xref: /freebsd/sys/kern/kern_proc.c (revision 4f6aec90ff8521301da9c6bc04310e1ab25a410c)
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_ktrace.h"
38 #include "opt_kstack_pages.h"
39 #include "opt_stack.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/elf.h>
44 #include <sys/exec.h>
45 #include <sys/kernel.h>
46 #include <sys/limits.h>
47 #include <sys/lock.h>
48 #include <sys/loginclass.h>
49 #include <sys/malloc.h>
50 #include <sys/mman.h>
51 #include <sys/mount.h>
52 #include <sys/mutex.h>
53 #include <sys/proc.h>
54 #include <sys/ptrace.h>
55 #include <sys/refcount.h>
56 #include <sys/resourcevar.h>
57 #include <sys/rwlock.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_DEFINE4(proc, kernel, ctor, entry, "struct proc *", "int",
95     "void *", "int");
96 SDT_PROBE_DEFINE4(proc, kernel, ctor, return, "struct proc *", "int",
97     "void *", "int");
98 SDT_PROBE_DEFINE4(proc, kernel, dtor, entry, "struct proc *", "int",
99     "void *", "struct thread *");
100 SDT_PROBE_DEFINE3(proc, kernel, dtor, return, "struct proc *", "int",
101     "void *");
102 SDT_PROBE_DEFINE3(proc, kernel, init, entry, "struct proc *", "int",
103     "int");
104 SDT_PROBE_DEFINE3(proc, kernel, init, return, "struct proc *", "int",
105     "int");
106 
107 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
108 MALLOC_DEFINE(M_SESSION, "session", "session header");
109 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
110 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
111 
112 static void doenterpgrp(struct proc *, struct pgrp *);
113 static void orphanpg(struct pgrp *pg);
114 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
115 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
116 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
117     int preferthread);
118 static void pgadjustjobc(struct pgrp *pgrp, int entering);
119 static void pgdelete(struct pgrp *);
120 static int proc_ctor(void *mem, int size, void *arg, int flags);
121 static void proc_dtor(void *mem, int size, void *arg);
122 static int proc_init(void *mem, int size, int flags);
123 static void proc_fini(void *mem, int size);
124 static void pargs_free(struct pargs *pa);
125 static struct proc *zpfind_locked(pid_t pid);
126 
127 /*
128  * Other process lists
129  */
130 struct pidhashhead *pidhashtbl;
131 u_long pidhash;
132 struct pgrphashhead *pgrphashtbl;
133 u_long pgrphash;
134 struct proclist allproc;
135 struct proclist zombproc;
136 struct sx allproc_lock;
137 struct sx proctree_lock;
138 struct mtx ppeers_lock;
139 uma_zone_t proc_zone;
140 
141 int kstack_pages = KSTACK_PAGES;
142 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
143     "Kernel stack size in pages");
144 static int vmmap_skip_res_cnt = 0;
145 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
146     &vmmap_skip_res_cnt, 0,
147     "Skip calculation of the pages resident count in kern.proc.vmmap");
148 
149 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
150 #ifdef COMPAT_FREEBSD32
151 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
152 #endif
153 
154 /*
155  * Initialize global process hashing structures.
156  */
157 void
158 procinit()
159 {
160 
161 	sx_init(&allproc_lock, "allproc");
162 	sx_init(&proctree_lock, "proctree");
163 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
164 	LIST_INIT(&allproc);
165 	LIST_INIT(&zombproc);
166 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
167 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
168 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
169 	    proc_ctor, proc_dtor, proc_init, proc_fini,
170 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
171 	uihashinit();
172 }
173 
174 /*
175  * Prepare a proc for use.
176  */
177 static int
178 proc_ctor(void *mem, int size, void *arg, int flags)
179 {
180 	struct proc *p;
181 
182 	p = (struct proc *)mem;
183 	SDT_PROBE(proc, kernel, ctor , entry, p, size, arg, flags, 0);
184 	EVENTHANDLER_INVOKE(process_ctor, p);
185 	SDT_PROBE(proc, kernel, ctor , return, p, size, arg, flags, 0);
186 	return (0);
187 }
188 
189 /*
190  * Reclaim a proc after use.
191  */
192 static void
193 proc_dtor(void *mem, int size, void *arg)
194 {
195 	struct proc *p;
196 	struct thread *td;
197 
198 	/* INVARIANTS checks go here */
199 	p = (struct proc *)mem;
200 	td = FIRST_THREAD_IN_PROC(p);
201 	SDT_PROBE(proc, kernel, dtor, entry, p, size, arg, td, 0);
202 	if (td != NULL) {
203 #ifdef INVARIANTS
204 		KASSERT((p->p_numthreads == 1),
205 		    ("bad number of threads in exiting process"));
206 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
207 #endif
208 		/* Free all OSD associated to this thread. */
209 		osd_thread_exit(td);
210 	}
211 	EVENTHANDLER_INVOKE(process_dtor, p);
212 	if (p->p_ksi != NULL)
213 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
214 	SDT_PROBE(proc, kernel, dtor, return, p, size, arg, 0, 0);
215 }
216 
217 /*
218  * Initialize type-stable parts of a proc (when newly created).
219  */
220 static int
221 proc_init(void *mem, int size, int flags)
222 {
223 	struct proc *p;
224 
225 	p = (struct proc *)mem;
226 	SDT_PROBE(proc, kernel, init, entry, p, size, flags, 0, 0);
227 	p->p_sched = (struct p_sched *)&p[1];
228 	bzero(&p->p_mtx, sizeof(struct mtx));
229 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK);
230 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_RECURSE);
231 	cv_init(&p->p_pwait, "ppwait");
232 	cv_init(&p->p_dbgwait, "dbgwait");
233 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
234 	EVENTHANDLER_INVOKE(process_init, p);
235 	p->p_stats = pstats_alloc();
236 	SDT_PROBE(proc, kernel, init, return, p, size, flags, 0, 0);
237 	return (0);
238 }
239 
240 /*
241  * UMA should ensure that this function is never called.
242  * Freeing a proc structure would violate type stability.
243  */
244 static void
245 proc_fini(void *mem, int size)
246 {
247 #ifdef notnow
248 	struct proc *p;
249 
250 	p = (struct proc *)mem;
251 	EVENTHANDLER_INVOKE(process_fini, p);
252 	pstats_free(p->p_stats);
253 	thread_free(FIRST_THREAD_IN_PROC(p));
254 	mtx_destroy(&p->p_mtx);
255 	if (p->p_ksi != NULL)
256 		ksiginfo_free(p->p_ksi);
257 #else
258 	panic("proc reclaimed");
259 #endif
260 }
261 
262 /*
263  * Is p an inferior of the current process?
264  */
265 int
266 inferior(struct proc *p)
267 {
268 
269 	sx_assert(&proctree_lock, SX_LOCKED);
270 	PROC_LOCK_ASSERT(p, MA_OWNED);
271 	for (; p != curproc; p = proc_realparent(p)) {
272 		if (p->p_pid == 0)
273 			return (0);
274 	}
275 	return (1);
276 }
277 
278 struct proc *
279 pfind_locked(pid_t pid)
280 {
281 	struct proc *p;
282 
283 	sx_assert(&allproc_lock, SX_LOCKED);
284 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
285 		if (p->p_pid == pid) {
286 			PROC_LOCK(p);
287 			if (p->p_state == PRS_NEW) {
288 				PROC_UNLOCK(p);
289 				p = NULL;
290 			}
291 			break;
292 		}
293 	}
294 	return (p);
295 }
296 
297 /*
298  * Locate a process by number; return only "live" processes -- i.e., neither
299  * zombies nor newly born but incompletely initialized processes.  By not
300  * returning processes in the PRS_NEW state, we allow callers to avoid
301  * testing for that condition to avoid dereferencing p_ucred, et al.
302  */
303 struct proc *
304 pfind(pid_t pid)
305 {
306 	struct proc *p;
307 
308 	sx_slock(&allproc_lock);
309 	p = pfind_locked(pid);
310 	sx_sunlock(&allproc_lock);
311 	return (p);
312 }
313 
314 static struct proc *
315 pfind_tid_locked(pid_t tid)
316 {
317 	struct proc *p;
318 	struct thread *td;
319 
320 	sx_assert(&allproc_lock, SX_LOCKED);
321 	FOREACH_PROC_IN_SYSTEM(p) {
322 		PROC_LOCK(p);
323 		if (p->p_state == PRS_NEW) {
324 			PROC_UNLOCK(p);
325 			continue;
326 		}
327 		FOREACH_THREAD_IN_PROC(p, td) {
328 			if (td->td_tid == tid)
329 				goto found;
330 		}
331 		PROC_UNLOCK(p);
332 	}
333 found:
334 	return (p);
335 }
336 
337 /*
338  * Locate a process group by number.
339  * The caller must hold proctree_lock.
340  */
341 struct pgrp *
342 pgfind(pgid)
343 	register pid_t pgid;
344 {
345 	register struct pgrp *pgrp;
346 
347 	sx_assert(&proctree_lock, SX_LOCKED);
348 
349 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
350 		if (pgrp->pg_id == pgid) {
351 			PGRP_LOCK(pgrp);
352 			return (pgrp);
353 		}
354 	}
355 	return (NULL);
356 }
357 
358 /*
359  * Locate process and do additional manipulations, depending on flags.
360  */
361 int
362 pget(pid_t pid, int flags, struct proc **pp)
363 {
364 	struct proc *p;
365 	int error;
366 
367 	sx_slock(&allproc_lock);
368 	if (pid <= PID_MAX) {
369 		p = pfind_locked(pid);
370 		if (p == NULL && (flags & PGET_NOTWEXIT) == 0)
371 			p = zpfind_locked(pid);
372 	} else if ((flags & PGET_NOTID) == 0) {
373 		p = pfind_tid_locked(pid);
374 	} else {
375 		p = NULL;
376 	}
377 	sx_sunlock(&allproc_lock);
378 	if (p == NULL)
379 		return (ESRCH);
380 	if ((flags & PGET_CANSEE) != 0) {
381 		error = p_cansee(curthread, p);
382 		if (error != 0)
383 			goto errout;
384 	}
385 	if ((flags & PGET_CANDEBUG) != 0) {
386 		error = p_candebug(curthread, p);
387 		if (error != 0)
388 			goto errout;
389 	}
390 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
391 		error = EPERM;
392 		goto errout;
393 	}
394 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
395 		error = ESRCH;
396 		goto errout;
397 	}
398 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
399 		/*
400 		 * XXXRW: Not clear ESRCH is the right error during proc
401 		 * execve().
402 		 */
403 		error = ESRCH;
404 		goto errout;
405 	}
406 	if ((flags & PGET_HOLD) != 0) {
407 		_PHOLD(p);
408 		PROC_UNLOCK(p);
409 	}
410 	*pp = p;
411 	return (0);
412 errout:
413 	PROC_UNLOCK(p);
414 	return (error);
415 }
416 
417 /*
418  * Create a new process group.
419  * pgid must be equal to the pid of p.
420  * Begin a new session if required.
421  */
422 int
423 enterpgrp(p, pgid, pgrp, sess)
424 	register struct proc *p;
425 	pid_t pgid;
426 	struct pgrp *pgrp;
427 	struct session *sess;
428 {
429 
430 	sx_assert(&proctree_lock, SX_XLOCKED);
431 
432 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
433 	KASSERT(p->p_pid == pgid,
434 	    ("enterpgrp: new pgrp and pid != pgid"));
435 	KASSERT(pgfind(pgid) == NULL,
436 	    ("enterpgrp: pgrp with pgid exists"));
437 	KASSERT(!SESS_LEADER(p),
438 	    ("enterpgrp: session leader attempted setpgrp"));
439 
440 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
441 
442 	if (sess != NULL) {
443 		/*
444 		 * new session
445 		 */
446 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
447 		PROC_LOCK(p);
448 		p->p_flag &= ~P_CONTROLT;
449 		PROC_UNLOCK(p);
450 		PGRP_LOCK(pgrp);
451 		sess->s_leader = p;
452 		sess->s_sid = p->p_pid;
453 		refcount_init(&sess->s_count, 1);
454 		sess->s_ttyvp = NULL;
455 		sess->s_ttydp = NULL;
456 		sess->s_ttyp = NULL;
457 		bcopy(p->p_session->s_login, sess->s_login,
458 			    sizeof(sess->s_login));
459 		pgrp->pg_session = sess;
460 		KASSERT(p == curproc,
461 		    ("enterpgrp: mksession and p != curproc"));
462 	} else {
463 		pgrp->pg_session = p->p_session;
464 		sess_hold(pgrp->pg_session);
465 		PGRP_LOCK(pgrp);
466 	}
467 	pgrp->pg_id = pgid;
468 	LIST_INIT(&pgrp->pg_members);
469 
470 	/*
471 	 * As we have an exclusive lock of proctree_lock,
472 	 * this should not deadlock.
473 	 */
474 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
475 	pgrp->pg_jobc = 0;
476 	SLIST_INIT(&pgrp->pg_sigiolst);
477 	PGRP_UNLOCK(pgrp);
478 
479 	doenterpgrp(p, pgrp);
480 
481 	return (0);
482 }
483 
484 /*
485  * Move p to an existing process group
486  */
487 int
488 enterthispgrp(p, pgrp)
489 	register struct proc *p;
490 	struct pgrp *pgrp;
491 {
492 
493 	sx_assert(&proctree_lock, SX_XLOCKED);
494 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
495 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
496 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
497 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
498 	KASSERT(pgrp->pg_session == p->p_session,
499 		("%s: pgrp's session %p, p->p_session %p.\n",
500 		__func__,
501 		pgrp->pg_session,
502 		p->p_session));
503 	KASSERT(pgrp != p->p_pgrp,
504 		("%s: p belongs to pgrp.", __func__));
505 
506 	doenterpgrp(p, pgrp);
507 
508 	return (0);
509 }
510 
511 /*
512  * Move p to a process group
513  */
514 static void
515 doenterpgrp(p, pgrp)
516 	struct proc *p;
517 	struct pgrp *pgrp;
518 {
519 	struct pgrp *savepgrp;
520 
521 	sx_assert(&proctree_lock, SX_XLOCKED);
522 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
523 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
524 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
525 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
526 
527 	savepgrp = p->p_pgrp;
528 
529 	/*
530 	 * Adjust eligibility of affected pgrps to participate in job control.
531 	 * Increment eligibility counts before decrementing, otherwise we
532 	 * could reach 0 spuriously during the first call.
533 	 */
534 	fixjobc(p, pgrp, 1);
535 	fixjobc(p, p->p_pgrp, 0);
536 
537 	PGRP_LOCK(pgrp);
538 	PGRP_LOCK(savepgrp);
539 	PROC_LOCK(p);
540 	LIST_REMOVE(p, p_pglist);
541 	p->p_pgrp = pgrp;
542 	PROC_UNLOCK(p);
543 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
544 	PGRP_UNLOCK(savepgrp);
545 	PGRP_UNLOCK(pgrp);
546 	if (LIST_EMPTY(&savepgrp->pg_members))
547 		pgdelete(savepgrp);
548 }
549 
550 /*
551  * remove process from process group
552  */
553 int
554 leavepgrp(p)
555 	register struct proc *p;
556 {
557 	struct pgrp *savepgrp;
558 
559 	sx_assert(&proctree_lock, SX_XLOCKED);
560 	savepgrp = p->p_pgrp;
561 	PGRP_LOCK(savepgrp);
562 	PROC_LOCK(p);
563 	LIST_REMOVE(p, p_pglist);
564 	p->p_pgrp = NULL;
565 	PROC_UNLOCK(p);
566 	PGRP_UNLOCK(savepgrp);
567 	if (LIST_EMPTY(&savepgrp->pg_members))
568 		pgdelete(savepgrp);
569 	return (0);
570 }
571 
572 /*
573  * delete a process group
574  */
575 static void
576 pgdelete(pgrp)
577 	register struct pgrp *pgrp;
578 {
579 	struct session *savesess;
580 	struct tty *tp;
581 
582 	sx_assert(&proctree_lock, SX_XLOCKED);
583 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
584 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
585 
586 	/*
587 	 * Reset any sigio structures pointing to us as a result of
588 	 * F_SETOWN with our pgid.
589 	 */
590 	funsetownlst(&pgrp->pg_sigiolst);
591 
592 	PGRP_LOCK(pgrp);
593 	tp = pgrp->pg_session->s_ttyp;
594 	LIST_REMOVE(pgrp, pg_hash);
595 	savesess = pgrp->pg_session;
596 	PGRP_UNLOCK(pgrp);
597 
598 	/* Remove the reference to the pgrp before deallocating it. */
599 	if (tp != NULL) {
600 		tty_lock(tp);
601 		tty_rel_pgrp(tp, pgrp);
602 	}
603 
604 	mtx_destroy(&pgrp->pg_mtx);
605 	free(pgrp, M_PGRP);
606 	sess_release(savesess);
607 }
608 
609 static void
610 pgadjustjobc(pgrp, entering)
611 	struct pgrp *pgrp;
612 	int entering;
613 {
614 
615 	PGRP_LOCK(pgrp);
616 	if (entering)
617 		pgrp->pg_jobc++;
618 	else {
619 		--pgrp->pg_jobc;
620 		if (pgrp->pg_jobc == 0)
621 			orphanpg(pgrp);
622 	}
623 	PGRP_UNLOCK(pgrp);
624 }
625 
626 /*
627  * Adjust pgrp jobc counters when specified process changes process group.
628  * We count the number of processes in each process group that "qualify"
629  * the group for terminal job control (those with a parent in a different
630  * process group of the same session).  If that count reaches zero, the
631  * process group becomes orphaned.  Check both the specified process'
632  * process group and that of its children.
633  * entering == 0 => p is leaving specified group.
634  * entering == 1 => p is entering specified group.
635  */
636 void
637 fixjobc(p, pgrp, entering)
638 	register struct proc *p;
639 	register struct pgrp *pgrp;
640 	int entering;
641 {
642 	register struct pgrp *hispgrp;
643 	register struct session *mysession;
644 
645 	sx_assert(&proctree_lock, SX_LOCKED);
646 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
647 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
648 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
649 
650 	/*
651 	 * Check p's parent to see whether p qualifies its own process
652 	 * group; if so, adjust count for p's process group.
653 	 */
654 	mysession = pgrp->pg_session;
655 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
656 	    hispgrp->pg_session == mysession)
657 		pgadjustjobc(pgrp, entering);
658 
659 	/*
660 	 * Check this process' children to see whether they qualify
661 	 * their process groups; if so, adjust counts for children's
662 	 * process groups.
663 	 */
664 	LIST_FOREACH(p, &p->p_children, p_sibling) {
665 		hispgrp = p->p_pgrp;
666 		if (hispgrp == pgrp ||
667 		    hispgrp->pg_session != mysession)
668 			continue;
669 		PROC_LOCK(p);
670 		if (p->p_state == PRS_ZOMBIE) {
671 			PROC_UNLOCK(p);
672 			continue;
673 		}
674 		PROC_UNLOCK(p);
675 		pgadjustjobc(hispgrp, entering);
676 	}
677 }
678 
679 /*
680  * A process group has become orphaned;
681  * if there are any stopped processes in the group,
682  * hang-up all process in that group.
683  */
684 static void
685 orphanpg(pg)
686 	struct pgrp *pg;
687 {
688 	register struct proc *p;
689 
690 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
691 
692 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
693 		PROC_LOCK(p);
694 		if (P_SHOULDSTOP(p)) {
695 			PROC_UNLOCK(p);
696 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
697 				PROC_LOCK(p);
698 				kern_psignal(p, SIGHUP);
699 				kern_psignal(p, SIGCONT);
700 				PROC_UNLOCK(p);
701 			}
702 			return;
703 		}
704 		PROC_UNLOCK(p);
705 	}
706 }
707 
708 void
709 sess_hold(struct session *s)
710 {
711 
712 	refcount_acquire(&s->s_count);
713 }
714 
715 void
716 sess_release(struct session *s)
717 {
718 
719 	if (refcount_release(&s->s_count)) {
720 		if (s->s_ttyp != NULL) {
721 			tty_lock(s->s_ttyp);
722 			tty_rel_sess(s->s_ttyp, s);
723 		}
724 		mtx_destroy(&s->s_mtx);
725 		free(s, M_SESSION);
726 	}
727 }
728 
729 #ifdef DDB
730 
731 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
732 {
733 	register struct pgrp *pgrp;
734 	register struct proc *p;
735 	register int i;
736 
737 	for (i = 0; i <= pgrphash; i++) {
738 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
739 			printf("\tindx %d\n", i);
740 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
741 				printf(
742 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
743 				    (void *)pgrp, (long)pgrp->pg_id,
744 				    (void *)pgrp->pg_session,
745 				    pgrp->pg_session->s_count,
746 				    (void *)LIST_FIRST(&pgrp->pg_members));
747 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
748 					printf("\t\tpid %ld addr %p pgrp %p\n",
749 					    (long)p->p_pid, (void *)p,
750 					    (void *)p->p_pgrp);
751 				}
752 			}
753 		}
754 	}
755 }
756 #endif /* DDB */
757 
758 /*
759  * Calculate the kinfo_proc members which contain process-wide
760  * informations.
761  * Must be called with the target process locked.
762  */
763 static void
764 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
765 {
766 	struct thread *td;
767 
768 	PROC_LOCK_ASSERT(p, MA_OWNED);
769 
770 	kp->ki_estcpu = 0;
771 	kp->ki_pctcpu = 0;
772 	FOREACH_THREAD_IN_PROC(p, td) {
773 		thread_lock(td);
774 		kp->ki_pctcpu += sched_pctcpu(td);
775 		kp->ki_estcpu += td->td_estcpu;
776 		thread_unlock(td);
777 	}
778 }
779 
780 /*
781  * Clear kinfo_proc and fill in any information that is common
782  * to all threads in the process.
783  * Must be called with the target process locked.
784  */
785 static void
786 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
787 {
788 	struct thread *td0;
789 	struct tty *tp;
790 	struct session *sp;
791 	struct ucred *cred;
792 	struct sigacts *ps;
793 
794 	/* For proc_realparent. */
795 	sx_assert(&proctree_lock, SX_LOCKED);
796 	PROC_LOCK_ASSERT(p, MA_OWNED);
797 	bzero(kp, sizeof(*kp));
798 
799 	kp->ki_structsize = sizeof(*kp);
800 	kp->ki_paddr = p;
801 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
802 	kp->ki_args = p->p_args;
803 	kp->ki_textvp = p->p_textvp;
804 #ifdef KTRACE
805 	kp->ki_tracep = p->p_tracevp;
806 	kp->ki_traceflag = p->p_traceflag;
807 #endif
808 	kp->ki_fd = p->p_fd;
809 	kp->ki_vmspace = p->p_vmspace;
810 	kp->ki_flag = p->p_flag;
811 	kp->ki_flag2 = p->p_flag2;
812 	cred = p->p_ucred;
813 	if (cred) {
814 		kp->ki_uid = cred->cr_uid;
815 		kp->ki_ruid = cred->cr_ruid;
816 		kp->ki_svuid = cred->cr_svuid;
817 		kp->ki_cr_flags = 0;
818 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
819 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
820 		/* XXX bde doesn't like KI_NGROUPS */
821 		if (cred->cr_ngroups > KI_NGROUPS) {
822 			kp->ki_ngroups = KI_NGROUPS;
823 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
824 		} else
825 			kp->ki_ngroups = cred->cr_ngroups;
826 		bcopy(cred->cr_groups, kp->ki_groups,
827 		    kp->ki_ngroups * sizeof(gid_t));
828 		kp->ki_rgid = cred->cr_rgid;
829 		kp->ki_svgid = cred->cr_svgid;
830 		/* If jailed(cred), emulate the old P_JAILED flag. */
831 		if (jailed(cred)) {
832 			kp->ki_flag |= P_JAILED;
833 			/* If inside the jail, use 0 as a jail ID. */
834 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
835 				kp->ki_jid = cred->cr_prison->pr_id;
836 		}
837 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
838 		    sizeof(kp->ki_loginclass));
839 	}
840 	ps = p->p_sigacts;
841 	if (ps) {
842 		mtx_lock(&ps->ps_mtx);
843 		kp->ki_sigignore = ps->ps_sigignore;
844 		kp->ki_sigcatch = ps->ps_sigcatch;
845 		mtx_unlock(&ps->ps_mtx);
846 	}
847 	if (p->p_state != PRS_NEW &&
848 	    p->p_state != PRS_ZOMBIE &&
849 	    p->p_vmspace != NULL) {
850 		struct vmspace *vm = p->p_vmspace;
851 
852 		kp->ki_size = vm->vm_map.size;
853 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
854 		FOREACH_THREAD_IN_PROC(p, td0) {
855 			if (!TD_IS_SWAPPED(td0))
856 				kp->ki_rssize += td0->td_kstack_pages;
857 		}
858 		kp->ki_swrss = vm->vm_swrss;
859 		kp->ki_tsize = vm->vm_tsize;
860 		kp->ki_dsize = vm->vm_dsize;
861 		kp->ki_ssize = vm->vm_ssize;
862 	} else if (p->p_state == PRS_ZOMBIE)
863 		kp->ki_stat = SZOMB;
864 	if (kp->ki_flag & P_INMEM)
865 		kp->ki_sflag = PS_INMEM;
866 	else
867 		kp->ki_sflag = 0;
868 	/* Calculate legacy swtime as seconds since 'swtick'. */
869 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
870 	kp->ki_pid = p->p_pid;
871 	kp->ki_nice = p->p_nice;
872 	kp->ki_fibnum = p->p_fibnum;
873 	kp->ki_start = p->p_stats->p_start;
874 	timevaladd(&kp->ki_start, &boottime);
875 	PROC_SLOCK(p);
876 	rufetch(p, &kp->ki_rusage);
877 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
878 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
879 	PROC_SUNLOCK(p);
880 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
881 	/* Some callers want child times in a single value. */
882 	kp->ki_childtime = kp->ki_childstime;
883 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
884 
885 	FOREACH_THREAD_IN_PROC(p, td0)
886 		kp->ki_cow += td0->td_cow;
887 
888 	tp = NULL;
889 	if (p->p_pgrp) {
890 		kp->ki_pgid = p->p_pgrp->pg_id;
891 		kp->ki_jobc = p->p_pgrp->pg_jobc;
892 		sp = p->p_pgrp->pg_session;
893 
894 		if (sp != NULL) {
895 			kp->ki_sid = sp->s_sid;
896 			SESS_LOCK(sp);
897 			strlcpy(kp->ki_login, sp->s_login,
898 			    sizeof(kp->ki_login));
899 			if (sp->s_ttyvp)
900 				kp->ki_kiflag |= KI_CTTY;
901 			if (SESS_LEADER(p))
902 				kp->ki_kiflag |= KI_SLEADER;
903 			/* XXX proctree_lock */
904 			tp = sp->s_ttyp;
905 			SESS_UNLOCK(sp);
906 		}
907 	}
908 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
909 		kp->ki_tdev = tty_udev(tp);
910 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
911 		if (tp->t_session)
912 			kp->ki_tsid = tp->t_session->s_sid;
913 	} else
914 		kp->ki_tdev = NODEV;
915 	if (p->p_comm[0] != '\0')
916 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
917 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
918 	    p->p_sysent->sv_name[0] != '\0')
919 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
920 	kp->ki_siglist = p->p_siglist;
921 	kp->ki_xstat = p->p_xstat;
922 	kp->ki_acflag = p->p_acflag;
923 	kp->ki_lock = p->p_lock;
924 	if (p->p_pptr)
925 		kp->ki_ppid = proc_realparent(p)->p_pid;
926 	if (p->p_flag & P_TRACED)
927 		kp->ki_tracer = p->p_pptr->p_pid;
928 }
929 
930 /*
931  * Fill in information that is thread specific.  Must be called with
932  * target process locked.  If 'preferthread' is set, overwrite certain
933  * process-related fields that are maintained for both threads and
934  * processes.
935  */
936 static void
937 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
938 {
939 	struct proc *p;
940 
941 	p = td->td_proc;
942 	kp->ki_tdaddr = td;
943 	PROC_LOCK_ASSERT(p, MA_OWNED);
944 
945 	if (preferthread)
946 		PROC_SLOCK(p);
947 	thread_lock(td);
948 	if (td->td_wmesg != NULL)
949 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
950 	else
951 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
952 	strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname));
953 	if (TD_ON_LOCK(td)) {
954 		kp->ki_kiflag |= KI_LOCKBLOCK;
955 		strlcpy(kp->ki_lockname, td->td_lockname,
956 		    sizeof(kp->ki_lockname));
957 	} else {
958 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
959 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
960 	}
961 
962 	if (p->p_state == PRS_NORMAL) { /* approximate. */
963 		if (TD_ON_RUNQ(td) ||
964 		    TD_CAN_RUN(td) ||
965 		    TD_IS_RUNNING(td)) {
966 			kp->ki_stat = SRUN;
967 		} else if (P_SHOULDSTOP(p)) {
968 			kp->ki_stat = SSTOP;
969 		} else if (TD_IS_SLEEPING(td)) {
970 			kp->ki_stat = SSLEEP;
971 		} else if (TD_ON_LOCK(td)) {
972 			kp->ki_stat = SLOCK;
973 		} else {
974 			kp->ki_stat = SWAIT;
975 		}
976 	} else if (p->p_state == PRS_ZOMBIE) {
977 		kp->ki_stat = SZOMB;
978 	} else {
979 		kp->ki_stat = SIDL;
980 	}
981 
982 	/* Things in the thread */
983 	kp->ki_wchan = td->td_wchan;
984 	kp->ki_pri.pri_level = td->td_priority;
985 	kp->ki_pri.pri_native = td->td_base_pri;
986 	kp->ki_lastcpu = td->td_lastcpu;
987 	kp->ki_oncpu = td->td_oncpu;
988 	kp->ki_tdflags = td->td_flags;
989 	kp->ki_tid = td->td_tid;
990 	kp->ki_numthreads = p->p_numthreads;
991 	kp->ki_pcb = td->td_pcb;
992 	kp->ki_kstack = (void *)td->td_kstack;
993 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
994 	kp->ki_pri.pri_class = td->td_pri_class;
995 	kp->ki_pri.pri_user = td->td_user_pri;
996 
997 	if (preferthread) {
998 		rufetchtd(td, &kp->ki_rusage);
999 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1000 		kp->ki_pctcpu = sched_pctcpu(td);
1001 		kp->ki_estcpu = td->td_estcpu;
1002 		kp->ki_cow = td->td_cow;
1003 	}
1004 
1005 	/* We can't get this anymore but ps etc never used it anyway. */
1006 	kp->ki_rqindex = 0;
1007 
1008 	if (preferthread)
1009 		kp->ki_siglist = td->td_siglist;
1010 	kp->ki_sigmask = td->td_sigmask;
1011 	thread_unlock(td);
1012 	if (preferthread)
1013 		PROC_SUNLOCK(p);
1014 }
1015 
1016 /*
1017  * Fill in a kinfo_proc structure for the specified process.
1018  * Must be called with the target process locked.
1019  */
1020 void
1021 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1022 {
1023 
1024 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1025 
1026 	fill_kinfo_proc_only(p, kp);
1027 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1028 	fill_kinfo_aggregate(p, kp);
1029 }
1030 
1031 struct pstats *
1032 pstats_alloc(void)
1033 {
1034 
1035 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1036 }
1037 
1038 /*
1039  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1040  */
1041 void
1042 pstats_fork(struct pstats *src, struct pstats *dst)
1043 {
1044 
1045 	bzero(&dst->pstat_startzero,
1046 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1047 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1048 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1049 }
1050 
1051 void
1052 pstats_free(struct pstats *ps)
1053 {
1054 
1055 	free(ps, M_SUBPROC);
1056 }
1057 
1058 static struct proc *
1059 zpfind_locked(pid_t pid)
1060 {
1061 	struct proc *p;
1062 
1063 	sx_assert(&allproc_lock, SX_LOCKED);
1064 	LIST_FOREACH(p, &zombproc, p_list) {
1065 		if (p->p_pid == pid) {
1066 			PROC_LOCK(p);
1067 			break;
1068 		}
1069 	}
1070 	return (p);
1071 }
1072 
1073 /*
1074  * Locate a zombie process by number
1075  */
1076 struct proc *
1077 zpfind(pid_t pid)
1078 {
1079 	struct proc *p;
1080 
1081 	sx_slock(&allproc_lock);
1082 	p = zpfind_locked(pid);
1083 	sx_sunlock(&allproc_lock);
1084 	return (p);
1085 }
1086 
1087 #ifdef COMPAT_FREEBSD32
1088 
1089 /*
1090  * This function is typically used to copy out the kernel address, so
1091  * it can be replaced by assignment of zero.
1092  */
1093 static inline uint32_t
1094 ptr32_trim(void *ptr)
1095 {
1096 	uintptr_t uptr;
1097 
1098 	uptr = (uintptr_t)ptr;
1099 	return ((uptr > UINT_MAX) ? 0 : uptr);
1100 }
1101 
1102 #define PTRTRIM_CP(src,dst,fld) \
1103 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1104 
1105 static void
1106 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1107 {
1108 	int i;
1109 
1110 	bzero(ki32, sizeof(struct kinfo_proc32));
1111 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1112 	CP(*ki, *ki32, ki_layout);
1113 	PTRTRIM_CP(*ki, *ki32, ki_args);
1114 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1115 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1116 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1117 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1118 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1119 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1120 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1121 	CP(*ki, *ki32, ki_pid);
1122 	CP(*ki, *ki32, ki_ppid);
1123 	CP(*ki, *ki32, ki_pgid);
1124 	CP(*ki, *ki32, ki_tpgid);
1125 	CP(*ki, *ki32, ki_sid);
1126 	CP(*ki, *ki32, ki_tsid);
1127 	CP(*ki, *ki32, ki_jobc);
1128 	CP(*ki, *ki32, ki_tdev);
1129 	CP(*ki, *ki32, ki_siglist);
1130 	CP(*ki, *ki32, ki_sigmask);
1131 	CP(*ki, *ki32, ki_sigignore);
1132 	CP(*ki, *ki32, ki_sigcatch);
1133 	CP(*ki, *ki32, ki_uid);
1134 	CP(*ki, *ki32, ki_ruid);
1135 	CP(*ki, *ki32, ki_svuid);
1136 	CP(*ki, *ki32, ki_rgid);
1137 	CP(*ki, *ki32, ki_svgid);
1138 	CP(*ki, *ki32, ki_ngroups);
1139 	for (i = 0; i < KI_NGROUPS; i++)
1140 		CP(*ki, *ki32, ki_groups[i]);
1141 	CP(*ki, *ki32, ki_size);
1142 	CP(*ki, *ki32, ki_rssize);
1143 	CP(*ki, *ki32, ki_swrss);
1144 	CP(*ki, *ki32, ki_tsize);
1145 	CP(*ki, *ki32, ki_dsize);
1146 	CP(*ki, *ki32, ki_ssize);
1147 	CP(*ki, *ki32, ki_xstat);
1148 	CP(*ki, *ki32, ki_acflag);
1149 	CP(*ki, *ki32, ki_pctcpu);
1150 	CP(*ki, *ki32, ki_estcpu);
1151 	CP(*ki, *ki32, ki_slptime);
1152 	CP(*ki, *ki32, ki_swtime);
1153 	CP(*ki, *ki32, ki_cow);
1154 	CP(*ki, *ki32, ki_runtime);
1155 	TV_CP(*ki, *ki32, ki_start);
1156 	TV_CP(*ki, *ki32, ki_childtime);
1157 	CP(*ki, *ki32, ki_flag);
1158 	CP(*ki, *ki32, ki_kiflag);
1159 	CP(*ki, *ki32, ki_traceflag);
1160 	CP(*ki, *ki32, ki_stat);
1161 	CP(*ki, *ki32, ki_nice);
1162 	CP(*ki, *ki32, ki_lock);
1163 	CP(*ki, *ki32, ki_rqindex);
1164 	CP(*ki, *ki32, ki_oncpu);
1165 	CP(*ki, *ki32, ki_lastcpu);
1166 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1167 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1168 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1169 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1170 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1171 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1172 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1173 	CP(*ki, *ki32, ki_tracer);
1174 	CP(*ki, *ki32, ki_flag2);
1175 	CP(*ki, *ki32, ki_fibnum);
1176 	CP(*ki, *ki32, ki_cr_flags);
1177 	CP(*ki, *ki32, ki_jid);
1178 	CP(*ki, *ki32, ki_numthreads);
1179 	CP(*ki, *ki32, ki_tid);
1180 	CP(*ki, *ki32, ki_pri);
1181 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1182 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1183 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1184 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1185 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1186 	CP(*ki, *ki32, ki_sflag);
1187 	CP(*ki, *ki32, ki_tdflags);
1188 }
1189 #endif
1190 
1191 int
1192 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1193 {
1194 	struct thread *td;
1195 	struct kinfo_proc ki;
1196 #ifdef COMPAT_FREEBSD32
1197 	struct kinfo_proc32 ki32;
1198 #endif
1199 	int error;
1200 
1201 	PROC_LOCK_ASSERT(p, MA_OWNED);
1202 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1203 
1204 	error = 0;
1205 	fill_kinfo_proc(p, &ki);
1206 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1207 #ifdef COMPAT_FREEBSD32
1208 		if ((flags & KERN_PROC_MASK32) != 0) {
1209 			freebsd32_kinfo_proc_out(&ki, &ki32);
1210 			error = sbuf_bcat(sb, &ki32, sizeof(ki32));
1211 		} else
1212 #endif
1213 			error = sbuf_bcat(sb, &ki, sizeof(ki));
1214 	} else {
1215 		FOREACH_THREAD_IN_PROC(p, td) {
1216 			fill_kinfo_thread(td, &ki, 1);
1217 #ifdef COMPAT_FREEBSD32
1218 			if ((flags & KERN_PROC_MASK32) != 0) {
1219 				freebsd32_kinfo_proc_out(&ki, &ki32);
1220 				error = sbuf_bcat(sb, &ki32, sizeof(ki32));
1221 			} else
1222 #endif
1223 				error = sbuf_bcat(sb, &ki, sizeof(ki));
1224 			if (error)
1225 				break;
1226 		}
1227 	}
1228 	PROC_UNLOCK(p);
1229 	return (error);
1230 }
1231 
1232 static int
1233 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags,
1234     int doingzomb)
1235 {
1236 	struct sbuf sb;
1237 	struct kinfo_proc ki;
1238 	struct proc *np;
1239 	int error, error2;
1240 	pid_t pid;
1241 
1242 	pid = p->p_pid;
1243 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1244 	error = kern_proc_out(p, &sb, flags);
1245 	error2 = sbuf_finish(&sb);
1246 	sbuf_delete(&sb);
1247 	if (error != 0)
1248 		return (error);
1249 	else if (error2 != 0)
1250 		return (error2);
1251 	if (doingzomb)
1252 		np = zpfind(pid);
1253 	else {
1254 		if (pid == 0)
1255 			return (0);
1256 		np = pfind(pid);
1257 	}
1258 	if (np == NULL)
1259 		return (ESRCH);
1260 	if (np != p) {
1261 		PROC_UNLOCK(np);
1262 		return (ESRCH);
1263 	}
1264 	PROC_UNLOCK(np);
1265 	return (0);
1266 }
1267 
1268 static int
1269 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1270 {
1271 	int *name = (int *)arg1;
1272 	u_int namelen = arg2;
1273 	struct proc *p;
1274 	int flags, doingzomb, oid_number;
1275 	int error = 0;
1276 
1277 	oid_number = oidp->oid_number;
1278 	if (oid_number != KERN_PROC_ALL &&
1279 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1280 		flags = KERN_PROC_NOTHREADS;
1281 	else {
1282 		flags = 0;
1283 		oid_number &= ~KERN_PROC_INC_THREAD;
1284 	}
1285 #ifdef COMPAT_FREEBSD32
1286 	if (req->flags & SCTL_MASK32)
1287 		flags |= KERN_PROC_MASK32;
1288 #endif
1289 	if (oid_number == KERN_PROC_PID) {
1290 		if (namelen != 1)
1291 			return (EINVAL);
1292 		error = sysctl_wire_old_buffer(req, 0);
1293 		if (error)
1294 			return (error);
1295 		sx_slock(&proctree_lock);
1296 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1297 		if (error == 0)
1298 			error = sysctl_out_proc(p, req, flags, 0);
1299 		sx_sunlock(&proctree_lock);
1300 		return (error);
1301 	}
1302 
1303 	switch (oid_number) {
1304 	case KERN_PROC_ALL:
1305 		if (namelen != 0)
1306 			return (EINVAL);
1307 		break;
1308 	case KERN_PROC_PROC:
1309 		if (namelen != 0 && namelen != 1)
1310 			return (EINVAL);
1311 		break;
1312 	default:
1313 		if (namelen != 1)
1314 			return (EINVAL);
1315 		break;
1316 	}
1317 
1318 	if (!req->oldptr) {
1319 		/* overestimate by 5 procs */
1320 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1321 		if (error)
1322 			return (error);
1323 	}
1324 	error = sysctl_wire_old_buffer(req, 0);
1325 	if (error != 0)
1326 		return (error);
1327 	sx_slock(&proctree_lock);
1328 	sx_slock(&allproc_lock);
1329 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1330 		if (!doingzomb)
1331 			p = LIST_FIRST(&allproc);
1332 		else
1333 			p = LIST_FIRST(&zombproc);
1334 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
1335 			/*
1336 			 * Skip embryonic processes.
1337 			 */
1338 			PROC_LOCK(p);
1339 			if (p->p_state == PRS_NEW) {
1340 				PROC_UNLOCK(p);
1341 				continue;
1342 			}
1343 			KASSERT(p->p_ucred != NULL,
1344 			    ("process credential is NULL for non-NEW proc"));
1345 			/*
1346 			 * Show a user only appropriate processes.
1347 			 */
1348 			if (p_cansee(curthread, p)) {
1349 				PROC_UNLOCK(p);
1350 				continue;
1351 			}
1352 			/*
1353 			 * TODO - make more efficient (see notes below).
1354 			 * do by session.
1355 			 */
1356 			switch (oid_number) {
1357 
1358 			case KERN_PROC_GID:
1359 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1360 					PROC_UNLOCK(p);
1361 					continue;
1362 				}
1363 				break;
1364 
1365 			case KERN_PROC_PGRP:
1366 				/* could do this by traversing pgrp */
1367 				if (p->p_pgrp == NULL ||
1368 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1369 					PROC_UNLOCK(p);
1370 					continue;
1371 				}
1372 				break;
1373 
1374 			case KERN_PROC_RGID:
1375 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1376 					PROC_UNLOCK(p);
1377 					continue;
1378 				}
1379 				break;
1380 
1381 			case KERN_PROC_SESSION:
1382 				if (p->p_session == NULL ||
1383 				    p->p_session->s_sid != (pid_t)name[0]) {
1384 					PROC_UNLOCK(p);
1385 					continue;
1386 				}
1387 				break;
1388 
1389 			case KERN_PROC_TTY:
1390 				if ((p->p_flag & P_CONTROLT) == 0 ||
1391 				    p->p_session == NULL) {
1392 					PROC_UNLOCK(p);
1393 					continue;
1394 				}
1395 				/* XXX proctree_lock */
1396 				SESS_LOCK(p->p_session);
1397 				if (p->p_session->s_ttyp == NULL ||
1398 				    tty_udev(p->p_session->s_ttyp) !=
1399 				    (dev_t)name[0]) {
1400 					SESS_UNLOCK(p->p_session);
1401 					PROC_UNLOCK(p);
1402 					continue;
1403 				}
1404 				SESS_UNLOCK(p->p_session);
1405 				break;
1406 
1407 			case KERN_PROC_UID:
1408 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1409 					PROC_UNLOCK(p);
1410 					continue;
1411 				}
1412 				break;
1413 
1414 			case KERN_PROC_RUID:
1415 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1416 					PROC_UNLOCK(p);
1417 					continue;
1418 				}
1419 				break;
1420 
1421 			case KERN_PROC_PROC:
1422 				break;
1423 
1424 			default:
1425 				break;
1426 
1427 			}
1428 
1429 			error = sysctl_out_proc(p, req, flags, doingzomb);
1430 			if (error) {
1431 				sx_sunlock(&allproc_lock);
1432 				sx_sunlock(&proctree_lock);
1433 				return (error);
1434 			}
1435 		}
1436 	}
1437 	sx_sunlock(&allproc_lock);
1438 	sx_sunlock(&proctree_lock);
1439 	return (0);
1440 }
1441 
1442 struct pargs *
1443 pargs_alloc(int len)
1444 {
1445 	struct pargs *pa;
1446 
1447 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1448 		M_WAITOK);
1449 	refcount_init(&pa->ar_ref, 1);
1450 	pa->ar_length = len;
1451 	return (pa);
1452 }
1453 
1454 static void
1455 pargs_free(struct pargs *pa)
1456 {
1457 
1458 	free(pa, M_PARGS);
1459 }
1460 
1461 void
1462 pargs_hold(struct pargs *pa)
1463 {
1464 
1465 	if (pa == NULL)
1466 		return;
1467 	refcount_acquire(&pa->ar_ref);
1468 }
1469 
1470 void
1471 pargs_drop(struct pargs *pa)
1472 {
1473 
1474 	if (pa == NULL)
1475 		return;
1476 	if (refcount_release(&pa->ar_ref))
1477 		pargs_free(pa);
1478 }
1479 
1480 static int
1481 proc_read_mem(struct thread *td, struct proc *p, vm_offset_t offset, void* buf,
1482     size_t len)
1483 {
1484 	struct iovec iov;
1485 	struct uio uio;
1486 
1487 	iov.iov_base = (caddr_t)buf;
1488 	iov.iov_len = len;
1489 	uio.uio_iov = &iov;
1490 	uio.uio_iovcnt = 1;
1491 	uio.uio_offset = offset;
1492 	uio.uio_resid = (ssize_t)len;
1493 	uio.uio_segflg = UIO_SYSSPACE;
1494 	uio.uio_rw = UIO_READ;
1495 	uio.uio_td = td;
1496 
1497 	return (proc_rwmem(p, &uio));
1498 }
1499 
1500 static int
1501 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1502     size_t len)
1503 {
1504 	size_t i;
1505 	int error;
1506 
1507 	error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, len);
1508 	/*
1509 	 * Reading the chunk may validly return EFAULT if the string is shorter
1510 	 * than the chunk and is aligned at the end of the page, assuming the
1511 	 * next page is not mapped.  So if EFAULT is returned do a fallback to
1512 	 * one byte read loop.
1513 	 */
1514 	if (error == EFAULT) {
1515 		for (i = 0; i < len; i++, buf++, sptr++) {
1516 			error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, 1);
1517 			if (error != 0)
1518 				return (error);
1519 			if (*buf == '\0')
1520 				break;
1521 		}
1522 		error = 0;
1523 	}
1524 	return (error);
1525 }
1526 
1527 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1528 
1529 enum proc_vector_type {
1530 	PROC_ARG,
1531 	PROC_ENV,
1532 	PROC_AUX,
1533 };
1534 
1535 #ifdef COMPAT_FREEBSD32
1536 static int
1537 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1538     size_t *vsizep, enum proc_vector_type type)
1539 {
1540 	struct freebsd32_ps_strings pss;
1541 	Elf32_Auxinfo aux;
1542 	vm_offset_t vptr, ptr;
1543 	uint32_t *proc_vector32;
1544 	char **proc_vector;
1545 	size_t vsize, size;
1546 	int i, error;
1547 
1548 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1549 	    &pss, sizeof(pss));
1550 	if (error != 0)
1551 		return (error);
1552 	switch (type) {
1553 	case PROC_ARG:
1554 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1555 		vsize = pss.ps_nargvstr;
1556 		if (vsize > ARG_MAX)
1557 			return (ENOEXEC);
1558 		size = vsize * sizeof(int32_t);
1559 		break;
1560 	case PROC_ENV:
1561 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1562 		vsize = pss.ps_nenvstr;
1563 		if (vsize > ARG_MAX)
1564 			return (ENOEXEC);
1565 		size = vsize * sizeof(int32_t);
1566 		break;
1567 	case PROC_AUX:
1568 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1569 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1570 		if (vptr % 4 != 0)
1571 			return (ENOEXEC);
1572 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1573 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1574 			if (error != 0)
1575 				return (error);
1576 			if (aux.a_type == AT_NULL)
1577 				break;
1578 			ptr += sizeof(aux);
1579 		}
1580 		if (aux.a_type != AT_NULL)
1581 			return (ENOEXEC);
1582 		vsize = i + 1;
1583 		size = vsize * sizeof(aux);
1584 		break;
1585 	default:
1586 		KASSERT(0, ("Wrong proc vector type: %d", type));
1587 		return (EINVAL);
1588 	}
1589 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1590 	error = proc_read_mem(td, p, vptr, proc_vector32, size);
1591 	if (error != 0)
1592 		goto done;
1593 	if (type == PROC_AUX) {
1594 		*proc_vectorp = (char **)proc_vector32;
1595 		*vsizep = vsize;
1596 		return (0);
1597 	}
1598 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1599 	for (i = 0; i < (int)vsize; i++)
1600 		proc_vector[i] = PTRIN(proc_vector32[i]);
1601 	*proc_vectorp = proc_vector;
1602 	*vsizep = vsize;
1603 done:
1604 	free(proc_vector32, M_TEMP);
1605 	return (error);
1606 }
1607 #endif
1608 
1609 static int
1610 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1611     size_t *vsizep, enum proc_vector_type type)
1612 {
1613 	struct ps_strings pss;
1614 	Elf_Auxinfo aux;
1615 	vm_offset_t vptr, ptr;
1616 	char **proc_vector;
1617 	size_t vsize, size;
1618 	int error, i;
1619 
1620 #ifdef COMPAT_FREEBSD32
1621 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1622 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1623 #endif
1624 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1625 	    &pss, sizeof(pss));
1626 	if (error != 0)
1627 		return (error);
1628 	switch (type) {
1629 	case PROC_ARG:
1630 		vptr = (vm_offset_t)pss.ps_argvstr;
1631 		vsize = pss.ps_nargvstr;
1632 		if (vsize > ARG_MAX)
1633 			return (ENOEXEC);
1634 		size = vsize * sizeof(char *);
1635 		break;
1636 	case PROC_ENV:
1637 		vptr = (vm_offset_t)pss.ps_envstr;
1638 		vsize = pss.ps_nenvstr;
1639 		if (vsize > ARG_MAX)
1640 			return (ENOEXEC);
1641 		size = vsize * sizeof(char *);
1642 		break;
1643 	case PROC_AUX:
1644 		/*
1645 		 * The aux array is just above env array on the stack. Check
1646 		 * that the address is naturally aligned.
1647 		 */
1648 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1649 		    * sizeof(char *);
1650 #if __ELF_WORD_SIZE == 64
1651 		if (vptr % sizeof(uint64_t) != 0)
1652 #else
1653 		if (vptr % sizeof(uint32_t) != 0)
1654 #endif
1655 			return (ENOEXEC);
1656 		/*
1657 		 * We count the array size reading the aux vectors from the
1658 		 * stack until AT_NULL vector is returned.  So (to keep the code
1659 		 * simple) we read the process stack twice: the first time here
1660 		 * to find the size and the second time when copying the vectors
1661 		 * to the allocated proc_vector.
1662 		 */
1663 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1664 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1665 			if (error != 0)
1666 				return (error);
1667 			if (aux.a_type == AT_NULL)
1668 				break;
1669 			ptr += sizeof(aux);
1670 		}
1671 		/*
1672 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1673 		 * not reached AT_NULL, it is most likely we are reading wrong
1674 		 * data: either the process doesn't have auxv array or data has
1675 		 * been modified. Return the error in this case.
1676 		 */
1677 		if (aux.a_type != AT_NULL)
1678 			return (ENOEXEC);
1679 		vsize = i + 1;
1680 		size = vsize * sizeof(aux);
1681 		break;
1682 	default:
1683 		KASSERT(0, ("Wrong proc vector type: %d", type));
1684 		return (EINVAL); /* In case we are built without INVARIANTS. */
1685 	}
1686 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1687 	if (proc_vector == NULL)
1688 		return (ENOMEM);
1689 	error = proc_read_mem(td, p, vptr, proc_vector, size);
1690 	if (error != 0) {
1691 		free(proc_vector, M_TEMP);
1692 		return (error);
1693 	}
1694 	*proc_vectorp = proc_vector;
1695 	*vsizep = vsize;
1696 
1697 	return (0);
1698 }
1699 
1700 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1701 
1702 static int
1703 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1704     enum proc_vector_type type)
1705 {
1706 	size_t done, len, nchr, vsize;
1707 	int error, i;
1708 	char **proc_vector, *sptr;
1709 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1710 
1711 	PROC_ASSERT_HELD(p);
1712 
1713 	/*
1714 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1715 	 */
1716 	nchr = 2 * (PATH_MAX + ARG_MAX);
1717 
1718 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1719 	if (error != 0)
1720 		return (error);
1721 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1722 		/*
1723 		 * The program may have scribbled into its argv array, e.g. to
1724 		 * remove some arguments.  If that has happened, break out
1725 		 * before trying to read from NULL.
1726 		 */
1727 		if (proc_vector[i] == NULL)
1728 			break;
1729 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1730 			error = proc_read_string(td, p, sptr, pss_string,
1731 			    sizeof(pss_string));
1732 			if (error != 0)
1733 				goto done;
1734 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1735 			if (done + len >= nchr)
1736 				len = nchr - done - 1;
1737 			sbuf_bcat(sb, pss_string, len);
1738 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1739 				break;
1740 			done += GET_PS_STRINGS_CHUNK_SZ;
1741 		}
1742 		sbuf_bcat(sb, "", 1);
1743 		done += len + 1;
1744 	}
1745 done:
1746 	free(proc_vector, M_TEMP);
1747 	return (error);
1748 }
1749 
1750 int
1751 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1752 {
1753 
1754 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1755 }
1756 
1757 int
1758 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1759 {
1760 
1761 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1762 }
1763 
1764 int
1765 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1766 {
1767 	size_t vsize, size;
1768 	char **auxv;
1769 	int error;
1770 
1771 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1772 	if (error == 0) {
1773 #ifdef COMPAT_FREEBSD32
1774 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1775 			size = vsize * sizeof(Elf32_Auxinfo);
1776 		else
1777 #endif
1778 			size = vsize * sizeof(Elf_Auxinfo);
1779 		error = sbuf_bcat(sb, auxv, size);
1780 		free(auxv, M_TEMP);
1781 	}
1782 	return (error);
1783 }
1784 
1785 /*
1786  * This sysctl allows a process to retrieve the argument list or process
1787  * title for another process without groping around in the address space
1788  * of the other process.  It also allow a process to set its own "process
1789  * title to a string of its own choice.
1790  */
1791 static int
1792 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1793 {
1794 	int *name = (int *)arg1;
1795 	u_int namelen = arg2;
1796 	struct pargs *newpa, *pa;
1797 	struct proc *p;
1798 	struct sbuf sb;
1799 	int flags, error = 0, error2;
1800 
1801 	if (namelen != 1)
1802 		return (EINVAL);
1803 
1804 	flags = PGET_CANSEE;
1805 	if (req->newptr != NULL)
1806 		flags |= PGET_ISCURRENT;
1807 	error = pget((pid_t)name[0], flags, &p);
1808 	if (error)
1809 		return (error);
1810 
1811 	pa = p->p_args;
1812 	if (pa != NULL) {
1813 		pargs_hold(pa);
1814 		PROC_UNLOCK(p);
1815 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1816 		pargs_drop(pa);
1817 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1818 		_PHOLD(p);
1819 		PROC_UNLOCK(p);
1820 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1821 		error = proc_getargv(curthread, p, &sb);
1822 		error2 = sbuf_finish(&sb);
1823 		PRELE(p);
1824 		sbuf_delete(&sb);
1825 		if (error == 0 && error2 != 0)
1826 			error = error2;
1827 	} else {
1828 		PROC_UNLOCK(p);
1829 	}
1830 	if (error != 0 || req->newptr == NULL)
1831 		return (error);
1832 
1833 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1834 		return (ENOMEM);
1835 	newpa = pargs_alloc(req->newlen);
1836 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1837 	if (error != 0) {
1838 		pargs_free(newpa);
1839 		return (error);
1840 	}
1841 	PROC_LOCK(p);
1842 	pa = p->p_args;
1843 	p->p_args = newpa;
1844 	PROC_UNLOCK(p);
1845 	pargs_drop(pa);
1846 	return (0);
1847 }
1848 
1849 /*
1850  * This sysctl allows a process to retrieve environment of another process.
1851  */
1852 static int
1853 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
1854 {
1855 	int *name = (int *)arg1;
1856 	u_int namelen = arg2;
1857 	struct proc *p;
1858 	struct sbuf sb;
1859 	int error, error2;
1860 
1861 	if (namelen != 1)
1862 		return (EINVAL);
1863 
1864 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1865 	if (error != 0)
1866 		return (error);
1867 	if ((p->p_flag & P_SYSTEM) != 0) {
1868 		PRELE(p);
1869 		return (0);
1870 	}
1871 
1872 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1873 	error = proc_getenvv(curthread, p, &sb);
1874 	error2 = sbuf_finish(&sb);
1875 	PRELE(p);
1876 	sbuf_delete(&sb);
1877 	return (error != 0 ? error : error2);
1878 }
1879 
1880 /*
1881  * This sysctl allows a process to retrieve ELF auxiliary vector of
1882  * another process.
1883  */
1884 static int
1885 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
1886 {
1887 	int *name = (int *)arg1;
1888 	u_int namelen = arg2;
1889 	struct proc *p;
1890 	struct sbuf sb;
1891 	int error, error2;
1892 
1893 	if (namelen != 1)
1894 		return (EINVAL);
1895 
1896 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1897 	if (error != 0)
1898 		return (error);
1899 	if ((p->p_flag & P_SYSTEM) != 0) {
1900 		PRELE(p);
1901 		return (0);
1902 	}
1903 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1904 	error = proc_getauxv(curthread, p, &sb);
1905 	error2 = sbuf_finish(&sb);
1906 	PRELE(p);
1907 	sbuf_delete(&sb);
1908 	return (error != 0 ? error : error2);
1909 }
1910 
1911 /*
1912  * This sysctl allows a process to retrieve the path of the executable for
1913  * itself or another process.
1914  */
1915 static int
1916 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1917 {
1918 	pid_t *pidp = (pid_t *)arg1;
1919 	unsigned int arglen = arg2;
1920 	struct proc *p;
1921 	struct vnode *vp;
1922 	char *retbuf, *freebuf;
1923 	int error;
1924 
1925 	if (arglen != 1)
1926 		return (EINVAL);
1927 	if (*pidp == -1) {	/* -1 means this process */
1928 		p = req->td->td_proc;
1929 	} else {
1930 		error = pget(*pidp, PGET_CANSEE, &p);
1931 		if (error != 0)
1932 			return (error);
1933 	}
1934 
1935 	vp = p->p_textvp;
1936 	if (vp == NULL) {
1937 		if (*pidp != -1)
1938 			PROC_UNLOCK(p);
1939 		return (0);
1940 	}
1941 	vref(vp);
1942 	if (*pidp != -1)
1943 		PROC_UNLOCK(p);
1944 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1945 	vrele(vp);
1946 	if (error)
1947 		return (error);
1948 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1949 	free(freebuf, M_TEMP);
1950 	return (error);
1951 }
1952 
1953 static int
1954 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1955 {
1956 	struct proc *p;
1957 	char *sv_name;
1958 	int *name;
1959 	int namelen;
1960 	int error;
1961 
1962 	namelen = arg2;
1963 	if (namelen != 1)
1964 		return (EINVAL);
1965 
1966 	name = (int *)arg1;
1967 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
1968 	if (error != 0)
1969 		return (error);
1970 	sv_name = p->p_sysent->sv_name;
1971 	PROC_UNLOCK(p);
1972 	return (sysctl_handle_string(oidp, sv_name, 0, req));
1973 }
1974 
1975 #ifdef KINFO_OVMENTRY_SIZE
1976 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
1977 #endif
1978 
1979 #ifdef COMPAT_FREEBSD7
1980 static int
1981 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
1982 {
1983 	vm_map_entry_t entry, tmp_entry;
1984 	unsigned int last_timestamp;
1985 	char *fullpath, *freepath;
1986 	struct kinfo_ovmentry *kve;
1987 	struct vattr va;
1988 	struct ucred *cred;
1989 	int error, *name;
1990 	struct vnode *vp;
1991 	struct proc *p;
1992 	vm_map_t map;
1993 	struct vmspace *vm;
1994 
1995 	name = (int *)arg1;
1996 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1997 	if (error != 0)
1998 		return (error);
1999 	vm = vmspace_acquire_ref(p);
2000 	if (vm == NULL) {
2001 		PRELE(p);
2002 		return (ESRCH);
2003 	}
2004 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2005 
2006 	map = &vm->vm_map;
2007 	vm_map_lock_read(map);
2008 	for (entry = map->header.next; entry != &map->header;
2009 	    entry = entry->next) {
2010 		vm_object_t obj, tobj, lobj;
2011 		vm_offset_t addr;
2012 
2013 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2014 			continue;
2015 
2016 		bzero(kve, sizeof(*kve));
2017 		kve->kve_structsize = sizeof(*kve);
2018 
2019 		kve->kve_private_resident = 0;
2020 		obj = entry->object.vm_object;
2021 		if (obj != NULL) {
2022 			VM_OBJECT_RLOCK(obj);
2023 			if (obj->shadow_count == 1)
2024 				kve->kve_private_resident =
2025 				    obj->resident_page_count;
2026 		}
2027 		kve->kve_resident = 0;
2028 		addr = entry->start;
2029 		while (addr < entry->end) {
2030 			if (pmap_extract(map->pmap, addr))
2031 				kve->kve_resident++;
2032 			addr += PAGE_SIZE;
2033 		}
2034 
2035 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2036 			if (tobj != obj)
2037 				VM_OBJECT_RLOCK(tobj);
2038 			if (lobj != obj)
2039 				VM_OBJECT_RUNLOCK(lobj);
2040 			lobj = tobj;
2041 		}
2042 
2043 		kve->kve_start = (void*)entry->start;
2044 		kve->kve_end = (void*)entry->end;
2045 		kve->kve_offset = (off_t)entry->offset;
2046 
2047 		if (entry->protection & VM_PROT_READ)
2048 			kve->kve_protection |= KVME_PROT_READ;
2049 		if (entry->protection & VM_PROT_WRITE)
2050 			kve->kve_protection |= KVME_PROT_WRITE;
2051 		if (entry->protection & VM_PROT_EXECUTE)
2052 			kve->kve_protection |= KVME_PROT_EXEC;
2053 
2054 		if (entry->eflags & MAP_ENTRY_COW)
2055 			kve->kve_flags |= KVME_FLAG_COW;
2056 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2057 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2058 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2059 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2060 
2061 		last_timestamp = map->timestamp;
2062 		vm_map_unlock_read(map);
2063 
2064 		kve->kve_fileid = 0;
2065 		kve->kve_fsid = 0;
2066 		freepath = NULL;
2067 		fullpath = "";
2068 		if (lobj) {
2069 			vp = NULL;
2070 			switch (lobj->type) {
2071 			case OBJT_DEFAULT:
2072 				kve->kve_type = KVME_TYPE_DEFAULT;
2073 				break;
2074 			case OBJT_VNODE:
2075 				kve->kve_type = KVME_TYPE_VNODE;
2076 				vp = lobj->handle;
2077 				vref(vp);
2078 				break;
2079 			case OBJT_SWAP:
2080 				kve->kve_type = KVME_TYPE_SWAP;
2081 				break;
2082 			case OBJT_DEVICE:
2083 				kve->kve_type = KVME_TYPE_DEVICE;
2084 				break;
2085 			case OBJT_PHYS:
2086 				kve->kve_type = KVME_TYPE_PHYS;
2087 				break;
2088 			case OBJT_DEAD:
2089 				kve->kve_type = KVME_TYPE_DEAD;
2090 				break;
2091 			case OBJT_SG:
2092 				kve->kve_type = KVME_TYPE_SG;
2093 				break;
2094 			default:
2095 				kve->kve_type = KVME_TYPE_UNKNOWN;
2096 				break;
2097 			}
2098 			if (lobj != obj)
2099 				VM_OBJECT_RUNLOCK(lobj);
2100 
2101 			kve->kve_ref_count = obj->ref_count;
2102 			kve->kve_shadow_count = obj->shadow_count;
2103 			VM_OBJECT_RUNLOCK(obj);
2104 			if (vp != NULL) {
2105 				vn_fullpath(curthread, vp, &fullpath,
2106 				    &freepath);
2107 				cred = curthread->td_ucred;
2108 				vn_lock(vp, LK_SHARED | LK_RETRY);
2109 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2110 					kve->kve_fileid = va.va_fileid;
2111 					kve->kve_fsid = va.va_fsid;
2112 				}
2113 				vput(vp);
2114 			}
2115 		} else {
2116 			kve->kve_type = KVME_TYPE_NONE;
2117 			kve->kve_ref_count = 0;
2118 			kve->kve_shadow_count = 0;
2119 		}
2120 
2121 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2122 		if (freepath != NULL)
2123 			free(freepath, M_TEMP);
2124 
2125 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2126 		vm_map_lock_read(map);
2127 		if (error)
2128 			break;
2129 		if (last_timestamp != map->timestamp) {
2130 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2131 			entry = tmp_entry;
2132 		}
2133 	}
2134 	vm_map_unlock_read(map);
2135 	vmspace_free(vm);
2136 	PRELE(p);
2137 	free(kve, M_TEMP);
2138 	return (error);
2139 }
2140 #endif	/* COMPAT_FREEBSD7 */
2141 
2142 #ifdef KINFO_VMENTRY_SIZE
2143 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2144 #endif
2145 
2146 static void
2147 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2148     struct kinfo_vmentry *kve)
2149 {
2150 	vm_object_t obj, tobj;
2151 	vm_page_t m, m_adv;
2152 	vm_offset_t addr;
2153 	vm_paddr_t locked_pa;
2154 	vm_pindex_t pi, pi_adv, pindex;
2155 
2156 	locked_pa = 0;
2157 	obj = entry->object.vm_object;
2158 	addr = entry->start;
2159 	m_adv = NULL;
2160 	pi = OFF_TO_IDX(entry->offset);
2161 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2162 		if (m_adv != NULL) {
2163 			m = m_adv;
2164 		} else {
2165 			pi_adv = OFF_TO_IDX(entry->end - addr);
2166 			pindex = pi;
2167 			for (tobj = obj;; tobj = tobj->backing_object) {
2168 				m = vm_page_find_least(tobj, pindex);
2169 				if (m != NULL) {
2170 					if (m->pindex == pindex)
2171 						break;
2172 					if (pi_adv > m->pindex - pindex) {
2173 						pi_adv = m->pindex - pindex;
2174 						m_adv = m;
2175 					}
2176 				}
2177 				if (tobj->backing_object == NULL)
2178 					goto next;
2179 				pindex += OFF_TO_IDX(tobj->
2180 				    backing_object_offset);
2181 			}
2182 		}
2183 		m_adv = NULL;
2184 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2185 		    (addr & (pagesizes[1] - 1)) == 0 &&
2186 		    (pmap_mincore(map->pmap, addr, &locked_pa) &
2187 		    MINCORE_SUPER) != 0) {
2188 			kve->kve_flags |= KVME_FLAG_SUPER;
2189 			pi_adv = OFF_TO_IDX(pagesizes[1]);
2190 		} else {
2191 			/*
2192 			 * We do not test the found page on validity.
2193 			 * Either the page is busy and being paged in,
2194 			 * or it was invalidated.  The first case
2195 			 * should be counted as resident, the second
2196 			 * is not so clear; we do account both.
2197 			 */
2198 			pi_adv = 1;
2199 		}
2200 		kve->kve_resident += pi_adv;
2201 next:;
2202 	}
2203 	PA_UNLOCK_COND(locked_pa);
2204 }
2205 
2206 /*
2207  * Must be called with the process locked and will return unlocked.
2208  */
2209 int
2210 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb)
2211 {
2212 	vm_map_entry_t entry, tmp_entry;
2213 	struct vattr va;
2214 	vm_map_t map;
2215 	vm_object_t obj, tobj, lobj;
2216 	char *fullpath, *freepath;
2217 	struct kinfo_vmentry *kve;
2218 	struct ucred *cred;
2219 	struct vnode *vp;
2220 	struct vmspace *vm;
2221 	vm_offset_t addr;
2222 	unsigned int last_timestamp;
2223 	int error;
2224 
2225 	PROC_LOCK_ASSERT(p, MA_OWNED);
2226 
2227 	_PHOLD(p);
2228 	PROC_UNLOCK(p);
2229 	vm = vmspace_acquire_ref(p);
2230 	if (vm == NULL) {
2231 		PRELE(p);
2232 		return (ESRCH);
2233 	}
2234 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2235 
2236 	error = 0;
2237 	map = &vm->vm_map;
2238 	vm_map_lock_read(map);
2239 	for (entry = map->header.next; entry != &map->header;
2240 	    entry = entry->next) {
2241 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2242 			continue;
2243 
2244 		addr = entry->end;
2245 		bzero(kve, sizeof(*kve));
2246 		obj = entry->object.vm_object;
2247 		if (obj != NULL) {
2248 			for (tobj = obj; tobj != NULL;
2249 			    tobj = tobj->backing_object) {
2250 				VM_OBJECT_RLOCK(tobj);
2251 				lobj = tobj;
2252 			}
2253 			if (obj->backing_object == NULL)
2254 				kve->kve_private_resident =
2255 				    obj->resident_page_count;
2256 			if (!vmmap_skip_res_cnt)
2257 				kern_proc_vmmap_resident(map, entry, kve);
2258 			for (tobj = obj; tobj != NULL;
2259 			    tobj = tobj->backing_object) {
2260 				if (tobj != obj && tobj != lobj)
2261 					VM_OBJECT_RUNLOCK(tobj);
2262 			}
2263 		} else {
2264 			lobj = NULL;
2265 		}
2266 
2267 		kve->kve_start = entry->start;
2268 		kve->kve_end = entry->end;
2269 		kve->kve_offset = entry->offset;
2270 
2271 		if (entry->protection & VM_PROT_READ)
2272 			kve->kve_protection |= KVME_PROT_READ;
2273 		if (entry->protection & VM_PROT_WRITE)
2274 			kve->kve_protection |= KVME_PROT_WRITE;
2275 		if (entry->protection & VM_PROT_EXECUTE)
2276 			kve->kve_protection |= KVME_PROT_EXEC;
2277 
2278 		if (entry->eflags & MAP_ENTRY_COW)
2279 			kve->kve_flags |= KVME_FLAG_COW;
2280 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2281 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2282 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2283 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2284 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2285 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2286 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2287 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2288 
2289 		last_timestamp = map->timestamp;
2290 		vm_map_unlock_read(map);
2291 
2292 		freepath = NULL;
2293 		fullpath = "";
2294 		if (lobj != NULL) {
2295 			vp = NULL;
2296 			switch (lobj->type) {
2297 			case OBJT_DEFAULT:
2298 				kve->kve_type = KVME_TYPE_DEFAULT;
2299 				break;
2300 			case OBJT_VNODE:
2301 				kve->kve_type = KVME_TYPE_VNODE;
2302 				vp = lobj->handle;
2303 				vref(vp);
2304 				break;
2305 			case OBJT_SWAP:
2306 				kve->kve_type = KVME_TYPE_SWAP;
2307 				break;
2308 			case OBJT_DEVICE:
2309 				kve->kve_type = KVME_TYPE_DEVICE;
2310 				break;
2311 			case OBJT_PHYS:
2312 				kve->kve_type = KVME_TYPE_PHYS;
2313 				break;
2314 			case OBJT_DEAD:
2315 				kve->kve_type = KVME_TYPE_DEAD;
2316 				break;
2317 			case OBJT_SG:
2318 				kve->kve_type = KVME_TYPE_SG;
2319 				break;
2320 			case OBJT_MGTDEVICE:
2321 				kve->kve_type = KVME_TYPE_MGTDEVICE;
2322 				break;
2323 			default:
2324 				kve->kve_type = KVME_TYPE_UNKNOWN;
2325 				break;
2326 			}
2327 			if (lobj != obj)
2328 				VM_OBJECT_RUNLOCK(lobj);
2329 
2330 			kve->kve_ref_count = obj->ref_count;
2331 			kve->kve_shadow_count = obj->shadow_count;
2332 			VM_OBJECT_RUNLOCK(obj);
2333 			if (vp != NULL) {
2334 				vn_fullpath(curthread, vp, &fullpath,
2335 				    &freepath);
2336 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2337 				cred = curthread->td_ucred;
2338 				vn_lock(vp, LK_SHARED | LK_RETRY);
2339 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2340 					kve->kve_vn_fileid = va.va_fileid;
2341 					kve->kve_vn_fsid = va.va_fsid;
2342 					kve->kve_vn_mode =
2343 					    MAKEIMODE(va.va_type, va.va_mode);
2344 					kve->kve_vn_size = va.va_size;
2345 					kve->kve_vn_rdev = va.va_rdev;
2346 					kve->kve_status = KF_ATTR_VALID;
2347 				}
2348 				vput(vp);
2349 			}
2350 		} else {
2351 			kve->kve_type = KVME_TYPE_NONE;
2352 			kve->kve_ref_count = 0;
2353 			kve->kve_shadow_count = 0;
2354 		}
2355 
2356 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2357 		if (freepath != NULL)
2358 			free(freepath, M_TEMP);
2359 
2360 		/* Pack record size down */
2361 		kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) +
2362 		    strlen(kve->kve_path) + 1;
2363 		kve->kve_structsize = roundup(kve->kve_structsize,
2364 		    sizeof(uint64_t));
2365 		error = sbuf_bcat(sb, kve, kve->kve_structsize);
2366 		vm_map_lock_read(map);
2367 		if (error)
2368 			break;
2369 		if (last_timestamp != map->timestamp) {
2370 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2371 			entry = tmp_entry;
2372 		}
2373 	}
2374 	vm_map_unlock_read(map);
2375 	vmspace_free(vm);
2376 	PRELE(p);
2377 	free(kve, M_TEMP);
2378 	return (error);
2379 }
2380 
2381 static int
2382 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2383 {
2384 	struct proc *p;
2385 	struct sbuf sb;
2386 	int error, error2, *name;
2387 
2388 	name = (int *)arg1;
2389 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2390 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2391 	if (error != 0) {
2392 		sbuf_delete(&sb);
2393 		return (error);
2394 	}
2395 	error = kern_proc_vmmap_out(p, &sb);
2396 	error2 = sbuf_finish(&sb);
2397 	sbuf_delete(&sb);
2398 	return (error != 0 ? error : error2);
2399 }
2400 
2401 #if defined(STACK) || defined(DDB)
2402 static int
2403 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2404 {
2405 	struct kinfo_kstack *kkstp;
2406 	int error, i, *name, numthreads;
2407 	lwpid_t *lwpidarray;
2408 	struct thread *td;
2409 	struct stack *st;
2410 	struct sbuf sb;
2411 	struct proc *p;
2412 
2413 	name = (int *)arg1;
2414 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2415 	if (error != 0)
2416 		return (error);
2417 
2418 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2419 	st = stack_create();
2420 
2421 	lwpidarray = NULL;
2422 	numthreads = 0;
2423 	PROC_LOCK(p);
2424 repeat:
2425 	if (numthreads < p->p_numthreads) {
2426 		if (lwpidarray != NULL) {
2427 			free(lwpidarray, M_TEMP);
2428 			lwpidarray = NULL;
2429 		}
2430 		numthreads = p->p_numthreads;
2431 		PROC_UNLOCK(p);
2432 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2433 		    M_WAITOK | M_ZERO);
2434 		PROC_LOCK(p);
2435 		goto repeat;
2436 	}
2437 	i = 0;
2438 
2439 	/*
2440 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2441 	 * of changes could have taken place.  Should we check to see if the
2442 	 * vmspace has been replaced, or the like, in order to prevent
2443 	 * giving a snapshot that spans, say, execve(2), with some threads
2444 	 * before and some after?  Among other things, the credentials could
2445 	 * have changed, in which case the right to extract debug info might
2446 	 * no longer be assured.
2447 	 */
2448 	FOREACH_THREAD_IN_PROC(p, td) {
2449 		KASSERT(i < numthreads,
2450 		    ("sysctl_kern_proc_kstack: numthreads"));
2451 		lwpidarray[i] = td->td_tid;
2452 		i++;
2453 	}
2454 	numthreads = i;
2455 	for (i = 0; i < numthreads; i++) {
2456 		td = thread_find(p, lwpidarray[i]);
2457 		if (td == NULL) {
2458 			continue;
2459 		}
2460 		bzero(kkstp, sizeof(*kkstp));
2461 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2462 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2463 		thread_lock(td);
2464 		kkstp->kkst_tid = td->td_tid;
2465 		if (TD_IS_SWAPPED(td))
2466 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2467 		else if (TD_IS_RUNNING(td))
2468 			kkstp->kkst_state = KKST_STATE_RUNNING;
2469 		else {
2470 			kkstp->kkst_state = KKST_STATE_STACKOK;
2471 			stack_save_td(st, td);
2472 		}
2473 		thread_unlock(td);
2474 		PROC_UNLOCK(p);
2475 		stack_sbuf_print(&sb, st);
2476 		sbuf_finish(&sb);
2477 		sbuf_delete(&sb);
2478 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2479 		PROC_LOCK(p);
2480 		if (error)
2481 			break;
2482 	}
2483 	_PRELE(p);
2484 	PROC_UNLOCK(p);
2485 	if (lwpidarray != NULL)
2486 		free(lwpidarray, M_TEMP);
2487 	stack_destroy(st);
2488 	free(kkstp, M_TEMP);
2489 	return (error);
2490 }
2491 #endif
2492 
2493 /*
2494  * This sysctl allows a process to retrieve the full list of groups from
2495  * itself or another process.
2496  */
2497 static int
2498 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2499 {
2500 	pid_t *pidp = (pid_t *)arg1;
2501 	unsigned int arglen = arg2;
2502 	struct proc *p;
2503 	struct ucred *cred;
2504 	int error;
2505 
2506 	if (arglen != 1)
2507 		return (EINVAL);
2508 	if (*pidp == -1) {	/* -1 means this process */
2509 		p = req->td->td_proc;
2510 	} else {
2511 		error = pget(*pidp, PGET_CANSEE, &p);
2512 		if (error != 0)
2513 			return (error);
2514 	}
2515 
2516 	cred = crhold(p->p_ucred);
2517 	if (*pidp != -1)
2518 		PROC_UNLOCK(p);
2519 
2520 	error = SYSCTL_OUT(req, cred->cr_groups,
2521 	    cred->cr_ngroups * sizeof(gid_t));
2522 	crfree(cred);
2523 	return (error);
2524 }
2525 
2526 /*
2527  * This sysctl allows a process to retrieve or/and set the resource limit for
2528  * another process.
2529  */
2530 static int
2531 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2532 {
2533 	int *name = (int *)arg1;
2534 	u_int namelen = arg2;
2535 	struct rlimit rlim;
2536 	struct proc *p;
2537 	u_int which;
2538 	int flags, error;
2539 
2540 	if (namelen != 2)
2541 		return (EINVAL);
2542 
2543 	which = (u_int)name[1];
2544 	if (which >= RLIM_NLIMITS)
2545 		return (EINVAL);
2546 
2547 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2548 		return (EINVAL);
2549 
2550 	flags = PGET_HOLD | PGET_NOTWEXIT;
2551 	if (req->newptr != NULL)
2552 		flags |= PGET_CANDEBUG;
2553 	else
2554 		flags |= PGET_CANSEE;
2555 	error = pget((pid_t)name[0], flags, &p);
2556 	if (error != 0)
2557 		return (error);
2558 
2559 	/*
2560 	 * Retrieve limit.
2561 	 */
2562 	if (req->oldptr != NULL) {
2563 		PROC_LOCK(p);
2564 		lim_rlimit(p, which, &rlim);
2565 		PROC_UNLOCK(p);
2566 	}
2567 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2568 	if (error != 0)
2569 		goto errout;
2570 
2571 	/*
2572 	 * Set limit.
2573 	 */
2574 	if (req->newptr != NULL) {
2575 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2576 		if (error == 0)
2577 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2578 	}
2579 
2580 errout:
2581 	PRELE(p);
2582 	return (error);
2583 }
2584 
2585 /*
2586  * This sysctl allows a process to retrieve ps_strings structure location of
2587  * another process.
2588  */
2589 static int
2590 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2591 {
2592 	int *name = (int *)arg1;
2593 	u_int namelen = arg2;
2594 	struct proc *p;
2595 	vm_offset_t ps_strings;
2596 	int error;
2597 #ifdef COMPAT_FREEBSD32
2598 	uint32_t ps_strings32;
2599 #endif
2600 
2601 	if (namelen != 1)
2602 		return (EINVAL);
2603 
2604 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2605 	if (error != 0)
2606 		return (error);
2607 #ifdef COMPAT_FREEBSD32
2608 	if ((req->flags & SCTL_MASK32) != 0) {
2609 		/*
2610 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2611 		 * process.
2612 		 */
2613 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2614 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2615 		PROC_UNLOCK(p);
2616 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2617 		return (error);
2618 	}
2619 #endif
2620 	ps_strings = p->p_sysent->sv_psstrings;
2621 	PROC_UNLOCK(p);
2622 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2623 	return (error);
2624 }
2625 
2626 /*
2627  * This sysctl allows a process to retrieve umask of another process.
2628  */
2629 static int
2630 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2631 {
2632 	int *name = (int *)arg1;
2633 	u_int namelen = arg2;
2634 	struct proc *p;
2635 	int error;
2636 	u_short fd_cmask;
2637 
2638 	if (namelen != 1)
2639 		return (EINVAL);
2640 
2641 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2642 	if (error != 0)
2643 		return (error);
2644 
2645 	FILEDESC_SLOCK(p->p_fd);
2646 	fd_cmask = p->p_fd->fd_cmask;
2647 	FILEDESC_SUNLOCK(p->p_fd);
2648 	PRELE(p);
2649 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2650 	return (error);
2651 }
2652 
2653 /*
2654  * This sysctl allows a process to set and retrieve binary osreldate of
2655  * another process.
2656  */
2657 static int
2658 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2659 {
2660 	int *name = (int *)arg1;
2661 	u_int namelen = arg2;
2662 	struct proc *p;
2663 	int flags, error, osrel;
2664 
2665 	if (namelen != 1)
2666 		return (EINVAL);
2667 
2668 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2669 		return (EINVAL);
2670 
2671 	flags = PGET_HOLD | PGET_NOTWEXIT;
2672 	if (req->newptr != NULL)
2673 		flags |= PGET_CANDEBUG;
2674 	else
2675 		flags |= PGET_CANSEE;
2676 	error = pget((pid_t)name[0], flags, &p);
2677 	if (error != 0)
2678 		return (error);
2679 
2680 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2681 	if (error != 0)
2682 		goto errout;
2683 
2684 	if (req->newptr != NULL) {
2685 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2686 		if (error != 0)
2687 			goto errout;
2688 		if (osrel < 0) {
2689 			error = EINVAL;
2690 			goto errout;
2691 		}
2692 		p->p_osrel = osrel;
2693 	}
2694 errout:
2695 	PRELE(p);
2696 	return (error);
2697 }
2698 
2699 static int
2700 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2701 {
2702 	int *name = (int *)arg1;
2703 	u_int namelen = arg2;
2704 	struct proc *p;
2705 	struct kinfo_sigtramp kst;
2706 	const struct sysentvec *sv;
2707 	int error;
2708 #ifdef COMPAT_FREEBSD32
2709 	struct kinfo_sigtramp32 kst32;
2710 #endif
2711 
2712 	if (namelen != 1)
2713 		return (EINVAL);
2714 
2715 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2716 	if (error != 0)
2717 		return (error);
2718 	sv = p->p_sysent;
2719 #ifdef COMPAT_FREEBSD32
2720 	if ((req->flags & SCTL_MASK32) != 0) {
2721 		bzero(&kst32, sizeof(kst32));
2722 		if (SV_PROC_FLAG(p, SV_ILP32)) {
2723 			if (sv->sv_sigcode_base != 0) {
2724 				kst32.ksigtramp_start = sv->sv_sigcode_base;
2725 				kst32.ksigtramp_end = sv->sv_sigcode_base +
2726 				    *sv->sv_szsigcode;
2727 			} else {
2728 				kst32.ksigtramp_start = sv->sv_psstrings -
2729 				    *sv->sv_szsigcode;
2730 				kst32.ksigtramp_end = sv->sv_psstrings;
2731 			}
2732 		}
2733 		PROC_UNLOCK(p);
2734 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2735 		return (error);
2736 	}
2737 #endif
2738 	bzero(&kst, sizeof(kst));
2739 	if (sv->sv_sigcode_base != 0) {
2740 		kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2741 		kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2742 		    *sv->sv_szsigcode;
2743 	} else {
2744 		kst.ksigtramp_start = (char *)sv->sv_psstrings -
2745 		    *sv->sv_szsigcode;
2746 		kst.ksigtramp_end = (char *)sv->sv_psstrings;
2747 	}
2748 	PROC_UNLOCK(p);
2749 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
2750 	return (error);
2751 }
2752 
2753 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2754 
2755 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2756 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2757 	"Return entire process table");
2758 
2759 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2760 	sysctl_kern_proc, "Process table");
2761 
2762 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2763 	sysctl_kern_proc, "Process table");
2764 
2765 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2766 	sysctl_kern_proc, "Process table");
2767 
2768 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2769 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2770 
2771 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2772 	sysctl_kern_proc, "Process table");
2773 
2774 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2775 	sysctl_kern_proc, "Process table");
2776 
2777 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2778 	sysctl_kern_proc, "Process table");
2779 
2780 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2781 	sysctl_kern_proc, "Process table");
2782 
2783 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2784 	sysctl_kern_proc, "Return process table, no threads");
2785 
2786 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2787 	CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2788 	sysctl_kern_proc_args, "Process argument list");
2789 
2790 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
2791 	sysctl_kern_proc_env, "Process environment");
2792 
2793 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
2794 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
2795 
2796 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2797 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2798 
2799 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2800 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2801 	"Process syscall vector name (ABI type)");
2802 
2803 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2804 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2805 
2806 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2807 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2808 
2809 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2810 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2811 
2812 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2813 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2814 
2815 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2816 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2817 
2818 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2819 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2820 
2821 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2822 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2823 
2824 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2825 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2826 
2827 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2828 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2829 	"Return process table, no threads");
2830 
2831 #ifdef COMPAT_FREEBSD7
2832 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2833 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2834 #endif
2835 
2836 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2837 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2838 
2839 #if defined(STACK) || defined(DDB)
2840 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2841 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2842 #endif
2843 
2844 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2845 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
2846 
2847 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
2848 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
2849 	"Process resource limits");
2850 
2851 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
2852 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
2853 	"Process ps_strings location");
2854 
2855 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
2856 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
2857 
2858 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
2859 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
2860 	"Process binary osreldate");
2861 
2862 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
2863 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
2864 	"Process signal trampoline location");
2865