1CLOCK_GETRES(2) Linux Programmer's Manual CLOCK_GETRES(2)
2
3
4
6 clock_getres, clock_gettime, clock_settime - clock and time functions
7
9 #include <time.h>
10
11 int clock_getres(clockid_t clockid, struct timespec *res);
12
13 int clock_gettime(clockid_t clockid, struct timespec *tp);
14
15 int clock_settime(clockid_t clockid, const struct timespec *tp);
16
17 Link with -lrt (only for glibc versions before 2.17).
18
19 Feature Test Macro Requirements for glibc (see feature_test_macros(7)):
20
21 clock_getres(), clock_gettime(), clock_settime():
22 _POSIX_C_SOURCE >= 199309L
23
25 The function clock_getres() finds the resolution (precision) of the
26 specified clock clockid, and, if res is non-NULL, stores it in the
27 struct timespec pointed to by res. The resolution of clocks depends on
28 the implementation and cannot be configured by a particular process.
29 If the time value pointed to by the argument tp of clock_settime() is
30 not a multiple of res, then it is truncated to a multiple of res.
31
32 The functions clock_gettime() and clock_settime() retrieve and set the
33 time of the specified clock clockid.
34
35 The res and tp arguments are timespec structures, as specified in
36 <time.h>:
37
38 struct timespec {
39 time_t tv_sec; /* seconds */
40 long tv_nsec; /* nanoseconds */
41 };
42
43 The clockid argument is the identifier of the particular clock on which
44 to act. A clock may be system-wide and hence visible for all pro‐
45 cesses, or per-process if it measures time only within a single
46 process.
47
48 All implementations support the system-wide real-time clock, which is
49 identified by CLOCK_REALTIME. Its time represents seconds and nanosec‐
50 onds since the Epoch. When its time is changed, timers for a relative
51 interval are unaffected, but timers for an absolute point in time are
52 affected.
53
54 More clocks may be implemented. The interpretation of the correspond‐
55 ing time values and the effect on timers is unspecified.
56
57 Sufficiently recent versions of glibc and the Linux kernel support the
58 following clocks:
59
60 CLOCK_REALTIME
61 A settable system-wide clock that measures real (i.e., wall-
62 clock) time. Setting this clock requires appropriate privi‐
63 leges. This clock is affected by discontinuous jumps in the
64 system time (e.g., if the system administrator manually changes
65 the clock), and by the incremental adjustments performed by adj‐
66 time(3) and NTP.
67
68 CLOCK_REALTIME_ALARM (since Linux 3.0; Linux-specific)
69 Like CLOCK_REALTIME, but not settable. See timer_create(2) for
70 further details.
71
72 CLOCK_REALTIME_COARSE (since Linux 2.6.32; Linux-specific)
73 A faster but less precise version of CLOCK_REALTIME. This clock
74 is not settable. Use when you need very fast, but not fine-
75 grained timestamps. Requires per-architecture support, and
76 probably also architecture support for this flag in the vdso(7).
77
78 CLOCK_TAI (since Linux 3.10; Linux-specific)
79 A nonsettable system-wide clock derived from wall-clock time but
80 ignoring leap seconds. This clock does not experience disconti‐
81 nuities and backwards jumps caused by NTP inserting leap seconds
82 as CLOCK_REALTIME does.
83
84 The acronym TAI refers to International Atomic Time.
85
86 CLOCK_MONOTONIC
87 A nonsettable system-wide clock that represents monotonic time
88 since—as described by POSIX—"some unspecified point in the
89 past". On Linux, that point corresponds to the number of sec‐
90 onds that the system has been running since it was booted.
91
92 The CLOCK_MONOTONIC clock is not affected by discontinuous jumps
93 in the system time (e.g., if the system administrator manually
94 changes the clock), but is affected by the incremental adjust‐
95 ments performed by adjtime(3) and NTP. This clock does not
96 count time that the system is suspended. All CLOCK_MONOTONIC
97 variants guarantee that the time returned by consecutive calls
98 will not go backwards, but successive calls may—depending on the
99 architecture—return identical (not-increased) time values.
100
101 CLOCK_MONOTONIC_COARSE (since Linux 2.6.32; Linux-specific)
102 A faster but less precise version of CLOCK_MONOTONIC. Use when
103 you need very fast, but not fine-grained timestamps. Requires
104 per-architecture support, and probably also architecture support
105 for this flag in the vdso(7).
106
107 CLOCK_MONOTONIC_RAW (since Linux 2.6.28; Linux-specific)
108 Similar to CLOCK_MONOTONIC, but provides access to a raw hard‐
109 ware-based time that is not subject to NTP adjustments or the
110 incremental adjustments performed by adjtime(3). This clock
111 does not count time that the system is suspended.
112
113 CLOCK_BOOTTIME (since Linux 2.6.39; Linux-specific)
114 A nonsettable system-wide clock that is identical to CLOCK_MONO‐
115 TONIC, except that it also includes any time that the system is
116 suspended. This allows applications to get a suspend-aware
117 monotonic clock without having to deal with the complications of
118 CLOCK_REALTIME, which may have discontinuities if the time is
119 changed using settimeofday(2) or similar.
120
121 CLOCK_BOOTTIME_ALARM (since Linux 3.0; Linux-specific)
122 Like CLOCK_BOOTTIME. See timer_create(2) for further details.
123
124 CLOCK_PROCESS_CPUTIME_ID (since Linux 2.6.12)
125 This is a clock that measures CPU time consumed by this process
126 (i.e., CPU time consumed by all threads in the process). On
127 Linux, this clock is not settable.
128
129 CLOCK_THREAD_CPUTIME_ID (since Linux 2.6.12)
130 This is a clock that measures CPU time consumed by this thread.
131 On Linux, this clock is not settable.
132
133 Linux also implements dynamic clock instances as described below.
134
135 Dynamic clocks
136 In addition to the hard-coded System-V style clock IDs described above,
137 Linux also supports POSIX clock operations on certain character de‐
138 vices. Such devices are called "dynamic" clocks, and are supported
139 since Linux 2.6.39.
140
141 Using the appropriate macros, open file descriptors may be converted
142 into clock IDs and passed to clock_gettime(), clock_settime(), and
143 clock_adjtime(2). The following example shows how to convert a file
144 descriptor into a dynamic clock ID.
145
146 #define CLOCKFD 3
147 #define FD_TO_CLOCKID(fd) ((~(clockid_t) (fd) << 3) | CLOCKFD)
148 #define CLOCKID_TO_FD(clk) ((unsigned int) ~((clk) >> 3))
149
150 struct timespec ts;
151 clockid_t clkid;
152 int fd;
153
154 fd = open("/dev/ptp0", O_RDWR);
155 clkid = FD_TO_CLOCKID(fd);
156 clock_gettime(clkid, &ts);
157
159 clock_gettime(), clock_settime(), and clock_getres() return 0 for suc‐
160 cess, or -1 for failure (in which case errno is set appropriately).
161
163 EACCES clock_settime() does not have write permission for the dynamic
164 POSIX clock device indicated.
165
166 EFAULT tp points outside the accessible address space.
167
168 EINVAL The clockid specified is invalid for one of two reasons. Either
169 the System-V style hard coded positive value is out of range, or
170 the dynamic clock ID does not refer to a valid instance of a
171 clock object.
172
173 EINVAL (clock_settime()): tp.tv_sec is negative or tp.tv_nsec is out‐
174 side the range [0..999,999,999].
175
176 EINVAL The clockid specified in a call to clock_settime() is not a set‐
177 table clock.
178
179 EINVAL (since Linux 4.3)
180 A call to clock_settime() with a clockid of CLOCK_REALTIME at‐
181 tempted to set the time to a value less than the current value
182 of the CLOCK_MONOTONIC clock.
183
184 ENODEV The hot-pluggable device (like USB for example) represented by a
185 dynamic clk_id has disappeared after its character device was
186 opened.
187
188 ENOTSUP
189 The operation is not supported by the dynamic POSIX clock device
190 specified.
191
192 EPERM clock_settime() does not have permission to set the clock indi‐
193 cated.
194
196 These system calls first appeared in Linux 2.6.
197
199 For an explanation of the terms used in this section, see at‐
200 tributes(7).
201
202 ┌─────────────────────────────────┬───────────────┬─────────┐
203 │Interface │ Attribute │ Value │
204 ├─────────────────────────────────┼───────────────┼─────────┤
205 │clock_getres(), clock_gettime(), │ Thread safety │ MT-Safe │
206 │clock_settime() │ │ │
207 └─────────────────────────────────┴───────────────┴─────────┘
208
210 POSIX.1-2001, POSIX.1-2008, SUSv2.
211
212 On POSIX systems on which these functions are available, the symbol
213 _POSIX_TIMERS is defined in <unistd.h> to a value greater than 0. The
214 symbols _POSIX_MONOTONIC_CLOCK, _POSIX_CPUTIME, _POSIX_THREAD_CPUTIME
215 indicate that CLOCK_MONOTONIC, CLOCK_PROCESS_CPUTIME_ID,
216 CLOCK_THREAD_CPUTIME_ID are available. (See also sysconf(3).)
217
219 POSIX.1 specifies the following:
220
221 Setting the value of the CLOCK_REALTIME clock via clock_set‐
222 time() shall have no effect on threads that are blocked waiting
223 for a relative time service based upon this clock, including the
224 nanosleep() function; nor on the expiration of relative timers
225 based upon this clock. Consequently, these time services shall
226 expire when the requested relative interval elapses, indepen‐
227 dently of the new or old value of the clock.
228
229 According to POSIX.1-2001, a process with "appropriate privileges" may
230 set the CLOCK_PROCESS_CPUTIME_ID and CLOCK_THREAD_CPUTIME_ID clocks us‐
231 ing clock_settime(). On Linux, these clocks are not settable (i.e., no
232 process has "appropriate privileges").
233
234 C library/kernel differences
235 On some architectures, an implementation of clock_gettime() is provided
236 in the vdso(7).
237
238 Historical note for SMP systems
239 Before Linux added kernel support for CLOCK_PROCESS_CPUTIME_ID and
240 CLOCK_THREAD_CPUTIME_ID, glibc implemented these clocks on many plat‐
241 forms using timer registers from the CPUs (TSC on i386, AR.ITC on Ita‐
242 nium). These registers may differ between CPUs and as a consequence
243 these clocks may return bogus results if a process is migrated to an‐
244 other CPU.
245
246 If the CPUs in an SMP system have different clock sources, then there
247 is no way to maintain a correlation between the timer registers since
248 each CPU will run at a slightly different frequency. If that is the
249 case, then clock_getcpuclockid(0) will return ENOENT to signify this
250 condition. The two clocks will then be useful only if it can be en‐
251 sured that a process stays on a certain CPU.
252
253 The processors in an SMP system do not start all at exactly the same
254 time and therefore the timer registers are typically running at an off‐
255 set. Some architectures include code that attempts to limit these off‐
256 sets on bootup. However, the code cannot guarantee to accurately tune
257 the offsets. Glibc contains no provisions to deal with these offsets
258 (unlike the Linux Kernel). Typically these offsets are small and
259 therefore the effects may be negligible in most cases.
260
261 Since glibc 2.4, the wrapper functions for the system calls described
262 in this page avoid the abovementioned problems by employing the kernel
263 implementation of CLOCK_PROCESS_CPUTIME_ID and CLOCK_THREAD_CPUTIME_ID,
264 on systems that provide such an implementation (i.e., Linux 2.6.12 and
265 later).
266
268 The program below demonstrates the use of clock_gettime() and clock_ge‐
269 tres() with various clocks. This is an example of what we might see
270 when running the program:
271
272 $ ./clock_times x
273 CLOCK_REALTIME : 1585985459.446 (18356 days + 7h 30m 59s)
274 resolution: 0.000000001
275 CLOCK_TAI : 1585985496.447 (18356 days + 7h 31m 36s)
276 resolution: 0.000000001
277 CLOCK_MONOTONIC: 52395.722 (14h 33m 15s)
278 resolution: 0.000000001
279 CLOCK_BOOTTIME : 72691.019 (20h 11m 31s)
280 resolution: 0.000000001
281
282 Program source
283
284 /* clock_times.c
285
286 Licensed under GNU General Public License v2 or later.
287 */
288 #define _XOPEN_SOURCE 600
289 #include <time.h>
290 #include <stdint.h>
291 #include <stdio.h>
292 #include <stdlib.h>
293 #include <stdbool.h>
294 #include <unistd.h>
295
296 #define SECS_IN_DAY (24 * 60 * 60)
297
298 static void
299 displayClock(clockid_t clock, const char *name, bool showRes)
300 {
301 struct timespec ts;
302
303 if (clock_gettime(clock, &ts) == -1) {
304 perror("clock_gettime");
305 exit(EXIT_FAILURE);
306 }
307
308 printf("%-15s: %10jd.%03ld (", name,
309 (intmax_t) ts.tv_sec, ts.tv_nsec / 1000000);
310
311 long days = ts.tv_sec / SECS_IN_DAY;
312 if (days > 0)
313 printf("%ld days + ", days);
314
315 printf("%2dh %2dm %2ds",
316 (int) (ts.tv_sec % SECS_IN_DAY) / 3600,
317 (int) (ts.tv_sec % 3600) / 60,
318 (int) ts.tv_sec % 60);
319 printf(")\n");
320
321 if (clock_getres(clock, &ts) == -1) {
322 perror("clock_getres");
323 exit(EXIT_FAILURE);
324 }
325
326 if (showRes)
327 printf(" resolution: %10jd.%09ld\n",
328 (intmax_t) ts.tv_sec, ts.tv_nsec);
329 }
330
331 int
332 main(int argc, char *argv[])
333 {
334 bool showRes = argc > 1;
335
336 displayClock(CLOCK_REALTIME, "CLOCK_REALTIME", showRes);
337 #ifdef CLOCK_TAI
338 displayClock(CLOCK_TAI, "CLOCK_TAI", showRes);
339 #endif
340 displayClock(CLOCK_MONOTONIC, "CLOCK_MONOTONIC", showRes);
341 #ifdef CLOCK_BOOTTIME
342 displayClock(CLOCK_BOOTTIME, "CLOCK_BOOTTIME", showRes);
343 #endif
344 exit(EXIT_SUCCESS);
345 }
346
348 date(1), gettimeofday(2), settimeofday(2), time(2), adjtime(3),
349 clock_getcpuclockid(3), ctime(3), ftime(3), pthread_getcpuclockid(3),
350 sysconf(3), time(7), time_namespaces(7), vdso(7), hwclock(8)
351
353 This page is part of release 5.10 of the Linux man-pages project. A
354 description of the project, information about reporting bugs, and the
355 latest version of this page, can be found at
356 https://www.kernel.org/doc/man-pages/.
357
358
359
360 2020-12-21 CLOCK_GETRES(2)