kern_procctl.c revision 276686
1/*-
2 * Copyright (c) 2014 John Baldwin
3 * Copyright (c) 2014 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: stable/10/sys/kern/kern_procctl.c 276686 2015-01-05 03:27:09Z kib $");
32
33#include <sys/param.h>
34#include <sys/systm.h>
35#include <sys/capability.h>
36#include <sys/lock.h>
37#include <sys/mutex.h>
38#include <sys/priv.h>
39#include <sys/proc.h>
40#include <sys/procctl.h>
41#include <sys/sx.h>
42#include <sys/syscallsubr.h>
43#include <sys/sysproto.h>
44#include <sys/wait.h>
45
46static int
47protect_setchild(struct thread *td, struct proc *p, int flags)
48{
49
50	PROC_LOCK_ASSERT(p, MA_OWNED);
51	if (p->p_flag & P_SYSTEM || p_cansched(td, p) != 0)
52		return (0);
53	if (flags & PPROT_SET) {
54		p->p_flag |= P_PROTECTED;
55		if (flags & PPROT_INHERIT)
56			p->p_flag2 |= P2_INHERIT_PROTECTED;
57	} else {
58		p->p_flag &= ~P_PROTECTED;
59		p->p_flag2 &= ~P2_INHERIT_PROTECTED;
60	}
61	return (1);
62}
63
64static int
65protect_setchildren(struct thread *td, struct proc *top, int flags)
66{
67	struct proc *p;
68	int ret;
69
70	p = top;
71	ret = 0;
72	sx_assert(&proctree_lock, SX_LOCKED);
73	for (;;) {
74		ret |= protect_setchild(td, p, flags);
75		PROC_UNLOCK(p);
76		/*
77		 * If this process has children, descend to them next,
78		 * otherwise do any siblings, and if done with this level,
79		 * follow back up the tree (but not past top).
80		 */
81		if (!LIST_EMPTY(&p->p_children))
82			p = LIST_FIRST(&p->p_children);
83		else for (;;) {
84			if (p == top) {
85				PROC_LOCK(p);
86				return (ret);
87			}
88			if (LIST_NEXT(p, p_sibling)) {
89				p = LIST_NEXT(p, p_sibling);
90				break;
91			}
92			p = p->p_pptr;
93		}
94		PROC_LOCK(p);
95	}
96}
97
98static int
99protect_set(struct thread *td, struct proc *p, int flags)
100{
101	int error, ret;
102
103	switch (PPROT_OP(flags)) {
104	case PPROT_SET:
105	case PPROT_CLEAR:
106		break;
107	default:
108		return (EINVAL);
109	}
110
111	if ((PPROT_FLAGS(flags) & ~(PPROT_DESCEND | PPROT_INHERIT)) != 0)
112		return (EINVAL);
113
114	error = priv_check(td, PRIV_VM_MADV_PROTECT);
115	if (error)
116		return (error);
117
118	if (flags & PPROT_DESCEND)
119		ret = protect_setchildren(td, p, flags);
120	else
121		ret = protect_setchild(td, p, flags);
122	if (ret == 0)
123		return (EPERM);
124	return (0);
125}
126
127static int
128reap_acquire(struct thread *td, struct proc *p)
129{
130
131	sx_assert(&proctree_lock, SX_XLOCKED);
132	if (p != curproc)
133		return (EPERM);
134	if ((p->p_treeflag & P_TREE_REAPER) != 0)
135		return (EBUSY);
136	p->p_treeflag |= P_TREE_REAPER;
137	/*
138	 * We do not reattach existing children and the whole tree
139	 * under them to us, since p->p_reaper already seen them.
140	 */
141	return (0);
142}
143
144static int
145reap_release(struct thread *td, struct proc *p)
146{
147
148	sx_assert(&proctree_lock, SX_XLOCKED);
149	if (p != curproc)
150		return (EPERM);
151	if (p == initproc)
152		return (EINVAL);
153	if ((p->p_treeflag & P_TREE_REAPER) == 0)
154		return (EINVAL);
155	reaper_abandon_children(p, false);
156	return (0);
157}
158
159static int
160reap_status(struct thread *td, struct proc *p,
161    struct procctl_reaper_status *rs)
162{
163	struct proc *reap, *p2;
164
165	sx_assert(&proctree_lock, SX_LOCKED);
166	bzero(rs, sizeof(*rs));
167	if ((p->p_treeflag & P_TREE_REAPER) == 0) {
168		reap = p->p_reaper;
169	} else {
170		reap = p;
171		rs->rs_flags |= REAPER_STATUS_OWNED;
172	}
173	if (reap == initproc)
174		rs->rs_flags |= REAPER_STATUS_REALINIT;
175	rs->rs_reaper = reap->p_pid;
176	rs->rs_descendants = 0;
177	rs->rs_children = 0;
178	if (!LIST_EMPTY(&reap->p_reaplist)) {
179		KASSERT(!LIST_EMPTY(&reap->p_children), ("no children"));
180		rs->rs_pid = LIST_FIRST(&reap->p_children)->p_pid;
181		LIST_FOREACH(p2, &reap->p_reaplist, p_reapsibling) {
182			if (proc_realparent(p2) == reap)
183				rs->rs_children++;
184			rs->rs_descendants++;
185		}
186	} else {
187		rs->rs_pid = -1;
188		KASSERT(LIST_EMPTY(&reap->p_reaplist), ("reap children list"));
189		KASSERT(LIST_EMPTY(&reap->p_children), ("children list"));
190	}
191	return (0);
192}
193
194static int
195reap_getpids(struct thread *td, struct proc *p, struct procctl_reaper_pids *rp)
196{
197	struct proc *reap, *p2;
198	struct procctl_reaper_pidinfo *pi, *pip;
199	u_int i, n;
200	int error;
201
202	sx_assert(&proctree_lock, SX_LOCKED);
203	PROC_UNLOCK(p);
204	reap = (p->p_treeflag & P_TREE_REAPER) == 0 ? p->p_reaper : p;
205	n = i = 0;
206	error = 0;
207	LIST_FOREACH(p2, &reap->p_reaplist, p_reapsibling)
208		n++;
209	sx_unlock(&proctree_lock);
210	if (rp->rp_count < n)
211		n = rp->rp_count;
212	pi = malloc(n * sizeof(*pi), M_TEMP, M_WAITOK);
213	sx_slock(&proctree_lock);
214	LIST_FOREACH(p2, &reap->p_reaplist, p_reapsibling) {
215		if (i == n)
216			break;
217		pip = &pi[i];
218		bzero(pip, sizeof(*pip));
219		pip->pi_pid = p2->p_pid;
220		pip->pi_subtree = p2->p_reapsubtree;
221		pip->pi_flags = REAPER_PIDINFO_VALID;
222		if (proc_realparent(p2) == reap)
223			pip->pi_flags |= REAPER_PIDINFO_CHILD;
224		i++;
225	}
226	sx_sunlock(&proctree_lock);
227	error = copyout(pi, rp->rp_pids, i * sizeof(*pi));
228	free(pi, M_TEMP);
229	sx_slock(&proctree_lock);
230	PROC_LOCK(p);
231	return (error);
232}
233
234static int
235reap_kill(struct thread *td, struct proc *p, struct procctl_reaper_kill *rk)
236{
237	struct proc *reap, *p2;
238	ksiginfo_t ksi;
239	int error, error1;
240
241	sx_assert(&proctree_lock, SX_LOCKED);
242	PROC_UNLOCK(p);
243	if (IN_CAPABILITY_MODE(td))
244		return (ECAPMODE);
245	if (rk->rk_sig <= 0 || rk->rk_sig > _SIG_MAXSIG)
246		return (EINVAL);
247	if ((rk->rk_flags & ~REAPER_KILL_CHILDREN) != 0)
248		return (EINVAL);
249	reap = (p->p_treeflag & P_TREE_REAPER) == 0 ? p->p_reaper : p;
250	ksiginfo_init(&ksi);
251	ksi.ksi_signo = rk->rk_sig;
252	ksi.ksi_code = SI_USER;
253	ksi.ksi_pid = td->td_proc->p_pid;
254	ksi.ksi_uid = td->td_ucred->cr_ruid;
255	error = ESRCH;
256	rk->rk_killed = 0;
257	rk->rk_fpid = -1;
258	for (p2 = (rk->rk_flags & REAPER_KILL_CHILDREN) != 0 ?
259	    LIST_FIRST(&reap->p_children) : LIST_FIRST(&reap->p_reaplist);
260	    p2 != NULL;
261	    p2 = (rk->rk_flags & REAPER_KILL_CHILDREN) != 0 ?
262	    LIST_NEXT(p2, p_sibling) : LIST_NEXT(p2, p_reapsibling)) {
263		if ((rk->rk_flags & REAPER_KILL_SUBTREE) != 0 &&
264		    p2->p_reapsubtree != rk->rk_subtree)
265			continue;
266		PROC_LOCK(p2);
267		error1 = p_cansignal(td, p2, rk->rk_sig);
268		if (error1 == 0) {
269			pksignal(p2, rk->rk_sig, &ksi);
270			rk->rk_killed++;
271			error = error1;
272		} else if (error == ESRCH) {
273			error = error1;
274			rk->rk_fpid = p2->p_pid;
275		}
276		PROC_UNLOCK(p2);
277		/* Do not end the loop on error, signal everything we can. */
278	}
279	PROC_LOCK(p);
280	return (error);
281}
282
283#ifndef _SYS_SYSPROTO_H_
284struct procctl_args {
285	idtype_t idtype;
286	id_t	id;
287	int	com;
288	void	*data;
289};
290#endif
291/* ARGSUSED */
292int
293sys_procctl(struct thread *td, struct procctl_args *uap)
294{
295	void *data;
296	union {
297		struct procctl_reaper_status rs;
298		struct procctl_reaper_pids rp;
299		struct procctl_reaper_kill rk;
300	} x;
301	int error, error1, flags;
302
303	switch (uap->com) {
304	case PROC_SPROTECT:
305		error = copyin(uap->data, &flags, sizeof(flags));
306		if (error != 0)
307			return (error);
308		data = &flags;
309		break;
310	case PROC_REAP_ACQUIRE:
311	case PROC_REAP_RELEASE:
312		if (uap->data != NULL)
313			return (EINVAL);
314		data = NULL;
315		break;
316	case PROC_REAP_STATUS:
317		data = &x.rs;
318		break;
319	case PROC_REAP_GETPIDS:
320		error = copyin(uap->data, &x.rp, sizeof(x.rp));
321		if (error != 0)
322			return (error);
323		data = &x.rp;
324		break;
325	case PROC_REAP_KILL:
326		error = copyin(uap->data, &x.rk, sizeof(x.rk));
327		if (error != 0)
328			return (error);
329		data = &x.rk;
330		break;
331	default:
332		return (EINVAL);
333	}
334	error = kern_procctl(td, uap->idtype, uap->id, uap->com, data);
335	switch (uap->com) {
336	case PROC_REAP_STATUS:
337		if (error == 0)
338			error = copyout(&x.rs, uap->data, sizeof(x.rs));
339		break;
340	case PROC_REAP_KILL:
341		error1 = copyout(&x.rk, uap->data, sizeof(x.rk));
342		if (error == 0)
343			error = error1;
344		break;
345	}
346	return (error);
347}
348
349static int
350kern_procctl_single(struct thread *td, struct proc *p, int com, void *data)
351{
352
353	PROC_LOCK_ASSERT(p, MA_OWNED);
354	switch (com) {
355	case PROC_SPROTECT:
356		return (protect_set(td, p, *(int *)data));
357	case PROC_REAP_ACQUIRE:
358		return (reap_acquire(td, p));
359	case PROC_REAP_RELEASE:
360		return (reap_release(td, p));
361	case PROC_REAP_STATUS:
362		return (reap_status(td, p, data));
363	case PROC_REAP_GETPIDS:
364		return (reap_getpids(td, p, data));
365	case PROC_REAP_KILL:
366		return (reap_kill(td, p, data));
367	default:
368		return (EINVAL);
369	}
370}
371
372int
373kern_procctl(struct thread *td, idtype_t idtype, id_t id, int com, void *data)
374{
375	struct pgrp *pg;
376	struct proc *p;
377	int error, first_error, ok;
378
379	switch (com) {
380	case PROC_REAP_ACQUIRE:
381	case PROC_REAP_RELEASE:
382	case PROC_REAP_STATUS:
383	case PROC_REAP_GETPIDS:
384	case PROC_REAP_KILL:
385		if (idtype != P_PID)
386			return (EINVAL);
387	}
388
389	switch (com) {
390	case PROC_SPROTECT:
391	case PROC_REAP_STATUS:
392	case PROC_REAP_GETPIDS:
393	case PROC_REAP_KILL:
394		sx_slock(&proctree_lock);
395		break;
396	case PROC_REAP_ACQUIRE:
397	case PROC_REAP_RELEASE:
398		sx_xlock(&proctree_lock);
399		break;
400	default:
401		return (EINVAL);
402	}
403
404	switch (idtype) {
405	case P_PID:
406		p = pfind(id);
407		if (p == NULL) {
408			error = ESRCH;
409			break;
410		}
411		error = p_cansee(td, p);
412		if (error == 0)
413			error = kern_procctl_single(td, p, com, data);
414		PROC_UNLOCK(p);
415		break;
416	case P_PGID:
417		/*
418		 * Attempt to apply the operation to all members of the
419		 * group.  Ignore processes in the group that can't be
420		 * seen.  Ignore errors so long as at least one process is
421		 * able to complete the request successfully.
422		 */
423		pg = pgfind(id);
424		if (pg == NULL) {
425			error = ESRCH;
426			break;
427		}
428		PGRP_UNLOCK(pg);
429		ok = 0;
430		first_error = 0;
431		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
432			PROC_LOCK(p);
433			if (p->p_state == PRS_NEW || p_cansee(td, p) != 0) {
434				PROC_UNLOCK(p);
435				continue;
436			}
437			error = kern_procctl_single(td, p, com, data);
438			PROC_UNLOCK(p);
439			if (error == 0)
440				ok = 1;
441			else if (first_error == 0)
442				first_error = error;
443		}
444		if (ok)
445			error = 0;
446		else if (first_error != 0)
447			error = first_error;
448		else
449			/*
450			 * Was not able to see any processes in the
451			 * process group.
452			 */
453			error = ESRCH;
454		break;
455	default:
456		error = EINVAL;
457		break;
458	}
459	sx_unlock(&proctree_lock);
460	return (error);
461}
462