1 /* SPDX-License-Identifier: GPL-2.0 */
3 * If TRACE_SYSTEM is defined, that will be the directory created
4 * in the ftrace directory under /sys/kernel/tracing/events/<system>
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
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
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
18 * #define TRACE_INCLUDE_FILE trace-events-sample
20 * As we do an the bottom of this file.
22 * Notice that TRACE_SYSTEM should be defined outside of #if
23 * protection, just like TRACE_INCLUDE_FILE.
26 #define TRACE_SYSTEM sample-trace
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:
33 #undef TRACE_SYSTEM_VAR
34 #define TRACE_SYSTEM_VAR sample_trace
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.
42 * The TRACE_SYSTEM_VAR is only used internally and not visible to
47 * Notice that this file is not protected like a normal header.
48 * We also must allow for rereading of this file. The
50 * || defined(TRACE_HEADER_MULTI_READ)
52 * serves this purpose.
54 #if !defined(_TRACE_EVENT_SAMPLE_H) || defined(TRACE_HEADER_MULTI_READ)
55 #define _TRACE_EVENT_SAMPLE_H
58 * All trace headers should include tracepoint.h, until we finally
59 * make it into a standard header.
61 #include <linux/tracepoint.h>
64 * The TRACE_EVENT macro is broken up into 5 parts.
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.
69 * proto: the prototype of the function trace_foo_bar()
70 * Here it is trace_foo_bar(char *foo, int bar).
72 * args: must match the arguments in the prototype.
73 * Here it is simply "foo, bar".
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
80 * Here are the currently defined types you can use:
82 * __field : Is broken up into type and name. Where type can be any
83 * primitive type (integer, long or pointer).
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.
94 * __field_struct(struct bar, foo)
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).
102 * __array( char, foo, 10) is the same as saying: char foo[10];
104 * Assigning arrays can be done like any array:
106 * __entry->foo[0] = 'a';
108 * memcpy(__entry->foo, bar, 10);
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.
119 * __dynamic_array( int, foo, bar) is similar to: int foo[bar];
121 * Note, unlike arrays, you must use the __get_dynamic_array() macro
122 * to access the array.
124 * memcpy(__get_dynamic_array(foo), bar, 10);
126 * Notice, that "__entry" is not needed here.
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
135 * __string(foo, bar) is similar to: strcpy(foo, bar)
137 * To assign a string, use the helper macro __assign_str().
139 * __assign_str(foo, bar);
141 * In most cases, the __assign_str() macro will take the same
142 * parameters as the __string() macro had to declare the string.
144 * __bitmask: This is another kind of __dynamic_array, but it expects
145 * an array of longs, and the number of bits to parse. It takes
146 * two parameters (name, nr_bits), where name is the name of the
147 * bitmask to save, and the nr_bits is the number of bits to record.
149 * __bitmask(target_cpu, nr_cpumask_bits)
151 * To assign a bitmask, use the __assign_bitmask() helper macro.
153 * __assign_bitmask(target_cpus, cpumask_bits(bar), nr_cpumask_bits);
156 * fast_assign: This is a C like function that is used to store the items
157 * into the ring buffer. A special variable called "__entry" will be the
158 * structure that points into the ring buffer and has the same fields as
159 * described by the struct part of TRACE_EVENT above.
161 * printk: This is a way to print out the data in pretty print. This is
162 * useful if the system crashes and you are logging via a serial line,
163 * the data can be printed to the console using this "printk" method.
164 * This is also used to print out the data from the trace files.
165 * Again, the __entry macro is used to access the data from the ring buffer.
167 * Note, __dynamic_array, __string, and __bitmask require special helpers
168 * to access the data.
170 * For __dynamic_array(int, foo, bar) use __get_dynamic_array(foo)
171 * Use __get_dynamic_array_len(foo) to get the length of the array
172 * saved. Note, __get_dynamic_array_len() returns the total allocated
173 * length of the dynamic array; __print_array() expects the second
174 * parameter to be the number of elements. To get that, the array length
175 * needs to be divided by the element size.
177 * For __string(foo, bar) use __get_str(foo)
179 * For __bitmask(target_cpus, nr_cpumask_bits) use __get_bitmask(target_cpus)
182 * Note, that for both the assign and the printk, __entry is the handler
183 * to the data structure in the ring buffer, and is defined by the
188 * It is OK to have helper functions in the file, but they need to be protected
189 * from being defined more than once. Remember, this file gets included more
192 #ifndef __TRACE_EVENT_SAMPLE_HELPER_FUNCTIONS
193 #define __TRACE_EVENT_SAMPLE_HELPER_FUNCTIONS
194 static inline int __length_of(const int *list)
201 for (i = 0; list[i]; i++)
207 TRACE_SAMPLE_FOO = 2,
208 TRACE_SAMPLE_BAR = 4,
209 TRACE_SAMPLE_ZOO = 8,
214 * If enums are used in the TP_printk(), their names will be shown in
215 * format files and not their values. This can cause problems with user
216 * space programs that parse the format files to know how to translate
217 * the raw binary trace output into human readable text.
219 * To help out user space programs, any enum that is used in the TP_printk()
220 * should be defined by TRACE_DEFINE_ENUM() macro. All that is needed to
221 * be done is to add this macro with the enum within it in the trace
222 * header file, and it will be converted in the output.
225 TRACE_DEFINE_ENUM(TRACE_SAMPLE_FOO);
226 TRACE_DEFINE_ENUM(TRACE_SAMPLE_BAR);
227 TRACE_DEFINE_ENUM(TRACE_SAMPLE_ZOO);
231 TP_PROTO(const char *foo, int bar, const int *lst,
232 const char *string, const struct cpumask *mask),
234 TP_ARGS(foo, bar, lst, string, mask),
237 __array( char, foo, 10 )
239 __dynamic_array(int, list, __length_of(lst))
240 __string( str, string )
241 __bitmask( cpus, num_possible_cpus() )
245 strlcpy(__entry->foo, foo, 10);
247 memcpy(__get_dynamic_array(list), lst,
248 __length_of(lst) * sizeof(int));
249 __assign_str(str, string);
250 __assign_bitmask(cpus, cpumask_bits(mask), num_possible_cpus());
253 TP_printk("foo %s %d %s %s %s %s (%s)", __entry->foo, __entry->bar,
256 * Notice here the use of some helper functions. This includes:
258 * __print_symbolic( variable, { value, "string" }, ... ),
260 * The variable is tested against each value of the { } pair. If
261 * the variable matches one of the values, then it will print the
262 * string in that pair. If non are matched, it returns a string
263 * version of the number (if __entry->bar == 7 then "7" is returned).
265 __print_symbolic(__entry->bar,
267 { TRACE_SAMPLE_FOO, "TWO" },
268 { TRACE_SAMPLE_BAR, "FOUR" },
269 { TRACE_SAMPLE_ZOO, "EIGHT" },
274 * __print_flags( variable, "delim", { value, "flag" }, ... ),
276 * This is similar to __print_symbolic, except that it tests the bits
277 * of the value. If ((FLAG & variable) == FLAG) then the string is
278 * printed. If more than one flag matches, then each one that does is
279 * also printed with delim in between them.
280 * If not all bits are accounted for, then the not found bits will be
281 * added in hex format: 0x506 will show BIT2|BIT4|0x500
283 __print_flags(__entry->bar, "|",
290 * __print_array( array, len, element_size )
292 * This prints out the array that is defined by __array in a nice format.
294 __print_array(__get_dynamic_array(list),
295 __get_dynamic_array_len(list) / sizeof(int),
297 __get_str(str), __get_bitmask(cpus))
301 * There may be a case where a tracepoint should only be called if
302 * some condition is set. Otherwise the tracepoint should not be called.
303 * But to do something like:
308 * Would cause a little overhead when tracing is not enabled, and that
309 * overhead, even if small, is not something we want. As tracepoints
310 * use static branch (aka jump_labels), where no branch is taken to
311 * skip the tracepoint when not enabled, and a jmp is placed to jump
312 * to the tracepoint code when it is enabled, having a if statement
313 * nullifies that optimization. It would be nice to place that
314 * condition within the static branch. This is where TRACE_EVENT_CONDITION
317 * TRACE_EVENT_CONDITION() is just like TRACE_EVENT, except it adds another
318 * parameter just after args. Where TRACE_EVENT has:
320 * TRACE_EVENT(name, proto, args, struct, assign, printk)
322 * the CONDITION version has:
324 * TRACE_EVENT_CONDITION(name, proto, args, cond, struct, assign, printk)
326 * Everything is the same as TRACE_EVENT except for the new cond. Think
327 * of the cond variable as:
330 * trace_foo_bar_with_cond();
332 * Except that the logic for the if branch is placed after the static branch.
333 * That is, the if statement that processes the condition will not be
334 * executed unless that traecpoint is enabled. Otherwise it still remains
337 TRACE_EVENT_CONDITION(foo_bar_with_cond,
339 TP_PROTO(const char *foo, int bar),
343 TP_CONDITION(!(bar % 10)),
351 __assign_str(foo, foo);
355 TP_printk("foo %s %d", __get_str(foo), __entry->bar)
358 int foo_bar_reg(void);
359 void foo_bar_unreg(void);
362 * Now in the case that some function needs to be called when the
363 * tracepoint is enabled and/or when it is disabled, the
364 * TRACE_EVENT_FN() serves this purpose. This is just like TRACE_EVENT()
365 * but adds two more parameters at the end:
367 * TRACE_EVENT_FN( name, proto, args, struct, assign, printk, reg, unreg)
369 * reg and unreg are functions with the prototype of:
373 * The reg function gets called before the tracepoint is enabled, and
374 * the unreg function gets called after the tracepoint is disabled.
376 * Note, reg and unreg are allowed to be NULL. If you only need to
377 * call a function before enabling, or after disabling, just set one
378 * function and pass in NULL for the other parameter.
380 TRACE_EVENT_FN(foo_bar_with_fn,
382 TP_PROTO(const char *foo, int bar),
392 __assign_str(foo, foo);
396 TP_printk("foo %s %d", __get_str(foo), __entry->bar),
398 foo_bar_reg, foo_bar_unreg
402 * Each TRACE_EVENT macro creates several helper functions to produce
403 * the code to add the tracepoint, create the files in the trace
404 * directory, hook it to perf, assign the values and to print out
405 * the raw data from the ring buffer. To prevent too much bloat,
406 * if there are more than one tracepoint that uses the same format
407 * for the proto, args, struct, assign and printk, and only the name
408 * is different, it is highly recommended to use the DECLARE_EVENT_CLASS
410 * DECLARE_EVENT_CLASS() macro creates most of the functions for the
411 * tracepoint. Then DEFINE_EVENT() is use to hook a tracepoint to those
412 * functions. This DEFINE_EVENT() is an instance of the class and can
413 * be enabled and disabled separately from other events (either TRACE_EVENT
414 * or other DEFINE_EVENT()s).
416 * Note, TRACE_EVENT() itself is simply defined as:
418 * #define TRACE_EVENT(name, proto, args, tstruct, assign, printk) \
419 * DEFINE_EVENT_CLASS(name, proto, args, tstruct, assign, printk); \
420 * DEFINE_EVENT(name, name, proto, args)
422 * The DEFINE_EVENT() also can be declared with conditions and reg functions:
424 * DEFINE_EVENT_CONDITION(template, name, proto, args, cond);
425 * DEFINE_EVENT_FN(template, name, proto, args, reg, unreg);
427 DECLARE_EVENT_CLASS(foo_template,
429 TP_PROTO(const char *foo, int bar),
439 __assign_str(foo, foo);
443 TP_printk("foo %s %d", __get_str(foo), __entry->bar)
447 * Here's a better way for the previous samples (except, the first
448 * example had more fields and could not be used here).
450 DEFINE_EVENT(foo_template, foo_with_template_simple,
451 TP_PROTO(const char *foo, int bar),
454 DEFINE_EVENT_CONDITION(foo_template, foo_with_template_cond,
455 TP_PROTO(const char *foo, int bar),
457 TP_CONDITION(!(bar % 8)));
460 DEFINE_EVENT_FN(foo_template, foo_with_template_fn,
461 TP_PROTO(const char *foo, int bar),
463 foo_bar_reg, foo_bar_unreg);
466 * Anytime two events share basically the same values and have
467 * the same output, use the DECLARE_EVENT_CLASS() and DEFINE_EVENT()
468 * when ever possible.
472 * If the event is similar to the DECLARE_EVENT_CLASS, but you need
473 * to have a different output, then use DEFINE_EVENT_PRINT() which
474 * lets you override the TP_printk() of the class.
477 DEFINE_EVENT_PRINT(foo_template, foo_with_template_print,
478 TP_PROTO(const char *foo, int bar),
480 TP_printk("bar %s %d", __get_str(foo), __entry->bar));
484 /***** NOTICE! The #if protection ends here. *****/
488 * There are several ways I could have done this. If I left out the
489 * TRACE_INCLUDE_PATH, then it would default to the kernel source
490 * include/trace/events directory.
492 * I could specify a path from the define_trace.h file back to this
495 * #define TRACE_INCLUDE_PATH ../../samples/trace_events
497 * But the safest and easiest way to simply make it use the directory
498 * that the file is in is to add in the Makefile:
500 * CFLAGS_trace-events-sample.o := -I$(src)
502 * This will make sure the current path is part of the include
503 * structure for our file so that define_trace.h can find it.
505 * I could have made only the top level directory the include:
507 * CFLAGS_trace-events-sample.o := -I$(PWD)
509 * And then let the path to this directory be the TRACE_INCLUDE_PATH:
511 * #define TRACE_INCLUDE_PATH samples/trace_events
513 * But then if something defines "samples" or "trace_events" as a macro
514 * then we could risk that being converted too, and give us an unexpected
517 #undef TRACE_INCLUDE_PATH
518 #undef TRACE_INCLUDE_FILE
519 #define TRACE_INCLUDE_PATH .
521 * TRACE_INCLUDE_FILE is not needed if the filename and TRACE_SYSTEM are equal
523 #define TRACE_INCLUDE_FILE trace-events-sample
524 #include <trace/define_trace.h>