xref: /freebsd/sys/kern/kern_proc.c (revision 6e908c6abc9b5cf9a2fef1409f51716d8b7dc978)
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 	else if ((flags & PGET_NOTID) == 0)
381 		p = pfind_tid_locked(pid);
382 	else
383 		p = NULL;
384 	if (p == NULL && (flags & PGET_NOTWEXIT) == 0)
385 		p = zpfind_locked(pid);
386 	sx_sunlock(&allproc_lock);
387 	if (p == NULL)
388 		return (ESRCH);
389 	if ((flags & PGET_CANSEE) != 0) {
390 		error = p_cansee(curthread, p);
391 		if (error != 0)
392 			goto errout;
393 	}
394 	if ((flags & PGET_CANDEBUG) != 0) {
395 		error = p_candebug(curthread, p);
396 		if (error != 0)
397 			goto errout;
398 	}
399 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
400 		error = EPERM;
401 		goto errout;
402 	}
403 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
404 		error = ESRCH;
405 		goto errout;
406 	}
407 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
408 		/*
409 		 * XXXRW: Not clear ESRCH is the right error during proc
410 		 * execve().
411 		 */
412 		error = ESRCH;
413 		goto errout;
414 	}
415 	if ((flags & PGET_HOLD) != 0) {
416 		_PHOLD(p);
417 		PROC_UNLOCK(p);
418 	}
419 	*pp = p;
420 	return (0);
421 errout:
422 	PROC_UNLOCK(p);
423 	return (error);
424 }
425 
426 /*
427  * Create a new process group.
428  * pgid must be equal to the pid of p.
429  * Begin a new session if required.
430  */
431 int
432 enterpgrp(p, pgid, pgrp, sess)
433 	register struct proc *p;
434 	pid_t pgid;
435 	struct pgrp *pgrp;
436 	struct session *sess;
437 {
438 
439 	sx_assert(&proctree_lock, SX_XLOCKED);
440 
441 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
442 	KASSERT(p->p_pid == pgid,
443 	    ("enterpgrp: new pgrp and pid != pgid"));
444 	KASSERT(pgfind(pgid) == NULL,
445 	    ("enterpgrp: pgrp with pgid exists"));
446 	KASSERT(!SESS_LEADER(p),
447 	    ("enterpgrp: session leader attempted setpgrp"));
448 
449 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
450 
451 	if (sess != NULL) {
452 		/*
453 		 * new session
454 		 */
455 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
456 		PROC_LOCK(p);
457 		p->p_flag &= ~P_CONTROLT;
458 		PROC_UNLOCK(p);
459 		PGRP_LOCK(pgrp);
460 		sess->s_leader = p;
461 		sess->s_sid = p->p_pid;
462 		refcount_init(&sess->s_count, 1);
463 		sess->s_ttyvp = NULL;
464 		sess->s_ttydp = NULL;
465 		sess->s_ttyp = NULL;
466 		bcopy(p->p_session->s_login, sess->s_login,
467 			    sizeof(sess->s_login));
468 		pgrp->pg_session = sess;
469 		KASSERT(p == curproc,
470 		    ("enterpgrp: mksession and p != curproc"));
471 	} else {
472 		pgrp->pg_session = p->p_session;
473 		sess_hold(pgrp->pg_session);
474 		PGRP_LOCK(pgrp);
475 	}
476 	pgrp->pg_id = pgid;
477 	LIST_INIT(&pgrp->pg_members);
478 
479 	/*
480 	 * As we have an exclusive lock of proctree_lock,
481 	 * this should not deadlock.
482 	 */
483 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
484 	pgrp->pg_jobc = 0;
485 	SLIST_INIT(&pgrp->pg_sigiolst);
486 	PGRP_UNLOCK(pgrp);
487 
488 	doenterpgrp(p, pgrp);
489 
490 	return (0);
491 }
492 
493 /*
494  * Move p to an existing process group
495  */
496 int
497 enterthispgrp(p, pgrp)
498 	register struct proc *p;
499 	struct pgrp *pgrp;
500 {
501 
502 	sx_assert(&proctree_lock, SX_XLOCKED);
503 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
504 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
505 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
506 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
507 	KASSERT(pgrp->pg_session == p->p_session,
508 		("%s: pgrp's session %p, p->p_session %p.\n",
509 		__func__,
510 		pgrp->pg_session,
511 		p->p_session));
512 	KASSERT(pgrp != p->p_pgrp,
513 		("%s: p belongs to pgrp.", __func__));
514 
515 	doenterpgrp(p, pgrp);
516 
517 	return (0);
518 }
519 
520 /*
521  * Move p to a process group
522  */
523 static void
524 doenterpgrp(p, pgrp)
525 	struct proc *p;
526 	struct pgrp *pgrp;
527 {
528 	struct pgrp *savepgrp;
529 
530 	sx_assert(&proctree_lock, SX_XLOCKED);
531 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
532 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
533 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
534 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
535 
536 	savepgrp = p->p_pgrp;
537 
538 	/*
539 	 * Adjust eligibility of affected pgrps to participate in job control.
540 	 * Increment eligibility counts before decrementing, otherwise we
541 	 * could reach 0 spuriously during the first call.
542 	 */
543 	fixjobc(p, pgrp, 1);
544 	fixjobc(p, p->p_pgrp, 0);
545 
546 	PGRP_LOCK(pgrp);
547 	PGRP_LOCK(savepgrp);
548 	PROC_LOCK(p);
549 	LIST_REMOVE(p, p_pglist);
550 	p->p_pgrp = pgrp;
551 	PROC_UNLOCK(p);
552 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
553 	PGRP_UNLOCK(savepgrp);
554 	PGRP_UNLOCK(pgrp);
555 	if (LIST_EMPTY(&savepgrp->pg_members))
556 		pgdelete(savepgrp);
557 }
558 
559 /*
560  * remove process from process group
561  */
562 int
563 leavepgrp(p)
564 	register struct proc *p;
565 {
566 	struct pgrp *savepgrp;
567 
568 	sx_assert(&proctree_lock, SX_XLOCKED);
569 	savepgrp = p->p_pgrp;
570 	PGRP_LOCK(savepgrp);
571 	PROC_LOCK(p);
572 	LIST_REMOVE(p, p_pglist);
573 	p->p_pgrp = NULL;
574 	PROC_UNLOCK(p);
575 	PGRP_UNLOCK(savepgrp);
576 	if (LIST_EMPTY(&savepgrp->pg_members))
577 		pgdelete(savepgrp);
578 	return (0);
579 }
580 
581 /*
582  * delete a process group
583  */
584 static void
585 pgdelete(pgrp)
586 	register struct pgrp *pgrp;
587 {
588 	struct session *savesess;
589 	struct tty *tp;
590 
591 	sx_assert(&proctree_lock, SX_XLOCKED);
592 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
593 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
594 
595 	/*
596 	 * Reset any sigio structures pointing to us as a result of
597 	 * F_SETOWN with our pgid.
598 	 */
599 	funsetownlst(&pgrp->pg_sigiolst);
600 
601 	PGRP_LOCK(pgrp);
602 	tp = pgrp->pg_session->s_ttyp;
603 	LIST_REMOVE(pgrp, pg_hash);
604 	savesess = pgrp->pg_session;
605 	PGRP_UNLOCK(pgrp);
606 
607 	/* Remove the reference to the pgrp before deallocating it. */
608 	if (tp != NULL) {
609 		tty_lock(tp);
610 		tty_rel_pgrp(tp, pgrp);
611 	}
612 
613 	mtx_destroy(&pgrp->pg_mtx);
614 	free(pgrp, M_PGRP);
615 	sess_release(savesess);
616 }
617 
618 static void
619 pgadjustjobc(pgrp, entering)
620 	struct pgrp *pgrp;
621 	int entering;
622 {
623 
624 	PGRP_LOCK(pgrp);
625 	if (entering)
626 		pgrp->pg_jobc++;
627 	else {
628 		--pgrp->pg_jobc;
629 		if (pgrp->pg_jobc == 0)
630 			orphanpg(pgrp);
631 	}
632 	PGRP_UNLOCK(pgrp);
633 }
634 
635 /*
636  * Adjust pgrp jobc counters when specified process changes process group.
637  * We count the number of processes in each process group that "qualify"
638  * the group for terminal job control (those with a parent in a different
639  * process group of the same session).  If that count reaches zero, the
640  * process group becomes orphaned.  Check both the specified process'
641  * process group and that of its children.
642  * entering == 0 => p is leaving specified group.
643  * entering == 1 => p is entering specified group.
644  */
645 void
646 fixjobc(p, pgrp, entering)
647 	register struct proc *p;
648 	register struct pgrp *pgrp;
649 	int entering;
650 {
651 	register struct pgrp *hispgrp;
652 	register struct session *mysession;
653 
654 	sx_assert(&proctree_lock, SX_LOCKED);
655 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
656 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
657 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
658 
659 	/*
660 	 * Check p's parent to see whether p qualifies its own process
661 	 * group; if so, adjust count for p's process group.
662 	 */
663 	mysession = pgrp->pg_session;
664 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
665 	    hispgrp->pg_session == mysession)
666 		pgadjustjobc(pgrp, entering);
667 
668 	/*
669 	 * Check this process' children to see whether they qualify
670 	 * their process groups; if so, adjust counts for children's
671 	 * process groups.
672 	 */
673 	LIST_FOREACH(p, &p->p_children, p_sibling) {
674 		hispgrp = p->p_pgrp;
675 		if (hispgrp == pgrp ||
676 		    hispgrp->pg_session != mysession)
677 			continue;
678 		PROC_LOCK(p);
679 		if (p->p_state == PRS_ZOMBIE) {
680 			PROC_UNLOCK(p);
681 			continue;
682 		}
683 		PROC_UNLOCK(p);
684 		pgadjustjobc(hispgrp, entering);
685 	}
686 }
687 
688 /*
689  * A process group has become orphaned;
690  * if there are any stopped processes in the group,
691  * hang-up all process in that group.
692  */
693 static void
694 orphanpg(pg)
695 	struct pgrp *pg;
696 {
697 	register struct proc *p;
698 
699 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
700 
701 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
702 		PROC_LOCK(p);
703 		if (P_SHOULDSTOP(p)) {
704 			PROC_UNLOCK(p);
705 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
706 				PROC_LOCK(p);
707 				kern_psignal(p, SIGHUP);
708 				kern_psignal(p, SIGCONT);
709 				PROC_UNLOCK(p);
710 			}
711 			return;
712 		}
713 		PROC_UNLOCK(p);
714 	}
715 }
716 
717 void
718 sess_hold(struct session *s)
719 {
720 
721 	refcount_acquire(&s->s_count);
722 }
723 
724 void
725 sess_release(struct session *s)
726 {
727 
728 	if (refcount_release(&s->s_count)) {
729 		if (s->s_ttyp != NULL) {
730 			tty_lock(s->s_ttyp);
731 			tty_rel_sess(s->s_ttyp, s);
732 		}
733 		mtx_destroy(&s->s_mtx);
734 		free(s, M_SESSION);
735 	}
736 }
737 
738 #ifdef DDB
739 
740 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
741 {
742 	register struct pgrp *pgrp;
743 	register struct proc *p;
744 	register int i;
745 
746 	for (i = 0; i <= pgrphash; i++) {
747 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
748 			printf("\tindx %d\n", i);
749 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
750 				printf(
751 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
752 				    (void *)pgrp, (long)pgrp->pg_id,
753 				    (void *)pgrp->pg_session,
754 				    pgrp->pg_session->s_count,
755 				    (void *)LIST_FIRST(&pgrp->pg_members));
756 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
757 					printf("\t\tpid %ld addr %p pgrp %p\n",
758 					    (long)p->p_pid, (void *)p,
759 					    (void *)p->p_pgrp);
760 				}
761 			}
762 		}
763 	}
764 }
765 #endif /* DDB */
766 
767 /*
768  * Calculate the kinfo_proc members which contain process-wide
769  * informations.
770  * Must be called with the target process locked.
771  */
772 static void
773 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
774 {
775 	struct thread *td;
776 
777 	PROC_LOCK_ASSERT(p, MA_OWNED);
778 
779 	kp->ki_estcpu = 0;
780 	kp->ki_pctcpu = 0;
781 	FOREACH_THREAD_IN_PROC(p, td) {
782 		thread_lock(td);
783 		kp->ki_pctcpu += sched_pctcpu(td);
784 		kp->ki_estcpu += td->td_estcpu;
785 		thread_unlock(td);
786 	}
787 }
788 
789 /*
790  * Clear kinfo_proc and fill in any information that is common
791  * to all threads in the process.
792  * Must be called with the target process locked.
793  */
794 static void
795 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
796 {
797 	struct thread *td0;
798 	struct tty *tp;
799 	struct session *sp;
800 	struct ucred *cred;
801 	struct sigacts *ps;
802 
803 	PROC_LOCK_ASSERT(p, MA_OWNED);
804 	bzero(kp, sizeof(*kp));
805 
806 	kp->ki_structsize = sizeof(*kp);
807 	kp->ki_paddr = p;
808 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
809 	kp->ki_args = p->p_args;
810 	kp->ki_textvp = p->p_textvp;
811 #ifdef KTRACE
812 	kp->ki_tracep = p->p_tracevp;
813 	kp->ki_traceflag = p->p_traceflag;
814 #endif
815 	kp->ki_fd = p->p_fd;
816 	kp->ki_vmspace = p->p_vmspace;
817 	kp->ki_flag = p->p_flag;
818 	cred = p->p_ucred;
819 	if (cred) {
820 		kp->ki_uid = cred->cr_uid;
821 		kp->ki_ruid = cred->cr_ruid;
822 		kp->ki_svuid = cred->cr_svuid;
823 		kp->ki_cr_flags = 0;
824 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
825 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
826 		/* XXX bde doesn't like KI_NGROUPS */
827 		if (cred->cr_ngroups > KI_NGROUPS) {
828 			kp->ki_ngroups = KI_NGROUPS;
829 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
830 		} else
831 			kp->ki_ngroups = cred->cr_ngroups;
832 		bcopy(cred->cr_groups, kp->ki_groups,
833 		    kp->ki_ngroups * sizeof(gid_t));
834 		kp->ki_rgid = cred->cr_rgid;
835 		kp->ki_svgid = cred->cr_svgid;
836 		/* If jailed(cred), emulate the old P_JAILED flag. */
837 		if (jailed(cred)) {
838 			kp->ki_flag |= P_JAILED;
839 			/* If inside the jail, use 0 as a jail ID. */
840 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
841 				kp->ki_jid = cred->cr_prison->pr_id;
842 		}
843 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
844 		    sizeof(kp->ki_loginclass));
845 	}
846 	ps = p->p_sigacts;
847 	if (ps) {
848 		mtx_lock(&ps->ps_mtx);
849 		kp->ki_sigignore = ps->ps_sigignore;
850 		kp->ki_sigcatch = ps->ps_sigcatch;
851 		mtx_unlock(&ps->ps_mtx);
852 	}
853 	if (p->p_state != PRS_NEW &&
854 	    p->p_state != PRS_ZOMBIE &&
855 	    p->p_vmspace != NULL) {
856 		struct vmspace *vm = p->p_vmspace;
857 
858 		kp->ki_size = vm->vm_map.size;
859 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
860 		FOREACH_THREAD_IN_PROC(p, td0) {
861 			if (!TD_IS_SWAPPED(td0))
862 				kp->ki_rssize += td0->td_kstack_pages;
863 		}
864 		kp->ki_swrss = vm->vm_swrss;
865 		kp->ki_tsize = vm->vm_tsize;
866 		kp->ki_dsize = vm->vm_dsize;
867 		kp->ki_ssize = vm->vm_ssize;
868 	} else if (p->p_state == PRS_ZOMBIE)
869 		kp->ki_stat = SZOMB;
870 	if (kp->ki_flag & P_INMEM)
871 		kp->ki_sflag = PS_INMEM;
872 	else
873 		kp->ki_sflag = 0;
874 	/* Calculate legacy swtime as seconds since 'swtick'. */
875 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
876 	kp->ki_pid = p->p_pid;
877 	kp->ki_nice = p->p_nice;
878 	kp->ki_start = p->p_stats->p_start;
879 	timevaladd(&kp->ki_start, &boottime);
880 	PROC_SLOCK(p);
881 	rufetch(p, &kp->ki_rusage);
882 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
883 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
884 	PROC_SUNLOCK(p);
885 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
886 	/* Some callers want child times in a single value. */
887 	kp->ki_childtime = kp->ki_childstime;
888 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
889 
890 	FOREACH_THREAD_IN_PROC(p, td0)
891 		kp->ki_cow += td0->td_cow;
892 
893 	tp = NULL;
894 	if (p->p_pgrp) {
895 		kp->ki_pgid = p->p_pgrp->pg_id;
896 		kp->ki_jobc = p->p_pgrp->pg_jobc;
897 		sp = p->p_pgrp->pg_session;
898 
899 		if (sp != NULL) {
900 			kp->ki_sid = sp->s_sid;
901 			SESS_LOCK(sp);
902 			strlcpy(kp->ki_login, sp->s_login,
903 			    sizeof(kp->ki_login));
904 			if (sp->s_ttyvp)
905 				kp->ki_kiflag |= KI_CTTY;
906 			if (SESS_LEADER(p))
907 				kp->ki_kiflag |= KI_SLEADER;
908 			/* XXX proctree_lock */
909 			tp = sp->s_ttyp;
910 			SESS_UNLOCK(sp);
911 		}
912 	}
913 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
914 		kp->ki_tdev = tty_udev(tp);
915 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
916 		if (tp->t_session)
917 			kp->ki_tsid = tp->t_session->s_sid;
918 	} else
919 		kp->ki_tdev = NODEV;
920 	if (p->p_comm[0] != '\0')
921 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
922 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
923 	    p->p_sysent->sv_name[0] != '\0')
924 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
925 	kp->ki_siglist = p->p_siglist;
926 	kp->ki_xstat = p->p_xstat;
927 	kp->ki_acflag = p->p_acflag;
928 	kp->ki_lock = p->p_lock;
929 	if (p->p_pptr)
930 		kp->ki_ppid = p->p_pptr->p_pid;
931 }
932 
933 /*
934  * Fill in information that is thread specific.  Must be called with
935  * target process locked.  If 'preferthread' is set, overwrite certain
936  * process-related fields that are maintained for both threads and
937  * processes.
938  */
939 static void
940 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
941 {
942 	struct proc *p;
943 
944 	p = td->td_proc;
945 	kp->ki_tdaddr = td;
946 	PROC_LOCK_ASSERT(p, MA_OWNED);
947 
948 	if (preferthread)
949 		PROC_SLOCK(p);
950 	thread_lock(td);
951 	if (td->td_wmesg != NULL)
952 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
953 	else
954 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
955 	strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname));
956 	if (TD_ON_LOCK(td)) {
957 		kp->ki_kiflag |= KI_LOCKBLOCK;
958 		strlcpy(kp->ki_lockname, td->td_lockname,
959 		    sizeof(kp->ki_lockname));
960 	} else {
961 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
962 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
963 	}
964 
965 	if (p->p_state == PRS_NORMAL) { /* approximate. */
966 		if (TD_ON_RUNQ(td) ||
967 		    TD_CAN_RUN(td) ||
968 		    TD_IS_RUNNING(td)) {
969 			kp->ki_stat = SRUN;
970 		} else if (P_SHOULDSTOP(p)) {
971 			kp->ki_stat = SSTOP;
972 		} else if (TD_IS_SLEEPING(td)) {
973 			kp->ki_stat = SSLEEP;
974 		} else if (TD_ON_LOCK(td)) {
975 			kp->ki_stat = SLOCK;
976 		} else {
977 			kp->ki_stat = SWAIT;
978 		}
979 	} else if (p->p_state == PRS_ZOMBIE) {
980 		kp->ki_stat = SZOMB;
981 	} else {
982 		kp->ki_stat = SIDL;
983 	}
984 
985 	/* Things in the thread */
986 	kp->ki_wchan = td->td_wchan;
987 	kp->ki_pri.pri_level = td->td_priority;
988 	kp->ki_pri.pri_native = td->td_base_pri;
989 	kp->ki_lastcpu = td->td_lastcpu;
990 	kp->ki_oncpu = td->td_oncpu;
991 	kp->ki_tdflags = td->td_flags;
992 	kp->ki_tid = td->td_tid;
993 	kp->ki_numthreads = p->p_numthreads;
994 	kp->ki_pcb = td->td_pcb;
995 	kp->ki_kstack = (void *)td->td_kstack;
996 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
997 	kp->ki_pri.pri_class = td->td_pri_class;
998 	kp->ki_pri.pri_user = td->td_user_pri;
999 
1000 	if (preferthread) {
1001 		rufetchtd(td, &kp->ki_rusage);
1002 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1003 		kp->ki_pctcpu = sched_pctcpu(td);
1004 		kp->ki_estcpu = td->td_estcpu;
1005 		kp->ki_cow = td->td_cow;
1006 	}
1007 
1008 	/* We can't get this anymore but ps etc never used it anyway. */
1009 	kp->ki_rqindex = 0;
1010 
1011 	if (preferthread)
1012 		kp->ki_siglist = td->td_siglist;
1013 	kp->ki_sigmask = td->td_sigmask;
1014 	thread_unlock(td);
1015 	if (preferthread)
1016 		PROC_SUNLOCK(p);
1017 }
1018 
1019 /*
1020  * Fill in a kinfo_proc structure for the specified process.
1021  * Must be called with the target process locked.
1022  */
1023 void
1024 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1025 {
1026 
1027 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1028 
1029 	fill_kinfo_proc_only(p, kp);
1030 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1031 	fill_kinfo_aggregate(p, kp);
1032 }
1033 
1034 struct pstats *
1035 pstats_alloc(void)
1036 {
1037 
1038 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1039 }
1040 
1041 /*
1042  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1043  */
1044 void
1045 pstats_fork(struct pstats *src, struct pstats *dst)
1046 {
1047 
1048 	bzero(&dst->pstat_startzero,
1049 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1050 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1051 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1052 }
1053 
1054 void
1055 pstats_free(struct pstats *ps)
1056 {
1057 
1058 	free(ps, M_SUBPROC);
1059 }
1060 
1061 static struct proc *
1062 zpfind_locked(pid_t pid)
1063 {
1064 	struct proc *p;
1065 
1066 	sx_assert(&allproc_lock, SX_LOCKED);
1067 	LIST_FOREACH(p, &zombproc, p_list) {
1068 		if (p->p_pid == pid) {
1069 			PROC_LOCK(p);
1070 			break;
1071 		}
1072 	}
1073 	return (p);
1074 }
1075 
1076 /*
1077  * Locate a zombie process by number
1078  */
1079 struct proc *
1080 zpfind(pid_t pid)
1081 {
1082 	struct proc *p;
1083 
1084 	sx_slock(&allproc_lock);
1085 	p = zpfind_locked(pid);
1086 	sx_sunlock(&allproc_lock);
1087 	return (p);
1088 }
1089 
1090 #define KERN_PROC_ZOMBMASK	0x3
1091 #define KERN_PROC_NOTHREADS	0x4
1092 
1093 #ifdef COMPAT_FREEBSD32
1094 
1095 /*
1096  * This function is typically used to copy out the kernel address, so
1097  * it can be replaced by assignment of zero.
1098  */
1099 static inline uint32_t
1100 ptr32_trim(void *ptr)
1101 {
1102 	uintptr_t uptr;
1103 
1104 	uptr = (uintptr_t)ptr;
1105 	return ((uptr > UINT_MAX) ? 0 : uptr);
1106 }
1107 
1108 #define PTRTRIM_CP(src,dst,fld) \
1109 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1110 
1111 static void
1112 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1113 {
1114 	int i;
1115 
1116 	bzero(ki32, sizeof(struct kinfo_proc32));
1117 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1118 	CP(*ki, *ki32, ki_layout);
1119 	PTRTRIM_CP(*ki, *ki32, ki_args);
1120 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1121 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1122 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1123 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1124 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1125 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1126 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1127 	CP(*ki, *ki32, ki_pid);
1128 	CP(*ki, *ki32, ki_ppid);
1129 	CP(*ki, *ki32, ki_pgid);
1130 	CP(*ki, *ki32, ki_tpgid);
1131 	CP(*ki, *ki32, ki_sid);
1132 	CP(*ki, *ki32, ki_tsid);
1133 	CP(*ki, *ki32, ki_jobc);
1134 	CP(*ki, *ki32, ki_tdev);
1135 	CP(*ki, *ki32, ki_siglist);
1136 	CP(*ki, *ki32, ki_sigmask);
1137 	CP(*ki, *ki32, ki_sigignore);
1138 	CP(*ki, *ki32, ki_sigcatch);
1139 	CP(*ki, *ki32, ki_uid);
1140 	CP(*ki, *ki32, ki_ruid);
1141 	CP(*ki, *ki32, ki_svuid);
1142 	CP(*ki, *ki32, ki_rgid);
1143 	CP(*ki, *ki32, ki_svgid);
1144 	CP(*ki, *ki32, ki_ngroups);
1145 	for (i = 0; i < KI_NGROUPS; i++)
1146 		CP(*ki, *ki32, ki_groups[i]);
1147 	CP(*ki, *ki32, ki_size);
1148 	CP(*ki, *ki32, ki_rssize);
1149 	CP(*ki, *ki32, ki_swrss);
1150 	CP(*ki, *ki32, ki_tsize);
1151 	CP(*ki, *ki32, ki_dsize);
1152 	CP(*ki, *ki32, ki_ssize);
1153 	CP(*ki, *ki32, ki_xstat);
1154 	CP(*ki, *ki32, ki_acflag);
1155 	CP(*ki, *ki32, ki_pctcpu);
1156 	CP(*ki, *ki32, ki_estcpu);
1157 	CP(*ki, *ki32, ki_slptime);
1158 	CP(*ki, *ki32, ki_swtime);
1159 	CP(*ki, *ki32, ki_cow);
1160 	CP(*ki, *ki32, ki_runtime);
1161 	TV_CP(*ki, *ki32, ki_start);
1162 	TV_CP(*ki, *ki32, ki_childtime);
1163 	CP(*ki, *ki32, ki_flag);
1164 	CP(*ki, *ki32, ki_kiflag);
1165 	CP(*ki, *ki32, ki_traceflag);
1166 	CP(*ki, *ki32, ki_stat);
1167 	CP(*ki, *ki32, ki_nice);
1168 	CP(*ki, *ki32, ki_lock);
1169 	CP(*ki, *ki32, ki_rqindex);
1170 	CP(*ki, *ki32, ki_oncpu);
1171 	CP(*ki, *ki32, ki_lastcpu);
1172 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1173 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1174 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1175 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1176 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1177 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1178 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1179 	CP(*ki, *ki32, ki_cr_flags);
1180 	CP(*ki, *ki32, ki_jid);
1181 	CP(*ki, *ki32, ki_numthreads);
1182 	CP(*ki, *ki32, ki_tid);
1183 	CP(*ki, *ki32, ki_pri);
1184 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1185 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1186 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1187 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1188 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1189 	CP(*ki, *ki32, ki_sflag);
1190 	CP(*ki, *ki32, ki_tdflags);
1191 }
1192 
1193 static int
1194 sysctl_out_proc_copyout(struct kinfo_proc *ki, struct sysctl_req *req)
1195 {
1196 	struct kinfo_proc32 ki32;
1197 	int error;
1198 
1199 	if (req->flags & SCTL_MASK32) {
1200 		freebsd32_kinfo_proc_out(ki, &ki32);
1201 		error = SYSCTL_OUT(req, &ki32, sizeof(struct kinfo_proc32));
1202 	} else
1203 		error = SYSCTL_OUT(req, ki, sizeof(struct kinfo_proc));
1204 	return (error);
1205 }
1206 #else
1207 static int
1208 sysctl_out_proc_copyout(struct kinfo_proc *ki, struct sysctl_req *req)
1209 {
1210 
1211 	return (SYSCTL_OUT(req, ki, sizeof(struct kinfo_proc)));
1212 }
1213 #endif
1214 
1215 /*
1216  * Must be called with the process locked and will return with it unlocked.
1217  */
1218 static int
1219 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1220 {
1221 	struct thread *td;
1222 	struct kinfo_proc kinfo_proc;
1223 	int error = 0;
1224 	struct proc *np;
1225 	pid_t pid = p->p_pid;
1226 
1227 	PROC_LOCK_ASSERT(p, MA_OWNED);
1228 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1229 
1230 	fill_kinfo_proc(p, &kinfo_proc);
1231 	if (flags & KERN_PROC_NOTHREADS)
1232 		error = sysctl_out_proc_copyout(&kinfo_proc, req);
1233 	else {
1234 		FOREACH_THREAD_IN_PROC(p, td) {
1235 			fill_kinfo_thread(td, &kinfo_proc, 1);
1236 			error = sysctl_out_proc_copyout(&kinfo_proc, req);
1237 			if (error)
1238 				break;
1239 		}
1240 	}
1241 	PROC_UNLOCK(p);
1242 	if (error)
1243 		return (error);
1244 	if (flags & KERN_PROC_ZOMBMASK)
1245 		np = zpfind(pid);
1246 	else {
1247 		if (pid == 0)
1248 			return (0);
1249 		np = pfind(pid);
1250 	}
1251 	if (np == NULL)
1252 		return (ESRCH);
1253 	if (np != p) {
1254 		PROC_UNLOCK(np);
1255 		return (ESRCH);
1256 	}
1257 	PROC_UNLOCK(np);
1258 	return (0);
1259 }
1260 
1261 static int
1262 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1263 {
1264 	int *name = (int *)arg1;
1265 	u_int namelen = arg2;
1266 	struct proc *p;
1267 	int flags, doingzomb, oid_number;
1268 	int error = 0;
1269 
1270 	oid_number = oidp->oid_number;
1271 	if (oid_number != KERN_PROC_ALL &&
1272 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1273 		flags = KERN_PROC_NOTHREADS;
1274 	else {
1275 		flags = 0;
1276 		oid_number &= ~KERN_PROC_INC_THREAD;
1277 	}
1278 	if (oid_number == KERN_PROC_PID) {
1279 		if (namelen != 1)
1280 			return (EINVAL);
1281 		error = sysctl_wire_old_buffer(req, 0);
1282 		if (error)
1283 			return (error);
1284 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1285 		if (error != 0)
1286 			return (error);
1287 		error = sysctl_out_proc(p, req, flags);
1288 		return (error);
1289 	}
1290 
1291 	switch (oid_number) {
1292 	case KERN_PROC_ALL:
1293 		if (namelen != 0)
1294 			return (EINVAL);
1295 		break;
1296 	case KERN_PROC_PROC:
1297 		if (namelen != 0 && namelen != 1)
1298 			return (EINVAL);
1299 		break;
1300 	default:
1301 		if (namelen != 1)
1302 			return (EINVAL);
1303 		break;
1304 	}
1305 
1306 	if (!req->oldptr) {
1307 		/* overestimate by 5 procs */
1308 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1309 		if (error)
1310 			return (error);
1311 	}
1312 	error = sysctl_wire_old_buffer(req, 0);
1313 	if (error != 0)
1314 		return (error);
1315 	sx_slock(&allproc_lock);
1316 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1317 		if (!doingzomb)
1318 			p = LIST_FIRST(&allproc);
1319 		else
1320 			p = LIST_FIRST(&zombproc);
1321 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
1322 			/*
1323 			 * Skip embryonic processes.
1324 			 */
1325 			PROC_LOCK(p);
1326 			if (p->p_state == PRS_NEW) {
1327 				PROC_UNLOCK(p);
1328 				continue;
1329 			}
1330 			KASSERT(p->p_ucred != NULL,
1331 			    ("process credential is NULL for non-NEW proc"));
1332 			/*
1333 			 * Show a user only appropriate processes.
1334 			 */
1335 			if (p_cansee(curthread, p)) {
1336 				PROC_UNLOCK(p);
1337 				continue;
1338 			}
1339 			/*
1340 			 * TODO - make more efficient (see notes below).
1341 			 * do by session.
1342 			 */
1343 			switch (oid_number) {
1344 
1345 			case KERN_PROC_GID:
1346 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1347 					PROC_UNLOCK(p);
1348 					continue;
1349 				}
1350 				break;
1351 
1352 			case KERN_PROC_PGRP:
1353 				/* could do this by traversing pgrp */
1354 				if (p->p_pgrp == NULL ||
1355 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1356 					PROC_UNLOCK(p);
1357 					continue;
1358 				}
1359 				break;
1360 
1361 			case KERN_PROC_RGID:
1362 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1363 					PROC_UNLOCK(p);
1364 					continue;
1365 				}
1366 				break;
1367 
1368 			case KERN_PROC_SESSION:
1369 				if (p->p_session == NULL ||
1370 				    p->p_session->s_sid != (pid_t)name[0]) {
1371 					PROC_UNLOCK(p);
1372 					continue;
1373 				}
1374 				break;
1375 
1376 			case KERN_PROC_TTY:
1377 				if ((p->p_flag & P_CONTROLT) == 0 ||
1378 				    p->p_session == NULL) {
1379 					PROC_UNLOCK(p);
1380 					continue;
1381 				}
1382 				/* XXX proctree_lock */
1383 				SESS_LOCK(p->p_session);
1384 				if (p->p_session->s_ttyp == NULL ||
1385 				    tty_udev(p->p_session->s_ttyp) !=
1386 				    (dev_t)name[0]) {
1387 					SESS_UNLOCK(p->p_session);
1388 					PROC_UNLOCK(p);
1389 					continue;
1390 				}
1391 				SESS_UNLOCK(p->p_session);
1392 				break;
1393 
1394 			case KERN_PROC_UID:
1395 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1396 					PROC_UNLOCK(p);
1397 					continue;
1398 				}
1399 				break;
1400 
1401 			case KERN_PROC_RUID:
1402 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1403 					PROC_UNLOCK(p);
1404 					continue;
1405 				}
1406 				break;
1407 
1408 			case KERN_PROC_PROC:
1409 				break;
1410 
1411 			default:
1412 				break;
1413 
1414 			}
1415 
1416 			error = sysctl_out_proc(p, req, flags | doingzomb);
1417 			if (error) {
1418 				sx_sunlock(&allproc_lock);
1419 				return (error);
1420 			}
1421 		}
1422 	}
1423 	sx_sunlock(&allproc_lock);
1424 	return (0);
1425 }
1426 
1427 struct pargs *
1428 pargs_alloc(int len)
1429 {
1430 	struct pargs *pa;
1431 
1432 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1433 		M_WAITOK);
1434 	refcount_init(&pa->ar_ref, 1);
1435 	pa->ar_length = len;
1436 	return (pa);
1437 }
1438 
1439 static void
1440 pargs_free(struct pargs *pa)
1441 {
1442 
1443 	free(pa, M_PARGS);
1444 }
1445 
1446 void
1447 pargs_hold(struct pargs *pa)
1448 {
1449 
1450 	if (pa == NULL)
1451 		return;
1452 	refcount_acquire(&pa->ar_ref);
1453 }
1454 
1455 void
1456 pargs_drop(struct pargs *pa)
1457 {
1458 
1459 	if (pa == NULL)
1460 		return;
1461 	if (refcount_release(&pa->ar_ref))
1462 		pargs_free(pa);
1463 }
1464 
1465 static int
1466 proc_read_mem(struct thread *td, struct proc *p, vm_offset_t offset, void* buf,
1467     size_t len)
1468 {
1469 	struct iovec iov;
1470 	struct uio uio;
1471 
1472 	iov.iov_base = (caddr_t)buf;
1473 	iov.iov_len = len;
1474 	uio.uio_iov = &iov;
1475 	uio.uio_iovcnt = 1;
1476 	uio.uio_offset = offset;
1477 	uio.uio_resid = (ssize_t)len;
1478 	uio.uio_segflg = UIO_SYSSPACE;
1479 	uio.uio_rw = UIO_READ;
1480 	uio.uio_td = td;
1481 
1482 	return (proc_rwmem(p, &uio));
1483 }
1484 
1485 static int
1486 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1487     size_t len)
1488 {
1489 	size_t i;
1490 	int error;
1491 
1492 	error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, len);
1493 	/*
1494 	 * Reading the chunk may validly return EFAULT if the string is shorter
1495 	 * than the chunk and is aligned at the end of the page, assuming the
1496 	 * next page is not mapped.  So if EFAULT is returned do a fallback to
1497 	 * one byte read loop.
1498 	 */
1499 	if (error == EFAULT) {
1500 		for (i = 0; i < len; i++, buf++, sptr++) {
1501 			error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, 1);
1502 			if (error != 0)
1503 				return (error);
1504 			if (*buf == '\0')
1505 				break;
1506 		}
1507 		error = 0;
1508 	}
1509 	return (error);
1510 }
1511 
1512 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1513 
1514 enum proc_vector_type {
1515 	PROC_ARG,
1516 	PROC_ENV,
1517 	PROC_AUX,
1518 };
1519 
1520 #ifdef COMPAT_FREEBSD32
1521 static int
1522 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1523     size_t *vsizep, enum proc_vector_type type)
1524 {
1525 	struct freebsd32_ps_strings pss;
1526 	Elf32_Auxinfo aux;
1527 	vm_offset_t vptr, ptr;
1528 	uint32_t *proc_vector32;
1529 	char **proc_vector;
1530 	size_t vsize, size;
1531 	int i, error;
1532 
1533 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1534 	    &pss, sizeof(pss));
1535 	if (error != 0)
1536 		return (error);
1537 	switch (type) {
1538 	case PROC_ARG:
1539 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1540 		vsize = pss.ps_nargvstr;
1541 		if (vsize > ARG_MAX)
1542 			return (ENOEXEC);
1543 		size = vsize * sizeof(int32_t);
1544 		break;
1545 	case PROC_ENV:
1546 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1547 		vsize = pss.ps_nenvstr;
1548 		if (vsize > ARG_MAX)
1549 			return (ENOEXEC);
1550 		size = vsize * sizeof(int32_t);
1551 		break;
1552 	case PROC_AUX:
1553 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1554 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1555 		if (vptr % 4 != 0)
1556 			return (ENOEXEC);
1557 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1558 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1559 			if (error != 0)
1560 				return (error);
1561 			if (aux.a_type == AT_NULL)
1562 				break;
1563 			ptr += sizeof(aux);
1564 		}
1565 		if (aux.a_type != AT_NULL)
1566 			return (ENOEXEC);
1567 		vsize = i + 1;
1568 		size = vsize * sizeof(aux);
1569 		break;
1570 	default:
1571 		KASSERT(0, ("Wrong proc vector type: %d", type));
1572 		return (EINVAL);
1573 	}
1574 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1575 	error = proc_read_mem(td, p, vptr, proc_vector32, size);
1576 	if (error != 0)
1577 		goto done;
1578 	if (type == PROC_AUX) {
1579 		*proc_vectorp = (char **)proc_vector32;
1580 		*vsizep = vsize;
1581 		return (0);
1582 	}
1583 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1584 	for (i = 0; i < (int)vsize; i++)
1585 		proc_vector[i] = PTRIN(proc_vector32[i]);
1586 	*proc_vectorp = proc_vector;
1587 	*vsizep = vsize;
1588 done:
1589 	free(proc_vector32, M_TEMP);
1590 	return (error);
1591 }
1592 #endif
1593 
1594 static int
1595 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1596     size_t *vsizep, enum proc_vector_type type)
1597 {
1598 	struct ps_strings pss;
1599 	Elf_Auxinfo aux;
1600 	vm_offset_t vptr, ptr;
1601 	char **proc_vector;
1602 	size_t vsize, size;
1603 	int error, i;
1604 
1605 #ifdef COMPAT_FREEBSD32
1606 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1607 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1608 #endif
1609 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1610 	    &pss, sizeof(pss));
1611 	if (error != 0)
1612 		return (error);
1613 	switch (type) {
1614 	case PROC_ARG:
1615 		vptr = (vm_offset_t)pss.ps_argvstr;
1616 		vsize = pss.ps_nargvstr;
1617 		if (vsize > ARG_MAX)
1618 			return (ENOEXEC);
1619 		size = vsize * sizeof(char *);
1620 		break;
1621 	case PROC_ENV:
1622 		vptr = (vm_offset_t)pss.ps_envstr;
1623 		vsize = pss.ps_nenvstr;
1624 		if (vsize > ARG_MAX)
1625 			return (ENOEXEC);
1626 		size = vsize * sizeof(char *);
1627 		break;
1628 	case PROC_AUX:
1629 		/*
1630 		 * The aux array is just above env array on the stack. Check
1631 		 * that the address is naturally aligned.
1632 		 */
1633 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1634 		    * sizeof(char *);
1635 #if __ELF_WORD_SIZE == 64
1636 		if (vptr % sizeof(uint64_t) != 0)
1637 #else
1638 		if (vptr % sizeof(uint32_t) != 0)
1639 #endif
1640 			return (ENOEXEC);
1641 		/*
1642 		 * We count the array size reading the aux vectors from the
1643 		 * stack until AT_NULL vector is returned.  So (to keep the code
1644 		 * simple) we read the process stack twice: the first time here
1645 		 * to find the size and the second time when copying the vectors
1646 		 * to the allocated proc_vector.
1647 		 */
1648 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1649 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1650 			if (error != 0)
1651 				return (error);
1652 			if (aux.a_type == AT_NULL)
1653 				break;
1654 			ptr += sizeof(aux);
1655 		}
1656 		/*
1657 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1658 		 * not reached AT_NULL, it is most likely we are reading wrong
1659 		 * data: either the process doesn't have auxv array or data has
1660 		 * been modified. Return the error in this case.
1661 		 */
1662 		if (aux.a_type != AT_NULL)
1663 			return (ENOEXEC);
1664 		vsize = i + 1;
1665 		size = vsize * sizeof(aux);
1666 		break;
1667 	default:
1668 		KASSERT(0, ("Wrong proc vector type: %d", type));
1669 		return (EINVAL); /* In case we are built without INVARIANTS. */
1670 	}
1671 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1672 	if (proc_vector == NULL)
1673 		return (ENOMEM);
1674 	error = proc_read_mem(td, p, vptr, proc_vector, size);
1675 	if (error != 0) {
1676 		free(proc_vector, M_TEMP);
1677 		return (error);
1678 	}
1679 	*proc_vectorp = proc_vector;
1680 	*vsizep = vsize;
1681 
1682 	return (0);
1683 }
1684 
1685 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1686 
1687 static int
1688 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1689     enum proc_vector_type type)
1690 {
1691 	size_t done, len, nchr, vsize;
1692 	int error, i;
1693 	char **proc_vector, *sptr;
1694 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1695 
1696 	PROC_ASSERT_HELD(p);
1697 
1698 	/*
1699 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1700 	 */
1701 	nchr = 2 * (PATH_MAX + ARG_MAX);
1702 
1703 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1704 	if (error != 0)
1705 		return (error);
1706 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1707 		/*
1708 		 * The program may have scribbled into its argv array, e.g. to
1709 		 * remove some arguments.  If that has happened, break out
1710 		 * before trying to read from NULL.
1711 		 */
1712 		if (proc_vector[i] == NULL)
1713 			break;
1714 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1715 			error = proc_read_string(td, p, sptr, pss_string,
1716 			    sizeof(pss_string));
1717 			if (error != 0)
1718 				goto done;
1719 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1720 			if (done + len >= nchr)
1721 				len = nchr - done - 1;
1722 			sbuf_bcat(sb, pss_string, len);
1723 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1724 				break;
1725 			done += GET_PS_STRINGS_CHUNK_SZ;
1726 		}
1727 		sbuf_bcat(sb, "", 1);
1728 		done += len + 1;
1729 	}
1730 done:
1731 	free(proc_vector, M_TEMP);
1732 	return (error);
1733 }
1734 
1735 int
1736 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1737 {
1738 
1739 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1740 }
1741 
1742 int
1743 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1744 {
1745 
1746 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1747 }
1748 
1749 /*
1750  * This sysctl allows a process to retrieve the argument list or process
1751  * title for another process without groping around in the address space
1752  * of the other process.  It also allow a process to set its own "process
1753  * title to a string of its own choice.
1754  */
1755 static int
1756 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1757 {
1758 	int *name = (int *)arg1;
1759 	u_int namelen = arg2;
1760 	struct pargs *newpa, *pa;
1761 	struct proc *p;
1762 	struct sbuf sb;
1763 	int flags, error = 0, error2;
1764 
1765 	if (namelen != 1)
1766 		return (EINVAL);
1767 
1768 	flags = PGET_CANSEE;
1769 	if (req->newptr != NULL)
1770 		flags |= PGET_ISCURRENT;
1771 	error = pget((pid_t)name[0], flags, &p);
1772 	if (error)
1773 		return (error);
1774 
1775 	pa = p->p_args;
1776 	if (pa != NULL) {
1777 		pargs_hold(pa);
1778 		PROC_UNLOCK(p);
1779 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1780 		pargs_drop(pa);
1781 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1782 		_PHOLD(p);
1783 		PROC_UNLOCK(p);
1784 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1785 		error = proc_getargv(curthread, p, &sb);
1786 		error2 = sbuf_finish(&sb);
1787 		PRELE(p);
1788 		sbuf_delete(&sb);
1789 		if (error == 0 && error2 != 0)
1790 			error = error2;
1791 	} else {
1792 		PROC_UNLOCK(p);
1793 	}
1794 	if (error != 0 || req->newptr == NULL)
1795 		return (error);
1796 
1797 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1798 		return (ENOMEM);
1799 	newpa = pargs_alloc(req->newlen);
1800 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1801 	if (error != 0) {
1802 		pargs_free(newpa);
1803 		return (error);
1804 	}
1805 	PROC_LOCK(p);
1806 	pa = p->p_args;
1807 	p->p_args = newpa;
1808 	PROC_UNLOCK(p);
1809 	pargs_drop(pa);
1810 	return (0);
1811 }
1812 
1813 /*
1814  * This sysctl allows a process to retrieve environment of another process.
1815  */
1816 static int
1817 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
1818 {
1819 	int *name = (int *)arg1;
1820 	u_int namelen = arg2;
1821 	struct proc *p;
1822 	struct sbuf sb;
1823 	int error, error2;
1824 
1825 	if (namelen != 1)
1826 		return (EINVAL);
1827 
1828 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1829 	if (error != 0)
1830 		return (error);
1831 	if ((p->p_flag & P_SYSTEM) != 0) {
1832 		PRELE(p);
1833 		return (0);
1834 	}
1835 
1836 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1837 	error = proc_getenvv(curthread, p, &sb);
1838 	error2 = sbuf_finish(&sb);
1839 	PRELE(p);
1840 	sbuf_delete(&sb);
1841 	return (error != 0 ? error : error2);
1842 }
1843 
1844 /*
1845  * This sysctl allows a process to retrieve ELF auxiliary vector of
1846  * another process.
1847  */
1848 static int
1849 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
1850 {
1851 	int *name = (int *)arg1;
1852 	u_int namelen = arg2;
1853 	struct proc *p;
1854 	size_t vsize, size;
1855 	char **auxv;
1856 	int error;
1857 
1858 	if (namelen != 1)
1859 		return (EINVAL);
1860 
1861 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1862 	if (error != 0)
1863 		return (error);
1864 	if ((p->p_flag & P_SYSTEM) != 0) {
1865 		PRELE(p);
1866 		return (0);
1867 	}
1868 	error = get_proc_vector(curthread, p, &auxv, &vsize, PROC_AUX);
1869 	if (error == 0) {
1870 #ifdef COMPAT_FREEBSD32
1871 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1872 			size = vsize * sizeof(Elf32_Auxinfo);
1873 		else
1874 #endif
1875 		size = vsize * sizeof(Elf_Auxinfo);
1876 		PRELE(p);
1877 		error = SYSCTL_OUT(req, auxv, size);
1878 		free(auxv, M_TEMP);
1879 	} else {
1880 		PRELE(p);
1881 	}
1882 	return (error);
1883 }
1884 
1885 /*
1886  * This sysctl allows a process to retrieve the path of the executable for
1887  * itself or another process.
1888  */
1889 static int
1890 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1891 {
1892 	pid_t *pidp = (pid_t *)arg1;
1893 	unsigned int arglen = arg2;
1894 	struct proc *p;
1895 	struct vnode *vp;
1896 	char *retbuf, *freebuf;
1897 	int error;
1898 
1899 	if (arglen != 1)
1900 		return (EINVAL);
1901 	if (*pidp == -1) {	/* -1 means this process */
1902 		p = req->td->td_proc;
1903 	} else {
1904 		error = pget(*pidp, PGET_CANSEE, &p);
1905 		if (error != 0)
1906 			return (error);
1907 	}
1908 
1909 	vp = p->p_textvp;
1910 	if (vp == NULL) {
1911 		if (*pidp != -1)
1912 			PROC_UNLOCK(p);
1913 		return (0);
1914 	}
1915 	vref(vp);
1916 	if (*pidp != -1)
1917 		PROC_UNLOCK(p);
1918 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1919 	vrele(vp);
1920 	if (error)
1921 		return (error);
1922 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1923 	free(freebuf, M_TEMP);
1924 	return (error);
1925 }
1926 
1927 static int
1928 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1929 {
1930 	struct proc *p;
1931 	char *sv_name;
1932 	int *name;
1933 	int namelen;
1934 	int error;
1935 
1936 	namelen = arg2;
1937 	if (namelen != 1)
1938 		return (EINVAL);
1939 
1940 	name = (int *)arg1;
1941 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
1942 	if (error != 0)
1943 		return (error);
1944 	sv_name = p->p_sysent->sv_name;
1945 	PROC_UNLOCK(p);
1946 	return (sysctl_handle_string(oidp, sv_name, 0, req));
1947 }
1948 
1949 #ifdef KINFO_OVMENTRY_SIZE
1950 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
1951 #endif
1952 
1953 #ifdef COMPAT_FREEBSD7
1954 static int
1955 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
1956 {
1957 	vm_map_entry_t entry, tmp_entry;
1958 	unsigned int last_timestamp;
1959 	char *fullpath, *freepath;
1960 	struct kinfo_ovmentry *kve;
1961 	struct vattr va;
1962 	struct ucred *cred;
1963 	int error, *name;
1964 	struct vnode *vp;
1965 	struct proc *p;
1966 	vm_map_t map;
1967 	struct vmspace *vm;
1968 
1969 	name = (int *)arg1;
1970 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1971 	if (error != 0)
1972 		return (error);
1973 	vm = vmspace_acquire_ref(p);
1974 	if (vm == NULL) {
1975 		PRELE(p);
1976 		return (ESRCH);
1977 	}
1978 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1979 
1980 	map = &vm->vm_map;
1981 	vm_map_lock_read(map);
1982 	for (entry = map->header.next; entry != &map->header;
1983 	    entry = entry->next) {
1984 		vm_object_t obj, tobj, lobj;
1985 		vm_offset_t addr;
1986 
1987 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
1988 			continue;
1989 
1990 		bzero(kve, sizeof(*kve));
1991 		kve->kve_structsize = sizeof(*kve);
1992 
1993 		kve->kve_private_resident = 0;
1994 		obj = entry->object.vm_object;
1995 		if (obj != NULL) {
1996 			VM_OBJECT_LOCK(obj);
1997 			if (obj->shadow_count == 1)
1998 				kve->kve_private_resident =
1999 				    obj->resident_page_count;
2000 		}
2001 		kve->kve_resident = 0;
2002 		addr = entry->start;
2003 		while (addr < entry->end) {
2004 			if (pmap_extract(map->pmap, addr))
2005 				kve->kve_resident++;
2006 			addr += PAGE_SIZE;
2007 		}
2008 
2009 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2010 			if (tobj != obj)
2011 				VM_OBJECT_LOCK(tobj);
2012 			if (lobj != obj)
2013 				VM_OBJECT_UNLOCK(lobj);
2014 			lobj = tobj;
2015 		}
2016 
2017 		kve->kve_start = (void*)entry->start;
2018 		kve->kve_end = (void*)entry->end;
2019 		kve->kve_offset = (off_t)entry->offset;
2020 
2021 		if (entry->protection & VM_PROT_READ)
2022 			kve->kve_protection |= KVME_PROT_READ;
2023 		if (entry->protection & VM_PROT_WRITE)
2024 			kve->kve_protection |= KVME_PROT_WRITE;
2025 		if (entry->protection & VM_PROT_EXECUTE)
2026 			kve->kve_protection |= KVME_PROT_EXEC;
2027 
2028 		if (entry->eflags & MAP_ENTRY_COW)
2029 			kve->kve_flags |= KVME_FLAG_COW;
2030 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2031 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2032 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2033 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2034 
2035 		last_timestamp = map->timestamp;
2036 		vm_map_unlock_read(map);
2037 
2038 		kve->kve_fileid = 0;
2039 		kve->kve_fsid = 0;
2040 		freepath = NULL;
2041 		fullpath = "";
2042 		if (lobj) {
2043 			vp = NULL;
2044 			switch (lobj->type) {
2045 			case OBJT_DEFAULT:
2046 				kve->kve_type = KVME_TYPE_DEFAULT;
2047 				break;
2048 			case OBJT_VNODE:
2049 				kve->kve_type = KVME_TYPE_VNODE;
2050 				vp = lobj->handle;
2051 				vref(vp);
2052 				break;
2053 			case OBJT_SWAP:
2054 				kve->kve_type = KVME_TYPE_SWAP;
2055 				break;
2056 			case OBJT_DEVICE:
2057 				kve->kve_type = KVME_TYPE_DEVICE;
2058 				break;
2059 			case OBJT_PHYS:
2060 				kve->kve_type = KVME_TYPE_PHYS;
2061 				break;
2062 			case OBJT_DEAD:
2063 				kve->kve_type = KVME_TYPE_DEAD;
2064 				break;
2065 			case OBJT_SG:
2066 				kve->kve_type = KVME_TYPE_SG;
2067 				break;
2068 			default:
2069 				kve->kve_type = KVME_TYPE_UNKNOWN;
2070 				break;
2071 			}
2072 			if (lobj != obj)
2073 				VM_OBJECT_UNLOCK(lobj);
2074 
2075 			kve->kve_ref_count = obj->ref_count;
2076 			kve->kve_shadow_count = obj->shadow_count;
2077 			VM_OBJECT_UNLOCK(obj);
2078 			if (vp != NULL) {
2079 				vn_fullpath(curthread, vp, &fullpath,
2080 				    &freepath);
2081 				cred = curthread->td_ucred;
2082 				vn_lock(vp, LK_SHARED | LK_RETRY);
2083 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2084 					kve->kve_fileid = va.va_fileid;
2085 					kve->kve_fsid = va.va_fsid;
2086 				}
2087 				vput(vp);
2088 			}
2089 		} else {
2090 			kve->kve_type = KVME_TYPE_NONE;
2091 			kve->kve_ref_count = 0;
2092 			kve->kve_shadow_count = 0;
2093 		}
2094 
2095 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2096 		if (freepath != NULL)
2097 			free(freepath, M_TEMP);
2098 
2099 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2100 		vm_map_lock_read(map);
2101 		if (error)
2102 			break;
2103 		if (last_timestamp != map->timestamp) {
2104 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2105 			entry = tmp_entry;
2106 		}
2107 	}
2108 	vm_map_unlock_read(map);
2109 	vmspace_free(vm);
2110 	PRELE(p);
2111 	free(kve, M_TEMP);
2112 	return (error);
2113 }
2114 #endif	/* COMPAT_FREEBSD7 */
2115 
2116 #ifdef KINFO_VMENTRY_SIZE
2117 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2118 #endif
2119 
2120 static int
2121 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2122 {
2123 	vm_map_entry_t entry, tmp_entry;
2124 	unsigned int last_timestamp;
2125 	char *fullpath, *freepath;
2126 	struct kinfo_vmentry *kve;
2127 	struct vattr va;
2128 	struct ucred *cred;
2129 	int error, *name;
2130 	struct vnode *vp;
2131 	struct proc *p;
2132 	struct vmspace *vm;
2133 	vm_map_t map;
2134 
2135 	name = (int *)arg1;
2136 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2137 	if (error != 0)
2138 		return (error);
2139 	vm = vmspace_acquire_ref(p);
2140 	if (vm == NULL) {
2141 		PRELE(p);
2142 		return (ESRCH);
2143 	}
2144 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2145 
2146 	map = &vm->vm_map;
2147 	vm_map_lock_read(map);
2148 	for (entry = map->header.next; entry != &map->header;
2149 	    entry = entry->next) {
2150 		vm_object_t obj, tobj, lobj;
2151 		vm_offset_t addr;
2152 		vm_paddr_t locked_pa;
2153 		int mincoreinfo;
2154 
2155 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2156 			continue;
2157 
2158 		bzero(kve, sizeof(*kve));
2159 
2160 		kve->kve_private_resident = 0;
2161 		obj = entry->object.vm_object;
2162 		if (obj != NULL) {
2163 			VM_OBJECT_LOCK(obj);
2164 			if (obj->shadow_count == 1)
2165 				kve->kve_private_resident =
2166 				    obj->resident_page_count;
2167 		}
2168 		kve->kve_resident = 0;
2169 		addr = entry->start;
2170 		while (addr < entry->end) {
2171 			locked_pa = 0;
2172 			mincoreinfo = pmap_mincore(map->pmap, addr, &locked_pa);
2173 			if (locked_pa != 0)
2174 				vm_page_unlock(PHYS_TO_VM_PAGE(locked_pa));
2175 			if (mincoreinfo & MINCORE_INCORE)
2176 				kve->kve_resident++;
2177 			if (mincoreinfo & MINCORE_SUPER)
2178 				kve->kve_flags |= KVME_FLAG_SUPER;
2179 			addr += PAGE_SIZE;
2180 		}
2181 
2182 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2183 			if (tobj != obj)
2184 				VM_OBJECT_LOCK(tobj);
2185 			if (lobj != obj)
2186 				VM_OBJECT_UNLOCK(lobj);
2187 			lobj = tobj;
2188 		}
2189 
2190 		kve->kve_start = entry->start;
2191 		kve->kve_end = entry->end;
2192 		kve->kve_offset = entry->offset;
2193 
2194 		if (entry->protection & VM_PROT_READ)
2195 			kve->kve_protection |= KVME_PROT_READ;
2196 		if (entry->protection & VM_PROT_WRITE)
2197 			kve->kve_protection |= KVME_PROT_WRITE;
2198 		if (entry->protection & VM_PROT_EXECUTE)
2199 			kve->kve_protection |= KVME_PROT_EXEC;
2200 
2201 		if (entry->eflags & MAP_ENTRY_COW)
2202 			kve->kve_flags |= KVME_FLAG_COW;
2203 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2204 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2205 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2206 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2207 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2208 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2209 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2210 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2211 
2212 		last_timestamp = map->timestamp;
2213 		vm_map_unlock_read(map);
2214 
2215 		freepath = NULL;
2216 		fullpath = "";
2217 		if (lobj) {
2218 			vp = NULL;
2219 			switch (lobj->type) {
2220 			case OBJT_DEFAULT:
2221 				kve->kve_type = KVME_TYPE_DEFAULT;
2222 				break;
2223 			case OBJT_VNODE:
2224 				kve->kve_type = KVME_TYPE_VNODE;
2225 				vp = lobj->handle;
2226 				vref(vp);
2227 				break;
2228 			case OBJT_SWAP:
2229 				kve->kve_type = KVME_TYPE_SWAP;
2230 				break;
2231 			case OBJT_DEVICE:
2232 				kve->kve_type = KVME_TYPE_DEVICE;
2233 				break;
2234 			case OBJT_PHYS:
2235 				kve->kve_type = KVME_TYPE_PHYS;
2236 				break;
2237 			case OBJT_DEAD:
2238 				kve->kve_type = KVME_TYPE_DEAD;
2239 				break;
2240 			case OBJT_SG:
2241 				kve->kve_type = KVME_TYPE_SG;
2242 				break;
2243 			default:
2244 				kve->kve_type = KVME_TYPE_UNKNOWN;
2245 				break;
2246 			}
2247 			if (lobj != obj)
2248 				VM_OBJECT_UNLOCK(lobj);
2249 
2250 			kve->kve_ref_count = obj->ref_count;
2251 			kve->kve_shadow_count = obj->shadow_count;
2252 			VM_OBJECT_UNLOCK(obj);
2253 			if (vp != NULL) {
2254 				vn_fullpath(curthread, vp, &fullpath,
2255 				    &freepath);
2256 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2257 				cred = curthread->td_ucred;
2258 				vn_lock(vp, LK_SHARED | LK_RETRY);
2259 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2260 					kve->kve_vn_fileid = va.va_fileid;
2261 					kve->kve_vn_fsid = va.va_fsid;
2262 					kve->kve_vn_mode =
2263 					    MAKEIMODE(va.va_type, va.va_mode);
2264 					kve->kve_vn_size = va.va_size;
2265 					kve->kve_vn_rdev = va.va_rdev;
2266 					kve->kve_status = KF_ATTR_VALID;
2267 				}
2268 				vput(vp);
2269 			}
2270 		} else {
2271 			kve->kve_type = KVME_TYPE_NONE;
2272 			kve->kve_ref_count = 0;
2273 			kve->kve_shadow_count = 0;
2274 		}
2275 
2276 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2277 		if (freepath != NULL)
2278 			free(freepath, M_TEMP);
2279 
2280 		/* Pack record size down */
2281 		kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) +
2282 		    strlen(kve->kve_path) + 1;
2283 		kve->kve_structsize = roundup(kve->kve_structsize,
2284 		    sizeof(uint64_t));
2285 		error = SYSCTL_OUT(req, kve, kve->kve_structsize);
2286 		vm_map_lock_read(map);
2287 		if (error)
2288 			break;
2289 		if (last_timestamp != map->timestamp) {
2290 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2291 			entry = tmp_entry;
2292 		}
2293 	}
2294 	vm_map_unlock_read(map);
2295 	vmspace_free(vm);
2296 	PRELE(p);
2297 	free(kve, M_TEMP);
2298 	return (error);
2299 }
2300 
2301 #if defined(STACK) || defined(DDB)
2302 static int
2303 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2304 {
2305 	struct kinfo_kstack *kkstp;
2306 	int error, i, *name, numthreads;
2307 	lwpid_t *lwpidarray;
2308 	struct thread *td;
2309 	struct stack *st;
2310 	struct sbuf sb;
2311 	struct proc *p;
2312 
2313 	name = (int *)arg1;
2314 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2315 	if (error != 0)
2316 		return (error);
2317 
2318 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2319 	st = stack_create();
2320 
2321 	lwpidarray = NULL;
2322 	numthreads = 0;
2323 	PROC_LOCK(p);
2324 repeat:
2325 	if (numthreads < p->p_numthreads) {
2326 		if (lwpidarray != NULL) {
2327 			free(lwpidarray, M_TEMP);
2328 			lwpidarray = NULL;
2329 		}
2330 		numthreads = p->p_numthreads;
2331 		PROC_UNLOCK(p);
2332 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2333 		    M_WAITOK | M_ZERO);
2334 		PROC_LOCK(p);
2335 		goto repeat;
2336 	}
2337 	i = 0;
2338 
2339 	/*
2340 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2341 	 * of changes could have taken place.  Should we check to see if the
2342 	 * vmspace has been replaced, or the like, in order to prevent
2343 	 * giving a snapshot that spans, say, execve(2), with some threads
2344 	 * before and some after?  Among other things, the credentials could
2345 	 * have changed, in which case the right to extract debug info might
2346 	 * no longer be assured.
2347 	 */
2348 	FOREACH_THREAD_IN_PROC(p, td) {
2349 		KASSERT(i < numthreads,
2350 		    ("sysctl_kern_proc_kstack: numthreads"));
2351 		lwpidarray[i] = td->td_tid;
2352 		i++;
2353 	}
2354 	numthreads = i;
2355 	for (i = 0; i < numthreads; i++) {
2356 		td = thread_find(p, lwpidarray[i]);
2357 		if (td == NULL) {
2358 			continue;
2359 		}
2360 		bzero(kkstp, sizeof(*kkstp));
2361 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2362 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2363 		thread_lock(td);
2364 		kkstp->kkst_tid = td->td_tid;
2365 		if (TD_IS_SWAPPED(td))
2366 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2367 		else if (TD_IS_RUNNING(td))
2368 			kkstp->kkst_state = KKST_STATE_RUNNING;
2369 		else {
2370 			kkstp->kkst_state = KKST_STATE_STACKOK;
2371 			stack_save_td(st, td);
2372 		}
2373 		thread_unlock(td);
2374 		PROC_UNLOCK(p);
2375 		stack_sbuf_print(&sb, st);
2376 		sbuf_finish(&sb);
2377 		sbuf_delete(&sb);
2378 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2379 		PROC_LOCK(p);
2380 		if (error)
2381 			break;
2382 	}
2383 	_PRELE(p);
2384 	PROC_UNLOCK(p);
2385 	if (lwpidarray != NULL)
2386 		free(lwpidarray, M_TEMP);
2387 	stack_destroy(st);
2388 	free(kkstp, M_TEMP);
2389 	return (error);
2390 }
2391 #endif
2392 
2393 /*
2394  * This sysctl allows a process to retrieve the full list of groups from
2395  * itself or another process.
2396  */
2397 static int
2398 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2399 {
2400 	pid_t *pidp = (pid_t *)arg1;
2401 	unsigned int arglen = arg2;
2402 	struct proc *p;
2403 	struct ucred *cred;
2404 	int error;
2405 
2406 	if (arglen != 1)
2407 		return (EINVAL);
2408 	if (*pidp == -1) {	/* -1 means this process */
2409 		p = req->td->td_proc;
2410 	} else {
2411 		error = pget(*pidp, PGET_CANSEE, &p);
2412 		if (error != 0)
2413 			return (error);
2414 	}
2415 
2416 	cred = crhold(p->p_ucred);
2417 	if (*pidp != -1)
2418 		PROC_UNLOCK(p);
2419 
2420 	error = SYSCTL_OUT(req, cred->cr_groups,
2421 	    cred->cr_ngroups * sizeof(gid_t));
2422 	crfree(cred);
2423 	return (error);
2424 }
2425 
2426 /*
2427  * This sysctl allows a process to retrieve or/and set the resource limit for
2428  * another process.
2429  */
2430 static int
2431 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2432 {
2433 	int *name = (int *)arg1;
2434 	u_int namelen = arg2;
2435 	struct rlimit rlim;
2436 	struct proc *p;
2437 	u_int which;
2438 	int flags, error;
2439 
2440 	if (namelen != 2)
2441 		return (EINVAL);
2442 
2443 	which = (u_int)name[1];
2444 	if (which >= RLIM_NLIMITS)
2445 		return (EINVAL);
2446 
2447 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2448 		return (EINVAL);
2449 
2450 	flags = PGET_HOLD | PGET_NOTWEXIT;
2451 	if (req->newptr != NULL)
2452 		flags |= PGET_CANDEBUG;
2453 	else
2454 		flags |= PGET_CANSEE;
2455 	error = pget((pid_t)name[0], flags, &p);
2456 	if (error != 0)
2457 		return (error);
2458 
2459 	/*
2460 	 * Retrieve limit.
2461 	 */
2462 	if (req->oldptr != NULL) {
2463 		PROC_LOCK(p);
2464 		lim_rlimit(p, which, &rlim);
2465 		PROC_UNLOCK(p);
2466 	}
2467 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2468 	if (error != 0)
2469 		goto errout;
2470 
2471 	/*
2472 	 * Set limit.
2473 	 */
2474 	if (req->newptr != NULL) {
2475 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2476 		if (error == 0)
2477 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2478 	}
2479 
2480 errout:
2481 	PRELE(p);
2482 	return (error);
2483 }
2484 
2485 /*
2486  * This sysctl allows a process to retrieve ps_strings structure location of
2487  * another process.
2488  */
2489 static int
2490 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2491 {
2492 	int *name = (int *)arg1;
2493 	u_int namelen = arg2;
2494 	struct proc *p;
2495 	vm_offset_t ps_strings;
2496 	int error;
2497 #ifdef COMPAT_FREEBSD32
2498 	uint32_t ps_strings32;
2499 #endif
2500 
2501 	if (namelen != 1)
2502 		return (EINVAL);
2503 
2504 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2505 	if (error != 0)
2506 		return (error);
2507 #ifdef COMPAT_FREEBSD32
2508 	if ((req->flags & SCTL_MASK32) != 0) {
2509 		/*
2510 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2511 		 * process.
2512 		 */
2513 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2514 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2515 		PROC_UNLOCK(p);
2516 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2517 		return (error);
2518 	}
2519 #endif
2520 	ps_strings = p->p_sysent->sv_psstrings;
2521 	PROC_UNLOCK(p);
2522 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2523 	return (error);
2524 }
2525 
2526 /*
2527  * This sysctl allows a process to retrieve umask of another process.
2528  */
2529 static int
2530 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2531 {
2532 	int *name = (int *)arg1;
2533 	u_int namelen = arg2;
2534 	struct proc *p;
2535 	int error;
2536 	u_short fd_cmask;
2537 
2538 	if (namelen != 1)
2539 		return (EINVAL);
2540 
2541 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2542 	if (error != 0)
2543 		return (error);
2544 
2545 	FILEDESC_SLOCK(p->p_fd);
2546 	fd_cmask = p->p_fd->fd_cmask;
2547 	FILEDESC_SUNLOCK(p->p_fd);
2548 	PRELE(p);
2549 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2550 	return (error);
2551 }
2552 
2553 /*
2554  * This sysctl allows a process to set and retrieve binary osreldate of
2555  * another process.
2556  */
2557 static int
2558 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2559 {
2560 	int *name = (int *)arg1;
2561 	u_int namelen = arg2;
2562 	struct proc *p;
2563 	int flags, error, osrel;
2564 
2565 	if (namelen != 1)
2566 		return (EINVAL);
2567 
2568 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2569 		return (EINVAL);
2570 
2571 	flags = PGET_HOLD | PGET_NOTWEXIT;
2572 	if (req->newptr != NULL)
2573 		flags |= PGET_CANDEBUG;
2574 	else
2575 		flags |= PGET_CANSEE;
2576 	error = pget((pid_t)name[0], flags, &p);
2577 	if (error != 0)
2578 		return (error);
2579 
2580 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2581 	if (error != 0)
2582 		goto errout;
2583 
2584 	if (req->newptr != NULL) {
2585 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2586 		if (error != 0)
2587 			goto errout;
2588 		if (osrel < 0) {
2589 			error = EINVAL;
2590 			goto errout;
2591 		}
2592 		p->p_osrel = osrel;
2593 	}
2594 errout:
2595 	PRELE(p);
2596 	return (error);
2597 }
2598 
2599 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2600 
2601 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2602 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2603 	"Return entire process table");
2604 
2605 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2606 	sysctl_kern_proc, "Process table");
2607 
2608 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2609 	sysctl_kern_proc, "Process table");
2610 
2611 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2612 	sysctl_kern_proc, "Process table");
2613 
2614 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2615 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2616 
2617 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2618 	sysctl_kern_proc, "Process table");
2619 
2620 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2621 	sysctl_kern_proc, "Process table");
2622 
2623 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2624 	sysctl_kern_proc, "Process table");
2625 
2626 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2627 	sysctl_kern_proc, "Process table");
2628 
2629 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2630 	sysctl_kern_proc, "Return process table, no threads");
2631 
2632 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2633 	CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2634 	sysctl_kern_proc_args, "Process argument list");
2635 
2636 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
2637 	sysctl_kern_proc_env, "Process environment");
2638 
2639 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
2640 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
2641 
2642 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2643 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2644 
2645 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2646 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2647 	"Process syscall vector name (ABI type)");
2648 
2649 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2650 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2651 
2652 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2653 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2654 
2655 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2656 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2657 
2658 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2659 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2660 
2661 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2662 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2663 
2664 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2665 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2666 
2667 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2668 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2669 
2670 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2671 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2672 
2673 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2674 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2675 	"Return process table, no threads");
2676 
2677 #ifdef COMPAT_FREEBSD7
2678 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2679 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2680 #endif
2681 
2682 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2683 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2684 
2685 #if defined(STACK) || defined(DDB)
2686 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2687 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2688 #endif
2689 
2690 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2691 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
2692 
2693 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
2694 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
2695 	"Process resource limits");
2696 
2697 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
2698 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
2699 	"Process ps_strings location");
2700 
2701 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
2702 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
2703 
2704 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
2705 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
2706 	"Process binary osreldate");
2707