| 1 |
.\" Copyright 1996, Markus Kuhn <[email protected]> |
| 2 |
.\" Copyright, the authors of the Linux man-pages project |
| 3 |
.\" |
| 4 |
.\" SPDX-License-Identifier: GPL-2.0-or-later |
| 5 |
.\" |
| 6 |
.TH nanosleep 2 2026-02-08 "Linux man-pages 6.18" |
| 7 |
.SH NAME |
| 8 |
nanosleep \- high-resolution sleep |
| 9 |
.SH LIBRARY |
| 10 |
Standard C library |
| 11 |
.RI ( libc ,\~ \-lc ) |
| 12 |
.SH SYNOPSIS |
| 13 |
.nf |
| 14 |
.B #include <time.h> |
| 15 |
.P |
| 16 |
.BI "int nanosleep(const struct timespec *" duration , |
| 17 |
.BI " struct timespec *_Nullable " rem ); |
| 18 |
.fi |
| 19 |
.P |
| 20 |
.RS -4 |
| 21 |
Feature Test Macro Requirements for glibc (see |
| 22 |
.BR feature_test_macros (7)): |
| 23 |
.RE |
| 24 |
.P |
| 25 |
.BR nanosleep (): |
| 26 |
.nf |
| 27 |
_POSIX_C_SOURCE >= 199309L |
| 28 |
.fi |
| 29 |
.SH DESCRIPTION |
| 30 |
.BR nanosleep () |
| 31 |
suspends the execution of the calling thread |
| 32 |
until either at least the time specified in |
| 33 |
.I *duration |
| 34 |
has elapsed, or the delivery of a signal |
| 35 |
that triggers the invocation of a handler in the calling thread or |
| 36 |
that terminates the process. |
| 37 |
.P |
| 38 |
If the call is interrupted by a signal handler, |
| 39 |
.BR nanosleep () |
| 40 |
returns \-1, sets |
| 41 |
.I errno |
| 42 |
to |
| 43 |
.BR EINTR , |
| 44 |
and writes the remaining time into the structure pointed to by |
| 45 |
.I rem |
| 46 |
unless |
| 47 |
.I rem |
| 48 |
is NULL. |
| 49 |
The value of |
| 50 |
.I *rem |
| 51 |
can then be used to call |
| 52 |
.BR nanosleep () |
| 53 |
again and complete the specified pause (but see NOTES). |
| 54 |
.P |
| 55 |
The |
| 56 |
.BR timespec (3) |
| 57 |
structure |
| 58 |
is used to specify intervals of time with nanosecond precision. |
| 59 |
.P |
| 60 |
The value of the nanoseconds field must be in the range [0, 999999999]. |
| 61 |
.P |
| 62 |
Compared to |
| 63 |
.BR sleep (3) |
| 64 |
and |
| 65 |
.BR usleep (3), |
| 66 |
.BR nanosleep () |
| 67 |
has the following advantages: |
| 68 |
it provides a higher resolution for specifying the sleep interval; |
| 69 |
POSIX.1 explicitly specifies that it |
| 70 |
does not interact with signals; |
| 71 |
and it makes the task of resuming a sleep that has been |
| 72 |
interrupted by a signal handler easier. |
| 73 |
.SH RETURN VALUE |
| 74 |
On successfully sleeping for the requested duration, |
| 75 |
.BR nanosleep () |
| 76 |
returns 0. |
| 77 |
If the call is interrupted by a signal handler or encounters an error, |
| 78 |
then it returns \-1, with |
| 79 |
.I errno |
| 80 |
set to indicate the error. |
| 81 |
.SH ERRORS |
| 82 |
.TP |
| 83 |
.B EFAULT |
| 84 |
Problem with copying information from user space. |
| 85 |
.TP |
| 86 |
.B EINTR |
| 87 |
The pause has been interrupted by a signal that was |
| 88 |
delivered to the thread (see |
| 89 |
.BR signal (7)). |
| 90 |
The remaining sleep time has been written |
| 91 |
into |
| 92 |
.I *rem |
| 93 |
so that the thread can easily call |
| 94 |
.BR nanosleep () |
| 95 |
again and continue with the pause. |
| 96 |
.TP |
| 97 |
.B EINVAL |
| 98 |
The value in the |
| 99 |
.I tv_nsec |
| 100 |
field was not in the range [0, 999999999] or |
| 101 |
.I tv_sec |
| 102 |
was negative. |
| 103 |
.SH VERSIONS |
| 104 |
POSIX.1 specifies that |
| 105 |
.BR nanosleep () |
| 106 |
should measure time against the |
| 107 |
.B CLOCK_REALTIME |
| 108 |
clock. |
| 109 |
However, Linux measures the time using the |
| 110 |
.B CLOCK_MONOTONIC |
| 111 |
clock. |
| 112 |
.\" See also http://thread.gmane.org/gmane.linux.kernel/696854/ |
| 113 |
.\" Subject: nanosleep() uses CLOCK_MONOTONIC, should be CLOCK_REALTIME? |
| 114 |
.\" Date: 2008-06-22 07:35:41 GMT |
| 115 |
This probably does not matter, since the POSIX.1 specification for |
| 116 |
.BR clock_settime (2) |
| 117 |
says that discontinuous changes in |
| 118 |
.B CLOCK_REALTIME |
| 119 |
should not affect |
| 120 |
.BR nanosleep (): |
| 121 |
.RS |
| 122 |
.P |
| 123 |
Setting the value of the |
| 124 |
.B CLOCK_REALTIME |
| 125 |
clock via |
| 126 |
.BR clock_settime (2) |
| 127 |
shall |
| 128 |
have no effect on threads that are blocked waiting for a relative time |
| 129 |
service based upon this clock, including the |
| 130 |
.BR nanosleep () |
| 131 |
function. |
| 132 |
Consequently, |
| 133 |
these time services shall expire when the requested duration elapses, |
| 134 |
independently of the new or old value of the clock. |
| 135 |
.RE |
| 136 |
.SH STANDARDS |
| 137 |
POSIX.1-2024. |
| 138 |
.SH HISTORY |
| 139 |
POSIX.1-2001. |
| 140 |
.P |
| 141 |
In order to support applications requiring much more precise pauses |
| 142 |
(e.g., in order to control some time-critical hardware), |
| 143 |
.BR nanosleep () |
| 144 |
would handle pauses of up to 2 milliseconds by busy waiting with microsecond |
| 145 |
precision when called from a thread scheduled under a real-time policy |
| 146 |
like |
| 147 |
.B SCHED_FIFO |
| 148 |
or |
| 149 |
.BR SCHED_RR . |
| 150 |
This special extension was removed in Linux 2.5.39, |
| 151 |
and is thus not available in Linux 2.6.0 and later kernels. |
| 152 |
.SH NOTES |
| 153 |
If the |
| 154 |
.I duration |
| 155 |
is not an exact multiple of the granularity underlying clock (see |
| 156 |
.BR time (7)), |
| 157 |
then the interval will be rounded up to the next multiple. |
| 158 |
Furthermore, after the sleep completes, there may still be a delay before |
| 159 |
the CPU becomes free to once again execute the calling thread. |
| 160 |
.P |
| 161 |
The fact that |
| 162 |
.BR nanosleep () |
| 163 |
sleeps for a relative interval can be problematic if the call |
| 164 |
is repeatedly restarted after being interrupted by signals, |
| 165 |
since the time between the interruptions and restarts of the call |
| 166 |
will lead to drift in the time when the sleep finally completes. |
| 167 |
This problem can be avoided by using |
| 168 |
.BR clock_nanosleep (2) |
| 169 |
with an absolute time value. |
| 170 |
.SH BUGS |
| 171 |
If a program that catches signals and uses |
| 172 |
.BR nanosleep () |
| 173 |
receives signals at a very high rate, |
| 174 |
then scheduling delays and rounding errors in the kernel's |
| 175 |
calculation of the sleep interval and the returned |
| 176 |
.I remain |
| 177 |
value mean that the |
| 178 |
.I remain |
| 179 |
value may steadily |
| 180 |
.I increase |
| 181 |
on successive restarts of the |
| 182 |
.BR nanosleep () |
| 183 |
call. |
| 184 |
To avoid such problems, use |
| 185 |
.BR clock_nanosleep (2) |
| 186 |
with the |
| 187 |
.B TIMER_ABSTIME |
| 188 |
flag to sleep to an absolute deadline. |
| 189 |
.P |
| 190 |
In Linux 2.4, if |
| 191 |
.BR nanosleep () |
| 192 |
is stopped by a signal (e.g., |
| 193 |
.BR SIGTSTP ), |
| 194 |
then the call fails with the error |
| 195 |
.B EINTR |
| 196 |
after the thread is resumed by a |
| 197 |
.B SIGCONT |
| 198 |
signal. |
| 199 |
If the system call is subsequently restarted, |
| 200 |
then the time that the thread spent in the stopped state is |
| 201 |
.I not |
| 202 |
counted against the sleep interval. |
| 203 |
This problem is fixed in Linux 2.6.0 and later kernels. |
| 204 |
.SH SEE ALSO |
| 205 |
.BR clock_nanosleep (2), |
| 206 |
.BR restart_syscall (2), |
| 207 |
.BR sched_setscheduler (2), |
| 208 |
.BR timer_create (2), |
| 209 |
.BR sleep (3), |
| 210 |
.BR timespec (3), |
| 211 |
.BR usleep (3), |
| 212 |
.BR time (7) |