]>
Commit | Line | Data |
---|---|---|
fea681da | 1 | .\" Copyright (c) 1992 Drew Eckhardt <drew@cs.colorado.edu>, March 28, 1992 |
fb1fa92b | 2 | .\" and Copyright (c) Michael Kerrisk, 2001, 2002, 2005, 2013, 2019 |
2297bf0e | 3 | .\" |
fd0fc519 | 4 | .\" %%%LICENSE_START(GPL_NOVERSION_ONELINE) |
fea681da | 5 | .\" May be distributed under the GNU General Public License. |
fd0fc519 | 6 | .\" %%%LICENSE_END |
dccaff1e | 7 | .\" |
fea681da MK |
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 | |
c13182ef | 18 | .\" Added CLONE_PARENT and revised entire page remove ambiguity |
fea681da MK |
19 | .\" between "calling process" and "parent process" |
20 | .\" Added CLONE_PTRACE and CLONE_VFORK | |
21 | .\" Added EPERM and EINVAL error codes | |
fd8a5be4 | 22 | .\" Renamed "__clone" to "clone" (which is the prototype in <sched.h>) |
fea681da | 23 | .\" various other minor tidy ups and clarifications. |
c11b1abf | 24 | .\" Modified 26 Jun 2001 by Michael Kerrisk <mtk.manpages@gmail.com> |
d9bfdb9c | 25 | .\" Updated notes for 2.4.7+ behavior of CLONE_THREAD |
c11b1abf | 26 | .\" Modified 15 Oct 2002 by Michael Kerrisk <mtk.manpages@gmail.com> |
fea681da MK |
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 | |
0967c11f | 31 | .\" Modified 2004-09-10 - added CLONE_PARENT_SETTID etc. - aeb. |
d9bfdb9c | 32 | .\" 2005-04-12, mtk, noted the PID caching behavior of NPTL's getpid() |
31830ef0 | 33 | .\" wrapper under BUGS. |
fd8a5be4 MK |
34 | .\" 2005-05-10, mtk, added CLONE_SYSVSEM, CLONE_UNTRACED, CLONE_STOPPED. |
35 | .\" 2005-05-17, mtk, Substantially enhanced discussion of CLONE_THREAD. | |
4e836144 | 36 | .\" 2008-11-18, mtk, order CLONE_* flags alphabetically |
82ee147a | 37 | .\" 2008-11-18, mtk, document CLONE_NEWPID |
43ce9dda | 38 | .\" 2008-11-19, mtk, document CLONE_NEWUTS |
667417b3 | 39 | .\" 2008-11-19, mtk, document CLONE_NEWIPC |
cfdc761b | 40 | .\" 2008-11-19, Jens Axboe, mtk, document CLONE_IO |
fea681da | 41 | .\" |
a5409de9 | 42 | .TH CLONE 2 2019-11-19 "Linux" "Linux Programmer's Manual" |
fea681da | 43 | .SH NAME |
a9e52b43 | 44 | clone, __clone2, clone3 \- create a child process |
fea681da | 45 | .SH SYNOPSIS |
c10859eb | 46 | .nf |
81f10dad | 47 | /* Prototype for the glibc wrapper function */ |
dbfe9c70 | 48 | .PP |
4f71ba5d | 49 | .B #define _GNU_SOURCE |
fea681da | 50 | .B #include <sched.h> |
dbfe9c70 | 51 | .PP |
81c2368f MK |
52 | .BI "int clone(int (*" "fn" ")(void *), void *" stack \ |
53 | ", int " flags ", void *" "arg" ", ... " | |
54 | .BI " /* pid_t *" parent_tid ", void *" tls \ | |
55 | ", pid_t *" child_tid " */ );" | |
dbfe9c70 | 56 | .PP |
faa0e55a MK |
57 | /* For the prototype of the raw clone() system call, see NOTES */ |
58 | .PP | |
75e28eba | 59 | .BI "long clone3(struct clone_args *" cl_args ", size_t " size ); |
c10859eb | 60 | .fi |
faa0e55a MK |
61 | .PP |
62 | .IR Note : | |
63 | There is not yet a glibc wrapper for | |
64 | .BR clone3 (); | |
65 | see NOTES. | |
fea681da | 66 | .SH DESCRIPTION |
faa0e55a | 67 | These system calls |
8dd6b0bc | 68 | create a new ("child") process, in a manner similar to |
fea681da | 69 | .BR fork (2). |
efeece04 | 70 | .PP |
8dd6b0bc | 71 | By contrast with |
fea681da | 72 | .BR fork (2), |
225f5da8 | 73 | these system calls provide more precise control over what pieces of execution |
10337567 MK |
74 | context are shared between the calling process and the child process. |
75 | For example, using these system calls, the caller can control whether | |
76 | or not the two processes share the virtual address space, | |
77 | the table of file descriptors, and the table of signal handlers. | |
225f5da8 | 78 | These system calls also allow the new child process to be placed |
10337567 MK |
79 | in separate |
80 | .BR namespaces (7). | |
81 | .PP | |
82 | Note that in this manual | |
c13182ef | 83 | page, "calling process" normally corresponds to "parent process". |
a10c5a33 | 84 | But see the descriptions of |
c13182ef | 85 | .B CLONE_PARENT |
a10c5a33 MK |
86 | and |
87 | .B CLONE_THREAD | |
10337567 | 88 | below. |
efeece04 | 89 | .PP |
faa0e55a MK |
90 | This page describes the following interfaces: |
91 | .IP * 3 | |
92 | The glibc | |
93 | .BR clone () | |
94 | wrapper function and the underlying system call on which it is based. | |
95 | The main text describes the wrapper function; | |
96 | the differences for the raw system call | |
97 | are described toward the end of this page. | |
98 | .IP * | |
99 | The newer | |
100 | .BR clone3 () | |
101 | system call. | |
5261b0fe MK |
102 | .PP |
103 | In the remainder of this page, the terminology "the clone call" is used | |
324f6154 | 104 | when noting details that apply to all of these interfaces, |
faa0e55a MK |
105 | .\" |
106 | .SS The clone() wrapper function | |
107 | .PP | |
108 | When the child process is created with the | |
109 | .BR clone () | |
110 | wrapper function, | |
7495cbc7 MK |
111 | it commences execution by calling the function pointed to by the argument |
112 | .IR fn . | |
fea681da | 113 | (This differs from |
c13182ef | 114 | .BR fork (2), |
fea681da | 115 | where execution continues in the child from the point |
c13182ef MK |
116 | of the |
117 | .BR fork (2) | |
fea681da MK |
118 | call.) |
119 | The | |
fea681da | 120 | .I arg |
7495cbc7 MK |
121 | argument is passed as the argument of the function |
122 | .IR fn . | |
efeece04 | 123 | .PP |
c13182ef | 124 | When the |
fea681da | 125 | .IR fn ( arg ) |
4ba17a6d | 126 | function returns, the child process terminates. |
c13182ef | 127 | The integer returned by |
fea681da | 128 | .I fn |
4ba17a6d | 129 | is the exit status for the child process. |
c13182ef | 130 | The child process may also terminate explicitly by calling |
fea681da MK |
131 | .BR exit (2) |
132 | or after receiving a fatal signal. | |
efeece04 | 133 | .PP |
fea681da | 134 | The |
81c2368f | 135 | .I stack |
c13182ef MK |
136 | argument specifies the location of the stack used by the child process. |
137 | Since the child and calling process may share memory, | |
fea681da | 138 | it is not possible for the child process to execute in the |
c13182ef MK |
139 | same stack as the calling process. |
140 | The calling process must therefore | |
fea681da MK |
141 | set up memory space for the child stack and pass a pointer to this |
142 | space to | |
edcc65ff | 143 | .BR clone (). |
5fab2e7c | 144 | Stacks grow downward on all processors that run Linux |
fea681da | 145 | (except the HP PA processors), so |
81c2368f | 146 | .I stack |
fea681da MK |
147 | usually points to the topmost address of the memory space set up for |
148 | the child stack. | |
faa0e55a MK |
149 | Note that |
150 | .BR clone () | |
151 | does not provide a means whereby the caller can inform the kernel of the | |
152 | size of the stack area. | |
153 | .PP | |
154 | The remaining arguments to | |
155 | .BR clone () | |
156 | are discussed below. | |
157 | .\" | |
158 | .SS clone3() | |
159 | .PP | |
160 | The | |
161 | .BR clone3 () | |
162 | system call provides a superset of the functionality of the older | |
163 | .BR clone () | |
164 | interface. | |
165 | It also provides a number of API improvements, including: | |
166 | space for additional flags bits; | |
167 | cleaner separation in the use of various arguments; | |
168 | and the ability to specify the size of the child's stack area. | |
169 | .PP | |
170 | As with | |
171 | .BR fork (2), | |
172 | .BR clone3 () | |
173 | returns in both the parent and the child. | |
174 | It returns 0 in the child process and returns the PID of the child | |
175 | in the parent. | |
176 | .PP | |
177 | The | |
178 | .I cl_args | |
179 | argument of | |
180 | .BR clone3 () | |
181 | is a structure of the following form: | |
182 | .PP | |
183 | .in +4n | |
184 | .EX | |
185 | struct clone_args { | |
186 | u64 flags; /* Flags bit mask */ | |
187 | u64 pidfd; /* Where to store PID file descriptor | |
be66dbc7 | 188 | (\fIpid_t *\fP) */ |
faa0e55a | 189 | u64 child_tid; /* Where to store child TID, |
be66dbc7 | 190 | in child's memory (\fIpid_t *\fP) */ |
faa0e55a MK |
191 | u64 parent_tid; /* Where to store child TID, |
192 | in parent's memory (\fIint *\fP) */ | |
193 | u64 exit_signal; /* Signal to deliver to parent on | |
194 | child termination */ | |
195 | u64 stack; /* Pointer to lowest byte of stack */ | |
196 | u64 stack_size; /* Size of stack */ | |
197 | u64 tls; /* Location of new TLS */ | |
f7d5e082 MK |
198 | u64 set_tid; /* Pointer to a \fIpid_t\fP array |
199 | (since Linux 5.5) */ | |
200 | u64 set_tid_size; /* Number of elements in \fIset_tid\fP | |
201 | (since Linux 5.5) */ | |
ed7c1377 MK |
202 | u64 cgroup; /* File descriptor for target cgroup |
203 | of child (since Linux 5.7) */ | |
faa0e55a MK |
204 | }; |
205 | .EE | |
206 | .in | |
207 | .PP | |
208 | The | |
209 | .I size | |
210 | argument that is supplied to | |
211 | .BR clone3 () | |
212 | should be initialized to the size of this structure. | |
213 | (The existence of the | |
214 | .I size | |
215 | argument permits future extensions to the | |
216 | .IR clone_args | |
217 | structure.) | |
218 | .PP | |
219 | The stack for the child process is specified via | |
220 | .IR cl_args.stack , | |
221 | which points to the lowest byte of the stack area, | |
222 | and | |
223 | .IR cl_args.stack_size , | |
224 | which specifies the size of the stack in bytes. | |
225 | In the case where the | |
226 | .BR CLONE_VM | |
227 | flag (see below) is specified, a stack must be explicitly allocated | |
228 | and specified. | |
229 | Otherwise, these two fields can be specified as NULL and 0, | |
230 | which causes the child to use the same stack area as the parent | |
231 | (in the child's own virtual address space). | |
232 | .PP | |
233 | The remaining fields in the | |
234 | .I cl_args | |
235 | argument are discussed below. | |
236 | .\" | |
237 | .SS Equivalence between clone() and clone3() arguments | |
238 | .PP | |
239 | Unlike the older | |
240 | .BR clone () | |
241 | interface, where arguments are passed individually, in the newer | |
242 | .BR clone3 () | |
243 | interface the arguments are packaged into the | |
244 | .I clone_args | |
245 | structure shown above. | |
246 | This structure allows for a superset of the information passed via the | |
247 | .BR clone () | |
248 | arguments. | |
249 | .PP | |
250 | The following table shows the equivalence between the arguments of | |
251 | .BR clone () | |
252 | and the fields in the | |
253 | .I clone_args | |
254 | argument supplied to | |
255 | .BR clone3 (): | |
256 | .RS | |
257 | .TS | |
258 | lb lb lb | |
259 | l l l | |
260 | li li l. | |
97883fae | 261 | clone() clone3() Notes |
faa0e55a | 262 | \fIcl_args\fP field |
f5d5180f | 263 | flags & ~0xff flags For most flags; details below |
faa0e55a MK |
264 | parent_tid pidfd See CLONE_PIDFD |
265 | child_tid child_tid See CLONE_CHILD_SETTID | |
266 | parent_tid parent_tid See CLONE_PARENT_SETTID | |
267 | flags & 0xff exit_signal | |
268 | stack stack | |
269 | \fP---\fP stack_size | |
270 | tls tls See CLONE_SETTLS | |
bf031aaa AR |
271 | \fP---\fP set_tid See below for details |
272 | \fP---\fP set_tid_size | |
ed7c1377 | 273 | \fP---\fP cgroup See CLONE_INTO_CGROUP |
faa0e55a MK |
274 | .TE |
275 | .RE | |
5fbce8f2 MK |
276 | .\" |
277 | .SS The child termination signal | |
efeece04 | 278 | .PP |
faa0e55a MK |
279 | When the child process terminates, a signal may be sent to the parent. |
280 | The termination signal is specified in the low byte of | |
fea681da | 281 | .I flags |
faa0e55a MK |
282 | .RB ( clone ()) |
283 | or in | |
284 | .I cl_args.exit_signal | |
285 | .RB ( clone3 ()). | |
fd8a5be4 | 286 | If this signal is specified as anything other than |
fea681da MK |
287 | .BR SIGCHLD , |
288 | then the parent process must specify the | |
c13182ef MK |
289 | .B __WALL |
290 | or | |
fea681da | 291 | .B __WCLONE |
c13182ef MK |
292 | options when waiting for the child with |
293 | .BR wait (2). | |
faa0e55a | 294 | If no signal (i.e., zero) is specified, then the parent process is not signaled |
fea681da | 295 | when the child terminates. |
5fbce8f2 | 296 | .\" |
bf031aaa AR |
297 | .SS The set_tid array |
298 | .PP | |
299 | By default, the kernel chooses the next sequential PID for the new | |
300 | process in each of the PID namespaces where it is present. | |
301 | When creating a process with | |
302 | .BR clone3 (), | |
303 | the | |
304 | .I set_tid | |
b386cee3 MK |
305 | array (available since Linux 5.5) |
306 | can be used to select specific PIDs for the process in some | |
bf031aaa | 307 | or all of the PID namespaces where it is present. |
ee8bb310 | 308 | If the PID of the newly created process should be set only for the current |
bf031aaa AR |
309 | PID namespace or in the newly created PID namespace (if |
310 | .I flags | |
311 | contains | |
312 | .BR CLONE_NEWPID ) | |
313 | then the first element in the | |
314 | .I set_tid | |
315 | array has to be the desired PID and | |
316 | .I set_tid_size | |
317 | needs to be 1. | |
318 | .PP | |
319 | If the PID of the newly created process should have a certain value in | |
ee8bb310 | 320 | multiple PID namespaces, then the |
bf031aaa | 321 | .I set_tid |
09007c4b MK |
322 | array can have multiple entries. |
323 | The first entry defines the PID in the most | |
ee8bb310 MK |
324 | deeply nested PID namespace and each of the following entries contains |
325 | the PID in the | |
326 | corresponding ancestor PID namespace. | |
09007c4b | 327 | The number of PID namespaces in which a PID |
bf031aaa AR |
328 | should be set is defined by |
329 | .I set_tid_size | |
330 | which cannot be larger than the number of currently nested PID namespaces. | |
331 | .PP | |
332 | To create a process with the following PIDs in a PID namespace hierarchy: | |
333 | .RS | |
334 | .TS | |
ee8bb310 MK |
335 | lb lb lb |
336 | l l l. | |
337 | PID NS level Requested PID Notes | |
338 | 0 31496 Outermost PID namespace | |
bf031aaa | 339 | 1 42 |
ee8bb310 | 340 | 2 7 Innermost PID namespace |
bf031aaa AR |
341 | .TE |
342 | .RE | |
343 | .PP | |
344 | Set the array to: | |
345 | .PP | |
ee8bb310 | 346 | .in +4n |
bf031aaa | 347 | .EX |
ee8bb310 MK |
348 | set_tid[0] = 7; |
349 | set_tid[1] = 42; | |
350 | set_tid[2] = 31496; | |
351 | set_tid_size = 3; | |
bf031aaa | 352 | .EE |
ee8bb310 | 353 | .in |
bf031aaa AR |
354 | .PP |
355 | If only the PIDs in the two innermost PID namespaces | |
356 | need to be specified, set the array to: | |
357 | .PP | |
ee8bb310 | 358 | .in +4n |
bf031aaa | 359 | .EX |
ee8bb310 MK |
360 | set_tid[0] = 7; |
361 | set_tid[1] = 42; | |
362 | set_tid_size = 2; | |
bf031aaa | 363 | .EE |
ee8bb310 | 364 | .in |
bf031aaa AR |
365 | .PP |
366 | The PID in the PID namespaces outside the two innermost PID namespaces | |
367 | will be selected the same way as any other PID is selected. | |
368 | .PP | |
369 | The | |
370 | .I set_tid | |
371 | feature requires | |
ee8bb310 | 372 | .BR CAP_SYS_ADMIN |
bf031aaa AR |
373 | in all owning user namespaces of the target PID namespaces. |
374 | .PP | |
ee8bb310 MK |
375 | Callers may only choose a PID greater than 1 in a given PID namespace |
376 | if an | |
377 | .BR init | |
378 | process (i.e., a process with PID 1) already exists in that namespace. | |
09007c4b | 379 | Otherwise the PID |
bf031aaa AR |
380 | entry for this PID namespace must be 1. |
381 | .\" | |
16853a31 | 382 | .SS The flags mask |
efeece04 | 383 | .PP |
faa0e55a MK |
384 | Both |
385 | .BR clone () | |
386 | and | |
387 | .BR clone3 () | |
388 | allow a flags bit mask that modifies their behavior | |
389 | and allows the caller to specify what is shared between the calling process | |
390 | and the child process. | |
5261b0fe MK |
391 | This bit mask\(emthe |
392 | .I flags | |
393 | argument of | |
394 | .BR clone () | |
395 | or the | |
396 | .I cl_args.flags | |
397 | field passed to | |
398 | .BR clone3 ()\(emis | |
16853a31 MK |
399 | referred to as the |
400 | .I flags | |
401 | mask in the remainder of this page. | |
402 | .PP | |
403 | The | |
404 | .I flags | |
405 | mask is specified as a bitwise-OR of zero or more of | |
406 | the constants listed below. | |
5261b0fe | 407 | Except as noted below, these flags are available |
faa0e55a MK |
408 | (and have the same effect) in both |
409 | .BR clone () | |
410 | and | |
411 | .BR clone3 (). | |
fea681da | 412 | .TP |
f5dbc7c8 | 413 | .BR CLONE_CHILD_CLEARTID " (since Linux 2.5.49)" |
e2bf1234 | 414 | Clear (zero) the child thread ID at the location pointed to by |
81c2368f | 415 | .I child_tid |
faa0e55a MK |
416 | .RB ( clone ()) |
417 | or | |
418 | .I cl_args.child_tid | |
419 | .RB ( clone3 ()) | |
f5dbc7c8 MK |
420 | in child memory when the child exits, and do a wakeup on the futex |
421 | at that address. | |
422 | The address involved may be changed by the | |
423 | .BR set_tid_address (2) | |
424 | system call. | |
425 | This is used by threading libraries. | |
426 | .TP | |
427 | .BR CLONE_CHILD_SETTID " (since Linux 2.5.49)" | |
e2bf1234 | 428 | Store the child thread ID at the location pointed to by |
81c2368f | 429 | .I child_tid |
faa0e55a MK |
430 | .RB ( clone ()) |
431 | or | |
432 | .I cl_args.child_tid | |
433 | .RB ( clone3 ()) | |
8ef021ea | 434 | in the child's memory. |
5261b0fe | 435 | The store operation completes before the clone call |
6ab62ed8 | 436 | returns control to user space in the child process. |
5261b0fe | 437 | (Note that the store operation may not have completed before the clone call |
6ab62ed8 MK |
438 | returns in the parent process, which will be relevant if the |
439 | .BR CLONE_VM | |
440 | flag is also employed.) | |
f5dbc7c8 | 441 | .TP |
27f14b44 MK |
442 | .BR CLONE_CLEAR_SIGHAND " (since Linux 5.5)" |
443 | .\" commit b612e5df4587c934bd056bf05f4a1deca4de4f75 | |
444 | By default, signal dispositions in the child thread are the same as | |
445 | in the parent. | |
446 | If this flag is specified, | |
447 | then all signals that are handled in the parent | |
448 | are reset to their default dispositions | |
449 | .RB ( SIG_DFL ) | |
450 | in the child. | |
451 | .IP | |
452 | Specifying this flag together with | |
453 | .B CLONE_SIGHAND | |
454 | is nonsensical and disallowed. | |
455 | .TP | |
baa435c6 MK |
456 | .BR CLONE_DETACHED " (historical)" |
457 | For a while (during the Linux 2.5 development series) | |
458 | .\" added in 2.5.32; removed in 2.6.0-test4 | |
459 | there was a | |
460 | .B CLONE_DETACHED | |
461 | flag, | |
462 | which caused the parent not to receive a signal when the child terminated. | |
463 | Ultimately, the effect of this flag was subsumed under the | |
464 | .BR CLONE_THREAD | |
465 | flag and by the time Linux 2.6.0 was released, this flag had no effect. | |
466 | Starting in Linux 2.6.2, the need to give this flag together with | |
467 | .B CLONE_THREAD | |
468 | disappeared. | |
469 | .IP | |
470 | This flag is still defined, but it is usually ignored when calling | |
471 | .BR clone (). | |
472 | However, see the description of | |
473 | .BR CLONE_PIDFD | |
474 | for some exceptions. | |
475 | .TP | |
1603d6a1 | 476 | .BR CLONE_FILES " (since Linux 2.0)" |
fea681da | 477 | If |
f5dbc7c8 MK |
478 | .B CLONE_FILES |
479 | is set, the calling process and the child process share the same file | |
480 | descriptor table. | |
481 | Any file descriptor created by the calling process or by the child | |
482 | process is also valid in the other process. | |
483 | Similarly, if one of the processes closes a file descriptor, | |
484 | or changes its associated flags (using the | |
485 | .BR fcntl (2) | |
486 | .B F_SETFD | |
487 | operation), the other process is also affected. | |
8a76b19e KE |
488 | If a process sharing a file descriptor table calls |
489 | .BR execve (2), | |
490 | its file descriptor table is duplicated (unshared). | |
efeece04 | 491 | .IP |
fea681da | 492 | If |
f5dbc7c8 MK |
493 | .B CLONE_FILES |
494 | is not set, the child process inherits a copy of all file descriptors | |
5261b0fe | 495 | opened in the calling process at the time of the clone call. |
f5dbc7c8 MK |
496 | Subsequent operations that open or close file descriptors, |
497 | or change file descriptor flags, | |
498 | performed by either the calling | |
499 | process or the child process do not affect the other process. | |
db8ba2b4 | 500 | Note, however, |
839d161f MK |
501 | that the duplicated file descriptors in the child refer to the same |
502 | open file descriptions as the corresponding file descriptors | |
503 | in the calling process, | |
2433365b | 504 | and thus share file offsets and file status flags (see |
db8ba2b4 | 505 | .BR open (2)). |
fea681da | 506 | .TP |
1603d6a1 | 507 | .BR CLONE_FS " (since Linux 2.0)" |
fea681da MK |
508 | If |
509 | .B CLONE_FS | |
9ee4a2b6 | 510 | is set, the caller and the child process share the same filesystem |
c13182ef | 511 | information. |
9ee4a2b6 | 512 | This includes the root of the filesystem, the current |
c13182ef MK |
513 | working directory, and the umask. |
514 | Any call to | |
fea681da MK |
515 | .BR chroot (2), |
516 | .BR chdir (2), | |
517 | or | |
518 | .BR umask (2) | |
edcc65ff | 519 | performed by the calling process or the child process also affects the |
fea681da | 520 | other process. |
efeece04 | 521 | .IP |
c13182ef | 522 | If |
fea681da | 523 | .B CLONE_FS |
9ee4a2b6 | 524 | is not set, the child process works on a copy of the filesystem |
5261b0fe | 525 | information of the calling process at the time of the clone call. |
fea681da MK |
526 | Calls to |
527 | .BR chroot (2), | |
528 | .BR chdir (2), | |
4ba17a6d | 529 | or |
fea681da MK |
530 | .BR umask (2) |
531 | performed later by one of the processes do not affect the other process. | |
fea681da | 532 | .TP |
edc1b9fc MK |
533 | .BR CLONE_INTO_CGROUP " (since Linux 5.7)" |
534 | .\" commit ef2c41cf38a7559bbf91af42d5b6a4429db8fc68 | |
535 | By default, a child process is placed in the same version 2 | |
536 | cgroup as its parent. | |
537 | The | |
538 | .B CLONE_INTO_CGROUP | |
17d86030 | 539 | flag allows the child process to be created in a different version 2 cgroup. |
edc1b9fc MK |
540 | (Note that |
541 | .BR CLONE_INTO_CGROUP | |
542 | has effect only for version 2 cgroups.) | |
543 | .IP | |
544 | In order to place the child process in a different cgroup, | |
545 | the caller specifies | |
546 | .BR CLONE_INTO_CGROUP | |
547 | in | |
548 | .I cl_args.flags | |
549 | and passes a file descriptor that refers to a version 2 cgroup in the | |
550 | .I cl_args.cgroup | |
551 | field. | |
17d86030 | 552 | (This file descriptor can be obtained by opening a cgroup v2 directory |
edc1b9fc MK |
553 | using either the |
554 | .B O_RDONLY | |
555 | or the | |
556 | .B O_PATH | |
557 | flag.) | |
558 | Note that all of the usual restrictions (described in | |
559 | .BR cgroups (7)) | |
560 | on placing a process into a version 2 cgroup apply. | |
561 | .IP | |
562 | Spawning a process into a cgroup different from the parent's cgroup | |
563 | makes it possible for a service manager to directly spawn new | |
564 | services into dedicated cgroups. | |
565 | This eliminates the accounting | |
566 | jitter that would be caused if the child process was first created in the | |
567 | same cgroup as the parent and then | |
568 | moved into the target cgroup. | |
569 | The | |
570 | .BR CLONE_INTO_CGROUP | |
571 | flag also allows the creation of | |
572 | frozen child processes by spawning them into a frozen cgroup. | |
573 | (See | |
574 | .BR cgroups (7) | |
575 | for a description of the freezer controller.) | |
576 | For threaded applications (or even thread implementations which | |
577 | make use of cgroups to limit individual threads), it is possible to | |
578 | establish a fixed cgroup layout before spawning each thread | |
579 | directly into its target cgroup. | |
580 | .TP | |
a4cc375e | 581 | .BR CLONE_IO " (since Linux 2.6.25)" |
11f27a1c JA |
582 | If |
583 | .B CLONE_IO | |
584 | is set, then the new process shares an I/O context with | |
585 | the calling process. | |
586 | If this flag is not set, then (as with | |
587 | .BR fork (2)) | |
588 | the new process has its own I/O context. | |
efeece04 | 589 | .IP |
11f27a1c | 590 | .\" The following based on text from Jens Axboe |
d1f84ed7 | 591 | The I/O context is the I/O scope of the disk scheduler (i.e., |
11f27a1c JA |
592 | what the I/O scheduler uses to model scheduling of a process's I/O). |
593 | If processes share the same I/O context, | |
594 | they are treated as one by the I/O scheduler. | |
595 | As a consequence, they get to share disk time. | |
596 | For some I/O schedulers, | |
597 | .\" the anticipatory and CFQ scheduler | |
598 | if two processes share an I/O context, | |
599 | they will be allowed to interleave their disk access. | |
600 | If several threads are doing I/O on behalf of the same process | |
601 | .RB ( aio_read (3), | |
602 | for instance), they should employ | |
603 | .BR CLONE_IO | |
604 | to get better I/O performance. | |
605 | .\" with CFQ and AS. | |
efeece04 | 606 | .IP |
11f27a1c JA |
607 | If the kernel is not configured with the |
608 | .B CONFIG_BLOCK | |
609 | option, this flag is a no-op. | |
610 | .TP | |
c5af0674 MK |
611 | .BR CLONE_NEWCGROUP " (since Linux 4.6)" |
612 | Create the process in a new cgroup namespace. | |
613 | If this flag is not set, then (as with | |
614 | .BR fork (2)) | |
615 | the process is created in the same cgroup namespaces as the calling process. | |
efeece04 | 616 | .IP |
c5af0674 | 617 | For further information on cgroup namespaces, see |
b9fe4bc3 | 618 | .BR cgroup_namespaces (7). |
efeece04 | 619 | .IP |
c5af0674 MK |
620 | Only a privileged process |
621 | .RB ( CAP_SYS_ADMIN ) | |
622 | can employ | |
623 | .BR CLONE_NEWCGROUP . | |
624 | .\" | |
625 | .TP | |
8722311b | 626 | .BR CLONE_NEWIPC " (since Linux 2.6.19)" |
667417b3 MK |
627 | If |
628 | .B CLONE_NEWIPC | |
629 | is set, then create the process in a new IPC namespace. | |
630 | If this flag is not set, then (as with | |
06b30458 | 631 | .BR fork (2)), |
667417b3 MK |
632 | the process is created in the same IPC namespace as |
633 | the calling process. | |
efeece04 | 634 | .IP |
981eda4a MK |
635 | For further information on IPC namespaces, see |
636 | .BR ipc_namespaces (7). | |
637 | .IP | |
ab5dd83f MK |
638 | Only a privileged process |
639 | .RB ( CAP_SYS_ADMIN ) | |
640 | can employ | |
641 | .BR CLONE_NEWIPC . | |
667417b3 MK |
642 | This flag can't be specified in conjunction with |
643 | .BR CLONE_SYSVSEM . | |
644 | .TP | |
163bf178 | 645 | .BR CLONE_NEWNET " (since Linux 2.6.24)" |
33a0ccb2 | 646 | (The implementation of this flag was completed only |
9108d867 | 647 | by about kernel version 2.6.29.) |
efeece04 | 648 | .IP |
163bf178 MK |
649 | If |
650 | .B CLONE_NEWNET | |
651 | is set, then create the process in a new network namespace. | |
652 | If this flag is not set, then (as with | |
57ef8c39 | 653 | .BR fork (2)) |
163bf178 MK |
654 | the process is created in the same network namespace as |
655 | the calling process. | |
efeece04 | 656 | .IP |
73680728 | 657 | For further information on network namespaces, see |
40002795 | 658 | .BR network_namespaces (7). |
efeece04 | 659 | .IP |
ab5dd83f MK |
660 | Only a privileged process |
661 | .RB ( CAP_SYS_ADMIN ) | |
662 | can employ | |
663 | .BR CLONE_NEWNET . | |
163bf178 | 664 | .TP |
c10859eb | 665 | .BR CLONE_NEWNS " (since Linux 2.4.19)" |
3dd2331c MK |
666 | If |
667 | .B CLONE_NEWNS | |
668 | is set, the cloned child is started in a new mount namespace, | |
669 | initialized with a copy of the namespace of the parent. | |
670 | If | |
fea681da | 671 | .B CLONE_NEWNS |
3dd2331c | 672 | is not set, the child lives in the same mount |
4df2eb09 | 673 | namespace as the parent. |
efeece04 | 674 | .IP |
981eda4a MK |
675 | For further information on mount namespaces, see |
676 | .BR namespaces (7) | |
677 | and | |
678 | .BR mount_namespaces (7). | |
679 | .IP | |
ab5dd83f MK |
680 | Only a privileged process |
681 | .RB ( CAP_SYS_ADMIN ) | |
682 | can employ | |
683 | .BR CLONE_NEWNS . | |
fea681da MK |
684 | It is not permitted to specify both |
685 | .B CLONE_NEWNS | |
686 | and | |
687 | .B CLONE_FS | |
9219d208 | 688 | .\" See https://lwn.net/Articles/543273/ |
5261b0fe | 689 | in the same clone call. |
9d005472 MK |
690 | .TP |
691 | .BR CLONE_NEWPID " (since Linux 2.6.24)" | |
692 | .\" This explanation draws a lot of details from | |
693 | .\" http://lwn.net/Articles/259217/ | |
694 | .\" Authors: Pavel Emelyanov <xemul@openvz.org> | |
695 | .\" and Kir Kolyshkin <kir@openvz.org> | |
696 | .\" | |
697 | .\" The primary kernel commit is 30e49c263e36341b60b735cbef5ca37912549264 | |
698 | .\" Author: Pavel Emelyanov <xemul@openvz.org> | |
699 | If | |
700 | .B CLONE_NEWPID | |
701 | is set, then create the process in a new PID namespace. | |
702 | If this flag is not set, then (as with | |
703 | .BR fork (2)) | |
704 | the process is created in the same PID namespace as | |
705 | the calling process. | |
efeece04 | 706 | .IP |
9d005472 | 707 | For further information on PID namespaces, see |
7e0e902b MK |
708 | .BR namespaces (7) |
709 | and | |
39b3f005 | 710 | .BR pid_namespaces (7). |
efeece04 | 711 | .IP |
ab5dd83f MK |
712 | Only a privileged process |
713 | .RB ( CAP_SYS_ADMIN ) | |
714 | can employ | |
715 | .BR CLONE_NEWPID . | |
9d005472 | 716 | This flag can't be specified in conjunction with |
f0007192 MK |
717 | .BR CLONE_THREAD |
718 | or | |
719 | .BR CLONE_PARENT . | |
70d21f17 | 720 | .TP |
06b30458 MK |
721 | .BR CLONE_NEWUSER |
722 | (This flag first became meaningful for | |
723 | .BR clone () | |
4d2b3ed7 MK |
724 | in Linux 2.6.23, |
725 | the current | |
11a38815 | 726 | .BR clone () |
4d2b3ed7 MK |
727 | semantics were merged in Linux 3.5, |
728 | and the final pieces to make the user namespaces completely usable were | |
729 | merged in Linux 3.8.) | |
efeece04 | 730 | .IP |
70d21f17 EB |
731 | If |
732 | .B CLONE_NEWUSER | |
06b30458 MK |
733 | is set, then create the process in a new user namespace. |
734 | If this flag is not set, then (as with | |
57ef8c39 | 735 | .BR fork (2)) |
70d21f17 | 736 | the process is created in the same user namespace as the calling process. |
efeece04 | 737 | .IP |
981eda4a MK |
738 | For further information on user namespaces, see |
739 | .BR namespaces (7) | |
740 | and | |
741 | .BR user_namespaces (7). | |
742 | .IP | |
fefbcba8 MK |
743 | Before Linux 3.8, use of |
744 | .BR CLONE_NEWUSER | |
745 | required that the caller have three capabilities: | |
746 | .BR CAP_SYS_ADMIN , | |
747 | .BR CAP_SETUID , | |
748 | and | |
749 | .BR CAP_SETGID . | |
750 | .\" Before Linux 2.6.29, it appears that only CAP_SYS_ADMIN was needed | |
06b30458 | 751 | Starting with Linux 3.8, |
9d005472 | 752 | no privileges are needed to create a user namespace. |
efeece04 | 753 | .IP |
5e72cf7d MK |
754 | This flag can't be specified in conjunction with |
755 | .BR CLONE_THREAD | |
756 | or | |
757 | .BR CLONE_PARENT . | |
758 | For security reasons, | |
759 | .\" commit e66eded8309ebf679d3d3c1f5820d1f2ca332c71 | |
760 | .\" https://lwn.net/Articles/543273/ | |
761 | .\" The fix actually went into 3.9 and into 3.8.3. However, user namespaces | |
762 | .\" were, for practical purposes, unusable in earlier 3.8.x because of the | |
ab3311aa | 763 | .\" various filesystems that didn't support userns. |
f0007192 MK |
764 | .BR CLONE_NEWUSER |
765 | cannot be specified in conjunction with | |
5e72cf7d | 766 | .BR CLONE_FS . |
82ee147a | 767 | .TP |
43ce9dda MK |
768 | .BR CLONE_NEWUTS " (since Linux 2.6.19)" |
769 | If | |
770 | .B CLONE_NEWUTS | |
e1b11906 MK |
771 | is set, then create the process in a new UTS namespace, |
772 | whose identifiers are initialized by duplicating the identifiers | |
773 | from the UTS namespace of the calling process. | |
43ce9dda | 774 | If this flag is not set, then (as with |
57ef8c39 | 775 | .BR fork (2)) |
43ce9dda MK |
776 | the process is created in the same UTS namespace as |
777 | the calling process. | |
efeece04 | 778 | .IP |
981eda4a MK |
779 | For further information on UTS namespaces, see |
780 | .BR uts_namespaces (7). | |
781 | .IP | |
ab5dd83f MK |
782 | Only a privileged process |
783 | .RB ( CAP_SYS_ADMIN ) | |
784 | can employ | |
785 | .BR CLONE_NEWUTS . | |
43ce9dda | 786 | .TP |
f5dbc7c8 MK |
787 | .BR CLONE_PARENT " (since Linux 2.3.12)" |
788 | If | |
789 | .B CLONE_PARENT | |
790 | is set, then the parent of the new child (as returned by | |
791 | .BR getppid (2)) | |
792 | will be the same as that of the calling process. | |
efeece04 | 793 | .IP |
f5dbc7c8 MK |
794 | If |
795 | .B CLONE_PARENT | |
796 | is not set, then (as with | |
797 | .BR fork (2)) | |
798 | the child's parent is the calling process. | |
efeece04 | 799 | .IP |
f5dbc7c8 MK |
800 | Note that it is the parent process, as returned by |
801 | .BR getppid (2), | |
802 | which is signaled when the child terminates, so that | |
803 | if | |
804 | .B CLONE_PARENT | |
805 | is set, then the parent of the calling process, rather than the | |
806 | calling process itself, will be signaled. | |
a17b9d28 | 807 | .IP |
4269a6ab | 808 | The |
a17b9d28 | 809 | .B CLONE_PARENT |
4269a6ab MK |
810 | flag can't be used in clone calls by the |
811 | global init process (PID 1 in the initial PID namespace) | |
812 | and init processes in other PID namespaces. | |
813 | This restriction prevents the creation of multi-rooted process trees | |
814 | as well as the creation of unreapable zombies in the initial PID namespace. | |
f5dbc7c8 MK |
815 | .TP |
816 | .BR CLONE_PARENT_SETTID " (since Linux 2.5.49)" | |
e2bf1234 | 817 | Store the child thread ID at the location pointed to by |
81c2368f | 818 | .I parent_tid |
faa0e55a MK |
819 | .RB ( clone ()) |
820 | or | |
d5d482ec | 821 | .I cl_args.parent_tid |
faa0e55a | 822 | .RB ( clone3 ()) |
8ef021ea | 823 | in the parent's memory. |
f5dbc7c8 MK |
824 | (In Linux 2.5.32-2.5.48 there was a flag |
825 | .B CLONE_SETTID | |
826 | that did this.) | |
5261b0fe | 827 | The store operation completes before the clone call |
b5da2f91 | 828 | returns control to user space. |
f5dbc7c8 | 829 | .TP |
1c173eb3 | 830 | .BR CLONE_PID " (Linux 2.0 to 2.5.15)" |
f5dbc7c8 MK |
831 | If |
832 | .B CLONE_PID | |
833 | is set, the child process is created with the same process ID as | |
834 | the calling process. | |
835 | This is good for hacking the system, but otherwise | |
836 | of not much use. | |
1c173eb3 | 837 | From Linux 2.3.21 onward, this flag could be |
f5dbc7c8 | 838 | specified only by the system boot process (PID 0). |
1c173eb3 | 839 | The flag disappeared completely from the kernel sources in Linux 2.5.16. |
f5d5180f | 840 | Subsequently, the kernel silently ignored this bit if it was specified in the |
16853a31 MK |
841 | .IR flags |
842 | mask. | |
f5d5180f MK |
843 | Much later, the same bit was recycled for use as the |
844 | .B CLONE_PIDFD | |
845 | flag. | |
f5dbc7c8 | 846 | .TP |
9f938981 | 847 | .BR CLONE_PIDFD " (since Linux 5.2)" |
4e98b074 | 848 | .\" commit b3e5838252665ee4cfa76b82bdf1198dca81e5be |
faa0e55a MK |
849 | If this flag is specified, |
850 | a PID file descriptor referring to the child process is allocated | |
851 | and placed at a specified location in the parent's memory. | |
b4ebffb2 | 852 | The close-on-exec flag is set on this new file descriptor. |
34a975f8 MK |
853 | PID file descriptors can be used for the purposes described in |
854 | .BR pidfd_open (2). | |
faa0e55a MK |
855 | .RS |
856 | .IP * 3 | |
857 | When using | |
858 | .BR clone3 (), | |
859 | the PID file descriptor is placed at the location pointed to by | |
860 | .IR cl_args.pidfd . | |
861 | .IP * | |
862 | When using | |
863 | .BR clone (), | |
864 | the PID file descriptor is placed at the location pointed to by | |
865 | .IR parent_tid . | |
9f938981 | 866 | Since the |
81c2368f | 867 | .I parent_tid |
b97cc7ae | 868 | argument is used to return the PID file descriptor, |
9f938981 CB |
869 | .B CLONE_PIDFD |
870 | cannot be used with | |
faa0e55a MK |
871 | .B CLONE_PARENT_SETTID |
872 | when calling | |
873 | .BR clone (). | |
874 | .RE | |
9f938981 CB |
875 | .IP |
876 | It is currently not possible to use this flag together with | |
877 | .B CLONE_THREAD. | |
b97cc7ae | 878 | This means that the process identified by the PID file descriptor |
f6183e5b | 879 | will always be a thread group leader. |
9f938981 | 880 | .IP |
baa435c6 | 881 | If the obsolete |
9f938981 | 882 | .B CLONE_DETACHED |
baa435c6 MK |
883 | flag is specified alongside |
884 | .BR CLONE_PIDFD | |
885 | when calling | |
886 | .BR clone (), | |
4e98b074 | 887 | an error is returned. |
baa435c6 MK |
888 | An error also results if |
889 | .B CLONE_DETACHED | |
890 | is specified when calling | |
891 | .BR clone3 (). | |
892 | This error behavior ensures that the bit corresponding to | |
893 | .BR CLONE_DETACHED | |
894 | can be reused for further PID file descriptor features in the future. | |
9f938981 | 895 | .TP |
1603d6a1 | 896 | .BR CLONE_PTRACE " (since Linux 2.2)" |
f5dbc7c8 MK |
897 | If |
898 | .B CLONE_PTRACE | |
899 | is specified, and the calling process is being traced, | |
900 | then trace the child also (see | |
901 | .BR ptrace (2)). | |
902 | .TP | |
903 | .BR CLONE_SETTLS " (since Linux 2.5.32)" | |
dd6d3d2e | 904 | The TLS (Thread Local Storage) descriptor is set to |
81c2368f | 905 | .IR tls . |
efeece04 | 906 | .IP |
dd6d3d2e | 907 | The interpretation of |
81c2368f | 908 | .I tls |
dd6d3d2e KF |
909 | and the resulting effect is architecture dependent. |
910 | On x86, | |
81c2368f | 911 | .I tls |
dd6d3d2e | 912 | is interpreted as a |
2551f801 | 913 | .IR "struct user_desc\ *" |
35bf8cb4 | 914 | (see |
dd6d3d2e | 915 | .BR set_thread_area (2)). |
9ea5bc66 | 916 | On x86-64 it is the new value to be set for the %fs base register |
35bf8cb4 | 917 | (see the |
2551f801 | 918 | .B ARCH_SET_FS |
dd6d3d2e KF |
919 | argument to |
920 | .BR arch_prctl (2)). | |
921 | On architectures with a dedicated TLS register, it is the new value | |
922 | of that register. | |
f5d5180f MK |
923 | .IP |
924 | Use of this flag requires detailed knowledge and generally it | |
925 | should not be used except in libraries implementing threading. | |
f5dbc7c8 | 926 | .TP |
1603d6a1 | 927 | .BR CLONE_SIGHAND " (since Linux 2.0)" |
fea681da MK |
928 | If |
929 | .B CLONE_SIGHAND | |
314c8ff4 | 930 | is set, the calling process and the child process share the same table of |
c13182ef MK |
931 | signal handlers. |
932 | If the calling process or child process calls | |
fea681da | 933 | .BR sigaction (2) |
c13182ef MK |
934 | to change the behavior associated with a signal, the behavior is |
935 | changed in the other process as well. | |
936 | However, the calling process and child | |
fea681da | 937 | processes still have distinct signal masks and sets of pending |
c13182ef | 938 | signals. |
4ba17a6d | 939 | So, one of them may block or unblock signals using |
fea681da MK |
940 | .BR sigprocmask (2) |
941 | without affecting the other process. | |
efeece04 | 942 | .IP |
fea681da MK |
943 | If |
944 | .B CLONE_SIGHAND | |
945 | is not set, the child process inherits a copy of the signal handlers | |
5261b0fe | 946 | of the calling process at the time of the clone call. |
c13182ef | 947 | Calls to |
fea681da MK |
948 | .BR sigaction (2) |
949 | performed later by one of the processes have no effect on the other | |
950 | process. | |
efeece04 | 951 | .IP |
d6bec36e MK |
952 | Since Linux 2.6.0, |
953 | .\" Precisely: Linux 2.6.0-test6 | |
16853a31 | 954 | the |
29546c24 | 955 | .I flags |
16853a31 | 956 | mask must also include |
29546c24 MK |
957 | .B CLONE_VM |
958 | if | |
959 | .B CLONE_SIGHAND | |
960 | is specified | |
fea681da | 961 | .TP |
d6bec36e MK |
962 | .BR CLONE_STOPPED " (since Linux 2.6.0)" |
963 | .\" Precisely: Linux 2.6.0-test2 | |
a69b6bda MK |
964 | If |
965 | .B CLONE_STOPPED | |
966 | is set, then the child is initially stopped (as though it was sent a | |
967 | .B SIGSTOP | |
968 | signal), and must be resumed by sending it a | |
969 | .B SIGCONT | |
970 | signal. | |
efeece04 | 971 | .IP |
a60450a9 MK |
972 | This flag was |
973 | .I deprecated | |
974 | from Linux 2.6.25 onward, | |
975 | and was | |
976 | .I removed | |
28b44abc MK |
977 | altogether in Linux 2.6.38. |
978 | Since then, the kernel silently ignores it without error. | |
a5a061ee | 979 | .\" glibc 2.8 removed this defn from bits/sched.h |
c5af0674 MK |
980 | Starting with Linux 4.6, the same bit was reused for the |
981 | .BR CLONE_NEWCGROUP | |
982 | flag. | |
a69b6bda | 983 | .TP |
f5dbc7c8 | 984 | .BR CLONE_SYSVSEM " (since Linux 2.5.10)" |
fea681da | 985 | If |
f5dbc7c8 MK |
986 | .B CLONE_SYSVSEM |
987 | is set, then the child and the calling process share | |
5ada4b94 MK |
988 | a single list of System V semaphore adjustment |
989 | .RI ( semadj ) | |
990 | values (see | |
f5dbc7c8 | 991 | .BR semop (2)). |
5ada4b94 MK |
992 | In this case, the shared list accumulates |
993 | .I semadj | |
994 | values across all processes sharing the list, | |
995 | and semaphore adjustments are performed only when the last process | |
996 | that is sharing the list terminates (or ceases sharing the list using | |
997 | .BR unshare (2)). | |
f5d401dd | 998 | If this flag is not set, then the child has a separate |
5ada4b94 MK |
999 | .I semadj |
1000 | list that is initially empty. | |
fea681da | 1001 | .TP |
d6bec36e MK |
1002 | .BR CLONE_THREAD " (since Linux 2.4.0)" |
1003 | .\" Precisely: Linux 2.6.0-test8 | |
fea681da MK |
1004 | If |
1005 | .B CLONE_THREAD | |
1006 | is set, the child is placed in the same thread group as the calling process. | |
fd8a5be4 MK |
1007 | To make the remainder of the discussion of |
1008 | .B CLONE_THREAD | |
1009 | more readable, the term "thread" is used to refer to the | |
1010 | processes within a thread group. | |
efeece04 | 1011 | .IP |
fd8a5be4 MK |
1012 | Thread groups were a feature added in Linux 2.4 to support the |
1013 | POSIX threads notion of a set of threads that share a single PID. | |
1014 | Internally, this shared PID is the so-called | |
1015 | thread group identifier (TGID) for the thread group. | |
c13182ef | 1016 | Since Linux 2.4, calls to |
fea681da | 1017 | .BR getpid (2) |
fd8a5be4 | 1018 | return the TGID of the caller. |
efeece04 | 1019 | .IP |
fd8a5be4 MK |
1020 | The threads within a group can be distinguished by their (system-wide) |
1021 | unique thread IDs (TID). | |
1022 | A new thread's TID is available as the function result | |
5261b0fe | 1023 | returned to the caller, |
fd8a5be4 MK |
1024 | and a thread can obtain |
1025 | its own TID using | |
1026 | .BR gettid (2). | |
efeece04 | 1027 | .IP |
5261b0fe | 1028 | When a clone call is made without specifying |
fd8a5be4 MK |
1029 | .BR CLONE_THREAD , |
1030 | then the resulting thread is placed in a new thread group | |
1031 | whose TGID is the same as the thread's TID. | |
1032 | This thread is the | |
1033 | .I leader | |
1034 | of the new thread group. | |
efeece04 | 1035 | .IP |
fd8a5be4 MK |
1036 | A new thread created with |
1037 | .B CLONE_THREAD | |
5261b0fe | 1038 | has the same parent process as the process that made the clone call |
c13182ef | 1039 | (i.e., like |
fd8a5be4 MK |
1040 | .BR CLONE_PARENT ), |
1041 | so that calls to | |
1042 | .BR getppid (2) | |
1043 | return the same value for all of the threads in a thread group. | |
1044 | When a | |
c13182ef | 1045 | .B CLONE_THREAD |
5261b0fe | 1046 | thread terminates, the thread that created it is not sent a |
fd8a5be4 MK |
1047 | .B SIGCHLD |
1048 | (or other termination) signal; | |
1049 | nor can the status of such a thread be obtained | |
1050 | using | |
1051 | .BR wait (2). | |
1052 | (The thread is said to be | |
1053 | .IR detached .) | |
efeece04 | 1054 | .IP |
e2fbf61d MK |
1055 | After all of the threads in a thread group terminate |
1056 | the parent process of the thread group is sent a | |
fd8a5be4 MK |
1057 | .B SIGCHLD |
1058 | (or other termination) signal. | |
efeece04 | 1059 | .IP |
fd8a5be4 MK |
1060 | If any of the threads in a thread group performs an |
1061 | .BR execve (2), | |
1062 | then all threads other than the thread group leader are terminated, | |
1063 | and the new program is executed in the thread group leader. | |
efeece04 | 1064 | .IP |
f7110f60 MK |
1065 | If one of the threads in a thread group creates a child using |
1066 | .BR fork (2), | |
1067 | then any thread in the group can | |
1068 | .BR wait (2) | |
1069 | for that child. | |
efeece04 | 1070 | .IP |
16853a31 | 1071 | Since Linux 2.5.35, the |
fd8a5be4 | 1072 | .I flags |
16853a31 | 1073 | mask must also include |
fd8a5be4 MK |
1074 | .B CLONE_SIGHAND |
1075 | if | |
1076 | .B CLONE_THREAD | |
6fd69f33 | 1077 | is specified |
d6bec36e MK |
1078 | (and note that, since Linux 2.6.0, |
1079 | .\" Precisely: Linux 2.6.0-test6 | |
6fd69f33 MK |
1080 | .BR CLONE_SIGHAND |
1081 | also requires | |
1082 | .BR CLONE_VM | |
1083 | to be included). | |
efeece04 | 1084 | .IP |
e2fbf61d MK |
1085 | Signal dispositions and actions are process-wide: |
1086 | if an unhandled signal is delivered to a thread, then | |
1087 | it will affect (terminate, stop, continue, be ignored in) | |
1088 | all members of the thread group. | |
efeece04 | 1089 | .IP |
99408a60 | 1090 | Each thread has its own signal mask, as set by |
f957eebd MK |
1091 | .BR sigprocmask (2). |
1092 | .IP | |
1093 | A signal may be process-directed or thread-directed. | |
1094 | A process-directed signal is targeted at a thread group (i.e., a TGID), | |
1095 | and is delivered to an arbitrarily selected thread from among those | |
1096 | that are not blocking the signal. | |
ed4f87f0 | 1097 | A signal may be process-directed because it was generated by the kernel |
f957eebd MK |
1098 | for reasons other than a hardware exception, or because it was sent using |
1099 | .BR kill (2) | |
1100 | or | |
1101 | .BR sigqueue (3). | |
1102 | A thread-directed signal is targeted at (i.e., delivered to) | |
1103 | a specific thread. | |
1104 | A signal may be thread directed because it was sent using | |
1105 | .BR tgkill (2) | |
1106 | or | |
1107 | .BR pthread_sigqueue (3), | |
1108 | or because the thread executed a machine language instruction that triggered | |
1109 | a hardware exception | |
1110 | (e.g., invalid memory access triggering | |
1111 | .BR SIGSEGV | |
1112 | or a floating-point exception triggering | |
1113 | .BR SIGFPE ). | |
1114 | .IP | |
99408a60 MK |
1115 | A call to |
1116 | .BR sigpending (2) | |
f957eebd MK |
1117 | returns a signal set that is the union of the pending process-directed |
1118 | signals and the signals that are pending for the calling thread. | |
efeece04 | 1119 | .IP |
475c2753 | 1120 | If a process-directed signal is delivered to a thread group, |
e2fbf61d MK |
1121 | and the thread group has installed a handler for the signal, then |
1122 | the handler will be invoked in exactly one, arbitrarily selected | |
1123 | member of the thread group that has not blocked the signal. | |
c13182ef | 1124 | If multiple threads in a group are waiting to accept the same signal using |
e2fbf61d MK |
1125 | .BR sigwaitinfo (2), |
1126 | the kernel will arbitrarily select one of these threads | |
475c2753 | 1127 | to receive the signal. |
a69b6bda | 1128 | .TP |
f5dbc7c8 | 1129 | .BR CLONE_UNTRACED " (since Linux 2.5.46)" |
a69b6bda | 1130 | If |
f5dbc7c8 MK |
1131 | .B CLONE_UNTRACED |
1132 | is specified, then a tracing process cannot force | |
1133 | .B CLONE_PTRACE | |
1134 | on this child process. | |
fea681da | 1135 | .TP |
1603d6a1 | 1136 | .BR CLONE_VFORK " (since Linux 2.2)" |
f5dbc7c8 MK |
1137 | If |
1138 | .B CLONE_VFORK | |
1139 | is set, the execution of the calling process is suspended | |
1140 | until the child releases its virtual memory | |
1141 | resources via a call to | |
1142 | .BR execve (2) | |
1143 | or | |
1144 | .BR _exit (2) | |
1145 | (as with | |
1146 | .BR vfork (2)). | |
efeece04 | 1147 | .IP |
f5dbc7c8 MK |
1148 | If |
1149 | .B CLONE_VFORK | |
4b4a853a | 1150 | is not set, then both the calling process and the child are schedulable |
f5dbc7c8 MK |
1151 | after the call, and an application should not rely on execution occurring |
1152 | in any particular order. | |
fea681da | 1153 | .TP |
1603d6a1 | 1154 | .BR CLONE_VM " (since Linux 2.0)" |
f5dbc7c8 MK |
1155 | If |
1156 | .B CLONE_VM | |
1157 | is set, the calling process and the child process run in the same memory | |
1158 | space. | |
1159 | In particular, memory writes performed by the calling process | |
1160 | or by the child process are also visible in the other process. | |
1161 | Moreover, any memory mapping or unmapping performed with | |
1162 | .BR mmap (2) | |
1163 | or | |
1164 | .BR munmap (2) | |
1165 | by the child or calling process also affects the other process. | |
efeece04 | 1166 | .IP |
f5dbc7c8 MK |
1167 | If |
1168 | .B CLONE_VM | |
1169 | is not set, the child process runs in a separate copy of the memory | |
5261b0fe | 1170 | space of the calling process at the time of the clone call. |
f5dbc7c8 MK |
1171 | Memory writes or file mappings/unmappings performed by one of the |
1172 | processes do not affect the other, as with | |
1173 | .BR fork (2). | |
47297adb | 1174 | .SH RETURN VALUE |
0bfa087b MK |
1175 | .\" gettid(2) returns current->pid; |
1176 | .\" getpid(2) returns current->tgid; | |
fea681da | 1177 | On success, the thread ID of the child process is returned |
c13182ef | 1178 | in the caller's thread of execution. |
84811e86 | 1179 | On failure, \-1 is returned |
fea681da MK |
1180 | in the caller's context, no child process will be created, and |
1181 | .I errno | |
1182 | will be set appropriately. | |
fea681da MK |
1183 | .SH ERRORS |
1184 | .TP | |
1185 | .B EAGAIN | |
e1b6e186 MK |
1186 | Too many processes are already running; see |
1187 | .BR fork (2). | |
fea681da | 1188 | .TP |
6ba79da9 MK |
1189 | .BR EBUSY " (" clone3 "() only)" |
1190 | .B CLONE_INTO_CGROUP | |
1191 | was specified in | |
1192 | .IR cl_args.flags , | |
1193 | but the file descriptor specified in | |
1194 | .IR cl_args.cgroup | |
1195 | refers to a version 2 cgroup in which a domain controller is enabled. | |
1196 | .TP | |
bf031aaa | 1197 | .BR EEXIST " (" clone3 "() only)" |
ee8bb310 | 1198 | One (or more) of the PIDs specified in |
bf031aaa AR |
1199 | .I set_tid |
1200 | already exists in the corresponding PID namespace. | |
1201 | .TP | |
fea681da | 1202 | .B EINVAL |
27f14b44 MK |
1203 | Both |
1204 | .B CLONE_SIGHAND | |
1205 | and | |
1206 | .B CLONE_CLEAR_SIGHAND | |
1207 | were specified in the | |
1208 | .I flags | |
1209 | mask. | |
1210 | .TP | |
1211 | .B EINVAL | |
fea681da | 1212 | .B CLONE_SIGHAND |
16853a31 MK |
1213 | was specified in the |
1214 | .I flags | |
1215 | mask, but | |
fea681da | 1216 | .B CLONE_VM |
2e8a7fb3 | 1217 | was not. |
d6bec36e MK |
1218 | (Since Linux 2.6.0.) |
1219 | .\" Precisely: Linux 2.6.0-test6 | |
fea681da MK |
1220 | .TP |
1221 | .B EINVAL | |
1222 | .B CLONE_THREAD | |
16853a31 MK |
1223 | was specified in the |
1224 | .I flags | |
1225 | mask, but | |
fea681da | 1226 | .B CLONE_SIGHAND |
6387216b MK |
1227 | was not. |
1228 | (Since Linux 2.5.35.) | |
29546c24 MK |
1229 | .\" .TP |
1230 | .\" .B EINVAL | |
1231 | .\" Precisely one of | |
1232 | .\" .B CLONE_DETACHED | |
1233 | .\" and | |
1234 | .\" .B CLONE_THREAD | |
6387216b MK |
1235 | .\" was specified. |
1236 | .\" (Since Linux 2.6.0-test6.) | |
fea681da MK |
1237 | .TP |
1238 | .B EINVAL | |
d6868c69 | 1239 | .B CLONE_THREAD |
16853a31 MK |
1240 | was specified in the |
1241 | .I flags | |
1242 | mask, but the current process previously called | |
d6868c69 JH |
1243 | .BR unshare (2) |
1244 | with the | |
1245 | .B CLONE_NEWPID | |
1246 | flag or used | |
1247 | .BR setns (2) | |
1248 | to reassociate itself with a PID namespace. | |
1249 | .TP | |
1250 | .B EINVAL | |
d34e5645 | 1251 | .\" commit e66eded8309ebf679d3d3c1f5820d1f2ca332c71 |
fea681da MK |
1252 | Both |
1253 | .B CLONE_FS | |
1254 | and | |
1255 | .B CLONE_NEWNS | |
16853a31 MK |
1256 | were specified in the |
1257 | .IR flags | |
1258 | mask. | |
fea681da | 1259 | .TP |
d34e5645 MK |
1260 | .BR EINVAL " (since Linux 3.9)" |
1261 | Both | |
1262 | .B CLONE_NEWUSER | |
1263 | and | |
1264 | .B CLONE_FS | |
16853a31 MK |
1265 | were specified in the |
1266 | .IR flags | |
1267 | mask. | |
d34e5645 | 1268 | .TP |
fea681da | 1269 | .B EINVAL |
82ee147a | 1270 | Both |
667417b3 MK |
1271 | .B CLONE_NEWIPC |
1272 | and | |
1273 | .B CLONE_SYSVSEM | |
16853a31 MK |
1274 | were specified in the |
1275 | .IR flags | |
1276 | mask. | |
667417b3 MK |
1277 | .TP |
1278 | .B EINVAL | |
f0007192 | 1279 | One (or both) of |
82ee147a | 1280 | .BR CLONE_NEWPID |
f0007192 MK |
1281 | or |
1282 | .BR CLONE_NEWUSER | |
1283 | and one (or both) of | |
82ee147a | 1284 | .BR CLONE_THREAD |
f0007192 MK |
1285 | or |
1286 | .BR CLONE_PARENT | |
16853a31 MK |
1287 | were specified in the |
1288 | .IR flags | |
1289 | mask. | |
82ee147a | 1290 | .TP |
be479fdf MK |
1291 | .BR EINVAL " (since Linux 2.6.32)" |
1292 | .\" commit 123be07b0b399670a7cc3d82fef0cb4f93ef885c | |
1293 | .BR CLONE_PARENT | |
1294 | was specified, and the caller is an init process. | |
1295 | .TP | |
82ee147a | 1296 | .B EINVAL |
d4748fad | 1297 | Returned by the glibc |
edcc65ff | 1298 | .BR clone () |
d4748fad MK |
1299 | wrapper function when |
1300 | .IR fn | |
1301 | or | |
81c2368f | 1302 | .IR stack |
d4748fad | 1303 | is specified as NULL. |
fea681da | 1304 | .TP |
28cad2c1 | 1305 | .B EINVAL |
667417b3 | 1306 | .BR CLONE_NEWIPC |
16853a31 MK |
1307 | was specified in the |
1308 | .IR flags | |
1309 | mask, | |
667417b3 MK |
1310 | but the kernel was not configured with the |
1311 | .B CONFIG_SYSVIPC | |
1312 | and | |
1313 | .BR CONFIG_IPC_NS | |
1314 | options. | |
1315 | .TP | |
1316 | .B EINVAL | |
163bf178 | 1317 | .BR CLONE_NEWNET |
16853a31 MK |
1318 | was specified in the |
1319 | .IR flags | |
1320 | mask, | |
163bf178 MK |
1321 | but the kernel was not configured with the |
1322 | .B CONFIG_NET_NS | |
1323 | option. | |
1324 | .TP | |
1325 | .B EINVAL | |
28cad2c1 | 1326 | .BR CLONE_NEWPID |
16853a31 MK |
1327 | was specified in the |
1328 | .IR flags | |
1329 | mask, | |
28cad2c1 MK |
1330 | but the kernel was not configured with the |
1331 | .B CONFIG_PID_NS | |
1332 | option. | |
1333 | .TP | |
43ce9dda | 1334 | .B EINVAL |
231d0bbe | 1335 | .BR CLONE_NEWUSER |
16853a31 MK |
1336 | was specified in the |
1337 | .IR flags | |
1338 | mask, | |
231d0bbe MK |
1339 | but the kernel was not configured with the |
1340 | .B CONFIG_USER_NS | |
1341 | option. | |
1342 | .TP | |
1343 | .B EINVAL | |
43ce9dda | 1344 | .BR CLONE_NEWUTS |
16853a31 MK |
1345 | was specified in the |
1346 | .IR flags | |
1347 | mask, | |
43ce9dda | 1348 | but the kernel was not configured with the |
832fe8ea | 1349 | .B CONFIG_UTS_NS |
43ce9dda MK |
1350 | option. |
1351 | .TP | |
c550a897 | 1352 | .B EINVAL |
81c2368f | 1353 | .I stack |
c550a897 MK |
1354 | is not aligned to a suitable boundary for this architecture. |
1355 | For example, on aarch64, | |
81c2368f | 1356 | .I stack |
c550a897 MK |
1357 | must be a multiple of 16. |
1358 | .TP | |
bc03b116 | 1359 | .BR EINVAL " (" clone3 "() only)" |
baa435c6 MK |
1360 | .B CLONE_DETACHED |
1361 | was specified in the | |
1362 | .I flags | |
1363 | mask. | |
1364 | .TP | |
bc03b116 | 1365 | .BR EINVAL " (" clone "() only)" |
9f938981 CB |
1366 | .B CLONE_PIDFD |
1367 | was specified together with | |
16853a31 MK |
1368 | .B CLONE_DETACHED |
1369 | in the | |
1370 | .I flags | |
1371 | mask. | |
9f938981 CB |
1372 | .TP |
1373 | .B EINVAL | |
1374 | .B CLONE_PIDFD | |
1375 | was specified together with | |
16853a31 MK |
1376 | .B CLONE_THREAD |
1377 | in the | |
1378 | .I flags | |
1379 | mask. | |
9f938981 | 1380 | .TP |
faa0e55a | 1381 | .BR "EINVAL " "(" clone "() only)" |
9f938981 CB |
1382 | .B CLONE_PIDFD |
1383 | was specified together with | |
16853a31 MK |
1384 | .B CLONE_PARENT_SETTID |
1385 | in the | |
1386 | .I flags | |
1387 | mask. | |
9f938981 | 1388 | .TP |
bf031aaa AR |
1389 | .BR EINVAL " (" clone3 "() only)" |
1390 | .I set_tid_size | |
ee8bb310 | 1391 | is greater than the number of nested PID namespaces. |
bf031aaa AR |
1392 | .TP |
1393 | .BR EINVAL " (" clone3 "() only)" | |
2a2b2a5d | 1394 | One of the PIDs specified in |
bf031aaa | 1395 | .I set_tid |
2a2b2a5d | 1396 | was an invalid. |
bf031aaa | 1397 | .TP |
ba9ae75d MK |
1398 | .BR EINVAL " (AArch64 only, Linux 4.6 and earlier)" |
1399 | .I stack | |
1400 | was not aligned to a 126-bit boundary. | |
1401 | .TP | |
fea681da MK |
1402 | .B ENOMEM |
1403 | Cannot allocate sufficient memory to allocate a task structure for the | |
1404 | child, or to copy those parts of the caller's context that need to be | |
1405 | copied. | |
1406 | .TP | |
b20e22ae MK |
1407 | .BR ENOSPC " (since Linux 3.7)" |
1408 | .\" commit f2302505775fd13ba93f034206f1e2a587017929 | |
1409 | .B CLONE_NEWPID | |
16853a31 MK |
1410 | was specified in the |
1411 | .I flags | |
1412 | mask, | |
b20e22ae MK |
1413 | but the limit on the nesting depth of PID namespaces |
1414 | would have been exceeded; see | |
1415 | .BR pid_namespaces (7). | |
1416 | .TP | |
b5742ecc MK |
1417 | .BR ENOSPC " (since Linux 4.9; beforehand " EUSERS ) |
1418 | .B CLONE_NEWUSER | |
16853a31 MK |
1419 | was specified in the |
1420 | .IR flags | |
1421 | mask, and the call would cause the limit on the number of | |
b5742ecc MK |
1422 | nested user namespaces to be exceeded. |
1423 | See | |
1424 | .BR user_namespaces (7). | |
efeece04 | 1425 | .IP |
b5742ecc MK |
1426 | From Linux 3.11 to Linux 4.8, the error diagnosed in this case was |
1427 | .BR EUSERS . | |
1428 | .TP | |
2f7a331e | 1429 | .BR ENOSPC " (since Linux 4.9)" |
16853a31 | 1430 | One of the values in the |
2f7a331e | 1431 | .I flags |
16853a31 | 1432 | mask specified the creation of a new user namespace, |
2f7a331e MK |
1433 | but doing so would have caused the limit defined by the corresponding file in |
1434 | .IR /proc/sys/user | |
1435 | to be exceeded. | |
1436 | For further details, see | |
1437 | .BR namespaces (7). | |
1438 | .TP | |
6ba79da9 MK |
1439 | .BR EOPNOTSUP " (" clone3 "() only)" |
1440 | .B CLONE_INTO_CGROUP | |
1441 | was specified in | |
1442 | .IR cl_args.flags , | |
1443 | but the file descriptor specified in | |
1444 | .IR cl_args.cgroup | |
1445 | refers to a version 2 cgroup that is in the | |
1446 | .IR "domain invalid" | |
1447 | state. | |
1448 | .TP | |
fea681da | 1449 | .B EPERM |
aa825b59 | 1450 | .BR CLONE_NEWCGROUP , |
667417b3 | 1451 | .BR CLONE_NEWIPC , |
163bf178 | 1452 | .BR CLONE_NEWNET , |
43ce9dda MK |
1453 | .BR CLONE_NEWNS , |
1454 | .BR CLONE_NEWPID , | |
82ee147a | 1455 | or |
43ce9dda | 1456 | .BR CLONE_NEWUTS |
00b08db3 | 1457 | was specified by an unprivileged process (process without \fBCAP_SYS_ADMIN\fP). |
fea681da MK |
1458 | .TP |
1459 | .B EPERM | |
1460 | .B CLONE_PID | |
1461 | was specified by a process other than process 0. | |
1c173eb3 | 1462 | (This error occurs only on Linux 2.5.15 and earlier.) |
365d292a MK |
1463 | .TP |
1464 | .B EPERM | |
1465 | .BR CLONE_NEWUSER | |
16853a31 MK |
1466 | was specified in the |
1467 | .IR flags | |
1468 | mask, | |
365d292a MK |
1469 | but either the effective user ID or the effective group ID of the caller |
1470 | does not have a mapping in the parent namespace (see | |
f58fb24f | 1471 | .BR user_namespaces (7)). |
6fd119e7 | 1472 | .TP |
ac007938 MK |
1473 | .BR EPERM " (since Linux 3.9)" |
1474 | .\" commit 3151527ee007b73a0ebd296010f1c0454a919c7d | |
11a38815 | 1475 | .B CLONE_NEWUSER |
16853a31 | 1476 | was specified in the |
ac007938 | 1477 | .I flags |
16853a31 | 1478 | mask and the caller is in a chroot environment |
ac007938 MK |
1479 | .\" FIXME What is the rationale for this restriction? |
1480 | (i.e., the caller's root directory does not match the root directory | |
1481 | of the mount namespace in which it resides). | |
1482 | .TP | |
bf031aaa AR |
1483 | .BR EPERM " (" clone3 "() only)" |
1484 | .I set_tid_size | |
1485 | was greater than zero, and the caller lacks the | |
1486 | .B CAP_SYS_ADMIN | |
1487 | capability in one or more of the user namespaces that own the | |
1488 | corresponding PID namespaces. | |
1489 | .TP | |
6717ee86 MK |
1490 | .BR ERESTARTNOINTR " (since Linux 2.6.17)" |
1491 | .\" commit 4a2c7a7837da1b91468e50426066d988050e4d56 | |
1492 | System call was interrupted by a signal and will be restarted. | |
1493 | (This can be seen only during a trace.) | |
1494 | .TP | |
b5742ecc | 1495 | .BR EUSERS " (Linux 3.11 to Linux 4.8)" |
6fd119e7 | 1496 | .B CLONE_NEWUSER |
16853a31 MK |
1497 | was specified in the |
1498 | .IR flags | |
1499 | mask, | |
b5742ecc MK |
1500 | and the limit on the number of nested user namespaces would be exceeded. |
1501 | See the discussion of the | |
1502 | .BR ENOSPC | |
1503 | error above. | |
faa0e55a MK |
1504 | .SH VERSIONS |
1505 | The | |
1506 | .BR clone3 () | |
1507 | system call first appeared in Linux 5.3. | |
92b72224 MK |
1508 | .\" There is no entry for |
1509 | .\" .BR clone () | |
1510 | .\" in libc5. | |
1511 | .\" glibc2 provides | |
1512 | .\" .BR clone () | |
1513 | .\" as described in this manual page. | |
47297adb | 1514 | .SH CONFORMING TO |
faa0e55a MK |
1515 | These system calls |
1516 | are Linux-specific and should not be used in programs | |
a1d5f77c | 1517 | intended to be portable. |
fea681da | 1518 | .SH NOTES |
673d16da MK |
1519 | .PP |
1520 | One use of these systems calls | |
1521 | is to implement threads: multiple flows of control in a program that | |
1522 | run concurrently in a shared address space. | |
1523 | .PP | |
1524 | Glibc does not provide a wrapper for | |
1525 | .BR clone3 (); | |
1526 | call it using | |
1527 | .BR syscall (2). | |
1528 | .PP | |
1529 | Note that the glibc | |
1530 | .BR clone () | |
1531 | wrapper function makes some changes | |
1532 | in the memory pointed to by | |
1533 | .I stack | |
1534 | (changes required to set the stack up correctly for the child) | |
1535 | .I before | |
1536 | invoking the | |
1537 | .BR clone () | |
1538 | system call. | |
1539 | So, in cases where | |
1540 | .BR clone () | |
1541 | is used to recursively create children, | |
1542 | do not use the buffer employed for the parent's stack | |
1543 | as the stack of the child. | |
1544 | .PP | |
79bdcc4a MK |
1545 | The |
1546 | .BR kcmp (2) | |
1547 | system call can be used to test whether two processes share various | |
49dba87f | 1548 | resources such as a file descriptor table, |
79bdcc4a | 1549 | System V semaphore undo operations, or a virtual address space. |
efeece04 MK |
1550 | .PP |
1551 | .PP | |
c471c363 MK |
1552 | Handlers registered using |
1553 | .BR pthread_atfork (3) | |
5261b0fe | 1554 | are not executed during a clone call. |
efeece04 | 1555 | .PP |
ca8b1e32 | 1556 | In the Linux 2.4.x series, |
fd8a5be4 MK |
1557 | .B CLONE_THREAD |
1558 | generally does not make the parent of the new thread the same | |
1559 | as the parent of the calling process. | |
1560 | However, for kernel versions 2.4.7 to 2.4.18 the | |
1561 | .B CLONE_THREAD | |
1562 | flag implied the | |
c13182ef | 1563 | .B CLONE_PARENT |
ca8b1e32 | 1564 | flag (as in Linux 2.6.0 and later). |
efeece04 | 1565 | .PP |
34ccb744 | 1566 | On i386, |
a5a997ca MK |
1567 | .BR clone () |
1568 | should not be called through vsyscall, but directly through | |
1569 | .IR "int $0x80" . | |
673d16da MK |
1570 | .\" |
1571 | .SS C library/kernel differences | |
1572 | The raw | |
1573 | .BR clone () | |
1574 | system call corresponds more closely to | |
1575 | .BR fork (2) | |
1576 | in that execution in the child continues from the point of the | |
1577 | call. | |
1578 | As such, the | |
1579 | .I fn | |
1580 | and | |
1581 | .I arg | |
1582 | arguments of the | |
1583 | .BR clone () | |
1584 | wrapper function are omitted. | |
1585 | .PP | |
1586 | In contrast to the glibc wrapper, the raw | |
1587 | .BR clone () | |
1588 | system call accepts NULL as a | |
1589 | .I stack | |
1590 | argument (and | |
1591 | .BR clone3 () | |
1592 | likewise allows | |
1593 | .I cl_args.stack | |
1594 | to be NULL). | |
1595 | In this case, the child uses a duplicate of the parent's stack. | |
1596 | (Copy-on-write semantics ensure that the child gets separate copies | |
1597 | of stack pages when either process modifies the stack.) | |
1598 | In this case, for correct operation, the | |
1599 | .B CLONE_VM | |
1600 | option should not be specified. | |
1601 | (If the child | |
1602 | .I shares | |
1603 | the parent's memory because of the use of the | |
1604 | .BR CLONE_VM | |
1605 | flag, | |
1606 | then no copy-on-write duplication occurs and chaos is likely to result.) | |
1607 | .PP | |
1608 | The order of the arguments also differs in the raw system call, | |
1609 | and there are variations in the arguments across architectures, | |
1610 | as detailed in the following paragraphs. | |
1611 | .PP | |
1612 | The raw system call interface on x86-64 and some other architectures | |
1613 | (including sh, tile, and alpha) is: | |
1614 | .PP | |
1615 | .in +4 | |
1616 | .EX | |
1617 | .BI "long clone(unsigned long " flags ", void *" stack , | |
1618 | .BI " int *" parent_tid ", int *" child_tid , | |
1619 | .BI " unsigned long " tls ); | |
1620 | .EE | |
1621 | .in | |
1622 | .PP | |
1623 | On x86-32, and several other common architectures | |
1624 | (including score, ARM, ARM 64, PA-RISC, arc, Power PC, xtensa, | |
1625 | and MIPS), | |
1626 | .\" CONFIG_CLONE_BACKWARDS | |
1627 | the order of the last two arguments is reversed: | |
1628 | .PP | |
1629 | .in +4 | |
1630 | .EX | |
1631 | .BI "long clone(unsigned long " flags ", void *" stack , | |
1632 | .BI " int *" parent_tid ", unsigned long " tls , | |
1633 | .BI " int *" child_tid ); | |
1634 | .EE | |
1635 | .in | |
1636 | .PP | |
1637 | On the cris and s390 architectures, | |
1638 | .\" CONFIG_CLONE_BACKWARDS2 | |
1639 | the order of the first two arguments is reversed: | |
1640 | .PP | |
1641 | .in +4 | |
1642 | .EX | |
1643 | .BI "long clone(void *" stack ", unsigned long " flags , | |
1644 | .BI " int *" parent_tid ", int *" child_tid , | |
1645 | .BI " unsigned long " tls ); | |
1646 | .EE | |
1647 | .in | |
1648 | .PP | |
1649 | On the microblaze architecture, | |
1650 | .\" CONFIG_CLONE_BACKWARDS3 | |
1651 | an additional argument is supplied: | |
1652 | .PP | |
1653 | .in +4 | |
1654 | .EX | |
1655 | .BI "long clone(unsigned long " flags ", void *" stack , | |
1656 | .BI " int " stack_size , "\fR /* Size of stack */" | |
1657 | .BI " int *" parent_tid ", int *" child_tid , | |
1658 | .BI " unsigned long " tls ); | |
1659 | .EE | |
1660 | .in | |
1661 | .\" | |
1662 | .SS blackfin, m68k, and sparc | |
1663 | .\" Mike Frysinger noted in a 2013 mail: | |
1664 | .\" these arches don't define __ARCH_WANT_SYS_CLONE: | |
1665 | .\" blackfin ia64 m68k sparc | |
1666 | The argument-passing conventions on | |
1667 | blackfin, m68k, and sparc are different from the descriptions above. | |
1668 | For details, see the kernel (and glibc) source. | |
1669 | .SS ia64 | |
1670 | On ia64, a different interface is used: | |
1671 | .PP | |
1672 | .in +4 | |
1673 | .EX | |
1674 | .BI "int __clone2(int (*" "fn" ")(void *), " | |
1675 | .BI " void *" stack_base ", size_t " stack_size , | |
1676 | .BI " int " flags ", void *" "arg" ", ... " | |
1677 | .BI " /* pid_t *" parent_tid ", struct user_desc *" tls , | |
1678 | .BI " pid_t *" child_tid " */ );" | |
1679 | .EE | |
1680 | .in | |
1681 | .PP | |
1682 | The prototype shown above is for the glibc wrapper function; | |
1683 | for the system call itself, | |
1684 | the prototype can be described as follows (it is identical to the | |
1685 | .BR clone () | |
1686 | prototype on microblaze): | |
1687 | .PP | |
1688 | .in +4 | |
1689 | .EX | |
1690 | .BI "long clone2(unsigned long " flags ", void *" stack_base , | |
1691 | .BI " int " stack_size , "\fR /* Size of stack */" | |
1692 | .BI " int *" parent_tid ", int *" child_tid , | |
1693 | .BI " unsigned long " tls ); | |
1694 | .EE | |
1695 | .in | |
1696 | .PP | |
1697 | .BR __clone2 () | |
1698 | operates in the same way as | |
1699 | .BR clone (), | |
1700 | except that | |
1701 | .I stack_base | |
1702 | points to the lowest address of the child's stack area, | |
1703 | and | |
1704 | .I stack_size | |
1705 | specifies the size of the stack pointed to by | |
1706 | .IR stack_base . | |
1707 | .SS Linux 2.4 and earlier | |
1708 | In Linux 2.4 and earlier, | |
1709 | .BR clone () | |
1710 | does not take arguments | |
1711 | .IR parent_tid , | |
1712 | .IR tls , | |
1713 | and | |
1714 | .IR child_tid . | |
31830ef0 | 1715 | .SH BUGS |
abcf3b1d MK |
1716 | GNU C library versions 2.3.4 up to and including 2.24 |
1717 | contained a wrapper function for | |
0bfa087b | 1718 | .BR getpid (2) |
abcf3b1d MK |
1719 | that performed caching of PIDs. |
1720 | This caching relied on support in the glibc wrapper for | |
c60237c9 | 1721 | .BR clone (), |
abcf3b1d MK |
1722 | but limitations in the implementation |
1723 | meant that the cache was not up to date in some circumstances. | |
c60237c9 | 1724 | In particular, |
abcf3b1d | 1725 | if a signal was delivered to the child immediately after the |
c60237c9 MK |
1726 | .BR clone () |
1727 | call, then a call to | |
0b80cf56 | 1728 | .BR getpid (2) |
abcf3b1d | 1729 | in a handler for the signal could return the PID |
c60237c9 | 1730 | of the calling process ("the parent"), |
abcf3b1d | 1731 | if the clone wrapper had not yet had a chance to update the PID |
c60237c9 MK |
1732 | cache in the child. |
1733 | (This discussion ignores the case where the child was created using | |
9291ce36 | 1734 | .BR CLONE_THREAD , |
c60237c9 | 1735 | when |
0b80cf56 | 1736 | .BR getpid (2) |
c60237c9 MK |
1737 | .I should |
1738 | return the same value in the child and in the process that called | |
1739 | .BR clone (), | |
a1d48abb | 1740 | since the caller and the child are in the same thread group. |
e7d807b7 | 1741 | The stale-cache problem also does not occur if the |
a1d48abb JR |
1742 | .I flags |
1743 | argument includes | |
1744 | .BR CLONE_VM .) | |
abcf3b1d MK |
1745 | To get the truth, it was sometimes necessary to use code such as the following: |
1746 | .PP | |
47f743f1 MK |
1747 | .in +4n |
1748 | .EX | |
1749 | #include <syscall.h> | |
31830ef0 | 1750 | |
47f743f1 | 1751 | pid_t mypid; |
31830ef0 | 1752 | |
47f743f1 MK |
1753 | mypid = syscall(SYS_getpid); |
1754 | .EE | |
1755 | .in | |
c60237c9 MK |
1756 | .\" See also the following bug reports |
1757 | .\" https://bugzilla.redhat.com/show_bug.cgi?id=417521 | |
1758 | .\" http://sourceware.org/bugzilla/show_bug.cgi?id=6910 | |
abcf3b1d MK |
1759 | .PP |
1760 | Because of the stale-cache problem, as well as other problems noted in | |
1761 | .BR getpid (2), | |
1762 | the PID caching feature was removed in glibc 2.25. | |
8c7b566c | 1763 | .SH EXAMPLE |
8c7b566c | 1764 | The following program demonstrates the use of |
9c13072a | 1765 | .BR clone () |
8c7b566c MK |
1766 | to create a child process that executes in a separate UTS namespace. |
1767 | The child changes the hostname in its UTS namespace. | |
1768 | Both parent and child then display the system hostname, | |
1769 | making it possible to see that the hostname | |
1770 | differs in the UTS namespaces of the parent and child. | |
1771 | For an example of the use of this program, see | |
1772 | .BR setns (2). | |
99c3a000 MK |
1773 | .PP |
1774 | Within the sample program, we allocate the memory that is to | |
1775 | be used for the child's stack using | |
1776 | .BR mmap (2) | |
1777 | rather than | |
1778 | .BR malloc (3) | |
1779 | for the following reasons: | |
1780 | .IP * 3 | |
1781 | .BR mmap (2) | |
1782 | allocates a block of memory that starts on a page | |
1783 | boundary and is a multiple of the page size. | |
1784 | This is useful if we want to establish a guard page (a page with protection | |
1785 | .BR PROT_NONE ) | |
1786 | at the end of the stack using | |
1787 | .BR mprotect (2). | |
1788 | .IP * | |
1789 | We can specify the | |
1790 | .BR MAP_STACK | |
1791 | flag to request a mapping that is suitable for a stack. | |
1792 | For the moment, this flag is a no-op on Linux, | |
1793 | but it exists and has effect on some other systems, | |
1794 | so we should include it for portability. | |
f30b7415 | 1795 | .SS Program source |
e7d0bb47 | 1796 | .EX |
8c7b566c MK |
1797 | #define _GNU_SOURCE |
1798 | #include <sys/wait.h> | |
1799 | #include <sys/utsname.h> | |
1800 | #include <sched.h> | |
1801 | #include <string.h> | |
1802 | #include <stdio.h> | |
1803 | #include <stdlib.h> | |
1804 | #include <unistd.h> | |
99c3a000 | 1805 | #include <sys/mman.h> |
8c7b566c | 1806 | |
d1a71985 | 1807 | #define errExit(msg) do { perror(msg); exit(EXIT_FAILURE); \e |
8c7b566c MK |
1808 | } while (0) |
1809 | ||
1810 | static int /* Start function for cloned child */ | |
1811 | childFunc(void *arg) | |
1812 | { | |
1813 | struct utsname uts; | |
1814 | ||
1815 | /* Change hostname in UTS namespace of child */ | |
1816 | ||
1817 | if (sethostname(arg, strlen(arg)) == \-1) | |
1818 | errExit("sethostname"); | |
1819 | ||
07d4e6ea | 1820 | /* Retrieve and display hostname */ |
8c7b566c MK |
1821 | |
1822 | if (uname(&uts) == \-1) | |
1823 | errExit("uname"); | |
d1a71985 | 1824 | printf("uts.nodename in child: %s\en", uts.nodename); |
8c7b566c MK |
1825 | |
1826 | /* Keep the namespace open for a while, by sleeping. | |
1827 | This allows some experimentation\-\-for example, another | |
1828 | process might join the namespace. */ | |
9f1b9726 | 1829 | |
8c7b566c MK |
1830 | sleep(200); |
1831 | ||
1832 | return 0; /* Child terminates now */ | |
1833 | } | |
1834 | ||
1835 | #define STACK_SIZE (1024 * 1024) /* Stack size for cloned child */ | |
1836 | ||
1837 | int | |
1838 | main(int argc, char *argv[]) | |
1839 | { | |
1840 | char *stack; /* Start of stack buffer */ | |
1841 | char *stackTop; /* End of stack buffer */ | |
1842 | pid_t pid; | |
1843 | struct utsname uts; | |
1844 | ||
1845 | if (argc < 2) { | |
d1a71985 | 1846 | fprintf(stderr, "Usage: %s <child\-hostname>\en", argv[0]); |
8c7b566c MK |
1847 | exit(EXIT_SUCCESS); |
1848 | } | |
1849 | ||
99c3a000 | 1850 | /* Allocate memory to be used for the stack of the child */ |
8c7b566c | 1851 | |
99c3a000 MK |
1852 | stack = mmap(NULL, STACK_SIZE, PROT_READ | PROT_WRITE, |
1853 | MAP_PRIVATE | MAP_ANONYMOUS | MAP_STACK, \-1, 0); | |
8eea66b8 | 1854 | if (stack == MAP_FAILED) |
99c3a000 MK |
1855 | errExit("mmap"); |
1856 | ||
8c7b566c MK |
1857 | stackTop = stack + STACK_SIZE; /* Assume stack grows downward */ |
1858 | ||
1859 | /* Create child that has its own UTS namespace; | |
1860 | child commences execution in childFunc() */ | |
1861 | ||
1862 | pid = clone(childFunc, stackTop, CLONE_NEWUTS | SIGCHLD, argv[1]); | |
1863 | if (pid == \-1) | |
1864 | errExit("clone"); | |
d1a71985 | 1865 | printf("clone() returned %ld\en", (long) pid); |
8c7b566c MK |
1866 | |
1867 | /* Parent falls through to here */ | |
1868 | ||
1869 | sleep(1); /* Give child time to change its hostname */ | |
1870 | ||
9f1b9726 | 1871 | /* Display hostname in parent\(aqs UTS namespace. This will be |
8c7b566c MK |
1872 | different from hostname in child\(aqs UTS namespace. */ |
1873 | ||
1874 | if (uname(&uts) == \-1) | |
1875 | errExit("uname"); | |
d1a71985 | 1876 | printf("uts.nodename in parent: %s\en", uts.nodename); |
8c7b566c MK |
1877 | |
1878 | if (waitpid(pid, NULL, 0) == \-1) /* Wait for child */ | |
1879 | errExit("waitpid"); | |
d1a71985 | 1880 | printf("child has terminated\en"); |
8c7b566c MK |
1881 | |
1882 | exit(EXIT_SUCCESS); | |
1883 | } | |
e7d0bb47 | 1884 | .EE |
47297adb | 1885 | .SH SEE ALSO |
fea681da | 1886 | .BR fork (2), |
2b44301c | 1887 | .BR futex (2), |
fea681da MK |
1888 | .BR getpid (2), |
1889 | .BR gettid (2), | |
6f8746e4 | 1890 | .BR kcmp (2), |
99c3a000 | 1891 | .BR mmap (2), |
d8837668 | 1892 | .BR pidfd_open (2), |
f2d0bbf1 | 1893 | .BR set_thread_area (2), |
2b44301c | 1894 | .BR set_tid_address (2), |
8403481f | 1895 | .BR setns (2), |
f2d0bbf1 | 1896 | .BR tkill (2), |
5cc01e9c | 1897 | .BR unshare (2), |
fea681da | 1898 | .BR wait (2), |
3616b7c0 | 1899 | .BR capabilities (7), |
41096af1 | 1900 | .BR namespaces (7), |
3616b7c0 | 1901 | .BR pthreads (7) |