man2/vfork.2

   1 .\" Copyright (c) 1999 Andries Brouwer (aeb@cwi.nl), 1 Nov 1999
   2 .\"
   3 .\" %%%LICENSE_START(VERBATIM)
   4 .\" Permission is granted to make and distribute verbatim copies of this
   5 .\" manual provided the copyright notice and this permission notice are
   6 .\" preserved on all copies.
   7 .\"
   8 .\" Permission is granted to copy and distribute modified versions of this
   9 .\" manual under the conditions for verbatim copying, provided that the
  10 .\" entire resulting derived work is distributed under the terms of a
  11 .\" permission notice identical to this one.
  12 .\"
  13 .\" Since the Linux kernel and libraries are constantly changing, this
  14 .\" manual page may be incorrect or out-of-date.  The author(s) assume no
  15 .\" responsibility for errors or omissions, or for damages resulting from
  16 .\" the use of the information contained herein.  The author(s) may not
  17 .\" have taken the same level of care in the production of this manual,
  18 .\" which is licensed free of charge, as they might when working
  19 .\" professionally.
  20 .\"
  21 .\" Formatted or processed versions of this manual, if unaccompanied by
  22 .\" the source, must acknowledge the copyright and authors of this work.
  23 .\" %%%LICENSE_END
  24 .\"
  25 .\" 1999-11-10: Merged text taken from the page contributed by
  26 .\" Reed H. Petty (rhp@draper.net)
  27 .\"
  28 .TH VFORK 2 2016-03-15 "Linux" "Linux Programmer's Manual"
  29 .SH NAME
  30 vfork \- create a child process and block parent
  31 .SH SYNOPSIS
  32 .B #include <sys/types.h>
  33 .br
  34 .B #include <unistd.h>
  35 .sp
  36 .B pid_t vfork(void);
  37 .sp
  38 .in -4n
  39 Feature Test Macro Requirements for glibc (see
  40 .BR feature_test_macros (7)):
  41 .in
  42 .sp
  43 .BR vfork ():
  44 .ad l
  45 .RS 4
  46 .PD 0
  47 .TP 4
  48 Since glibc 2.12:
  49 .nf
  50 (_XOPEN_SOURCE\ >=\ 500) && ! (_POSIX_C_SOURCE\ >=\ 200809L)
  51     || /* Since glibc 2.19: */ _DEFAULT_SOURCE
  52     || /* Glibc versions <= 2.19: */ _BSD_SOURCE
  53 .TP 4
  54 .fi
  55 Before glibc 2.12:
  56 _BSD_SOURCE || _XOPEN_SOURCE\ >=\ 500
  57 .\"     || _XOPEN_SOURCE\ &&\ _XOPEN_SOURCE_EXTENDED
  58 .PD
  59 .RE
  60 .ad b
  61 .SH DESCRIPTION
  62 .SS Standard description
  63 (From POSIX.1)
  64 The
  65 .BR vfork ()
  66 function has the same effect as
  67 .BR fork (2),
  68 except that the behavior is undefined if the process created by
  69 .BR vfork ()
  70 either modifies any data other than a variable of type
  71 .I pid_t
  72 used to store the return value from
  73 .BR vfork (),
  74 or returns from the function in which
  75 .BR vfork ()
  76 was called, or calls any other function before successfully calling
  77 .BR _exit (2)
  78 or one of the
  79 .BR exec (3)
  80 family of functions.
  81 .SS Linux description
  82 .BR vfork (),
  83 just like
  84 .BR fork (2),
  85 creates a child process of the calling process.
  86 For details and return value and errors, see
  87 .BR fork (2).
  88 .PP
  89 .BR vfork ()
  90 is a special case of
  91 .BR clone (2).
  92 It is used to create new processes without copying the page tables of
  93 the parent process.
  94 It may be useful in performance-sensitive applications
  95 where a child is created which then immediately issues an
  96 .BR execve (2).
  97 .PP
  98 .BR vfork ()
  99 differs from
 100 .BR fork (2)
 101 in that the calling thread is suspended until the child terminates
 102 (either normally,
 103 by calling
 104 .BR _exit (2),
 105 or abnormally, after delivery of a fatal signal),
 106 or it makes a call to
 107 .BR execve (2).
 108 Until that point, the child shares all memory with its parent,
 109 including the stack.
 110 The child must not return from the current function or call
 111 .BR exit (3),
 112 but may call
 113 .BR _exit (2).
 114
 115 As with
 116 .BR fork (2),
 117 the child process created by
 118 .BR vfork ()
 119 inherits copies of various of the caller's process attributes
 120 (e.g., file descriptors, signal dispositions, and current working directory);
 121 the
 122 .BR vfork ()
 123 call differs only in the treatment of the virtual address space,
 124 as described above.
 125
 126 Signals sent to the parent
 127 arrive after the child releases the parent's memory
 128 (i.e., after the child terminates
 129 or calls
 130 .BR execve (2)).
 131 .SS Historic description
 132 Under Linux,
 133 .BR fork (2)
 134 is implemented using copy-on-write pages, so the only penalty incurred by
 135 .BR fork (2)
 136 is the time and memory required to duplicate the parent's page tables,
 137 and to create a unique task structure for the child.
 138 However, in the bad old days a
 139 .BR fork (2)
 140 would require making a complete copy of the caller's data space,
 141 often needlessly, since usually immediately afterward an
 142 .BR exec (3)
 143 is done.
 144 Thus, for greater efficiency, BSD introduced the
 145 .BR vfork ()
 146 system call, which did not fully copy the address space of
 147 the parent process, but borrowed the parent's memory and thread
 148 of control until a call to
 149 .BR execve (2)
 150 or an exit occurred.
 151 The parent process was suspended while the
 152 child was using its resources.
 153 The use of
 154 .BR vfork ()
 155 was tricky: for example, not modifying data
 156 in the parent process depended on knowing which variables were
 157 held in a register.
 158 .SH CONFORMING TO
 159 4.3BSD; POSIX.1-2001 (but marked OBSOLETE).
 160 POSIX.1-2008 removes the specification of
 161 .BR vfork ().
 162
 163 The requirements put on
 164 .BR vfork ()
 165 by the standards are weaker than those put on
 166 .BR fork (2),
 167 so an implementation where the two are synonymous is compliant.
 168 In particular, the programmer cannot rely on the parent
 169 remaining blocked until the child either terminates or calls
 170 .BR execve (2),
 171 and cannot rely on any specific behavior with respect to shared memory.
 172 .\" In AIXv3.1 vfork is equivalent to fork.
 173 .SH NOTES
 174 .PP
 175 Some consider the semantics of
 176 .BR vfork ()
 177 to be an architectural blemish, and the 4.2BSD man page stated:
 178 "This system call will be eliminated when proper system sharing mechanisms
 179 are implemented.
 180 Users should not depend on the memory sharing semantics of
 181 .BR vfork ()
 182 as it will, in that case, be made synonymous to
 183 .BR fork (2).\c
 184 "
 185 However, even though modern memory management hardware
 186 has decreased the performance difference between
 187 .BR fork (2)
 188 and
 189 .BR vfork (),
 190 there are various reasons why Linux and other systems have retained
 191 .BR vfork ():
 192 .IP * 3
 193 Some performance-critical applications require the small performance
 194 advantage conferred by
 195 .BR vfork ().
 196 .IP *
 197 .BR vfork ()
 198 can be implemented on systems that lack a memory-management unit (MMU), but
 199 .BR fork (2)
 200 can't be implemented on such systems.
 201 (POSIX.1-2008 removed
 202 .BR vfork ()
 203 from the standard; the POSIX rationale for the
 204 .BR posix_spawn (3)
 205 function notes that that function,
 206 which provides functionality equivalent to
 207 .BR fork (2)+ exec (3),
 208 is designed to be implementable on systems that lack an MMU.)
 209 .\" http://stackoverflow.com/questions/4259629/what-is-the-difference-between-fork-and-vfork
 210 .\" http://developers.sun.com/solaris/articles/subprocess/subprocess.html
 211 .\" http://mailman.uclinux.org/pipermail/uclinux-dev/2009-April/000684.html
 212 .SS Linux notes
 213 Fork handlers established using
 214 .BR pthread_atfork (3)
 215 are not called when a multithreaded program employing
 216 the NPTL threading library calls
 217 .BR vfork ().
 218 Fork handlers are called in this case in a program using the
 219 LinuxThreads threading library.
 220 (See
 221 .BR pthreads (7)
 222 for a description of Linux threading libraries.)
 223
 224 A call to
 225 .BR vfork ()
 226 is equivalent to calling
 227 .BR clone (2)
 228 with
 229 .I flags
 230 specified as:
 231
 232      CLONE_VM | CLONE_VFORK | SIGCHLD
 233 .SS History
 234 The
 235 .BR vfork ()
 236 system call appeared in 3.0BSD.
 237 .\" In the release notes for 4.2BSD Sam Leffler wrote: `vfork: Is still
 238 .\" present, but definitely on its way out'.
 239 In 4.4BSD it was made synonymous to
 240 .BR fork (2)
 241 but NetBSD introduced it again,
 242 cf.
 243 .UR http://www.netbsd.org\:/Documentation\:/kernel\:/vfork.html
 244 .UE .
 245 In Linux, it has been equivalent to
 246 .BR fork (2)
 247 until 2.2.0-pre6 or so.
 248 Since 2.2.0-pre9 (on i386, somewhat later on
 249 other architectures) it is an independent system call.
 250 Support was added in glibc 2.0.112.
 251 .SH BUGS
 252 .PP
 253 Details of the signal handling are obscure and differ between systems.
 254 The BSD man page states:
 255 "To avoid a possible deadlock situation, processes that are children
 256 in the middle of a
 257 .BR vfork ()
 258 are never sent
 259 .B SIGTTOU
 260 or
 261 .B SIGTTIN
 262 signals; rather, output or
 263 .IR ioctl s
 264 are allowed and input attempts result in an end-of-file indication."
 265 .\"
 266 .\" As far as I can tell, the following is not true in 2.6.19:
 267 .\" Currently (Linux 2.3.25),
 268 .\" .BR strace (1)
 269 .\" cannot follow
 270 .\" .BR vfork ()
 271 .\" and requires a kernel patch.
 272 .SH SEE ALSO
 273 .BR clone (2),
 274 .BR execve (2),
 275 .BR fork (2),
 276 .BR unshare (2),
 277 .BR wait (2)