1CLOCK_GETRES(2) Linux Programmer's Manual CLOCK_GETRES(2)
2
6 clock_getres, clock_gettime, clock_settime - clock and time functions
7
9 #include <time.h>
10 int clock_getres(clockid_t clockid, struct timespec *res);
12 int clock_gettime(clockid_t clockid, struct timespec *tp);
14 int clock_settime(clockid_t clockid, const struct timespec *tp);
15 Link with -lrt (only for glibc versions before 2.17).
17 Feature Test Macro Requirements for glibc (see feature_test_macros(7)):
19 clock_getres(), clock_gettime(), clock_settime():
21 _POSIX_C_SOURCE >= 199309L
22
24 The function clock_getres() finds the resolution (precision) of the
25 specified clock clockid, and, if res is non-NULL, stores it in the
26 struct timespec pointed to by res. The resolution of clocks depends on
27 the implementation and cannot be configured by a particular process.
28 If the time value pointed to by the argument tp of clock_settime() is
29 not a multiple of res, then it is truncated to a multiple of res.
30 The functions clock_gettime() and clock_settime() retrieve and set the
32 time of the specified clock clockid.
33 The res and tp arguments are timespec structures, as specified in
35 <time.h>:
36 struct timespec {
38 time_t tv_sec; /* seconds */
39 long tv_nsec; /* nanoseconds */
40 };
41 The clockid argument is the identifier of the particular clock on which
43 to act. A clock may be system-wide and hence visible for all pro‐
44 cesses, or per-process if it measures time only within a single
45 process.
46 All implementations support the system-wide real-time clock, which is
48 identified by CLOCK_REALTIME. Its time represents seconds and nanosec‐
49 onds since the Epoch. When its time is changed, timers for a relative
50 interval are unaffected, but timers for an absolute point in time are
51 affected.
52 More clocks may be implemented. The interpretation of the correspond‐
54 ing time values and the effect on timers is unspecified.
55 Sufficiently recent versions of glibc and the Linux kernel support the
57 following clocks:
58 CLOCK_REALTIME
60 A settable system-wide clock that measures real (i.e., wall-
61 clock) time. Setting this clock requires appropriate privi‐
62 leges. This clock is affected by discontinuous jumps in the
63 system time (e.g., if the system administrator manually changes
64 the clock), and by the incremental adjustments performed by adj‐
65 time(3) and NTP.
66 CLOCK_REALTIME_ALARM (since Linux 3.0; Linux-specific)
68 Like CLOCK_REALTIME, but not settable. See timer_create(2) for
69 further details.
70 CLOCK_REALTIME_COARSE (since Linux 2.6.32; Linux-specific)
72 A faster but less precise version of CLOCK_REALTIME. This clock
73 is not settable. Use when you need very fast, but not fine-
74 grained timestamps. Requires per-architecture support, and
75 probably also architecture support for this flag in the vdso(7).
76 CLOCK_TAI (since Linux 3.10; Linux-specific)
78 A nonsettable system-wide clock derived from wall-clock time but
79 ignoring leap seconds. This clock does not experience disconti‐
80 nuities and backwards jumps caused by NTP inserting leap seconds
81 as CLOCK_REALTIME does.
82 The acronym TAI refers to International Atomic Time.
84 CLOCK_MONOTONIC
86 A nonsettable system-wide clock that represents monotonic time
87 since—as described by POSIX—"some unspecified point in the
88 past". On Linux, that point corresponds to the number of sec‐
89 onds that the system has been running since it was booted.
90 The CLOCK_MONOTONIC clock is not affected by discontinuous jumps
92 in the system time (e.g., if the system administrator manually
93 changes the clock), but is affected by the incremental adjust‐
94 ments performed by adjtime(3) and NTP. This clock does not
95 count time that the system is suspended. All CLOCK_MONOTONIC
96 variants guarantee that the time returned by consecutive calls
97 will not go backwards, but successive calls may—depending on the
98 architecture—return identical (not-increased) time values.
99 CLOCK_MONOTONIC_COARSE (since Linux 2.6.32; Linux-specific)
101 A faster but less precise version of CLOCK_MONOTONIC. Use when
102 you need very fast, but not fine-grained timestamps. Requires
103 per-architecture support, and probably also architecture support
104 for this flag in the vdso(7).
105 CLOCK_MONOTONIC_RAW (since Linux 2.6.28; Linux-specific)
107 Similar to CLOCK_MONOTONIC, but provides access to a raw hard‐
108 ware-based time that is not subject to NTP adjustments or the
109 incremental adjustments performed by adjtime(3). This clock
110 does not count time that the system is suspended.
111 CLOCK_BOOTTIME (since Linux 2.6.39; Linux-specific)
113 A nonsettable system-wide clock that is identical to CLOCK_MONO‐
114 TONIC, except that it also includes any time that the system is
115 suspended. This allows applications to get a suspend-aware
116 monotonic clock without having to deal with the complications of
117 CLOCK_REALTIME, which may have discontinuities if the time is
118 changed using settimeofday(2) or similar.
119 CLOCK_BOOTTIME_ALARM (since Linux 3.0; Linux-specific)
121 Like CLOCK_BOOTTIME. See timer_create(2) for further details.
122 CLOCK_PROCESS_CPUTIME_ID (since Linux 2.6.12)
124 This is a clock that measures CPU time consumed by this process
125 (i.e., CPU time consumed by all threads in the process). On
126 Linux, this clock is not settable.
127 CLOCK_THREAD_CPUTIME_ID (since Linux 2.6.12)
129 This is a clock that measures CPU time consumed by this thread.
130 On Linux, this clock is not settable.
131 Linux also implements dynamic clock instances as described below.
133 Dynamic clocks
135 In addition to the hard-coded System-V style clock IDs described above,
136 Linux also supports POSIX clock operations on certain character de‐
137 vices. Such devices are called "dynamic" clocks, and are supported
138 since Linux 2.6.39.
139 Using the appropriate macros, open file descriptors may be converted
141 into clock IDs and passed to clock_gettime(), clock_settime(), and
142 clock_adjtime(2). The following example shows how to convert a file
143 descriptor into a dynamic clock ID.
144 #define CLOCKFD 3
146 #define FD_TO_CLOCKID(fd) ((~(clockid_t) (fd) << 3) | CLOCKFD)
147 #define CLOCKID_TO_FD(clk) ((unsigned int) ~((clk) >> 3))
148 struct timespec ts;
150 clockid_t clkid;
151 int fd;
152 fd = open("/dev/ptp0", O_RDWR);
154 clkid = FD_TO_CLOCKID(fd);
155 clock_gettime(clkid, &ts);
156
158 clock_gettime(), clock_settime(), and clock_getres() return 0 for suc‐
159 cess. On error, -1 is returned and errno is set to indicate the error.
160
162 EACCES clock_settime() does not have write permission for the dynamic
163 POSIX clock device indicated.
164 EFAULT tp points outside the accessible address space.
166 EINVAL The clockid specified is invalid for one of two reasons. Either
168 the System-V style hard coded positive value is out of range, or
169 the dynamic clock ID does not refer to a valid instance of a
170 clock object.
171 EINVAL (clock_settime()): tp.tv_sec is negative or tp.tv_nsec is out‐
173 side the range [0..999,999,999].
174 EINVAL The clockid specified in a call to clock_settime() is not a set‐
176 table clock.
177 EINVAL (since Linux 4.3)
179 A call to clock_settime() with a clockid of CLOCK_REALTIME at‐
180 tempted to set the time to a value less than the current value
181 of the CLOCK_MONOTONIC clock.
182 ENODEV The hot-pluggable device (like USB for example) represented by a
184 dynamic clk_id has disappeared after its character device was
185 opened.
186 ENOTSUP
188 The operation is not supported by the dynamic POSIX clock device
189 specified.
190 EPERM clock_settime() does not have permission to set the clock indi‐
192 cated.
193
195 These system calls first appeared in Linux 2.6.
196
198 For an explanation of the terms used in this section, see at‐
199 tributes(7).
200 ┌────────────────────────────────────────────┬───────────────┬─────────┐
202 │Interface │ Attribute │ Value │
203 ├────────────────────────────────────────────┼───────────────┼─────────┤
204 │clock_getres(), clock_gettime(), │ Thread safety │ MT-Safe │
205 │clock_settime() │ │ │
206 └────────────────────────────────────────────┴───────────────┴─────────┘
207
209 POSIX.1-2001, POSIX.1-2008, SUSv2.
210 On POSIX systems on which these functions are available, the symbol
212 _POSIX_TIMERS is defined in <unistd.h> to a value greater than 0. The
213 symbols _POSIX_MONOTONIC_CLOCK, _POSIX_CPUTIME, _POSIX_THREAD_CPUTIME
214 indicate that CLOCK_MONOTONIC, CLOCK_PROCESS_CPUTIME_ID,
215 CLOCK_THREAD_CPUTIME_ID are available. (See also sysconf(3).)
216
218 POSIX.1 specifies the following:
219 Setting the value of the CLOCK_REALTIME clock via clock_set‐
221 time() shall have no effect on threads that are blocked waiting
222 for a relative time service based upon this clock, including the
223 nanosleep() function; nor on the expiration of relative timers
224 based upon this clock. Consequently, these time services shall
225 expire when the requested relative interval elapses, indepen‐
226 dently of the new or old value of the clock.
227 According to POSIX.1-2001, a process with "appropriate privileges" may
229 set the CLOCK_PROCESS_CPUTIME_ID and CLOCK_THREAD_CPUTIME_ID clocks us‐
230 ing clock_settime(). On Linux, these clocks are not settable (i.e., no
231 process has "appropriate privileges").
232 C library/kernel differences
234 On some architectures, an implementation of clock_gettime() is provided
235 in the vdso(7).
236 Historical note for SMP systems
238 Before Linux added kernel support for CLOCK_PROCESS_CPUTIME_ID and
239 CLOCK_THREAD_CPUTIME_ID, glibc implemented these clocks on many plat‐
240 forms using timer registers from the CPUs (TSC on i386, AR.ITC on Ita‐
241 nium). These registers may differ between CPUs and as a consequence
242 these clocks may return bogus results if a process is migrated to an‐
243 other CPU.
244 If the CPUs in an SMP system have different clock sources, then there
246 is no way to maintain a correlation between the timer registers since
247 each CPU will run at a slightly different frequency. If that is the
248 case, then clock_getcpuclockid(0) will return ENOENT to signify this
249 condition. The two clocks will then be useful only if it can be en‐
250 sured that a process stays on a certain CPU.
251 The processors in an SMP system do not start all at exactly the same
253 time and therefore the timer registers are typically running at an off‐
254 set. Some architectures include code that attempts to limit these off‐
255 sets on bootup. However, the code cannot guarantee to accurately tune
256 the offsets. Glibc contains no provisions to deal with these offsets
257 (unlike the Linux Kernel). Typically these offsets are small and
258 therefore the effects may be negligible in most cases.
259 Since glibc 2.4, the wrapper functions for the system calls described
261 in this page avoid the abovementioned problems by employing the kernel
262 implementation of CLOCK_PROCESS_CPUTIME_ID and CLOCK_THREAD_CPUTIME_ID,
263 on systems that provide such an implementation (i.e., Linux 2.6.12 and
264 later).
265
267 The program below demonstrates the use of clock_gettime() and clock_ge‐
268 tres() with various clocks. This is an example of what we might see
269 when running the program:
270 $ ./clock_times x
272 CLOCK_REALTIME : 1585985459.446 (18356 days + 7h 30m 59s)
273 resolution: 0.000000001
274 CLOCK_TAI : 1585985496.447 (18356 days + 7h 31m 36s)
275 resolution: 0.000000001
276 CLOCK_MONOTONIC: 52395.722 (14h 33m 15s)
277 resolution: 0.000000001
278 CLOCK_BOOTTIME : 72691.019 (20h 11m 31s)
279 resolution: 0.000000001
280 Program source
282 /* clock_times.c
284 Licensed under GNU General Public License v2 or later.
286 */
287 #define _XOPEN_SOURCE 600
288 #include <time.h>
289 #include <stdint.h>
290 #include <stdio.h>
291 #include <stdlib.h>
292 #include <stdbool.h>
293 #include <unistd.h>
294 #define SECS_IN_DAY (24 * 60 * 60)
296 static void
298 displayClock(clockid_t clock, const char *name, bool showRes)
299 {
300 struct timespec ts;
301 if (clock_gettime(clock, &ts) == -1) {
303 perror("clock_gettime");
304 exit(EXIT_FAILURE);
305 }
306 printf("%-15s: %10jd.%03ld (", name,
308 (intmax_t) ts.tv_sec, ts.tv_nsec / 1000000);
309 long days = ts.tv_sec / SECS_IN_DAY;
311 if (days > 0)
312 printf("%ld days + ", days);
313 printf("%2dh %2dm %2ds",
315 (int) (ts.tv_sec % SECS_IN_DAY) / 3600,
316 (int) (ts.tv_sec % 3600) / 60,
317 (int) ts.tv_sec % 60);
318 printf(")\n");
319 if (clock_getres(clock, &ts) == -1) {
321 perror("clock_getres");
322 exit(EXIT_FAILURE);
323 }
324 if (showRes)
326 printf(" resolution: %10jd.%09ld\n",
327 (intmax_t) ts.tv_sec, ts.tv_nsec);
328 }
329 int
331 main(int argc, char *argv[])
332 {
333 bool showRes = argc > 1;
334 displayClock(CLOCK_REALTIME, "CLOCK_REALTIME", showRes);
336 #ifdef CLOCK_TAI
337 displayClock(CLOCK_TAI, "CLOCK_TAI", showRes);
338 #endif
339 displayClock(CLOCK_MONOTONIC, "CLOCK_MONOTONIC", showRes);
340 #ifdef CLOCK_BOOTTIME
341 displayClock(CLOCK_BOOTTIME, "CLOCK_BOOTTIME", showRes);
342 #endif
343 exit(EXIT_SUCCESS);
344 }
345
347 date(1), gettimeofday(2), settimeofday(2), time(2), adjtime(3),
348 clock_getcpuclockid(3), ctime(3), ftime(3), pthread_getcpuclockid(3),
349 sysconf(3), time(7), time_namespaces(7), vdso(7), hwclock(8)
350
352 This page is part of release 5.12 of the Linux man-pages project. A
353 description of the project, information about reporting bugs, and the
354 latest version of this page, can be found at
355 https://www.kernel.org/doc/man-pages/.
356 2021-03-22 CLOCK_GETRES(2)