xref: /freebsd/sys/kern/kern_proc.c (revision f53355131f65d64e7643d734dbcd4fb2a5de20ed)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1989, 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)kern_proc.c	8.7 (Berkeley) 2/14/95
32  */
33 
34 #include <sys/cdefs.h>
35 #include "opt_ddb.h"
36 #include "opt_ktrace.h"
37 #include "opt_kstack_pages.h"
38 #include "opt_stack.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/bitstring.h>
43 #include <sys/elf.h>
44 #include <sys/eventhandler.h>
45 #include <sys/exec.h>
46 #include <sys/fcntl.h>
47 #include <sys/jail.h>
48 #include <sys/kernel.h>
49 #include <sys/limits.h>
50 #include <sys/lock.h>
51 #include <sys/loginclass.h>
52 #include <sys/malloc.h>
53 #include <sys/mman.h>
54 #include <sys/mount.h>
55 #include <sys/mutex.h>
56 #include <sys/namei.h>
57 #include <sys/proc.h>
58 #include <sys/ptrace.h>
59 #include <sys/refcount.h>
60 #include <sys/resourcevar.h>
61 #include <sys/rwlock.h>
62 #include <sys/sbuf.h>
63 #include <sys/sysent.h>
64 #include <sys/sched.h>
65 #include <sys/smp.h>
66 #include <sys/stack.h>
67 #include <sys/stat.h>
68 #include <sys/dtrace_bsd.h>
69 #include <sys/sysctl.h>
70 #include <sys/filedesc.h>
71 #include <sys/tty.h>
72 #include <sys/signalvar.h>
73 #include <sys/sdt.h>
74 #include <sys/sx.h>
75 #include <sys/user.h>
76 #include <sys/vnode.h>
77 #include <sys/wait.h>
78 #ifdef KTRACE
79 #include <sys/ktrace.h>
80 #endif
81 
82 #ifdef DDB
83 #include <ddb/ddb.h>
84 #endif
85 
86 #include <vm/vm.h>
87 #include <vm/vm_param.h>
88 #include <vm/vm_extern.h>
89 #include <vm/pmap.h>
90 #include <vm/vm_map.h>
91 #include <vm/vm_object.h>
92 #include <vm/vm_page.h>
93 #include <vm/uma.h>
94 
95 #include <fs/devfs/devfs.h>
96 
97 #ifdef COMPAT_FREEBSD32
98 #include <compat/freebsd32/freebsd32.h>
99 #include <compat/freebsd32/freebsd32_util.h>
100 #endif
101 
102 SDT_PROVIDER_DEFINE(proc);
103 
104 MALLOC_DEFINE(M_SESSION, "session", "session header");
105 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
106 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
107 
108 static void doenterpgrp(struct proc *, struct pgrp *);
109 static void orphanpg(struct pgrp *pg);
110 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
111 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
112 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
113     int preferthread);
114 static void pgdelete(struct pgrp *);
115 static int pgrp_init(void *mem, int size, int flags);
116 static int proc_ctor(void *mem, int size, void *arg, int flags);
117 static void proc_dtor(void *mem, int size, void *arg);
118 static int proc_init(void *mem, int size, int flags);
119 static void proc_fini(void *mem, int size);
120 static void pargs_free(struct pargs *pa);
121 
122 /*
123  * Other process lists
124  */
125 struct pidhashhead *pidhashtbl = NULL;
126 struct sx *pidhashtbl_lock;
127 u_long pidhash;
128 u_long pidhashlock;
129 struct pgrphashhead *pgrphashtbl;
130 u_long pgrphash;
131 struct proclist allproc = LIST_HEAD_INITIALIZER(allproc);
132 struct sx __exclusive_cache_line allproc_lock;
133 struct sx __exclusive_cache_line proctree_lock;
134 struct mtx __exclusive_cache_line ppeers_lock;
135 struct mtx __exclusive_cache_line procid_lock;
136 uma_zone_t proc_zone;
137 uma_zone_t pgrp_zone;
138 
139 /*
140  * The offset of various fields in struct proc and struct thread.
141  * These are used by kernel debuggers to enumerate kernel threads and
142  * processes.
143  */
144 const int proc_off_p_pid = offsetof(struct proc, p_pid);
145 const int proc_off_p_comm = offsetof(struct proc, p_comm);
146 const int proc_off_p_list = offsetof(struct proc, p_list);
147 const int proc_off_p_hash = offsetof(struct proc, p_hash);
148 const int proc_off_p_threads = offsetof(struct proc, p_threads);
149 const int thread_off_td_tid = offsetof(struct thread, td_tid);
150 const int thread_off_td_name = offsetof(struct thread, td_name);
151 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
152 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
153 const int thread_off_td_plist = offsetof(struct thread, td_plist);
154 
155 EVENTHANDLER_LIST_DEFINE(process_ctor);
156 EVENTHANDLER_LIST_DEFINE(process_dtor);
157 EVENTHANDLER_LIST_DEFINE(process_init);
158 EVENTHANDLER_LIST_DEFINE(process_fini);
159 EVENTHANDLER_LIST_DEFINE(process_exit);
160 EVENTHANDLER_LIST_DEFINE(process_fork);
161 EVENTHANDLER_LIST_DEFINE(process_exec);
162 
163 int kstack_pages = KSTACK_PAGES;
164 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RDTUN | CTLFLAG_NOFETCH,
165     &kstack_pages, 0,
166     "Kernel stack size in pages");
167 static int vmmap_skip_res_cnt = 0;
168 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
169     &vmmap_skip_res_cnt, 0,
170     "Skip calculation of the pages resident count in kern.proc.vmmap");
171 
172 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
173 #ifdef COMPAT_FREEBSD32
174 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
175 #endif
176 
177 /*
178  * Initialize global process hashing structures.
179  */
180 void
181 procinit(void)
182 {
183 	u_long i;
184 
185 	sx_init(&allproc_lock, "allproc");
186 	sx_init(&proctree_lock, "proctree");
187 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
188 	mtx_init(&procid_lock, "procid", NULL, MTX_DEF);
189 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
190 	pidhashlock = (pidhash + 1) / 64;
191 	if (pidhashlock > 0)
192 		pidhashlock--;
193 	pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1),
194 	    M_PROC, M_WAITOK | M_ZERO);
195 	for (i = 0; i < pidhashlock + 1; i++)
196 		sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK);
197 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
198 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
199 	    proc_ctor, proc_dtor, proc_init, proc_fini,
200 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
201 	pgrp_zone = uma_zcreate("PGRP", sizeof(struct pgrp), NULL, NULL,
202 	    pgrp_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
203 	uihashinit();
204 }
205 
206 /*
207  * Prepare a proc for use.
208  */
209 static int
210 proc_ctor(void *mem, int size, void *arg, int flags)
211 {
212 	struct proc *p;
213 	struct thread *td;
214 
215 	p = (struct proc *)mem;
216 #ifdef KDTRACE_HOOKS
217 	kdtrace_proc_ctor(p);
218 #endif
219 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
220 	td = FIRST_THREAD_IN_PROC(p);
221 	if (td != NULL) {
222 		/* Make sure all thread constructors are executed */
223 		EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
224 	}
225 	return (0);
226 }
227 
228 /*
229  * Reclaim a proc after use.
230  */
231 static void
232 proc_dtor(void *mem, int size, void *arg)
233 {
234 	struct proc *p;
235 	struct thread *td;
236 
237 	/* INVARIANTS checks go here */
238 	p = (struct proc *)mem;
239 	td = FIRST_THREAD_IN_PROC(p);
240 	if (td != NULL) {
241 #ifdef INVARIANTS
242 		KASSERT((p->p_numthreads == 1),
243 		    ("bad number of threads in exiting process"));
244 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
245 #endif
246 		/* Free all OSD associated to this thread. */
247 		osd_thread_exit(td);
248 		ast_kclear(td);
249 
250 		/* Make sure all thread destructors are executed */
251 		EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
252 	}
253 	EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
254 #ifdef KDTRACE_HOOKS
255 	kdtrace_proc_dtor(p);
256 #endif
257 	if (p->p_ksi != NULL)
258 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
259 }
260 
261 /*
262  * Initialize type-stable parts of a proc (when newly created).
263  */
264 static int
265 proc_init(void *mem, int size, int flags)
266 {
267 	struct proc *p;
268 
269 	p = (struct proc *)mem;
270 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
271 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
272 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
273 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
274 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
275 	cv_init(&p->p_pwait, "ppwait");
276 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
277 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
278 	p->p_stats = pstats_alloc();
279 	p->p_pgrp = NULL;
280 	return (0);
281 }
282 
283 /*
284  * UMA should ensure that this function is never called.
285  * Freeing a proc structure would violate type stability.
286  */
287 static void
288 proc_fini(void *mem, int size)
289 {
290 #ifdef notnow
291 	struct proc *p;
292 
293 	p = (struct proc *)mem;
294 	EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
295 	pstats_free(p->p_stats);
296 	thread_free(FIRST_THREAD_IN_PROC(p));
297 	mtx_destroy(&p->p_mtx);
298 	if (p->p_ksi != NULL)
299 		ksiginfo_free(p->p_ksi);
300 #else
301 	panic("proc reclaimed");
302 #endif
303 }
304 
305 static int
306 pgrp_init(void *mem, int size, int flags)
307 {
308 	struct pgrp *pg;
309 
310 	pg = mem;
311 	mtx_init(&pg->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
312 	sx_init(&pg->pg_killsx, "killpg racer");
313 	return (0);
314 }
315 
316 /*
317  * PID space management.
318  *
319  * These bitmaps are used by fork_findpid.
320  */
321 bitstr_t bit_decl(proc_id_pidmap, PID_MAX);
322 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX);
323 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX);
324 bitstr_t bit_decl(proc_id_reapmap, PID_MAX);
325 
326 static bitstr_t *proc_id_array[] = {
327 	proc_id_pidmap,
328 	proc_id_grpidmap,
329 	proc_id_sessidmap,
330 	proc_id_reapmap,
331 };
332 
333 void
334 proc_id_set(int type, pid_t id)
335 {
336 
337 	KASSERT(type >= 0 && type < nitems(proc_id_array),
338 	    ("invalid type %d\n", type));
339 	mtx_lock(&procid_lock);
340 	KASSERT(bit_test(proc_id_array[type], id) == 0,
341 	    ("bit %d already set in %d\n", id, type));
342 	bit_set(proc_id_array[type], id);
343 	mtx_unlock(&procid_lock);
344 }
345 
346 void
347 proc_id_set_cond(int type, pid_t id)
348 {
349 
350 	KASSERT(type >= 0 && type < nitems(proc_id_array),
351 	    ("invalid type %d\n", type));
352 	if (bit_test(proc_id_array[type], id))
353 		return;
354 	mtx_lock(&procid_lock);
355 	bit_set(proc_id_array[type], id);
356 	mtx_unlock(&procid_lock);
357 }
358 
359 void
360 proc_id_clear(int type, pid_t id)
361 {
362 
363 	KASSERT(type >= 0 && type < nitems(proc_id_array),
364 	    ("invalid type %d\n", type));
365 	mtx_lock(&procid_lock);
366 	KASSERT(bit_test(proc_id_array[type], id) != 0,
367 	    ("bit %d not set in %d\n", id, type));
368 	bit_clear(proc_id_array[type], id);
369 	mtx_unlock(&procid_lock);
370 }
371 
372 /*
373  * Is p an inferior of the current process?
374  */
375 int
376 inferior(struct proc *p)
377 {
378 
379 	sx_assert(&proctree_lock, SX_LOCKED);
380 	PROC_LOCK_ASSERT(p, MA_OWNED);
381 	for (; p != curproc; p = proc_realparent(p)) {
382 		if (p->p_pid == 0)
383 			return (0);
384 	}
385 	return (1);
386 }
387 
388 /*
389  * Shared lock all the pid hash lists.
390  */
391 void
392 pidhash_slockall(void)
393 {
394 	u_long i;
395 
396 	for (i = 0; i < pidhashlock + 1; i++)
397 		sx_slock(&pidhashtbl_lock[i]);
398 }
399 
400 /*
401  * Shared unlock all the pid hash lists.
402  */
403 void
404 pidhash_sunlockall(void)
405 {
406 	u_long i;
407 
408 	for (i = 0; i < pidhashlock + 1; i++)
409 		sx_sunlock(&pidhashtbl_lock[i]);
410 }
411 
412 /*
413  * Similar to pfind_any(), this function finds zombies.
414  */
415 struct proc *
416 pfind_any_locked(pid_t pid)
417 {
418 	struct proc *p;
419 
420 	sx_assert(PIDHASHLOCK(pid), SX_LOCKED);
421 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
422 		if (p->p_pid == pid) {
423 			PROC_LOCK(p);
424 			if (p->p_state == PRS_NEW) {
425 				PROC_UNLOCK(p);
426 				p = NULL;
427 			}
428 			break;
429 		}
430 	}
431 	return (p);
432 }
433 
434 /*
435  * Locate a process by number.
436  *
437  * By not returning processes in the PRS_NEW state, we allow callers to avoid
438  * testing for that condition to avoid dereferencing p_ucred, et al.
439  */
440 static __always_inline struct proc *
441 _pfind(pid_t pid, bool zombie)
442 {
443 	struct proc *p;
444 
445 	p = curproc;
446 	if (p->p_pid == pid) {
447 		PROC_LOCK(p);
448 		return (p);
449 	}
450 	sx_slock(PIDHASHLOCK(pid));
451 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
452 		if (p->p_pid == pid) {
453 			PROC_LOCK(p);
454 			if (p->p_state == PRS_NEW ||
455 			    (!zombie && p->p_state == PRS_ZOMBIE)) {
456 				PROC_UNLOCK(p);
457 				p = NULL;
458 			}
459 			break;
460 		}
461 	}
462 	sx_sunlock(PIDHASHLOCK(pid));
463 	return (p);
464 }
465 
466 struct proc *
467 pfind(pid_t pid)
468 {
469 
470 	return (_pfind(pid, false));
471 }
472 
473 /*
474  * Same as pfind but allow zombies.
475  */
476 struct proc *
477 pfind_any(pid_t pid)
478 {
479 
480 	return (_pfind(pid, true));
481 }
482 
483 /*
484  * Locate a process group by number.
485  * The caller must hold proctree_lock.
486  */
487 struct pgrp *
488 pgfind(pid_t pgid)
489 {
490 	struct pgrp *pgrp;
491 
492 	sx_assert(&proctree_lock, SX_LOCKED);
493 
494 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
495 		if (pgrp->pg_id == pgid) {
496 			PGRP_LOCK(pgrp);
497 			return (pgrp);
498 		}
499 	}
500 	return (NULL);
501 }
502 
503 /*
504  * Locate process and do additional manipulations, depending on flags.
505  */
506 int
507 pget(pid_t pid, int flags, struct proc **pp)
508 {
509 	struct proc *p;
510 	struct thread *td1;
511 	int error;
512 
513 	p = curproc;
514 	if (p->p_pid == pid) {
515 		PROC_LOCK(p);
516 	} else {
517 		p = NULL;
518 		if (pid <= PID_MAX) {
519 			if ((flags & PGET_NOTWEXIT) == 0)
520 				p = pfind_any(pid);
521 			else
522 				p = pfind(pid);
523 		} else if ((flags & PGET_NOTID) == 0) {
524 			td1 = tdfind(pid, -1);
525 			if (td1 != NULL)
526 				p = td1->td_proc;
527 		}
528 		if (p == NULL)
529 			return (ESRCH);
530 		if ((flags & PGET_CANSEE) != 0) {
531 			error = p_cansee(curthread, p);
532 			if (error != 0)
533 				goto errout;
534 		}
535 	}
536 	if ((flags & PGET_CANDEBUG) != 0) {
537 		error = p_candebug(curthread, p);
538 		if (error != 0)
539 			goto errout;
540 	}
541 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
542 		error = EPERM;
543 		goto errout;
544 	}
545 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
546 		error = ESRCH;
547 		goto errout;
548 	}
549 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
550 		/*
551 		 * XXXRW: Not clear ESRCH is the right error during proc
552 		 * execve().
553 		 */
554 		error = ESRCH;
555 		goto errout;
556 	}
557 	if ((flags & PGET_HOLD) != 0) {
558 		_PHOLD(p);
559 		PROC_UNLOCK(p);
560 	}
561 	*pp = p;
562 	return (0);
563 errout:
564 	PROC_UNLOCK(p);
565 	return (error);
566 }
567 
568 /*
569  * Create a new process group.
570  * pgid must be equal to the pid of p.
571  * Begin a new session if required.
572  */
573 int
574 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
575 {
576 	struct pgrp *old_pgrp;
577 
578 	sx_assert(&proctree_lock, SX_XLOCKED);
579 
580 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
581 	KASSERT(p->p_pid == pgid,
582 	    ("enterpgrp: new pgrp and pid != pgid"));
583 	KASSERT(pgfind(pgid) == NULL,
584 	    ("enterpgrp: pgrp with pgid exists"));
585 	KASSERT(!SESS_LEADER(p),
586 	    ("enterpgrp: session leader attempted setpgrp"));
587 
588 	old_pgrp = p->p_pgrp;
589 	if (!sx_try_xlock(&old_pgrp->pg_killsx)) {
590 		sx_xunlock(&proctree_lock);
591 		sx_xlock(&old_pgrp->pg_killsx);
592 		sx_xunlock(&old_pgrp->pg_killsx);
593 		return (ERESTART);
594 	}
595 	MPASS(old_pgrp == p->p_pgrp);
596 
597 	if (sess != NULL) {
598 		/*
599 		 * new session
600 		 */
601 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
602 		PROC_LOCK(p);
603 		p->p_flag &= ~P_CONTROLT;
604 		PROC_UNLOCK(p);
605 		PGRP_LOCK(pgrp);
606 		sess->s_leader = p;
607 		sess->s_sid = p->p_pid;
608 		proc_id_set(PROC_ID_SESSION, p->p_pid);
609 		refcount_init(&sess->s_count, 1);
610 		sess->s_ttyvp = NULL;
611 		sess->s_ttydp = NULL;
612 		sess->s_ttyp = NULL;
613 		bcopy(p->p_session->s_login, sess->s_login,
614 			    sizeof(sess->s_login));
615 		pgrp->pg_session = sess;
616 		KASSERT(p == curproc,
617 		    ("enterpgrp: mksession and p != curproc"));
618 	} else {
619 		pgrp->pg_session = p->p_session;
620 		sess_hold(pgrp->pg_session);
621 		PGRP_LOCK(pgrp);
622 	}
623 	pgrp->pg_id = pgid;
624 	proc_id_set(PROC_ID_GROUP, p->p_pid);
625 	LIST_INIT(&pgrp->pg_members);
626 	pgrp->pg_flags = 0;
627 
628 	/*
629 	 * As we have an exclusive lock of proctree_lock,
630 	 * this should not deadlock.
631 	 */
632 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
633 	SLIST_INIT(&pgrp->pg_sigiolst);
634 	PGRP_UNLOCK(pgrp);
635 
636 	doenterpgrp(p, pgrp);
637 
638 	sx_xunlock(&old_pgrp->pg_killsx);
639 	return (0);
640 }
641 
642 /*
643  * Move p to an existing process group
644  */
645 int
646 enterthispgrp(struct proc *p, struct pgrp *pgrp)
647 {
648 	struct pgrp *old_pgrp;
649 
650 	sx_assert(&proctree_lock, SX_XLOCKED);
651 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
652 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
653 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
654 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
655 	KASSERT(pgrp->pg_session == p->p_session,
656 	    ("%s: pgrp's session %p, p->p_session %p proc %p\n",
657 	    __func__, pgrp->pg_session, p->p_session, p));
658 	KASSERT(pgrp != p->p_pgrp,
659 	    ("%s: p %p belongs to pgrp %p", __func__, p, pgrp));
660 
661 	old_pgrp = p->p_pgrp;
662 	if (!sx_try_xlock(&old_pgrp->pg_killsx)) {
663 		sx_xunlock(&proctree_lock);
664 		sx_xlock(&old_pgrp->pg_killsx);
665 		sx_xunlock(&old_pgrp->pg_killsx);
666 		return (ERESTART);
667 	}
668 	MPASS(old_pgrp == p->p_pgrp);
669 	if (!sx_try_xlock(&pgrp->pg_killsx)) {
670 		sx_xunlock(&old_pgrp->pg_killsx);
671 		sx_xunlock(&proctree_lock);
672 		sx_xlock(&pgrp->pg_killsx);
673 		sx_xunlock(&pgrp->pg_killsx);
674 		return (ERESTART);
675 	}
676 
677 	doenterpgrp(p, pgrp);
678 
679 	sx_xunlock(&pgrp->pg_killsx);
680 	sx_xunlock(&old_pgrp->pg_killsx);
681 	return (0);
682 }
683 
684 /*
685  * If true, any child of q which belongs to group pgrp, qualifies the
686  * process group pgrp as not orphaned.
687  */
688 static bool
689 isjobproc(struct proc *q, struct pgrp *pgrp)
690 {
691 	sx_assert(&proctree_lock, SX_LOCKED);
692 
693 	return (q->p_pgrp != pgrp &&
694 	    q->p_pgrp->pg_session == pgrp->pg_session);
695 }
696 
697 static struct proc *
698 jobc_reaper(struct proc *p)
699 {
700 	struct proc *pp;
701 
702 	sx_assert(&proctree_lock, SA_LOCKED);
703 
704 	for (pp = p;;) {
705 		pp = pp->p_reaper;
706 		if (pp->p_reaper == pp ||
707 		    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
708 			return (pp);
709 	}
710 }
711 
712 static struct proc *
713 jobc_parent(struct proc *p, struct proc *p_exiting)
714 {
715 	struct proc *pp;
716 
717 	sx_assert(&proctree_lock, SA_LOCKED);
718 
719 	pp = proc_realparent(p);
720 	if (pp->p_pptr == NULL || pp == p_exiting ||
721 	    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
722 		return (pp);
723 	return (jobc_reaper(pp));
724 }
725 
726 static int
727 pgrp_calc_jobc(struct pgrp *pgrp)
728 {
729 	struct proc *q;
730 	int cnt;
731 
732 #ifdef INVARIANTS
733 	if (!mtx_owned(&pgrp->pg_mtx))
734 		sx_assert(&proctree_lock, SA_LOCKED);
735 #endif
736 
737 	cnt = 0;
738 	LIST_FOREACH(q, &pgrp->pg_members, p_pglist) {
739 		if ((q->p_treeflag & P_TREE_GRPEXITED) != 0 ||
740 		    q->p_pptr == NULL)
741 			continue;
742 		if (isjobproc(jobc_parent(q, NULL), pgrp))
743 			cnt++;
744 	}
745 	return (cnt);
746 }
747 
748 /*
749  * Move p to a process group
750  */
751 static void
752 doenterpgrp(struct proc *p, struct pgrp *pgrp)
753 {
754 	struct pgrp *savepgrp;
755 	struct proc *pp;
756 
757 	sx_assert(&proctree_lock, SX_XLOCKED);
758 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
759 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
760 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
761 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
762 
763 	savepgrp = p->p_pgrp;
764 	pp = jobc_parent(p, NULL);
765 
766 	PGRP_LOCK(pgrp);
767 	PGRP_LOCK(savepgrp);
768 	if (isjobproc(pp, savepgrp) && pgrp_calc_jobc(savepgrp) == 1)
769 		orphanpg(savepgrp);
770 	PROC_LOCK(p);
771 	LIST_REMOVE(p, p_pglist);
772 	p->p_pgrp = pgrp;
773 	PROC_UNLOCK(p);
774 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
775 	if (isjobproc(pp, pgrp))
776 		pgrp->pg_flags &= ~PGRP_ORPHANED;
777 	PGRP_UNLOCK(savepgrp);
778 	PGRP_UNLOCK(pgrp);
779 	if (LIST_EMPTY(&savepgrp->pg_members))
780 		pgdelete(savepgrp);
781 }
782 
783 /*
784  * remove process from process group
785  */
786 int
787 leavepgrp(struct proc *p)
788 {
789 	struct pgrp *savepgrp;
790 
791 	sx_assert(&proctree_lock, SX_XLOCKED);
792 	savepgrp = p->p_pgrp;
793 	PGRP_LOCK(savepgrp);
794 	PROC_LOCK(p);
795 	LIST_REMOVE(p, p_pglist);
796 	p->p_pgrp = NULL;
797 	PROC_UNLOCK(p);
798 	PGRP_UNLOCK(savepgrp);
799 	if (LIST_EMPTY(&savepgrp->pg_members))
800 		pgdelete(savepgrp);
801 	return (0);
802 }
803 
804 /*
805  * delete a process group
806  */
807 static void
808 pgdelete(struct pgrp *pgrp)
809 {
810 	struct session *savesess;
811 	struct tty *tp;
812 
813 	sx_assert(&proctree_lock, SX_XLOCKED);
814 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
815 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
816 
817 	/*
818 	 * Reset any sigio structures pointing to us as a result of
819 	 * F_SETOWN with our pgid.  The proctree lock ensures that
820 	 * new sigio structures will not be added after this point.
821 	 */
822 	funsetownlst(&pgrp->pg_sigiolst);
823 
824 	PGRP_LOCK(pgrp);
825 	tp = pgrp->pg_session->s_ttyp;
826 	LIST_REMOVE(pgrp, pg_hash);
827 	savesess = pgrp->pg_session;
828 	PGRP_UNLOCK(pgrp);
829 
830 	/* Remove the reference to the pgrp before deallocating it. */
831 	if (tp != NULL) {
832 		tty_lock(tp);
833 		tty_rel_pgrp(tp, pgrp);
834 	}
835 
836 	proc_id_clear(PROC_ID_GROUP, pgrp->pg_id);
837 	uma_zfree(pgrp_zone, pgrp);
838 	sess_release(savesess);
839 }
840 
841 
842 static void
843 fixjobc_kill(struct proc *p)
844 {
845 	struct proc *q;
846 	struct pgrp *pgrp;
847 
848 	sx_assert(&proctree_lock, SX_LOCKED);
849 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
850 	pgrp = p->p_pgrp;
851 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
852 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
853 
854 	/*
855 	 * p no longer affects process group orphanage for children.
856 	 * It is marked by the flag because p is only physically
857 	 * removed from its process group on wait(2).
858 	 */
859 	MPASS((p->p_treeflag & P_TREE_GRPEXITED) == 0);
860 	p->p_treeflag |= P_TREE_GRPEXITED;
861 
862 	/*
863 	 * Check if exiting p orphans its own group.
864 	 */
865 	pgrp = p->p_pgrp;
866 	if (isjobproc(jobc_parent(p, NULL), pgrp)) {
867 		PGRP_LOCK(pgrp);
868 		if (pgrp_calc_jobc(pgrp) == 0)
869 			orphanpg(pgrp);
870 		PGRP_UNLOCK(pgrp);
871 	}
872 
873 	/*
874 	 * Check this process' children to see whether they qualify
875 	 * their process groups after reparenting to reaper.
876 	 */
877 	LIST_FOREACH(q, &p->p_children, p_sibling) {
878 		pgrp = q->p_pgrp;
879 		PGRP_LOCK(pgrp);
880 		if (pgrp_calc_jobc(pgrp) == 0) {
881 			/*
882 			 * We want to handle exactly the children that
883 			 * has p as realparent.  Then, when calculating
884 			 * jobc_parent for children, we should ignore
885 			 * P_TREE_GRPEXITED flag already set on p.
886 			 */
887 			if (jobc_parent(q, p) == p && isjobproc(p, pgrp))
888 				orphanpg(pgrp);
889 		} else
890 			pgrp->pg_flags &= ~PGRP_ORPHANED;
891 		PGRP_UNLOCK(pgrp);
892 	}
893 	LIST_FOREACH(q, &p->p_orphans, p_orphan) {
894 		pgrp = q->p_pgrp;
895 		PGRP_LOCK(pgrp);
896 		if (pgrp_calc_jobc(pgrp) == 0) {
897 			if (isjobproc(p, pgrp))
898 				orphanpg(pgrp);
899 		} else
900 			pgrp->pg_flags &= ~PGRP_ORPHANED;
901 		PGRP_UNLOCK(pgrp);
902 	}
903 }
904 
905 void
906 killjobc(void)
907 {
908 	struct session *sp;
909 	struct tty *tp;
910 	struct proc *p;
911 	struct vnode *ttyvp;
912 
913 	p = curproc;
914 	MPASS(p->p_flag & P_WEXIT);
915 	sx_assert(&proctree_lock, SX_LOCKED);
916 
917 	if (SESS_LEADER(p)) {
918 		sp = p->p_session;
919 
920 		/*
921 		 * s_ttyp is not zero'd; we use this to indicate that
922 		 * the session once had a controlling terminal. (for
923 		 * logging and informational purposes)
924 		 */
925 		SESS_LOCK(sp);
926 		ttyvp = sp->s_ttyvp;
927 		tp = sp->s_ttyp;
928 		sp->s_ttyvp = NULL;
929 		sp->s_ttydp = NULL;
930 		sp->s_leader = NULL;
931 		SESS_UNLOCK(sp);
932 
933 		/*
934 		 * Signal foreground pgrp and revoke access to
935 		 * controlling terminal if it has not been revoked
936 		 * already.
937 		 *
938 		 * Because the TTY may have been revoked in the mean
939 		 * time and could already have a new session associated
940 		 * with it, make sure we don't send a SIGHUP to a
941 		 * foreground process group that does not belong to this
942 		 * session.
943 		 */
944 
945 		if (tp != NULL) {
946 			tty_lock(tp);
947 			if (tp->t_session == sp)
948 				tty_signal_pgrp(tp, SIGHUP);
949 			tty_unlock(tp);
950 		}
951 
952 		if (ttyvp != NULL) {
953 			sx_xunlock(&proctree_lock);
954 			if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
955 				VOP_REVOKE(ttyvp, REVOKEALL);
956 				VOP_UNLOCK(ttyvp);
957 			}
958 			devfs_ctty_unref(ttyvp);
959 			sx_xlock(&proctree_lock);
960 		}
961 	}
962 	fixjobc_kill(p);
963 }
964 
965 /*
966  * A process group has become orphaned, mark it as such for signal
967  * delivery code.  If there are any stopped processes in the group,
968  * hang-up all process in that group.
969  */
970 static void
971 orphanpg(struct pgrp *pg)
972 {
973 	struct proc *p;
974 
975 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
976 
977 	pg->pg_flags |= PGRP_ORPHANED;
978 
979 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
980 		PROC_LOCK(p);
981 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
982 			PROC_UNLOCK(p);
983 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
984 				PROC_LOCK(p);
985 				kern_psignal(p, SIGHUP);
986 				kern_psignal(p, SIGCONT);
987 				PROC_UNLOCK(p);
988 			}
989 			return;
990 		}
991 		PROC_UNLOCK(p);
992 	}
993 }
994 
995 void
996 sess_hold(struct session *s)
997 {
998 
999 	refcount_acquire(&s->s_count);
1000 }
1001 
1002 void
1003 sess_release(struct session *s)
1004 {
1005 
1006 	if (refcount_release(&s->s_count)) {
1007 		if (s->s_ttyp != NULL) {
1008 			tty_lock(s->s_ttyp);
1009 			tty_rel_sess(s->s_ttyp, s);
1010 		}
1011 		proc_id_clear(PROC_ID_SESSION, s->s_sid);
1012 		mtx_destroy(&s->s_mtx);
1013 		free(s, M_SESSION);
1014 	}
1015 }
1016 
1017 #ifdef DDB
1018 
1019 static void
1020 db_print_pgrp_one(struct pgrp *pgrp, struct proc *p)
1021 {
1022 	db_printf(
1023 	    "    pid %d at %p pr %d pgrp %p e %d jc %d\n",
1024 	    p->p_pid, p, p->p_pptr == NULL ? -1 : p->p_pptr->p_pid,
1025 	    p->p_pgrp, (p->p_treeflag & P_TREE_GRPEXITED) != 0,
1026 	    p->p_pptr == NULL ? 0 : isjobproc(p->p_pptr, pgrp));
1027 }
1028 
1029 DB_SHOW_COMMAND_FLAGS(pgrpdump, pgrpdump, DB_CMD_MEMSAFE)
1030 {
1031 	struct pgrp *pgrp;
1032 	struct proc *p;
1033 	int i;
1034 
1035 	for (i = 0; i <= pgrphash; i++) {
1036 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
1037 			db_printf("indx %d\n", i);
1038 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
1039 				db_printf(
1040 			"  pgrp %p, pgid %d, sess %p, sesscnt %d, mem %p\n",
1041 				    pgrp, (int)pgrp->pg_id, pgrp->pg_session,
1042 				    pgrp->pg_session->s_count,
1043 				    LIST_FIRST(&pgrp->pg_members));
1044 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist)
1045 					db_print_pgrp_one(pgrp, p);
1046 			}
1047 		}
1048 	}
1049 }
1050 #endif /* DDB */
1051 
1052 /*
1053  * Calculate the kinfo_proc members which contain process-wide
1054  * informations.
1055  * Must be called with the target process locked.
1056  */
1057 static void
1058 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
1059 {
1060 	struct thread *td;
1061 
1062 	PROC_LOCK_ASSERT(p, MA_OWNED);
1063 
1064 	kp->ki_estcpu = 0;
1065 	kp->ki_pctcpu = 0;
1066 	FOREACH_THREAD_IN_PROC(p, td) {
1067 		thread_lock(td);
1068 		kp->ki_pctcpu += sched_pctcpu(td);
1069 		kp->ki_estcpu += sched_estcpu(td);
1070 		thread_unlock(td);
1071 	}
1072 }
1073 
1074 /*
1075  * Fill in any information that is common to all threads in the process.
1076  * Must be called with the target process locked.
1077  */
1078 static void
1079 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
1080 {
1081 	struct thread *td0;
1082 	struct ucred *cred;
1083 	struct sigacts *ps;
1084 	struct timeval boottime;
1085 
1086 	PROC_LOCK_ASSERT(p, MA_OWNED);
1087 
1088 	kp->ki_structsize = sizeof(*kp);
1089 	kp->ki_paddr = p;
1090 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
1091 	kp->ki_args = p->p_args;
1092 	kp->ki_textvp = p->p_textvp;
1093 #ifdef KTRACE
1094 	kp->ki_tracep = ktr_get_tracevp(p, false);
1095 	kp->ki_traceflag = p->p_traceflag;
1096 #endif
1097 	kp->ki_fd = p->p_fd;
1098 	kp->ki_pd = p->p_pd;
1099 	kp->ki_vmspace = p->p_vmspace;
1100 	kp->ki_flag = p->p_flag;
1101 	kp->ki_flag2 = p->p_flag2;
1102 	cred = p->p_ucred;
1103 	if (cred) {
1104 		kp->ki_uid = cred->cr_uid;
1105 		kp->ki_ruid = cred->cr_ruid;
1106 		kp->ki_svuid = cred->cr_svuid;
1107 		kp->ki_cr_flags = 0;
1108 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
1109 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
1110 		/* XXX bde doesn't like KI_NGROUPS */
1111 		if (cred->cr_ngroups > KI_NGROUPS) {
1112 			kp->ki_ngroups = KI_NGROUPS;
1113 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
1114 		} else
1115 			kp->ki_ngroups = cred->cr_ngroups;
1116 		bcopy(cred->cr_groups, kp->ki_groups,
1117 		    kp->ki_ngroups * sizeof(gid_t));
1118 		kp->ki_rgid = cred->cr_rgid;
1119 		kp->ki_svgid = cred->cr_svgid;
1120 		/* If jailed(cred), emulate the old P_JAILED flag. */
1121 		if (jailed(cred)) {
1122 			kp->ki_flag |= P_JAILED;
1123 			/* If inside the jail, use 0 as a jail ID. */
1124 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
1125 				kp->ki_jid = cred->cr_prison->pr_id;
1126 		}
1127 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
1128 		    sizeof(kp->ki_loginclass));
1129 	}
1130 	ps = p->p_sigacts;
1131 	if (ps) {
1132 		mtx_lock(&ps->ps_mtx);
1133 		kp->ki_sigignore = ps->ps_sigignore;
1134 		kp->ki_sigcatch = ps->ps_sigcatch;
1135 		mtx_unlock(&ps->ps_mtx);
1136 	}
1137 	if (p->p_state != PRS_NEW &&
1138 	    p->p_state != PRS_ZOMBIE &&
1139 	    p->p_vmspace != NULL) {
1140 		struct vmspace *vm = p->p_vmspace;
1141 
1142 		kp->ki_size = vm->vm_map.size;
1143 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
1144 		FOREACH_THREAD_IN_PROC(p, td0) {
1145 			if (!TD_IS_SWAPPED(td0))
1146 				kp->ki_rssize += td0->td_kstack_pages;
1147 		}
1148 		kp->ki_swrss = vm->vm_swrss;
1149 		kp->ki_tsize = vm->vm_tsize;
1150 		kp->ki_dsize = vm->vm_dsize;
1151 		kp->ki_ssize = vm->vm_ssize;
1152 	} else if (p->p_state == PRS_ZOMBIE)
1153 		kp->ki_stat = SZOMB;
1154 	if (kp->ki_flag & P_INMEM)
1155 		kp->ki_sflag = PS_INMEM;
1156 	else
1157 		kp->ki_sflag = 0;
1158 	/* Calculate legacy swtime as seconds since 'swtick'. */
1159 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
1160 	kp->ki_pid = p->p_pid;
1161 	kp->ki_nice = p->p_nice;
1162 	kp->ki_fibnum = p->p_fibnum;
1163 	kp->ki_start = p->p_stats->p_start;
1164 	getboottime(&boottime);
1165 	timevaladd(&kp->ki_start, &boottime);
1166 	PROC_STATLOCK(p);
1167 	rufetch(p, &kp->ki_rusage);
1168 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
1169 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
1170 	PROC_STATUNLOCK(p);
1171 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1172 	/* Some callers want child times in a single value. */
1173 	kp->ki_childtime = kp->ki_childstime;
1174 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1175 
1176 	FOREACH_THREAD_IN_PROC(p, td0)
1177 		kp->ki_cow += td0->td_cow;
1178 
1179 	if (p->p_comm[0] != '\0')
1180 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1181 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1182 	    p->p_sysent->sv_name[0] != '\0')
1183 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1184 	kp->ki_siglist = p->p_siglist;
1185 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1186 	kp->ki_acflag = p->p_acflag;
1187 	kp->ki_lock = p->p_lock;
1188 	if (p->p_pptr) {
1189 		kp->ki_ppid = p->p_oppid;
1190 		if (p->p_flag & P_TRACED)
1191 			kp->ki_tracer = p->p_pptr->p_pid;
1192 	}
1193 }
1194 
1195 /*
1196  * Fill job-related process information.
1197  */
1198 static void
1199 fill_kinfo_proc_pgrp(struct proc *p, struct kinfo_proc *kp)
1200 {
1201 	struct tty *tp;
1202 	struct session *sp;
1203 	struct pgrp *pgrp;
1204 
1205 	sx_assert(&proctree_lock, SA_LOCKED);
1206 	PROC_LOCK_ASSERT(p, MA_OWNED);
1207 
1208 	pgrp = p->p_pgrp;
1209 	if (pgrp == NULL)
1210 		return;
1211 
1212 	kp->ki_pgid = pgrp->pg_id;
1213 	kp->ki_jobc = pgrp_calc_jobc(pgrp);
1214 
1215 	sp = pgrp->pg_session;
1216 	tp = NULL;
1217 
1218 	if (sp != NULL) {
1219 		kp->ki_sid = sp->s_sid;
1220 		SESS_LOCK(sp);
1221 		strlcpy(kp->ki_login, sp->s_login, sizeof(kp->ki_login));
1222 		if (sp->s_ttyvp)
1223 			kp->ki_kiflag |= KI_CTTY;
1224 		if (SESS_LEADER(p))
1225 			kp->ki_kiflag |= KI_SLEADER;
1226 		tp = sp->s_ttyp;
1227 		SESS_UNLOCK(sp);
1228 	}
1229 
1230 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1231 		kp->ki_tdev = tty_udev(tp);
1232 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1233 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1234 		if (tp->t_session)
1235 			kp->ki_tsid = tp->t_session->s_sid;
1236 	} else {
1237 		kp->ki_tdev = NODEV;
1238 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1239 	}
1240 }
1241 
1242 /*
1243  * Fill in information that is thread specific.  Must be called with
1244  * target process locked.  If 'preferthread' is set, overwrite certain
1245  * process-related fields that are maintained for both threads and
1246  * processes.
1247  */
1248 static void
1249 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1250 {
1251 	struct proc *p;
1252 
1253 	p = td->td_proc;
1254 	kp->ki_tdaddr = td;
1255 	PROC_LOCK_ASSERT(p, MA_OWNED);
1256 
1257 	if (preferthread)
1258 		PROC_STATLOCK(p);
1259 	thread_lock(td);
1260 	if (td->td_wmesg != NULL)
1261 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1262 	else
1263 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1264 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1265 	    sizeof(kp->ki_tdname)) {
1266 		strlcpy(kp->ki_moretdname,
1267 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1268 		    sizeof(kp->ki_moretdname));
1269 	} else {
1270 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1271 	}
1272 	if (TD_ON_LOCK(td)) {
1273 		kp->ki_kiflag |= KI_LOCKBLOCK;
1274 		strlcpy(kp->ki_lockname, td->td_lockname,
1275 		    sizeof(kp->ki_lockname));
1276 	} else {
1277 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1278 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1279 	}
1280 
1281 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1282 		if (TD_ON_RUNQ(td) ||
1283 		    TD_CAN_RUN(td) ||
1284 		    TD_IS_RUNNING(td)) {
1285 			kp->ki_stat = SRUN;
1286 		} else if (P_SHOULDSTOP(p)) {
1287 			kp->ki_stat = SSTOP;
1288 		} else if (TD_IS_SLEEPING(td)) {
1289 			kp->ki_stat = SSLEEP;
1290 		} else if (TD_ON_LOCK(td)) {
1291 			kp->ki_stat = SLOCK;
1292 		} else {
1293 			kp->ki_stat = SWAIT;
1294 		}
1295 	} else if (p->p_state == PRS_ZOMBIE) {
1296 		kp->ki_stat = SZOMB;
1297 	} else {
1298 		kp->ki_stat = SIDL;
1299 	}
1300 
1301 	/* Things in the thread */
1302 	kp->ki_wchan = td->td_wchan;
1303 	kp->ki_pri.pri_level = td->td_priority;
1304 	kp->ki_pri.pri_native = td->td_base_pri;
1305 
1306 	/*
1307 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1308 	 * the maximum u_char CPU value.
1309 	 */
1310 	if (td->td_lastcpu == NOCPU)
1311 		kp->ki_lastcpu_old = NOCPU_OLD;
1312 	else if (td->td_lastcpu > MAXCPU_OLD)
1313 		kp->ki_lastcpu_old = MAXCPU_OLD;
1314 	else
1315 		kp->ki_lastcpu_old = td->td_lastcpu;
1316 
1317 	if (td->td_oncpu == NOCPU)
1318 		kp->ki_oncpu_old = NOCPU_OLD;
1319 	else if (td->td_oncpu > MAXCPU_OLD)
1320 		kp->ki_oncpu_old = MAXCPU_OLD;
1321 	else
1322 		kp->ki_oncpu_old = td->td_oncpu;
1323 
1324 	kp->ki_lastcpu = td->td_lastcpu;
1325 	kp->ki_oncpu = td->td_oncpu;
1326 	kp->ki_tdflags = td->td_flags;
1327 	kp->ki_tid = td->td_tid;
1328 	kp->ki_numthreads = p->p_numthreads;
1329 	kp->ki_pcb = td->td_pcb;
1330 	kp->ki_kstack = (void *)td->td_kstack;
1331 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1332 	kp->ki_pri.pri_class = td->td_pri_class;
1333 	kp->ki_pri.pri_user = td->td_user_pri;
1334 
1335 	if (preferthread) {
1336 		rufetchtd(td, &kp->ki_rusage);
1337 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1338 		kp->ki_pctcpu = sched_pctcpu(td);
1339 		kp->ki_estcpu = sched_estcpu(td);
1340 		kp->ki_cow = td->td_cow;
1341 	}
1342 
1343 	/* We can't get this anymore but ps etc never used it anyway. */
1344 	kp->ki_rqindex = 0;
1345 
1346 	if (preferthread)
1347 		kp->ki_siglist = td->td_siglist;
1348 	kp->ki_sigmask = td->td_sigmask;
1349 	thread_unlock(td);
1350 	if (preferthread)
1351 		PROC_STATUNLOCK(p);
1352 }
1353 
1354 /*
1355  * Fill in a kinfo_proc structure for the specified process.
1356  * Must be called with the target process locked.
1357  */
1358 void
1359 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1360 {
1361 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1362 
1363 	bzero(kp, sizeof(*kp));
1364 
1365 	fill_kinfo_proc_pgrp(p,kp);
1366 	fill_kinfo_proc_only(p, kp);
1367 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1368 	fill_kinfo_aggregate(p, kp);
1369 }
1370 
1371 struct pstats *
1372 pstats_alloc(void)
1373 {
1374 
1375 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1376 }
1377 
1378 /*
1379  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1380  */
1381 void
1382 pstats_fork(struct pstats *src, struct pstats *dst)
1383 {
1384 
1385 	bzero(&dst->pstat_startzero,
1386 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1387 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1388 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1389 }
1390 
1391 void
1392 pstats_free(struct pstats *ps)
1393 {
1394 
1395 	free(ps, M_SUBPROC);
1396 }
1397 
1398 #ifdef COMPAT_FREEBSD32
1399 
1400 /*
1401  * This function is typically used to copy out the kernel address, so
1402  * it can be replaced by assignment of zero.
1403  */
1404 static inline uint32_t
1405 ptr32_trim(const void *ptr)
1406 {
1407 	uintptr_t uptr;
1408 
1409 	uptr = (uintptr_t)ptr;
1410 	return ((uptr > UINT_MAX) ? 0 : uptr);
1411 }
1412 
1413 #define PTRTRIM_CP(src,dst,fld) \
1414 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1415 
1416 static void
1417 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1418 {
1419 	int i;
1420 
1421 	bzero(ki32, sizeof(struct kinfo_proc32));
1422 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1423 	CP(*ki, *ki32, ki_layout);
1424 	PTRTRIM_CP(*ki, *ki32, ki_args);
1425 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1426 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1427 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1428 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1429 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1430 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1431 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1432 	CP(*ki, *ki32, ki_pid);
1433 	CP(*ki, *ki32, ki_ppid);
1434 	CP(*ki, *ki32, ki_pgid);
1435 	CP(*ki, *ki32, ki_tpgid);
1436 	CP(*ki, *ki32, ki_sid);
1437 	CP(*ki, *ki32, ki_tsid);
1438 	CP(*ki, *ki32, ki_jobc);
1439 	CP(*ki, *ki32, ki_tdev);
1440 	CP(*ki, *ki32, ki_tdev_freebsd11);
1441 	CP(*ki, *ki32, ki_siglist);
1442 	CP(*ki, *ki32, ki_sigmask);
1443 	CP(*ki, *ki32, ki_sigignore);
1444 	CP(*ki, *ki32, ki_sigcatch);
1445 	CP(*ki, *ki32, ki_uid);
1446 	CP(*ki, *ki32, ki_ruid);
1447 	CP(*ki, *ki32, ki_svuid);
1448 	CP(*ki, *ki32, ki_rgid);
1449 	CP(*ki, *ki32, ki_svgid);
1450 	CP(*ki, *ki32, ki_ngroups);
1451 	for (i = 0; i < KI_NGROUPS; i++)
1452 		CP(*ki, *ki32, ki_groups[i]);
1453 	CP(*ki, *ki32, ki_size);
1454 	CP(*ki, *ki32, ki_rssize);
1455 	CP(*ki, *ki32, ki_swrss);
1456 	CP(*ki, *ki32, ki_tsize);
1457 	CP(*ki, *ki32, ki_dsize);
1458 	CP(*ki, *ki32, ki_ssize);
1459 	CP(*ki, *ki32, ki_xstat);
1460 	CP(*ki, *ki32, ki_acflag);
1461 	CP(*ki, *ki32, ki_pctcpu);
1462 	CP(*ki, *ki32, ki_estcpu);
1463 	CP(*ki, *ki32, ki_slptime);
1464 	CP(*ki, *ki32, ki_swtime);
1465 	CP(*ki, *ki32, ki_cow);
1466 	CP(*ki, *ki32, ki_runtime);
1467 	TV_CP(*ki, *ki32, ki_start);
1468 	TV_CP(*ki, *ki32, ki_childtime);
1469 	CP(*ki, *ki32, ki_flag);
1470 	CP(*ki, *ki32, ki_kiflag);
1471 	CP(*ki, *ki32, ki_traceflag);
1472 	CP(*ki, *ki32, ki_stat);
1473 	CP(*ki, *ki32, ki_nice);
1474 	CP(*ki, *ki32, ki_lock);
1475 	CP(*ki, *ki32, ki_rqindex);
1476 	CP(*ki, *ki32, ki_oncpu);
1477 	CP(*ki, *ki32, ki_lastcpu);
1478 
1479 	/* XXX TODO: wrap cpu value as appropriate */
1480 	CP(*ki, *ki32, ki_oncpu_old);
1481 	CP(*ki, *ki32, ki_lastcpu_old);
1482 
1483 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1484 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1485 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1486 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1487 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1488 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1489 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1490 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1491 	CP(*ki, *ki32, ki_tracer);
1492 	CP(*ki, *ki32, ki_flag2);
1493 	CP(*ki, *ki32, ki_fibnum);
1494 	CP(*ki, *ki32, ki_cr_flags);
1495 	CP(*ki, *ki32, ki_jid);
1496 	CP(*ki, *ki32, ki_numthreads);
1497 	CP(*ki, *ki32, ki_tid);
1498 	CP(*ki, *ki32, ki_pri);
1499 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1500 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1501 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1502 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1503 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1504 	PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1505 	CP(*ki, *ki32, ki_sflag);
1506 	CP(*ki, *ki32, ki_tdflags);
1507 }
1508 #endif
1509 
1510 static ssize_t
1511 kern_proc_out_size(struct proc *p, int flags)
1512 {
1513 	ssize_t size = 0;
1514 
1515 	PROC_LOCK_ASSERT(p, MA_OWNED);
1516 
1517 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1518 #ifdef COMPAT_FREEBSD32
1519 		if ((flags & KERN_PROC_MASK32) != 0) {
1520 			size += sizeof(struct kinfo_proc32);
1521 		} else
1522 #endif
1523 			size += sizeof(struct kinfo_proc);
1524 	} else {
1525 #ifdef COMPAT_FREEBSD32
1526 		if ((flags & KERN_PROC_MASK32) != 0)
1527 			size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1528 		else
1529 #endif
1530 			size += sizeof(struct kinfo_proc) * p->p_numthreads;
1531 	}
1532 	PROC_UNLOCK(p);
1533 	return (size);
1534 }
1535 
1536 int
1537 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1538 {
1539 	struct thread *td;
1540 	struct kinfo_proc ki;
1541 #ifdef COMPAT_FREEBSD32
1542 	struct kinfo_proc32 ki32;
1543 #endif
1544 	int error;
1545 
1546 	PROC_LOCK_ASSERT(p, MA_OWNED);
1547 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1548 
1549 	error = 0;
1550 	fill_kinfo_proc(p, &ki);
1551 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1552 #ifdef COMPAT_FREEBSD32
1553 		if ((flags & KERN_PROC_MASK32) != 0) {
1554 			freebsd32_kinfo_proc_out(&ki, &ki32);
1555 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1556 				error = ENOMEM;
1557 		} else
1558 #endif
1559 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1560 				error = ENOMEM;
1561 	} else {
1562 		FOREACH_THREAD_IN_PROC(p, td) {
1563 			fill_kinfo_thread(td, &ki, 1);
1564 #ifdef COMPAT_FREEBSD32
1565 			if ((flags & KERN_PROC_MASK32) != 0) {
1566 				freebsd32_kinfo_proc_out(&ki, &ki32);
1567 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1568 					error = ENOMEM;
1569 			} else
1570 #endif
1571 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1572 					error = ENOMEM;
1573 			if (error != 0)
1574 				break;
1575 		}
1576 	}
1577 	PROC_UNLOCK(p);
1578 	return (error);
1579 }
1580 
1581 static int
1582 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1583 {
1584 	struct sbuf sb;
1585 	struct kinfo_proc ki;
1586 	int error, error2;
1587 
1588 	if (req->oldptr == NULL)
1589 		return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1590 
1591 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1592 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1593 	error = kern_proc_out(p, &sb, flags);
1594 	error2 = sbuf_finish(&sb);
1595 	sbuf_delete(&sb);
1596 	if (error != 0)
1597 		return (error);
1598 	else if (error2 != 0)
1599 		return (error2);
1600 	return (0);
1601 }
1602 
1603 int
1604 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg)
1605 {
1606 	struct proc *p;
1607 	int error, i, j;
1608 
1609 	for (i = 0; i < pidhashlock + 1; i++) {
1610 		sx_slock(&proctree_lock);
1611 		sx_slock(&pidhashtbl_lock[i]);
1612 		for (j = i; j <= pidhash; j += pidhashlock + 1) {
1613 			LIST_FOREACH(p, &pidhashtbl[j], p_hash) {
1614 				if (p->p_state == PRS_NEW)
1615 					continue;
1616 				error = cb(p, cbarg);
1617 				PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1618 				if (error != 0) {
1619 					sx_sunlock(&pidhashtbl_lock[i]);
1620 					sx_sunlock(&proctree_lock);
1621 					return (error);
1622 				}
1623 			}
1624 		}
1625 		sx_sunlock(&pidhashtbl_lock[i]);
1626 		sx_sunlock(&proctree_lock);
1627 	}
1628 	return (0);
1629 }
1630 
1631 struct kern_proc_out_args {
1632 	struct sysctl_req *req;
1633 	int flags;
1634 	int oid_number;
1635 	int *name;
1636 };
1637 
1638 static int
1639 sysctl_kern_proc_iterate(struct proc *p, void *origarg)
1640 {
1641 	struct kern_proc_out_args *arg = origarg;
1642 	int *name = arg->name;
1643 	int oid_number = arg->oid_number;
1644 	int flags = arg->flags;
1645 	struct sysctl_req *req = arg->req;
1646 	int error = 0;
1647 
1648 	PROC_LOCK(p);
1649 
1650 	KASSERT(p->p_ucred != NULL,
1651 	    ("process credential is NULL for non-NEW proc"));
1652 	/*
1653 	 * Show a user only appropriate processes.
1654 	 */
1655 	if (p_cansee(curthread, p))
1656 		goto skip;
1657 	/*
1658 	 * TODO - make more efficient (see notes below).
1659 	 * do by session.
1660 	 */
1661 	switch (oid_number) {
1662 	case KERN_PROC_GID:
1663 		if (p->p_ucred->cr_gid != (gid_t)name[0])
1664 			goto skip;
1665 		break;
1666 
1667 	case KERN_PROC_PGRP:
1668 		/* could do this by traversing pgrp */
1669 		if (p->p_pgrp == NULL ||
1670 		    p->p_pgrp->pg_id != (pid_t)name[0])
1671 			goto skip;
1672 		break;
1673 
1674 	case KERN_PROC_RGID:
1675 		if (p->p_ucred->cr_rgid != (gid_t)name[0])
1676 			goto skip;
1677 		break;
1678 
1679 	case KERN_PROC_SESSION:
1680 		if (p->p_session == NULL ||
1681 		    p->p_session->s_sid != (pid_t)name[0])
1682 			goto skip;
1683 		break;
1684 
1685 	case KERN_PROC_TTY:
1686 		if ((p->p_flag & P_CONTROLT) == 0 ||
1687 		    p->p_session == NULL)
1688 			goto skip;
1689 		/* XXX proctree_lock */
1690 		SESS_LOCK(p->p_session);
1691 		if (p->p_session->s_ttyp == NULL ||
1692 		    tty_udev(p->p_session->s_ttyp) !=
1693 		    (dev_t)name[0]) {
1694 			SESS_UNLOCK(p->p_session);
1695 			goto skip;
1696 		}
1697 		SESS_UNLOCK(p->p_session);
1698 		break;
1699 
1700 	case KERN_PROC_UID:
1701 		if (p->p_ucred->cr_uid != (uid_t)name[0])
1702 			goto skip;
1703 		break;
1704 
1705 	case KERN_PROC_RUID:
1706 		if (p->p_ucred->cr_ruid != (uid_t)name[0])
1707 			goto skip;
1708 		break;
1709 
1710 	case KERN_PROC_PROC:
1711 		break;
1712 
1713 	default:
1714 		break;
1715 	}
1716 	error = sysctl_out_proc(p, req, flags);
1717 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1718 	return (error);
1719 skip:
1720 	PROC_UNLOCK(p);
1721 	return (0);
1722 }
1723 
1724 static int
1725 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1726 {
1727 	struct kern_proc_out_args iterarg;
1728 	int *name = (int *)arg1;
1729 	u_int namelen = arg2;
1730 	struct proc *p;
1731 	int flags, oid_number;
1732 	int error = 0;
1733 
1734 	oid_number = oidp->oid_number;
1735 	if (oid_number != KERN_PROC_ALL &&
1736 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1737 		flags = KERN_PROC_NOTHREADS;
1738 	else {
1739 		flags = 0;
1740 		oid_number &= ~KERN_PROC_INC_THREAD;
1741 	}
1742 #ifdef COMPAT_FREEBSD32
1743 	if (req->flags & SCTL_MASK32)
1744 		flags |= KERN_PROC_MASK32;
1745 #endif
1746 	if (oid_number == KERN_PROC_PID) {
1747 		if (namelen != 1)
1748 			return (EINVAL);
1749 		error = sysctl_wire_old_buffer(req, 0);
1750 		if (error)
1751 			return (error);
1752 		sx_slock(&proctree_lock);
1753 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1754 		if (error == 0)
1755 			error = sysctl_out_proc(p, req, flags);
1756 		sx_sunlock(&proctree_lock);
1757 		return (error);
1758 	}
1759 
1760 	switch (oid_number) {
1761 	case KERN_PROC_ALL:
1762 		if (namelen != 0)
1763 			return (EINVAL);
1764 		break;
1765 	case KERN_PROC_PROC:
1766 		if (namelen != 0 && namelen != 1)
1767 			return (EINVAL);
1768 		break;
1769 	default:
1770 		if (namelen != 1)
1771 			return (EINVAL);
1772 		break;
1773 	}
1774 
1775 	if (req->oldptr == NULL) {
1776 		/* overestimate by 5 procs */
1777 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1778 		if (error)
1779 			return (error);
1780 	} else {
1781 		error = sysctl_wire_old_buffer(req, 0);
1782 		if (error != 0)
1783 			return (error);
1784 	}
1785 	iterarg.flags = flags;
1786 	iterarg.oid_number = oid_number;
1787 	iterarg.req = req;
1788 	iterarg.name = name;
1789 	error = proc_iterate(sysctl_kern_proc_iterate, &iterarg);
1790 	return (error);
1791 }
1792 
1793 struct pargs *
1794 pargs_alloc(int len)
1795 {
1796 	struct pargs *pa;
1797 
1798 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1799 		M_WAITOK);
1800 	refcount_init(&pa->ar_ref, 1);
1801 	pa->ar_length = len;
1802 	return (pa);
1803 }
1804 
1805 static void
1806 pargs_free(struct pargs *pa)
1807 {
1808 
1809 	free(pa, M_PARGS);
1810 }
1811 
1812 void
1813 pargs_hold(struct pargs *pa)
1814 {
1815 
1816 	if (pa == NULL)
1817 		return;
1818 	refcount_acquire(&pa->ar_ref);
1819 }
1820 
1821 void
1822 pargs_drop(struct pargs *pa)
1823 {
1824 
1825 	if (pa == NULL)
1826 		return;
1827 	if (refcount_release(&pa->ar_ref))
1828 		pargs_free(pa);
1829 }
1830 
1831 static int
1832 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1833     size_t len)
1834 {
1835 	ssize_t n;
1836 
1837 	/*
1838 	 * This may return a short read if the string is shorter than the chunk
1839 	 * and is aligned at the end of the page, and the following page is not
1840 	 * mapped.
1841 	 */
1842 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1843 	if (n <= 0)
1844 		return (ENOMEM);
1845 	return (0);
1846 }
1847 
1848 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1849 
1850 enum proc_vector_type {
1851 	PROC_ARG,
1852 	PROC_ENV,
1853 	PROC_AUX,
1854 };
1855 
1856 #ifdef COMPAT_FREEBSD32
1857 static int
1858 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1859     size_t *vsizep, enum proc_vector_type type)
1860 {
1861 	struct freebsd32_ps_strings pss;
1862 	Elf32_Auxinfo aux;
1863 	vm_offset_t vptr, ptr;
1864 	uint32_t *proc_vector32;
1865 	char **proc_vector;
1866 	size_t vsize, size;
1867 	int i, error;
1868 
1869 	error = 0;
1870 	if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) !=
1871 	    sizeof(pss))
1872 		return (ENOMEM);
1873 	switch (type) {
1874 	case PROC_ARG:
1875 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1876 		vsize = pss.ps_nargvstr;
1877 		if (vsize > ARG_MAX)
1878 			return (ENOEXEC);
1879 		size = vsize * sizeof(int32_t);
1880 		break;
1881 	case PROC_ENV:
1882 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1883 		vsize = pss.ps_nenvstr;
1884 		if (vsize > ARG_MAX)
1885 			return (ENOEXEC);
1886 		size = vsize * sizeof(int32_t);
1887 		break;
1888 	case PROC_AUX:
1889 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1890 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1891 		if (vptr % 4 != 0)
1892 			return (ENOEXEC);
1893 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1894 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1895 			    sizeof(aux))
1896 				return (ENOMEM);
1897 			if (aux.a_type == AT_NULL)
1898 				break;
1899 			ptr += sizeof(aux);
1900 		}
1901 		if (aux.a_type != AT_NULL)
1902 			return (ENOEXEC);
1903 		vsize = i + 1;
1904 		size = vsize * sizeof(aux);
1905 		break;
1906 	default:
1907 		KASSERT(0, ("Wrong proc vector type: %d", type));
1908 		return (EINVAL);
1909 	}
1910 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1911 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1912 		error = ENOMEM;
1913 		goto done;
1914 	}
1915 	if (type == PROC_AUX) {
1916 		*proc_vectorp = (char **)proc_vector32;
1917 		*vsizep = vsize;
1918 		return (0);
1919 	}
1920 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1921 	for (i = 0; i < (int)vsize; i++)
1922 		proc_vector[i] = PTRIN(proc_vector32[i]);
1923 	*proc_vectorp = proc_vector;
1924 	*vsizep = vsize;
1925 done:
1926 	free(proc_vector32, M_TEMP);
1927 	return (error);
1928 }
1929 #endif
1930 
1931 static int
1932 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1933     size_t *vsizep, enum proc_vector_type type)
1934 {
1935 	struct ps_strings pss;
1936 	Elf_Auxinfo aux;
1937 	vm_offset_t vptr, ptr;
1938 	char **proc_vector;
1939 	size_t vsize, size;
1940 	int i;
1941 
1942 #ifdef COMPAT_FREEBSD32
1943 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1944 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1945 #endif
1946 	if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) !=
1947 	    sizeof(pss))
1948 		return (ENOMEM);
1949 	switch (type) {
1950 	case PROC_ARG:
1951 		vptr = (vm_offset_t)pss.ps_argvstr;
1952 		vsize = pss.ps_nargvstr;
1953 		if (vsize > ARG_MAX)
1954 			return (ENOEXEC);
1955 		size = vsize * sizeof(char *);
1956 		break;
1957 	case PROC_ENV:
1958 		vptr = (vm_offset_t)pss.ps_envstr;
1959 		vsize = pss.ps_nenvstr;
1960 		if (vsize > ARG_MAX)
1961 			return (ENOEXEC);
1962 		size = vsize * sizeof(char *);
1963 		break;
1964 	case PROC_AUX:
1965 		/*
1966 		 * The aux array is just above env array on the stack. Check
1967 		 * that the address is naturally aligned.
1968 		 */
1969 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1970 		    * sizeof(char *);
1971 #if __ELF_WORD_SIZE == 64
1972 		if (vptr % sizeof(uint64_t) != 0)
1973 #else
1974 		if (vptr % sizeof(uint32_t) != 0)
1975 #endif
1976 			return (ENOEXEC);
1977 		/*
1978 		 * We count the array size reading the aux vectors from the
1979 		 * stack until AT_NULL vector is returned.  So (to keep the code
1980 		 * simple) we read the process stack twice: the first time here
1981 		 * to find the size and the second time when copying the vectors
1982 		 * to the allocated proc_vector.
1983 		 */
1984 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1985 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1986 			    sizeof(aux))
1987 				return (ENOMEM);
1988 			if (aux.a_type == AT_NULL)
1989 				break;
1990 			ptr += sizeof(aux);
1991 		}
1992 		/*
1993 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1994 		 * not reached AT_NULL, it is most likely we are reading wrong
1995 		 * data: either the process doesn't have auxv array or data has
1996 		 * been modified. Return the error in this case.
1997 		 */
1998 		if (aux.a_type != AT_NULL)
1999 			return (ENOEXEC);
2000 		vsize = i + 1;
2001 		size = vsize * sizeof(aux);
2002 		break;
2003 	default:
2004 		KASSERT(0, ("Wrong proc vector type: %d", type));
2005 		return (EINVAL); /* In case we are built without INVARIANTS. */
2006 	}
2007 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
2008 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
2009 		free(proc_vector, M_TEMP);
2010 		return (ENOMEM);
2011 	}
2012 	*proc_vectorp = proc_vector;
2013 	*vsizep = vsize;
2014 
2015 	return (0);
2016 }
2017 
2018 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
2019 
2020 static int
2021 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
2022     enum proc_vector_type type)
2023 {
2024 	size_t done, len, nchr, vsize;
2025 	int error, i;
2026 	char **proc_vector, *sptr;
2027 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
2028 
2029 	PROC_ASSERT_HELD(p);
2030 
2031 	/*
2032 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
2033 	 */
2034 	nchr = 2 * (PATH_MAX + ARG_MAX);
2035 
2036 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
2037 	if (error != 0)
2038 		return (error);
2039 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
2040 		/*
2041 		 * The program may have scribbled into its argv array, e.g. to
2042 		 * remove some arguments.  If that has happened, break out
2043 		 * before trying to read from NULL.
2044 		 */
2045 		if (proc_vector[i] == NULL)
2046 			break;
2047 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
2048 			error = proc_read_string(td, p, sptr, pss_string,
2049 			    sizeof(pss_string));
2050 			if (error != 0)
2051 				goto done;
2052 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
2053 			if (done + len >= nchr)
2054 				len = nchr - done - 1;
2055 			sbuf_bcat(sb, pss_string, len);
2056 			if (len != GET_PS_STRINGS_CHUNK_SZ)
2057 				break;
2058 			done += GET_PS_STRINGS_CHUNK_SZ;
2059 		}
2060 		sbuf_bcat(sb, "", 1);
2061 		done += len + 1;
2062 	}
2063 done:
2064 	free(proc_vector, M_TEMP);
2065 	return (error);
2066 }
2067 
2068 int
2069 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
2070 {
2071 
2072 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
2073 }
2074 
2075 int
2076 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
2077 {
2078 
2079 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
2080 }
2081 
2082 int
2083 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
2084 {
2085 	size_t vsize, size;
2086 	char **auxv;
2087 	int error;
2088 
2089 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
2090 	if (error == 0) {
2091 #ifdef COMPAT_FREEBSD32
2092 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
2093 			size = vsize * sizeof(Elf32_Auxinfo);
2094 		else
2095 #endif
2096 			size = vsize * sizeof(Elf_Auxinfo);
2097 		if (sbuf_bcat(sb, auxv, size) != 0)
2098 			error = ENOMEM;
2099 		free(auxv, M_TEMP);
2100 	}
2101 	return (error);
2102 }
2103 
2104 /*
2105  * This sysctl allows a process to retrieve the argument list or process
2106  * title for another process without groping around in the address space
2107  * of the other process.  It also allow a process to set its own "process
2108  * title to a string of its own choice.
2109  */
2110 static int
2111 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
2112 {
2113 	int *name = (int *)arg1;
2114 	u_int namelen = arg2;
2115 	struct pargs *newpa, *pa;
2116 	struct proc *p;
2117 	struct sbuf sb;
2118 	int flags, error = 0, error2;
2119 	pid_t pid;
2120 
2121 	if (namelen != 1)
2122 		return (EINVAL);
2123 
2124 	p = curproc;
2125 	pid = (pid_t)name[0];
2126 	if (pid == -1) {
2127 		pid = p->p_pid;
2128 	}
2129 
2130 	/*
2131 	 * If the query is for this process and it is single-threaded, there
2132 	 * is nobody to modify pargs, thus we can just read.
2133 	 */
2134 	if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
2135 	    (pa = p->p_args) != NULL)
2136 		return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
2137 
2138 	flags = PGET_CANSEE;
2139 	if (req->newptr != NULL)
2140 		flags |= PGET_ISCURRENT;
2141 	error = pget(pid, flags, &p);
2142 	if (error)
2143 		return (error);
2144 
2145 	pa = p->p_args;
2146 	if (pa != NULL) {
2147 		pargs_hold(pa);
2148 		PROC_UNLOCK(p);
2149 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
2150 		pargs_drop(pa);
2151 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
2152 		_PHOLD(p);
2153 		PROC_UNLOCK(p);
2154 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2155 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2156 		error = proc_getargv(curthread, p, &sb);
2157 		error2 = sbuf_finish(&sb);
2158 		PRELE(p);
2159 		sbuf_delete(&sb);
2160 		if (error == 0 && error2 != 0)
2161 			error = error2;
2162 	} else {
2163 		PROC_UNLOCK(p);
2164 	}
2165 	if (error != 0 || req->newptr == NULL)
2166 		return (error);
2167 
2168 	if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
2169 		return (ENOMEM);
2170 
2171 	if (req->newlen == 0) {
2172 		/*
2173 		 * Clear the argument pointer, so that we'll fetch arguments
2174 		 * with proc_getargv() until further notice.
2175 		 */
2176 		newpa = NULL;
2177 	} else {
2178 		newpa = pargs_alloc(req->newlen);
2179 		error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
2180 		if (error != 0) {
2181 			pargs_free(newpa);
2182 			return (error);
2183 		}
2184 	}
2185 	PROC_LOCK(p);
2186 	pa = p->p_args;
2187 	p->p_args = newpa;
2188 	PROC_UNLOCK(p);
2189 	pargs_drop(pa);
2190 	return (0);
2191 }
2192 
2193 /*
2194  * This sysctl allows a process to retrieve environment of another process.
2195  */
2196 static int
2197 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2198 {
2199 	int *name = (int *)arg1;
2200 	u_int namelen = arg2;
2201 	struct proc *p;
2202 	struct sbuf sb;
2203 	int error, error2;
2204 
2205 	if (namelen != 1)
2206 		return (EINVAL);
2207 
2208 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2209 	if (error != 0)
2210 		return (error);
2211 	if ((p->p_flag & P_SYSTEM) != 0) {
2212 		PRELE(p);
2213 		return (0);
2214 	}
2215 
2216 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2217 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2218 	error = proc_getenvv(curthread, p, &sb);
2219 	error2 = sbuf_finish(&sb);
2220 	PRELE(p);
2221 	sbuf_delete(&sb);
2222 	return (error != 0 ? error : error2);
2223 }
2224 
2225 /*
2226  * This sysctl allows a process to retrieve ELF auxiliary vector of
2227  * another process.
2228  */
2229 static int
2230 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2231 {
2232 	int *name = (int *)arg1;
2233 	u_int namelen = arg2;
2234 	struct proc *p;
2235 	struct sbuf sb;
2236 	int error, error2;
2237 
2238 	if (namelen != 1)
2239 		return (EINVAL);
2240 
2241 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2242 	if (error != 0)
2243 		return (error);
2244 	if ((p->p_flag & P_SYSTEM) != 0) {
2245 		PRELE(p);
2246 		return (0);
2247 	}
2248 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2249 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2250 	error = proc_getauxv(curthread, p, &sb);
2251 	error2 = sbuf_finish(&sb);
2252 	PRELE(p);
2253 	sbuf_delete(&sb);
2254 	return (error != 0 ? error : error2);
2255 }
2256 
2257 /*
2258  * Look up the canonical executable path running in the specified process.
2259  * It tries to return the same hardlink name as was used for execve(2).
2260  * This allows the programs that modify their behavior based on their progname,
2261  * to operate correctly.
2262  *
2263  * Result is returned in retbuf, it must not be freed, similar to vn_fullpath()
2264  *   calling conventions.
2265  * binname is a pointer to temporary string buffer of length MAXPATHLEN,
2266  *   allocated and freed by caller.
2267  * freebuf should be freed by caller, from the M_TEMP malloc type.
2268  */
2269 int
2270 proc_get_binpath(struct proc *p, char *binname, char **retbuf,
2271     char **freebuf)
2272 {
2273 	struct nameidata nd;
2274 	struct vnode *vp, *dvp;
2275 	size_t freepath_size;
2276 	int error;
2277 	bool do_fullpath;
2278 
2279 	PROC_LOCK_ASSERT(p, MA_OWNED);
2280 
2281 	vp = p->p_textvp;
2282 	if (vp == NULL) {
2283 		PROC_UNLOCK(p);
2284 		*retbuf = "";
2285 		*freebuf = NULL;
2286 		return (0);
2287 	}
2288 	vref(vp);
2289 	dvp = p->p_textdvp;
2290 	if (dvp != NULL)
2291 		vref(dvp);
2292 	if (p->p_binname != NULL)
2293 		strlcpy(binname, p->p_binname, MAXPATHLEN);
2294 	PROC_UNLOCK(p);
2295 
2296 	do_fullpath = true;
2297 	*freebuf = NULL;
2298 	if (dvp != NULL && binname[0] != '\0') {
2299 		freepath_size = MAXPATHLEN;
2300 		if (vn_fullpath_hardlink(vp, dvp, binname, strlen(binname),
2301 		    retbuf, freebuf, &freepath_size) == 0) {
2302 			/*
2303 			 * Recheck the looked up path.  The binary
2304 			 * might have been renamed or replaced, in
2305 			 * which case we should not report old name.
2306 			 */
2307 			NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, *retbuf);
2308 			error = namei(&nd);
2309 			if (error == 0) {
2310 				if (nd.ni_vp == vp)
2311 					do_fullpath = false;
2312 				vrele(nd.ni_vp);
2313 				NDFREE_PNBUF(&nd);
2314 			}
2315 		}
2316 	}
2317 	if (do_fullpath) {
2318 		free(*freebuf, M_TEMP);
2319 		*freebuf = NULL;
2320 		error = vn_fullpath(vp, retbuf, freebuf);
2321 	}
2322 	vrele(vp);
2323 	if (dvp != NULL)
2324 		vrele(dvp);
2325 	return (error);
2326 }
2327 
2328 /*
2329  * This sysctl allows a process to retrieve the path of the executable for
2330  * itself or another process.
2331  */
2332 static int
2333 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2334 {
2335 	pid_t *pidp = (pid_t *)arg1;
2336 	unsigned int arglen = arg2;
2337 	struct proc *p;
2338 	char *retbuf, *freebuf, *binname;
2339 	int error;
2340 
2341 	if (arglen != 1)
2342 		return (EINVAL);
2343 	binname = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
2344 	binname[0] = '\0';
2345 	if (*pidp == -1) {	/* -1 means this process */
2346 		error = 0;
2347 		p = req->td->td_proc;
2348 		PROC_LOCK(p);
2349 	} else {
2350 		error = pget(*pidp, PGET_CANSEE, &p);
2351 	}
2352 
2353 	if (error == 0)
2354 		error = proc_get_binpath(p, binname, &retbuf, &freebuf);
2355 	free(binname, M_TEMP);
2356 	if (error != 0)
2357 		return (error);
2358 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2359 	free(freebuf, M_TEMP);
2360 	return (error);
2361 }
2362 
2363 static int
2364 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2365 {
2366 	struct proc *p;
2367 	char *sv_name;
2368 	int *name;
2369 	int namelen;
2370 	int error;
2371 
2372 	namelen = arg2;
2373 	if (namelen != 1)
2374 		return (EINVAL);
2375 
2376 	name = (int *)arg1;
2377 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2378 	if (error != 0)
2379 		return (error);
2380 	sv_name = p->p_sysent->sv_name;
2381 	PROC_UNLOCK(p);
2382 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2383 }
2384 
2385 #ifdef KINFO_OVMENTRY_SIZE
2386 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2387 #endif
2388 
2389 #ifdef COMPAT_FREEBSD7
2390 static int
2391 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2392 {
2393 	vm_map_entry_t entry, tmp_entry;
2394 	unsigned int last_timestamp, namelen;
2395 	char *fullpath, *freepath;
2396 	struct kinfo_ovmentry *kve;
2397 	struct vattr va;
2398 	struct ucred *cred;
2399 	int error, *name;
2400 	struct vnode *vp;
2401 	struct proc *p;
2402 	vm_map_t map;
2403 	struct vmspace *vm;
2404 
2405 	namelen = arg2;
2406 	if (namelen != 1)
2407 		return (EINVAL);
2408 
2409 	name = (int *)arg1;
2410 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2411 	if (error != 0)
2412 		return (error);
2413 	vm = vmspace_acquire_ref(p);
2414 	if (vm == NULL) {
2415 		PRELE(p);
2416 		return (ESRCH);
2417 	}
2418 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2419 
2420 	map = &vm->vm_map;
2421 	vm_map_lock_read(map);
2422 	VM_MAP_ENTRY_FOREACH(entry, map) {
2423 		vm_object_t obj, tobj, lobj;
2424 		vm_offset_t addr;
2425 
2426 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2427 			continue;
2428 
2429 		bzero(kve, sizeof(*kve));
2430 		kve->kve_structsize = sizeof(*kve);
2431 
2432 		kve->kve_private_resident = 0;
2433 		obj = entry->object.vm_object;
2434 		if (obj != NULL) {
2435 			VM_OBJECT_RLOCK(obj);
2436 			if (obj->shadow_count == 1)
2437 				kve->kve_private_resident =
2438 				    obj->resident_page_count;
2439 		}
2440 		kve->kve_resident = 0;
2441 		addr = entry->start;
2442 		while (addr < entry->end) {
2443 			if (pmap_extract(map->pmap, addr))
2444 				kve->kve_resident++;
2445 			addr += PAGE_SIZE;
2446 		}
2447 
2448 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2449 			if (tobj != obj) {
2450 				VM_OBJECT_RLOCK(tobj);
2451 				kve->kve_offset += tobj->backing_object_offset;
2452 			}
2453 			if (lobj != obj)
2454 				VM_OBJECT_RUNLOCK(lobj);
2455 			lobj = tobj;
2456 		}
2457 
2458 		kve->kve_start = (void*)entry->start;
2459 		kve->kve_end = (void*)entry->end;
2460 		kve->kve_offset += (off_t)entry->offset;
2461 
2462 		if (entry->protection & VM_PROT_READ)
2463 			kve->kve_protection |= KVME_PROT_READ;
2464 		if (entry->protection & VM_PROT_WRITE)
2465 			kve->kve_protection |= KVME_PROT_WRITE;
2466 		if (entry->protection & VM_PROT_EXECUTE)
2467 			kve->kve_protection |= KVME_PROT_EXEC;
2468 
2469 		if (entry->eflags & MAP_ENTRY_COW)
2470 			kve->kve_flags |= KVME_FLAG_COW;
2471 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2472 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2473 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2474 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2475 
2476 		last_timestamp = map->timestamp;
2477 		vm_map_unlock_read(map);
2478 
2479 		kve->kve_fileid = 0;
2480 		kve->kve_fsid = 0;
2481 		freepath = NULL;
2482 		fullpath = "";
2483 		if (lobj) {
2484 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2485 			if (kve->kve_type == KVME_TYPE_MGTDEVICE)
2486 				kve->kve_type = KVME_TYPE_UNKNOWN;
2487 			if (vp != NULL)
2488 				vref(vp);
2489 			if (lobj != obj)
2490 				VM_OBJECT_RUNLOCK(lobj);
2491 
2492 			kve->kve_ref_count = obj->ref_count;
2493 			kve->kve_shadow_count = obj->shadow_count;
2494 			VM_OBJECT_RUNLOCK(obj);
2495 			if (vp != NULL) {
2496 				vn_fullpath(vp, &fullpath, &freepath);
2497 				cred = curthread->td_ucred;
2498 				vn_lock(vp, LK_SHARED | LK_RETRY);
2499 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2500 					kve->kve_fileid = va.va_fileid;
2501 					/* truncate */
2502 					kve->kve_fsid = va.va_fsid;
2503 				}
2504 				vput(vp);
2505 			}
2506 		} else {
2507 			kve->kve_type = KVME_TYPE_NONE;
2508 			kve->kve_ref_count = 0;
2509 			kve->kve_shadow_count = 0;
2510 		}
2511 
2512 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2513 		if (freepath != NULL)
2514 			free(freepath, M_TEMP);
2515 
2516 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2517 		vm_map_lock_read(map);
2518 		if (error)
2519 			break;
2520 		if (last_timestamp != map->timestamp) {
2521 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2522 			entry = tmp_entry;
2523 		}
2524 	}
2525 	vm_map_unlock_read(map);
2526 	vmspace_free(vm);
2527 	PRELE(p);
2528 	free(kve, M_TEMP);
2529 	return (error);
2530 }
2531 #endif	/* COMPAT_FREEBSD7 */
2532 
2533 #ifdef KINFO_VMENTRY_SIZE
2534 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2535 #endif
2536 
2537 void
2538 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2539     int *resident_count, bool *super)
2540 {
2541 	vm_object_t obj, tobj;
2542 	vm_page_t m, m_adv;
2543 	vm_offset_t addr;
2544 	vm_paddr_t pa;
2545 	vm_pindex_t pi, pi_adv, pindex;
2546 
2547 	*super = false;
2548 	*resident_count = 0;
2549 	if (vmmap_skip_res_cnt)
2550 		return;
2551 
2552 	pa = 0;
2553 	obj = entry->object.vm_object;
2554 	addr = entry->start;
2555 	m_adv = NULL;
2556 	pi = OFF_TO_IDX(entry->offset);
2557 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2558 		if (m_adv != NULL) {
2559 			m = m_adv;
2560 		} else {
2561 			pi_adv = atop(entry->end - addr);
2562 			pindex = pi;
2563 			for (tobj = obj;; tobj = tobj->backing_object) {
2564 				m = vm_page_find_least(tobj, pindex);
2565 				if (m != NULL) {
2566 					if (m->pindex == pindex)
2567 						break;
2568 					if (pi_adv > m->pindex - pindex) {
2569 						pi_adv = m->pindex - pindex;
2570 						m_adv = m;
2571 					}
2572 				}
2573 				if (tobj->backing_object == NULL)
2574 					goto next;
2575 				pindex += OFF_TO_IDX(tobj->
2576 				    backing_object_offset);
2577 			}
2578 		}
2579 		m_adv = NULL;
2580 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2581 		    (addr & (pagesizes[1] - 1)) == 0 &&
2582 		    (pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) {
2583 			*super = true;
2584 			pi_adv = atop(pagesizes[1]);
2585 		} else {
2586 			/*
2587 			 * We do not test the found page on validity.
2588 			 * Either the page is busy and being paged in,
2589 			 * or it was invalidated.  The first case
2590 			 * should be counted as resident, the second
2591 			 * is not so clear; we do account both.
2592 			 */
2593 			pi_adv = 1;
2594 		}
2595 		*resident_count += pi_adv;
2596 next:;
2597 	}
2598 }
2599 
2600 /*
2601  * Must be called with the process locked and will return unlocked.
2602  */
2603 int
2604 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2605 {
2606 	vm_map_entry_t entry, tmp_entry;
2607 	struct vattr va;
2608 	vm_map_t map;
2609 	vm_object_t lobj, nobj, obj, tobj;
2610 	char *fullpath, *freepath;
2611 	struct kinfo_vmentry *kve;
2612 	struct ucred *cred;
2613 	struct vnode *vp;
2614 	struct vmspace *vm;
2615 	vm_offset_t addr;
2616 	unsigned int last_timestamp;
2617 	int error;
2618 	bool guard, super;
2619 
2620 	PROC_LOCK_ASSERT(p, MA_OWNED);
2621 
2622 	_PHOLD(p);
2623 	PROC_UNLOCK(p);
2624 	vm = vmspace_acquire_ref(p);
2625 	if (vm == NULL) {
2626 		PRELE(p);
2627 		return (ESRCH);
2628 	}
2629 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2630 
2631 	error = 0;
2632 	map = &vm->vm_map;
2633 	vm_map_lock_read(map);
2634 	VM_MAP_ENTRY_FOREACH(entry, map) {
2635 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2636 			continue;
2637 
2638 		addr = entry->end;
2639 		bzero(kve, sizeof(*kve));
2640 		obj = entry->object.vm_object;
2641 		if (obj != NULL) {
2642 			if ((obj->flags & OBJ_ANON) != 0)
2643 				kve->kve_obj = (uintptr_t)obj;
2644 
2645 			for (tobj = obj; tobj != NULL;
2646 			    tobj = tobj->backing_object) {
2647 				VM_OBJECT_RLOCK(tobj);
2648 				kve->kve_offset += tobj->backing_object_offset;
2649 				lobj = tobj;
2650 			}
2651 			if (obj->backing_object == NULL)
2652 				kve->kve_private_resident =
2653 				    obj->resident_page_count;
2654 			kern_proc_vmmap_resident(map, entry,
2655 			    &kve->kve_resident, &super);
2656 			if (super)
2657 				kve->kve_flags |= KVME_FLAG_SUPER;
2658 			for (tobj = obj; tobj != NULL; tobj = nobj) {
2659 				nobj = tobj->backing_object;
2660 				if (tobj != obj && tobj != lobj)
2661 					VM_OBJECT_RUNLOCK(tobj);
2662 			}
2663 		} else {
2664 			lobj = NULL;
2665 		}
2666 
2667 		kve->kve_start = entry->start;
2668 		kve->kve_end = entry->end;
2669 		kve->kve_offset += entry->offset;
2670 
2671 		if (entry->protection & VM_PROT_READ)
2672 			kve->kve_protection |= KVME_PROT_READ;
2673 		if (entry->protection & VM_PROT_WRITE)
2674 			kve->kve_protection |= KVME_PROT_WRITE;
2675 		if (entry->protection & VM_PROT_EXECUTE)
2676 			kve->kve_protection |= KVME_PROT_EXEC;
2677 
2678 		if (entry->eflags & MAP_ENTRY_COW)
2679 			kve->kve_flags |= KVME_FLAG_COW;
2680 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2681 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2682 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2683 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2684 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2685 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2686 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2687 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2688 		if (entry->eflags & MAP_ENTRY_USER_WIRED)
2689 			kve->kve_flags |= KVME_FLAG_USER_WIRED;
2690 
2691 		guard = (entry->eflags & MAP_ENTRY_GUARD) != 0;
2692 
2693 		last_timestamp = map->timestamp;
2694 		vm_map_unlock_read(map);
2695 
2696 		freepath = NULL;
2697 		fullpath = "";
2698 		if (lobj != NULL) {
2699 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2700 			if (vp != NULL)
2701 				vref(vp);
2702 			if (lobj != obj)
2703 				VM_OBJECT_RUNLOCK(lobj);
2704 
2705 			kve->kve_ref_count = obj->ref_count;
2706 			kve->kve_shadow_count = obj->shadow_count;
2707 			VM_OBJECT_RUNLOCK(obj);
2708 			if (vp != NULL) {
2709 				vn_fullpath(vp, &fullpath, &freepath);
2710 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2711 				cred = curthread->td_ucred;
2712 				vn_lock(vp, LK_SHARED | LK_RETRY);
2713 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2714 					kve->kve_vn_fileid = va.va_fileid;
2715 					kve->kve_vn_fsid = va.va_fsid;
2716 					kve->kve_vn_fsid_freebsd11 =
2717 					    kve->kve_vn_fsid; /* truncate */
2718 					kve->kve_vn_mode =
2719 					    MAKEIMODE(va.va_type, va.va_mode);
2720 					kve->kve_vn_size = va.va_size;
2721 					kve->kve_vn_rdev = va.va_rdev;
2722 					kve->kve_vn_rdev_freebsd11 =
2723 					    kve->kve_vn_rdev; /* truncate */
2724 					kve->kve_status = KF_ATTR_VALID;
2725 				}
2726 				vput(vp);
2727 			}
2728 		} else {
2729 			kve->kve_type = guard ? KVME_TYPE_GUARD :
2730 			    KVME_TYPE_NONE;
2731 			kve->kve_ref_count = 0;
2732 			kve->kve_shadow_count = 0;
2733 		}
2734 
2735 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2736 		if (freepath != NULL)
2737 			free(freepath, M_TEMP);
2738 
2739 		/* Pack record size down */
2740 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2741 			kve->kve_structsize =
2742 			    offsetof(struct kinfo_vmentry, kve_path) +
2743 			    strlen(kve->kve_path) + 1;
2744 		else
2745 			kve->kve_structsize = sizeof(*kve);
2746 		kve->kve_structsize = roundup(kve->kve_structsize,
2747 		    sizeof(uint64_t));
2748 
2749 		/* Halt filling and truncate rather than exceeding maxlen */
2750 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2751 			error = 0;
2752 			vm_map_lock_read(map);
2753 			break;
2754 		} else if (maxlen != -1)
2755 			maxlen -= kve->kve_structsize;
2756 
2757 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2758 			error = ENOMEM;
2759 		vm_map_lock_read(map);
2760 		if (error != 0)
2761 			break;
2762 		if (last_timestamp != map->timestamp) {
2763 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2764 			entry = tmp_entry;
2765 		}
2766 	}
2767 	vm_map_unlock_read(map);
2768 	vmspace_free(vm);
2769 	PRELE(p);
2770 	free(kve, M_TEMP);
2771 	return (error);
2772 }
2773 
2774 static int
2775 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2776 {
2777 	struct proc *p;
2778 	struct sbuf sb;
2779 	u_int namelen;
2780 	int error, error2, *name;
2781 
2782 	namelen = arg2;
2783 	if (namelen != 1)
2784 		return (EINVAL);
2785 
2786 	name = (int *)arg1;
2787 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2788 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2789 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2790 	if (error != 0) {
2791 		sbuf_delete(&sb);
2792 		return (error);
2793 	}
2794 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2795 	error2 = sbuf_finish(&sb);
2796 	sbuf_delete(&sb);
2797 	return (error != 0 ? error : error2);
2798 }
2799 
2800 #if defined(STACK) || defined(DDB)
2801 static int
2802 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2803 {
2804 	struct kinfo_kstack *kkstp;
2805 	int error, i, *name, numthreads;
2806 	lwpid_t *lwpidarray;
2807 	struct thread *td;
2808 	struct stack *st;
2809 	struct sbuf sb;
2810 	struct proc *p;
2811 	u_int namelen;
2812 
2813 	namelen = arg2;
2814 	if (namelen != 1)
2815 		return (EINVAL);
2816 
2817 	name = (int *)arg1;
2818 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2819 	if (error != 0)
2820 		return (error);
2821 
2822 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2823 	st = stack_create(M_WAITOK);
2824 
2825 	lwpidarray = NULL;
2826 	PROC_LOCK(p);
2827 	do {
2828 		if (lwpidarray != NULL) {
2829 			free(lwpidarray, M_TEMP);
2830 			lwpidarray = NULL;
2831 		}
2832 		numthreads = p->p_numthreads;
2833 		PROC_UNLOCK(p);
2834 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2835 		    M_WAITOK | M_ZERO);
2836 		PROC_LOCK(p);
2837 	} while (numthreads < p->p_numthreads);
2838 
2839 	/*
2840 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2841 	 * of changes could have taken place.  Should we check to see if the
2842 	 * vmspace has been replaced, or the like, in order to prevent
2843 	 * giving a snapshot that spans, say, execve(2), with some threads
2844 	 * before and some after?  Among other things, the credentials could
2845 	 * have changed, in which case the right to extract debug info might
2846 	 * no longer be assured.
2847 	 */
2848 	i = 0;
2849 	FOREACH_THREAD_IN_PROC(p, td) {
2850 		KASSERT(i < numthreads,
2851 		    ("sysctl_kern_proc_kstack: numthreads"));
2852 		lwpidarray[i] = td->td_tid;
2853 		i++;
2854 	}
2855 	PROC_UNLOCK(p);
2856 	numthreads = i;
2857 	for (i = 0; i < numthreads; i++) {
2858 		td = tdfind(lwpidarray[i], p->p_pid);
2859 		if (td == NULL) {
2860 			continue;
2861 		}
2862 		bzero(kkstp, sizeof(*kkstp));
2863 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2864 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2865 		thread_lock(td);
2866 		kkstp->kkst_tid = td->td_tid;
2867 		if (TD_IS_SWAPPED(td))
2868 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2869 		else if (stack_save_td(st, td) == 0)
2870 			kkstp->kkst_state = KKST_STATE_STACKOK;
2871 		else
2872 			kkstp->kkst_state = KKST_STATE_RUNNING;
2873 		thread_unlock(td);
2874 		PROC_UNLOCK(p);
2875 		stack_sbuf_print(&sb, st);
2876 		sbuf_finish(&sb);
2877 		sbuf_delete(&sb);
2878 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2879 		if (error)
2880 			break;
2881 	}
2882 	PRELE(p);
2883 	if (lwpidarray != NULL)
2884 		free(lwpidarray, M_TEMP);
2885 	stack_destroy(st);
2886 	free(kkstp, M_TEMP);
2887 	return (error);
2888 }
2889 #endif
2890 
2891 /*
2892  * This sysctl allows a process to retrieve the full list of groups from
2893  * itself or another process.
2894  */
2895 static int
2896 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2897 {
2898 	pid_t *pidp = (pid_t *)arg1;
2899 	unsigned int arglen = arg2;
2900 	struct proc *p;
2901 	struct ucred *cred;
2902 	int error;
2903 
2904 	if (arglen != 1)
2905 		return (EINVAL);
2906 	if (*pidp == -1) {	/* -1 means this process */
2907 		p = req->td->td_proc;
2908 		PROC_LOCK(p);
2909 	} else {
2910 		error = pget(*pidp, PGET_CANSEE, &p);
2911 		if (error != 0)
2912 			return (error);
2913 	}
2914 
2915 	cred = crhold(p->p_ucred);
2916 	PROC_UNLOCK(p);
2917 
2918 	error = SYSCTL_OUT(req, cred->cr_groups,
2919 	    cred->cr_ngroups * sizeof(gid_t));
2920 	crfree(cred);
2921 	return (error);
2922 }
2923 
2924 /*
2925  * This sysctl allows a process to retrieve or/and set the resource limit for
2926  * another process.
2927  */
2928 static int
2929 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2930 {
2931 	int *name = (int *)arg1;
2932 	u_int namelen = arg2;
2933 	struct rlimit rlim;
2934 	struct proc *p;
2935 	u_int which;
2936 	int flags, error;
2937 
2938 	if (namelen != 2)
2939 		return (EINVAL);
2940 
2941 	which = (u_int)name[1];
2942 	if (which >= RLIM_NLIMITS)
2943 		return (EINVAL);
2944 
2945 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2946 		return (EINVAL);
2947 
2948 	flags = PGET_HOLD | PGET_NOTWEXIT;
2949 	if (req->newptr != NULL)
2950 		flags |= PGET_CANDEBUG;
2951 	else
2952 		flags |= PGET_CANSEE;
2953 	error = pget((pid_t)name[0], flags, &p);
2954 	if (error != 0)
2955 		return (error);
2956 
2957 	/*
2958 	 * Retrieve limit.
2959 	 */
2960 	if (req->oldptr != NULL) {
2961 		PROC_LOCK(p);
2962 		lim_rlimit_proc(p, which, &rlim);
2963 		PROC_UNLOCK(p);
2964 	}
2965 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2966 	if (error != 0)
2967 		goto errout;
2968 
2969 	/*
2970 	 * Set limit.
2971 	 */
2972 	if (req->newptr != NULL) {
2973 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2974 		if (error == 0)
2975 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2976 	}
2977 
2978 errout:
2979 	PRELE(p);
2980 	return (error);
2981 }
2982 
2983 /*
2984  * This sysctl allows a process to retrieve ps_strings structure location of
2985  * another process.
2986  */
2987 static int
2988 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2989 {
2990 	int *name = (int *)arg1;
2991 	u_int namelen = arg2;
2992 	struct proc *p;
2993 	vm_offset_t ps_strings;
2994 	int error;
2995 #ifdef COMPAT_FREEBSD32
2996 	uint32_t ps_strings32;
2997 #endif
2998 
2999 	if (namelen != 1)
3000 		return (EINVAL);
3001 
3002 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3003 	if (error != 0)
3004 		return (error);
3005 #ifdef COMPAT_FREEBSD32
3006 	if ((req->flags & SCTL_MASK32) != 0) {
3007 		/*
3008 		 * We return 0 if the 32 bit emulation request is for a 64 bit
3009 		 * process.
3010 		 */
3011 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
3012 		    PTROUT(PROC_PS_STRINGS(p)) : 0;
3013 		PROC_UNLOCK(p);
3014 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
3015 		return (error);
3016 	}
3017 #endif
3018 	ps_strings = PROC_PS_STRINGS(p);
3019 	PROC_UNLOCK(p);
3020 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
3021 	return (error);
3022 }
3023 
3024 /*
3025  * This sysctl allows a process to retrieve umask of another process.
3026  */
3027 static int
3028 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
3029 {
3030 	int *name = (int *)arg1;
3031 	u_int namelen = arg2;
3032 	struct proc *p;
3033 	int error;
3034 	u_short cmask;
3035 	pid_t pid;
3036 
3037 	if (namelen != 1)
3038 		return (EINVAL);
3039 
3040 	pid = (pid_t)name[0];
3041 	p = curproc;
3042 	if (pid == p->p_pid || pid == 0) {
3043 		cmask = p->p_pd->pd_cmask;
3044 		goto out;
3045 	}
3046 
3047 	error = pget(pid, PGET_WANTREAD, &p);
3048 	if (error != 0)
3049 		return (error);
3050 
3051 	cmask = p->p_pd->pd_cmask;
3052 	PRELE(p);
3053 out:
3054 	error = SYSCTL_OUT(req, &cmask, sizeof(cmask));
3055 	return (error);
3056 }
3057 
3058 /*
3059  * This sysctl allows a process to set and retrieve binary osreldate of
3060  * another process.
3061  */
3062 static int
3063 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
3064 {
3065 	int *name = (int *)arg1;
3066 	u_int namelen = arg2;
3067 	struct proc *p;
3068 	int flags, error, osrel;
3069 
3070 	if (namelen != 1)
3071 		return (EINVAL);
3072 
3073 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
3074 		return (EINVAL);
3075 
3076 	flags = PGET_HOLD | PGET_NOTWEXIT;
3077 	if (req->newptr != NULL)
3078 		flags |= PGET_CANDEBUG;
3079 	else
3080 		flags |= PGET_CANSEE;
3081 	error = pget((pid_t)name[0], flags, &p);
3082 	if (error != 0)
3083 		return (error);
3084 
3085 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
3086 	if (error != 0)
3087 		goto errout;
3088 
3089 	if (req->newptr != NULL) {
3090 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
3091 		if (error != 0)
3092 			goto errout;
3093 		if (osrel < 0) {
3094 			error = EINVAL;
3095 			goto errout;
3096 		}
3097 		p->p_osrel = osrel;
3098 	}
3099 errout:
3100 	PRELE(p);
3101 	return (error);
3102 }
3103 
3104 static int
3105 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
3106 {
3107 	int *name = (int *)arg1;
3108 	u_int namelen = arg2;
3109 	struct proc *p;
3110 	struct kinfo_sigtramp kst;
3111 	const struct sysentvec *sv;
3112 	int error;
3113 #ifdef COMPAT_FREEBSD32
3114 	struct kinfo_sigtramp32 kst32;
3115 #endif
3116 
3117 	if (namelen != 1)
3118 		return (EINVAL);
3119 
3120 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3121 	if (error != 0)
3122 		return (error);
3123 	sv = p->p_sysent;
3124 #ifdef COMPAT_FREEBSD32
3125 	if ((req->flags & SCTL_MASK32) != 0) {
3126 		bzero(&kst32, sizeof(kst32));
3127 		if (SV_PROC_FLAG(p, SV_ILP32)) {
3128 			if (PROC_HAS_SHP(p)) {
3129 				kst32.ksigtramp_start = PROC_SIGCODE(p);
3130 				kst32.ksigtramp_end = kst32.ksigtramp_start +
3131 				    ((sv->sv_flags & SV_DSO_SIG) == 0 ?
3132 				    *sv->sv_szsigcode :
3133 				    (uintptr_t)sv->sv_szsigcode);
3134 			} else {
3135 				kst32.ksigtramp_start = PROC_PS_STRINGS(p) -
3136 				    *sv->sv_szsigcode;
3137 				kst32.ksigtramp_end = PROC_PS_STRINGS(p);
3138 			}
3139 		}
3140 		PROC_UNLOCK(p);
3141 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
3142 		return (error);
3143 	}
3144 #endif
3145 	bzero(&kst, sizeof(kst));
3146 	if (PROC_HAS_SHP(p)) {
3147 		kst.ksigtramp_start = (char *)PROC_SIGCODE(p);
3148 		kst.ksigtramp_end = (char *)kst.ksigtramp_start +
3149 		    ((sv->sv_flags & SV_DSO_SIG) == 0 ? *sv->sv_szsigcode :
3150 		    (uintptr_t)sv->sv_szsigcode);
3151 	} else {
3152 		kst.ksigtramp_start = (char *)PROC_PS_STRINGS(p) -
3153 		    *sv->sv_szsigcode;
3154 		kst.ksigtramp_end = (char *)PROC_PS_STRINGS(p);
3155 	}
3156 	PROC_UNLOCK(p);
3157 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
3158 	return (error);
3159 }
3160 
3161 static int
3162 sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS)
3163 {
3164 	int *name = (int *)arg1;
3165 	u_int namelen = arg2;
3166 	pid_t pid;
3167 	struct proc *p;
3168 	struct thread *td1;
3169 	uintptr_t addr;
3170 #ifdef COMPAT_FREEBSD32
3171 	uint32_t addr32;
3172 #endif
3173 	int error;
3174 
3175 	if (namelen != 1 || req->newptr != NULL)
3176 		return (EINVAL);
3177 
3178 	pid = (pid_t)name[0];
3179 	error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p);
3180 	if (error != 0)
3181 		return (error);
3182 
3183 	PROC_LOCK(p);
3184 #ifdef COMPAT_FREEBSD32
3185 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3186 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3187 			error = EINVAL;
3188 			goto errlocked;
3189 		}
3190 	}
3191 #endif
3192 	if (pid <= PID_MAX) {
3193 		td1 = FIRST_THREAD_IN_PROC(p);
3194 	} else {
3195 		FOREACH_THREAD_IN_PROC(p, td1) {
3196 			if (td1->td_tid == pid)
3197 				break;
3198 		}
3199 	}
3200 	if (td1 == NULL) {
3201 		error = ESRCH;
3202 		goto errlocked;
3203 	}
3204 	/*
3205 	 * The access to the private thread flags.  It is fine as far
3206 	 * as no out-of-thin-air values are read from td_pflags, and
3207 	 * usermode read of the td_sigblock_ptr is racy inherently,
3208 	 * since target process might have already changed it
3209 	 * meantime.
3210 	 */
3211 	if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0)
3212 		addr = (uintptr_t)td1->td_sigblock_ptr;
3213 	else
3214 		error = ENOTTY;
3215 
3216 errlocked:
3217 	_PRELE(p);
3218 	PROC_UNLOCK(p);
3219 	if (error != 0)
3220 		return (error);
3221 
3222 #ifdef COMPAT_FREEBSD32
3223 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3224 		addr32 = addr;
3225 		error = SYSCTL_OUT(req, &addr32, sizeof(addr32));
3226 	} else
3227 #endif
3228 		error = SYSCTL_OUT(req, &addr, sizeof(addr));
3229 	return (error);
3230 }
3231 
3232 static int
3233 sysctl_kern_proc_vm_layout(SYSCTL_HANDLER_ARGS)
3234 {
3235 	struct kinfo_vm_layout kvm;
3236 	struct proc *p;
3237 	struct vmspace *vmspace;
3238 	int error, *name;
3239 
3240 	name = (int *)arg1;
3241 	if ((u_int)arg2 != 1)
3242 		return (EINVAL);
3243 
3244 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3245 	if (error != 0)
3246 		return (error);
3247 #ifdef COMPAT_FREEBSD32
3248 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3249 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3250 			PROC_UNLOCK(p);
3251 			return (EINVAL);
3252 		}
3253 	}
3254 #endif
3255 	vmspace = vmspace_acquire_ref(p);
3256 	PROC_UNLOCK(p);
3257 
3258 	memset(&kvm, 0, sizeof(kvm));
3259 	kvm.kvm_min_user_addr = vm_map_min(&vmspace->vm_map);
3260 	kvm.kvm_max_user_addr = vm_map_max(&vmspace->vm_map);
3261 	kvm.kvm_text_addr = (uintptr_t)vmspace->vm_taddr;
3262 	kvm.kvm_text_size = vmspace->vm_tsize;
3263 	kvm.kvm_data_addr = (uintptr_t)vmspace->vm_daddr;
3264 	kvm.kvm_data_size = vmspace->vm_dsize;
3265 	kvm.kvm_stack_addr = (uintptr_t)vmspace->vm_maxsaddr;
3266 	kvm.kvm_stack_size = vmspace->vm_ssize;
3267 	kvm.kvm_shp_addr = vmspace->vm_shp_base;
3268 	kvm.kvm_shp_size = p->p_sysent->sv_shared_page_len;
3269 	if ((vmspace->vm_map.flags & MAP_WIREFUTURE) != 0)
3270 		kvm.kvm_map_flags |= KMAP_FLAG_WIREFUTURE;
3271 	if ((vmspace->vm_map.flags & MAP_ASLR) != 0)
3272 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR;
3273 	if ((vmspace->vm_map.flags & MAP_ASLR_IGNSTART) != 0)
3274 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_IGNSTART;
3275 	if ((vmspace->vm_map.flags & MAP_WXORX) != 0)
3276 		kvm.kvm_map_flags |= KMAP_FLAG_WXORX;
3277 	if ((vmspace->vm_map.flags & MAP_ASLR_STACK) != 0)
3278 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_STACK;
3279 	if (vmspace->vm_shp_base != p->p_sysent->sv_shared_page_base &&
3280 	    PROC_HAS_SHP(p))
3281 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_SHARED_PAGE;
3282 
3283 #ifdef COMPAT_FREEBSD32
3284 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3285 		struct kinfo_vm_layout32 kvm32;
3286 
3287 		memset(&kvm32, 0, sizeof(kvm32));
3288 		kvm32.kvm_min_user_addr = (uint32_t)kvm.kvm_min_user_addr;
3289 		kvm32.kvm_max_user_addr = (uint32_t)kvm.kvm_max_user_addr;
3290 		kvm32.kvm_text_addr = (uint32_t)kvm.kvm_text_addr;
3291 		kvm32.kvm_text_size = (uint32_t)kvm.kvm_text_size;
3292 		kvm32.kvm_data_addr = (uint32_t)kvm.kvm_data_addr;
3293 		kvm32.kvm_data_size = (uint32_t)kvm.kvm_data_size;
3294 		kvm32.kvm_stack_addr = (uint32_t)kvm.kvm_stack_addr;
3295 		kvm32.kvm_stack_size = (uint32_t)kvm.kvm_stack_size;
3296 		kvm32.kvm_shp_addr = (uint32_t)kvm.kvm_shp_addr;
3297 		kvm32.kvm_shp_size = (uint32_t)kvm.kvm_shp_size;
3298 		kvm32.kvm_map_flags = kvm.kvm_map_flags;
3299 		error = SYSCTL_OUT(req, &kvm32, sizeof(kvm32));
3300 		goto out;
3301 	}
3302 #endif
3303 
3304 	error = SYSCTL_OUT(req, &kvm, sizeof(kvm));
3305 #ifdef COMPAT_FREEBSD32
3306 out:
3307 #endif
3308 	vmspace_free(vmspace);
3309 	return (error);
3310 }
3311 
3312 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,  0,
3313     "Process table");
3314 
3315 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
3316 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
3317 	"Return entire process table");
3318 
3319 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3320 	sysctl_kern_proc, "Process table");
3321 
3322 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
3323 	sysctl_kern_proc, "Process table");
3324 
3325 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3326 	sysctl_kern_proc, "Process table");
3327 
3328 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
3329 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3330 
3331 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
3332 	sysctl_kern_proc, "Process table");
3333 
3334 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3335 	sysctl_kern_proc, "Process table");
3336 
3337 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3338 	sysctl_kern_proc, "Process table");
3339 
3340 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3341 	sysctl_kern_proc, "Process table");
3342 
3343 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
3344 	sysctl_kern_proc, "Return process table, no threads");
3345 
3346 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
3347 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
3348 	sysctl_kern_proc_args, "Process argument list");
3349 
3350 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
3351 	sysctl_kern_proc_env, "Process environment");
3352 
3353 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
3354 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3355 
3356 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3357 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3358 
3359 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3360 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3361 	"Process syscall vector name (ABI type)");
3362 
3363 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3364 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3365 
3366 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3367 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3368 
3369 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3370 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3371 
3372 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3373 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3374 
3375 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3376 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3377 
3378 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3379 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3380 
3381 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3382 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3383 
3384 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3385 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3386 
3387 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3388 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3389 	"Return process table, including threads");
3390 
3391 #ifdef COMPAT_FREEBSD7
3392 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3393 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3394 #endif
3395 
3396 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3397 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3398 
3399 #if defined(STACK) || defined(DDB)
3400 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3401 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3402 #endif
3403 
3404 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3405 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3406 
3407 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3408 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3409 	"Process resource limits");
3410 
3411 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3412 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3413 	"Process ps_strings location");
3414 
3415 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3416 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3417 
3418 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3419 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3420 	"Process binary osreldate");
3421 
3422 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3423 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3424 	"Process signal trampoline location");
3425 
3426 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD |
3427 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk,
3428 	"Thread sigfastblock address");
3429 
3430 static SYSCTL_NODE(_kern_proc, KERN_PROC_VM_LAYOUT, vm_layout, CTLFLAG_RD |
3431 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_vm_layout,
3432 	"Process virtual address space layout info");
3433 
3434 static struct sx stop_all_proc_blocker;
3435 SX_SYSINIT(stop_all_proc_blocker, &stop_all_proc_blocker, "sapblk");
3436 
3437 bool
3438 stop_all_proc_block(void)
3439 {
3440 	return (sx_xlock_sig(&stop_all_proc_blocker) == 0);
3441 }
3442 
3443 void
3444 stop_all_proc_unblock(void)
3445 {
3446 	sx_xunlock(&stop_all_proc_blocker);
3447 }
3448 
3449 int allproc_gen;
3450 
3451 /*
3452  * stop_all_proc() purpose is to stop all process which have usermode,
3453  * except current process for obvious reasons.  This makes it somewhat
3454  * unreliable when invoked from multithreaded process.  The service
3455  * must not be user-callable anyway.
3456  */
3457 void
3458 stop_all_proc(void)
3459 {
3460 	struct proc *cp, *p;
3461 	int r, gen;
3462 	bool restart, seen_stopped, seen_exiting, stopped_some;
3463 
3464 	if (!stop_all_proc_block())
3465 		return;
3466 
3467 	cp = curproc;
3468 allproc_loop:
3469 	sx_xlock(&allproc_lock);
3470 	gen = allproc_gen;
3471 	seen_exiting = seen_stopped = stopped_some = restart = false;
3472 	LIST_REMOVE(cp, p_list);
3473 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3474 	for (;;) {
3475 		p = LIST_NEXT(cp, p_list);
3476 		if (p == NULL)
3477 			break;
3478 		LIST_REMOVE(cp, p_list);
3479 		LIST_INSERT_AFTER(p, cp, p_list);
3480 		PROC_LOCK(p);
3481 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) {
3482 			PROC_UNLOCK(p);
3483 			continue;
3484 		}
3485 		if ((p->p_flag2 & P2_WEXIT) != 0) {
3486 			seen_exiting = true;
3487 			PROC_UNLOCK(p);
3488 			continue;
3489 		}
3490 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3491 			/*
3492 			 * Stopped processes are tolerated when there
3493 			 * are no other processes which might continue
3494 			 * them.  P_STOPPED_SINGLE but not
3495 			 * P_TOTAL_STOP process still has at least one
3496 			 * thread running.
3497 			 */
3498 			seen_stopped = true;
3499 			PROC_UNLOCK(p);
3500 			continue;
3501 		}
3502 		sx_xunlock(&allproc_lock);
3503 		_PHOLD(p);
3504 		r = thread_single(p, SINGLE_ALLPROC);
3505 		if (r != 0)
3506 			restart = true;
3507 		else
3508 			stopped_some = true;
3509 		_PRELE(p);
3510 		PROC_UNLOCK(p);
3511 		sx_xlock(&allproc_lock);
3512 	}
3513 	/* Catch forked children we did not see in iteration. */
3514 	if (gen != allproc_gen)
3515 		restart = true;
3516 	sx_xunlock(&allproc_lock);
3517 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3518 		kern_yield(PRI_USER);
3519 		goto allproc_loop;
3520 	}
3521 }
3522 
3523 void
3524 resume_all_proc(void)
3525 {
3526 	struct proc *cp, *p;
3527 
3528 	cp = curproc;
3529 	sx_xlock(&allproc_lock);
3530 again:
3531 	LIST_REMOVE(cp, p_list);
3532 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3533 	for (;;) {
3534 		p = LIST_NEXT(cp, p_list);
3535 		if (p == NULL)
3536 			break;
3537 		LIST_REMOVE(cp, p_list);
3538 		LIST_INSERT_AFTER(p, cp, p_list);
3539 		PROC_LOCK(p);
3540 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3541 			sx_xunlock(&allproc_lock);
3542 			_PHOLD(p);
3543 			thread_single_end(p, SINGLE_ALLPROC);
3544 			_PRELE(p);
3545 			PROC_UNLOCK(p);
3546 			sx_xlock(&allproc_lock);
3547 		} else {
3548 			PROC_UNLOCK(p);
3549 		}
3550 	}
3551 	/*  Did the loop above missed any stopped process ? */
3552 	FOREACH_PROC_IN_SYSTEM(p) {
3553 		/* No need for proc lock. */
3554 		if ((p->p_flag & P_TOTAL_STOP) != 0)
3555 			goto again;
3556 	}
3557 	sx_xunlock(&allproc_lock);
3558 
3559 	stop_all_proc_unblock();
3560 }
3561 
3562 /* #define	TOTAL_STOP_DEBUG	1 */
3563 #ifdef TOTAL_STOP_DEBUG
3564 volatile static int ap_resume;
3565 #include <sys/mount.h>
3566 
3567 static int
3568 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3569 {
3570 	int error, val;
3571 
3572 	val = 0;
3573 	ap_resume = 0;
3574 	error = sysctl_handle_int(oidp, &val, 0, req);
3575 	if (error != 0 || req->newptr == NULL)
3576 		return (error);
3577 	if (val != 0) {
3578 		stop_all_proc();
3579 		syncer_suspend();
3580 		while (ap_resume == 0)
3581 			;
3582 		syncer_resume();
3583 		resume_all_proc();
3584 	}
3585 	return (0);
3586 }
3587 
3588 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3589     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3590     sysctl_debug_stop_all_proc, "I",
3591     "");
3592 #endif
3593