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