xref: /freebsd/lib/libsys/rfork.2 (revision 6580f5c38dd5b01aeeaed16b370f1a12423437f0)
1.\"
2.\" This manual page is taken directly from Plan9, and modified to
3.\" describe the actual BSD implementation. Permission for
4.\" use of this page comes from Rob Pike <rob@plan9.att.com>.
5.\"
6.Dd September 25, 2019
7.Dt RFORK 2
8.Os
9.Sh NAME
10.Nm rfork
11.Nd manipulate process resources
12.Sh LIBRARY
13.Lb libc
14.Sh SYNOPSIS
15.In unistd.h
16.Ft pid_t
17.Fn rfork "int flags"
18.Sh DESCRIPTION
19Forking, vforking or rforking are the only ways new processes are created.
20The
21.Fa flags
22argument to
23.Fn rfork
24selects which resources of the
25invoking process (parent) are shared
26by the new process (child) or initialized to
27their default values.
28The resources include
29the open file descriptor table (which, when shared, permits processes
30to open and close files for other processes),
31and open files.
32The
33.Fa flags
34argument
35is either
36.Dv RFSPAWN
37or the logical OR of some subset of:
38.Bl -tag -width ".Dv RFLINUXTHPN"
39.It Dv RFPROC
40If set a new process is created; otherwise changes affect the
41current process.
42.It Dv RFNOWAIT
43If set, the child process will be dissociated from the parent.
44Upon
45exit the child will not leave a status for the parent to collect.
46See
47.Xr wait 2 .
48.It Dv RFFDG
49If set, the invoker's file descriptor table (see
50.Xr intro 2 )
51is copied; otherwise the two processes share a
52single table.
53.It Dv RFCFDG
54If set, the new process starts with a clean file descriptor table.
55Is mutually exclusive with
56.Dv RFFDG .
57.It Dv RFTHREAD
58If set, the new process shares file descriptor to process leaders table
59with its parent.
60Only applies when neither
61.Dv RFFDG
62nor
63.Dv RFCFDG
64are set.
65.It Dv RFMEM
66If set, the kernel will force sharing of the entire address space,
67typically by sharing the hardware page table directly.
68The child
69will thus inherit and share all the segments the parent process owns,
70whether they are normally shareable or not.
71The stack segment is
72not split (both the parent and child return on the same stack) and thus
73.Fn rfork
74with the RFMEM flag may not generally be called directly from high level
75languages including C.
76May be set only with
77.Dv RFPROC .
78A helper function is provided to assist with this problem and will cause
79the new process to run on the provided stack.
80See
81.Xr rfork_thread 3
82for information.
83Note that a lot of code will not run correctly in such an environment.
84.It Dv RFSIGSHARE
85If set, the kernel will force sharing the sigacts structure between the
86child and the parent.
87.It Dv RFTSIGZMB
88If set, the kernel will deliver a specified signal to the parent
89upon the child exit, instead of default SIGCHLD.
90The signal number
91.Dv signum
92is specified by oring the
93.Dv RFTSIGFLAGS(signum)
94expression into
95.Fa flags .
96Specifying signal number 0 disables signal delivery upon the child exit.
97.It Dv RFLINUXTHPN
98If set, the kernel will deliver SIGUSR1 instead of SIGCHLD upon thread
99exit for the child.
100This is intended to mimic certain Linux clone behaviour.
101.El
102.Pp
103File descriptors in a shared file descriptor table are kept
104open until either they are explicitly closed
105or all processes sharing the table exit.
106.Pp
107If
108.Dv RFSPAWN
109is passed,
110.Nm
111will use
112.Xr vfork 2
113semantics but reset all signal actions in the child to default.
114This flag is used by the
115.Xr posix_spawn 3
116implementation in libc.
117.Pp
118If
119.Dv RFPROC
120is set, the
121value returned in the parent process
122is the process id
123of the child process; the value returned in the child is zero.
124Without
125.Dv RFPROC ,
126the return value is zero.
127Process id's range from 1 to the maximum integer
128.Ft ( int )
129value.
130The
131.Fn rfork
132system call
133will sleep, if necessary, until required process resources are available.
134.Pp
135The
136.Fn fork
137system call
138can be implemented as a call to
139.Fn rfork "RFFDG | RFPROC"
140but is not for backwards compatibility.
141.Sh RETURN VALUES
142Upon successful completion,
143.Fn rfork
144returns a value
145of 0 to the child process and returns the process ID of the child
146process to the parent process.
147Otherwise, a value of -1 is returned
148to the parent process, no child process is created, and the global
149variable
150.Va errno
151is set to indicate the error.
152.Sh ERRORS
153The
154.Fn rfork
155system call
156will fail and no child process will be created if:
157.Bl -tag -width Er
158.It Bq Er EAGAIN
159The system-imposed limit on the total
160number of processes under execution would be exceeded.
161The limit is given by the
162.Xr sysctl 3
163MIB variable
164.Dv KERN_MAXPROC .
165(The limit is actually ten less than this
166except for the super user).
167.It Bq Er EAGAIN
168The user is not the super user, and
169the system-imposed limit
170on the total number of
171processes under execution by a single user would be exceeded.
172The limit is given by the
173.Xr sysctl 3
174MIB variable
175.Dv KERN_MAXPROCPERUID .
176.It Bq Er EAGAIN
177The user is not the super user, and
178the soft resource limit corresponding to the
179.Fa resource
180argument
181.Dv RLIMIT_NOFILE
182would be exceeded (see
183.Xr getrlimit 2 ) .
184.It Bq Er EINVAL
185Both the RFFDG and the RFCFDG flags were specified.
186.It Bq Er EINVAL
187Any flags not listed above were specified.
188.It Bq Er EINVAL
189An invalid signal number was specified.
190.It Bq Er ENOMEM
191There is insufficient swap space for the new process.
192.El
193.Sh SEE ALSO
194.Xr fork 2 ,
195.Xr intro 2 ,
196.Xr minherit 2 ,
197.Xr vfork 2 ,
198.Xr pthread_create 3 ,
199.Xr rfork_thread 3
200.Sh HISTORY
201The
202.Fn rfork
203function first appeared in Plan9.
204