xref: /freebsd/sys/kern/kern_proc.c (revision d184218c18d067f8fd47203f54ab02a7b2ed9b11)
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_DEFINE(proc, kernel, ctor, entry, entry);
96 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 0, "struct proc *");
97 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 1, "int");
98 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 2, "void *");
99 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 3, "int");
100 SDT_PROBE_DEFINE(proc, kernel, ctor, return, return);
101 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 0, "struct proc *");
102 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 1, "int");
103 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 2, "void *");
104 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 3, "int");
105 SDT_PROBE_DEFINE(proc, kernel, dtor, entry, entry);
106 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 0, "struct proc *");
107 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 1, "int");
108 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 2, "void *");
109 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 3, "struct thread *");
110 SDT_PROBE_DEFINE(proc, kernel, dtor, return, return);
111 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 0, "struct proc *");
112 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 1, "int");
113 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 2, "void *");
114 SDT_PROBE_DEFINE(proc, kernel, init, entry, entry);
115 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 0, "struct proc *");
116 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 1, "int");
117 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 2, "int");
118 SDT_PROBE_DEFINE(proc, kernel, init, return, return);
119 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 0, "struct proc *");
120 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 1, "int");
121 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 2, "int");
122 
123 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
124 MALLOC_DEFINE(M_SESSION, "session", "session header");
125 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
126 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
127 
128 static void doenterpgrp(struct proc *, struct pgrp *);
129 static void orphanpg(struct pgrp *pg);
130 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
131 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
132 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
133     int preferthread);
134 static void pgadjustjobc(struct pgrp *pgrp, int entering);
135 static void pgdelete(struct pgrp *);
136 static int proc_ctor(void *mem, int size, void *arg, int flags);
137 static void proc_dtor(void *mem, int size, void *arg);
138 static int proc_init(void *mem, int size, int flags);
139 static void proc_fini(void *mem, int size);
140 static void pargs_free(struct pargs *pa);
141 static struct proc *zpfind_locked(pid_t pid);
142 
143 /*
144  * Other process lists
145  */
146 struct pidhashhead *pidhashtbl;
147 u_long pidhash;
148 struct pgrphashhead *pgrphashtbl;
149 u_long pgrphash;
150 struct proclist allproc;
151 struct proclist zombproc;
152 struct sx allproc_lock;
153 struct sx proctree_lock;
154 struct mtx ppeers_lock;
155 uma_zone_t proc_zone;
156 
157 int kstack_pages = KSTACK_PAGES;
158 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
159     "Kernel stack size in pages");
160 
161 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
162 #ifdef COMPAT_FREEBSD32
163 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
164 #endif
165 
166 /*
167  * Initialize global process hashing structures.
168  */
169 void
170 procinit()
171 {
172 
173 	sx_init(&allproc_lock, "allproc");
174 	sx_init(&proctree_lock, "proctree");
175 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
176 	LIST_INIT(&allproc);
177 	LIST_INIT(&zombproc);
178 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
179 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
180 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
181 	    proc_ctor, proc_dtor, proc_init, proc_fini,
182 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
183 	uihashinit();
184 }
185 
186 /*
187  * Prepare a proc for use.
188  */
189 static int
190 proc_ctor(void *mem, int size, void *arg, int flags)
191 {
192 	struct proc *p;
193 
194 	p = (struct proc *)mem;
195 	SDT_PROBE(proc, kernel, ctor , entry, p, size, arg, flags, 0);
196 	EVENTHANDLER_INVOKE(process_ctor, p);
197 	SDT_PROBE(proc, kernel, ctor , return, p, size, arg, flags, 0);
198 	return (0);
199 }
200 
201 /*
202  * Reclaim a proc after use.
203  */
204 static void
205 proc_dtor(void *mem, int size, void *arg)
206 {
207 	struct proc *p;
208 	struct thread *td;
209 
210 	/* INVARIANTS checks go here */
211 	p = (struct proc *)mem;
212 	td = FIRST_THREAD_IN_PROC(p);
213 	SDT_PROBE(proc, kernel, dtor, entry, p, size, arg, td, 0);
214 	if (td != NULL) {
215 #ifdef INVARIANTS
216 		KASSERT((p->p_numthreads == 1),
217 		    ("bad number of threads in exiting process"));
218 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
219 #endif
220 		/* Free all OSD associated to this thread. */
221 		osd_thread_exit(td);
222 	}
223 	EVENTHANDLER_INVOKE(process_dtor, p);
224 	if (p->p_ksi != NULL)
225 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
226 	SDT_PROBE(proc, kernel, dtor, return, p, size, arg, 0, 0);
227 }
228 
229 /*
230  * Initialize type-stable parts of a proc (when newly created).
231  */
232 static int
233 proc_init(void *mem, int size, int flags)
234 {
235 	struct proc *p;
236 
237 	p = (struct proc *)mem;
238 	SDT_PROBE(proc, kernel, init, entry, p, size, flags, 0, 0);
239 	p->p_sched = (struct p_sched *)&p[1];
240 	bzero(&p->p_mtx, sizeof(struct mtx));
241 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK);
242 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_RECURSE);
243 	cv_init(&p->p_pwait, "ppwait");
244 	cv_init(&p->p_dbgwait, "dbgwait");
245 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
246 	EVENTHANDLER_INVOKE(process_init, p);
247 	p->p_stats = pstats_alloc();
248 	SDT_PROBE(proc, kernel, init, return, p, size, flags, 0, 0);
249 	return (0);
250 }
251 
252 /*
253  * UMA should ensure that this function is never called.
254  * Freeing a proc structure would violate type stability.
255  */
256 static void
257 proc_fini(void *mem, int size)
258 {
259 #ifdef notnow
260 	struct proc *p;
261 
262 	p = (struct proc *)mem;
263 	EVENTHANDLER_INVOKE(process_fini, p);
264 	pstats_free(p->p_stats);
265 	thread_free(FIRST_THREAD_IN_PROC(p));
266 	mtx_destroy(&p->p_mtx);
267 	if (p->p_ksi != NULL)
268 		ksiginfo_free(p->p_ksi);
269 #else
270 	panic("proc reclaimed");
271 #endif
272 }
273 
274 /*
275  * Is p an inferior of the current process?
276  */
277 int
278 inferior(p)
279 	register struct proc *p;
280 {
281 
282 	sx_assert(&proctree_lock, SX_LOCKED);
283 	for (; p != curproc; p = p->p_pptr)
284 		if (p->p_pid == 0)
285 			return (0);
286 	return (1);
287 }
288 
289 struct proc *
290 pfind_locked(pid_t pid)
291 {
292 	struct proc *p;
293 
294 	sx_assert(&allproc_lock, SX_LOCKED);
295 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
296 		if (p->p_pid == pid) {
297 			PROC_LOCK(p);
298 			if (p->p_state == PRS_NEW) {
299 				PROC_UNLOCK(p);
300 				p = NULL;
301 			}
302 			break;
303 		}
304 	}
305 	return (p);
306 }
307 
308 /*
309  * Locate a process by number; return only "live" processes -- i.e., neither
310  * zombies nor newly born but incompletely initialized processes.  By not
311  * returning processes in the PRS_NEW state, we allow callers to avoid
312  * testing for that condition to avoid dereferencing p_ucred, et al.
313  */
314 struct proc *
315 pfind(pid_t pid)
316 {
317 	struct proc *p;
318 
319 	sx_slock(&allproc_lock);
320 	p = pfind_locked(pid);
321 	sx_sunlock(&allproc_lock);
322 	return (p);
323 }
324 
325 static struct proc *
326 pfind_tid_locked(pid_t tid)
327 {
328 	struct proc *p;
329 	struct thread *td;
330 
331 	sx_assert(&allproc_lock, SX_LOCKED);
332 	FOREACH_PROC_IN_SYSTEM(p) {
333 		PROC_LOCK(p);
334 		if (p->p_state == PRS_NEW) {
335 			PROC_UNLOCK(p);
336 			continue;
337 		}
338 		FOREACH_THREAD_IN_PROC(p, td) {
339 			if (td->td_tid == tid)
340 				goto found;
341 		}
342 		PROC_UNLOCK(p);
343 	}
344 found:
345 	return (p);
346 }
347 
348 /*
349  * Locate a process group by number.
350  * The caller must hold proctree_lock.
351  */
352 struct pgrp *
353 pgfind(pgid)
354 	register pid_t pgid;
355 {
356 	register struct pgrp *pgrp;
357 
358 	sx_assert(&proctree_lock, SX_LOCKED);
359 
360 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
361 		if (pgrp->pg_id == pgid) {
362 			PGRP_LOCK(pgrp);
363 			return (pgrp);
364 		}
365 	}
366 	return (NULL);
367 }
368 
369 /*
370  * Locate process and do additional manipulations, depending on flags.
371  */
372 int
373 pget(pid_t pid, int flags, struct proc **pp)
374 {
375 	struct proc *p;
376 	int error;
377 
378 	sx_slock(&allproc_lock);
379 	if (pid <= PID_MAX) {
380 		p = pfind_locked(pid);
381 		if (p == NULL && (flags & PGET_NOTWEXIT) == 0)
382 			p = zpfind_locked(pid);
383 	} else if ((flags & PGET_NOTID) == 0) {
384 		p = pfind_tid_locked(pid);
385 	} else {
386 		p = NULL;
387 	}
388 	sx_sunlock(&allproc_lock);
389 	if (p == NULL)
390 		return (ESRCH);
391 	if ((flags & PGET_CANSEE) != 0) {
392 		error = p_cansee(curthread, p);
393 		if (error != 0)
394 			goto errout;
395 	}
396 	if ((flags & PGET_CANDEBUG) != 0) {
397 		error = p_candebug(curthread, p);
398 		if (error != 0)
399 			goto errout;
400 	}
401 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
402 		error = EPERM;
403 		goto errout;
404 	}
405 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
406 		error = ESRCH;
407 		goto errout;
408 	}
409 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
410 		/*
411 		 * XXXRW: Not clear ESRCH is the right error during proc
412 		 * execve().
413 		 */
414 		error = ESRCH;
415 		goto errout;
416 	}
417 	if ((flags & PGET_HOLD) != 0) {
418 		_PHOLD(p);
419 		PROC_UNLOCK(p);
420 	}
421 	*pp = p;
422 	return (0);
423 errout:
424 	PROC_UNLOCK(p);
425 	return (error);
426 }
427 
428 /*
429  * Create a new process group.
430  * pgid must be equal to the pid of p.
431  * Begin a new session if required.
432  */
433 int
434 enterpgrp(p, pgid, pgrp, sess)
435 	register struct proc *p;
436 	pid_t pgid;
437 	struct pgrp *pgrp;
438 	struct session *sess;
439 {
440 
441 	sx_assert(&proctree_lock, SX_XLOCKED);
442 
443 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
444 	KASSERT(p->p_pid == pgid,
445 	    ("enterpgrp: new pgrp and pid != pgid"));
446 	KASSERT(pgfind(pgid) == NULL,
447 	    ("enterpgrp: pgrp with pgid exists"));
448 	KASSERT(!SESS_LEADER(p),
449 	    ("enterpgrp: session leader attempted setpgrp"));
450 
451 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
452 
453 	if (sess != NULL) {
454 		/*
455 		 * new session
456 		 */
457 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
458 		PROC_LOCK(p);
459 		p->p_flag &= ~P_CONTROLT;
460 		PROC_UNLOCK(p);
461 		PGRP_LOCK(pgrp);
462 		sess->s_leader = p;
463 		sess->s_sid = p->p_pid;
464 		refcount_init(&sess->s_count, 1);
465 		sess->s_ttyvp = NULL;
466 		sess->s_ttydp = NULL;
467 		sess->s_ttyp = NULL;
468 		bcopy(p->p_session->s_login, sess->s_login,
469 			    sizeof(sess->s_login));
470 		pgrp->pg_session = sess;
471 		KASSERT(p == curproc,
472 		    ("enterpgrp: mksession and p != curproc"));
473 	} else {
474 		pgrp->pg_session = p->p_session;
475 		sess_hold(pgrp->pg_session);
476 		PGRP_LOCK(pgrp);
477 	}
478 	pgrp->pg_id = pgid;
479 	LIST_INIT(&pgrp->pg_members);
480 
481 	/*
482 	 * As we have an exclusive lock of proctree_lock,
483 	 * this should not deadlock.
484 	 */
485 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
486 	pgrp->pg_jobc = 0;
487 	SLIST_INIT(&pgrp->pg_sigiolst);
488 	PGRP_UNLOCK(pgrp);
489 
490 	doenterpgrp(p, pgrp);
491 
492 	return (0);
493 }
494 
495 /*
496  * Move p to an existing process group
497  */
498 int
499 enterthispgrp(p, pgrp)
500 	register struct proc *p;
501 	struct pgrp *pgrp;
502 {
503 
504 	sx_assert(&proctree_lock, SX_XLOCKED);
505 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
506 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
507 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
508 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
509 	KASSERT(pgrp->pg_session == p->p_session,
510 		("%s: pgrp's session %p, p->p_session %p.\n",
511 		__func__,
512 		pgrp->pg_session,
513 		p->p_session));
514 	KASSERT(pgrp != p->p_pgrp,
515 		("%s: p belongs to pgrp.", __func__));
516 
517 	doenterpgrp(p, pgrp);
518 
519 	return (0);
520 }
521 
522 /*
523  * Move p to a process group
524  */
525 static void
526 doenterpgrp(p, pgrp)
527 	struct proc *p;
528 	struct pgrp *pgrp;
529 {
530 	struct pgrp *savepgrp;
531 
532 	sx_assert(&proctree_lock, SX_XLOCKED);
533 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
534 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
535 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
536 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
537 
538 	savepgrp = p->p_pgrp;
539 
540 	/*
541 	 * Adjust eligibility of affected pgrps to participate in job control.
542 	 * Increment eligibility counts before decrementing, otherwise we
543 	 * could reach 0 spuriously during the first call.
544 	 */
545 	fixjobc(p, pgrp, 1);
546 	fixjobc(p, p->p_pgrp, 0);
547 
548 	PGRP_LOCK(pgrp);
549 	PGRP_LOCK(savepgrp);
550 	PROC_LOCK(p);
551 	LIST_REMOVE(p, p_pglist);
552 	p->p_pgrp = pgrp;
553 	PROC_UNLOCK(p);
554 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
555 	PGRP_UNLOCK(savepgrp);
556 	PGRP_UNLOCK(pgrp);
557 	if (LIST_EMPTY(&savepgrp->pg_members))
558 		pgdelete(savepgrp);
559 }
560 
561 /*
562  * remove process from process group
563  */
564 int
565 leavepgrp(p)
566 	register struct proc *p;
567 {
568 	struct pgrp *savepgrp;
569 
570 	sx_assert(&proctree_lock, SX_XLOCKED);
571 	savepgrp = p->p_pgrp;
572 	PGRP_LOCK(savepgrp);
573 	PROC_LOCK(p);
574 	LIST_REMOVE(p, p_pglist);
575 	p->p_pgrp = NULL;
576 	PROC_UNLOCK(p);
577 	PGRP_UNLOCK(savepgrp);
578 	if (LIST_EMPTY(&savepgrp->pg_members))
579 		pgdelete(savepgrp);
580 	return (0);
581 }
582 
583 /*
584  * delete a process group
585  */
586 static void
587 pgdelete(pgrp)
588 	register struct pgrp *pgrp;
589 {
590 	struct session *savesess;
591 	struct tty *tp;
592 
593 	sx_assert(&proctree_lock, SX_XLOCKED);
594 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
595 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
596 
597 	/*
598 	 * Reset any sigio structures pointing to us as a result of
599 	 * F_SETOWN with our pgid.
600 	 */
601 	funsetownlst(&pgrp->pg_sigiolst);
602 
603 	PGRP_LOCK(pgrp);
604 	tp = pgrp->pg_session->s_ttyp;
605 	LIST_REMOVE(pgrp, pg_hash);
606 	savesess = pgrp->pg_session;
607 	PGRP_UNLOCK(pgrp);
608 
609 	/* Remove the reference to the pgrp before deallocating it. */
610 	if (tp != NULL) {
611 		tty_lock(tp);
612 		tty_rel_pgrp(tp, pgrp);
613 	}
614 
615 	mtx_destroy(&pgrp->pg_mtx);
616 	free(pgrp, M_PGRP);
617 	sess_release(savesess);
618 }
619 
620 static void
621 pgadjustjobc(pgrp, entering)
622 	struct pgrp *pgrp;
623 	int entering;
624 {
625 
626 	PGRP_LOCK(pgrp);
627 	if (entering)
628 		pgrp->pg_jobc++;
629 	else {
630 		--pgrp->pg_jobc;
631 		if (pgrp->pg_jobc == 0)
632 			orphanpg(pgrp);
633 	}
634 	PGRP_UNLOCK(pgrp);
635 }
636 
637 /*
638  * Adjust pgrp jobc counters when specified process changes process group.
639  * We count the number of processes in each process group that "qualify"
640  * the group for terminal job control (those with a parent in a different
641  * process group of the same session).  If that count reaches zero, the
642  * process group becomes orphaned.  Check both the specified process'
643  * process group and that of its children.
644  * entering == 0 => p is leaving specified group.
645  * entering == 1 => p is entering specified group.
646  */
647 void
648 fixjobc(p, pgrp, entering)
649 	register struct proc *p;
650 	register struct pgrp *pgrp;
651 	int entering;
652 {
653 	register struct pgrp *hispgrp;
654 	register struct session *mysession;
655 
656 	sx_assert(&proctree_lock, SX_LOCKED);
657 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
658 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
659 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
660 
661 	/*
662 	 * Check p's parent to see whether p qualifies its own process
663 	 * group; if so, adjust count for p's process group.
664 	 */
665 	mysession = pgrp->pg_session;
666 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
667 	    hispgrp->pg_session == mysession)
668 		pgadjustjobc(pgrp, entering);
669 
670 	/*
671 	 * Check this process' children to see whether they qualify
672 	 * their process groups; if so, adjust counts for children's
673 	 * process groups.
674 	 */
675 	LIST_FOREACH(p, &p->p_children, p_sibling) {
676 		hispgrp = p->p_pgrp;
677 		if (hispgrp == pgrp ||
678 		    hispgrp->pg_session != mysession)
679 			continue;
680 		PROC_LOCK(p);
681 		if (p->p_state == PRS_ZOMBIE) {
682 			PROC_UNLOCK(p);
683 			continue;
684 		}
685 		PROC_UNLOCK(p);
686 		pgadjustjobc(hispgrp, entering);
687 	}
688 }
689 
690 /*
691  * A process group has become orphaned;
692  * if there are any stopped processes in the group,
693  * hang-up all process in that group.
694  */
695 static void
696 orphanpg(pg)
697 	struct pgrp *pg;
698 {
699 	register struct proc *p;
700 
701 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
702 
703 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
704 		PROC_LOCK(p);
705 		if (P_SHOULDSTOP(p)) {
706 			PROC_UNLOCK(p);
707 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
708 				PROC_LOCK(p);
709 				kern_psignal(p, SIGHUP);
710 				kern_psignal(p, SIGCONT);
711 				PROC_UNLOCK(p);
712 			}
713 			return;
714 		}
715 		PROC_UNLOCK(p);
716 	}
717 }
718 
719 void
720 sess_hold(struct session *s)
721 {
722 
723 	refcount_acquire(&s->s_count);
724 }
725 
726 void
727 sess_release(struct session *s)
728 {
729 
730 	if (refcount_release(&s->s_count)) {
731 		if (s->s_ttyp != NULL) {
732 			tty_lock(s->s_ttyp);
733 			tty_rel_sess(s->s_ttyp, s);
734 		}
735 		mtx_destroy(&s->s_mtx);
736 		free(s, M_SESSION);
737 	}
738 }
739 
740 #ifdef DDB
741 
742 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
743 {
744 	register struct pgrp *pgrp;
745 	register struct proc *p;
746 	register int i;
747 
748 	for (i = 0; i <= pgrphash; i++) {
749 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
750 			printf("\tindx %d\n", i);
751 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
752 				printf(
753 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
754 				    (void *)pgrp, (long)pgrp->pg_id,
755 				    (void *)pgrp->pg_session,
756 				    pgrp->pg_session->s_count,
757 				    (void *)LIST_FIRST(&pgrp->pg_members));
758 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
759 					printf("\t\tpid %ld addr %p pgrp %p\n",
760 					    (long)p->p_pid, (void *)p,
761 					    (void *)p->p_pgrp);
762 				}
763 			}
764 		}
765 	}
766 }
767 #endif /* DDB */
768 
769 /*
770  * Calculate the kinfo_proc members which contain process-wide
771  * informations.
772  * Must be called with the target process locked.
773  */
774 static void
775 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
776 {
777 	struct thread *td;
778 
779 	PROC_LOCK_ASSERT(p, MA_OWNED);
780 
781 	kp->ki_estcpu = 0;
782 	kp->ki_pctcpu = 0;
783 	FOREACH_THREAD_IN_PROC(p, td) {
784 		thread_lock(td);
785 		kp->ki_pctcpu += sched_pctcpu(td);
786 		kp->ki_estcpu += td->td_estcpu;
787 		thread_unlock(td);
788 	}
789 }
790 
791 /*
792  * Clear kinfo_proc and fill in any information that is common
793  * to all threads in the process.
794  * Must be called with the target process locked.
795  */
796 static void
797 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
798 {
799 	struct thread *td0;
800 	struct tty *tp;
801 	struct session *sp;
802 	struct ucred *cred;
803 	struct sigacts *ps;
804 
805 	PROC_LOCK_ASSERT(p, MA_OWNED);
806 	bzero(kp, sizeof(*kp));
807 
808 	kp->ki_structsize = sizeof(*kp);
809 	kp->ki_paddr = p;
810 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
811 	kp->ki_args = p->p_args;
812 	kp->ki_textvp = p->p_textvp;
813 #ifdef KTRACE
814 	kp->ki_tracep = p->p_tracevp;
815 	kp->ki_traceflag = p->p_traceflag;
816 #endif
817 	kp->ki_fd = p->p_fd;
818 	kp->ki_vmspace = p->p_vmspace;
819 	kp->ki_flag = p->p_flag;
820 	cred = p->p_ucred;
821 	if (cred) {
822 		kp->ki_uid = cred->cr_uid;
823 		kp->ki_ruid = cred->cr_ruid;
824 		kp->ki_svuid = cred->cr_svuid;
825 		kp->ki_cr_flags = 0;
826 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
827 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
828 		/* XXX bde doesn't like KI_NGROUPS */
829 		if (cred->cr_ngroups > KI_NGROUPS) {
830 			kp->ki_ngroups = KI_NGROUPS;
831 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
832 		} else
833 			kp->ki_ngroups = cred->cr_ngroups;
834 		bcopy(cred->cr_groups, kp->ki_groups,
835 		    kp->ki_ngroups * sizeof(gid_t));
836 		kp->ki_rgid = cred->cr_rgid;
837 		kp->ki_svgid = cred->cr_svgid;
838 		/* If jailed(cred), emulate the old P_JAILED flag. */
839 		if (jailed(cred)) {
840 			kp->ki_flag |= P_JAILED;
841 			/* If inside the jail, use 0 as a jail ID. */
842 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
843 				kp->ki_jid = cred->cr_prison->pr_id;
844 		}
845 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
846 		    sizeof(kp->ki_loginclass));
847 	}
848 	ps = p->p_sigacts;
849 	if (ps) {
850 		mtx_lock(&ps->ps_mtx);
851 		kp->ki_sigignore = ps->ps_sigignore;
852 		kp->ki_sigcatch = ps->ps_sigcatch;
853 		mtx_unlock(&ps->ps_mtx);
854 	}
855 	if (p->p_state != PRS_NEW &&
856 	    p->p_state != PRS_ZOMBIE &&
857 	    p->p_vmspace != NULL) {
858 		struct vmspace *vm = p->p_vmspace;
859 
860 		kp->ki_size = vm->vm_map.size;
861 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
862 		FOREACH_THREAD_IN_PROC(p, td0) {
863 			if (!TD_IS_SWAPPED(td0))
864 				kp->ki_rssize += td0->td_kstack_pages;
865 		}
866 		kp->ki_swrss = vm->vm_swrss;
867 		kp->ki_tsize = vm->vm_tsize;
868 		kp->ki_dsize = vm->vm_dsize;
869 		kp->ki_ssize = vm->vm_ssize;
870 	} else if (p->p_state == PRS_ZOMBIE)
871 		kp->ki_stat = SZOMB;
872 	if (kp->ki_flag & P_INMEM)
873 		kp->ki_sflag = PS_INMEM;
874 	else
875 		kp->ki_sflag = 0;
876 	/* Calculate legacy swtime as seconds since 'swtick'. */
877 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
878 	kp->ki_pid = p->p_pid;
879 	kp->ki_nice = p->p_nice;
880 	kp->ki_start = p->p_stats->p_start;
881 	timevaladd(&kp->ki_start, &boottime);
882 	PROC_SLOCK(p);
883 	rufetch(p, &kp->ki_rusage);
884 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
885 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
886 	PROC_SUNLOCK(p);
887 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
888 	/* Some callers want child times in a single value. */
889 	kp->ki_childtime = kp->ki_childstime;
890 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
891 
892 	FOREACH_THREAD_IN_PROC(p, td0)
893 		kp->ki_cow += td0->td_cow;
894 
895 	tp = NULL;
896 	if (p->p_pgrp) {
897 		kp->ki_pgid = p->p_pgrp->pg_id;
898 		kp->ki_jobc = p->p_pgrp->pg_jobc;
899 		sp = p->p_pgrp->pg_session;
900 
901 		if (sp != NULL) {
902 			kp->ki_sid = sp->s_sid;
903 			SESS_LOCK(sp);
904 			strlcpy(kp->ki_login, sp->s_login,
905 			    sizeof(kp->ki_login));
906 			if (sp->s_ttyvp)
907 				kp->ki_kiflag |= KI_CTTY;
908 			if (SESS_LEADER(p))
909 				kp->ki_kiflag |= KI_SLEADER;
910 			/* XXX proctree_lock */
911 			tp = sp->s_ttyp;
912 			SESS_UNLOCK(sp);
913 		}
914 	}
915 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
916 		kp->ki_tdev = tty_udev(tp);
917 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
918 		if (tp->t_session)
919 			kp->ki_tsid = tp->t_session->s_sid;
920 	} else
921 		kp->ki_tdev = NODEV;
922 	if (p->p_comm[0] != '\0')
923 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
924 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
925 	    p->p_sysent->sv_name[0] != '\0')
926 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
927 	kp->ki_siglist = p->p_siglist;
928 	kp->ki_xstat = p->p_xstat;
929 	kp->ki_acflag = p->p_acflag;
930 	kp->ki_lock = p->p_lock;
931 	if (p->p_pptr)
932 		kp->ki_ppid = p->p_pptr->p_pid;
933 }
934 
935 /*
936  * Fill in information that is thread specific.  Must be called with
937  * target process locked.  If 'preferthread' is set, overwrite certain
938  * process-related fields that are maintained for both threads and
939  * processes.
940  */
941 static void
942 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
943 {
944 	struct proc *p;
945 
946 	p = td->td_proc;
947 	kp->ki_tdaddr = td;
948 	PROC_LOCK_ASSERT(p, MA_OWNED);
949 
950 	if (preferthread)
951 		PROC_SLOCK(p);
952 	thread_lock(td);
953 	if (td->td_wmesg != NULL)
954 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
955 	else
956 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
957 	strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname));
958 	if (TD_ON_LOCK(td)) {
959 		kp->ki_kiflag |= KI_LOCKBLOCK;
960 		strlcpy(kp->ki_lockname, td->td_lockname,
961 		    sizeof(kp->ki_lockname));
962 	} else {
963 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
964 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
965 	}
966 
967 	if (p->p_state == PRS_NORMAL) { /* approximate. */
968 		if (TD_ON_RUNQ(td) ||
969 		    TD_CAN_RUN(td) ||
970 		    TD_IS_RUNNING(td)) {
971 			kp->ki_stat = SRUN;
972 		} else if (P_SHOULDSTOP(p)) {
973 			kp->ki_stat = SSTOP;
974 		} else if (TD_IS_SLEEPING(td)) {
975 			kp->ki_stat = SSLEEP;
976 		} else if (TD_ON_LOCK(td)) {
977 			kp->ki_stat = SLOCK;
978 		} else {
979 			kp->ki_stat = SWAIT;
980 		}
981 	} else if (p->p_state == PRS_ZOMBIE) {
982 		kp->ki_stat = SZOMB;
983 	} else {
984 		kp->ki_stat = SIDL;
985 	}
986 
987 	/* Things in the thread */
988 	kp->ki_wchan = td->td_wchan;
989 	kp->ki_pri.pri_level = td->td_priority;
990 	kp->ki_pri.pri_native = td->td_base_pri;
991 	kp->ki_lastcpu = td->td_lastcpu;
992 	kp->ki_oncpu = td->td_oncpu;
993 	kp->ki_tdflags = td->td_flags;
994 	kp->ki_tid = td->td_tid;
995 	kp->ki_numthreads = p->p_numthreads;
996 	kp->ki_pcb = td->td_pcb;
997 	kp->ki_kstack = (void *)td->td_kstack;
998 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
999 	kp->ki_pri.pri_class = td->td_pri_class;
1000 	kp->ki_pri.pri_user = td->td_user_pri;
1001 
1002 	if (preferthread) {
1003 		rufetchtd(td, &kp->ki_rusage);
1004 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1005 		kp->ki_pctcpu = sched_pctcpu(td);
1006 		kp->ki_estcpu = td->td_estcpu;
1007 		kp->ki_cow = td->td_cow;
1008 	}
1009 
1010 	/* We can't get this anymore but ps etc never used it anyway. */
1011 	kp->ki_rqindex = 0;
1012 
1013 	if (preferthread)
1014 		kp->ki_siglist = td->td_siglist;
1015 	kp->ki_sigmask = td->td_sigmask;
1016 	thread_unlock(td);
1017 	if (preferthread)
1018 		PROC_SUNLOCK(p);
1019 }
1020 
1021 /*
1022  * Fill in a kinfo_proc structure for the specified process.
1023  * Must be called with the target process locked.
1024  */
1025 void
1026 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1027 {
1028 
1029 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1030 
1031 	fill_kinfo_proc_only(p, kp);
1032 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1033 	fill_kinfo_aggregate(p, kp);
1034 }
1035 
1036 struct pstats *
1037 pstats_alloc(void)
1038 {
1039 
1040 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1041 }
1042 
1043 /*
1044  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1045  */
1046 void
1047 pstats_fork(struct pstats *src, struct pstats *dst)
1048 {
1049 
1050 	bzero(&dst->pstat_startzero,
1051 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1052 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1053 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1054 }
1055 
1056 void
1057 pstats_free(struct pstats *ps)
1058 {
1059 
1060 	free(ps, M_SUBPROC);
1061 }
1062 
1063 static struct proc *
1064 zpfind_locked(pid_t pid)
1065 {
1066 	struct proc *p;
1067 
1068 	sx_assert(&allproc_lock, SX_LOCKED);
1069 	LIST_FOREACH(p, &zombproc, p_list) {
1070 		if (p->p_pid == pid) {
1071 			PROC_LOCK(p);
1072 			break;
1073 		}
1074 	}
1075 	return (p);
1076 }
1077 
1078 /*
1079  * Locate a zombie process by number
1080  */
1081 struct proc *
1082 zpfind(pid_t pid)
1083 {
1084 	struct proc *p;
1085 
1086 	sx_slock(&allproc_lock);
1087 	p = zpfind_locked(pid);
1088 	sx_sunlock(&allproc_lock);
1089 	return (p);
1090 }
1091 
1092 #define KERN_PROC_ZOMBMASK	0x3
1093 #define KERN_PROC_NOTHREADS	0x4
1094 
1095 #ifdef COMPAT_FREEBSD32
1096 
1097 /*
1098  * This function is typically used to copy out the kernel address, so
1099  * it can be replaced by assignment of zero.
1100  */
1101 static inline uint32_t
1102 ptr32_trim(void *ptr)
1103 {
1104 	uintptr_t uptr;
1105 
1106 	uptr = (uintptr_t)ptr;
1107 	return ((uptr > UINT_MAX) ? 0 : uptr);
1108 }
1109 
1110 #define PTRTRIM_CP(src,dst,fld) \
1111 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1112 
1113 static void
1114 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1115 {
1116 	int i;
1117 
1118 	bzero(ki32, sizeof(struct kinfo_proc32));
1119 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1120 	CP(*ki, *ki32, ki_layout);
1121 	PTRTRIM_CP(*ki, *ki32, ki_args);
1122 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1123 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1124 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1125 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1126 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1127 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1128 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1129 	CP(*ki, *ki32, ki_pid);
1130 	CP(*ki, *ki32, ki_ppid);
1131 	CP(*ki, *ki32, ki_pgid);
1132 	CP(*ki, *ki32, ki_tpgid);
1133 	CP(*ki, *ki32, ki_sid);
1134 	CP(*ki, *ki32, ki_tsid);
1135 	CP(*ki, *ki32, ki_jobc);
1136 	CP(*ki, *ki32, ki_tdev);
1137 	CP(*ki, *ki32, ki_siglist);
1138 	CP(*ki, *ki32, ki_sigmask);
1139 	CP(*ki, *ki32, ki_sigignore);
1140 	CP(*ki, *ki32, ki_sigcatch);
1141 	CP(*ki, *ki32, ki_uid);
1142 	CP(*ki, *ki32, ki_ruid);
1143 	CP(*ki, *ki32, ki_svuid);
1144 	CP(*ki, *ki32, ki_rgid);
1145 	CP(*ki, *ki32, ki_svgid);
1146 	CP(*ki, *ki32, ki_ngroups);
1147 	for (i = 0; i < KI_NGROUPS; i++)
1148 		CP(*ki, *ki32, ki_groups[i]);
1149 	CP(*ki, *ki32, ki_size);
1150 	CP(*ki, *ki32, ki_rssize);
1151 	CP(*ki, *ki32, ki_swrss);
1152 	CP(*ki, *ki32, ki_tsize);
1153 	CP(*ki, *ki32, ki_dsize);
1154 	CP(*ki, *ki32, ki_ssize);
1155 	CP(*ki, *ki32, ki_xstat);
1156 	CP(*ki, *ki32, ki_acflag);
1157 	CP(*ki, *ki32, ki_pctcpu);
1158 	CP(*ki, *ki32, ki_estcpu);
1159 	CP(*ki, *ki32, ki_slptime);
1160 	CP(*ki, *ki32, ki_swtime);
1161 	CP(*ki, *ki32, ki_cow);
1162 	CP(*ki, *ki32, ki_runtime);
1163 	TV_CP(*ki, *ki32, ki_start);
1164 	TV_CP(*ki, *ki32, ki_childtime);
1165 	CP(*ki, *ki32, ki_flag);
1166 	CP(*ki, *ki32, ki_kiflag);
1167 	CP(*ki, *ki32, ki_traceflag);
1168 	CP(*ki, *ki32, ki_stat);
1169 	CP(*ki, *ki32, ki_nice);
1170 	CP(*ki, *ki32, ki_lock);
1171 	CP(*ki, *ki32, ki_rqindex);
1172 	CP(*ki, *ki32, ki_oncpu);
1173 	CP(*ki, *ki32, ki_lastcpu);
1174 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1175 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1176 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1177 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1178 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1179 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1180 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1181 	CP(*ki, *ki32, ki_cr_flags);
1182 	CP(*ki, *ki32, ki_jid);
1183 	CP(*ki, *ki32, ki_numthreads);
1184 	CP(*ki, *ki32, ki_tid);
1185 	CP(*ki, *ki32, ki_pri);
1186 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1187 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1188 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1189 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1190 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1191 	CP(*ki, *ki32, ki_sflag);
1192 	CP(*ki, *ki32, ki_tdflags);
1193 }
1194 
1195 static int
1196 sysctl_out_proc_copyout(struct kinfo_proc *ki, struct sysctl_req *req)
1197 {
1198 	struct kinfo_proc32 ki32;
1199 	int error;
1200 
1201 	if (req->flags & SCTL_MASK32) {
1202 		freebsd32_kinfo_proc_out(ki, &ki32);
1203 		error = SYSCTL_OUT(req, &ki32, sizeof(struct kinfo_proc32));
1204 	} else
1205 		error = SYSCTL_OUT(req, ki, sizeof(struct kinfo_proc));
1206 	return (error);
1207 }
1208 #else
1209 static int
1210 sysctl_out_proc_copyout(struct kinfo_proc *ki, struct sysctl_req *req)
1211 {
1212 
1213 	return (SYSCTL_OUT(req, ki, sizeof(struct kinfo_proc)));
1214 }
1215 #endif
1216 
1217 /*
1218  * Must be called with the process locked and will return with it unlocked.
1219  */
1220 static int
1221 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1222 {
1223 	struct thread *td;
1224 	struct kinfo_proc kinfo_proc;
1225 	int error = 0;
1226 	struct proc *np;
1227 	pid_t pid = p->p_pid;
1228 
1229 	PROC_LOCK_ASSERT(p, MA_OWNED);
1230 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1231 
1232 	fill_kinfo_proc(p, &kinfo_proc);
1233 	if (flags & KERN_PROC_NOTHREADS)
1234 		error = sysctl_out_proc_copyout(&kinfo_proc, req);
1235 	else {
1236 		FOREACH_THREAD_IN_PROC(p, td) {
1237 			fill_kinfo_thread(td, &kinfo_proc, 1);
1238 			error = sysctl_out_proc_copyout(&kinfo_proc, req);
1239 			if (error)
1240 				break;
1241 		}
1242 	}
1243 	PROC_UNLOCK(p);
1244 	if (error)
1245 		return (error);
1246 	if (flags & KERN_PROC_ZOMBMASK)
1247 		np = zpfind(pid);
1248 	else {
1249 		if (pid == 0)
1250 			return (0);
1251 		np = pfind(pid);
1252 	}
1253 	if (np == NULL)
1254 		return (ESRCH);
1255 	if (np != p) {
1256 		PROC_UNLOCK(np);
1257 		return (ESRCH);
1258 	}
1259 	PROC_UNLOCK(np);
1260 	return (0);
1261 }
1262 
1263 static int
1264 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1265 {
1266 	int *name = (int *)arg1;
1267 	u_int namelen = arg2;
1268 	struct proc *p;
1269 	int flags, doingzomb, oid_number;
1270 	int error = 0;
1271 
1272 	oid_number = oidp->oid_number;
1273 	if (oid_number != KERN_PROC_ALL &&
1274 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1275 		flags = KERN_PROC_NOTHREADS;
1276 	else {
1277 		flags = 0;
1278 		oid_number &= ~KERN_PROC_INC_THREAD;
1279 	}
1280 	if (oid_number == KERN_PROC_PID) {
1281 		if (namelen != 1)
1282 			return (EINVAL);
1283 		error = sysctl_wire_old_buffer(req, 0);
1284 		if (error)
1285 			return (error);
1286 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1287 		if (error != 0)
1288 			return (error);
1289 		error = sysctl_out_proc(p, req, flags);
1290 		return (error);
1291 	}
1292 
1293 	switch (oid_number) {
1294 	case KERN_PROC_ALL:
1295 		if (namelen != 0)
1296 			return (EINVAL);
1297 		break;
1298 	case KERN_PROC_PROC:
1299 		if (namelen != 0 && namelen != 1)
1300 			return (EINVAL);
1301 		break;
1302 	default:
1303 		if (namelen != 1)
1304 			return (EINVAL);
1305 		break;
1306 	}
1307 
1308 	if (!req->oldptr) {
1309 		/* overestimate by 5 procs */
1310 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1311 		if (error)
1312 			return (error);
1313 	}
1314 	error = sysctl_wire_old_buffer(req, 0);
1315 	if (error != 0)
1316 		return (error);
1317 	sx_slock(&allproc_lock);
1318 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1319 		if (!doingzomb)
1320 			p = LIST_FIRST(&allproc);
1321 		else
1322 			p = LIST_FIRST(&zombproc);
1323 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
1324 			/*
1325 			 * Skip embryonic processes.
1326 			 */
1327 			PROC_LOCK(p);
1328 			if (p->p_state == PRS_NEW) {
1329 				PROC_UNLOCK(p);
1330 				continue;
1331 			}
1332 			KASSERT(p->p_ucred != NULL,
1333 			    ("process credential is NULL for non-NEW proc"));
1334 			/*
1335 			 * Show a user only appropriate processes.
1336 			 */
1337 			if (p_cansee(curthread, p)) {
1338 				PROC_UNLOCK(p);
1339 				continue;
1340 			}
1341 			/*
1342 			 * TODO - make more efficient (see notes below).
1343 			 * do by session.
1344 			 */
1345 			switch (oid_number) {
1346 
1347 			case KERN_PROC_GID:
1348 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1349 					PROC_UNLOCK(p);
1350 					continue;
1351 				}
1352 				break;
1353 
1354 			case KERN_PROC_PGRP:
1355 				/* could do this by traversing pgrp */
1356 				if (p->p_pgrp == NULL ||
1357 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1358 					PROC_UNLOCK(p);
1359 					continue;
1360 				}
1361 				break;
1362 
1363 			case KERN_PROC_RGID:
1364 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1365 					PROC_UNLOCK(p);
1366 					continue;
1367 				}
1368 				break;
1369 
1370 			case KERN_PROC_SESSION:
1371 				if (p->p_session == NULL ||
1372 				    p->p_session->s_sid != (pid_t)name[0]) {
1373 					PROC_UNLOCK(p);
1374 					continue;
1375 				}
1376 				break;
1377 
1378 			case KERN_PROC_TTY:
1379 				if ((p->p_flag & P_CONTROLT) == 0 ||
1380 				    p->p_session == NULL) {
1381 					PROC_UNLOCK(p);
1382 					continue;
1383 				}
1384 				/* XXX proctree_lock */
1385 				SESS_LOCK(p->p_session);
1386 				if (p->p_session->s_ttyp == NULL ||
1387 				    tty_udev(p->p_session->s_ttyp) !=
1388 				    (dev_t)name[0]) {
1389 					SESS_UNLOCK(p->p_session);
1390 					PROC_UNLOCK(p);
1391 					continue;
1392 				}
1393 				SESS_UNLOCK(p->p_session);
1394 				break;
1395 
1396 			case KERN_PROC_UID:
1397 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1398 					PROC_UNLOCK(p);
1399 					continue;
1400 				}
1401 				break;
1402 
1403 			case KERN_PROC_RUID:
1404 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1405 					PROC_UNLOCK(p);
1406 					continue;
1407 				}
1408 				break;
1409 
1410 			case KERN_PROC_PROC:
1411 				break;
1412 
1413 			default:
1414 				break;
1415 
1416 			}
1417 
1418 			error = sysctl_out_proc(p, req, flags | doingzomb);
1419 			if (error) {
1420 				sx_sunlock(&allproc_lock);
1421 				return (error);
1422 			}
1423 		}
1424 	}
1425 	sx_sunlock(&allproc_lock);
1426 	return (0);
1427 }
1428 
1429 struct pargs *
1430 pargs_alloc(int len)
1431 {
1432 	struct pargs *pa;
1433 
1434 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1435 		M_WAITOK);
1436 	refcount_init(&pa->ar_ref, 1);
1437 	pa->ar_length = len;
1438 	return (pa);
1439 }
1440 
1441 static void
1442 pargs_free(struct pargs *pa)
1443 {
1444 
1445 	free(pa, M_PARGS);
1446 }
1447 
1448 void
1449 pargs_hold(struct pargs *pa)
1450 {
1451 
1452 	if (pa == NULL)
1453 		return;
1454 	refcount_acquire(&pa->ar_ref);
1455 }
1456 
1457 void
1458 pargs_drop(struct pargs *pa)
1459 {
1460 
1461 	if (pa == NULL)
1462 		return;
1463 	if (refcount_release(&pa->ar_ref))
1464 		pargs_free(pa);
1465 }
1466 
1467 static int
1468 proc_read_mem(struct thread *td, struct proc *p, vm_offset_t offset, void* buf,
1469     size_t len)
1470 {
1471 	struct iovec iov;
1472 	struct uio uio;
1473 
1474 	iov.iov_base = (caddr_t)buf;
1475 	iov.iov_len = len;
1476 	uio.uio_iov = &iov;
1477 	uio.uio_iovcnt = 1;
1478 	uio.uio_offset = offset;
1479 	uio.uio_resid = (ssize_t)len;
1480 	uio.uio_segflg = UIO_SYSSPACE;
1481 	uio.uio_rw = UIO_READ;
1482 	uio.uio_td = td;
1483 
1484 	return (proc_rwmem(p, &uio));
1485 }
1486 
1487 static int
1488 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1489     size_t len)
1490 {
1491 	size_t i;
1492 	int error;
1493 
1494 	error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, len);
1495 	/*
1496 	 * Reading the chunk may validly return EFAULT if the string is shorter
1497 	 * than the chunk and is aligned at the end of the page, assuming the
1498 	 * next page is not mapped.  So if EFAULT is returned do a fallback to
1499 	 * one byte read loop.
1500 	 */
1501 	if (error == EFAULT) {
1502 		for (i = 0; i < len; i++, buf++, sptr++) {
1503 			error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, 1);
1504 			if (error != 0)
1505 				return (error);
1506 			if (*buf == '\0')
1507 				break;
1508 		}
1509 		error = 0;
1510 	}
1511 	return (error);
1512 }
1513 
1514 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1515 
1516 enum proc_vector_type {
1517 	PROC_ARG,
1518 	PROC_ENV,
1519 	PROC_AUX,
1520 };
1521 
1522 #ifdef COMPAT_FREEBSD32
1523 static int
1524 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1525     size_t *vsizep, enum proc_vector_type type)
1526 {
1527 	struct freebsd32_ps_strings pss;
1528 	Elf32_Auxinfo aux;
1529 	vm_offset_t vptr, ptr;
1530 	uint32_t *proc_vector32;
1531 	char **proc_vector;
1532 	size_t vsize, size;
1533 	int i, error;
1534 
1535 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1536 	    &pss, sizeof(pss));
1537 	if (error != 0)
1538 		return (error);
1539 	switch (type) {
1540 	case PROC_ARG:
1541 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1542 		vsize = pss.ps_nargvstr;
1543 		if (vsize > ARG_MAX)
1544 			return (ENOEXEC);
1545 		size = vsize * sizeof(int32_t);
1546 		break;
1547 	case PROC_ENV:
1548 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1549 		vsize = pss.ps_nenvstr;
1550 		if (vsize > ARG_MAX)
1551 			return (ENOEXEC);
1552 		size = vsize * sizeof(int32_t);
1553 		break;
1554 	case PROC_AUX:
1555 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1556 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1557 		if (vptr % 4 != 0)
1558 			return (ENOEXEC);
1559 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1560 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1561 			if (error != 0)
1562 				return (error);
1563 			if (aux.a_type == AT_NULL)
1564 				break;
1565 			ptr += sizeof(aux);
1566 		}
1567 		if (aux.a_type != AT_NULL)
1568 			return (ENOEXEC);
1569 		vsize = i + 1;
1570 		size = vsize * sizeof(aux);
1571 		break;
1572 	default:
1573 		KASSERT(0, ("Wrong proc vector type: %d", type));
1574 		return (EINVAL);
1575 	}
1576 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1577 	error = proc_read_mem(td, p, vptr, proc_vector32, size);
1578 	if (error != 0)
1579 		goto done;
1580 	if (type == PROC_AUX) {
1581 		*proc_vectorp = (char **)proc_vector32;
1582 		*vsizep = vsize;
1583 		return (0);
1584 	}
1585 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1586 	for (i = 0; i < (int)vsize; i++)
1587 		proc_vector[i] = PTRIN(proc_vector32[i]);
1588 	*proc_vectorp = proc_vector;
1589 	*vsizep = vsize;
1590 done:
1591 	free(proc_vector32, M_TEMP);
1592 	return (error);
1593 }
1594 #endif
1595 
1596 static int
1597 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1598     size_t *vsizep, enum proc_vector_type type)
1599 {
1600 	struct ps_strings pss;
1601 	Elf_Auxinfo aux;
1602 	vm_offset_t vptr, ptr;
1603 	char **proc_vector;
1604 	size_t vsize, size;
1605 	int error, i;
1606 
1607 #ifdef COMPAT_FREEBSD32
1608 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1609 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1610 #endif
1611 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1612 	    &pss, sizeof(pss));
1613 	if (error != 0)
1614 		return (error);
1615 	switch (type) {
1616 	case PROC_ARG:
1617 		vptr = (vm_offset_t)pss.ps_argvstr;
1618 		vsize = pss.ps_nargvstr;
1619 		if (vsize > ARG_MAX)
1620 			return (ENOEXEC);
1621 		size = vsize * sizeof(char *);
1622 		break;
1623 	case PROC_ENV:
1624 		vptr = (vm_offset_t)pss.ps_envstr;
1625 		vsize = pss.ps_nenvstr;
1626 		if (vsize > ARG_MAX)
1627 			return (ENOEXEC);
1628 		size = vsize * sizeof(char *);
1629 		break;
1630 	case PROC_AUX:
1631 		/*
1632 		 * The aux array is just above env array on the stack. Check
1633 		 * that the address is naturally aligned.
1634 		 */
1635 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1636 		    * sizeof(char *);
1637 #if __ELF_WORD_SIZE == 64
1638 		if (vptr % sizeof(uint64_t) != 0)
1639 #else
1640 		if (vptr % sizeof(uint32_t) != 0)
1641 #endif
1642 			return (ENOEXEC);
1643 		/*
1644 		 * We count the array size reading the aux vectors from the
1645 		 * stack until AT_NULL vector is returned.  So (to keep the code
1646 		 * simple) we read the process stack twice: the first time here
1647 		 * to find the size and the second time when copying the vectors
1648 		 * to the allocated proc_vector.
1649 		 */
1650 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1651 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1652 			if (error != 0)
1653 				return (error);
1654 			if (aux.a_type == AT_NULL)
1655 				break;
1656 			ptr += sizeof(aux);
1657 		}
1658 		/*
1659 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1660 		 * not reached AT_NULL, it is most likely we are reading wrong
1661 		 * data: either the process doesn't have auxv array or data has
1662 		 * been modified. Return the error in this case.
1663 		 */
1664 		if (aux.a_type != AT_NULL)
1665 			return (ENOEXEC);
1666 		vsize = i + 1;
1667 		size = vsize * sizeof(aux);
1668 		break;
1669 	default:
1670 		KASSERT(0, ("Wrong proc vector type: %d", type));
1671 		return (EINVAL); /* In case we are built without INVARIANTS. */
1672 	}
1673 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1674 	if (proc_vector == NULL)
1675 		return (ENOMEM);
1676 	error = proc_read_mem(td, p, vptr, proc_vector, size);
1677 	if (error != 0) {
1678 		free(proc_vector, M_TEMP);
1679 		return (error);
1680 	}
1681 	*proc_vectorp = proc_vector;
1682 	*vsizep = vsize;
1683 
1684 	return (0);
1685 }
1686 
1687 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1688 
1689 static int
1690 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1691     enum proc_vector_type type)
1692 {
1693 	size_t done, len, nchr, vsize;
1694 	int error, i;
1695 	char **proc_vector, *sptr;
1696 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1697 
1698 	PROC_ASSERT_HELD(p);
1699 
1700 	/*
1701 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1702 	 */
1703 	nchr = 2 * (PATH_MAX + ARG_MAX);
1704 
1705 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1706 	if (error != 0)
1707 		return (error);
1708 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1709 		/*
1710 		 * The program may have scribbled into its argv array, e.g. to
1711 		 * remove some arguments.  If that has happened, break out
1712 		 * before trying to read from NULL.
1713 		 */
1714 		if (proc_vector[i] == NULL)
1715 			break;
1716 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1717 			error = proc_read_string(td, p, sptr, pss_string,
1718 			    sizeof(pss_string));
1719 			if (error != 0)
1720 				goto done;
1721 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1722 			if (done + len >= nchr)
1723 				len = nchr - done - 1;
1724 			sbuf_bcat(sb, pss_string, len);
1725 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1726 				break;
1727 			done += GET_PS_STRINGS_CHUNK_SZ;
1728 		}
1729 		sbuf_bcat(sb, "", 1);
1730 		done += len + 1;
1731 	}
1732 done:
1733 	free(proc_vector, M_TEMP);
1734 	return (error);
1735 }
1736 
1737 int
1738 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1739 {
1740 
1741 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1742 }
1743 
1744 int
1745 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1746 {
1747 
1748 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1749 }
1750 
1751 /*
1752  * This sysctl allows a process to retrieve the argument list or process
1753  * title for another process without groping around in the address space
1754  * of the other process.  It also allow a process to set its own "process
1755  * title to a string of its own choice.
1756  */
1757 static int
1758 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1759 {
1760 	int *name = (int *)arg1;
1761 	u_int namelen = arg2;
1762 	struct pargs *newpa, *pa;
1763 	struct proc *p;
1764 	struct sbuf sb;
1765 	int flags, error = 0, error2;
1766 
1767 	if (namelen != 1)
1768 		return (EINVAL);
1769 
1770 	flags = PGET_CANSEE;
1771 	if (req->newptr != NULL)
1772 		flags |= PGET_ISCURRENT;
1773 	error = pget((pid_t)name[0], flags, &p);
1774 	if (error)
1775 		return (error);
1776 
1777 	pa = p->p_args;
1778 	if (pa != NULL) {
1779 		pargs_hold(pa);
1780 		PROC_UNLOCK(p);
1781 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1782 		pargs_drop(pa);
1783 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1784 		_PHOLD(p);
1785 		PROC_UNLOCK(p);
1786 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1787 		error = proc_getargv(curthread, p, &sb);
1788 		error2 = sbuf_finish(&sb);
1789 		PRELE(p);
1790 		sbuf_delete(&sb);
1791 		if (error == 0 && error2 != 0)
1792 			error = error2;
1793 	} else {
1794 		PROC_UNLOCK(p);
1795 	}
1796 	if (error != 0 || req->newptr == NULL)
1797 		return (error);
1798 
1799 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1800 		return (ENOMEM);
1801 	newpa = pargs_alloc(req->newlen);
1802 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1803 	if (error != 0) {
1804 		pargs_free(newpa);
1805 		return (error);
1806 	}
1807 	PROC_LOCK(p);
1808 	pa = p->p_args;
1809 	p->p_args = newpa;
1810 	PROC_UNLOCK(p);
1811 	pargs_drop(pa);
1812 	return (0);
1813 }
1814 
1815 /*
1816  * This sysctl allows a process to retrieve environment of another process.
1817  */
1818 static int
1819 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
1820 {
1821 	int *name = (int *)arg1;
1822 	u_int namelen = arg2;
1823 	struct proc *p;
1824 	struct sbuf sb;
1825 	int error, error2;
1826 
1827 	if (namelen != 1)
1828 		return (EINVAL);
1829 
1830 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1831 	if (error != 0)
1832 		return (error);
1833 	if ((p->p_flag & P_SYSTEM) != 0) {
1834 		PRELE(p);
1835 		return (0);
1836 	}
1837 
1838 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1839 	error = proc_getenvv(curthread, p, &sb);
1840 	error2 = sbuf_finish(&sb);
1841 	PRELE(p);
1842 	sbuf_delete(&sb);
1843 	return (error != 0 ? error : error2);
1844 }
1845 
1846 /*
1847  * This sysctl allows a process to retrieve ELF auxiliary vector of
1848  * another process.
1849  */
1850 static int
1851 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
1852 {
1853 	int *name = (int *)arg1;
1854 	u_int namelen = arg2;
1855 	struct proc *p;
1856 	size_t vsize, size;
1857 	char **auxv;
1858 	int error;
1859 
1860 	if (namelen != 1)
1861 		return (EINVAL);
1862 
1863 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1864 	if (error != 0)
1865 		return (error);
1866 	if ((p->p_flag & P_SYSTEM) != 0) {
1867 		PRELE(p);
1868 		return (0);
1869 	}
1870 	error = get_proc_vector(curthread, p, &auxv, &vsize, PROC_AUX);
1871 	if (error == 0) {
1872 #ifdef COMPAT_FREEBSD32
1873 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1874 			size = vsize * sizeof(Elf32_Auxinfo);
1875 		else
1876 #endif
1877 		size = vsize * sizeof(Elf_Auxinfo);
1878 		PRELE(p);
1879 		error = SYSCTL_OUT(req, auxv, size);
1880 		free(auxv, M_TEMP);
1881 	} else {
1882 		PRELE(p);
1883 	}
1884 	return (error);
1885 }
1886 
1887 /*
1888  * This sysctl allows a process to retrieve the path of the executable for
1889  * itself or another process.
1890  */
1891 static int
1892 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1893 {
1894 	pid_t *pidp = (pid_t *)arg1;
1895 	unsigned int arglen = arg2;
1896 	struct proc *p;
1897 	struct vnode *vp;
1898 	char *retbuf, *freebuf;
1899 	int error;
1900 
1901 	if (arglen != 1)
1902 		return (EINVAL);
1903 	if (*pidp == -1) {	/* -1 means this process */
1904 		p = req->td->td_proc;
1905 	} else {
1906 		error = pget(*pidp, PGET_CANSEE, &p);
1907 		if (error != 0)
1908 			return (error);
1909 	}
1910 
1911 	vp = p->p_textvp;
1912 	if (vp == NULL) {
1913 		if (*pidp != -1)
1914 			PROC_UNLOCK(p);
1915 		return (0);
1916 	}
1917 	vref(vp);
1918 	if (*pidp != -1)
1919 		PROC_UNLOCK(p);
1920 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1921 	vrele(vp);
1922 	if (error)
1923 		return (error);
1924 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1925 	free(freebuf, M_TEMP);
1926 	return (error);
1927 }
1928 
1929 static int
1930 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1931 {
1932 	struct proc *p;
1933 	char *sv_name;
1934 	int *name;
1935 	int namelen;
1936 	int error;
1937 
1938 	namelen = arg2;
1939 	if (namelen != 1)
1940 		return (EINVAL);
1941 
1942 	name = (int *)arg1;
1943 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
1944 	if (error != 0)
1945 		return (error);
1946 	sv_name = p->p_sysent->sv_name;
1947 	PROC_UNLOCK(p);
1948 	return (sysctl_handle_string(oidp, sv_name, 0, req));
1949 }
1950 
1951 #ifdef KINFO_OVMENTRY_SIZE
1952 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
1953 #endif
1954 
1955 #ifdef COMPAT_FREEBSD7
1956 static int
1957 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
1958 {
1959 	vm_map_entry_t entry, tmp_entry;
1960 	unsigned int last_timestamp;
1961 	char *fullpath, *freepath;
1962 	struct kinfo_ovmentry *kve;
1963 	struct vattr va;
1964 	struct ucred *cred;
1965 	int error, *name;
1966 	struct vnode *vp;
1967 	struct proc *p;
1968 	vm_map_t map;
1969 	struct vmspace *vm;
1970 
1971 	name = (int *)arg1;
1972 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1973 	if (error != 0)
1974 		return (error);
1975 	vm = vmspace_acquire_ref(p);
1976 	if (vm == NULL) {
1977 		PRELE(p);
1978 		return (ESRCH);
1979 	}
1980 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1981 
1982 	map = &vm->vm_map;
1983 	vm_map_lock_read(map);
1984 	for (entry = map->header.next; entry != &map->header;
1985 	    entry = entry->next) {
1986 		vm_object_t obj, tobj, lobj;
1987 		vm_offset_t addr;
1988 
1989 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
1990 			continue;
1991 
1992 		bzero(kve, sizeof(*kve));
1993 		kve->kve_structsize = sizeof(*kve);
1994 
1995 		kve->kve_private_resident = 0;
1996 		obj = entry->object.vm_object;
1997 		if (obj != NULL) {
1998 			VM_OBJECT_RLOCK(obj);
1999 			if (obj->shadow_count == 1)
2000 				kve->kve_private_resident =
2001 				    obj->resident_page_count;
2002 		}
2003 		kve->kve_resident = 0;
2004 		addr = entry->start;
2005 		while (addr < entry->end) {
2006 			if (pmap_extract(map->pmap, addr))
2007 				kve->kve_resident++;
2008 			addr += PAGE_SIZE;
2009 		}
2010 
2011 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2012 			if (tobj != obj)
2013 				VM_OBJECT_RLOCK(tobj);
2014 			if (lobj != obj)
2015 				VM_OBJECT_RUNLOCK(lobj);
2016 			lobj = tobj;
2017 		}
2018 
2019 		kve->kve_start = (void*)entry->start;
2020 		kve->kve_end = (void*)entry->end;
2021 		kve->kve_offset = (off_t)entry->offset;
2022 
2023 		if (entry->protection & VM_PROT_READ)
2024 			kve->kve_protection |= KVME_PROT_READ;
2025 		if (entry->protection & VM_PROT_WRITE)
2026 			kve->kve_protection |= KVME_PROT_WRITE;
2027 		if (entry->protection & VM_PROT_EXECUTE)
2028 			kve->kve_protection |= KVME_PROT_EXEC;
2029 
2030 		if (entry->eflags & MAP_ENTRY_COW)
2031 			kve->kve_flags |= KVME_FLAG_COW;
2032 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2033 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2034 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2035 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2036 
2037 		last_timestamp = map->timestamp;
2038 		vm_map_unlock_read(map);
2039 
2040 		kve->kve_fileid = 0;
2041 		kve->kve_fsid = 0;
2042 		freepath = NULL;
2043 		fullpath = "";
2044 		if (lobj) {
2045 			vp = NULL;
2046 			switch (lobj->type) {
2047 			case OBJT_DEFAULT:
2048 				kve->kve_type = KVME_TYPE_DEFAULT;
2049 				break;
2050 			case OBJT_VNODE:
2051 				kve->kve_type = KVME_TYPE_VNODE;
2052 				vp = lobj->handle;
2053 				vref(vp);
2054 				break;
2055 			case OBJT_SWAP:
2056 				kve->kve_type = KVME_TYPE_SWAP;
2057 				break;
2058 			case OBJT_DEVICE:
2059 				kve->kve_type = KVME_TYPE_DEVICE;
2060 				break;
2061 			case OBJT_PHYS:
2062 				kve->kve_type = KVME_TYPE_PHYS;
2063 				break;
2064 			case OBJT_DEAD:
2065 				kve->kve_type = KVME_TYPE_DEAD;
2066 				break;
2067 			case OBJT_SG:
2068 				kve->kve_type = KVME_TYPE_SG;
2069 				break;
2070 			default:
2071 				kve->kve_type = KVME_TYPE_UNKNOWN;
2072 				break;
2073 			}
2074 			if (lobj != obj)
2075 				VM_OBJECT_RUNLOCK(lobj);
2076 
2077 			kve->kve_ref_count = obj->ref_count;
2078 			kve->kve_shadow_count = obj->shadow_count;
2079 			VM_OBJECT_RUNLOCK(obj);
2080 			if (vp != NULL) {
2081 				vn_fullpath(curthread, vp, &fullpath,
2082 				    &freepath);
2083 				cred = curthread->td_ucred;
2084 				vn_lock(vp, LK_SHARED | LK_RETRY);
2085 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2086 					kve->kve_fileid = va.va_fileid;
2087 					kve->kve_fsid = va.va_fsid;
2088 				}
2089 				vput(vp);
2090 			}
2091 		} else {
2092 			kve->kve_type = KVME_TYPE_NONE;
2093 			kve->kve_ref_count = 0;
2094 			kve->kve_shadow_count = 0;
2095 		}
2096 
2097 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2098 		if (freepath != NULL)
2099 			free(freepath, M_TEMP);
2100 
2101 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2102 		vm_map_lock_read(map);
2103 		if (error)
2104 			break;
2105 		if (last_timestamp != map->timestamp) {
2106 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2107 			entry = tmp_entry;
2108 		}
2109 	}
2110 	vm_map_unlock_read(map);
2111 	vmspace_free(vm);
2112 	PRELE(p);
2113 	free(kve, M_TEMP);
2114 	return (error);
2115 }
2116 #endif	/* COMPAT_FREEBSD7 */
2117 
2118 #ifdef KINFO_VMENTRY_SIZE
2119 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2120 #endif
2121 
2122 static int
2123 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2124 {
2125 	vm_map_entry_t entry, tmp_entry;
2126 	unsigned int last_timestamp;
2127 	char *fullpath, *freepath;
2128 	struct kinfo_vmentry *kve;
2129 	struct vattr va;
2130 	struct ucred *cred;
2131 	int error, *name;
2132 	struct vnode *vp;
2133 	struct proc *p;
2134 	struct vmspace *vm;
2135 	vm_map_t map;
2136 
2137 	name = (int *)arg1;
2138 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2139 	if (error != 0)
2140 		return (error);
2141 	vm = vmspace_acquire_ref(p);
2142 	if (vm == NULL) {
2143 		PRELE(p);
2144 		return (ESRCH);
2145 	}
2146 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2147 
2148 	map = &vm->vm_map;
2149 	vm_map_lock_read(map);
2150 	for (entry = map->header.next; entry != &map->header;
2151 	    entry = entry->next) {
2152 		vm_object_t obj, tobj, lobj;
2153 		vm_offset_t addr;
2154 		vm_paddr_t locked_pa;
2155 		int mincoreinfo;
2156 
2157 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2158 			continue;
2159 
2160 		bzero(kve, sizeof(*kve));
2161 
2162 		kve->kve_private_resident = 0;
2163 		obj = entry->object.vm_object;
2164 		if (obj != NULL) {
2165 			VM_OBJECT_RLOCK(obj);
2166 			if (obj->shadow_count == 1)
2167 				kve->kve_private_resident =
2168 				    obj->resident_page_count;
2169 		}
2170 		kve->kve_resident = 0;
2171 		addr = entry->start;
2172 		while (addr < entry->end) {
2173 			locked_pa = 0;
2174 			mincoreinfo = pmap_mincore(map->pmap, addr, &locked_pa);
2175 			if (locked_pa != 0)
2176 				vm_page_unlock(PHYS_TO_VM_PAGE(locked_pa));
2177 			if (mincoreinfo & MINCORE_INCORE)
2178 				kve->kve_resident++;
2179 			if (mincoreinfo & MINCORE_SUPER)
2180 				kve->kve_flags |= KVME_FLAG_SUPER;
2181 			addr += PAGE_SIZE;
2182 		}
2183 
2184 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2185 			if (tobj != obj)
2186 				VM_OBJECT_RLOCK(tobj);
2187 			if (lobj != obj)
2188 				VM_OBJECT_RUNLOCK(lobj);
2189 			lobj = tobj;
2190 		}
2191 
2192 		kve->kve_start = entry->start;
2193 		kve->kve_end = entry->end;
2194 		kve->kve_offset = entry->offset;
2195 
2196 		if (entry->protection & VM_PROT_READ)
2197 			kve->kve_protection |= KVME_PROT_READ;
2198 		if (entry->protection & VM_PROT_WRITE)
2199 			kve->kve_protection |= KVME_PROT_WRITE;
2200 		if (entry->protection & VM_PROT_EXECUTE)
2201 			kve->kve_protection |= KVME_PROT_EXEC;
2202 
2203 		if (entry->eflags & MAP_ENTRY_COW)
2204 			kve->kve_flags |= KVME_FLAG_COW;
2205 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2206 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2207 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2208 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2209 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2210 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2211 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2212 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2213 
2214 		last_timestamp = map->timestamp;
2215 		vm_map_unlock_read(map);
2216 
2217 		freepath = NULL;
2218 		fullpath = "";
2219 		if (lobj) {
2220 			vp = NULL;
2221 			switch (lobj->type) {
2222 			case OBJT_DEFAULT:
2223 				kve->kve_type = KVME_TYPE_DEFAULT;
2224 				break;
2225 			case OBJT_VNODE:
2226 				kve->kve_type = KVME_TYPE_VNODE;
2227 				vp = lobj->handle;
2228 				vref(vp);
2229 				break;
2230 			case OBJT_SWAP:
2231 				kve->kve_type = KVME_TYPE_SWAP;
2232 				break;
2233 			case OBJT_DEVICE:
2234 				kve->kve_type = KVME_TYPE_DEVICE;
2235 				break;
2236 			case OBJT_PHYS:
2237 				kve->kve_type = KVME_TYPE_PHYS;
2238 				break;
2239 			case OBJT_DEAD:
2240 				kve->kve_type = KVME_TYPE_DEAD;
2241 				break;
2242 			case OBJT_SG:
2243 				kve->kve_type = KVME_TYPE_SG;
2244 				break;
2245 			default:
2246 				kve->kve_type = KVME_TYPE_UNKNOWN;
2247 				break;
2248 			}
2249 			if (lobj != obj)
2250 				VM_OBJECT_RUNLOCK(lobj);
2251 
2252 			kve->kve_ref_count = obj->ref_count;
2253 			kve->kve_shadow_count = obj->shadow_count;
2254 			VM_OBJECT_RUNLOCK(obj);
2255 			if (vp != NULL) {
2256 				vn_fullpath(curthread, vp, &fullpath,
2257 				    &freepath);
2258 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2259 				cred = curthread->td_ucred;
2260 				vn_lock(vp, LK_SHARED | LK_RETRY);
2261 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2262 					kve->kve_vn_fileid = va.va_fileid;
2263 					kve->kve_vn_fsid = va.va_fsid;
2264 					kve->kve_vn_mode =
2265 					    MAKEIMODE(va.va_type, va.va_mode);
2266 					kve->kve_vn_size = va.va_size;
2267 					kve->kve_vn_rdev = va.va_rdev;
2268 					kve->kve_status = KF_ATTR_VALID;
2269 				}
2270 				vput(vp);
2271 			}
2272 		} else {
2273 			kve->kve_type = KVME_TYPE_NONE;
2274 			kve->kve_ref_count = 0;
2275 			kve->kve_shadow_count = 0;
2276 		}
2277 
2278 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2279 		if (freepath != NULL)
2280 			free(freepath, M_TEMP);
2281 
2282 		/* Pack record size down */
2283 		kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) +
2284 		    strlen(kve->kve_path) + 1;
2285 		kve->kve_structsize = roundup(kve->kve_structsize,
2286 		    sizeof(uint64_t));
2287 		error = SYSCTL_OUT(req, kve, kve->kve_structsize);
2288 		vm_map_lock_read(map);
2289 		if (error)
2290 			break;
2291 		if (last_timestamp != map->timestamp) {
2292 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2293 			entry = tmp_entry;
2294 		}
2295 	}
2296 	vm_map_unlock_read(map);
2297 	vmspace_free(vm);
2298 	PRELE(p);
2299 	free(kve, M_TEMP);
2300 	return (error);
2301 }
2302 
2303 #if defined(STACK) || defined(DDB)
2304 static int
2305 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2306 {
2307 	struct kinfo_kstack *kkstp;
2308 	int error, i, *name, numthreads;
2309 	lwpid_t *lwpidarray;
2310 	struct thread *td;
2311 	struct stack *st;
2312 	struct sbuf sb;
2313 	struct proc *p;
2314 
2315 	name = (int *)arg1;
2316 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2317 	if (error != 0)
2318 		return (error);
2319 
2320 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2321 	st = stack_create();
2322 
2323 	lwpidarray = NULL;
2324 	numthreads = 0;
2325 	PROC_LOCK(p);
2326 repeat:
2327 	if (numthreads < p->p_numthreads) {
2328 		if (lwpidarray != NULL) {
2329 			free(lwpidarray, M_TEMP);
2330 			lwpidarray = NULL;
2331 		}
2332 		numthreads = p->p_numthreads;
2333 		PROC_UNLOCK(p);
2334 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2335 		    M_WAITOK | M_ZERO);
2336 		PROC_LOCK(p);
2337 		goto repeat;
2338 	}
2339 	i = 0;
2340 
2341 	/*
2342 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2343 	 * of changes could have taken place.  Should we check to see if the
2344 	 * vmspace has been replaced, or the like, in order to prevent
2345 	 * giving a snapshot that spans, say, execve(2), with some threads
2346 	 * before and some after?  Among other things, the credentials could
2347 	 * have changed, in which case the right to extract debug info might
2348 	 * no longer be assured.
2349 	 */
2350 	FOREACH_THREAD_IN_PROC(p, td) {
2351 		KASSERT(i < numthreads,
2352 		    ("sysctl_kern_proc_kstack: numthreads"));
2353 		lwpidarray[i] = td->td_tid;
2354 		i++;
2355 	}
2356 	numthreads = i;
2357 	for (i = 0; i < numthreads; i++) {
2358 		td = thread_find(p, lwpidarray[i]);
2359 		if (td == NULL) {
2360 			continue;
2361 		}
2362 		bzero(kkstp, sizeof(*kkstp));
2363 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2364 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2365 		thread_lock(td);
2366 		kkstp->kkst_tid = td->td_tid;
2367 		if (TD_IS_SWAPPED(td))
2368 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2369 		else if (TD_IS_RUNNING(td))
2370 			kkstp->kkst_state = KKST_STATE_RUNNING;
2371 		else {
2372 			kkstp->kkst_state = KKST_STATE_STACKOK;
2373 			stack_save_td(st, td);
2374 		}
2375 		thread_unlock(td);
2376 		PROC_UNLOCK(p);
2377 		stack_sbuf_print(&sb, st);
2378 		sbuf_finish(&sb);
2379 		sbuf_delete(&sb);
2380 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2381 		PROC_LOCK(p);
2382 		if (error)
2383 			break;
2384 	}
2385 	_PRELE(p);
2386 	PROC_UNLOCK(p);
2387 	if (lwpidarray != NULL)
2388 		free(lwpidarray, M_TEMP);
2389 	stack_destroy(st);
2390 	free(kkstp, M_TEMP);
2391 	return (error);
2392 }
2393 #endif
2394 
2395 /*
2396  * This sysctl allows a process to retrieve the full list of groups from
2397  * itself or another process.
2398  */
2399 static int
2400 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2401 {
2402 	pid_t *pidp = (pid_t *)arg1;
2403 	unsigned int arglen = arg2;
2404 	struct proc *p;
2405 	struct ucred *cred;
2406 	int error;
2407 
2408 	if (arglen != 1)
2409 		return (EINVAL);
2410 	if (*pidp == -1) {	/* -1 means this process */
2411 		p = req->td->td_proc;
2412 	} else {
2413 		error = pget(*pidp, PGET_CANSEE, &p);
2414 		if (error != 0)
2415 			return (error);
2416 	}
2417 
2418 	cred = crhold(p->p_ucred);
2419 	if (*pidp != -1)
2420 		PROC_UNLOCK(p);
2421 
2422 	error = SYSCTL_OUT(req, cred->cr_groups,
2423 	    cred->cr_ngroups * sizeof(gid_t));
2424 	crfree(cred);
2425 	return (error);
2426 }
2427 
2428 /*
2429  * This sysctl allows a process to retrieve or/and set the resource limit for
2430  * another process.
2431  */
2432 static int
2433 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2434 {
2435 	int *name = (int *)arg1;
2436 	u_int namelen = arg2;
2437 	struct rlimit rlim;
2438 	struct proc *p;
2439 	u_int which;
2440 	int flags, error;
2441 
2442 	if (namelen != 2)
2443 		return (EINVAL);
2444 
2445 	which = (u_int)name[1];
2446 	if (which >= RLIM_NLIMITS)
2447 		return (EINVAL);
2448 
2449 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2450 		return (EINVAL);
2451 
2452 	flags = PGET_HOLD | PGET_NOTWEXIT;
2453 	if (req->newptr != NULL)
2454 		flags |= PGET_CANDEBUG;
2455 	else
2456 		flags |= PGET_CANSEE;
2457 	error = pget((pid_t)name[0], flags, &p);
2458 	if (error != 0)
2459 		return (error);
2460 
2461 	/*
2462 	 * Retrieve limit.
2463 	 */
2464 	if (req->oldptr != NULL) {
2465 		PROC_LOCK(p);
2466 		lim_rlimit(p, which, &rlim);
2467 		PROC_UNLOCK(p);
2468 	}
2469 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2470 	if (error != 0)
2471 		goto errout;
2472 
2473 	/*
2474 	 * Set limit.
2475 	 */
2476 	if (req->newptr != NULL) {
2477 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2478 		if (error == 0)
2479 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2480 	}
2481 
2482 errout:
2483 	PRELE(p);
2484 	return (error);
2485 }
2486 
2487 /*
2488  * This sysctl allows a process to retrieve ps_strings structure location of
2489  * another process.
2490  */
2491 static int
2492 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2493 {
2494 	int *name = (int *)arg1;
2495 	u_int namelen = arg2;
2496 	struct proc *p;
2497 	vm_offset_t ps_strings;
2498 	int error;
2499 #ifdef COMPAT_FREEBSD32
2500 	uint32_t ps_strings32;
2501 #endif
2502 
2503 	if (namelen != 1)
2504 		return (EINVAL);
2505 
2506 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2507 	if (error != 0)
2508 		return (error);
2509 #ifdef COMPAT_FREEBSD32
2510 	if ((req->flags & SCTL_MASK32) != 0) {
2511 		/*
2512 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2513 		 * process.
2514 		 */
2515 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2516 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2517 		PROC_UNLOCK(p);
2518 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2519 		return (error);
2520 	}
2521 #endif
2522 	ps_strings = p->p_sysent->sv_psstrings;
2523 	PROC_UNLOCK(p);
2524 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2525 	return (error);
2526 }
2527 
2528 /*
2529  * This sysctl allows a process to retrieve umask of another process.
2530  */
2531 static int
2532 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2533 {
2534 	int *name = (int *)arg1;
2535 	u_int namelen = arg2;
2536 	struct proc *p;
2537 	int error;
2538 	u_short fd_cmask;
2539 
2540 	if (namelen != 1)
2541 		return (EINVAL);
2542 
2543 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2544 	if (error != 0)
2545 		return (error);
2546 
2547 	FILEDESC_SLOCK(p->p_fd);
2548 	fd_cmask = p->p_fd->fd_cmask;
2549 	FILEDESC_SUNLOCK(p->p_fd);
2550 	PRELE(p);
2551 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2552 	return (error);
2553 }
2554 
2555 /*
2556  * This sysctl allows a process to set and retrieve binary osreldate of
2557  * another process.
2558  */
2559 static int
2560 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2561 {
2562 	int *name = (int *)arg1;
2563 	u_int namelen = arg2;
2564 	struct proc *p;
2565 	int flags, error, osrel;
2566 
2567 	if (namelen != 1)
2568 		return (EINVAL);
2569 
2570 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2571 		return (EINVAL);
2572 
2573 	flags = PGET_HOLD | PGET_NOTWEXIT;
2574 	if (req->newptr != NULL)
2575 		flags |= PGET_CANDEBUG;
2576 	else
2577 		flags |= PGET_CANSEE;
2578 	error = pget((pid_t)name[0], flags, &p);
2579 	if (error != 0)
2580 		return (error);
2581 
2582 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2583 	if (error != 0)
2584 		goto errout;
2585 
2586 	if (req->newptr != NULL) {
2587 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2588 		if (error != 0)
2589 			goto errout;
2590 		if (osrel < 0) {
2591 			error = EINVAL;
2592 			goto errout;
2593 		}
2594 		p->p_osrel = osrel;
2595 	}
2596 errout:
2597 	PRELE(p);
2598 	return (error);
2599 }
2600 
2601 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2602 
2603 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2604 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2605 	"Return entire process table");
2606 
2607 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2608 	sysctl_kern_proc, "Process table");
2609 
2610 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2611 	sysctl_kern_proc, "Process table");
2612 
2613 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2614 	sysctl_kern_proc, "Process table");
2615 
2616 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2617 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2618 
2619 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2620 	sysctl_kern_proc, "Process table");
2621 
2622 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2623 	sysctl_kern_proc, "Process table");
2624 
2625 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2626 	sysctl_kern_proc, "Process table");
2627 
2628 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2629 	sysctl_kern_proc, "Process table");
2630 
2631 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2632 	sysctl_kern_proc, "Return process table, no threads");
2633 
2634 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2635 	CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2636 	sysctl_kern_proc_args, "Process argument list");
2637 
2638 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
2639 	sysctl_kern_proc_env, "Process environment");
2640 
2641 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
2642 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
2643 
2644 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2645 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2646 
2647 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2648 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2649 	"Process syscall vector name (ABI type)");
2650 
2651 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2652 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2653 
2654 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2655 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2656 
2657 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2658 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2659 
2660 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2661 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2662 
2663 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2664 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2665 
2666 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2667 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2668 
2669 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2670 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2671 
2672 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2673 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2674 
2675 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2676 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2677 	"Return process table, no threads");
2678 
2679 #ifdef COMPAT_FREEBSD7
2680 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2681 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2682 #endif
2683 
2684 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2685 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2686 
2687 #if defined(STACK) || defined(DDB)
2688 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2689 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2690 #endif
2691 
2692 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2693 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
2694 
2695 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
2696 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
2697 	"Process resource limits");
2698 
2699 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
2700 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
2701 	"Process ps_strings location");
2702 
2703 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
2704 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
2705 
2706 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
2707 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
2708 	"Process binary osreldate");
2709