xref: /freebsd/sys/kern/kern_proc.c (revision 8d20be1e22095c27faf8fe8b2f0d089739cc742e)
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_kdtrace.h"
38 #include "opt_ktrace.h"
39 #include "opt_kstack_pages.h"
40 #include "opt_stack.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/elf.h>
45 #include <sys/exec.h>
46 #include <sys/kernel.h>
47 #include <sys/limits.h>
48 #include <sys/lock.h>
49 #include <sys/loginclass.h>
50 #include <sys/malloc.h>
51 #include <sys/mman.h>
52 #include <sys/mount.h>
53 #include <sys/mutex.h>
54 #include <sys/proc.h>
55 #include <sys/ptrace.h>
56 #include <sys/refcount.h>
57 #include <sys/resourcevar.h>
58 #include <sys/rwlock.h>
59 #include <sys/sbuf.h>
60 #include <sys/sysent.h>
61 #include <sys/sched.h>
62 #include <sys/smp.h>
63 #include <sys/stack.h>
64 #include <sys/stat.h>
65 #include <sys/sysctl.h>
66 #include <sys/filedesc.h>
67 #include <sys/tty.h>
68 #include <sys/signalvar.h>
69 #include <sys/sdt.h>
70 #include <sys/sx.h>
71 #include <sys/user.h>
72 #include <sys/jail.h>
73 #include <sys/vnode.h>
74 #include <sys/eventhandler.h>
75 
76 #ifdef DDB
77 #include <ddb/ddb.h>
78 #endif
79 
80 #include <vm/vm.h>
81 #include <vm/vm_param.h>
82 #include <vm/vm_extern.h>
83 #include <vm/pmap.h>
84 #include <vm/vm_map.h>
85 #include <vm/vm_object.h>
86 #include <vm/vm_page.h>
87 #include <vm/uma.h>
88 
89 #ifdef COMPAT_FREEBSD32
90 #include <compat/freebsd32/freebsd32.h>
91 #include <compat/freebsd32/freebsd32_util.h>
92 #endif
93 
94 SDT_PROVIDER_DEFINE(proc);
95 SDT_PROBE_DEFINE4(proc, kernel, ctor, entry, entry, "struct proc *", "int",
96     "void *", "int");
97 SDT_PROBE_DEFINE4(proc, kernel, ctor, return, return, "struct proc *", "int",
98     "void *", "int");
99 SDT_PROBE_DEFINE4(proc, kernel, dtor, entry, entry, "struct proc *", "int",
100     "void *", "struct thread *");
101 SDT_PROBE_DEFINE3(proc, kernel, dtor, return, return, "struct proc *", "int",
102     "void *");
103 SDT_PROBE_DEFINE3(proc, kernel, init, entry, entry, "struct proc *", "int",
104     "int");
105 SDT_PROBE_DEFINE3(proc, kernel, init, return, return, "struct proc *", "int",
106     "int");
107 
108 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
109 MALLOC_DEFINE(M_SESSION, "session", "session header");
110 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
111 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
112 
113 static void doenterpgrp(struct proc *, struct pgrp *);
114 static void orphanpg(struct pgrp *pg);
115 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
116 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
117 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
118     int preferthread);
119 static void pgadjustjobc(struct pgrp *pgrp, int entering);
120 static void pgdelete(struct pgrp *);
121 static int proc_ctor(void *mem, int size, void *arg, int flags);
122 static void proc_dtor(void *mem, int size, void *arg);
123 static int proc_init(void *mem, int size, int flags);
124 static void proc_fini(void *mem, int size);
125 static void pargs_free(struct pargs *pa);
126 static struct proc *zpfind_locked(pid_t pid);
127 
128 /*
129  * Other process lists
130  */
131 struct pidhashhead *pidhashtbl;
132 u_long pidhash;
133 struct pgrphashhead *pgrphashtbl;
134 u_long pgrphash;
135 struct proclist allproc;
136 struct proclist zombproc;
137 struct sx allproc_lock;
138 struct sx proctree_lock;
139 struct mtx ppeers_lock;
140 uma_zone_t proc_zone;
141 
142 int kstack_pages = KSTACK_PAGES;
143 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
144     "Kernel stack size in pages");
145 
146 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
147 #ifdef COMPAT_FREEBSD32
148 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
149 #endif
150 
151 /*
152  * Initialize global process hashing structures.
153  */
154 void
155 procinit()
156 {
157 
158 	sx_init(&allproc_lock, "allproc");
159 	sx_init(&proctree_lock, "proctree");
160 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
161 	LIST_INIT(&allproc);
162 	LIST_INIT(&zombproc);
163 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
164 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
165 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
166 	    proc_ctor, proc_dtor, proc_init, proc_fini,
167 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
168 	uihashinit();
169 }
170 
171 /*
172  * Prepare a proc for use.
173  */
174 static int
175 proc_ctor(void *mem, int size, void *arg, int flags)
176 {
177 	struct proc *p;
178 
179 	p = (struct proc *)mem;
180 	SDT_PROBE(proc, kernel, ctor , entry, p, size, arg, flags, 0);
181 	EVENTHANDLER_INVOKE(process_ctor, p);
182 	SDT_PROBE(proc, kernel, ctor , return, p, size, arg, flags, 0);
183 	return (0);
184 }
185 
186 /*
187  * Reclaim a proc after use.
188  */
189 static void
190 proc_dtor(void *mem, int size, void *arg)
191 {
192 	struct proc *p;
193 	struct thread *td;
194 
195 	/* INVARIANTS checks go here */
196 	p = (struct proc *)mem;
197 	td = FIRST_THREAD_IN_PROC(p);
198 	SDT_PROBE(proc, kernel, dtor, entry, p, size, arg, td, 0);
199 	if (td != NULL) {
200 #ifdef INVARIANTS
201 		KASSERT((p->p_numthreads == 1),
202 		    ("bad number of threads in exiting process"));
203 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
204 #endif
205 		/* Free all OSD associated to this thread. */
206 		osd_thread_exit(td);
207 	}
208 	EVENTHANDLER_INVOKE(process_dtor, p);
209 	if (p->p_ksi != NULL)
210 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
211 	SDT_PROBE(proc, kernel, dtor, return, p, size, arg, 0, 0);
212 }
213 
214 /*
215  * Initialize type-stable parts of a proc (when newly created).
216  */
217 static int
218 proc_init(void *mem, int size, int flags)
219 {
220 	struct proc *p;
221 
222 	p = (struct proc *)mem;
223 	SDT_PROBE(proc, kernel, init, entry, p, size, flags, 0, 0);
224 	p->p_sched = (struct p_sched *)&p[1];
225 	bzero(&p->p_mtx, sizeof(struct mtx));
226 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK);
227 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_RECURSE);
228 	cv_init(&p->p_pwait, "ppwait");
229 	cv_init(&p->p_dbgwait, "dbgwait");
230 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
231 	EVENTHANDLER_INVOKE(process_init, p);
232 	p->p_stats = pstats_alloc();
233 	SDT_PROBE(proc, kernel, init, return, p, size, flags, 0, 0);
234 	return (0);
235 }
236 
237 /*
238  * UMA should ensure that this function is never called.
239  * Freeing a proc structure would violate type stability.
240  */
241 static void
242 proc_fini(void *mem, int size)
243 {
244 #ifdef notnow
245 	struct proc *p;
246 
247 	p = (struct proc *)mem;
248 	EVENTHANDLER_INVOKE(process_fini, p);
249 	pstats_free(p->p_stats);
250 	thread_free(FIRST_THREAD_IN_PROC(p));
251 	mtx_destroy(&p->p_mtx);
252 	if (p->p_ksi != NULL)
253 		ksiginfo_free(p->p_ksi);
254 #else
255 	panic("proc reclaimed");
256 #endif
257 }
258 
259 /*
260  * Is p an inferior of the current process?
261  */
262 int
263 inferior(p)
264 	register struct proc *p;
265 {
266 
267 	sx_assert(&proctree_lock, SX_LOCKED);
268 	for (; p != curproc; p = p->p_pptr)
269 		if (p->p_pid == 0)
270 			return (0);
271 	return (1);
272 }
273 
274 struct proc *
275 pfind_locked(pid_t pid)
276 {
277 	struct proc *p;
278 
279 	sx_assert(&allproc_lock, SX_LOCKED);
280 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
281 		if (p->p_pid == pid) {
282 			PROC_LOCK(p);
283 			if (p->p_state == PRS_NEW) {
284 				PROC_UNLOCK(p);
285 				p = NULL;
286 			}
287 			break;
288 		}
289 	}
290 	return (p);
291 }
292 
293 /*
294  * Locate a process by number; return only "live" processes -- i.e., neither
295  * zombies nor newly born but incompletely initialized processes.  By not
296  * returning processes in the PRS_NEW state, we allow callers to avoid
297  * testing for that condition to avoid dereferencing p_ucred, et al.
298  */
299 struct proc *
300 pfind(pid_t pid)
301 {
302 	struct proc *p;
303 
304 	sx_slock(&allproc_lock);
305 	p = pfind_locked(pid);
306 	sx_sunlock(&allproc_lock);
307 	return (p);
308 }
309 
310 static struct proc *
311 pfind_tid_locked(pid_t tid)
312 {
313 	struct proc *p;
314 	struct thread *td;
315 
316 	sx_assert(&allproc_lock, SX_LOCKED);
317 	FOREACH_PROC_IN_SYSTEM(p) {
318 		PROC_LOCK(p);
319 		if (p->p_state == PRS_NEW) {
320 			PROC_UNLOCK(p);
321 			continue;
322 		}
323 		FOREACH_THREAD_IN_PROC(p, td) {
324 			if (td->td_tid == tid)
325 				goto found;
326 		}
327 		PROC_UNLOCK(p);
328 	}
329 found:
330 	return (p);
331 }
332 
333 /*
334  * Locate a process group by number.
335  * The caller must hold proctree_lock.
336  */
337 struct pgrp *
338 pgfind(pgid)
339 	register pid_t pgid;
340 {
341 	register struct pgrp *pgrp;
342 
343 	sx_assert(&proctree_lock, SX_LOCKED);
344 
345 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
346 		if (pgrp->pg_id == pgid) {
347 			PGRP_LOCK(pgrp);
348 			return (pgrp);
349 		}
350 	}
351 	return (NULL);
352 }
353 
354 /*
355  * Locate process and do additional manipulations, depending on flags.
356  */
357 int
358 pget(pid_t pid, int flags, struct proc **pp)
359 {
360 	struct proc *p;
361 	int error;
362 
363 	sx_slock(&allproc_lock);
364 	if (pid <= PID_MAX) {
365 		p = pfind_locked(pid);
366 		if (p == NULL && (flags & PGET_NOTWEXIT) == 0)
367 			p = zpfind_locked(pid);
368 	} else if ((flags & PGET_NOTID) == 0) {
369 		p = pfind_tid_locked(pid);
370 	} else {
371 		p = NULL;
372 	}
373 	sx_sunlock(&allproc_lock);
374 	if (p == NULL)
375 		return (ESRCH);
376 	if ((flags & PGET_CANSEE) != 0) {
377 		error = p_cansee(curthread, p);
378 		if (error != 0)
379 			goto errout;
380 	}
381 	if ((flags & PGET_CANDEBUG) != 0) {
382 		error = p_candebug(curthread, p);
383 		if (error != 0)
384 			goto errout;
385 	}
386 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
387 		error = EPERM;
388 		goto errout;
389 	}
390 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
391 		error = ESRCH;
392 		goto errout;
393 	}
394 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
395 		/*
396 		 * XXXRW: Not clear ESRCH is the right error during proc
397 		 * execve().
398 		 */
399 		error = ESRCH;
400 		goto errout;
401 	}
402 	if ((flags & PGET_HOLD) != 0) {
403 		_PHOLD(p);
404 		PROC_UNLOCK(p);
405 	}
406 	*pp = p;
407 	return (0);
408 errout:
409 	PROC_UNLOCK(p);
410 	return (error);
411 }
412 
413 /*
414  * Create a new process group.
415  * pgid must be equal to the pid of p.
416  * Begin a new session if required.
417  */
418 int
419 enterpgrp(p, pgid, pgrp, sess)
420 	register struct proc *p;
421 	pid_t pgid;
422 	struct pgrp *pgrp;
423 	struct session *sess;
424 {
425 
426 	sx_assert(&proctree_lock, SX_XLOCKED);
427 
428 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
429 	KASSERT(p->p_pid == pgid,
430 	    ("enterpgrp: new pgrp and pid != pgid"));
431 	KASSERT(pgfind(pgid) == NULL,
432 	    ("enterpgrp: pgrp with pgid exists"));
433 	KASSERT(!SESS_LEADER(p),
434 	    ("enterpgrp: session leader attempted setpgrp"));
435 
436 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
437 
438 	if (sess != NULL) {
439 		/*
440 		 * new session
441 		 */
442 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
443 		PROC_LOCK(p);
444 		p->p_flag &= ~P_CONTROLT;
445 		PROC_UNLOCK(p);
446 		PGRP_LOCK(pgrp);
447 		sess->s_leader = p;
448 		sess->s_sid = p->p_pid;
449 		refcount_init(&sess->s_count, 1);
450 		sess->s_ttyvp = NULL;
451 		sess->s_ttydp = NULL;
452 		sess->s_ttyp = NULL;
453 		bcopy(p->p_session->s_login, sess->s_login,
454 			    sizeof(sess->s_login));
455 		pgrp->pg_session = sess;
456 		KASSERT(p == curproc,
457 		    ("enterpgrp: mksession and p != curproc"));
458 	} else {
459 		pgrp->pg_session = p->p_session;
460 		sess_hold(pgrp->pg_session);
461 		PGRP_LOCK(pgrp);
462 	}
463 	pgrp->pg_id = pgid;
464 	LIST_INIT(&pgrp->pg_members);
465 
466 	/*
467 	 * As we have an exclusive lock of proctree_lock,
468 	 * this should not deadlock.
469 	 */
470 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
471 	pgrp->pg_jobc = 0;
472 	SLIST_INIT(&pgrp->pg_sigiolst);
473 	PGRP_UNLOCK(pgrp);
474 
475 	doenterpgrp(p, pgrp);
476 
477 	return (0);
478 }
479 
480 /*
481  * Move p to an existing process group
482  */
483 int
484 enterthispgrp(p, pgrp)
485 	register struct proc *p;
486 	struct pgrp *pgrp;
487 {
488 
489 	sx_assert(&proctree_lock, SX_XLOCKED);
490 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
491 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
492 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
493 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
494 	KASSERT(pgrp->pg_session == p->p_session,
495 		("%s: pgrp's session %p, p->p_session %p.\n",
496 		__func__,
497 		pgrp->pg_session,
498 		p->p_session));
499 	KASSERT(pgrp != p->p_pgrp,
500 		("%s: p belongs to pgrp.", __func__));
501 
502 	doenterpgrp(p, pgrp);
503 
504 	return (0);
505 }
506 
507 /*
508  * Move p to a process group
509  */
510 static void
511 doenterpgrp(p, pgrp)
512 	struct proc *p;
513 	struct pgrp *pgrp;
514 {
515 	struct pgrp *savepgrp;
516 
517 	sx_assert(&proctree_lock, SX_XLOCKED);
518 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
519 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
520 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
521 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
522 
523 	savepgrp = p->p_pgrp;
524 
525 	/*
526 	 * Adjust eligibility of affected pgrps to participate in job control.
527 	 * Increment eligibility counts before decrementing, otherwise we
528 	 * could reach 0 spuriously during the first call.
529 	 */
530 	fixjobc(p, pgrp, 1);
531 	fixjobc(p, p->p_pgrp, 0);
532 
533 	PGRP_LOCK(pgrp);
534 	PGRP_LOCK(savepgrp);
535 	PROC_LOCK(p);
536 	LIST_REMOVE(p, p_pglist);
537 	p->p_pgrp = pgrp;
538 	PROC_UNLOCK(p);
539 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
540 	PGRP_UNLOCK(savepgrp);
541 	PGRP_UNLOCK(pgrp);
542 	if (LIST_EMPTY(&savepgrp->pg_members))
543 		pgdelete(savepgrp);
544 }
545 
546 /*
547  * remove process from process group
548  */
549 int
550 leavepgrp(p)
551 	register struct proc *p;
552 {
553 	struct pgrp *savepgrp;
554 
555 	sx_assert(&proctree_lock, SX_XLOCKED);
556 	savepgrp = p->p_pgrp;
557 	PGRP_LOCK(savepgrp);
558 	PROC_LOCK(p);
559 	LIST_REMOVE(p, p_pglist);
560 	p->p_pgrp = NULL;
561 	PROC_UNLOCK(p);
562 	PGRP_UNLOCK(savepgrp);
563 	if (LIST_EMPTY(&savepgrp->pg_members))
564 		pgdelete(savepgrp);
565 	return (0);
566 }
567 
568 /*
569  * delete a process group
570  */
571 static void
572 pgdelete(pgrp)
573 	register struct pgrp *pgrp;
574 {
575 	struct session *savesess;
576 	struct tty *tp;
577 
578 	sx_assert(&proctree_lock, SX_XLOCKED);
579 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
580 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
581 
582 	/*
583 	 * Reset any sigio structures pointing to us as a result of
584 	 * F_SETOWN with our pgid.
585 	 */
586 	funsetownlst(&pgrp->pg_sigiolst);
587 
588 	PGRP_LOCK(pgrp);
589 	tp = pgrp->pg_session->s_ttyp;
590 	LIST_REMOVE(pgrp, pg_hash);
591 	savesess = pgrp->pg_session;
592 	PGRP_UNLOCK(pgrp);
593 
594 	/* Remove the reference to the pgrp before deallocating it. */
595 	if (tp != NULL) {
596 		tty_lock(tp);
597 		tty_rel_pgrp(tp, pgrp);
598 	}
599 
600 	mtx_destroy(&pgrp->pg_mtx);
601 	free(pgrp, M_PGRP);
602 	sess_release(savesess);
603 }
604 
605 static void
606 pgadjustjobc(pgrp, entering)
607 	struct pgrp *pgrp;
608 	int entering;
609 {
610 
611 	PGRP_LOCK(pgrp);
612 	if (entering)
613 		pgrp->pg_jobc++;
614 	else {
615 		--pgrp->pg_jobc;
616 		if (pgrp->pg_jobc == 0)
617 			orphanpg(pgrp);
618 	}
619 	PGRP_UNLOCK(pgrp);
620 }
621 
622 /*
623  * Adjust pgrp jobc counters when specified process changes process group.
624  * We count the number of processes in each process group that "qualify"
625  * the group for terminal job control (those with a parent in a different
626  * process group of the same session).  If that count reaches zero, the
627  * process group becomes orphaned.  Check both the specified process'
628  * process group and that of its children.
629  * entering == 0 => p is leaving specified group.
630  * entering == 1 => p is entering specified group.
631  */
632 void
633 fixjobc(p, pgrp, entering)
634 	register struct proc *p;
635 	register struct pgrp *pgrp;
636 	int entering;
637 {
638 	register struct pgrp *hispgrp;
639 	register struct session *mysession;
640 
641 	sx_assert(&proctree_lock, SX_LOCKED);
642 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
643 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
644 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
645 
646 	/*
647 	 * Check p's parent to see whether p qualifies its own process
648 	 * group; if so, adjust count for p's process group.
649 	 */
650 	mysession = pgrp->pg_session;
651 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
652 	    hispgrp->pg_session == mysession)
653 		pgadjustjobc(pgrp, entering);
654 
655 	/*
656 	 * Check this process' children to see whether they qualify
657 	 * their process groups; if so, adjust counts for children's
658 	 * process groups.
659 	 */
660 	LIST_FOREACH(p, &p->p_children, p_sibling) {
661 		hispgrp = p->p_pgrp;
662 		if (hispgrp == pgrp ||
663 		    hispgrp->pg_session != mysession)
664 			continue;
665 		PROC_LOCK(p);
666 		if (p->p_state == PRS_ZOMBIE) {
667 			PROC_UNLOCK(p);
668 			continue;
669 		}
670 		PROC_UNLOCK(p);
671 		pgadjustjobc(hispgrp, entering);
672 	}
673 }
674 
675 /*
676  * A process group has become orphaned;
677  * if there are any stopped processes in the group,
678  * hang-up all process in that group.
679  */
680 static void
681 orphanpg(pg)
682 	struct pgrp *pg;
683 {
684 	register struct proc *p;
685 
686 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
687 
688 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
689 		PROC_LOCK(p);
690 		if (P_SHOULDSTOP(p)) {
691 			PROC_UNLOCK(p);
692 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
693 				PROC_LOCK(p);
694 				kern_psignal(p, SIGHUP);
695 				kern_psignal(p, SIGCONT);
696 				PROC_UNLOCK(p);
697 			}
698 			return;
699 		}
700 		PROC_UNLOCK(p);
701 	}
702 }
703 
704 void
705 sess_hold(struct session *s)
706 {
707 
708 	refcount_acquire(&s->s_count);
709 }
710 
711 void
712 sess_release(struct session *s)
713 {
714 
715 	if (refcount_release(&s->s_count)) {
716 		if (s->s_ttyp != NULL) {
717 			tty_lock(s->s_ttyp);
718 			tty_rel_sess(s->s_ttyp, s);
719 		}
720 		mtx_destroy(&s->s_mtx);
721 		free(s, M_SESSION);
722 	}
723 }
724 
725 #ifdef DDB
726 
727 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
728 {
729 	register struct pgrp *pgrp;
730 	register struct proc *p;
731 	register int i;
732 
733 	for (i = 0; i <= pgrphash; i++) {
734 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
735 			printf("\tindx %d\n", i);
736 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
737 				printf(
738 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
739 				    (void *)pgrp, (long)pgrp->pg_id,
740 				    (void *)pgrp->pg_session,
741 				    pgrp->pg_session->s_count,
742 				    (void *)LIST_FIRST(&pgrp->pg_members));
743 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
744 					printf("\t\tpid %ld addr %p pgrp %p\n",
745 					    (long)p->p_pid, (void *)p,
746 					    (void *)p->p_pgrp);
747 				}
748 			}
749 		}
750 	}
751 }
752 #endif /* DDB */
753 
754 /*
755  * Calculate the kinfo_proc members which contain process-wide
756  * informations.
757  * Must be called with the target process locked.
758  */
759 static void
760 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
761 {
762 	struct thread *td;
763 
764 	PROC_LOCK_ASSERT(p, MA_OWNED);
765 
766 	kp->ki_estcpu = 0;
767 	kp->ki_pctcpu = 0;
768 	FOREACH_THREAD_IN_PROC(p, td) {
769 		thread_lock(td);
770 		kp->ki_pctcpu += sched_pctcpu(td);
771 		kp->ki_estcpu += td->td_estcpu;
772 		thread_unlock(td);
773 	}
774 }
775 
776 /*
777  * Clear kinfo_proc and fill in any information that is common
778  * to all threads in the process.
779  * Must be called with the target process locked.
780  */
781 static void
782 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
783 {
784 	struct thread *td0;
785 	struct tty *tp;
786 	struct session *sp;
787 	struct ucred *cred;
788 	struct sigacts *ps;
789 
790 	PROC_LOCK_ASSERT(p, MA_OWNED);
791 	bzero(kp, sizeof(*kp));
792 
793 	kp->ki_structsize = sizeof(*kp);
794 	kp->ki_paddr = p;
795 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
796 	kp->ki_args = p->p_args;
797 	kp->ki_textvp = p->p_textvp;
798 #ifdef KTRACE
799 	kp->ki_tracep = p->p_tracevp;
800 	kp->ki_traceflag = p->p_traceflag;
801 #endif
802 	kp->ki_fd = p->p_fd;
803 	kp->ki_vmspace = p->p_vmspace;
804 	kp->ki_flag = p->p_flag;
805 	kp->ki_flag2 = p->p_flag2;
806 	cred = p->p_ucred;
807 	if (cred) {
808 		kp->ki_uid = cred->cr_uid;
809 		kp->ki_ruid = cred->cr_ruid;
810 		kp->ki_svuid = cred->cr_svuid;
811 		kp->ki_cr_flags = 0;
812 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
813 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
814 		/* XXX bde doesn't like KI_NGROUPS */
815 		if (cred->cr_ngroups > KI_NGROUPS) {
816 			kp->ki_ngroups = KI_NGROUPS;
817 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
818 		} else
819 			kp->ki_ngroups = cred->cr_ngroups;
820 		bcopy(cred->cr_groups, kp->ki_groups,
821 		    kp->ki_ngroups * sizeof(gid_t));
822 		kp->ki_rgid = cred->cr_rgid;
823 		kp->ki_svgid = cred->cr_svgid;
824 		/* If jailed(cred), emulate the old P_JAILED flag. */
825 		if (jailed(cred)) {
826 			kp->ki_flag |= P_JAILED;
827 			/* If inside the jail, use 0 as a jail ID. */
828 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
829 				kp->ki_jid = cred->cr_prison->pr_id;
830 		}
831 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
832 		    sizeof(kp->ki_loginclass));
833 	}
834 	ps = p->p_sigacts;
835 	if (ps) {
836 		mtx_lock(&ps->ps_mtx);
837 		kp->ki_sigignore = ps->ps_sigignore;
838 		kp->ki_sigcatch = ps->ps_sigcatch;
839 		mtx_unlock(&ps->ps_mtx);
840 	}
841 	if (p->p_state != PRS_NEW &&
842 	    p->p_state != PRS_ZOMBIE &&
843 	    p->p_vmspace != NULL) {
844 		struct vmspace *vm = p->p_vmspace;
845 
846 		kp->ki_size = vm->vm_map.size;
847 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
848 		FOREACH_THREAD_IN_PROC(p, td0) {
849 			if (!TD_IS_SWAPPED(td0))
850 				kp->ki_rssize += td0->td_kstack_pages;
851 		}
852 		kp->ki_swrss = vm->vm_swrss;
853 		kp->ki_tsize = vm->vm_tsize;
854 		kp->ki_dsize = vm->vm_dsize;
855 		kp->ki_ssize = vm->vm_ssize;
856 	} else if (p->p_state == PRS_ZOMBIE)
857 		kp->ki_stat = SZOMB;
858 	if (kp->ki_flag & P_INMEM)
859 		kp->ki_sflag = PS_INMEM;
860 	else
861 		kp->ki_sflag = 0;
862 	/* Calculate legacy swtime as seconds since 'swtick'. */
863 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
864 	kp->ki_pid = p->p_pid;
865 	kp->ki_nice = p->p_nice;
866 	kp->ki_fibnum = p->p_fibnum;
867 	kp->ki_start = p->p_stats->p_start;
868 	timevaladd(&kp->ki_start, &boottime);
869 	PROC_SLOCK(p);
870 	rufetch(p, &kp->ki_rusage);
871 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
872 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
873 	PROC_SUNLOCK(p);
874 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
875 	/* Some callers want child times in a single value. */
876 	kp->ki_childtime = kp->ki_childstime;
877 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
878 
879 	FOREACH_THREAD_IN_PROC(p, td0)
880 		kp->ki_cow += td0->td_cow;
881 
882 	tp = NULL;
883 	if (p->p_pgrp) {
884 		kp->ki_pgid = p->p_pgrp->pg_id;
885 		kp->ki_jobc = p->p_pgrp->pg_jobc;
886 		sp = p->p_pgrp->pg_session;
887 
888 		if (sp != NULL) {
889 			kp->ki_sid = sp->s_sid;
890 			SESS_LOCK(sp);
891 			strlcpy(kp->ki_login, sp->s_login,
892 			    sizeof(kp->ki_login));
893 			if (sp->s_ttyvp)
894 				kp->ki_kiflag |= KI_CTTY;
895 			if (SESS_LEADER(p))
896 				kp->ki_kiflag |= KI_SLEADER;
897 			/* XXX proctree_lock */
898 			tp = sp->s_ttyp;
899 			SESS_UNLOCK(sp);
900 		}
901 	}
902 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
903 		kp->ki_tdev = tty_udev(tp);
904 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
905 		if (tp->t_session)
906 			kp->ki_tsid = tp->t_session->s_sid;
907 	} else
908 		kp->ki_tdev = NODEV;
909 	if (p->p_comm[0] != '\0')
910 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
911 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
912 	    p->p_sysent->sv_name[0] != '\0')
913 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
914 	kp->ki_siglist = p->p_siglist;
915 	kp->ki_xstat = p->p_xstat;
916 	kp->ki_acflag = p->p_acflag;
917 	kp->ki_lock = p->p_lock;
918 	if (p->p_pptr)
919 		kp->ki_ppid = p->p_pptr->p_pid;
920 }
921 
922 /*
923  * Fill in information that is thread specific.  Must be called with
924  * target process locked.  If 'preferthread' is set, overwrite certain
925  * process-related fields that are maintained for both threads and
926  * processes.
927  */
928 static void
929 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
930 {
931 	struct proc *p;
932 
933 	p = td->td_proc;
934 	kp->ki_tdaddr = td;
935 	PROC_LOCK_ASSERT(p, MA_OWNED);
936 
937 	if (preferthread)
938 		PROC_SLOCK(p);
939 	thread_lock(td);
940 	if (td->td_wmesg != NULL)
941 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
942 	else
943 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
944 	strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname));
945 	if (TD_ON_LOCK(td)) {
946 		kp->ki_kiflag |= KI_LOCKBLOCK;
947 		strlcpy(kp->ki_lockname, td->td_lockname,
948 		    sizeof(kp->ki_lockname));
949 	} else {
950 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
951 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
952 	}
953 
954 	if (p->p_state == PRS_NORMAL) { /* approximate. */
955 		if (TD_ON_RUNQ(td) ||
956 		    TD_CAN_RUN(td) ||
957 		    TD_IS_RUNNING(td)) {
958 			kp->ki_stat = SRUN;
959 		} else if (P_SHOULDSTOP(p)) {
960 			kp->ki_stat = SSTOP;
961 		} else if (TD_IS_SLEEPING(td)) {
962 			kp->ki_stat = SSLEEP;
963 		} else if (TD_ON_LOCK(td)) {
964 			kp->ki_stat = SLOCK;
965 		} else {
966 			kp->ki_stat = SWAIT;
967 		}
968 	} else if (p->p_state == PRS_ZOMBIE) {
969 		kp->ki_stat = SZOMB;
970 	} else {
971 		kp->ki_stat = SIDL;
972 	}
973 
974 	/* Things in the thread */
975 	kp->ki_wchan = td->td_wchan;
976 	kp->ki_pri.pri_level = td->td_priority;
977 	kp->ki_pri.pri_native = td->td_base_pri;
978 	kp->ki_lastcpu = td->td_lastcpu;
979 	kp->ki_oncpu = td->td_oncpu;
980 	kp->ki_tdflags = td->td_flags;
981 	kp->ki_tid = td->td_tid;
982 	kp->ki_numthreads = p->p_numthreads;
983 	kp->ki_pcb = td->td_pcb;
984 	kp->ki_kstack = (void *)td->td_kstack;
985 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
986 	kp->ki_pri.pri_class = td->td_pri_class;
987 	kp->ki_pri.pri_user = td->td_user_pri;
988 
989 	if (preferthread) {
990 		rufetchtd(td, &kp->ki_rusage);
991 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
992 		kp->ki_pctcpu = sched_pctcpu(td);
993 		kp->ki_estcpu = td->td_estcpu;
994 		kp->ki_cow = td->td_cow;
995 	}
996 
997 	/* We can't get this anymore but ps etc never used it anyway. */
998 	kp->ki_rqindex = 0;
999 
1000 	if (preferthread)
1001 		kp->ki_siglist = td->td_siglist;
1002 	kp->ki_sigmask = td->td_sigmask;
1003 	thread_unlock(td);
1004 	if (preferthread)
1005 		PROC_SUNLOCK(p);
1006 }
1007 
1008 /*
1009  * Fill in a kinfo_proc structure for the specified process.
1010  * Must be called with the target process locked.
1011  */
1012 void
1013 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1014 {
1015 
1016 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1017 
1018 	fill_kinfo_proc_only(p, kp);
1019 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1020 	fill_kinfo_aggregate(p, kp);
1021 }
1022 
1023 struct pstats *
1024 pstats_alloc(void)
1025 {
1026 
1027 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1028 }
1029 
1030 /*
1031  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1032  */
1033 void
1034 pstats_fork(struct pstats *src, struct pstats *dst)
1035 {
1036 
1037 	bzero(&dst->pstat_startzero,
1038 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1039 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1040 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1041 }
1042 
1043 void
1044 pstats_free(struct pstats *ps)
1045 {
1046 
1047 	free(ps, M_SUBPROC);
1048 }
1049 
1050 static struct proc *
1051 zpfind_locked(pid_t pid)
1052 {
1053 	struct proc *p;
1054 
1055 	sx_assert(&allproc_lock, SX_LOCKED);
1056 	LIST_FOREACH(p, &zombproc, p_list) {
1057 		if (p->p_pid == pid) {
1058 			PROC_LOCK(p);
1059 			break;
1060 		}
1061 	}
1062 	return (p);
1063 }
1064 
1065 /*
1066  * Locate a zombie process by number
1067  */
1068 struct proc *
1069 zpfind(pid_t pid)
1070 {
1071 	struct proc *p;
1072 
1073 	sx_slock(&allproc_lock);
1074 	p = zpfind_locked(pid);
1075 	sx_sunlock(&allproc_lock);
1076 	return (p);
1077 }
1078 
1079 #ifdef COMPAT_FREEBSD32
1080 
1081 /*
1082  * This function is typically used to copy out the kernel address, so
1083  * it can be replaced by assignment of zero.
1084  */
1085 static inline uint32_t
1086 ptr32_trim(void *ptr)
1087 {
1088 	uintptr_t uptr;
1089 
1090 	uptr = (uintptr_t)ptr;
1091 	return ((uptr > UINT_MAX) ? 0 : uptr);
1092 }
1093 
1094 #define PTRTRIM_CP(src,dst,fld) \
1095 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1096 
1097 static void
1098 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1099 {
1100 	int i;
1101 
1102 	bzero(ki32, sizeof(struct kinfo_proc32));
1103 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1104 	CP(*ki, *ki32, ki_layout);
1105 	PTRTRIM_CP(*ki, *ki32, ki_args);
1106 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1107 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1108 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1109 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1110 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1111 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1112 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1113 	CP(*ki, *ki32, ki_pid);
1114 	CP(*ki, *ki32, ki_ppid);
1115 	CP(*ki, *ki32, ki_pgid);
1116 	CP(*ki, *ki32, ki_tpgid);
1117 	CP(*ki, *ki32, ki_sid);
1118 	CP(*ki, *ki32, ki_tsid);
1119 	CP(*ki, *ki32, ki_jobc);
1120 	CP(*ki, *ki32, ki_tdev);
1121 	CP(*ki, *ki32, ki_siglist);
1122 	CP(*ki, *ki32, ki_sigmask);
1123 	CP(*ki, *ki32, ki_sigignore);
1124 	CP(*ki, *ki32, ki_sigcatch);
1125 	CP(*ki, *ki32, ki_uid);
1126 	CP(*ki, *ki32, ki_ruid);
1127 	CP(*ki, *ki32, ki_svuid);
1128 	CP(*ki, *ki32, ki_rgid);
1129 	CP(*ki, *ki32, ki_svgid);
1130 	CP(*ki, *ki32, ki_ngroups);
1131 	for (i = 0; i < KI_NGROUPS; i++)
1132 		CP(*ki, *ki32, ki_groups[i]);
1133 	CP(*ki, *ki32, ki_size);
1134 	CP(*ki, *ki32, ki_rssize);
1135 	CP(*ki, *ki32, ki_swrss);
1136 	CP(*ki, *ki32, ki_tsize);
1137 	CP(*ki, *ki32, ki_dsize);
1138 	CP(*ki, *ki32, ki_ssize);
1139 	CP(*ki, *ki32, ki_xstat);
1140 	CP(*ki, *ki32, ki_acflag);
1141 	CP(*ki, *ki32, ki_pctcpu);
1142 	CP(*ki, *ki32, ki_estcpu);
1143 	CP(*ki, *ki32, ki_slptime);
1144 	CP(*ki, *ki32, ki_swtime);
1145 	CP(*ki, *ki32, ki_cow);
1146 	CP(*ki, *ki32, ki_runtime);
1147 	TV_CP(*ki, *ki32, ki_start);
1148 	TV_CP(*ki, *ki32, ki_childtime);
1149 	CP(*ki, *ki32, ki_flag);
1150 	CP(*ki, *ki32, ki_kiflag);
1151 	CP(*ki, *ki32, ki_traceflag);
1152 	CP(*ki, *ki32, ki_stat);
1153 	CP(*ki, *ki32, ki_nice);
1154 	CP(*ki, *ki32, ki_lock);
1155 	CP(*ki, *ki32, ki_rqindex);
1156 	CP(*ki, *ki32, ki_oncpu);
1157 	CP(*ki, *ki32, ki_lastcpu);
1158 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1159 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1160 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1161 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1162 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1163 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1164 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1165 	CP(*ki, *ki32, ki_flag2);
1166 	CP(*ki, *ki32, ki_fibnum);
1167 	CP(*ki, *ki32, ki_cr_flags);
1168 	CP(*ki, *ki32, ki_jid);
1169 	CP(*ki, *ki32, ki_numthreads);
1170 	CP(*ki, *ki32, ki_tid);
1171 	CP(*ki, *ki32, ki_pri);
1172 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1173 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1174 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1175 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1176 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1177 	CP(*ki, *ki32, ki_sflag);
1178 	CP(*ki, *ki32, ki_tdflags);
1179 }
1180 #endif
1181 
1182 int
1183 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1184 {
1185 	struct thread *td;
1186 	struct kinfo_proc ki;
1187 #ifdef COMPAT_FREEBSD32
1188 	struct kinfo_proc32 ki32;
1189 #endif
1190 	int error;
1191 
1192 	PROC_LOCK_ASSERT(p, MA_OWNED);
1193 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1194 
1195 	error = 0;
1196 	fill_kinfo_proc(p, &ki);
1197 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1198 #ifdef COMPAT_FREEBSD32
1199 		if ((flags & KERN_PROC_MASK32) != 0) {
1200 			freebsd32_kinfo_proc_out(&ki, &ki32);
1201 			error = sbuf_bcat(sb, &ki32, sizeof(ki32));
1202 		} else
1203 #endif
1204 			error = sbuf_bcat(sb, &ki, sizeof(ki));
1205 	} else {
1206 		FOREACH_THREAD_IN_PROC(p, td) {
1207 			fill_kinfo_thread(td, &ki, 1);
1208 #ifdef COMPAT_FREEBSD32
1209 			if ((flags & KERN_PROC_MASK32) != 0) {
1210 				freebsd32_kinfo_proc_out(&ki, &ki32);
1211 				error = sbuf_bcat(sb, &ki32, sizeof(ki32));
1212 			} else
1213 #endif
1214 				error = sbuf_bcat(sb, &ki, sizeof(ki));
1215 			if (error)
1216 				break;
1217 		}
1218 	}
1219 	PROC_UNLOCK(p);
1220 	return (error);
1221 }
1222 
1223 static int
1224 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags,
1225     int doingzomb)
1226 {
1227 	struct sbuf sb;
1228 	struct kinfo_proc ki;
1229 	struct proc *np;
1230 	int error, error2;
1231 	pid_t pid;
1232 
1233 	pid = p->p_pid;
1234 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1235 	error = kern_proc_out(p, &sb, flags);
1236 	error2 = sbuf_finish(&sb);
1237 	sbuf_delete(&sb);
1238 	if (error != 0)
1239 		return (error);
1240 	else if (error2 != 0)
1241 		return (error2);
1242 	if (doingzomb)
1243 		np = zpfind(pid);
1244 	else {
1245 		if (pid == 0)
1246 			return (0);
1247 		np = pfind(pid);
1248 	}
1249 	if (np == NULL)
1250 		return (ESRCH);
1251 	if (np != p) {
1252 		PROC_UNLOCK(np);
1253 		return (ESRCH);
1254 	}
1255 	PROC_UNLOCK(np);
1256 	return (0);
1257 }
1258 
1259 static int
1260 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1261 {
1262 	int *name = (int *)arg1;
1263 	u_int namelen = arg2;
1264 	struct proc *p;
1265 	int flags, doingzomb, oid_number;
1266 	int error = 0;
1267 
1268 	oid_number = oidp->oid_number;
1269 	if (oid_number != KERN_PROC_ALL &&
1270 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1271 		flags = KERN_PROC_NOTHREADS;
1272 	else {
1273 		flags = 0;
1274 		oid_number &= ~KERN_PROC_INC_THREAD;
1275 	}
1276 #ifdef COMPAT_FREEBSD32
1277 	if (req->flags & SCTL_MASK32)
1278 		flags |= KERN_PROC_MASK32;
1279 #endif
1280 	if (oid_number == KERN_PROC_PID) {
1281 		if (namelen != 1)
1282 			return (EINVAL);
1283 		error = sysctl_wire_old_buffer(req, 0);
1284 		if (error)
1285 			return (error);
1286 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1287 		if (error != 0)
1288 			return (error);
1289 		error = sysctl_out_proc(p, req, flags, 0);
1290 		return (error);
1291 	}
1292 
1293 	switch (oid_number) {
1294 	case KERN_PROC_ALL:
1295 		if (namelen != 0)
1296 			return (EINVAL);
1297 		break;
1298 	case KERN_PROC_PROC:
1299 		if (namelen != 0 && namelen != 1)
1300 			return (EINVAL);
1301 		break;
1302 	default:
1303 		if (namelen != 1)
1304 			return (EINVAL);
1305 		break;
1306 	}
1307 
1308 	if (!req->oldptr) {
1309 		/* overestimate by 5 procs */
1310 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1311 		if (error)
1312 			return (error);
1313 	}
1314 	error = sysctl_wire_old_buffer(req, 0);
1315 	if (error != 0)
1316 		return (error);
1317 	sx_slock(&allproc_lock);
1318 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1319 		if (!doingzomb)
1320 			p = LIST_FIRST(&allproc);
1321 		else
1322 			p = LIST_FIRST(&zombproc);
1323 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
1324 			/*
1325 			 * Skip embryonic processes.
1326 			 */
1327 			PROC_LOCK(p);
1328 			if (p->p_state == PRS_NEW) {
1329 				PROC_UNLOCK(p);
1330 				continue;
1331 			}
1332 			KASSERT(p->p_ucred != NULL,
1333 			    ("process credential is NULL for non-NEW proc"));
1334 			/*
1335 			 * Show a user only appropriate processes.
1336 			 */
1337 			if (p_cansee(curthread, p)) {
1338 				PROC_UNLOCK(p);
1339 				continue;
1340 			}
1341 			/*
1342 			 * TODO - make more efficient (see notes below).
1343 			 * do by session.
1344 			 */
1345 			switch (oid_number) {
1346 
1347 			case KERN_PROC_GID:
1348 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1349 					PROC_UNLOCK(p);
1350 					continue;
1351 				}
1352 				break;
1353 
1354 			case KERN_PROC_PGRP:
1355 				/* could do this by traversing pgrp */
1356 				if (p->p_pgrp == NULL ||
1357 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1358 					PROC_UNLOCK(p);
1359 					continue;
1360 				}
1361 				break;
1362 
1363 			case KERN_PROC_RGID:
1364 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1365 					PROC_UNLOCK(p);
1366 					continue;
1367 				}
1368 				break;
1369 
1370 			case KERN_PROC_SESSION:
1371 				if (p->p_session == NULL ||
1372 				    p->p_session->s_sid != (pid_t)name[0]) {
1373 					PROC_UNLOCK(p);
1374 					continue;
1375 				}
1376 				break;
1377 
1378 			case KERN_PROC_TTY:
1379 				if ((p->p_flag & P_CONTROLT) == 0 ||
1380 				    p->p_session == NULL) {
1381 					PROC_UNLOCK(p);
1382 					continue;
1383 				}
1384 				/* XXX proctree_lock */
1385 				SESS_LOCK(p->p_session);
1386 				if (p->p_session->s_ttyp == NULL ||
1387 				    tty_udev(p->p_session->s_ttyp) !=
1388 				    (dev_t)name[0]) {
1389 					SESS_UNLOCK(p->p_session);
1390 					PROC_UNLOCK(p);
1391 					continue;
1392 				}
1393 				SESS_UNLOCK(p->p_session);
1394 				break;
1395 
1396 			case KERN_PROC_UID:
1397 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1398 					PROC_UNLOCK(p);
1399 					continue;
1400 				}
1401 				break;
1402 
1403 			case KERN_PROC_RUID:
1404 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1405 					PROC_UNLOCK(p);
1406 					continue;
1407 				}
1408 				break;
1409 
1410 			case KERN_PROC_PROC:
1411 				break;
1412 
1413 			default:
1414 				break;
1415 
1416 			}
1417 
1418 			error = sysctl_out_proc(p, req, flags, doingzomb);
1419 			if (error) {
1420 				sx_sunlock(&allproc_lock);
1421 				return (error);
1422 			}
1423 		}
1424 	}
1425 	sx_sunlock(&allproc_lock);
1426 	return (0);
1427 }
1428 
1429 struct pargs *
1430 pargs_alloc(int len)
1431 {
1432 	struct pargs *pa;
1433 
1434 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1435 		M_WAITOK);
1436 	refcount_init(&pa->ar_ref, 1);
1437 	pa->ar_length = len;
1438 	return (pa);
1439 }
1440 
1441 static void
1442 pargs_free(struct pargs *pa)
1443 {
1444 
1445 	free(pa, M_PARGS);
1446 }
1447 
1448 void
1449 pargs_hold(struct pargs *pa)
1450 {
1451 
1452 	if (pa == NULL)
1453 		return;
1454 	refcount_acquire(&pa->ar_ref);
1455 }
1456 
1457 void
1458 pargs_drop(struct pargs *pa)
1459 {
1460 
1461 	if (pa == NULL)
1462 		return;
1463 	if (refcount_release(&pa->ar_ref))
1464 		pargs_free(pa);
1465 }
1466 
1467 static int
1468 proc_read_mem(struct thread *td, struct proc *p, vm_offset_t offset, void* buf,
1469     size_t len)
1470 {
1471 	struct iovec iov;
1472 	struct uio uio;
1473 
1474 	iov.iov_base = (caddr_t)buf;
1475 	iov.iov_len = len;
1476 	uio.uio_iov = &iov;
1477 	uio.uio_iovcnt = 1;
1478 	uio.uio_offset = offset;
1479 	uio.uio_resid = (ssize_t)len;
1480 	uio.uio_segflg = UIO_SYSSPACE;
1481 	uio.uio_rw = UIO_READ;
1482 	uio.uio_td = td;
1483 
1484 	return (proc_rwmem(p, &uio));
1485 }
1486 
1487 static int
1488 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1489     size_t len)
1490 {
1491 	size_t i;
1492 	int error;
1493 
1494 	error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, len);
1495 	/*
1496 	 * Reading the chunk may validly return EFAULT if the string is shorter
1497 	 * than the chunk and is aligned at the end of the page, assuming the
1498 	 * next page is not mapped.  So if EFAULT is returned do a fallback to
1499 	 * one byte read loop.
1500 	 */
1501 	if (error == EFAULT) {
1502 		for (i = 0; i < len; i++, buf++, sptr++) {
1503 			error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, 1);
1504 			if (error != 0)
1505 				return (error);
1506 			if (*buf == '\0')
1507 				break;
1508 		}
1509 		error = 0;
1510 	}
1511 	return (error);
1512 }
1513 
1514 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1515 
1516 enum proc_vector_type {
1517 	PROC_ARG,
1518 	PROC_ENV,
1519 	PROC_AUX,
1520 };
1521 
1522 #ifdef COMPAT_FREEBSD32
1523 static int
1524 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1525     size_t *vsizep, enum proc_vector_type type)
1526 {
1527 	struct freebsd32_ps_strings pss;
1528 	Elf32_Auxinfo aux;
1529 	vm_offset_t vptr, ptr;
1530 	uint32_t *proc_vector32;
1531 	char **proc_vector;
1532 	size_t vsize, size;
1533 	int i, error;
1534 
1535 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1536 	    &pss, sizeof(pss));
1537 	if (error != 0)
1538 		return (error);
1539 	switch (type) {
1540 	case PROC_ARG:
1541 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1542 		vsize = pss.ps_nargvstr;
1543 		if (vsize > ARG_MAX)
1544 			return (ENOEXEC);
1545 		size = vsize * sizeof(int32_t);
1546 		break;
1547 	case PROC_ENV:
1548 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1549 		vsize = pss.ps_nenvstr;
1550 		if (vsize > ARG_MAX)
1551 			return (ENOEXEC);
1552 		size = vsize * sizeof(int32_t);
1553 		break;
1554 	case PROC_AUX:
1555 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1556 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1557 		if (vptr % 4 != 0)
1558 			return (ENOEXEC);
1559 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1560 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1561 			if (error != 0)
1562 				return (error);
1563 			if (aux.a_type == AT_NULL)
1564 				break;
1565 			ptr += sizeof(aux);
1566 		}
1567 		if (aux.a_type != AT_NULL)
1568 			return (ENOEXEC);
1569 		vsize = i + 1;
1570 		size = vsize * sizeof(aux);
1571 		break;
1572 	default:
1573 		KASSERT(0, ("Wrong proc vector type: %d", type));
1574 		return (EINVAL);
1575 	}
1576 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1577 	error = proc_read_mem(td, p, vptr, proc_vector32, size);
1578 	if (error != 0)
1579 		goto done;
1580 	if (type == PROC_AUX) {
1581 		*proc_vectorp = (char **)proc_vector32;
1582 		*vsizep = vsize;
1583 		return (0);
1584 	}
1585 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1586 	for (i = 0; i < (int)vsize; i++)
1587 		proc_vector[i] = PTRIN(proc_vector32[i]);
1588 	*proc_vectorp = proc_vector;
1589 	*vsizep = vsize;
1590 done:
1591 	free(proc_vector32, M_TEMP);
1592 	return (error);
1593 }
1594 #endif
1595 
1596 static int
1597 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1598     size_t *vsizep, enum proc_vector_type type)
1599 {
1600 	struct ps_strings pss;
1601 	Elf_Auxinfo aux;
1602 	vm_offset_t vptr, ptr;
1603 	char **proc_vector;
1604 	size_t vsize, size;
1605 	int error, i;
1606 
1607 #ifdef COMPAT_FREEBSD32
1608 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1609 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1610 #endif
1611 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1612 	    &pss, sizeof(pss));
1613 	if (error != 0)
1614 		return (error);
1615 	switch (type) {
1616 	case PROC_ARG:
1617 		vptr = (vm_offset_t)pss.ps_argvstr;
1618 		vsize = pss.ps_nargvstr;
1619 		if (vsize > ARG_MAX)
1620 			return (ENOEXEC);
1621 		size = vsize * sizeof(char *);
1622 		break;
1623 	case PROC_ENV:
1624 		vptr = (vm_offset_t)pss.ps_envstr;
1625 		vsize = pss.ps_nenvstr;
1626 		if (vsize > ARG_MAX)
1627 			return (ENOEXEC);
1628 		size = vsize * sizeof(char *);
1629 		break;
1630 	case PROC_AUX:
1631 		/*
1632 		 * The aux array is just above env array on the stack. Check
1633 		 * that the address is naturally aligned.
1634 		 */
1635 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1636 		    * sizeof(char *);
1637 #if __ELF_WORD_SIZE == 64
1638 		if (vptr % sizeof(uint64_t) != 0)
1639 #else
1640 		if (vptr % sizeof(uint32_t) != 0)
1641 #endif
1642 			return (ENOEXEC);
1643 		/*
1644 		 * We count the array size reading the aux vectors from the
1645 		 * stack until AT_NULL vector is returned.  So (to keep the code
1646 		 * simple) we read the process stack twice: the first time here
1647 		 * to find the size and the second time when copying the vectors
1648 		 * to the allocated proc_vector.
1649 		 */
1650 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1651 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1652 			if (error != 0)
1653 				return (error);
1654 			if (aux.a_type == AT_NULL)
1655 				break;
1656 			ptr += sizeof(aux);
1657 		}
1658 		/*
1659 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1660 		 * not reached AT_NULL, it is most likely we are reading wrong
1661 		 * data: either the process doesn't have auxv array or data has
1662 		 * been modified. Return the error in this case.
1663 		 */
1664 		if (aux.a_type != AT_NULL)
1665 			return (ENOEXEC);
1666 		vsize = i + 1;
1667 		size = vsize * sizeof(aux);
1668 		break;
1669 	default:
1670 		KASSERT(0, ("Wrong proc vector type: %d", type));
1671 		return (EINVAL); /* In case we are built without INVARIANTS. */
1672 	}
1673 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1674 	if (proc_vector == NULL)
1675 		return (ENOMEM);
1676 	error = proc_read_mem(td, p, vptr, proc_vector, size);
1677 	if (error != 0) {
1678 		free(proc_vector, M_TEMP);
1679 		return (error);
1680 	}
1681 	*proc_vectorp = proc_vector;
1682 	*vsizep = vsize;
1683 
1684 	return (0);
1685 }
1686 
1687 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1688 
1689 static int
1690 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1691     enum proc_vector_type type)
1692 {
1693 	size_t done, len, nchr, vsize;
1694 	int error, i;
1695 	char **proc_vector, *sptr;
1696 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1697 
1698 	PROC_ASSERT_HELD(p);
1699 
1700 	/*
1701 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1702 	 */
1703 	nchr = 2 * (PATH_MAX + ARG_MAX);
1704 
1705 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1706 	if (error != 0)
1707 		return (error);
1708 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1709 		/*
1710 		 * The program may have scribbled into its argv array, e.g. to
1711 		 * remove some arguments.  If that has happened, break out
1712 		 * before trying to read from NULL.
1713 		 */
1714 		if (proc_vector[i] == NULL)
1715 			break;
1716 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1717 			error = proc_read_string(td, p, sptr, pss_string,
1718 			    sizeof(pss_string));
1719 			if (error != 0)
1720 				goto done;
1721 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1722 			if (done + len >= nchr)
1723 				len = nchr - done - 1;
1724 			sbuf_bcat(sb, pss_string, len);
1725 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1726 				break;
1727 			done += GET_PS_STRINGS_CHUNK_SZ;
1728 		}
1729 		sbuf_bcat(sb, "", 1);
1730 		done += len + 1;
1731 	}
1732 done:
1733 	free(proc_vector, M_TEMP);
1734 	return (error);
1735 }
1736 
1737 int
1738 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1739 {
1740 
1741 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1742 }
1743 
1744 int
1745 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1746 {
1747 
1748 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1749 }
1750 
1751 int
1752 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1753 {
1754 	size_t vsize, size;
1755 	char **auxv;
1756 	int error;
1757 
1758 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1759 	if (error == 0) {
1760 #ifdef COMPAT_FREEBSD32
1761 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1762 			size = vsize * sizeof(Elf32_Auxinfo);
1763 		else
1764 #endif
1765 			size = vsize * sizeof(Elf_Auxinfo);
1766 		error = sbuf_bcat(sb, auxv, size);
1767 		free(auxv, M_TEMP);
1768 	}
1769 	return (error);
1770 }
1771 
1772 /*
1773  * This sysctl allows a process to retrieve the argument list or process
1774  * title for another process without groping around in the address space
1775  * of the other process.  It also allow a process to set its own "process
1776  * title to a string of its own choice.
1777  */
1778 static int
1779 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1780 {
1781 	int *name = (int *)arg1;
1782 	u_int namelen = arg2;
1783 	struct pargs *newpa, *pa;
1784 	struct proc *p;
1785 	struct sbuf sb;
1786 	int flags, error = 0, error2;
1787 
1788 	if (namelen != 1)
1789 		return (EINVAL);
1790 
1791 	flags = PGET_CANSEE;
1792 	if (req->newptr != NULL)
1793 		flags |= PGET_ISCURRENT;
1794 	error = pget((pid_t)name[0], flags, &p);
1795 	if (error)
1796 		return (error);
1797 
1798 	pa = p->p_args;
1799 	if (pa != NULL) {
1800 		pargs_hold(pa);
1801 		PROC_UNLOCK(p);
1802 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1803 		pargs_drop(pa);
1804 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1805 		_PHOLD(p);
1806 		PROC_UNLOCK(p);
1807 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1808 		error = proc_getargv(curthread, p, &sb);
1809 		error2 = sbuf_finish(&sb);
1810 		PRELE(p);
1811 		sbuf_delete(&sb);
1812 		if (error == 0 && error2 != 0)
1813 			error = error2;
1814 	} else {
1815 		PROC_UNLOCK(p);
1816 	}
1817 	if (error != 0 || req->newptr == NULL)
1818 		return (error);
1819 
1820 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1821 		return (ENOMEM);
1822 	newpa = pargs_alloc(req->newlen);
1823 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1824 	if (error != 0) {
1825 		pargs_free(newpa);
1826 		return (error);
1827 	}
1828 	PROC_LOCK(p);
1829 	pa = p->p_args;
1830 	p->p_args = newpa;
1831 	PROC_UNLOCK(p);
1832 	pargs_drop(pa);
1833 	return (0);
1834 }
1835 
1836 /*
1837  * This sysctl allows a process to retrieve environment of another process.
1838  */
1839 static int
1840 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
1841 {
1842 	int *name = (int *)arg1;
1843 	u_int namelen = arg2;
1844 	struct proc *p;
1845 	struct sbuf sb;
1846 	int error, error2;
1847 
1848 	if (namelen != 1)
1849 		return (EINVAL);
1850 
1851 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1852 	if (error != 0)
1853 		return (error);
1854 	if ((p->p_flag & P_SYSTEM) != 0) {
1855 		PRELE(p);
1856 		return (0);
1857 	}
1858 
1859 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1860 	error = proc_getenvv(curthread, p, &sb);
1861 	error2 = sbuf_finish(&sb);
1862 	PRELE(p);
1863 	sbuf_delete(&sb);
1864 	return (error != 0 ? error : error2);
1865 }
1866 
1867 /*
1868  * This sysctl allows a process to retrieve ELF auxiliary vector of
1869  * another process.
1870  */
1871 static int
1872 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
1873 {
1874 	int *name = (int *)arg1;
1875 	u_int namelen = arg2;
1876 	struct proc *p;
1877 	struct sbuf sb;
1878 	int error, error2;
1879 
1880 	if (namelen != 1)
1881 		return (EINVAL);
1882 
1883 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1884 	if (error != 0)
1885 		return (error);
1886 	if ((p->p_flag & P_SYSTEM) != 0) {
1887 		PRELE(p);
1888 		return (0);
1889 	}
1890 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1891 	error = proc_getauxv(curthread, p, &sb);
1892 	error2 = sbuf_finish(&sb);
1893 	PRELE(p);
1894 	sbuf_delete(&sb);
1895 	return (error != 0 ? error : error2);
1896 }
1897 
1898 /*
1899  * This sysctl allows a process to retrieve the path of the executable for
1900  * itself or another process.
1901  */
1902 static int
1903 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1904 {
1905 	pid_t *pidp = (pid_t *)arg1;
1906 	unsigned int arglen = arg2;
1907 	struct proc *p;
1908 	struct vnode *vp;
1909 	char *retbuf, *freebuf;
1910 	int error;
1911 
1912 	if (arglen != 1)
1913 		return (EINVAL);
1914 	if (*pidp == -1) {	/* -1 means this process */
1915 		p = req->td->td_proc;
1916 	} else {
1917 		error = pget(*pidp, PGET_CANSEE, &p);
1918 		if (error != 0)
1919 			return (error);
1920 	}
1921 
1922 	vp = p->p_textvp;
1923 	if (vp == NULL) {
1924 		if (*pidp != -1)
1925 			PROC_UNLOCK(p);
1926 		return (0);
1927 	}
1928 	vref(vp);
1929 	if (*pidp != -1)
1930 		PROC_UNLOCK(p);
1931 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1932 	vrele(vp);
1933 	if (error)
1934 		return (error);
1935 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1936 	free(freebuf, M_TEMP);
1937 	return (error);
1938 }
1939 
1940 static int
1941 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1942 {
1943 	struct proc *p;
1944 	char *sv_name;
1945 	int *name;
1946 	int namelen;
1947 	int error;
1948 
1949 	namelen = arg2;
1950 	if (namelen != 1)
1951 		return (EINVAL);
1952 
1953 	name = (int *)arg1;
1954 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
1955 	if (error != 0)
1956 		return (error);
1957 	sv_name = p->p_sysent->sv_name;
1958 	PROC_UNLOCK(p);
1959 	return (sysctl_handle_string(oidp, sv_name, 0, req));
1960 }
1961 
1962 #ifdef KINFO_OVMENTRY_SIZE
1963 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
1964 #endif
1965 
1966 #ifdef COMPAT_FREEBSD7
1967 static int
1968 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
1969 {
1970 	vm_map_entry_t entry, tmp_entry;
1971 	unsigned int last_timestamp;
1972 	char *fullpath, *freepath;
1973 	struct kinfo_ovmentry *kve;
1974 	struct vattr va;
1975 	struct ucred *cred;
1976 	int error, *name;
1977 	struct vnode *vp;
1978 	struct proc *p;
1979 	vm_map_t map;
1980 	struct vmspace *vm;
1981 
1982 	name = (int *)arg1;
1983 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1984 	if (error != 0)
1985 		return (error);
1986 	vm = vmspace_acquire_ref(p);
1987 	if (vm == NULL) {
1988 		PRELE(p);
1989 		return (ESRCH);
1990 	}
1991 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1992 
1993 	map = &vm->vm_map;
1994 	vm_map_lock_read(map);
1995 	for (entry = map->header.next; entry != &map->header;
1996 	    entry = entry->next) {
1997 		vm_object_t obj, tobj, lobj;
1998 		vm_offset_t addr;
1999 
2000 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2001 			continue;
2002 
2003 		bzero(kve, sizeof(*kve));
2004 		kve->kve_structsize = sizeof(*kve);
2005 
2006 		kve->kve_private_resident = 0;
2007 		obj = entry->object.vm_object;
2008 		if (obj != NULL) {
2009 			VM_OBJECT_RLOCK(obj);
2010 			if (obj->shadow_count == 1)
2011 				kve->kve_private_resident =
2012 				    obj->resident_page_count;
2013 		}
2014 		kve->kve_resident = 0;
2015 		addr = entry->start;
2016 		while (addr < entry->end) {
2017 			if (pmap_extract(map->pmap, addr))
2018 				kve->kve_resident++;
2019 			addr += PAGE_SIZE;
2020 		}
2021 
2022 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2023 			if (tobj != obj)
2024 				VM_OBJECT_RLOCK(tobj);
2025 			if (lobj != obj)
2026 				VM_OBJECT_RUNLOCK(lobj);
2027 			lobj = tobj;
2028 		}
2029 
2030 		kve->kve_start = (void*)entry->start;
2031 		kve->kve_end = (void*)entry->end;
2032 		kve->kve_offset = (off_t)entry->offset;
2033 
2034 		if (entry->protection & VM_PROT_READ)
2035 			kve->kve_protection |= KVME_PROT_READ;
2036 		if (entry->protection & VM_PROT_WRITE)
2037 			kve->kve_protection |= KVME_PROT_WRITE;
2038 		if (entry->protection & VM_PROT_EXECUTE)
2039 			kve->kve_protection |= KVME_PROT_EXEC;
2040 
2041 		if (entry->eflags & MAP_ENTRY_COW)
2042 			kve->kve_flags |= KVME_FLAG_COW;
2043 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2044 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2045 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2046 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2047 
2048 		last_timestamp = map->timestamp;
2049 		vm_map_unlock_read(map);
2050 
2051 		kve->kve_fileid = 0;
2052 		kve->kve_fsid = 0;
2053 		freepath = NULL;
2054 		fullpath = "";
2055 		if (lobj) {
2056 			vp = NULL;
2057 			switch (lobj->type) {
2058 			case OBJT_DEFAULT:
2059 				kve->kve_type = KVME_TYPE_DEFAULT;
2060 				break;
2061 			case OBJT_VNODE:
2062 				kve->kve_type = KVME_TYPE_VNODE;
2063 				vp = lobj->handle;
2064 				vref(vp);
2065 				break;
2066 			case OBJT_SWAP:
2067 				kve->kve_type = KVME_TYPE_SWAP;
2068 				break;
2069 			case OBJT_DEVICE:
2070 				kve->kve_type = KVME_TYPE_DEVICE;
2071 				break;
2072 			case OBJT_PHYS:
2073 				kve->kve_type = KVME_TYPE_PHYS;
2074 				break;
2075 			case OBJT_DEAD:
2076 				kve->kve_type = KVME_TYPE_DEAD;
2077 				break;
2078 			case OBJT_SG:
2079 				kve->kve_type = KVME_TYPE_SG;
2080 				break;
2081 			default:
2082 				kve->kve_type = KVME_TYPE_UNKNOWN;
2083 				break;
2084 			}
2085 			if (lobj != obj)
2086 				VM_OBJECT_RUNLOCK(lobj);
2087 
2088 			kve->kve_ref_count = obj->ref_count;
2089 			kve->kve_shadow_count = obj->shadow_count;
2090 			VM_OBJECT_RUNLOCK(obj);
2091 			if (vp != NULL) {
2092 				vn_fullpath(curthread, vp, &fullpath,
2093 				    &freepath);
2094 				cred = curthread->td_ucred;
2095 				vn_lock(vp, LK_SHARED | LK_RETRY);
2096 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2097 					kve->kve_fileid = va.va_fileid;
2098 					kve->kve_fsid = va.va_fsid;
2099 				}
2100 				vput(vp);
2101 			}
2102 		} else {
2103 			kve->kve_type = KVME_TYPE_NONE;
2104 			kve->kve_ref_count = 0;
2105 			kve->kve_shadow_count = 0;
2106 		}
2107 
2108 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2109 		if (freepath != NULL)
2110 			free(freepath, M_TEMP);
2111 
2112 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2113 		vm_map_lock_read(map);
2114 		if (error)
2115 			break;
2116 		if (last_timestamp != map->timestamp) {
2117 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2118 			entry = tmp_entry;
2119 		}
2120 	}
2121 	vm_map_unlock_read(map);
2122 	vmspace_free(vm);
2123 	PRELE(p);
2124 	free(kve, M_TEMP);
2125 	return (error);
2126 }
2127 #endif	/* COMPAT_FREEBSD7 */
2128 
2129 #ifdef KINFO_VMENTRY_SIZE
2130 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2131 #endif
2132 
2133 /*
2134  * Must be called with the process locked and will return unlocked.
2135  */
2136 int
2137 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb)
2138 {
2139 	vm_map_entry_t entry, tmp_entry;
2140 	unsigned int last_timestamp;
2141 	char *fullpath, *freepath;
2142 	struct kinfo_vmentry *kve;
2143 	struct vattr va;
2144 	struct ucred *cred;
2145 	int error;
2146 	struct vnode *vp;
2147 	struct vmspace *vm;
2148 	vm_map_t map;
2149 
2150 	PROC_LOCK_ASSERT(p, MA_OWNED);
2151 
2152 	_PHOLD(p);
2153 	PROC_UNLOCK(p);
2154 	vm = vmspace_acquire_ref(p);
2155 	if (vm == NULL) {
2156 		PRELE(p);
2157 		return (ESRCH);
2158 	}
2159 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2160 
2161 	error = 0;
2162 	map = &vm->vm_map;
2163 	vm_map_lock_read(map);
2164 	for (entry = map->header.next; entry != &map->header;
2165 	    entry = entry->next) {
2166 		vm_object_t obj, tobj, lobj;
2167 		vm_offset_t addr;
2168 		vm_paddr_t locked_pa;
2169 		int mincoreinfo;
2170 
2171 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2172 			continue;
2173 
2174 		bzero(kve, sizeof(*kve));
2175 
2176 		kve->kve_private_resident = 0;
2177 		obj = entry->object.vm_object;
2178 		if (obj != NULL) {
2179 			VM_OBJECT_RLOCK(obj);
2180 			if (obj->shadow_count == 1)
2181 				kve->kve_private_resident =
2182 				    obj->resident_page_count;
2183 		}
2184 		kve->kve_resident = 0;
2185 		addr = entry->start;
2186 		while (addr < entry->end) {
2187 			locked_pa = 0;
2188 			mincoreinfo = pmap_mincore(map->pmap, addr, &locked_pa);
2189 			if (locked_pa != 0)
2190 				vm_page_unlock(PHYS_TO_VM_PAGE(locked_pa));
2191 			if (mincoreinfo & MINCORE_INCORE)
2192 				kve->kve_resident++;
2193 			if (mincoreinfo & MINCORE_SUPER)
2194 				kve->kve_flags |= KVME_FLAG_SUPER;
2195 			addr += PAGE_SIZE;
2196 		}
2197 
2198 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2199 			if (tobj != obj)
2200 				VM_OBJECT_RLOCK(tobj);
2201 			if (lobj != obj)
2202 				VM_OBJECT_RUNLOCK(lobj);
2203 			lobj = tobj;
2204 		}
2205 
2206 		kve->kve_start = entry->start;
2207 		kve->kve_end = entry->end;
2208 		kve->kve_offset = entry->offset;
2209 
2210 		if (entry->protection & VM_PROT_READ)
2211 			kve->kve_protection |= KVME_PROT_READ;
2212 		if (entry->protection & VM_PROT_WRITE)
2213 			kve->kve_protection |= KVME_PROT_WRITE;
2214 		if (entry->protection & VM_PROT_EXECUTE)
2215 			kve->kve_protection |= KVME_PROT_EXEC;
2216 
2217 		if (entry->eflags & MAP_ENTRY_COW)
2218 			kve->kve_flags |= KVME_FLAG_COW;
2219 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2220 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2221 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2222 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2223 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2224 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2225 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2226 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2227 
2228 		last_timestamp = map->timestamp;
2229 		vm_map_unlock_read(map);
2230 
2231 		freepath = NULL;
2232 		fullpath = "";
2233 		if (lobj) {
2234 			vp = NULL;
2235 			switch (lobj->type) {
2236 			case OBJT_DEFAULT:
2237 				kve->kve_type = KVME_TYPE_DEFAULT;
2238 				break;
2239 			case OBJT_VNODE:
2240 				kve->kve_type = KVME_TYPE_VNODE;
2241 				vp = lobj->handle;
2242 				vref(vp);
2243 				break;
2244 			case OBJT_SWAP:
2245 				kve->kve_type = KVME_TYPE_SWAP;
2246 				break;
2247 			case OBJT_DEVICE:
2248 				kve->kve_type = KVME_TYPE_DEVICE;
2249 				break;
2250 			case OBJT_PHYS:
2251 				kve->kve_type = KVME_TYPE_PHYS;
2252 				break;
2253 			case OBJT_DEAD:
2254 				kve->kve_type = KVME_TYPE_DEAD;
2255 				break;
2256 			case OBJT_SG:
2257 				kve->kve_type = KVME_TYPE_SG;
2258 				break;
2259 			default:
2260 				kve->kve_type = KVME_TYPE_UNKNOWN;
2261 				break;
2262 			}
2263 			if (lobj != obj)
2264 				VM_OBJECT_RUNLOCK(lobj);
2265 
2266 			kve->kve_ref_count = obj->ref_count;
2267 			kve->kve_shadow_count = obj->shadow_count;
2268 			VM_OBJECT_RUNLOCK(obj);
2269 			if (vp != NULL) {
2270 				vn_fullpath(curthread, vp, &fullpath,
2271 				    &freepath);
2272 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2273 				cred = curthread->td_ucred;
2274 				vn_lock(vp, LK_SHARED | LK_RETRY);
2275 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2276 					kve->kve_vn_fileid = va.va_fileid;
2277 					kve->kve_vn_fsid = va.va_fsid;
2278 					kve->kve_vn_mode =
2279 					    MAKEIMODE(va.va_type, va.va_mode);
2280 					kve->kve_vn_size = va.va_size;
2281 					kve->kve_vn_rdev = va.va_rdev;
2282 					kve->kve_status = KF_ATTR_VALID;
2283 				}
2284 				vput(vp);
2285 			}
2286 		} else {
2287 			kve->kve_type = KVME_TYPE_NONE;
2288 			kve->kve_ref_count = 0;
2289 			kve->kve_shadow_count = 0;
2290 		}
2291 
2292 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2293 		if (freepath != NULL)
2294 			free(freepath, M_TEMP);
2295 
2296 		/* Pack record size down */
2297 		kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) +
2298 		    strlen(kve->kve_path) + 1;
2299 		kve->kve_structsize = roundup(kve->kve_structsize,
2300 		    sizeof(uint64_t));
2301 		error = sbuf_bcat(sb, kve, kve->kve_structsize);
2302 		vm_map_lock_read(map);
2303 		if (error)
2304 			break;
2305 		if (last_timestamp != map->timestamp) {
2306 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2307 			entry = tmp_entry;
2308 		}
2309 	}
2310 	vm_map_unlock_read(map);
2311 	vmspace_free(vm);
2312 	PRELE(p);
2313 	free(kve, M_TEMP);
2314 	return (error);
2315 }
2316 
2317 static int
2318 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2319 {
2320 	struct proc *p;
2321 	struct sbuf sb;
2322 	int error, error2, *name;
2323 
2324 	name = (int *)arg1;
2325 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2326 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2327 	if (error != 0) {
2328 		sbuf_delete(&sb);
2329 		return (error);
2330 	}
2331 	error = kern_proc_vmmap_out(p, &sb);
2332 	error2 = sbuf_finish(&sb);
2333 	sbuf_delete(&sb);
2334 	return (error != 0 ? error : error2);
2335 }
2336 
2337 #if defined(STACK) || defined(DDB)
2338 static int
2339 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2340 {
2341 	struct kinfo_kstack *kkstp;
2342 	int error, i, *name, numthreads;
2343 	lwpid_t *lwpidarray;
2344 	struct thread *td;
2345 	struct stack *st;
2346 	struct sbuf sb;
2347 	struct proc *p;
2348 
2349 	name = (int *)arg1;
2350 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2351 	if (error != 0)
2352 		return (error);
2353 
2354 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2355 	st = stack_create();
2356 
2357 	lwpidarray = NULL;
2358 	numthreads = 0;
2359 	PROC_LOCK(p);
2360 repeat:
2361 	if (numthreads < p->p_numthreads) {
2362 		if (lwpidarray != NULL) {
2363 			free(lwpidarray, M_TEMP);
2364 			lwpidarray = NULL;
2365 		}
2366 		numthreads = p->p_numthreads;
2367 		PROC_UNLOCK(p);
2368 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2369 		    M_WAITOK | M_ZERO);
2370 		PROC_LOCK(p);
2371 		goto repeat;
2372 	}
2373 	i = 0;
2374 
2375 	/*
2376 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2377 	 * of changes could have taken place.  Should we check to see if the
2378 	 * vmspace has been replaced, or the like, in order to prevent
2379 	 * giving a snapshot that spans, say, execve(2), with some threads
2380 	 * before and some after?  Among other things, the credentials could
2381 	 * have changed, in which case the right to extract debug info might
2382 	 * no longer be assured.
2383 	 */
2384 	FOREACH_THREAD_IN_PROC(p, td) {
2385 		KASSERT(i < numthreads,
2386 		    ("sysctl_kern_proc_kstack: numthreads"));
2387 		lwpidarray[i] = td->td_tid;
2388 		i++;
2389 	}
2390 	numthreads = i;
2391 	for (i = 0; i < numthreads; i++) {
2392 		td = thread_find(p, lwpidarray[i]);
2393 		if (td == NULL) {
2394 			continue;
2395 		}
2396 		bzero(kkstp, sizeof(*kkstp));
2397 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2398 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2399 		thread_lock(td);
2400 		kkstp->kkst_tid = td->td_tid;
2401 		if (TD_IS_SWAPPED(td))
2402 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2403 		else if (TD_IS_RUNNING(td))
2404 			kkstp->kkst_state = KKST_STATE_RUNNING;
2405 		else {
2406 			kkstp->kkst_state = KKST_STATE_STACKOK;
2407 			stack_save_td(st, td);
2408 		}
2409 		thread_unlock(td);
2410 		PROC_UNLOCK(p);
2411 		stack_sbuf_print(&sb, st);
2412 		sbuf_finish(&sb);
2413 		sbuf_delete(&sb);
2414 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2415 		PROC_LOCK(p);
2416 		if (error)
2417 			break;
2418 	}
2419 	_PRELE(p);
2420 	PROC_UNLOCK(p);
2421 	if (lwpidarray != NULL)
2422 		free(lwpidarray, M_TEMP);
2423 	stack_destroy(st);
2424 	free(kkstp, M_TEMP);
2425 	return (error);
2426 }
2427 #endif
2428 
2429 /*
2430  * This sysctl allows a process to retrieve the full list of groups from
2431  * itself or another process.
2432  */
2433 static int
2434 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2435 {
2436 	pid_t *pidp = (pid_t *)arg1;
2437 	unsigned int arglen = arg2;
2438 	struct proc *p;
2439 	struct ucred *cred;
2440 	int error;
2441 
2442 	if (arglen != 1)
2443 		return (EINVAL);
2444 	if (*pidp == -1) {	/* -1 means this process */
2445 		p = req->td->td_proc;
2446 	} else {
2447 		error = pget(*pidp, PGET_CANSEE, &p);
2448 		if (error != 0)
2449 			return (error);
2450 	}
2451 
2452 	cred = crhold(p->p_ucred);
2453 	if (*pidp != -1)
2454 		PROC_UNLOCK(p);
2455 
2456 	error = SYSCTL_OUT(req, cred->cr_groups,
2457 	    cred->cr_ngroups * sizeof(gid_t));
2458 	crfree(cred);
2459 	return (error);
2460 }
2461 
2462 /*
2463  * This sysctl allows a process to retrieve or/and set the resource limit for
2464  * another process.
2465  */
2466 static int
2467 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2468 {
2469 	int *name = (int *)arg1;
2470 	u_int namelen = arg2;
2471 	struct rlimit rlim;
2472 	struct proc *p;
2473 	u_int which;
2474 	int flags, error;
2475 
2476 	if (namelen != 2)
2477 		return (EINVAL);
2478 
2479 	which = (u_int)name[1];
2480 	if (which >= RLIM_NLIMITS)
2481 		return (EINVAL);
2482 
2483 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2484 		return (EINVAL);
2485 
2486 	flags = PGET_HOLD | PGET_NOTWEXIT;
2487 	if (req->newptr != NULL)
2488 		flags |= PGET_CANDEBUG;
2489 	else
2490 		flags |= PGET_CANSEE;
2491 	error = pget((pid_t)name[0], flags, &p);
2492 	if (error != 0)
2493 		return (error);
2494 
2495 	/*
2496 	 * Retrieve limit.
2497 	 */
2498 	if (req->oldptr != NULL) {
2499 		PROC_LOCK(p);
2500 		lim_rlimit(p, which, &rlim);
2501 		PROC_UNLOCK(p);
2502 	}
2503 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2504 	if (error != 0)
2505 		goto errout;
2506 
2507 	/*
2508 	 * Set limit.
2509 	 */
2510 	if (req->newptr != NULL) {
2511 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2512 		if (error == 0)
2513 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2514 	}
2515 
2516 errout:
2517 	PRELE(p);
2518 	return (error);
2519 }
2520 
2521 /*
2522  * This sysctl allows a process to retrieve ps_strings structure location of
2523  * another process.
2524  */
2525 static int
2526 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2527 {
2528 	int *name = (int *)arg1;
2529 	u_int namelen = arg2;
2530 	struct proc *p;
2531 	vm_offset_t ps_strings;
2532 	int error;
2533 #ifdef COMPAT_FREEBSD32
2534 	uint32_t ps_strings32;
2535 #endif
2536 
2537 	if (namelen != 1)
2538 		return (EINVAL);
2539 
2540 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2541 	if (error != 0)
2542 		return (error);
2543 #ifdef COMPAT_FREEBSD32
2544 	if ((req->flags & SCTL_MASK32) != 0) {
2545 		/*
2546 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2547 		 * process.
2548 		 */
2549 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2550 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2551 		PROC_UNLOCK(p);
2552 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2553 		return (error);
2554 	}
2555 #endif
2556 	ps_strings = p->p_sysent->sv_psstrings;
2557 	PROC_UNLOCK(p);
2558 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2559 	return (error);
2560 }
2561 
2562 /*
2563  * This sysctl allows a process to retrieve umask of another process.
2564  */
2565 static int
2566 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2567 {
2568 	int *name = (int *)arg1;
2569 	u_int namelen = arg2;
2570 	struct proc *p;
2571 	int error;
2572 	u_short fd_cmask;
2573 
2574 	if (namelen != 1)
2575 		return (EINVAL);
2576 
2577 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2578 	if (error != 0)
2579 		return (error);
2580 
2581 	FILEDESC_SLOCK(p->p_fd);
2582 	fd_cmask = p->p_fd->fd_cmask;
2583 	FILEDESC_SUNLOCK(p->p_fd);
2584 	PRELE(p);
2585 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2586 	return (error);
2587 }
2588 
2589 /*
2590  * This sysctl allows a process to set and retrieve binary osreldate of
2591  * another process.
2592  */
2593 static int
2594 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2595 {
2596 	int *name = (int *)arg1;
2597 	u_int namelen = arg2;
2598 	struct proc *p;
2599 	int flags, error, osrel;
2600 
2601 	if (namelen != 1)
2602 		return (EINVAL);
2603 
2604 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2605 		return (EINVAL);
2606 
2607 	flags = PGET_HOLD | PGET_NOTWEXIT;
2608 	if (req->newptr != NULL)
2609 		flags |= PGET_CANDEBUG;
2610 	else
2611 		flags |= PGET_CANSEE;
2612 	error = pget((pid_t)name[0], flags, &p);
2613 	if (error != 0)
2614 		return (error);
2615 
2616 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2617 	if (error != 0)
2618 		goto errout;
2619 
2620 	if (req->newptr != NULL) {
2621 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2622 		if (error != 0)
2623 			goto errout;
2624 		if (osrel < 0) {
2625 			error = EINVAL;
2626 			goto errout;
2627 		}
2628 		p->p_osrel = osrel;
2629 	}
2630 errout:
2631 	PRELE(p);
2632 	return (error);
2633 }
2634 
2635 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2636 
2637 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2638 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2639 	"Return entire process table");
2640 
2641 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2642 	sysctl_kern_proc, "Process table");
2643 
2644 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2645 	sysctl_kern_proc, "Process table");
2646 
2647 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2648 	sysctl_kern_proc, "Process table");
2649 
2650 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2651 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2652 
2653 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2654 	sysctl_kern_proc, "Process table");
2655 
2656 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2657 	sysctl_kern_proc, "Process table");
2658 
2659 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2660 	sysctl_kern_proc, "Process table");
2661 
2662 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2663 	sysctl_kern_proc, "Process table");
2664 
2665 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2666 	sysctl_kern_proc, "Return process table, no threads");
2667 
2668 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2669 	CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2670 	sysctl_kern_proc_args, "Process argument list");
2671 
2672 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
2673 	sysctl_kern_proc_env, "Process environment");
2674 
2675 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
2676 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
2677 
2678 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2679 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2680 
2681 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2682 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2683 	"Process syscall vector name (ABI type)");
2684 
2685 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2686 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2687 
2688 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2689 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2690 
2691 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2692 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2693 
2694 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2695 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2696 
2697 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2698 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2699 
2700 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2701 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2702 
2703 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2704 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2705 
2706 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2707 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2708 
2709 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2710 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2711 	"Return process table, no threads");
2712 
2713 #ifdef COMPAT_FREEBSD7
2714 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2715 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2716 #endif
2717 
2718 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2719 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2720 
2721 #if defined(STACK) || defined(DDB)
2722 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2723 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2724 #endif
2725 
2726 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2727 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
2728 
2729 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
2730 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
2731 	"Process resource limits");
2732 
2733 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
2734 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
2735 	"Process ps_strings location");
2736 
2737 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
2738 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
2739 
2740 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
2741 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
2742 	"Process binary osreldate");
2743