xref: /freebsd/sys/kern/kern_procctl.c (revision 716fd348)
1 /*-
2  * Copyright (c) 2014 John Baldwin
3  * Copyright (c) 2014, 2016 The FreeBSD Foundation
4  *
5  * Portions of this software were developed by Konstantin Belousov
6  * under sponsorship from the FreeBSD Foundation.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  */
29 
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
32 
33 #include <sys/param.h>
34 #include <sys/_unrhdr.h>
35 #include <sys/systm.h>
36 #include <sys/capsicum.h>
37 #include <sys/lock.h>
38 #include <sys/mman.h>
39 #include <sys/mutex.h>
40 #include <sys/priv.h>
41 #include <sys/proc.h>
42 #include <sys/procctl.h>
43 #include <sys/sx.h>
44 #include <sys/syscallsubr.h>
45 #include <sys/sysproto.h>
46 #include <sys/wait.h>
47 
48 #include <vm/vm.h>
49 #include <vm/pmap.h>
50 #include <vm/vm_map.h>
51 #include <vm/vm_extern.h>
52 
53 static int
54 protect_setchild(struct thread *td, struct proc *p, int flags)
55 {
56 
57 	PROC_LOCK_ASSERT(p, MA_OWNED);
58 	if (p->p_flag & P_SYSTEM || p_cansched(td, p) != 0)
59 		return (0);
60 	if (flags & PPROT_SET) {
61 		p->p_flag |= P_PROTECTED;
62 		if (flags & PPROT_INHERIT)
63 			p->p_flag2 |= P2_INHERIT_PROTECTED;
64 	} else {
65 		p->p_flag &= ~P_PROTECTED;
66 		p->p_flag2 &= ~P2_INHERIT_PROTECTED;
67 	}
68 	return (1);
69 }
70 
71 static int
72 protect_setchildren(struct thread *td, struct proc *top, int flags)
73 {
74 	struct proc *p;
75 	int ret;
76 
77 	p = top;
78 	ret = 0;
79 	sx_assert(&proctree_lock, SX_LOCKED);
80 	for (;;) {
81 		ret |= protect_setchild(td, p, flags);
82 		PROC_UNLOCK(p);
83 		/*
84 		 * If this process has children, descend to them next,
85 		 * otherwise do any siblings, and if done with this level,
86 		 * follow back up the tree (but not past top).
87 		 */
88 		if (!LIST_EMPTY(&p->p_children))
89 			p = LIST_FIRST(&p->p_children);
90 		else for (;;) {
91 			if (p == top) {
92 				PROC_LOCK(p);
93 				return (ret);
94 			}
95 			if (LIST_NEXT(p, p_sibling)) {
96 				p = LIST_NEXT(p, p_sibling);
97 				break;
98 			}
99 			p = p->p_pptr;
100 		}
101 		PROC_LOCK(p);
102 	}
103 }
104 
105 static int
106 protect_set(struct thread *td, struct proc *p, void *data)
107 {
108 	int error, flags, ret;
109 
110 	flags = *(int *)data;
111 	switch (PPROT_OP(flags)) {
112 	case PPROT_SET:
113 	case PPROT_CLEAR:
114 		break;
115 	default:
116 		return (EINVAL);
117 	}
118 
119 	if ((PPROT_FLAGS(flags) & ~(PPROT_DESCEND | PPROT_INHERIT)) != 0)
120 		return (EINVAL);
121 
122 	error = priv_check(td, PRIV_VM_MADV_PROTECT);
123 	if (error)
124 		return (error);
125 
126 	if (flags & PPROT_DESCEND)
127 		ret = protect_setchildren(td, p, flags);
128 	else
129 		ret = protect_setchild(td, p, flags);
130 	if (ret == 0)
131 		return (EPERM);
132 	return (0);
133 }
134 
135 static int
136 reap_acquire(struct thread *td, struct proc *p, void *data __unused)
137 {
138 
139 	sx_assert(&proctree_lock, SX_XLOCKED);
140 	if (p != td->td_proc)
141 		return (EPERM);
142 	if ((p->p_treeflag & P_TREE_REAPER) != 0)
143 		return (EBUSY);
144 	p->p_treeflag |= P_TREE_REAPER;
145 	/*
146 	 * We do not reattach existing children and the whole tree
147 	 * under them to us, since p->p_reaper already seen them.
148 	 */
149 	return (0);
150 }
151 
152 static int
153 reap_release(struct thread *td, struct proc *p, void *data __unused)
154 {
155 
156 	sx_assert(&proctree_lock, SX_XLOCKED);
157 	if (p != td->td_proc)
158 		return (EPERM);
159 	if (p == initproc)
160 		return (EINVAL);
161 	if ((p->p_treeflag & P_TREE_REAPER) == 0)
162 		return (EINVAL);
163 	reaper_abandon_children(p, false);
164 	return (0);
165 }
166 
167 static int
168 reap_status(struct thread *td, struct proc *p, void *data)
169 {
170 	struct proc *reap, *p2, *first_p;
171 	struct procctl_reaper_status *rs;
172 
173 	rs = data;
174 	sx_assert(&proctree_lock, SX_LOCKED);
175 	if ((p->p_treeflag & P_TREE_REAPER) == 0) {
176 		reap = p->p_reaper;
177 	} else {
178 		reap = p;
179 		rs->rs_flags |= REAPER_STATUS_OWNED;
180 	}
181 	if (reap == initproc)
182 		rs->rs_flags |= REAPER_STATUS_REALINIT;
183 	rs->rs_reaper = reap->p_pid;
184 	rs->rs_descendants = 0;
185 	rs->rs_children = 0;
186 	if (!LIST_EMPTY(&reap->p_reaplist)) {
187 		first_p = LIST_FIRST(&reap->p_children);
188 		if (first_p == NULL)
189 			first_p = LIST_FIRST(&reap->p_reaplist);
190 		rs->rs_pid = first_p->p_pid;
191 		LIST_FOREACH(p2, &reap->p_reaplist, p_reapsibling) {
192 			if (proc_realparent(p2) == reap)
193 				rs->rs_children++;
194 			rs->rs_descendants++;
195 		}
196 	} else {
197 		rs->rs_pid = -1;
198 	}
199 	return (0);
200 }
201 
202 static int
203 reap_getpids(struct thread *td, struct proc *p, void *data)
204 {
205 	struct proc *reap, *p2;
206 	struct procctl_reaper_pidinfo *pi, *pip;
207 	struct procctl_reaper_pids *rp;
208 	u_int i, n;
209 	int error;
210 
211 	rp = data;
212 	sx_assert(&proctree_lock, SX_LOCKED);
213 	PROC_UNLOCK(p);
214 	reap = (p->p_treeflag & P_TREE_REAPER) == 0 ? p->p_reaper : p;
215 	n = i = 0;
216 	error = 0;
217 	LIST_FOREACH(p2, &reap->p_reaplist, p_reapsibling)
218 		n++;
219 	sx_unlock(&proctree_lock);
220 	if (rp->rp_count < n)
221 		n = rp->rp_count;
222 	pi = malloc(n * sizeof(*pi), M_TEMP, M_WAITOK);
223 	sx_slock(&proctree_lock);
224 	LIST_FOREACH(p2, &reap->p_reaplist, p_reapsibling) {
225 		if (i == n)
226 			break;
227 		pip = &pi[i];
228 		bzero(pip, sizeof(*pip));
229 		pip->pi_pid = p2->p_pid;
230 		pip->pi_subtree = p2->p_reapsubtree;
231 		pip->pi_flags = REAPER_PIDINFO_VALID;
232 		if (proc_realparent(p2) == reap)
233 			pip->pi_flags |= REAPER_PIDINFO_CHILD;
234 		if ((p2->p_treeflag & P_TREE_REAPER) != 0)
235 			pip->pi_flags |= REAPER_PIDINFO_REAPER;
236 		i++;
237 	}
238 	sx_sunlock(&proctree_lock);
239 	error = copyout(pi, rp->rp_pids, i * sizeof(*pi));
240 	free(pi, M_TEMP);
241 	sx_slock(&proctree_lock);
242 	PROC_LOCK(p);
243 	return (error);
244 }
245 
246 static void
247 reap_kill_proc_relock(struct proc *p, int xlocked)
248 {
249 	PROC_UNLOCK(p);
250 	if (xlocked)
251 		sx_xlock(&proctree_lock);
252 	else
253 		sx_slock(&proctree_lock);
254 	PROC_LOCK(p);
255 }
256 
257 static bool
258 reap_kill_proc_locked(struct thread *td, struct proc *p2,
259     ksiginfo_t *ksi, struct procctl_reaper_kill *rk, int *error)
260 {
261 	int error1, r, xlocked;
262 	bool need_stop;
263 
264 	PROC_LOCK_ASSERT(p2, MA_OWNED);
265 	PROC_ASSERT_HELD(p2);
266 
267 	error1 = p_cansignal(td, p2, rk->rk_sig);
268 	if (error1 != 0) {
269 		if (*error == ESRCH) {
270 			rk->rk_fpid = p2->p_pid;
271 			*error = error1;
272 		}
273 		return (true);
274 	}
275 
276 	/*
277 	 * The need_stop indicates if the target process needs to be
278 	 * suspended before being signalled.  This is needed when we
279 	 * guarantee that all processes in subtree are signalled,
280 	 * avoiding the race with some process not yet fully linked
281 	 * into all structures during fork, ignored by iterator, and
282 	 * then escaping signalling.
283 	 *
284 	 * If need_stop is true, then reap_kill_proc() returns true if
285 	 * the process was successfully stopped and signalled, and
286 	 * false if stopping failed and the signal was not sent.
287 	 *
288 	 * The thread cannot usefully stop itself anyway, and if other
289 	 * thread of the current process forks while the current
290 	 * thread signals the whole subtree, it is an application
291 	 * race.
292 	 */
293 	need_stop = p2 != td->td_proc &&
294 	    (p2->p_flag & (P_KPROC | P_SYSTEM)) == 0 &&
295 	    (rk->rk_flags & REAPER_KILL_CHILDREN) == 0;
296 
297 	if (need_stop) {
298 		if (P_SHOULDSTOP(p2) == P_STOPPED_SINGLE)
299 			return (false);	/* retry later */
300 		xlocked = sx_xlocked(&proctree_lock);
301 		sx_unlock(&proctree_lock);
302 		r = thread_single(p2, SINGLE_ALLPROC);
303 		if (r != 0) {
304 			reap_kill_proc_relock(p2, xlocked);
305 			return (false);
306 		}
307 	}
308 
309 	pksignal(p2, rk->rk_sig, ksi);
310 	rk->rk_killed++;
311 	*error = error1;
312 
313 	if (need_stop) {
314 		reap_kill_proc_relock(p2, xlocked);
315 		thread_single_end(p2, SINGLE_ALLPROC);
316 	}
317 	return (true);
318 }
319 
320 static bool
321 reap_kill_proc(struct thread *td, struct proc *p2, ksiginfo_t *ksi,
322     struct procctl_reaper_kill *rk, int *error)
323 {
324 	bool res;
325 
326 	res = true;
327 	PROC_LOCK(p2);
328 	if ((p2->p_flag & P_WEXIT) == 0) {
329 		_PHOLD_LITE(p2);
330 		res = reap_kill_proc_locked(td, p2, ksi, rk, error);
331 		_PRELE(p2);
332 	}
333 	PROC_UNLOCK(p2);
334 	return (res);
335 }
336 
337 struct reap_kill_tracker {
338 	struct proc *parent;
339 	TAILQ_ENTRY(reap_kill_tracker) link;
340 };
341 
342 TAILQ_HEAD(reap_kill_tracker_head, reap_kill_tracker);
343 
344 static void
345 reap_kill_sched(struct reap_kill_tracker_head *tracker, struct proc *p2)
346 {
347 	struct reap_kill_tracker *t;
348 
349 	t = malloc(sizeof(struct reap_kill_tracker), M_TEMP, M_WAITOK);
350 	t->parent = p2;
351 	TAILQ_INSERT_TAIL(tracker, t, link);
352 }
353 
354 static void
355 reap_kill_children(struct thread *td, struct proc *reaper,
356     struct procctl_reaper_kill *rk, ksiginfo_t *ksi, int *error)
357 {
358 	struct proc *p2;
359 
360 	LIST_FOREACH(p2, &reaper->p_children, p_sibling) {
361 		(void)reap_kill_proc(td, p2, ksi, rk, error);
362 		/*
363 		 * Do not end the loop on error, signal everything we
364 		 * can.
365 		 */
366 	}
367 }
368 
369 static bool
370 reap_kill_subtree_once(struct thread *td, struct proc *p, struct proc *reaper,
371     struct procctl_reaper_kill *rk, ksiginfo_t *ksi, int *error,
372     struct unrhdr *pids)
373 {
374 	struct reap_kill_tracker_head tracker;
375 	struct reap_kill_tracker *t;
376 	struct proc *p2;
377 	bool res;
378 
379 	res = false;
380 	TAILQ_INIT(&tracker);
381 	reap_kill_sched(&tracker, reaper);
382 	while ((t = TAILQ_FIRST(&tracker)) != NULL) {
383 		MPASS((t->parent->p_treeflag & P_TREE_REAPER) != 0);
384 		TAILQ_REMOVE(&tracker, t, link);
385 		LIST_FOREACH(p2, &t->parent->p_reaplist, p_reapsibling) {
386 			if (t->parent == reaper &&
387 			    (rk->rk_flags & REAPER_KILL_SUBTREE) != 0 &&
388 			    p2->p_reapsubtree != rk->rk_subtree)
389 				continue;
390 			if ((p2->p_treeflag & P_TREE_REAPER) != 0)
391 				reap_kill_sched(&tracker, p2);
392 			if (alloc_unr_specific(pids, p2->p_pid) != p2->p_pid)
393 				continue;
394 			if (!reap_kill_proc(td, p2, ksi, rk, error))
395 				free_unr(pids, p2->p_pid);
396 			res = true;
397 		}
398 		free(t, M_TEMP);
399 	}
400 	return (res);
401 }
402 
403 static void
404 reap_kill_subtree(struct thread *td, struct proc *p, struct proc *reaper,
405     struct procctl_reaper_kill *rk, ksiginfo_t *ksi, int *error)
406 {
407 	struct unrhdr pids;
408 
409 	/*
410 	 * pids records processes which were already signalled, to
411 	 * avoid doubling signals to them if iteration needs to be
412 	 * repeated.
413 	 */
414 	init_unrhdr(&pids, 1, PID_MAX, UNR_NO_MTX);
415 	while (reap_kill_subtree_once(td, p, reaper, rk, ksi, error, &pids))
416 	       ;
417 	clean_unrhdr(&pids);
418 	clear_unrhdr(&pids);
419 }
420 
421 static bool
422 reap_kill_sapblk(struct thread *td __unused, void *data)
423 {
424 	struct procctl_reaper_kill *rk;
425 
426 	rk = data;
427 	return ((rk->rk_flags & REAPER_KILL_CHILDREN) == 0);
428 }
429 
430 static int
431 reap_kill(struct thread *td, struct proc *p, void *data)
432 {
433 	struct proc *reaper;
434 	ksiginfo_t ksi;
435 	struct procctl_reaper_kill *rk;
436 	int error;
437 
438 	rk = data;
439 	sx_assert(&proctree_lock, SX_LOCKED);
440 	if (IN_CAPABILITY_MODE(td))
441 		return (ECAPMODE);
442 	if (rk->rk_sig <= 0 || rk->rk_sig > _SIG_MAXSIG ||
443 	    (rk->rk_flags & ~(REAPER_KILL_CHILDREN |
444 	    REAPER_KILL_SUBTREE)) != 0 || (rk->rk_flags &
445 	    (REAPER_KILL_CHILDREN | REAPER_KILL_SUBTREE)) ==
446 	    (REAPER_KILL_CHILDREN | REAPER_KILL_SUBTREE))
447 		return (EINVAL);
448 	PROC_UNLOCK(p);
449 	reaper = (p->p_treeflag & P_TREE_REAPER) == 0 ? p->p_reaper : p;
450 	ksiginfo_init(&ksi);
451 	ksi.ksi_signo = rk->rk_sig;
452 	ksi.ksi_code = SI_USER;
453 	ksi.ksi_pid = td->td_proc->p_pid;
454 	ksi.ksi_uid = td->td_ucred->cr_ruid;
455 	error = ESRCH;
456 	rk->rk_killed = 0;
457 	rk->rk_fpid = -1;
458 	if ((rk->rk_flags & REAPER_KILL_CHILDREN) != 0) {
459 		reap_kill_children(td, reaper, rk, &ksi, &error);
460 	} else {
461 		reap_kill_subtree(td, p, reaper, rk, &ksi, &error);
462 	}
463 	PROC_LOCK(p);
464 	return (error);
465 }
466 
467 static int
468 trace_ctl(struct thread *td, struct proc *p, void *data)
469 {
470 	int state;
471 
472 	PROC_LOCK_ASSERT(p, MA_OWNED);
473 	state = *(int *)data;
474 
475 	/*
476 	 * Ktrace changes p_traceflag from or to zero under the
477 	 * process lock, so the test does not need to acquire ktrace
478 	 * mutex.
479 	 */
480 	if ((p->p_flag & P_TRACED) != 0 || p->p_traceflag != 0)
481 		return (EBUSY);
482 
483 	switch (state) {
484 	case PROC_TRACE_CTL_ENABLE:
485 		if (td->td_proc != p)
486 			return (EPERM);
487 		p->p_flag2 &= ~(P2_NOTRACE | P2_NOTRACE_EXEC);
488 		break;
489 	case PROC_TRACE_CTL_DISABLE_EXEC:
490 		p->p_flag2 |= P2_NOTRACE_EXEC | P2_NOTRACE;
491 		break;
492 	case PROC_TRACE_CTL_DISABLE:
493 		if ((p->p_flag2 & P2_NOTRACE_EXEC) != 0) {
494 			KASSERT((p->p_flag2 & P2_NOTRACE) != 0,
495 			    ("dandling P2_NOTRACE_EXEC"));
496 			if (td->td_proc != p)
497 				return (EPERM);
498 			p->p_flag2 &= ~P2_NOTRACE_EXEC;
499 		} else {
500 			p->p_flag2 |= P2_NOTRACE;
501 		}
502 		break;
503 	default:
504 		return (EINVAL);
505 	}
506 	return (0);
507 }
508 
509 static int
510 trace_status(struct thread *td, struct proc *p, void *data)
511 {
512 	int *status;
513 
514 	status = data;
515 	if ((p->p_flag2 & P2_NOTRACE) != 0) {
516 		KASSERT((p->p_flag & P_TRACED) == 0,
517 		    ("%d traced but tracing disabled", p->p_pid));
518 		*status = -1;
519 	} else if ((p->p_flag & P_TRACED) != 0) {
520 		*status = p->p_pptr->p_pid;
521 	} else {
522 		*status = 0;
523 	}
524 	return (0);
525 }
526 
527 static int
528 trapcap_ctl(struct thread *td, struct proc *p, void *data)
529 {
530 	int state;
531 
532 	PROC_LOCK_ASSERT(p, MA_OWNED);
533 	state = *(int *)data;
534 
535 	switch (state) {
536 	case PROC_TRAPCAP_CTL_ENABLE:
537 		p->p_flag2 |= P2_TRAPCAP;
538 		break;
539 	case PROC_TRAPCAP_CTL_DISABLE:
540 		p->p_flag2 &= ~P2_TRAPCAP;
541 		break;
542 	default:
543 		return (EINVAL);
544 	}
545 	return (0);
546 }
547 
548 static int
549 trapcap_status(struct thread *td, struct proc *p, void *data)
550 {
551 	int *status;
552 
553 	status = data;
554 	*status = (p->p_flag2 & P2_TRAPCAP) != 0 ? PROC_TRAPCAP_CTL_ENABLE :
555 	    PROC_TRAPCAP_CTL_DISABLE;
556 	return (0);
557 }
558 
559 static int
560 no_new_privs_ctl(struct thread *td, struct proc *p, void *data)
561 {
562 	int state;
563 
564 	PROC_LOCK_ASSERT(p, MA_OWNED);
565 	state = *(int *)data;
566 
567 	if (state != PROC_NO_NEW_PRIVS_ENABLE)
568 		return (EINVAL);
569 	p->p_flag2 |= P2_NO_NEW_PRIVS;
570 	return (0);
571 }
572 
573 static int
574 no_new_privs_status(struct thread *td, struct proc *p, void *data)
575 {
576 
577 	*(int *)data = (p->p_flag2 & P2_NO_NEW_PRIVS) != 0 ?
578 	    PROC_NO_NEW_PRIVS_ENABLE : PROC_NO_NEW_PRIVS_DISABLE;
579 	return (0);
580 }
581 
582 static int
583 protmax_ctl(struct thread *td, struct proc *p, void *data)
584 {
585 	int state;
586 
587 	PROC_LOCK_ASSERT(p, MA_OWNED);
588 	state = *(int *)data;
589 
590 	switch (state) {
591 	case PROC_PROTMAX_FORCE_ENABLE:
592 		p->p_flag2 &= ~P2_PROTMAX_DISABLE;
593 		p->p_flag2 |= P2_PROTMAX_ENABLE;
594 		break;
595 	case PROC_PROTMAX_FORCE_DISABLE:
596 		p->p_flag2 |= P2_PROTMAX_DISABLE;
597 		p->p_flag2 &= ~P2_PROTMAX_ENABLE;
598 		break;
599 	case PROC_PROTMAX_NOFORCE:
600 		p->p_flag2 &= ~(P2_PROTMAX_ENABLE | P2_PROTMAX_DISABLE);
601 		break;
602 	default:
603 		return (EINVAL);
604 	}
605 	return (0);
606 }
607 
608 static int
609 protmax_status(struct thread *td, struct proc *p, void *data)
610 {
611 	int d;
612 
613 	switch (p->p_flag2 & (P2_PROTMAX_ENABLE | P2_PROTMAX_DISABLE)) {
614 	case 0:
615 		d = PROC_PROTMAX_NOFORCE;
616 		break;
617 	case P2_PROTMAX_ENABLE:
618 		d = PROC_PROTMAX_FORCE_ENABLE;
619 		break;
620 	case P2_PROTMAX_DISABLE:
621 		d = PROC_PROTMAX_FORCE_DISABLE;
622 		break;
623 	}
624 	if (kern_mmap_maxprot(p, PROT_READ) == PROT_READ)
625 		d |= PROC_PROTMAX_ACTIVE;
626 	*(int *)data = d;
627 	return (0);
628 }
629 
630 static int
631 aslr_ctl(struct thread *td, struct proc *p, void *data)
632 {
633 	int state;
634 
635 	PROC_LOCK_ASSERT(p, MA_OWNED);
636 	state = *(int *)data;
637 
638 	switch (state) {
639 	case PROC_ASLR_FORCE_ENABLE:
640 		p->p_flag2 &= ~P2_ASLR_DISABLE;
641 		p->p_flag2 |= P2_ASLR_ENABLE;
642 		break;
643 	case PROC_ASLR_FORCE_DISABLE:
644 		p->p_flag2 |= P2_ASLR_DISABLE;
645 		p->p_flag2 &= ~P2_ASLR_ENABLE;
646 		break;
647 	case PROC_ASLR_NOFORCE:
648 		p->p_flag2 &= ~(P2_ASLR_ENABLE | P2_ASLR_DISABLE);
649 		break;
650 	default:
651 		return (EINVAL);
652 	}
653 	return (0);
654 }
655 
656 static int
657 aslr_status(struct thread *td, struct proc *p, void *data)
658 {
659 	struct vmspace *vm;
660 	int d;
661 
662 	switch (p->p_flag2 & (P2_ASLR_ENABLE | P2_ASLR_DISABLE)) {
663 	case 0:
664 		d = PROC_ASLR_NOFORCE;
665 		break;
666 	case P2_ASLR_ENABLE:
667 		d = PROC_ASLR_FORCE_ENABLE;
668 		break;
669 	case P2_ASLR_DISABLE:
670 		d = PROC_ASLR_FORCE_DISABLE;
671 		break;
672 	}
673 	if ((p->p_flag & P_WEXIT) == 0) {
674 		_PHOLD(p);
675 		PROC_UNLOCK(p);
676 		vm = vmspace_acquire_ref(p);
677 		if (vm != NULL) {
678 			if ((vm->vm_map.flags & MAP_ASLR) != 0)
679 				d |= PROC_ASLR_ACTIVE;
680 			vmspace_free(vm);
681 		}
682 		PROC_LOCK(p);
683 		_PRELE(p);
684 	}
685 	*(int *)data = d;
686 	return (0);
687 }
688 
689 static int
690 stackgap_ctl(struct thread *td, struct proc *p, void *data)
691 {
692 	int state;
693 
694 	PROC_LOCK_ASSERT(p, MA_OWNED);
695 	state = *(int *)data;
696 
697 	if ((state & ~(PROC_STACKGAP_ENABLE | PROC_STACKGAP_DISABLE |
698 	    PROC_STACKGAP_ENABLE_EXEC | PROC_STACKGAP_DISABLE_EXEC)) != 0)
699 		return (EINVAL);
700 	switch (state & (PROC_STACKGAP_ENABLE | PROC_STACKGAP_DISABLE)) {
701 	case PROC_STACKGAP_ENABLE:
702 		if ((p->p_flag2 & P2_STKGAP_DISABLE) != 0)
703 			return (EINVAL);
704 		break;
705 	case PROC_STACKGAP_DISABLE:
706 		p->p_flag2 |= P2_STKGAP_DISABLE;
707 		break;
708 	case 0:
709 		break;
710 	default:
711 		return (EINVAL);
712 	}
713 	switch (state & (PROC_STACKGAP_ENABLE_EXEC |
714 	    PROC_STACKGAP_DISABLE_EXEC)) {
715 	case PROC_STACKGAP_ENABLE_EXEC:
716 		p->p_flag2 &= ~P2_STKGAP_DISABLE_EXEC;
717 		break;
718 	case PROC_STACKGAP_DISABLE_EXEC:
719 		p->p_flag2 |= P2_STKGAP_DISABLE_EXEC;
720 		break;
721 	case 0:
722 		break;
723 	default:
724 		return (EINVAL);
725 	}
726 	return (0);
727 }
728 
729 static int
730 stackgap_status(struct thread *td, struct proc *p, void *data)
731 {
732 	int d;
733 
734 	PROC_LOCK_ASSERT(p, MA_OWNED);
735 
736 	d = (p->p_flag2 & P2_STKGAP_DISABLE) != 0 ? PROC_STACKGAP_DISABLE :
737 	    PROC_STACKGAP_ENABLE;
738 	d |= (p->p_flag2 & P2_STKGAP_DISABLE_EXEC) != 0 ?
739 	    PROC_STACKGAP_DISABLE_EXEC : PROC_STACKGAP_ENABLE_EXEC;
740 	*(int *)data = d;
741 	return (0);
742 }
743 
744 static int
745 wxmap_ctl(struct thread *td, struct proc *p, void *data)
746 {
747 	struct vmspace *vm;
748 	vm_map_t map;
749 	int state;
750 
751 	PROC_LOCK_ASSERT(p, MA_OWNED);
752 	if ((p->p_flag & P_WEXIT) != 0)
753 		return (ESRCH);
754 	state = *(int *)data;
755 
756 	switch (state) {
757 	case PROC_WX_MAPPINGS_PERMIT:
758 		p->p_flag2 |= P2_WXORX_DISABLE;
759 		_PHOLD(p);
760 		PROC_UNLOCK(p);
761 		vm = vmspace_acquire_ref(p);
762 		if (vm != NULL) {
763 			map = &vm->vm_map;
764 			vm_map_lock(map);
765 			map->flags &= ~MAP_WXORX;
766 			vm_map_unlock(map);
767 			vmspace_free(vm);
768 		}
769 		PROC_LOCK(p);
770 		_PRELE(p);
771 		break;
772 	case PROC_WX_MAPPINGS_DISALLOW_EXEC:
773 		p->p_flag2 |= P2_WXORX_ENABLE_EXEC;
774 		break;
775 	default:
776 		return (EINVAL);
777 	}
778 
779 	return (0);
780 }
781 
782 static int
783 wxmap_status(struct thread *td, struct proc *p, void *data)
784 {
785 	struct vmspace *vm;
786 	int d;
787 
788 	PROC_LOCK_ASSERT(p, MA_OWNED);
789 	if ((p->p_flag & P_WEXIT) != 0)
790 		return (ESRCH);
791 
792 	d = 0;
793 	if ((p->p_flag2 & P2_WXORX_DISABLE) != 0)
794 		d |= PROC_WX_MAPPINGS_PERMIT;
795 	if ((p->p_flag2 & P2_WXORX_ENABLE_EXEC) != 0)
796 		d |= PROC_WX_MAPPINGS_DISALLOW_EXEC;
797 	_PHOLD(p);
798 	PROC_UNLOCK(p);
799 	vm = vmspace_acquire_ref(p);
800 	if (vm != NULL) {
801 		if ((vm->vm_map.flags & MAP_WXORX) != 0)
802 			d |= PROC_WXORX_ENFORCE;
803 		vmspace_free(vm);
804 	}
805 	PROC_LOCK(p);
806 	_PRELE(p);
807 	*(int *)data = d;
808 	return (0);
809 }
810 
811 static int
812 pdeathsig_ctl(struct thread *td, struct proc *p, void *data)
813 {
814 	int signum;
815 
816 	signum = *(int *)data;
817 	if (p != td->td_proc || (signum != 0 && !_SIG_VALID(signum)))
818 		return (EINVAL);
819 	p->p_pdeathsig = signum;
820 	return (0);
821 }
822 
823 static int
824 pdeathsig_status(struct thread *td, struct proc *p, void *data)
825 {
826 	if (p != td->td_proc)
827 		return (EINVAL);
828 	*(int *)data = p->p_pdeathsig;
829 	return (0);
830 }
831 
832 enum {
833 	PCTL_SLOCKED,
834 	PCTL_XLOCKED,
835 	PCTL_UNLOCKED,
836 };
837 
838 struct procctl_cmd_info {
839 	int lock_tree;
840 	bool one_proc : 1;
841 	bool esrch_is_einval : 1;
842 	bool copyout_on_error : 1;
843 	bool no_nonnull_data : 1;
844 	bool need_candebug : 1;
845 	int copyin_sz;
846 	int copyout_sz;
847 	int (*exec)(struct thread *, struct proc *, void *);
848 	bool (*sapblk)(struct thread *, void *);
849 };
850 static const struct procctl_cmd_info procctl_cmds_info[] = {
851 	[PROC_SPROTECT] =
852 	    { .lock_tree = PCTL_SLOCKED, .one_proc = false,
853 	      .esrch_is_einval = false, .no_nonnull_data = false,
854 	      .need_candebug = false,
855 	      .copyin_sz = sizeof(int), .copyout_sz = 0,
856 	      .exec = protect_set, .copyout_on_error = false, },
857 	[PROC_REAP_ACQUIRE] =
858 	    { .lock_tree = PCTL_XLOCKED, .one_proc = true,
859 	      .esrch_is_einval = false, .no_nonnull_data = true,
860 	      .need_candebug = false,
861 	      .copyin_sz = 0, .copyout_sz = 0,
862 	      .exec = reap_acquire, .copyout_on_error = false, },
863 	[PROC_REAP_RELEASE] =
864 	    { .lock_tree = PCTL_XLOCKED, .one_proc = true,
865 	      .esrch_is_einval = false, .no_nonnull_data = true,
866 	      .need_candebug = false,
867 	      .copyin_sz = 0, .copyout_sz = 0,
868 	      .exec = reap_release, .copyout_on_error = false, },
869 	[PROC_REAP_STATUS] =
870 	    { .lock_tree = PCTL_SLOCKED, .one_proc = true,
871 	      .esrch_is_einval = false, .no_nonnull_data = false,
872 	      .need_candebug = false,
873 	      .copyin_sz = 0,
874 	      .copyout_sz = sizeof(struct procctl_reaper_status),
875 	      .exec = reap_status, .copyout_on_error = false, },
876 	[PROC_REAP_GETPIDS] =
877 	    { .lock_tree = PCTL_SLOCKED, .one_proc = true,
878 	      .esrch_is_einval = false, .no_nonnull_data = false,
879 	      .need_candebug = false,
880 	      .copyin_sz = sizeof(struct procctl_reaper_pids),
881 	      .copyout_sz = 0,
882 	      .exec = reap_getpids, .copyout_on_error = false, },
883 	[PROC_REAP_KILL] =
884 	    { .lock_tree = PCTL_SLOCKED, .one_proc = true,
885 	      .esrch_is_einval = false, .no_nonnull_data = false,
886 	      .need_candebug = false,
887 	      .copyin_sz = sizeof(struct procctl_reaper_kill),
888 	      .copyout_sz = sizeof(struct procctl_reaper_kill),
889 	      .exec = reap_kill, .copyout_on_error = true,
890 	      .sapblk = reap_kill_sapblk, },
891 	[PROC_TRACE_CTL] =
892 	    { .lock_tree = PCTL_SLOCKED, .one_proc = false,
893 	      .esrch_is_einval = false, .no_nonnull_data = false,
894 	      .need_candebug = true,
895 	      .copyin_sz = sizeof(int), .copyout_sz = 0,
896 	      .exec = trace_ctl, .copyout_on_error = false, },
897 	[PROC_TRACE_STATUS] =
898 	    { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
899 	      .esrch_is_einval = false, .no_nonnull_data = false,
900 	      .need_candebug = false,
901 	      .copyin_sz = 0, .copyout_sz = sizeof(int),
902 	      .exec = trace_status, .copyout_on_error = false, },
903 	[PROC_TRAPCAP_CTL] =
904 	    { .lock_tree = PCTL_SLOCKED, .one_proc = false,
905 	      .esrch_is_einval = false, .no_nonnull_data = false,
906 	      .need_candebug = true,
907 	      .copyin_sz = sizeof(int), .copyout_sz = 0,
908 	      .exec = trapcap_ctl, .copyout_on_error = false, },
909 	[PROC_TRAPCAP_STATUS] =
910 	    { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
911 	      .esrch_is_einval = false, .no_nonnull_data = false,
912 	      .need_candebug = false,
913 	      .copyin_sz = 0, .copyout_sz = sizeof(int),
914 	      .exec = trapcap_status, .copyout_on_error = false, },
915 	[PROC_PDEATHSIG_CTL] =
916 	    { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
917 	      .esrch_is_einval = true, .no_nonnull_data = false,
918 	      .need_candebug = false,
919 	      .copyin_sz = sizeof(int), .copyout_sz = 0,
920 	      .exec = pdeathsig_ctl, .copyout_on_error = false, },
921 	[PROC_PDEATHSIG_STATUS] =
922 	    { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
923 	      .esrch_is_einval = true, .no_nonnull_data = false,
924 	      .need_candebug = false,
925 	      .copyin_sz = 0, .copyout_sz = sizeof(int),
926 	      .exec = pdeathsig_status, .copyout_on_error = false, },
927 	[PROC_ASLR_CTL] =
928 	    { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
929 	      .esrch_is_einval = false, .no_nonnull_data = false,
930 	      .need_candebug = true,
931 	      .copyin_sz = sizeof(int), .copyout_sz = 0,
932 	      .exec = aslr_ctl, .copyout_on_error = false, },
933 	[PROC_ASLR_STATUS] =
934 	    { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
935 	      .esrch_is_einval = false, .no_nonnull_data = false,
936 	      .need_candebug = false,
937 	      .copyin_sz = 0, .copyout_sz = sizeof(int),
938 	      .exec = aslr_status, .copyout_on_error = false, },
939 	[PROC_PROTMAX_CTL] =
940 	    { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
941 	      .esrch_is_einval = false, .no_nonnull_data = false,
942 	      .need_candebug = true,
943 	      .copyin_sz = sizeof(int), .copyout_sz = 0,
944 	      .exec = protmax_ctl, .copyout_on_error = false, },
945 	[PROC_PROTMAX_STATUS] =
946 	    { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
947 	      .esrch_is_einval = false, .no_nonnull_data = false,
948 	      .need_candebug = false,
949 	      .copyin_sz = 0, .copyout_sz = sizeof(int),
950 	      .exec = protmax_status, .copyout_on_error = false, },
951 	[PROC_STACKGAP_CTL] =
952 	    { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
953 	      .esrch_is_einval = false, .no_nonnull_data = false,
954 	      .need_candebug = true,
955 	      .copyin_sz = sizeof(int), .copyout_sz = 0,
956 	      .exec = stackgap_ctl, .copyout_on_error = false, },
957 	[PROC_STACKGAP_STATUS] =
958 	    { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
959 	      .esrch_is_einval = false, .no_nonnull_data = false,
960 	      .need_candebug = false,
961 	      .copyin_sz = 0, .copyout_sz = sizeof(int),
962 	      .exec = stackgap_status, .copyout_on_error = false, },
963 	[PROC_NO_NEW_PRIVS_CTL] =
964 	    { .lock_tree = PCTL_SLOCKED, .one_proc = true,
965 	      .esrch_is_einval = false, .no_nonnull_data = false,
966 	      .need_candebug = true,
967 	      .copyin_sz = sizeof(int), .copyout_sz = 0,
968 	      .exec = no_new_privs_ctl, .copyout_on_error = false, },
969 	[PROC_NO_NEW_PRIVS_STATUS] =
970 	    { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
971 	      .esrch_is_einval = false, .no_nonnull_data = false,
972 	      .need_candebug = false,
973 	      .copyin_sz = 0, .copyout_sz = sizeof(int),
974 	      .exec = no_new_privs_status, .copyout_on_error = false, },
975 	[PROC_WXMAP_CTL] =
976 	    { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
977 	      .esrch_is_einval = false, .no_nonnull_data = false,
978 	      .need_candebug = true,
979 	      .copyin_sz = sizeof(int), .copyout_sz = 0,
980 	      .exec = wxmap_ctl, .copyout_on_error = false, },
981 	[PROC_WXMAP_STATUS] =
982 	    { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
983 	      .esrch_is_einval = false, .no_nonnull_data = false,
984 	      .need_candebug = false,
985 	      .copyin_sz = 0, .copyout_sz = sizeof(int),
986 	      .exec = wxmap_status, .copyout_on_error = false, },
987 };
988 
989 int
990 sys_procctl(struct thread *td, struct procctl_args *uap)
991 {
992 	union {
993 		struct procctl_reaper_status rs;
994 		struct procctl_reaper_pids rp;
995 		struct procctl_reaper_kill rk;
996 		int flags;
997 	} x;
998 	const struct procctl_cmd_info *cmd_info;
999 	int error, error1;
1000 
1001 	if (uap->com >= PROC_PROCCTL_MD_MIN)
1002 		return (cpu_procctl(td, uap->idtype, uap->id,
1003 		    uap->com, uap->data));
1004 	if (uap->com == 0 || uap->com >= nitems(procctl_cmds_info))
1005 		return (EINVAL);
1006 	cmd_info = &procctl_cmds_info[uap->com];
1007 	bzero(&x, sizeof(x));
1008 
1009 	if (cmd_info->copyin_sz > 0) {
1010 		error = copyin(uap->data, &x, cmd_info->copyin_sz);
1011 		if (error != 0)
1012 			return (error);
1013 	} else if (cmd_info->no_nonnull_data && uap->data != NULL) {
1014 		return (EINVAL);
1015 	}
1016 
1017 	error = kern_procctl(td, uap->idtype, uap->id, uap->com, &x);
1018 
1019 	if (cmd_info->copyout_sz > 0 && (error == 0 ||
1020 	    cmd_info->copyout_on_error)) {
1021 		error1 = copyout(&x, uap->data, cmd_info->copyout_sz);
1022 		if (error == 0)
1023 			error = error1;
1024 	}
1025 	return (error);
1026 }
1027 
1028 static int
1029 kern_procctl_single(struct thread *td, struct proc *p, int com, void *data)
1030 {
1031 
1032 	PROC_LOCK_ASSERT(p, MA_OWNED);
1033 	return (procctl_cmds_info[com].exec(td, p, data));
1034 }
1035 
1036 int
1037 kern_procctl(struct thread *td, idtype_t idtype, id_t id, int com, void *data)
1038 {
1039 	struct pgrp *pg;
1040 	struct proc *p;
1041 	const struct procctl_cmd_info *cmd_info;
1042 	int error, first_error, ok;
1043 	bool sapblk;
1044 
1045 	MPASS(com > 0 && com < nitems(procctl_cmds_info));
1046 	cmd_info = &procctl_cmds_info[com];
1047 	if (idtype != P_PID && cmd_info->one_proc)
1048 		return (EINVAL);
1049 
1050 	sapblk = false;
1051 	if (cmd_info->sapblk != NULL) {
1052 		sapblk = cmd_info->sapblk(td, data);
1053 		if (sapblk)
1054 			stop_all_proc_block();
1055 	}
1056 
1057 	switch (cmd_info->lock_tree) {
1058 	case PCTL_XLOCKED:
1059 		sx_xlock(&proctree_lock);
1060 		break;
1061 	case PCTL_SLOCKED:
1062 		sx_slock(&proctree_lock);
1063 		break;
1064 	default:
1065 		break;
1066 	}
1067 
1068 	switch (idtype) {
1069 	case P_PID:
1070 		if (id == 0) {
1071 			p = td->td_proc;
1072 			error = 0;
1073 			PROC_LOCK(p);
1074 		} else {
1075 			p = pfind(id);
1076 			if (p == NULL) {
1077 				error = cmd_info->esrch_is_einval ?
1078 				    EINVAL : ESRCH;
1079 				break;
1080 			}
1081 			error = cmd_info->need_candebug ? p_candebug(td, p) :
1082 			    p_cansee(td, p);
1083 		}
1084 		if (error == 0)
1085 			error = kern_procctl_single(td, p, com, data);
1086 		PROC_UNLOCK(p);
1087 		break;
1088 	case P_PGID:
1089 		/*
1090 		 * Attempt to apply the operation to all members of the
1091 		 * group.  Ignore processes in the group that can't be
1092 		 * seen.  Ignore errors so long as at least one process is
1093 		 * able to complete the request successfully.
1094 		 */
1095 		pg = pgfind(id);
1096 		if (pg == NULL) {
1097 			error = ESRCH;
1098 			break;
1099 		}
1100 		PGRP_UNLOCK(pg);
1101 		ok = 0;
1102 		first_error = 0;
1103 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
1104 			PROC_LOCK(p);
1105 			if (p->p_state == PRS_NEW ||
1106 			    p->p_state == PRS_ZOMBIE ||
1107 			    (cmd_info->need_candebug ? p_candebug(td, p) :
1108 			    p_cansee(td, p)) != 0) {
1109 				PROC_UNLOCK(p);
1110 				continue;
1111 			}
1112 			error = kern_procctl_single(td, p, com, data);
1113 			PROC_UNLOCK(p);
1114 			if (error == 0)
1115 				ok = 1;
1116 			else if (first_error == 0)
1117 				first_error = error;
1118 		}
1119 		if (ok)
1120 			error = 0;
1121 		else if (first_error != 0)
1122 			error = first_error;
1123 		else
1124 			/*
1125 			 * Was not able to see any processes in the
1126 			 * process group.
1127 			 */
1128 			error = ESRCH;
1129 		break;
1130 	default:
1131 		error = EINVAL;
1132 		break;
1133 	}
1134 
1135 	switch (cmd_info->lock_tree) {
1136 	case PCTL_XLOCKED:
1137 		sx_xunlock(&proctree_lock);
1138 		break;
1139 	case PCTL_SLOCKED:
1140 		sx_sunlock(&proctree_lock);
1141 		break;
1142 	default:
1143 		break;
1144 	}
1145 	if (sapblk)
1146 		stop_all_proc_unblock();
1147 	return (error);
1148 }
1149