1timerfd_create(2) System Calls Manual timerfd_create(2)
2
6 timerfd_create, timerfd_settime, timerfd_gettime - timers that notify
7 via file descriptors
8
10 Standard C library (libc, -lc)
11
13 #include <sys/timerfd.h>
14 int timerfd_create(int clockid, int flags);
16 int timerfd_settime(int fd, int flags,
18 const struct itimerspec *new_value,
19 struct itimerspec *_Nullable old_value);
20 int timerfd_gettime(int fd, struct itimerspec *curr_value);
21
23 These system calls create and operate on a timer that delivers timer
24 expiration notifications via a file descriptor. They provide an alter‐
25 native to the use of setitimer(2) or timer_create(2), with the advan‐
26 tage that the file descriptor may be monitored by select(2), poll(2),
27 and epoll(7).
28 The use of these three system calls is analogous to the use of
30 timer_create(2), timer_settime(2), and timer_gettime(2). (There is no
31 analog of timer_getoverrun(2), since that functionality is provided by
32 read(2), as described below.)
33 timerfd_create()
35 timerfd_create() creates a new timer object, and returns a file de‐
36 scriptor that refers to that timer. The clockid argument specifies the
37 clock that is used to mark the progress of the timer, and must be one
38 of the following:
39 CLOCK_REALTIME
41 A settable system-wide real-time clock.
42 CLOCK_MONOTONIC
44 A nonsettable monotonically increasing clock that measures time
45 from some unspecified point in the past that does not change af‐
46 ter system startup.
47 CLOCK_BOOTTIME (Since Linux 3.15)
49 Like CLOCK_MONOTONIC, this is a monotonically increasing clock.
50 However, whereas the CLOCK_MONOTONIC clock does not measure the
51 time while a system is suspended, the CLOCK_BOOTTIME clock does
52 include the time during which the system is suspended. This is
53 useful for applications that need to be suspend-aware.
54 CLOCK_REALTIME is not suitable for such applications, since that
55 clock is affected by discontinuous changes to the system clock.
56 CLOCK_REALTIME_ALARM (since Linux 3.11)
58 This clock is like CLOCK_REALTIME, but will wake the system if
59 it is suspended. The caller must have the CAP_WAKE_ALARM capa‐
60 bility in order to set a timer against this clock.
61 CLOCK_BOOTTIME_ALARM (since Linux 3.11)
63 This clock is like CLOCK_BOOTTIME, but will wake the system if
64 it is suspended. The caller must have the CAP_WAKE_ALARM capa‐
65 bility in order to set a timer against this clock.
66 See clock_getres(2) for some further details on the above clocks.
68 The current value of each of these clocks can be retrieved using
70 clock_gettime(2).
71 Starting with Linux 2.6.27, the following values may be bitwise ORed in
73 flags to change the behavior of timerfd_create():
74 TFD_NONBLOCK Set the O_NONBLOCK file status flag on the open file de‐
76 scription (see open(2)) referred to by the new file de‐
77 scriptor. Using this flag saves extra calls to fcntl(2)
78 to achieve the same result.
79 TFD_CLOEXEC Set the close-on-exec (FD_CLOEXEC) flag on the new file
81 descriptor. See the description of the O_CLOEXEC flag in
82 open(2) for reasons why this may be useful.
83 In Linux versions up to and including 2.6.26, flags must be specified
85 as zero.
86 timerfd_settime()
88 timerfd_settime() arms (starts) or disarms (stops) the timer referred
89 to by the file descriptor fd.
90 The new_value argument specifies the initial expiration and interval
92 for the timer. The itimerspec structure used for this argument is de‐
93 scribed in itimerspec(3type).
94 new_value.it_value specifies the initial expiration of the timer, in
96 seconds and nanoseconds. Setting either field of new_value.it_value to
97 a nonzero value arms the timer. Setting both fields of
98 new_value.it_value to zero disarms the timer.
99 Setting one or both fields of new_value.it_interval to nonzero values
101 specifies the period, in seconds and nanoseconds, for repeated timer
102 expirations after the initial expiration. If both fields of
103 new_value.it_interval are zero, the timer expires just once, at the
104 time specified by new_value.it_value.
105 By default, the initial expiration time specified in new_value is in‐
107 terpreted relative to the current time on the timer's clock at the time
108 of the call (i.e., new_value.it_value specifies a time relative to the
109 current value of the clock specified by clockid). An absolute timeout
110 can be selected via the flags argument.
111 The flags argument is a bit mask that can include the following values:
113 TFD_TIMER_ABSTIME
115 Interpret new_value.it_value as an absolute value on the timer's
116 clock. The timer will expire when the value of the timer's
117 clock reaches the value specified in new_value.it_value.
118 TFD_TIMER_CANCEL_ON_SET
120 If this flag is specified along with TFD_TIMER_ABSTIME and the
121 clock for this timer is CLOCK_REALTIME or CLOCK_REALTIME_ALARM,
122 then mark this timer as cancelable if the real-time clock under‐
123 goes a discontinuous change (settimeofday(2), clock_settime(2),
124 or similar). When such changes occur, a current or future
125 read(2) from the file descriptor will fail with the error ECAN‐
126 CELED.
127 If the old_value argument is not NULL, then the itimerspec structure
129 that it points to is used to return the setting of the timer that was
130 current at the time of the call; see the description of timerfd_get‐
131 time() following.
132 timerfd_gettime()
134 timerfd_gettime() returns, in curr_value, an itimerspec structure that
135 contains the current setting of the timer referred to by the file de‐
136 scriptor fd.
137 The it_value field returns the amount of time until the timer will next
139 expire. If both fields of this structure are zero, then the timer is
140 currently disarmed. This field always contains a relative value, re‐
141 gardless of whether the TFD_TIMER_ABSTIME flag was specified when set‐
142 ting the timer.
143 The it_interval field returns the interval of the timer. If both
145 fields of this structure are zero, then the timer is set to expire just
146 once, at the time specified by curr_value.it_value.
147 Operating on a timer file descriptor
149 The file descriptor returned by timerfd_create() supports the following
150 additional operations:
151 read(2)
153 If the timer has already expired one or more times since its
154 settings were last modified using timerfd_settime(), or since
155 the last successful read(2), then the buffer given to read(2)
156 returns an unsigned 8-byte integer (uint64_t) containing the
157 number of expirations that have occurred. (The returned value
158 is in host byte order—that is, the native byte order for inte‐
159 gers on the host machine.)
160 If no timer expirations have occurred at the time of the
162 read(2), then the call either blocks until the next timer expi‐
163 ration, or fails with the error EAGAIN if the file descriptor
164 has been made nonblocking (via the use of the fcntl(2) F_SETFL
165 operation to set the O_NONBLOCK flag).
166 A read(2) fails with the error EINVAL if the size of the sup‐
168 plied buffer is less than 8 bytes.
169 If the associated clock is either CLOCK_REALTIME or CLOCK_REAL‐
171 TIME_ALARM, the timer is absolute (TFD_TIMER_ABSTIME), and the
172 flag TFD_TIMER_CANCEL_ON_SET was specified when calling
173 timerfd_settime(), then read(2) fails with the error ECANCELED
174 if the real-time clock undergoes a discontinuous change. (This
175 allows the reading application to discover such discontinuous
176 changes to the clock.)
177 If the associated clock is either CLOCK_REALTIME or CLOCK_REAL‐
179 TIME_ALARM, the timer is absolute (TFD_TIMER_ABSTIME), and the
180 flag TFD_TIMER_CANCEL_ON_SET was not specified when calling
181 timerfd_settime(), then a discontinuous negative change to the
182 clock (e.g., clock_settime(2)) may cause read(2) to unblock, but
183 return a value of 0 (i.e., no bytes read), if the clock change
184 occurs after the time expired, but before the read(2) on the
185 file descriptor.
186 poll(2), select(2) (and similar)
188 The file descriptor is readable (the select(2) readfds argument;
189 the poll(2) POLLIN flag) if one or more timer expirations have
190 occurred.
191 The file descriptor also supports the other file-descriptor mul‐
193 tiplexing APIs: pselect(2), ppoll(2), and epoll(7).
194 ioctl(2)
196 The following timerfd-specific command is supported:
197 TFD_IOC_SET_TICKS (since Linux 3.17)
199 Adjust the number of timer expirations that have oc‐
200 curred. The argument is a pointer to a nonzero 8-byte
201 integer (uint64_t*) containing the new number of expira‐
202 tions. Once the number is set, any waiter on the timer
203 is woken up. The only purpose of this command is to re‐
204 store the expirations for the purpose of checkpoint/re‐
205 store. This operation is available only if the kernel
206 was configured with the CONFIG_CHECKPOINT_RESTORE option.
207 close(2)
209 When the file descriptor is no longer required it should be
210 closed. When all file descriptors associated with the same
211 timer object have been closed, the timer is disarmed and its re‐
212 sources are freed by the kernel.
213 fork(2) semantics
215 After a fork(2), the child inherits a copy of the file descriptor cre‐
216 ated by timerfd_create(). The file descriptor refers to the same un‐
217 derlying timer object as the corresponding file descriptor in the par‐
218 ent, and read(2)s in the child will return information about expira‐
219 tions of the timer.
220 execve(2) semantics
222 A file descriptor created by timerfd_create() is preserved across ex‐
223 ecve(2), and continues to generate timer expirations if the timer was
224 armed.
225
227 On success, timerfd_create() returns a new file descriptor. On error,
228 -1 is returned and errno is set to indicate the error.
229 timerfd_settime() and timerfd_gettime() return 0 on success; on error
231 they return -1, and set errno to indicate the error.
232
234 timerfd_create() can fail with the following errors:
235 EINVAL The clockid is not valid.
237 EINVAL flags is invalid; or, in Linux 2.6.26 or earlier, flags is non‐
239 zero.
240 EMFILE The per-process limit on the number of open file descriptors has
242 been reached.
243 ENFILE The system-wide limit on the total number of open files has been
245 reached.
246 ENODEV Could not mount (internal) anonymous inode device.
248 ENOMEM There was insufficient kernel memory to create the timer.
250 EPERM clockid was CLOCK_REALTIME_ALARM or CLOCK_BOOTTIME_ALARM but the
252 caller did not have the CAP_WAKE_ALARM capability.
253 timerfd_settime() and timerfd_gettime() can fail with the following er‐
255 rors:
256 EBADF fd is not a valid file descriptor.
258 EFAULT new_value, old_value, or curr_value is not a valid pointer.
260 EINVAL fd is not a valid timerfd file descriptor.
262 timerfd_settime() can also fail with the following errors:
264 ECANCELED
266 See NOTES.
267 EINVAL new_value is not properly initialized (one of the tv_nsec falls
269 outside the range zero to 999,999,999).
270 EINVAL flags is invalid.
272
274 Linux.
275
277 Linux 2.6.25, glibc 2.8.
278
280 Suppose the following scenario for CLOCK_REALTIME or CLOCK_REAL‐
281 TIME_ALARM timer that was created with timerfd_create():
282 (1) The timer has been started (timerfd_settime()) with the
284 TFD_TIMER_ABSTIME and TFD_TIMER_CANCEL_ON_SET flags;
285 (2) A discontinuous change (e.g., settimeofday(2)) is subsequently
287 made to the CLOCK_REALTIME clock; and
288 (3) the caller once more calls timerfd_settime() to rearm the timer
290 (without first doing a read(2) on the file descriptor).
291 In this case the following occurs:
293 • The timerfd_settime() returns -1 with errno set to ECANCELED. (This
295 enables the caller to know that the previous timer was affected by a
296 discontinuous change to the clock.)
297 • The timer is successfully rearmed with the settings provided in the
299 second timerfd_settime() call. (This was probably an implementation
300 accident, but won't be fixed now, in case there are applications
301 that depend on this behaviour.)
302
304 Currently, timerfd_create() supports fewer types of clock IDs than
305 timer_create(2).
306
308 The following program creates a timer and then monitors its progress.
309 The program accepts up to three command-line arguments. The first ar‐
310 gument specifies the number of seconds for the initial expiration of
311 the timer. The second argument specifies the interval for the timer,
312 in seconds. The third argument specifies the number of times the pro‐
313 gram should allow the timer to expire before terminating. The second
314 and third command-line arguments are optional.
315 The following shell session demonstrates the use of the program:
317 $ a.out 3 1 100
319 0.000: timer started
320 3.000: read: 1; total=1
321 4.000: read: 1; total=2
322 ^Z # type control-Z to suspend the program
323 [1]+ Stopped ./timerfd3_demo 3 1 100
324 $ fg # Resume execution after a few seconds
325 a.out 3 1 100
326 9.660: read: 5; total=7
327 10.000: read: 1; total=8
328 11.000: read: 1; total=9
329 ^C # type control-C to suspend the program
330 Program source
332 #include <err.h>
334 #include <inttypes.h>
335 #include <stdio.h>
336 #include <stdlib.h>
337 #include <sys/timerfd.h>
338 #include <time.h>
339 #include <unistd.h>
340 static void
342 print_elapsed_time(void)
343 {
344 int secs, nsecs;
345 static int first_call = 1;
346 struct timespec curr;
347 static struct timespec start;
348 if (first_call) {
350 first_call = 0;
351 if (clock_gettime(CLOCK_MONOTONIC, &start) == -1)
352 err(EXIT_FAILURE, "clock_gettime");
353 }
354 if (clock_gettime(CLOCK_MONOTONIC, &curr) == -1)
356 err(EXIT_FAILURE, "clock_gettime");
357 secs = curr.tv_sec - start.tv_sec;
359 nsecs = curr.tv_nsec - start.tv_nsec;
360 if (nsecs < 0) {
361 secs--;
362 nsecs += 1000000000;
363 }
364 printf("%d.%03d: ", secs, (nsecs + 500000) / 1000000);
365 }
366 int
368 main(int argc, char *argv[])
369 {
370 int fd;
371 ssize_t s;
372 uint64_t exp, tot_exp, max_exp;
373 struct timespec now;
374 struct itimerspec new_value;
375 if (argc != 2 && argc != 4) {
377 fprintf(stderr, "%s init-secs [interval-secs max-exp]\n",
378 argv[0]);
379 exit(EXIT_FAILURE);
380 }
381 if (clock_gettime(CLOCK_REALTIME, &now) == -1)
383 err(EXIT_FAILURE, "clock_gettime");
384 /* Create a CLOCK_REALTIME absolute timer with initial
386 expiration and interval as specified in command line. */
387 new_value.it_value.tv_sec = now.tv_sec + atoi(argv[1]);
389 new_value.it_value.tv_nsec = now.tv_nsec;
390 if (argc == 2) {
391 new_value.it_interval.tv_sec = 0;
392 max_exp = 1;
393 } else {
394 new_value.it_interval.tv_sec = atoi(argv[2]);
395 max_exp = atoi(argv[3]);
396 }
397 new_value.it_interval.tv_nsec = 0;
398 fd = timerfd_create(CLOCK_REALTIME, 0);
400 if (fd == -1)
401 err(EXIT_FAILURE, "timerfd_create");
402 if (timerfd_settime(fd, TFD_TIMER_ABSTIME, &new_value, NULL) == -1)
404 err(EXIT_FAILURE, "timerfd_settime");
405 print_elapsed_time();
407 printf("timer started\n");
408 for (tot_exp = 0; tot_exp < max_exp;) {
410 s = read(fd, &exp, sizeof(uint64_t));
411 if (s != sizeof(uint64_t))
412 err(EXIT_FAILURE, "read");
413 tot_exp += exp;
415 print_elapsed_time();
416 printf("read: %" PRIu64 "; total=%" PRIu64 "\n", exp, tot_exp);
417 }
418 exit(EXIT_SUCCESS);
420 }
421
423 eventfd(2), poll(2), read(2), select(2), setitimer(2), signalfd(2),
424 timer_create(2), timer_gettime(2), timer_settime(2), timespec(3),
425 epoll(7), time(7)
426Linux man-pages 6.05 2023-05-03 timerfd_create(2)