1execve(2) System Calls Manual execve(2)
2
6 execve - execute program
7
9 Standard C library (libc, -lc)
10
12 #include <unistd.h>
13 int execve(const char *pathname, char *const _Nullable argv[],
15 char *const _Nullable envp[]);
16
18 execve() executes the program referred to by pathname. This causes the
19 program that is currently being run by the calling process to be re‐
20 placed with a new program, with newly initialized stack, heap, and
21 (initialized and uninitialized) data segments.
22 pathname must be either a binary executable, or a script starting with
24 a line of the form:
25 #!interpreter [optional-arg]
27 For details of the latter case, see "Interpreter scripts" below.
29 argv is an array of pointers to strings passed to the new program as
31 its command-line arguments. By convention, the first of these strings
32 (i.e., argv[0]) should contain the filename associated with the file
33 being executed. The argv array must be terminated by a NULL pointer.
34 (Thus, in the new program, argv[argc] will be NULL.)
35 envp is an array of pointers to strings, conventionally of the form
37 key=value, which are passed as the environment of the new program. The
38 envp array must be terminated by a NULL pointer.
39 This manual page describes the Linux system call in detail; for an
41 overview of the nomenclature and the many, often preferable, standard‐
42 ised variants of this function provided by libc, including ones that
43 search the PATH environment variable, see exec(3).
44 The argument vector and environment can be accessed by the new pro‐
46 gram's main function, when it is defined as:
47 int main(int argc, char *argv[], char *envp[])
49 Note, however, that the use of a third argument to the main function is
51 not specified in POSIX.1; according to POSIX.1, the environment should
52 be accessed via the external variable environ(7).
53 execve() does not return on success, and the text, initialized data,
55 uninitialized data (bss), and stack of the calling process are over‐
56 written according to the contents of the newly loaded program.
57 If the current program is being ptraced, a SIGTRAP signal is sent to it
59 after a successful execve().
60 If the set-user-ID bit is set on the program file referred to by path‐
62 name, then the effective user ID of the calling process is changed to
63 that of the owner of the program file. Similarly, if the set-group-ID
64 bit is set on the program file, then the effective group ID of the
65 calling process is set to the group of the program file.
66 The aforementioned transformations of the effective IDs are not per‐
68 formed (i.e., the set-user-ID and set-group-ID bits are ignored) if any
69 of the following is true:
70 • the no_new_privs attribute is set for the calling thread (see
72 prctl(2));
73 • the underlying filesystem is mounted nosuid (the MS_NOSUID flag for
75 mount(2)); or
76 • the calling process is being ptraced.
78 The capabilities of the program file (see capabilities(7)) are also ig‐
80 nored if any of the above are true.
81 The effective user ID of the process is copied to the saved set-user-
83 ID; similarly, the effective group ID is copied to the saved set-group-
84 ID. This copying takes place after any effective ID changes that occur
85 because of the set-user-ID and set-group-ID mode bits.
86 The process's real UID and real GID, as well as its supplementary group
88 IDs, are unchanged by a call to execve().
89 If the executable is an a.out dynamically linked binary executable con‐
91 taining shared-library stubs, the Linux dynamic linker ld.so(8) is
92 called at the start of execution to bring needed shared objects into
93 memory and link the executable with them.
94 If the executable is a dynamically linked ELF executable, the inter‐
96 preter named in the PT_INTERP segment is used to load the needed shared
97 objects. This interpreter is typically /lib/ld-linux.so.2 for binaries
98 linked with glibc (see ld-linux.so(8)).
99 Effect on process attributes
101 All process attributes are preserved during an execve(), except the
102 following:
103 • The dispositions of any signals that are being caught are reset to
105 the default (signal(7)).
106 • Any alternate signal stack is not preserved (sigaltstack(2)).
108 • Memory mappings are not preserved (mmap(2)).
110 • Attached System V shared memory segments are detached (shmat(2)).
112 • POSIX shared memory regions are unmapped (shm_open(3)).
114 • Open POSIX message queue descriptors are closed (mq_overview(7)).
116 • Any open POSIX named semaphores are closed (sem_overview(7)).
118 • POSIX timers are not preserved (timer_create(2)).
120 • Any open directory streams are closed (opendir(3)).
122 • Memory locks are not preserved (mlock(2), mlockall(2)).
124 • Exit handlers are not preserved (atexit(3), on_exit(3)).
126 • The floating-point environment is reset to the default (see
128 fenv(3)).
129 The process attributes in the preceding list are all specified in
131 POSIX.1. The following Linux-specific process attributes are also not
132 preserved during an execve():
133 • The process's "dumpable" attribute is set to the value 1, unless a
135 set-user-ID program, a set-group-ID program, or a program with capa‐
136 bilities is being executed, in which case the dumpable flag may in‐
137 stead be reset to the value in /proc/sys/fs/suid_dumpable, in the
138 circumstances described under PR_SET_DUMPABLE in prctl(2). Note
139 that changes to the "dumpable" attribute may cause ownership of
140 files in the process's /proc/pid directory to change to root:root,
141 as described in proc(5).
142 • The prctl(2) PR_SET_KEEPCAPS flag is cleared.
144 • (Since Linux 2.4.36 / 2.6.23) If a set-user-ID or set-group-ID pro‐
146 gram is being executed, then the parent death signal set by prctl(2)
147 PR_SET_PDEATHSIG flag is cleared.
148 • The process name, as set by prctl(2) PR_SET_NAME (and displayed by
150 ps -o comm), is reset to the name of the new executable file.
151 • The SECBIT_KEEP_CAPS securebits flag is cleared. See capabili‐
153 ties(7).
154 • The termination signal is reset to SIGCHLD (see clone(2)).
156 • The file descriptor table is unshared, undoing the effect of the
158 CLONE_FILES flag of clone(2).
159 Note the following further points:
161 • All threads other than the calling thread are destroyed during an
163 execve(). Mutexes, condition variables, and other pthreads objects
164 are not preserved.
165 • The equivalent of setlocale(LC_ALL, "C") is executed at program
167 start-up.
168 • POSIX.1 specifies that the dispositions of any signals that are ig‐
170 nored or set to the default are left unchanged. POSIX.1 specifies
171 one exception: if SIGCHLD is being ignored, then an implementation
172 may leave the disposition unchanged or reset it to the default;
173 Linux does the former.
174 • Any outstanding asynchronous I/O operations are canceled
176 (aio_read(3), aio_write(3)).
177 • For the handling of capabilities during execve(), see capabili‐
179 ties(7).
180 • By default, file descriptors remain open across an execve(). File
182 descriptors that are marked close-on-exec are closed; see the de‐
183 scription of FD_CLOEXEC in fcntl(2). (If a file descriptor is
184 closed, this will cause the release of all record locks obtained on
185 the underlying file by this process. See fcntl(2) for details.)
186 POSIX.1 says that if file descriptors 0, 1, and 2 would otherwise be
187 closed after a successful execve(), and the process would gain priv‐
188 ilege because the set-user-ID or set-group-ID mode bit was set on
189 the executed file, then the system may open an unspecified file for
190 each of these file descriptors. As a general principle, no portable
191 program, whether privileged or not, can assume that these three file
192 descriptors will remain closed across an execve().
193 Interpreter scripts
195 An interpreter script is a text file that has execute permission en‐
196 abled and whose first line is of the form:
197 #!interpreter [optional-arg]
199 The interpreter must be a valid pathname for an executable file.
201 If the pathname argument of execve() specifies an interpreter script,
203 then interpreter will be invoked with the following arguments:
204 interpreter [optional-arg] pathname arg...
206 where pathname is the pathname of the file specified as the first argu‐
208 ment of execve(), and arg... is the series of words pointed to by the
209 argv argument of execve(), starting at argv[1]. Note that there is no
210 way to get the argv[0] that was passed to the execve() call.
211 For portable use, optional-arg should either be absent, or be specified
213 as a single word (i.e., it should not contain white space); see NOTES
214 below.
215 Since Linux 2.6.28, the kernel permits the interpreter of a script to
217 itself be a script. This permission is recursive, up to a limit of
218 four recursions, so that the interpreter may be a script which is in‐
219 terpreted by a script, and so on.
220 Limits on size of arguments and environment
222 Most UNIX implementations impose some limit on the total size of the
223 command-line argument (argv) and environment (envp) strings that may be
224 passed to a new program. POSIX.1 allows an implementation to advertise
225 this limit using the ARG_MAX constant (either defined in <limits.h> or
226 available at run time using the call sysconf(_SC_ARG_MAX)).
227 Before Linux 2.6.23, the memory used to store the environment and argu‐
229 ment strings was limited to 32 pages (defined by the kernel constant
230 MAX_ARG_PAGES). On architectures with a 4-kB page size, this yields a
231 maximum size of 128 kB.
232 On Linux 2.6.23 and later, most architectures support a size limit de‐
234 rived from the soft RLIMIT_STACK resource limit (see getrlimit(2)) that
235 is in force at the time of the execve() call. (Architectures with no
236 memory management unit are excepted: they maintain the limit that was
237 in effect before Linux 2.6.23.) This change allows programs to have a
238 much larger argument and/or environment list. For these architectures,
239 the total size is limited to 1/4 of the allowed stack size. (Imposing
240 the 1/4-limit ensures that the new program always has some stack
241 space.) Additionally, the total size is limited to 3/4 of the value of
242 the kernel constant _STK_LIM (8 MiB). Since Linux 2.6.25, the kernel
243 also places a floor of 32 pages on this size limit, so that, even when
244 RLIMIT_STACK is set very low, applications are guaranteed to have at
245 least as much argument and environment space as was provided by Linux
246 2.6.22 and earlier. (This guarantee was not provided in Linux 2.6.23
247 and 2.6.24.) Additionally, the limit per string is 32 pages (the ker‐
248 nel constant MAX_ARG_STRLEN), and the maximum number of strings is
249 0x7FFFFFFF.
250
252 On success, execve() does not return, on error -1 is returned, and er‐
253 rno is set to indicate the error.
254
256 E2BIG The total number of bytes in the environment (envp) and argument
257 list (argv) is too large.
258 EACCES Search permission is denied on a component of the path prefix of
260 pathname or the name of a script interpreter. (See also
261 path_resolution(7).)
262 EACCES The file or a script interpreter is not a regular file.
264 EACCES Execute permission is denied for the file or a script or ELF in‐
266 terpreter.
267 EACCES The filesystem is mounted noexec.
269 EAGAIN (since Linux 3.1)
271 Having changed its real UID using one of the set*uid() calls,
272 the caller was—and is now still—above its RLIMIT_NPROC resource
273 limit (see setrlimit(2)). For a more detailed explanation of
274 this error, see NOTES.
275 EFAULT pathname or one of the pointers in the vectors argv or envp
277 points outside your accessible address space.
278 EINVAL An ELF executable had more than one PT_INTERP segment (i.e.,
280 tried to name more than one interpreter).
281 EIO An I/O error occurred.
283 EISDIR An ELF interpreter was a directory.
285 ELIBBAD
287 An ELF interpreter was not in a recognized format.
288 ELOOP Too many symbolic links were encountered in resolving pathname
290 or the name of a script or ELF interpreter.
291 ELOOP The maximum recursion limit was reached during recursive script
293 interpretation (see "Interpreter scripts", above). Before Linux
294 3.8, the error produced for this case was ENOEXEC.
295 EMFILE The per-process limit on the number of open file descriptors has
297 been reached.
298 ENAMETOOLONG
300 pathname is too long.
301 ENFILE The system-wide limit on the total number of open files has been
303 reached.
304 ENOENT The file pathname or a script or ELF interpreter does not exist.
306 ENOEXEC
308 An executable is not in a recognized format, is for the wrong
309 architecture, or has some other format error that means it can‐
310 not be executed.
311 ENOMEM Insufficient kernel memory was available.
313 ENOTDIR
315 A component of the path prefix of pathname or a script or ELF
316 interpreter is not a directory.
317 EPERM The filesystem is mounted nosuid, the user is not the superuser,
319 and the file has the set-user-ID or set-group-ID bit set.
320 EPERM The process is being traced, the user is not the superuser and
322 the file has the set-user-ID or set-group-ID bit set.
323 EPERM A "capability-dumb" applications would not obtain the full set
325 of permitted capabilities granted by the executable file. See
326 capabilities(7).
327 ETXTBSY
329 The specified executable was open for writing by one or more
330 processes.
331
333 POSIX does not document the #! behavior, but it exists (with some vari‐
334 ations) on other UNIX systems.
335 On Linux, argv and envp can be specified as NULL. In both cases, this
337 has the same effect as specifying the argument as a pointer to a list
338 containing a single null pointer. Do not take advantage of this non‐
339 standard and nonportable misfeature! On many other UNIX systems, spec‐
340 ifying argv as NULL will result in an error (EFAULT). Some other UNIX
341 systems treat the envp==NULL case the same as Linux.
342 POSIX.1 says that values returned by sysconf(3) should be invariant
344 over the lifetime of a process. However, since Linux 2.6.23, if the
345 RLIMIT_STACK resource limit changes, then the value reported by
346 _SC_ARG_MAX will also change, to reflect the fact that the limit on
347 space for holding command-line arguments and environment variables has
348 changed.
349 Interpreter scripts
351 The kernel imposes a maximum length on the text that follows the "#!"
352 characters at the start of a script; characters beyond the limit are
353 ignored. Before Linux 5.1, the limit is 127 characters. Since Linux
354 5.1, the limit is 255 characters.
355 The semantics of the optional-arg argument of an interpreter script
357 vary across implementations. On Linux, the entire string following the
358 interpreter name is passed as a single argument to the interpreter, and
359 this string can include white space. However, behavior differs on some
360 other systems. Some systems use the first white space to terminate op‐
361 tional-arg. On some systems, an interpreter script can have multiple
362 arguments, and white spaces in optional-arg are used to delimit the ar‐
363 guments.
364 Linux (like most other modern UNIX systems) ignores the set-user-ID and
366 set-group-ID bits on scripts.
367
369 POSIX.1-2008.
370
372 POSIX.1-2001, SVr4, 4.3BSD.
373 With UNIX V6, the argument list of an exec() call was ended by 0, while
375 the argument list of main was ended by -1. Thus, this argument list
376 was not directly usable in a further exec() call. Since UNIX V7, both
377 are NULL.
378
380 One sometimes sees execve() (and the related functions described in
381 exec(3)) described as "executing a new process" (or similar). This is
382 a highly misleading description: there is no new process; many at‐
383 tributes of the calling process remain unchanged (in particular, its
384 PID). All that execve() does is arrange for an existing process (the
385 calling process) to execute a new program.
386 Set-user-ID and set-group-ID processes can not be ptrace(2)d.
388 The result of mounting a filesystem nosuid varies across Linux kernel
390 versions: some will refuse execution of set-user-ID and set-group-ID
391 executables when this would give the user powers they did not have al‐
392 ready (and return EPERM), some will just ignore the set-user-ID and
393 set-group-ID bits and exec() successfully.
394 In most cases where execve() fails, control returns to the original ex‐
396 ecutable image, and the caller of execve() can then handle the error.
397 However, in (rare) cases (typically caused by resource exhaustion),
398 failure may occur past the point of no return: the original executable
399 image has been torn down, but the new image could not be completely
400 built. In such cases, the kernel kills the process with a SIGSEGV
401 (SIGKILL until Linux 3.17) signal.
402 execve() and EAGAIN
404 A more detailed explanation of the EAGAIN error that can occur (since
405 Linux 3.1) when calling execve() is as follows.
406 The EAGAIN error can occur when a preceding call to setuid(2), se‐
408 treuid(2), or setresuid(2) caused the real user ID of the process to
409 change, and that change caused the process to exceed its RLIMIT_NPROC
410 resource limit (i.e., the number of processes belonging to the new real
411 UID exceeds the resource limit). From Linux 2.6.0 to Linux 3.0, this
412 caused the set*uid() call to fail. (Before Linux 2.6, the resource
413 limit was not imposed on processes that changed their user IDs.)
414 Since Linux 3.1, the scenario just described no longer causes the
416 set*uid() call to fail, because it too often led to security holes
417 where buggy applications didn't check the return status and assumed
418 that—if the caller had root privileges—the call would always succeed.
419 Instead, the set*uid() calls now successfully change the real UID, but
420 the kernel sets an internal flag, named PF_NPROC_EXCEEDED, to note that
421 the RLIMIT_NPROC resource limit has been exceeded. If the PF_NPROC_EX‐
422 CEEDED flag is set and the resource limit is still exceeded at the time
423 of a subsequent execve() call, that call fails with the error EAGAIN.
424 This kernel logic ensures that the RLIMIT_NPROC resource limit is still
425 enforced for the common privileged daemon workflow—namely, fork(2) +
426 set*uid() + execve().
427 If the resource limit was not still exceeded at the time of the ex‐
429 ecve() call (because other processes belonging to this real UID termi‐
430 nated between the set*uid() call and the execve() call), then the ex‐
431 ecve() call succeeds and the kernel clears the PF_NPROC_EXCEEDED
432 process flag. The flag is also cleared if a subsequent call to fork(2)
433 by this process succeeds.
434
436 The following program is designed to be execed by the second program
437 below. It just echoes its command-line arguments, one per line.
438 /* myecho.c */
440 #include <stdio.h>
442 #include <stdlib.h>
443 int
445 main(int argc, char *argv[])
446 {
447 for (size_t j = 0; j < argc; j++)
448 printf("argv[%zu]: %s\n", j, argv[j]);
449 exit(EXIT_SUCCESS);
451 }
452 This program can be used to exec the program named in its command-line
454 argument:
455 /* execve.c */
457 #include <stdio.h>
459 #include <stdlib.h>
460 #include <unistd.h>
461 int
463 main(int argc, char *argv[])
464 {
465 static char *newargv[] = { NULL, "hello", "world", NULL };
466 static char *newenviron[] = { NULL };
467 if (argc != 2) {
469 fprintf(stderr, "Usage: %s <file-to-exec>\n", argv[0]);
470 exit(EXIT_FAILURE);
471 }
472 newargv[0] = argv[1];
474 execve(argv[1], newargv, newenviron);
476 perror("execve"); /* execve() returns only on error */
477 exit(EXIT_FAILURE);
478 }
479 We can use the second program to exec the first as follows:
481 $ cc myecho.c -o myecho
483 $ cc execve.c -o execve
484 $ ./execve ./myecho
485 argv[0]: ./myecho
486 argv[1]: hello
487 argv[2]: world
488 We can also use these programs to demonstrate the use of a script in‐
490 terpreter. To do this we create a script whose "interpreter" is our
491 myecho program:
492 $ cat > script
494 #!./myecho script-arg
495 ^D
496 $ chmod +x script
497 We can then use our program to exec the script:
499 $ ./execve ./script
501 argv[0]: ./myecho
502 argv[1]: script-arg
503 argv[2]: ./script
504 argv[3]: hello
505 argv[4]: world
506
508 chmod(2), execveat(2), fork(2), get_robust_list(2), ptrace(2), exec(3),
509 fexecve(3), getauxval(3), getopt(3), system(3), capabilities(7), cre‐
510 dentials(7), environ(7), path_resolution(7), ld.so(8)
511Linux man-pages 6.05 2023-05-03 execve(2)