xref: /freebsd/sys/kern/kern_proc.c (revision 287472b39c7985d968be84ea145c3e75a3e6b875)
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 	cred = p->p_ucred;
806 	if (cred) {
807 		kp->ki_uid = cred->cr_uid;
808 		kp->ki_ruid = cred->cr_ruid;
809 		kp->ki_svuid = cred->cr_svuid;
810 		kp->ki_cr_flags = 0;
811 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
812 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
813 		/* XXX bde doesn't like KI_NGROUPS */
814 		if (cred->cr_ngroups > KI_NGROUPS) {
815 			kp->ki_ngroups = KI_NGROUPS;
816 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
817 		} else
818 			kp->ki_ngroups = cred->cr_ngroups;
819 		bcopy(cred->cr_groups, kp->ki_groups,
820 		    kp->ki_ngroups * sizeof(gid_t));
821 		kp->ki_rgid = cred->cr_rgid;
822 		kp->ki_svgid = cred->cr_svgid;
823 		/* If jailed(cred), emulate the old P_JAILED flag. */
824 		if (jailed(cred)) {
825 			kp->ki_flag |= P_JAILED;
826 			/* If inside the jail, use 0 as a jail ID. */
827 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
828 				kp->ki_jid = cred->cr_prison->pr_id;
829 		}
830 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
831 		    sizeof(kp->ki_loginclass));
832 	}
833 	ps = p->p_sigacts;
834 	if (ps) {
835 		mtx_lock(&ps->ps_mtx);
836 		kp->ki_sigignore = ps->ps_sigignore;
837 		kp->ki_sigcatch = ps->ps_sigcatch;
838 		mtx_unlock(&ps->ps_mtx);
839 	}
840 	if (p->p_state != PRS_NEW &&
841 	    p->p_state != PRS_ZOMBIE &&
842 	    p->p_vmspace != NULL) {
843 		struct vmspace *vm = p->p_vmspace;
844 
845 		kp->ki_size = vm->vm_map.size;
846 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
847 		FOREACH_THREAD_IN_PROC(p, td0) {
848 			if (!TD_IS_SWAPPED(td0))
849 				kp->ki_rssize += td0->td_kstack_pages;
850 		}
851 		kp->ki_swrss = vm->vm_swrss;
852 		kp->ki_tsize = vm->vm_tsize;
853 		kp->ki_dsize = vm->vm_dsize;
854 		kp->ki_ssize = vm->vm_ssize;
855 	} else if (p->p_state == PRS_ZOMBIE)
856 		kp->ki_stat = SZOMB;
857 	if (kp->ki_flag & P_INMEM)
858 		kp->ki_sflag = PS_INMEM;
859 	else
860 		kp->ki_sflag = 0;
861 	/* Calculate legacy swtime as seconds since 'swtick'. */
862 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
863 	kp->ki_pid = p->p_pid;
864 	kp->ki_nice = p->p_nice;
865 	kp->ki_start = p->p_stats->p_start;
866 	timevaladd(&kp->ki_start, &boottime);
867 	PROC_SLOCK(p);
868 	rufetch(p, &kp->ki_rusage);
869 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
870 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
871 	PROC_SUNLOCK(p);
872 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
873 	/* Some callers want child times in a single value. */
874 	kp->ki_childtime = kp->ki_childstime;
875 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
876 
877 	FOREACH_THREAD_IN_PROC(p, td0)
878 		kp->ki_cow += td0->td_cow;
879 
880 	tp = NULL;
881 	if (p->p_pgrp) {
882 		kp->ki_pgid = p->p_pgrp->pg_id;
883 		kp->ki_jobc = p->p_pgrp->pg_jobc;
884 		sp = p->p_pgrp->pg_session;
885 
886 		if (sp != NULL) {
887 			kp->ki_sid = sp->s_sid;
888 			SESS_LOCK(sp);
889 			strlcpy(kp->ki_login, sp->s_login,
890 			    sizeof(kp->ki_login));
891 			if (sp->s_ttyvp)
892 				kp->ki_kiflag |= KI_CTTY;
893 			if (SESS_LEADER(p))
894 				kp->ki_kiflag |= KI_SLEADER;
895 			/* XXX proctree_lock */
896 			tp = sp->s_ttyp;
897 			SESS_UNLOCK(sp);
898 		}
899 	}
900 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
901 		kp->ki_tdev = tty_udev(tp);
902 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
903 		if (tp->t_session)
904 			kp->ki_tsid = tp->t_session->s_sid;
905 	} else
906 		kp->ki_tdev = NODEV;
907 	if (p->p_comm[0] != '\0')
908 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
909 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
910 	    p->p_sysent->sv_name[0] != '\0')
911 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
912 	kp->ki_siglist = p->p_siglist;
913 	kp->ki_xstat = p->p_xstat;
914 	kp->ki_acflag = p->p_acflag;
915 	kp->ki_lock = p->p_lock;
916 	if (p->p_pptr)
917 		kp->ki_ppid = p->p_pptr->p_pid;
918 }
919 
920 /*
921  * Fill in information that is thread specific.  Must be called with
922  * target process locked.  If 'preferthread' is set, overwrite certain
923  * process-related fields that are maintained for both threads and
924  * processes.
925  */
926 static void
927 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
928 {
929 	struct proc *p;
930 
931 	p = td->td_proc;
932 	kp->ki_tdaddr = td;
933 	PROC_LOCK_ASSERT(p, MA_OWNED);
934 
935 	if (preferthread)
936 		PROC_SLOCK(p);
937 	thread_lock(td);
938 	if (td->td_wmesg != NULL)
939 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
940 	else
941 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
942 	strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname));
943 	if (TD_ON_LOCK(td)) {
944 		kp->ki_kiflag |= KI_LOCKBLOCK;
945 		strlcpy(kp->ki_lockname, td->td_lockname,
946 		    sizeof(kp->ki_lockname));
947 	} else {
948 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
949 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
950 	}
951 
952 	if (p->p_state == PRS_NORMAL) { /* approximate. */
953 		if (TD_ON_RUNQ(td) ||
954 		    TD_CAN_RUN(td) ||
955 		    TD_IS_RUNNING(td)) {
956 			kp->ki_stat = SRUN;
957 		} else if (P_SHOULDSTOP(p)) {
958 			kp->ki_stat = SSTOP;
959 		} else if (TD_IS_SLEEPING(td)) {
960 			kp->ki_stat = SSLEEP;
961 		} else if (TD_ON_LOCK(td)) {
962 			kp->ki_stat = SLOCK;
963 		} else {
964 			kp->ki_stat = SWAIT;
965 		}
966 	} else if (p->p_state == PRS_ZOMBIE) {
967 		kp->ki_stat = SZOMB;
968 	} else {
969 		kp->ki_stat = SIDL;
970 	}
971 
972 	/* Things in the thread */
973 	kp->ki_wchan = td->td_wchan;
974 	kp->ki_pri.pri_level = td->td_priority;
975 	kp->ki_pri.pri_native = td->td_base_pri;
976 	kp->ki_lastcpu = td->td_lastcpu;
977 	kp->ki_oncpu = td->td_oncpu;
978 	kp->ki_tdflags = td->td_flags;
979 	kp->ki_tid = td->td_tid;
980 	kp->ki_numthreads = p->p_numthreads;
981 	kp->ki_pcb = td->td_pcb;
982 	kp->ki_kstack = (void *)td->td_kstack;
983 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
984 	kp->ki_pri.pri_class = td->td_pri_class;
985 	kp->ki_pri.pri_user = td->td_user_pri;
986 
987 	if (preferthread) {
988 		rufetchtd(td, &kp->ki_rusage);
989 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
990 		kp->ki_pctcpu = sched_pctcpu(td);
991 		kp->ki_estcpu = td->td_estcpu;
992 		kp->ki_cow = td->td_cow;
993 	}
994 
995 	/* We can't get this anymore but ps etc never used it anyway. */
996 	kp->ki_rqindex = 0;
997 
998 	if (preferthread)
999 		kp->ki_siglist = td->td_siglist;
1000 	kp->ki_sigmask = td->td_sigmask;
1001 	thread_unlock(td);
1002 	if (preferthread)
1003 		PROC_SUNLOCK(p);
1004 }
1005 
1006 /*
1007  * Fill in a kinfo_proc structure for the specified process.
1008  * Must be called with the target process locked.
1009  */
1010 void
1011 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1012 {
1013 
1014 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1015 
1016 	fill_kinfo_proc_only(p, kp);
1017 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1018 	fill_kinfo_aggregate(p, kp);
1019 }
1020 
1021 struct pstats *
1022 pstats_alloc(void)
1023 {
1024 
1025 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1026 }
1027 
1028 /*
1029  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1030  */
1031 void
1032 pstats_fork(struct pstats *src, struct pstats *dst)
1033 {
1034 
1035 	bzero(&dst->pstat_startzero,
1036 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1037 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1038 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1039 }
1040 
1041 void
1042 pstats_free(struct pstats *ps)
1043 {
1044 
1045 	free(ps, M_SUBPROC);
1046 }
1047 
1048 static struct proc *
1049 zpfind_locked(pid_t pid)
1050 {
1051 	struct proc *p;
1052 
1053 	sx_assert(&allproc_lock, SX_LOCKED);
1054 	LIST_FOREACH(p, &zombproc, p_list) {
1055 		if (p->p_pid == pid) {
1056 			PROC_LOCK(p);
1057 			break;
1058 		}
1059 	}
1060 	return (p);
1061 }
1062 
1063 /*
1064  * Locate a zombie process by number
1065  */
1066 struct proc *
1067 zpfind(pid_t pid)
1068 {
1069 	struct proc *p;
1070 
1071 	sx_slock(&allproc_lock);
1072 	p = zpfind_locked(pid);
1073 	sx_sunlock(&allproc_lock);
1074 	return (p);
1075 }
1076 
1077 #ifdef COMPAT_FREEBSD32
1078 
1079 /*
1080  * This function is typically used to copy out the kernel address, so
1081  * it can be replaced by assignment of zero.
1082  */
1083 static inline uint32_t
1084 ptr32_trim(void *ptr)
1085 {
1086 	uintptr_t uptr;
1087 
1088 	uptr = (uintptr_t)ptr;
1089 	return ((uptr > UINT_MAX) ? 0 : uptr);
1090 }
1091 
1092 #define PTRTRIM_CP(src,dst,fld) \
1093 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1094 
1095 static void
1096 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1097 {
1098 	int i;
1099 
1100 	bzero(ki32, sizeof(struct kinfo_proc32));
1101 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1102 	CP(*ki, *ki32, ki_layout);
1103 	PTRTRIM_CP(*ki, *ki32, ki_args);
1104 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1105 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1106 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1107 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1108 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1109 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1110 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1111 	CP(*ki, *ki32, ki_pid);
1112 	CP(*ki, *ki32, ki_ppid);
1113 	CP(*ki, *ki32, ki_pgid);
1114 	CP(*ki, *ki32, ki_tpgid);
1115 	CP(*ki, *ki32, ki_sid);
1116 	CP(*ki, *ki32, ki_tsid);
1117 	CP(*ki, *ki32, ki_jobc);
1118 	CP(*ki, *ki32, ki_tdev);
1119 	CP(*ki, *ki32, ki_siglist);
1120 	CP(*ki, *ki32, ki_sigmask);
1121 	CP(*ki, *ki32, ki_sigignore);
1122 	CP(*ki, *ki32, ki_sigcatch);
1123 	CP(*ki, *ki32, ki_uid);
1124 	CP(*ki, *ki32, ki_ruid);
1125 	CP(*ki, *ki32, ki_svuid);
1126 	CP(*ki, *ki32, ki_rgid);
1127 	CP(*ki, *ki32, ki_svgid);
1128 	CP(*ki, *ki32, ki_ngroups);
1129 	for (i = 0; i < KI_NGROUPS; i++)
1130 		CP(*ki, *ki32, ki_groups[i]);
1131 	CP(*ki, *ki32, ki_size);
1132 	CP(*ki, *ki32, ki_rssize);
1133 	CP(*ki, *ki32, ki_swrss);
1134 	CP(*ki, *ki32, ki_tsize);
1135 	CP(*ki, *ki32, ki_dsize);
1136 	CP(*ki, *ki32, ki_ssize);
1137 	CP(*ki, *ki32, ki_xstat);
1138 	CP(*ki, *ki32, ki_acflag);
1139 	CP(*ki, *ki32, ki_pctcpu);
1140 	CP(*ki, *ki32, ki_estcpu);
1141 	CP(*ki, *ki32, ki_slptime);
1142 	CP(*ki, *ki32, ki_swtime);
1143 	CP(*ki, *ki32, ki_cow);
1144 	CP(*ki, *ki32, ki_runtime);
1145 	TV_CP(*ki, *ki32, ki_start);
1146 	TV_CP(*ki, *ki32, ki_childtime);
1147 	CP(*ki, *ki32, ki_flag);
1148 	CP(*ki, *ki32, ki_kiflag);
1149 	CP(*ki, *ki32, ki_traceflag);
1150 	CP(*ki, *ki32, ki_stat);
1151 	CP(*ki, *ki32, ki_nice);
1152 	CP(*ki, *ki32, ki_lock);
1153 	CP(*ki, *ki32, ki_rqindex);
1154 	CP(*ki, *ki32, ki_oncpu);
1155 	CP(*ki, *ki32, ki_lastcpu);
1156 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1157 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1158 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1159 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1160 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1161 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1162 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1163 	CP(*ki, *ki32, ki_cr_flags);
1164 	CP(*ki, *ki32, ki_jid);
1165 	CP(*ki, *ki32, ki_numthreads);
1166 	CP(*ki, *ki32, ki_tid);
1167 	CP(*ki, *ki32, ki_pri);
1168 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1169 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1170 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1171 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1172 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1173 	CP(*ki, *ki32, ki_sflag);
1174 	CP(*ki, *ki32, ki_tdflags);
1175 }
1176 #endif
1177 
1178 int
1179 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1180 {
1181 	struct thread *td;
1182 	struct kinfo_proc ki;
1183 #ifdef COMPAT_FREEBSD32
1184 	struct kinfo_proc32 ki32;
1185 #endif
1186 	int error;
1187 
1188 	PROC_LOCK_ASSERT(p, MA_OWNED);
1189 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1190 
1191 	error = 0;
1192 	fill_kinfo_proc(p, &ki);
1193 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1194 #ifdef COMPAT_FREEBSD32
1195 		if ((flags & KERN_PROC_MASK32) != 0) {
1196 			freebsd32_kinfo_proc_out(&ki, &ki32);
1197 			error = sbuf_bcat(sb, &ki32, sizeof(ki32));
1198 		} else
1199 #endif
1200 			error = sbuf_bcat(sb, &ki, sizeof(ki));
1201 	} else {
1202 		FOREACH_THREAD_IN_PROC(p, td) {
1203 			fill_kinfo_thread(td, &ki, 1);
1204 #ifdef COMPAT_FREEBSD32
1205 			if ((flags & KERN_PROC_MASK32) != 0) {
1206 				freebsd32_kinfo_proc_out(&ki, &ki32);
1207 				error = sbuf_bcat(sb, &ki32, sizeof(ki32));
1208 			} else
1209 #endif
1210 				error = sbuf_bcat(sb, &ki, sizeof(ki));
1211 			if (error)
1212 				break;
1213 		}
1214 	}
1215 	PROC_UNLOCK(p);
1216 	return (error);
1217 }
1218 
1219 static int
1220 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags,
1221     int doingzomb)
1222 {
1223 	struct sbuf sb;
1224 	struct kinfo_proc ki;
1225 	struct proc *np;
1226 	int error, error2;
1227 	pid_t pid;
1228 
1229 	pid = p->p_pid;
1230 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1231 	error = kern_proc_out(p, &sb, flags);
1232 	error2 = sbuf_finish(&sb);
1233 	sbuf_delete(&sb);
1234 	if (error != 0)
1235 		return (error);
1236 	else if (error2 != 0)
1237 		return (error2);
1238 	if (doingzomb)
1239 		np = zpfind(pid);
1240 	else {
1241 		if (pid == 0)
1242 			return (0);
1243 		np = pfind(pid);
1244 	}
1245 	if (np == NULL)
1246 		return (ESRCH);
1247 	if (np != p) {
1248 		PROC_UNLOCK(np);
1249 		return (ESRCH);
1250 	}
1251 	PROC_UNLOCK(np);
1252 	return (0);
1253 }
1254 
1255 static int
1256 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1257 {
1258 	int *name = (int *)arg1;
1259 	u_int namelen = arg2;
1260 	struct proc *p;
1261 	int flags, doingzomb, oid_number;
1262 	int error = 0;
1263 
1264 	oid_number = oidp->oid_number;
1265 	if (oid_number != KERN_PROC_ALL &&
1266 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1267 		flags = KERN_PROC_NOTHREADS;
1268 	else {
1269 		flags = 0;
1270 		oid_number &= ~KERN_PROC_INC_THREAD;
1271 	}
1272 #ifdef COMPAT_FREEBSD32
1273 	if (req->flags & SCTL_MASK32)
1274 		flags |= KERN_PROC_MASK32;
1275 #endif
1276 	if (oid_number == KERN_PROC_PID) {
1277 		if (namelen != 1)
1278 			return (EINVAL);
1279 		error = sysctl_wire_old_buffer(req, 0);
1280 		if (error)
1281 			return (error);
1282 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1283 		if (error != 0)
1284 			return (error);
1285 		error = sysctl_out_proc(p, req, flags, 0);
1286 		return (error);
1287 	}
1288 
1289 	switch (oid_number) {
1290 	case KERN_PROC_ALL:
1291 		if (namelen != 0)
1292 			return (EINVAL);
1293 		break;
1294 	case KERN_PROC_PROC:
1295 		if (namelen != 0 && namelen != 1)
1296 			return (EINVAL);
1297 		break;
1298 	default:
1299 		if (namelen != 1)
1300 			return (EINVAL);
1301 		break;
1302 	}
1303 
1304 	if (!req->oldptr) {
1305 		/* overestimate by 5 procs */
1306 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1307 		if (error)
1308 			return (error);
1309 	}
1310 	error = sysctl_wire_old_buffer(req, 0);
1311 	if (error != 0)
1312 		return (error);
1313 	sx_slock(&allproc_lock);
1314 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1315 		if (!doingzomb)
1316 			p = LIST_FIRST(&allproc);
1317 		else
1318 			p = LIST_FIRST(&zombproc);
1319 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
1320 			/*
1321 			 * Skip embryonic processes.
1322 			 */
1323 			PROC_LOCK(p);
1324 			if (p->p_state == PRS_NEW) {
1325 				PROC_UNLOCK(p);
1326 				continue;
1327 			}
1328 			KASSERT(p->p_ucred != NULL,
1329 			    ("process credential is NULL for non-NEW proc"));
1330 			/*
1331 			 * Show a user only appropriate processes.
1332 			 */
1333 			if (p_cansee(curthread, p)) {
1334 				PROC_UNLOCK(p);
1335 				continue;
1336 			}
1337 			/*
1338 			 * TODO - make more efficient (see notes below).
1339 			 * do by session.
1340 			 */
1341 			switch (oid_number) {
1342 
1343 			case KERN_PROC_GID:
1344 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1345 					PROC_UNLOCK(p);
1346 					continue;
1347 				}
1348 				break;
1349 
1350 			case KERN_PROC_PGRP:
1351 				/* could do this by traversing pgrp */
1352 				if (p->p_pgrp == NULL ||
1353 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1354 					PROC_UNLOCK(p);
1355 					continue;
1356 				}
1357 				break;
1358 
1359 			case KERN_PROC_RGID:
1360 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1361 					PROC_UNLOCK(p);
1362 					continue;
1363 				}
1364 				break;
1365 
1366 			case KERN_PROC_SESSION:
1367 				if (p->p_session == NULL ||
1368 				    p->p_session->s_sid != (pid_t)name[0]) {
1369 					PROC_UNLOCK(p);
1370 					continue;
1371 				}
1372 				break;
1373 
1374 			case KERN_PROC_TTY:
1375 				if ((p->p_flag & P_CONTROLT) == 0 ||
1376 				    p->p_session == NULL) {
1377 					PROC_UNLOCK(p);
1378 					continue;
1379 				}
1380 				/* XXX proctree_lock */
1381 				SESS_LOCK(p->p_session);
1382 				if (p->p_session->s_ttyp == NULL ||
1383 				    tty_udev(p->p_session->s_ttyp) !=
1384 				    (dev_t)name[0]) {
1385 					SESS_UNLOCK(p->p_session);
1386 					PROC_UNLOCK(p);
1387 					continue;
1388 				}
1389 				SESS_UNLOCK(p->p_session);
1390 				break;
1391 
1392 			case KERN_PROC_UID:
1393 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1394 					PROC_UNLOCK(p);
1395 					continue;
1396 				}
1397 				break;
1398 
1399 			case KERN_PROC_RUID:
1400 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1401 					PROC_UNLOCK(p);
1402 					continue;
1403 				}
1404 				break;
1405 
1406 			case KERN_PROC_PROC:
1407 				break;
1408 
1409 			default:
1410 				break;
1411 
1412 			}
1413 
1414 			error = sysctl_out_proc(p, req, flags, doingzomb);
1415 			if (error) {
1416 				sx_sunlock(&allproc_lock);
1417 				return (error);
1418 			}
1419 		}
1420 	}
1421 	sx_sunlock(&allproc_lock);
1422 	return (0);
1423 }
1424 
1425 struct pargs *
1426 pargs_alloc(int len)
1427 {
1428 	struct pargs *pa;
1429 
1430 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1431 		M_WAITOK);
1432 	refcount_init(&pa->ar_ref, 1);
1433 	pa->ar_length = len;
1434 	return (pa);
1435 }
1436 
1437 static void
1438 pargs_free(struct pargs *pa)
1439 {
1440 
1441 	free(pa, M_PARGS);
1442 }
1443 
1444 void
1445 pargs_hold(struct pargs *pa)
1446 {
1447 
1448 	if (pa == NULL)
1449 		return;
1450 	refcount_acquire(&pa->ar_ref);
1451 }
1452 
1453 void
1454 pargs_drop(struct pargs *pa)
1455 {
1456 
1457 	if (pa == NULL)
1458 		return;
1459 	if (refcount_release(&pa->ar_ref))
1460 		pargs_free(pa);
1461 }
1462 
1463 static int
1464 proc_read_mem(struct thread *td, struct proc *p, vm_offset_t offset, void* buf,
1465     size_t len)
1466 {
1467 	struct iovec iov;
1468 	struct uio uio;
1469 
1470 	iov.iov_base = (caddr_t)buf;
1471 	iov.iov_len = len;
1472 	uio.uio_iov = &iov;
1473 	uio.uio_iovcnt = 1;
1474 	uio.uio_offset = offset;
1475 	uio.uio_resid = (ssize_t)len;
1476 	uio.uio_segflg = UIO_SYSSPACE;
1477 	uio.uio_rw = UIO_READ;
1478 	uio.uio_td = td;
1479 
1480 	return (proc_rwmem(p, &uio));
1481 }
1482 
1483 static int
1484 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1485     size_t len)
1486 {
1487 	size_t i;
1488 	int error;
1489 
1490 	error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, len);
1491 	/*
1492 	 * Reading the chunk may validly return EFAULT if the string is shorter
1493 	 * than the chunk and is aligned at the end of the page, assuming the
1494 	 * next page is not mapped.  So if EFAULT is returned do a fallback to
1495 	 * one byte read loop.
1496 	 */
1497 	if (error == EFAULT) {
1498 		for (i = 0; i < len; i++, buf++, sptr++) {
1499 			error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, 1);
1500 			if (error != 0)
1501 				return (error);
1502 			if (*buf == '\0')
1503 				break;
1504 		}
1505 		error = 0;
1506 	}
1507 	return (error);
1508 }
1509 
1510 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1511 
1512 enum proc_vector_type {
1513 	PROC_ARG,
1514 	PROC_ENV,
1515 	PROC_AUX,
1516 };
1517 
1518 #ifdef COMPAT_FREEBSD32
1519 static int
1520 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1521     size_t *vsizep, enum proc_vector_type type)
1522 {
1523 	struct freebsd32_ps_strings pss;
1524 	Elf32_Auxinfo aux;
1525 	vm_offset_t vptr, ptr;
1526 	uint32_t *proc_vector32;
1527 	char **proc_vector;
1528 	size_t vsize, size;
1529 	int i, error;
1530 
1531 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1532 	    &pss, sizeof(pss));
1533 	if (error != 0)
1534 		return (error);
1535 	switch (type) {
1536 	case PROC_ARG:
1537 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1538 		vsize = pss.ps_nargvstr;
1539 		if (vsize > ARG_MAX)
1540 			return (ENOEXEC);
1541 		size = vsize * sizeof(int32_t);
1542 		break;
1543 	case PROC_ENV:
1544 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1545 		vsize = pss.ps_nenvstr;
1546 		if (vsize > ARG_MAX)
1547 			return (ENOEXEC);
1548 		size = vsize * sizeof(int32_t);
1549 		break;
1550 	case PROC_AUX:
1551 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1552 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1553 		if (vptr % 4 != 0)
1554 			return (ENOEXEC);
1555 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1556 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1557 			if (error != 0)
1558 				return (error);
1559 			if (aux.a_type == AT_NULL)
1560 				break;
1561 			ptr += sizeof(aux);
1562 		}
1563 		if (aux.a_type != AT_NULL)
1564 			return (ENOEXEC);
1565 		vsize = i + 1;
1566 		size = vsize * sizeof(aux);
1567 		break;
1568 	default:
1569 		KASSERT(0, ("Wrong proc vector type: %d", type));
1570 		return (EINVAL);
1571 	}
1572 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1573 	error = proc_read_mem(td, p, vptr, proc_vector32, size);
1574 	if (error != 0)
1575 		goto done;
1576 	if (type == PROC_AUX) {
1577 		*proc_vectorp = (char **)proc_vector32;
1578 		*vsizep = vsize;
1579 		return (0);
1580 	}
1581 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1582 	for (i = 0; i < (int)vsize; i++)
1583 		proc_vector[i] = PTRIN(proc_vector32[i]);
1584 	*proc_vectorp = proc_vector;
1585 	*vsizep = vsize;
1586 done:
1587 	free(proc_vector32, M_TEMP);
1588 	return (error);
1589 }
1590 #endif
1591 
1592 static int
1593 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1594     size_t *vsizep, enum proc_vector_type type)
1595 {
1596 	struct ps_strings pss;
1597 	Elf_Auxinfo aux;
1598 	vm_offset_t vptr, ptr;
1599 	char **proc_vector;
1600 	size_t vsize, size;
1601 	int error, i;
1602 
1603 #ifdef COMPAT_FREEBSD32
1604 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1605 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1606 #endif
1607 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1608 	    &pss, sizeof(pss));
1609 	if (error != 0)
1610 		return (error);
1611 	switch (type) {
1612 	case PROC_ARG:
1613 		vptr = (vm_offset_t)pss.ps_argvstr;
1614 		vsize = pss.ps_nargvstr;
1615 		if (vsize > ARG_MAX)
1616 			return (ENOEXEC);
1617 		size = vsize * sizeof(char *);
1618 		break;
1619 	case PROC_ENV:
1620 		vptr = (vm_offset_t)pss.ps_envstr;
1621 		vsize = pss.ps_nenvstr;
1622 		if (vsize > ARG_MAX)
1623 			return (ENOEXEC);
1624 		size = vsize * sizeof(char *);
1625 		break;
1626 	case PROC_AUX:
1627 		/*
1628 		 * The aux array is just above env array on the stack. Check
1629 		 * that the address is naturally aligned.
1630 		 */
1631 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1632 		    * sizeof(char *);
1633 #if __ELF_WORD_SIZE == 64
1634 		if (vptr % sizeof(uint64_t) != 0)
1635 #else
1636 		if (vptr % sizeof(uint32_t) != 0)
1637 #endif
1638 			return (ENOEXEC);
1639 		/*
1640 		 * We count the array size reading the aux vectors from the
1641 		 * stack until AT_NULL vector is returned.  So (to keep the code
1642 		 * simple) we read the process stack twice: the first time here
1643 		 * to find the size and the second time when copying the vectors
1644 		 * to the allocated proc_vector.
1645 		 */
1646 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1647 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1648 			if (error != 0)
1649 				return (error);
1650 			if (aux.a_type == AT_NULL)
1651 				break;
1652 			ptr += sizeof(aux);
1653 		}
1654 		/*
1655 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1656 		 * not reached AT_NULL, it is most likely we are reading wrong
1657 		 * data: either the process doesn't have auxv array or data has
1658 		 * been modified. Return the error in this case.
1659 		 */
1660 		if (aux.a_type != AT_NULL)
1661 			return (ENOEXEC);
1662 		vsize = i + 1;
1663 		size = vsize * sizeof(aux);
1664 		break;
1665 	default:
1666 		KASSERT(0, ("Wrong proc vector type: %d", type));
1667 		return (EINVAL); /* In case we are built without INVARIANTS. */
1668 	}
1669 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1670 	if (proc_vector == NULL)
1671 		return (ENOMEM);
1672 	error = proc_read_mem(td, p, vptr, proc_vector, size);
1673 	if (error != 0) {
1674 		free(proc_vector, M_TEMP);
1675 		return (error);
1676 	}
1677 	*proc_vectorp = proc_vector;
1678 	*vsizep = vsize;
1679 
1680 	return (0);
1681 }
1682 
1683 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1684 
1685 static int
1686 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1687     enum proc_vector_type type)
1688 {
1689 	size_t done, len, nchr, vsize;
1690 	int error, i;
1691 	char **proc_vector, *sptr;
1692 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1693 
1694 	PROC_ASSERT_HELD(p);
1695 
1696 	/*
1697 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1698 	 */
1699 	nchr = 2 * (PATH_MAX + ARG_MAX);
1700 
1701 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1702 	if (error != 0)
1703 		return (error);
1704 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1705 		/*
1706 		 * The program may have scribbled into its argv array, e.g. to
1707 		 * remove some arguments.  If that has happened, break out
1708 		 * before trying to read from NULL.
1709 		 */
1710 		if (proc_vector[i] == NULL)
1711 			break;
1712 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1713 			error = proc_read_string(td, p, sptr, pss_string,
1714 			    sizeof(pss_string));
1715 			if (error != 0)
1716 				goto done;
1717 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1718 			if (done + len >= nchr)
1719 				len = nchr - done - 1;
1720 			sbuf_bcat(sb, pss_string, len);
1721 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1722 				break;
1723 			done += GET_PS_STRINGS_CHUNK_SZ;
1724 		}
1725 		sbuf_bcat(sb, "", 1);
1726 		done += len + 1;
1727 	}
1728 done:
1729 	free(proc_vector, M_TEMP);
1730 	return (error);
1731 }
1732 
1733 int
1734 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1735 {
1736 
1737 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1738 }
1739 
1740 int
1741 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1742 {
1743 
1744 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1745 }
1746 
1747 int
1748 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1749 {
1750 	size_t vsize, size;
1751 	char **auxv;
1752 	int error;
1753 
1754 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1755 	if (error == 0) {
1756 #ifdef COMPAT_FREEBSD32
1757 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1758 			size = vsize * sizeof(Elf32_Auxinfo);
1759 		else
1760 #endif
1761 			size = vsize * sizeof(Elf_Auxinfo);
1762 		error = sbuf_bcat(sb, auxv, size);
1763 		free(auxv, M_TEMP);
1764 	}
1765 	return (error);
1766 }
1767 
1768 /*
1769  * This sysctl allows a process to retrieve the argument list or process
1770  * title for another process without groping around in the address space
1771  * of the other process.  It also allow a process to set its own "process
1772  * title to a string of its own choice.
1773  */
1774 static int
1775 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1776 {
1777 	int *name = (int *)arg1;
1778 	u_int namelen = arg2;
1779 	struct pargs *newpa, *pa;
1780 	struct proc *p;
1781 	struct sbuf sb;
1782 	int flags, error = 0, error2;
1783 
1784 	if (namelen != 1)
1785 		return (EINVAL);
1786 
1787 	flags = PGET_CANSEE;
1788 	if (req->newptr != NULL)
1789 		flags |= PGET_ISCURRENT;
1790 	error = pget((pid_t)name[0], flags, &p);
1791 	if (error)
1792 		return (error);
1793 
1794 	pa = p->p_args;
1795 	if (pa != NULL) {
1796 		pargs_hold(pa);
1797 		PROC_UNLOCK(p);
1798 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1799 		pargs_drop(pa);
1800 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1801 		_PHOLD(p);
1802 		PROC_UNLOCK(p);
1803 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1804 		error = proc_getargv(curthread, p, &sb);
1805 		error2 = sbuf_finish(&sb);
1806 		PRELE(p);
1807 		sbuf_delete(&sb);
1808 		if (error == 0 && error2 != 0)
1809 			error = error2;
1810 	} else {
1811 		PROC_UNLOCK(p);
1812 	}
1813 	if (error != 0 || req->newptr == NULL)
1814 		return (error);
1815 
1816 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1817 		return (ENOMEM);
1818 	newpa = pargs_alloc(req->newlen);
1819 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1820 	if (error != 0) {
1821 		pargs_free(newpa);
1822 		return (error);
1823 	}
1824 	PROC_LOCK(p);
1825 	pa = p->p_args;
1826 	p->p_args = newpa;
1827 	PROC_UNLOCK(p);
1828 	pargs_drop(pa);
1829 	return (0);
1830 }
1831 
1832 /*
1833  * This sysctl allows a process to retrieve environment of another process.
1834  */
1835 static int
1836 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
1837 {
1838 	int *name = (int *)arg1;
1839 	u_int namelen = arg2;
1840 	struct proc *p;
1841 	struct sbuf sb;
1842 	int error, error2;
1843 
1844 	if (namelen != 1)
1845 		return (EINVAL);
1846 
1847 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1848 	if (error != 0)
1849 		return (error);
1850 	if ((p->p_flag & P_SYSTEM) != 0) {
1851 		PRELE(p);
1852 		return (0);
1853 	}
1854 
1855 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1856 	error = proc_getenvv(curthread, p, &sb);
1857 	error2 = sbuf_finish(&sb);
1858 	PRELE(p);
1859 	sbuf_delete(&sb);
1860 	return (error != 0 ? error : error2);
1861 }
1862 
1863 /*
1864  * This sysctl allows a process to retrieve ELF auxiliary vector of
1865  * another process.
1866  */
1867 static int
1868 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
1869 {
1870 	int *name = (int *)arg1;
1871 	u_int namelen = arg2;
1872 	struct proc *p;
1873 	struct sbuf sb;
1874 	int error, error2;
1875 
1876 	if (namelen != 1)
1877 		return (EINVAL);
1878 
1879 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1880 	if (error != 0)
1881 		return (error);
1882 	if ((p->p_flag & P_SYSTEM) != 0) {
1883 		PRELE(p);
1884 		return (0);
1885 	}
1886 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1887 	error = proc_getauxv(curthread, p, &sb);
1888 	error2 = sbuf_finish(&sb);
1889 	PRELE(p);
1890 	sbuf_delete(&sb);
1891 	return (error != 0 ? error : error2);
1892 }
1893 
1894 /*
1895  * This sysctl allows a process to retrieve the path of the executable for
1896  * itself or another process.
1897  */
1898 static int
1899 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1900 {
1901 	pid_t *pidp = (pid_t *)arg1;
1902 	unsigned int arglen = arg2;
1903 	struct proc *p;
1904 	struct vnode *vp;
1905 	char *retbuf, *freebuf;
1906 	int error;
1907 
1908 	if (arglen != 1)
1909 		return (EINVAL);
1910 	if (*pidp == -1) {	/* -1 means this process */
1911 		p = req->td->td_proc;
1912 	} else {
1913 		error = pget(*pidp, PGET_CANSEE, &p);
1914 		if (error != 0)
1915 			return (error);
1916 	}
1917 
1918 	vp = p->p_textvp;
1919 	if (vp == NULL) {
1920 		if (*pidp != -1)
1921 			PROC_UNLOCK(p);
1922 		return (0);
1923 	}
1924 	vref(vp);
1925 	if (*pidp != -1)
1926 		PROC_UNLOCK(p);
1927 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1928 	vrele(vp);
1929 	if (error)
1930 		return (error);
1931 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1932 	free(freebuf, M_TEMP);
1933 	return (error);
1934 }
1935 
1936 static int
1937 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1938 {
1939 	struct proc *p;
1940 	char *sv_name;
1941 	int *name;
1942 	int namelen;
1943 	int error;
1944 
1945 	namelen = arg2;
1946 	if (namelen != 1)
1947 		return (EINVAL);
1948 
1949 	name = (int *)arg1;
1950 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
1951 	if (error != 0)
1952 		return (error);
1953 	sv_name = p->p_sysent->sv_name;
1954 	PROC_UNLOCK(p);
1955 	return (sysctl_handle_string(oidp, sv_name, 0, req));
1956 }
1957 
1958 #ifdef KINFO_OVMENTRY_SIZE
1959 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
1960 #endif
1961 
1962 #ifdef COMPAT_FREEBSD7
1963 static int
1964 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
1965 {
1966 	vm_map_entry_t entry, tmp_entry;
1967 	unsigned int last_timestamp;
1968 	char *fullpath, *freepath;
1969 	struct kinfo_ovmentry *kve;
1970 	struct vattr va;
1971 	struct ucred *cred;
1972 	int error, *name;
1973 	struct vnode *vp;
1974 	struct proc *p;
1975 	vm_map_t map;
1976 	struct vmspace *vm;
1977 
1978 	name = (int *)arg1;
1979 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1980 	if (error != 0)
1981 		return (error);
1982 	vm = vmspace_acquire_ref(p);
1983 	if (vm == NULL) {
1984 		PRELE(p);
1985 		return (ESRCH);
1986 	}
1987 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1988 
1989 	map = &vm->vm_map;
1990 	vm_map_lock_read(map);
1991 	for (entry = map->header.next; entry != &map->header;
1992 	    entry = entry->next) {
1993 		vm_object_t obj, tobj, lobj;
1994 		vm_offset_t addr;
1995 
1996 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
1997 			continue;
1998 
1999 		bzero(kve, sizeof(*kve));
2000 		kve->kve_structsize = sizeof(*kve);
2001 
2002 		kve->kve_private_resident = 0;
2003 		obj = entry->object.vm_object;
2004 		if (obj != NULL) {
2005 			VM_OBJECT_RLOCK(obj);
2006 			if (obj->shadow_count == 1)
2007 				kve->kve_private_resident =
2008 				    obj->resident_page_count;
2009 		}
2010 		kve->kve_resident = 0;
2011 		addr = entry->start;
2012 		while (addr < entry->end) {
2013 			if (pmap_extract(map->pmap, addr))
2014 				kve->kve_resident++;
2015 			addr += PAGE_SIZE;
2016 		}
2017 
2018 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2019 			if (tobj != obj)
2020 				VM_OBJECT_RLOCK(tobj);
2021 			if (lobj != obj)
2022 				VM_OBJECT_RUNLOCK(lobj);
2023 			lobj = tobj;
2024 		}
2025 
2026 		kve->kve_start = (void*)entry->start;
2027 		kve->kve_end = (void*)entry->end;
2028 		kve->kve_offset = (off_t)entry->offset;
2029 
2030 		if (entry->protection & VM_PROT_READ)
2031 			kve->kve_protection |= KVME_PROT_READ;
2032 		if (entry->protection & VM_PROT_WRITE)
2033 			kve->kve_protection |= KVME_PROT_WRITE;
2034 		if (entry->protection & VM_PROT_EXECUTE)
2035 			kve->kve_protection |= KVME_PROT_EXEC;
2036 
2037 		if (entry->eflags & MAP_ENTRY_COW)
2038 			kve->kve_flags |= KVME_FLAG_COW;
2039 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2040 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2041 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2042 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2043 
2044 		last_timestamp = map->timestamp;
2045 		vm_map_unlock_read(map);
2046 
2047 		kve->kve_fileid = 0;
2048 		kve->kve_fsid = 0;
2049 		freepath = NULL;
2050 		fullpath = "";
2051 		if (lobj) {
2052 			vp = NULL;
2053 			switch (lobj->type) {
2054 			case OBJT_DEFAULT:
2055 				kve->kve_type = KVME_TYPE_DEFAULT;
2056 				break;
2057 			case OBJT_VNODE:
2058 				kve->kve_type = KVME_TYPE_VNODE;
2059 				vp = lobj->handle;
2060 				vref(vp);
2061 				break;
2062 			case OBJT_SWAP:
2063 				kve->kve_type = KVME_TYPE_SWAP;
2064 				break;
2065 			case OBJT_DEVICE:
2066 				kve->kve_type = KVME_TYPE_DEVICE;
2067 				break;
2068 			case OBJT_PHYS:
2069 				kve->kve_type = KVME_TYPE_PHYS;
2070 				break;
2071 			case OBJT_DEAD:
2072 				kve->kve_type = KVME_TYPE_DEAD;
2073 				break;
2074 			case OBJT_SG:
2075 				kve->kve_type = KVME_TYPE_SG;
2076 				break;
2077 			default:
2078 				kve->kve_type = KVME_TYPE_UNKNOWN;
2079 				break;
2080 			}
2081 			if (lobj != obj)
2082 				VM_OBJECT_RUNLOCK(lobj);
2083 
2084 			kve->kve_ref_count = obj->ref_count;
2085 			kve->kve_shadow_count = obj->shadow_count;
2086 			VM_OBJECT_RUNLOCK(obj);
2087 			if (vp != NULL) {
2088 				vn_fullpath(curthread, vp, &fullpath,
2089 				    &freepath);
2090 				cred = curthread->td_ucred;
2091 				vn_lock(vp, LK_SHARED | LK_RETRY);
2092 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2093 					kve->kve_fileid = va.va_fileid;
2094 					kve->kve_fsid = va.va_fsid;
2095 				}
2096 				vput(vp);
2097 			}
2098 		} else {
2099 			kve->kve_type = KVME_TYPE_NONE;
2100 			kve->kve_ref_count = 0;
2101 			kve->kve_shadow_count = 0;
2102 		}
2103 
2104 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2105 		if (freepath != NULL)
2106 			free(freepath, M_TEMP);
2107 
2108 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2109 		vm_map_lock_read(map);
2110 		if (error)
2111 			break;
2112 		if (last_timestamp != map->timestamp) {
2113 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2114 			entry = tmp_entry;
2115 		}
2116 	}
2117 	vm_map_unlock_read(map);
2118 	vmspace_free(vm);
2119 	PRELE(p);
2120 	free(kve, M_TEMP);
2121 	return (error);
2122 }
2123 #endif	/* COMPAT_FREEBSD7 */
2124 
2125 #ifdef KINFO_VMENTRY_SIZE
2126 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2127 #endif
2128 
2129 /*
2130  * Must be called with the process locked and will return unlocked.
2131  */
2132 int
2133 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb)
2134 {
2135 	vm_map_entry_t entry, tmp_entry;
2136 	unsigned int last_timestamp;
2137 	char *fullpath, *freepath;
2138 	struct kinfo_vmentry *kve;
2139 	struct vattr va;
2140 	struct ucred *cred;
2141 	int error;
2142 	struct vnode *vp;
2143 	struct vmspace *vm;
2144 	vm_map_t map;
2145 
2146 	PROC_LOCK_ASSERT(p, MA_OWNED);
2147 
2148 	_PHOLD(p);
2149 	PROC_UNLOCK(p);
2150 	vm = vmspace_acquire_ref(p);
2151 	if (vm == NULL) {
2152 		PRELE(p);
2153 		return (ESRCH);
2154 	}
2155 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2156 
2157 	error = 0;
2158 	map = &vm->vm_map;
2159 	vm_map_lock_read(map);
2160 	for (entry = map->header.next; entry != &map->header;
2161 	    entry = entry->next) {
2162 		vm_object_t obj, tobj, lobj;
2163 		vm_offset_t addr;
2164 		vm_paddr_t locked_pa;
2165 		int mincoreinfo;
2166 
2167 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2168 			continue;
2169 
2170 		bzero(kve, sizeof(*kve));
2171 
2172 		kve->kve_private_resident = 0;
2173 		obj = entry->object.vm_object;
2174 		if (obj != NULL) {
2175 			VM_OBJECT_RLOCK(obj);
2176 			if (obj->shadow_count == 1)
2177 				kve->kve_private_resident =
2178 				    obj->resident_page_count;
2179 		}
2180 		kve->kve_resident = 0;
2181 		addr = entry->start;
2182 		while (addr < entry->end) {
2183 			locked_pa = 0;
2184 			mincoreinfo = pmap_mincore(map->pmap, addr, &locked_pa);
2185 			if (locked_pa != 0)
2186 				vm_page_unlock(PHYS_TO_VM_PAGE(locked_pa));
2187 			if (mincoreinfo & MINCORE_INCORE)
2188 				kve->kve_resident++;
2189 			if (mincoreinfo & MINCORE_SUPER)
2190 				kve->kve_flags |= KVME_FLAG_SUPER;
2191 			addr += PAGE_SIZE;
2192 		}
2193 
2194 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2195 			if (tobj != obj)
2196 				VM_OBJECT_RLOCK(tobj);
2197 			if (lobj != obj)
2198 				VM_OBJECT_RUNLOCK(lobj);
2199 			lobj = tobj;
2200 		}
2201 
2202 		kve->kve_start = entry->start;
2203 		kve->kve_end = entry->end;
2204 		kve->kve_offset = entry->offset;
2205 
2206 		if (entry->protection & VM_PROT_READ)
2207 			kve->kve_protection |= KVME_PROT_READ;
2208 		if (entry->protection & VM_PROT_WRITE)
2209 			kve->kve_protection |= KVME_PROT_WRITE;
2210 		if (entry->protection & VM_PROT_EXECUTE)
2211 			kve->kve_protection |= KVME_PROT_EXEC;
2212 
2213 		if (entry->eflags & MAP_ENTRY_COW)
2214 			kve->kve_flags |= KVME_FLAG_COW;
2215 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2216 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2217 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2218 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2219 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2220 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2221 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2222 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2223 
2224 		last_timestamp = map->timestamp;
2225 		vm_map_unlock_read(map);
2226 
2227 		freepath = NULL;
2228 		fullpath = "";
2229 		if (lobj) {
2230 			vp = NULL;
2231 			switch (lobj->type) {
2232 			case OBJT_DEFAULT:
2233 				kve->kve_type = KVME_TYPE_DEFAULT;
2234 				break;
2235 			case OBJT_VNODE:
2236 				kve->kve_type = KVME_TYPE_VNODE;
2237 				vp = lobj->handle;
2238 				vref(vp);
2239 				break;
2240 			case OBJT_SWAP:
2241 				kve->kve_type = KVME_TYPE_SWAP;
2242 				break;
2243 			case OBJT_DEVICE:
2244 				kve->kve_type = KVME_TYPE_DEVICE;
2245 				break;
2246 			case OBJT_PHYS:
2247 				kve->kve_type = KVME_TYPE_PHYS;
2248 				break;
2249 			case OBJT_DEAD:
2250 				kve->kve_type = KVME_TYPE_DEAD;
2251 				break;
2252 			case OBJT_SG:
2253 				kve->kve_type = KVME_TYPE_SG;
2254 				break;
2255 			default:
2256 				kve->kve_type = KVME_TYPE_UNKNOWN;
2257 				break;
2258 			}
2259 			if (lobj != obj)
2260 				VM_OBJECT_RUNLOCK(lobj);
2261 
2262 			kve->kve_ref_count = obj->ref_count;
2263 			kve->kve_shadow_count = obj->shadow_count;
2264 			VM_OBJECT_RUNLOCK(obj);
2265 			if (vp != NULL) {
2266 				vn_fullpath(curthread, vp, &fullpath,
2267 				    &freepath);
2268 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2269 				cred = curthread->td_ucred;
2270 				vn_lock(vp, LK_SHARED | LK_RETRY);
2271 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2272 					kve->kve_vn_fileid = va.va_fileid;
2273 					kve->kve_vn_fsid = va.va_fsid;
2274 					kve->kve_vn_mode =
2275 					    MAKEIMODE(va.va_type, va.va_mode);
2276 					kve->kve_vn_size = va.va_size;
2277 					kve->kve_vn_rdev = va.va_rdev;
2278 					kve->kve_status = KF_ATTR_VALID;
2279 				}
2280 				vput(vp);
2281 			}
2282 		} else {
2283 			kve->kve_type = KVME_TYPE_NONE;
2284 			kve->kve_ref_count = 0;
2285 			kve->kve_shadow_count = 0;
2286 		}
2287 
2288 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2289 		if (freepath != NULL)
2290 			free(freepath, M_TEMP);
2291 
2292 		/* Pack record size down */
2293 		kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) +
2294 		    strlen(kve->kve_path) + 1;
2295 		kve->kve_structsize = roundup(kve->kve_structsize,
2296 		    sizeof(uint64_t));
2297 		error = sbuf_bcat(sb, kve, kve->kve_structsize);
2298 		vm_map_lock_read(map);
2299 		if (error)
2300 			break;
2301 		if (last_timestamp != map->timestamp) {
2302 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2303 			entry = tmp_entry;
2304 		}
2305 	}
2306 	vm_map_unlock_read(map);
2307 	vmspace_free(vm);
2308 	PRELE(p);
2309 	free(kve, M_TEMP);
2310 	return (error);
2311 }
2312 
2313 static int
2314 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2315 {
2316 	struct proc *p;
2317 	struct sbuf sb;
2318 	int error, error2, *name;
2319 
2320 	name = (int *)arg1;
2321 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2322 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2323 	if (error != 0) {
2324 		sbuf_delete(&sb);
2325 		return (error);
2326 	}
2327 	error = kern_proc_vmmap_out(p, &sb);
2328 	error2 = sbuf_finish(&sb);
2329 	sbuf_delete(&sb);
2330 	return (error != 0 ? error : error2);
2331 }
2332 
2333 #if defined(STACK) || defined(DDB)
2334 static int
2335 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2336 {
2337 	struct kinfo_kstack *kkstp;
2338 	int error, i, *name, numthreads;
2339 	lwpid_t *lwpidarray;
2340 	struct thread *td;
2341 	struct stack *st;
2342 	struct sbuf sb;
2343 	struct proc *p;
2344 
2345 	name = (int *)arg1;
2346 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2347 	if (error != 0)
2348 		return (error);
2349 
2350 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2351 	st = stack_create();
2352 
2353 	lwpidarray = NULL;
2354 	numthreads = 0;
2355 	PROC_LOCK(p);
2356 repeat:
2357 	if (numthreads < p->p_numthreads) {
2358 		if (lwpidarray != NULL) {
2359 			free(lwpidarray, M_TEMP);
2360 			lwpidarray = NULL;
2361 		}
2362 		numthreads = p->p_numthreads;
2363 		PROC_UNLOCK(p);
2364 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2365 		    M_WAITOK | M_ZERO);
2366 		PROC_LOCK(p);
2367 		goto repeat;
2368 	}
2369 	i = 0;
2370 
2371 	/*
2372 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2373 	 * of changes could have taken place.  Should we check to see if the
2374 	 * vmspace has been replaced, or the like, in order to prevent
2375 	 * giving a snapshot that spans, say, execve(2), with some threads
2376 	 * before and some after?  Among other things, the credentials could
2377 	 * have changed, in which case the right to extract debug info might
2378 	 * no longer be assured.
2379 	 */
2380 	FOREACH_THREAD_IN_PROC(p, td) {
2381 		KASSERT(i < numthreads,
2382 		    ("sysctl_kern_proc_kstack: numthreads"));
2383 		lwpidarray[i] = td->td_tid;
2384 		i++;
2385 	}
2386 	numthreads = i;
2387 	for (i = 0; i < numthreads; i++) {
2388 		td = thread_find(p, lwpidarray[i]);
2389 		if (td == NULL) {
2390 			continue;
2391 		}
2392 		bzero(kkstp, sizeof(*kkstp));
2393 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2394 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2395 		thread_lock(td);
2396 		kkstp->kkst_tid = td->td_tid;
2397 		if (TD_IS_SWAPPED(td))
2398 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2399 		else if (TD_IS_RUNNING(td))
2400 			kkstp->kkst_state = KKST_STATE_RUNNING;
2401 		else {
2402 			kkstp->kkst_state = KKST_STATE_STACKOK;
2403 			stack_save_td(st, td);
2404 		}
2405 		thread_unlock(td);
2406 		PROC_UNLOCK(p);
2407 		stack_sbuf_print(&sb, st);
2408 		sbuf_finish(&sb);
2409 		sbuf_delete(&sb);
2410 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2411 		PROC_LOCK(p);
2412 		if (error)
2413 			break;
2414 	}
2415 	_PRELE(p);
2416 	PROC_UNLOCK(p);
2417 	if (lwpidarray != NULL)
2418 		free(lwpidarray, M_TEMP);
2419 	stack_destroy(st);
2420 	free(kkstp, M_TEMP);
2421 	return (error);
2422 }
2423 #endif
2424 
2425 /*
2426  * This sysctl allows a process to retrieve the full list of groups from
2427  * itself or another process.
2428  */
2429 static int
2430 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2431 {
2432 	pid_t *pidp = (pid_t *)arg1;
2433 	unsigned int arglen = arg2;
2434 	struct proc *p;
2435 	struct ucred *cred;
2436 	int error;
2437 
2438 	if (arglen != 1)
2439 		return (EINVAL);
2440 	if (*pidp == -1) {	/* -1 means this process */
2441 		p = req->td->td_proc;
2442 	} else {
2443 		error = pget(*pidp, PGET_CANSEE, &p);
2444 		if (error != 0)
2445 			return (error);
2446 	}
2447 
2448 	cred = crhold(p->p_ucred);
2449 	if (*pidp != -1)
2450 		PROC_UNLOCK(p);
2451 
2452 	error = SYSCTL_OUT(req, cred->cr_groups,
2453 	    cred->cr_ngroups * sizeof(gid_t));
2454 	crfree(cred);
2455 	return (error);
2456 }
2457 
2458 /*
2459  * This sysctl allows a process to retrieve or/and set the resource limit for
2460  * another process.
2461  */
2462 static int
2463 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2464 {
2465 	int *name = (int *)arg1;
2466 	u_int namelen = arg2;
2467 	struct rlimit rlim;
2468 	struct proc *p;
2469 	u_int which;
2470 	int flags, error;
2471 
2472 	if (namelen != 2)
2473 		return (EINVAL);
2474 
2475 	which = (u_int)name[1];
2476 	if (which >= RLIM_NLIMITS)
2477 		return (EINVAL);
2478 
2479 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2480 		return (EINVAL);
2481 
2482 	flags = PGET_HOLD | PGET_NOTWEXIT;
2483 	if (req->newptr != NULL)
2484 		flags |= PGET_CANDEBUG;
2485 	else
2486 		flags |= PGET_CANSEE;
2487 	error = pget((pid_t)name[0], flags, &p);
2488 	if (error != 0)
2489 		return (error);
2490 
2491 	/*
2492 	 * Retrieve limit.
2493 	 */
2494 	if (req->oldptr != NULL) {
2495 		PROC_LOCK(p);
2496 		lim_rlimit(p, which, &rlim);
2497 		PROC_UNLOCK(p);
2498 	}
2499 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2500 	if (error != 0)
2501 		goto errout;
2502 
2503 	/*
2504 	 * Set limit.
2505 	 */
2506 	if (req->newptr != NULL) {
2507 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2508 		if (error == 0)
2509 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2510 	}
2511 
2512 errout:
2513 	PRELE(p);
2514 	return (error);
2515 }
2516 
2517 /*
2518  * This sysctl allows a process to retrieve ps_strings structure location of
2519  * another process.
2520  */
2521 static int
2522 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2523 {
2524 	int *name = (int *)arg1;
2525 	u_int namelen = arg2;
2526 	struct proc *p;
2527 	vm_offset_t ps_strings;
2528 	int error;
2529 #ifdef COMPAT_FREEBSD32
2530 	uint32_t ps_strings32;
2531 #endif
2532 
2533 	if (namelen != 1)
2534 		return (EINVAL);
2535 
2536 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2537 	if (error != 0)
2538 		return (error);
2539 #ifdef COMPAT_FREEBSD32
2540 	if ((req->flags & SCTL_MASK32) != 0) {
2541 		/*
2542 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2543 		 * process.
2544 		 */
2545 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2546 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2547 		PROC_UNLOCK(p);
2548 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2549 		return (error);
2550 	}
2551 #endif
2552 	ps_strings = p->p_sysent->sv_psstrings;
2553 	PROC_UNLOCK(p);
2554 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2555 	return (error);
2556 }
2557 
2558 /*
2559  * This sysctl allows a process to retrieve umask of another process.
2560  */
2561 static int
2562 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2563 {
2564 	int *name = (int *)arg1;
2565 	u_int namelen = arg2;
2566 	struct proc *p;
2567 	int error;
2568 	u_short fd_cmask;
2569 
2570 	if (namelen != 1)
2571 		return (EINVAL);
2572 
2573 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2574 	if (error != 0)
2575 		return (error);
2576 
2577 	FILEDESC_SLOCK(p->p_fd);
2578 	fd_cmask = p->p_fd->fd_cmask;
2579 	FILEDESC_SUNLOCK(p->p_fd);
2580 	PRELE(p);
2581 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2582 	return (error);
2583 }
2584 
2585 /*
2586  * This sysctl allows a process to set and retrieve binary osreldate of
2587  * another process.
2588  */
2589 static int
2590 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2591 {
2592 	int *name = (int *)arg1;
2593 	u_int namelen = arg2;
2594 	struct proc *p;
2595 	int flags, error, osrel;
2596 
2597 	if (namelen != 1)
2598 		return (EINVAL);
2599 
2600 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2601 		return (EINVAL);
2602 
2603 	flags = PGET_HOLD | PGET_NOTWEXIT;
2604 	if (req->newptr != NULL)
2605 		flags |= PGET_CANDEBUG;
2606 	else
2607 		flags |= PGET_CANSEE;
2608 	error = pget((pid_t)name[0], flags, &p);
2609 	if (error != 0)
2610 		return (error);
2611 
2612 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2613 	if (error != 0)
2614 		goto errout;
2615 
2616 	if (req->newptr != NULL) {
2617 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2618 		if (error != 0)
2619 			goto errout;
2620 		if (osrel < 0) {
2621 			error = EINVAL;
2622 			goto errout;
2623 		}
2624 		p->p_osrel = osrel;
2625 	}
2626 errout:
2627 	PRELE(p);
2628 	return (error);
2629 }
2630 
2631 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2632 
2633 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2634 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2635 	"Return entire process table");
2636 
2637 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2638 	sysctl_kern_proc, "Process table");
2639 
2640 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2641 	sysctl_kern_proc, "Process table");
2642 
2643 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2644 	sysctl_kern_proc, "Process table");
2645 
2646 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2647 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2648 
2649 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2650 	sysctl_kern_proc, "Process table");
2651 
2652 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2653 	sysctl_kern_proc, "Process table");
2654 
2655 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2656 	sysctl_kern_proc, "Process table");
2657 
2658 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2659 	sysctl_kern_proc, "Process table");
2660 
2661 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2662 	sysctl_kern_proc, "Return process table, no threads");
2663 
2664 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2665 	CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2666 	sysctl_kern_proc_args, "Process argument list");
2667 
2668 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
2669 	sysctl_kern_proc_env, "Process environment");
2670 
2671 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
2672 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
2673 
2674 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2675 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2676 
2677 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2678 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2679 	"Process syscall vector name (ABI type)");
2680 
2681 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2682 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2683 
2684 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2685 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2686 
2687 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2688 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2689 
2690 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2691 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2692 
2693 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2694 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2695 
2696 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2697 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2698 
2699 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2700 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2701 
2702 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2703 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2704 
2705 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2706 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2707 	"Return process table, no threads");
2708 
2709 #ifdef COMPAT_FREEBSD7
2710 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2711 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2712 #endif
2713 
2714 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2715 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2716 
2717 #if defined(STACK) || defined(DDB)
2718 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2719 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2720 #endif
2721 
2722 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2723 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
2724 
2725 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
2726 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
2727 	"Process resource limits");
2728 
2729 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
2730 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
2731 	"Process ps_strings location");
2732 
2733 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
2734 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
2735 
2736 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
2737 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
2738 	"Process binary osreldate");
2739