xref: /freebsd/lib/libc/gen/posix_spawn.c (revision cc68614da8232d8baaca0ae0d0dd8f890f06623e)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 2008 Ed Schouten <ed@FreeBSD.org>
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28 
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
31 
32 #include "namespace.h"
33 #include <sys/param.h>
34 #include <sys/queue.h>
35 #include <sys/wait.h>
36 
37 #include <errno.h>
38 #include <fcntl.h>
39 #include <sched.h>
40 #include <spawn.h>
41 #include <signal.h>
42 #include <stdlib.h>
43 #include <string.h>
44 #include <unistd.h>
45 #include "un-namespace.h"
46 #include "libc_private.h"
47 
48 extern char **environ;
49 
50 struct __posix_spawnattr {
51 	short			sa_flags;
52 	pid_t			sa_pgroup;
53 	struct sched_param	sa_schedparam;
54 	int			sa_schedpolicy;
55 	sigset_t		sa_sigdefault;
56 	sigset_t		sa_sigmask;
57 };
58 
59 struct __posix_spawn_file_actions {
60 	STAILQ_HEAD(, __posix_spawn_file_actions_entry) fa_list;
61 };
62 
63 typedef struct __posix_spawn_file_actions_entry {
64 	STAILQ_ENTRY(__posix_spawn_file_actions_entry) fae_list;
65 	enum {
66 		FAE_OPEN,
67 		FAE_DUP2,
68 		FAE_CLOSE,
69 		FAE_CHDIR,
70 		FAE_FCHDIR,
71 		FAE_CLOSEFROM,
72 	} fae_action;
73 
74 	int fae_fildes;
75 	union {
76 		struct {
77 			char *path;
78 #define fae_path	fae_data.open.path
79 			int oflag;
80 #define fae_oflag	fae_data.open.oflag
81 			mode_t mode;
82 #define fae_mode	fae_data.open.mode
83 		} open;
84 		struct {
85 			int newfildes;
86 #define fae_newfildes	fae_data.dup2.newfildes
87 		} dup2;
88 	} fae_data;
89 } posix_spawn_file_actions_entry_t;
90 
91 /*
92  * Spawn routines
93  */
94 
95 static int
96 process_spawnattr(const posix_spawnattr_t sa)
97 {
98 	struct sigaction sigact = { .sa_flags = 0, .sa_handler = SIG_DFL };
99 	int i;
100 
101 	/*
102 	 * POSIX doesn't really describe in which order everything
103 	 * should be set. We'll just set them in the order in which they
104 	 * are mentioned.
105 	 */
106 
107 	/* Set process group */
108 	if (sa->sa_flags & POSIX_SPAWN_SETPGROUP) {
109 		if (setpgid(0, sa->sa_pgroup) != 0)
110 			return (errno);
111 	}
112 
113 	/* Set scheduler policy */
114 	if (sa->sa_flags & POSIX_SPAWN_SETSCHEDULER) {
115 		if (sched_setscheduler(0, sa->sa_schedpolicy,
116 		    &sa->sa_schedparam) != 0)
117 			return (errno);
118 	} else if (sa->sa_flags & POSIX_SPAWN_SETSCHEDPARAM) {
119 		if (sched_setparam(0, &sa->sa_schedparam) != 0)
120 			return (errno);
121 	}
122 
123 	/* Reset user ID's */
124 	if (sa->sa_flags & POSIX_SPAWN_RESETIDS) {
125 		if (setegid(getgid()) != 0)
126 			return (errno);
127 		if (seteuid(getuid()) != 0)
128 			return (errno);
129 	}
130 
131 	/*
132 	 * Set signal masks/defaults.
133 	 * Use unwrapped syscall, libthr is in undefined state after vfork().
134 	 */
135 	if (sa->sa_flags & POSIX_SPAWN_SETSIGMASK) {
136 		__sys_sigprocmask(SIG_SETMASK, &sa->sa_sigmask, NULL);
137 	}
138 
139 	if (sa->sa_flags & POSIX_SPAWN_SETSIGDEF) {
140 		for (i = 1; i <= _SIG_MAXSIG; i++) {
141 			if (sigismember(&sa->sa_sigdefault, i))
142 				if (__sys_sigaction(i, &sigact, NULL) != 0)
143 					return (errno);
144 		}
145 	}
146 
147 	return (0);
148 }
149 
150 static int
151 process_file_actions_entry(posix_spawn_file_actions_entry_t *fae)
152 {
153 	int fd, saved_errno;
154 
155 	switch (fae->fae_action) {
156 	case FAE_OPEN:
157 		/* Perform an open(), make it use the right fd */
158 		fd = _open(fae->fae_path, fae->fae_oflag, fae->fae_mode);
159 		if (fd < 0)
160 			return (errno);
161 		if (fd != fae->fae_fildes) {
162 			if (_dup2(fd, fae->fae_fildes) == -1) {
163 				saved_errno = errno;
164 				(void)_close(fd);
165 				return (saved_errno);
166 			}
167 			if (_close(fd) != 0) {
168 				if (errno == EBADF)
169 					return (EBADF);
170 			}
171 		}
172 		if (_fcntl(fae->fae_fildes, F_SETFD, 0) == -1)
173 			return (errno);
174 		break;
175 	case FAE_DUP2:
176 		/* Perform a dup2() */
177 		if (_dup2(fae->fae_fildes, fae->fae_newfildes) == -1)
178 			return (errno);
179 		if (_fcntl(fae->fae_newfildes, F_SETFD, 0) == -1)
180 			return (errno);
181 		break;
182 	case FAE_CLOSE:
183 		/* Perform a close(), do not fail if already closed */
184 		(void)_close(fae->fae_fildes);
185 		break;
186 	case FAE_CHDIR:
187 		if (chdir(fae->fae_path) != 0)
188 			return (errno);
189 		break;
190 	case FAE_FCHDIR:
191 		if (fchdir(fae->fae_fildes) != 0)
192 			return (errno);
193 		break;
194 	case FAE_CLOSEFROM:
195 		closefrom(fae->fae_fildes);
196 		break;
197 	}
198 	return (0);
199 }
200 
201 static int
202 process_file_actions(const posix_spawn_file_actions_t fa)
203 {
204 	posix_spawn_file_actions_entry_t *fae;
205 	int error;
206 
207 	/* Replay all file descriptor modifications */
208 	STAILQ_FOREACH(fae, &fa->fa_list, fae_list) {
209 		error = process_file_actions_entry(fae);
210 		if (error)
211 			return (error);
212 	}
213 	return (0);
214 }
215 
216 struct posix_spawn_args {
217 	const char *path;
218 	const posix_spawn_file_actions_t *fa;
219 	const posix_spawnattr_t *sa;
220 	char * const * argv;
221 	char * const * envp;
222 	int use_env_path;
223 	volatile int error;
224 };
225 
226 #define	PSPAWN_STACK_ALIGNMENT	16
227 #define	PSPAWN_STACK_ALIGNBYTES	(PSPAWN_STACK_ALIGNMENT - 1)
228 #define	PSPAWN_STACK_ALIGN(sz) \
229 	(((sz) + PSPAWN_STACK_ALIGNBYTES) & ~PSPAWN_STACK_ALIGNBYTES)
230 
231 #if defined(__i386__) || defined(__amd64__)
232 /*
233  * Below we'll assume that _RFORK_THREAD_STACK_SIZE is appropriately aligned for
234  * the posix_spawn() case where we do not end up calling _execvpe and won't ever
235  * try to allocate space on the stack for argv[].
236  */
237 #define	_RFORK_THREAD_STACK_SIZE	4096
238 _Static_assert((_RFORK_THREAD_STACK_SIZE % PSPAWN_STACK_ALIGNMENT) == 0,
239     "Inappropriate stack size alignment");
240 #endif
241 
242 static int
243 _posix_spawn_thr(void *data)
244 {
245 	struct posix_spawn_args *psa;
246 	char * const *envp;
247 
248 	psa = data;
249 	if (psa->sa != NULL) {
250 		psa->error = process_spawnattr(*psa->sa);
251 		if (psa->error)
252 			_exit(127);
253 	}
254 	if (psa->fa != NULL) {
255 		psa->error = process_file_actions(*psa->fa);
256 		if (psa->error)
257 			_exit(127);
258 	}
259 	envp = psa->envp != NULL ? psa->envp : environ;
260 	if (psa->use_env_path)
261 		_execvpe(psa->path, psa->argv, envp);
262 	else
263 		_execve(psa->path, psa->argv, envp);
264 	psa->error = errno;
265 
266 	/* This is called in such a way that it must not exit. */
267 	_exit(127);
268 }
269 
270 static int
271 do_posix_spawn(pid_t *pid, const char *path,
272     const posix_spawn_file_actions_t *fa,
273     const posix_spawnattr_t *sa,
274     char * const argv[], char * const envp[], int use_env_path)
275 {
276 	struct posix_spawn_args psa;
277 	pid_t p;
278 #ifdef _RFORK_THREAD_STACK_SIZE
279 	char *stack;
280 	size_t cnt, stacksz;
281 
282 	stacksz = _RFORK_THREAD_STACK_SIZE;
283 	if (use_env_path) {
284 		/*
285 		 * We need to make sure we have enough room on the stack for the
286 		 * potential alloca() in execvPe if it gets kicked back an
287 		 * ENOEXEC from execve(2), plus the original buffer we gave
288 		 * ourselves; this protects us in the event that the caller
289 		 * intentionally or inadvertently supplies enough arguments to
290 		 * make us blow past the stack we've allocated from it.
291 		 */
292 		for (cnt = 0; argv[cnt] != NULL; ++cnt)
293 			;
294 		stacksz += MAX(3, cnt + 2) * sizeof(char *);
295 		stacksz = PSPAWN_STACK_ALIGN(stacksz);
296 	}
297 
298 	/*
299 	 * aligned_alloc is not safe to use here, because we can't guarantee
300 	 * that aligned_alloc and free will be provided by the same
301 	 * implementation.  We've actively hit at least one application that
302 	 * will provide its own malloc/free but not aligned_alloc leading to
303 	 * a free by the wrong allocator.
304 	 */
305 	stack = malloc(stacksz);
306 	if (stack == NULL)
307 		return (ENOMEM);
308 	stacksz = (((uintptr_t)stack + stacksz) & ~PSPAWN_STACK_ALIGNBYTES) -
309 	    (uintptr_t)stack;
310 #endif
311 	psa.path = path;
312 	psa.fa = fa;
313 	psa.sa = sa;
314 	psa.argv = argv;
315 	psa.envp = envp;
316 	psa.use_env_path = use_env_path;
317 	psa.error = 0;
318 
319 	/*
320 	 * Passing RFSPAWN to rfork(2) gives us effectively a vfork that drops
321 	 * non-ignored signal handlers.  We'll fall back to the slightly less
322 	 * ideal vfork(2) if we get an EINVAL from rfork -- this should only
323 	 * happen with newer libc on older kernel that doesn't accept
324 	 * RFSPAWN.
325 	 */
326 #ifdef _RFORK_THREAD_STACK_SIZE
327 	/*
328 	 * x86 stores the return address on the stack, so rfork(2) cannot work
329 	 * as-is because the child would clobber the return address om the
330 	 * parent.  Because of this, we must use rfork_thread instead while
331 	 * almost every other arch stores the return address in a register.
332 	 */
333 	p = rfork_thread(RFSPAWN, stack + stacksz, _posix_spawn_thr, &psa);
334 	free(stack);
335 #else
336 	p = rfork(RFSPAWN);
337 	if (p == 0)
338 		/* _posix_spawn_thr does not return */
339 		_posix_spawn_thr(&psa);
340 #endif
341 	/*
342 	 * The above block should leave us in a state where we've either
343 	 * succeeded and we're ready to process the results, or we need to
344 	 * fallback to vfork() if the kernel didn't like RFSPAWN.
345 	 */
346 
347 	if (p == -1 && errno == EINVAL) {
348 		p = vfork();
349 		if (p == 0)
350 			/* _posix_spawn_thr does not return */
351 			_posix_spawn_thr(&psa);
352 	}
353 	if (p == -1)
354 		return (errno);
355 	if (psa.error != 0)
356 		/* Failed; ready to reap */
357 		_waitpid(p, NULL, WNOHANG);
358 	else if (pid != NULL)
359 		/* exec succeeded */
360 		*pid = p;
361 	return (psa.error);
362 }
363 
364 int
365 posix_spawn(pid_t *pid, const char *path,
366     const posix_spawn_file_actions_t *fa,
367     const posix_spawnattr_t *sa,
368     char * const argv[], char * const envp[])
369 {
370 	return (do_posix_spawn(pid, path, fa, sa, argv, envp, 0));
371 }
372 
373 int
374 posix_spawnp(pid_t *pid, const char *path,
375     const posix_spawn_file_actions_t *fa,
376     const posix_spawnattr_t *sa,
377     char * const argv[], char * const envp[])
378 {
379 	return (do_posix_spawn(pid, path, fa, sa, argv, envp, 1));
380 }
381 
382 /*
383  * File descriptor actions
384  */
385 
386 int
387 posix_spawn_file_actions_init(posix_spawn_file_actions_t *ret)
388 {
389 	posix_spawn_file_actions_t fa;
390 
391 	fa = malloc(sizeof(struct __posix_spawn_file_actions));
392 	if (fa == NULL)
393 		return (-1);
394 
395 	STAILQ_INIT(&fa->fa_list);
396 	*ret = fa;
397 	return (0);
398 }
399 
400 int
401 posix_spawn_file_actions_destroy(posix_spawn_file_actions_t *fa)
402 {
403 	posix_spawn_file_actions_entry_t *fae;
404 
405 	while ((fae = STAILQ_FIRST(&(*fa)->fa_list)) != NULL) {
406 		/* Remove file action entry from the queue */
407 		STAILQ_REMOVE_HEAD(&(*fa)->fa_list, fae_list);
408 
409 		/* Deallocate file action entry */
410 		if (fae->fae_action == FAE_OPEN ||
411 		    fae->fae_action == FAE_CHDIR)
412 			free(fae->fae_path);
413 		free(fae);
414 	}
415 
416 	free(*fa);
417 	return (0);
418 }
419 
420 int
421 posix_spawn_file_actions_addopen(posix_spawn_file_actions_t * __restrict fa,
422     int fildes, const char * __restrict path, int oflag, mode_t mode)
423 {
424 	posix_spawn_file_actions_entry_t *fae;
425 	int error;
426 
427 	if (fildes < 0)
428 		return (EBADF);
429 
430 	/* Allocate object */
431 	fae = malloc(sizeof(posix_spawn_file_actions_entry_t));
432 	if (fae == NULL)
433 		return (errno);
434 
435 	/* Set values and store in queue */
436 	fae->fae_action = FAE_OPEN;
437 	fae->fae_path = strdup(path);
438 	if (fae->fae_path == NULL) {
439 		error = errno;
440 		free(fae);
441 		return (error);
442 	}
443 	fae->fae_fildes = fildes;
444 	fae->fae_oflag = oflag;
445 	fae->fae_mode = mode;
446 
447 	STAILQ_INSERT_TAIL(&(*fa)->fa_list, fae, fae_list);
448 	return (0);
449 }
450 
451 int
452 posix_spawn_file_actions_adddup2(posix_spawn_file_actions_t *fa,
453     int fildes, int newfildes)
454 {
455 	posix_spawn_file_actions_entry_t *fae;
456 
457 	if (fildes < 0 || newfildes < 0)
458 		return (EBADF);
459 
460 	/* Allocate object */
461 	fae = malloc(sizeof(posix_spawn_file_actions_entry_t));
462 	if (fae == NULL)
463 		return (errno);
464 
465 	/* Set values and store in queue */
466 	fae->fae_action = FAE_DUP2;
467 	fae->fae_fildes = fildes;
468 	fae->fae_newfildes = newfildes;
469 
470 	STAILQ_INSERT_TAIL(&(*fa)->fa_list, fae, fae_list);
471 	return (0);
472 }
473 
474 int
475 posix_spawn_file_actions_addclose(posix_spawn_file_actions_t *fa,
476     int fildes)
477 {
478 	posix_spawn_file_actions_entry_t *fae;
479 
480 	if (fildes < 0)
481 		return (EBADF);
482 
483 	/* Allocate object */
484 	fae = malloc(sizeof(posix_spawn_file_actions_entry_t));
485 	if (fae == NULL)
486 		return (errno);
487 
488 	/* Set values and store in queue */
489 	fae->fae_action = FAE_CLOSE;
490 	fae->fae_fildes = fildes;
491 
492 	STAILQ_INSERT_TAIL(&(*fa)->fa_list, fae, fae_list);
493 	return (0);
494 }
495 
496 int
497 posix_spawn_file_actions_addchdir_np(posix_spawn_file_actions_t *
498     __restrict fa, const char *__restrict path)
499 {
500 	posix_spawn_file_actions_entry_t *fae;
501 	int error;
502 
503 	fae = malloc(sizeof(posix_spawn_file_actions_entry_t));
504 	if (fae == NULL)
505 		return (errno);
506 
507 	fae->fae_action = FAE_CHDIR;
508 	fae->fae_path = strdup(path);
509 	if (fae->fae_path == NULL) {
510 		error = errno;
511 		free(fae);
512 		return (error);
513 	}
514 
515 	STAILQ_INSERT_TAIL(&(*fa)->fa_list, fae, fae_list);
516 	return (0);
517 }
518 
519 int
520 posix_spawn_file_actions_addfchdir_np(posix_spawn_file_actions_t *__restrict fa,
521     int fildes)
522 {
523 	posix_spawn_file_actions_entry_t *fae;
524 
525 	if (fildes < 0)
526 		return (EBADF);
527 
528 	/* Allocate object */
529 	fae = malloc(sizeof(posix_spawn_file_actions_entry_t));
530 	if (fae == NULL)
531 		return (errno);
532 
533 	fae->fae_action = FAE_FCHDIR;
534 	fae->fae_fildes = fildes;
535 
536 	STAILQ_INSERT_TAIL(&(*fa)->fa_list, fae, fae_list);
537 	return (0);
538 }
539 
540 int
541 posix_spawn_file_actions_addclosefrom_np (posix_spawn_file_actions_t *
542     __restrict fa, int from)
543 {
544 	posix_spawn_file_actions_entry_t *fae;
545 
546 	if (from < 0)
547 		return (EBADF);
548 
549 	/* Allocate object */
550 	fae = malloc(sizeof(posix_spawn_file_actions_entry_t));
551 	if (fae == NULL)
552 		return (errno);
553 
554 	fae->fae_action = FAE_CLOSEFROM;
555 	fae->fae_fildes = from;
556 
557 	STAILQ_INSERT_TAIL(&(*fa)->fa_list, fae, fae_list);
558 	return (0);
559 }
560 
561 /*
562  * Spawn attributes
563  */
564 
565 int
566 posix_spawnattr_init(posix_spawnattr_t *ret)
567 {
568 	posix_spawnattr_t sa;
569 
570 	sa = calloc(1, sizeof(struct __posix_spawnattr));
571 	if (sa == NULL)
572 		return (errno);
573 
574 	/* Set defaults as specified by POSIX, cleared above */
575 	*ret = sa;
576 	return (0);
577 }
578 
579 int
580 posix_spawnattr_destroy(posix_spawnattr_t *sa)
581 {
582 	free(*sa);
583 	return (0);
584 }
585 
586 int
587 posix_spawnattr_getflags(const posix_spawnattr_t * __restrict sa,
588     short * __restrict flags)
589 {
590 	*flags = (*sa)->sa_flags;
591 	return (0);
592 }
593 
594 int
595 posix_spawnattr_getpgroup(const posix_spawnattr_t * __restrict sa,
596     pid_t * __restrict pgroup)
597 {
598 	*pgroup = (*sa)->sa_pgroup;
599 	return (0);
600 }
601 
602 int
603 posix_spawnattr_getschedparam(const posix_spawnattr_t * __restrict sa,
604     struct sched_param * __restrict schedparam)
605 {
606 	*schedparam = (*sa)->sa_schedparam;
607 	return (0);
608 }
609 
610 int
611 posix_spawnattr_getschedpolicy(const posix_spawnattr_t * __restrict sa,
612     int * __restrict schedpolicy)
613 {
614 	*schedpolicy = (*sa)->sa_schedpolicy;
615 	return (0);
616 }
617 
618 int
619 posix_spawnattr_getsigdefault(const posix_spawnattr_t * __restrict sa,
620     sigset_t * __restrict sigdefault)
621 {
622 	*sigdefault = (*sa)->sa_sigdefault;
623 	return (0);
624 }
625 
626 int
627 posix_spawnattr_getsigmask(const posix_spawnattr_t * __restrict sa,
628     sigset_t * __restrict sigmask)
629 {
630 	*sigmask = (*sa)->sa_sigmask;
631 	return (0);
632 }
633 
634 int
635 posix_spawnattr_setflags(posix_spawnattr_t *sa, short flags)
636 {
637 	(*sa)->sa_flags = flags;
638 	return (0);
639 }
640 
641 int
642 posix_spawnattr_setpgroup(posix_spawnattr_t *sa, pid_t pgroup)
643 {
644 	(*sa)->sa_pgroup = pgroup;
645 	return (0);
646 }
647 
648 int
649 posix_spawnattr_setschedparam(posix_spawnattr_t * __restrict sa,
650     const struct sched_param * __restrict schedparam)
651 {
652 	(*sa)->sa_schedparam = *schedparam;
653 	return (0);
654 }
655 
656 int
657 posix_spawnattr_setschedpolicy(posix_spawnattr_t *sa, int schedpolicy)
658 {
659 	(*sa)->sa_schedpolicy = schedpolicy;
660 	return (0);
661 }
662 
663 int
664 posix_spawnattr_setsigdefault(posix_spawnattr_t * __restrict sa,
665     const sigset_t * __restrict sigdefault)
666 {
667 	(*sa)->sa_sigdefault = *sigdefault;
668 	return (0);
669 }
670 
671 int
672 posix_spawnattr_setsigmask(posix_spawnattr_t * __restrict sa,
673     const sigset_t * __restrict sigmask)
674 {
675 	(*sa)->sa_sigmask = *sigmask;
676 	return (0);
677 }
678