xref: /freebsd/sys/kern/kern_proc.c (revision 526e1dc1c0d052b9d2a6cd6da7a16eb09c971c54)
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/sbuf.h>
59 #include <sys/sysent.h>
60 #include <sys/sched.h>
61 #include <sys/smp.h>
62 #include <sys/stack.h>
63 #include <sys/stat.h>
64 #include <sys/sysctl.h>
65 #include <sys/filedesc.h>
66 #include <sys/tty.h>
67 #include <sys/signalvar.h>
68 #include <sys/sdt.h>
69 #include <sys/sx.h>
70 #include <sys/user.h>
71 #include <sys/jail.h>
72 #include <sys/vnode.h>
73 #include <sys/eventhandler.h>
74 
75 #ifdef DDB
76 #include <ddb/ddb.h>
77 #endif
78 
79 #include <vm/vm.h>
80 #include <vm/vm_param.h>
81 #include <vm/vm_extern.h>
82 #include <vm/pmap.h>
83 #include <vm/vm_map.h>
84 #include <vm/vm_object.h>
85 #include <vm/vm_page.h>
86 #include <vm/uma.h>
87 
88 #ifdef COMPAT_FREEBSD32
89 #include <compat/freebsd32/freebsd32.h>
90 #include <compat/freebsd32/freebsd32_util.h>
91 #endif
92 
93 SDT_PROVIDER_DEFINE(proc);
94 SDT_PROBE_DEFINE(proc, kernel, ctor, entry, entry);
95 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 0, "struct proc *");
96 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 1, "int");
97 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 2, "void *");
98 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 3, "int");
99 SDT_PROBE_DEFINE(proc, kernel, ctor, return, return);
100 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 0, "struct proc *");
101 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 1, "int");
102 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 2, "void *");
103 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 3, "int");
104 SDT_PROBE_DEFINE(proc, kernel, dtor, entry, entry);
105 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 0, "struct proc *");
106 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 1, "int");
107 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 2, "void *");
108 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 3, "struct thread *");
109 SDT_PROBE_DEFINE(proc, kernel, dtor, return, return);
110 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 0, "struct proc *");
111 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 1, "int");
112 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 2, "void *");
113 SDT_PROBE_DEFINE(proc, kernel, init, entry, entry);
114 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 0, "struct proc *");
115 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 1, "int");
116 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 2, "int");
117 SDT_PROBE_DEFINE(proc, kernel, init, return, return);
118 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 0, "struct proc *");
119 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 1, "int");
120 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 2, "int");
121 
122 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
123 MALLOC_DEFINE(M_SESSION, "session", "session header");
124 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
125 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
126 
127 static void doenterpgrp(struct proc *, struct pgrp *);
128 static void orphanpg(struct pgrp *pg);
129 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
130 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
131 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
132     int preferthread);
133 static void pgadjustjobc(struct pgrp *pgrp, int entering);
134 static void pgdelete(struct pgrp *);
135 static int proc_ctor(void *mem, int size, void *arg, int flags);
136 static void proc_dtor(void *mem, int size, void *arg);
137 static int proc_init(void *mem, int size, int flags);
138 static void proc_fini(void *mem, int size);
139 static void pargs_free(struct pargs *pa);
140 static struct proc *zpfind_locked(pid_t pid);
141 
142 /*
143  * Other process lists
144  */
145 struct pidhashhead *pidhashtbl;
146 u_long pidhash;
147 struct pgrphashhead *pgrphashtbl;
148 u_long pgrphash;
149 struct proclist allproc;
150 struct proclist zombproc;
151 struct sx allproc_lock;
152 struct sx proctree_lock;
153 struct mtx ppeers_lock;
154 uma_zone_t proc_zone;
155 
156 int kstack_pages = KSTACK_PAGES;
157 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
158     "Kernel stack size in pages");
159 
160 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
161 #ifdef COMPAT_FREEBSD32
162 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
163 #endif
164 
165 /*
166  * Initialize global process hashing structures.
167  */
168 void
169 procinit()
170 {
171 
172 	sx_init(&allproc_lock, "allproc");
173 	sx_init(&proctree_lock, "proctree");
174 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
175 	LIST_INIT(&allproc);
176 	LIST_INIT(&zombproc);
177 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
178 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
179 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
180 	    proc_ctor, proc_dtor, proc_init, proc_fini,
181 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
182 	uihashinit();
183 }
184 
185 /*
186  * Prepare a proc for use.
187  */
188 static int
189 proc_ctor(void *mem, int size, void *arg, int flags)
190 {
191 	struct proc *p;
192 
193 	p = (struct proc *)mem;
194 	SDT_PROBE(proc, kernel, ctor , entry, p, size, arg, flags, 0);
195 	EVENTHANDLER_INVOKE(process_ctor, p);
196 	SDT_PROBE(proc, kernel, ctor , return, p, size, arg, flags, 0);
197 	return (0);
198 }
199 
200 /*
201  * Reclaim a proc after use.
202  */
203 static void
204 proc_dtor(void *mem, int size, void *arg)
205 {
206 	struct proc *p;
207 	struct thread *td;
208 
209 	/* INVARIANTS checks go here */
210 	p = (struct proc *)mem;
211 	td = FIRST_THREAD_IN_PROC(p);
212 	SDT_PROBE(proc, kernel, dtor, entry, p, size, arg, td, 0);
213 	if (td != NULL) {
214 #ifdef INVARIANTS
215 		KASSERT((p->p_numthreads == 1),
216 		    ("bad number of threads in exiting process"));
217 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
218 #endif
219 		/* Free all OSD associated to this thread. */
220 		osd_thread_exit(td);
221 	}
222 	EVENTHANDLER_INVOKE(process_dtor, p);
223 	if (p->p_ksi != NULL)
224 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
225 	SDT_PROBE(proc, kernel, dtor, return, p, size, arg, 0, 0);
226 }
227 
228 /*
229  * Initialize type-stable parts of a proc (when newly created).
230  */
231 static int
232 proc_init(void *mem, int size, int flags)
233 {
234 	struct proc *p;
235 
236 	p = (struct proc *)mem;
237 	SDT_PROBE(proc, kernel, init, entry, p, size, flags, 0, 0);
238 	p->p_sched = (struct p_sched *)&p[1];
239 	bzero(&p->p_mtx, sizeof(struct mtx));
240 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK);
241 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_RECURSE);
242 	cv_init(&p->p_pwait, "ppwait");
243 	cv_init(&p->p_dbgwait, "dbgwait");
244 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
245 	EVENTHANDLER_INVOKE(process_init, p);
246 	p->p_stats = pstats_alloc();
247 	SDT_PROBE(proc, kernel, init, return, p, size, flags, 0, 0);
248 	return (0);
249 }
250 
251 /*
252  * UMA should ensure that this function is never called.
253  * Freeing a proc structure would violate type stability.
254  */
255 static void
256 proc_fini(void *mem, int size)
257 {
258 #ifdef notnow
259 	struct proc *p;
260 
261 	p = (struct proc *)mem;
262 	EVENTHANDLER_INVOKE(process_fini, p);
263 	pstats_free(p->p_stats);
264 	thread_free(FIRST_THREAD_IN_PROC(p));
265 	mtx_destroy(&p->p_mtx);
266 	if (p->p_ksi != NULL)
267 		ksiginfo_free(p->p_ksi);
268 #else
269 	panic("proc reclaimed");
270 #endif
271 }
272 
273 /*
274  * Is p an inferior of the current process?
275  */
276 int
277 inferior(p)
278 	register struct proc *p;
279 {
280 
281 	sx_assert(&proctree_lock, SX_LOCKED);
282 	for (; p != curproc; p = p->p_pptr)
283 		if (p->p_pid == 0)
284 			return (0);
285 	return (1);
286 }
287 
288 struct proc *
289 pfind_locked(pid_t pid)
290 {
291 	struct proc *p;
292 
293 	sx_assert(&allproc_lock, SX_LOCKED);
294 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
295 		if (p->p_pid == pid) {
296 			PROC_LOCK(p);
297 			if (p->p_state == PRS_NEW) {
298 				PROC_UNLOCK(p);
299 				p = NULL;
300 			}
301 			break;
302 		}
303 	}
304 	return (p);
305 }
306 
307 /*
308  * Locate a process by number; return only "live" processes -- i.e., neither
309  * zombies nor newly born but incompletely initialized processes.  By not
310  * returning processes in the PRS_NEW state, we allow callers to avoid
311  * testing for that condition to avoid dereferencing p_ucred, et al.
312  */
313 struct proc *
314 pfind(pid_t pid)
315 {
316 	struct proc *p;
317 
318 	sx_slock(&allproc_lock);
319 	p = pfind_locked(pid);
320 	sx_sunlock(&allproc_lock);
321 	return (p);
322 }
323 
324 static struct proc *
325 pfind_tid_locked(pid_t tid)
326 {
327 	struct proc *p;
328 	struct thread *td;
329 
330 	sx_assert(&allproc_lock, SX_LOCKED);
331 	FOREACH_PROC_IN_SYSTEM(p) {
332 		PROC_LOCK(p);
333 		if (p->p_state == PRS_NEW) {
334 			PROC_UNLOCK(p);
335 			continue;
336 		}
337 		FOREACH_THREAD_IN_PROC(p, td) {
338 			if (td->td_tid == tid)
339 				goto found;
340 		}
341 		PROC_UNLOCK(p);
342 	}
343 found:
344 	return (p);
345 }
346 
347 /*
348  * Locate a process group by number.
349  * The caller must hold proctree_lock.
350  */
351 struct pgrp *
352 pgfind(pgid)
353 	register pid_t pgid;
354 {
355 	register struct pgrp *pgrp;
356 
357 	sx_assert(&proctree_lock, SX_LOCKED);
358 
359 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
360 		if (pgrp->pg_id == pgid) {
361 			PGRP_LOCK(pgrp);
362 			return (pgrp);
363 		}
364 	}
365 	return (NULL);
366 }
367 
368 /*
369  * Locate process and do additional manipulations, depending on flags.
370  */
371 int
372 pget(pid_t pid, int flags, struct proc **pp)
373 {
374 	struct proc *p;
375 	int error;
376 
377 	sx_slock(&allproc_lock);
378 	if (pid <= PID_MAX) {
379 		p = pfind_locked(pid);
380 		if (p == NULL && (flags & PGET_NOTWEXIT) == 0)
381 			p = zpfind_locked(pid);
382 	} else if ((flags & PGET_NOTID) == 0) {
383 		p = pfind_tid_locked(pid);
384 	} else {
385 		p = NULL;
386 	}
387 	sx_sunlock(&allproc_lock);
388 	if (p == NULL)
389 		return (ESRCH);
390 	if ((flags & PGET_CANSEE) != 0) {
391 		error = p_cansee(curthread, p);
392 		if (error != 0)
393 			goto errout;
394 	}
395 	if ((flags & PGET_CANDEBUG) != 0) {
396 		error = p_candebug(curthread, p);
397 		if (error != 0)
398 			goto errout;
399 	}
400 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
401 		error = EPERM;
402 		goto errout;
403 	}
404 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
405 		error = ESRCH;
406 		goto errout;
407 	}
408 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
409 		/*
410 		 * XXXRW: Not clear ESRCH is the right error during proc
411 		 * execve().
412 		 */
413 		error = ESRCH;
414 		goto errout;
415 	}
416 	if ((flags & PGET_HOLD) != 0) {
417 		_PHOLD(p);
418 		PROC_UNLOCK(p);
419 	}
420 	*pp = p;
421 	return (0);
422 errout:
423 	PROC_UNLOCK(p);
424 	return (error);
425 }
426 
427 /*
428  * Create a new process group.
429  * pgid must be equal to the pid of p.
430  * Begin a new session if required.
431  */
432 int
433 enterpgrp(p, pgid, pgrp, sess)
434 	register struct proc *p;
435 	pid_t pgid;
436 	struct pgrp *pgrp;
437 	struct session *sess;
438 {
439 
440 	sx_assert(&proctree_lock, SX_XLOCKED);
441 
442 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
443 	KASSERT(p->p_pid == pgid,
444 	    ("enterpgrp: new pgrp and pid != pgid"));
445 	KASSERT(pgfind(pgid) == NULL,
446 	    ("enterpgrp: pgrp with pgid exists"));
447 	KASSERT(!SESS_LEADER(p),
448 	    ("enterpgrp: session leader attempted setpgrp"));
449 
450 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
451 
452 	if (sess != NULL) {
453 		/*
454 		 * new session
455 		 */
456 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
457 		PROC_LOCK(p);
458 		p->p_flag &= ~P_CONTROLT;
459 		PROC_UNLOCK(p);
460 		PGRP_LOCK(pgrp);
461 		sess->s_leader = p;
462 		sess->s_sid = p->p_pid;
463 		refcount_init(&sess->s_count, 1);
464 		sess->s_ttyvp = NULL;
465 		sess->s_ttydp = NULL;
466 		sess->s_ttyp = NULL;
467 		bcopy(p->p_session->s_login, sess->s_login,
468 			    sizeof(sess->s_login));
469 		pgrp->pg_session = sess;
470 		KASSERT(p == curproc,
471 		    ("enterpgrp: mksession and p != curproc"));
472 	} else {
473 		pgrp->pg_session = p->p_session;
474 		sess_hold(pgrp->pg_session);
475 		PGRP_LOCK(pgrp);
476 	}
477 	pgrp->pg_id = pgid;
478 	LIST_INIT(&pgrp->pg_members);
479 
480 	/*
481 	 * As we have an exclusive lock of proctree_lock,
482 	 * this should not deadlock.
483 	 */
484 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
485 	pgrp->pg_jobc = 0;
486 	SLIST_INIT(&pgrp->pg_sigiolst);
487 	PGRP_UNLOCK(pgrp);
488 
489 	doenterpgrp(p, pgrp);
490 
491 	return (0);
492 }
493 
494 /*
495  * Move p to an existing process group
496  */
497 int
498 enterthispgrp(p, pgrp)
499 	register struct proc *p;
500 	struct pgrp *pgrp;
501 {
502 
503 	sx_assert(&proctree_lock, SX_XLOCKED);
504 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
505 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
506 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
507 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
508 	KASSERT(pgrp->pg_session == p->p_session,
509 		("%s: pgrp's session %p, p->p_session %p.\n",
510 		__func__,
511 		pgrp->pg_session,
512 		p->p_session));
513 	KASSERT(pgrp != p->p_pgrp,
514 		("%s: p belongs to pgrp.", __func__));
515 
516 	doenterpgrp(p, pgrp);
517 
518 	return (0);
519 }
520 
521 /*
522  * Move p to a process group
523  */
524 static void
525 doenterpgrp(p, pgrp)
526 	struct proc *p;
527 	struct pgrp *pgrp;
528 {
529 	struct pgrp *savepgrp;
530 
531 	sx_assert(&proctree_lock, SX_XLOCKED);
532 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
533 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
534 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
535 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
536 
537 	savepgrp = p->p_pgrp;
538 
539 	/*
540 	 * Adjust eligibility of affected pgrps to participate in job control.
541 	 * Increment eligibility counts before decrementing, otherwise we
542 	 * could reach 0 spuriously during the first call.
543 	 */
544 	fixjobc(p, pgrp, 1);
545 	fixjobc(p, p->p_pgrp, 0);
546 
547 	PGRP_LOCK(pgrp);
548 	PGRP_LOCK(savepgrp);
549 	PROC_LOCK(p);
550 	LIST_REMOVE(p, p_pglist);
551 	p->p_pgrp = pgrp;
552 	PROC_UNLOCK(p);
553 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
554 	PGRP_UNLOCK(savepgrp);
555 	PGRP_UNLOCK(pgrp);
556 	if (LIST_EMPTY(&savepgrp->pg_members))
557 		pgdelete(savepgrp);
558 }
559 
560 /*
561  * remove process from process group
562  */
563 int
564 leavepgrp(p)
565 	register struct proc *p;
566 {
567 	struct pgrp *savepgrp;
568 
569 	sx_assert(&proctree_lock, SX_XLOCKED);
570 	savepgrp = p->p_pgrp;
571 	PGRP_LOCK(savepgrp);
572 	PROC_LOCK(p);
573 	LIST_REMOVE(p, p_pglist);
574 	p->p_pgrp = NULL;
575 	PROC_UNLOCK(p);
576 	PGRP_UNLOCK(savepgrp);
577 	if (LIST_EMPTY(&savepgrp->pg_members))
578 		pgdelete(savepgrp);
579 	return (0);
580 }
581 
582 /*
583  * delete a process group
584  */
585 static void
586 pgdelete(pgrp)
587 	register struct pgrp *pgrp;
588 {
589 	struct session *savesess;
590 	struct tty *tp;
591 
592 	sx_assert(&proctree_lock, SX_XLOCKED);
593 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
594 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
595 
596 	/*
597 	 * Reset any sigio structures pointing to us as a result of
598 	 * F_SETOWN with our pgid.
599 	 */
600 	funsetownlst(&pgrp->pg_sigiolst);
601 
602 	PGRP_LOCK(pgrp);
603 	tp = pgrp->pg_session->s_ttyp;
604 	LIST_REMOVE(pgrp, pg_hash);
605 	savesess = pgrp->pg_session;
606 	PGRP_UNLOCK(pgrp);
607 
608 	/* Remove the reference to the pgrp before deallocating it. */
609 	if (tp != NULL) {
610 		tty_lock(tp);
611 		tty_rel_pgrp(tp, pgrp);
612 	}
613 
614 	mtx_destroy(&pgrp->pg_mtx);
615 	free(pgrp, M_PGRP);
616 	sess_release(savesess);
617 }
618 
619 static void
620 pgadjustjobc(pgrp, entering)
621 	struct pgrp *pgrp;
622 	int entering;
623 {
624 
625 	PGRP_LOCK(pgrp);
626 	if (entering)
627 		pgrp->pg_jobc++;
628 	else {
629 		--pgrp->pg_jobc;
630 		if (pgrp->pg_jobc == 0)
631 			orphanpg(pgrp);
632 	}
633 	PGRP_UNLOCK(pgrp);
634 }
635 
636 /*
637  * Adjust pgrp jobc counters when specified process changes process group.
638  * We count the number of processes in each process group that "qualify"
639  * the group for terminal job control (those with a parent in a different
640  * process group of the same session).  If that count reaches zero, the
641  * process group becomes orphaned.  Check both the specified process'
642  * process group and that of its children.
643  * entering == 0 => p is leaving specified group.
644  * entering == 1 => p is entering specified group.
645  */
646 void
647 fixjobc(p, pgrp, entering)
648 	register struct proc *p;
649 	register struct pgrp *pgrp;
650 	int entering;
651 {
652 	register struct pgrp *hispgrp;
653 	register struct session *mysession;
654 
655 	sx_assert(&proctree_lock, SX_LOCKED);
656 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
657 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
658 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
659 
660 	/*
661 	 * Check p's parent to see whether p qualifies its own process
662 	 * group; if so, adjust count for p's process group.
663 	 */
664 	mysession = pgrp->pg_session;
665 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
666 	    hispgrp->pg_session == mysession)
667 		pgadjustjobc(pgrp, entering);
668 
669 	/*
670 	 * Check this process' children to see whether they qualify
671 	 * their process groups; if so, adjust counts for children's
672 	 * process groups.
673 	 */
674 	LIST_FOREACH(p, &p->p_children, p_sibling) {
675 		hispgrp = p->p_pgrp;
676 		if (hispgrp == pgrp ||
677 		    hispgrp->pg_session != mysession)
678 			continue;
679 		PROC_LOCK(p);
680 		if (p->p_state == PRS_ZOMBIE) {
681 			PROC_UNLOCK(p);
682 			continue;
683 		}
684 		PROC_UNLOCK(p);
685 		pgadjustjobc(hispgrp, entering);
686 	}
687 }
688 
689 /*
690  * A process group has become orphaned;
691  * if there are any stopped processes in the group,
692  * hang-up all process in that group.
693  */
694 static void
695 orphanpg(pg)
696 	struct pgrp *pg;
697 {
698 	register struct proc *p;
699 
700 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
701 
702 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
703 		PROC_LOCK(p);
704 		if (P_SHOULDSTOP(p)) {
705 			PROC_UNLOCK(p);
706 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
707 				PROC_LOCK(p);
708 				kern_psignal(p, SIGHUP);
709 				kern_psignal(p, SIGCONT);
710 				PROC_UNLOCK(p);
711 			}
712 			return;
713 		}
714 		PROC_UNLOCK(p);
715 	}
716 }
717 
718 void
719 sess_hold(struct session *s)
720 {
721 
722 	refcount_acquire(&s->s_count);
723 }
724 
725 void
726 sess_release(struct session *s)
727 {
728 
729 	if (refcount_release(&s->s_count)) {
730 		if (s->s_ttyp != NULL) {
731 			tty_lock(s->s_ttyp);
732 			tty_rel_sess(s->s_ttyp, s);
733 		}
734 		mtx_destroy(&s->s_mtx);
735 		free(s, M_SESSION);
736 	}
737 }
738 
739 #ifdef DDB
740 
741 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
742 {
743 	register struct pgrp *pgrp;
744 	register struct proc *p;
745 	register int i;
746 
747 	for (i = 0; i <= pgrphash; i++) {
748 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
749 			printf("\tindx %d\n", i);
750 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
751 				printf(
752 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
753 				    (void *)pgrp, (long)pgrp->pg_id,
754 				    (void *)pgrp->pg_session,
755 				    pgrp->pg_session->s_count,
756 				    (void *)LIST_FIRST(&pgrp->pg_members));
757 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
758 					printf("\t\tpid %ld addr %p pgrp %p\n",
759 					    (long)p->p_pid, (void *)p,
760 					    (void *)p->p_pgrp);
761 				}
762 			}
763 		}
764 	}
765 }
766 #endif /* DDB */
767 
768 /*
769  * Calculate the kinfo_proc members which contain process-wide
770  * informations.
771  * Must be called with the target process locked.
772  */
773 static void
774 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
775 {
776 	struct thread *td;
777 
778 	PROC_LOCK_ASSERT(p, MA_OWNED);
779 
780 	kp->ki_estcpu = 0;
781 	kp->ki_pctcpu = 0;
782 	FOREACH_THREAD_IN_PROC(p, td) {
783 		thread_lock(td);
784 		kp->ki_pctcpu += sched_pctcpu(td);
785 		kp->ki_estcpu += td->td_estcpu;
786 		thread_unlock(td);
787 	}
788 }
789 
790 /*
791  * Clear kinfo_proc and fill in any information that is common
792  * to all threads in the process.
793  * Must be called with the target process locked.
794  */
795 static void
796 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
797 {
798 	struct thread *td0;
799 	struct tty *tp;
800 	struct session *sp;
801 	struct ucred *cred;
802 	struct sigacts *ps;
803 
804 	PROC_LOCK_ASSERT(p, MA_OWNED);
805 	bzero(kp, sizeof(*kp));
806 
807 	kp->ki_structsize = sizeof(*kp);
808 	kp->ki_paddr = p;
809 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
810 	kp->ki_args = p->p_args;
811 	kp->ki_textvp = p->p_textvp;
812 #ifdef KTRACE
813 	kp->ki_tracep = p->p_tracevp;
814 	kp->ki_traceflag = p->p_traceflag;
815 #endif
816 	kp->ki_fd = p->p_fd;
817 	kp->ki_vmspace = p->p_vmspace;
818 	kp->ki_flag = p->p_flag;
819 	cred = p->p_ucred;
820 	if (cred) {
821 		kp->ki_uid = cred->cr_uid;
822 		kp->ki_ruid = cred->cr_ruid;
823 		kp->ki_svuid = cred->cr_svuid;
824 		kp->ki_cr_flags = 0;
825 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
826 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
827 		/* XXX bde doesn't like KI_NGROUPS */
828 		if (cred->cr_ngroups > KI_NGROUPS) {
829 			kp->ki_ngroups = KI_NGROUPS;
830 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
831 		} else
832 			kp->ki_ngroups = cred->cr_ngroups;
833 		bcopy(cred->cr_groups, kp->ki_groups,
834 		    kp->ki_ngroups * sizeof(gid_t));
835 		kp->ki_rgid = cred->cr_rgid;
836 		kp->ki_svgid = cred->cr_svgid;
837 		/* If jailed(cred), emulate the old P_JAILED flag. */
838 		if (jailed(cred)) {
839 			kp->ki_flag |= P_JAILED;
840 			/* If inside the jail, use 0 as a jail ID. */
841 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
842 				kp->ki_jid = cred->cr_prison->pr_id;
843 		}
844 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
845 		    sizeof(kp->ki_loginclass));
846 	}
847 	ps = p->p_sigacts;
848 	if (ps) {
849 		mtx_lock(&ps->ps_mtx);
850 		kp->ki_sigignore = ps->ps_sigignore;
851 		kp->ki_sigcatch = ps->ps_sigcatch;
852 		mtx_unlock(&ps->ps_mtx);
853 	}
854 	if (p->p_state != PRS_NEW &&
855 	    p->p_state != PRS_ZOMBIE &&
856 	    p->p_vmspace != NULL) {
857 		struct vmspace *vm = p->p_vmspace;
858 
859 		kp->ki_size = vm->vm_map.size;
860 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
861 		FOREACH_THREAD_IN_PROC(p, td0) {
862 			if (!TD_IS_SWAPPED(td0))
863 				kp->ki_rssize += td0->td_kstack_pages;
864 		}
865 		kp->ki_swrss = vm->vm_swrss;
866 		kp->ki_tsize = vm->vm_tsize;
867 		kp->ki_dsize = vm->vm_dsize;
868 		kp->ki_ssize = vm->vm_ssize;
869 	} else if (p->p_state == PRS_ZOMBIE)
870 		kp->ki_stat = SZOMB;
871 	if (kp->ki_flag & P_INMEM)
872 		kp->ki_sflag = PS_INMEM;
873 	else
874 		kp->ki_sflag = 0;
875 	/* Calculate legacy swtime as seconds since 'swtick'. */
876 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
877 	kp->ki_pid = p->p_pid;
878 	kp->ki_nice = p->p_nice;
879 	kp->ki_start = p->p_stats->p_start;
880 	timevaladd(&kp->ki_start, &boottime);
881 	PROC_SLOCK(p);
882 	rufetch(p, &kp->ki_rusage);
883 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
884 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
885 	PROC_SUNLOCK(p);
886 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
887 	/* Some callers want child times in a single value. */
888 	kp->ki_childtime = kp->ki_childstime;
889 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
890 
891 	FOREACH_THREAD_IN_PROC(p, td0)
892 		kp->ki_cow += td0->td_cow;
893 
894 	tp = NULL;
895 	if (p->p_pgrp) {
896 		kp->ki_pgid = p->p_pgrp->pg_id;
897 		kp->ki_jobc = p->p_pgrp->pg_jobc;
898 		sp = p->p_pgrp->pg_session;
899 
900 		if (sp != NULL) {
901 			kp->ki_sid = sp->s_sid;
902 			SESS_LOCK(sp);
903 			strlcpy(kp->ki_login, sp->s_login,
904 			    sizeof(kp->ki_login));
905 			if (sp->s_ttyvp)
906 				kp->ki_kiflag |= KI_CTTY;
907 			if (SESS_LEADER(p))
908 				kp->ki_kiflag |= KI_SLEADER;
909 			/* XXX proctree_lock */
910 			tp = sp->s_ttyp;
911 			SESS_UNLOCK(sp);
912 		}
913 	}
914 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
915 		kp->ki_tdev = tty_udev(tp);
916 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
917 		if (tp->t_session)
918 			kp->ki_tsid = tp->t_session->s_sid;
919 	} else
920 		kp->ki_tdev = NODEV;
921 	if (p->p_comm[0] != '\0')
922 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
923 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
924 	    p->p_sysent->sv_name[0] != '\0')
925 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
926 	kp->ki_siglist = p->p_siglist;
927 	kp->ki_xstat = p->p_xstat;
928 	kp->ki_acflag = p->p_acflag;
929 	kp->ki_lock = p->p_lock;
930 	if (p->p_pptr)
931 		kp->ki_ppid = p->p_pptr->p_pid;
932 }
933 
934 /*
935  * Fill in information that is thread specific.  Must be called with
936  * target process locked.  If 'preferthread' is set, overwrite certain
937  * process-related fields that are maintained for both threads and
938  * processes.
939  */
940 static void
941 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
942 {
943 	struct proc *p;
944 
945 	p = td->td_proc;
946 	kp->ki_tdaddr = td;
947 	PROC_LOCK_ASSERT(p, MA_OWNED);
948 
949 	if (preferthread)
950 		PROC_SLOCK(p);
951 	thread_lock(td);
952 	if (td->td_wmesg != NULL)
953 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
954 	else
955 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
956 	strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname));
957 	if (TD_ON_LOCK(td)) {
958 		kp->ki_kiflag |= KI_LOCKBLOCK;
959 		strlcpy(kp->ki_lockname, td->td_lockname,
960 		    sizeof(kp->ki_lockname));
961 	} else {
962 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
963 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
964 	}
965 
966 	if (p->p_state == PRS_NORMAL) { /* approximate. */
967 		if (TD_ON_RUNQ(td) ||
968 		    TD_CAN_RUN(td) ||
969 		    TD_IS_RUNNING(td)) {
970 			kp->ki_stat = SRUN;
971 		} else if (P_SHOULDSTOP(p)) {
972 			kp->ki_stat = SSTOP;
973 		} else if (TD_IS_SLEEPING(td)) {
974 			kp->ki_stat = SSLEEP;
975 		} else if (TD_ON_LOCK(td)) {
976 			kp->ki_stat = SLOCK;
977 		} else {
978 			kp->ki_stat = SWAIT;
979 		}
980 	} else if (p->p_state == PRS_ZOMBIE) {
981 		kp->ki_stat = SZOMB;
982 	} else {
983 		kp->ki_stat = SIDL;
984 	}
985 
986 	/* Things in the thread */
987 	kp->ki_wchan = td->td_wchan;
988 	kp->ki_pri.pri_level = td->td_priority;
989 	kp->ki_pri.pri_native = td->td_base_pri;
990 	kp->ki_lastcpu = td->td_lastcpu;
991 	kp->ki_oncpu = td->td_oncpu;
992 	kp->ki_tdflags = td->td_flags;
993 	kp->ki_tid = td->td_tid;
994 	kp->ki_numthreads = p->p_numthreads;
995 	kp->ki_pcb = td->td_pcb;
996 	kp->ki_kstack = (void *)td->td_kstack;
997 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
998 	kp->ki_pri.pri_class = td->td_pri_class;
999 	kp->ki_pri.pri_user = td->td_user_pri;
1000 
1001 	if (preferthread) {
1002 		rufetchtd(td, &kp->ki_rusage);
1003 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1004 		kp->ki_pctcpu = sched_pctcpu(td);
1005 		kp->ki_estcpu = td->td_estcpu;
1006 		kp->ki_cow = td->td_cow;
1007 	}
1008 
1009 	/* We can't get this anymore but ps etc never used it anyway. */
1010 	kp->ki_rqindex = 0;
1011 
1012 	if (preferthread)
1013 		kp->ki_siglist = td->td_siglist;
1014 	kp->ki_sigmask = td->td_sigmask;
1015 	thread_unlock(td);
1016 	if (preferthread)
1017 		PROC_SUNLOCK(p);
1018 }
1019 
1020 /*
1021  * Fill in a kinfo_proc structure for the specified process.
1022  * Must be called with the target process locked.
1023  */
1024 void
1025 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1026 {
1027 
1028 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1029 
1030 	fill_kinfo_proc_only(p, kp);
1031 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1032 	fill_kinfo_aggregate(p, kp);
1033 }
1034 
1035 struct pstats *
1036 pstats_alloc(void)
1037 {
1038 
1039 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1040 }
1041 
1042 /*
1043  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1044  */
1045 void
1046 pstats_fork(struct pstats *src, struct pstats *dst)
1047 {
1048 
1049 	bzero(&dst->pstat_startzero,
1050 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1051 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1052 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1053 }
1054 
1055 void
1056 pstats_free(struct pstats *ps)
1057 {
1058 
1059 	free(ps, M_SUBPROC);
1060 }
1061 
1062 static struct proc *
1063 zpfind_locked(pid_t pid)
1064 {
1065 	struct proc *p;
1066 
1067 	sx_assert(&allproc_lock, SX_LOCKED);
1068 	LIST_FOREACH(p, &zombproc, p_list) {
1069 		if (p->p_pid == pid) {
1070 			PROC_LOCK(p);
1071 			break;
1072 		}
1073 	}
1074 	return (p);
1075 }
1076 
1077 /*
1078  * Locate a zombie process by number
1079  */
1080 struct proc *
1081 zpfind(pid_t pid)
1082 {
1083 	struct proc *p;
1084 
1085 	sx_slock(&allproc_lock);
1086 	p = zpfind_locked(pid);
1087 	sx_sunlock(&allproc_lock);
1088 	return (p);
1089 }
1090 
1091 #define KERN_PROC_ZOMBMASK	0x3
1092 #define KERN_PROC_NOTHREADS	0x4
1093 
1094 #ifdef COMPAT_FREEBSD32
1095 
1096 /*
1097  * This function is typically used to copy out the kernel address, so
1098  * it can be replaced by assignment of zero.
1099  */
1100 static inline uint32_t
1101 ptr32_trim(void *ptr)
1102 {
1103 	uintptr_t uptr;
1104 
1105 	uptr = (uintptr_t)ptr;
1106 	return ((uptr > UINT_MAX) ? 0 : uptr);
1107 }
1108 
1109 #define PTRTRIM_CP(src,dst,fld) \
1110 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1111 
1112 static void
1113 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1114 {
1115 	int i;
1116 
1117 	bzero(ki32, sizeof(struct kinfo_proc32));
1118 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1119 	CP(*ki, *ki32, ki_layout);
1120 	PTRTRIM_CP(*ki, *ki32, ki_args);
1121 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1122 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1123 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1124 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1125 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1126 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1127 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1128 	CP(*ki, *ki32, ki_pid);
1129 	CP(*ki, *ki32, ki_ppid);
1130 	CP(*ki, *ki32, ki_pgid);
1131 	CP(*ki, *ki32, ki_tpgid);
1132 	CP(*ki, *ki32, ki_sid);
1133 	CP(*ki, *ki32, ki_tsid);
1134 	CP(*ki, *ki32, ki_jobc);
1135 	CP(*ki, *ki32, ki_tdev);
1136 	CP(*ki, *ki32, ki_siglist);
1137 	CP(*ki, *ki32, ki_sigmask);
1138 	CP(*ki, *ki32, ki_sigignore);
1139 	CP(*ki, *ki32, ki_sigcatch);
1140 	CP(*ki, *ki32, ki_uid);
1141 	CP(*ki, *ki32, ki_ruid);
1142 	CP(*ki, *ki32, ki_svuid);
1143 	CP(*ki, *ki32, ki_rgid);
1144 	CP(*ki, *ki32, ki_svgid);
1145 	CP(*ki, *ki32, ki_ngroups);
1146 	for (i = 0; i < KI_NGROUPS; i++)
1147 		CP(*ki, *ki32, ki_groups[i]);
1148 	CP(*ki, *ki32, ki_size);
1149 	CP(*ki, *ki32, ki_rssize);
1150 	CP(*ki, *ki32, ki_swrss);
1151 	CP(*ki, *ki32, ki_tsize);
1152 	CP(*ki, *ki32, ki_dsize);
1153 	CP(*ki, *ki32, ki_ssize);
1154 	CP(*ki, *ki32, ki_xstat);
1155 	CP(*ki, *ki32, ki_acflag);
1156 	CP(*ki, *ki32, ki_pctcpu);
1157 	CP(*ki, *ki32, ki_estcpu);
1158 	CP(*ki, *ki32, ki_slptime);
1159 	CP(*ki, *ki32, ki_swtime);
1160 	CP(*ki, *ki32, ki_cow);
1161 	CP(*ki, *ki32, ki_runtime);
1162 	TV_CP(*ki, *ki32, ki_start);
1163 	TV_CP(*ki, *ki32, ki_childtime);
1164 	CP(*ki, *ki32, ki_flag);
1165 	CP(*ki, *ki32, ki_kiflag);
1166 	CP(*ki, *ki32, ki_traceflag);
1167 	CP(*ki, *ki32, ki_stat);
1168 	CP(*ki, *ki32, ki_nice);
1169 	CP(*ki, *ki32, ki_lock);
1170 	CP(*ki, *ki32, ki_rqindex);
1171 	CP(*ki, *ki32, ki_oncpu);
1172 	CP(*ki, *ki32, ki_lastcpu);
1173 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1174 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1175 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1176 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1177 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1178 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1179 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1180 	CP(*ki, *ki32, ki_cr_flags);
1181 	CP(*ki, *ki32, ki_jid);
1182 	CP(*ki, *ki32, ki_numthreads);
1183 	CP(*ki, *ki32, ki_tid);
1184 	CP(*ki, *ki32, ki_pri);
1185 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1186 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1187 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1188 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1189 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1190 	CP(*ki, *ki32, ki_sflag);
1191 	CP(*ki, *ki32, ki_tdflags);
1192 }
1193 
1194 static int
1195 sysctl_out_proc_copyout(struct kinfo_proc *ki, struct sysctl_req *req)
1196 {
1197 	struct kinfo_proc32 ki32;
1198 	int error;
1199 
1200 	if (req->flags & SCTL_MASK32) {
1201 		freebsd32_kinfo_proc_out(ki, &ki32);
1202 		error = SYSCTL_OUT(req, &ki32, sizeof(struct kinfo_proc32));
1203 	} else
1204 		error = SYSCTL_OUT(req, ki, sizeof(struct kinfo_proc));
1205 	return (error);
1206 }
1207 #else
1208 static int
1209 sysctl_out_proc_copyout(struct kinfo_proc *ki, struct sysctl_req *req)
1210 {
1211 
1212 	return (SYSCTL_OUT(req, ki, sizeof(struct kinfo_proc)));
1213 }
1214 #endif
1215 
1216 /*
1217  * Must be called with the process locked and will return with it unlocked.
1218  */
1219 static int
1220 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1221 {
1222 	struct thread *td;
1223 	struct kinfo_proc kinfo_proc;
1224 	int error = 0;
1225 	struct proc *np;
1226 	pid_t pid = p->p_pid;
1227 
1228 	PROC_LOCK_ASSERT(p, MA_OWNED);
1229 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1230 
1231 	fill_kinfo_proc(p, &kinfo_proc);
1232 	if (flags & KERN_PROC_NOTHREADS)
1233 		error = sysctl_out_proc_copyout(&kinfo_proc, req);
1234 	else {
1235 		FOREACH_THREAD_IN_PROC(p, td) {
1236 			fill_kinfo_thread(td, &kinfo_proc, 1);
1237 			error = sysctl_out_proc_copyout(&kinfo_proc, req);
1238 			if (error)
1239 				break;
1240 		}
1241 	}
1242 	PROC_UNLOCK(p);
1243 	if (error)
1244 		return (error);
1245 	if (flags & KERN_PROC_ZOMBMASK)
1246 		np = zpfind(pid);
1247 	else {
1248 		if (pid == 0)
1249 			return (0);
1250 		np = pfind(pid);
1251 	}
1252 	if (np == NULL)
1253 		return (ESRCH);
1254 	if (np != p) {
1255 		PROC_UNLOCK(np);
1256 		return (ESRCH);
1257 	}
1258 	PROC_UNLOCK(np);
1259 	return (0);
1260 }
1261 
1262 static int
1263 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1264 {
1265 	int *name = (int *)arg1;
1266 	u_int namelen = arg2;
1267 	struct proc *p;
1268 	int flags, doingzomb, oid_number;
1269 	int error = 0;
1270 
1271 	oid_number = oidp->oid_number;
1272 	if (oid_number != KERN_PROC_ALL &&
1273 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1274 		flags = KERN_PROC_NOTHREADS;
1275 	else {
1276 		flags = 0;
1277 		oid_number &= ~KERN_PROC_INC_THREAD;
1278 	}
1279 	if (oid_number == KERN_PROC_PID) {
1280 		if (namelen != 1)
1281 			return (EINVAL);
1282 		error = sysctl_wire_old_buffer(req, 0);
1283 		if (error)
1284 			return (error);
1285 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1286 		if (error != 0)
1287 			return (error);
1288 		error = sysctl_out_proc(p, req, flags);
1289 		return (error);
1290 	}
1291 
1292 	switch (oid_number) {
1293 	case KERN_PROC_ALL:
1294 		if (namelen != 0)
1295 			return (EINVAL);
1296 		break;
1297 	case KERN_PROC_PROC:
1298 		if (namelen != 0 && namelen != 1)
1299 			return (EINVAL);
1300 		break;
1301 	default:
1302 		if (namelen != 1)
1303 			return (EINVAL);
1304 		break;
1305 	}
1306 
1307 	if (!req->oldptr) {
1308 		/* overestimate by 5 procs */
1309 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1310 		if (error)
1311 			return (error);
1312 	}
1313 	error = sysctl_wire_old_buffer(req, 0);
1314 	if (error != 0)
1315 		return (error);
1316 	sx_slock(&allproc_lock);
1317 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1318 		if (!doingzomb)
1319 			p = LIST_FIRST(&allproc);
1320 		else
1321 			p = LIST_FIRST(&zombproc);
1322 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
1323 			/*
1324 			 * Skip embryonic processes.
1325 			 */
1326 			PROC_LOCK(p);
1327 			if (p->p_state == PRS_NEW) {
1328 				PROC_UNLOCK(p);
1329 				continue;
1330 			}
1331 			KASSERT(p->p_ucred != NULL,
1332 			    ("process credential is NULL for non-NEW proc"));
1333 			/*
1334 			 * Show a user only appropriate processes.
1335 			 */
1336 			if (p_cansee(curthread, p)) {
1337 				PROC_UNLOCK(p);
1338 				continue;
1339 			}
1340 			/*
1341 			 * TODO - make more efficient (see notes below).
1342 			 * do by session.
1343 			 */
1344 			switch (oid_number) {
1345 
1346 			case KERN_PROC_GID:
1347 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1348 					PROC_UNLOCK(p);
1349 					continue;
1350 				}
1351 				break;
1352 
1353 			case KERN_PROC_PGRP:
1354 				/* could do this by traversing pgrp */
1355 				if (p->p_pgrp == NULL ||
1356 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1357 					PROC_UNLOCK(p);
1358 					continue;
1359 				}
1360 				break;
1361 
1362 			case KERN_PROC_RGID:
1363 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1364 					PROC_UNLOCK(p);
1365 					continue;
1366 				}
1367 				break;
1368 
1369 			case KERN_PROC_SESSION:
1370 				if (p->p_session == NULL ||
1371 				    p->p_session->s_sid != (pid_t)name[0]) {
1372 					PROC_UNLOCK(p);
1373 					continue;
1374 				}
1375 				break;
1376 
1377 			case KERN_PROC_TTY:
1378 				if ((p->p_flag & P_CONTROLT) == 0 ||
1379 				    p->p_session == NULL) {
1380 					PROC_UNLOCK(p);
1381 					continue;
1382 				}
1383 				/* XXX proctree_lock */
1384 				SESS_LOCK(p->p_session);
1385 				if (p->p_session->s_ttyp == NULL ||
1386 				    tty_udev(p->p_session->s_ttyp) !=
1387 				    (dev_t)name[0]) {
1388 					SESS_UNLOCK(p->p_session);
1389 					PROC_UNLOCK(p);
1390 					continue;
1391 				}
1392 				SESS_UNLOCK(p->p_session);
1393 				break;
1394 
1395 			case KERN_PROC_UID:
1396 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1397 					PROC_UNLOCK(p);
1398 					continue;
1399 				}
1400 				break;
1401 
1402 			case KERN_PROC_RUID:
1403 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1404 					PROC_UNLOCK(p);
1405 					continue;
1406 				}
1407 				break;
1408 
1409 			case KERN_PROC_PROC:
1410 				break;
1411 
1412 			default:
1413 				break;
1414 
1415 			}
1416 
1417 			error = sysctl_out_proc(p, req, flags | doingzomb);
1418 			if (error) {
1419 				sx_sunlock(&allproc_lock);
1420 				return (error);
1421 			}
1422 		}
1423 	}
1424 	sx_sunlock(&allproc_lock);
1425 	return (0);
1426 }
1427 
1428 struct pargs *
1429 pargs_alloc(int len)
1430 {
1431 	struct pargs *pa;
1432 
1433 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1434 		M_WAITOK);
1435 	refcount_init(&pa->ar_ref, 1);
1436 	pa->ar_length = len;
1437 	return (pa);
1438 }
1439 
1440 static void
1441 pargs_free(struct pargs *pa)
1442 {
1443 
1444 	free(pa, M_PARGS);
1445 }
1446 
1447 void
1448 pargs_hold(struct pargs *pa)
1449 {
1450 
1451 	if (pa == NULL)
1452 		return;
1453 	refcount_acquire(&pa->ar_ref);
1454 }
1455 
1456 void
1457 pargs_drop(struct pargs *pa)
1458 {
1459 
1460 	if (pa == NULL)
1461 		return;
1462 	if (refcount_release(&pa->ar_ref))
1463 		pargs_free(pa);
1464 }
1465 
1466 static int
1467 proc_read_mem(struct thread *td, struct proc *p, vm_offset_t offset, void* buf,
1468     size_t len)
1469 {
1470 	struct iovec iov;
1471 	struct uio uio;
1472 
1473 	iov.iov_base = (caddr_t)buf;
1474 	iov.iov_len = len;
1475 	uio.uio_iov = &iov;
1476 	uio.uio_iovcnt = 1;
1477 	uio.uio_offset = offset;
1478 	uio.uio_resid = (ssize_t)len;
1479 	uio.uio_segflg = UIO_SYSSPACE;
1480 	uio.uio_rw = UIO_READ;
1481 	uio.uio_td = td;
1482 
1483 	return (proc_rwmem(p, &uio));
1484 }
1485 
1486 static int
1487 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1488     size_t len)
1489 {
1490 	size_t i;
1491 	int error;
1492 
1493 	error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, len);
1494 	/*
1495 	 * Reading the chunk may validly return EFAULT if the string is shorter
1496 	 * than the chunk and is aligned at the end of the page, assuming the
1497 	 * next page is not mapped.  So if EFAULT is returned do a fallback to
1498 	 * one byte read loop.
1499 	 */
1500 	if (error == EFAULT) {
1501 		for (i = 0; i < len; i++, buf++, sptr++) {
1502 			error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, 1);
1503 			if (error != 0)
1504 				return (error);
1505 			if (*buf == '\0')
1506 				break;
1507 		}
1508 		error = 0;
1509 	}
1510 	return (error);
1511 }
1512 
1513 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1514 
1515 enum proc_vector_type {
1516 	PROC_ARG,
1517 	PROC_ENV,
1518 	PROC_AUX,
1519 };
1520 
1521 #ifdef COMPAT_FREEBSD32
1522 static int
1523 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1524     size_t *vsizep, enum proc_vector_type type)
1525 {
1526 	struct freebsd32_ps_strings pss;
1527 	Elf32_Auxinfo aux;
1528 	vm_offset_t vptr, ptr;
1529 	uint32_t *proc_vector32;
1530 	char **proc_vector;
1531 	size_t vsize, size;
1532 	int i, error;
1533 
1534 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1535 	    &pss, sizeof(pss));
1536 	if (error != 0)
1537 		return (error);
1538 	switch (type) {
1539 	case PROC_ARG:
1540 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1541 		vsize = pss.ps_nargvstr;
1542 		if (vsize > ARG_MAX)
1543 			return (ENOEXEC);
1544 		size = vsize * sizeof(int32_t);
1545 		break;
1546 	case PROC_ENV:
1547 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1548 		vsize = pss.ps_nenvstr;
1549 		if (vsize > ARG_MAX)
1550 			return (ENOEXEC);
1551 		size = vsize * sizeof(int32_t);
1552 		break;
1553 	case PROC_AUX:
1554 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1555 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1556 		if (vptr % 4 != 0)
1557 			return (ENOEXEC);
1558 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1559 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1560 			if (error != 0)
1561 				return (error);
1562 			if (aux.a_type == AT_NULL)
1563 				break;
1564 			ptr += sizeof(aux);
1565 		}
1566 		if (aux.a_type != AT_NULL)
1567 			return (ENOEXEC);
1568 		vsize = i + 1;
1569 		size = vsize * sizeof(aux);
1570 		break;
1571 	default:
1572 		KASSERT(0, ("Wrong proc vector type: %d", type));
1573 		return (EINVAL);
1574 	}
1575 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1576 	error = proc_read_mem(td, p, vptr, proc_vector32, size);
1577 	if (error != 0)
1578 		goto done;
1579 	if (type == PROC_AUX) {
1580 		*proc_vectorp = (char **)proc_vector32;
1581 		*vsizep = vsize;
1582 		return (0);
1583 	}
1584 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1585 	for (i = 0; i < (int)vsize; i++)
1586 		proc_vector[i] = PTRIN(proc_vector32[i]);
1587 	*proc_vectorp = proc_vector;
1588 	*vsizep = vsize;
1589 done:
1590 	free(proc_vector32, M_TEMP);
1591 	return (error);
1592 }
1593 #endif
1594 
1595 static int
1596 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1597     size_t *vsizep, enum proc_vector_type type)
1598 {
1599 	struct ps_strings pss;
1600 	Elf_Auxinfo aux;
1601 	vm_offset_t vptr, ptr;
1602 	char **proc_vector;
1603 	size_t vsize, size;
1604 	int error, i;
1605 
1606 #ifdef COMPAT_FREEBSD32
1607 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1608 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1609 #endif
1610 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1611 	    &pss, sizeof(pss));
1612 	if (error != 0)
1613 		return (error);
1614 	switch (type) {
1615 	case PROC_ARG:
1616 		vptr = (vm_offset_t)pss.ps_argvstr;
1617 		vsize = pss.ps_nargvstr;
1618 		if (vsize > ARG_MAX)
1619 			return (ENOEXEC);
1620 		size = vsize * sizeof(char *);
1621 		break;
1622 	case PROC_ENV:
1623 		vptr = (vm_offset_t)pss.ps_envstr;
1624 		vsize = pss.ps_nenvstr;
1625 		if (vsize > ARG_MAX)
1626 			return (ENOEXEC);
1627 		size = vsize * sizeof(char *);
1628 		break;
1629 	case PROC_AUX:
1630 		/*
1631 		 * The aux array is just above env array on the stack. Check
1632 		 * that the address is naturally aligned.
1633 		 */
1634 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1635 		    * sizeof(char *);
1636 #if __ELF_WORD_SIZE == 64
1637 		if (vptr % sizeof(uint64_t) != 0)
1638 #else
1639 		if (vptr % sizeof(uint32_t) != 0)
1640 #endif
1641 			return (ENOEXEC);
1642 		/*
1643 		 * We count the array size reading the aux vectors from the
1644 		 * stack until AT_NULL vector is returned.  So (to keep the code
1645 		 * simple) we read the process stack twice: the first time here
1646 		 * to find the size and the second time when copying the vectors
1647 		 * to the allocated proc_vector.
1648 		 */
1649 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1650 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1651 			if (error != 0)
1652 				return (error);
1653 			if (aux.a_type == AT_NULL)
1654 				break;
1655 			ptr += sizeof(aux);
1656 		}
1657 		/*
1658 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1659 		 * not reached AT_NULL, it is most likely we are reading wrong
1660 		 * data: either the process doesn't have auxv array or data has
1661 		 * been modified. Return the error in this case.
1662 		 */
1663 		if (aux.a_type != AT_NULL)
1664 			return (ENOEXEC);
1665 		vsize = i + 1;
1666 		size = vsize * sizeof(aux);
1667 		break;
1668 	default:
1669 		KASSERT(0, ("Wrong proc vector type: %d", type));
1670 		return (EINVAL); /* In case we are built without INVARIANTS. */
1671 	}
1672 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1673 	if (proc_vector == NULL)
1674 		return (ENOMEM);
1675 	error = proc_read_mem(td, p, vptr, proc_vector, size);
1676 	if (error != 0) {
1677 		free(proc_vector, M_TEMP);
1678 		return (error);
1679 	}
1680 	*proc_vectorp = proc_vector;
1681 	*vsizep = vsize;
1682 
1683 	return (0);
1684 }
1685 
1686 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1687 
1688 static int
1689 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1690     enum proc_vector_type type)
1691 {
1692 	size_t done, len, nchr, vsize;
1693 	int error, i;
1694 	char **proc_vector, *sptr;
1695 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1696 
1697 	PROC_ASSERT_HELD(p);
1698 
1699 	/*
1700 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1701 	 */
1702 	nchr = 2 * (PATH_MAX + ARG_MAX);
1703 
1704 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1705 	if (error != 0)
1706 		return (error);
1707 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1708 		/*
1709 		 * The program may have scribbled into its argv array, e.g. to
1710 		 * remove some arguments.  If that has happened, break out
1711 		 * before trying to read from NULL.
1712 		 */
1713 		if (proc_vector[i] == NULL)
1714 			break;
1715 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1716 			error = proc_read_string(td, p, sptr, pss_string,
1717 			    sizeof(pss_string));
1718 			if (error != 0)
1719 				goto done;
1720 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1721 			if (done + len >= nchr)
1722 				len = nchr - done - 1;
1723 			sbuf_bcat(sb, pss_string, len);
1724 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1725 				break;
1726 			done += GET_PS_STRINGS_CHUNK_SZ;
1727 		}
1728 		sbuf_bcat(sb, "", 1);
1729 		done += len + 1;
1730 	}
1731 done:
1732 	free(proc_vector, M_TEMP);
1733 	return (error);
1734 }
1735 
1736 int
1737 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1738 {
1739 
1740 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1741 }
1742 
1743 int
1744 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1745 {
1746 
1747 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1748 }
1749 
1750 /*
1751  * This sysctl allows a process to retrieve the argument list or process
1752  * title for another process without groping around in the address space
1753  * of the other process.  It also allow a process to set its own "process
1754  * title to a string of its own choice.
1755  */
1756 static int
1757 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1758 {
1759 	int *name = (int *)arg1;
1760 	u_int namelen = arg2;
1761 	struct pargs *newpa, *pa;
1762 	struct proc *p;
1763 	struct sbuf sb;
1764 	int flags, error = 0, error2;
1765 
1766 	if (namelen != 1)
1767 		return (EINVAL);
1768 
1769 	flags = PGET_CANSEE;
1770 	if (req->newptr != NULL)
1771 		flags |= PGET_ISCURRENT;
1772 	error = pget((pid_t)name[0], flags, &p);
1773 	if (error)
1774 		return (error);
1775 
1776 	pa = p->p_args;
1777 	if (pa != NULL) {
1778 		pargs_hold(pa);
1779 		PROC_UNLOCK(p);
1780 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1781 		pargs_drop(pa);
1782 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1783 		_PHOLD(p);
1784 		PROC_UNLOCK(p);
1785 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1786 		error = proc_getargv(curthread, p, &sb);
1787 		error2 = sbuf_finish(&sb);
1788 		PRELE(p);
1789 		sbuf_delete(&sb);
1790 		if (error == 0 && error2 != 0)
1791 			error = error2;
1792 	} else {
1793 		PROC_UNLOCK(p);
1794 	}
1795 	if (error != 0 || req->newptr == NULL)
1796 		return (error);
1797 
1798 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1799 		return (ENOMEM);
1800 	newpa = pargs_alloc(req->newlen);
1801 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1802 	if (error != 0) {
1803 		pargs_free(newpa);
1804 		return (error);
1805 	}
1806 	PROC_LOCK(p);
1807 	pa = p->p_args;
1808 	p->p_args = newpa;
1809 	PROC_UNLOCK(p);
1810 	pargs_drop(pa);
1811 	return (0);
1812 }
1813 
1814 /*
1815  * This sysctl allows a process to retrieve environment of another process.
1816  */
1817 static int
1818 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
1819 {
1820 	int *name = (int *)arg1;
1821 	u_int namelen = arg2;
1822 	struct proc *p;
1823 	struct sbuf sb;
1824 	int error, error2;
1825 
1826 	if (namelen != 1)
1827 		return (EINVAL);
1828 
1829 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1830 	if (error != 0)
1831 		return (error);
1832 	if ((p->p_flag & P_SYSTEM) != 0) {
1833 		PRELE(p);
1834 		return (0);
1835 	}
1836 
1837 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1838 	error = proc_getenvv(curthread, p, &sb);
1839 	error2 = sbuf_finish(&sb);
1840 	PRELE(p);
1841 	sbuf_delete(&sb);
1842 	return (error != 0 ? error : error2);
1843 }
1844 
1845 /*
1846  * This sysctl allows a process to retrieve ELF auxiliary vector of
1847  * another process.
1848  */
1849 static int
1850 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
1851 {
1852 	int *name = (int *)arg1;
1853 	u_int namelen = arg2;
1854 	struct proc *p;
1855 	size_t vsize, size;
1856 	char **auxv;
1857 	int error;
1858 
1859 	if (namelen != 1)
1860 		return (EINVAL);
1861 
1862 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1863 	if (error != 0)
1864 		return (error);
1865 	if ((p->p_flag & P_SYSTEM) != 0) {
1866 		PRELE(p);
1867 		return (0);
1868 	}
1869 	error = get_proc_vector(curthread, p, &auxv, &vsize, PROC_AUX);
1870 	if (error == 0) {
1871 #ifdef COMPAT_FREEBSD32
1872 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1873 			size = vsize * sizeof(Elf32_Auxinfo);
1874 		else
1875 #endif
1876 		size = vsize * sizeof(Elf_Auxinfo);
1877 		PRELE(p);
1878 		error = SYSCTL_OUT(req, auxv, size);
1879 		free(auxv, M_TEMP);
1880 	} else {
1881 		PRELE(p);
1882 	}
1883 	return (error);
1884 }
1885 
1886 /*
1887  * This sysctl allows a process to retrieve the path of the executable for
1888  * itself or another process.
1889  */
1890 static int
1891 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1892 {
1893 	pid_t *pidp = (pid_t *)arg1;
1894 	unsigned int arglen = arg2;
1895 	struct proc *p;
1896 	struct vnode *vp;
1897 	char *retbuf, *freebuf;
1898 	int error;
1899 
1900 	if (arglen != 1)
1901 		return (EINVAL);
1902 	if (*pidp == -1) {	/* -1 means this process */
1903 		p = req->td->td_proc;
1904 	} else {
1905 		error = pget(*pidp, PGET_CANSEE, &p);
1906 		if (error != 0)
1907 			return (error);
1908 	}
1909 
1910 	vp = p->p_textvp;
1911 	if (vp == NULL) {
1912 		if (*pidp != -1)
1913 			PROC_UNLOCK(p);
1914 		return (0);
1915 	}
1916 	vref(vp);
1917 	if (*pidp != -1)
1918 		PROC_UNLOCK(p);
1919 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1920 	vrele(vp);
1921 	if (error)
1922 		return (error);
1923 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1924 	free(freebuf, M_TEMP);
1925 	return (error);
1926 }
1927 
1928 static int
1929 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1930 {
1931 	struct proc *p;
1932 	char *sv_name;
1933 	int *name;
1934 	int namelen;
1935 	int error;
1936 
1937 	namelen = arg2;
1938 	if (namelen != 1)
1939 		return (EINVAL);
1940 
1941 	name = (int *)arg1;
1942 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
1943 	if (error != 0)
1944 		return (error);
1945 	sv_name = p->p_sysent->sv_name;
1946 	PROC_UNLOCK(p);
1947 	return (sysctl_handle_string(oidp, sv_name, 0, req));
1948 }
1949 
1950 #ifdef KINFO_OVMENTRY_SIZE
1951 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
1952 #endif
1953 
1954 #ifdef COMPAT_FREEBSD7
1955 static int
1956 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
1957 {
1958 	vm_map_entry_t entry, tmp_entry;
1959 	unsigned int last_timestamp;
1960 	char *fullpath, *freepath;
1961 	struct kinfo_ovmentry *kve;
1962 	struct vattr va;
1963 	struct ucred *cred;
1964 	int error, *name;
1965 	struct vnode *vp;
1966 	struct proc *p;
1967 	vm_map_t map;
1968 	struct vmspace *vm;
1969 
1970 	name = (int *)arg1;
1971 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1972 	if (error != 0)
1973 		return (error);
1974 	vm = vmspace_acquire_ref(p);
1975 	if (vm == NULL) {
1976 		PRELE(p);
1977 		return (ESRCH);
1978 	}
1979 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1980 
1981 	map = &vm->vm_map;
1982 	vm_map_lock_read(map);
1983 	for (entry = map->header.next; entry != &map->header;
1984 	    entry = entry->next) {
1985 		vm_object_t obj, tobj, lobj;
1986 		vm_offset_t addr;
1987 
1988 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
1989 			continue;
1990 
1991 		bzero(kve, sizeof(*kve));
1992 		kve->kve_structsize = sizeof(*kve);
1993 
1994 		kve->kve_private_resident = 0;
1995 		obj = entry->object.vm_object;
1996 		if (obj != NULL) {
1997 			VM_OBJECT_LOCK(obj);
1998 			if (obj->shadow_count == 1)
1999 				kve->kve_private_resident =
2000 				    obj->resident_page_count;
2001 		}
2002 		kve->kve_resident = 0;
2003 		addr = entry->start;
2004 		while (addr < entry->end) {
2005 			if (pmap_extract(map->pmap, addr))
2006 				kve->kve_resident++;
2007 			addr += PAGE_SIZE;
2008 		}
2009 
2010 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2011 			if (tobj != obj)
2012 				VM_OBJECT_LOCK(tobj);
2013 			if (lobj != obj)
2014 				VM_OBJECT_UNLOCK(lobj);
2015 			lobj = tobj;
2016 		}
2017 
2018 		kve->kve_start = (void*)entry->start;
2019 		kve->kve_end = (void*)entry->end;
2020 		kve->kve_offset = (off_t)entry->offset;
2021 
2022 		if (entry->protection & VM_PROT_READ)
2023 			kve->kve_protection |= KVME_PROT_READ;
2024 		if (entry->protection & VM_PROT_WRITE)
2025 			kve->kve_protection |= KVME_PROT_WRITE;
2026 		if (entry->protection & VM_PROT_EXECUTE)
2027 			kve->kve_protection |= KVME_PROT_EXEC;
2028 
2029 		if (entry->eflags & MAP_ENTRY_COW)
2030 			kve->kve_flags |= KVME_FLAG_COW;
2031 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2032 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2033 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2034 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2035 
2036 		last_timestamp = map->timestamp;
2037 		vm_map_unlock_read(map);
2038 
2039 		kve->kve_fileid = 0;
2040 		kve->kve_fsid = 0;
2041 		freepath = NULL;
2042 		fullpath = "";
2043 		if (lobj) {
2044 			vp = NULL;
2045 			switch (lobj->type) {
2046 			case OBJT_DEFAULT:
2047 				kve->kve_type = KVME_TYPE_DEFAULT;
2048 				break;
2049 			case OBJT_VNODE:
2050 				kve->kve_type = KVME_TYPE_VNODE;
2051 				vp = lobj->handle;
2052 				vref(vp);
2053 				break;
2054 			case OBJT_SWAP:
2055 				kve->kve_type = KVME_TYPE_SWAP;
2056 				break;
2057 			case OBJT_DEVICE:
2058 				kve->kve_type = KVME_TYPE_DEVICE;
2059 				break;
2060 			case OBJT_PHYS:
2061 				kve->kve_type = KVME_TYPE_PHYS;
2062 				break;
2063 			case OBJT_DEAD:
2064 				kve->kve_type = KVME_TYPE_DEAD;
2065 				break;
2066 			case OBJT_SG:
2067 				kve->kve_type = KVME_TYPE_SG;
2068 				break;
2069 			default:
2070 				kve->kve_type = KVME_TYPE_UNKNOWN;
2071 				break;
2072 			}
2073 			if (lobj != obj)
2074 				VM_OBJECT_UNLOCK(lobj);
2075 
2076 			kve->kve_ref_count = obj->ref_count;
2077 			kve->kve_shadow_count = obj->shadow_count;
2078 			VM_OBJECT_UNLOCK(obj);
2079 			if (vp != NULL) {
2080 				vn_fullpath(curthread, vp, &fullpath,
2081 				    &freepath);
2082 				cred = curthread->td_ucred;
2083 				vn_lock(vp, LK_SHARED | LK_RETRY);
2084 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2085 					kve->kve_fileid = va.va_fileid;
2086 					kve->kve_fsid = va.va_fsid;
2087 				}
2088 				vput(vp);
2089 			}
2090 		} else {
2091 			kve->kve_type = KVME_TYPE_NONE;
2092 			kve->kve_ref_count = 0;
2093 			kve->kve_shadow_count = 0;
2094 		}
2095 
2096 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2097 		if (freepath != NULL)
2098 			free(freepath, M_TEMP);
2099 
2100 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2101 		vm_map_lock_read(map);
2102 		if (error)
2103 			break;
2104 		if (last_timestamp != map->timestamp) {
2105 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2106 			entry = tmp_entry;
2107 		}
2108 	}
2109 	vm_map_unlock_read(map);
2110 	vmspace_free(vm);
2111 	PRELE(p);
2112 	free(kve, M_TEMP);
2113 	return (error);
2114 }
2115 #endif	/* COMPAT_FREEBSD7 */
2116 
2117 #ifdef KINFO_VMENTRY_SIZE
2118 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2119 #endif
2120 
2121 static int
2122 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2123 {
2124 	vm_map_entry_t entry, tmp_entry;
2125 	unsigned int last_timestamp;
2126 	char *fullpath, *freepath;
2127 	struct kinfo_vmentry *kve;
2128 	struct vattr va;
2129 	struct ucred *cred;
2130 	int error, *name;
2131 	struct vnode *vp;
2132 	struct proc *p;
2133 	struct vmspace *vm;
2134 	vm_map_t map;
2135 
2136 	name = (int *)arg1;
2137 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2138 	if (error != 0)
2139 		return (error);
2140 	vm = vmspace_acquire_ref(p);
2141 	if (vm == NULL) {
2142 		PRELE(p);
2143 		return (ESRCH);
2144 	}
2145 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2146 
2147 	map = &vm->vm_map;
2148 	vm_map_lock_read(map);
2149 	for (entry = map->header.next; entry != &map->header;
2150 	    entry = entry->next) {
2151 		vm_object_t obj, tobj, lobj;
2152 		vm_offset_t addr;
2153 		vm_paddr_t locked_pa;
2154 		int mincoreinfo;
2155 
2156 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2157 			continue;
2158 
2159 		bzero(kve, sizeof(*kve));
2160 
2161 		kve->kve_private_resident = 0;
2162 		obj = entry->object.vm_object;
2163 		if (obj != NULL) {
2164 			VM_OBJECT_LOCK(obj);
2165 			if (obj->shadow_count == 1)
2166 				kve->kve_private_resident =
2167 				    obj->resident_page_count;
2168 		}
2169 		kve->kve_resident = 0;
2170 		addr = entry->start;
2171 		while (addr < entry->end) {
2172 			locked_pa = 0;
2173 			mincoreinfo = pmap_mincore(map->pmap, addr, &locked_pa);
2174 			if (locked_pa != 0)
2175 				vm_page_unlock(PHYS_TO_VM_PAGE(locked_pa));
2176 			if (mincoreinfo & MINCORE_INCORE)
2177 				kve->kve_resident++;
2178 			if (mincoreinfo & MINCORE_SUPER)
2179 				kve->kve_flags |= KVME_FLAG_SUPER;
2180 			addr += PAGE_SIZE;
2181 		}
2182 
2183 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2184 			if (tobj != obj)
2185 				VM_OBJECT_LOCK(tobj);
2186 			if (lobj != obj)
2187 				VM_OBJECT_UNLOCK(lobj);
2188 			lobj = tobj;
2189 		}
2190 
2191 		kve->kve_start = entry->start;
2192 		kve->kve_end = entry->end;
2193 		kve->kve_offset = entry->offset;
2194 
2195 		if (entry->protection & VM_PROT_READ)
2196 			kve->kve_protection |= KVME_PROT_READ;
2197 		if (entry->protection & VM_PROT_WRITE)
2198 			kve->kve_protection |= KVME_PROT_WRITE;
2199 		if (entry->protection & VM_PROT_EXECUTE)
2200 			kve->kve_protection |= KVME_PROT_EXEC;
2201 
2202 		if (entry->eflags & MAP_ENTRY_COW)
2203 			kve->kve_flags |= KVME_FLAG_COW;
2204 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2205 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2206 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2207 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2208 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2209 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2210 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2211 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2212 
2213 		last_timestamp = map->timestamp;
2214 		vm_map_unlock_read(map);
2215 
2216 		freepath = NULL;
2217 		fullpath = "";
2218 		if (lobj) {
2219 			vp = NULL;
2220 			switch (lobj->type) {
2221 			case OBJT_DEFAULT:
2222 				kve->kve_type = KVME_TYPE_DEFAULT;
2223 				break;
2224 			case OBJT_VNODE:
2225 				kve->kve_type = KVME_TYPE_VNODE;
2226 				vp = lobj->handle;
2227 				vref(vp);
2228 				break;
2229 			case OBJT_SWAP:
2230 				kve->kve_type = KVME_TYPE_SWAP;
2231 				break;
2232 			case OBJT_DEVICE:
2233 				kve->kve_type = KVME_TYPE_DEVICE;
2234 				break;
2235 			case OBJT_PHYS:
2236 				kve->kve_type = KVME_TYPE_PHYS;
2237 				break;
2238 			case OBJT_DEAD:
2239 				kve->kve_type = KVME_TYPE_DEAD;
2240 				break;
2241 			case OBJT_SG:
2242 				kve->kve_type = KVME_TYPE_SG;
2243 				break;
2244 			default:
2245 				kve->kve_type = KVME_TYPE_UNKNOWN;
2246 				break;
2247 			}
2248 			if (lobj != obj)
2249 				VM_OBJECT_UNLOCK(lobj);
2250 
2251 			kve->kve_ref_count = obj->ref_count;
2252 			kve->kve_shadow_count = obj->shadow_count;
2253 			VM_OBJECT_UNLOCK(obj);
2254 			if (vp != NULL) {
2255 				vn_fullpath(curthread, vp, &fullpath,
2256 				    &freepath);
2257 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2258 				cred = curthread->td_ucred;
2259 				vn_lock(vp, LK_SHARED | LK_RETRY);
2260 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2261 					kve->kve_vn_fileid = va.va_fileid;
2262 					kve->kve_vn_fsid = va.va_fsid;
2263 					kve->kve_vn_mode =
2264 					    MAKEIMODE(va.va_type, va.va_mode);
2265 					kve->kve_vn_size = va.va_size;
2266 					kve->kve_vn_rdev = va.va_rdev;
2267 					kve->kve_status = KF_ATTR_VALID;
2268 				}
2269 				vput(vp);
2270 			}
2271 		} else {
2272 			kve->kve_type = KVME_TYPE_NONE;
2273 			kve->kve_ref_count = 0;
2274 			kve->kve_shadow_count = 0;
2275 		}
2276 
2277 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2278 		if (freepath != NULL)
2279 			free(freepath, M_TEMP);
2280 
2281 		/* Pack record size down */
2282 		kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) +
2283 		    strlen(kve->kve_path) + 1;
2284 		kve->kve_structsize = roundup(kve->kve_structsize,
2285 		    sizeof(uint64_t));
2286 		error = SYSCTL_OUT(req, kve, kve->kve_structsize);
2287 		vm_map_lock_read(map);
2288 		if (error)
2289 			break;
2290 		if (last_timestamp != map->timestamp) {
2291 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2292 			entry = tmp_entry;
2293 		}
2294 	}
2295 	vm_map_unlock_read(map);
2296 	vmspace_free(vm);
2297 	PRELE(p);
2298 	free(kve, M_TEMP);
2299 	return (error);
2300 }
2301 
2302 #if defined(STACK) || defined(DDB)
2303 static int
2304 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2305 {
2306 	struct kinfo_kstack *kkstp;
2307 	int error, i, *name, numthreads;
2308 	lwpid_t *lwpidarray;
2309 	struct thread *td;
2310 	struct stack *st;
2311 	struct sbuf sb;
2312 	struct proc *p;
2313 
2314 	name = (int *)arg1;
2315 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2316 	if (error != 0)
2317 		return (error);
2318 
2319 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2320 	st = stack_create();
2321 
2322 	lwpidarray = NULL;
2323 	numthreads = 0;
2324 	PROC_LOCK(p);
2325 repeat:
2326 	if (numthreads < p->p_numthreads) {
2327 		if (lwpidarray != NULL) {
2328 			free(lwpidarray, M_TEMP);
2329 			lwpidarray = NULL;
2330 		}
2331 		numthreads = p->p_numthreads;
2332 		PROC_UNLOCK(p);
2333 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2334 		    M_WAITOK | M_ZERO);
2335 		PROC_LOCK(p);
2336 		goto repeat;
2337 	}
2338 	i = 0;
2339 
2340 	/*
2341 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2342 	 * of changes could have taken place.  Should we check to see if the
2343 	 * vmspace has been replaced, or the like, in order to prevent
2344 	 * giving a snapshot that spans, say, execve(2), with some threads
2345 	 * before and some after?  Among other things, the credentials could
2346 	 * have changed, in which case the right to extract debug info might
2347 	 * no longer be assured.
2348 	 */
2349 	FOREACH_THREAD_IN_PROC(p, td) {
2350 		KASSERT(i < numthreads,
2351 		    ("sysctl_kern_proc_kstack: numthreads"));
2352 		lwpidarray[i] = td->td_tid;
2353 		i++;
2354 	}
2355 	numthreads = i;
2356 	for (i = 0; i < numthreads; i++) {
2357 		td = thread_find(p, lwpidarray[i]);
2358 		if (td == NULL) {
2359 			continue;
2360 		}
2361 		bzero(kkstp, sizeof(*kkstp));
2362 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2363 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2364 		thread_lock(td);
2365 		kkstp->kkst_tid = td->td_tid;
2366 		if (TD_IS_SWAPPED(td))
2367 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2368 		else if (TD_IS_RUNNING(td))
2369 			kkstp->kkst_state = KKST_STATE_RUNNING;
2370 		else {
2371 			kkstp->kkst_state = KKST_STATE_STACKOK;
2372 			stack_save_td(st, td);
2373 		}
2374 		thread_unlock(td);
2375 		PROC_UNLOCK(p);
2376 		stack_sbuf_print(&sb, st);
2377 		sbuf_finish(&sb);
2378 		sbuf_delete(&sb);
2379 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2380 		PROC_LOCK(p);
2381 		if (error)
2382 			break;
2383 	}
2384 	_PRELE(p);
2385 	PROC_UNLOCK(p);
2386 	if (lwpidarray != NULL)
2387 		free(lwpidarray, M_TEMP);
2388 	stack_destroy(st);
2389 	free(kkstp, M_TEMP);
2390 	return (error);
2391 }
2392 #endif
2393 
2394 /*
2395  * This sysctl allows a process to retrieve the full list of groups from
2396  * itself or another process.
2397  */
2398 static int
2399 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2400 {
2401 	pid_t *pidp = (pid_t *)arg1;
2402 	unsigned int arglen = arg2;
2403 	struct proc *p;
2404 	struct ucred *cred;
2405 	int error;
2406 
2407 	if (arglen != 1)
2408 		return (EINVAL);
2409 	if (*pidp == -1) {	/* -1 means this process */
2410 		p = req->td->td_proc;
2411 	} else {
2412 		error = pget(*pidp, PGET_CANSEE, &p);
2413 		if (error != 0)
2414 			return (error);
2415 	}
2416 
2417 	cred = crhold(p->p_ucred);
2418 	if (*pidp != -1)
2419 		PROC_UNLOCK(p);
2420 
2421 	error = SYSCTL_OUT(req, cred->cr_groups,
2422 	    cred->cr_ngroups * sizeof(gid_t));
2423 	crfree(cred);
2424 	return (error);
2425 }
2426 
2427 /*
2428  * This sysctl allows a process to retrieve or/and set the resource limit for
2429  * another process.
2430  */
2431 static int
2432 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2433 {
2434 	int *name = (int *)arg1;
2435 	u_int namelen = arg2;
2436 	struct rlimit rlim;
2437 	struct proc *p;
2438 	u_int which;
2439 	int flags, error;
2440 
2441 	if (namelen != 2)
2442 		return (EINVAL);
2443 
2444 	which = (u_int)name[1];
2445 	if (which >= RLIM_NLIMITS)
2446 		return (EINVAL);
2447 
2448 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2449 		return (EINVAL);
2450 
2451 	flags = PGET_HOLD | PGET_NOTWEXIT;
2452 	if (req->newptr != NULL)
2453 		flags |= PGET_CANDEBUG;
2454 	else
2455 		flags |= PGET_CANSEE;
2456 	error = pget((pid_t)name[0], flags, &p);
2457 	if (error != 0)
2458 		return (error);
2459 
2460 	/*
2461 	 * Retrieve limit.
2462 	 */
2463 	if (req->oldptr != NULL) {
2464 		PROC_LOCK(p);
2465 		lim_rlimit(p, which, &rlim);
2466 		PROC_UNLOCK(p);
2467 	}
2468 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2469 	if (error != 0)
2470 		goto errout;
2471 
2472 	/*
2473 	 * Set limit.
2474 	 */
2475 	if (req->newptr != NULL) {
2476 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2477 		if (error == 0)
2478 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2479 	}
2480 
2481 errout:
2482 	PRELE(p);
2483 	return (error);
2484 }
2485 
2486 /*
2487  * This sysctl allows a process to retrieve ps_strings structure location of
2488  * another process.
2489  */
2490 static int
2491 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2492 {
2493 	int *name = (int *)arg1;
2494 	u_int namelen = arg2;
2495 	struct proc *p;
2496 	vm_offset_t ps_strings;
2497 	int error;
2498 #ifdef COMPAT_FREEBSD32
2499 	uint32_t ps_strings32;
2500 #endif
2501 
2502 	if (namelen != 1)
2503 		return (EINVAL);
2504 
2505 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2506 	if (error != 0)
2507 		return (error);
2508 #ifdef COMPAT_FREEBSD32
2509 	if ((req->flags & SCTL_MASK32) != 0) {
2510 		/*
2511 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2512 		 * process.
2513 		 */
2514 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2515 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2516 		PROC_UNLOCK(p);
2517 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2518 		return (error);
2519 	}
2520 #endif
2521 	ps_strings = p->p_sysent->sv_psstrings;
2522 	PROC_UNLOCK(p);
2523 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2524 	return (error);
2525 }
2526 
2527 /*
2528  * This sysctl allows a process to retrieve umask of another process.
2529  */
2530 static int
2531 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2532 {
2533 	int *name = (int *)arg1;
2534 	u_int namelen = arg2;
2535 	struct proc *p;
2536 	int error;
2537 	u_short fd_cmask;
2538 
2539 	if (namelen != 1)
2540 		return (EINVAL);
2541 
2542 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2543 	if (error != 0)
2544 		return (error);
2545 
2546 	FILEDESC_SLOCK(p->p_fd);
2547 	fd_cmask = p->p_fd->fd_cmask;
2548 	FILEDESC_SUNLOCK(p->p_fd);
2549 	PRELE(p);
2550 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2551 	return (error);
2552 }
2553 
2554 /*
2555  * This sysctl allows a process to set and retrieve binary osreldate of
2556  * another process.
2557  */
2558 static int
2559 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2560 {
2561 	int *name = (int *)arg1;
2562 	u_int namelen = arg2;
2563 	struct proc *p;
2564 	int flags, error, osrel;
2565 
2566 	if (namelen != 1)
2567 		return (EINVAL);
2568 
2569 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2570 		return (EINVAL);
2571 
2572 	flags = PGET_HOLD | PGET_NOTWEXIT;
2573 	if (req->newptr != NULL)
2574 		flags |= PGET_CANDEBUG;
2575 	else
2576 		flags |= PGET_CANSEE;
2577 	error = pget((pid_t)name[0], flags, &p);
2578 	if (error != 0)
2579 		return (error);
2580 
2581 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2582 	if (error != 0)
2583 		goto errout;
2584 
2585 	if (req->newptr != NULL) {
2586 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2587 		if (error != 0)
2588 			goto errout;
2589 		if (osrel < 0) {
2590 			error = EINVAL;
2591 			goto errout;
2592 		}
2593 		p->p_osrel = osrel;
2594 	}
2595 errout:
2596 	PRELE(p);
2597 	return (error);
2598 }
2599 
2600 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2601 
2602 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2603 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2604 	"Return entire process table");
2605 
2606 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2607 	sysctl_kern_proc, "Process table");
2608 
2609 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2610 	sysctl_kern_proc, "Process table");
2611 
2612 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2613 	sysctl_kern_proc, "Process table");
2614 
2615 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2616 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2617 
2618 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2619 	sysctl_kern_proc, "Process table");
2620 
2621 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2622 	sysctl_kern_proc, "Process table");
2623 
2624 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2625 	sysctl_kern_proc, "Process table");
2626 
2627 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2628 	sysctl_kern_proc, "Process table");
2629 
2630 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2631 	sysctl_kern_proc, "Return process table, no threads");
2632 
2633 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2634 	CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2635 	sysctl_kern_proc_args, "Process argument list");
2636 
2637 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
2638 	sysctl_kern_proc_env, "Process environment");
2639 
2640 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
2641 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
2642 
2643 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2644 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2645 
2646 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2647 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2648 	"Process syscall vector name (ABI type)");
2649 
2650 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2651 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2652 
2653 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2654 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2655 
2656 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2657 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2658 
2659 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2660 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2661 
2662 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2663 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2664 
2665 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2666 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2667 
2668 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2669 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2670 
2671 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2672 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2673 
2674 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2675 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2676 	"Return process table, no threads");
2677 
2678 #ifdef COMPAT_FREEBSD7
2679 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2680 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2681 #endif
2682 
2683 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2684 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2685 
2686 #if defined(STACK) || defined(DDB)
2687 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2688 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2689 #endif
2690 
2691 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2692 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
2693 
2694 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
2695 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
2696 	"Process resource limits");
2697 
2698 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
2699 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
2700 	"Process ps_strings location");
2701 
2702 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
2703 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
2704 
2705 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
2706 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
2707 	"Process binary osreldate");
2708