpha-numeric 37 * and underscored. By default, TRACE_SYSTEM_VAR will be equal to 38 * TRACE_SYSTEM. As TRACE_SYSTEM_VAR must be alpha-numeric, if 39 * TRACE_SYSTEM is not, then TRACE_SYSTEM_VAR must be defined with 40 * only alpha-numeric and underscores. 41 * 42 * The TRACE_SYSTEM_VAR is only used internally and not visible to 43 * user space. 44 */ 45 46 /* 47 * Notice that this file is not protected like a normal header. 48 * We also must allow for rereading of this file. The 49 * 50 * || defined(TRACE_HEADER_MULTI_READ) 51 * 52 * serves this purpose. 53 */ 54 #if !defined(_TRACE_EVENT_SAMPLE_H) || defined(TRACE_HEADER_MULTI_READ) 55 #define _TRACE_EVENT_SAMPLE_H 56 57 /* 58 * All trace headers should include tracepoint.h, until we finally 59 * make it into a standard header. 60 */ 61 #include <linux/tracepoint.h> 62 63 /* 64 * The TRACE_EVENT macro is broken up into 5 parts. 65 * 66 * name: name of the trace point. This is also how to enable the tracepoint. 67 * A function called trace_foo_bar() will be created. 68 * 69 * proto: the prototype of the function trace_foo_bar() 70 * Here it is trace_foo_bar(char *foo, int bar). 71 * 72 * args: must match the arguments in the prototype. 73 * Here it is simply "foo, bar". 74 * 75 * struct: This defines the way the data will be stored in the ring buffer. 76 * The items declared here become part of a special structure 77 * called "__entry", which can be used in the fast_assign part of the 78 * TRACE_EVENT macro. 79 * 80 * Here are the currently defined types you can use: 81 * 82 * __field : Is broken up into type and name. Where type can be any 83 * primitive type (integer, long or pointer). 84 * 85 * __field(int, foo) 86 * 87 * __entry->foo = 5; 88 * 89 * __field_struct : This can be any static complex data type (struct, union 90 * but not an array). Be careful using complex types, as each 91 * event is limited in size, and copying large amounts of data 92 * into the ring buffer can slow things down. 93 * 94 * __field_struct(struct bar, foo) 95 * 96 * __entry->bar.x = y; 97 98 * __array: There are three fields (type, name, size). The type is the 99 * type of elements in the array, the name is the name of the array. 100 * size is the number of items in the array (not the total size). 101 * 102 * __array( char, foo, 10) is the same as saying: char foo[10]; 103 * 104 * Assigning arrays can be done like any array: 105 * 106 * __entry->foo[0] = 'a'; 107 * 108 * memcpy(__entry->foo, bar, 10); 109 * 110 * __dynamic_array: This is similar to array, but can vary its size from 111 * instance to instance of the tracepoint being called. 112 * Like __array, this too has three elements (type, name, size); 113 * type is the type of the element, name is the name of the array. 114 * The size is different than __array. It is not a static number, 115 * but the algorithm to figure out the length of the array for the 116 * specific instance of tracepoint. Again, size is the number of 117 * items in the array, not the total length in bytes. 118 * 119 * __dynamic_array( int, foo, bar) is similar to: int foo[bar]; 120 * 121 * Note, unlike arrays, you must use the __get_dynamic_array() macro 122 * to access the array. 123 * 124 * memcpy(__get_dynamic_array(foo), bar, 10); 125 * 126 * Notice, that "__entry" is not needed here. 127 * 128 * __string: This is a special kind of __dynamic_array. It expects to 129 * have a null terminated character array passed to it (it allows 130 * for NULL too, which would be converted into "(null)"). __string 131 * takes two parameter (name, src), where name is the name of 132 * the string saved, and src is the string to copy into the 133 * ring buffer. 134 * 135 * __string(foo, bar) is similar to: strcpy(foo, bar) 136 * 137 * To assign a string, use the helper macro __assign_str(). 138 * 139 * __assign_str(foo); 140 * 141 * The __string() macro saves off the string that is passed into 142 * the second parameter, and the __assign_str() will store than 143 * saved string into the "foo" field. 144 * 145 * __vstring: This is similar to __string() but instead of taking a 146 * dynamic length, it takes a variable list va_list 'va' variable. 147 * Some event callers already have a message from parameters saved 148 * in a va_list. Passing in the format and the va_list variable 149 * will save just enough on the ring buffer for that string. 150 * Note, the va variable used is a pointer to a va_list, not 151 * to the va_list directly. 152 * 153 * (va_list *va) 154 * 155 * __vstring(foo, fmt, va) is similar to: vsnprintf(foo, fmt, va) 156 * 157 * To assign the string, use the helper macro __assign_vstr(). 158 * 159 * __assign_vstr(foo, fmt, va); 160 * 161 * In most cases, the __assign_vstr() macro will take the same 162 * parameters as the __vstring() macro had to declare the string. 163 * Use __get_str() to retrieve the __vstring() just like it would for 164 * __string(). 165 * 166 * __string_len: This is a helper to a __dynamic_array, but it understands 167 * that the array has characters in it, it will allocate 'len' + 1 bytes 168 * in the ring buffer and add a '\0' to the string. This is 169 * useful if the string being saved has no terminating '\0' byte. 170 * It requires that the length of the string is known as it acts 171 * like a memcpy(). 172 * 173 * Declared with: 174 * 175 * __string_len(foo, bar, len) 176 * 177 * To assign this string, use the helper macro __assign_str(). 178 * The length is saved via the __string_len() and is retrieved in 179 * __assign_str(). 180 * 181 * __assign_str(foo); 182 * 183 * Then len + 1 is allocated to the ring buffer, and a nul terminating 184 * byte is added. This is similar to: 185 * 186 * memcpy(__get_str(foo), bar, len); 187 * __get_str(foo)[len] = 0; 188 * 189 * The advantage of using this over __dynamic_array, is that it 190 * takes care of allocating the extra byte on the ring buffer 191 * for the '\0' terminating byte, and __get_str(foo) can be used 192 * in the TP_printk(). 193 * 194 * __bitmask: This is another kind of __dynamic_array, but it expects 195 * an array of longs, and the number of bits to parse. It takes 196 * two parameters (name, nr_bits), where name is the name of the 197 * bitmask to save, and the nr_bits is the number of bits to record. 198 * 199 * __bitmask(target_cpu, nr_cpumask_bits) 200 * 201 * To assign a bitmask, use the __assign_bitmask() helper macro. 202 * 203 * __assign_bitmask(target_cpus, cpumask_bits(bar), nr_cpumask_bits); 204 * 205 * __cpumask: This is pretty much the same as __bitmask but is specific for 206 * CPU masks. The type displayed to the user via the format files will 207 * be "cpumaks_t" such that user space may deal with them differently 208 * if they choose to do so, and the bits is always set to nr_cpumask_bits. 209 * 210 * __cpumask(target_cpu) 211 * 212 * To assign a cpumask, use the __assign_cpumask() helper macro. 213 * 214 * __assign_cpumask(target_cpus, cpumask_bits(bar)); 215 * 216 * fast_assign: This is a C like function that is used to store the items 217 * into the ring buffer. A special variable called "__entry" will be the 218 * structure that points into the ring buffer and has the same fields as 219 * described by the struct part of TRACE_EVENT above. 220 * 221 * printk: This is a way to print out the data in pretty print. This is 222 * useful if the system crashes and you are logging via a serial line, 223 * the data can be printed to the console using this "printk" method. 224 * This is also used to print out the data from the trace files. 225 * Again, the __entry macro is used to access the data from the ring buffer. 226 * 227 * Note, __dynamic_array, __string, __bitmask and __cpumask require special 228 * helpers to access the data. 229 * 230 * For __dynamic_array(int, foo, bar) use __get_dynamic_array(foo) 231 * Use __get_dynamic_array_len(foo) to get the length of the array 232 * saved. Note, __get_dynamic_array_len() returns the total allocated 233 * length of the dynamic array; __print_array() expects the second 234 * parameter to be the number of elements. To get that, the array length 235 * needs to be divided by the element size. 236 * 237 * For __string(foo, bar) use __get_str(foo) 238 * 239 * For __bitmask(target_cpus, nr_cpumask_bits) use __get_bitmask(target_cpus) 240 * 241 * For __cpumask(target_cpus) use __get_cpumask(target_cpus) 242 * 243 * 244 * Note, that for both the assign and the printk, __entry is the handler 245 * to the data structure in the ring buffer, and is defined by the 246 * TP_STRUCT__entry. 247 */ 248 249 /* 250 * It is OK to have helper functions in the file, but they need to be protected 251 * from being defined more than once. Remember, this file gets included more 252 * than once. 253 */ 254 #ifndef __TRACE_EVENT_SAMPLE_HELPER_FUNCTIONS 255 #define __TRACE_EVENT_SAMPLE_HELPER_FUNCTIONS 256 static inline int __length_of(const int *list) 257 { 258 int i; 259 260 if (!list) 261 return 0; 262 263 for (i = 0; list[i]; i++) 264 ; 265 return i; 266 } 267 268 enum { 269 TRACE_SAMPLE_FOO = 2, 270 TRACE_SAMPLE_BAR = 4, 271 TRACE_SAMPLE_ZOO = 8, 272 }; 273 #endif 274 275 /* 276 * If enums are used in the TP_printk(), their names will be shown in 277 * format files and not their values. This can cause problems with user 278 * space programs that parse the format files to know how to translate 279 * the raw binary trace output into human readable text. 280 * 281 * To help out user space programs, any enum that is used in the TP_printk() 282 * should be defined by TRACE_DEFINE_ENUM() macro. All that is needed to 283 * be done is to add this macro with the enum within it in the trace 284 * header file, and it will be converted in the output. 285 */ 286 287 TRACE_DEFINE_ENUM(TRACE_SAMPLE_FOO); 288 TRACE_DEFINE_ENUM(TRACE_SAMPLE_BAR); 289 TRACE_DEFINE_ENUM(TRACE_SAMPLE_ZOO); 290 291 TRACE_EVENT(foo_bar, 292 293 TP_PROTO(const char *foo, int bar, const int *lst, 294 const char *string, const struct cpumask *mask, 295 const char *fmt, va_list *va), 296 297 TP_ARGS(foo, bar, lst, string, mask, fmt, va), 298 299 TP_STRUCT__entry( 300 __array( char, foo, 10 ) 301 __field( int, bar ) 302 __dynamic_array(int, list, __length_of(lst)) 303 __string( str, string ) 304 __bitmask( cpus, num_possible_cpus() ) 305 __cpumask( cpum ) 306 __vstring( vstr, fmt, va ) 307 __string_len( lstr, foo, bar / 2 < strlen(foo) ? bar / 2 : strlen(foo) ) 308 ), 309 310 TP_fast_assign( 311 strscpy(__entry->foo, foo, 10); 312 __entry->bar = bar; 313 memcpy(__get_dynamic_array(list), lst, 314 __length_of(lst) * sizeof(int)); 315 __assign_str(str); 316 __assign_str(lstr); 317 __assign_vstr(vstr, fmt, va); 318 __assign_bitmask(cpus, cpumask_bits(mask), num_possible_cpus()); 319 __assign_cpumask(cpum, cpumask_bits(mask)); 320 ), 321 322 TP_printk("foo %s %d %s %s %s %s %s (%s) (%s) %s", __entry->foo, __entry->bar, 323 324 /* 325 * Notice here the use of some helper functions. This includes: 326 * 327 * __print_symbolic( variable, { value, "string" }, ... ), 328 * 329 * The variable is tested against each value of the { } pair. If 330 * the variable matches one of the values, then it will print the 331 * string in that pair. If non are matched, it returns a string 332 * version of the number (if __entry->bar == 7 then "7" is returned). 333 */ 334 __print_symbolic(__entry->bar, 335 { 0, "zero" }, 336 { TRACE_SAMPLE_FOO, "TWO" }, 337 { TRACE_SAMPLE_BAR, "FOUR" }, 338 { TRACE_SAMPLE_ZOO, "EIGHT" }, 339 { 10, "TEN" } 340 ), 341 342 /* 343 * __print_flags( variable, "delim", { value, "flag" }, ... ), 344 * 345 * This is similar to __print_symbolic, except that it tests the bits 346 * of the value. If ((FLAG & variable) == FLAG) then the string is 347 * printed. If more than one flag matches, then each one that does is 348 * also printed with delim in between them. 349 * If not all bits are accounted for, then the not found bits will be 350 * added in hex format: 0x506 will show BIT2|BIT4|0x500 351 */ 352 __print_flags(__entry->bar, "|", 353 { 1, "BIT1" }, 354 { 2, "BIT2" }, 355 { 4, "BIT3" }, 356 { 8, "BIT4" } 357 ), 358 /* 359 * __print_array( array, len, element_size ) 360 * 361 * This prints out the array that is defined by __array in a nice format. 362 */ 363 __print_array(__get_dynamic_array(list), 364 __get_dynamic_array_len(list) / sizeof(int), 365 sizeof(int)), 366 __get_str(str), __get_str(lstr), 367 __get_bitmask(cpus), __get_cpumask(cpum), 368 __get_str(vstr)) 369 ); 370 371 /* 372 * There may be a case where a tracepoint should only be called if 373 * some condition is set. Otherwise the tracepoint should not be called. 374 * But to do something like: 375 * 376 * if (cond) 377 * trace_foo(); 378 * 379 * Would cause a little overhead when tracing is not enabled, and that 380 * overhead, even if small, is not something we want. As tracepoints 381 * use static branch (aka jump_labels), where no branch is taken to 382 * skip the tracepoint when not enabled, and a jmp is placed to jump 383 * to the tracepoint code when it is enabled, having a if statement 384 * nullifies that optimization. It would be nice to place that 385 * condition within the static branch. This is where TRACE_EVENT_CONDITION 386 * comes in. 387 * 388 * TRACE_EVENT_CONDITION() is just like TRACE_EVENT, except it adds another 389 * parameter just after args. Where TRACE_EVENT has: 390 * 391 * TRACE_EVENT(name, proto, args, struct, assign, printk) 392 * 393 * the CONDITION version has: 394 * 395 * TRACE_EVENT_CONDITION(name, proto, args, cond, struct, assign, printk) 396 * 397 * Everything is the same as TRACE_EVENT except for the new cond. Think 398 * of the cond variable as: 399 * 400 * if (cond) 401 * trace_foo_bar_with_cond(); 402 * 403 * Except that the logic for the if branch is placed after the static branch. 404 * That is, the if statement that processes the condition will not be 405 * executed unless that traecpoint is enabled. Otherwise it still remains 406 * a nop. 407 */ 408 TRACE_EVENT_CONDITION(foo_bar_with_cond, 409 410 TP_PROTO(const char *foo, int bar), 411 412 TP_ARGS(foo, bar), 413 414 TP_CONDITION(!(bar % 10)), 415 416 TP_STRUCT__entry( 417 __string( foo, foo ) 418 __field( int, bar ) 419 ), 420 421 TP_fast_assign( 422 __assign_str(foo); 423 __entry->bar = bar; 424 ), 425 426 TP_printk("foo %s %d", __get_str(foo), __entry->bar) 427 ); 428 429 int foo_bar_reg(void); 430 void foo_bar_unreg(void); 431 432 /* 433 * Now in the case that some function needs to be called when the 434 * tracepoint is enabled and/or when it is disabled, the 435 * TRACE_EVENT_FN() serves this purpose. This is just like TRACE_EVENT() 436 * but adds two more parameters at the end: 437 * 438 * TRACE_EVENT_FN( name, proto, args, struct, assign, printk, reg, unreg) 439 * 440 * reg and unreg are functions with the prototype of: 441 * 442 * void reg(void) 443 * 444 * The reg function gets called before the tracepoint is enabled, and 445 * the unreg function gets called after the tracepoint is disabled. 446 * 447 * Note, reg and unreg are allowed to be NULL. If you only need to 448 * call a function before enabling, or after disabling, just set one 449 * function and pass in NULL for the other parameter. 450 */ 451 TRACE_EVENT_FN(foo_bar_with_fn, 452 453 TP_PROTO(const char *foo, int bar), 454 455 TP_ARGS(foo, bar), 456 457 TP_STRUCT__entry( 458 __string( foo, foo ) 459 __field( int, bar ) 460 ), 461 462 TP_fast_assign( 463 __assign_str(foo); 464 __entry->bar = bar; 465 ), 466 467 TP_printk("foo %s %d", __get_str(foo), __entry->bar), 468 469 foo_bar_reg, foo_bar_unreg 470 ); 471 472 /* 473 * Each TRACE_EVENT macro creates several helper functions to produce 474 * the code to add the tracepoint, create the files in the trace 475 * directory, hook it to perf, assign the values and to print out 476 * the raw data from the ring buffer. To prevent too much bloat, 477 * if there are more than one tracepoint that uses the same format 478 * for the proto, args, struct, assign and printk, and only the name 479 * is different, it is highly recommended to use the DECLARE_EVENT_CLASS 480 * 481 * DECLARE_EVENT_CLASS() macro creates most of the functions for the 482 * tracepoint. Then DEFINE_EVENT() is use to hook a tracepoint to those 483 * functions. This DEFINE_EVENT() is an instance of the class and can 484 * be enabled and disabled separately from other events (either TRACE_EVENT 485 * or other DEFINE_EVENT()s). 486 * 487 * Note, TRACE_EVENT() itself is simply defined as: 488 * 489 * #define TRACE_EVENT(name, proto, args, tstruct, assign, printk) \ 490 * DECLARE_EVENT_CLASS(name, proto, args, tstruct, assign, printk); \ 491 * DEFINE_EVENT(name, name, proto, args) 492 * 493 * The DEFINE_EVENT() also can be declared with conditions and reg functions: 494 * 495 * DEFINE_EVENT_CONDITION(template, name, proto, args, cond); 496 * DEFINE_EVENT_FN(template, name, proto, args, reg, unreg); 497 */ 498 DECLARE_EVENT_CLASS(foo_template, 499 500 TP_PROTO(const char *foo, int bar), 501 502 TP_ARGS(foo, bar), 503 504 TP_STRUCT__entry( 505 __string( foo, foo ) 506 __field( int, bar ) 507 ), 508 509 TP_fast_assign( 510 __assign_str(foo); 511 __entry->bar = bar; 512 ), 513 514 TP_printk("foo %s %d", __get_str(foo), __entry->bar) 515 ); 516 517 /* 518 * Here's a better way for the previous samples (except, the first 519 * example had more fields and could not be used here). 520 */ 521 DEFINE_EVENT(foo_template, foo_with_template_simple, 522 TP_PROTO(const char *foo, int bar), 523 TP_ARGS(foo, bar)); 524 525 DEFINE_EVENT_CONDITION(foo_template, foo_with_template_cond, 526 TP_PROTO(const char *foo, int bar), 527 TP_ARGS(foo, bar), 528 TP_CONDITION(!(bar % 8))); 529 530 531 DEFINE_EVENT_FN(foo_template, foo_with_template_fn, 532 TP_PROTO(const char *foo, int bar), 533 TP_ARGS(foo, bar), 534 foo_bar_reg, foo_bar_unreg); 535 536 /* 537 * Anytime two events share basically the same values and have 538 * the same output, use the DECLARE_EVENT_CLASS() and DEFINE_EVENT() 539 * when ever possible. 540 */ 541 542 /* 543 * If the event is similar to the DECLARE_EVENT_CLASS, but you need 544 * to have a different output, then use DEFINE_EVENT_PRINT() which 545 * lets you override the TP_printk() of the class. 546 */ 547 548 DEFINE_EVENT_PRINT(foo_template, foo_with_template_print, 549 TP_PROTO(const char *foo, int bar), 550 TP_ARGS(foo, bar), 551 TP_printk("bar %s %d", __get_str(foo), __entry->bar)); 552 553 /* 554 * There are yet another __rel_loc dynamic data attribute. If you 555 * use __rel_dynamic_array() and __rel_string() etc. macros, you 556 * can use this attribute. There is no difference from the viewpoint 557 * of functionality with/without 'rel' but the encoding is a bit 558 * different. This is expected to be used with user-space event, 559 * there is no reason that the kernel event use this, but only for 560 * testing. 561 */ 562 563 TRACE_EVENT(foo_rel_loc, 564 565 TP_PROTO(const char *foo, int bar, unsigned long *mask, const cpumask_t *cpus), 566 567 TP_ARGS(foo, bar, mask, cpus), 568 569 TP_STRUCT__entry( 570 __rel_string( foo, foo ) 571 __field( int, bar ) 572 __rel_bitmask( bitmask, 573 BITS_PER_BYTE * sizeof(unsigned long) ) 574 __rel_cpumask( cpumask ) 575 ), 576 577 TP_fast_assign( 578 __assign_rel_str(foo); 579 __entry->bar = bar; 580 __assign_rel_bitmask(bitmask, mask, 581 BITS_PER_BYTE * sizeof(unsigned long)); 582 __assign_rel_cpumask(cpumask, cpus); 583 ), 584 585 TP_printk("foo_rel_loc %s, %d, %s, %s", __get_rel_str(foo), __entry->bar, 586 __get_rel_bitmask(bitmask), 587 __get_rel_cpumask(cpumask)) 588 ); 589 #endif 590 591 /***** NOTICE! The #if protection ends here. *****/ 592 593 594 /* 595 * There are several ways I could have done this. If I left out the 596 * TRACE_INCLUDE_PATH, then it would default to the kernel source 597 * include/trace/events directory. 598 * 599 * I could specify a path from the define_trace.h file back to this 600 * file. 601 * 602 * #define TRACE_INCLUDE_PATH ../../samples/trace_events 603 * 604 * But the safest and easiest way to simply make it use the directory 605 * that the file is in is to add in the Makefile: 606 * 607 * CFLAGS_trace-events-sample.o := -I$(src) 608 * 609 * This will make sure the current path is part of the include 610 * structure for our file so that define_trace.h can find it. 611 * 612 * I could have made only the top level directory the include: 613 * 614 * CFLAGS_trace-events-sample.o := -I$(PWD) 615 * 616 * And then let the path to this directory be the TRACE_INCLUDE_PATH: 617 * 618 * #define TRACE_INCLUDE_PATH samples/trace_events 619 * 620 * But then if something defines "samples" or "trace_events" as a macro 621 * then we could risk that being converted too, and give us an unexpected 622 * result. 623 */ 624 #undef TRACE_INCLUDE_PATH 625 #undef TRACE_INCLUDE_FILE 626 #define TRACE_INCLUDE_PATH . 627 /* 628 * TRACE_INCLUDE_FILE is not needed if the filename and TRACE_SYSTEM are equal 629 */ 630 #define TRACE_INCLUDE_FILE trace-events-sample 631 #include <trace/define_trace.h> 632