man2/vfork.2

   1 .\" Copyright (c) 1999 Andries Brouwer (aeb@cwi.nl), 1 Nov 1999
   2 .\"
   3 .\" Permission is granted to make and distribute verbatim copies of this
   4 .\" manual provided the copyright notice and this permission notice are
   5 .\" preserved on all copies.
   6 .\"
   7 .\" Permission is granted to copy and distribute modified versions of this
   8 .\" manual under the conditions for verbatim copying, provided that the
   9 .\" entire resulting derived work is distributed under the terms of a
  10 .\" permission notice identical to this one.
  11 .\"
  12 .\" Since the Linux kernel and libraries are constantly changing, this
  13 .\" manual page may be incorrect or out-of-date.  The author(s) assume no
  14 .\" responsibility for errors or omissions, or for damages resulting from
  15 .\" the use of the information contained herein.  The author(s) may not
  16 .\" have taken the same level of care in the production of this manual,
  17 .\" which is licensed free of charge, as they might when working
  18 .\" professionally.
  19 .\"
  20 .\" Formatted or processed versions of this manual, if unaccompanied by
  21 .\" the source, must acknowledge the copyright and authors of this work.
  22 .\"
  23 .\" 1999-11-10: Merged text taken from the page contributed by
  24 .\" Reed H. Petty (rhp@draper.net)
  25 .\"
  26 .TH VFORK 2 2009-06-21 "Linux" "Linux Programmer's Manual"
  27 .SH NAME
  28 vfork \- create a child process and block parent
  29 .SH SYNOPSIS
  30 .B #include <sys/types.h>
  31 .br
  32 .B #include <unistd.h>
  33 .sp
  34 .B pid_t vfork(void);
  35 .sp
  36 .in -4n
  37 Feature Test Macro Requirements for glibc (see
  38 .BR feature_test_macros (7)):
  39 .in
  40 .sp
  41 .BR vfork ():
  42 .ad l
  43 .RS 4
  44 .PD 0
  45 .TP 4
  46 Since glibc 2.12:
  47 .nf
  48 _BSD_SOURCE ||
  49     (_XOPEN_SOURCE >= 500 ||
  50         _XOPEN_SOURCE && _XOPEN_SOURCE_EXTENDED) &&
  51     !(_POSIX_C_SOURCE >= 200809L || _XOPEN_SOURCE >= 700)
  52 .TP 4
  53 .fi
  54 Before glibc 2.12:
  55 _BSD_SOURCE || _XOPEN_SOURCE >= 500 || _XOPEN_SOURCE && _XOPEN_SOURCE_EXTENDED
  56 .PD
  57 .RE
  58 .ad b
  59 .SH DESCRIPTION
  60 .SS "Standard Description"
  61 (From POSIX.1)
  62 The
  63 .BR vfork ()
  64 function has the same effect as
  65 .BR fork (2),
  66 except that the behavior is undefined if the process created by
  67 .BR vfork ()
  68 either modifies any data other than a variable of type
  69 .I pid_t
  70 used to store the return value from
  71 .BR vfork (),
  72 or returns from the function in which
  73 .BR vfork ()
  74 was called, or calls any other function before successfully calling
  75 .BR _exit (2)
  76 or one of the
  77 .BR exec (3)
  78 family of functions.
  79 .SS "Linux Description"
  80 .BR vfork (),
  81 just like
  82 .BR fork (2),
  83 creates a child process of the calling process.
  84 For details and return value and errors, see
  85 .BR fork (2).
  86 .PP
  87 .BR vfork ()
  88 is a special case of
  89 .BR clone (2).
  90 It is used to create new processes without copying the page tables of
  91 the parent process.
  92 It may be useful in performance-sensitive applications
  93 where a child is created which then immediately issues an
  94 .BR execve (2).
  95 .PP
  96 .BR vfork ()
  97 differs from
  98 .BR fork (2)
  99 in that the parent is suspended until the child terminates
 100 (either normally,
 101 by calling
 102 .BR _exit (2),
 103 or abnormally, after delivery of a fatal signal),
 104 or it makes a call to
 105 .BR execve (2).
 106 Until that point, the child shares all memory with its parent,
 107 including the stack.
 108 The child must not return from the current function or call
 109 .BR exit (3),
 110 but may call
 111 .BR _exit (2).
 112 .PP
 113 Signal handlers are inherited, but not shared.
 114 Signals to the parent
 115 arrive after the child releases the parent's memory
 116 (i.e., after the child terminates
 117 or calls
 118 .BR execve (2)).
 119 .SS "Historic Description"
 120 Under Linux,
 121 .BR fork (2)
 122 is implemented using copy-on-write pages, so the only penalty incurred by
 123 .BR fork (2)
 124 is the time and memory required to duplicate the parent's page tables,
 125 and to create a unique task structure for the child.
 126 However, in the bad old days a
 127 .BR fork (2)
 128 would require making a complete copy of the caller's data space,
 129 often needlessly, since usually immediately afterwards an
 130 .BR exec (3)
 131 is done.
 132 Thus, for greater efficiency, BSD introduced the
 133 .BR vfork ()
 134 system call, which did not fully copy the address space of
 135 the parent process, but borrowed the parent's memory and thread
 136 of control until a call to
 137 .BR execve (2)
 138 or an exit occurred.
 139 The parent process was suspended while the
 140 child was using its resources.
 141 The use of
 142 .BR vfork ()
 143 was tricky: for example, not modifying data
 144 in the parent process depended on knowing which variables were
 145 held in a register.
 146 .SH "CONFORMING TO"
 147 4.3BSD, POSIX.1-2001.
 148 POSIX.1-2008 removes the specification of
 149 .BR vfork ().
 150 The requirements put on
 151 .BR vfork ()
 152 by the standards are weaker than those put on
 153 .BR fork (2),
 154 so an implementation where the two are synonymous is compliant.
 155 In particular, the programmer cannot rely on the parent
 156 remaining blocked until the child either terminates or calls
 157 .BR execve (2),
 158 and cannot rely on any specific behavior with respect to shared memory.
 159 .\" In AIXv3.1 vfork is equivalent to fork.
 160 .SH NOTES
 161 .SS Linux Notes
 162 Fork handlers established using
 163 .BR pthread_atfork (3)
 164 are not called when a multithreaded program employing
 165 the NPTL threading library calls
 166 .BR vfork ().
 167 Fork handlers are called in this case in a program using the
 168 LinuxThreads threading library.
 169 (See
 170 .BR pthreads (7)
 171 for a description of Linux threading libraries.)
 172 .SS History
 173 The
 174 .BR vfork ()
 175 system call appeared in 3.0BSD.
 176 .\" In the release notes for 4.2BSD Sam Leffler wrote: `vfork: Is still
 177 .\" present, but definitely on its way out'.
 178 In 4.4BSD it was made synonymous to
 179 .BR fork (2)
 180 but NetBSD introduced it again,
 181 cf. http://www.netbsd.org/Documentation/kernel/vfork.html .
 182 In Linux, it has been equivalent to
 183 .BR fork (2)
 184 until 2.2.0-pre6 or so.
 185 Since 2.2.0-pre9 (on i386, somewhat later on
 186 other architectures) it is an independent system call.
 187 Support was added in glibc 2.0.112.
 188 .SH BUGS
 189 It is rather unfortunate that Linux revived this specter from the past.
 190 The BSD man page states:
 191 "This system call will be eliminated when proper system sharing mechanisms
 192 are implemented.
 193 Users should not depend on the memory sharing semantics of
 194 .BR vfork ()
 195 as it will, in that case, be made synonymous to
 196 .BR fork (2).\c
 197 "
 198
 199 Details of the signal handling are obscure and differ between systems.
 200 The BSD man page states:
 201 "To avoid a possible deadlock situation, processes that are children
 202 in the middle of a
 203 .BR vfork ()
 204 are never sent
 205 .B SIGTTOU
 206 or
 207 .B SIGTTIN
 208 signals; rather, output or
 209 .IR ioctl s
 210 are allowed and input attempts result in an end-of-file indication."
 211 .\"
 212 .\" As far as I can tell, the following is not true in 2.6.19:
 213 .\" Currently (Linux 2.3.25),
 214 .\" .BR strace (1)
 215 .\" cannot follow
 216 .\" .BR vfork ()
 217 .\" and requires a kernel patch.
 218 .SH "SEE ALSO"
 219 .BR clone (2),
 220 .BR execve (2),
 221 .BR fork (2),
 222 .BR unshare (2),
 223 .BR wait (2)