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15 .\" manual page may be incorrect or out-of-date.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 (i.e., the set-user-ID and set-group-ID bits are ignored)
103 if any of the following is true:
107 attribute is set for the calling thread (see
110 the underlying filesystem is mounted
118 the calling process is being ptraced.
120 The effective user ID of the process is copied to the saved set-user-ID;
121 similarly, the effective group ID is copied to the saved set-group-ID.
122 This copying takes place after any effective ID changes that occur
123 because of the set-user-ID and set-group-ID mode bits.
125 The process's read UID and real GID, as well its supplementary group IDs,
126 are unchanged by a call to
129 If the executable is an a.out dynamically linked
130 binary executable containing
131 shared-library stubs, the Linux dynamic linker
133 is called at the start of execution to bring
134 needed shared objects into memory
135 and link the executable with them.
137 If the executable is a dynamically linked ELF executable, the
138 interpreter named in the PT_INTERP segment is used to load the needed
140 This interpreter is typically
141 .I /lib/ld-linux.so.2
142 for binaries linked with glibc (see
143 .BR ld-linux.so (8)).
145 All process attributes are preserved during an
147 except the following:
149 The dispositions of any signals that are being caught are
153 Any alternate signal stack is not preserved
154 .RB ( sigaltstack (2)).
156 Memory mappings are not preserved
159 Attached System\ V shared memory segments are detached
162 POSIX shared memory regions are unmapped
165 Open POSIX message queue descriptors are closed
166 .RB ( mq_overview (7)).
168 Any open POSIX named semaphores are closed
169 .RB ( sem_overview (7)).
171 POSIX timers are not preserved
172 .RB ( timer_create (2)).
174 Any open directory streams are closed
177 Memory locks are not preserved
181 Exit handlers are not preserved
185 The floating-point environment is reset to the default (see
188 The process attributes in the preceding list are all specified
190 The following Linux-specific process attributes are also
191 not preserved during an
198 unless a set-user-ID or set-group ID program is being executed,
199 in which case it is cleared.
206 (Since Linux 2.4.36 / 2.6.23)
207 If a set-user-ID or set-group-ID program is being executed,
208 then the parent death signal set by
213 The process name, as set by
217 .IR "ps\ \-o comm" ),
218 is reset to the name of the new executable file.
225 .BR capabilities (7).
227 The termination signal is reset to
232 The file descriptor table is unshared, undoing the effect of the
237 Note the following further points:
239 All threads other than the calling thread are destroyed during an
241 Mutexes, condition variables, and other pthreads objects are not preserved.
243 The equivalent of \fIsetlocale(LC_ALL, "C")\fP
244 is executed at program start-up.
246 POSIX.1 specifies that the dispositions of any signals that
247 are ignored or set to the default are left unchanged.
248 POSIX.1 specifies one exception: if
251 then an implementation may leave the disposition unchanged or
252 reset it to the default; Linux does the former.
254 Any outstanding asynchronous I/O operations are canceled
258 For the handling of capabilities during
261 .BR capabilities (7).
263 By default, file descriptors remain open across an
265 File descriptors that are marked close-on-exec are closed;
266 see the description of
270 (If a file descriptor is closed, this will cause the release
271 of all record locks obtained on the underlying file by this process.
275 POSIX.1 says that if file descriptors 0, 1, and 2 would
276 otherwise be closed after a successful
278 and the process would gain privilege because the set-user-ID or
279 set-group_ID mode bit was set on the executed file,
280 then the system may open an unspecified file for each of these
282 As a general principle, no portable program, whether privileged or not,
283 can assume that these three file descriptors will remain
286 .\" On Linux it appears that these file descriptors are
287 .\" always open after an execve(), and it looks like
288 .\" Solaris 8 and FreeBSD 6.1 are the same. -- mtk, 30 Apr 2007
289 .SS Interpreter scripts
290 An interpreter script is a text file that has execute
291 permission enabled and whose first line is of the form:
295 \fB#!\fP \fIinterpreter \fP[optional-arg]
301 must be a valid pathname for an executable file.
306 specifies an interpreter script, then
308 will be invoked with the following arguments:
312 \fIinterpreter\fP [optional-arg] \fIfilename\fP arg...
318 is the series of words pointed to by the
327 should either be absent, or be specified as a single word (i.e., it
328 should not contain white space); see NOTES below.
331 .\" commit bf2a9a39639b8b51377905397a5005f444e9a892
332 the kernel permits the interpreter of a script to itself be a script.
333 This permission is recursive, up to a limit of four recursions,
334 so that the interpreter may be a script which is interpreted by a script,
336 .SS Limits on size of arguments and environment
337 Most UNIX implementations impose some limit on the total size
338 of the command-line argument
342 strings that may be passed to a new program.
343 POSIX.1 allows an implementation to advertise this limit using the
345 constant (either defined in
347 or available at run time using the call
348 .IR "sysconf(_SC_ARG_MAX)" ).
350 On Linux prior to kernel 2.6.23, the memory used to store the
351 environment and argument strings was limited to 32 pages
352 (defined by the kernel constant
354 On architectures with a 4-kB page size,
355 this yields a maximum size of 128 kB.
357 On kernel 2.6.23 and later, most architectures support a size limit
358 derived from the soft
362 that is in force at the time of the
365 (Architectures with no memory management unit are excepted:
366 they maintain the limit that was in effect before kernel 2.6.23.)
367 This change allows programs to have a much larger
368 argument and/or environment list.
369 .\" For some background on the changes to ARG_MAX in kernels 2.6.23 and
371 .\" http://sourceware.org/bugzilla/show_bug.cgi?id=5786
372 .\" http://bugzilla.kernel.org/show_bug.cgi?id=10095
373 .\" http://thread.gmane.org/gmane.linux.kernel/646709/focus=648101,
374 .\" checked into 2.6.25 as commit a64e715fc74b1a7dcc5944f848acc38b2c4d4ee2.
375 For these architectures, the total size is limited to 1/4 of the allowed
377 (Imposing the 1/4-limit
378 ensures that the new program always has some stack space.)
379 .\" Ollie: That doesn't include the lists of pointers, though,
380 .\" so the actual usage is a bit higher (1 pointer per argument).
382 the kernel places a floor of 32 pages on this size limit,
386 applications are guaranteed to have at least as much argument and
387 environment space as was provided by Linux 2.6.23 and earlier.
388 (This guarantee was not provided in Linux 2.6.23 and 2.6.24.)
389 Additionally, the limit per string is 32 pages (the kernel constant
390 .BR MAX_ARG_STRLEN ),
391 and the maximum number of strings is 0x7FFFFFFF.
395 does not return, on error \-1 is returned, and
397 is set appropriately.
401 The total number of bytes in the environment
408 Search permission is denied on a component of the path prefix of
410 or the name of a script interpreter.
412 .BR path_resolution (7).)
415 The file or a script interpreter is not a regular file.
418 Execute permission is denied for the file or a script or ELF interpreter.
421 The filesystem is mounted
424 .BR EAGAIN " (since Linux 3.1)"
425 .\" commit 72fa59970f8698023045ab0713d66f3f4f96945c
426 Having changed its real UID using one of the
428 calls, the caller was\(emand is now still\(emabove its
432 For a more detailed explanation of this error, see NOTES.
436 or one of the pointers in the vectors
440 points outside your accessible address space.
443 An ELF executable had more than one PT_INTERP segment (i.e., tried to
444 name more than one interpreter).
447 An I/O error occurred.
450 An ELF interpreter was a directory.
453 An ELF interpreter was not in a recognized format.
456 Too many symbolic links were encountered in resolving
458 or the name of a script or ELF interpreter.
461 The maximum recursion limit was reached during recursive script
462 interpretation (see "Interpreter scripts", above).
464 .\" commit d740269867021faf4ce38a449353d2b986c34a67
465 the error produced for this case was
469 The per-process limit on the number of open file descriptors has been reached.
476 The system-wide limit on the total number of open files has been reached.
481 or a script or ELF interpreter does not exist, or a shared library
482 .\" FIXME but see http://sourceware.org/bugzilla/show_bug.cgi?id=12241
483 needed for the file or interpreter cannot be found.
486 An executable is not in a recognized format, is for the wrong
487 architecture, or has some other format error that means it cannot be
491 Insufficient kernel memory was available.
494 A component of the path prefix of
496 or a script or ELF interpreter is not a directory.
499 The filesystem is mounted
501 the user is not the superuser,
502 and the file has the set-user-ID or set-group-ID bit set.
505 The process is being traced, the user is not the superuser and the
506 file has the set-user-ID or set-group-ID bit set.
509 A "capability-dumb" applications would not obtain the full set of
510 permitted capabilities granted by the executable file.
512 .BR capabilities (7).
515 The specified executable was open for writing by one or more processes.
517 POSIX.1-2001, POSIX.1-2008, SVr4, 4.3BSD.
518 POSIX does not document the #! behavior, but it exists
519 (with some variations) on other UNIX systems.
520 .\" SVr4 documents additional error
521 .\" conditions EAGAIN, EINTR, ELIBACC, ENOLINK, EMULTIHOP; POSIX does not
522 .\" document ETXTBSY, EPERM, EFAULT, ELOOP, EIO, ENFILE, EMFILE, EINVAL,
523 .\" EISDIR or ELIBBAD error conditions.
525 Set-user-ID and set-group-ID processes can not be
528 The result of mounting a filesystem
530 varies across Linux kernel versions:
531 some will refuse execution of set-user-ID and set-group-ID
532 executables when this would
533 give the user powers she did not have already (and return
535 some will just ignore the set-user-ID and set-group-ID bits and
543 can be specified as NULL.
544 In both cases, this has the same effect as specifying the argument
545 as a pointer to a list containing a single null pointer.
546 .B "Do not take advantage of this nonstandard and nonportable misfeature!"
547 On many other UNIX systems, specifying
549 as NULL will result in an error
552 other UNIX systems treat the
554 case the same as Linux.
555 .\" e.g., EFAULT on Solaris 8 and FreeBSD 6.1; but
556 .\" HP-UX 11 is like Linux -- mtk, Apr 2007
557 .\" Bug filed 30 Apr 2007: http://bugzilla.kernel.org/show_bug.cgi?id=8408
558 .\" Bug rejected (because fix would constitute an ABI change).
561 POSIX.1 says that values returned by
563 should be invariant over the lifetime of a process.
564 However, since Linux 2.6.23, if the
566 resource limit changes, then the value reported by
569 to reflect the fact that the limit on space for holding
570 command-line arguments and environment variables has changed.
574 fails, control returns to the original executable image,
577 can then handle the error.
578 However, in (rare) cases (typically caused by resource exhaustion),
579 failure may occur past the point of no return:
580 the original executable image has been torn down,
581 but the new image could not be completely built.
582 In such cases, the kernel kills the process with a
586 .SS Interpreter scripts
587 A maximum line length of 127 characters is allowed for the first line in
588 an interpreter scripts.
592 argument of an interpreter script vary across implementations.
593 On Linux, the entire string following the
595 name is passed as a single argument to the interpreter,
596 and this string can include white space.
597 However, behavior differs on some other systems.
600 use the first white space to terminate
603 .\" e.g., FreeBSD before 6.0, but not FreeBSD 6.0 onward
604 an interpreter script can have multiple arguments,
607 are used to delimit the arguments.
609 Linux ignores the set-user-ID and set-group-ID bits on scripts.
612 .\" Some Linux versions have failed to check permissions on ELF
613 .\" interpreters. This is a security hole, because it allows users to
614 .\" open any file, such as a rewinding tape device, for reading. Some
615 .\" Linux versions have also had other security holes in
617 .\" that could be exploited for denial of service by a suitably crafted
618 .\" ELF binary. There are no known problems with 2.0.34 or 2.2.15.
619 .SS execve() and EAGAIN
620 A more detailed explanation of the
622 error that can occur (since Linux 3.1) when calling
628 error can occur when a
635 caused the real user ID of the process to change,
636 and that change caused the process to exceed its
638 resource limit (i.e., the number of processes belonging
639 to the new real UID exceeds the resource limit).
640 From Linux 2.6.0 to 3.0, this caused the
644 .\" commit 909cc4ae86f3380152a18e2a3c44523893ee11c4
645 the resource limit was not imposed on processes that
646 changed their user IDs.)
648 Since Linux 3.1, the scenario just described no longer causes the
651 because it too often led to security holes where buggy applications
652 didn't check the return status and assumed
653 that\(emif the caller had root privileges\(emthe call would always succeed.
656 calls now successfully change the real UID,
657 but the kernel sets an internal flag, named
658 .BR PF_NPROC_EXCEEDED ,
661 resource limit has been exceeded.
663 .BR PF_NPROC_EXCEEDED
664 flag is set and the resource limit is still
665 exceeded at the time of a subsequent
667 call, that call fails with the error
669 This kernel logic ensures that the
671 resource limit is still enforced for the
672 common privileged daemon workflow\(emnamely,
679 If the resource limit was not still exceeded at the time of the
682 (because other processes belonging to this real UID terminated between the
688 call succeeds and the kernel clears the
689 .BR PF_NPROC_EXCEEDED
691 The flag is also cleared if a subsequent call to
693 by this process succeeds.
695 With UNIX\ V6, the argument list of an
698 while the argument list of
701 Thus, this argument list was not directly usable in a further
704 Since UNIX\ V7, both are NULL.
707 .\" Some Linux versions have failed to check permissions on ELF
708 .\" interpreters. This is a security hole, because it allows users to
709 .\" open any file, such as a rewinding tape device, for reading. Some
710 .\" Linux versions have also had other security holes in
712 .\" that could be exploited for denial of service by a suitably crafted
713 .\" ELF binary. There are no known problems with 2.0.34 or 2.2.15.
715 The following program is designed to be execed by the second program below.
716 It just echoes its command-line arguments, one per line.
726 main(int argc, char *argv[])
730 for (j = 0; j < argc; j++)
731 printf("argv[%d]: %s\\n", j, argv[j]);
738 This program can be used to exec the program named in its command-line
750 main(int argc, char *argv[])
752 char *newargv[] = { NULL, "hello", "world", NULL };
753 char *newenviron[] = { NULL };
756 fprintf(stderr, "Usage: %s <file\-to\-exec>\\n", argv[0]);
760 newargv[0] = argv[1];
762 execve(argv[1], newargv, newenviron);
763 perror("execve"); /* execve() returns only on error */
769 We can use the second program to exec the first as follows:
773 .RB "$" " cc myecho.c \-o myecho"
774 .RB "$" " cc execve.c \-o execve"
775 .RB "$" " ./execve ./myecho"
782 We can also use these programs to demonstrate the use of a script
784 To do this we create a script whose "interpreter" is our
790 .RB "$" " cat > script"
791 .B #!./myecho script-arg
793 .RB "$" " chmod +x script"
797 We can then use our program to exec the script:
801 .RB "$" " ./execve ./script"
820 .BR path_resolution (7),