1 .\" Copyright (c) 1992 Drew Eckhardt (drew@cs.colorado.edu), March 28, 1992
2 .\" and Copyright (c) 2006 Michael Kerrisk <mtk.manpages@gmail.com>
4 .\" %%%LICENSE_START(VERBATIM)
5 .\" Permission is granted to make and distribute verbatim copies of this
6 .\" manual provided the copyright notice and this permission notice are
7 .\" preserved on all copies.
9 .\" Permission is granted to copy and distribute modified versions of this
10 .\" manual under the conditions for verbatim copying, provided that the
11 .\" entire resulting derived work is distributed under the terms of a
12 .\" permission notice identical to this one.
14 .\" Since the Linux kernel and libraries are constantly changing, this
15 .\" manual page may be incorrect or out-of-date. The author(s) assume no
16 .\" responsibility for errors or omissions, or for damages resulting from
17 .\" the use of the information contained herein. The author(s) may not
18 .\" have taken the same level of care in the production of this manual,
19 .\" which is licensed free of charge, as they might when working
22 .\" Formatted or processed versions of this manual, if unaccompanied by
23 .\" the source, must acknowledge the copyright and authors of this work.
26 .\" Modified by Michael Haardt <michael@moria.de>
27 .\" Modified 1993-07-21 by Rik Faith <faith@cs.unc.edu>
28 .\" Modified 1994-08-21 by Michael Chastain <mec@shell.portal.com>:
29 .\" Modified 1997-01-31 by Eric S. Raymond <esr@thyrsus.com>
30 .\" Modified 1999-11-12 by Urs Thuermann <urs@isnogud.escape.de>
31 .\" Modified 2004-06-23 by Michael Kerrisk <mtk.manpages@gmail.com>
32 .\" 2006-09-04 Michael Kerrisk <mtk.manpages@gmail.com>
33 .\" Added list of process attributes that are not preserved on exec().
34 .\" 2007-09-14 Ollie Wild <aaw@google.com>, mtk
35 .\" Add text describing limits on command-line arguments + environment
37 .TH EXECVE 2 2017-03-13 "Linux" "Linux Programmer's Manual"
39 execve \- execute program
41 .B #include <unistd.h>
43 .BI "int execve(const char *" filename ", char *const " argv "[], "
45 .BI " char *const " envp []);
48 executes the program pointed to by \fIfilename\fP.
49 \fIfilename\fP must be either a binary executable, or a script
50 starting with a line of the form:
54 \fB#!\fP \fIinterpreter \fP[optional-arg]
58 For details of the latter case, see "Interpreter scripts" below.
60 \fIargv\fP is an array of argument strings passed to the new program.
61 By convention, the first of these strings (i.e.,
63 should contain the filename associated with the file being executed.
64 \fIenvp\fP is an array of strings, conventionally of the form
65 \fBkey=value\fP, which are passed as environment to the new program.
66 The \fIargv\fP and \fIenvp\fP arrays must each include a null pointer
67 at the end of the array.
69 The argument vector and environment can be accessed by the
70 called program's main function, when it is defined as:
74 int main(int argc, char *argv[], char *envp[])
78 Note, however, that the use of a third argument to the main function
79 is not specified in POSIX.1;
81 the environment should be accessed via the external variable
85 does not return on success, and the text, initialized data,
86 uninitialized data (bss), and stack of the calling process are overwritten
87 according to the contents of the newly loaded program.
89 If the current program is being ptraced, a \fBSIGTRAP\fP signal is sent to it
93 If the set-user-ID bit is set on the program file pointed to by
95 then the effective user ID of the calling process is changed
96 to that of the owner of the program file.
97 Similarly, when the set-group-ID
98 bit of the program file is set the effective group ID of the calling
99 process is set to the group of the program file.
101 The aforementioned transformations of the effective IDs are
103 performed (i.e., the set-user-ID and set-group-ID bits are ignored)
104 if any of the following is true:
108 attribute is set for the calling thread (see
111 the underlying filesystem is mounted
119 the calling process is being ptraced.
121 The capabilities of the program file (see
122 .BR capabilities (7))
123 are also ignored if any of the above are true.
125 The effective user ID of the process is copied to the saved set-user-ID;
126 similarly, the effective group ID is copied to the saved set-group-ID.
127 This copying takes place after any effective ID changes that occur
128 because of the set-user-ID and set-group-ID mode bits.
130 The process's real UID and real GID, as well its supplementary group IDs,
131 are unchanged by a call to
134 If the executable is an a.out dynamically linked
135 binary executable containing
136 shared-library stubs, the Linux dynamic linker
138 is called at the start of execution to bring
139 needed shared objects into memory
140 and link the executable with them.
142 If the executable is a dynamically linked ELF executable, the
143 interpreter named in the PT_INTERP segment is used to load the needed
145 This interpreter is typically
146 .I /lib/ld-linux.so.2
147 for binaries linked with glibc (see
148 .BR ld-linux.so (8)).
150 All process attributes are preserved during an
152 except the following:
154 The dispositions of any signals that are being caught are
158 Any alternate signal stack is not preserved
159 .RB ( sigaltstack (2)).
161 Memory mappings are not preserved
164 Attached System\ V shared memory segments are detached
167 POSIX shared memory regions are unmapped
170 Open POSIX message queue descriptors are closed
171 .RB ( mq_overview (7)).
173 Any open POSIX named semaphores are closed
174 .RB ( sem_overview (7)).
176 POSIX timers are not preserved
177 .RB ( timer_create (2)).
179 Any open directory streams are closed
182 Memory locks are not preserved
186 Exit handlers are not preserved
190 The floating-point environment is reset to the default (see
193 The process attributes in the preceding list are all specified
195 The following Linux-specific process attributes are also
196 not preserved during an
203 unless a set-user-ID or set-group ID program is being executed,
204 in which case it is cleared.
211 (Since Linux 2.4.36 / 2.6.23)
212 If a set-user-ID or set-group-ID program is being executed,
213 then the parent death signal set by
218 The process name, as set by
222 .IR "ps\ \-o comm" ),
223 is reset to the name of the new executable file.
230 .BR capabilities (7).
232 The termination signal is reset to
237 The file descriptor table is unshared, undoing the effect of the
242 Note the following further points:
244 All threads other than the calling thread are destroyed during an
246 Mutexes, condition variables, and other pthreads objects are not preserved.
248 The equivalent of \fIsetlocale(LC_ALL, "C")\fP
249 is executed at program start-up.
251 POSIX.1 specifies that the dispositions of any signals that
252 are ignored or set to the default are left unchanged.
253 POSIX.1 specifies one exception: if
256 then an implementation may leave the disposition unchanged or
257 reset it to the default; Linux does the former.
259 Any outstanding asynchronous I/O operations are canceled
263 For the handling of capabilities during
266 .BR capabilities (7).
268 By default, file descriptors remain open across an
270 File descriptors that are marked close-on-exec are closed;
271 see the description of
275 (If a file descriptor is closed, this will cause the release
276 of all record locks obtained on the underlying file by this process.
280 POSIX.1 says that if file descriptors 0, 1, and 2 would
281 otherwise be closed after a successful
283 and the process would gain privilege because the set-user-ID or
284 set-group_ID mode bit was set on the executed file,
285 then the system may open an unspecified file for each of these
287 As a general principle, no portable program, whether privileged or not,
288 can assume that these three file descriptors will remain
291 .\" On Linux it appears that these file descriptors are
292 .\" always open after an execve(), and it looks like
293 .\" Solaris 8 and FreeBSD 6.1 are the same. -- mtk, 30 Apr 2007
294 .SS Interpreter scripts
295 An interpreter script is a text file that has execute
296 permission enabled and whose first line is of the form:
300 \fB#!\fP \fIinterpreter \fP[optional-arg]
306 must be a valid pathname for an executable file.
311 specifies an interpreter script, then
313 will be invoked with the following arguments:
317 \fIinterpreter\fP [optional-arg] \fIfilename\fP arg...
323 is the series of words pointed to by the
332 should either be absent, or be specified as a single word (i.e., it
333 should not contain white space); see NOTES below.
336 .\" commit bf2a9a39639b8b51377905397a5005f444e9a892
337 the kernel permits the interpreter of a script to itself be a script.
338 This permission is recursive, up to a limit of four recursions,
339 so that the interpreter may be a script which is interpreted by a script,
341 .SS Limits on size of arguments and environment
342 Most UNIX implementations impose some limit on the total size
343 of the command-line argument
347 strings that may be passed to a new program.
348 POSIX.1 allows an implementation to advertise this limit using the
350 constant (either defined in
352 or available at run time using the call
353 .IR "sysconf(_SC_ARG_MAX)" ).
355 On Linux prior to kernel 2.6.23, the memory used to store the
356 environment and argument strings was limited to 32 pages
357 (defined by the kernel constant
359 On architectures with a 4-kB page size,
360 this yields a maximum size of 128\ kB.
362 On kernel 2.6.23 and later, most architectures support a size limit
363 derived from the soft
367 that is in force at the time of the
370 (Architectures with no memory management unit are excepted:
371 they maintain the limit that was in effect before kernel 2.6.23.)
372 This change allows programs to have a much larger
373 argument and/or environment list.
374 .\" For some background on the changes to ARG_MAX in kernels 2.6.23 and
376 .\" http://sourceware.org/bugzilla/show_bug.cgi?id=5786
377 .\" http://bugzilla.kernel.org/show_bug.cgi?id=10095
378 .\" http://thread.gmane.org/gmane.linux.kernel/646709/focus=648101,
379 .\" checked into 2.6.25 as commit a64e715fc74b1a7dcc5944f848acc38b2c4d4ee2.
380 For these architectures, the total size is limited to 1/4 of the allowed
382 (Imposing the 1/4-limit
383 ensures that the new program always has some stack space.)
384 .\" Ollie: That doesn't include the lists of pointers, though,
385 .\" so the actual usage is a bit higher (1 pointer per argument).
387 the kernel places a floor of 32 pages on this size limit,
391 applications are guaranteed to have at least as much argument and
392 environment space as was provided by Linux 2.6.23 and earlier.
393 (This guarantee was not provided in Linux 2.6.23 and 2.6.24.)
394 Additionally, the limit per string is 32 pages (the kernel constant
395 .BR MAX_ARG_STRLEN ),
396 and the maximum number of strings is 0x7FFFFFFF.
400 does not return, on error \-1 is returned, and
402 is set appropriately.
406 The total number of bytes in the environment
413 Search permission is denied on a component of the path prefix of
415 or the name of a script interpreter.
417 .BR path_resolution (7).)
420 The file or a script interpreter is not a regular file.
423 Execute permission is denied for the file or a script or ELF interpreter.
426 The filesystem is mounted
429 .BR EAGAIN " (since Linux 3.1)"
430 .\" commit 72fa59970f8698023045ab0713d66f3f4f96945c
431 Having changed its real UID using one of the
433 calls, the caller was\(emand is now still\(emabove its
437 For a more detailed explanation of this error, see NOTES.
441 or one of the pointers in the vectors
445 points outside your accessible address space.
448 An ELF executable had more than one PT_INTERP segment (i.e., tried to
449 name more than one interpreter).
452 An I/O error occurred.
455 An ELF interpreter was a directory.
458 An ELF interpreter was not in a recognized format.
461 Too many symbolic links were encountered in resolving
463 or the name of a script or ELF interpreter.
466 The maximum recursion limit was reached during recursive script
467 interpretation (see "Interpreter scripts", above).
469 .\" commit d740269867021faf4ce38a449353d2b986c34a67
470 the error produced for this case was
474 The per-process limit on the number of open file descriptors has been reached.
481 The system-wide limit on the total number of open files has been reached.
486 or a script or ELF interpreter does not exist, or a shared library
487 .\" FIXME but see http://sourceware.org/bugzilla/show_bug.cgi?id=12241
488 needed for the file or interpreter cannot be found.
491 An executable is not in a recognized format, is for the wrong
492 architecture, or has some other format error that means it cannot be
496 Insufficient kernel memory was available.
499 A component of the path prefix of
501 or a script or ELF interpreter is not a directory.
504 The filesystem is mounted
506 the user is not the superuser,
507 and the file has the set-user-ID or set-group-ID bit set.
510 The process is being traced, the user is not the superuser and the
511 file has the set-user-ID or set-group-ID bit set.
514 A "capability-dumb" applications would not obtain the full set of
515 permitted capabilities granted by the executable file.
517 .BR capabilities (7).
520 The specified executable was open for writing by one or more processes.
522 POSIX.1-2001, POSIX.1-2008, SVr4, 4.3BSD.
523 POSIX does not document the #! behavior, but it exists
524 (with some variations) on other UNIX systems.
525 .\" SVr4 documents additional error
526 .\" conditions EAGAIN, EINTR, ELIBACC, ENOLINK, EMULTIHOP; POSIX does not
527 .\" document ETXTBSY, EPERM, EFAULT, ELOOP, EIO, ENFILE, EMFILE, EINVAL,
528 .\" EISDIR or ELIBBAD error conditions.
530 Set-user-ID and set-group-ID processes can not be
533 The result of mounting a filesystem
535 varies across Linux kernel versions:
536 some will refuse execution of set-user-ID and set-group-ID
537 executables when this would
538 give the user powers she did not have already (and return
540 some will just ignore the set-user-ID and set-group-ID bits and
548 can be specified as NULL.
549 In both cases, this has the same effect as specifying the argument
550 as a pointer to a list containing a single null pointer.
551 .B "Do not take advantage of this nonstandard and nonportable misfeature!"
552 On many other UNIX systems, specifying
554 as NULL will result in an error
557 other UNIX systems treat the
559 case the same as Linux.
560 .\" e.g., EFAULT on Solaris 8 and FreeBSD 6.1; but
561 .\" HP-UX 11 is like Linux -- mtk, Apr 2007
562 .\" Bug filed 30 Apr 2007: http://bugzilla.kernel.org/show_bug.cgi?id=8408
563 .\" Bug rejected (because fix would constitute an ABI change).
566 POSIX.1 says that values returned by
568 should be invariant over the lifetime of a process.
569 However, since Linux 2.6.23, if the
571 resource limit changes, then the value reported by
574 to reflect the fact that the limit on space for holding
575 command-line arguments and environment variables has changed.
579 fails, control returns to the original executable image,
582 can then handle the error.
583 However, in (rare) cases (typically caused by resource exhaustion),
584 failure may occur past the point of no return:
585 the original executable image has been torn down,
586 but the new image could not be completely built.
587 In such cases, the kernel kills the process with a
591 .SS Interpreter scripts
592 A maximum line length of 127 characters is allowed for the first line in
593 an interpreter script.
597 argument of an interpreter script vary across implementations.
598 On Linux, the entire string following the
600 name is passed as a single argument to the interpreter,
601 and this string can include white space.
602 However, behavior differs on some other systems.
605 use the first white space to terminate
608 .\" e.g., FreeBSD before 6.0, but not FreeBSD 6.0 onward
609 an interpreter script can have multiple arguments,
612 are used to delimit the arguments.
614 Linux ignores the set-user-ID and set-group-ID bits on scripts.
617 .\" Some Linux versions have failed to check permissions on ELF
618 .\" interpreters. This is a security hole, because it allows users to
619 .\" open any file, such as a rewinding tape device, for reading. Some
620 .\" Linux versions have also had other security holes in
622 .\" that could be exploited for denial of service by a suitably crafted
623 .\" ELF binary. There are no known problems with 2.0.34 or 2.2.15.
624 .SS execve() and EAGAIN
625 A more detailed explanation of the
627 error that can occur (since Linux 3.1) when calling
633 error can occur when a
640 caused the real user ID of the process to change,
641 and that change caused the process to exceed its
643 resource limit (i.e., the number of processes belonging
644 to the new real UID exceeds the resource limit).
645 From Linux 2.6.0 to 3.0, this caused the
649 .\" commit 909cc4ae86f3380152a18e2a3c44523893ee11c4
650 the resource limit was not imposed on processes that
651 changed their user IDs.)
653 Since Linux 3.1, the scenario just described no longer causes the
656 because it too often led to security holes where buggy applications
657 didn't check the return status and assumed
658 that\(emif the caller had root privileges\(emthe call would always succeed.
661 calls now successfully change the real UID,
662 but the kernel sets an internal flag, named
663 .BR PF_NPROC_EXCEEDED ,
666 resource limit has been exceeded.
668 .BR PF_NPROC_EXCEEDED
669 flag is set and the resource limit is still
670 exceeded at the time of a subsequent
672 call, that call fails with the error
674 This kernel logic ensures that the
676 resource limit is still enforced for the
677 common privileged daemon workflow\(emnamely,
684 If the resource limit was not still exceeded at the time of the
687 (because other processes belonging to this real UID terminated between the
693 call succeeds and the kernel clears the
694 .BR PF_NPROC_EXCEEDED
696 The flag is also cleared if a subsequent call to
698 by this process succeeds.
700 With UNIX\ V6, the argument list of an
703 while the argument list of
706 Thus, this argument list was not directly usable in a further
709 Since UNIX\ V7, both are NULL.
712 .\" Some Linux versions have failed to check permissions on ELF
713 .\" interpreters. This is a security hole, because it allows users to
714 .\" open any file, such as a rewinding tape device, for reading. Some
715 .\" Linux versions have also had other security holes in
717 .\" that could be exploited for denial of service by a suitably crafted
718 .\" ELF binary. There are no known problems with 2.0.34 or 2.2.15.
720 The following program is designed to be execed by the second program below.
721 It just echoes its command-line arguments, one per line.
731 main(int argc, char *argv[])
735 for (j = 0; j < argc; j++)
736 printf("argv[%d]: %s\\n", j, argv[j]);
743 This program can be used to exec the program named in its command-line
755 main(int argc, char *argv[])
757 char *newargv[] = { NULL, "hello", "world", NULL };
758 char *newenviron[] = { NULL };
761 fprintf(stderr, "Usage: %s <file\-to\-exec>\\n", argv[0]);
765 newargv[0] = argv[1];
767 execve(argv[1], newargv, newenviron);
768 perror("execve"); /* execve() returns only on error */
774 We can use the second program to exec the first as follows:
778 .RB "$" " cc myecho.c \-o myecho"
779 .RB "$" " cc execve.c \-o execve"
780 .RB "$" " ./execve ./myecho"
787 We can also use these programs to demonstrate the use of a script
789 To do this we create a script whose "interpreter" is our
795 .RB "$" " cat > script"
796 .B #!./myecho script-arg
798 .RB "$" " chmod +x script"
802 We can then use our program to exec the script:
806 .RB "$" " ./execve ./script"
825 .BR path_resolution (7),