1 .\" Copyright (c) 1992 Drew Eckhardt <drew@cs.colorado.edu>, March 28, 1992
2 .\" and Copyright (c) Michael Kerrisk, 2001, 2002, 2005, 2013
4 .\" %%%LICENSE_START(GPL_NOVERSION_ONELINE)
5 .\" May be distributed under the GNU General Public License.
8 .\" Modified by Michael Haardt <michael@moria.de>
9 .\" Modified 24 Jul 1993 by Rik Faith <faith@cs.unc.edu>
10 .\" Modified 21 Aug 1994 by Michael Chastain <mec@shell.portal.com>:
11 .\" New man page (copied from 'fork.2').
12 .\" Modified 10 June 1995 by Andries Brouwer <aeb@cwi.nl>
13 .\" Modified 25 April 1998 by Xavier Leroy <Xavier.Leroy@inria.fr>
14 .\" Modified 26 Jun 2001 by Michael Kerrisk
15 .\" Mostly upgraded to 2.4.x
16 .\" Added prototype for sys_clone() plus description
17 .\" Added CLONE_THREAD with a brief description of thread groups
18 .\" Added CLONE_PARENT and revised entire page remove ambiguity
19 .\" between "calling process" and "parent process"
20 .\" Added CLONE_PTRACE and CLONE_VFORK
21 .\" Added EPERM and EINVAL error codes
22 .\" Renamed "__clone" to "clone" (which is the prototype in <sched.h>)
23 .\" various other minor tidy ups and clarifications.
24 .\" Modified 26 Jun 2001 by Michael Kerrisk <mtk.manpages@gmail.com>
25 .\" Updated notes for 2.4.7+ behavior of CLONE_THREAD
26 .\" Modified 15 Oct 2002 by Michael Kerrisk <mtk.manpages@gmail.com>
27 .\" Added description for CLONE_NEWNS, which was added in 2.4.19
28 .\" Slightly rephrased, aeb.
29 .\" Modified 1 Feb 2003 - added CLONE_SIGHAND restriction, aeb.
30 .\" Modified 1 Jan 2004 - various updates, aeb
31 .\" Modified 2004-09-10 - added CLONE_PARENT_SETTID etc. - aeb.
32 .\" 2005-04-12, mtk, noted the PID caching behavior of NPTL's getpid()
33 .\" wrapper under BUGS.
34 .\" 2005-05-10, mtk, added CLONE_SYSVSEM, CLONE_UNTRACED, CLONE_STOPPED.
35 .\" 2005-05-17, mtk, Substantially enhanced discussion of CLONE_THREAD.
36 .\" 2008-11-18, mtk, order CLONE_* flags alphabetically
37 .\" 2008-11-18, mtk, document CLONE_NEWPID
38 .\" 2008-11-19, mtk, document CLONE_NEWUTS
39 .\" 2008-11-19, mtk, document CLONE_NEWIPC
40 .\" 2008-11-19, Jens Axboe, mtk, document CLONE_IO
42 .TH CLONE 2 2017-09-15 "Linux" "Linux Programmer's Manual"
44 clone, __clone2 \- create a child process
47 /* Prototype for the glibc wrapper function */
49 .B #define _GNU_SOURCE
52 .BI "int clone(int (*" "fn" ")(void *), void *" child_stack ,
53 .BI " int " flags ", void *" "arg" ", ... "
54 .BI " /* pid_t *" ptid ", void *" newtls \
55 ", pid_t *" ctid " */ );"
57 /* For the prototype of the raw system call, see NOTES */
61 creates a new process, in a manner similar to
64 This page describes both the glibc
66 wrapper function and the underlying system call on which it is based.
67 The main text describes the wrapper function;
68 the differences for the raw system call
69 are described toward the end of this page.
74 allows the child process to share parts of its execution context with
75 the calling process, such as the virtual address space, the table of file
76 descriptors, and the table of signal handlers.
77 (Note that on this manual
78 page, "calling process" normally corresponds to "parent process".
79 But see the description of
85 is to implement threads: multiple flows of control in a program that
86 run concurrently in a shared address space.
88 When the child process is created with
90 it commences execution by calling the function pointed to by the argument
94 where execution continues in the child from the point
100 argument is passed as the argument of the function
105 function returns, the child process terminates.
106 The integer returned by
108 is the exit status for the child process.
109 The child process may also terminate explicitly by calling
111 or after receiving a fatal signal.
115 argument specifies the location of the stack used by the child process.
116 Since the child and calling process may share memory,
117 it is not possible for the child process to execute in the
118 same stack as the calling process.
119 The calling process must therefore
120 set up memory space for the child stack and pass a pointer to this
123 Stacks grow downward on all processors that run Linux
124 (except the HP PA processors), so
126 usually points to the topmost address of the memory space set up for
131 contains the number of the
132 .I "termination signal"
133 sent to the parent when the child dies.
134 If this signal is specified as anything other than
136 then the parent process must specify the
140 options when waiting for the child with
142 If no signal is specified, then the parent process is not signaled
143 when the child terminates.
146 may also be bitwise-ORed with zero or more of the following constants,
147 in order to specify what is shared between the calling process
148 and the child process:
150 .BR CLONE_CHILD_CLEARTID " (since Linux 2.5.49)"
151 Clear (zero) the child thread ID at the location
153 in child memory when the child exits, and do a wakeup on the futex
155 The address involved may be changed by the
156 .BR set_tid_address (2)
158 This is used by threading libraries.
160 .BR CLONE_CHILD_SETTID " (since Linux 2.5.49)"
161 Store the child thread ID at the location
163 in the child's memory.
164 The store operation completes before
166 returns control to user space.
168 .BR CLONE_FILES " (since Linux 2.0)"
171 is set, the calling process and the child process share the same file
173 Any file descriptor created by the calling process or by the child
174 process is also valid in the other process.
175 Similarly, if one of the processes closes a file descriptor,
176 or changes its associated flags (using the
179 operation), the other process is also affected.
180 If a process sharing a file descriptor table calls
182 its file descriptor table is duplicated (unshared).
186 is not set, the child process inherits a copy of all file descriptors
187 opened in the calling process at the time of
189 Subsequent operations that open or close file descriptors,
190 or change file descriptor flags,
191 performed by either the calling
192 process or the child process do not affect the other process.
194 that the duplicated file descriptors in the child refer to the same open file
195 descriptions as the corresponding file descriptors in the calling process,
196 and thus share file offsets and file status flags (see
199 .BR CLONE_FS " (since Linux 2.0)"
202 is set, the caller and the child process share the same filesystem
204 This includes the root of the filesystem, the current
205 working directory, and the umask.
211 performed by the calling process or the child process also affects the
216 is not set, the child process works on a copy of the filesystem
217 information of the calling process at the time of the
225 performed later by one of the processes do not affect the other process.
227 .BR CLONE_IO " (since Linux 2.6.25)"
230 is set, then the new process shares an I/O context with
232 If this flag is not set, then (as with
234 the new process has its own I/O context.
236 .\" The following based on text from Jens Axboe
237 The I/O context is the I/O scope of the disk scheduler (i.e.,
238 what the I/O scheduler uses to model scheduling of a process's I/O).
239 If processes share the same I/O context,
240 they are treated as one by the I/O scheduler.
241 As a consequence, they get to share disk time.
242 For some I/O schedulers,
243 .\" the anticipatory and CFQ scheduler
244 if two processes share an I/O context,
245 they will be allowed to interleave their disk access.
246 If several threads are doing I/O on behalf of the same process
248 for instance), they should employ
250 to get better I/O performance.
253 If the kernel is not configured with the
255 option, this flag is a no-op.
257 .BR CLONE_NEWCGROUP " (since Linux 4.6)"
258 Create the process in a new cgroup namespace.
259 If this flag is not set, then (as with
261 the process is created in the same cgroup namespaces as the calling process.
262 This flag is intended for the implementation of containers.
264 For further information on cgroup namespaces, see
265 .BR cgroup_namespaces (7).
267 Only a privileged process
268 .RB ( CAP_SYS_ADMIN )
270 .BR CLONE_NEWCGROUP .
273 .BR CLONE_NEWIPC " (since Linux 2.6.19)"
276 is set, then create the process in a new IPC namespace.
277 If this flag is not set, then (as with
279 the process is created in the same IPC namespace as
281 This flag is intended for the implementation of containers.
283 An IPC namespace provides an isolated view of System\ V IPC objects (see
285 and (since Linux 2.6.30)
286 .\" commit 7eafd7c74c3f2e67c27621b987b28397110d643f
287 .\" https://lwn.net/Articles/312232/
290 .BR mq_overview (7)).
291 The common characteristic of these IPC mechanisms is that IPC
292 objects are identified by mechanisms other than filesystem
295 Objects created in an IPC namespace are visible to all other processes
296 that are members of that namespace,
297 but are not visible to processes in other IPC namespaces.
299 When an IPC namespace is destroyed
300 (i.e., when the last process that is a member of the namespace terminates),
301 all IPC objects in the namespace are automatically destroyed.
303 Only a privileged process
304 .RB ( CAP_SYS_ADMIN )
307 This flag can't be specified in conjunction with
310 For further information on IPC namespaces, see
313 .BR CLONE_NEWNET " (since Linux 2.6.24)"
314 (The implementation of this flag was completed only
315 by about kernel version 2.6.29.)
319 is set, then create the process in a new network namespace.
320 If this flag is not set, then (as with
322 the process is created in the same network namespace as
324 This flag is intended for the implementation of containers.
326 A network namespace provides an isolated view of the networking stack
327 (network device interfaces, IPv4 and IPv6 protocol stacks,
328 IP routing tables, firewall rules, the
332 directory trees, sockets, etc.).
333 A physical network device can live in exactly one
335 A virtual network device ("veth") pair provides a pipe-like abstraction
336 .\" FIXME . Add pointer to veth(4) page when it is eventually completed
337 that can be used to create tunnels between network namespaces,
338 and can be used to create a bridge to a physical network device
339 in another namespace.
341 When a network namespace is freed
342 (i.e., when the last process in the namespace terminates),
343 its physical network devices are moved back to the
344 initial network namespace (not to the parent of the process).
345 For further information on network namespaces, see
348 Only a privileged process
349 .RB ( CAP_SYS_ADMIN )
353 .BR CLONE_NEWNS " (since Linux 2.4.19)"
356 is set, the cloned child is started in a new mount namespace,
357 initialized with a copy of the namespace of the parent.
360 is not set, the child lives in the same mount
361 namespace as the parent.
363 Only a privileged process
364 .RB ( CAP_SYS_ADMIN )
367 It is not permitted to specify both
371 .\" See https://lwn.net/Articles/543273/
376 For further information on mount namespaces, see
379 .BR mount_namespaces (7).
381 .BR CLONE_NEWPID " (since Linux 2.6.24)"
382 .\" This explanation draws a lot of details from
383 .\" http://lwn.net/Articles/259217/
384 .\" Authors: Pavel Emelyanov <xemul@openvz.org>
385 .\" and Kir Kolyshkin <kir@openvz.org>
387 .\" The primary kernel commit is 30e49c263e36341b60b735cbef5ca37912549264
388 .\" Author: Pavel Emelyanov <xemul@openvz.org>
391 is set, then create the process in a new PID namespace.
392 If this flag is not set, then (as with
394 the process is created in the same PID namespace as
396 This flag is intended for the implementation of containers.
398 For further information on PID namespaces, see
401 .BR pid_namespaces (7).
403 Only a privileged process
404 .RB ( CAP_SYS_ADMIN )
407 This flag can't be specified in conjunction with
413 (This flag first became meaningful for
418 semantics were merged in Linux 3.5,
419 and the final pieces to make the user namespaces completely usable were
420 merged in Linux 3.8.)
424 is set, then create the process in a new user namespace.
425 If this flag is not set, then (as with
427 the process is created in the same user namespace as the calling process.
429 Before Linux 3.8, use of
431 required that the caller have three capabilities:
436 .\" Before Linux 2.6.29, it appears that only CAP_SYS_ADMIN was needed
437 Starting with Linux 3.8,
438 no privileges are needed to create a user namespace.
440 This flag can't be specified in conjunction with
444 For security reasons,
445 .\" commit e66eded8309ebf679d3d3c1f5820d1f2ca332c71
446 .\" https://lwn.net/Articles/543273/
447 .\" The fix actually went into 3.9 and into 3.8.3. However, user namespaces
448 .\" were, for practical purposes, unusable in earlier 3.8.x because of the
449 .\" various filesystems that didn't support userns.
451 cannot be specified in conjunction with
454 For further information on user namespaces, see
457 .BR user_namespaces (7).
459 .BR CLONE_NEWUTS " (since Linux 2.6.19)"
462 is set, then create the process in a new UTS namespace,
463 whose identifiers are initialized by duplicating the identifiers
464 from the UTS namespace of the calling process.
465 If this flag is not set, then (as with
467 the process is created in the same UTS namespace as
469 This flag is intended for the implementation of containers.
471 A UTS namespace is the set of identifiers returned by
473 among these, the domain name and the hostname can be modified by
474 .BR setdomainname (2)
478 Changes made to the identifiers in a UTS namespace
479 are visible to all other processes in the same namespace,
480 but are not visible to processes in other UTS namespaces.
482 Only a privileged process
483 .RB ( CAP_SYS_ADMIN )
487 For further information on UTS namespaces, see
490 .BR CLONE_PARENT " (since Linux 2.3.12)"
493 is set, then the parent of the new child (as returned by
495 will be the same as that of the calling process.
499 is not set, then (as with
501 the child's parent is the calling process.
503 Note that it is the parent process, as returned by
505 which is signaled when the child terminates, so that
508 is set, then the parent of the calling process, rather than the
509 calling process itself, will be signaled.
511 .BR CLONE_PARENT_SETTID " (since Linux 2.5.49)"
512 Store the child thread ID at the location
514 in the parent's memory.
515 (In Linux 2.5.32-2.5.48 there was a flag
518 The store operation completes before
520 returns control to user space.
522 .BR CLONE_PID " (Linux 2.0 to 2.5.15)"
525 is set, the child process is created with the same process ID as
527 This is good for hacking the system, but otherwise
529 From Linux 2.3.21 onward, this flag could be
530 specified only by the system boot process (PID 0).
531 The flag disappeared completely from the kernel sources in Linux 2.5.16.
532 Since then, the kernel silently ignores this bit if it is specified in
535 .BR CLONE_PTRACE " (since Linux 2.2)"
538 is specified, and the calling process is being traced,
539 then trace the child also (see
542 .BR CLONE_SETTLS " (since Linux 2.5.32)"
543 The TLS (Thread Local Storage) descriptor is set to
546 The interpretation of
548 and the resulting effect is architecture dependent.
552 .IR "struct user_desc\ *"
554 .BR set_thread_area (2)).
555 On x86-64 it is the new value to be set for the %fs base register
560 On architectures with a dedicated TLS register, it is the new value
563 .BR CLONE_SIGHAND " (since Linux 2.0)"
566 is set, the calling process and the child process share the same table of
568 If the calling process or child process calls
570 to change the behavior associated with a signal, the behavior is
571 changed in the other process as well.
572 However, the calling process and child
573 processes still have distinct signal masks and sets of pending
575 So, one of them may block or unblock signals using
577 without affecting the other process.
581 is not set, the child process inherits a copy of the signal handlers
582 of the calling process at the time
587 performed later by one of the processes have no effect on the other
590 Since Linux 2.6.0-test6,
598 .BR CLONE_STOPPED " (since Linux 2.6.0-test2)"
601 is set, then the child is initially stopped (as though it was sent a
603 signal), and must be resumed by sending it a
609 from Linux 2.6.25 onward,
612 altogether in Linux 2.6.38.
613 Since then, the kernel silently ignores it without error.
614 .\" glibc 2.8 removed this defn from bits/sched.h
615 Starting with Linux 4.6, the same bit was reused for the
619 .BR CLONE_SYSVSEM " (since Linux 2.5.10)"
622 is set, then the child and the calling process share
623 a single list of System V semaphore adjustment
627 In this case, the shared list accumulates
629 values across all processes sharing the list,
630 and semaphore adjustments are performed only when the last process
631 that is sharing the list terminates (or ceases sharing the list using
633 If this flag is not set, then the child has a separate
635 list that is initially empty.
637 .BR CLONE_THREAD " (since Linux 2.4.0-test8)"
640 is set, the child is placed in the same thread group as the calling process.
641 To make the remainder of the discussion of
643 more readable, the term "thread" is used to refer to the
644 processes within a thread group.
646 Thread groups were a feature added in Linux 2.4 to support the
647 POSIX threads notion of a set of threads that share a single PID.
648 Internally, this shared PID is the so-called
649 thread group identifier (TGID) for the thread group.
650 Since Linux 2.4, calls to
652 return the TGID of the caller.
654 The threads within a group can be distinguished by their (system-wide)
655 unique thread IDs (TID).
656 A new thread's TID is available as the function result
657 returned to the caller of
659 and a thread can obtain
663 When a call is made to
667 then the resulting thread is placed in a new thread group
668 whose TGID is the same as the thread's TID.
671 of the new thread group.
673 A new thread created with
675 has the same parent process as the caller of
681 return the same value for all of the threads in a thread group.
684 thread terminates, the thread that created it using
688 (or other termination) signal;
689 nor can the status of such a thread be obtained
692 (The thread is said to be
695 After all of the threads in a thread group terminate
696 the parent process of the thread group is sent a
698 (or other termination) signal.
700 If any of the threads in a thread group performs an
702 then all threads other than the thread group leader are terminated,
703 and the new program is executed in the thread group leader.
705 If one of the threads in a thread group creates a child using
707 then any thread in the group can
718 (and note that, since Linux 2.6.0-test6,
724 Signals may be sent to a thread group as a whole (i.e., a TGID) using
726 or to a specific thread (i.e., TID) using
729 Signal dispositions and actions are process-wide:
730 if an unhandled signal is delivered to a thread, then
731 it will affect (terminate, stop, continue, be ignored in)
732 all members of the thread group.
734 Each thread has its own signal mask, as set by
736 but signals can be pending either: for the whole process
737 (i.e., deliverable to any member of the thread group),
740 or for an individual thread, when sent with
744 returns a signal set that is the union of the signals pending for the
745 whole process and the signals that are pending for the calling thread.
749 is used to send a signal to a thread group,
750 and the thread group has installed a handler for the signal, then
751 the handler will be invoked in exactly one, arbitrarily selected
752 member of the thread group that has not blocked the signal.
753 If multiple threads in a group are waiting to accept the same signal using
755 the kernel will arbitrarily select one of these threads
756 to receive a signal sent using
759 .BR CLONE_UNTRACED " (since Linux 2.5.46)"
762 is specified, then a tracing process cannot force
764 on this child process.
766 .BR CLONE_VFORK " (since Linux 2.2)"
769 is set, the execution of the calling process is suspended
770 until the child releases its virtual memory
771 resources via a call to
780 is not set, then both the calling process and the child are schedulable
781 after the call, and an application should not rely on execution occurring
782 in any particular order.
784 .BR CLONE_VM " (since Linux 2.0)"
787 is set, the calling process and the child process run in the same memory
789 In particular, memory writes performed by the calling process
790 or by the child process are also visible in the other process.
791 Moreover, any memory mapping or unmapping performed with
795 by the child or calling process also affects the other process.
799 is not set, the child process runs in a separate copy of the memory
800 space of the calling process at the time of
802 Memory writes or file mappings/unmappings performed by one of the
803 processes do not affect the other, as with
808 wrapper function makes some changes
809 in the memory pointed to by
811 (changes required to set the stack up correctly for the child)
818 is used to recursively create children,
819 do not use the buffer employed for the parent's stack
820 as the stack of the child.
822 .SS C library/kernel differences
825 system call corresponds more closely to
827 in that execution in the child continues from the point of the
835 wrapper function are omitted.
837 Another difference for the raw
839 system call is that the
841 argument may be zero,
842 in which case the child uses a duplicate of the parent's stack.
843 (Copy-on-write semantics ensure that the child gets separate copies
844 of stack pages when either process modifies the stack.)
845 In this case, for correct operation, the
847 option should not be specified.
850 the parent's memory because of the use of the
853 then no copy-on-write duplication occurs and chaos is likely to result.)
855 The order of the arguments also differs in the raw system call,
856 and there are variations in the arguments across architectures,
857 as detailed in the following paragraphs.
859 The raw system call interface on x86-64 and some other architectures
860 (including sh, tile, and alpha) is roughly:
864 .BI "long clone(unsigned long " flags ", void *" child_stack ,
865 .BI " int *" ptid ", int *" ctid ,
866 .BI " unsigned long " newtls );
870 On x86-32, and several other common architectures
871 (including score, ARM, ARM 64, PA-RISC, arc, Power PC, xtensa,
873 .\" CONFIG_CLONE_BACKWARDS
874 the order of the last two arguments is reversed:
878 .BI "long clone(unsigned long " flags ", void *" child_stack ,
879 .BI " int *" ptid ", unsigned long " newtls ,
884 On the cris and s390 architectures,
885 .\" CONFIG_CLONE_BACKWARDS2
886 the order of the first two arguments is reversed:
890 .BI "long clone(void *" child_stack ", unsigned long " flags ,
891 .BI " int *" ptid ", int *" ctid ,
892 .BI " unsigned long " newtls );
896 On the microblaze architecture,
897 .\" CONFIG_CLONE_BACKWARDS3
898 an additional argument is supplied:
902 .BI "long clone(unsigned long " flags ", void *" child_stack ,
903 .BI " int " stack_size , "\fR /* Size of stack */"
904 .BI " int *" ptid ", int *" ctid ,
905 .BI " unsigned long " newtls );
909 .SS blackfin, m68k, and sparc
910 .\" Mike Frysinger noted in a 2013 mail:
911 .\" these arches don't define __ARCH_WANT_SYS_CLONE:
912 .\" blackfin ia64 m68k sparc
913 The argument-passing conventions on
914 blackfin, m68k, and sparc are different from the descriptions above.
915 For details, see the kernel (and glibc) source.
917 On ia64, a different interface is used:
920 .BI "int __clone2(int (*" "fn" ")(void *), "
921 .BI " void *" child_stack_base ", size_t " stack_size ,
922 .BI " int " flags ", void *" "arg" ", ... "
923 .BI " /* pid_t *" ptid ", struct user_desc *" tls \
924 ", pid_t *" ctid " */ );"
927 The prototype shown above is for the glibc wrapper function;
928 the raw system call interface has no
932 argument, and changes the order of the arguments so that
934 is the first argument, and
936 is the last argument.
939 operates in the same way as
943 points to the lowest address of the child's stack area,
946 specifies the size of the stack pointed to by
947 .IR child_stack_base .
948 .SS Linux 2.4 and earlier
949 In Linux 2.4 and earlier,
951 does not take arguments
957 .\" gettid(2) returns current->pid;
958 .\" getpid(2) returns current->tgid;
959 On success, the thread ID of the child process is returned
960 in the caller's thread of execution.
961 On failure, \-1 is returned
962 in the caller's context, no child process will be created, and
964 will be set appropriately.
968 Too many processes are already running; see
976 (Since Linux 2.6.0-test6.)
983 (Since Linux 2.5.35.)
987 .\" .B CLONE_DETACHED
991 .\" (Since Linux 2.6.0-test6.)
994 .\" commit e66eded8309ebf679d3d3c1f5820d1f2ca332c71
1002 .BR EINVAL " (since Linux 3.9)"
1023 and one (or both) of
1031 Returned by the glibc
1033 wrapper function when
1037 is specified as NULL.
1043 but the kernel was not configured with the
1053 but the kernel was not configured with the
1061 but the kernel was not configured with the
1069 but the kernel was not configured with the
1075 is not aligned to a suitable boundary for this architecture.
1076 For example, on aarch64,
1078 must be a multiple of 16.
1081 Cannot allocate sufficient memory to allocate a task structure for the
1082 child, or to copy those parts of the caller's context that need to be
1085 .BR ENOSPC " (since Linux 3.7)"
1086 .\" commit f2302505775fd13ba93f034206f1e2a587017929
1088 was specified in flags,
1089 but the limit on the nesting depth of PID namespaces
1090 would have been exceeded; see
1091 .BR pid_namespaces (7).
1093 .BR ENOSPC " (since Linux 4.9; beforehand " EUSERS )
1097 and the call would cause the limit on the number of
1098 nested user namespaces to be exceeded.
1100 .BR user_namespaces (7).
1102 From Linux 3.11 to Linux 4.8, the error diagnosed in this case was
1105 .BR ENOSPC " (since Linux 4.9)"
1106 One of the values in
1108 specified the creation of a new user namespace,
1109 but doing so would have caused the limit defined by the corresponding file in
1112 For further details, see
1116 .BR CLONE_NEWCGROUP ,
1123 was specified by an unprivileged process (process without \fBCAP_SYS_ADMIN\fP).
1127 was specified by a process other than process 0.
1128 (This error occurs only on Linux 2.5.15 and earlier.)
1134 but either the effective user ID or the effective group ID of the caller
1135 does not have a mapping in the parent namespace (see
1136 .BR user_namespaces (7)).
1138 .BR EPERM " (since Linux 3.9)"
1139 .\" commit 3151527ee007b73a0ebd296010f1c0454a919c7d
1143 and the caller is in a chroot environment
1144 .\" FIXME What is the rationale for this restriction?
1145 (i.e., the caller's root directory does not match the root directory
1146 of the mount namespace in which it resides).
1148 .BR ERESTARTNOINTR " (since Linux 2.6.17)"
1149 .\" commit 4a2c7a7837da1b91468e50426066d988050e4d56
1150 System call was interrupted by a signal and will be restarted.
1151 (This can be seen only during a trace.)
1153 .BR EUSERS " (Linux 3.11 to Linux 4.8)"
1157 and the limit on the number of nested user namespaces would be exceeded.
1158 See the discussion of the
1162 .\" There is no entry for
1167 .\" as described in this manual page.
1170 is Linux-specific and should not be used in programs
1171 intended to be portable.
1175 system call can be used to test whether two processes share various
1176 resources such as a file descriptor table,
1177 System V semaphore undo operations, or a virtual address space.
1180 Handlers registered using
1181 .BR pthread_atfork (3)
1182 are not executed during a call to
1185 In the Linux 2.4.x series,
1187 generally does not make the parent of the new thread the same
1188 as the parent of the calling process.
1189 However, for kernel versions 2.4.7 to 2.4.18 the
1193 flag (as in Linux 2.6.0 and later).
1195 For a while there was
1197 (introduced in 2.5.32):
1198 parent wants no child-exit signal.
1199 In Linux 2.6.2, the need to give this flag together with
1202 This flag is still defined, but has no effect.
1206 should not be called through vsyscall, but directly through
1209 GNU C library versions 2.3.4 up to and including 2.24
1210 contained a wrapper function for
1212 that performed caching of PIDs.
1213 This caching relied on support in the glibc wrapper for
1215 but limitations in the implementation
1216 meant that the cache was not up to date in some circumstances.
1218 if a signal was delivered to the child immediately after the
1220 call, then a call to
1222 in a handler for the signal could return the PID
1223 of the calling process ("the parent"),
1224 if the clone wrapper had not yet had a chance to update the PID
1226 (This discussion ignores the case where the child was created using
1231 return the same value in the child and in the process that called
1233 since the caller and the child are in the same thread group.
1234 The stale-cache problem also does not occur if the
1238 To get the truth, it was sometimes necessary to use code such as the following:
1242 #include <syscall.h>
1246 mypid = syscall(SYS_getpid);
1249 .\" See also the following bug reports
1250 .\" https://bugzilla.redhat.com/show_bug.cgi?id=417521
1251 .\" http://sourceware.org/bugzilla/show_bug.cgi?id=6910
1253 Because of the stale-cache problem, as well as other problems noted in
1255 the PID caching feature was removed in glibc 2.25.
1257 The following program demonstrates the use of
1259 to create a child process that executes in a separate UTS namespace.
1260 The child changes the hostname in its UTS namespace.
1261 Both parent and child then display the system hostname,
1262 making it possible to see that the hostname
1263 differs in the UTS namespaces of the parent and child.
1264 For an example of the use of this program, see
1269 #include <sys/wait.h>
1270 #include <sys/utsname.h>
1277 #define errExit(msg) do { perror(msg); exit(EXIT_FAILURE); \\
1280 static int /* Start function for cloned child */
1281 childFunc(void *arg)
1285 /* Change hostname in UTS namespace of child */
1287 if (sethostname(arg, strlen(arg)) == \-1)
1288 errExit("sethostname");
1290 /* Retrieve and display hostname */
1292 if (uname(&uts) == \-1)
1294 printf("uts.nodename in child: %s\\n", uts.nodename);
1296 /* Keep the namespace open for a while, by sleeping.
1297 This allows some experimentation\-\-for example, another
1298 process might join the namespace. */
1302 return 0; /* Child terminates now */
1305 #define STACK_SIZE (1024 * 1024) /* Stack size for cloned child */
1308 main(int argc, char *argv[])
1310 char *stack; /* Start of stack buffer */
1311 char *stackTop; /* End of stack buffer */
1316 fprintf(stderr, "Usage: %s <child\-hostname>\\n", argv[0]);
1320 /* Allocate stack for child */
1322 stack = malloc(STACK_SIZE);
1325 stackTop = stack + STACK_SIZE; /* Assume stack grows downward */
1327 /* Create child that has its own UTS namespace;
1328 child commences execution in childFunc() */
1330 pid = clone(childFunc, stackTop, CLONE_NEWUTS | SIGCHLD, argv[1]);
1333 printf("clone() returned %ld\\n", (long) pid);
1335 /* Parent falls through to here */
1337 sleep(1); /* Give child time to change its hostname */
1339 /* Display hostname in parent\(aqs UTS namespace. This will be
1340 different from hostname in child\(aqs UTS namespace. */
1342 if (uname(&uts) == \-1)
1344 printf("uts.nodename in parent: %s\\n", uts.nodename);
1346 if (waitpid(pid, NULL, 0) == \-1) /* Wait for child */
1348 printf("child has terminated\\n");
1359 .BR set_thread_area (2),
1360 .BR set_tid_address (2),
1365 .BR capabilities (7),