man2/clone.2

   1 .\" Copyright (c) 1992 Drew Eckhardt <drew@cs.colorado.edu>, March 28, 1992
   2 .\" and Copyright (c) Michael Kerrisk, 2001, 2002, 2005, 2013
   3 .\"
   4 .\" %%%LICENSE_START(GPL_NOVERSION_ONELINE)
   5 .\" May be distributed under the GNU General Public License.
   6 .\" %%%LICENSE_END
   7 .\"
   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
  41 .\"
  42 .TH CLONE 2 2015-02-21 "Linux" "Linux Programmer's Manual"
  43 .SH NAME
  44 clone, __clone2 \- create a child process
  45 .SH SYNOPSIS
  46 .nf
  47 /* Prototype for the glibc wrapper function */
  48
  49 .B #include <sched.h>
  50
  51 .BI "int clone(int (*" "fn" ")(void *), void *" child_stack ,
  52 .BI "          int " flags ", void *" "arg" ", ... "
  53 .BI "          /* pid_t *" ptid ", struct user_desc *" tls \
  54 ", pid_t *" ctid " */ );"
  55
  56 /* Prototype for the raw system call */
  57
  58 .BI "long clone(unsigned long " flags ", void *" child_stack ,
  59 .BI "          void *" ptid ", void *" ctid ,
  60 .BI "          struct pt_regs *" regs );
  61 .fi
  62 .sp
  63 .in -4n
  64 Feature Test Macro Requirements for glibc wrapper function (see
  65 .BR feature_test_macros (7)):
  66 .in
  67 .sp
  68 .BR clone ():
  69 .ad l
  70 .RS 4
  71 .PD 0
  72 .TP 4
  73 Since glibc 2.14:
  74 _GNU_SOURCE
  75 .TP 4
  76 .\" See http://sources.redhat.com/bugzilla/show_bug.cgi?id=4749
  77 Before glibc 2.14:
  78 _BSD_SOURCE || _SVID_SOURCE
  79     /* _GNU_SOURCE also suffices */
  80 .PD
  81 .RE
  82 .ad b
  83 .SH DESCRIPTION
  84 .BR clone ()
  85 creates a new process, in a manner similar to
  86 .BR fork (2).
  87
  88 This page describes both the glibc
  89 .BR clone ()
  90 wrapper function and the underlying system call on which it is based.
  91 The main text describes the wrapper function;
  92 the differences for the raw system call
  93 are described toward the end of this page.
  94
  95 Unlike
  96 .BR fork (2),
  97 .BR clone ()
  98 allows the child process to share parts of its execution context with
  99 the calling process, such as the memory space, the table of file
 100 descriptors, and the table of signal handlers.
 101 (Note that on this manual
 102 page, "calling process" normally corresponds to "parent process".
 103 But see the description of
 104 .B CLONE_PARENT
 105 below.)
 106
 107 The main use of
 108 .BR clone ()
 109 is to implement threads: multiple threads of control in a program that
 110 run concurrently in a shared memory space.
 111
 112 When the child process is created with
 113 .BR clone (),
 114 it executes the function
 115 .IR fn ( arg ).
 116 (This differs from
 117 .BR fork (2),
 118 where execution continues in the child from the point
 119 of the
 120 .BR fork (2)
 121 call.)
 122 The
 123 .I fn
 124 argument is a pointer to a function that is called by the child
 125 process at the beginning of its execution.
 126 The
 127 .I arg
 128 argument is passed to the
 129 .I fn
 130 function.
 131
 132 When the
 133 .IR fn ( arg )
 134 function application returns, the child process terminates.
 135 The integer returned by
 136 .I fn
 137 is the exit code for the child process.
 138 The child process may also terminate explicitly by calling
 139 .BR exit (2)
 140 or after receiving a fatal signal.
 141
 142 The
 143 .I child_stack
 144 argument specifies the location of the stack used by the child process.
 145 Since the child and calling process may share memory,
 146 it is not possible for the child process to execute in the
 147 same stack as the calling process.
 148 The calling process must therefore
 149 set up memory space for the child stack and pass a pointer to this
 150 space to
 151 .BR clone ().
 152 Stacks grow downward on all processors that run Linux
 153 (except the HP PA processors), so
 154 .I child_stack
 155 usually points to the topmost address of the memory space set up for
 156 the child stack.
 157
 158 The low byte of
 159 .I flags
 160 contains the number of the
 161 .I "termination signal"
 162 sent to the parent when the child dies.
 163 If this signal is specified as anything other than
 164 .BR SIGCHLD ,
 165 then the parent process must specify the
 166 .B __WALL
 167 or
 168 .B __WCLONE
 169 options when waiting for the child with
 170 .BR wait (2).
 171 If no signal is specified, then the parent process is not signaled
 172 when the child terminates.
 173
 174 .I flags
 175 may also be bitwise-or'ed with zero or more of the following constants,
 176 in order to specify what is shared between the calling process
 177 and the child process:
 178 .TP
 179 .BR CLONE_CHILD_CLEARTID " (since Linux 2.5.49)"
 180 Erase child thread ID at location
 181 .I ctid
 182 in child memory when the child exits, and do a wakeup on the futex
 183 at that address.
 184 The address involved may be changed by the
 185 .BR set_tid_address (2)
 186 system call.
 187 This is used by threading libraries.
 188 .TP
 189 .BR CLONE_CHILD_SETTID " (since Linux 2.5.49)"
 190 Store child thread ID at location
 191 .I ctid
 192 in child memory.
 193 .TP
 194 .BR CLONE_FILES " (since Linux 2.0)"
 195 If
 196 .B CLONE_FILES
 197 is set, the calling process and the child process share the same file
 198 descriptor table.
 199 Any file descriptor created by the calling process or by the child
 200 process is also valid in the other process.
 201 Similarly, if one of the processes closes a file descriptor,
 202 or changes its associated flags (using the
 203 .BR fcntl (2)
 204 .B F_SETFD
 205 operation), the other process is also affected.
 206 If a process sharing a file descriptor table calls
 207 .BR execve (2),
 208 its file descriptor table is duplicated (unshared).
 209
 210 If
 211 .B CLONE_FILES
 212 is not set, the child process inherits a copy of all file descriptors
 213 opened in the calling process at the time of
 214 .BR clone ().
 215 (The duplicated file descriptors in the child refer to the
 216 same open file descriptions (see
 217 .BR open (2))
 218 as the corresponding file descriptors in the calling process.)
 219 Subsequent operations that open or close file descriptors,
 220 or change file descriptor flags,
 221 performed by either the calling
 222 process or the child process do not affect the other process.
 223 .TP
 224 .BR CLONE_FS " (since Linux 2.0)"
 225 If
 226 .B CLONE_FS
 227 is set, the caller and the child process share the same filesystem
 228 information.
 229 This includes the root of the filesystem, the current
 230 working directory, and the umask.
 231 Any call to
 232 .BR chroot (2),
 233 .BR chdir (2),
 234 or
 235 .BR umask (2)
 236 performed by the calling process or the child process also affects the
 237 other process.
 238
 239 If
 240 .B CLONE_FS
 241 is not set, the child process works on a copy of the filesystem
 242 information of the calling process at the time of the
 243 .BR clone ()
 244 call.
 245 Calls to
 246 .BR chroot (2),
 247 .BR chdir (2),
 248 .BR umask (2)
 249 performed later by one of the processes do not affect the other process.
 250 .TP
 251 .BR CLONE_IO " (since Linux 2.6.25)"
 252 If
 253 .B CLONE_IO
 254 is set, then the new process shares an I/O context with
 255 the calling process.
 256 If this flag is not set, then (as with
 257 .BR fork (2))
 258 the new process has its own I/O context.
 259
 260 .\" The following based on text from Jens Axboe
 261 The I/O context is the I/O scope of the disk scheduler (i.e,
 262 what the I/O scheduler uses to model scheduling of a process's I/O).
 263 If processes share the same I/O context,
 264 they are treated as one by the I/O scheduler.
 265 As a consequence, they get to share disk time.
 266 For some I/O schedulers,
 267 .\" the anticipatory and CFQ scheduler
 268 if two processes share an I/O context,
 269 they will be allowed to interleave their disk access.
 270 If several threads are doing I/O on behalf of the same process
 271 .RB ( aio_read (3),
 272 for instance), they should employ
 273 .BR CLONE_IO
 274 to get better I/O performance.
 275 .\" with CFQ and AS.
 276
 277 If the kernel is not configured with the
 278 .B CONFIG_BLOCK
 279 option, this flag is a no-op.
 280 .TP
 281 .BR CLONE_NEWIPC " (since Linux 2.6.19)"
 282 If
 283 .B CLONE_NEWIPC
 284 is set, then create the process in a new IPC namespace.
 285 If this flag is not set, then (as with
 286 .BR fork (2)),
 287 the process is created in the same IPC namespace as
 288 the calling process.
 289 This flag is intended for the implementation of containers.
 290
 291 An IPC namespace provides an isolated view of System\ V IPC objects (see
 292 .BR svipc (7))
 293 and (since Linux 2.6.30)
 294 .\" commit 7eafd7c74c3f2e67c27621b987b28397110d643f
 295 .\" https://lwn.net/Articles/312232/
 296 POSIX message queues
 297 (see
 298 .BR mq_overview (7)).
 299 The common characteristic of these IPC mechanisms is that IPC
 300 objects are identified by mechanisms other than filesystem
 301 pathnames.
 302
 303 Objects created in an IPC namespace are visible to all other processes
 304 that are members of that namespace,
 305 but are not visible to processes in other IPC namespaces.
 306
 307 When an IPC namespace is destroyed
 308 (i.e., when the last process that is a member of the namespace terminates),
 309 all IPC objects in the namespace are automatically destroyed.
 310
 311 Only a privileged process
 312 .RB ( CAP_SYS_ADMIN )
 313 can employ
 314 .BR CLONE_NEWIPC .
 315 This flag can't be specified in conjunction with
 316 .BR CLONE_SYSVSEM .
 317
 318 For further information on IPC namespaces, see
 319 .BR namespaces (7).
 320 .TP
 321 .BR CLONE_NEWNET " (since Linux 2.6.24)"
 322 (The implementation of this flag was completed only
 323 by about kernel version 2.6.29.)
 324
 325 If
 326 .B CLONE_NEWNET
 327 is set, then create the process in a new network namespace.
 328 If this flag is not set, then (as with
 329 .BR fork (2))
 330 the process is created in the same network namespace as
 331 the calling process.
 332 This flag is intended for the implementation of containers.
 333
 334 A network namespace provides an isolated view of the networking stack
 335 (network device interfaces, IPv4 and IPv6 protocol stacks,
 336 IP routing tables, firewall rules, the
 337 .I /proc/net
 338 and
 339 .I /sys/class/net
 340 directory trees, sockets, etc.).
 341 A physical network device can live in exactly one
 342 network namespace.
 343 A virtual network device ("veth") pair provides a pipe-like abstraction
 344 .\" FIXME . Add pointer to veth(4) page when it is eventually completed
 345 that can be used to create tunnels between network namespaces,
 346 and can be used to create a bridge to a physical network device
 347 in another namespace.
 348
 349 When a network namespace is freed
 350 (i.e., when the last process in the namespace terminates),
 351 its physical network devices are moved back to the
 352 initial network namespace (not to the parent of the process).
 353 For further information on network namespaces, see
 354 .BR namespaces (7).
 355
 356 Only a privileged process
 357 .RB ( CAP_SYS_ADMIN )
 358 can employ
 359 .BR CLONE_NEWNET .
 360 .TP
 361 .BR CLONE_NEWNS " (since Linux 2.4.19)"
 362 If
 363 .B CLONE_NEWNS
 364 is set, the cloned child is started in a new mount namespace,
 365 initialized with a copy of the namespace of the parent.
 366 If
 367 .B CLONE_NEWNS
 368 is not set, the child lives in the same mount
 369 namespace as the parent.
 370
 371 For further information on mount namespaces, see
 372 .BR namespaces (7).
 373
 374 Only a privileged process
 375 .RB ( CAP_SYS_ADMIN )
 376 can employ
 377 .BR CLONE_NEWNS .
 378 It is not permitted to specify both
 379 .B CLONE_NEWNS
 380 and
 381 .B CLONE_FS
 382 .\" See https://lwn.net/Articles/543273/
 383 in the same
 384 .BR clone ()
 385 call.
 386 .TP
 387 .BR CLONE_NEWPID " (since Linux 2.6.24)"
 388 .\" This explanation draws a lot of details from
 389 .\" http://lwn.net/Articles/259217/
 390 .\" Authors: Pavel Emelyanov <xemul@openvz.org>
 391 .\" and Kir Kolyshkin <kir@openvz.org>
 392 .\"
 393 .\" The primary kernel commit is 30e49c263e36341b60b735cbef5ca37912549264
 394 .\" Author: Pavel Emelyanov <xemul@openvz.org>
 395 If
 396 .B CLONE_NEWPID
 397 is set, then create the process in a new PID namespace.
 398 If this flag is not set, then (as with
 399 .BR fork (2))
 400 the process is created in the same PID namespace as
 401 the calling process.
 402 This flag is intended for the implementation of containers.
 403
 404 For further information on PID namespaces, see
 405 .BR namespaces (7)
 406 and
 407 .BR pid_namespaces (7)
 408
 409 Only a privileged process
 410 .RB ( CAP_SYS_ADMIN )
 411 can employ
 412 .BR CLONE_NEWPID .
 413 This flag can't be specified in conjunction with
 414 .BR CLONE_THREAD
 415 or
 416 .BR CLONE_PARENT .
 417 .TP
 418 .BR CLONE_NEWUSER
 419 (This flag first became meaningful for
 420 .BR clone ()
 421 in Linux 2.6.23,
 422 the current
 423 .BR clone ()
 424 semantics were merged in Linux 3.5,
 425 and the final pieces to make the user namespaces completely usable were
 426 merged in Linux 3.8.)
 427
 428 If
 429 .B CLONE_NEWUSER
 430 is set, then create the process in a new user namespace.
 431 If this flag is not set, then (as with
 432 .BR fork (2))
 433 the process is created in the same user namespace as the calling process.
 434
 435 For further information on user namespaces, see
 436 .BR namespaces (7)
 437 and
 438 .BR user_namespaces (7)
 439
 440 Before Linux 3.8, use of
 441 .BR CLONE_NEWUSER
 442 required that the caller have three capabilities:
 443 .BR CAP_SYS_ADMIN ,
 444 .BR CAP_SETUID ,
 445 and
 446 .BR CAP_SETGID .
 447 .\" Before Linux 2.6.29, it appears that only CAP_SYS_ADMIN was needed
 448 Starting with Linux 3.8,
 449 no privileges are needed to create a user namespace.
 450
 451 This flag can't be specified in conjunction with
 452 .BR CLONE_THREAD
 453 or
 454 .BR CLONE_PARENT .
 455 For security reasons,
 456 .\" commit e66eded8309ebf679d3d3c1f5820d1f2ca332c71
 457 .\" https://lwn.net/Articles/543273/
 458 .\" The fix actually went into 3.9 and into 3.8.3. However, user namespaces
 459 .\" were, for practical purposes, unusable in earlier 3.8.x because of the
 460 .\" various filesystems that didn't support userns.
 461 .BR CLONE_NEWUSER
 462 cannot be specified in conjunction with
 463 .BR CLONE_FS .
 464
 465 For further information on user namespaces, see
 466 .BR user_namespaces (7).
 467 .TP
 468 .BR CLONE_NEWUTS " (since Linux 2.6.19)"
 469 If
 470 .B CLONE_NEWUTS
 471 is set, then create the process in a new UTS namespace,
 472 whose identifiers are initialized by duplicating the identifiers
 473 from the UTS namespace of the calling process.
 474 If this flag is not set, then (as with
 475 .BR fork (2))
 476 the process is created in the same UTS namespace as
 477 the calling process.
 478 This flag is intended for the implementation of containers.
 479
 480 A UTS namespace is the set of identifiers returned by
 481 .BR uname (2);
 482 among these, the domain name and the hostname can be modified by
 483 .BR setdomainname (2)
 484 and
 485 .BR sethostname (2),
 486 respectively.
 487 Changes made to the identifiers in a UTS namespace
 488 are visible to all other processes in the same namespace,
 489 but are not visible to processes in other UTS namespaces.
 490
 491 Only a privileged process
 492 .RB ( CAP_SYS_ADMIN )
 493 can employ
 494 .BR CLONE_NEWUTS .
 495
 496 For further information on UTS namespaces, see
 497 .BR namespaces (7).
 498 .TP
 499 .BR CLONE_PARENT " (since Linux 2.3.12)"
 500 If
 501 .B CLONE_PARENT
 502 is set, then the parent of the new child (as returned by
 503 .BR getppid (2))
 504 will be the same as that of the calling process.
 505
 506 If
 507 .B CLONE_PARENT
 508 is not set, then (as with
 509 .BR fork (2))
 510 the child's parent is the calling process.
 511
 512 Note that it is the parent process, as returned by
 513 .BR getppid (2),
 514 which is signaled when the child terminates, so that
 515 if
 516 .B CLONE_PARENT
 517 is set, then the parent of the calling process, rather than the
 518 calling process itself, will be signaled.
 519 .TP
 520 .BR CLONE_PARENT_SETTID " (since Linux 2.5.49)"
 521 Store child thread ID at location
 522 .I ptid
 523 in parent memory.
 524 (In Linux 2.5.32-2.5.48 there was a flag
 525 .B CLONE_SETTID
 526 that did this.)
 527 .TP
 528 .BR CLONE_PID " (obsolete)"
 529 If
 530 .B CLONE_PID
 531 is set, the child process is created with the same process ID as
 532 the calling process.
 533 This is good for hacking the system, but otherwise
 534 of not much use.
 535 Since 2.3.21 this flag can be
 536 specified only by the system boot process (PID 0).
 537 It disappeared in Linux 2.5.16.
 538 .TP
 539 .BR CLONE_PTRACE " (since Linux 2.2)"
 540 If
 541 .B CLONE_PTRACE
 542 is specified, and the calling process is being traced,
 543 then trace the child also (see
 544 .BR ptrace (2)).
 545 .TP
 546 .BR CLONE_SETTLS " (since Linux 2.5.32)"
 547 The
 548 .I newtls
 549 argument is the new TLS (Thread Local Storage) descriptor.
 550 (See
 551 .BR set_thread_area (2).)
 552 .TP
 553 .BR CLONE_SIGHAND " (since Linux 2.0)"
 554 If
 555 .B CLONE_SIGHAND
 556 is set, the calling process and the child process share the same table of
 557 signal handlers.
 558 If the calling process or child process calls
 559 .BR sigaction (2)
 560 to change the behavior associated with a signal, the behavior is
 561 changed in the other process as well.
 562 However, the calling process and child
 563 processes still have distinct signal masks and sets of pending
 564 signals.
 565 So, one of them may block or unblock some signals using
 566 .BR sigprocmask (2)
 567 without affecting the other process.
 568
 569 If
 570 .B CLONE_SIGHAND
 571 is not set, the child process inherits a copy of the signal handlers
 572 of the calling process at the time
 573 .BR clone ()
 574 is called.
 575 Calls to
 576 .BR sigaction (2)
 577 performed later by one of the processes have no effect on the other
 578 process.
 579
 580 Since Linux 2.6.0-test6,
 581 .I flags
 582 must also include
 583 .B CLONE_VM
 584 if
 585 .B CLONE_SIGHAND
 586 is specified
 587 .TP
 588 .BR CLONE_STOPPED " (since Linux 2.6.0-test2)"
 589 If
 590 .B CLONE_STOPPED
 591 is set, then the child is initially stopped (as though it was sent a
 592 .B SIGSTOP
 593 signal), and must be resumed by sending it a
 594 .B SIGCONT
 595 signal.
 596
 597 This flag was
 598 .I deprecated
 599 from Linux 2.6.25 onward,
 600 and was
 601 .I removed
 602 altogether in Linux 2.6.38.
 603 .\" glibc 2.8 removed this defn from bits/sched.h
 604 .TP
 605 .BR CLONE_SYSVSEM " (since Linux 2.5.10)"
 606 If
 607 .B CLONE_SYSVSEM
 608 is set, then the child and the calling process share
 609 a single list of System V semaphore adjustment
 610 .RI ( semadj )
 611 values (see
 612 .BR semop (2)).
 613 In this case, the shared list accumulates
 614 .I semadj
 615 values across all processes sharing the list,
 616 and semaphore adjustments are performed only when the last process
 617 that is sharing the list terminates (or ceases sharing the list using
 618 .BR unshare (2)).
 619 If this flag is not set, then the child has a separate
 620 .I semadj
 621 list that is initially empty.
 622 .TP
 623 .BR CLONE_THREAD " (since Linux 2.4.0-test8)"
 624 If
 625 .B CLONE_THREAD
 626 is set, the child is placed in the same thread group as the calling process.
 627 To make the remainder of the discussion of
 628 .B CLONE_THREAD
 629 more readable, the term "thread" is used to refer to the
 630 processes within a thread group.
 631
 632 Thread groups were a feature added in Linux 2.4 to support the
 633 POSIX threads notion of a set of threads that share a single PID.
 634 Internally, this shared PID is the so-called
 635 thread group identifier (TGID) for the thread group.
 636 Since Linux 2.4, calls to
 637 .BR getpid (2)
 638 return the TGID of the caller.
 639
 640 The threads within a group can be distinguished by their (system-wide)
 641 unique thread IDs (TID).
 642 A new thread's TID is available as the function result
 643 returned to the caller of
 644 .BR clone (),
 645 and a thread can obtain
 646 its own TID using
 647 .BR gettid (2).
 648
 649 When a call is made to
 650 .BR clone ()
 651 without specifying
 652 .BR CLONE_THREAD ,
 653 then the resulting thread is placed in a new thread group
 654 whose TGID is the same as the thread's TID.
 655 This thread is the
 656 .I leader
 657 of the new thread group.
 658
 659 A new thread created with
 660 .B CLONE_THREAD
 661 has the same parent process as the caller of
 662 .BR clone ()
 663 (i.e., like
 664 .BR CLONE_PARENT ),
 665 so that calls to
 666 .BR getppid (2)
 667 return the same value for all of the threads in a thread group.
 668 When a
 669 .B CLONE_THREAD
 670 thread terminates, the thread that created it using
 671 .BR clone ()
 672 is not sent a
 673 .B SIGCHLD
 674 (or other termination) signal;
 675 nor can the status of such a thread be obtained
 676 using
 677 .BR wait (2).
 678 (The thread is said to be
 679 .IR detached .)
 680
 681 After all of the threads in a thread group terminate
 682 the parent process of the thread group is sent a
 683 .B SIGCHLD
 684 (or other termination) signal.
 685
 686 If any of the threads in a thread group performs an
 687 .BR execve (2),
 688 then all threads other than the thread group leader are terminated,
 689 and the new program is executed in the thread group leader.
 690
 691 If one of the threads in a thread group creates a child using
 692 .BR fork (2),
 693 then any thread in the group can
 694 .BR wait (2)
 695 for that child.
 696
 697 Since Linux 2.5.35,
 698 .I flags
 699 must also include
 700 .B CLONE_SIGHAND
 701 if
 702 .B CLONE_THREAD
 703 is specified
 704 (and note that, since Linux 2.6.0-test6,
 705 .BR CLONE_SIGHAND
 706 also requires
 707 .BR CLONE_VM
 708 to be included).
 709
 710 Signals may be sent to a thread group as a whole (i.e., a TGID) using
 711 .BR kill (2),
 712 or to a specific thread (i.e., TID) using
 713 .BR tgkill (2).
 714
 715 Signal dispositions and actions are process-wide:
 716 if an unhandled signal is delivered to a thread, then
 717 it will affect (terminate, stop, continue, be ignored in)
 718 all members of the thread group.
 719
 720 Each thread has its own signal mask, as set by
 721 .BR sigprocmask (2),
 722 but signals can be pending either: for the whole process
 723 (i.e., deliverable to any member of the thread group),
 724 when sent with
 725 .BR kill (2);
 726 or for an individual thread, when sent with
 727 .BR tgkill (2).
 728 A call to
 729 .BR sigpending (2)
 730 returns a signal set that is the union of the signals pending for the
 731 whole process and the signals that are pending for the calling thread.
 732
 733 If
 734 .BR kill (2)
 735 is used to send a signal to a thread group,
 736 and the thread group has installed a handler for the signal, then
 737 the handler will be invoked in exactly one, arbitrarily selected
 738 member of the thread group that has not blocked the signal.
 739 If multiple threads in a group are waiting to accept the same signal using
 740 .BR sigwaitinfo (2),
 741 the kernel will arbitrarily select one of these threads
 742 to receive a signal sent using
 743 .BR kill (2).
 744 .TP
 745 .BR CLONE_UNTRACED " (since Linux 2.5.46)"
 746 If
 747 .B CLONE_UNTRACED
 748 is specified, then a tracing process cannot force
 749 .B CLONE_PTRACE
 750 on this child process.
 751 .TP
 752 .BR CLONE_VFORK " (since Linux 2.2)"
 753 If
 754 .B CLONE_VFORK
 755 is set, the execution of the calling process is suspended
 756 until the child releases its virtual memory
 757 resources via a call to
 758 .BR execve (2)
 759 or
 760 .BR _exit (2)
 761 (as with
 762 .BR vfork (2)).
 763
 764 If
 765 .B CLONE_VFORK
 766 is not set, then both the calling process and the child are schedulable
 767 after the call, and an application should not rely on execution occurring
 768 in any particular order.
 769 .TP
 770 .BR CLONE_VM " (since Linux 2.0)"
 771 If
 772 .B CLONE_VM
 773 is set, the calling process and the child process run in the same memory
 774 space.
 775 In particular, memory writes performed by the calling process
 776 or by the child process are also visible in the other process.
 777 Moreover, any memory mapping or unmapping performed with
 778 .BR mmap (2)
 779 or
 780 .BR munmap (2)
 781 by the child or calling process also affects the other process.
 782
 783 If
 784 .B CLONE_VM
 785 is not set, the child process runs in a separate copy of the memory
 786 space of the calling process at the time of
 787 .BR clone ().
 788 Memory writes or file mappings/unmappings performed by one of the
 789 processes do not affect the other, as with
 790 .BR fork (2).
 791 .SS C library/kernel ABI differences
 792 The raw
 793 .BR clone ()
 794 system call corresponds more closely to
 795 .BR fork (2)
 796 in that execution in the child continues from the point of the
 797 call.
 798 As such, the
 799 .I fn
 800 and
 801 .I arg
 802 arguments of the
 803 .BR clone ()
 804 wrapper function are omitted.
 805 Furthermore, the argument order changes.
 806 The raw system call interface on x86 and many other architectures is roughly:
 807 .in +4
 808 .nf
 809
 810 .BI "long clone(unsigned long " flags ", void *" child_stack ,
 811 .BI "           void *" ptid ", void *" ctid ,
 812 .BI "           struct pt_regs *" regs );
 813
 814 .fi
 815 .in
 816 Another difference for the raw system call is that the
 817 .I child_stack
 818 argument may be zero, in which case copy-on-write semantics ensure that the
 819 child gets separate copies of stack pages when either process modifies
 820 the stack.
 821 In this case, for correct operation, the
 822 .B CLONE_VM
 823 option should not be specified.
 824
 825 For some architectures, the order of the arguments for the system call
 826 differs from that shown above.
 827 On the score, microblaze, ARM, ARM 64, PA-RISC, arc, Power PC, xtensa,
 828 and MIPS architectures,
 829 the order of the fourth and fifth arguments is reversed.
 830 On the cris and s390 architectures,
 831 the order of the first and second arguments is reversed.
 832 .SS blackfin, m68k, and sparc
 833 The argument-passing conventions on
 834 blackfin, m68k, and sparc are different from the descriptions above.
 835 For details, see the kernel (and glibc) source.
 836 .SS ia64
 837 On ia64, a different interface is used:
 838 .nf
 839
 840 .BI "int __clone2(int (*" "fn" ")(void *), "
 841 .BI "             void *" child_stack_base ", size_t " stack_size ,
 842 .BI "             int " flags ", void *" "arg" ", ... "
 843 .BI "          /* pid_t *" ptid ", struct user_desc *" tls \
 844 ", pid_t *" ctid " */ );"
 845 .fi
 846 .PP
 847 The prototype shown above is for the glibc wrapper function;
 848 the raw system call interface has no
 849 .I fn
 850 or
 851 .I arg
 852 argument, and changes the order of the arguments so that
 853 .I flags
 854 is the first argument, and
 855 .I tls
 856 is the last argument.
 857 .PP
 858 .BR __clone2 ()
 859 operates in the same way as
 860 .BR clone (),
 861 except that
 862 .I child_stack_base
 863 points to the lowest address of the child's stack area,
 864 and
 865 .I stack_size
 866 specifies the size of the stack pointed to by
 867 .IR child_stack_base .
 868 .SS Linux 2.4 and earlier
 869 In Linux 2.4 and earlier,
 870 .BR clone ()
 871 does not take arguments
 872 .IR ptid ,
 873 .IR tls ,
 874 and
 875 .IR ctid .
 876 .SH RETURN VALUE
 877 .\" gettid(2) returns current->pid;
 878 .\" getpid(2) returns current->tgid;
 879 On success, the thread ID of the child process is returned
 880 in the caller's thread of execution.
 881 On failure, \-1 is returned
 882 in the caller's context, no child process will be created, and
 883 .I errno
 884 will be set appropriately.
 885 .SH ERRORS
 886 .TP
 887 .B EAGAIN
 888 Too many processes are already running; see
 889 .BR fork (2).
 890 .TP
 891 .B EINVAL
 892 .B CLONE_SIGHAND
 893 was specified, but
 894 .B CLONE_VM
 895 was not.
 896 (Since Linux 2.6.0-test6.)
 897 .TP
 898 .B EINVAL
 899 .B CLONE_THREAD
 900 was specified, but
 901 .B CLONE_SIGHAND
 902 was not.
 903 (Since Linux 2.5.35.)
 904 .\" .TP
 905 .\" .B EINVAL
 906 .\" Precisely one of
 907 .\" .B CLONE_DETACHED
 908 .\" and
 909 .\" .B CLONE_THREAD
 910 .\" was specified.
 911 .\" (Since Linux 2.6.0-test6.)
 912 .TP
 913 .B EINVAL
 914 .\" commit e66eded8309ebf679d3d3c1f5820d1f2ca332c71
 915 Both
 916 .B CLONE_FS
 917 and
 918 .B CLONE_NEWNS
 919 were specified in
 920 .IR flags .
 921 .TP
 922 .BR EINVAL " (since Linux 3.9)"
 923 Both
 924 .B CLONE_NEWUSER
 925 and
 926 .B CLONE_FS
 927 were specified in
 928 .IR flags .
 929 .TP
 930 .B EINVAL
 931 Both
 932 .B CLONE_NEWIPC
 933 and
 934 .B CLONE_SYSVSEM
 935 were specified in
 936 .IR flags .
 937 .TP
 938 .B EINVAL
 939 One (or both) of
 940 .BR CLONE_NEWPID
 941 or
 942 .BR CLONE_NEWUSER
 943 and one (or both) of
 944 .BR CLONE_THREAD
 945 or
 946 .BR CLONE_PARENT
 947 were specified in
 948 .IR flags .
 949 .TP
 950 .B EINVAL
 951 Returned by
 952 .BR clone ()
 953 when a zero value is specified for
 954 .IR child_stack .
 955 .TP
 956 .B EINVAL
 957 .BR CLONE_NEWIPC
 958 was specified in
 959 .IR flags ,
 960 but the kernel was not configured with the
 961 .B CONFIG_SYSVIPC
 962 and
 963 .BR CONFIG_IPC_NS
 964 options.
 965 .TP
 966 .B EINVAL
 967 .BR CLONE_NEWNET
 968 was specified in
 969 .IR flags ,
 970 but the kernel was not configured with the
 971 .B CONFIG_NET_NS
 972 option.
 973 .TP
 974 .B EINVAL
 975 .BR CLONE_NEWPID
 976 was specified in
 977 .IR flags ,
 978 but the kernel was not configured with the
 979 .B CONFIG_PID_NS
 980 option.
 981 .TP
 982 .B EINVAL
 983 .BR CLONE_NEWUTS
 984 was specified in
 985 .IR flags ,
 986 but the kernel was not configured with the
 987 .B CONFIG_UTS
 988 option.
 989 .TP
 990 .B ENOMEM
 991 Cannot allocate sufficient memory to allocate a task structure for the
 992 child, or to copy those parts of the caller's context that need to be
 993 copied.
 994 .TP
 995 .B EPERM
 996 .BR CLONE_NEWIPC ,
 997 .BR CLONE_NEWNET ,
 998 .BR CLONE_NEWNS ,
 999 .BR CLONE_NEWPID ,
1000 or
1001 .BR CLONE_NEWUTS
1002 was specified by an unprivileged process (process without \fBCAP_SYS_ADMIN\fP).
1003 .TP
1004 .B EPERM
1005 .B CLONE_PID
1006 was specified by a process other than process 0.
1007 .TP
1008 .B EPERM
1009 .BR CLONE_NEWUSER
1010 was specified in
1011 .IR flags ,
1012 but either the effective user ID or the effective group ID of the caller
1013 does not have a mapping in the parent namespace (see
1014 .BR user_namespaces (7)).
1015 .TP
1016 .BR EPERM " (since Linux 3.9)"
1017 .\" commit 3151527ee007b73a0ebd296010f1c0454a919c7d
1018 .B CLONE_NEWUSER
1019 was specified in
1020 .I flags
1021 and the caller is in a chroot environment
1022 .\" FIXME What is the rationale for this restriction?
1023 (i.e., the caller's root directory does not match the root directory
1024 of the mount namespace in which it resides).
1025 .TP
1026 .BR EUSERS " (since Linux 3.11)"
1027 .B CLONE_NEWUSER
1028 was specified in
1029 .IR flags ,
1030 and the call would cause the limit on the number of
1031 nested user namespaces to be exceeded.
1032 See
1033 .BR user_namespaces (7).
1034 .SH VERSIONS
1035 There is no entry for
1036 .BR clone ()
1037 in libc5.
1038 glibc2 provides
1039 .BR clone ()
1040 as described in this manual page.
1041 .SH CONFORMING TO
1042 .BR clone ()
1043 is Linux-specific and should not be used in programs
1044 intended to be portable.
1045 .SH NOTES
1046 In the kernel 2.4.x series,
1047 .B CLONE_THREAD
1048 generally does not make the parent of the new thread the same
1049 as the parent of the calling process.
1050 However, for kernel versions 2.4.7 to 2.4.18 the
1051 .B CLONE_THREAD
1052 flag implied the
1053 .B CLONE_PARENT
1054 flag (as in kernel 2.6).
1055
1056 For a while there was
1057 .B CLONE_DETACHED
1058 (introduced in 2.5.32):
1059 parent wants no child-exit signal.
1060 In 2.6.2 the need to give this
1061 together with
1062 .B CLONE_THREAD
1063 disappeared.
1064 This flag is still defined, but has no effect.
1065
1066 On i386,
1067 .BR clone ()
1068 should not be called through vsyscall, but directly through
1069 .IR "int $0x80" .
1070 .SH BUGS
1071 Versions of the GNU C library that include the NPTL threading library
1072 contain a wrapper function for
1073 .BR getpid (2)
1074 that performs caching of PIDs.
1075 This caching relies on support in the glibc wrapper for
1076 .BR clone (),
1077 but as currently implemented,
1078 the cache may not be up to date in some circumstances.
1079 In particular,
1080 if a signal is delivered to the child immediately after the
1081 .BR clone ()
1082 call, then a call to
1083 .BR getpid (2)
1084 in a handler for the signal may return the PID
1085 of the calling process ("the parent"),
1086 if the clone wrapper has not yet had a chance to update the PID
1087 cache in the child.
1088 (This discussion ignores the case where the child was created using
1089 .BR CLONE_THREAD ,
1090 when
1091 .BR getpid (2)
1092 .I should
1093 return the same value in the child and in the process that called
1094 .BR clone (),
1095 since the caller and the child are in the same thread group.
1096 The stale-cache problem also does not occur if the
1097 .I flags
1098 argument includes
1099 .BR CLONE_VM .)
1100 To get the truth, it may be necessary to use code such as the following:
1101 .nf
1102
1103     #include <syscall.h>
1104
1105     pid_t mypid;
1106
1107     mypid = syscall(SYS_getpid);
1108 .fi
1109 .\" See also the following bug reports
1110 .\" https://bugzilla.redhat.com/show_bug.cgi?id=417521
1111 .\" http://sourceware.org/bugzilla/show_bug.cgi?id=6910
1112 .SH EXAMPLE
1113 The following program demonstrates the use of
1114 .BR clone ()
1115 to create a child process that executes in a separate UTS namespace.
1116 The child changes the hostname in its UTS namespace.
1117 Both parent and child then display the system hostname,
1118 making it possible to see that the hostname
1119 differs in the UTS namespaces of the parent and child.
1120 For an example of the use of this program, see
1121 .BR setns (2).
1122 .SS Program source
1123 .nf
1124 #define _GNU_SOURCE
1125 #include <sys/wait.h>
1126 #include <sys/utsname.h>
1127 #include <sched.h>
1128 #include <string.h>
1129 #include <stdio.h>
1130 #include <stdlib.h>
1131 #include <unistd.h>
1132
1133 #define errExit(msg)    do { perror(msg); exit(EXIT_FAILURE); \\
1134                         } while (0)
1135
1136 static int              /* Start function for cloned child */
1137 childFunc(void *arg)
1138 {
1139     struct utsname uts;
1140
1141     /* Change hostname in UTS namespace of child */
1142
1143     if (sethostname(arg, strlen(arg)) == \-1)
1144         errExit("sethostname");
1145
1146     /* Retrieve and display hostname */
1147
1148     if (uname(&uts) == \-1)
1149         errExit("uname");
1150     printf("uts.nodename in child:  %s\\n", uts.nodename);
1151
1152     /* Keep the namespace open for a while, by sleeping.
1153        This allows some experimentation\-\-for example, another
1154        process might join the namespace. */
1155
1156     sleep(200);
1157
1158     return 0;           /* Child terminates now */
1159 }
1160
1161 #define STACK_SIZE (1024 * 1024)    /* Stack size for cloned child */
1162
1163 int
1164 main(int argc, char *argv[])
1165 {
1166     char *stack;                    /* Start of stack buffer */
1167     char *stackTop;                 /* End of stack buffer */
1168     pid_t pid;
1169     struct utsname uts;
1170
1171     if (argc < 2) {
1172         fprintf(stderr, "Usage: %s <child\-hostname>\\n", argv[0]);
1173         exit(EXIT_SUCCESS);
1174     }
1175
1176     /* Allocate stack for child */
1177
1178     stack = malloc(STACK_SIZE);
1179     if (stack == NULL)
1180         errExit("malloc");
1181     stackTop = stack + STACK_SIZE;  /* Assume stack grows downward */
1182
1183     /* Create child that has its own UTS namespace;
1184        child commences execution in childFunc() */
1185
1186     pid = clone(childFunc, stackTop, CLONE_NEWUTS | SIGCHLD, argv[1]);
1187     if (pid == \-1)
1188         errExit("clone");
1189     printf("clone() returned %ld\\n", (long) pid);
1190
1191     /* Parent falls through to here */
1192
1193     sleep(1);           /* Give child time to change its hostname */
1194
1195     /* Display hostname in parent\(aqs UTS namespace. This will be
1196        different from hostname in child\(aqs UTS namespace. */
1197
1198     if (uname(&uts) == \-1)
1199         errExit("uname");
1200     printf("uts.nodename in parent: %s\\n", uts.nodename);
1201
1202     if (waitpid(pid, NULL, 0) == \-1)    /* Wait for child */
1203         errExit("waitpid");
1204     printf("child has terminated\\n");
1205
1206     exit(EXIT_SUCCESS);
1207 }
1208 .fi
1209 .SH SEE ALSO
1210 .BR fork (2),
1211 .BR futex (2),
1212 .BR getpid (2),
1213 .BR gettid (2),
1214 .BR kcmp (2),
1215 .BR set_thread_area (2),
1216 .BR set_tid_address (2),
1217 .BR setns (2),
1218 .BR tkill (2),
1219 .BR unshare (2),
1220 .BR wait (2),
1221 .BR capabilities (7),
1222 .BR namespaces (7),
1223 .BR pthreads (7)