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15 .\" manual page may be incorrect or out-of-date.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 2016-10-08 "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.
100 The aforementioned transformations of the effective IDs are
102 performed if any of the following is true:
104 the underlying filesystem is mounted
112 the calling process is being ptraced.
114 The effective user ID of the process is copied to the saved set-user-ID;
115 similarly, the effective group ID is copied to the saved set-group-ID.
116 This copying takes place after any effective ID changes that occur
117 because of the set-user-ID and set-group-ID mode bits.
119 The process's read UID and real GID, as well its supplementary group IDs,
120 are unchanged by a call to
123 If the executable is an a.out dynamically linked
124 binary executable containing
125 shared-library stubs, the Linux dynamic linker
127 is called at the start of execution to bring
128 needed shared objects into memory
129 and link the executable with them.
131 If the executable is a dynamically linked ELF executable, the
132 interpreter named in the PT_INTERP segment is used to load the needed
134 This interpreter is typically
135 .I /lib/ld-linux.so.2
136 for binaries linked with glibc (see
137 .BR ld-linux.so (8)).
139 All process attributes are preserved during an
141 except the following:
143 The dispositions of any signals that are being caught are
147 Any alternate signal stack is not preserved
148 .RB ( sigaltstack (2)).
150 Memory mappings are not preserved
153 Attached System\ V shared memory segments are detached
156 POSIX shared memory regions are unmapped
159 Open POSIX message queue descriptors are closed
160 .RB ( mq_overview (7)).
162 Any open POSIX named semaphores are closed
163 .RB ( sem_overview (7)).
165 POSIX timers are not preserved
166 .RB ( timer_create (2)).
168 Any open directory streams are closed
171 Memory locks are not preserved
175 Exit handlers are not preserved
179 The floating-point environment is reset to the default (see
182 The process attributes in the preceding list are all specified
184 The following Linux-specific process attributes are also
185 not preserved during an
192 unless a set-user-ID or set-group ID program is being executed,
193 in which case it is cleared.
200 (Since Linux 2.4.36 / 2.6.23)
201 If a set-user-ID or set-group-ID program is being executed,
202 then the parent death signal set by
207 The process name, as set by
211 .IR "ps\ \-o comm" ),
212 is reset to the name of the new executable file.
219 .BR capabilities (7).
221 The termination signal is reset to
226 The file descriptor table is unshared, undoing the effect of the
231 Note the following further points:
233 All threads other than the calling thread are destroyed during an
235 Mutexes, condition variables, and other pthreads objects are not preserved.
237 The equivalent of \fIsetlocale(LC_ALL, "C")\fP
238 is executed at program start-up.
240 POSIX.1 specifies that the dispositions of any signals that
241 are ignored or set to the default are left unchanged.
242 POSIX.1 specifies one exception: if
245 then an implementation may leave the disposition unchanged or
246 reset it to the default; Linux does the former.
248 Any outstanding asynchronous I/O operations are canceled
252 For the handling of capabilities during
255 .BR capabilities (7).
257 By default, file descriptors remain open across an
259 File descriptors that are marked close-on-exec are closed;
260 see the description of
264 (If a file descriptor is closed, this will cause the release
265 of all record locks obtained on the underlying file by this process.
269 POSIX.1 says that if file descriptors 0, 1, and 2 would
270 otherwise be closed after a successful
272 and the process would gain privilege because the set-user-ID or
273 set-group_ID mode bit was set on the executed file,
274 then the system may open an unspecified file for each of these
276 As a general principle, no portable program, whether privileged or not,
277 can assume that these three file descriptors will remain
280 .\" On Linux it appears that these file descriptors are
281 .\" always open after an execve(), and it looks like
282 .\" Solaris 8 and FreeBSD 6.1 are the same. -- mtk, 30 Apr 2007
283 .SS Interpreter scripts
284 An interpreter script is a text file that has execute
285 permission enabled and whose first line is of the form:
289 \fB#!\fP \fIinterpreter \fP[optional-arg]
295 must be a valid pathname for an executable file.
300 specifies an interpreter script, then
302 will be invoked with the following arguments:
306 \fIinterpreter\fP [optional-arg] \fIfilename\fP arg...
312 is the series of words pointed to by the
321 should either be absent, or be specified as a single word (i.e., it
322 should not contain white space); see NOTES below.
325 .\" commit bf2a9a39639b8b51377905397a5005f444e9a892
326 the kernel permits the interpreter of a script to itself be a script.
327 This permission is recursive, up to a limit of four recursions,
328 so that the interpreter may be a script which is interpreted by a script,
330 .SS Limits on size of arguments and environment
331 Most UNIX implementations impose some limit on the total size
332 of the command-line argument
336 strings that may be passed to a new program.
337 POSIX.1 allows an implementation to advertise this limit using the
339 constant (either defined in
341 or available at run time using the call
342 .IR "sysconf(_SC_ARG_MAX)" ).
344 On Linux prior to kernel 2.6.23, the memory used to store the
345 environment and argument strings was limited to 32 pages
346 (defined by the kernel constant
348 On architectures with a 4-kB page size,
349 this yields a maximum size of 128 kB.
351 On kernel 2.6.23 and later, most architectures support a size limit
352 derived from the soft
356 that is in force at the time of the
359 (Architectures with no memory management unit are excepted:
360 they maintain the limit that was in effect before kernel 2.6.23.)
361 This change allows programs to have a much larger
362 argument and/or environment list.
363 .\" For some background on the changes to ARG_MAX in kernels 2.6.23 and
365 .\" http://sourceware.org/bugzilla/show_bug.cgi?id=5786
366 .\" http://bugzilla.kernel.org/show_bug.cgi?id=10095
367 .\" http://thread.gmane.org/gmane.linux.kernel/646709/focus=648101,
368 .\" checked into 2.6.25 as commit a64e715fc74b1a7dcc5944f848acc38b2c4d4ee2.
369 For these architectures, the total size is limited to 1/4 of the allowed
371 (Imposing the 1/4-limit
372 ensures that the new program always has some stack space.)
373 .\" Ollie: That doesn't include the lists of pointers, though,
374 .\" so the actual usage is a bit higher (1 pointer per argument).
376 the kernel places a floor of 32 pages on this size limit,
380 applications are guaranteed to have at least as much argument and
381 environment space as was provided by Linux 2.6.23 and earlier.
382 (This guarantee was not provided in Linux 2.6.23 and 2.6.24.)
383 Additionally, the limit per string is 32 pages (the kernel constant
384 .BR MAX_ARG_STRLEN ),
385 and the maximum number of strings is 0x7FFFFFFF.
389 does not return, on error \-1 is returned, and
391 is set appropriately.
395 The total number of bytes in the environment
402 Search permission is denied on a component of the path prefix of
404 or the name of a script interpreter.
406 .BR path_resolution (7).)
409 The file or a script interpreter is not a regular file.
412 Execute permission is denied for the file or a script or ELF interpreter.
415 The filesystem is mounted
418 .BR EAGAIN " (since Linux 3.1)"
419 .\" commit 72fa59970f8698023045ab0713d66f3f4f96945c
420 Having changed its real UID using one of the
422 calls, the caller was\(emand is now still\(emabove its
426 For a more detailed explanation of this error, see NOTES.
430 or one of the pointers in the vectors
434 points outside your accessible address space.
437 An ELF executable had more than one PT_INTERP segment (i.e., tried to
438 name more than one interpreter).
441 An I/O error occurred.
444 An ELF interpreter was a directory.
447 An ELF interpreter was not in a recognized format.
450 Too many symbolic links were encountered in resolving
452 or the name of a script or ELF interpreter.
455 The maximum recursion limit was reached during recursive script
456 interpretation (see "Interpreter scripts", above).
458 .\" commit d740269867021faf4ce38a449353d2b986c34a67
459 the error produced for this case was
463 The per-process limit on the number of open file descriptors has been reached.
470 The system-wide limit on the total number of open files has been reached.
475 or a script or ELF interpreter does not exist, or a shared library
476 .\" FIXME but see http://sourceware.org/bugzilla/show_bug.cgi?id=12241
477 needed for the file or interpreter cannot be found.
480 An executable is not in a recognized format, is for the wrong
481 architecture, or has some other format error that means it cannot be
485 Insufficient kernel memory was available.
488 A component of the path prefix of
490 or a script or ELF interpreter is not a directory.
493 The filesystem is mounted
495 the user is not the superuser,
496 and the file has the set-user-ID or set-group-ID bit set.
499 The process is being traced, the user is not the superuser and the
500 file has the set-user-ID or set-group-ID bit set.
503 A "capability-dumb" applications would not obtain the full set of
504 permitted capabilities granted by the executable file.
506 .BR capabilities (7).
509 The specified executable was open for writing by one or more processes.
511 POSIX.1-2001, POSIX.1-2008, SVr4, 4.3BSD.
512 POSIX does not document the #! behavior, but it exists
513 (with some variations) on other UNIX systems.
514 .\" SVr4 documents additional error
515 .\" conditions EAGAIN, EINTR, ELIBACC, ENOLINK, EMULTIHOP; POSIX does not
516 .\" document ETXTBSY, EPERM, EFAULT, ELOOP, EIO, ENFILE, EMFILE, EINVAL,
517 .\" EISDIR or ELIBBAD error conditions.
519 Set-user-ID and set-group-ID processes can not be
522 The result of mounting a filesystem
524 varies across Linux kernel versions:
525 some will refuse execution of set-user-ID and set-group-ID
526 executables when this would
527 give the user powers she did not have already (and return
529 some will just ignore the set-user-ID and set-group-ID bits and
537 can be specified as NULL.
538 In both cases, this has the same effect as specifying the argument
539 as a pointer to a list containing a single null pointer.
540 .B "Do not take advantage of this nonstandard and nonportable misfeature!"
541 On many other UNIX systems, specifying
543 as NULL will result in an error
546 other UNIX systems treat the
548 case the same as Linux.
549 .\" e.g., EFAULT on Solaris 8 and FreeBSD 6.1; but
550 .\" HP-UX 11 is like Linux -- mtk, Apr 2007
551 .\" Bug filed 30 Apr 2007: http://bugzilla.kernel.org/show_bug.cgi?id=8408
552 .\" Bug rejected (because fix would constitute an ABI change).
555 POSIX.1 says that values returned by
557 should be invariant over the lifetime of a process.
558 However, since Linux 2.6.23, if the
560 resource limit changes, then the value reported by
563 to reflect the fact that the limit on space for holding
564 command-line arguments and environment variables has changed.
568 fails, control returns to the original executable image,
571 can then handle the error.
572 However, in (rare) cases (typically caused by resource exhaustion),
573 failure may occur past the point of no return:
574 the original executable image has been torn down,
575 but the new image could not be completely built.
576 In such cases, the kernel kills the process with a
580 .SS Interpreter scripts
581 A maximum line length of 127 characters is allowed for the first line in
582 an interpreter scripts.
586 argument of an interpreter script vary across implementations.
587 On Linux, the entire string following the
589 name is passed as a single argument to the interpreter,
590 and this string can include white space.
591 However, behavior differs on some other systems.
594 use the first white space to terminate
597 .\" e.g., FreeBSD before 6.0, but not FreeBSD 6.0 onward
598 an interpreter script can have multiple arguments,
601 are used to delimit the arguments.
603 Linux ignores the set-user-ID and set-group-ID bits on scripts.
606 .\" Some Linux versions have failed to check permissions on ELF
607 .\" interpreters. This is a security hole, because it allows users to
608 .\" open any file, such as a rewinding tape device, for reading. Some
609 .\" Linux versions have also had other security holes in
611 .\" that could be exploited for denial of service by a suitably crafted
612 .\" ELF binary. There are no known problems with 2.0.34 or 2.2.15.
613 .SS execve() and EAGAIN
614 A more detailed explanation of the
616 error that can occur (since Linux 3.1) when calling
622 error can occur when a
629 caused the real user ID of the process to change,
630 and that change caused the process to exceed its
632 resource limit (i.e., the number of processes belonging
633 to the new real UID exceeds the resource limit).
634 From Linux 2.6.0 to 3.0, this caused the
638 .\" commit 909cc4ae86f3380152a18e2a3c44523893ee11c4
639 the resource limit was not imposed on processes that
640 changed their user IDs.)
642 Since Linux 3.1, the scenario just described no longer causes the
645 because it too often led to security holes where buggy applications
646 didn't check the return status and assumed
647 that\(emif the caller had root privileges\(emthe call would always succeed.
650 calls now successfully change the real UID,
651 but the kernel sets an internal flag, named
652 .BR PF_NPROC_EXCEEDED ,
655 resource limit has been exceeded.
657 .BR PF_NPROC_EXCEEDED
658 flag is set and the resource limit is still
659 exceeded at the time of a subsequent
661 call, that call fails with the error
663 This kernel logic ensures that the
665 resource limit is still enforced for the
666 common privileged daemon workflow\(emnamely,
673 If the resource limit was not still exceeded at the time of the
676 (because other processes belonging to this real UID terminated between the
682 call succeeds and the kernel clears the
683 .BR PF_NPROC_EXCEEDED
685 The flag is also cleared if a subsequent call to
687 by this process succeeds.
689 With UNIX\ V6, the argument list of an
692 while the argument list of
695 Thus, this argument list was not directly usable in a further
698 Since UNIX\ V7, both are NULL.
701 .\" Some Linux versions have failed to check permissions on ELF
702 .\" interpreters. This is a security hole, because it allows users to
703 .\" open any file, such as a rewinding tape device, for reading. Some
704 .\" Linux versions have also had other security holes in
706 .\" that could be exploited for denial of service by a suitably crafted
707 .\" ELF binary. There are no known problems with 2.0.34 or 2.2.15.
709 The following program is designed to be execed by the second program below.
710 It just echoes its command-line arguments, one per line.
720 main(int argc, char *argv[])
724 for (j = 0; j < argc; j++)
725 printf("argv[%d]: %s\\n", j, argv[j]);
732 This program can be used to exec the program named in its command-line
744 main(int argc, char *argv[])
746 char *newargv[] = { NULL, "hello", "world", NULL };
747 char *newenviron[] = { NULL };
750 fprintf(stderr, "Usage: %s <file\-to\-exec>\\n", argv[0]);
754 newargv[0] = argv[1];
756 execve(argv[1], newargv, newenviron);
757 perror("execve"); /* execve() returns only on error */
763 We can use the second program to exec the first as follows:
767 .RB "$" " cc myecho.c \-o myecho"
768 .RB "$" " cc execve.c \-o execve"
769 .RB "$" " ./execve ./myecho"
776 We can also use these programs to demonstrate the use of a script
778 To do this we create a script whose "interpreter" is our
784 .RB "$" " cat > script"
785 .B #!./myecho script-arg
787 .RB "$" " chmod +x script"
791 We can then use our program to exec the script:
795 .RB "$" " ./execve ./script"
814 .BR path_resolution (7),