1EXECVE(2) Linux Programmer's Manual EXECVE(2)
2
6 execve - execute program
7
9 #include <unistd.h>
10 int execve(const char *pathname, char *const argv[],
12 char *const envp[]);
13
15 execve() executes the program referred to by pathname. This causes the
16 program that is currently being run by the calling process to be
17 replaced with a new program, with newly initialized stack, heap, and
18 (initialized and uninitialized) data segments.
19 pathname must be either a binary executable, or a script starting with
21 a line of the form:
22 #!interpreter [optional-arg]
24 For details of the latter case, see "Interpreter scripts" below.
26 argv is an array of argument strings passed to the new program. By
28 convention, the first of these strings (i.e., argv[0]) should contain
29 the filename associated with the file being executed. envp is an array
30 of strings, conventionally of the form key=value, which are passed as
31 environment to the new program. The argv and envp arrays must each
32 include a null pointer at the end of the array.
33 The argument vector and environment can be accessed by the called pro‐
35 gram's main function, when it is defined as:
36 int main(int argc, char *argv[], char *envp[])
38 Note, however, that the use of a third argument to the main function is
40 not specified in POSIX.1; according to POSIX.1, the environment should
41 be accessed via the external variable environ(7).
42 execve() does not return on success, and the text, initialized data,
44 uninitialized data (bss), and stack of the calling process are over‐
45 written according to the contents of the newly loaded program.
46 If the current program is being ptraced, a SIGTRAP signal is sent to it
48 after a successful execve().
49 If the set-user-ID bit is set on the program file referred to by path‐
51 name, then the effective user ID of the calling process is changed to
52 that of the owner of the program file. Similarly, when the set-group-
53 ID bit of the program file is set the effective group ID of the calling
54 process is set to the group of the program file.
55 The aforementioned transformations of the effective IDs are not per‐
57 formed (i.e., the set-user-ID and set-group-ID bits are ignored) if any
58 of the following is true:
59 * the no_new_privs attribute is set for the calling thread (see
61 prctl(2));
62 * the underlying filesystem is mounted nosuid (the MS_NOSUID flag for
64 mount(2)); or
65 * the calling process is being ptraced.
67 The capabilities of the program file (see capabilities(7)) are also
69 ignored if any of the above are true.
70 The effective user ID of the process is copied to the saved set-user-
72 ID; similarly, the effective group ID is copied to the saved set-group-
73 ID. This copying takes place after any effective ID changes that occur
74 because of the set-user-ID and set-group-ID mode bits.
75 The process's real UID and real GID, as well its supplementary group
77 IDs, are unchanged by a call to execve().
78 If the executable is an a.out dynamically linked binary executable con‐
80 taining shared-library stubs, the Linux dynamic linker ld.so(8) is
81 called at the start of execution to bring needed shared objects into
82 memory and link the executable with them.
83 If the executable is a dynamically linked ELF executable, the inter‐
85 preter named in the PT_INTERP segment is used to load the needed shared
86 objects. This interpreter is typically /lib/ld-linux.so.2 for binaries
87 linked with glibc (see ld-linux.so(8)).
88 All process attributes are preserved during an execve(), except the
90 following:
91 * The dispositions of any signals that are being caught are reset to
93 the default (signal(7)).
94 * Any alternate signal stack is not preserved (sigaltstack(2)).
96 * Memory mappings are not preserved (mmap(2)).
98 * Attached System V shared memory segments are detached (shmat(2)).
100 * POSIX shared memory regions are unmapped (shm_open(3)).
102 * Open POSIX message queue descriptors are closed (mq_overview(7)).
104 * Any open POSIX named semaphores are closed (sem_overview(7)).
106 * POSIX timers are not preserved (timer_create(2)).
108 * Any open directory streams are closed (opendir(3)).
110 * Memory locks are not preserved (mlock(2), mlockall(2)).
112 * Exit handlers are not preserved (atexit(3), on_exit(3)).
114 * The floating-point environment is reset to the default (see
116 fenv(3)).
117 The process attributes in the preceding list are all specified in
119 POSIX.1. The following Linux-specific process attributes are also not
120 preserved during an execve():
121 * The prctl(2) PR_SET_DUMPABLE flag is set, unless a set-user-ID or
123 set-group ID program is being executed, in which case it is cleared.
124 * The prctl(2) PR_SET_KEEPCAPS flag is cleared.
126 * (Since Linux 2.4.36 / 2.6.23) If a set-user-ID or set-group-ID pro‐
128 gram is being executed, then the parent death signal set by prctl(2)
129 PR_SET_PDEATHSIG flag is cleared.
130 * The process name, as set by prctl(2) PR_SET_NAME (and displayed by
132 ps -o comm), is reset to the name of the new executable file.
133 * The SECBIT_KEEP_CAPS securebits flag is cleared. See capabili‐
135 ties(7).
136 * The termination signal is reset to SIGCHLD (see clone(2)).
138 * The file descriptor table is unshared, undoing the effect of the
140 CLONE_FILES flag of clone(2).
141 Note the following further points:
143 * All threads other than the calling thread are destroyed during an
145 execve(). Mutexes, condition variables, and other pthreads objects
146 are not preserved.
147 * The equivalent of setlocale(LC_ALL, "C") is executed at program
149 start-up.
150 * POSIX.1 specifies that the dispositions of any signals that are
152 ignored or set to the default are left unchanged. POSIX.1 specifies
153 one exception: if SIGCHLD is being ignored, then an implementation
154 may leave the disposition unchanged or reset it to the default;
155 Linux does the former.
156 * Any outstanding asynchronous I/O operations are canceled
158 (aio_read(3), aio_write(3)).
159 * For the handling of capabilities during execve(), see capabili‐
161 ties(7).
162 * By default, file descriptors remain open across an execve(). File
164 descriptors that are marked close-on-exec are closed; see the
165 description of FD_CLOEXEC in fcntl(2). (If a file descriptor is
166 closed, this will cause the release of all record locks obtained on
167 the underlying file by this process. See fcntl(2) for details.)
168 POSIX.1 says that if file descriptors 0, 1, and 2 would otherwise be
169 closed after a successful execve(), and the process would gain priv‐
170 ilege because the set-user-ID or set-group_ID mode bit was set on
171 the executed file, then the system may open an unspecified file for
172 each of these file descriptors. As a general principle, no portable
173 program, whether privileged or not, can assume that these three file
174 descriptors will remain closed across an execve().
175 Interpreter scripts
177 An interpreter script is a text file that has execute permission
178 enabled and whose first line is of the form:
179 #!interpreter [optional-arg]
181 The interpreter must be a valid pathname for an executable file.
183 If the pathname argument of execve() specifies an interpreter script,
185 then interpreter will be invoked with the following arguments:
186 interpreter [optional-arg] pathname arg...
188 where pathname is the absolute pathname of the file specified as the
190 first argument of execve(), and arg... is the series of words pointed
191 to by the argv argument of execve(), starting at argv[1]. Note that
192 there is no way to get the argv[0] that was passed to the execve()
193 call.
194 For portable use, optional-arg should either be absent, or be specified
196 as a single word (i.e., it should not contain white space); see NOTES
197 below.
198 Since Linux 2.6.28, the kernel permits the interpreter of a script to
200 itself be a script. This permission is recursive, up to a limit of
201 four recursions, so that the interpreter may be a script which is
202 interpreted by a script, and so on.
203 Limits on size of arguments and environment
205 Most UNIX implementations impose some limit on the total size of the
206 command-line argument (argv) and environment (envp) strings that may be
207 passed to a new program. POSIX.1 allows an implementation to advertise
208 this limit using the ARG_MAX constant (either defined in <limits.h> or
209 available at run time using the call sysconf(_SC_ARG_MAX)).
210 On Linux prior to kernel 2.6.23, the memory used to store the environ‐
212 ment and argument strings was limited to 32 pages (defined by the ker‐
213 nel constant MAX_ARG_PAGES). On architectures with a 4-kB page size,
214 this yields a maximum size of 128 kB.
215 On kernel 2.6.23 and later, most architectures support a size limit
217 derived from the soft RLIMIT_STACK resource limit (see getrlimit(2))
218 that is in force at the time of the execve() call. (Architectures with
219 no memory management unit are excepted: they maintain the limit that
220 was in effect before kernel 2.6.23.) This change allows programs to
221 have a much larger argument and/or environment list. For these archi‐
222 tectures, the total size is limited to 1/4 of the allowed stack size.
223 (Imposing the 1/4-limit ensures that the new program always has some
224 stack space.) Additionally, the total size is limited to 3/4 of the
225 value of the kernel constant _STK_LIM (8 Mibibytes). Since Linux
226 2.6.25, the kernel also places a floor of 32 pages on this size limit,
227 so that, even when RLIMIT_STACK is set very low, applications are guar‐
228 anteed to have at least as much argument and environment space as was
229 provided by Linux 2.6.23 and earlier. (This guarantee was not provided
230 in Linux 2.6.23 and 2.6.24.) Additionally, the limit per string is 32
231 pages (the kernel constant MAX_ARG_STRLEN), and the maximum number of
232 strings is 0x7FFFFFFF.
233
235 On success, execve() does not return, on error -1 is returned, and
236 errno is set appropriately.
237
239 E2BIG The total number of bytes in the environment (envp) and argument
240 list (argv) is too large.
241 EACCES Search permission is denied on a component of the path prefix of
243 pathname or the name of a script interpreter. (See also
244 path_resolution(7).)
245 EACCES The file or a script interpreter is not a regular file.
247 EACCES Execute permission is denied for the file or a script or ELF
249 interpreter.
250 EACCES The filesystem is mounted noexec.
252 EAGAIN (since Linux 3.1)
254 Having changed its real UID using one of the set*uid() calls,
255 the caller was—and is now still—above its RLIMIT_NPROC resource
256 limit (see setrlimit(2)). For a more detailed explanation of
257 this error, see NOTES.
258 EFAULT pathname or one of the pointers in the vectors argv or envp
260 points outside your accessible address space.
261 EINVAL An ELF executable had more than one PT_INTERP segment (i.e.,
263 tried to name more than one interpreter).
264 EIO An I/O error occurred.
266 EISDIR An ELF interpreter was a directory.
268 ELIBBAD
270 An ELF interpreter was not in a recognized format.
271 ELOOP Too many symbolic links were encountered in resolving pathname
273 or the name of a script or ELF interpreter.
274 ELOOP The maximum recursion limit was reached during recursive script
276 interpretation (see "Interpreter scripts", above). Before Linux
277 3.8, the error produced for this case was ENOEXEC.
278 EMFILE The per-process limit on the number of open file descriptors has
280 been reached.
281 ENAMETOOLONG
283 pathname is too long.
284 ENFILE The system-wide limit on the total number of open files has been
286 reached.
287 ENOENT The file pathname or a script or ELF interpreter does not exist,
289 or a shared library needed for the file or interpreter cannot be
290 found.
291 ENOEXEC
293 An executable is not in a recognized format, is for the wrong
294 architecture, or has some other format error that means it can‐
295 not be executed.
296 ENOMEM Insufficient kernel memory was available.
298 ENOTDIR
300 A component of the path prefix of pathname or a script or ELF
301 interpreter is not a directory.
302 EPERM The filesystem is mounted nosuid, the user is not the superuser,
304 and the file has the set-user-ID or set-group-ID bit set.
305 EPERM The process is being traced, the user is not the superuser and
307 the file has the set-user-ID or set-group-ID bit set.
308 EPERM A "capability-dumb" applications would not obtain the full set
310 of permitted capabilities granted by the executable file. See
311 capabilities(7).
312 ETXTBSY
314 The specified executable was open for writing by one or more
315 processes.
316
318 POSIX.1-2001, POSIX.1-2008, SVr4, 4.3BSD. POSIX does not document the
319 #! behavior, but it exists (with some variations) on other UNIX sys‐
320 tems.
321
323 One sometimes sees execve() (and the related functions described in
324 exec(3)) described as "executing a new process" (or similar). This is
325 a highly misleading description: there is no new process; many
326 attributes of the calling process remain unchanged (in particular, its
327 PID). All that execve() does is arrange for an existing process (the
328 calling process) to execute a new program.
329 Set-user-ID and set-group-ID processes can not be ptrace(2)d.
331 The result of mounting a filesystem nosuid varies across Linux kernel
333 versions: some will refuse execution of set-user-ID and set-group-ID
334 executables when this would give the user powers they did not have
335 already (and return EPERM), some will just ignore the set-user-ID and
336 set-group-ID bits and exec() successfully.
337 On Linux, argv and envp can be specified as NULL. In both cases, this
339 has the same effect as specifying the argument as a pointer to a list
340 containing a single null pointer. Do not take advantage of this non‐
341 standard and nonportable misfeature! On many other UNIX systems, spec‐
342 ifying argv as NULL will result in an error (EFAULT). Some other UNIX
343 systems treat the envp==NULL case the same as Linux.
344 POSIX.1 says that values returned by sysconf(3) should be invariant
346 over the lifetime of a process. However, since Linux 2.6.23, if the
347 RLIMIT_STACK resource limit changes, then the value reported by
348 _SC_ARG_MAX will also change, to reflect the fact that the limit on
349 space for holding command-line arguments and environment variables has
350 changed.
351 In most cases where execve() fails, control returns to the original
353 executable image, and the caller of execve() can then handle the error.
354 However, in (rare) cases (typically caused by resource exhaustion),
355 failure may occur past the point of no return: the original executable
356 image has been torn down, but the new image could not be completely
357 built. In such cases, the kernel kills the process with a SIGKILL sig‐
358 nal.
359 Interpreter scripts
361 The kernel imposes a maximum length on the text that follows the "#!"
362 characters at the start of a script; characters beyond the limit are
363 ignored. Before Linux 5.1, the limit is 127 characters. Since Linux
364 5.1, the limit is 255 characters.
365 The semantics of the optional-arg argument of an interpreter script
367 vary across implementations. On Linux, the entire string following the
368 interpreter name is passed as a single argument to the interpreter, and
369 this string can include white space. However, behavior differs on some
370 other systems. Some systems use the first white space to terminate
371 optional-arg. On some systems, an interpreter script can have multiple
372 arguments, and white spaces in optional-arg are used to delimit the
373 arguments.
374 Linux (like most other modern UNIX systems) ignores the set-user-ID and
376 set-group-ID bits on scripts.
377 execve() and EAGAIN
379 A more detailed explanation of the EAGAIN error that can occur (since
380 Linux 3.1) when calling execve() is as follows.
381 The EAGAIN error can occur when a preceding call to setuid(2),
383 setreuid(2), or setresuid(2) caused the real user ID of the process to
384 change, and that change caused the process to exceed its RLIMIT_NPROC
385 resource limit (i.e., the number of processes belonging to the new real
386 UID exceeds the resource limit). From Linux 2.6.0 to 3.0, this caused
387 the set*uid() call to fail. (Prior to 2.6, the resource limit was not
388 imposed on processes that changed their user IDs.)
389 Since Linux 3.1, the scenario just described no longer causes the
391 set*uid() call to fail, because it too often led to security holes
392 where buggy applications didn't check the return status and assumed
393 that—if the caller had root privileges—the call would always succeed.
394 Instead, the set*uid() calls now successfully change the real UID, but
395 the kernel sets an internal flag, named PF_NPROC_EXCEEDED, to note that
396 the RLIMIT_NPROC resource limit has been exceeded. If the
397 PF_NPROC_EXCEEDED flag is set and the resource limit is still exceeded
398 at the time of a subsequent execve() call, that call fails with the
399 error EAGAIN. This kernel logic ensures that the RLIMIT_NPROC resource
400 limit is still enforced for the common privileged daemon workflow—
401 namely, fork(2) + set*uid() + execve().
402 If the resource limit was not still exceeded at the time of the
404 execve() call (because other processes belonging to this real UID ter‐
405 minated between the set*uid() call and the execve() call), then the
406 execve() call succeeds and the kernel clears the PF_NPROC_EXCEEDED
407 process flag. The flag is also cleared if a subsequent call to fork(2)
408 by this process succeeds.
409 Historical
411 With UNIX V6, the argument list of an exec() call was ended by 0, while
412 the argument list of main was ended by -1. Thus, this argument list
413 was not directly usable in a further exec() call. Since UNIX V7, both
414 are NULL.
415
417 The following program is designed to be execed by the second program
418 below. It just echoes its command-line arguments, one per line.
419 /* myecho.c */
421 #include <stdio.h>
423 #include <stdlib.h>
424 int
426 main(int argc, char *argv[])
427 {
428 int j;
429 for (j = 0; j < argc; j++)
431 printf("argv[%d]: %s\n", j, argv[j]);
432 exit(EXIT_SUCCESS);
434 }
435 This program can be used to exec the program named in its command-line
437 argument:
438 /* execve.c */
440 #include <stdio.h>
442 #include <stdlib.h>
443 #include <unistd.h>
444 int
446 main(int argc, char *argv[])
447 {
448 char *newargv[] = { NULL, "hello", "world", NULL };
449 char *newenviron[] = { NULL };
450 if (argc != 2) {
452 fprintf(stderr, "Usage: %s <file-to-exec>\n", argv[0]);
453 exit(EXIT_FAILURE);
454 }
455 newargv[0] = argv[1];
457 execve(argv[1], newargv, newenviron);
459 perror("execve"); /* execve() returns only on error */
460 exit(EXIT_FAILURE);
461 }
462 We can use the second program to exec the first as follows:
464 $ cc myecho.c -o myecho
466 $ cc execve.c -o execve
467 $ ./execve ./myecho
468 argv[0]: ./myecho
469 argv[1]: hello
470 argv[2]: world
471 We can also use these programs to demonstrate the use of a script
473 interpreter. To do this we create a script whose "interpreter" is our
474 myecho program:
475 $ cat > script
477 #!./myecho script-arg
478 ^D
479 $ chmod +x script
480 We can then use our program to exec the script:
482 $ ./execve ./script
484 argv[0]: ./myecho
485 argv[1]: script-arg
486 argv[2]: ./script
487 argv[3]: hello
488 argv[4]: world
489
491 chmod(2), execveat(2), fork(2), get_robust_list(2), ptrace(2), exec(3),
492 fexecve(3), getopt(3), system(3), credentials(7), environ(7), path_res‐
493 olution(7), ld.so(8)
494
496 This page is part of release 5.04 of the Linux man-pages project. A
497 description of the project, information about reporting bugs, and the
498 latest version of this page, can be found at
499 https://www.kernel.org/doc/man-pages/.
500Linux 2019-10-10 EXECVE(2)