xref: /netbsd-src/lib/libc/sys/execve.2 (revision 2f85358c65c9c151f0643ed2dc431dbe549ce62b)
1.\"	$NetBSD: execve.2,v 1.47 2024/04/28 23:10:26 uwe Exp $
2.\"
3.\" Copyright (c) 1980, 1991, 1993
4.\"	The Regents of the University of California.  All rights reserved.
5.\"
6.\" Redistribution and use in source and binary forms, with or without
7.\" modification, are permitted provided that the following conditions
8.\" are met:
9.\" 1. Redistributions of source code must retain the above copyright
10.\"    notice, this list of conditions and the following disclaimer.
11.\" 2. Redistributions in binary form must reproduce the above copyright
12.\"    notice, this list of conditions and the following disclaimer in the
13.\"    documentation and/or other materials provided with the distribution.
14.\" 3. Neither the name of the University nor the names of its contributors
15.\"    may be used to endorse or promote products derived from this software
16.\"    without specific prior written permission.
17.\"
18.\" THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
19.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21.\" ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
22.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28.\" SUCH DAMAGE.
29.\"
30.\"     @(#)execve.2	8.5 (Berkeley) 6/1/94
31.\"
32.Dd September 16, 2019
33.Dt EXECVE 2
34.Os
35.Sh NAME
36.Nm execve ,
37.Nm fexecve
38.Nd execute a file
39.Sh LIBRARY
40.Lb libc
41.Sh SYNOPSIS
42.In unistd.h
43.Ft int
44.Fn execve "const char *path" "char *const argv[]" "char *const envp[]"
45.Ft int
46.Fn fexecve "int fd" "char *const argv[]" "char *const envp[]"
47.Sh DESCRIPTION
48The
49.Fn execve
50system call
51transforms the calling process into a new process.
52The new process is constructed from an ordinary file,
53whose name is pointed to by
54.Fa path ,
55called the
56.Em new process file .
57The
58.Fn fexecve
59system call is equivalent to
60.Fn execve
61except that the file to be executed is determined by the file
62descriptor
63.Fa fd
64instead of a
65.Fa path .
66This file is either an executable object file,
67or a file of data for an interpreter.
68An executable object file consists of an identifying header,
69followed by pages of data representing the initial program (text)
70and initialized data pages.
71Additional pages may be specified
72by the header to be initialized with zero data; see
73.Xr elf 5
74and
75.Xr a.out 5 .
76.Pp
77An interpreter file begins with a line of the form:
78.Bd -ragged -offset indent
79.Ic \&#! Ns Ar interpreter Op Ar arg
80.Ed
81.Pp
82When an interpreter file is
83.Nm Ap d ,
84the system actually
85.Nm Ap s
86the specified
87.Ar interpreter .
88If the optional
89.Ar arg
90is specified, it becomes the first argument to the
91.Ar interpreter ,
92and the name of the originally
93.Nm Ap d
94file becomes the second argument;
95otherwise, the name of the originally
96.Nm Ap d
97file becomes the first argument.
98The original arguments are shifted over to become the subsequent arguments.
99The zeroth argument, normally the name of the
100.Nm Ap d
101file, is left unchanged.
102The interpreter named by
103.Ar interpreter
104must not itself be an interpreter file.
105(See
106.Xr script 7
107for a detailed discussion of interpreter file execution.)
108.Pp
109The argument
110.Fa argv
111is a pointer to a null-terminated array of
112character pointers to null-terminated character strings.
113These strings construct the argument list to be made available to the new
114process.
115At least one argument must be present in
116the array; by custom, the first element should be
117the name of the executed program (for example, the last component of
118.Fa path ) .
119.Pp
120The argument
121.Fa envp
122is also a pointer to a null-terminated array of
123character pointers to null-terminated strings.
124A pointer to this array is normally stored in the global variable
125.Va environ .
126These strings pass information to the
127new process that is not directly an argument to the command
128.Pq see Xr environ 7 .
129.Pp
130File descriptors open in the calling process image remain open in
131the new process image, except for those for which the close-on-exec
132flag is set (see
133.Xr close 2
134and
135.Xr fcntl 2 ) .
136Descriptors that remain open are unaffected by
137.Fn execve .
138.Pp
139In the case of a new setuid or setgid executable being executed, if
140file descriptors 0, 1, or 2
141.Po
142representing
143.Em stdin , stdout ,
144and
145.Em stderr
146.Pc
147are currently unallocated, these descriptors will be opened to point to
148some system file like
149.Pa /dev/null .
150The intent is to ensure these descriptors are not unallocated, since
151many libraries make assumptions about the use of these three file descriptors.
152.Pp
153Signals set to be ignored in the calling process are set to be ignored in
154the new process.
155Signals which are set to be caught in the calling process image
156are set to default action in the new process image.
157Blocked signals remain blocked regardless of changes to the signal action.
158The signal stack is reset to be undefined (see
159.Xr sigaction 2
160for more information).
161.Pp
162If the set-user-ID mode bit of the new process image file is set
163.Pq see Xr chmod 2 ,
164the effective user ID of the new process image is set to the owner ID
165of the new process image file.
166If the set-group-ID mode bit of the new process image file is set,
167the effective group ID of the new process image is set to the group ID
168of the new process image file.
169(The effective group ID is the first element of the group list.)
170The real user ID, real group ID and
171other group IDs of the new process image remain the same as the calling
172process image.
173After any set-user-ID and set-group-ID processing,
174the effective user ID is recorded as the saved set-user-ID,
175and the effective group ID is recorded as the saved set-group-ID.
176These values may be used in changing the effective IDs later
177.Pq see Xr setuid 2 .
178The set-ID bits are not honored if the respective file system has the
179.Cm nosuid
180option enabled or if the new process file is an interpreter file.
181Syscall
182tracing is disabled if effective IDs are changed.
183.Pp
184The new process also inherits the following attributes from
185the calling process:
186.Pp
187.Bl -column "parent process ID" -offset indent -compact
188.It process ID        Ta see Xr getpid 2
189.It parent process ID Ta see Xr getppid 2
190.It process group ID  Ta see Xr getpgrp 2
191.It access groups     Ta see Xr getgroups 2
192.It working directory Ta see Xr chdir 2
193.It root directory    Ta see Xr chroot 2
194.It control terminal  Ta see Xr termios 4
195.It resource usages   Ta see Xr getrusage 2
196.It interval timers   Ta see Xr getitimer 2
197.It resource limits   Ta see Xr getrlimit 2
198.It file mode mask    Ta see Xr umask 2
199.It signal mask       Ta see Xr sigaction 2 , Xr sigprocmask 2
200.El
201.Pp
202When a program is executed as a result of an
203.Fn execve
204system call, it is entered as follows:
205.Bd -literal -offset indent
206main(argc, argv, envp)
207int argc;
208char **argv, **envp;
209.Ed
210.Pp
211where
212.Fa argc
213.Pq the Dq arg count
214is the number of elements in
215.Fa argv ,
216and
217.Fa argv
218points to the array of character pointers
219to the arguments themselves.
220.Pp
221The
222.Fn fexecve
223function ignores the file offset of
224.Fa fd .
225Since execute permission is checked by
226.Fn fexecve ,
227the file descriptor
228.Fa fd
229need not have been opened with the
230.Dv O_EXEC
231flag.
232However, if the file to be executed denies read permission for the process
233preparing to do the exec, the only way to provide the
234.Fa fd
235to
236.Fn fexecve
237is to use the
238.Dv O_EXEC
239flag when opening
240.Fa fd .
241Note that the file to be executed can not be open for writing.
242.Sh RETURN VALUES
243As the
244.Fn execve
245system call overlays the current process image
246with a new process image the successful call
247has no process to return to.
248If
249.Fn execve
250does return to the calling process an error has occurred; the
251return value will be \-1 and the global variable
252.Va errno
253is set to indicate the error.
254.Sh ERRORS
255The
256.Fn execve
257system call
258will fail and return to the calling process if:
259.Bl -tag -width Er
260.It Bq Er E2BIG
261The number of bytes in the new process' argument list
262is larger than the system-imposed limit.
263The default compile time limit is 262144 bytes and is specified
264in the variable
265.Dv NCARGS
266in
267.In sys/param.h
268and get be read from the
269.Xr sysctl 3
270MIB variable
271.Dv KERN_ARGMAX .
272.It Bq Er EACCES
273Search permission is denied for a component of the path prefix,
274the new process file is not an ordinary file,
275its file mode denies execute permission, or
276it is on a file system mounted with execution
277disabled
278.Po
279.Dv MNT_NOEXEC
280in
281.In sys/mount.h
282.Pc .
283.It Bq Er EAGAIN
284A
285.Xr setuid 7
286process has exceeded the current resource limit for the number of
287processes it is allowed to run concurrently.
288.It Bq Er EFAULT
289The new process file is not as long as indicated by
290the size values in its header; or
291the
292.Fa path ,
293.Fa argv ,
294or
295.Fa envp
296arguments point to an illegal address.
297.It Bq Er EIO
298An I/O error occurred while reading from the file system.
299.It Bq Er ELOOP
300Too many symbolic links were encountered in translating the pathname.
301.It Bq Er ENAMETOOLONG
302A component of a pathname exceeded
303.Brq Dv NAME_MAX
304characters, or an entire path name exceeded
305.Brq Dv PATH_MAX
306characters.
307.It Bq Er ENOENT
308The new process file does not exist, or
309the new process file is a script starting with
310.Ql #!
311and the script interpreter does not exist.
312.It Bq Er ENOEXEC
313The new process file has the appropriate access
314permission, but has an invalid magic number in its header.
315.It Bq Er ENOMEM
316The new process requires more virtual memory than
317is allowed by the imposed maximum
318.Pq Xr getrlimit 2 .
319.It Bq Er ENOTDIR
320A component of the path prefix is not a directory.
321.It Bq Er ETXTBSY
322The new process file is a pure procedure (shared text)
323file that is currently open for writing or reading by some process.
324.El
325.Pp
326In addition, the
327.Fn fexecve
328will fail and return to the calling process if:
329.Bl -tag -width Er
330.It Bq Er EBADF
331The
332.Fa fd
333argument is not a valid file descriptor open for executing.
334.El
335.Sh SEE ALSO
336.Xr _exit 2 ,
337.Xr fork 2 ,
338.Xr open 2 ,
339.Xr execl 3 ,
340.Xr exit 3 ,
341.Xr sysctl 3 ,
342.Xr a.out 5 ,
343.Xr elf 5 ,
344.\" .Xr fdescfs 5 ,
345.Xr environ 7 ,
346.Xr script 7 ,
347.Xr mount 8
348.Sh STANDARDS
349The
350.Fn execve
351system call conforms to
352.St -p1003.1-2001 .
353with the exception of reopening descriptors 0, 1, and/or 2 in certain
354circumstances.
355A future update of the Standard is expected to require this behavior,
356and it may become the default for non-privileged processes as well.
357.\" NB: update this caveat when TC1 is blessed.
358The support for executing interpreted programs is an extension.
359The
360.Fn fexecve
361system call conforms to The Open Group Extended API Set 2 specification.
362.Sh HISTORY
363The
364.Fn execve
365function call first appeared in
366.At v7 .
367The
368.Fn fexecve
369system call appeared in
370.Nx 10.0 .
371.Sh BUGS
372If a program is
373.Em setuid
374to a non-super-user, but is executed when
375the real
376.Em uid
377is
378.Dq root ,
379then the program has some of the powers of a super-user as well.
380.\" .Pp
381.\" When executing an interpreted program through
382.\" .Fn fexecve ,
383.\" kernel supplies
384.\" .Pa /dev/fd/n
385.\" as a second argument to the interpreter,
386.\" where
387.\" .Ar n
388.\" is the file descriptor passed in the
389.\" .Fa fd
390.\" argument to
391.\" .Fn fexecve .
392.\" For this construction to work correctly, the
393.\" .Xr fdescfs 5
394.\" filesystem shall be mounted on
395.\" .Pa /dev/fd .
396