xref: /freebsd/sys/kern/kern_proc.c (revision a0e793cbf1951d07fc47a0d9ea389d7dacba5213)
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 	PROC_LOCK_ASSERT(p, MA_OWNED);
795 	bzero(kp, sizeof(*kp));
796 
797 	kp->ki_structsize = sizeof(*kp);
798 	kp->ki_paddr = p;
799 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
800 	kp->ki_args = p->p_args;
801 	kp->ki_textvp = p->p_textvp;
802 #ifdef KTRACE
803 	kp->ki_tracep = p->p_tracevp;
804 	kp->ki_traceflag = p->p_traceflag;
805 #endif
806 	kp->ki_fd = p->p_fd;
807 	kp->ki_vmspace = p->p_vmspace;
808 	kp->ki_flag = p->p_flag;
809 	kp->ki_flag2 = p->p_flag2;
810 	cred = p->p_ucred;
811 	if (cred) {
812 		kp->ki_uid = cred->cr_uid;
813 		kp->ki_ruid = cred->cr_ruid;
814 		kp->ki_svuid = cred->cr_svuid;
815 		kp->ki_cr_flags = 0;
816 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
817 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
818 		/* XXX bde doesn't like KI_NGROUPS */
819 		if (cred->cr_ngroups > KI_NGROUPS) {
820 			kp->ki_ngroups = KI_NGROUPS;
821 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
822 		} else
823 			kp->ki_ngroups = cred->cr_ngroups;
824 		bcopy(cred->cr_groups, kp->ki_groups,
825 		    kp->ki_ngroups * sizeof(gid_t));
826 		kp->ki_rgid = cred->cr_rgid;
827 		kp->ki_svgid = cred->cr_svgid;
828 		/* If jailed(cred), emulate the old P_JAILED flag. */
829 		if (jailed(cred)) {
830 			kp->ki_flag |= P_JAILED;
831 			/* If inside the jail, use 0 as a jail ID. */
832 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
833 				kp->ki_jid = cred->cr_prison->pr_id;
834 		}
835 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
836 		    sizeof(kp->ki_loginclass));
837 	}
838 	ps = p->p_sigacts;
839 	if (ps) {
840 		mtx_lock(&ps->ps_mtx);
841 		kp->ki_sigignore = ps->ps_sigignore;
842 		kp->ki_sigcatch = ps->ps_sigcatch;
843 		mtx_unlock(&ps->ps_mtx);
844 	}
845 	if (p->p_state != PRS_NEW &&
846 	    p->p_state != PRS_ZOMBIE &&
847 	    p->p_vmspace != NULL) {
848 		struct vmspace *vm = p->p_vmspace;
849 
850 		kp->ki_size = vm->vm_map.size;
851 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
852 		FOREACH_THREAD_IN_PROC(p, td0) {
853 			if (!TD_IS_SWAPPED(td0))
854 				kp->ki_rssize += td0->td_kstack_pages;
855 		}
856 		kp->ki_swrss = vm->vm_swrss;
857 		kp->ki_tsize = vm->vm_tsize;
858 		kp->ki_dsize = vm->vm_dsize;
859 		kp->ki_ssize = vm->vm_ssize;
860 	} else if (p->p_state == PRS_ZOMBIE)
861 		kp->ki_stat = SZOMB;
862 	if (kp->ki_flag & P_INMEM)
863 		kp->ki_sflag = PS_INMEM;
864 	else
865 		kp->ki_sflag = 0;
866 	/* Calculate legacy swtime as seconds since 'swtick'. */
867 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
868 	kp->ki_pid = p->p_pid;
869 	kp->ki_nice = p->p_nice;
870 	kp->ki_fibnum = p->p_fibnum;
871 	kp->ki_start = p->p_stats->p_start;
872 	timevaladd(&kp->ki_start, &boottime);
873 	PROC_SLOCK(p);
874 	rufetch(p, &kp->ki_rusage);
875 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
876 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
877 	PROC_SUNLOCK(p);
878 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
879 	/* Some callers want child times in a single value. */
880 	kp->ki_childtime = kp->ki_childstime;
881 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
882 
883 	FOREACH_THREAD_IN_PROC(p, td0)
884 		kp->ki_cow += td0->td_cow;
885 
886 	tp = NULL;
887 	if (p->p_pgrp) {
888 		kp->ki_pgid = p->p_pgrp->pg_id;
889 		kp->ki_jobc = p->p_pgrp->pg_jobc;
890 		sp = p->p_pgrp->pg_session;
891 
892 		if (sp != NULL) {
893 			kp->ki_sid = sp->s_sid;
894 			SESS_LOCK(sp);
895 			strlcpy(kp->ki_login, sp->s_login,
896 			    sizeof(kp->ki_login));
897 			if (sp->s_ttyvp)
898 				kp->ki_kiflag |= KI_CTTY;
899 			if (SESS_LEADER(p))
900 				kp->ki_kiflag |= KI_SLEADER;
901 			/* XXX proctree_lock */
902 			tp = sp->s_ttyp;
903 			SESS_UNLOCK(sp);
904 		}
905 	}
906 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
907 		kp->ki_tdev = tty_udev(tp);
908 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
909 		if (tp->t_session)
910 			kp->ki_tsid = tp->t_session->s_sid;
911 	} else
912 		kp->ki_tdev = NODEV;
913 	if (p->p_comm[0] != '\0')
914 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
915 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
916 	    p->p_sysent->sv_name[0] != '\0')
917 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
918 	kp->ki_siglist = p->p_siglist;
919 	kp->ki_xstat = p->p_xstat;
920 	kp->ki_acflag = p->p_acflag;
921 	kp->ki_lock = p->p_lock;
922 	if (p->p_pptr)
923 		kp->ki_ppid = p->p_pptr->p_pid;
924 }
925 
926 /*
927  * Fill in information that is thread specific.  Must be called with
928  * target process locked.  If 'preferthread' is set, overwrite certain
929  * process-related fields that are maintained for both threads and
930  * processes.
931  */
932 static void
933 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
934 {
935 	struct proc *p;
936 
937 	p = td->td_proc;
938 	kp->ki_tdaddr = td;
939 	PROC_LOCK_ASSERT(p, MA_OWNED);
940 
941 	if (preferthread)
942 		PROC_SLOCK(p);
943 	thread_lock(td);
944 	if (td->td_wmesg != NULL)
945 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
946 	else
947 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
948 	strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname));
949 	if (TD_ON_LOCK(td)) {
950 		kp->ki_kiflag |= KI_LOCKBLOCK;
951 		strlcpy(kp->ki_lockname, td->td_lockname,
952 		    sizeof(kp->ki_lockname));
953 	} else {
954 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
955 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
956 	}
957 
958 	if (p->p_state == PRS_NORMAL) { /* approximate. */
959 		if (TD_ON_RUNQ(td) ||
960 		    TD_CAN_RUN(td) ||
961 		    TD_IS_RUNNING(td)) {
962 			kp->ki_stat = SRUN;
963 		} else if (P_SHOULDSTOP(p)) {
964 			kp->ki_stat = SSTOP;
965 		} else if (TD_IS_SLEEPING(td)) {
966 			kp->ki_stat = SSLEEP;
967 		} else if (TD_ON_LOCK(td)) {
968 			kp->ki_stat = SLOCK;
969 		} else {
970 			kp->ki_stat = SWAIT;
971 		}
972 	} else if (p->p_state == PRS_ZOMBIE) {
973 		kp->ki_stat = SZOMB;
974 	} else {
975 		kp->ki_stat = SIDL;
976 	}
977 
978 	/* Things in the thread */
979 	kp->ki_wchan = td->td_wchan;
980 	kp->ki_pri.pri_level = td->td_priority;
981 	kp->ki_pri.pri_native = td->td_base_pri;
982 	kp->ki_lastcpu = td->td_lastcpu;
983 	kp->ki_oncpu = td->td_oncpu;
984 	kp->ki_tdflags = td->td_flags;
985 	kp->ki_tid = td->td_tid;
986 	kp->ki_numthreads = p->p_numthreads;
987 	kp->ki_pcb = td->td_pcb;
988 	kp->ki_kstack = (void *)td->td_kstack;
989 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
990 	kp->ki_pri.pri_class = td->td_pri_class;
991 	kp->ki_pri.pri_user = td->td_user_pri;
992 
993 	if (preferthread) {
994 		rufetchtd(td, &kp->ki_rusage);
995 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
996 		kp->ki_pctcpu = sched_pctcpu(td);
997 		kp->ki_estcpu = td->td_estcpu;
998 		kp->ki_cow = td->td_cow;
999 	}
1000 
1001 	/* We can't get this anymore but ps etc never used it anyway. */
1002 	kp->ki_rqindex = 0;
1003 
1004 	if (preferthread)
1005 		kp->ki_siglist = td->td_siglist;
1006 	kp->ki_sigmask = td->td_sigmask;
1007 	thread_unlock(td);
1008 	if (preferthread)
1009 		PROC_SUNLOCK(p);
1010 }
1011 
1012 /*
1013  * Fill in a kinfo_proc structure for the specified process.
1014  * Must be called with the target process locked.
1015  */
1016 void
1017 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1018 {
1019 
1020 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1021 
1022 	fill_kinfo_proc_only(p, kp);
1023 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1024 	fill_kinfo_aggregate(p, kp);
1025 }
1026 
1027 struct pstats *
1028 pstats_alloc(void)
1029 {
1030 
1031 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1032 }
1033 
1034 /*
1035  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1036  */
1037 void
1038 pstats_fork(struct pstats *src, struct pstats *dst)
1039 {
1040 
1041 	bzero(&dst->pstat_startzero,
1042 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1043 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1044 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1045 }
1046 
1047 void
1048 pstats_free(struct pstats *ps)
1049 {
1050 
1051 	free(ps, M_SUBPROC);
1052 }
1053 
1054 static struct proc *
1055 zpfind_locked(pid_t pid)
1056 {
1057 	struct proc *p;
1058 
1059 	sx_assert(&allproc_lock, SX_LOCKED);
1060 	LIST_FOREACH(p, &zombproc, p_list) {
1061 		if (p->p_pid == pid) {
1062 			PROC_LOCK(p);
1063 			break;
1064 		}
1065 	}
1066 	return (p);
1067 }
1068 
1069 /*
1070  * Locate a zombie process by number
1071  */
1072 struct proc *
1073 zpfind(pid_t pid)
1074 {
1075 	struct proc *p;
1076 
1077 	sx_slock(&allproc_lock);
1078 	p = zpfind_locked(pid);
1079 	sx_sunlock(&allproc_lock);
1080 	return (p);
1081 }
1082 
1083 #ifdef COMPAT_FREEBSD32
1084 
1085 /*
1086  * This function is typically used to copy out the kernel address, so
1087  * it can be replaced by assignment of zero.
1088  */
1089 static inline uint32_t
1090 ptr32_trim(void *ptr)
1091 {
1092 	uintptr_t uptr;
1093 
1094 	uptr = (uintptr_t)ptr;
1095 	return ((uptr > UINT_MAX) ? 0 : uptr);
1096 }
1097 
1098 #define PTRTRIM_CP(src,dst,fld) \
1099 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1100 
1101 static void
1102 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1103 {
1104 	int i;
1105 
1106 	bzero(ki32, sizeof(struct kinfo_proc32));
1107 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1108 	CP(*ki, *ki32, ki_layout);
1109 	PTRTRIM_CP(*ki, *ki32, ki_args);
1110 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1111 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1112 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1113 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1114 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1115 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1116 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1117 	CP(*ki, *ki32, ki_pid);
1118 	CP(*ki, *ki32, ki_ppid);
1119 	CP(*ki, *ki32, ki_pgid);
1120 	CP(*ki, *ki32, ki_tpgid);
1121 	CP(*ki, *ki32, ki_sid);
1122 	CP(*ki, *ki32, ki_tsid);
1123 	CP(*ki, *ki32, ki_jobc);
1124 	CP(*ki, *ki32, ki_tdev);
1125 	CP(*ki, *ki32, ki_siglist);
1126 	CP(*ki, *ki32, ki_sigmask);
1127 	CP(*ki, *ki32, ki_sigignore);
1128 	CP(*ki, *ki32, ki_sigcatch);
1129 	CP(*ki, *ki32, ki_uid);
1130 	CP(*ki, *ki32, ki_ruid);
1131 	CP(*ki, *ki32, ki_svuid);
1132 	CP(*ki, *ki32, ki_rgid);
1133 	CP(*ki, *ki32, ki_svgid);
1134 	CP(*ki, *ki32, ki_ngroups);
1135 	for (i = 0; i < KI_NGROUPS; i++)
1136 		CP(*ki, *ki32, ki_groups[i]);
1137 	CP(*ki, *ki32, ki_size);
1138 	CP(*ki, *ki32, ki_rssize);
1139 	CP(*ki, *ki32, ki_swrss);
1140 	CP(*ki, *ki32, ki_tsize);
1141 	CP(*ki, *ki32, ki_dsize);
1142 	CP(*ki, *ki32, ki_ssize);
1143 	CP(*ki, *ki32, ki_xstat);
1144 	CP(*ki, *ki32, ki_acflag);
1145 	CP(*ki, *ki32, ki_pctcpu);
1146 	CP(*ki, *ki32, ki_estcpu);
1147 	CP(*ki, *ki32, ki_slptime);
1148 	CP(*ki, *ki32, ki_swtime);
1149 	CP(*ki, *ki32, ki_cow);
1150 	CP(*ki, *ki32, ki_runtime);
1151 	TV_CP(*ki, *ki32, ki_start);
1152 	TV_CP(*ki, *ki32, ki_childtime);
1153 	CP(*ki, *ki32, ki_flag);
1154 	CP(*ki, *ki32, ki_kiflag);
1155 	CP(*ki, *ki32, ki_traceflag);
1156 	CP(*ki, *ki32, ki_stat);
1157 	CP(*ki, *ki32, ki_nice);
1158 	CP(*ki, *ki32, ki_lock);
1159 	CP(*ki, *ki32, ki_rqindex);
1160 	CP(*ki, *ki32, ki_oncpu);
1161 	CP(*ki, *ki32, ki_lastcpu);
1162 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1163 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1164 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1165 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1166 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1167 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1168 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1169 	CP(*ki, *ki32, ki_flag2);
1170 	CP(*ki, *ki32, ki_fibnum);
1171 	CP(*ki, *ki32, ki_cr_flags);
1172 	CP(*ki, *ki32, ki_jid);
1173 	CP(*ki, *ki32, ki_numthreads);
1174 	CP(*ki, *ki32, ki_tid);
1175 	CP(*ki, *ki32, ki_pri);
1176 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1177 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1178 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1179 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1180 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1181 	CP(*ki, *ki32, ki_sflag);
1182 	CP(*ki, *ki32, ki_tdflags);
1183 }
1184 #endif
1185 
1186 int
1187 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1188 {
1189 	struct thread *td;
1190 	struct kinfo_proc ki;
1191 #ifdef COMPAT_FREEBSD32
1192 	struct kinfo_proc32 ki32;
1193 #endif
1194 	int error;
1195 
1196 	PROC_LOCK_ASSERT(p, MA_OWNED);
1197 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1198 
1199 	error = 0;
1200 	fill_kinfo_proc(p, &ki);
1201 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1202 #ifdef COMPAT_FREEBSD32
1203 		if ((flags & KERN_PROC_MASK32) != 0) {
1204 			freebsd32_kinfo_proc_out(&ki, &ki32);
1205 			error = sbuf_bcat(sb, &ki32, sizeof(ki32));
1206 		} else
1207 #endif
1208 			error = sbuf_bcat(sb, &ki, sizeof(ki));
1209 	} else {
1210 		FOREACH_THREAD_IN_PROC(p, td) {
1211 			fill_kinfo_thread(td, &ki, 1);
1212 #ifdef COMPAT_FREEBSD32
1213 			if ((flags & KERN_PROC_MASK32) != 0) {
1214 				freebsd32_kinfo_proc_out(&ki, &ki32);
1215 				error = sbuf_bcat(sb, &ki32, sizeof(ki32));
1216 			} else
1217 #endif
1218 				error = sbuf_bcat(sb, &ki, sizeof(ki));
1219 			if (error)
1220 				break;
1221 		}
1222 	}
1223 	PROC_UNLOCK(p);
1224 	return (error);
1225 }
1226 
1227 static int
1228 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags,
1229     int doingzomb)
1230 {
1231 	struct sbuf sb;
1232 	struct kinfo_proc ki;
1233 	struct proc *np;
1234 	int error, error2;
1235 	pid_t pid;
1236 
1237 	pid = p->p_pid;
1238 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1239 	error = kern_proc_out(p, &sb, flags);
1240 	error2 = sbuf_finish(&sb);
1241 	sbuf_delete(&sb);
1242 	if (error != 0)
1243 		return (error);
1244 	else if (error2 != 0)
1245 		return (error2);
1246 	if (doingzomb)
1247 		np = zpfind(pid);
1248 	else {
1249 		if (pid == 0)
1250 			return (0);
1251 		np = pfind(pid);
1252 	}
1253 	if (np == NULL)
1254 		return (ESRCH);
1255 	if (np != p) {
1256 		PROC_UNLOCK(np);
1257 		return (ESRCH);
1258 	}
1259 	PROC_UNLOCK(np);
1260 	return (0);
1261 }
1262 
1263 static int
1264 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1265 {
1266 	int *name = (int *)arg1;
1267 	u_int namelen = arg2;
1268 	struct proc *p;
1269 	int flags, doingzomb, oid_number;
1270 	int error = 0;
1271 
1272 	oid_number = oidp->oid_number;
1273 	if (oid_number != KERN_PROC_ALL &&
1274 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1275 		flags = KERN_PROC_NOTHREADS;
1276 	else {
1277 		flags = 0;
1278 		oid_number &= ~KERN_PROC_INC_THREAD;
1279 	}
1280 #ifdef COMPAT_FREEBSD32
1281 	if (req->flags & SCTL_MASK32)
1282 		flags |= KERN_PROC_MASK32;
1283 #endif
1284 	if (oid_number == KERN_PROC_PID) {
1285 		if (namelen != 1)
1286 			return (EINVAL);
1287 		error = sysctl_wire_old_buffer(req, 0);
1288 		if (error)
1289 			return (error);
1290 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1291 		if (error != 0)
1292 			return (error);
1293 		error = sysctl_out_proc(p, req, flags, 0);
1294 		return (error);
1295 	}
1296 
1297 	switch (oid_number) {
1298 	case KERN_PROC_ALL:
1299 		if (namelen != 0)
1300 			return (EINVAL);
1301 		break;
1302 	case KERN_PROC_PROC:
1303 		if (namelen != 0 && namelen != 1)
1304 			return (EINVAL);
1305 		break;
1306 	default:
1307 		if (namelen != 1)
1308 			return (EINVAL);
1309 		break;
1310 	}
1311 
1312 	if (!req->oldptr) {
1313 		/* overestimate by 5 procs */
1314 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1315 		if (error)
1316 			return (error);
1317 	}
1318 	error = sysctl_wire_old_buffer(req, 0);
1319 	if (error != 0)
1320 		return (error);
1321 	sx_slock(&allproc_lock);
1322 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1323 		if (!doingzomb)
1324 			p = LIST_FIRST(&allproc);
1325 		else
1326 			p = LIST_FIRST(&zombproc);
1327 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
1328 			/*
1329 			 * Skip embryonic processes.
1330 			 */
1331 			PROC_LOCK(p);
1332 			if (p->p_state == PRS_NEW) {
1333 				PROC_UNLOCK(p);
1334 				continue;
1335 			}
1336 			KASSERT(p->p_ucred != NULL,
1337 			    ("process credential is NULL for non-NEW proc"));
1338 			/*
1339 			 * Show a user only appropriate processes.
1340 			 */
1341 			if (p_cansee(curthread, p)) {
1342 				PROC_UNLOCK(p);
1343 				continue;
1344 			}
1345 			/*
1346 			 * TODO - make more efficient (see notes below).
1347 			 * do by session.
1348 			 */
1349 			switch (oid_number) {
1350 
1351 			case KERN_PROC_GID:
1352 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1353 					PROC_UNLOCK(p);
1354 					continue;
1355 				}
1356 				break;
1357 
1358 			case KERN_PROC_PGRP:
1359 				/* could do this by traversing pgrp */
1360 				if (p->p_pgrp == NULL ||
1361 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1362 					PROC_UNLOCK(p);
1363 					continue;
1364 				}
1365 				break;
1366 
1367 			case KERN_PROC_RGID:
1368 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1369 					PROC_UNLOCK(p);
1370 					continue;
1371 				}
1372 				break;
1373 
1374 			case KERN_PROC_SESSION:
1375 				if (p->p_session == NULL ||
1376 				    p->p_session->s_sid != (pid_t)name[0]) {
1377 					PROC_UNLOCK(p);
1378 					continue;
1379 				}
1380 				break;
1381 
1382 			case KERN_PROC_TTY:
1383 				if ((p->p_flag & P_CONTROLT) == 0 ||
1384 				    p->p_session == NULL) {
1385 					PROC_UNLOCK(p);
1386 					continue;
1387 				}
1388 				/* XXX proctree_lock */
1389 				SESS_LOCK(p->p_session);
1390 				if (p->p_session->s_ttyp == NULL ||
1391 				    tty_udev(p->p_session->s_ttyp) !=
1392 				    (dev_t)name[0]) {
1393 					SESS_UNLOCK(p->p_session);
1394 					PROC_UNLOCK(p);
1395 					continue;
1396 				}
1397 				SESS_UNLOCK(p->p_session);
1398 				break;
1399 
1400 			case KERN_PROC_UID:
1401 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1402 					PROC_UNLOCK(p);
1403 					continue;
1404 				}
1405 				break;
1406 
1407 			case KERN_PROC_RUID:
1408 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1409 					PROC_UNLOCK(p);
1410 					continue;
1411 				}
1412 				break;
1413 
1414 			case KERN_PROC_PROC:
1415 				break;
1416 
1417 			default:
1418 				break;
1419 
1420 			}
1421 
1422 			error = sysctl_out_proc(p, req, flags, doingzomb);
1423 			if (error) {
1424 				sx_sunlock(&allproc_lock);
1425 				return (error);
1426 			}
1427 		}
1428 	}
1429 	sx_sunlock(&allproc_lock);
1430 	return (0);
1431 }
1432 
1433 struct pargs *
1434 pargs_alloc(int len)
1435 {
1436 	struct pargs *pa;
1437 
1438 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1439 		M_WAITOK);
1440 	refcount_init(&pa->ar_ref, 1);
1441 	pa->ar_length = len;
1442 	return (pa);
1443 }
1444 
1445 static void
1446 pargs_free(struct pargs *pa)
1447 {
1448 
1449 	free(pa, M_PARGS);
1450 }
1451 
1452 void
1453 pargs_hold(struct pargs *pa)
1454 {
1455 
1456 	if (pa == NULL)
1457 		return;
1458 	refcount_acquire(&pa->ar_ref);
1459 }
1460 
1461 void
1462 pargs_drop(struct pargs *pa)
1463 {
1464 
1465 	if (pa == NULL)
1466 		return;
1467 	if (refcount_release(&pa->ar_ref))
1468 		pargs_free(pa);
1469 }
1470 
1471 static int
1472 proc_read_mem(struct thread *td, struct proc *p, vm_offset_t offset, void* buf,
1473     size_t len)
1474 {
1475 	struct iovec iov;
1476 	struct uio uio;
1477 
1478 	iov.iov_base = (caddr_t)buf;
1479 	iov.iov_len = len;
1480 	uio.uio_iov = &iov;
1481 	uio.uio_iovcnt = 1;
1482 	uio.uio_offset = offset;
1483 	uio.uio_resid = (ssize_t)len;
1484 	uio.uio_segflg = UIO_SYSSPACE;
1485 	uio.uio_rw = UIO_READ;
1486 	uio.uio_td = td;
1487 
1488 	return (proc_rwmem(p, &uio));
1489 }
1490 
1491 static int
1492 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1493     size_t len)
1494 {
1495 	size_t i;
1496 	int error;
1497 
1498 	error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, len);
1499 	/*
1500 	 * Reading the chunk may validly return EFAULT if the string is shorter
1501 	 * than the chunk and is aligned at the end of the page, assuming the
1502 	 * next page is not mapped.  So if EFAULT is returned do a fallback to
1503 	 * one byte read loop.
1504 	 */
1505 	if (error == EFAULT) {
1506 		for (i = 0; i < len; i++, buf++, sptr++) {
1507 			error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, 1);
1508 			if (error != 0)
1509 				return (error);
1510 			if (*buf == '\0')
1511 				break;
1512 		}
1513 		error = 0;
1514 	}
1515 	return (error);
1516 }
1517 
1518 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1519 
1520 enum proc_vector_type {
1521 	PROC_ARG,
1522 	PROC_ENV,
1523 	PROC_AUX,
1524 };
1525 
1526 #ifdef COMPAT_FREEBSD32
1527 static int
1528 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1529     size_t *vsizep, enum proc_vector_type type)
1530 {
1531 	struct freebsd32_ps_strings pss;
1532 	Elf32_Auxinfo aux;
1533 	vm_offset_t vptr, ptr;
1534 	uint32_t *proc_vector32;
1535 	char **proc_vector;
1536 	size_t vsize, size;
1537 	int i, error;
1538 
1539 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1540 	    &pss, sizeof(pss));
1541 	if (error != 0)
1542 		return (error);
1543 	switch (type) {
1544 	case PROC_ARG:
1545 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1546 		vsize = pss.ps_nargvstr;
1547 		if (vsize > ARG_MAX)
1548 			return (ENOEXEC);
1549 		size = vsize * sizeof(int32_t);
1550 		break;
1551 	case PROC_ENV:
1552 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1553 		vsize = pss.ps_nenvstr;
1554 		if (vsize > ARG_MAX)
1555 			return (ENOEXEC);
1556 		size = vsize * sizeof(int32_t);
1557 		break;
1558 	case PROC_AUX:
1559 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1560 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1561 		if (vptr % 4 != 0)
1562 			return (ENOEXEC);
1563 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1564 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1565 			if (error != 0)
1566 				return (error);
1567 			if (aux.a_type == AT_NULL)
1568 				break;
1569 			ptr += sizeof(aux);
1570 		}
1571 		if (aux.a_type != AT_NULL)
1572 			return (ENOEXEC);
1573 		vsize = i + 1;
1574 		size = vsize * sizeof(aux);
1575 		break;
1576 	default:
1577 		KASSERT(0, ("Wrong proc vector type: %d", type));
1578 		return (EINVAL);
1579 	}
1580 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1581 	error = proc_read_mem(td, p, vptr, proc_vector32, size);
1582 	if (error != 0)
1583 		goto done;
1584 	if (type == PROC_AUX) {
1585 		*proc_vectorp = (char **)proc_vector32;
1586 		*vsizep = vsize;
1587 		return (0);
1588 	}
1589 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1590 	for (i = 0; i < (int)vsize; i++)
1591 		proc_vector[i] = PTRIN(proc_vector32[i]);
1592 	*proc_vectorp = proc_vector;
1593 	*vsizep = vsize;
1594 done:
1595 	free(proc_vector32, M_TEMP);
1596 	return (error);
1597 }
1598 #endif
1599 
1600 static int
1601 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1602     size_t *vsizep, enum proc_vector_type type)
1603 {
1604 	struct ps_strings pss;
1605 	Elf_Auxinfo aux;
1606 	vm_offset_t vptr, ptr;
1607 	char **proc_vector;
1608 	size_t vsize, size;
1609 	int error, i;
1610 
1611 #ifdef COMPAT_FREEBSD32
1612 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1613 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1614 #endif
1615 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1616 	    &pss, sizeof(pss));
1617 	if (error != 0)
1618 		return (error);
1619 	switch (type) {
1620 	case PROC_ARG:
1621 		vptr = (vm_offset_t)pss.ps_argvstr;
1622 		vsize = pss.ps_nargvstr;
1623 		if (vsize > ARG_MAX)
1624 			return (ENOEXEC);
1625 		size = vsize * sizeof(char *);
1626 		break;
1627 	case PROC_ENV:
1628 		vptr = (vm_offset_t)pss.ps_envstr;
1629 		vsize = pss.ps_nenvstr;
1630 		if (vsize > ARG_MAX)
1631 			return (ENOEXEC);
1632 		size = vsize * sizeof(char *);
1633 		break;
1634 	case PROC_AUX:
1635 		/*
1636 		 * The aux array is just above env array on the stack. Check
1637 		 * that the address is naturally aligned.
1638 		 */
1639 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1640 		    * sizeof(char *);
1641 #if __ELF_WORD_SIZE == 64
1642 		if (vptr % sizeof(uint64_t) != 0)
1643 #else
1644 		if (vptr % sizeof(uint32_t) != 0)
1645 #endif
1646 			return (ENOEXEC);
1647 		/*
1648 		 * We count the array size reading the aux vectors from the
1649 		 * stack until AT_NULL vector is returned.  So (to keep the code
1650 		 * simple) we read the process stack twice: the first time here
1651 		 * to find the size and the second time when copying the vectors
1652 		 * to the allocated proc_vector.
1653 		 */
1654 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1655 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1656 			if (error != 0)
1657 				return (error);
1658 			if (aux.a_type == AT_NULL)
1659 				break;
1660 			ptr += sizeof(aux);
1661 		}
1662 		/*
1663 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1664 		 * not reached AT_NULL, it is most likely we are reading wrong
1665 		 * data: either the process doesn't have auxv array or data has
1666 		 * been modified. Return the error in this case.
1667 		 */
1668 		if (aux.a_type != AT_NULL)
1669 			return (ENOEXEC);
1670 		vsize = i + 1;
1671 		size = vsize * sizeof(aux);
1672 		break;
1673 	default:
1674 		KASSERT(0, ("Wrong proc vector type: %d", type));
1675 		return (EINVAL); /* In case we are built without INVARIANTS. */
1676 	}
1677 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1678 	if (proc_vector == NULL)
1679 		return (ENOMEM);
1680 	error = proc_read_mem(td, p, vptr, proc_vector, size);
1681 	if (error != 0) {
1682 		free(proc_vector, M_TEMP);
1683 		return (error);
1684 	}
1685 	*proc_vectorp = proc_vector;
1686 	*vsizep = vsize;
1687 
1688 	return (0);
1689 }
1690 
1691 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1692 
1693 static int
1694 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1695     enum proc_vector_type type)
1696 {
1697 	size_t done, len, nchr, vsize;
1698 	int error, i;
1699 	char **proc_vector, *sptr;
1700 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1701 
1702 	PROC_ASSERT_HELD(p);
1703 
1704 	/*
1705 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1706 	 */
1707 	nchr = 2 * (PATH_MAX + ARG_MAX);
1708 
1709 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1710 	if (error != 0)
1711 		return (error);
1712 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1713 		/*
1714 		 * The program may have scribbled into its argv array, e.g. to
1715 		 * remove some arguments.  If that has happened, break out
1716 		 * before trying to read from NULL.
1717 		 */
1718 		if (proc_vector[i] == NULL)
1719 			break;
1720 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1721 			error = proc_read_string(td, p, sptr, pss_string,
1722 			    sizeof(pss_string));
1723 			if (error != 0)
1724 				goto done;
1725 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1726 			if (done + len >= nchr)
1727 				len = nchr - done - 1;
1728 			sbuf_bcat(sb, pss_string, len);
1729 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1730 				break;
1731 			done += GET_PS_STRINGS_CHUNK_SZ;
1732 		}
1733 		sbuf_bcat(sb, "", 1);
1734 		done += len + 1;
1735 	}
1736 done:
1737 	free(proc_vector, M_TEMP);
1738 	return (error);
1739 }
1740 
1741 int
1742 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1743 {
1744 
1745 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1746 }
1747 
1748 int
1749 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1750 {
1751 
1752 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1753 }
1754 
1755 int
1756 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1757 {
1758 	size_t vsize, size;
1759 	char **auxv;
1760 	int error;
1761 
1762 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1763 	if (error == 0) {
1764 #ifdef COMPAT_FREEBSD32
1765 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1766 			size = vsize * sizeof(Elf32_Auxinfo);
1767 		else
1768 #endif
1769 			size = vsize * sizeof(Elf_Auxinfo);
1770 		error = sbuf_bcat(sb, auxv, size);
1771 		free(auxv, M_TEMP);
1772 	}
1773 	return (error);
1774 }
1775 
1776 /*
1777  * This sysctl allows a process to retrieve the argument list or process
1778  * title for another process without groping around in the address space
1779  * of the other process.  It also allow a process to set its own "process
1780  * title to a string of its own choice.
1781  */
1782 static int
1783 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1784 {
1785 	int *name = (int *)arg1;
1786 	u_int namelen = arg2;
1787 	struct pargs *newpa, *pa;
1788 	struct proc *p;
1789 	struct sbuf sb;
1790 	int flags, error = 0, error2;
1791 
1792 	if (namelen != 1)
1793 		return (EINVAL);
1794 
1795 	flags = PGET_CANSEE;
1796 	if (req->newptr != NULL)
1797 		flags |= PGET_ISCURRENT;
1798 	error = pget((pid_t)name[0], flags, &p);
1799 	if (error)
1800 		return (error);
1801 
1802 	pa = p->p_args;
1803 	if (pa != NULL) {
1804 		pargs_hold(pa);
1805 		PROC_UNLOCK(p);
1806 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1807 		pargs_drop(pa);
1808 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1809 		_PHOLD(p);
1810 		PROC_UNLOCK(p);
1811 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1812 		error = proc_getargv(curthread, p, &sb);
1813 		error2 = sbuf_finish(&sb);
1814 		PRELE(p);
1815 		sbuf_delete(&sb);
1816 		if (error == 0 && error2 != 0)
1817 			error = error2;
1818 	} else {
1819 		PROC_UNLOCK(p);
1820 	}
1821 	if (error != 0 || req->newptr == NULL)
1822 		return (error);
1823 
1824 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1825 		return (ENOMEM);
1826 	newpa = pargs_alloc(req->newlen);
1827 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1828 	if (error != 0) {
1829 		pargs_free(newpa);
1830 		return (error);
1831 	}
1832 	PROC_LOCK(p);
1833 	pa = p->p_args;
1834 	p->p_args = newpa;
1835 	PROC_UNLOCK(p);
1836 	pargs_drop(pa);
1837 	return (0);
1838 }
1839 
1840 /*
1841  * This sysctl allows a process to retrieve environment of another process.
1842  */
1843 static int
1844 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
1845 {
1846 	int *name = (int *)arg1;
1847 	u_int namelen = arg2;
1848 	struct proc *p;
1849 	struct sbuf sb;
1850 	int error, error2;
1851 
1852 	if (namelen != 1)
1853 		return (EINVAL);
1854 
1855 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1856 	if (error != 0)
1857 		return (error);
1858 	if ((p->p_flag & P_SYSTEM) != 0) {
1859 		PRELE(p);
1860 		return (0);
1861 	}
1862 
1863 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1864 	error = proc_getenvv(curthread, p, &sb);
1865 	error2 = sbuf_finish(&sb);
1866 	PRELE(p);
1867 	sbuf_delete(&sb);
1868 	return (error != 0 ? error : error2);
1869 }
1870 
1871 /*
1872  * This sysctl allows a process to retrieve ELF auxiliary vector of
1873  * another process.
1874  */
1875 static int
1876 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
1877 {
1878 	int *name = (int *)arg1;
1879 	u_int namelen = arg2;
1880 	struct proc *p;
1881 	struct sbuf sb;
1882 	int error, error2;
1883 
1884 	if (namelen != 1)
1885 		return (EINVAL);
1886 
1887 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1888 	if (error != 0)
1889 		return (error);
1890 	if ((p->p_flag & P_SYSTEM) != 0) {
1891 		PRELE(p);
1892 		return (0);
1893 	}
1894 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1895 	error = proc_getauxv(curthread, p, &sb);
1896 	error2 = sbuf_finish(&sb);
1897 	PRELE(p);
1898 	sbuf_delete(&sb);
1899 	return (error != 0 ? error : error2);
1900 }
1901 
1902 /*
1903  * This sysctl allows a process to retrieve the path of the executable for
1904  * itself or another process.
1905  */
1906 static int
1907 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1908 {
1909 	pid_t *pidp = (pid_t *)arg1;
1910 	unsigned int arglen = arg2;
1911 	struct proc *p;
1912 	struct vnode *vp;
1913 	char *retbuf, *freebuf;
1914 	int error;
1915 
1916 	if (arglen != 1)
1917 		return (EINVAL);
1918 	if (*pidp == -1) {	/* -1 means this process */
1919 		p = req->td->td_proc;
1920 	} else {
1921 		error = pget(*pidp, PGET_CANSEE, &p);
1922 		if (error != 0)
1923 			return (error);
1924 	}
1925 
1926 	vp = p->p_textvp;
1927 	if (vp == NULL) {
1928 		if (*pidp != -1)
1929 			PROC_UNLOCK(p);
1930 		return (0);
1931 	}
1932 	vref(vp);
1933 	if (*pidp != -1)
1934 		PROC_UNLOCK(p);
1935 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1936 	vrele(vp);
1937 	if (error)
1938 		return (error);
1939 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1940 	free(freebuf, M_TEMP);
1941 	return (error);
1942 }
1943 
1944 static int
1945 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1946 {
1947 	struct proc *p;
1948 	char *sv_name;
1949 	int *name;
1950 	int namelen;
1951 	int error;
1952 
1953 	namelen = arg2;
1954 	if (namelen != 1)
1955 		return (EINVAL);
1956 
1957 	name = (int *)arg1;
1958 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
1959 	if (error != 0)
1960 		return (error);
1961 	sv_name = p->p_sysent->sv_name;
1962 	PROC_UNLOCK(p);
1963 	return (sysctl_handle_string(oidp, sv_name, 0, req));
1964 }
1965 
1966 #ifdef KINFO_OVMENTRY_SIZE
1967 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
1968 #endif
1969 
1970 #ifdef COMPAT_FREEBSD7
1971 static int
1972 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
1973 {
1974 	vm_map_entry_t entry, tmp_entry;
1975 	unsigned int last_timestamp;
1976 	char *fullpath, *freepath;
1977 	struct kinfo_ovmentry *kve;
1978 	struct vattr va;
1979 	struct ucred *cred;
1980 	int error, *name;
1981 	struct vnode *vp;
1982 	struct proc *p;
1983 	vm_map_t map;
1984 	struct vmspace *vm;
1985 
1986 	name = (int *)arg1;
1987 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1988 	if (error != 0)
1989 		return (error);
1990 	vm = vmspace_acquire_ref(p);
1991 	if (vm == NULL) {
1992 		PRELE(p);
1993 		return (ESRCH);
1994 	}
1995 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1996 
1997 	map = &vm->vm_map;
1998 	vm_map_lock_read(map);
1999 	for (entry = map->header.next; entry != &map->header;
2000 	    entry = entry->next) {
2001 		vm_object_t obj, tobj, lobj;
2002 		vm_offset_t addr;
2003 
2004 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2005 			continue;
2006 
2007 		bzero(kve, sizeof(*kve));
2008 		kve->kve_structsize = sizeof(*kve);
2009 
2010 		kve->kve_private_resident = 0;
2011 		obj = entry->object.vm_object;
2012 		if (obj != NULL) {
2013 			VM_OBJECT_RLOCK(obj);
2014 			if (obj->shadow_count == 1)
2015 				kve->kve_private_resident =
2016 				    obj->resident_page_count;
2017 		}
2018 		kve->kve_resident = 0;
2019 		addr = entry->start;
2020 		while (addr < entry->end) {
2021 			if (pmap_extract(map->pmap, addr))
2022 				kve->kve_resident++;
2023 			addr += PAGE_SIZE;
2024 		}
2025 
2026 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2027 			if (tobj != obj)
2028 				VM_OBJECT_RLOCK(tobj);
2029 			if (lobj != obj)
2030 				VM_OBJECT_RUNLOCK(lobj);
2031 			lobj = tobj;
2032 		}
2033 
2034 		kve->kve_start = (void*)entry->start;
2035 		kve->kve_end = (void*)entry->end;
2036 		kve->kve_offset = (off_t)entry->offset;
2037 
2038 		if (entry->protection & VM_PROT_READ)
2039 			kve->kve_protection |= KVME_PROT_READ;
2040 		if (entry->protection & VM_PROT_WRITE)
2041 			kve->kve_protection |= KVME_PROT_WRITE;
2042 		if (entry->protection & VM_PROT_EXECUTE)
2043 			kve->kve_protection |= KVME_PROT_EXEC;
2044 
2045 		if (entry->eflags & MAP_ENTRY_COW)
2046 			kve->kve_flags |= KVME_FLAG_COW;
2047 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2048 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2049 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2050 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2051 
2052 		last_timestamp = map->timestamp;
2053 		vm_map_unlock_read(map);
2054 
2055 		kve->kve_fileid = 0;
2056 		kve->kve_fsid = 0;
2057 		freepath = NULL;
2058 		fullpath = "";
2059 		if (lobj) {
2060 			vp = NULL;
2061 			switch (lobj->type) {
2062 			case OBJT_DEFAULT:
2063 				kve->kve_type = KVME_TYPE_DEFAULT;
2064 				break;
2065 			case OBJT_VNODE:
2066 				kve->kve_type = KVME_TYPE_VNODE;
2067 				vp = lobj->handle;
2068 				vref(vp);
2069 				break;
2070 			case OBJT_SWAP:
2071 				kve->kve_type = KVME_TYPE_SWAP;
2072 				break;
2073 			case OBJT_DEVICE:
2074 				kve->kve_type = KVME_TYPE_DEVICE;
2075 				break;
2076 			case OBJT_PHYS:
2077 				kve->kve_type = KVME_TYPE_PHYS;
2078 				break;
2079 			case OBJT_DEAD:
2080 				kve->kve_type = KVME_TYPE_DEAD;
2081 				break;
2082 			case OBJT_SG:
2083 				kve->kve_type = KVME_TYPE_SG;
2084 				break;
2085 			default:
2086 				kve->kve_type = KVME_TYPE_UNKNOWN;
2087 				break;
2088 			}
2089 			if (lobj != obj)
2090 				VM_OBJECT_RUNLOCK(lobj);
2091 
2092 			kve->kve_ref_count = obj->ref_count;
2093 			kve->kve_shadow_count = obj->shadow_count;
2094 			VM_OBJECT_RUNLOCK(obj);
2095 			if (vp != NULL) {
2096 				vn_fullpath(curthread, vp, &fullpath,
2097 				    &freepath);
2098 				cred = curthread->td_ucred;
2099 				vn_lock(vp, LK_SHARED | LK_RETRY);
2100 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2101 					kve->kve_fileid = va.va_fileid;
2102 					kve->kve_fsid = va.va_fsid;
2103 				}
2104 				vput(vp);
2105 			}
2106 		} else {
2107 			kve->kve_type = KVME_TYPE_NONE;
2108 			kve->kve_ref_count = 0;
2109 			kve->kve_shadow_count = 0;
2110 		}
2111 
2112 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2113 		if (freepath != NULL)
2114 			free(freepath, M_TEMP);
2115 
2116 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2117 		vm_map_lock_read(map);
2118 		if (error)
2119 			break;
2120 		if (last_timestamp != map->timestamp) {
2121 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2122 			entry = tmp_entry;
2123 		}
2124 	}
2125 	vm_map_unlock_read(map);
2126 	vmspace_free(vm);
2127 	PRELE(p);
2128 	free(kve, M_TEMP);
2129 	return (error);
2130 }
2131 #endif	/* COMPAT_FREEBSD7 */
2132 
2133 #ifdef KINFO_VMENTRY_SIZE
2134 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2135 #endif
2136 
2137 static void
2138 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2139     struct kinfo_vmentry *kve)
2140 {
2141 	vm_object_t obj, tobj;
2142 	vm_page_t m, m_adv;
2143 	vm_offset_t addr;
2144 	vm_paddr_t locked_pa;
2145 	vm_pindex_t pi, pi_adv, pindex;
2146 
2147 	locked_pa = 0;
2148 	obj = entry->object.vm_object;
2149 	addr = entry->start;
2150 	m_adv = NULL;
2151 	pi = OFF_TO_IDX(entry->offset);
2152 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2153 		if (m_adv != NULL) {
2154 			m = m_adv;
2155 		} else {
2156 			pi_adv = OFF_TO_IDX(entry->end - addr);
2157 			pindex = pi;
2158 			for (tobj = obj;; tobj = tobj->backing_object) {
2159 				m = vm_page_find_least(tobj, pindex);
2160 				if (m != NULL) {
2161 					if (m->pindex == pindex)
2162 						break;
2163 					if (pi_adv > m->pindex - pindex) {
2164 						pi_adv = m->pindex - pindex;
2165 						m_adv = m;
2166 					}
2167 				}
2168 				if (tobj->backing_object == NULL)
2169 					goto next;
2170 				pindex += OFF_TO_IDX(tobj->
2171 				    backing_object_offset);
2172 			}
2173 		}
2174 		m_adv = NULL;
2175 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2176 		    (addr & (pagesizes[1] - 1)) == 0 &&
2177 		    (pmap_mincore(map->pmap, addr, &locked_pa) &
2178 		    MINCORE_SUPER) != 0) {
2179 			kve->kve_flags |= KVME_FLAG_SUPER;
2180 			pi_adv = OFF_TO_IDX(pagesizes[1]);
2181 		} else {
2182 			/*
2183 			 * We do not test the found page on validity.
2184 			 * Either the page is busy and being paged in,
2185 			 * or it was invalidated.  The first case
2186 			 * should be counted as resident, the second
2187 			 * is not so clear; we do account both.
2188 			 */
2189 			pi_adv = 1;
2190 		}
2191 		kve->kve_resident += pi_adv;
2192 next:;
2193 	}
2194 	PA_UNLOCK_COND(locked_pa);
2195 }
2196 
2197 /*
2198  * Must be called with the process locked and will return unlocked.
2199  */
2200 int
2201 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb)
2202 {
2203 	vm_map_entry_t entry, tmp_entry;
2204 	struct vattr va;
2205 	vm_map_t map;
2206 	vm_object_t obj, tobj, lobj;
2207 	char *fullpath, *freepath;
2208 	struct kinfo_vmentry *kve;
2209 	struct ucred *cred;
2210 	struct vnode *vp;
2211 	struct vmspace *vm;
2212 	vm_offset_t addr;
2213 	unsigned int last_timestamp;
2214 	int error;
2215 
2216 	PROC_LOCK_ASSERT(p, MA_OWNED);
2217 
2218 	_PHOLD(p);
2219 	PROC_UNLOCK(p);
2220 	vm = vmspace_acquire_ref(p);
2221 	if (vm == NULL) {
2222 		PRELE(p);
2223 		return (ESRCH);
2224 	}
2225 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2226 
2227 	error = 0;
2228 	map = &vm->vm_map;
2229 	vm_map_lock_read(map);
2230 	for (entry = map->header.next; entry != &map->header;
2231 	    entry = entry->next) {
2232 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2233 			continue;
2234 
2235 		addr = entry->end;
2236 		bzero(kve, sizeof(*kve));
2237 		obj = entry->object.vm_object;
2238 		if (obj != NULL) {
2239 			for (tobj = obj; tobj != NULL;
2240 			    tobj = tobj->backing_object) {
2241 				VM_OBJECT_RLOCK(tobj);
2242 				lobj = tobj;
2243 			}
2244 			if (obj->backing_object == NULL)
2245 				kve->kve_private_resident =
2246 				    obj->resident_page_count;
2247 			if (!vmmap_skip_res_cnt)
2248 				kern_proc_vmmap_resident(map, entry, kve);
2249 			for (tobj = obj; tobj != NULL;
2250 			    tobj = tobj->backing_object) {
2251 				if (tobj != obj && tobj != lobj)
2252 					VM_OBJECT_RUNLOCK(tobj);
2253 			}
2254 		} else {
2255 			lobj = NULL;
2256 		}
2257 
2258 		kve->kve_start = entry->start;
2259 		kve->kve_end = entry->end;
2260 		kve->kve_offset = entry->offset;
2261 
2262 		if (entry->protection & VM_PROT_READ)
2263 			kve->kve_protection |= KVME_PROT_READ;
2264 		if (entry->protection & VM_PROT_WRITE)
2265 			kve->kve_protection |= KVME_PROT_WRITE;
2266 		if (entry->protection & VM_PROT_EXECUTE)
2267 			kve->kve_protection |= KVME_PROT_EXEC;
2268 
2269 		if (entry->eflags & MAP_ENTRY_COW)
2270 			kve->kve_flags |= KVME_FLAG_COW;
2271 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2272 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2273 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2274 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2275 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2276 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2277 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2278 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2279 
2280 		last_timestamp = map->timestamp;
2281 		vm_map_unlock_read(map);
2282 
2283 		freepath = NULL;
2284 		fullpath = "";
2285 		if (lobj != NULL) {
2286 			vp = NULL;
2287 			switch (lobj->type) {
2288 			case OBJT_DEFAULT:
2289 				kve->kve_type = KVME_TYPE_DEFAULT;
2290 				break;
2291 			case OBJT_VNODE:
2292 				kve->kve_type = KVME_TYPE_VNODE;
2293 				vp = lobj->handle;
2294 				vref(vp);
2295 				break;
2296 			case OBJT_SWAP:
2297 				kve->kve_type = KVME_TYPE_SWAP;
2298 				break;
2299 			case OBJT_DEVICE:
2300 				kve->kve_type = KVME_TYPE_DEVICE;
2301 				break;
2302 			case OBJT_PHYS:
2303 				kve->kve_type = KVME_TYPE_PHYS;
2304 				break;
2305 			case OBJT_DEAD:
2306 				kve->kve_type = KVME_TYPE_DEAD;
2307 				break;
2308 			case OBJT_SG:
2309 				kve->kve_type = KVME_TYPE_SG;
2310 				break;
2311 			case OBJT_MGTDEVICE:
2312 				kve->kve_type = KVME_TYPE_MGTDEVICE;
2313 				break;
2314 			default:
2315 				kve->kve_type = KVME_TYPE_UNKNOWN;
2316 				break;
2317 			}
2318 			if (lobj != obj)
2319 				VM_OBJECT_RUNLOCK(lobj);
2320 
2321 			kve->kve_ref_count = obj->ref_count;
2322 			kve->kve_shadow_count = obj->shadow_count;
2323 			VM_OBJECT_RUNLOCK(obj);
2324 			if (vp != NULL) {
2325 				vn_fullpath(curthread, vp, &fullpath,
2326 				    &freepath);
2327 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2328 				cred = curthread->td_ucred;
2329 				vn_lock(vp, LK_SHARED | LK_RETRY);
2330 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2331 					kve->kve_vn_fileid = va.va_fileid;
2332 					kve->kve_vn_fsid = va.va_fsid;
2333 					kve->kve_vn_mode =
2334 					    MAKEIMODE(va.va_type, va.va_mode);
2335 					kve->kve_vn_size = va.va_size;
2336 					kve->kve_vn_rdev = va.va_rdev;
2337 					kve->kve_status = KF_ATTR_VALID;
2338 				}
2339 				vput(vp);
2340 			}
2341 		} else {
2342 			kve->kve_type = KVME_TYPE_NONE;
2343 			kve->kve_ref_count = 0;
2344 			kve->kve_shadow_count = 0;
2345 		}
2346 
2347 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2348 		if (freepath != NULL)
2349 			free(freepath, M_TEMP);
2350 
2351 		/* Pack record size down */
2352 		kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) +
2353 		    strlen(kve->kve_path) + 1;
2354 		kve->kve_structsize = roundup(kve->kve_structsize,
2355 		    sizeof(uint64_t));
2356 		error = sbuf_bcat(sb, kve, kve->kve_structsize);
2357 		vm_map_lock_read(map);
2358 		if (error)
2359 			break;
2360 		if (last_timestamp != map->timestamp) {
2361 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2362 			entry = tmp_entry;
2363 		}
2364 	}
2365 	vm_map_unlock_read(map);
2366 	vmspace_free(vm);
2367 	PRELE(p);
2368 	free(kve, M_TEMP);
2369 	return (error);
2370 }
2371 
2372 static int
2373 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2374 {
2375 	struct proc *p;
2376 	struct sbuf sb;
2377 	int error, error2, *name;
2378 
2379 	name = (int *)arg1;
2380 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2381 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2382 	if (error != 0) {
2383 		sbuf_delete(&sb);
2384 		return (error);
2385 	}
2386 	error = kern_proc_vmmap_out(p, &sb);
2387 	error2 = sbuf_finish(&sb);
2388 	sbuf_delete(&sb);
2389 	return (error != 0 ? error : error2);
2390 }
2391 
2392 #if defined(STACK) || defined(DDB)
2393 static int
2394 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2395 {
2396 	struct kinfo_kstack *kkstp;
2397 	int error, i, *name, numthreads;
2398 	lwpid_t *lwpidarray;
2399 	struct thread *td;
2400 	struct stack *st;
2401 	struct sbuf sb;
2402 	struct proc *p;
2403 
2404 	name = (int *)arg1;
2405 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2406 	if (error != 0)
2407 		return (error);
2408 
2409 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2410 	st = stack_create();
2411 
2412 	lwpidarray = NULL;
2413 	numthreads = 0;
2414 	PROC_LOCK(p);
2415 repeat:
2416 	if (numthreads < p->p_numthreads) {
2417 		if (lwpidarray != NULL) {
2418 			free(lwpidarray, M_TEMP);
2419 			lwpidarray = NULL;
2420 		}
2421 		numthreads = p->p_numthreads;
2422 		PROC_UNLOCK(p);
2423 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2424 		    M_WAITOK | M_ZERO);
2425 		PROC_LOCK(p);
2426 		goto repeat;
2427 	}
2428 	i = 0;
2429 
2430 	/*
2431 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2432 	 * of changes could have taken place.  Should we check to see if the
2433 	 * vmspace has been replaced, or the like, in order to prevent
2434 	 * giving a snapshot that spans, say, execve(2), with some threads
2435 	 * before and some after?  Among other things, the credentials could
2436 	 * have changed, in which case the right to extract debug info might
2437 	 * no longer be assured.
2438 	 */
2439 	FOREACH_THREAD_IN_PROC(p, td) {
2440 		KASSERT(i < numthreads,
2441 		    ("sysctl_kern_proc_kstack: numthreads"));
2442 		lwpidarray[i] = td->td_tid;
2443 		i++;
2444 	}
2445 	numthreads = i;
2446 	for (i = 0; i < numthreads; i++) {
2447 		td = thread_find(p, lwpidarray[i]);
2448 		if (td == NULL) {
2449 			continue;
2450 		}
2451 		bzero(kkstp, sizeof(*kkstp));
2452 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2453 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2454 		thread_lock(td);
2455 		kkstp->kkst_tid = td->td_tid;
2456 		if (TD_IS_SWAPPED(td))
2457 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2458 		else if (TD_IS_RUNNING(td))
2459 			kkstp->kkst_state = KKST_STATE_RUNNING;
2460 		else {
2461 			kkstp->kkst_state = KKST_STATE_STACKOK;
2462 			stack_save_td(st, td);
2463 		}
2464 		thread_unlock(td);
2465 		PROC_UNLOCK(p);
2466 		stack_sbuf_print(&sb, st);
2467 		sbuf_finish(&sb);
2468 		sbuf_delete(&sb);
2469 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2470 		PROC_LOCK(p);
2471 		if (error)
2472 			break;
2473 	}
2474 	_PRELE(p);
2475 	PROC_UNLOCK(p);
2476 	if (lwpidarray != NULL)
2477 		free(lwpidarray, M_TEMP);
2478 	stack_destroy(st);
2479 	free(kkstp, M_TEMP);
2480 	return (error);
2481 }
2482 #endif
2483 
2484 /*
2485  * This sysctl allows a process to retrieve the full list of groups from
2486  * itself or another process.
2487  */
2488 static int
2489 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2490 {
2491 	pid_t *pidp = (pid_t *)arg1;
2492 	unsigned int arglen = arg2;
2493 	struct proc *p;
2494 	struct ucred *cred;
2495 	int error;
2496 
2497 	if (arglen != 1)
2498 		return (EINVAL);
2499 	if (*pidp == -1) {	/* -1 means this process */
2500 		p = req->td->td_proc;
2501 	} else {
2502 		error = pget(*pidp, PGET_CANSEE, &p);
2503 		if (error != 0)
2504 			return (error);
2505 	}
2506 
2507 	cred = crhold(p->p_ucred);
2508 	if (*pidp != -1)
2509 		PROC_UNLOCK(p);
2510 
2511 	error = SYSCTL_OUT(req, cred->cr_groups,
2512 	    cred->cr_ngroups * sizeof(gid_t));
2513 	crfree(cred);
2514 	return (error);
2515 }
2516 
2517 /*
2518  * This sysctl allows a process to retrieve or/and set the resource limit for
2519  * another process.
2520  */
2521 static int
2522 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2523 {
2524 	int *name = (int *)arg1;
2525 	u_int namelen = arg2;
2526 	struct rlimit rlim;
2527 	struct proc *p;
2528 	u_int which;
2529 	int flags, error;
2530 
2531 	if (namelen != 2)
2532 		return (EINVAL);
2533 
2534 	which = (u_int)name[1];
2535 	if (which >= RLIM_NLIMITS)
2536 		return (EINVAL);
2537 
2538 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2539 		return (EINVAL);
2540 
2541 	flags = PGET_HOLD | PGET_NOTWEXIT;
2542 	if (req->newptr != NULL)
2543 		flags |= PGET_CANDEBUG;
2544 	else
2545 		flags |= PGET_CANSEE;
2546 	error = pget((pid_t)name[0], flags, &p);
2547 	if (error != 0)
2548 		return (error);
2549 
2550 	/*
2551 	 * Retrieve limit.
2552 	 */
2553 	if (req->oldptr != NULL) {
2554 		PROC_LOCK(p);
2555 		lim_rlimit(p, which, &rlim);
2556 		PROC_UNLOCK(p);
2557 	}
2558 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2559 	if (error != 0)
2560 		goto errout;
2561 
2562 	/*
2563 	 * Set limit.
2564 	 */
2565 	if (req->newptr != NULL) {
2566 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2567 		if (error == 0)
2568 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2569 	}
2570 
2571 errout:
2572 	PRELE(p);
2573 	return (error);
2574 }
2575 
2576 /*
2577  * This sysctl allows a process to retrieve ps_strings structure location of
2578  * another process.
2579  */
2580 static int
2581 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2582 {
2583 	int *name = (int *)arg1;
2584 	u_int namelen = arg2;
2585 	struct proc *p;
2586 	vm_offset_t ps_strings;
2587 	int error;
2588 #ifdef COMPAT_FREEBSD32
2589 	uint32_t ps_strings32;
2590 #endif
2591 
2592 	if (namelen != 1)
2593 		return (EINVAL);
2594 
2595 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2596 	if (error != 0)
2597 		return (error);
2598 #ifdef COMPAT_FREEBSD32
2599 	if ((req->flags & SCTL_MASK32) != 0) {
2600 		/*
2601 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2602 		 * process.
2603 		 */
2604 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2605 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2606 		PROC_UNLOCK(p);
2607 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2608 		return (error);
2609 	}
2610 #endif
2611 	ps_strings = p->p_sysent->sv_psstrings;
2612 	PROC_UNLOCK(p);
2613 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2614 	return (error);
2615 }
2616 
2617 /*
2618  * This sysctl allows a process to retrieve umask of another process.
2619  */
2620 static int
2621 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2622 {
2623 	int *name = (int *)arg1;
2624 	u_int namelen = arg2;
2625 	struct proc *p;
2626 	int error;
2627 	u_short fd_cmask;
2628 
2629 	if (namelen != 1)
2630 		return (EINVAL);
2631 
2632 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2633 	if (error != 0)
2634 		return (error);
2635 
2636 	FILEDESC_SLOCK(p->p_fd);
2637 	fd_cmask = p->p_fd->fd_cmask;
2638 	FILEDESC_SUNLOCK(p->p_fd);
2639 	PRELE(p);
2640 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2641 	return (error);
2642 }
2643 
2644 /*
2645  * This sysctl allows a process to set and retrieve binary osreldate of
2646  * another process.
2647  */
2648 static int
2649 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2650 {
2651 	int *name = (int *)arg1;
2652 	u_int namelen = arg2;
2653 	struct proc *p;
2654 	int flags, error, osrel;
2655 
2656 	if (namelen != 1)
2657 		return (EINVAL);
2658 
2659 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2660 		return (EINVAL);
2661 
2662 	flags = PGET_HOLD | PGET_NOTWEXIT;
2663 	if (req->newptr != NULL)
2664 		flags |= PGET_CANDEBUG;
2665 	else
2666 		flags |= PGET_CANSEE;
2667 	error = pget((pid_t)name[0], flags, &p);
2668 	if (error != 0)
2669 		return (error);
2670 
2671 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2672 	if (error != 0)
2673 		goto errout;
2674 
2675 	if (req->newptr != NULL) {
2676 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2677 		if (error != 0)
2678 			goto errout;
2679 		if (osrel < 0) {
2680 			error = EINVAL;
2681 			goto errout;
2682 		}
2683 		p->p_osrel = osrel;
2684 	}
2685 errout:
2686 	PRELE(p);
2687 	return (error);
2688 }
2689 
2690 static int
2691 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2692 {
2693 	int *name = (int *)arg1;
2694 	u_int namelen = arg2;
2695 	struct proc *p;
2696 	struct kinfo_sigtramp kst;
2697 	const struct sysentvec *sv;
2698 	int error;
2699 #ifdef COMPAT_FREEBSD32
2700 	struct kinfo_sigtramp32 kst32;
2701 #endif
2702 
2703 	if (namelen != 1)
2704 		return (EINVAL);
2705 
2706 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2707 	if (error != 0)
2708 		return (error);
2709 	sv = p->p_sysent;
2710 #ifdef COMPAT_FREEBSD32
2711 	if ((req->flags & SCTL_MASK32) != 0) {
2712 		bzero(&kst32, sizeof(kst32));
2713 		if (SV_PROC_FLAG(p, SV_ILP32)) {
2714 			if (sv->sv_sigcode_base != 0) {
2715 				kst32.ksigtramp_start = sv->sv_sigcode_base;
2716 				kst32.ksigtramp_end = sv->sv_sigcode_base +
2717 				    *sv->sv_szsigcode;
2718 			} else {
2719 				kst32.ksigtramp_start = sv->sv_psstrings -
2720 				    *sv->sv_szsigcode;
2721 				kst32.ksigtramp_end = sv->sv_psstrings;
2722 			}
2723 		}
2724 		PROC_UNLOCK(p);
2725 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2726 		return (error);
2727 	}
2728 #endif
2729 	bzero(&kst, sizeof(kst));
2730 	if (sv->sv_sigcode_base != 0) {
2731 		kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2732 		kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2733 		    *sv->sv_szsigcode;
2734 	} else {
2735 		kst.ksigtramp_start = (char *)sv->sv_psstrings -
2736 		    *sv->sv_szsigcode;
2737 		kst.ksigtramp_end = (char *)sv->sv_psstrings;
2738 	}
2739 	PROC_UNLOCK(p);
2740 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
2741 	return (error);
2742 }
2743 
2744 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2745 
2746 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2747 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2748 	"Return entire process table");
2749 
2750 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2751 	sysctl_kern_proc, "Process table");
2752 
2753 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2754 	sysctl_kern_proc, "Process table");
2755 
2756 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2757 	sysctl_kern_proc, "Process table");
2758 
2759 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2760 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2761 
2762 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2763 	sysctl_kern_proc, "Process table");
2764 
2765 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2766 	sysctl_kern_proc, "Process table");
2767 
2768 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2769 	sysctl_kern_proc, "Process table");
2770 
2771 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2772 	sysctl_kern_proc, "Process table");
2773 
2774 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2775 	sysctl_kern_proc, "Return process table, no threads");
2776 
2777 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2778 	CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2779 	sysctl_kern_proc_args, "Process argument list");
2780 
2781 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
2782 	sysctl_kern_proc_env, "Process environment");
2783 
2784 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
2785 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
2786 
2787 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2788 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2789 
2790 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2791 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2792 	"Process syscall vector name (ABI type)");
2793 
2794 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2795 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2796 
2797 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2798 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2799 
2800 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2801 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2802 
2803 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2804 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2805 
2806 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2807 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2808 
2809 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2810 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2811 
2812 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2813 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2814 
2815 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2816 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2817 
2818 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2819 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2820 	"Return process table, no threads");
2821 
2822 #ifdef COMPAT_FREEBSD7
2823 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2824 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2825 #endif
2826 
2827 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2828 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2829 
2830 #if defined(STACK) || defined(DDB)
2831 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2832 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2833 #endif
2834 
2835 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2836 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
2837 
2838 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
2839 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
2840 	"Process resource limits");
2841 
2842 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
2843 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
2844 	"Process ps_strings location");
2845 
2846 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
2847 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
2848 
2849 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
2850 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
2851 	"Process binary osreldate");
2852 
2853 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
2854 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
2855 	"Process signal trampoline location");
2856