xref: /freebsd/sys/kern/kern_proc.c (revision eb69d1f144a6fcc765d1b9d44a5ae8082353e70b)
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 __FBSDID("$FreeBSD$");
36 
37 #include "opt_compat.h"
38 #include "opt_ddb.h"
39 #include "opt_ktrace.h"
40 #include "opt_kstack_pages.h"
41 #include "opt_stack.h"
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/elf.h>
46 #include <sys/eventhandler.h>
47 #include <sys/exec.h>
48 #include <sys/jail.h>
49 #include <sys/kernel.h>
50 #include <sys/limits.h>
51 #include <sys/lock.h>
52 #include <sys/loginclass.h>
53 #include <sys/malloc.h>
54 #include <sys/mman.h>
55 #include <sys/mount.h>
56 #include <sys/mutex.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/sysctl.h>
69 #include <sys/filedesc.h>
70 #include <sys/tty.h>
71 #include <sys/signalvar.h>
72 #include <sys/sdt.h>
73 #include <sys/sx.h>
74 #include <sys/user.h>
75 #include <sys/vnode.h>
76 #include <sys/wait.h>
77 
78 #ifdef DDB
79 #include <ddb/ddb.h>
80 #endif
81 
82 #include <vm/vm.h>
83 #include <vm/vm_param.h>
84 #include <vm/vm_extern.h>
85 #include <vm/pmap.h>
86 #include <vm/vm_map.h>
87 #include <vm/vm_object.h>
88 #include <vm/vm_page.h>
89 #include <vm/uma.h>
90 
91 #ifdef COMPAT_FREEBSD32
92 #include <compat/freebsd32/freebsd32.h>
93 #include <compat/freebsd32/freebsd32_util.h>
94 #endif
95 
96 SDT_PROVIDER_DEFINE(proc);
97 SDT_PROBE_DEFINE4(proc, , ctor, entry, "struct proc *", "int", "void *",
98     "int");
99 SDT_PROBE_DEFINE4(proc, , ctor, return, "struct proc *", "int", "void *",
100     "int");
101 SDT_PROBE_DEFINE4(proc, , dtor, entry, "struct proc *", "int", "void *",
102     "struct thread *");
103 SDT_PROBE_DEFINE3(proc, , dtor, return, "struct proc *", "int", "void *");
104 SDT_PROBE_DEFINE3(proc, , init, entry, "struct proc *", "int", "int");
105 SDT_PROBE_DEFINE3(proc, , init, return, "struct proc *", "int", "int");
106 
107 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
108 MALLOC_DEFINE(M_SESSION, "session", "session header");
109 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
110 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
111 
112 static void doenterpgrp(struct proc *, struct pgrp *);
113 static void orphanpg(struct pgrp *pg);
114 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
115 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
116 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
117     int preferthread);
118 static void pgadjustjobc(struct pgrp *pgrp, int entering);
119 static void pgdelete(struct pgrp *);
120 static int proc_ctor(void *mem, int size, void *arg, int flags);
121 static void proc_dtor(void *mem, int size, void *arg);
122 static int proc_init(void *mem, int size, int flags);
123 static void proc_fini(void *mem, int size);
124 static void pargs_free(struct pargs *pa);
125 static struct proc *zpfind_locked(pid_t pid);
126 
127 /*
128  * Other process lists
129  */
130 struct pidhashhead *pidhashtbl;
131 u_long pidhash;
132 struct pgrphashhead *pgrphashtbl;
133 u_long pgrphash;
134 struct proclist allproc;
135 struct proclist zombproc;
136 struct sx __exclusive_cache_line allproc_lock;
137 struct sx __exclusive_cache_line proctree_lock;
138 struct mtx __exclusive_cache_line ppeers_lock;
139 uma_zone_t proc_zone;
140 
141 /*
142  * The offset of various fields in struct proc and struct thread.
143  * These are used by kernel debuggers to enumerate kernel threads and
144  * processes.
145  */
146 const int proc_off_p_pid = offsetof(struct proc, p_pid);
147 const int proc_off_p_comm = offsetof(struct proc, p_comm);
148 const int proc_off_p_list = offsetof(struct proc, p_list);
149 const int proc_off_p_threads = offsetof(struct proc, p_threads);
150 const int thread_off_td_tid = offsetof(struct thread, td_tid);
151 const int thread_off_td_name = offsetof(struct thread, td_name);
152 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
153 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
154 const int thread_off_td_plist = offsetof(struct thread, td_plist);
155 
156 EVENTHANDLER_LIST_DEFINE(process_ctor);
157 EVENTHANDLER_LIST_DEFINE(process_dtor);
158 EVENTHANDLER_LIST_DEFINE(process_init);
159 EVENTHANDLER_LIST_DEFINE(process_fini);
160 EVENTHANDLER_LIST_DEFINE(process_exit);
161 EVENTHANDLER_LIST_DEFINE(process_fork);
162 EVENTHANDLER_LIST_DEFINE(process_exec);
163 
164 EVENTHANDLER_LIST_DECLARE(thread_ctor);
165 EVENTHANDLER_LIST_DECLARE(thread_dtor);
166 
167 int kstack_pages = KSTACK_PAGES;
168 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
169     "Kernel stack size in pages");
170 static int vmmap_skip_res_cnt = 0;
171 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
172     &vmmap_skip_res_cnt, 0,
173     "Skip calculation of the pages resident count in kern.proc.vmmap");
174 
175 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
176 #ifdef COMPAT_FREEBSD32
177 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
178 #endif
179 
180 /*
181  * Initialize global process hashing structures.
182  */
183 void
184 procinit(void)
185 {
186 
187 	sx_init(&allproc_lock, "allproc");
188 	sx_init(&proctree_lock, "proctree");
189 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
190 	LIST_INIT(&allproc);
191 	LIST_INIT(&zombproc);
192 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
193 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
194 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
195 	    proc_ctor, proc_dtor, proc_init, proc_fini,
196 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
197 	uihashinit();
198 }
199 
200 /*
201  * Prepare a proc for use.
202  */
203 static int
204 proc_ctor(void *mem, int size, void *arg, int flags)
205 {
206 	struct proc *p;
207 	struct thread *td;
208 
209 	p = (struct proc *)mem;
210 	SDT_PROBE4(proc, , ctor , entry, p, size, arg, flags);
211 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
212 	SDT_PROBE4(proc, , ctor , return, p, size, arg, flags);
213 	td = FIRST_THREAD_IN_PROC(p);
214 	if (td != NULL) {
215 		/* Make sure all thread constructors are executed */
216 		EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
217 	}
218 	return (0);
219 }
220 
221 /*
222  * Reclaim a proc after use.
223  */
224 static void
225 proc_dtor(void *mem, int size, void *arg)
226 {
227 	struct proc *p;
228 	struct thread *td;
229 
230 	/* INVARIANTS checks go here */
231 	p = (struct proc *)mem;
232 	td = FIRST_THREAD_IN_PROC(p);
233 	SDT_PROBE4(proc, , dtor, entry, p, size, arg, td);
234 	if (td != NULL) {
235 #ifdef INVARIANTS
236 		KASSERT((p->p_numthreads == 1),
237 		    ("bad number of threads in exiting process"));
238 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
239 #endif
240 		/* Free all OSD associated to this thread. */
241 		osd_thread_exit(td);
242 		td_softdep_cleanup(td);
243 		MPASS(td->td_su == NULL);
244 
245 		/* Make sure all thread destructors are executed */
246 		EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
247 	}
248 	EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
249 	if (p->p_ksi != NULL)
250 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
251 	SDT_PROBE3(proc, , dtor, return, p, size, arg);
252 }
253 
254 /*
255  * Initialize type-stable parts of a proc (when newly created).
256  */
257 static int
258 proc_init(void *mem, int size, int flags)
259 {
260 	struct proc *p;
261 
262 	p = (struct proc *)mem;
263 	SDT_PROBE3(proc, , init, entry, p, size, flags);
264 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
265 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
266 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
267 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
268 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
269 	cv_init(&p->p_pwait, "ppwait");
270 	cv_init(&p->p_dbgwait, "dbgwait");
271 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
272 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
273 	p->p_stats = pstats_alloc();
274 	p->p_pgrp = NULL;
275 	SDT_PROBE3(proc, , init, return, p, size, flags);
276 	return (0);
277 }
278 
279 /*
280  * UMA should ensure that this function is never called.
281  * Freeing a proc structure would violate type stability.
282  */
283 static void
284 proc_fini(void *mem, int size)
285 {
286 #ifdef notnow
287 	struct proc *p;
288 
289 	p = (struct proc *)mem;
290 	EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
291 	pstats_free(p->p_stats);
292 	thread_free(FIRST_THREAD_IN_PROC(p));
293 	mtx_destroy(&p->p_mtx);
294 	if (p->p_ksi != NULL)
295 		ksiginfo_free(p->p_ksi);
296 #else
297 	panic("proc reclaimed");
298 #endif
299 }
300 
301 /*
302  * Is p an inferior of the current process?
303  */
304 int
305 inferior(struct proc *p)
306 {
307 
308 	sx_assert(&proctree_lock, SX_LOCKED);
309 	PROC_LOCK_ASSERT(p, MA_OWNED);
310 	for (; p != curproc; p = proc_realparent(p)) {
311 		if (p->p_pid == 0)
312 			return (0);
313 	}
314 	return (1);
315 }
316 
317 struct proc *
318 pfind_locked(pid_t pid)
319 {
320 	struct proc *p;
321 
322 	sx_assert(&allproc_lock, SX_LOCKED);
323 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
324 		if (p->p_pid == pid) {
325 			PROC_LOCK(p);
326 			if (p->p_state == PRS_NEW) {
327 				PROC_UNLOCK(p);
328 				p = NULL;
329 			}
330 			break;
331 		}
332 	}
333 	return (p);
334 }
335 
336 /*
337  * Locate a process by number; return only "live" processes -- i.e., neither
338  * zombies nor newly born but incompletely initialized processes.  By not
339  * returning processes in the PRS_NEW state, we allow callers to avoid
340  * testing for that condition to avoid dereferencing p_ucred, et al.
341  */
342 struct proc *
343 pfind(pid_t pid)
344 {
345 	struct proc *p;
346 
347 	p = curproc;
348 	if (p->p_pid == pid) {
349 		PROC_LOCK(p);
350 		return (p);
351 	}
352 	sx_slock(&allproc_lock);
353 	p = pfind_locked(pid);
354 	sx_sunlock(&allproc_lock);
355 	return (p);
356 }
357 
358 /*
359  * Same as pfind but allow zombies.
360  */
361 struct proc *
362 pfind_any(pid_t pid)
363 {
364 	struct proc *p;
365 
366 	sx_slock(&allproc_lock);
367 	p = pfind_locked(pid);
368 	if (p == NULL)
369 		p = zpfind_locked(pid);
370 	sx_sunlock(&allproc_lock);
371 
372 	return (p);
373 }
374 
375 static struct proc *
376 pfind_tid_locked(pid_t tid)
377 {
378 	struct proc *p;
379 	struct thread *td;
380 
381 	sx_assert(&allproc_lock, SX_LOCKED);
382 	FOREACH_PROC_IN_SYSTEM(p) {
383 		PROC_LOCK(p);
384 		if (p->p_state == PRS_NEW) {
385 			PROC_UNLOCK(p);
386 			continue;
387 		}
388 		FOREACH_THREAD_IN_PROC(p, td) {
389 			if (td->td_tid == tid)
390 				goto found;
391 		}
392 		PROC_UNLOCK(p);
393 	}
394 found:
395 	return (p);
396 }
397 
398 /*
399  * Locate a process group by number.
400  * The caller must hold proctree_lock.
401  */
402 struct pgrp *
403 pgfind(pid_t pgid)
404 {
405 	struct pgrp *pgrp;
406 
407 	sx_assert(&proctree_lock, SX_LOCKED);
408 
409 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
410 		if (pgrp->pg_id == pgid) {
411 			PGRP_LOCK(pgrp);
412 			return (pgrp);
413 		}
414 	}
415 	return (NULL);
416 }
417 
418 /*
419  * Locate process and do additional manipulations, depending on flags.
420  */
421 int
422 pget(pid_t pid, int flags, struct proc **pp)
423 {
424 	struct proc *p;
425 	int error;
426 
427 	p = curproc;
428 	if (p->p_pid == pid) {
429 		PROC_LOCK(p);
430 	} else {
431 		sx_slock(&allproc_lock);
432 		if (pid <= PID_MAX) {
433 			p = pfind_locked(pid);
434 			if (p == NULL && (flags & PGET_NOTWEXIT) == 0)
435 				p = zpfind_locked(pid);
436 		} else if ((flags & PGET_NOTID) == 0) {
437 			p = pfind_tid_locked(pid);
438 		} else {
439 			p = NULL;
440 		}
441 		sx_sunlock(&allproc_lock);
442 		if (p == NULL)
443 			return (ESRCH);
444 		if ((flags & PGET_CANSEE) != 0) {
445 			error = p_cansee(curthread, p);
446 			if (error != 0)
447 				goto errout;
448 		}
449 	}
450 	if ((flags & PGET_CANDEBUG) != 0) {
451 		error = p_candebug(curthread, p);
452 		if (error != 0)
453 			goto errout;
454 	}
455 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
456 		error = EPERM;
457 		goto errout;
458 	}
459 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
460 		error = ESRCH;
461 		goto errout;
462 	}
463 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
464 		/*
465 		 * XXXRW: Not clear ESRCH is the right error during proc
466 		 * execve().
467 		 */
468 		error = ESRCH;
469 		goto errout;
470 	}
471 	if ((flags & PGET_HOLD) != 0) {
472 		_PHOLD(p);
473 		PROC_UNLOCK(p);
474 	}
475 	*pp = p;
476 	return (0);
477 errout:
478 	PROC_UNLOCK(p);
479 	return (error);
480 }
481 
482 /*
483  * Create a new process group.
484  * pgid must be equal to the pid of p.
485  * Begin a new session if required.
486  */
487 int
488 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
489 {
490 
491 	sx_assert(&proctree_lock, SX_XLOCKED);
492 
493 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
494 	KASSERT(p->p_pid == pgid,
495 	    ("enterpgrp: new pgrp and pid != pgid"));
496 	KASSERT(pgfind(pgid) == NULL,
497 	    ("enterpgrp: pgrp with pgid exists"));
498 	KASSERT(!SESS_LEADER(p),
499 	    ("enterpgrp: session leader attempted setpgrp"));
500 
501 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
502 
503 	if (sess != NULL) {
504 		/*
505 		 * new session
506 		 */
507 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
508 		PROC_LOCK(p);
509 		p->p_flag &= ~P_CONTROLT;
510 		PROC_UNLOCK(p);
511 		PGRP_LOCK(pgrp);
512 		sess->s_leader = p;
513 		sess->s_sid = p->p_pid;
514 		refcount_init(&sess->s_count, 1);
515 		sess->s_ttyvp = NULL;
516 		sess->s_ttydp = NULL;
517 		sess->s_ttyp = NULL;
518 		bcopy(p->p_session->s_login, sess->s_login,
519 			    sizeof(sess->s_login));
520 		pgrp->pg_session = sess;
521 		KASSERT(p == curproc,
522 		    ("enterpgrp: mksession and p != curproc"));
523 	} else {
524 		pgrp->pg_session = p->p_session;
525 		sess_hold(pgrp->pg_session);
526 		PGRP_LOCK(pgrp);
527 	}
528 	pgrp->pg_id = pgid;
529 	LIST_INIT(&pgrp->pg_members);
530 
531 	/*
532 	 * As we have an exclusive lock of proctree_lock,
533 	 * this should not deadlock.
534 	 */
535 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
536 	pgrp->pg_jobc = 0;
537 	SLIST_INIT(&pgrp->pg_sigiolst);
538 	PGRP_UNLOCK(pgrp);
539 
540 	doenterpgrp(p, pgrp);
541 
542 	return (0);
543 }
544 
545 /*
546  * Move p to an existing process group
547  */
548 int
549 enterthispgrp(struct proc *p, struct pgrp *pgrp)
550 {
551 
552 	sx_assert(&proctree_lock, SX_XLOCKED);
553 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
554 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
555 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
556 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
557 	KASSERT(pgrp->pg_session == p->p_session,
558 		("%s: pgrp's session %p, p->p_session %p.\n",
559 		__func__,
560 		pgrp->pg_session,
561 		p->p_session));
562 	KASSERT(pgrp != p->p_pgrp,
563 		("%s: p belongs to pgrp.", __func__));
564 
565 	doenterpgrp(p, pgrp);
566 
567 	return (0);
568 }
569 
570 /*
571  * Move p to a process group
572  */
573 static void
574 doenterpgrp(struct proc *p, struct pgrp *pgrp)
575 {
576 	struct pgrp *savepgrp;
577 
578 	sx_assert(&proctree_lock, SX_XLOCKED);
579 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
580 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
581 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
582 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
583 
584 	savepgrp = p->p_pgrp;
585 
586 	/*
587 	 * Adjust eligibility of affected pgrps to participate in job control.
588 	 * Increment eligibility counts before decrementing, otherwise we
589 	 * could reach 0 spuriously during the first call.
590 	 */
591 	fixjobc(p, pgrp, 1);
592 	fixjobc(p, p->p_pgrp, 0);
593 
594 	PGRP_LOCK(pgrp);
595 	PGRP_LOCK(savepgrp);
596 	PROC_LOCK(p);
597 	LIST_REMOVE(p, p_pglist);
598 	p->p_pgrp = pgrp;
599 	PROC_UNLOCK(p);
600 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
601 	PGRP_UNLOCK(savepgrp);
602 	PGRP_UNLOCK(pgrp);
603 	if (LIST_EMPTY(&savepgrp->pg_members))
604 		pgdelete(savepgrp);
605 }
606 
607 /*
608  * remove process from process group
609  */
610 int
611 leavepgrp(struct proc *p)
612 {
613 	struct pgrp *savepgrp;
614 
615 	sx_assert(&proctree_lock, SX_XLOCKED);
616 	savepgrp = p->p_pgrp;
617 	PGRP_LOCK(savepgrp);
618 	PROC_LOCK(p);
619 	LIST_REMOVE(p, p_pglist);
620 	p->p_pgrp = NULL;
621 	PROC_UNLOCK(p);
622 	PGRP_UNLOCK(savepgrp);
623 	if (LIST_EMPTY(&savepgrp->pg_members))
624 		pgdelete(savepgrp);
625 	return (0);
626 }
627 
628 /*
629  * delete a process group
630  */
631 static void
632 pgdelete(struct pgrp *pgrp)
633 {
634 	struct session *savesess;
635 	struct tty *tp;
636 
637 	sx_assert(&proctree_lock, SX_XLOCKED);
638 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
639 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
640 
641 	/*
642 	 * Reset any sigio structures pointing to us as a result of
643 	 * F_SETOWN with our pgid.
644 	 */
645 	funsetownlst(&pgrp->pg_sigiolst);
646 
647 	PGRP_LOCK(pgrp);
648 	tp = pgrp->pg_session->s_ttyp;
649 	LIST_REMOVE(pgrp, pg_hash);
650 	savesess = pgrp->pg_session;
651 	PGRP_UNLOCK(pgrp);
652 
653 	/* Remove the reference to the pgrp before deallocating it. */
654 	if (tp != NULL) {
655 		tty_lock(tp);
656 		tty_rel_pgrp(tp, pgrp);
657 	}
658 
659 	mtx_destroy(&pgrp->pg_mtx);
660 	free(pgrp, M_PGRP);
661 	sess_release(savesess);
662 }
663 
664 static void
665 pgadjustjobc(struct pgrp *pgrp, int entering)
666 {
667 
668 	PGRP_LOCK(pgrp);
669 	if (entering)
670 		pgrp->pg_jobc++;
671 	else {
672 		--pgrp->pg_jobc;
673 		if (pgrp->pg_jobc == 0)
674 			orphanpg(pgrp);
675 	}
676 	PGRP_UNLOCK(pgrp);
677 }
678 
679 /*
680  * Adjust pgrp jobc counters when specified process changes process group.
681  * We count the number of processes in each process group that "qualify"
682  * the group for terminal job control (those with a parent in a different
683  * process group of the same session).  If that count reaches zero, the
684  * process group becomes orphaned.  Check both the specified process'
685  * process group and that of its children.
686  * entering == 0 => p is leaving specified group.
687  * entering == 1 => p is entering specified group.
688  */
689 void
690 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
691 {
692 	struct pgrp *hispgrp;
693 	struct session *mysession;
694 	struct proc *q;
695 
696 	sx_assert(&proctree_lock, SX_LOCKED);
697 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
698 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
699 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
700 
701 	/*
702 	 * Check p's parent to see whether p qualifies its own process
703 	 * group; if so, adjust count for p's process group.
704 	 */
705 	mysession = pgrp->pg_session;
706 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
707 	    hispgrp->pg_session == mysession)
708 		pgadjustjobc(pgrp, entering);
709 
710 	/*
711 	 * Check this process' children to see whether they qualify
712 	 * their process groups; if so, adjust counts for children's
713 	 * process groups.
714 	 */
715 	LIST_FOREACH(q, &p->p_children, p_sibling) {
716 		hispgrp = q->p_pgrp;
717 		if (hispgrp == pgrp ||
718 		    hispgrp->pg_session != mysession)
719 			continue;
720 		if (q->p_state == PRS_ZOMBIE)
721 			continue;
722 		pgadjustjobc(hispgrp, entering);
723 	}
724 }
725 
726 void
727 killjobc(void)
728 {
729 	struct session *sp;
730 	struct tty *tp;
731 	struct proc *p;
732 	struct vnode *ttyvp;
733 
734 	p = curproc;
735 	MPASS(p->p_flag & P_WEXIT);
736 	/*
737 	 * Do a quick check to see if there is anything to do with the
738 	 * proctree_lock held. pgrp and LIST_EMPTY checks are for fixjobc().
739 	 */
740 	PROC_LOCK(p);
741 	if (!SESS_LEADER(p) &&
742 	    (p->p_pgrp == p->p_pptr->p_pgrp) &&
743 	    LIST_EMPTY(&p->p_children)) {
744 		PROC_UNLOCK(p);
745 		return;
746 	}
747 	PROC_UNLOCK(p);
748 
749 	sx_xlock(&proctree_lock);
750 	if (SESS_LEADER(p)) {
751 		sp = p->p_session;
752 
753 		/*
754 		 * s_ttyp is not zero'd; we use this to indicate that
755 		 * the session once had a controlling terminal. (for
756 		 * logging and informational purposes)
757 		 */
758 		SESS_LOCK(sp);
759 		ttyvp = sp->s_ttyvp;
760 		tp = sp->s_ttyp;
761 		sp->s_ttyvp = NULL;
762 		sp->s_ttydp = NULL;
763 		sp->s_leader = NULL;
764 		SESS_UNLOCK(sp);
765 
766 		/*
767 		 * Signal foreground pgrp and revoke access to
768 		 * controlling terminal if it has not been revoked
769 		 * already.
770 		 *
771 		 * Because the TTY may have been revoked in the mean
772 		 * time and could already have a new session associated
773 		 * with it, make sure we don't send a SIGHUP to a
774 		 * foreground process group that does not belong to this
775 		 * session.
776 		 */
777 
778 		if (tp != NULL) {
779 			tty_lock(tp);
780 			if (tp->t_session == sp)
781 				tty_signal_pgrp(tp, SIGHUP);
782 			tty_unlock(tp);
783 		}
784 
785 		if (ttyvp != NULL) {
786 			sx_xunlock(&proctree_lock);
787 			if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
788 				VOP_REVOKE(ttyvp, REVOKEALL);
789 				VOP_UNLOCK(ttyvp, 0);
790 			}
791 			vrele(ttyvp);
792 			sx_xlock(&proctree_lock);
793 		}
794 	}
795 	fixjobc(p, p->p_pgrp, 0);
796 	sx_xunlock(&proctree_lock);
797 }
798 
799 /*
800  * A process group has become orphaned;
801  * if there are any stopped processes in the group,
802  * hang-up all process in that group.
803  */
804 static void
805 orphanpg(struct pgrp *pg)
806 {
807 	struct proc *p;
808 
809 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
810 
811 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
812 		PROC_LOCK(p);
813 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
814 			PROC_UNLOCK(p);
815 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
816 				PROC_LOCK(p);
817 				kern_psignal(p, SIGHUP);
818 				kern_psignal(p, SIGCONT);
819 				PROC_UNLOCK(p);
820 			}
821 			return;
822 		}
823 		PROC_UNLOCK(p);
824 	}
825 }
826 
827 void
828 sess_hold(struct session *s)
829 {
830 
831 	refcount_acquire(&s->s_count);
832 }
833 
834 void
835 sess_release(struct session *s)
836 {
837 
838 	if (refcount_release(&s->s_count)) {
839 		if (s->s_ttyp != NULL) {
840 			tty_lock(s->s_ttyp);
841 			tty_rel_sess(s->s_ttyp, s);
842 		}
843 		mtx_destroy(&s->s_mtx);
844 		free(s, M_SESSION);
845 	}
846 }
847 
848 #ifdef DDB
849 
850 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
851 {
852 	struct pgrp *pgrp;
853 	struct proc *p;
854 	int i;
855 
856 	for (i = 0; i <= pgrphash; i++) {
857 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
858 			printf("\tindx %d\n", i);
859 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
860 				printf(
861 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
862 				    (void *)pgrp, (long)pgrp->pg_id,
863 				    (void *)pgrp->pg_session,
864 				    pgrp->pg_session->s_count,
865 				    (void *)LIST_FIRST(&pgrp->pg_members));
866 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
867 					printf("\t\tpid %ld addr %p pgrp %p\n",
868 					    (long)p->p_pid, (void *)p,
869 					    (void *)p->p_pgrp);
870 				}
871 			}
872 		}
873 	}
874 }
875 #endif /* DDB */
876 
877 /*
878  * Calculate the kinfo_proc members which contain process-wide
879  * informations.
880  * Must be called with the target process locked.
881  */
882 static void
883 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
884 {
885 	struct thread *td;
886 
887 	PROC_LOCK_ASSERT(p, MA_OWNED);
888 
889 	kp->ki_estcpu = 0;
890 	kp->ki_pctcpu = 0;
891 	FOREACH_THREAD_IN_PROC(p, td) {
892 		thread_lock(td);
893 		kp->ki_pctcpu += sched_pctcpu(td);
894 		kp->ki_estcpu += sched_estcpu(td);
895 		thread_unlock(td);
896 	}
897 }
898 
899 /*
900  * Clear kinfo_proc and fill in any information that is common
901  * to all threads in the process.
902  * Must be called with the target process locked.
903  */
904 static void
905 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
906 {
907 	struct thread *td0;
908 	struct tty *tp;
909 	struct session *sp;
910 	struct ucred *cred;
911 	struct sigacts *ps;
912 	struct timeval boottime;
913 
914 	/* For proc_realparent. */
915 	sx_assert(&proctree_lock, SX_LOCKED);
916 	PROC_LOCK_ASSERT(p, MA_OWNED);
917 	bzero(kp, sizeof(*kp));
918 
919 	kp->ki_structsize = sizeof(*kp);
920 	kp->ki_paddr = p;
921 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
922 	kp->ki_args = p->p_args;
923 	kp->ki_textvp = p->p_textvp;
924 #ifdef KTRACE
925 	kp->ki_tracep = p->p_tracevp;
926 	kp->ki_traceflag = p->p_traceflag;
927 #endif
928 	kp->ki_fd = p->p_fd;
929 	kp->ki_vmspace = p->p_vmspace;
930 	kp->ki_flag = p->p_flag;
931 	kp->ki_flag2 = p->p_flag2;
932 	cred = p->p_ucred;
933 	if (cred) {
934 		kp->ki_uid = cred->cr_uid;
935 		kp->ki_ruid = cred->cr_ruid;
936 		kp->ki_svuid = cred->cr_svuid;
937 		kp->ki_cr_flags = 0;
938 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
939 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
940 		/* XXX bde doesn't like KI_NGROUPS */
941 		if (cred->cr_ngroups > KI_NGROUPS) {
942 			kp->ki_ngroups = KI_NGROUPS;
943 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
944 		} else
945 			kp->ki_ngroups = cred->cr_ngroups;
946 		bcopy(cred->cr_groups, kp->ki_groups,
947 		    kp->ki_ngroups * sizeof(gid_t));
948 		kp->ki_rgid = cred->cr_rgid;
949 		kp->ki_svgid = cred->cr_svgid;
950 		/* If jailed(cred), emulate the old P_JAILED flag. */
951 		if (jailed(cred)) {
952 			kp->ki_flag |= P_JAILED;
953 			/* If inside the jail, use 0 as a jail ID. */
954 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
955 				kp->ki_jid = cred->cr_prison->pr_id;
956 		}
957 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
958 		    sizeof(kp->ki_loginclass));
959 	}
960 	ps = p->p_sigacts;
961 	if (ps) {
962 		mtx_lock(&ps->ps_mtx);
963 		kp->ki_sigignore = ps->ps_sigignore;
964 		kp->ki_sigcatch = ps->ps_sigcatch;
965 		mtx_unlock(&ps->ps_mtx);
966 	}
967 	if (p->p_state != PRS_NEW &&
968 	    p->p_state != PRS_ZOMBIE &&
969 	    p->p_vmspace != NULL) {
970 		struct vmspace *vm = p->p_vmspace;
971 
972 		kp->ki_size = vm->vm_map.size;
973 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
974 		FOREACH_THREAD_IN_PROC(p, td0) {
975 			if (!TD_IS_SWAPPED(td0))
976 				kp->ki_rssize += td0->td_kstack_pages;
977 		}
978 		kp->ki_swrss = vm->vm_swrss;
979 		kp->ki_tsize = vm->vm_tsize;
980 		kp->ki_dsize = vm->vm_dsize;
981 		kp->ki_ssize = vm->vm_ssize;
982 	} else if (p->p_state == PRS_ZOMBIE)
983 		kp->ki_stat = SZOMB;
984 	if (kp->ki_flag & P_INMEM)
985 		kp->ki_sflag = PS_INMEM;
986 	else
987 		kp->ki_sflag = 0;
988 	/* Calculate legacy swtime as seconds since 'swtick'. */
989 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
990 	kp->ki_pid = p->p_pid;
991 	kp->ki_nice = p->p_nice;
992 	kp->ki_fibnum = p->p_fibnum;
993 	kp->ki_start = p->p_stats->p_start;
994 	getboottime(&boottime);
995 	timevaladd(&kp->ki_start, &boottime);
996 	PROC_STATLOCK(p);
997 	rufetch(p, &kp->ki_rusage);
998 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
999 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
1000 	PROC_STATUNLOCK(p);
1001 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1002 	/* Some callers want child times in a single value. */
1003 	kp->ki_childtime = kp->ki_childstime;
1004 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1005 
1006 	FOREACH_THREAD_IN_PROC(p, td0)
1007 		kp->ki_cow += td0->td_cow;
1008 
1009 	tp = NULL;
1010 	if (p->p_pgrp) {
1011 		kp->ki_pgid = p->p_pgrp->pg_id;
1012 		kp->ki_jobc = p->p_pgrp->pg_jobc;
1013 		sp = p->p_pgrp->pg_session;
1014 
1015 		if (sp != NULL) {
1016 			kp->ki_sid = sp->s_sid;
1017 			SESS_LOCK(sp);
1018 			strlcpy(kp->ki_login, sp->s_login,
1019 			    sizeof(kp->ki_login));
1020 			if (sp->s_ttyvp)
1021 				kp->ki_kiflag |= KI_CTTY;
1022 			if (SESS_LEADER(p))
1023 				kp->ki_kiflag |= KI_SLEADER;
1024 			/* XXX proctree_lock */
1025 			tp = sp->s_ttyp;
1026 			SESS_UNLOCK(sp);
1027 		}
1028 	}
1029 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1030 		kp->ki_tdev = tty_udev(tp);
1031 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1032 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1033 		if (tp->t_session)
1034 			kp->ki_tsid = tp->t_session->s_sid;
1035 	} else {
1036 		kp->ki_tdev = NODEV;
1037 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1038 	}
1039 	if (p->p_comm[0] != '\0')
1040 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1041 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1042 	    p->p_sysent->sv_name[0] != '\0')
1043 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1044 	kp->ki_siglist = p->p_siglist;
1045 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1046 	kp->ki_acflag = p->p_acflag;
1047 	kp->ki_lock = p->p_lock;
1048 	if (p->p_pptr) {
1049 		kp->ki_ppid = proc_realparent(p)->p_pid;
1050 		if (p->p_flag & P_TRACED)
1051 			kp->ki_tracer = p->p_pptr->p_pid;
1052 	}
1053 }
1054 
1055 /*
1056  * Fill in information that is thread specific.  Must be called with
1057  * target process locked.  If 'preferthread' is set, overwrite certain
1058  * process-related fields that are maintained for both threads and
1059  * processes.
1060  */
1061 static void
1062 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1063 {
1064 	struct proc *p;
1065 
1066 	p = td->td_proc;
1067 	kp->ki_tdaddr = td;
1068 	PROC_LOCK_ASSERT(p, MA_OWNED);
1069 
1070 	if (preferthread)
1071 		PROC_STATLOCK(p);
1072 	thread_lock(td);
1073 	if (td->td_wmesg != NULL)
1074 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1075 	else
1076 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1077 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1078 	    sizeof(kp->ki_tdname)) {
1079 		strlcpy(kp->ki_moretdname,
1080 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1081 		    sizeof(kp->ki_moretdname));
1082 	} else {
1083 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1084 	}
1085 	if (TD_ON_LOCK(td)) {
1086 		kp->ki_kiflag |= KI_LOCKBLOCK;
1087 		strlcpy(kp->ki_lockname, td->td_lockname,
1088 		    sizeof(kp->ki_lockname));
1089 	} else {
1090 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1091 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1092 	}
1093 
1094 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1095 		if (TD_ON_RUNQ(td) ||
1096 		    TD_CAN_RUN(td) ||
1097 		    TD_IS_RUNNING(td)) {
1098 			kp->ki_stat = SRUN;
1099 		} else if (P_SHOULDSTOP(p)) {
1100 			kp->ki_stat = SSTOP;
1101 		} else if (TD_IS_SLEEPING(td)) {
1102 			kp->ki_stat = SSLEEP;
1103 		} else if (TD_ON_LOCK(td)) {
1104 			kp->ki_stat = SLOCK;
1105 		} else {
1106 			kp->ki_stat = SWAIT;
1107 		}
1108 	} else if (p->p_state == PRS_ZOMBIE) {
1109 		kp->ki_stat = SZOMB;
1110 	} else {
1111 		kp->ki_stat = SIDL;
1112 	}
1113 
1114 	/* Things in the thread */
1115 	kp->ki_wchan = td->td_wchan;
1116 	kp->ki_pri.pri_level = td->td_priority;
1117 	kp->ki_pri.pri_native = td->td_base_pri;
1118 
1119 	/*
1120 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1121 	 * the maximum u_char CPU value.
1122 	 */
1123 	if (td->td_lastcpu == NOCPU)
1124 		kp->ki_lastcpu_old = NOCPU_OLD;
1125 	else if (td->td_lastcpu > MAXCPU_OLD)
1126 		kp->ki_lastcpu_old = MAXCPU_OLD;
1127 	else
1128 		kp->ki_lastcpu_old = td->td_lastcpu;
1129 
1130 	if (td->td_oncpu == NOCPU)
1131 		kp->ki_oncpu_old = NOCPU_OLD;
1132 	else if (td->td_oncpu > MAXCPU_OLD)
1133 		kp->ki_oncpu_old = MAXCPU_OLD;
1134 	else
1135 		kp->ki_oncpu_old = td->td_oncpu;
1136 
1137 	kp->ki_lastcpu = td->td_lastcpu;
1138 	kp->ki_oncpu = td->td_oncpu;
1139 	kp->ki_tdflags = td->td_flags;
1140 	kp->ki_tid = td->td_tid;
1141 	kp->ki_numthreads = p->p_numthreads;
1142 	kp->ki_pcb = td->td_pcb;
1143 	kp->ki_kstack = (void *)td->td_kstack;
1144 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1145 	kp->ki_pri.pri_class = td->td_pri_class;
1146 	kp->ki_pri.pri_user = td->td_user_pri;
1147 
1148 	if (preferthread) {
1149 		rufetchtd(td, &kp->ki_rusage);
1150 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1151 		kp->ki_pctcpu = sched_pctcpu(td);
1152 		kp->ki_estcpu = sched_estcpu(td);
1153 		kp->ki_cow = td->td_cow;
1154 	}
1155 
1156 	/* We can't get this anymore but ps etc never used it anyway. */
1157 	kp->ki_rqindex = 0;
1158 
1159 	if (preferthread)
1160 		kp->ki_siglist = td->td_siglist;
1161 	kp->ki_sigmask = td->td_sigmask;
1162 	thread_unlock(td);
1163 	if (preferthread)
1164 		PROC_STATUNLOCK(p);
1165 }
1166 
1167 /*
1168  * Fill in a kinfo_proc structure for the specified process.
1169  * Must be called with the target process locked.
1170  */
1171 void
1172 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1173 {
1174 
1175 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1176 
1177 	fill_kinfo_proc_only(p, kp);
1178 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1179 	fill_kinfo_aggregate(p, kp);
1180 }
1181 
1182 struct pstats *
1183 pstats_alloc(void)
1184 {
1185 
1186 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1187 }
1188 
1189 /*
1190  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1191  */
1192 void
1193 pstats_fork(struct pstats *src, struct pstats *dst)
1194 {
1195 
1196 	bzero(&dst->pstat_startzero,
1197 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1198 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1199 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1200 }
1201 
1202 void
1203 pstats_free(struct pstats *ps)
1204 {
1205 
1206 	free(ps, M_SUBPROC);
1207 }
1208 
1209 static struct proc *
1210 zpfind_locked(pid_t pid)
1211 {
1212 	struct proc *p;
1213 
1214 	sx_assert(&allproc_lock, SX_LOCKED);
1215 	LIST_FOREACH(p, &zombproc, p_list) {
1216 		if (p->p_pid == pid) {
1217 			PROC_LOCK(p);
1218 			break;
1219 		}
1220 	}
1221 	return (p);
1222 }
1223 
1224 /*
1225  * Locate a zombie process by number
1226  */
1227 struct proc *
1228 zpfind(pid_t pid)
1229 {
1230 	struct proc *p;
1231 
1232 	sx_slock(&allproc_lock);
1233 	p = zpfind_locked(pid);
1234 	sx_sunlock(&allproc_lock);
1235 	return (p);
1236 }
1237 
1238 #ifdef COMPAT_FREEBSD32
1239 
1240 /*
1241  * This function is typically used to copy out the kernel address, so
1242  * it can be replaced by assignment of zero.
1243  */
1244 static inline uint32_t
1245 ptr32_trim(void *ptr)
1246 {
1247 	uintptr_t uptr;
1248 
1249 	uptr = (uintptr_t)ptr;
1250 	return ((uptr > UINT_MAX) ? 0 : uptr);
1251 }
1252 
1253 #define PTRTRIM_CP(src,dst,fld) \
1254 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1255 
1256 static void
1257 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1258 {
1259 	int i;
1260 
1261 	bzero(ki32, sizeof(struct kinfo_proc32));
1262 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1263 	CP(*ki, *ki32, ki_layout);
1264 	PTRTRIM_CP(*ki, *ki32, ki_args);
1265 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1266 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1267 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1268 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1269 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1270 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1271 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1272 	CP(*ki, *ki32, ki_pid);
1273 	CP(*ki, *ki32, ki_ppid);
1274 	CP(*ki, *ki32, ki_pgid);
1275 	CP(*ki, *ki32, ki_tpgid);
1276 	CP(*ki, *ki32, ki_sid);
1277 	CP(*ki, *ki32, ki_tsid);
1278 	CP(*ki, *ki32, ki_jobc);
1279 	CP(*ki, *ki32, ki_tdev);
1280 	CP(*ki, *ki32, ki_tdev_freebsd11);
1281 	CP(*ki, *ki32, ki_siglist);
1282 	CP(*ki, *ki32, ki_sigmask);
1283 	CP(*ki, *ki32, ki_sigignore);
1284 	CP(*ki, *ki32, ki_sigcatch);
1285 	CP(*ki, *ki32, ki_uid);
1286 	CP(*ki, *ki32, ki_ruid);
1287 	CP(*ki, *ki32, ki_svuid);
1288 	CP(*ki, *ki32, ki_rgid);
1289 	CP(*ki, *ki32, ki_svgid);
1290 	CP(*ki, *ki32, ki_ngroups);
1291 	for (i = 0; i < KI_NGROUPS; i++)
1292 		CP(*ki, *ki32, ki_groups[i]);
1293 	CP(*ki, *ki32, ki_size);
1294 	CP(*ki, *ki32, ki_rssize);
1295 	CP(*ki, *ki32, ki_swrss);
1296 	CP(*ki, *ki32, ki_tsize);
1297 	CP(*ki, *ki32, ki_dsize);
1298 	CP(*ki, *ki32, ki_ssize);
1299 	CP(*ki, *ki32, ki_xstat);
1300 	CP(*ki, *ki32, ki_acflag);
1301 	CP(*ki, *ki32, ki_pctcpu);
1302 	CP(*ki, *ki32, ki_estcpu);
1303 	CP(*ki, *ki32, ki_slptime);
1304 	CP(*ki, *ki32, ki_swtime);
1305 	CP(*ki, *ki32, ki_cow);
1306 	CP(*ki, *ki32, ki_runtime);
1307 	TV_CP(*ki, *ki32, ki_start);
1308 	TV_CP(*ki, *ki32, ki_childtime);
1309 	CP(*ki, *ki32, ki_flag);
1310 	CP(*ki, *ki32, ki_kiflag);
1311 	CP(*ki, *ki32, ki_traceflag);
1312 	CP(*ki, *ki32, ki_stat);
1313 	CP(*ki, *ki32, ki_nice);
1314 	CP(*ki, *ki32, ki_lock);
1315 	CP(*ki, *ki32, ki_rqindex);
1316 	CP(*ki, *ki32, ki_oncpu);
1317 	CP(*ki, *ki32, ki_lastcpu);
1318 
1319 	/* XXX TODO: wrap cpu value as appropriate */
1320 	CP(*ki, *ki32, ki_oncpu_old);
1321 	CP(*ki, *ki32, ki_lastcpu_old);
1322 
1323 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1324 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1325 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1326 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1327 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1328 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1329 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1330 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1331 	CP(*ki, *ki32, ki_tracer);
1332 	CP(*ki, *ki32, ki_flag2);
1333 	CP(*ki, *ki32, ki_fibnum);
1334 	CP(*ki, *ki32, ki_cr_flags);
1335 	CP(*ki, *ki32, ki_jid);
1336 	CP(*ki, *ki32, ki_numthreads);
1337 	CP(*ki, *ki32, ki_tid);
1338 	CP(*ki, *ki32, ki_pri);
1339 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1340 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1341 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1342 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1343 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1344 	PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1345 	CP(*ki, *ki32, ki_sflag);
1346 	CP(*ki, *ki32, ki_tdflags);
1347 }
1348 #endif
1349 
1350 int
1351 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1352 {
1353 	struct thread *td;
1354 	struct kinfo_proc ki;
1355 #ifdef COMPAT_FREEBSD32
1356 	struct kinfo_proc32 ki32;
1357 #endif
1358 	int error;
1359 
1360 	PROC_LOCK_ASSERT(p, MA_OWNED);
1361 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1362 
1363 	error = 0;
1364 	fill_kinfo_proc(p, &ki);
1365 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1366 #ifdef COMPAT_FREEBSD32
1367 		if ((flags & KERN_PROC_MASK32) != 0) {
1368 			freebsd32_kinfo_proc_out(&ki, &ki32);
1369 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1370 				error = ENOMEM;
1371 		} else
1372 #endif
1373 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1374 				error = ENOMEM;
1375 	} else {
1376 		FOREACH_THREAD_IN_PROC(p, td) {
1377 			fill_kinfo_thread(td, &ki, 1);
1378 #ifdef COMPAT_FREEBSD32
1379 			if ((flags & KERN_PROC_MASK32) != 0) {
1380 				freebsd32_kinfo_proc_out(&ki, &ki32);
1381 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1382 					error = ENOMEM;
1383 			} else
1384 #endif
1385 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1386 					error = ENOMEM;
1387 			if (error != 0)
1388 				break;
1389 		}
1390 	}
1391 	PROC_UNLOCK(p);
1392 	return (error);
1393 }
1394 
1395 static int
1396 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1397 {
1398 	struct sbuf sb;
1399 	struct kinfo_proc ki;
1400 	int error, error2;
1401 
1402 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1403 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1404 	error = kern_proc_out(p, &sb, flags);
1405 	error2 = sbuf_finish(&sb);
1406 	sbuf_delete(&sb);
1407 	if (error != 0)
1408 		return (error);
1409 	else if (error2 != 0)
1410 		return (error2);
1411 	return (0);
1412 }
1413 
1414 static int
1415 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1416 {
1417 	int *name = (int *)arg1;
1418 	u_int namelen = arg2;
1419 	struct proc *p;
1420 	int flags, doingzomb, oid_number;
1421 	int error = 0;
1422 
1423 	oid_number = oidp->oid_number;
1424 	if (oid_number != KERN_PROC_ALL &&
1425 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1426 		flags = KERN_PROC_NOTHREADS;
1427 	else {
1428 		flags = 0;
1429 		oid_number &= ~KERN_PROC_INC_THREAD;
1430 	}
1431 #ifdef COMPAT_FREEBSD32
1432 	if (req->flags & SCTL_MASK32)
1433 		flags |= KERN_PROC_MASK32;
1434 #endif
1435 	if (oid_number == KERN_PROC_PID) {
1436 		if (namelen != 1)
1437 			return (EINVAL);
1438 		error = sysctl_wire_old_buffer(req, 0);
1439 		if (error)
1440 			return (error);
1441 		sx_slock(&proctree_lock);
1442 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1443 		if (error == 0)
1444 			error = sysctl_out_proc(p, req, flags);
1445 		sx_sunlock(&proctree_lock);
1446 		return (error);
1447 	}
1448 
1449 	switch (oid_number) {
1450 	case KERN_PROC_ALL:
1451 		if (namelen != 0)
1452 			return (EINVAL);
1453 		break;
1454 	case KERN_PROC_PROC:
1455 		if (namelen != 0 && namelen != 1)
1456 			return (EINVAL);
1457 		break;
1458 	default:
1459 		if (namelen != 1)
1460 			return (EINVAL);
1461 		break;
1462 	}
1463 
1464 	if (!req->oldptr) {
1465 		/* overestimate by 5 procs */
1466 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1467 		if (error)
1468 			return (error);
1469 	}
1470 	error = sysctl_wire_old_buffer(req, 0);
1471 	if (error != 0)
1472 		return (error);
1473 	sx_slock(&proctree_lock);
1474 	sx_slock(&allproc_lock);
1475 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1476 		if (!doingzomb)
1477 			p = LIST_FIRST(&allproc);
1478 		else
1479 			p = LIST_FIRST(&zombproc);
1480 		for (; p != NULL; p = LIST_NEXT(p, p_list)) {
1481 			/*
1482 			 * Skip embryonic processes.
1483 			 */
1484 			if (p->p_state == PRS_NEW)
1485 				continue;
1486 			PROC_LOCK(p);
1487 			KASSERT(p->p_ucred != NULL,
1488 			    ("process credential is NULL for non-NEW proc"));
1489 			/*
1490 			 * Show a user only appropriate processes.
1491 			 */
1492 			if (p_cansee(curthread, p)) {
1493 				PROC_UNLOCK(p);
1494 				continue;
1495 			}
1496 			/*
1497 			 * TODO - make more efficient (see notes below).
1498 			 * do by session.
1499 			 */
1500 			switch (oid_number) {
1501 
1502 			case KERN_PROC_GID:
1503 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1504 					PROC_UNLOCK(p);
1505 					continue;
1506 				}
1507 				break;
1508 
1509 			case KERN_PROC_PGRP:
1510 				/* could do this by traversing pgrp */
1511 				if (p->p_pgrp == NULL ||
1512 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1513 					PROC_UNLOCK(p);
1514 					continue;
1515 				}
1516 				break;
1517 
1518 			case KERN_PROC_RGID:
1519 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1520 					PROC_UNLOCK(p);
1521 					continue;
1522 				}
1523 				break;
1524 
1525 			case KERN_PROC_SESSION:
1526 				if (p->p_session == NULL ||
1527 				    p->p_session->s_sid != (pid_t)name[0]) {
1528 					PROC_UNLOCK(p);
1529 					continue;
1530 				}
1531 				break;
1532 
1533 			case KERN_PROC_TTY:
1534 				if ((p->p_flag & P_CONTROLT) == 0 ||
1535 				    p->p_session == NULL) {
1536 					PROC_UNLOCK(p);
1537 					continue;
1538 				}
1539 				/* XXX proctree_lock */
1540 				SESS_LOCK(p->p_session);
1541 				if (p->p_session->s_ttyp == NULL ||
1542 				    tty_udev(p->p_session->s_ttyp) !=
1543 				    (dev_t)name[0]) {
1544 					SESS_UNLOCK(p->p_session);
1545 					PROC_UNLOCK(p);
1546 					continue;
1547 				}
1548 				SESS_UNLOCK(p->p_session);
1549 				break;
1550 
1551 			case KERN_PROC_UID:
1552 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1553 					PROC_UNLOCK(p);
1554 					continue;
1555 				}
1556 				break;
1557 
1558 			case KERN_PROC_RUID:
1559 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1560 					PROC_UNLOCK(p);
1561 					continue;
1562 				}
1563 				break;
1564 
1565 			case KERN_PROC_PROC:
1566 				break;
1567 
1568 			default:
1569 				break;
1570 
1571 			}
1572 
1573 			error = sysctl_out_proc(p, req, flags);
1574 			if (error) {
1575 				sx_sunlock(&allproc_lock);
1576 				sx_sunlock(&proctree_lock);
1577 				return (error);
1578 			}
1579 		}
1580 	}
1581 	sx_sunlock(&allproc_lock);
1582 	sx_sunlock(&proctree_lock);
1583 	return (0);
1584 }
1585 
1586 struct pargs *
1587 pargs_alloc(int len)
1588 {
1589 	struct pargs *pa;
1590 
1591 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1592 		M_WAITOK);
1593 	refcount_init(&pa->ar_ref, 1);
1594 	pa->ar_length = len;
1595 	return (pa);
1596 }
1597 
1598 static void
1599 pargs_free(struct pargs *pa)
1600 {
1601 
1602 	free(pa, M_PARGS);
1603 }
1604 
1605 void
1606 pargs_hold(struct pargs *pa)
1607 {
1608 
1609 	if (pa == NULL)
1610 		return;
1611 	refcount_acquire(&pa->ar_ref);
1612 }
1613 
1614 void
1615 pargs_drop(struct pargs *pa)
1616 {
1617 
1618 	if (pa == NULL)
1619 		return;
1620 	if (refcount_release(&pa->ar_ref))
1621 		pargs_free(pa);
1622 }
1623 
1624 static int
1625 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1626     size_t len)
1627 {
1628 	ssize_t n;
1629 
1630 	/*
1631 	 * This may return a short read if the string is shorter than the chunk
1632 	 * and is aligned at the end of the page, and the following page is not
1633 	 * mapped.
1634 	 */
1635 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1636 	if (n <= 0)
1637 		return (ENOMEM);
1638 	return (0);
1639 }
1640 
1641 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1642 
1643 enum proc_vector_type {
1644 	PROC_ARG,
1645 	PROC_ENV,
1646 	PROC_AUX,
1647 };
1648 
1649 #ifdef COMPAT_FREEBSD32
1650 static int
1651 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1652     size_t *vsizep, enum proc_vector_type type)
1653 {
1654 	struct freebsd32_ps_strings pss;
1655 	Elf32_Auxinfo aux;
1656 	vm_offset_t vptr, ptr;
1657 	uint32_t *proc_vector32;
1658 	char **proc_vector;
1659 	size_t vsize, size;
1660 	int i, error;
1661 
1662 	error = 0;
1663 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1664 	    sizeof(pss)) != sizeof(pss))
1665 		return (ENOMEM);
1666 	switch (type) {
1667 	case PROC_ARG:
1668 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1669 		vsize = pss.ps_nargvstr;
1670 		if (vsize > ARG_MAX)
1671 			return (ENOEXEC);
1672 		size = vsize * sizeof(int32_t);
1673 		break;
1674 	case PROC_ENV:
1675 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1676 		vsize = pss.ps_nenvstr;
1677 		if (vsize > ARG_MAX)
1678 			return (ENOEXEC);
1679 		size = vsize * sizeof(int32_t);
1680 		break;
1681 	case PROC_AUX:
1682 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1683 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1684 		if (vptr % 4 != 0)
1685 			return (ENOEXEC);
1686 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1687 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1688 			    sizeof(aux))
1689 				return (ENOMEM);
1690 			if (aux.a_type == AT_NULL)
1691 				break;
1692 			ptr += sizeof(aux);
1693 		}
1694 		if (aux.a_type != AT_NULL)
1695 			return (ENOEXEC);
1696 		vsize = i + 1;
1697 		size = vsize * sizeof(aux);
1698 		break;
1699 	default:
1700 		KASSERT(0, ("Wrong proc vector type: %d", type));
1701 		return (EINVAL);
1702 	}
1703 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1704 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1705 		error = ENOMEM;
1706 		goto done;
1707 	}
1708 	if (type == PROC_AUX) {
1709 		*proc_vectorp = (char **)proc_vector32;
1710 		*vsizep = vsize;
1711 		return (0);
1712 	}
1713 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1714 	for (i = 0; i < (int)vsize; i++)
1715 		proc_vector[i] = PTRIN(proc_vector32[i]);
1716 	*proc_vectorp = proc_vector;
1717 	*vsizep = vsize;
1718 done:
1719 	free(proc_vector32, M_TEMP);
1720 	return (error);
1721 }
1722 #endif
1723 
1724 static int
1725 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1726     size_t *vsizep, enum proc_vector_type type)
1727 {
1728 	struct ps_strings pss;
1729 	Elf_Auxinfo aux;
1730 	vm_offset_t vptr, ptr;
1731 	char **proc_vector;
1732 	size_t vsize, size;
1733 	int i;
1734 
1735 #ifdef COMPAT_FREEBSD32
1736 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1737 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1738 #endif
1739 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1740 	    sizeof(pss)) != sizeof(pss))
1741 		return (ENOMEM);
1742 	switch (type) {
1743 	case PROC_ARG:
1744 		vptr = (vm_offset_t)pss.ps_argvstr;
1745 		vsize = pss.ps_nargvstr;
1746 		if (vsize > ARG_MAX)
1747 			return (ENOEXEC);
1748 		size = vsize * sizeof(char *);
1749 		break;
1750 	case PROC_ENV:
1751 		vptr = (vm_offset_t)pss.ps_envstr;
1752 		vsize = pss.ps_nenvstr;
1753 		if (vsize > ARG_MAX)
1754 			return (ENOEXEC);
1755 		size = vsize * sizeof(char *);
1756 		break;
1757 	case PROC_AUX:
1758 		/*
1759 		 * The aux array is just above env array on the stack. Check
1760 		 * that the address is naturally aligned.
1761 		 */
1762 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1763 		    * sizeof(char *);
1764 #if __ELF_WORD_SIZE == 64
1765 		if (vptr % sizeof(uint64_t) != 0)
1766 #else
1767 		if (vptr % sizeof(uint32_t) != 0)
1768 #endif
1769 			return (ENOEXEC);
1770 		/*
1771 		 * We count the array size reading the aux vectors from the
1772 		 * stack until AT_NULL vector is returned.  So (to keep the code
1773 		 * simple) we read the process stack twice: the first time here
1774 		 * to find the size and the second time when copying the vectors
1775 		 * to the allocated proc_vector.
1776 		 */
1777 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1778 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1779 			    sizeof(aux))
1780 				return (ENOMEM);
1781 			if (aux.a_type == AT_NULL)
1782 				break;
1783 			ptr += sizeof(aux);
1784 		}
1785 		/*
1786 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1787 		 * not reached AT_NULL, it is most likely we are reading wrong
1788 		 * data: either the process doesn't have auxv array or data has
1789 		 * been modified. Return the error in this case.
1790 		 */
1791 		if (aux.a_type != AT_NULL)
1792 			return (ENOEXEC);
1793 		vsize = i + 1;
1794 		size = vsize * sizeof(aux);
1795 		break;
1796 	default:
1797 		KASSERT(0, ("Wrong proc vector type: %d", type));
1798 		return (EINVAL); /* In case we are built without INVARIANTS. */
1799 	}
1800 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1801 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
1802 		free(proc_vector, M_TEMP);
1803 		return (ENOMEM);
1804 	}
1805 	*proc_vectorp = proc_vector;
1806 	*vsizep = vsize;
1807 
1808 	return (0);
1809 }
1810 
1811 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1812 
1813 static int
1814 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1815     enum proc_vector_type type)
1816 {
1817 	size_t done, len, nchr, vsize;
1818 	int error, i;
1819 	char **proc_vector, *sptr;
1820 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1821 
1822 	PROC_ASSERT_HELD(p);
1823 
1824 	/*
1825 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1826 	 */
1827 	nchr = 2 * (PATH_MAX + ARG_MAX);
1828 
1829 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1830 	if (error != 0)
1831 		return (error);
1832 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1833 		/*
1834 		 * The program may have scribbled into its argv array, e.g. to
1835 		 * remove some arguments.  If that has happened, break out
1836 		 * before trying to read from NULL.
1837 		 */
1838 		if (proc_vector[i] == NULL)
1839 			break;
1840 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1841 			error = proc_read_string(td, p, sptr, pss_string,
1842 			    sizeof(pss_string));
1843 			if (error != 0)
1844 				goto done;
1845 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1846 			if (done + len >= nchr)
1847 				len = nchr - done - 1;
1848 			sbuf_bcat(sb, pss_string, len);
1849 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1850 				break;
1851 			done += GET_PS_STRINGS_CHUNK_SZ;
1852 		}
1853 		sbuf_bcat(sb, "", 1);
1854 		done += len + 1;
1855 	}
1856 done:
1857 	free(proc_vector, M_TEMP);
1858 	return (error);
1859 }
1860 
1861 int
1862 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1863 {
1864 
1865 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1866 }
1867 
1868 int
1869 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1870 {
1871 
1872 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1873 }
1874 
1875 int
1876 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1877 {
1878 	size_t vsize, size;
1879 	char **auxv;
1880 	int error;
1881 
1882 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1883 	if (error == 0) {
1884 #ifdef COMPAT_FREEBSD32
1885 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1886 			size = vsize * sizeof(Elf32_Auxinfo);
1887 		else
1888 #endif
1889 			size = vsize * sizeof(Elf_Auxinfo);
1890 		if (sbuf_bcat(sb, auxv, size) != 0)
1891 			error = ENOMEM;
1892 		free(auxv, M_TEMP);
1893 	}
1894 	return (error);
1895 }
1896 
1897 /*
1898  * This sysctl allows a process to retrieve the argument list or process
1899  * title for another process without groping around in the address space
1900  * of the other process.  It also allow a process to set its own "process
1901  * title to a string of its own choice.
1902  */
1903 static int
1904 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1905 {
1906 	int *name = (int *)arg1;
1907 	u_int namelen = arg2;
1908 	struct pargs *newpa, *pa;
1909 	struct proc *p;
1910 	struct sbuf sb;
1911 	int flags, error = 0, error2;
1912 	pid_t pid;
1913 
1914 	if (namelen != 1)
1915 		return (EINVAL);
1916 
1917 	pid = (pid_t)name[0];
1918 	/*
1919 	 * If the query is for this process and it is single-threaded, there
1920 	 * is nobody to modify pargs, thus we can just read.
1921 	 */
1922 	p = curproc;
1923 	if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
1924 	    (pa = p->p_args) != NULL)
1925 		return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
1926 
1927 	flags = PGET_CANSEE;
1928 	if (req->newptr != NULL)
1929 		flags |= PGET_ISCURRENT;
1930 	error = pget(pid, flags, &p);
1931 	if (error)
1932 		return (error);
1933 
1934 	pa = p->p_args;
1935 	if (pa != NULL) {
1936 		pargs_hold(pa);
1937 		PROC_UNLOCK(p);
1938 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1939 		pargs_drop(pa);
1940 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1941 		_PHOLD(p);
1942 		PROC_UNLOCK(p);
1943 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1944 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1945 		error = proc_getargv(curthread, p, &sb);
1946 		error2 = sbuf_finish(&sb);
1947 		PRELE(p);
1948 		sbuf_delete(&sb);
1949 		if (error == 0 && error2 != 0)
1950 			error = error2;
1951 	} else {
1952 		PROC_UNLOCK(p);
1953 	}
1954 	if (error != 0 || req->newptr == NULL)
1955 		return (error);
1956 
1957 	if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
1958 		return (ENOMEM);
1959 	newpa = pargs_alloc(req->newlen);
1960 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1961 	if (error != 0) {
1962 		pargs_free(newpa);
1963 		return (error);
1964 	}
1965 	PROC_LOCK(p);
1966 	pa = p->p_args;
1967 	p->p_args = newpa;
1968 	PROC_UNLOCK(p);
1969 	pargs_drop(pa);
1970 	return (0);
1971 }
1972 
1973 /*
1974  * This sysctl allows a process to retrieve environment of another process.
1975  */
1976 static int
1977 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
1978 {
1979 	int *name = (int *)arg1;
1980 	u_int namelen = arg2;
1981 	struct proc *p;
1982 	struct sbuf sb;
1983 	int error, error2;
1984 
1985 	if (namelen != 1)
1986 		return (EINVAL);
1987 
1988 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1989 	if (error != 0)
1990 		return (error);
1991 	if ((p->p_flag & P_SYSTEM) != 0) {
1992 		PRELE(p);
1993 		return (0);
1994 	}
1995 
1996 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1997 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1998 	error = proc_getenvv(curthread, p, &sb);
1999 	error2 = sbuf_finish(&sb);
2000 	PRELE(p);
2001 	sbuf_delete(&sb);
2002 	return (error != 0 ? error : error2);
2003 }
2004 
2005 /*
2006  * This sysctl allows a process to retrieve ELF auxiliary vector of
2007  * another process.
2008  */
2009 static int
2010 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2011 {
2012 	int *name = (int *)arg1;
2013 	u_int namelen = arg2;
2014 	struct proc *p;
2015 	struct sbuf sb;
2016 	int error, error2;
2017 
2018 	if (namelen != 1)
2019 		return (EINVAL);
2020 
2021 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2022 	if (error != 0)
2023 		return (error);
2024 	if ((p->p_flag & P_SYSTEM) != 0) {
2025 		PRELE(p);
2026 		return (0);
2027 	}
2028 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2029 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2030 	error = proc_getauxv(curthread, p, &sb);
2031 	error2 = sbuf_finish(&sb);
2032 	PRELE(p);
2033 	sbuf_delete(&sb);
2034 	return (error != 0 ? error : error2);
2035 }
2036 
2037 /*
2038  * This sysctl allows a process to retrieve the path of the executable for
2039  * itself or another process.
2040  */
2041 static int
2042 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2043 {
2044 	pid_t *pidp = (pid_t *)arg1;
2045 	unsigned int arglen = arg2;
2046 	struct proc *p;
2047 	struct vnode *vp;
2048 	char *retbuf, *freebuf;
2049 	int error;
2050 
2051 	if (arglen != 1)
2052 		return (EINVAL);
2053 	if (*pidp == -1) {	/* -1 means this process */
2054 		p = req->td->td_proc;
2055 	} else {
2056 		error = pget(*pidp, PGET_CANSEE, &p);
2057 		if (error != 0)
2058 			return (error);
2059 	}
2060 
2061 	vp = p->p_textvp;
2062 	if (vp == NULL) {
2063 		if (*pidp != -1)
2064 			PROC_UNLOCK(p);
2065 		return (0);
2066 	}
2067 	vref(vp);
2068 	if (*pidp != -1)
2069 		PROC_UNLOCK(p);
2070 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
2071 	vrele(vp);
2072 	if (error)
2073 		return (error);
2074 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2075 	free(freebuf, M_TEMP);
2076 	return (error);
2077 }
2078 
2079 static int
2080 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2081 {
2082 	struct proc *p;
2083 	char *sv_name;
2084 	int *name;
2085 	int namelen;
2086 	int error;
2087 
2088 	namelen = arg2;
2089 	if (namelen != 1)
2090 		return (EINVAL);
2091 
2092 	name = (int *)arg1;
2093 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2094 	if (error != 0)
2095 		return (error);
2096 	sv_name = p->p_sysent->sv_name;
2097 	PROC_UNLOCK(p);
2098 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2099 }
2100 
2101 #ifdef KINFO_OVMENTRY_SIZE
2102 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2103 #endif
2104 
2105 #ifdef COMPAT_FREEBSD7
2106 static int
2107 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2108 {
2109 	vm_map_entry_t entry, tmp_entry;
2110 	unsigned int last_timestamp;
2111 	char *fullpath, *freepath;
2112 	struct kinfo_ovmentry *kve;
2113 	struct vattr va;
2114 	struct ucred *cred;
2115 	int error, *name;
2116 	struct vnode *vp;
2117 	struct proc *p;
2118 	vm_map_t map;
2119 	struct vmspace *vm;
2120 
2121 	name = (int *)arg1;
2122 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2123 	if (error != 0)
2124 		return (error);
2125 	vm = vmspace_acquire_ref(p);
2126 	if (vm == NULL) {
2127 		PRELE(p);
2128 		return (ESRCH);
2129 	}
2130 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2131 
2132 	map = &vm->vm_map;
2133 	vm_map_lock_read(map);
2134 	for (entry = map->header.next; entry != &map->header;
2135 	    entry = entry->next) {
2136 		vm_object_t obj, tobj, lobj;
2137 		vm_offset_t addr;
2138 
2139 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2140 			continue;
2141 
2142 		bzero(kve, sizeof(*kve));
2143 		kve->kve_structsize = sizeof(*kve);
2144 
2145 		kve->kve_private_resident = 0;
2146 		obj = entry->object.vm_object;
2147 		if (obj != NULL) {
2148 			VM_OBJECT_RLOCK(obj);
2149 			if (obj->shadow_count == 1)
2150 				kve->kve_private_resident =
2151 				    obj->resident_page_count;
2152 		}
2153 		kve->kve_resident = 0;
2154 		addr = entry->start;
2155 		while (addr < entry->end) {
2156 			if (pmap_extract(map->pmap, addr))
2157 				kve->kve_resident++;
2158 			addr += PAGE_SIZE;
2159 		}
2160 
2161 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2162 			if (tobj != obj) {
2163 				VM_OBJECT_RLOCK(tobj);
2164 				kve->kve_offset += tobj->backing_object_offset;
2165 			}
2166 			if (lobj != obj)
2167 				VM_OBJECT_RUNLOCK(lobj);
2168 			lobj = tobj;
2169 		}
2170 
2171 		kve->kve_start = (void*)entry->start;
2172 		kve->kve_end = (void*)entry->end;
2173 		kve->kve_offset += (off_t)entry->offset;
2174 
2175 		if (entry->protection & VM_PROT_READ)
2176 			kve->kve_protection |= KVME_PROT_READ;
2177 		if (entry->protection & VM_PROT_WRITE)
2178 			kve->kve_protection |= KVME_PROT_WRITE;
2179 		if (entry->protection & VM_PROT_EXECUTE)
2180 			kve->kve_protection |= KVME_PROT_EXEC;
2181 
2182 		if (entry->eflags & MAP_ENTRY_COW)
2183 			kve->kve_flags |= KVME_FLAG_COW;
2184 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2185 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2186 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2187 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2188 
2189 		last_timestamp = map->timestamp;
2190 		vm_map_unlock_read(map);
2191 
2192 		kve->kve_fileid = 0;
2193 		kve->kve_fsid = 0;
2194 		freepath = NULL;
2195 		fullpath = "";
2196 		if (lobj) {
2197 			vp = NULL;
2198 			switch (lobj->type) {
2199 			case OBJT_DEFAULT:
2200 				kve->kve_type = KVME_TYPE_DEFAULT;
2201 				break;
2202 			case OBJT_VNODE:
2203 				kve->kve_type = KVME_TYPE_VNODE;
2204 				vp = lobj->handle;
2205 				vref(vp);
2206 				break;
2207 			case OBJT_SWAP:
2208 				if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2209 					kve->kve_type = KVME_TYPE_VNODE;
2210 					if ((lobj->flags & OBJ_TMPFS) != 0) {
2211 						vp = lobj->un_pager.swp.swp_tmpfs;
2212 						vref(vp);
2213 					}
2214 				} else {
2215 					kve->kve_type = KVME_TYPE_SWAP;
2216 				}
2217 				break;
2218 			case OBJT_DEVICE:
2219 				kve->kve_type = KVME_TYPE_DEVICE;
2220 				break;
2221 			case OBJT_PHYS:
2222 				kve->kve_type = KVME_TYPE_PHYS;
2223 				break;
2224 			case OBJT_DEAD:
2225 				kve->kve_type = KVME_TYPE_DEAD;
2226 				break;
2227 			case OBJT_SG:
2228 				kve->kve_type = KVME_TYPE_SG;
2229 				break;
2230 			default:
2231 				kve->kve_type = KVME_TYPE_UNKNOWN;
2232 				break;
2233 			}
2234 			if (lobj != obj)
2235 				VM_OBJECT_RUNLOCK(lobj);
2236 
2237 			kve->kve_ref_count = obj->ref_count;
2238 			kve->kve_shadow_count = obj->shadow_count;
2239 			VM_OBJECT_RUNLOCK(obj);
2240 			if (vp != NULL) {
2241 				vn_fullpath(curthread, vp, &fullpath,
2242 				    &freepath);
2243 				cred = curthread->td_ucred;
2244 				vn_lock(vp, LK_SHARED | LK_RETRY);
2245 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2246 					kve->kve_fileid = va.va_fileid;
2247 					/* truncate */
2248 					kve->kve_fsid = va.va_fsid;
2249 				}
2250 				vput(vp);
2251 			}
2252 		} else {
2253 			kve->kve_type = KVME_TYPE_NONE;
2254 			kve->kve_ref_count = 0;
2255 			kve->kve_shadow_count = 0;
2256 		}
2257 
2258 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2259 		if (freepath != NULL)
2260 			free(freepath, M_TEMP);
2261 
2262 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2263 		vm_map_lock_read(map);
2264 		if (error)
2265 			break;
2266 		if (last_timestamp != map->timestamp) {
2267 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2268 			entry = tmp_entry;
2269 		}
2270 	}
2271 	vm_map_unlock_read(map);
2272 	vmspace_free(vm);
2273 	PRELE(p);
2274 	free(kve, M_TEMP);
2275 	return (error);
2276 }
2277 #endif	/* COMPAT_FREEBSD7 */
2278 
2279 #ifdef KINFO_VMENTRY_SIZE
2280 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2281 #endif
2282 
2283 void
2284 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2285     int *resident_count, bool *super)
2286 {
2287 	vm_object_t obj, tobj;
2288 	vm_page_t m, m_adv;
2289 	vm_offset_t addr;
2290 	vm_paddr_t locked_pa;
2291 	vm_pindex_t pi, pi_adv, pindex;
2292 
2293 	*super = false;
2294 	*resident_count = 0;
2295 	if (vmmap_skip_res_cnt)
2296 		return;
2297 
2298 	locked_pa = 0;
2299 	obj = entry->object.vm_object;
2300 	addr = entry->start;
2301 	m_adv = NULL;
2302 	pi = OFF_TO_IDX(entry->offset);
2303 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2304 		if (m_adv != NULL) {
2305 			m = m_adv;
2306 		} else {
2307 			pi_adv = atop(entry->end - addr);
2308 			pindex = pi;
2309 			for (tobj = obj;; tobj = tobj->backing_object) {
2310 				m = vm_page_find_least(tobj, pindex);
2311 				if (m != NULL) {
2312 					if (m->pindex == pindex)
2313 						break;
2314 					if (pi_adv > m->pindex - pindex) {
2315 						pi_adv = m->pindex - pindex;
2316 						m_adv = m;
2317 					}
2318 				}
2319 				if (tobj->backing_object == NULL)
2320 					goto next;
2321 				pindex += OFF_TO_IDX(tobj->
2322 				    backing_object_offset);
2323 			}
2324 		}
2325 		m_adv = NULL;
2326 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2327 		    (addr & (pagesizes[1] - 1)) == 0 &&
2328 		    (pmap_mincore(map->pmap, addr, &locked_pa) &
2329 		    MINCORE_SUPER) != 0) {
2330 			*super = true;
2331 			pi_adv = atop(pagesizes[1]);
2332 		} else {
2333 			/*
2334 			 * We do not test the found page on validity.
2335 			 * Either the page is busy and being paged in,
2336 			 * or it was invalidated.  The first case
2337 			 * should be counted as resident, the second
2338 			 * is not so clear; we do account both.
2339 			 */
2340 			pi_adv = 1;
2341 		}
2342 		*resident_count += pi_adv;
2343 next:;
2344 	}
2345 	PA_UNLOCK_COND(locked_pa);
2346 }
2347 
2348 /*
2349  * Must be called with the process locked and will return unlocked.
2350  */
2351 int
2352 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2353 {
2354 	vm_map_entry_t entry, tmp_entry;
2355 	struct vattr va;
2356 	vm_map_t map;
2357 	vm_object_t obj, tobj, lobj;
2358 	char *fullpath, *freepath;
2359 	struct kinfo_vmentry *kve;
2360 	struct ucred *cred;
2361 	struct vnode *vp;
2362 	struct vmspace *vm;
2363 	vm_offset_t addr;
2364 	unsigned int last_timestamp;
2365 	int error;
2366 	bool super;
2367 
2368 	PROC_LOCK_ASSERT(p, MA_OWNED);
2369 
2370 	_PHOLD(p);
2371 	PROC_UNLOCK(p);
2372 	vm = vmspace_acquire_ref(p);
2373 	if (vm == NULL) {
2374 		PRELE(p);
2375 		return (ESRCH);
2376 	}
2377 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2378 
2379 	error = 0;
2380 	map = &vm->vm_map;
2381 	vm_map_lock_read(map);
2382 	for (entry = map->header.next; entry != &map->header;
2383 	    entry = entry->next) {
2384 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2385 			continue;
2386 
2387 		addr = entry->end;
2388 		bzero(kve, sizeof(*kve));
2389 		obj = entry->object.vm_object;
2390 		if (obj != NULL) {
2391 			for (tobj = obj; tobj != NULL;
2392 			    tobj = tobj->backing_object) {
2393 				VM_OBJECT_RLOCK(tobj);
2394 				kve->kve_offset += tobj->backing_object_offset;
2395 				lobj = tobj;
2396 			}
2397 			if (obj->backing_object == NULL)
2398 				kve->kve_private_resident =
2399 				    obj->resident_page_count;
2400 			kern_proc_vmmap_resident(map, entry,
2401 			    &kve->kve_resident, &super);
2402 			if (super)
2403 				kve->kve_flags |= KVME_FLAG_SUPER;
2404 			for (tobj = obj; tobj != NULL;
2405 			    tobj = tobj->backing_object) {
2406 				if (tobj != obj && tobj != lobj)
2407 					VM_OBJECT_RUNLOCK(tobj);
2408 			}
2409 		} else {
2410 			lobj = NULL;
2411 		}
2412 
2413 		kve->kve_start = entry->start;
2414 		kve->kve_end = entry->end;
2415 		kve->kve_offset += entry->offset;
2416 
2417 		if (entry->protection & VM_PROT_READ)
2418 			kve->kve_protection |= KVME_PROT_READ;
2419 		if (entry->protection & VM_PROT_WRITE)
2420 			kve->kve_protection |= KVME_PROT_WRITE;
2421 		if (entry->protection & VM_PROT_EXECUTE)
2422 			kve->kve_protection |= KVME_PROT_EXEC;
2423 
2424 		if (entry->eflags & MAP_ENTRY_COW)
2425 			kve->kve_flags |= KVME_FLAG_COW;
2426 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2427 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2428 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2429 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2430 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2431 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2432 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2433 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2434 
2435 		last_timestamp = map->timestamp;
2436 		vm_map_unlock_read(map);
2437 
2438 		freepath = NULL;
2439 		fullpath = "";
2440 		if (lobj != NULL) {
2441 			vp = NULL;
2442 			switch (lobj->type) {
2443 			case OBJT_DEFAULT:
2444 				kve->kve_type = KVME_TYPE_DEFAULT;
2445 				break;
2446 			case OBJT_VNODE:
2447 				kve->kve_type = KVME_TYPE_VNODE;
2448 				vp = lobj->handle;
2449 				vref(vp);
2450 				break;
2451 			case OBJT_SWAP:
2452 				if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2453 					kve->kve_type = KVME_TYPE_VNODE;
2454 					if ((lobj->flags & OBJ_TMPFS) != 0) {
2455 						vp = lobj->un_pager.swp.swp_tmpfs;
2456 						vref(vp);
2457 					}
2458 				} else {
2459 					kve->kve_type = KVME_TYPE_SWAP;
2460 				}
2461 				break;
2462 			case OBJT_DEVICE:
2463 				kve->kve_type = KVME_TYPE_DEVICE;
2464 				break;
2465 			case OBJT_PHYS:
2466 				kve->kve_type = KVME_TYPE_PHYS;
2467 				break;
2468 			case OBJT_DEAD:
2469 				kve->kve_type = KVME_TYPE_DEAD;
2470 				break;
2471 			case OBJT_SG:
2472 				kve->kve_type = KVME_TYPE_SG;
2473 				break;
2474 			case OBJT_MGTDEVICE:
2475 				kve->kve_type = KVME_TYPE_MGTDEVICE;
2476 				break;
2477 			default:
2478 				kve->kve_type = KVME_TYPE_UNKNOWN;
2479 				break;
2480 			}
2481 			if (lobj != obj)
2482 				VM_OBJECT_RUNLOCK(lobj);
2483 
2484 			kve->kve_ref_count = obj->ref_count;
2485 			kve->kve_shadow_count = obj->shadow_count;
2486 			VM_OBJECT_RUNLOCK(obj);
2487 			if (vp != NULL) {
2488 				vn_fullpath(curthread, vp, &fullpath,
2489 				    &freepath);
2490 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2491 				cred = curthread->td_ucred;
2492 				vn_lock(vp, LK_SHARED | LK_RETRY);
2493 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2494 					kve->kve_vn_fileid = va.va_fileid;
2495 					kve->kve_vn_fsid = va.va_fsid;
2496 					kve->kve_vn_fsid_freebsd11 =
2497 					    kve->kve_vn_fsid; /* truncate */
2498 					kve->kve_vn_mode =
2499 					    MAKEIMODE(va.va_type, va.va_mode);
2500 					kve->kve_vn_size = va.va_size;
2501 					kve->kve_vn_rdev = va.va_rdev;
2502 					kve->kve_vn_rdev_freebsd11 =
2503 					    kve->kve_vn_rdev; /* truncate */
2504 					kve->kve_status = KF_ATTR_VALID;
2505 				}
2506 				vput(vp);
2507 			}
2508 		} else {
2509 			kve->kve_type = KVME_TYPE_NONE;
2510 			kve->kve_ref_count = 0;
2511 			kve->kve_shadow_count = 0;
2512 		}
2513 
2514 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2515 		if (freepath != NULL)
2516 			free(freepath, M_TEMP);
2517 
2518 		/* Pack record size down */
2519 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2520 			kve->kve_structsize =
2521 			    offsetof(struct kinfo_vmentry, kve_path) +
2522 			    strlen(kve->kve_path) + 1;
2523 		else
2524 			kve->kve_structsize = sizeof(*kve);
2525 		kve->kve_structsize = roundup(kve->kve_structsize,
2526 		    sizeof(uint64_t));
2527 
2528 		/* Halt filling and truncate rather than exceeding maxlen */
2529 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2530 			error = 0;
2531 			vm_map_lock_read(map);
2532 			break;
2533 		} else if (maxlen != -1)
2534 			maxlen -= kve->kve_structsize;
2535 
2536 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2537 			error = ENOMEM;
2538 		vm_map_lock_read(map);
2539 		if (error != 0)
2540 			break;
2541 		if (last_timestamp != map->timestamp) {
2542 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2543 			entry = tmp_entry;
2544 		}
2545 	}
2546 	vm_map_unlock_read(map);
2547 	vmspace_free(vm);
2548 	PRELE(p);
2549 	free(kve, M_TEMP);
2550 	return (error);
2551 }
2552 
2553 static int
2554 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2555 {
2556 	struct proc *p;
2557 	struct sbuf sb;
2558 	int error, error2, *name;
2559 
2560 	name = (int *)arg1;
2561 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2562 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2563 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2564 	if (error != 0) {
2565 		sbuf_delete(&sb);
2566 		return (error);
2567 	}
2568 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2569 	error2 = sbuf_finish(&sb);
2570 	sbuf_delete(&sb);
2571 	return (error != 0 ? error : error2);
2572 }
2573 
2574 #if defined(STACK) || defined(DDB)
2575 static int
2576 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2577 {
2578 	struct kinfo_kstack *kkstp;
2579 	int error, i, *name, numthreads;
2580 	lwpid_t *lwpidarray;
2581 	struct thread *td;
2582 	struct stack *st;
2583 	struct sbuf sb;
2584 	struct proc *p;
2585 
2586 	name = (int *)arg1;
2587 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2588 	if (error != 0)
2589 		return (error);
2590 
2591 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2592 	st = stack_create(M_WAITOK);
2593 
2594 	lwpidarray = NULL;
2595 	PROC_LOCK(p);
2596 	do {
2597 		if (lwpidarray != NULL) {
2598 			free(lwpidarray, M_TEMP);
2599 			lwpidarray = NULL;
2600 		}
2601 		numthreads = p->p_numthreads;
2602 		PROC_UNLOCK(p);
2603 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2604 		    M_WAITOK | M_ZERO);
2605 		PROC_LOCK(p);
2606 	} while (numthreads < p->p_numthreads);
2607 
2608 	/*
2609 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2610 	 * of changes could have taken place.  Should we check to see if the
2611 	 * vmspace has been replaced, or the like, in order to prevent
2612 	 * giving a snapshot that spans, say, execve(2), with some threads
2613 	 * before and some after?  Among other things, the credentials could
2614 	 * have changed, in which case the right to extract debug info might
2615 	 * no longer be assured.
2616 	 */
2617 	i = 0;
2618 	FOREACH_THREAD_IN_PROC(p, td) {
2619 		KASSERT(i < numthreads,
2620 		    ("sysctl_kern_proc_kstack: numthreads"));
2621 		lwpidarray[i] = td->td_tid;
2622 		i++;
2623 	}
2624 	numthreads = i;
2625 	for (i = 0; i < numthreads; i++) {
2626 		td = thread_find(p, lwpidarray[i]);
2627 		if (td == NULL) {
2628 			continue;
2629 		}
2630 		bzero(kkstp, sizeof(*kkstp));
2631 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2632 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2633 		thread_lock(td);
2634 		kkstp->kkst_tid = td->td_tid;
2635 		if (TD_IS_SWAPPED(td)) {
2636 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2637 		} else if (TD_IS_RUNNING(td)) {
2638 			if (stack_save_td_running(st, td) == 0)
2639 				kkstp->kkst_state = KKST_STATE_STACKOK;
2640 			else
2641 				kkstp->kkst_state = KKST_STATE_RUNNING;
2642 		} else {
2643 			kkstp->kkst_state = KKST_STATE_STACKOK;
2644 			stack_save_td(st, td);
2645 		}
2646 		thread_unlock(td);
2647 		PROC_UNLOCK(p);
2648 		stack_sbuf_print(&sb, st);
2649 		sbuf_finish(&sb);
2650 		sbuf_delete(&sb);
2651 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2652 		PROC_LOCK(p);
2653 		if (error)
2654 			break;
2655 	}
2656 	_PRELE(p);
2657 	PROC_UNLOCK(p);
2658 	if (lwpidarray != NULL)
2659 		free(lwpidarray, M_TEMP);
2660 	stack_destroy(st);
2661 	free(kkstp, M_TEMP);
2662 	return (error);
2663 }
2664 #endif
2665 
2666 /*
2667  * This sysctl allows a process to retrieve the full list of groups from
2668  * itself or another process.
2669  */
2670 static int
2671 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2672 {
2673 	pid_t *pidp = (pid_t *)arg1;
2674 	unsigned int arglen = arg2;
2675 	struct proc *p;
2676 	struct ucred *cred;
2677 	int error;
2678 
2679 	if (arglen != 1)
2680 		return (EINVAL);
2681 	if (*pidp == -1) {	/* -1 means this process */
2682 		p = req->td->td_proc;
2683 		PROC_LOCK(p);
2684 	} else {
2685 		error = pget(*pidp, PGET_CANSEE, &p);
2686 		if (error != 0)
2687 			return (error);
2688 	}
2689 
2690 	cred = crhold(p->p_ucred);
2691 	PROC_UNLOCK(p);
2692 
2693 	error = SYSCTL_OUT(req, cred->cr_groups,
2694 	    cred->cr_ngroups * sizeof(gid_t));
2695 	crfree(cred);
2696 	return (error);
2697 }
2698 
2699 /*
2700  * This sysctl allows a process to retrieve or/and set the resource limit for
2701  * another process.
2702  */
2703 static int
2704 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2705 {
2706 	int *name = (int *)arg1;
2707 	u_int namelen = arg2;
2708 	struct rlimit rlim;
2709 	struct proc *p;
2710 	u_int which;
2711 	int flags, error;
2712 
2713 	if (namelen != 2)
2714 		return (EINVAL);
2715 
2716 	which = (u_int)name[1];
2717 	if (which >= RLIM_NLIMITS)
2718 		return (EINVAL);
2719 
2720 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2721 		return (EINVAL);
2722 
2723 	flags = PGET_HOLD | PGET_NOTWEXIT;
2724 	if (req->newptr != NULL)
2725 		flags |= PGET_CANDEBUG;
2726 	else
2727 		flags |= PGET_CANSEE;
2728 	error = pget((pid_t)name[0], flags, &p);
2729 	if (error != 0)
2730 		return (error);
2731 
2732 	/*
2733 	 * Retrieve limit.
2734 	 */
2735 	if (req->oldptr != NULL) {
2736 		PROC_LOCK(p);
2737 		lim_rlimit_proc(p, which, &rlim);
2738 		PROC_UNLOCK(p);
2739 	}
2740 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2741 	if (error != 0)
2742 		goto errout;
2743 
2744 	/*
2745 	 * Set limit.
2746 	 */
2747 	if (req->newptr != NULL) {
2748 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2749 		if (error == 0)
2750 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2751 	}
2752 
2753 errout:
2754 	PRELE(p);
2755 	return (error);
2756 }
2757 
2758 /*
2759  * This sysctl allows a process to retrieve ps_strings structure location of
2760  * another process.
2761  */
2762 static int
2763 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2764 {
2765 	int *name = (int *)arg1;
2766 	u_int namelen = arg2;
2767 	struct proc *p;
2768 	vm_offset_t ps_strings;
2769 	int error;
2770 #ifdef COMPAT_FREEBSD32
2771 	uint32_t ps_strings32;
2772 #endif
2773 
2774 	if (namelen != 1)
2775 		return (EINVAL);
2776 
2777 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2778 	if (error != 0)
2779 		return (error);
2780 #ifdef COMPAT_FREEBSD32
2781 	if ((req->flags & SCTL_MASK32) != 0) {
2782 		/*
2783 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2784 		 * process.
2785 		 */
2786 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2787 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2788 		PROC_UNLOCK(p);
2789 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2790 		return (error);
2791 	}
2792 #endif
2793 	ps_strings = p->p_sysent->sv_psstrings;
2794 	PROC_UNLOCK(p);
2795 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2796 	return (error);
2797 }
2798 
2799 /*
2800  * This sysctl allows a process to retrieve umask of another process.
2801  */
2802 static int
2803 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2804 {
2805 	int *name = (int *)arg1;
2806 	u_int namelen = arg2;
2807 	struct proc *p;
2808 	int error;
2809 	u_short fd_cmask;
2810 	pid_t pid;
2811 
2812 	if (namelen != 1)
2813 		return (EINVAL);
2814 
2815 	pid = (pid_t)name[0];
2816 	p = curproc;
2817 	if (pid == p->p_pid || pid == 0) {
2818 		fd_cmask = p->p_fd->fd_cmask;
2819 		goto out;
2820 	}
2821 
2822 	error = pget(pid, PGET_WANTREAD, &p);
2823 	if (error != 0)
2824 		return (error);
2825 
2826 	fd_cmask = p->p_fd->fd_cmask;
2827 	PRELE(p);
2828 out:
2829 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2830 	return (error);
2831 }
2832 
2833 /*
2834  * This sysctl allows a process to set and retrieve binary osreldate of
2835  * another process.
2836  */
2837 static int
2838 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2839 {
2840 	int *name = (int *)arg1;
2841 	u_int namelen = arg2;
2842 	struct proc *p;
2843 	int flags, error, osrel;
2844 
2845 	if (namelen != 1)
2846 		return (EINVAL);
2847 
2848 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2849 		return (EINVAL);
2850 
2851 	flags = PGET_HOLD | PGET_NOTWEXIT;
2852 	if (req->newptr != NULL)
2853 		flags |= PGET_CANDEBUG;
2854 	else
2855 		flags |= PGET_CANSEE;
2856 	error = pget((pid_t)name[0], flags, &p);
2857 	if (error != 0)
2858 		return (error);
2859 
2860 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2861 	if (error != 0)
2862 		goto errout;
2863 
2864 	if (req->newptr != NULL) {
2865 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2866 		if (error != 0)
2867 			goto errout;
2868 		if (osrel < 0) {
2869 			error = EINVAL;
2870 			goto errout;
2871 		}
2872 		p->p_osrel = osrel;
2873 	}
2874 errout:
2875 	PRELE(p);
2876 	return (error);
2877 }
2878 
2879 static int
2880 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2881 {
2882 	int *name = (int *)arg1;
2883 	u_int namelen = arg2;
2884 	struct proc *p;
2885 	struct kinfo_sigtramp kst;
2886 	const struct sysentvec *sv;
2887 	int error;
2888 #ifdef COMPAT_FREEBSD32
2889 	struct kinfo_sigtramp32 kst32;
2890 #endif
2891 
2892 	if (namelen != 1)
2893 		return (EINVAL);
2894 
2895 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2896 	if (error != 0)
2897 		return (error);
2898 	sv = p->p_sysent;
2899 #ifdef COMPAT_FREEBSD32
2900 	if ((req->flags & SCTL_MASK32) != 0) {
2901 		bzero(&kst32, sizeof(kst32));
2902 		if (SV_PROC_FLAG(p, SV_ILP32)) {
2903 			if (sv->sv_sigcode_base != 0) {
2904 				kst32.ksigtramp_start = sv->sv_sigcode_base;
2905 				kst32.ksigtramp_end = sv->sv_sigcode_base +
2906 				    *sv->sv_szsigcode;
2907 			} else {
2908 				kst32.ksigtramp_start = sv->sv_psstrings -
2909 				    *sv->sv_szsigcode;
2910 				kst32.ksigtramp_end = sv->sv_psstrings;
2911 			}
2912 		}
2913 		PROC_UNLOCK(p);
2914 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2915 		return (error);
2916 	}
2917 #endif
2918 	bzero(&kst, sizeof(kst));
2919 	if (sv->sv_sigcode_base != 0) {
2920 		kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2921 		kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2922 		    *sv->sv_szsigcode;
2923 	} else {
2924 		kst.ksigtramp_start = (char *)sv->sv_psstrings -
2925 		    *sv->sv_szsigcode;
2926 		kst.ksigtramp_end = (char *)sv->sv_psstrings;
2927 	}
2928 	PROC_UNLOCK(p);
2929 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
2930 	return (error);
2931 }
2932 
2933 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2934 
2935 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2936 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2937 	"Return entire process table");
2938 
2939 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2940 	sysctl_kern_proc, "Process table");
2941 
2942 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2943 	sysctl_kern_proc, "Process table");
2944 
2945 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2946 	sysctl_kern_proc, "Process table");
2947 
2948 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2949 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2950 
2951 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2952 	sysctl_kern_proc, "Process table");
2953 
2954 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2955 	sysctl_kern_proc, "Process table");
2956 
2957 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2958 	sysctl_kern_proc, "Process table");
2959 
2960 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2961 	sysctl_kern_proc, "Process table");
2962 
2963 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2964 	sysctl_kern_proc, "Return process table, no threads");
2965 
2966 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2967 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2968 	sysctl_kern_proc_args, "Process argument list");
2969 
2970 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
2971 	sysctl_kern_proc_env, "Process environment");
2972 
2973 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
2974 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
2975 
2976 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2977 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2978 
2979 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2980 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2981 	"Process syscall vector name (ABI type)");
2982 
2983 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2984 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2985 
2986 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2987 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2988 
2989 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2990 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2991 
2992 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2993 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2994 
2995 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2996 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2997 
2998 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2999 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3000 
3001 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3002 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3003 
3004 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3005 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3006 
3007 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3008 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3009 	"Return process table, no threads");
3010 
3011 #ifdef COMPAT_FREEBSD7
3012 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3013 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3014 #endif
3015 
3016 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3017 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3018 
3019 #if defined(STACK) || defined(DDB)
3020 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3021 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3022 #endif
3023 
3024 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3025 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3026 
3027 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3028 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3029 	"Process resource limits");
3030 
3031 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3032 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3033 	"Process ps_strings location");
3034 
3035 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3036 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3037 
3038 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3039 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3040 	"Process binary osreldate");
3041 
3042 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3043 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3044 	"Process signal trampoline location");
3045 
3046 int allproc_gen;
3047 
3048 /*
3049  * stop_all_proc() purpose is to stop all process which have usermode,
3050  * except current process for obvious reasons.  This makes it somewhat
3051  * unreliable when invoked from multithreaded process.  The service
3052  * must not be user-callable anyway.
3053  */
3054 void
3055 stop_all_proc(void)
3056 {
3057 	struct proc *cp, *p;
3058 	int r, gen;
3059 	bool restart, seen_stopped, seen_exiting, stopped_some;
3060 
3061 	cp = curproc;
3062 allproc_loop:
3063 	sx_xlock(&allproc_lock);
3064 	gen = allproc_gen;
3065 	seen_exiting = seen_stopped = stopped_some = restart = false;
3066 	LIST_REMOVE(cp, p_list);
3067 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3068 	for (;;) {
3069 		p = LIST_NEXT(cp, p_list);
3070 		if (p == NULL)
3071 			break;
3072 		LIST_REMOVE(cp, p_list);
3073 		LIST_INSERT_AFTER(p, cp, p_list);
3074 		PROC_LOCK(p);
3075 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) {
3076 			PROC_UNLOCK(p);
3077 			continue;
3078 		}
3079 		if ((p->p_flag & P_WEXIT) != 0) {
3080 			seen_exiting = true;
3081 			PROC_UNLOCK(p);
3082 			continue;
3083 		}
3084 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3085 			/*
3086 			 * Stopped processes are tolerated when there
3087 			 * are no other processes which might continue
3088 			 * them.  P_STOPPED_SINGLE but not
3089 			 * P_TOTAL_STOP process still has at least one
3090 			 * thread running.
3091 			 */
3092 			seen_stopped = true;
3093 			PROC_UNLOCK(p);
3094 			continue;
3095 		}
3096 		_PHOLD(p);
3097 		sx_xunlock(&allproc_lock);
3098 		r = thread_single(p, SINGLE_ALLPROC);
3099 		if (r != 0)
3100 			restart = true;
3101 		else
3102 			stopped_some = true;
3103 		_PRELE(p);
3104 		PROC_UNLOCK(p);
3105 		sx_xlock(&allproc_lock);
3106 	}
3107 	/* Catch forked children we did not see in iteration. */
3108 	if (gen != allproc_gen)
3109 		restart = true;
3110 	sx_xunlock(&allproc_lock);
3111 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3112 		kern_yield(PRI_USER);
3113 		goto allproc_loop;
3114 	}
3115 }
3116 
3117 void
3118 resume_all_proc(void)
3119 {
3120 	struct proc *cp, *p;
3121 
3122 	cp = curproc;
3123 	sx_xlock(&allproc_lock);
3124 again:
3125 	LIST_REMOVE(cp, p_list);
3126 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3127 	for (;;) {
3128 		p = LIST_NEXT(cp, p_list);
3129 		if (p == NULL)
3130 			break;
3131 		LIST_REMOVE(cp, p_list);
3132 		LIST_INSERT_AFTER(p, cp, p_list);
3133 		PROC_LOCK(p);
3134 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3135 			sx_xunlock(&allproc_lock);
3136 			_PHOLD(p);
3137 			thread_single_end(p, SINGLE_ALLPROC);
3138 			_PRELE(p);
3139 			PROC_UNLOCK(p);
3140 			sx_xlock(&allproc_lock);
3141 		} else {
3142 			PROC_UNLOCK(p);
3143 		}
3144 	}
3145 	/*  Did the loop above missed any stopped process ? */
3146 	LIST_FOREACH(p, &allproc, p_list) {
3147 		/* No need for proc lock. */
3148 		if ((p->p_flag & P_TOTAL_STOP) != 0)
3149 			goto again;
3150 	}
3151 	sx_xunlock(&allproc_lock);
3152 }
3153 
3154 /* #define	TOTAL_STOP_DEBUG	1 */
3155 #ifdef TOTAL_STOP_DEBUG
3156 volatile static int ap_resume;
3157 #include <sys/mount.h>
3158 
3159 static int
3160 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3161 {
3162 	int error, val;
3163 
3164 	val = 0;
3165 	ap_resume = 0;
3166 	error = sysctl_handle_int(oidp, &val, 0, req);
3167 	if (error != 0 || req->newptr == NULL)
3168 		return (error);
3169 	if (val != 0) {
3170 		stop_all_proc();
3171 		syncer_suspend();
3172 		while (ap_resume == 0)
3173 			;
3174 		syncer_resume();
3175 		resume_all_proc();
3176 	}
3177 	return (0);
3178 }
3179 
3180 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3181     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3182     sysctl_debug_stop_all_proc, "I",
3183     "");
3184 #endif
3185