]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/blob - sys/kern/kern_procctl.c
kern_jail.c: Remove #ifdefs for VNET_NFSD
[FreeBSD/FreeBSD.git] / sys / kern / kern_procctl.c
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/taskqueue.h>
47 #include <sys/wait.h>
48
49 #include <vm/vm.h>
50 #include <vm/pmap.h>
51 #include <vm/vm_map.h>
52 #include <vm/vm_extern.h>
53
54 static int
55 protect_setchild(struct thread *td, struct proc *p, int flags)
56 {
57
58         PROC_LOCK_ASSERT(p, MA_OWNED);
59         if (p->p_flag & P_SYSTEM || p_cansched(td, p) != 0)
60                 return (0);
61         if (flags & PPROT_SET) {
62                 p->p_flag |= P_PROTECTED;
63                 if (flags & PPROT_INHERIT)
64                         p->p_flag2 |= P2_INHERIT_PROTECTED;
65         } else {
66                 p->p_flag &= ~P_PROTECTED;
67                 p->p_flag2 &= ~P2_INHERIT_PROTECTED;
68         }
69         return (1);
70 }
71
72 static int
73 protect_setchildren(struct thread *td, struct proc *top, int flags)
74 {
75         struct proc *p;
76         int ret;
77
78         p = top;
79         ret = 0;
80         sx_assert(&proctree_lock, SX_LOCKED);
81         for (;;) {
82                 ret |= protect_setchild(td, p, flags);
83                 PROC_UNLOCK(p);
84                 /*
85                  * If this process has children, descend to them next,
86                  * otherwise do any siblings, and if done with this level,
87                  * follow back up the tree (but not past top).
88                  */
89                 if (!LIST_EMPTY(&p->p_children))
90                         p = LIST_FIRST(&p->p_children);
91                 else for (;;) {
92                         if (p == top) {
93                                 PROC_LOCK(p);
94                                 return (ret);
95                         }
96                         if (LIST_NEXT(p, p_sibling)) {
97                                 p = LIST_NEXT(p, p_sibling);
98                                 break;
99                         }
100                         p = p->p_pptr;
101                 }
102                 PROC_LOCK(p);
103         }
104 }
105
106 static int
107 protect_set(struct thread *td, struct proc *p, void *data)
108 {
109         int error, flags, ret;
110
111         flags = *(int *)data;
112         switch (PPROT_OP(flags)) {
113         case PPROT_SET:
114         case PPROT_CLEAR:
115                 break;
116         default:
117                 return (EINVAL);
118         }
119
120         if ((PPROT_FLAGS(flags) & ~(PPROT_DESCEND | PPROT_INHERIT)) != 0)
121                 return (EINVAL);
122
123         error = priv_check(td, PRIV_VM_MADV_PROTECT);
124         if (error)
125                 return (error);
126
127         if (flags & PPROT_DESCEND)
128                 ret = protect_setchildren(td, p, flags);
129         else
130                 ret = protect_setchild(td, p, flags);
131         if (ret == 0)
132                 return (EPERM);
133         return (0);
134 }
135
136 static int
137 reap_acquire(struct thread *td, struct proc *p, void *data __unused)
138 {
139
140         sx_assert(&proctree_lock, SX_XLOCKED);
141         if (p != td->td_proc)
142                 return (EPERM);
143         if ((p->p_treeflag & P_TREE_REAPER) != 0)
144                 return (EBUSY);
145         p->p_treeflag |= P_TREE_REAPER;
146         /*
147          * We do not reattach existing children and the whole tree
148          * under them to us, since p->p_reaper already seen them.
149          */
150         return (0);
151 }
152
153 static int
154 reap_release(struct thread *td, struct proc *p, void *data __unused)
155 {
156
157         sx_assert(&proctree_lock, SX_XLOCKED);
158         if (p != td->td_proc)
159                 return (EPERM);
160         if (p == initproc)
161                 return (EINVAL);
162         if ((p->p_treeflag & P_TREE_REAPER) == 0)
163                 return (EINVAL);
164         reaper_abandon_children(p, false);
165         return (0);
166 }
167
168 static int
169 reap_status(struct thread *td, struct proc *p, void *data)
170 {
171         struct proc *reap, *p2, *first_p;
172         struct procctl_reaper_status *rs;
173
174         rs = data;
175         sx_assert(&proctree_lock, SX_LOCKED);
176         if ((p->p_treeflag & P_TREE_REAPER) == 0) {
177                 reap = p->p_reaper;
178         } else {
179                 reap = p;
180                 rs->rs_flags |= REAPER_STATUS_OWNED;
181         }
182         if (reap == initproc)
183                 rs->rs_flags |= REAPER_STATUS_REALINIT;
184         rs->rs_reaper = reap->p_pid;
185         rs->rs_descendants = 0;
186         rs->rs_children = 0;
187         if (!LIST_EMPTY(&reap->p_reaplist)) {
188                 first_p = LIST_FIRST(&reap->p_children);
189                 if (first_p == NULL)
190                         first_p = LIST_FIRST(&reap->p_reaplist);
191                 rs->rs_pid = first_p->p_pid;
192                 LIST_FOREACH(p2, &reap->p_reaplist, p_reapsibling) {
193                         if (proc_realparent(p2) == reap)
194                                 rs->rs_children++;
195                         rs->rs_descendants++;
196                 }
197         } else {
198                 rs->rs_pid = -1;
199         }
200         return (0);
201 }
202
203 static int
204 reap_getpids(struct thread *td, struct proc *p, void *data)
205 {
206         struct proc *reap, *p2;
207         struct procctl_reaper_pidinfo *pi, *pip;
208         struct procctl_reaper_pids *rp;
209         u_int i, n;
210         int error;
211
212         rp = data;
213         sx_assert(&proctree_lock, SX_LOCKED);
214         PROC_UNLOCK(p);
215         reap = (p->p_treeflag & P_TREE_REAPER) == 0 ? p->p_reaper : p;
216         n = i = 0;
217         error = 0;
218         LIST_FOREACH(p2, &reap->p_reaplist, p_reapsibling)
219                 n++;
220         sx_unlock(&proctree_lock);
221         if (rp->rp_count < n)
222                 n = rp->rp_count;
223         pi = malloc(n * sizeof(*pi), M_TEMP, M_WAITOK);
224         sx_slock(&proctree_lock);
225         LIST_FOREACH(p2, &reap->p_reaplist, p_reapsibling) {
226                 if (i == n)
227                         break;
228                 pip = &pi[i];
229                 bzero(pip, sizeof(*pip));
230                 pip->pi_pid = p2->p_pid;
231                 pip->pi_subtree = p2->p_reapsubtree;
232                 pip->pi_flags = REAPER_PIDINFO_VALID;
233                 if (proc_realparent(p2) == reap)
234                         pip->pi_flags |= REAPER_PIDINFO_CHILD;
235                 if ((p2->p_treeflag & P_TREE_REAPER) != 0)
236                         pip->pi_flags |= REAPER_PIDINFO_REAPER;
237                 if ((p2->p_flag & P_STOPPED) != 0)
238                         pip->pi_flags |= REAPER_PIDINFO_STOPPED;
239                 if (p2->p_state == PRS_ZOMBIE)
240                         pip->pi_flags |= REAPER_PIDINFO_ZOMBIE;
241                 else if ((p2->p_flag & P_WEXIT) != 0)
242                         pip->pi_flags |= REAPER_PIDINFO_EXITING;
243                 i++;
244         }
245         sx_sunlock(&proctree_lock);
246         error = copyout(pi, rp->rp_pids, i * sizeof(*pi));
247         free(pi, M_TEMP);
248         sx_slock(&proctree_lock);
249         PROC_LOCK(p);
250         return (error);
251 }
252
253 struct reap_kill_proc_work {
254         struct ucred *cr;
255         struct proc *target;
256         ksiginfo_t *ksi;
257         struct procctl_reaper_kill *rk;
258         int *error;
259         struct task t;
260 };
261
262 static void
263 reap_kill_proc_locked(struct reap_kill_proc_work *w)
264 {
265         int error1;
266         bool need_stop;
267
268         PROC_LOCK_ASSERT(w->target, MA_OWNED);
269         PROC_ASSERT_HELD(w->target);
270
271         error1 = cr_cansignal(w->cr, w->target, w->rk->rk_sig);
272         if (error1 != 0) {
273                 if (*w->error == ESRCH) {
274                         w->rk->rk_fpid = w->target->p_pid;
275                         *w->error = error1;
276                 }
277                 return;
278         }
279
280         /*
281          * The need_stop indicates if the target process needs to be
282          * suspended before being signalled.  This is needed when we
283          * guarantee that all processes in subtree are signalled,
284          * avoiding the race with some process not yet fully linked
285          * into all structures during fork, ignored by iterator, and
286          * then escaping signalling.
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         if ((w->target->p_flag & (P_KPROC | P_SYSTEM | P_STOPPED)) == 0)
294                 need_stop = thread_single(w->target, SINGLE_ALLPROC) == 0;
295         else
296                 need_stop = false;
297
298         (void)pksignal(w->target, w->rk->rk_sig, w->ksi);
299         w->rk->rk_killed++;
300         *w->error = error1;
301
302         if (need_stop)
303                 thread_single_end(w->target, SINGLE_ALLPROC);
304 }
305
306 static void
307 reap_kill_proc_work(void *arg, int pending __unused)
308 {
309         struct reap_kill_proc_work *w;
310
311         w = arg;
312         PROC_LOCK(w->target);
313         if ((w->target->p_flag2 & P2_WEXIT) == 0)
314                 reap_kill_proc_locked(w);
315         PROC_UNLOCK(w->target);
316
317         sx_xlock(&proctree_lock);
318         w->target = NULL;
319         wakeup(&w->target);
320         sx_xunlock(&proctree_lock);
321 }
322
323 struct reap_kill_tracker {
324         struct proc *parent;
325         TAILQ_ENTRY(reap_kill_tracker) link;
326 };
327
328 TAILQ_HEAD(reap_kill_tracker_head, reap_kill_tracker);
329
330 static void
331 reap_kill_sched(struct reap_kill_tracker_head *tracker, struct proc *p2)
332 {
333         struct reap_kill_tracker *t;
334
335         PROC_LOCK(p2);
336         if ((p2->p_flag2 & P2_WEXIT) != 0) {
337                 PROC_UNLOCK(p2);
338                 return;
339         }
340         _PHOLD_LITE(p2);
341         PROC_UNLOCK(p2);
342         t = malloc(sizeof(struct reap_kill_tracker), M_TEMP, M_WAITOK);
343         t->parent = p2;
344         TAILQ_INSERT_TAIL(tracker, t, link);
345 }
346
347 static void
348 reap_kill_sched_free(struct reap_kill_tracker *t)
349 {
350         PRELE(t->parent);
351         free(t, M_TEMP);
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         int error1;
360
361         LIST_FOREACH(p2, &reaper->p_children, p_sibling) {
362                 PROC_LOCK(p2);
363                 if ((p2->p_flag2 & P2_WEXIT) == 0) {
364                         error1 = p_cansignal(td, p2, rk->rk_sig);
365                         if (error1 != 0) {
366                                 if (*error == ESRCH) {
367                                         rk->rk_fpid = p2->p_pid;
368                                         *error = error1;
369                                 }
370
371                                 /*
372                                  * Do not end the loop on error,
373                                  * signal everything we can.
374                                  */
375                         } else {
376                                 (void)pksignal(p2, rk->rk_sig, ksi);
377                                 rk->rk_killed++;
378                         }
379                 }
380                 PROC_UNLOCK(p2);
381         }
382 }
383
384 static bool
385 reap_kill_subtree_once(struct thread *td, struct proc *p, struct proc *reaper,
386     struct unrhdr *pids, struct reap_kill_proc_work *w)
387 {
388         struct reap_kill_tracker_head tracker;
389         struct reap_kill_tracker *t;
390         struct proc *p2;
391         int r, xlocked;
392         bool res, st;
393
394         res = false;
395         TAILQ_INIT(&tracker);
396         reap_kill_sched(&tracker, reaper);
397         while ((t = TAILQ_FIRST(&tracker)) != NULL) {
398                 TAILQ_REMOVE(&tracker, t, link);
399
400                 /*
401                  * Since reap_kill_proc() drops proctree_lock sx, it
402                  * is possible that the tracked reaper is no longer.
403                  * In this case the subtree is reparented to the new
404                  * reaper, which should handle it.
405                  */
406                 if ((t->parent->p_treeflag & P_TREE_REAPER) == 0) {
407                         reap_kill_sched_free(t);
408                         res = true;
409                         continue;
410                 }
411
412                 LIST_FOREACH(p2, &t->parent->p_reaplist, p_reapsibling) {
413                         if (t->parent == reaper &&
414                             (w->rk->rk_flags & REAPER_KILL_SUBTREE) != 0 &&
415                             p2->p_reapsubtree != w->rk->rk_subtree)
416                                 continue;
417                         if ((p2->p_treeflag & P_TREE_REAPER) != 0)
418                                 reap_kill_sched(&tracker, p2);
419                         if (alloc_unr_specific(pids, p2->p_pid) != p2->p_pid)
420                                 continue;
421                         if (p2 == td->td_proc) {
422                                 if ((p2->p_flag & P_HADTHREADS) != 0 &&
423                                     (p2->p_flag2 & P2_WEXIT) == 0) {
424                                         xlocked = sx_xlocked(&proctree_lock);
425                                         sx_unlock(&proctree_lock);
426                                         st = true;
427                                 } else {
428                                         st = false;
429                                 }
430                                 PROC_LOCK(p2);
431                                 if (st)
432                                         r = thread_single(p2, SINGLE_NO_EXIT);
433                                 (void)pksignal(p2, w->rk->rk_sig, w->ksi);
434                                 w->rk->rk_killed++;
435                                 if (st && r == 0)
436                                         thread_single_end(p2, SINGLE_NO_EXIT);
437                                 PROC_UNLOCK(p2);
438                                 if (st) {
439                                         if (xlocked)
440                                                 sx_xlock(&proctree_lock);
441                                         else
442                                                 sx_slock(&proctree_lock);
443                                 }
444                         } else {
445                                 PROC_LOCK(p2);
446                                 if ((p2->p_flag2 & P2_WEXIT) == 0) {
447                                         _PHOLD_LITE(p2);
448                                         PROC_UNLOCK(p2);
449                                         w->target = p2;
450                                         taskqueue_enqueue(taskqueue_thread,
451                                             &w->t);
452                                         while (w->target != NULL) {
453                                                 sx_sleep(&w->target,
454                                                     &proctree_lock, PWAIT,
455                                                     "reapst", 0);
456                                         }
457                                         PROC_LOCK(p2);
458                                         _PRELE(p2);
459                                 }
460                                 PROC_UNLOCK(p2);
461                         }
462                         res = true;
463                 }
464                 reap_kill_sched_free(t);
465         }
466         return (res);
467 }
468
469 static void
470 reap_kill_subtree(struct thread *td, struct proc *p, struct proc *reaper,
471     struct reap_kill_proc_work *w)
472 {
473         struct unrhdr pids;
474
475         /*
476          * pids records processes which were already signalled, to
477          * avoid doubling signals to them if iteration needs to be
478          * repeated.
479          */
480         init_unrhdr(&pids, 1, PID_MAX, UNR_NO_MTX);
481         PROC_LOCK(td->td_proc);
482         if ((td->td_proc->p_flag2 & P2_WEXIT) != 0) {
483                 PROC_UNLOCK(td->td_proc);
484                 goto out;
485         }
486         PROC_UNLOCK(td->td_proc);
487         while (reap_kill_subtree_once(td, p, reaper, &pids, w))
488                ;
489 out:
490         clean_unrhdr(&pids);
491         clear_unrhdr(&pids);
492 }
493
494 static bool
495 reap_kill_sapblk(struct thread *td __unused, void *data)
496 {
497         struct procctl_reaper_kill *rk;
498
499         rk = data;
500         return ((rk->rk_flags & REAPER_KILL_CHILDREN) == 0);
501 }
502
503 static int
504 reap_kill(struct thread *td, struct proc *p, void *data)
505 {
506         struct reap_kill_proc_work w;
507         struct proc *reaper;
508         ksiginfo_t ksi;
509         struct procctl_reaper_kill *rk;
510         int error;
511
512         rk = data;
513         sx_assert(&proctree_lock, SX_LOCKED);
514         if (IN_CAPABILITY_MODE(td))
515                 return (ECAPMODE);
516         if (rk->rk_sig <= 0 || rk->rk_sig > _SIG_MAXSIG ||
517             (rk->rk_flags & ~(REAPER_KILL_CHILDREN |
518             REAPER_KILL_SUBTREE)) != 0 || (rk->rk_flags &
519             (REAPER_KILL_CHILDREN | REAPER_KILL_SUBTREE)) ==
520             (REAPER_KILL_CHILDREN | REAPER_KILL_SUBTREE))
521                 return (EINVAL);
522         PROC_UNLOCK(p);
523         reaper = (p->p_treeflag & P_TREE_REAPER) == 0 ? p->p_reaper : p;
524         ksiginfo_init(&ksi);
525         ksi.ksi_signo = rk->rk_sig;
526         ksi.ksi_code = SI_USER;
527         ksi.ksi_pid = td->td_proc->p_pid;
528         ksi.ksi_uid = td->td_ucred->cr_ruid;
529         error = ESRCH;
530         rk->rk_killed = 0;
531         rk->rk_fpid = -1;
532         if ((rk->rk_flags & REAPER_KILL_CHILDREN) != 0) {
533                 reap_kill_children(td, reaper, rk, &ksi, &error);
534         } else {
535                 w.cr = crhold(td->td_ucred);
536                 w.ksi = &ksi;
537                 w.rk = rk;
538                 w.error = &error;
539                 TASK_INIT(&w.t, 0, reap_kill_proc_work, &w);
540
541                 /*
542                  * Prevent swapout, since w, ksi, and possibly rk, are
543                  * allocated on the stack.  We sleep in
544                  * reap_kill_subtree_once() waiting for task to
545                  * complete single-threading.
546                  */
547                 PHOLD(td->td_proc);
548
549                 reap_kill_subtree(td, p, reaper, &w);
550                 PRELE(td->td_proc);
551                 crfree(w.cr);
552         }
553         PROC_LOCK(p);
554         return (error);
555 }
556
557 static int
558 trace_ctl(struct thread *td, struct proc *p, void *data)
559 {
560         int state;
561
562         PROC_LOCK_ASSERT(p, MA_OWNED);
563         state = *(int *)data;
564
565         /*
566          * Ktrace changes p_traceflag from or to zero under the
567          * process lock, so the test does not need to acquire ktrace
568          * mutex.
569          */
570         if ((p->p_flag & P_TRACED) != 0 || p->p_traceflag != 0)
571                 return (EBUSY);
572
573         switch (state) {
574         case PROC_TRACE_CTL_ENABLE:
575                 if (td->td_proc != p)
576                         return (EPERM);
577                 p->p_flag2 &= ~(P2_NOTRACE | P2_NOTRACE_EXEC);
578                 break;
579         case PROC_TRACE_CTL_DISABLE_EXEC:
580                 p->p_flag2 |= P2_NOTRACE_EXEC | P2_NOTRACE;
581                 break;
582         case PROC_TRACE_CTL_DISABLE:
583                 if ((p->p_flag2 & P2_NOTRACE_EXEC) != 0) {
584                         KASSERT((p->p_flag2 & P2_NOTRACE) != 0,
585                             ("dandling P2_NOTRACE_EXEC"));
586                         if (td->td_proc != p)
587                                 return (EPERM);
588                         p->p_flag2 &= ~P2_NOTRACE_EXEC;
589                 } else {
590                         p->p_flag2 |= P2_NOTRACE;
591                 }
592                 break;
593         default:
594                 return (EINVAL);
595         }
596         return (0);
597 }
598
599 static int
600 trace_status(struct thread *td, struct proc *p, void *data)
601 {
602         int *status;
603
604         status = data;
605         if ((p->p_flag2 & P2_NOTRACE) != 0) {
606                 KASSERT((p->p_flag & P_TRACED) == 0,
607                     ("%d traced but tracing disabled", p->p_pid));
608                 *status = -1;
609         } else if ((p->p_flag & P_TRACED) != 0) {
610                 *status = p->p_pptr->p_pid;
611         } else {
612                 *status = 0;
613         }
614         return (0);
615 }
616
617 static int
618 trapcap_ctl(struct thread *td, struct proc *p, void *data)
619 {
620         int state;
621
622         PROC_LOCK_ASSERT(p, MA_OWNED);
623         state = *(int *)data;
624
625         switch (state) {
626         case PROC_TRAPCAP_CTL_ENABLE:
627                 p->p_flag2 |= P2_TRAPCAP;
628                 break;
629         case PROC_TRAPCAP_CTL_DISABLE:
630                 p->p_flag2 &= ~P2_TRAPCAP;
631                 break;
632         default:
633                 return (EINVAL);
634         }
635         return (0);
636 }
637
638 static int
639 trapcap_status(struct thread *td, struct proc *p, void *data)
640 {
641         int *status;
642
643         status = data;
644         *status = (p->p_flag2 & P2_TRAPCAP) != 0 ? PROC_TRAPCAP_CTL_ENABLE :
645             PROC_TRAPCAP_CTL_DISABLE;
646         return (0);
647 }
648
649 static int
650 no_new_privs_ctl(struct thread *td, struct proc *p, void *data)
651 {
652         int state;
653
654         PROC_LOCK_ASSERT(p, MA_OWNED);
655         state = *(int *)data;
656
657         if (state != PROC_NO_NEW_PRIVS_ENABLE)
658                 return (EINVAL);
659         p->p_flag2 |= P2_NO_NEW_PRIVS;
660         return (0);
661 }
662
663 static int
664 no_new_privs_status(struct thread *td, struct proc *p, void *data)
665 {
666
667         *(int *)data = (p->p_flag2 & P2_NO_NEW_PRIVS) != 0 ?
668             PROC_NO_NEW_PRIVS_ENABLE : PROC_NO_NEW_PRIVS_DISABLE;
669         return (0);
670 }
671
672 static int
673 protmax_ctl(struct thread *td, struct proc *p, void *data)
674 {
675         int state;
676
677         PROC_LOCK_ASSERT(p, MA_OWNED);
678         state = *(int *)data;
679
680         switch (state) {
681         case PROC_PROTMAX_FORCE_ENABLE:
682                 p->p_flag2 &= ~P2_PROTMAX_DISABLE;
683                 p->p_flag2 |= P2_PROTMAX_ENABLE;
684                 break;
685         case PROC_PROTMAX_FORCE_DISABLE:
686                 p->p_flag2 |= P2_PROTMAX_DISABLE;
687                 p->p_flag2 &= ~P2_PROTMAX_ENABLE;
688                 break;
689         case PROC_PROTMAX_NOFORCE:
690                 p->p_flag2 &= ~(P2_PROTMAX_ENABLE | P2_PROTMAX_DISABLE);
691                 break;
692         default:
693                 return (EINVAL);
694         }
695         return (0);
696 }
697
698 static int
699 protmax_status(struct thread *td, struct proc *p, void *data)
700 {
701         int d;
702
703         switch (p->p_flag2 & (P2_PROTMAX_ENABLE | P2_PROTMAX_DISABLE)) {
704         case 0:
705                 d = PROC_PROTMAX_NOFORCE;
706                 break;
707         case P2_PROTMAX_ENABLE:
708                 d = PROC_PROTMAX_FORCE_ENABLE;
709                 break;
710         case P2_PROTMAX_DISABLE:
711                 d = PROC_PROTMAX_FORCE_DISABLE;
712                 break;
713         }
714         if (kern_mmap_maxprot(p, PROT_READ) == PROT_READ)
715                 d |= PROC_PROTMAX_ACTIVE;
716         *(int *)data = d;
717         return (0);
718 }
719
720 static int
721 aslr_ctl(struct thread *td, struct proc *p, void *data)
722 {
723         int state;
724
725         PROC_LOCK_ASSERT(p, MA_OWNED);
726         state = *(int *)data;
727
728         switch (state) {
729         case PROC_ASLR_FORCE_ENABLE:
730                 p->p_flag2 &= ~P2_ASLR_DISABLE;
731                 p->p_flag2 |= P2_ASLR_ENABLE;
732                 break;
733         case PROC_ASLR_FORCE_DISABLE:
734                 p->p_flag2 |= P2_ASLR_DISABLE;
735                 p->p_flag2 &= ~P2_ASLR_ENABLE;
736                 break;
737         case PROC_ASLR_NOFORCE:
738                 p->p_flag2 &= ~(P2_ASLR_ENABLE | P2_ASLR_DISABLE);
739                 break;
740         default:
741                 return (EINVAL);
742         }
743         return (0);
744 }
745
746 static int
747 aslr_status(struct thread *td, struct proc *p, void *data)
748 {
749         struct vmspace *vm;
750         int d;
751
752         switch (p->p_flag2 & (P2_ASLR_ENABLE | P2_ASLR_DISABLE)) {
753         case 0:
754                 d = PROC_ASLR_NOFORCE;
755                 break;
756         case P2_ASLR_ENABLE:
757                 d = PROC_ASLR_FORCE_ENABLE;
758                 break;
759         case P2_ASLR_DISABLE:
760                 d = PROC_ASLR_FORCE_DISABLE;
761                 break;
762         }
763         if ((p->p_flag & P_WEXIT) == 0) {
764                 _PHOLD(p);
765                 PROC_UNLOCK(p);
766                 vm = vmspace_acquire_ref(p);
767                 if (vm != NULL) {
768                         if ((vm->vm_map.flags & MAP_ASLR) != 0)
769                                 d |= PROC_ASLR_ACTIVE;
770                         vmspace_free(vm);
771                 }
772                 PROC_LOCK(p);
773                 _PRELE(p);
774         }
775         *(int *)data = d;
776         return (0);
777 }
778
779 static int
780 stackgap_ctl(struct thread *td, struct proc *p, void *data)
781 {
782         int state;
783
784         PROC_LOCK_ASSERT(p, MA_OWNED);
785         state = *(int *)data;
786
787         if ((state & ~(PROC_STACKGAP_ENABLE | PROC_STACKGAP_DISABLE |
788             PROC_STACKGAP_ENABLE_EXEC | PROC_STACKGAP_DISABLE_EXEC)) != 0)
789                 return (EINVAL);
790         switch (state & (PROC_STACKGAP_ENABLE | PROC_STACKGAP_DISABLE)) {
791         case PROC_STACKGAP_ENABLE:
792                 if ((p->p_flag2 & P2_STKGAP_DISABLE) != 0)
793                         return (EINVAL);
794                 break;
795         case PROC_STACKGAP_DISABLE:
796                 p->p_flag2 |= P2_STKGAP_DISABLE;
797                 break;
798         case 0:
799                 break;
800         default:
801                 return (EINVAL);
802         }
803         switch (state & (PROC_STACKGAP_ENABLE_EXEC |
804             PROC_STACKGAP_DISABLE_EXEC)) {
805         case PROC_STACKGAP_ENABLE_EXEC:
806                 p->p_flag2 &= ~P2_STKGAP_DISABLE_EXEC;
807                 break;
808         case PROC_STACKGAP_DISABLE_EXEC:
809                 p->p_flag2 |= P2_STKGAP_DISABLE_EXEC;
810                 break;
811         case 0:
812                 break;
813         default:
814                 return (EINVAL);
815         }
816         return (0);
817 }
818
819 static int
820 stackgap_status(struct thread *td, struct proc *p, void *data)
821 {
822         int d;
823
824         PROC_LOCK_ASSERT(p, MA_OWNED);
825
826         d = (p->p_flag2 & P2_STKGAP_DISABLE) != 0 ? PROC_STACKGAP_DISABLE :
827             PROC_STACKGAP_ENABLE;
828         d |= (p->p_flag2 & P2_STKGAP_DISABLE_EXEC) != 0 ?
829             PROC_STACKGAP_DISABLE_EXEC : PROC_STACKGAP_ENABLE_EXEC;
830         *(int *)data = d;
831         return (0);
832 }
833
834 static int
835 wxmap_ctl(struct thread *td, struct proc *p, void *data)
836 {
837         struct vmspace *vm;
838         vm_map_t map;
839         int state;
840
841         PROC_LOCK_ASSERT(p, MA_OWNED);
842         if ((p->p_flag & P_WEXIT) != 0)
843                 return (ESRCH);
844         state = *(int *)data;
845
846         switch (state) {
847         case PROC_WX_MAPPINGS_PERMIT:
848                 p->p_flag2 |= P2_WXORX_DISABLE;
849                 _PHOLD(p);
850                 PROC_UNLOCK(p);
851                 vm = vmspace_acquire_ref(p);
852                 if (vm != NULL) {
853                         map = &vm->vm_map;
854                         vm_map_lock(map);
855                         map->flags &= ~MAP_WXORX;
856                         vm_map_unlock(map);
857                         vmspace_free(vm);
858                 }
859                 PROC_LOCK(p);
860                 _PRELE(p);
861                 break;
862         case PROC_WX_MAPPINGS_DISALLOW_EXEC:
863                 p->p_flag2 |= P2_WXORX_ENABLE_EXEC;
864                 break;
865         default:
866                 return (EINVAL);
867         }
868
869         return (0);
870 }
871
872 static int
873 wxmap_status(struct thread *td, struct proc *p, void *data)
874 {
875         struct vmspace *vm;
876         int d;
877
878         PROC_LOCK_ASSERT(p, MA_OWNED);
879         if ((p->p_flag & P_WEXIT) != 0)
880                 return (ESRCH);
881
882         d = 0;
883         if ((p->p_flag2 & P2_WXORX_DISABLE) != 0)
884                 d |= PROC_WX_MAPPINGS_PERMIT;
885         if ((p->p_flag2 & P2_WXORX_ENABLE_EXEC) != 0)
886                 d |= PROC_WX_MAPPINGS_DISALLOW_EXEC;
887         _PHOLD(p);
888         PROC_UNLOCK(p);
889         vm = vmspace_acquire_ref(p);
890         if (vm != NULL) {
891                 if ((vm->vm_map.flags & MAP_WXORX) != 0)
892                         d |= PROC_WXORX_ENFORCE;
893                 vmspace_free(vm);
894         }
895         PROC_LOCK(p);
896         _PRELE(p);
897         *(int *)data = d;
898         return (0);
899 }
900
901 static int
902 pdeathsig_ctl(struct thread *td, struct proc *p, void *data)
903 {
904         int signum;
905
906         signum = *(int *)data;
907         if (p != td->td_proc || (signum != 0 && !_SIG_VALID(signum)))
908                 return (EINVAL);
909         p->p_pdeathsig = signum;
910         return (0);
911 }
912
913 static int
914 pdeathsig_status(struct thread *td, struct proc *p, void *data)
915 {
916         if (p != td->td_proc)
917                 return (EINVAL);
918         *(int *)data = p->p_pdeathsig;
919         return (0);
920 }
921
922 enum {
923         PCTL_SLOCKED,
924         PCTL_XLOCKED,
925         PCTL_UNLOCKED,
926 };
927
928 struct procctl_cmd_info {
929         int lock_tree;
930         bool one_proc : 1;
931         bool esrch_is_einval : 1;
932         bool copyout_on_error : 1;
933         bool no_nonnull_data : 1;
934         bool need_candebug : 1;
935         int copyin_sz;
936         int copyout_sz;
937         int (*exec)(struct thread *, struct proc *, void *);
938         bool (*sapblk)(struct thread *, void *);
939 };
940 static const struct procctl_cmd_info procctl_cmds_info[] = {
941         [PROC_SPROTECT] =
942             { .lock_tree = PCTL_SLOCKED, .one_proc = false,
943               .esrch_is_einval = false, .no_nonnull_data = false,
944               .need_candebug = false,
945               .copyin_sz = sizeof(int), .copyout_sz = 0,
946               .exec = protect_set, .copyout_on_error = false, },
947         [PROC_REAP_ACQUIRE] =
948             { .lock_tree = PCTL_XLOCKED, .one_proc = true,
949               .esrch_is_einval = false, .no_nonnull_data = true,
950               .need_candebug = false,
951               .copyin_sz = 0, .copyout_sz = 0,
952               .exec = reap_acquire, .copyout_on_error = false, },
953         [PROC_REAP_RELEASE] =
954             { .lock_tree = PCTL_XLOCKED, .one_proc = true,
955               .esrch_is_einval = false, .no_nonnull_data = true,
956               .need_candebug = false,
957               .copyin_sz = 0, .copyout_sz = 0,
958               .exec = reap_release, .copyout_on_error = false, },
959         [PROC_REAP_STATUS] =
960             { .lock_tree = PCTL_SLOCKED, .one_proc = true,
961               .esrch_is_einval = false, .no_nonnull_data = false,
962               .need_candebug = false,
963               .copyin_sz = 0,
964               .copyout_sz = sizeof(struct procctl_reaper_status),
965               .exec = reap_status, .copyout_on_error = false, },
966         [PROC_REAP_GETPIDS] =
967             { .lock_tree = PCTL_SLOCKED, .one_proc = true,
968               .esrch_is_einval = false, .no_nonnull_data = false,
969               .need_candebug = false,
970               .copyin_sz = sizeof(struct procctl_reaper_pids),
971               .copyout_sz = 0,
972               .exec = reap_getpids, .copyout_on_error = false, },
973         [PROC_REAP_KILL] =
974             { .lock_tree = PCTL_SLOCKED, .one_proc = true,
975               .esrch_is_einval = false, .no_nonnull_data = false,
976               .need_candebug = false,
977               .copyin_sz = sizeof(struct procctl_reaper_kill),
978               .copyout_sz = sizeof(struct procctl_reaper_kill),
979               .exec = reap_kill, .copyout_on_error = true,
980               .sapblk = reap_kill_sapblk, },
981         [PROC_TRACE_CTL] =
982             { .lock_tree = PCTL_SLOCKED, .one_proc = false,
983               .esrch_is_einval = false, .no_nonnull_data = false,
984               .need_candebug = true,
985               .copyin_sz = sizeof(int), .copyout_sz = 0,
986               .exec = trace_ctl, .copyout_on_error = false, },
987         [PROC_TRACE_STATUS] =
988             { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
989               .esrch_is_einval = false, .no_nonnull_data = false,
990               .need_candebug = false,
991               .copyin_sz = 0, .copyout_sz = sizeof(int),
992               .exec = trace_status, .copyout_on_error = false, },
993         [PROC_TRAPCAP_CTL] =
994             { .lock_tree = PCTL_SLOCKED, .one_proc = false,
995               .esrch_is_einval = false, .no_nonnull_data = false,
996               .need_candebug = true,
997               .copyin_sz = sizeof(int), .copyout_sz = 0,
998               .exec = trapcap_ctl, .copyout_on_error = false, },
999         [PROC_TRAPCAP_STATUS] =
1000             { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
1001               .esrch_is_einval = false, .no_nonnull_data = false,
1002               .need_candebug = false,
1003               .copyin_sz = 0, .copyout_sz = sizeof(int),
1004               .exec = trapcap_status, .copyout_on_error = false, },
1005         [PROC_PDEATHSIG_CTL] =
1006             { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
1007               .esrch_is_einval = true, .no_nonnull_data = false,
1008               .need_candebug = false,
1009               .copyin_sz = sizeof(int), .copyout_sz = 0,
1010               .exec = pdeathsig_ctl, .copyout_on_error = false, },
1011         [PROC_PDEATHSIG_STATUS] =
1012             { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
1013               .esrch_is_einval = true, .no_nonnull_data = false,
1014               .need_candebug = false,
1015               .copyin_sz = 0, .copyout_sz = sizeof(int),
1016               .exec = pdeathsig_status, .copyout_on_error = false, },
1017         [PROC_ASLR_CTL] =
1018             { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
1019               .esrch_is_einval = false, .no_nonnull_data = false,
1020               .need_candebug = true,
1021               .copyin_sz = sizeof(int), .copyout_sz = 0,
1022               .exec = aslr_ctl, .copyout_on_error = false, },
1023         [PROC_ASLR_STATUS] =
1024             { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
1025               .esrch_is_einval = false, .no_nonnull_data = false,
1026               .need_candebug = false,
1027               .copyin_sz = 0, .copyout_sz = sizeof(int),
1028               .exec = aslr_status, .copyout_on_error = false, },
1029         [PROC_PROTMAX_CTL] =
1030             { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
1031               .esrch_is_einval = false, .no_nonnull_data = false,
1032               .need_candebug = true,
1033               .copyin_sz = sizeof(int), .copyout_sz = 0,
1034               .exec = protmax_ctl, .copyout_on_error = false, },
1035         [PROC_PROTMAX_STATUS] =
1036             { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
1037               .esrch_is_einval = false, .no_nonnull_data = false,
1038               .need_candebug = false,
1039               .copyin_sz = 0, .copyout_sz = sizeof(int),
1040               .exec = protmax_status, .copyout_on_error = false, },
1041         [PROC_STACKGAP_CTL] =
1042             { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
1043               .esrch_is_einval = false, .no_nonnull_data = false,
1044               .need_candebug = true,
1045               .copyin_sz = sizeof(int), .copyout_sz = 0,
1046               .exec = stackgap_ctl, .copyout_on_error = false, },
1047         [PROC_STACKGAP_STATUS] =
1048             { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
1049               .esrch_is_einval = false, .no_nonnull_data = false,
1050               .need_candebug = false,
1051               .copyin_sz = 0, .copyout_sz = sizeof(int),
1052               .exec = stackgap_status, .copyout_on_error = false, },
1053         [PROC_NO_NEW_PRIVS_CTL] =
1054             { .lock_tree = PCTL_SLOCKED, .one_proc = true,
1055               .esrch_is_einval = false, .no_nonnull_data = false,
1056               .need_candebug = true,
1057               .copyin_sz = sizeof(int), .copyout_sz = 0,
1058               .exec = no_new_privs_ctl, .copyout_on_error = false, },
1059         [PROC_NO_NEW_PRIVS_STATUS] =
1060             { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
1061               .esrch_is_einval = false, .no_nonnull_data = false,
1062               .need_candebug = false,
1063               .copyin_sz = 0, .copyout_sz = sizeof(int),
1064               .exec = no_new_privs_status, .copyout_on_error = false, },
1065         [PROC_WXMAP_CTL] =
1066             { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
1067               .esrch_is_einval = false, .no_nonnull_data = false,
1068               .need_candebug = true,
1069               .copyin_sz = sizeof(int), .copyout_sz = 0,
1070               .exec = wxmap_ctl, .copyout_on_error = false, },
1071         [PROC_WXMAP_STATUS] =
1072             { .lock_tree = PCTL_UNLOCKED, .one_proc = true,
1073               .esrch_is_einval = false, .no_nonnull_data = false,
1074               .need_candebug = false,
1075               .copyin_sz = 0, .copyout_sz = sizeof(int),
1076               .exec = wxmap_status, .copyout_on_error = false, },
1077 };
1078
1079 int
1080 sys_procctl(struct thread *td, struct procctl_args *uap)
1081 {
1082         union {
1083                 struct procctl_reaper_status rs;
1084                 struct procctl_reaper_pids rp;
1085                 struct procctl_reaper_kill rk;
1086                 int flags;
1087         } x;
1088         const struct procctl_cmd_info *cmd_info;
1089         int error, error1;
1090
1091         if (uap->com >= PROC_PROCCTL_MD_MIN)
1092                 return (cpu_procctl(td, uap->idtype, uap->id,
1093                     uap->com, uap->data));
1094         if (uap->com == 0 || uap->com >= nitems(procctl_cmds_info))
1095                 return (EINVAL);
1096         cmd_info = &procctl_cmds_info[uap->com];
1097         bzero(&x, sizeof(x));
1098
1099         if (cmd_info->copyin_sz > 0) {
1100                 error = copyin(uap->data, &x, cmd_info->copyin_sz);
1101                 if (error != 0)
1102                         return (error);
1103         } else if (cmd_info->no_nonnull_data && uap->data != NULL) {
1104                 return (EINVAL);
1105         }
1106
1107         error = kern_procctl(td, uap->idtype, uap->id, uap->com, &x);
1108
1109         if (cmd_info->copyout_sz > 0 && (error == 0 ||
1110             cmd_info->copyout_on_error)) {
1111                 error1 = copyout(&x, uap->data, cmd_info->copyout_sz);
1112                 if (error == 0)
1113                         error = error1;
1114         }
1115         return (error);
1116 }
1117
1118 static int
1119 kern_procctl_single(struct thread *td, struct proc *p, int com, void *data)
1120 {
1121
1122         PROC_LOCK_ASSERT(p, MA_OWNED);
1123         return (procctl_cmds_info[com].exec(td, p, data));
1124 }
1125
1126 int
1127 kern_procctl(struct thread *td, idtype_t idtype, id_t id, int com, void *data)
1128 {
1129         struct pgrp *pg;
1130         struct proc *p;
1131         const struct procctl_cmd_info *cmd_info;
1132         int error, first_error, ok;
1133         bool sapblk;
1134
1135         MPASS(com > 0 && com < nitems(procctl_cmds_info));
1136         cmd_info = &procctl_cmds_info[com];
1137         if (idtype != P_PID && cmd_info->one_proc)
1138                 return (EINVAL);
1139
1140         sapblk = false;
1141         if (cmd_info->sapblk != NULL) {
1142                 sapblk = cmd_info->sapblk(td, data);
1143                 if (sapblk && !stop_all_proc_block())
1144                         return (ERESTART);
1145         }
1146
1147         switch (cmd_info->lock_tree) {
1148         case PCTL_XLOCKED:
1149                 sx_xlock(&proctree_lock);
1150                 break;
1151         case PCTL_SLOCKED:
1152                 sx_slock(&proctree_lock);
1153                 break;
1154         default:
1155                 break;
1156         }
1157
1158         switch (idtype) {
1159         case P_PID:
1160                 if (id == 0) {
1161                         p = td->td_proc;
1162                         error = 0;
1163                         PROC_LOCK(p);
1164                 } else {
1165                         p = pfind(id);
1166                         if (p == NULL) {
1167                                 error = cmd_info->esrch_is_einval ?
1168                                     EINVAL : ESRCH;
1169                                 break;
1170                         }
1171                         error = cmd_info->need_candebug ? p_candebug(td, p) :
1172                             p_cansee(td, p);
1173                 }
1174                 if (error == 0)
1175                         error = kern_procctl_single(td, p, com, data);
1176                 PROC_UNLOCK(p);
1177                 break;
1178         case P_PGID:
1179                 /*
1180                  * Attempt to apply the operation to all members of the
1181                  * group.  Ignore processes in the group that can't be
1182                  * seen.  Ignore errors so long as at least one process is
1183                  * able to complete the request successfully.
1184                  */
1185                 pg = pgfind(id);
1186                 if (pg == NULL) {
1187                         error = ESRCH;
1188                         break;
1189                 }
1190                 PGRP_UNLOCK(pg);
1191                 ok = 0;
1192                 first_error = 0;
1193                 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
1194                         PROC_LOCK(p);
1195                         if (p->p_state == PRS_NEW ||
1196                             p->p_state == PRS_ZOMBIE ||
1197                             (cmd_info->need_candebug ? p_candebug(td, p) :
1198                             p_cansee(td, p)) != 0) {
1199                                 PROC_UNLOCK(p);
1200                                 continue;
1201                         }
1202                         error = kern_procctl_single(td, p, com, data);
1203                         PROC_UNLOCK(p);
1204                         if (error == 0)
1205                                 ok = 1;
1206                         else if (first_error == 0)
1207                                 first_error = error;
1208                 }
1209                 if (ok)
1210                         error = 0;
1211                 else if (first_error != 0)
1212                         error = first_error;
1213                 else
1214                         /*
1215                          * Was not able to see any processes in the
1216                          * process group.
1217                          */
1218                         error = ESRCH;
1219                 break;
1220         default:
1221                 error = EINVAL;
1222                 break;
1223         }
1224
1225         switch (cmd_info->lock_tree) {
1226         case PCTL_XLOCKED:
1227                 sx_xunlock(&proctree_lock);
1228                 break;
1229         case PCTL_SLOCKED:
1230                 sx_sunlock(&proctree_lock);
1231                 break;
1232         default:
1233                 break;
1234         }
1235         if (sapblk)
1236                 stop_all_proc_unblock();
1237         return (error);
1238 }