1NANOSLEEP(2) Linux Programmer's Manual NANOSLEEP(2)
2
6 nanosleep - high-resolution sleep
7
9 #include <time.h>
10 int nanosleep(const struct timespec *req, struct timespec *rem);
12 Feature Test Macro Requirements for glibc (see feature_test_macros(7)):
14 nanosleep(): _POSIX_C_SOURCE >= 199309L
16
18 nanosleep() suspends the execution of the calling thread until either
19 at least the time specified in *req has elapsed, or the delivery of a
20 signal that triggers the invocation of a handler in the calling thread
21 or that terminates the process.
22 If the call is interrupted by a signal handler, nanosleep() returns -1,
24 sets errno to EINTR, and writes the remaining time into the structure
25 pointed to by rem unless rem is NULL. The value of *rem can then be
26 used to call nanosleep() again and complete the specified pause (but
27 see NOTES).
28 The structure timespec is used to specify intervals of time with
30 nanosecond precision. It is defined as follows:
31 struct timespec {
33 time_t tv_sec; /* seconds */
34 long tv_nsec; /* nanoseconds */
35 };
36 The value of the nanoseconds field must be in the range 0 to 999999999.
38 Compared to sleep(3) and usleep(3), nanosleep() has the following
40 advantages: it provides a higher resolution for specifying the sleep
41 interval; POSIX.1 explicitly specifies that it does not interact with
42 signals; and it makes the task of resuming a sleep that has been inter‐
43 rupted by a signal handler easier.
44
46 On successfully sleeping for the requested interval, nanosleep()
47 returns 0. If the call is interrupted by a signal handler or encoun‐
48 ters an error, then it returns -1, with errno set to indicate the
49 error.
50
52 EFAULT Problem with copying information from user space.
53 EINTR The pause has been interrupted by a signal that was delivered to
55 the thread (see signal(7)). The remaining sleep time has been
56 written into *rem so that the thread can easily call nanosleep()
57 again and continue with the pause.
58 EINVAL The value in the tv_nsec field was not in the range 0 to
60 999999999 or tv_sec was negative.
61
63 POSIX.1-2001, POSIX.1-2008.
64
66 If the interval specified in req is not an exact multiple of the granu‐
67 larity underlying clock (see time(7)), then the interval will be
68 rounded up to the next multiple. Furthermore, after the sleep com‐
69 pletes, there may still be a delay before the CPU becomes free to once
70 again execute the calling thread.
71 The fact that nanosleep() sleeps for a relative interval can be prob‐
73 lematic if the call is repeatedly restarted after being interrupted by
74 signals, since the time between the interruptions and restarts of the
75 call will lead to drift in the time when the sleep finally completes.
76 This problem can be avoided by using clock_nanosleep(2) with an abso‐
77 lute time value.
78 POSIX.1 specifies that nanosleep() should measure time against the
80 CLOCK_REALTIME clock. However, Linux measures the time using the
81 CLOCK_MONOTONIC clock. This probably does not matter, since the
82 POSIX.1 specification for clock_settime(2) says that discontinuous
83 changes in CLOCK_REALTIME should not affect nanosleep():
84 Setting the value of the CLOCK_REALTIME clock via clock_set‐
86 time(2) shall have no effect on threads that are blocked waiting
87 for a relative time service based upon this clock, including the
88 nanosleep() function; ... Consequently, these time services
89 shall expire when the requested relative interval elapses, inde‐
90 pendently of the new or old value of the clock.
91 Old behavior
93 In order to support applications requiring much more precise pauses
94 (e.g., in order to control some time-critical hardware), nanosleep()
95 would handle pauses of up to 2 milliseconds by busy waiting with
96 microsecond precision when called from a thread scheduled under a real-
97 time policy like SCHED_FIFO or SCHED_RR. This special extension was
98 removed in kernel 2.5.39, and is thus not available in Linux 2.6.0 and
99 later kernels.
100
102 If a program that catches signals and uses nanosleep() receives signals
103 at a very high rate, then scheduling delays and rounding errors in the
104 kernel's calculation of the sleep interval and the returned remain
105 value mean that the remain value may steadily increase on successive
106 restarts of the nanosleep() call. To avoid such problems, use
107 clock_nanosleep(2) with the TIMER_ABSTIME flag to sleep to an absolute
108 deadline.
109 In Linux 2.4, if nanosleep() is stopped by a signal (e.g., SIGTSTP),
111 then the call fails with the error EINTR after the thread is resumed by
112 a SIGCONT signal. If the system call is subsequently restarted, then
113 the time that the thread spent in the stopped state is not counted
114 against the sleep interval. This problem is fixed in Linux 2.6.0 and
115 later kernels.
116
118 clock_nanosleep(2), restart_syscall(2), sched_setscheduler(2),
119 timer_create(2), sleep(3), usleep(3), time(7)
120
122 This page is part of release 5.02 of the Linux man-pages project. A
123 description of the project, information about reporting bugs, and the
124 latest version of this page, can be found at
125 https://www.kernel.org/doc/man-pages/.
126Linux 2017-09-15 NANOSLEEP(2)