xref: /freebsd/sys/kern/kern_proc.c (revision fcb560670601b2a4d87bb31d7531c8dcc37ee71b)
1 /*-
2  * Copyright (c) 1982, 1986, 1989, 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)kern_proc.c	8.7 (Berkeley) 2/14/95
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_compat.h"
36 #include "opt_ddb.h"
37 #include "opt_ktrace.h"
38 #include "opt_kstack_pages.h"
39 #include "opt_stack.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/elf.h>
44 #include <sys/exec.h>
45 #include <sys/kernel.h>
46 #include <sys/limits.h>
47 #include <sys/lock.h>
48 #include <sys/loginclass.h>
49 #include <sys/malloc.h>
50 #include <sys/mman.h>
51 #include <sys/mount.h>
52 #include <sys/mutex.h>
53 #include <sys/proc.h>
54 #include <sys/ptrace.h>
55 #include <sys/refcount.h>
56 #include <sys/resourcevar.h>
57 #include <sys/rwlock.h>
58 #include <sys/sbuf.h>
59 #include <sys/sysent.h>
60 #include <sys/sched.h>
61 #include <sys/smp.h>
62 #include <sys/stack.h>
63 #include <sys/stat.h>
64 #include <sys/sysctl.h>
65 #include <sys/filedesc.h>
66 #include <sys/tty.h>
67 #include <sys/signalvar.h>
68 #include <sys/sdt.h>
69 #include <sys/sx.h>
70 #include <sys/user.h>
71 #include <sys/jail.h>
72 #include <sys/vnode.h>
73 #include <sys/eventhandler.h>
74 
75 #ifdef DDB
76 #include <ddb/ddb.h>
77 #endif
78 
79 #include <vm/vm.h>
80 #include <vm/vm_param.h>
81 #include <vm/vm_extern.h>
82 #include <vm/pmap.h>
83 #include <vm/vm_map.h>
84 #include <vm/vm_object.h>
85 #include <vm/vm_page.h>
86 #include <vm/uma.h>
87 
88 #ifdef COMPAT_FREEBSD32
89 #include <compat/freebsd32/freebsd32.h>
90 #include <compat/freebsd32/freebsd32_util.h>
91 #endif
92 
93 SDT_PROVIDER_DEFINE(proc);
94 SDT_PROBE_DEFINE4(proc, kernel, ctor, entry, "struct proc *", "int",
95     "void *", "int");
96 SDT_PROBE_DEFINE4(proc, kernel, ctor, return, "struct proc *", "int",
97     "void *", "int");
98 SDT_PROBE_DEFINE4(proc, kernel, dtor, entry, "struct proc *", "int",
99     "void *", "struct thread *");
100 SDT_PROBE_DEFINE3(proc, kernel, dtor, return, "struct proc *", "int",
101     "void *");
102 SDT_PROBE_DEFINE3(proc, kernel, init, entry, "struct proc *", "int",
103     "int");
104 SDT_PROBE_DEFINE3(proc, kernel, init, return, "struct proc *", "int",
105     "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 allproc_lock;
137 struct sx proctree_lock;
138 struct mtx ppeers_lock;
139 uma_zone_t proc_zone;
140 
141 int kstack_pages = KSTACK_PAGES;
142 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
143     "Kernel stack size in pages");
144 static int vmmap_skip_res_cnt = 0;
145 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
146     &vmmap_skip_res_cnt, 0,
147     "Skip calculation of the pages resident count in kern.proc.vmmap");
148 
149 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
150 #ifdef COMPAT_FREEBSD32
151 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
152 #endif
153 
154 /*
155  * Initialize global process hashing structures.
156  */
157 void
158 procinit()
159 {
160 
161 	sx_init(&allproc_lock, "allproc");
162 	sx_init(&proctree_lock, "proctree");
163 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
164 	LIST_INIT(&allproc);
165 	LIST_INIT(&zombproc);
166 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
167 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
168 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
169 	    proc_ctor, proc_dtor, proc_init, proc_fini,
170 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
171 	uihashinit();
172 }
173 
174 /*
175  * Prepare a proc for use.
176  */
177 static int
178 proc_ctor(void *mem, int size, void *arg, int flags)
179 {
180 	struct proc *p;
181 
182 	p = (struct proc *)mem;
183 	SDT_PROBE(proc, kernel, ctor , entry, p, size, arg, flags, 0);
184 	EVENTHANDLER_INVOKE(process_ctor, p);
185 	SDT_PROBE(proc, kernel, ctor , return, p, size, arg, flags, 0);
186 	return (0);
187 }
188 
189 /*
190  * Reclaim a proc after use.
191  */
192 static void
193 proc_dtor(void *mem, int size, void *arg)
194 {
195 	struct proc *p;
196 	struct thread *td;
197 
198 	/* INVARIANTS checks go here */
199 	p = (struct proc *)mem;
200 	td = FIRST_THREAD_IN_PROC(p);
201 	SDT_PROBE(proc, kernel, dtor, entry, p, size, arg, td, 0);
202 	if (td != NULL) {
203 #ifdef INVARIANTS
204 		KASSERT((p->p_numthreads == 1),
205 		    ("bad number of threads in exiting process"));
206 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
207 #endif
208 		/* Free all OSD associated to this thread. */
209 		osd_thread_exit(td);
210 	}
211 	EVENTHANDLER_INVOKE(process_dtor, p);
212 	if (p->p_ksi != NULL)
213 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
214 	SDT_PROBE(proc, kernel, dtor, return, p, size, arg, 0, 0);
215 }
216 
217 /*
218  * Initialize type-stable parts of a proc (when newly created).
219  */
220 static int
221 proc_init(void *mem, int size, int flags)
222 {
223 	struct proc *p;
224 
225 	p = (struct proc *)mem;
226 	SDT_PROBE(proc, kernel, init, entry, p, size, flags, 0, 0);
227 	p->p_sched = (struct p_sched *)&p[1];
228 	bzero(&p->p_mtx, sizeof(struct mtx));
229 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK);
230 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN);
231 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN);
232 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN);
233 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN);
234 	cv_init(&p->p_pwait, "ppwait");
235 	cv_init(&p->p_dbgwait, "dbgwait");
236 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
237 	EVENTHANDLER_INVOKE(process_init, p);
238 	p->p_stats = pstats_alloc();
239 	SDT_PROBE(proc, kernel, init, return, p, size, flags, 0, 0);
240 	return (0);
241 }
242 
243 /*
244  * UMA should ensure that this function is never called.
245  * Freeing a proc structure would violate type stability.
246  */
247 static void
248 proc_fini(void *mem, int size)
249 {
250 #ifdef notnow
251 	struct proc *p;
252 
253 	p = (struct proc *)mem;
254 	EVENTHANDLER_INVOKE(process_fini, p);
255 	pstats_free(p->p_stats);
256 	thread_free(FIRST_THREAD_IN_PROC(p));
257 	mtx_destroy(&p->p_mtx);
258 	if (p->p_ksi != NULL)
259 		ksiginfo_free(p->p_ksi);
260 #else
261 	panic("proc reclaimed");
262 #endif
263 }
264 
265 /*
266  * Is p an inferior of the current process?
267  */
268 int
269 inferior(struct proc *p)
270 {
271 
272 	sx_assert(&proctree_lock, SX_LOCKED);
273 	PROC_LOCK_ASSERT(p, MA_OWNED);
274 	for (; p != curproc; p = proc_realparent(p)) {
275 		if (p->p_pid == 0)
276 			return (0);
277 	}
278 	return (1);
279 }
280 
281 struct proc *
282 pfind_locked(pid_t pid)
283 {
284 	struct proc *p;
285 
286 	sx_assert(&allproc_lock, SX_LOCKED);
287 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
288 		if (p->p_pid == pid) {
289 			PROC_LOCK(p);
290 			if (p->p_state == PRS_NEW) {
291 				PROC_UNLOCK(p);
292 				p = NULL;
293 			}
294 			break;
295 		}
296 	}
297 	return (p);
298 }
299 
300 /*
301  * Locate a process by number; return only "live" processes -- i.e., neither
302  * zombies nor newly born but incompletely initialized processes.  By not
303  * returning processes in the PRS_NEW state, we allow callers to avoid
304  * testing for that condition to avoid dereferencing p_ucred, et al.
305  */
306 struct proc *
307 pfind(pid_t pid)
308 {
309 	struct proc *p;
310 
311 	sx_slock(&allproc_lock);
312 	p = pfind_locked(pid);
313 	sx_sunlock(&allproc_lock);
314 	return (p);
315 }
316 
317 static struct proc *
318 pfind_tid_locked(pid_t tid)
319 {
320 	struct proc *p;
321 	struct thread *td;
322 
323 	sx_assert(&allproc_lock, SX_LOCKED);
324 	FOREACH_PROC_IN_SYSTEM(p) {
325 		PROC_LOCK(p);
326 		if (p->p_state == PRS_NEW) {
327 			PROC_UNLOCK(p);
328 			continue;
329 		}
330 		FOREACH_THREAD_IN_PROC(p, td) {
331 			if (td->td_tid == tid)
332 				goto found;
333 		}
334 		PROC_UNLOCK(p);
335 	}
336 found:
337 	return (p);
338 }
339 
340 /*
341  * Locate a process group by number.
342  * The caller must hold proctree_lock.
343  */
344 struct pgrp *
345 pgfind(pgid)
346 	register pid_t pgid;
347 {
348 	register struct pgrp *pgrp;
349 
350 	sx_assert(&proctree_lock, SX_LOCKED);
351 
352 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
353 		if (pgrp->pg_id == pgid) {
354 			PGRP_LOCK(pgrp);
355 			return (pgrp);
356 		}
357 	}
358 	return (NULL);
359 }
360 
361 /*
362  * Locate process and do additional manipulations, depending on flags.
363  */
364 int
365 pget(pid_t pid, int flags, struct proc **pp)
366 {
367 	struct proc *p;
368 	int error;
369 
370 	sx_slock(&allproc_lock);
371 	if (pid <= PID_MAX) {
372 		p = pfind_locked(pid);
373 		if (p == NULL && (flags & PGET_NOTWEXIT) == 0)
374 			p = zpfind_locked(pid);
375 	} else if ((flags & PGET_NOTID) == 0) {
376 		p = pfind_tid_locked(pid);
377 	} else {
378 		p = NULL;
379 	}
380 	sx_sunlock(&allproc_lock);
381 	if (p == NULL)
382 		return (ESRCH);
383 	if ((flags & PGET_CANSEE) != 0) {
384 		error = p_cansee(curthread, p);
385 		if (error != 0)
386 			goto errout;
387 	}
388 	if ((flags & PGET_CANDEBUG) != 0) {
389 		error = p_candebug(curthread, p);
390 		if (error != 0)
391 			goto errout;
392 	}
393 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
394 		error = EPERM;
395 		goto errout;
396 	}
397 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
398 		error = ESRCH;
399 		goto errout;
400 	}
401 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
402 		/*
403 		 * XXXRW: Not clear ESRCH is the right error during proc
404 		 * execve().
405 		 */
406 		error = ESRCH;
407 		goto errout;
408 	}
409 	if ((flags & PGET_HOLD) != 0) {
410 		_PHOLD(p);
411 		PROC_UNLOCK(p);
412 	}
413 	*pp = p;
414 	return (0);
415 errout:
416 	PROC_UNLOCK(p);
417 	return (error);
418 }
419 
420 /*
421  * Create a new process group.
422  * pgid must be equal to the pid of p.
423  * Begin a new session if required.
424  */
425 int
426 enterpgrp(p, pgid, pgrp, sess)
427 	register struct proc *p;
428 	pid_t pgid;
429 	struct pgrp *pgrp;
430 	struct session *sess;
431 {
432 
433 	sx_assert(&proctree_lock, SX_XLOCKED);
434 
435 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
436 	KASSERT(p->p_pid == pgid,
437 	    ("enterpgrp: new pgrp and pid != pgid"));
438 	KASSERT(pgfind(pgid) == NULL,
439 	    ("enterpgrp: pgrp with pgid exists"));
440 	KASSERT(!SESS_LEADER(p),
441 	    ("enterpgrp: session leader attempted setpgrp"));
442 
443 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
444 
445 	if (sess != NULL) {
446 		/*
447 		 * new session
448 		 */
449 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
450 		PROC_LOCK(p);
451 		p->p_flag &= ~P_CONTROLT;
452 		PROC_UNLOCK(p);
453 		PGRP_LOCK(pgrp);
454 		sess->s_leader = p;
455 		sess->s_sid = p->p_pid;
456 		refcount_init(&sess->s_count, 1);
457 		sess->s_ttyvp = NULL;
458 		sess->s_ttydp = NULL;
459 		sess->s_ttyp = NULL;
460 		bcopy(p->p_session->s_login, sess->s_login,
461 			    sizeof(sess->s_login));
462 		pgrp->pg_session = sess;
463 		KASSERT(p == curproc,
464 		    ("enterpgrp: mksession and p != curproc"));
465 	} else {
466 		pgrp->pg_session = p->p_session;
467 		sess_hold(pgrp->pg_session);
468 		PGRP_LOCK(pgrp);
469 	}
470 	pgrp->pg_id = pgid;
471 	LIST_INIT(&pgrp->pg_members);
472 
473 	/*
474 	 * As we have an exclusive lock of proctree_lock,
475 	 * this should not deadlock.
476 	 */
477 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
478 	pgrp->pg_jobc = 0;
479 	SLIST_INIT(&pgrp->pg_sigiolst);
480 	PGRP_UNLOCK(pgrp);
481 
482 	doenterpgrp(p, pgrp);
483 
484 	return (0);
485 }
486 
487 /*
488  * Move p to an existing process group
489  */
490 int
491 enterthispgrp(p, pgrp)
492 	register struct proc *p;
493 	struct pgrp *pgrp;
494 {
495 
496 	sx_assert(&proctree_lock, SX_XLOCKED);
497 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
498 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
499 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
500 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
501 	KASSERT(pgrp->pg_session == p->p_session,
502 		("%s: pgrp's session %p, p->p_session %p.\n",
503 		__func__,
504 		pgrp->pg_session,
505 		p->p_session));
506 	KASSERT(pgrp != p->p_pgrp,
507 		("%s: p belongs to pgrp.", __func__));
508 
509 	doenterpgrp(p, pgrp);
510 
511 	return (0);
512 }
513 
514 /*
515  * Move p to a process group
516  */
517 static void
518 doenterpgrp(p, pgrp)
519 	struct proc *p;
520 	struct pgrp *pgrp;
521 {
522 	struct pgrp *savepgrp;
523 
524 	sx_assert(&proctree_lock, SX_XLOCKED);
525 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
526 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
527 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
528 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
529 
530 	savepgrp = p->p_pgrp;
531 
532 	/*
533 	 * Adjust eligibility of affected pgrps to participate in job control.
534 	 * Increment eligibility counts before decrementing, otherwise we
535 	 * could reach 0 spuriously during the first call.
536 	 */
537 	fixjobc(p, pgrp, 1);
538 	fixjobc(p, p->p_pgrp, 0);
539 
540 	PGRP_LOCK(pgrp);
541 	PGRP_LOCK(savepgrp);
542 	PROC_LOCK(p);
543 	LIST_REMOVE(p, p_pglist);
544 	p->p_pgrp = pgrp;
545 	PROC_UNLOCK(p);
546 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
547 	PGRP_UNLOCK(savepgrp);
548 	PGRP_UNLOCK(pgrp);
549 	if (LIST_EMPTY(&savepgrp->pg_members))
550 		pgdelete(savepgrp);
551 }
552 
553 /*
554  * remove process from process group
555  */
556 int
557 leavepgrp(p)
558 	register struct proc *p;
559 {
560 	struct pgrp *savepgrp;
561 
562 	sx_assert(&proctree_lock, SX_XLOCKED);
563 	savepgrp = p->p_pgrp;
564 	PGRP_LOCK(savepgrp);
565 	PROC_LOCK(p);
566 	LIST_REMOVE(p, p_pglist);
567 	p->p_pgrp = NULL;
568 	PROC_UNLOCK(p);
569 	PGRP_UNLOCK(savepgrp);
570 	if (LIST_EMPTY(&savepgrp->pg_members))
571 		pgdelete(savepgrp);
572 	return (0);
573 }
574 
575 /*
576  * delete a process group
577  */
578 static void
579 pgdelete(pgrp)
580 	register struct pgrp *pgrp;
581 {
582 	struct session *savesess;
583 	struct tty *tp;
584 
585 	sx_assert(&proctree_lock, SX_XLOCKED);
586 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
587 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
588 
589 	/*
590 	 * Reset any sigio structures pointing to us as a result of
591 	 * F_SETOWN with our pgid.
592 	 */
593 	funsetownlst(&pgrp->pg_sigiolst);
594 
595 	PGRP_LOCK(pgrp);
596 	tp = pgrp->pg_session->s_ttyp;
597 	LIST_REMOVE(pgrp, pg_hash);
598 	savesess = pgrp->pg_session;
599 	PGRP_UNLOCK(pgrp);
600 
601 	/* Remove the reference to the pgrp before deallocating it. */
602 	if (tp != NULL) {
603 		tty_lock(tp);
604 		tty_rel_pgrp(tp, pgrp);
605 	}
606 
607 	mtx_destroy(&pgrp->pg_mtx);
608 	free(pgrp, M_PGRP);
609 	sess_release(savesess);
610 }
611 
612 static void
613 pgadjustjobc(pgrp, entering)
614 	struct pgrp *pgrp;
615 	int entering;
616 {
617 
618 	PGRP_LOCK(pgrp);
619 	if (entering)
620 		pgrp->pg_jobc++;
621 	else {
622 		--pgrp->pg_jobc;
623 		if (pgrp->pg_jobc == 0)
624 			orphanpg(pgrp);
625 	}
626 	PGRP_UNLOCK(pgrp);
627 }
628 
629 /*
630  * Adjust pgrp jobc counters when specified process changes process group.
631  * We count the number of processes in each process group that "qualify"
632  * the group for terminal job control (those with a parent in a different
633  * process group of the same session).  If that count reaches zero, the
634  * process group becomes orphaned.  Check both the specified process'
635  * process group and that of its children.
636  * entering == 0 => p is leaving specified group.
637  * entering == 1 => p is entering specified group.
638  */
639 void
640 fixjobc(p, pgrp, entering)
641 	register struct proc *p;
642 	register struct pgrp *pgrp;
643 	int entering;
644 {
645 	register struct pgrp *hispgrp;
646 	register struct session *mysession;
647 
648 	sx_assert(&proctree_lock, SX_LOCKED);
649 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
650 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
651 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
652 
653 	/*
654 	 * Check p's parent to see whether p qualifies its own process
655 	 * group; if so, adjust count for p's process group.
656 	 */
657 	mysession = pgrp->pg_session;
658 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
659 	    hispgrp->pg_session == mysession)
660 		pgadjustjobc(pgrp, entering);
661 
662 	/*
663 	 * Check this process' children to see whether they qualify
664 	 * their process groups; if so, adjust counts for children's
665 	 * process groups.
666 	 */
667 	LIST_FOREACH(p, &p->p_children, p_sibling) {
668 		hispgrp = p->p_pgrp;
669 		if (hispgrp == pgrp ||
670 		    hispgrp->pg_session != mysession)
671 			continue;
672 		PROC_LOCK(p);
673 		if (p->p_state == PRS_ZOMBIE) {
674 			PROC_UNLOCK(p);
675 			continue;
676 		}
677 		PROC_UNLOCK(p);
678 		pgadjustjobc(hispgrp, entering);
679 	}
680 }
681 
682 /*
683  * A process group has become orphaned;
684  * if there are any stopped processes in the group,
685  * hang-up all process in that group.
686  */
687 static void
688 orphanpg(pg)
689 	struct pgrp *pg;
690 {
691 	register struct proc *p;
692 
693 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
694 
695 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
696 		PROC_LOCK(p);
697 		if (P_SHOULDSTOP(p)) {
698 			PROC_UNLOCK(p);
699 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
700 				PROC_LOCK(p);
701 				kern_psignal(p, SIGHUP);
702 				kern_psignal(p, SIGCONT);
703 				PROC_UNLOCK(p);
704 			}
705 			return;
706 		}
707 		PROC_UNLOCK(p);
708 	}
709 }
710 
711 void
712 sess_hold(struct session *s)
713 {
714 
715 	refcount_acquire(&s->s_count);
716 }
717 
718 void
719 sess_release(struct session *s)
720 {
721 
722 	if (refcount_release(&s->s_count)) {
723 		if (s->s_ttyp != NULL) {
724 			tty_lock(s->s_ttyp);
725 			tty_rel_sess(s->s_ttyp, s);
726 		}
727 		mtx_destroy(&s->s_mtx);
728 		free(s, M_SESSION);
729 	}
730 }
731 
732 #ifdef DDB
733 
734 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
735 {
736 	register struct pgrp *pgrp;
737 	register struct proc *p;
738 	register int i;
739 
740 	for (i = 0; i <= pgrphash; i++) {
741 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
742 			printf("\tindx %d\n", i);
743 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
744 				printf(
745 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
746 				    (void *)pgrp, (long)pgrp->pg_id,
747 				    (void *)pgrp->pg_session,
748 				    pgrp->pg_session->s_count,
749 				    (void *)LIST_FIRST(&pgrp->pg_members));
750 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
751 					printf("\t\tpid %ld addr %p pgrp %p\n",
752 					    (long)p->p_pid, (void *)p,
753 					    (void *)p->p_pgrp);
754 				}
755 			}
756 		}
757 	}
758 }
759 #endif /* DDB */
760 
761 /*
762  * Calculate the kinfo_proc members which contain process-wide
763  * informations.
764  * Must be called with the target process locked.
765  */
766 static void
767 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
768 {
769 	struct thread *td;
770 
771 	PROC_LOCK_ASSERT(p, MA_OWNED);
772 
773 	kp->ki_estcpu = 0;
774 	kp->ki_pctcpu = 0;
775 	FOREACH_THREAD_IN_PROC(p, td) {
776 		thread_lock(td);
777 		kp->ki_pctcpu += sched_pctcpu(td);
778 		kp->ki_estcpu += td->td_estcpu;
779 		thread_unlock(td);
780 	}
781 }
782 
783 /*
784  * Clear kinfo_proc and fill in any information that is common
785  * to all threads in the process.
786  * Must be called with the target process locked.
787  */
788 static void
789 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
790 {
791 	struct thread *td0;
792 	struct tty *tp;
793 	struct session *sp;
794 	struct ucred *cred;
795 	struct sigacts *ps;
796 
797 	/* For proc_realparent. */
798 	sx_assert(&proctree_lock, SX_LOCKED);
799 	PROC_LOCK_ASSERT(p, MA_OWNED);
800 	bzero(kp, sizeof(*kp));
801 
802 	kp->ki_structsize = sizeof(*kp);
803 	kp->ki_paddr = p;
804 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
805 	kp->ki_args = p->p_args;
806 	kp->ki_textvp = p->p_textvp;
807 #ifdef KTRACE
808 	kp->ki_tracep = p->p_tracevp;
809 	kp->ki_traceflag = p->p_traceflag;
810 #endif
811 	kp->ki_fd = p->p_fd;
812 	kp->ki_vmspace = p->p_vmspace;
813 	kp->ki_flag = p->p_flag;
814 	kp->ki_flag2 = p->p_flag2;
815 	cred = p->p_ucred;
816 	if (cred) {
817 		kp->ki_uid = cred->cr_uid;
818 		kp->ki_ruid = cred->cr_ruid;
819 		kp->ki_svuid = cred->cr_svuid;
820 		kp->ki_cr_flags = 0;
821 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
822 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
823 		/* XXX bde doesn't like KI_NGROUPS */
824 		if (cred->cr_ngroups > KI_NGROUPS) {
825 			kp->ki_ngroups = KI_NGROUPS;
826 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
827 		} else
828 			kp->ki_ngroups = cred->cr_ngroups;
829 		bcopy(cred->cr_groups, kp->ki_groups,
830 		    kp->ki_ngroups * sizeof(gid_t));
831 		kp->ki_rgid = cred->cr_rgid;
832 		kp->ki_svgid = cred->cr_svgid;
833 		/* If jailed(cred), emulate the old P_JAILED flag. */
834 		if (jailed(cred)) {
835 			kp->ki_flag |= P_JAILED;
836 			/* If inside the jail, use 0 as a jail ID. */
837 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
838 				kp->ki_jid = cred->cr_prison->pr_id;
839 		}
840 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
841 		    sizeof(kp->ki_loginclass));
842 	}
843 	ps = p->p_sigacts;
844 	if (ps) {
845 		mtx_lock(&ps->ps_mtx);
846 		kp->ki_sigignore = ps->ps_sigignore;
847 		kp->ki_sigcatch = ps->ps_sigcatch;
848 		mtx_unlock(&ps->ps_mtx);
849 	}
850 	if (p->p_state != PRS_NEW &&
851 	    p->p_state != PRS_ZOMBIE &&
852 	    p->p_vmspace != NULL) {
853 		struct vmspace *vm = p->p_vmspace;
854 
855 		kp->ki_size = vm->vm_map.size;
856 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
857 		FOREACH_THREAD_IN_PROC(p, td0) {
858 			if (!TD_IS_SWAPPED(td0))
859 				kp->ki_rssize += td0->td_kstack_pages;
860 		}
861 		kp->ki_swrss = vm->vm_swrss;
862 		kp->ki_tsize = vm->vm_tsize;
863 		kp->ki_dsize = vm->vm_dsize;
864 		kp->ki_ssize = vm->vm_ssize;
865 	} else if (p->p_state == PRS_ZOMBIE)
866 		kp->ki_stat = SZOMB;
867 	if (kp->ki_flag & P_INMEM)
868 		kp->ki_sflag = PS_INMEM;
869 	else
870 		kp->ki_sflag = 0;
871 	/* Calculate legacy swtime as seconds since 'swtick'. */
872 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
873 	kp->ki_pid = p->p_pid;
874 	kp->ki_nice = p->p_nice;
875 	kp->ki_fibnum = p->p_fibnum;
876 	kp->ki_start = p->p_stats->p_start;
877 	timevaladd(&kp->ki_start, &boottime);
878 	PROC_STATLOCK(p);
879 	rufetch(p, &kp->ki_rusage);
880 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
881 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
882 	PROC_STATUNLOCK(p);
883 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
884 	/* Some callers want child times in a single value. */
885 	kp->ki_childtime = kp->ki_childstime;
886 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
887 
888 	FOREACH_THREAD_IN_PROC(p, td0)
889 		kp->ki_cow += td0->td_cow;
890 
891 	tp = NULL;
892 	if (p->p_pgrp) {
893 		kp->ki_pgid = p->p_pgrp->pg_id;
894 		kp->ki_jobc = p->p_pgrp->pg_jobc;
895 		sp = p->p_pgrp->pg_session;
896 
897 		if (sp != NULL) {
898 			kp->ki_sid = sp->s_sid;
899 			SESS_LOCK(sp);
900 			strlcpy(kp->ki_login, sp->s_login,
901 			    sizeof(kp->ki_login));
902 			if (sp->s_ttyvp)
903 				kp->ki_kiflag |= KI_CTTY;
904 			if (SESS_LEADER(p))
905 				kp->ki_kiflag |= KI_SLEADER;
906 			/* XXX proctree_lock */
907 			tp = sp->s_ttyp;
908 			SESS_UNLOCK(sp);
909 		}
910 	}
911 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
912 		kp->ki_tdev = tty_udev(tp);
913 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
914 		if (tp->t_session)
915 			kp->ki_tsid = tp->t_session->s_sid;
916 	} else
917 		kp->ki_tdev = NODEV;
918 	if (p->p_comm[0] != '\0')
919 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
920 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
921 	    p->p_sysent->sv_name[0] != '\0')
922 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
923 	kp->ki_siglist = p->p_siglist;
924 	kp->ki_xstat = p->p_xstat;
925 	kp->ki_acflag = p->p_acflag;
926 	kp->ki_lock = p->p_lock;
927 	if (p->p_pptr) {
928 		kp->ki_ppid = proc_realparent(p)->p_pid;
929 		if (p->p_flag & P_TRACED)
930 			kp->ki_tracer = p->p_pptr->p_pid;
931 	}
932 }
933 
934 /*
935  * Fill in information that is thread specific.  Must be called with
936  * target process locked.  If 'preferthread' is set, overwrite certain
937  * process-related fields that are maintained for both threads and
938  * processes.
939  */
940 static void
941 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
942 {
943 	struct proc *p;
944 
945 	p = td->td_proc;
946 	kp->ki_tdaddr = td;
947 	PROC_LOCK_ASSERT(p, MA_OWNED);
948 
949 	if (preferthread)
950 		PROC_STATLOCK(p);
951 	thread_lock(td);
952 	if (td->td_wmesg != NULL)
953 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
954 	else
955 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
956 	strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname));
957 	if (TD_ON_LOCK(td)) {
958 		kp->ki_kiflag |= KI_LOCKBLOCK;
959 		strlcpy(kp->ki_lockname, td->td_lockname,
960 		    sizeof(kp->ki_lockname));
961 	} else {
962 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
963 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
964 	}
965 
966 	if (p->p_state == PRS_NORMAL) { /* approximate. */
967 		if (TD_ON_RUNQ(td) ||
968 		    TD_CAN_RUN(td) ||
969 		    TD_IS_RUNNING(td)) {
970 			kp->ki_stat = SRUN;
971 		} else if (P_SHOULDSTOP(p)) {
972 			kp->ki_stat = SSTOP;
973 		} else if (TD_IS_SLEEPING(td)) {
974 			kp->ki_stat = SSLEEP;
975 		} else if (TD_ON_LOCK(td)) {
976 			kp->ki_stat = SLOCK;
977 		} else {
978 			kp->ki_stat = SWAIT;
979 		}
980 	} else if (p->p_state == PRS_ZOMBIE) {
981 		kp->ki_stat = SZOMB;
982 	} else {
983 		kp->ki_stat = SIDL;
984 	}
985 
986 	/* Things in the thread */
987 	kp->ki_wchan = td->td_wchan;
988 	kp->ki_pri.pri_level = td->td_priority;
989 	kp->ki_pri.pri_native = td->td_base_pri;
990 
991 	/*
992 	 * Note: legacy fields; clamp at the old NOCPU value and/or
993 	 * the maximum u_char CPU value.
994 	 */
995 	if (td->td_lastcpu == NOCPU)
996 		kp->ki_lastcpu_old = NOCPU_OLD;
997 	else if (td->td_lastcpu > MAXCPU_OLD)
998 		kp->ki_lastcpu_old = MAXCPU_OLD;
999 	else
1000 		kp->ki_lastcpu_old = td->td_lastcpu;
1001 
1002 	if (td->td_oncpu == NOCPU)
1003 		kp->ki_oncpu_old = NOCPU_OLD;
1004 	else if (td->td_oncpu > MAXCPU_OLD)
1005 		kp->ki_oncpu_old = MAXCPU_OLD;
1006 	else
1007 		kp->ki_oncpu_old = td->td_oncpu;
1008 
1009 	kp->ki_lastcpu = td->td_lastcpu;
1010 	kp->ki_oncpu = td->td_oncpu;
1011 	kp->ki_tdflags = td->td_flags;
1012 	kp->ki_tid = td->td_tid;
1013 	kp->ki_numthreads = p->p_numthreads;
1014 	kp->ki_pcb = td->td_pcb;
1015 	kp->ki_kstack = (void *)td->td_kstack;
1016 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1017 	kp->ki_pri.pri_class = td->td_pri_class;
1018 	kp->ki_pri.pri_user = td->td_user_pri;
1019 
1020 	if (preferthread) {
1021 		rufetchtd(td, &kp->ki_rusage);
1022 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1023 		kp->ki_pctcpu = sched_pctcpu(td);
1024 		kp->ki_estcpu = td->td_estcpu;
1025 		kp->ki_cow = td->td_cow;
1026 	}
1027 
1028 	/* We can't get this anymore but ps etc never used it anyway. */
1029 	kp->ki_rqindex = 0;
1030 
1031 	if (preferthread)
1032 		kp->ki_siglist = td->td_siglist;
1033 	kp->ki_sigmask = td->td_sigmask;
1034 	thread_unlock(td);
1035 	if (preferthread)
1036 		PROC_STATUNLOCK(p);
1037 }
1038 
1039 /*
1040  * Fill in a kinfo_proc structure for the specified process.
1041  * Must be called with the target process locked.
1042  */
1043 void
1044 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1045 {
1046 
1047 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1048 
1049 	fill_kinfo_proc_only(p, kp);
1050 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1051 	fill_kinfo_aggregate(p, kp);
1052 }
1053 
1054 struct pstats *
1055 pstats_alloc(void)
1056 {
1057 
1058 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1059 }
1060 
1061 /*
1062  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1063  */
1064 void
1065 pstats_fork(struct pstats *src, struct pstats *dst)
1066 {
1067 
1068 	bzero(&dst->pstat_startzero,
1069 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1070 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1071 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1072 }
1073 
1074 void
1075 pstats_free(struct pstats *ps)
1076 {
1077 
1078 	free(ps, M_SUBPROC);
1079 }
1080 
1081 static struct proc *
1082 zpfind_locked(pid_t pid)
1083 {
1084 	struct proc *p;
1085 
1086 	sx_assert(&allproc_lock, SX_LOCKED);
1087 	LIST_FOREACH(p, &zombproc, p_list) {
1088 		if (p->p_pid == pid) {
1089 			PROC_LOCK(p);
1090 			break;
1091 		}
1092 	}
1093 	return (p);
1094 }
1095 
1096 /*
1097  * Locate a zombie process by number
1098  */
1099 struct proc *
1100 zpfind(pid_t pid)
1101 {
1102 	struct proc *p;
1103 
1104 	sx_slock(&allproc_lock);
1105 	p = zpfind_locked(pid);
1106 	sx_sunlock(&allproc_lock);
1107 	return (p);
1108 }
1109 
1110 #ifdef COMPAT_FREEBSD32
1111 
1112 /*
1113  * This function is typically used to copy out the kernel address, so
1114  * it can be replaced by assignment of zero.
1115  */
1116 static inline uint32_t
1117 ptr32_trim(void *ptr)
1118 {
1119 	uintptr_t uptr;
1120 
1121 	uptr = (uintptr_t)ptr;
1122 	return ((uptr > UINT_MAX) ? 0 : uptr);
1123 }
1124 
1125 #define PTRTRIM_CP(src,dst,fld) \
1126 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1127 
1128 static void
1129 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1130 {
1131 	int i;
1132 
1133 	bzero(ki32, sizeof(struct kinfo_proc32));
1134 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1135 	CP(*ki, *ki32, ki_layout);
1136 	PTRTRIM_CP(*ki, *ki32, ki_args);
1137 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1138 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1139 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1140 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1141 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1142 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1143 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1144 	CP(*ki, *ki32, ki_pid);
1145 	CP(*ki, *ki32, ki_ppid);
1146 	CP(*ki, *ki32, ki_pgid);
1147 	CP(*ki, *ki32, ki_tpgid);
1148 	CP(*ki, *ki32, ki_sid);
1149 	CP(*ki, *ki32, ki_tsid);
1150 	CP(*ki, *ki32, ki_jobc);
1151 	CP(*ki, *ki32, ki_tdev);
1152 	CP(*ki, *ki32, ki_siglist);
1153 	CP(*ki, *ki32, ki_sigmask);
1154 	CP(*ki, *ki32, ki_sigignore);
1155 	CP(*ki, *ki32, ki_sigcatch);
1156 	CP(*ki, *ki32, ki_uid);
1157 	CP(*ki, *ki32, ki_ruid);
1158 	CP(*ki, *ki32, ki_svuid);
1159 	CP(*ki, *ki32, ki_rgid);
1160 	CP(*ki, *ki32, ki_svgid);
1161 	CP(*ki, *ki32, ki_ngroups);
1162 	for (i = 0; i < KI_NGROUPS; i++)
1163 		CP(*ki, *ki32, ki_groups[i]);
1164 	CP(*ki, *ki32, ki_size);
1165 	CP(*ki, *ki32, ki_rssize);
1166 	CP(*ki, *ki32, ki_swrss);
1167 	CP(*ki, *ki32, ki_tsize);
1168 	CP(*ki, *ki32, ki_dsize);
1169 	CP(*ki, *ki32, ki_ssize);
1170 	CP(*ki, *ki32, ki_xstat);
1171 	CP(*ki, *ki32, ki_acflag);
1172 	CP(*ki, *ki32, ki_pctcpu);
1173 	CP(*ki, *ki32, ki_estcpu);
1174 	CP(*ki, *ki32, ki_slptime);
1175 	CP(*ki, *ki32, ki_swtime);
1176 	CP(*ki, *ki32, ki_cow);
1177 	CP(*ki, *ki32, ki_runtime);
1178 	TV_CP(*ki, *ki32, ki_start);
1179 	TV_CP(*ki, *ki32, ki_childtime);
1180 	CP(*ki, *ki32, ki_flag);
1181 	CP(*ki, *ki32, ki_kiflag);
1182 	CP(*ki, *ki32, ki_traceflag);
1183 	CP(*ki, *ki32, ki_stat);
1184 	CP(*ki, *ki32, ki_nice);
1185 	CP(*ki, *ki32, ki_lock);
1186 	CP(*ki, *ki32, ki_rqindex);
1187 	CP(*ki, *ki32, ki_oncpu);
1188 	CP(*ki, *ki32, ki_lastcpu);
1189 
1190 	/* XXX TODO: wrap cpu value as appropriate */
1191 	CP(*ki, *ki32, ki_oncpu_old);
1192 	CP(*ki, *ki32, ki_lastcpu_old);
1193 
1194 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1195 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1196 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1197 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1198 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1199 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1200 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1201 	CP(*ki, *ki32, ki_tracer);
1202 	CP(*ki, *ki32, ki_flag2);
1203 	CP(*ki, *ki32, ki_fibnum);
1204 	CP(*ki, *ki32, ki_cr_flags);
1205 	CP(*ki, *ki32, ki_jid);
1206 	CP(*ki, *ki32, ki_numthreads);
1207 	CP(*ki, *ki32, ki_tid);
1208 	CP(*ki, *ki32, ki_pri);
1209 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1210 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1211 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1212 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1213 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1214 	CP(*ki, *ki32, ki_sflag);
1215 	CP(*ki, *ki32, ki_tdflags);
1216 }
1217 #endif
1218 
1219 int
1220 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1221 {
1222 	struct thread *td;
1223 	struct kinfo_proc ki;
1224 #ifdef COMPAT_FREEBSD32
1225 	struct kinfo_proc32 ki32;
1226 #endif
1227 	int error;
1228 
1229 	PROC_LOCK_ASSERT(p, MA_OWNED);
1230 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1231 
1232 	error = 0;
1233 	fill_kinfo_proc(p, &ki);
1234 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1235 #ifdef COMPAT_FREEBSD32
1236 		if ((flags & KERN_PROC_MASK32) != 0) {
1237 			freebsd32_kinfo_proc_out(&ki, &ki32);
1238 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1239 				error = ENOMEM;
1240 		} else
1241 #endif
1242 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1243 				error = ENOMEM;
1244 	} else {
1245 		FOREACH_THREAD_IN_PROC(p, td) {
1246 			fill_kinfo_thread(td, &ki, 1);
1247 #ifdef COMPAT_FREEBSD32
1248 			if ((flags & KERN_PROC_MASK32) != 0) {
1249 				freebsd32_kinfo_proc_out(&ki, &ki32);
1250 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1251 					error = ENOMEM;
1252 			} else
1253 #endif
1254 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1255 					error = ENOMEM;
1256 			if (error != 0)
1257 				break;
1258 		}
1259 	}
1260 	PROC_UNLOCK(p);
1261 	return (error);
1262 }
1263 
1264 static int
1265 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags,
1266     int doingzomb)
1267 {
1268 	struct sbuf sb;
1269 	struct kinfo_proc ki;
1270 	struct proc *np;
1271 	int error, error2;
1272 	pid_t pid;
1273 
1274 	pid = p->p_pid;
1275 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1276 	error = kern_proc_out(p, &sb, flags);
1277 	error2 = sbuf_finish(&sb);
1278 	sbuf_delete(&sb);
1279 	if (error != 0)
1280 		return (error);
1281 	else if (error2 != 0)
1282 		return (error2);
1283 	if (doingzomb)
1284 		np = zpfind(pid);
1285 	else {
1286 		if (pid == 0)
1287 			return (0);
1288 		np = pfind(pid);
1289 	}
1290 	if (np == NULL)
1291 		return (ESRCH);
1292 	if (np != p) {
1293 		PROC_UNLOCK(np);
1294 		return (ESRCH);
1295 	}
1296 	PROC_UNLOCK(np);
1297 	return (0);
1298 }
1299 
1300 static int
1301 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1302 {
1303 	int *name = (int *)arg1;
1304 	u_int namelen = arg2;
1305 	struct proc *p;
1306 	int flags, doingzomb, oid_number;
1307 	int error = 0;
1308 
1309 	oid_number = oidp->oid_number;
1310 	if (oid_number != KERN_PROC_ALL &&
1311 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1312 		flags = KERN_PROC_NOTHREADS;
1313 	else {
1314 		flags = 0;
1315 		oid_number &= ~KERN_PROC_INC_THREAD;
1316 	}
1317 #ifdef COMPAT_FREEBSD32
1318 	if (req->flags & SCTL_MASK32)
1319 		flags |= KERN_PROC_MASK32;
1320 #endif
1321 	if (oid_number == KERN_PROC_PID) {
1322 		if (namelen != 1)
1323 			return (EINVAL);
1324 		error = sysctl_wire_old_buffer(req, 0);
1325 		if (error)
1326 			return (error);
1327 		sx_slock(&proctree_lock);
1328 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1329 		if (error == 0)
1330 			error = sysctl_out_proc(p, req, flags, 0);
1331 		sx_sunlock(&proctree_lock);
1332 		return (error);
1333 	}
1334 
1335 	switch (oid_number) {
1336 	case KERN_PROC_ALL:
1337 		if (namelen != 0)
1338 			return (EINVAL);
1339 		break;
1340 	case KERN_PROC_PROC:
1341 		if (namelen != 0 && namelen != 1)
1342 			return (EINVAL);
1343 		break;
1344 	default:
1345 		if (namelen != 1)
1346 			return (EINVAL);
1347 		break;
1348 	}
1349 
1350 	if (!req->oldptr) {
1351 		/* overestimate by 5 procs */
1352 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1353 		if (error)
1354 			return (error);
1355 	}
1356 	error = sysctl_wire_old_buffer(req, 0);
1357 	if (error != 0)
1358 		return (error);
1359 	sx_slock(&proctree_lock);
1360 	sx_slock(&allproc_lock);
1361 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1362 		if (!doingzomb)
1363 			p = LIST_FIRST(&allproc);
1364 		else
1365 			p = LIST_FIRST(&zombproc);
1366 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
1367 			/*
1368 			 * Skip embryonic processes.
1369 			 */
1370 			PROC_LOCK(p);
1371 			if (p->p_state == PRS_NEW) {
1372 				PROC_UNLOCK(p);
1373 				continue;
1374 			}
1375 			KASSERT(p->p_ucred != NULL,
1376 			    ("process credential is NULL for non-NEW proc"));
1377 			/*
1378 			 * Show a user only appropriate processes.
1379 			 */
1380 			if (p_cansee(curthread, p)) {
1381 				PROC_UNLOCK(p);
1382 				continue;
1383 			}
1384 			/*
1385 			 * TODO - make more efficient (see notes below).
1386 			 * do by session.
1387 			 */
1388 			switch (oid_number) {
1389 
1390 			case KERN_PROC_GID:
1391 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1392 					PROC_UNLOCK(p);
1393 					continue;
1394 				}
1395 				break;
1396 
1397 			case KERN_PROC_PGRP:
1398 				/* could do this by traversing pgrp */
1399 				if (p->p_pgrp == NULL ||
1400 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1401 					PROC_UNLOCK(p);
1402 					continue;
1403 				}
1404 				break;
1405 
1406 			case KERN_PROC_RGID:
1407 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1408 					PROC_UNLOCK(p);
1409 					continue;
1410 				}
1411 				break;
1412 
1413 			case KERN_PROC_SESSION:
1414 				if (p->p_session == NULL ||
1415 				    p->p_session->s_sid != (pid_t)name[0]) {
1416 					PROC_UNLOCK(p);
1417 					continue;
1418 				}
1419 				break;
1420 
1421 			case KERN_PROC_TTY:
1422 				if ((p->p_flag & P_CONTROLT) == 0 ||
1423 				    p->p_session == NULL) {
1424 					PROC_UNLOCK(p);
1425 					continue;
1426 				}
1427 				/* XXX proctree_lock */
1428 				SESS_LOCK(p->p_session);
1429 				if (p->p_session->s_ttyp == NULL ||
1430 				    tty_udev(p->p_session->s_ttyp) !=
1431 				    (dev_t)name[0]) {
1432 					SESS_UNLOCK(p->p_session);
1433 					PROC_UNLOCK(p);
1434 					continue;
1435 				}
1436 				SESS_UNLOCK(p->p_session);
1437 				break;
1438 
1439 			case KERN_PROC_UID:
1440 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1441 					PROC_UNLOCK(p);
1442 					continue;
1443 				}
1444 				break;
1445 
1446 			case KERN_PROC_RUID:
1447 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1448 					PROC_UNLOCK(p);
1449 					continue;
1450 				}
1451 				break;
1452 
1453 			case KERN_PROC_PROC:
1454 				break;
1455 
1456 			default:
1457 				break;
1458 
1459 			}
1460 
1461 			error = sysctl_out_proc(p, req, flags, doingzomb);
1462 			if (error) {
1463 				sx_sunlock(&allproc_lock);
1464 				sx_sunlock(&proctree_lock);
1465 				return (error);
1466 			}
1467 		}
1468 	}
1469 	sx_sunlock(&allproc_lock);
1470 	sx_sunlock(&proctree_lock);
1471 	return (0);
1472 }
1473 
1474 struct pargs *
1475 pargs_alloc(int len)
1476 {
1477 	struct pargs *pa;
1478 
1479 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1480 		M_WAITOK);
1481 	refcount_init(&pa->ar_ref, 1);
1482 	pa->ar_length = len;
1483 	return (pa);
1484 }
1485 
1486 static void
1487 pargs_free(struct pargs *pa)
1488 {
1489 
1490 	free(pa, M_PARGS);
1491 }
1492 
1493 void
1494 pargs_hold(struct pargs *pa)
1495 {
1496 
1497 	if (pa == NULL)
1498 		return;
1499 	refcount_acquire(&pa->ar_ref);
1500 }
1501 
1502 void
1503 pargs_drop(struct pargs *pa)
1504 {
1505 
1506 	if (pa == NULL)
1507 		return;
1508 	if (refcount_release(&pa->ar_ref))
1509 		pargs_free(pa);
1510 }
1511 
1512 static int
1513 proc_read_mem(struct thread *td, struct proc *p, vm_offset_t offset, void* buf,
1514     size_t len)
1515 {
1516 	struct iovec iov;
1517 	struct uio uio;
1518 
1519 	iov.iov_base = (caddr_t)buf;
1520 	iov.iov_len = len;
1521 	uio.uio_iov = &iov;
1522 	uio.uio_iovcnt = 1;
1523 	uio.uio_offset = offset;
1524 	uio.uio_resid = (ssize_t)len;
1525 	uio.uio_segflg = UIO_SYSSPACE;
1526 	uio.uio_rw = UIO_READ;
1527 	uio.uio_td = td;
1528 
1529 	return (proc_rwmem(p, &uio));
1530 }
1531 
1532 static int
1533 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1534     size_t len)
1535 {
1536 	size_t i;
1537 	int error;
1538 
1539 	error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, len);
1540 	/*
1541 	 * Reading the chunk may validly return EFAULT if the string is shorter
1542 	 * than the chunk and is aligned at the end of the page, assuming the
1543 	 * next page is not mapped.  So if EFAULT is returned do a fallback to
1544 	 * one byte read loop.
1545 	 */
1546 	if (error == EFAULT) {
1547 		for (i = 0; i < len; i++, buf++, sptr++) {
1548 			error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, 1);
1549 			if (error != 0)
1550 				return (error);
1551 			if (*buf == '\0')
1552 				break;
1553 		}
1554 		error = 0;
1555 	}
1556 	return (error);
1557 }
1558 
1559 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1560 
1561 enum proc_vector_type {
1562 	PROC_ARG,
1563 	PROC_ENV,
1564 	PROC_AUX,
1565 };
1566 
1567 #ifdef COMPAT_FREEBSD32
1568 static int
1569 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1570     size_t *vsizep, enum proc_vector_type type)
1571 {
1572 	struct freebsd32_ps_strings pss;
1573 	Elf32_Auxinfo aux;
1574 	vm_offset_t vptr, ptr;
1575 	uint32_t *proc_vector32;
1576 	char **proc_vector;
1577 	size_t vsize, size;
1578 	int i, error;
1579 
1580 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1581 	    &pss, sizeof(pss));
1582 	if (error != 0)
1583 		return (error);
1584 	switch (type) {
1585 	case PROC_ARG:
1586 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1587 		vsize = pss.ps_nargvstr;
1588 		if (vsize > ARG_MAX)
1589 			return (ENOEXEC);
1590 		size = vsize * sizeof(int32_t);
1591 		break;
1592 	case PROC_ENV:
1593 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1594 		vsize = pss.ps_nenvstr;
1595 		if (vsize > ARG_MAX)
1596 			return (ENOEXEC);
1597 		size = vsize * sizeof(int32_t);
1598 		break;
1599 	case PROC_AUX:
1600 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1601 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1602 		if (vptr % 4 != 0)
1603 			return (ENOEXEC);
1604 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1605 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1606 			if (error != 0)
1607 				return (error);
1608 			if (aux.a_type == AT_NULL)
1609 				break;
1610 			ptr += sizeof(aux);
1611 		}
1612 		if (aux.a_type != AT_NULL)
1613 			return (ENOEXEC);
1614 		vsize = i + 1;
1615 		size = vsize * sizeof(aux);
1616 		break;
1617 	default:
1618 		KASSERT(0, ("Wrong proc vector type: %d", type));
1619 		return (EINVAL);
1620 	}
1621 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1622 	error = proc_read_mem(td, p, vptr, proc_vector32, size);
1623 	if (error != 0)
1624 		goto done;
1625 	if (type == PROC_AUX) {
1626 		*proc_vectorp = (char **)proc_vector32;
1627 		*vsizep = vsize;
1628 		return (0);
1629 	}
1630 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1631 	for (i = 0; i < (int)vsize; i++)
1632 		proc_vector[i] = PTRIN(proc_vector32[i]);
1633 	*proc_vectorp = proc_vector;
1634 	*vsizep = vsize;
1635 done:
1636 	free(proc_vector32, M_TEMP);
1637 	return (error);
1638 }
1639 #endif
1640 
1641 static int
1642 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1643     size_t *vsizep, enum proc_vector_type type)
1644 {
1645 	struct ps_strings pss;
1646 	Elf_Auxinfo aux;
1647 	vm_offset_t vptr, ptr;
1648 	char **proc_vector;
1649 	size_t vsize, size;
1650 	int error, i;
1651 
1652 #ifdef COMPAT_FREEBSD32
1653 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1654 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1655 #endif
1656 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1657 	    &pss, sizeof(pss));
1658 	if (error != 0)
1659 		return (error);
1660 	switch (type) {
1661 	case PROC_ARG:
1662 		vptr = (vm_offset_t)pss.ps_argvstr;
1663 		vsize = pss.ps_nargvstr;
1664 		if (vsize > ARG_MAX)
1665 			return (ENOEXEC);
1666 		size = vsize * sizeof(char *);
1667 		break;
1668 	case PROC_ENV:
1669 		vptr = (vm_offset_t)pss.ps_envstr;
1670 		vsize = pss.ps_nenvstr;
1671 		if (vsize > ARG_MAX)
1672 			return (ENOEXEC);
1673 		size = vsize * sizeof(char *);
1674 		break;
1675 	case PROC_AUX:
1676 		/*
1677 		 * The aux array is just above env array on the stack. Check
1678 		 * that the address is naturally aligned.
1679 		 */
1680 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1681 		    * sizeof(char *);
1682 #if __ELF_WORD_SIZE == 64
1683 		if (vptr % sizeof(uint64_t) != 0)
1684 #else
1685 		if (vptr % sizeof(uint32_t) != 0)
1686 #endif
1687 			return (ENOEXEC);
1688 		/*
1689 		 * We count the array size reading the aux vectors from the
1690 		 * stack until AT_NULL vector is returned.  So (to keep the code
1691 		 * simple) we read the process stack twice: the first time here
1692 		 * to find the size and the second time when copying the vectors
1693 		 * to the allocated proc_vector.
1694 		 */
1695 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1696 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1697 			if (error != 0)
1698 				return (error);
1699 			if (aux.a_type == AT_NULL)
1700 				break;
1701 			ptr += sizeof(aux);
1702 		}
1703 		/*
1704 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1705 		 * not reached AT_NULL, it is most likely we are reading wrong
1706 		 * data: either the process doesn't have auxv array or data has
1707 		 * been modified. Return the error in this case.
1708 		 */
1709 		if (aux.a_type != AT_NULL)
1710 			return (ENOEXEC);
1711 		vsize = i + 1;
1712 		size = vsize * sizeof(aux);
1713 		break;
1714 	default:
1715 		KASSERT(0, ("Wrong proc vector type: %d", type));
1716 		return (EINVAL); /* In case we are built without INVARIANTS. */
1717 	}
1718 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1719 	if (proc_vector == NULL)
1720 		return (ENOMEM);
1721 	error = proc_read_mem(td, p, vptr, proc_vector, size);
1722 	if (error != 0) {
1723 		free(proc_vector, M_TEMP);
1724 		return (error);
1725 	}
1726 	*proc_vectorp = proc_vector;
1727 	*vsizep = vsize;
1728 
1729 	return (0);
1730 }
1731 
1732 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1733 
1734 static int
1735 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1736     enum proc_vector_type type)
1737 {
1738 	size_t done, len, nchr, vsize;
1739 	int error, i;
1740 	char **proc_vector, *sptr;
1741 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1742 
1743 	PROC_ASSERT_HELD(p);
1744 
1745 	/*
1746 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1747 	 */
1748 	nchr = 2 * (PATH_MAX + ARG_MAX);
1749 
1750 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1751 	if (error != 0)
1752 		return (error);
1753 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1754 		/*
1755 		 * The program may have scribbled into its argv array, e.g. to
1756 		 * remove some arguments.  If that has happened, break out
1757 		 * before trying to read from NULL.
1758 		 */
1759 		if (proc_vector[i] == NULL)
1760 			break;
1761 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1762 			error = proc_read_string(td, p, sptr, pss_string,
1763 			    sizeof(pss_string));
1764 			if (error != 0)
1765 				goto done;
1766 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1767 			if (done + len >= nchr)
1768 				len = nchr - done - 1;
1769 			sbuf_bcat(sb, pss_string, len);
1770 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1771 				break;
1772 			done += GET_PS_STRINGS_CHUNK_SZ;
1773 		}
1774 		sbuf_bcat(sb, "", 1);
1775 		done += len + 1;
1776 	}
1777 done:
1778 	free(proc_vector, M_TEMP);
1779 	return (error);
1780 }
1781 
1782 int
1783 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1784 {
1785 
1786 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1787 }
1788 
1789 int
1790 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1791 {
1792 
1793 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1794 }
1795 
1796 int
1797 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1798 {
1799 	size_t vsize, size;
1800 	char **auxv;
1801 	int error;
1802 
1803 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1804 	if (error == 0) {
1805 #ifdef COMPAT_FREEBSD32
1806 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1807 			size = vsize * sizeof(Elf32_Auxinfo);
1808 		else
1809 #endif
1810 			size = vsize * sizeof(Elf_Auxinfo);
1811 		if (sbuf_bcat(sb, auxv, size) != 0)
1812 			error = ENOMEM;
1813 		free(auxv, M_TEMP);
1814 	}
1815 	return (error);
1816 }
1817 
1818 /*
1819  * This sysctl allows a process to retrieve the argument list or process
1820  * title for another process without groping around in the address space
1821  * of the other process.  It also allow a process to set its own "process
1822  * title to a string of its own choice.
1823  */
1824 static int
1825 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1826 {
1827 	int *name = (int *)arg1;
1828 	u_int namelen = arg2;
1829 	struct pargs *newpa, *pa;
1830 	struct proc *p;
1831 	struct sbuf sb;
1832 	int flags, error = 0, error2;
1833 
1834 	if (namelen != 1)
1835 		return (EINVAL);
1836 
1837 	flags = PGET_CANSEE;
1838 	if (req->newptr != NULL)
1839 		flags |= PGET_ISCURRENT;
1840 	error = pget((pid_t)name[0], flags, &p);
1841 	if (error)
1842 		return (error);
1843 
1844 	pa = p->p_args;
1845 	if (pa != NULL) {
1846 		pargs_hold(pa);
1847 		PROC_UNLOCK(p);
1848 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1849 		pargs_drop(pa);
1850 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1851 		_PHOLD(p);
1852 		PROC_UNLOCK(p);
1853 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1854 		error = proc_getargv(curthread, p, &sb);
1855 		error2 = sbuf_finish(&sb);
1856 		PRELE(p);
1857 		sbuf_delete(&sb);
1858 		if (error == 0 && error2 != 0)
1859 			error = error2;
1860 	} else {
1861 		PROC_UNLOCK(p);
1862 	}
1863 	if (error != 0 || req->newptr == NULL)
1864 		return (error);
1865 
1866 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1867 		return (ENOMEM);
1868 	newpa = pargs_alloc(req->newlen);
1869 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1870 	if (error != 0) {
1871 		pargs_free(newpa);
1872 		return (error);
1873 	}
1874 	PROC_LOCK(p);
1875 	pa = p->p_args;
1876 	p->p_args = newpa;
1877 	PROC_UNLOCK(p);
1878 	pargs_drop(pa);
1879 	return (0);
1880 }
1881 
1882 /*
1883  * This sysctl allows a process to retrieve environment of another process.
1884  */
1885 static int
1886 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
1887 {
1888 	int *name = (int *)arg1;
1889 	u_int namelen = arg2;
1890 	struct proc *p;
1891 	struct sbuf sb;
1892 	int error, error2;
1893 
1894 	if (namelen != 1)
1895 		return (EINVAL);
1896 
1897 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1898 	if (error != 0)
1899 		return (error);
1900 	if ((p->p_flag & P_SYSTEM) != 0) {
1901 		PRELE(p);
1902 		return (0);
1903 	}
1904 
1905 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1906 	error = proc_getenvv(curthread, p, &sb);
1907 	error2 = sbuf_finish(&sb);
1908 	PRELE(p);
1909 	sbuf_delete(&sb);
1910 	return (error != 0 ? error : error2);
1911 }
1912 
1913 /*
1914  * This sysctl allows a process to retrieve ELF auxiliary vector of
1915  * another process.
1916  */
1917 static int
1918 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
1919 {
1920 	int *name = (int *)arg1;
1921 	u_int namelen = arg2;
1922 	struct proc *p;
1923 	struct sbuf sb;
1924 	int error, error2;
1925 
1926 	if (namelen != 1)
1927 		return (EINVAL);
1928 
1929 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1930 	if (error != 0)
1931 		return (error);
1932 	if ((p->p_flag & P_SYSTEM) != 0) {
1933 		PRELE(p);
1934 		return (0);
1935 	}
1936 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1937 	error = proc_getauxv(curthread, p, &sb);
1938 	error2 = sbuf_finish(&sb);
1939 	PRELE(p);
1940 	sbuf_delete(&sb);
1941 	return (error != 0 ? error : error2);
1942 }
1943 
1944 /*
1945  * This sysctl allows a process to retrieve the path of the executable for
1946  * itself or another process.
1947  */
1948 static int
1949 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1950 {
1951 	pid_t *pidp = (pid_t *)arg1;
1952 	unsigned int arglen = arg2;
1953 	struct proc *p;
1954 	struct vnode *vp;
1955 	char *retbuf, *freebuf;
1956 	int error;
1957 
1958 	if (arglen != 1)
1959 		return (EINVAL);
1960 	if (*pidp == -1) {	/* -1 means this process */
1961 		p = req->td->td_proc;
1962 	} else {
1963 		error = pget(*pidp, PGET_CANSEE, &p);
1964 		if (error != 0)
1965 			return (error);
1966 	}
1967 
1968 	vp = p->p_textvp;
1969 	if (vp == NULL) {
1970 		if (*pidp != -1)
1971 			PROC_UNLOCK(p);
1972 		return (0);
1973 	}
1974 	vref(vp);
1975 	if (*pidp != -1)
1976 		PROC_UNLOCK(p);
1977 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1978 	vrele(vp);
1979 	if (error)
1980 		return (error);
1981 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1982 	free(freebuf, M_TEMP);
1983 	return (error);
1984 }
1985 
1986 static int
1987 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1988 {
1989 	struct proc *p;
1990 	char *sv_name;
1991 	int *name;
1992 	int namelen;
1993 	int error;
1994 
1995 	namelen = arg2;
1996 	if (namelen != 1)
1997 		return (EINVAL);
1998 
1999 	name = (int *)arg1;
2000 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2001 	if (error != 0)
2002 		return (error);
2003 	sv_name = p->p_sysent->sv_name;
2004 	PROC_UNLOCK(p);
2005 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2006 }
2007 
2008 #ifdef KINFO_OVMENTRY_SIZE
2009 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2010 #endif
2011 
2012 #ifdef COMPAT_FREEBSD7
2013 static int
2014 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2015 {
2016 	vm_map_entry_t entry, tmp_entry;
2017 	unsigned int last_timestamp;
2018 	char *fullpath, *freepath;
2019 	struct kinfo_ovmentry *kve;
2020 	struct vattr va;
2021 	struct ucred *cred;
2022 	int error, *name;
2023 	struct vnode *vp;
2024 	struct proc *p;
2025 	vm_map_t map;
2026 	struct vmspace *vm;
2027 
2028 	name = (int *)arg1;
2029 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2030 	if (error != 0)
2031 		return (error);
2032 	vm = vmspace_acquire_ref(p);
2033 	if (vm == NULL) {
2034 		PRELE(p);
2035 		return (ESRCH);
2036 	}
2037 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2038 
2039 	map = &vm->vm_map;
2040 	vm_map_lock_read(map);
2041 	for (entry = map->header.next; entry != &map->header;
2042 	    entry = entry->next) {
2043 		vm_object_t obj, tobj, lobj;
2044 		vm_offset_t addr;
2045 
2046 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2047 			continue;
2048 
2049 		bzero(kve, sizeof(*kve));
2050 		kve->kve_structsize = sizeof(*kve);
2051 
2052 		kve->kve_private_resident = 0;
2053 		obj = entry->object.vm_object;
2054 		if (obj != NULL) {
2055 			VM_OBJECT_RLOCK(obj);
2056 			if (obj->shadow_count == 1)
2057 				kve->kve_private_resident =
2058 				    obj->resident_page_count;
2059 		}
2060 		kve->kve_resident = 0;
2061 		addr = entry->start;
2062 		while (addr < entry->end) {
2063 			if (pmap_extract(map->pmap, addr))
2064 				kve->kve_resident++;
2065 			addr += PAGE_SIZE;
2066 		}
2067 
2068 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2069 			if (tobj != obj)
2070 				VM_OBJECT_RLOCK(tobj);
2071 			if (lobj != obj)
2072 				VM_OBJECT_RUNLOCK(lobj);
2073 			lobj = tobj;
2074 		}
2075 
2076 		kve->kve_start = (void*)entry->start;
2077 		kve->kve_end = (void*)entry->end;
2078 		kve->kve_offset = (off_t)entry->offset;
2079 
2080 		if (entry->protection & VM_PROT_READ)
2081 			kve->kve_protection |= KVME_PROT_READ;
2082 		if (entry->protection & VM_PROT_WRITE)
2083 			kve->kve_protection |= KVME_PROT_WRITE;
2084 		if (entry->protection & VM_PROT_EXECUTE)
2085 			kve->kve_protection |= KVME_PROT_EXEC;
2086 
2087 		if (entry->eflags & MAP_ENTRY_COW)
2088 			kve->kve_flags |= KVME_FLAG_COW;
2089 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2090 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2091 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2092 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2093 
2094 		last_timestamp = map->timestamp;
2095 		vm_map_unlock_read(map);
2096 
2097 		kve->kve_fileid = 0;
2098 		kve->kve_fsid = 0;
2099 		freepath = NULL;
2100 		fullpath = "";
2101 		if (lobj) {
2102 			vp = NULL;
2103 			switch (lobj->type) {
2104 			case OBJT_DEFAULT:
2105 				kve->kve_type = KVME_TYPE_DEFAULT;
2106 				break;
2107 			case OBJT_VNODE:
2108 				kve->kve_type = KVME_TYPE_VNODE;
2109 				vp = lobj->handle;
2110 				vref(vp);
2111 				break;
2112 			case OBJT_SWAP:
2113 				kve->kve_type = KVME_TYPE_SWAP;
2114 				break;
2115 			case OBJT_DEVICE:
2116 				kve->kve_type = KVME_TYPE_DEVICE;
2117 				break;
2118 			case OBJT_PHYS:
2119 				kve->kve_type = KVME_TYPE_PHYS;
2120 				break;
2121 			case OBJT_DEAD:
2122 				kve->kve_type = KVME_TYPE_DEAD;
2123 				break;
2124 			case OBJT_SG:
2125 				kve->kve_type = KVME_TYPE_SG;
2126 				break;
2127 			default:
2128 				kve->kve_type = KVME_TYPE_UNKNOWN;
2129 				break;
2130 			}
2131 			if (lobj != obj)
2132 				VM_OBJECT_RUNLOCK(lobj);
2133 
2134 			kve->kve_ref_count = obj->ref_count;
2135 			kve->kve_shadow_count = obj->shadow_count;
2136 			VM_OBJECT_RUNLOCK(obj);
2137 			if (vp != NULL) {
2138 				vn_fullpath(curthread, vp, &fullpath,
2139 				    &freepath);
2140 				cred = curthread->td_ucred;
2141 				vn_lock(vp, LK_SHARED | LK_RETRY);
2142 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2143 					kve->kve_fileid = va.va_fileid;
2144 					kve->kve_fsid = va.va_fsid;
2145 				}
2146 				vput(vp);
2147 			}
2148 		} else {
2149 			kve->kve_type = KVME_TYPE_NONE;
2150 			kve->kve_ref_count = 0;
2151 			kve->kve_shadow_count = 0;
2152 		}
2153 
2154 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2155 		if (freepath != NULL)
2156 			free(freepath, M_TEMP);
2157 
2158 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2159 		vm_map_lock_read(map);
2160 		if (error)
2161 			break;
2162 		if (last_timestamp != map->timestamp) {
2163 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2164 			entry = tmp_entry;
2165 		}
2166 	}
2167 	vm_map_unlock_read(map);
2168 	vmspace_free(vm);
2169 	PRELE(p);
2170 	free(kve, M_TEMP);
2171 	return (error);
2172 }
2173 #endif	/* COMPAT_FREEBSD7 */
2174 
2175 #ifdef KINFO_VMENTRY_SIZE
2176 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2177 #endif
2178 
2179 static void
2180 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2181     struct kinfo_vmentry *kve)
2182 {
2183 	vm_object_t obj, tobj;
2184 	vm_page_t m, m_adv;
2185 	vm_offset_t addr;
2186 	vm_paddr_t locked_pa;
2187 	vm_pindex_t pi, pi_adv, pindex;
2188 
2189 	locked_pa = 0;
2190 	obj = entry->object.vm_object;
2191 	addr = entry->start;
2192 	m_adv = NULL;
2193 	pi = OFF_TO_IDX(entry->offset);
2194 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2195 		if (m_adv != NULL) {
2196 			m = m_adv;
2197 		} else {
2198 			pi_adv = OFF_TO_IDX(entry->end - addr);
2199 			pindex = pi;
2200 			for (tobj = obj;; tobj = tobj->backing_object) {
2201 				m = vm_page_find_least(tobj, pindex);
2202 				if (m != NULL) {
2203 					if (m->pindex == pindex)
2204 						break;
2205 					if (pi_adv > m->pindex - pindex) {
2206 						pi_adv = m->pindex - pindex;
2207 						m_adv = m;
2208 					}
2209 				}
2210 				if (tobj->backing_object == NULL)
2211 					goto next;
2212 				pindex += OFF_TO_IDX(tobj->
2213 				    backing_object_offset);
2214 			}
2215 		}
2216 		m_adv = NULL;
2217 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2218 		    (addr & (pagesizes[1] - 1)) == 0 &&
2219 		    (pmap_mincore(map->pmap, addr, &locked_pa) &
2220 		    MINCORE_SUPER) != 0) {
2221 			kve->kve_flags |= KVME_FLAG_SUPER;
2222 			pi_adv = OFF_TO_IDX(pagesizes[1]);
2223 		} else {
2224 			/*
2225 			 * We do not test the found page on validity.
2226 			 * Either the page is busy and being paged in,
2227 			 * or it was invalidated.  The first case
2228 			 * should be counted as resident, the second
2229 			 * is not so clear; we do account both.
2230 			 */
2231 			pi_adv = 1;
2232 		}
2233 		kve->kve_resident += pi_adv;
2234 next:;
2235 	}
2236 	PA_UNLOCK_COND(locked_pa);
2237 }
2238 
2239 /*
2240  * Must be called with the process locked and will return unlocked.
2241  */
2242 int
2243 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb)
2244 {
2245 	vm_map_entry_t entry, tmp_entry;
2246 	struct vattr va;
2247 	vm_map_t map;
2248 	vm_object_t obj, tobj, lobj;
2249 	char *fullpath, *freepath;
2250 	struct kinfo_vmentry *kve;
2251 	struct ucred *cred;
2252 	struct vnode *vp;
2253 	struct vmspace *vm;
2254 	vm_offset_t addr;
2255 	unsigned int last_timestamp;
2256 	int error;
2257 
2258 	PROC_LOCK_ASSERT(p, MA_OWNED);
2259 
2260 	_PHOLD(p);
2261 	PROC_UNLOCK(p);
2262 	vm = vmspace_acquire_ref(p);
2263 	if (vm == NULL) {
2264 		PRELE(p);
2265 		return (ESRCH);
2266 	}
2267 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2268 
2269 	error = 0;
2270 	map = &vm->vm_map;
2271 	vm_map_lock_read(map);
2272 	for (entry = map->header.next; entry != &map->header;
2273 	    entry = entry->next) {
2274 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2275 			continue;
2276 
2277 		addr = entry->end;
2278 		bzero(kve, sizeof(*kve));
2279 		obj = entry->object.vm_object;
2280 		if (obj != NULL) {
2281 			for (tobj = obj; tobj != NULL;
2282 			    tobj = tobj->backing_object) {
2283 				VM_OBJECT_RLOCK(tobj);
2284 				lobj = tobj;
2285 			}
2286 			if (obj->backing_object == NULL)
2287 				kve->kve_private_resident =
2288 				    obj->resident_page_count;
2289 			if (!vmmap_skip_res_cnt)
2290 				kern_proc_vmmap_resident(map, entry, kve);
2291 			for (tobj = obj; tobj != NULL;
2292 			    tobj = tobj->backing_object) {
2293 				if (tobj != obj && tobj != lobj)
2294 					VM_OBJECT_RUNLOCK(tobj);
2295 			}
2296 		} else {
2297 			lobj = NULL;
2298 		}
2299 
2300 		kve->kve_start = entry->start;
2301 		kve->kve_end = entry->end;
2302 		kve->kve_offset = entry->offset;
2303 
2304 		if (entry->protection & VM_PROT_READ)
2305 			kve->kve_protection |= KVME_PROT_READ;
2306 		if (entry->protection & VM_PROT_WRITE)
2307 			kve->kve_protection |= KVME_PROT_WRITE;
2308 		if (entry->protection & VM_PROT_EXECUTE)
2309 			kve->kve_protection |= KVME_PROT_EXEC;
2310 
2311 		if (entry->eflags & MAP_ENTRY_COW)
2312 			kve->kve_flags |= KVME_FLAG_COW;
2313 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2314 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2315 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2316 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2317 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2318 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2319 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2320 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2321 
2322 		last_timestamp = map->timestamp;
2323 		vm_map_unlock_read(map);
2324 
2325 		freepath = NULL;
2326 		fullpath = "";
2327 		if (lobj != NULL) {
2328 			vp = NULL;
2329 			switch (lobj->type) {
2330 			case OBJT_DEFAULT:
2331 				kve->kve_type = KVME_TYPE_DEFAULT;
2332 				break;
2333 			case OBJT_VNODE:
2334 				kve->kve_type = KVME_TYPE_VNODE;
2335 				vp = lobj->handle;
2336 				vref(vp);
2337 				break;
2338 			case OBJT_SWAP:
2339 				kve->kve_type = KVME_TYPE_SWAP;
2340 				break;
2341 			case OBJT_DEVICE:
2342 				kve->kve_type = KVME_TYPE_DEVICE;
2343 				break;
2344 			case OBJT_PHYS:
2345 				kve->kve_type = KVME_TYPE_PHYS;
2346 				break;
2347 			case OBJT_DEAD:
2348 				kve->kve_type = KVME_TYPE_DEAD;
2349 				break;
2350 			case OBJT_SG:
2351 				kve->kve_type = KVME_TYPE_SG;
2352 				break;
2353 			case OBJT_MGTDEVICE:
2354 				kve->kve_type = KVME_TYPE_MGTDEVICE;
2355 				break;
2356 			default:
2357 				kve->kve_type = KVME_TYPE_UNKNOWN;
2358 				break;
2359 			}
2360 			if (lobj != obj)
2361 				VM_OBJECT_RUNLOCK(lobj);
2362 
2363 			kve->kve_ref_count = obj->ref_count;
2364 			kve->kve_shadow_count = obj->shadow_count;
2365 			VM_OBJECT_RUNLOCK(obj);
2366 			if (vp != NULL) {
2367 				vn_fullpath(curthread, vp, &fullpath,
2368 				    &freepath);
2369 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2370 				cred = curthread->td_ucred;
2371 				vn_lock(vp, LK_SHARED | LK_RETRY);
2372 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2373 					kve->kve_vn_fileid = va.va_fileid;
2374 					kve->kve_vn_fsid = va.va_fsid;
2375 					kve->kve_vn_mode =
2376 					    MAKEIMODE(va.va_type, va.va_mode);
2377 					kve->kve_vn_size = va.va_size;
2378 					kve->kve_vn_rdev = va.va_rdev;
2379 					kve->kve_status = KF_ATTR_VALID;
2380 				}
2381 				vput(vp);
2382 			}
2383 		} else {
2384 			kve->kve_type = KVME_TYPE_NONE;
2385 			kve->kve_ref_count = 0;
2386 			kve->kve_shadow_count = 0;
2387 		}
2388 
2389 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2390 		if (freepath != NULL)
2391 			free(freepath, M_TEMP);
2392 
2393 		/* Pack record size down */
2394 		kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) +
2395 		    strlen(kve->kve_path) + 1;
2396 		kve->kve_structsize = roundup(kve->kve_structsize,
2397 		    sizeof(uint64_t));
2398 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2399 			error = ENOMEM;
2400 		vm_map_lock_read(map);
2401 		if (error != 0)
2402 			break;
2403 		if (last_timestamp != map->timestamp) {
2404 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2405 			entry = tmp_entry;
2406 		}
2407 	}
2408 	vm_map_unlock_read(map);
2409 	vmspace_free(vm);
2410 	PRELE(p);
2411 	free(kve, M_TEMP);
2412 	return (error);
2413 }
2414 
2415 static int
2416 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2417 {
2418 	struct proc *p;
2419 	struct sbuf sb;
2420 	int error, error2, *name;
2421 
2422 	name = (int *)arg1;
2423 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2424 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2425 	if (error != 0) {
2426 		sbuf_delete(&sb);
2427 		return (error);
2428 	}
2429 	error = kern_proc_vmmap_out(p, &sb);
2430 	error2 = sbuf_finish(&sb);
2431 	sbuf_delete(&sb);
2432 	return (error != 0 ? error : error2);
2433 }
2434 
2435 #if defined(STACK) || defined(DDB)
2436 static int
2437 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2438 {
2439 	struct kinfo_kstack *kkstp;
2440 	int error, i, *name, numthreads;
2441 	lwpid_t *lwpidarray;
2442 	struct thread *td;
2443 	struct stack *st;
2444 	struct sbuf sb;
2445 	struct proc *p;
2446 
2447 	name = (int *)arg1;
2448 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2449 	if (error != 0)
2450 		return (error);
2451 
2452 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2453 	st = stack_create();
2454 
2455 	lwpidarray = NULL;
2456 	numthreads = 0;
2457 	PROC_LOCK(p);
2458 repeat:
2459 	if (numthreads < p->p_numthreads) {
2460 		if (lwpidarray != NULL) {
2461 			free(lwpidarray, M_TEMP);
2462 			lwpidarray = NULL;
2463 		}
2464 		numthreads = p->p_numthreads;
2465 		PROC_UNLOCK(p);
2466 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2467 		    M_WAITOK | M_ZERO);
2468 		PROC_LOCK(p);
2469 		goto repeat;
2470 	}
2471 	i = 0;
2472 
2473 	/*
2474 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2475 	 * of changes could have taken place.  Should we check to see if the
2476 	 * vmspace has been replaced, or the like, in order to prevent
2477 	 * giving a snapshot that spans, say, execve(2), with some threads
2478 	 * before and some after?  Among other things, the credentials could
2479 	 * have changed, in which case the right to extract debug info might
2480 	 * no longer be assured.
2481 	 */
2482 	FOREACH_THREAD_IN_PROC(p, td) {
2483 		KASSERT(i < numthreads,
2484 		    ("sysctl_kern_proc_kstack: numthreads"));
2485 		lwpidarray[i] = td->td_tid;
2486 		i++;
2487 	}
2488 	numthreads = i;
2489 	for (i = 0; i < numthreads; i++) {
2490 		td = thread_find(p, lwpidarray[i]);
2491 		if (td == NULL) {
2492 			continue;
2493 		}
2494 		bzero(kkstp, sizeof(*kkstp));
2495 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2496 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2497 		thread_lock(td);
2498 		kkstp->kkst_tid = td->td_tid;
2499 		if (TD_IS_SWAPPED(td))
2500 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2501 		else if (TD_IS_RUNNING(td))
2502 			kkstp->kkst_state = KKST_STATE_RUNNING;
2503 		else {
2504 			kkstp->kkst_state = KKST_STATE_STACKOK;
2505 			stack_save_td(st, td);
2506 		}
2507 		thread_unlock(td);
2508 		PROC_UNLOCK(p);
2509 		stack_sbuf_print(&sb, st);
2510 		sbuf_finish(&sb);
2511 		sbuf_delete(&sb);
2512 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2513 		PROC_LOCK(p);
2514 		if (error)
2515 			break;
2516 	}
2517 	_PRELE(p);
2518 	PROC_UNLOCK(p);
2519 	if (lwpidarray != NULL)
2520 		free(lwpidarray, M_TEMP);
2521 	stack_destroy(st);
2522 	free(kkstp, M_TEMP);
2523 	return (error);
2524 }
2525 #endif
2526 
2527 /*
2528  * This sysctl allows a process to retrieve the full list of groups from
2529  * itself or another process.
2530  */
2531 static int
2532 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2533 {
2534 	pid_t *pidp = (pid_t *)arg1;
2535 	unsigned int arglen = arg2;
2536 	struct proc *p;
2537 	struct ucred *cred;
2538 	int error;
2539 
2540 	if (arglen != 1)
2541 		return (EINVAL);
2542 	if (*pidp == -1) {	/* -1 means this process */
2543 		p = req->td->td_proc;
2544 		PROC_LOCK(p);
2545 	} else {
2546 		error = pget(*pidp, PGET_CANSEE, &p);
2547 		if (error != 0)
2548 			return (error);
2549 	}
2550 
2551 	cred = crhold(p->p_ucred);
2552 	PROC_UNLOCK(p);
2553 
2554 	error = SYSCTL_OUT(req, cred->cr_groups,
2555 	    cred->cr_ngroups * sizeof(gid_t));
2556 	crfree(cred);
2557 	return (error);
2558 }
2559 
2560 /*
2561  * This sysctl allows a process to retrieve or/and set the resource limit for
2562  * another process.
2563  */
2564 static int
2565 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2566 {
2567 	int *name = (int *)arg1;
2568 	u_int namelen = arg2;
2569 	struct rlimit rlim;
2570 	struct proc *p;
2571 	u_int which;
2572 	int flags, error;
2573 
2574 	if (namelen != 2)
2575 		return (EINVAL);
2576 
2577 	which = (u_int)name[1];
2578 	if (which >= RLIM_NLIMITS)
2579 		return (EINVAL);
2580 
2581 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2582 		return (EINVAL);
2583 
2584 	flags = PGET_HOLD | PGET_NOTWEXIT;
2585 	if (req->newptr != NULL)
2586 		flags |= PGET_CANDEBUG;
2587 	else
2588 		flags |= PGET_CANSEE;
2589 	error = pget((pid_t)name[0], flags, &p);
2590 	if (error != 0)
2591 		return (error);
2592 
2593 	/*
2594 	 * Retrieve limit.
2595 	 */
2596 	if (req->oldptr != NULL) {
2597 		PROC_LOCK(p);
2598 		lim_rlimit(p, which, &rlim);
2599 		PROC_UNLOCK(p);
2600 	}
2601 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2602 	if (error != 0)
2603 		goto errout;
2604 
2605 	/*
2606 	 * Set limit.
2607 	 */
2608 	if (req->newptr != NULL) {
2609 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2610 		if (error == 0)
2611 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2612 	}
2613 
2614 errout:
2615 	PRELE(p);
2616 	return (error);
2617 }
2618 
2619 /*
2620  * This sysctl allows a process to retrieve ps_strings structure location of
2621  * another process.
2622  */
2623 static int
2624 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2625 {
2626 	int *name = (int *)arg1;
2627 	u_int namelen = arg2;
2628 	struct proc *p;
2629 	vm_offset_t ps_strings;
2630 	int error;
2631 #ifdef COMPAT_FREEBSD32
2632 	uint32_t ps_strings32;
2633 #endif
2634 
2635 	if (namelen != 1)
2636 		return (EINVAL);
2637 
2638 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2639 	if (error != 0)
2640 		return (error);
2641 #ifdef COMPAT_FREEBSD32
2642 	if ((req->flags & SCTL_MASK32) != 0) {
2643 		/*
2644 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2645 		 * process.
2646 		 */
2647 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2648 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2649 		PROC_UNLOCK(p);
2650 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2651 		return (error);
2652 	}
2653 #endif
2654 	ps_strings = p->p_sysent->sv_psstrings;
2655 	PROC_UNLOCK(p);
2656 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2657 	return (error);
2658 }
2659 
2660 /*
2661  * This sysctl allows a process to retrieve umask of another process.
2662  */
2663 static int
2664 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2665 {
2666 	int *name = (int *)arg1;
2667 	u_int namelen = arg2;
2668 	struct proc *p;
2669 	int error;
2670 	u_short fd_cmask;
2671 
2672 	if (namelen != 1)
2673 		return (EINVAL);
2674 
2675 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2676 	if (error != 0)
2677 		return (error);
2678 
2679 	FILEDESC_SLOCK(p->p_fd);
2680 	fd_cmask = p->p_fd->fd_cmask;
2681 	FILEDESC_SUNLOCK(p->p_fd);
2682 	PRELE(p);
2683 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2684 	return (error);
2685 }
2686 
2687 /*
2688  * This sysctl allows a process to set and retrieve binary osreldate of
2689  * another process.
2690  */
2691 static int
2692 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2693 {
2694 	int *name = (int *)arg1;
2695 	u_int namelen = arg2;
2696 	struct proc *p;
2697 	int flags, error, osrel;
2698 
2699 	if (namelen != 1)
2700 		return (EINVAL);
2701 
2702 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2703 		return (EINVAL);
2704 
2705 	flags = PGET_HOLD | PGET_NOTWEXIT;
2706 	if (req->newptr != NULL)
2707 		flags |= PGET_CANDEBUG;
2708 	else
2709 		flags |= PGET_CANSEE;
2710 	error = pget((pid_t)name[0], flags, &p);
2711 	if (error != 0)
2712 		return (error);
2713 
2714 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2715 	if (error != 0)
2716 		goto errout;
2717 
2718 	if (req->newptr != NULL) {
2719 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2720 		if (error != 0)
2721 			goto errout;
2722 		if (osrel < 0) {
2723 			error = EINVAL;
2724 			goto errout;
2725 		}
2726 		p->p_osrel = osrel;
2727 	}
2728 errout:
2729 	PRELE(p);
2730 	return (error);
2731 }
2732 
2733 static int
2734 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2735 {
2736 	int *name = (int *)arg1;
2737 	u_int namelen = arg2;
2738 	struct proc *p;
2739 	struct kinfo_sigtramp kst;
2740 	const struct sysentvec *sv;
2741 	int error;
2742 #ifdef COMPAT_FREEBSD32
2743 	struct kinfo_sigtramp32 kst32;
2744 #endif
2745 
2746 	if (namelen != 1)
2747 		return (EINVAL);
2748 
2749 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2750 	if (error != 0)
2751 		return (error);
2752 	sv = p->p_sysent;
2753 #ifdef COMPAT_FREEBSD32
2754 	if ((req->flags & SCTL_MASK32) != 0) {
2755 		bzero(&kst32, sizeof(kst32));
2756 		if (SV_PROC_FLAG(p, SV_ILP32)) {
2757 			if (sv->sv_sigcode_base != 0) {
2758 				kst32.ksigtramp_start = sv->sv_sigcode_base;
2759 				kst32.ksigtramp_end = sv->sv_sigcode_base +
2760 				    *sv->sv_szsigcode;
2761 			} else {
2762 				kst32.ksigtramp_start = sv->sv_psstrings -
2763 				    *sv->sv_szsigcode;
2764 				kst32.ksigtramp_end = sv->sv_psstrings;
2765 			}
2766 		}
2767 		PROC_UNLOCK(p);
2768 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2769 		return (error);
2770 	}
2771 #endif
2772 	bzero(&kst, sizeof(kst));
2773 	if (sv->sv_sigcode_base != 0) {
2774 		kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2775 		kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2776 		    *sv->sv_szsigcode;
2777 	} else {
2778 		kst.ksigtramp_start = (char *)sv->sv_psstrings -
2779 		    *sv->sv_szsigcode;
2780 		kst.ksigtramp_end = (char *)sv->sv_psstrings;
2781 	}
2782 	PROC_UNLOCK(p);
2783 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
2784 	return (error);
2785 }
2786 
2787 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2788 
2789 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2790 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2791 	"Return entire process table");
2792 
2793 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2794 	sysctl_kern_proc, "Process table");
2795 
2796 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2797 	sysctl_kern_proc, "Process table");
2798 
2799 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2800 	sysctl_kern_proc, "Process table");
2801 
2802 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2803 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2804 
2805 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2806 	sysctl_kern_proc, "Process table");
2807 
2808 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2809 	sysctl_kern_proc, "Process table");
2810 
2811 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2812 	sysctl_kern_proc, "Process table");
2813 
2814 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2815 	sysctl_kern_proc, "Process table");
2816 
2817 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2818 	sysctl_kern_proc, "Return process table, no threads");
2819 
2820 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2821 	CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2822 	sysctl_kern_proc_args, "Process argument list");
2823 
2824 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
2825 	sysctl_kern_proc_env, "Process environment");
2826 
2827 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
2828 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
2829 
2830 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2831 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2832 
2833 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2834 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2835 	"Process syscall vector name (ABI type)");
2836 
2837 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2838 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2839 
2840 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2841 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2842 
2843 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2844 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2845 
2846 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2847 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2848 
2849 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2850 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2851 
2852 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2853 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2854 
2855 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2856 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2857 
2858 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2859 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2860 
2861 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2862 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2863 	"Return process table, no threads");
2864 
2865 #ifdef COMPAT_FREEBSD7
2866 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2867 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2868 #endif
2869 
2870 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2871 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2872 
2873 #if defined(STACK) || defined(DDB)
2874 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2875 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2876 #endif
2877 
2878 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2879 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
2880 
2881 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
2882 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
2883 	"Process resource limits");
2884 
2885 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
2886 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
2887 	"Process ps_strings location");
2888 
2889 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
2890 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
2891 
2892 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
2893 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
2894 	"Process binary osreldate");
2895 
2896 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
2897 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
2898 	"Process signal trampoline location");
2899 
2900 int allproc_gen;
2901 
2902 void
2903 stop_all_proc(void)
2904 {
2905 	struct proc *cp, *p;
2906 	int r, gen;
2907 	bool restart, seen_stopped, seen_exiting, stopped_some;
2908 
2909 	cp = curproc;
2910 	/*
2911 	 * stop_all_proc() assumes that all process which have
2912 	 * usermode must be stopped, except current process, for
2913 	 * obvious reasons.  Since other threads in the process
2914 	 * establishing global stop could unstop something, disable
2915 	 * calls from multithreaded processes as precaution.  The
2916 	 * service must not be user-callable anyway.
2917 	 */
2918 	KASSERT((cp->p_flag & P_HADTHREADS) == 0 ||
2919 	    (cp->p_flag & P_KTHREAD) != 0, ("mt stop_all_proc"));
2920 
2921 allproc_loop:
2922 	sx_xlock(&allproc_lock);
2923 	gen = allproc_gen;
2924 	seen_exiting = seen_stopped = stopped_some = restart = false;
2925 	LIST_REMOVE(cp, p_list);
2926 	LIST_INSERT_HEAD(&allproc, cp, p_list);
2927 	for (;;) {
2928 		p = LIST_NEXT(cp, p_list);
2929 		if (p == NULL)
2930 			break;
2931 		LIST_REMOVE(cp, p_list);
2932 		LIST_INSERT_AFTER(p, cp, p_list);
2933 		PROC_LOCK(p);
2934 		if ((p->p_flag & (P_KTHREAD | P_SYSTEM |
2935 		    P_TOTAL_STOP)) != 0) {
2936 			PROC_UNLOCK(p);
2937 			continue;
2938 		}
2939 		if ((p->p_flag & P_WEXIT) != 0) {
2940 			seen_exiting = true;
2941 			PROC_UNLOCK(p);
2942 			continue;
2943 		}
2944 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
2945 			/*
2946 			 * Stopped processes are tolerated when there
2947 			 * are no other processes which might continue
2948 			 * them.  P_STOPPED_SINGLE but not
2949 			 * P_TOTAL_STOP process still has at least one
2950 			 * thread running.
2951 			 */
2952 			seen_stopped = true;
2953 			PROC_UNLOCK(p);
2954 			continue;
2955 		}
2956 		_PHOLD(p);
2957 		sx_xunlock(&allproc_lock);
2958 		r = thread_single(p, SINGLE_ALLPROC);
2959 		if (r != 0)
2960 			restart = true;
2961 		else
2962 			stopped_some = true;
2963 		_PRELE(p);
2964 		PROC_UNLOCK(p);
2965 		sx_xlock(&allproc_lock);
2966 	}
2967 	/* Catch forked children we did not see in iteration. */
2968 	if (gen != allproc_gen)
2969 		restart = true;
2970 	sx_xunlock(&allproc_lock);
2971 	if (restart || stopped_some || seen_exiting || seen_stopped) {
2972 		kern_yield(PRI_USER);
2973 		goto allproc_loop;
2974 	}
2975 }
2976 
2977 void
2978 resume_all_proc(void)
2979 {
2980 	struct proc *cp, *p;
2981 
2982 	cp = curproc;
2983 	sx_xlock(&allproc_lock);
2984 	LIST_REMOVE(cp, p_list);
2985 	LIST_INSERT_HEAD(&allproc, cp, p_list);
2986 	for (;;) {
2987 		p = LIST_NEXT(cp, p_list);
2988 		if (p == NULL)
2989 			break;
2990 		LIST_REMOVE(cp, p_list);
2991 		LIST_INSERT_AFTER(p, cp, p_list);
2992 		PROC_LOCK(p);
2993 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
2994 			sx_xunlock(&allproc_lock);
2995 			_PHOLD(p);
2996 			thread_single_end(p, SINGLE_ALLPROC);
2997 			_PRELE(p);
2998 			PROC_UNLOCK(p);
2999 			sx_xlock(&allproc_lock);
3000 		} else {
3001 			PROC_UNLOCK(p);
3002 		}
3003 	}
3004 	sx_xunlock(&allproc_lock);
3005 }
3006 
3007 #define	TOTAL_STOP_DEBUG	1
3008 #ifdef TOTAL_STOP_DEBUG
3009 volatile static int ap_resume;
3010 #include <sys/mount.h>
3011 
3012 static int
3013 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3014 {
3015 	int error, val;
3016 
3017 	val = 0;
3018 	ap_resume = 0;
3019 	error = sysctl_handle_int(oidp, &val, 0, req);
3020 	if (error != 0 || req->newptr == NULL)
3021 		return (error);
3022 	if (val != 0) {
3023 		stop_all_proc();
3024 		syncer_suspend();
3025 		while (ap_resume == 0)
3026 			;
3027 		syncer_resume();
3028 		resume_all_proc();
3029 	}
3030 	return (0);
3031 }
3032 
3033 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3034     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3035     sysctl_debug_stop_all_proc, "I",
3036     "");
3037 #endif
3038