| 1 |
.\" Copyright 2006, Andi Kleen <[email protected]> |
| 2 |
.\" Copyright, the authors of the Linux man-pages project |
| 3 |
.\" |
| 4 |
.\" SPDX-License-Identifier: Linux-man-pages-1-para |
| 5 |
.\" |
| 6 |
.TH getcpu 2 2026-02-08 "Linux man-pages 6.18" |
| 7 |
.SH NAME |
| 8 |
getcpu \- determine CPU and NUMA node on which the calling thread is running |
| 9 |
.SH LIBRARY |
| 10 |
Standard C library |
| 11 |
.RI ( libc ,\~ \-lc ) |
| 12 |
.SH SYNOPSIS |
| 13 |
.nf |
| 14 |
.BR "#define _GNU_SOURCE" " /* See feature_test_macros(7) */" |
| 15 |
.B #include <sched.h> |
| 16 |
.P |
| 17 |
.BI "int getcpu(unsigned int *_Nullable " cpu ", unsigned int *_Nullable " node ); |
| 18 |
.fi |
| 19 |
.SH DESCRIPTION |
| 20 |
The |
| 21 |
.BR getcpu () |
| 22 |
system call identifies the processor and node on which the calling |
| 23 |
thread or process is currently running and writes them into the |
| 24 |
integers pointed to by the |
| 25 |
.I cpu |
| 26 |
and |
| 27 |
.I node |
| 28 |
arguments. |
| 29 |
The processor is a unique small integer identifying a CPU. |
| 30 |
The node is a unique small identifier identifying a NUMA node. |
| 31 |
When either |
| 32 |
.I cpu |
| 33 |
or |
| 34 |
.I node |
| 35 |
is NULL nothing is written to the respective pointer. |
| 36 |
.P |
| 37 |
The information placed in |
| 38 |
.I cpu |
| 39 |
is guaranteed to be current only at the time of the call: |
| 40 |
unless the CPU affinity has been fixed using |
| 41 |
.BR sched_setaffinity (2), |
| 42 |
the kernel might change the CPU at any time. |
| 43 |
(Normally this does not happen |
| 44 |
because the scheduler tries to minimize movements between CPUs to |
| 45 |
keep caches hot, but it is possible.) |
| 46 |
The caller must allow for the possibility that the information returned in |
| 47 |
.I cpu |
| 48 |
and |
| 49 |
.I node |
| 50 |
is no longer current by the time the call returns. |
| 51 |
.SH RETURN VALUE |
| 52 |
On success, 0 is returned. |
| 53 |
On error, \-1 is returned, and |
| 54 |
.I errno |
| 55 |
is set to indicate the error. |
| 56 |
.SH ERRORS |
| 57 |
.TP |
| 58 |
.B EFAULT |
| 59 |
Arguments point outside the calling process's address space. |
| 60 |
.SH STANDARDS |
| 61 |
Linux. |
| 62 |
.SH HISTORY |
| 63 |
Linux 2.6.19 (x86-64 and i386), |
| 64 |
glibc 2.29. |
| 65 |
.\" |
| 66 |
.SS C library/kernel differences |
| 67 |
The kernel system call has a third argument: |
| 68 |
.P |
| 69 |
.in +4n |
| 70 |
.nf |
| 71 |
.BI "int getcpu(unsigned int *" cpu ", unsigned int *" node , |
| 72 |
.BI " struct getcpu_cache *" tcache ); |
| 73 |
.fi |
| 74 |
.in |
| 75 |
.P |
| 76 |
The |
| 77 |
.I tcache |
| 78 |
argument is unused since Linux 2.6.24, |
| 79 |
and (when invoking the system call directly) |
| 80 |
should be specified as NULL, |
| 81 |
unless portability to Linux 2.6.23 or earlier is required. |
| 82 |
.P |
| 83 |
.\" commit 4307d1e5ada595c87f9a4d16db16ba5edb70dcb1 |
| 84 |
.\" Author: Ingo Molnar <[email protected]> |
| 85 |
.\" Date: Wed Nov 7 18:37:48 2007 +0100 |
| 86 |
.\" x86: ignore the sys_getcpu() tcache parameter |
| 87 |
In Linux 2.6.23 and earlier, if the |
| 88 |
.I tcache |
| 89 |
argument was non-NULL, |
| 90 |
then it specified a pointer to a caller-allocated buffer in thread-local |
| 91 |
storage that was used to provide a caching mechanism for |
| 92 |
.BR getcpu (). |
| 93 |
Use of the cache could speed |
| 94 |
.BR getcpu () |
| 95 |
calls, at the cost that there was a very small chance that |
| 96 |
the returned information would be out of date. |
| 97 |
The caching mechanism was considered to cause problems when |
| 98 |
migrating threads between CPUs, and so the argument is now ignored. |
| 99 |
.\" |
| 100 |
.\" ===== Before Linux 2.6.24: ===== |
| 101 |
.\" .I tcache |
| 102 |
.\" is a pointer to a |
| 103 |
.\" .IR "struct getcpu_cache" |
| 104 |
.\" that is used as a cache by |
| 105 |
.\" .BR getcpu (). |
| 106 |
.\" The caller should put the cache into a thread-local variable |
| 107 |
.\" if the process is multithreaded, |
| 108 |
.\" because the cache cannot be shared between different threads. |
| 109 |
.\" .I tcache |
| 110 |
.\" can be NULL. |
| 111 |
.\" If it is not NULL |
| 112 |
.\" .BR getcpu () |
| 113 |
.\" will use it to speed up operation. |
| 114 |
.\" The information inside the cache is private to the system call |
| 115 |
.\" and should not be accessed by the user program. |
| 116 |
.\" The information placed in the cache can change between Linux releases. |
| 117 |
.\" |
| 118 |
.\" When no cache is specified |
| 119 |
.\" .BR getcpu () |
| 120 |
.\" will be slower, |
| 121 |
.\" but always retrieve the current CPU and node information. |
| 122 |
.\" With a cache |
| 123 |
.\" .BR getcpu () |
| 124 |
.\" is faster. |
| 125 |
.\" However, the cached information is updated only once per jiffy (see |
| 126 |
.\" .BR time (7)). |
| 127 |
.\" This means that the information could theoretically be out of date, |
| 128 |
.\" although in practice the scheduler's attempt to maintain |
| 129 |
.\" soft CPU affinity means that the information is unlikely to change |
| 130 |
.\" over the course of the caching interval. |
| 131 |
.SH NOTES |
| 132 |
Linux makes a best effort to make this call as fast as possible. |
| 133 |
(On some architectures, this is done via an implementation in the |
| 134 |
.BR vdso (7).) |
| 135 |
The intention of |
| 136 |
.BR getcpu () |
| 137 |
is to allow programs to make optimizations with per-CPU data |
| 138 |
or for NUMA optimization. |
| 139 |
.SH SEE ALSO |
| 140 |
.BR mbind (2), |
| 141 |
.BR sched_setaffinity (2), |
| 142 |
.BR set_mempolicy (2), |
| 143 |
.BR sched_getcpu (3), |
| 144 |
.BR cpuset (7), |
| 145 |
.BR vdso (7) |