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15 .\" manual page may be incorrect or out-of-date.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 2019-05-09 "Linux" "Linux Programmer's Manual"
39 execve \- execute program
41 .B #include <unistd.h>
43 .BI "int execve(const char *" pathname ", char *const " argv "[], "
45 .BI " char *const " envp []);
48 executes the program referred to by \fIpathname\fP.
49 This causes the program that is currently being run by the calling process
50 to be replaced with a new program, with newly initialized stack, heap,
51 and (initialized and uninitialized) data segments.
53 \fIpathname\fP must be either a binary executable, or a script
54 starting with a line of the form:
58 \fB#!\fP\fIinterpreter \fP[optional-arg]
62 For details of the latter case, see "Interpreter scripts" below.
64 \fIargv\fP is an array of argument strings passed to the new program.
65 By convention, the first of these strings (i.e.,
67 should contain the filename associated with the file being executed.
68 \fIenvp\fP is an array of strings, conventionally of the form
69 \fBkey=value\fP, which are passed as environment to the new program.
70 The \fIargv\fP and \fIenvp\fP arrays must each include a null pointer
71 at the end of the array.
73 The argument vector and environment can be accessed by the
74 called program's main function, when it is defined as:
78 int main(int argc, char *argv[], char *envp[])
82 Note, however, that the use of a third argument to the main function
83 is not specified in POSIX.1;
85 the environment should be accessed via the external variable
89 does not return on success, and the text, initialized data,
90 uninitialized data (bss), and stack of the calling process are overwritten
91 according to the contents of the newly loaded program.
93 If the current program is being ptraced, a \fBSIGTRAP\fP signal is sent to it
97 If the set-user-ID bit is set on the program file referred to by
99 then the effective user ID of the calling process is changed
100 to that of the owner of the program file.
101 Similarly, when the set-group-ID
102 bit of the program file is set the effective group ID of the calling
103 process is set to the group of the program file.
105 The aforementioned transformations of the effective IDs are
107 performed (i.e., the set-user-ID and set-group-ID bits are ignored)
108 if any of the following is true:
112 attribute is set for the calling thread (see
115 the underlying filesystem is mounted
123 the calling process is being ptraced.
125 The capabilities of the program file (see
126 .BR capabilities (7))
127 are also ignored if any of the above are true.
129 The effective user ID of the process is copied to the saved set-user-ID;
130 similarly, the effective group ID is copied to the saved set-group-ID.
131 This copying takes place after any effective ID changes that occur
132 because of the set-user-ID and set-group-ID mode bits.
134 The process's real UID and real GID, as well its supplementary group IDs,
135 are unchanged by a call to
138 If the executable is an a.out dynamically linked
139 binary executable containing
140 shared-library stubs, the Linux dynamic linker
142 is called at the start of execution to bring
143 needed shared objects into memory
144 and link the executable with them.
146 If the executable is a dynamically linked ELF executable, the
147 interpreter named in the PT_INTERP segment is used to load the needed
149 This interpreter is typically
150 .I /lib/ld-linux.so.2
151 for binaries linked with glibc (see
152 .BR ld-linux.so (8)).
154 All process attributes are preserved during an
156 except the following:
158 The dispositions of any signals that are being caught are
162 Any alternate signal stack is not preserved
163 .RB ( sigaltstack (2)).
165 Memory mappings are not preserved
168 Attached System\ V shared memory segments are detached
171 POSIX shared memory regions are unmapped
174 Open POSIX message queue descriptors are closed
175 .RB ( mq_overview (7)).
177 Any open POSIX named semaphores are closed
178 .RB ( sem_overview (7)).
180 POSIX timers are not preserved
181 .RB ( timer_create (2)).
183 Any open directory streams are closed
186 Memory locks are not preserved
190 Exit handlers are not preserved
194 The floating-point environment is reset to the default (see
197 The process attributes in the preceding list are all specified
199 The following Linux-specific process attributes are also
200 not preserved during an
207 unless a set-user-ID or set-group ID program is being executed,
208 in which case it is cleared.
215 (Since Linux 2.4.36 / 2.6.23)
216 If a set-user-ID or set-group-ID program is being executed,
217 then the parent death signal set by
222 The process name, as set by
226 .IR "ps\ \-o comm" ),
227 is reset to the name of the new executable file.
234 .BR capabilities (7).
236 The termination signal is reset to
241 The file descriptor table is unshared, undoing the effect of the
246 Note the following further points:
248 All threads other than the calling thread are destroyed during an
250 Mutexes, condition variables, and other pthreads objects are not preserved.
252 The equivalent of \fIsetlocale(LC_ALL, "C")\fP
253 is executed at program start-up.
255 POSIX.1 specifies that the dispositions of any signals that
256 are ignored or set to the default are left unchanged.
257 POSIX.1 specifies one exception: if
260 then an implementation may leave the disposition unchanged or
261 reset it to the default; Linux does the former.
263 Any outstanding asynchronous I/O operations are canceled
267 For the handling of capabilities during
270 .BR capabilities (7).
272 By default, file descriptors remain open across an
274 File descriptors that are marked close-on-exec are closed;
275 see the description of
279 (If a file descriptor is closed, this will cause the release
280 of all record locks obtained on the underlying file by this process.
284 POSIX.1 says that if file descriptors 0, 1, and 2 would
285 otherwise be closed after a successful
287 and the process would gain privilege because the set-user-ID or
288 set-group_ID mode bit was set on the executed file,
289 then the system may open an unspecified file for each of these
291 As a general principle, no portable program, whether privileged or not,
292 can assume that these three file descriptors will remain
295 .\" On Linux it appears that these file descriptors are
296 .\" always open after an execve(), and it looks like
297 .\" Solaris 8 and FreeBSD 6.1 are the same. -- mtk, 30 Apr 2007
298 .SS Interpreter scripts
299 An interpreter script is a text file that has execute
300 permission enabled and whose first line is of the form:
304 \fB#!\fP\fIinterpreter \fP[optional-arg]
310 must be a valid pathname for an executable file.
316 specifies an interpreter script, then
318 will be invoked with the following arguments:
322 \fIinterpreter\fP [optional-arg] \fIpathname\fP arg...
328 is the absolute pathname of the file specified as the first argument of
332 is the series of words pointed to by the
338 Note that there is no way to get the
340 that was passed to the
343 .\" See the P - preserve-argv[0] option.
344 .\" Documentation/admin-guide/binfmt-misc.rst
345 .\" https://www.kernel.org/doc/html/latest/admin-guide/binfmt-misc.html
349 should either be absent, or be specified as a single word (i.e., it
350 should not contain white space); see NOTES below.
353 .\" commit bf2a9a39639b8b51377905397a5005f444e9a892
354 the kernel permits the interpreter of a script to itself be a script.
355 This permission is recursive, up to a limit of four recursions,
356 so that the interpreter may be a script which is interpreted by a script,
358 .SS Limits on size of arguments and environment
359 Most UNIX implementations impose some limit on the total size
360 of the command-line argument
364 strings that may be passed to a new program.
365 POSIX.1 allows an implementation to advertise this limit using the
367 constant (either defined in
369 or available at run time using the call
370 .IR "sysconf(_SC_ARG_MAX)" ).
372 On Linux prior to kernel 2.6.23, the memory used to store the
373 environment and argument strings was limited to 32 pages
374 (defined by the kernel constant
376 On architectures with a 4-kB page size,
377 this yields a maximum size of 128\ kB.
379 On kernel 2.6.23 and later, most architectures support a size limit
380 derived from the soft
384 that is in force at the time of the
387 (Architectures with no memory management unit are excepted:
388 they maintain the limit that was in effect before kernel 2.6.23.)
389 This change allows programs to have a much larger
390 argument and/or environment list.
391 .\" For some background on the changes to ARG_MAX in kernels 2.6.23 and
393 .\" http://sourceware.org/bugzilla/show_bug.cgi?id=5786
394 .\" http://bugzilla.kernel.org/show_bug.cgi?id=10095
395 .\" http://thread.gmane.org/gmane.linux.kernel/646709/focus=648101,
396 .\" checked into 2.6.25 as commit a64e715fc74b1a7dcc5944f848acc38b2c4d4ee2.
397 For these architectures, the total size is limited to 1/4 of the allowed
399 (Imposing the 1/4-limit
400 ensures that the new program always has some stack space.)
401 .\" Ollie: That doesn't include the lists of pointers, though,
402 .\" so the actual usage is a bit higher (1 pointer per argument).
403 Additionally, the total size is limited to 3/4 of the value
404 of the kernel constant
408 the kernel also places a floor of 32 pages on this size limit,
412 applications are guaranteed to have at least as much argument and
413 environment space as was provided by Linux 2.6.23 and earlier.
414 (This guarantee was not provided in Linux 2.6.23 and 2.6.24.)
415 Additionally, the limit per string is 32 pages (the kernel constant
416 .BR MAX_ARG_STRLEN ),
417 and the maximum number of strings is 0x7FFFFFFF.
421 does not return, on error \-1 is returned, and
423 is set appropriately.
427 The total number of bytes in the environment
434 Search permission is denied on a component of the path prefix of
436 or the name of a script interpreter.
438 .BR path_resolution (7).)
441 The file or a script interpreter is not a regular file.
444 Execute permission is denied for the file or a script or ELF interpreter.
447 The filesystem is mounted
450 .BR EAGAIN " (since Linux 3.1)"
451 .\" commit 72fa59970f8698023045ab0713d66f3f4f96945c
452 Having changed its real UID using one of the
454 calls, the caller was\(emand is now still\(emabove its
458 For a more detailed explanation of this error, see NOTES.
462 or one of the pointers in the vectors
466 points outside your accessible address space.
469 An ELF executable had more than one PT_INTERP segment (i.e., tried to
470 name more than one interpreter).
473 An I/O error occurred.
476 An ELF interpreter was a directory.
479 An ELF interpreter was not in a recognized format.
482 Too many symbolic links were encountered in resolving
484 or the name of a script or ELF interpreter.
487 The maximum recursion limit was reached during recursive script
488 interpretation (see "Interpreter scripts", above).
490 .\" commit d740269867021faf4ce38a449353d2b986c34a67
491 the error produced for this case was
495 The per-process limit on the number of open file descriptors has been reached.
502 The system-wide limit on the total number of open files has been reached.
507 or a script or ELF interpreter does not exist, or a shared library
508 .\" FIXME but see http://sourceware.org/bugzilla/show_bug.cgi?id=12241
509 needed for the file or interpreter cannot be found.
512 An executable is not in a recognized format, is for the wrong
513 architecture, or has some other format error that means it cannot be
517 Insufficient kernel memory was available.
520 A component of the path prefix of
522 or a script or ELF interpreter is not a directory.
525 The filesystem is mounted
527 the user is not the superuser,
528 and the file has the set-user-ID or set-group-ID bit set.
531 The process is being traced, the user is not the superuser and the
532 file has the set-user-ID or set-group-ID bit set.
535 A "capability-dumb" applications would not obtain the full set of
536 permitted capabilities granted by the executable file.
538 .BR capabilities (7).
541 The specified executable was open for writing by one or more processes.
543 POSIX.1-2001, POSIX.1-2008, SVr4, 4.3BSD.
544 POSIX does not document the #! behavior, but it exists
545 (with some variations) on other UNIX systems.
546 .\" SVr4 documents additional error
547 .\" conditions EAGAIN, EINTR, ELIBACC, ENOLINK, EMULTIHOP; POSIX does not
548 .\" document ETXTBSY, EPERM, EFAULT, ELOOP, EIO, ENFILE, EMFILE, EINVAL,
549 .\" EISDIR or ELIBBAD error conditions.
553 (and the related functions described in
555 described as "executing a
557 process" (or similar).
558 This is a highly misleading description:
559 there is no new process;
560 many attributes of the calling process remain unchanged
561 (in particular, its PID).
564 does is arrange for an existing process (the calling process)
565 to execute a new program.
567 Set-user-ID and set-group-ID processes can not be
570 The result of mounting a filesystem
572 varies across Linux kernel versions:
573 some will refuse execution of set-user-ID and set-group-ID
574 executables when this would
575 give the user powers they did not have already (and return
577 some will just ignore the set-user-ID and set-group-ID bits and
585 can be specified as NULL.
586 In both cases, this has the same effect as specifying the argument
587 as a pointer to a list containing a single null pointer.
588 .B "Do not take advantage of this nonstandard and nonportable misfeature!"
589 On many other UNIX systems, specifying
591 as NULL will result in an error
594 other UNIX systems treat the
596 case the same as Linux.
597 .\" e.g., EFAULT on Solaris 8 and FreeBSD 6.1; but
598 .\" HP-UX 11 is like Linux -- mtk, Apr 2007
599 .\" Bug filed 30 Apr 2007: http://bugzilla.kernel.org/show_bug.cgi?id=8408
600 .\" Bug rejected (because fix would constitute an ABI change).
603 POSIX.1 says that values returned by
605 should be invariant over the lifetime of a process.
606 However, since Linux 2.6.23, if the
608 resource limit changes, then the value reported by
611 to reflect the fact that the limit on space for holding
612 command-line arguments and environment variables has changed.
616 fails, control returns to the original executable image,
619 can then handle the error.
620 However, in (rare) cases (typically caused by resource exhaustion),
621 failure may occur past the point of no return:
622 the original executable image has been torn down,
623 but the new image could not be completely built.
624 In such cases, the kernel kills the process with a
628 .SS Interpreter scripts
629 The kernel imposes a maximum length on the text that follows the
630 "#!" characters at the start of a script;
631 characters beyond the limit are ignored.
632 Before Linux 5.1, the limit is 127 characters.
634 .\" commit 6eb3c3d0a52dca337e327ae8868ca1f44a712e02
635 the limit is 255 characters.
639 argument of an interpreter script vary across implementations.
640 On Linux, the entire string following the
642 name is passed as a single argument to the interpreter,
643 and this string can include white space.
644 However, behavior differs on some other systems.
647 use the first white space to terminate
650 .\" e.g., FreeBSD before 6.0, but not FreeBSD 6.0 onward
651 an interpreter script can have multiple arguments,
654 are used to delimit the arguments.
656 Linux (like most other modern UNIX systems)
657 ignores the set-user-ID and set-group-ID bits on scripts.
660 .\" Some Linux versions have failed to check permissions on ELF
661 .\" interpreters. This is a security hole, because it allows users to
662 .\" open any file, such as a rewinding tape device, for reading. Some
663 .\" Linux versions have also had other security holes in
665 .\" that could be exploited for denial of service by a suitably crafted
666 .\" ELF binary. There are no known problems with 2.0.34 or 2.2.15.
667 .SS execve() and EAGAIN
668 A more detailed explanation of the
670 error that can occur (since Linux 3.1) when calling
676 error can occur when a
683 caused the real user ID of the process to change,
684 and that change caused the process to exceed its
686 resource limit (i.e., the number of processes belonging
687 to the new real UID exceeds the resource limit).
688 From Linux 2.6.0 to 3.0, this caused the
692 .\" commit 909cc4ae86f3380152a18e2a3c44523893ee11c4
693 the resource limit was not imposed on processes that
694 changed their user IDs.)
696 Since Linux 3.1, the scenario just described no longer causes the
699 because it too often led to security holes where buggy applications
700 didn't check the return status and assumed
701 that\(emif the caller had root privileges\(emthe call would always succeed.
704 calls now successfully change the real UID,
705 but the kernel sets an internal flag, named
706 .BR PF_NPROC_EXCEEDED ,
709 resource limit has been exceeded.
711 .BR PF_NPROC_EXCEEDED
712 flag is set and the resource limit is still
713 exceeded at the time of a subsequent
715 call, that call fails with the error
717 This kernel logic ensures that the
719 resource limit is still enforced for the
720 common privileged daemon workflow\(emnamely,
727 If the resource limit was not still exceeded at the time of the
730 (because other processes belonging to this real UID terminated between the
736 call succeeds and the kernel clears the
737 .BR PF_NPROC_EXCEEDED
739 The flag is also cleared if a subsequent call to
741 by this process succeeds.
743 With UNIX\ V6, the argument list of an
746 while the argument list of
749 Thus, this argument list was not directly usable in a further
752 Since UNIX\ V7, both are NULL.
755 .\" Some Linux versions have failed to check permissions on ELF
756 .\" interpreters. This is a security hole, because it allows users to
757 .\" open any file, such as a rewinding tape device, for reading. Some
758 .\" Linux versions have also had other security holes in
760 .\" that could be exploited for denial of service by a suitably crafted
761 .\" ELF binary. There are no known problems with 2.0.34 or 2.2.15.
763 The following program is designed to be execed by the second program below.
764 It just echoes its command-line arguments, one per line.
774 main(int argc, char *argv[])
778 for (j = 0; j < argc; j++)
779 printf("argv[%d]: %s\en", j, argv[j]);
786 This program can be used to exec the program named in its command-line
798 main(int argc, char *argv[])
800 char *newargv[] = { NULL, "hello", "world", NULL };
801 char *newenviron[] = { NULL };
804 fprintf(stderr, "Usage: %s <file\-to\-exec>\en", argv[0]);
808 newargv[0] = argv[1];
810 execve(argv[1], newargv, newenviron);
811 perror("execve"); /* execve() returns only on error */
817 We can use the second program to exec the first as follows:
821 .RB "$" " cc myecho.c \-o myecho"
822 .RB "$" " cc execve.c \-o execve"
823 .RB "$" " ./execve ./myecho"
830 We can also use these programs to demonstrate the use of a script
832 To do this we create a script whose "interpreter" is our
838 .RB "$" " cat > script"
839 .B #!./myecho script-arg
841 .RB "$" " chmod +x script"
845 We can then use our program to exec the script:
849 .RB "$" " ./execve ./script"
861 .BR get_robust_list (2),
869 .BR path_resolution (7),