xref: /freebsd/sys/kern/kern_proc.c (revision 6574b8ed19b093f0af09501d2c9676c28993cb97)
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 /*
932  * Fill in information that is thread specific.  Must be called with
933  * target process locked.  If 'preferthread' is set, overwrite certain
934  * process-related fields that are maintained for both threads and
935  * processes.
936  */
937 static void
938 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
939 {
940 	struct proc *p;
941 
942 	p = td->td_proc;
943 	kp->ki_tdaddr = td;
944 	PROC_LOCK_ASSERT(p, MA_OWNED);
945 
946 	if (preferthread)
947 		PROC_SLOCK(p);
948 	thread_lock(td);
949 	if (td->td_wmesg != NULL)
950 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
951 	else
952 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
953 	strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname));
954 	if (TD_ON_LOCK(td)) {
955 		kp->ki_kiflag |= KI_LOCKBLOCK;
956 		strlcpy(kp->ki_lockname, td->td_lockname,
957 		    sizeof(kp->ki_lockname));
958 	} else {
959 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
960 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
961 	}
962 
963 	if (p->p_state == PRS_NORMAL) { /* approximate. */
964 		if (TD_ON_RUNQ(td) ||
965 		    TD_CAN_RUN(td) ||
966 		    TD_IS_RUNNING(td)) {
967 			kp->ki_stat = SRUN;
968 		} else if (P_SHOULDSTOP(p)) {
969 			kp->ki_stat = SSTOP;
970 		} else if (TD_IS_SLEEPING(td)) {
971 			kp->ki_stat = SSLEEP;
972 		} else if (TD_ON_LOCK(td)) {
973 			kp->ki_stat = SLOCK;
974 		} else {
975 			kp->ki_stat = SWAIT;
976 		}
977 	} else if (p->p_state == PRS_ZOMBIE) {
978 		kp->ki_stat = SZOMB;
979 	} else {
980 		kp->ki_stat = SIDL;
981 	}
982 
983 	/* Things in the thread */
984 	kp->ki_wchan = td->td_wchan;
985 	kp->ki_pri.pri_level = td->td_priority;
986 	kp->ki_pri.pri_native = td->td_base_pri;
987 	kp->ki_lastcpu = td->td_lastcpu;
988 	kp->ki_oncpu = td->td_oncpu;
989 	kp->ki_tdflags = td->td_flags;
990 	kp->ki_tid = td->td_tid;
991 	kp->ki_numthreads = p->p_numthreads;
992 	kp->ki_pcb = td->td_pcb;
993 	kp->ki_kstack = (void *)td->td_kstack;
994 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
995 	kp->ki_pri.pri_class = td->td_pri_class;
996 	kp->ki_pri.pri_user = td->td_user_pri;
997 
998 	if (preferthread) {
999 		rufetchtd(td, &kp->ki_rusage);
1000 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1001 		kp->ki_pctcpu = sched_pctcpu(td);
1002 		kp->ki_estcpu = td->td_estcpu;
1003 		kp->ki_cow = td->td_cow;
1004 	}
1005 
1006 	/* We can't get this anymore but ps etc never used it anyway. */
1007 	kp->ki_rqindex = 0;
1008 
1009 	if (preferthread)
1010 		kp->ki_siglist = td->td_siglist;
1011 	kp->ki_sigmask = td->td_sigmask;
1012 	thread_unlock(td);
1013 	if (preferthread)
1014 		PROC_SUNLOCK(p);
1015 }
1016 
1017 /*
1018  * Fill in a kinfo_proc structure for the specified process.
1019  * Must be called with the target process locked.
1020  */
1021 void
1022 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1023 {
1024 
1025 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1026 
1027 	fill_kinfo_proc_only(p, kp);
1028 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1029 	fill_kinfo_aggregate(p, kp);
1030 }
1031 
1032 struct pstats *
1033 pstats_alloc(void)
1034 {
1035 
1036 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1037 }
1038 
1039 /*
1040  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1041  */
1042 void
1043 pstats_fork(struct pstats *src, struct pstats *dst)
1044 {
1045 
1046 	bzero(&dst->pstat_startzero,
1047 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1048 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1049 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1050 }
1051 
1052 void
1053 pstats_free(struct pstats *ps)
1054 {
1055 
1056 	free(ps, M_SUBPROC);
1057 }
1058 
1059 static struct proc *
1060 zpfind_locked(pid_t pid)
1061 {
1062 	struct proc *p;
1063 
1064 	sx_assert(&allproc_lock, SX_LOCKED);
1065 	LIST_FOREACH(p, &zombproc, p_list) {
1066 		if (p->p_pid == pid) {
1067 			PROC_LOCK(p);
1068 			break;
1069 		}
1070 	}
1071 	return (p);
1072 }
1073 
1074 /*
1075  * Locate a zombie process by number
1076  */
1077 struct proc *
1078 zpfind(pid_t pid)
1079 {
1080 	struct proc *p;
1081 
1082 	sx_slock(&allproc_lock);
1083 	p = zpfind_locked(pid);
1084 	sx_sunlock(&allproc_lock);
1085 	return (p);
1086 }
1087 
1088 #ifdef COMPAT_FREEBSD32
1089 
1090 /*
1091  * This function is typically used to copy out the kernel address, so
1092  * it can be replaced by assignment of zero.
1093  */
1094 static inline uint32_t
1095 ptr32_trim(void *ptr)
1096 {
1097 	uintptr_t uptr;
1098 
1099 	uptr = (uintptr_t)ptr;
1100 	return ((uptr > UINT_MAX) ? 0 : uptr);
1101 }
1102 
1103 #define PTRTRIM_CP(src,dst,fld) \
1104 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1105 
1106 static void
1107 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1108 {
1109 	int i;
1110 
1111 	bzero(ki32, sizeof(struct kinfo_proc32));
1112 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1113 	CP(*ki, *ki32, ki_layout);
1114 	PTRTRIM_CP(*ki, *ki32, ki_args);
1115 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1116 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1117 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1118 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1119 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1120 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1121 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1122 	CP(*ki, *ki32, ki_pid);
1123 	CP(*ki, *ki32, ki_ppid);
1124 	CP(*ki, *ki32, ki_pgid);
1125 	CP(*ki, *ki32, ki_tpgid);
1126 	CP(*ki, *ki32, ki_sid);
1127 	CP(*ki, *ki32, ki_tsid);
1128 	CP(*ki, *ki32, ki_jobc);
1129 	CP(*ki, *ki32, ki_tdev);
1130 	CP(*ki, *ki32, ki_siglist);
1131 	CP(*ki, *ki32, ki_sigmask);
1132 	CP(*ki, *ki32, ki_sigignore);
1133 	CP(*ki, *ki32, ki_sigcatch);
1134 	CP(*ki, *ki32, ki_uid);
1135 	CP(*ki, *ki32, ki_ruid);
1136 	CP(*ki, *ki32, ki_svuid);
1137 	CP(*ki, *ki32, ki_rgid);
1138 	CP(*ki, *ki32, ki_svgid);
1139 	CP(*ki, *ki32, ki_ngroups);
1140 	for (i = 0; i < KI_NGROUPS; i++)
1141 		CP(*ki, *ki32, ki_groups[i]);
1142 	CP(*ki, *ki32, ki_size);
1143 	CP(*ki, *ki32, ki_rssize);
1144 	CP(*ki, *ki32, ki_swrss);
1145 	CP(*ki, *ki32, ki_tsize);
1146 	CP(*ki, *ki32, ki_dsize);
1147 	CP(*ki, *ki32, ki_ssize);
1148 	CP(*ki, *ki32, ki_xstat);
1149 	CP(*ki, *ki32, ki_acflag);
1150 	CP(*ki, *ki32, ki_pctcpu);
1151 	CP(*ki, *ki32, ki_estcpu);
1152 	CP(*ki, *ki32, ki_slptime);
1153 	CP(*ki, *ki32, ki_swtime);
1154 	CP(*ki, *ki32, ki_cow);
1155 	CP(*ki, *ki32, ki_runtime);
1156 	TV_CP(*ki, *ki32, ki_start);
1157 	TV_CP(*ki, *ki32, ki_childtime);
1158 	CP(*ki, *ki32, ki_flag);
1159 	CP(*ki, *ki32, ki_kiflag);
1160 	CP(*ki, *ki32, ki_traceflag);
1161 	CP(*ki, *ki32, ki_stat);
1162 	CP(*ki, *ki32, ki_nice);
1163 	CP(*ki, *ki32, ki_lock);
1164 	CP(*ki, *ki32, ki_rqindex);
1165 	CP(*ki, *ki32, ki_oncpu);
1166 	CP(*ki, *ki32, ki_lastcpu);
1167 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1168 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1169 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1170 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1171 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1172 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1173 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1174 	CP(*ki, *ki32, ki_tracer);
1175 	CP(*ki, *ki32, ki_flag2);
1176 	CP(*ki, *ki32, ki_fibnum);
1177 	CP(*ki, *ki32, ki_cr_flags);
1178 	CP(*ki, *ki32, ki_jid);
1179 	CP(*ki, *ki32, ki_numthreads);
1180 	CP(*ki, *ki32, ki_tid);
1181 	CP(*ki, *ki32, ki_pri);
1182 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1183 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1184 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1185 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1186 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1187 	CP(*ki, *ki32, ki_sflag);
1188 	CP(*ki, *ki32, ki_tdflags);
1189 }
1190 #endif
1191 
1192 int
1193 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1194 {
1195 	struct thread *td;
1196 	struct kinfo_proc ki;
1197 #ifdef COMPAT_FREEBSD32
1198 	struct kinfo_proc32 ki32;
1199 #endif
1200 	int error;
1201 
1202 	PROC_LOCK_ASSERT(p, MA_OWNED);
1203 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1204 
1205 	error = 0;
1206 	fill_kinfo_proc(p, &ki);
1207 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1208 #ifdef COMPAT_FREEBSD32
1209 		if ((flags & KERN_PROC_MASK32) != 0) {
1210 			freebsd32_kinfo_proc_out(&ki, &ki32);
1211 			error = sbuf_bcat(sb, &ki32, sizeof(ki32));
1212 		} else
1213 #endif
1214 			error = sbuf_bcat(sb, &ki, sizeof(ki));
1215 	} else {
1216 		FOREACH_THREAD_IN_PROC(p, td) {
1217 			fill_kinfo_thread(td, &ki, 1);
1218 #ifdef COMPAT_FREEBSD32
1219 			if ((flags & KERN_PROC_MASK32) != 0) {
1220 				freebsd32_kinfo_proc_out(&ki, &ki32);
1221 				error = sbuf_bcat(sb, &ki32, sizeof(ki32));
1222 			} else
1223 #endif
1224 				error = sbuf_bcat(sb, &ki, sizeof(ki));
1225 			if (error)
1226 				break;
1227 		}
1228 	}
1229 	PROC_UNLOCK(p);
1230 	return (error);
1231 }
1232 
1233 static int
1234 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags,
1235     int doingzomb)
1236 {
1237 	struct sbuf sb;
1238 	struct kinfo_proc ki;
1239 	struct proc *np;
1240 	int error, error2;
1241 	pid_t pid;
1242 
1243 	pid = p->p_pid;
1244 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1245 	error = kern_proc_out(p, &sb, flags);
1246 	error2 = sbuf_finish(&sb);
1247 	sbuf_delete(&sb);
1248 	if (error != 0)
1249 		return (error);
1250 	else if (error2 != 0)
1251 		return (error2);
1252 	if (doingzomb)
1253 		np = zpfind(pid);
1254 	else {
1255 		if (pid == 0)
1256 			return (0);
1257 		np = pfind(pid);
1258 	}
1259 	if (np == NULL)
1260 		return (ESRCH);
1261 	if (np != p) {
1262 		PROC_UNLOCK(np);
1263 		return (ESRCH);
1264 	}
1265 	PROC_UNLOCK(np);
1266 	return (0);
1267 }
1268 
1269 static int
1270 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1271 {
1272 	int *name = (int *)arg1;
1273 	u_int namelen = arg2;
1274 	struct proc *p;
1275 	int flags, doingzomb, oid_number;
1276 	int error = 0;
1277 
1278 	oid_number = oidp->oid_number;
1279 	if (oid_number != KERN_PROC_ALL &&
1280 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1281 		flags = KERN_PROC_NOTHREADS;
1282 	else {
1283 		flags = 0;
1284 		oid_number &= ~KERN_PROC_INC_THREAD;
1285 	}
1286 #ifdef COMPAT_FREEBSD32
1287 	if (req->flags & SCTL_MASK32)
1288 		flags |= KERN_PROC_MASK32;
1289 #endif
1290 	if (oid_number == KERN_PROC_PID) {
1291 		if (namelen != 1)
1292 			return (EINVAL);
1293 		error = sysctl_wire_old_buffer(req, 0);
1294 		if (error)
1295 			return (error);
1296 		sx_slock(&proctree_lock);
1297 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1298 		if (error == 0)
1299 			error = sysctl_out_proc(p, req, flags, 0);
1300 		sx_sunlock(&proctree_lock);
1301 		return (error);
1302 	}
1303 
1304 	switch (oid_number) {
1305 	case KERN_PROC_ALL:
1306 		if (namelen != 0)
1307 			return (EINVAL);
1308 		break;
1309 	case KERN_PROC_PROC:
1310 		if (namelen != 0 && namelen != 1)
1311 			return (EINVAL);
1312 		break;
1313 	default:
1314 		if (namelen != 1)
1315 			return (EINVAL);
1316 		break;
1317 	}
1318 
1319 	if (!req->oldptr) {
1320 		/* overestimate by 5 procs */
1321 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1322 		if (error)
1323 			return (error);
1324 	}
1325 	error = sysctl_wire_old_buffer(req, 0);
1326 	if (error != 0)
1327 		return (error);
1328 	sx_slock(&proctree_lock);
1329 	sx_slock(&allproc_lock);
1330 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1331 		if (!doingzomb)
1332 			p = LIST_FIRST(&allproc);
1333 		else
1334 			p = LIST_FIRST(&zombproc);
1335 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
1336 			/*
1337 			 * Skip embryonic processes.
1338 			 */
1339 			PROC_LOCK(p);
1340 			if (p->p_state == PRS_NEW) {
1341 				PROC_UNLOCK(p);
1342 				continue;
1343 			}
1344 			KASSERT(p->p_ucred != NULL,
1345 			    ("process credential is NULL for non-NEW proc"));
1346 			/*
1347 			 * Show a user only appropriate processes.
1348 			 */
1349 			if (p_cansee(curthread, p)) {
1350 				PROC_UNLOCK(p);
1351 				continue;
1352 			}
1353 			/*
1354 			 * TODO - make more efficient (see notes below).
1355 			 * do by session.
1356 			 */
1357 			switch (oid_number) {
1358 
1359 			case KERN_PROC_GID:
1360 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1361 					PROC_UNLOCK(p);
1362 					continue;
1363 				}
1364 				break;
1365 
1366 			case KERN_PROC_PGRP:
1367 				/* could do this by traversing pgrp */
1368 				if (p->p_pgrp == NULL ||
1369 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1370 					PROC_UNLOCK(p);
1371 					continue;
1372 				}
1373 				break;
1374 
1375 			case KERN_PROC_RGID:
1376 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1377 					PROC_UNLOCK(p);
1378 					continue;
1379 				}
1380 				break;
1381 
1382 			case KERN_PROC_SESSION:
1383 				if (p->p_session == NULL ||
1384 				    p->p_session->s_sid != (pid_t)name[0]) {
1385 					PROC_UNLOCK(p);
1386 					continue;
1387 				}
1388 				break;
1389 
1390 			case KERN_PROC_TTY:
1391 				if ((p->p_flag & P_CONTROLT) == 0 ||
1392 				    p->p_session == NULL) {
1393 					PROC_UNLOCK(p);
1394 					continue;
1395 				}
1396 				/* XXX proctree_lock */
1397 				SESS_LOCK(p->p_session);
1398 				if (p->p_session->s_ttyp == NULL ||
1399 				    tty_udev(p->p_session->s_ttyp) !=
1400 				    (dev_t)name[0]) {
1401 					SESS_UNLOCK(p->p_session);
1402 					PROC_UNLOCK(p);
1403 					continue;
1404 				}
1405 				SESS_UNLOCK(p->p_session);
1406 				break;
1407 
1408 			case KERN_PROC_UID:
1409 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1410 					PROC_UNLOCK(p);
1411 					continue;
1412 				}
1413 				break;
1414 
1415 			case KERN_PROC_RUID:
1416 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1417 					PROC_UNLOCK(p);
1418 					continue;
1419 				}
1420 				break;
1421 
1422 			case KERN_PROC_PROC:
1423 				break;
1424 
1425 			default:
1426 				break;
1427 
1428 			}
1429 
1430 			error = sysctl_out_proc(p, req, flags, doingzomb);
1431 			if (error) {
1432 				sx_sunlock(&allproc_lock);
1433 				sx_sunlock(&proctree_lock);
1434 				return (error);
1435 			}
1436 		}
1437 	}
1438 	sx_sunlock(&allproc_lock);
1439 	sx_sunlock(&proctree_lock);
1440 	return (0);
1441 }
1442 
1443 struct pargs *
1444 pargs_alloc(int len)
1445 {
1446 	struct pargs *pa;
1447 
1448 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1449 		M_WAITOK);
1450 	refcount_init(&pa->ar_ref, 1);
1451 	pa->ar_length = len;
1452 	return (pa);
1453 }
1454 
1455 static void
1456 pargs_free(struct pargs *pa)
1457 {
1458 
1459 	free(pa, M_PARGS);
1460 }
1461 
1462 void
1463 pargs_hold(struct pargs *pa)
1464 {
1465 
1466 	if (pa == NULL)
1467 		return;
1468 	refcount_acquire(&pa->ar_ref);
1469 }
1470 
1471 void
1472 pargs_drop(struct pargs *pa)
1473 {
1474 
1475 	if (pa == NULL)
1476 		return;
1477 	if (refcount_release(&pa->ar_ref))
1478 		pargs_free(pa);
1479 }
1480 
1481 static int
1482 proc_read_mem(struct thread *td, struct proc *p, vm_offset_t offset, void* buf,
1483     size_t len)
1484 {
1485 	struct iovec iov;
1486 	struct uio uio;
1487 
1488 	iov.iov_base = (caddr_t)buf;
1489 	iov.iov_len = len;
1490 	uio.uio_iov = &iov;
1491 	uio.uio_iovcnt = 1;
1492 	uio.uio_offset = offset;
1493 	uio.uio_resid = (ssize_t)len;
1494 	uio.uio_segflg = UIO_SYSSPACE;
1495 	uio.uio_rw = UIO_READ;
1496 	uio.uio_td = td;
1497 
1498 	return (proc_rwmem(p, &uio));
1499 }
1500 
1501 static int
1502 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1503     size_t len)
1504 {
1505 	size_t i;
1506 	int error;
1507 
1508 	error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, len);
1509 	/*
1510 	 * Reading the chunk may validly return EFAULT if the string is shorter
1511 	 * than the chunk and is aligned at the end of the page, assuming the
1512 	 * next page is not mapped.  So if EFAULT is returned do a fallback to
1513 	 * one byte read loop.
1514 	 */
1515 	if (error == EFAULT) {
1516 		for (i = 0; i < len; i++, buf++, sptr++) {
1517 			error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, 1);
1518 			if (error != 0)
1519 				return (error);
1520 			if (*buf == '\0')
1521 				break;
1522 		}
1523 		error = 0;
1524 	}
1525 	return (error);
1526 }
1527 
1528 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1529 
1530 enum proc_vector_type {
1531 	PROC_ARG,
1532 	PROC_ENV,
1533 	PROC_AUX,
1534 };
1535 
1536 #ifdef COMPAT_FREEBSD32
1537 static int
1538 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1539     size_t *vsizep, enum proc_vector_type type)
1540 {
1541 	struct freebsd32_ps_strings pss;
1542 	Elf32_Auxinfo aux;
1543 	vm_offset_t vptr, ptr;
1544 	uint32_t *proc_vector32;
1545 	char **proc_vector;
1546 	size_t vsize, size;
1547 	int i, error;
1548 
1549 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1550 	    &pss, sizeof(pss));
1551 	if (error != 0)
1552 		return (error);
1553 	switch (type) {
1554 	case PROC_ARG:
1555 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1556 		vsize = pss.ps_nargvstr;
1557 		if (vsize > ARG_MAX)
1558 			return (ENOEXEC);
1559 		size = vsize * sizeof(int32_t);
1560 		break;
1561 	case PROC_ENV:
1562 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1563 		vsize = pss.ps_nenvstr;
1564 		if (vsize > ARG_MAX)
1565 			return (ENOEXEC);
1566 		size = vsize * sizeof(int32_t);
1567 		break;
1568 	case PROC_AUX:
1569 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1570 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1571 		if (vptr % 4 != 0)
1572 			return (ENOEXEC);
1573 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1574 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1575 			if (error != 0)
1576 				return (error);
1577 			if (aux.a_type == AT_NULL)
1578 				break;
1579 			ptr += sizeof(aux);
1580 		}
1581 		if (aux.a_type != AT_NULL)
1582 			return (ENOEXEC);
1583 		vsize = i + 1;
1584 		size = vsize * sizeof(aux);
1585 		break;
1586 	default:
1587 		KASSERT(0, ("Wrong proc vector type: %d", type));
1588 		return (EINVAL);
1589 	}
1590 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1591 	error = proc_read_mem(td, p, vptr, proc_vector32, size);
1592 	if (error != 0)
1593 		goto done;
1594 	if (type == PROC_AUX) {
1595 		*proc_vectorp = (char **)proc_vector32;
1596 		*vsizep = vsize;
1597 		return (0);
1598 	}
1599 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1600 	for (i = 0; i < (int)vsize; i++)
1601 		proc_vector[i] = PTRIN(proc_vector32[i]);
1602 	*proc_vectorp = proc_vector;
1603 	*vsizep = vsize;
1604 done:
1605 	free(proc_vector32, M_TEMP);
1606 	return (error);
1607 }
1608 #endif
1609 
1610 static int
1611 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1612     size_t *vsizep, enum proc_vector_type type)
1613 {
1614 	struct ps_strings pss;
1615 	Elf_Auxinfo aux;
1616 	vm_offset_t vptr, ptr;
1617 	char **proc_vector;
1618 	size_t vsize, size;
1619 	int error, i;
1620 
1621 #ifdef COMPAT_FREEBSD32
1622 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1623 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1624 #endif
1625 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1626 	    &pss, sizeof(pss));
1627 	if (error != 0)
1628 		return (error);
1629 	switch (type) {
1630 	case PROC_ARG:
1631 		vptr = (vm_offset_t)pss.ps_argvstr;
1632 		vsize = pss.ps_nargvstr;
1633 		if (vsize > ARG_MAX)
1634 			return (ENOEXEC);
1635 		size = vsize * sizeof(char *);
1636 		break;
1637 	case PROC_ENV:
1638 		vptr = (vm_offset_t)pss.ps_envstr;
1639 		vsize = pss.ps_nenvstr;
1640 		if (vsize > ARG_MAX)
1641 			return (ENOEXEC);
1642 		size = vsize * sizeof(char *);
1643 		break;
1644 	case PROC_AUX:
1645 		/*
1646 		 * The aux array is just above env array on the stack. Check
1647 		 * that the address is naturally aligned.
1648 		 */
1649 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1650 		    * sizeof(char *);
1651 #if __ELF_WORD_SIZE == 64
1652 		if (vptr % sizeof(uint64_t) != 0)
1653 #else
1654 		if (vptr % sizeof(uint32_t) != 0)
1655 #endif
1656 			return (ENOEXEC);
1657 		/*
1658 		 * We count the array size reading the aux vectors from the
1659 		 * stack until AT_NULL vector is returned.  So (to keep the code
1660 		 * simple) we read the process stack twice: the first time here
1661 		 * to find the size and the second time when copying the vectors
1662 		 * to the allocated proc_vector.
1663 		 */
1664 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1665 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1666 			if (error != 0)
1667 				return (error);
1668 			if (aux.a_type == AT_NULL)
1669 				break;
1670 			ptr += sizeof(aux);
1671 		}
1672 		/*
1673 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1674 		 * not reached AT_NULL, it is most likely we are reading wrong
1675 		 * data: either the process doesn't have auxv array or data has
1676 		 * been modified. Return the error in this case.
1677 		 */
1678 		if (aux.a_type != AT_NULL)
1679 			return (ENOEXEC);
1680 		vsize = i + 1;
1681 		size = vsize * sizeof(aux);
1682 		break;
1683 	default:
1684 		KASSERT(0, ("Wrong proc vector type: %d", type));
1685 		return (EINVAL); /* In case we are built without INVARIANTS. */
1686 	}
1687 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1688 	if (proc_vector == NULL)
1689 		return (ENOMEM);
1690 	error = proc_read_mem(td, p, vptr, proc_vector, size);
1691 	if (error != 0) {
1692 		free(proc_vector, M_TEMP);
1693 		return (error);
1694 	}
1695 	*proc_vectorp = proc_vector;
1696 	*vsizep = vsize;
1697 
1698 	return (0);
1699 }
1700 
1701 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1702 
1703 static int
1704 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1705     enum proc_vector_type type)
1706 {
1707 	size_t done, len, nchr, vsize;
1708 	int error, i;
1709 	char **proc_vector, *sptr;
1710 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1711 
1712 	PROC_ASSERT_HELD(p);
1713 
1714 	/*
1715 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1716 	 */
1717 	nchr = 2 * (PATH_MAX + ARG_MAX);
1718 
1719 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1720 	if (error != 0)
1721 		return (error);
1722 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1723 		/*
1724 		 * The program may have scribbled into its argv array, e.g. to
1725 		 * remove some arguments.  If that has happened, break out
1726 		 * before trying to read from NULL.
1727 		 */
1728 		if (proc_vector[i] == NULL)
1729 			break;
1730 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1731 			error = proc_read_string(td, p, sptr, pss_string,
1732 			    sizeof(pss_string));
1733 			if (error != 0)
1734 				goto done;
1735 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1736 			if (done + len >= nchr)
1737 				len = nchr - done - 1;
1738 			sbuf_bcat(sb, pss_string, len);
1739 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1740 				break;
1741 			done += GET_PS_STRINGS_CHUNK_SZ;
1742 		}
1743 		sbuf_bcat(sb, "", 1);
1744 		done += len + 1;
1745 	}
1746 done:
1747 	free(proc_vector, M_TEMP);
1748 	return (error);
1749 }
1750 
1751 int
1752 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1753 {
1754 
1755 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1756 }
1757 
1758 int
1759 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1760 {
1761 
1762 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1763 }
1764 
1765 int
1766 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1767 {
1768 	size_t vsize, size;
1769 	char **auxv;
1770 	int error;
1771 
1772 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1773 	if (error == 0) {
1774 #ifdef COMPAT_FREEBSD32
1775 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1776 			size = vsize * sizeof(Elf32_Auxinfo);
1777 		else
1778 #endif
1779 			size = vsize * sizeof(Elf_Auxinfo);
1780 		error = sbuf_bcat(sb, auxv, size);
1781 		free(auxv, M_TEMP);
1782 	}
1783 	return (error);
1784 }
1785 
1786 /*
1787  * This sysctl allows a process to retrieve the argument list or process
1788  * title for another process without groping around in the address space
1789  * of the other process.  It also allow a process to set its own "process
1790  * title to a string of its own choice.
1791  */
1792 static int
1793 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1794 {
1795 	int *name = (int *)arg1;
1796 	u_int namelen = arg2;
1797 	struct pargs *newpa, *pa;
1798 	struct proc *p;
1799 	struct sbuf sb;
1800 	int flags, error = 0, error2;
1801 
1802 	if (namelen != 1)
1803 		return (EINVAL);
1804 
1805 	flags = PGET_CANSEE;
1806 	if (req->newptr != NULL)
1807 		flags |= PGET_ISCURRENT;
1808 	error = pget((pid_t)name[0], flags, &p);
1809 	if (error)
1810 		return (error);
1811 
1812 	pa = p->p_args;
1813 	if (pa != NULL) {
1814 		pargs_hold(pa);
1815 		PROC_UNLOCK(p);
1816 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1817 		pargs_drop(pa);
1818 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1819 		_PHOLD(p);
1820 		PROC_UNLOCK(p);
1821 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1822 		error = proc_getargv(curthread, p, &sb);
1823 		error2 = sbuf_finish(&sb);
1824 		PRELE(p);
1825 		sbuf_delete(&sb);
1826 		if (error == 0 && error2 != 0)
1827 			error = error2;
1828 	} else {
1829 		PROC_UNLOCK(p);
1830 	}
1831 	if (error != 0 || req->newptr == NULL)
1832 		return (error);
1833 
1834 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1835 		return (ENOMEM);
1836 	newpa = pargs_alloc(req->newlen);
1837 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1838 	if (error != 0) {
1839 		pargs_free(newpa);
1840 		return (error);
1841 	}
1842 	PROC_LOCK(p);
1843 	pa = p->p_args;
1844 	p->p_args = newpa;
1845 	PROC_UNLOCK(p);
1846 	pargs_drop(pa);
1847 	return (0);
1848 }
1849 
1850 /*
1851  * This sysctl allows a process to retrieve environment of another process.
1852  */
1853 static int
1854 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
1855 {
1856 	int *name = (int *)arg1;
1857 	u_int namelen = arg2;
1858 	struct proc *p;
1859 	struct sbuf sb;
1860 	int error, error2;
1861 
1862 	if (namelen != 1)
1863 		return (EINVAL);
1864 
1865 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1866 	if (error != 0)
1867 		return (error);
1868 	if ((p->p_flag & P_SYSTEM) != 0) {
1869 		PRELE(p);
1870 		return (0);
1871 	}
1872 
1873 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1874 	error = proc_getenvv(curthread, p, &sb);
1875 	error2 = sbuf_finish(&sb);
1876 	PRELE(p);
1877 	sbuf_delete(&sb);
1878 	return (error != 0 ? error : error2);
1879 }
1880 
1881 /*
1882  * This sysctl allows a process to retrieve ELF auxiliary vector of
1883  * another process.
1884  */
1885 static int
1886 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
1887 {
1888 	int *name = (int *)arg1;
1889 	u_int namelen = arg2;
1890 	struct proc *p;
1891 	struct sbuf sb;
1892 	int error, error2;
1893 
1894 	if (namelen != 1)
1895 		return (EINVAL);
1896 
1897 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1898 	if (error != 0)
1899 		return (error);
1900 	if ((p->p_flag & P_SYSTEM) != 0) {
1901 		PRELE(p);
1902 		return (0);
1903 	}
1904 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1905 	error = proc_getauxv(curthread, p, &sb);
1906 	error2 = sbuf_finish(&sb);
1907 	PRELE(p);
1908 	sbuf_delete(&sb);
1909 	return (error != 0 ? error : error2);
1910 }
1911 
1912 /*
1913  * This sysctl allows a process to retrieve the path of the executable for
1914  * itself or another process.
1915  */
1916 static int
1917 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1918 {
1919 	pid_t *pidp = (pid_t *)arg1;
1920 	unsigned int arglen = arg2;
1921 	struct proc *p;
1922 	struct vnode *vp;
1923 	char *retbuf, *freebuf;
1924 	int error;
1925 
1926 	if (arglen != 1)
1927 		return (EINVAL);
1928 	if (*pidp == -1) {	/* -1 means this process */
1929 		p = req->td->td_proc;
1930 	} else {
1931 		error = pget(*pidp, PGET_CANSEE, &p);
1932 		if (error != 0)
1933 			return (error);
1934 	}
1935 
1936 	vp = p->p_textvp;
1937 	if (vp == NULL) {
1938 		if (*pidp != -1)
1939 			PROC_UNLOCK(p);
1940 		return (0);
1941 	}
1942 	vref(vp);
1943 	if (*pidp != -1)
1944 		PROC_UNLOCK(p);
1945 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1946 	vrele(vp);
1947 	if (error)
1948 		return (error);
1949 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1950 	free(freebuf, M_TEMP);
1951 	return (error);
1952 }
1953 
1954 static int
1955 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1956 {
1957 	struct proc *p;
1958 	char *sv_name;
1959 	int *name;
1960 	int namelen;
1961 	int error;
1962 
1963 	namelen = arg2;
1964 	if (namelen != 1)
1965 		return (EINVAL);
1966 
1967 	name = (int *)arg1;
1968 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
1969 	if (error != 0)
1970 		return (error);
1971 	sv_name = p->p_sysent->sv_name;
1972 	PROC_UNLOCK(p);
1973 	return (sysctl_handle_string(oidp, sv_name, 0, req));
1974 }
1975 
1976 #ifdef KINFO_OVMENTRY_SIZE
1977 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
1978 #endif
1979 
1980 #ifdef COMPAT_FREEBSD7
1981 static int
1982 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
1983 {
1984 	vm_map_entry_t entry, tmp_entry;
1985 	unsigned int last_timestamp;
1986 	char *fullpath, *freepath;
1987 	struct kinfo_ovmentry *kve;
1988 	struct vattr va;
1989 	struct ucred *cred;
1990 	int error, *name;
1991 	struct vnode *vp;
1992 	struct proc *p;
1993 	vm_map_t map;
1994 	struct vmspace *vm;
1995 
1996 	name = (int *)arg1;
1997 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1998 	if (error != 0)
1999 		return (error);
2000 	vm = vmspace_acquire_ref(p);
2001 	if (vm == NULL) {
2002 		PRELE(p);
2003 		return (ESRCH);
2004 	}
2005 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2006 
2007 	map = &vm->vm_map;
2008 	vm_map_lock_read(map);
2009 	for (entry = map->header.next; entry != &map->header;
2010 	    entry = entry->next) {
2011 		vm_object_t obj, tobj, lobj;
2012 		vm_offset_t addr;
2013 
2014 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2015 			continue;
2016 
2017 		bzero(kve, sizeof(*kve));
2018 		kve->kve_structsize = sizeof(*kve);
2019 
2020 		kve->kve_private_resident = 0;
2021 		obj = entry->object.vm_object;
2022 		if (obj != NULL) {
2023 			VM_OBJECT_RLOCK(obj);
2024 			if (obj->shadow_count == 1)
2025 				kve->kve_private_resident =
2026 				    obj->resident_page_count;
2027 		}
2028 		kve->kve_resident = 0;
2029 		addr = entry->start;
2030 		while (addr < entry->end) {
2031 			if (pmap_extract(map->pmap, addr))
2032 				kve->kve_resident++;
2033 			addr += PAGE_SIZE;
2034 		}
2035 
2036 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2037 			if (tobj != obj)
2038 				VM_OBJECT_RLOCK(tobj);
2039 			if (lobj != obj)
2040 				VM_OBJECT_RUNLOCK(lobj);
2041 			lobj = tobj;
2042 		}
2043 
2044 		kve->kve_start = (void*)entry->start;
2045 		kve->kve_end = (void*)entry->end;
2046 		kve->kve_offset = (off_t)entry->offset;
2047 
2048 		if (entry->protection & VM_PROT_READ)
2049 			kve->kve_protection |= KVME_PROT_READ;
2050 		if (entry->protection & VM_PROT_WRITE)
2051 			kve->kve_protection |= KVME_PROT_WRITE;
2052 		if (entry->protection & VM_PROT_EXECUTE)
2053 			kve->kve_protection |= KVME_PROT_EXEC;
2054 
2055 		if (entry->eflags & MAP_ENTRY_COW)
2056 			kve->kve_flags |= KVME_FLAG_COW;
2057 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2058 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2059 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2060 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2061 
2062 		last_timestamp = map->timestamp;
2063 		vm_map_unlock_read(map);
2064 
2065 		kve->kve_fileid = 0;
2066 		kve->kve_fsid = 0;
2067 		freepath = NULL;
2068 		fullpath = "";
2069 		if (lobj) {
2070 			vp = NULL;
2071 			switch (lobj->type) {
2072 			case OBJT_DEFAULT:
2073 				kve->kve_type = KVME_TYPE_DEFAULT;
2074 				break;
2075 			case OBJT_VNODE:
2076 				kve->kve_type = KVME_TYPE_VNODE;
2077 				vp = lobj->handle;
2078 				vref(vp);
2079 				break;
2080 			case OBJT_SWAP:
2081 				kve->kve_type = KVME_TYPE_SWAP;
2082 				break;
2083 			case OBJT_DEVICE:
2084 				kve->kve_type = KVME_TYPE_DEVICE;
2085 				break;
2086 			case OBJT_PHYS:
2087 				kve->kve_type = KVME_TYPE_PHYS;
2088 				break;
2089 			case OBJT_DEAD:
2090 				kve->kve_type = KVME_TYPE_DEAD;
2091 				break;
2092 			case OBJT_SG:
2093 				kve->kve_type = KVME_TYPE_SG;
2094 				break;
2095 			default:
2096 				kve->kve_type = KVME_TYPE_UNKNOWN;
2097 				break;
2098 			}
2099 			if (lobj != obj)
2100 				VM_OBJECT_RUNLOCK(lobj);
2101 
2102 			kve->kve_ref_count = obj->ref_count;
2103 			kve->kve_shadow_count = obj->shadow_count;
2104 			VM_OBJECT_RUNLOCK(obj);
2105 			if (vp != NULL) {
2106 				vn_fullpath(curthread, vp, &fullpath,
2107 				    &freepath);
2108 				cred = curthread->td_ucred;
2109 				vn_lock(vp, LK_SHARED | LK_RETRY);
2110 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2111 					kve->kve_fileid = va.va_fileid;
2112 					kve->kve_fsid = va.va_fsid;
2113 				}
2114 				vput(vp);
2115 			}
2116 		} else {
2117 			kve->kve_type = KVME_TYPE_NONE;
2118 			kve->kve_ref_count = 0;
2119 			kve->kve_shadow_count = 0;
2120 		}
2121 
2122 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2123 		if (freepath != NULL)
2124 			free(freepath, M_TEMP);
2125 
2126 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2127 		vm_map_lock_read(map);
2128 		if (error)
2129 			break;
2130 		if (last_timestamp != map->timestamp) {
2131 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2132 			entry = tmp_entry;
2133 		}
2134 	}
2135 	vm_map_unlock_read(map);
2136 	vmspace_free(vm);
2137 	PRELE(p);
2138 	free(kve, M_TEMP);
2139 	return (error);
2140 }
2141 #endif	/* COMPAT_FREEBSD7 */
2142 
2143 #ifdef KINFO_VMENTRY_SIZE
2144 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2145 #endif
2146 
2147 static void
2148 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2149     struct kinfo_vmentry *kve)
2150 {
2151 	vm_object_t obj, tobj;
2152 	vm_page_t m, m_adv;
2153 	vm_offset_t addr;
2154 	vm_paddr_t locked_pa;
2155 	vm_pindex_t pi, pi_adv, pindex;
2156 
2157 	locked_pa = 0;
2158 	obj = entry->object.vm_object;
2159 	addr = entry->start;
2160 	m_adv = NULL;
2161 	pi = OFF_TO_IDX(entry->offset);
2162 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2163 		if (m_adv != NULL) {
2164 			m = m_adv;
2165 		} else {
2166 			pi_adv = OFF_TO_IDX(entry->end - addr);
2167 			pindex = pi;
2168 			for (tobj = obj;; tobj = tobj->backing_object) {
2169 				m = vm_page_find_least(tobj, pindex);
2170 				if (m != NULL) {
2171 					if (m->pindex == pindex)
2172 						break;
2173 					if (pi_adv > m->pindex - pindex) {
2174 						pi_adv = m->pindex - pindex;
2175 						m_adv = m;
2176 					}
2177 				}
2178 				if (tobj->backing_object == NULL)
2179 					goto next;
2180 				pindex += OFF_TO_IDX(tobj->
2181 				    backing_object_offset);
2182 			}
2183 		}
2184 		m_adv = NULL;
2185 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2186 		    (addr & (pagesizes[1] - 1)) == 0 &&
2187 		    (pmap_mincore(map->pmap, addr, &locked_pa) &
2188 		    MINCORE_SUPER) != 0) {
2189 			kve->kve_flags |= KVME_FLAG_SUPER;
2190 			pi_adv = OFF_TO_IDX(pagesizes[1]);
2191 		} else {
2192 			/*
2193 			 * We do not test the found page on validity.
2194 			 * Either the page is busy and being paged in,
2195 			 * or it was invalidated.  The first case
2196 			 * should be counted as resident, the second
2197 			 * is not so clear; we do account both.
2198 			 */
2199 			pi_adv = 1;
2200 		}
2201 		kve->kve_resident += pi_adv;
2202 next:;
2203 	}
2204 	PA_UNLOCK_COND(locked_pa);
2205 }
2206 
2207 /*
2208  * Must be called with the process locked and will return unlocked.
2209  */
2210 int
2211 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb)
2212 {
2213 	vm_map_entry_t entry, tmp_entry;
2214 	struct vattr va;
2215 	vm_map_t map;
2216 	vm_object_t obj, tobj, lobj;
2217 	char *fullpath, *freepath;
2218 	struct kinfo_vmentry *kve;
2219 	struct ucred *cred;
2220 	struct vnode *vp;
2221 	struct vmspace *vm;
2222 	vm_offset_t addr;
2223 	unsigned int last_timestamp;
2224 	int error;
2225 
2226 	PROC_LOCK_ASSERT(p, MA_OWNED);
2227 
2228 	_PHOLD(p);
2229 	PROC_UNLOCK(p);
2230 	vm = vmspace_acquire_ref(p);
2231 	if (vm == NULL) {
2232 		PRELE(p);
2233 		return (ESRCH);
2234 	}
2235 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2236 
2237 	error = 0;
2238 	map = &vm->vm_map;
2239 	vm_map_lock_read(map);
2240 	for (entry = map->header.next; entry != &map->header;
2241 	    entry = entry->next) {
2242 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2243 			continue;
2244 
2245 		addr = entry->end;
2246 		bzero(kve, sizeof(*kve));
2247 		obj = entry->object.vm_object;
2248 		if (obj != NULL) {
2249 			for (tobj = obj; tobj != NULL;
2250 			    tobj = tobj->backing_object) {
2251 				VM_OBJECT_RLOCK(tobj);
2252 				lobj = tobj;
2253 			}
2254 			if (obj->backing_object == NULL)
2255 				kve->kve_private_resident =
2256 				    obj->resident_page_count;
2257 			if (!vmmap_skip_res_cnt)
2258 				kern_proc_vmmap_resident(map, entry, kve);
2259 			for (tobj = obj; tobj != NULL;
2260 			    tobj = tobj->backing_object) {
2261 				if (tobj != obj && tobj != lobj)
2262 					VM_OBJECT_RUNLOCK(tobj);
2263 			}
2264 		} else {
2265 			lobj = NULL;
2266 		}
2267 
2268 		kve->kve_start = entry->start;
2269 		kve->kve_end = entry->end;
2270 		kve->kve_offset = entry->offset;
2271 
2272 		if (entry->protection & VM_PROT_READ)
2273 			kve->kve_protection |= KVME_PROT_READ;
2274 		if (entry->protection & VM_PROT_WRITE)
2275 			kve->kve_protection |= KVME_PROT_WRITE;
2276 		if (entry->protection & VM_PROT_EXECUTE)
2277 			kve->kve_protection |= KVME_PROT_EXEC;
2278 
2279 		if (entry->eflags & MAP_ENTRY_COW)
2280 			kve->kve_flags |= KVME_FLAG_COW;
2281 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2282 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2283 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2284 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2285 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2286 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2287 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2288 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2289 
2290 		last_timestamp = map->timestamp;
2291 		vm_map_unlock_read(map);
2292 
2293 		freepath = NULL;
2294 		fullpath = "";
2295 		if (lobj != NULL) {
2296 			vp = NULL;
2297 			switch (lobj->type) {
2298 			case OBJT_DEFAULT:
2299 				kve->kve_type = KVME_TYPE_DEFAULT;
2300 				break;
2301 			case OBJT_VNODE:
2302 				kve->kve_type = KVME_TYPE_VNODE;
2303 				vp = lobj->handle;
2304 				vref(vp);
2305 				break;
2306 			case OBJT_SWAP:
2307 				kve->kve_type = KVME_TYPE_SWAP;
2308 				break;
2309 			case OBJT_DEVICE:
2310 				kve->kve_type = KVME_TYPE_DEVICE;
2311 				break;
2312 			case OBJT_PHYS:
2313 				kve->kve_type = KVME_TYPE_PHYS;
2314 				break;
2315 			case OBJT_DEAD:
2316 				kve->kve_type = KVME_TYPE_DEAD;
2317 				break;
2318 			case OBJT_SG:
2319 				kve->kve_type = KVME_TYPE_SG;
2320 				break;
2321 			case OBJT_MGTDEVICE:
2322 				kve->kve_type = KVME_TYPE_MGTDEVICE;
2323 				break;
2324 			default:
2325 				kve->kve_type = KVME_TYPE_UNKNOWN;
2326 				break;
2327 			}
2328 			if (lobj != obj)
2329 				VM_OBJECT_RUNLOCK(lobj);
2330 
2331 			kve->kve_ref_count = obj->ref_count;
2332 			kve->kve_shadow_count = obj->shadow_count;
2333 			VM_OBJECT_RUNLOCK(obj);
2334 			if (vp != NULL) {
2335 				vn_fullpath(curthread, vp, &fullpath,
2336 				    &freepath);
2337 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2338 				cred = curthread->td_ucred;
2339 				vn_lock(vp, LK_SHARED | LK_RETRY);
2340 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2341 					kve->kve_vn_fileid = va.va_fileid;
2342 					kve->kve_vn_fsid = va.va_fsid;
2343 					kve->kve_vn_mode =
2344 					    MAKEIMODE(va.va_type, va.va_mode);
2345 					kve->kve_vn_size = va.va_size;
2346 					kve->kve_vn_rdev = va.va_rdev;
2347 					kve->kve_status = KF_ATTR_VALID;
2348 				}
2349 				vput(vp);
2350 			}
2351 		} else {
2352 			kve->kve_type = KVME_TYPE_NONE;
2353 			kve->kve_ref_count = 0;
2354 			kve->kve_shadow_count = 0;
2355 		}
2356 
2357 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2358 		if (freepath != NULL)
2359 			free(freepath, M_TEMP);
2360 
2361 		/* Pack record size down */
2362 		kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) +
2363 		    strlen(kve->kve_path) + 1;
2364 		kve->kve_structsize = roundup(kve->kve_structsize,
2365 		    sizeof(uint64_t));
2366 		error = sbuf_bcat(sb, kve, kve->kve_structsize);
2367 		vm_map_lock_read(map);
2368 		if (error)
2369 			break;
2370 		if (last_timestamp != map->timestamp) {
2371 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2372 			entry = tmp_entry;
2373 		}
2374 	}
2375 	vm_map_unlock_read(map);
2376 	vmspace_free(vm);
2377 	PRELE(p);
2378 	free(kve, M_TEMP);
2379 	return (error);
2380 }
2381 
2382 static int
2383 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2384 {
2385 	struct proc *p;
2386 	struct sbuf sb;
2387 	int error, error2, *name;
2388 
2389 	name = (int *)arg1;
2390 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2391 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2392 	if (error != 0) {
2393 		sbuf_delete(&sb);
2394 		return (error);
2395 	}
2396 	error = kern_proc_vmmap_out(p, &sb);
2397 	error2 = sbuf_finish(&sb);
2398 	sbuf_delete(&sb);
2399 	return (error != 0 ? error : error2);
2400 }
2401 
2402 #if defined(STACK) || defined(DDB)
2403 static int
2404 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2405 {
2406 	struct kinfo_kstack *kkstp;
2407 	int error, i, *name, numthreads;
2408 	lwpid_t *lwpidarray;
2409 	struct thread *td;
2410 	struct stack *st;
2411 	struct sbuf sb;
2412 	struct proc *p;
2413 
2414 	name = (int *)arg1;
2415 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2416 	if (error != 0)
2417 		return (error);
2418 
2419 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2420 	st = stack_create();
2421 
2422 	lwpidarray = NULL;
2423 	numthreads = 0;
2424 	PROC_LOCK(p);
2425 repeat:
2426 	if (numthreads < p->p_numthreads) {
2427 		if (lwpidarray != NULL) {
2428 			free(lwpidarray, M_TEMP);
2429 			lwpidarray = NULL;
2430 		}
2431 		numthreads = p->p_numthreads;
2432 		PROC_UNLOCK(p);
2433 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2434 		    M_WAITOK | M_ZERO);
2435 		PROC_LOCK(p);
2436 		goto repeat;
2437 	}
2438 	i = 0;
2439 
2440 	/*
2441 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2442 	 * of changes could have taken place.  Should we check to see if the
2443 	 * vmspace has been replaced, or the like, in order to prevent
2444 	 * giving a snapshot that spans, say, execve(2), with some threads
2445 	 * before and some after?  Among other things, the credentials could
2446 	 * have changed, in which case the right to extract debug info might
2447 	 * no longer be assured.
2448 	 */
2449 	FOREACH_THREAD_IN_PROC(p, td) {
2450 		KASSERT(i < numthreads,
2451 		    ("sysctl_kern_proc_kstack: numthreads"));
2452 		lwpidarray[i] = td->td_tid;
2453 		i++;
2454 	}
2455 	numthreads = i;
2456 	for (i = 0; i < numthreads; i++) {
2457 		td = thread_find(p, lwpidarray[i]);
2458 		if (td == NULL) {
2459 			continue;
2460 		}
2461 		bzero(kkstp, sizeof(*kkstp));
2462 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2463 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2464 		thread_lock(td);
2465 		kkstp->kkst_tid = td->td_tid;
2466 		if (TD_IS_SWAPPED(td))
2467 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2468 		else if (TD_IS_RUNNING(td))
2469 			kkstp->kkst_state = KKST_STATE_RUNNING;
2470 		else {
2471 			kkstp->kkst_state = KKST_STATE_STACKOK;
2472 			stack_save_td(st, td);
2473 		}
2474 		thread_unlock(td);
2475 		PROC_UNLOCK(p);
2476 		stack_sbuf_print(&sb, st);
2477 		sbuf_finish(&sb);
2478 		sbuf_delete(&sb);
2479 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2480 		PROC_LOCK(p);
2481 		if (error)
2482 			break;
2483 	}
2484 	_PRELE(p);
2485 	PROC_UNLOCK(p);
2486 	if (lwpidarray != NULL)
2487 		free(lwpidarray, M_TEMP);
2488 	stack_destroy(st);
2489 	free(kkstp, M_TEMP);
2490 	return (error);
2491 }
2492 #endif
2493 
2494 /*
2495  * This sysctl allows a process to retrieve the full list of groups from
2496  * itself or another process.
2497  */
2498 static int
2499 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2500 {
2501 	pid_t *pidp = (pid_t *)arg1;
2502 	unsigned int arglen = arg2;
2503 	struct proc *p;
2504 	struct ucred *cred;
2505 	int error;
2506 
2507 	if (arglen != 1)
2508 		return (EINVAL);
2509 	if (*pidp == -1) {	/* -1 means this process */
2510 		p = req->td->td_proc;
2511 		PROC_LOCK(p);
2512 	} else {
2513 		error = pget(*pidp, PGET_CANSEE, &p);
2514 		if (error != 0)
2515 			return (error);
2516 	}
2517 
2518 	cred = crhold(p->p_ucred);
2519 	PROC_UNLOCK(p);
2520 
2521 	error = SYSCTL_OUT(req, cred->cr_groups,
2522 	    cred->cr_ngroups * sizeof(gid_t));
2523 	crfree(cred);
2524 	return (error);
2525 }
2526 
2527 /*
2528  * This sysctl allows a process to retrieve or/and set the resource limit for
2529  * another process.
2530  */
2531 static int
2532 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2533 {
2534 	int *name = (int *)arg1;
2535 	u_int namelen = arg2;
2536 	struct rlimit rlim;
2537 	struct proc *p;
2538 	u_int which;
2539 	int flags, error;
2540 
2541 	if (namelen != 2)
2542 		return (EINVAL);
2543 
2544 	which = (u_int)name[1];
2545 	if (which >= RLIM_NLIMITS)
2546 		return (EINVAL);
2547 
2548 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2549 		return (EINVAL);
2550 
2551 	flags = PGET_HOLD | PGET_NOTWEXIT;
2552 	if (req->newptr != NULL)
2553 		flags |= PGET_CANDEBUG;
2554 	else
2555 		flags |= PGET_CANSEE;
2556 	error = pget((pid_t)name[0], flags, &p);
2557 	if (error != 0)
2558 		return (error);
2559 
2560 	/*
2561 	 * Retrieve limit.
2562 	 */
2563 	if (req->oldptr != NULL) {
2564 		PROC_LOCK(p);
2565 		lim_rlimit(p, which, &rlim);
2566 		PROC_UNLOCK(p);
2567 	}
2568 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2569 	if (error != 0)
2570 		goto errout;
2571 
2572 	/*
2573 	 * Set limit.
2574 	 */
2575 	if (req->newptr != NULL) {
2576 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2577 		if (error == 0)
2578 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2579 	}
2580 
2581 errout:
2582 	PRELE(p);
2583 	return (error);
2584 }
2585 
2586 /*
2587  * This sysctl allows a process to retrieve ps_strings structure location of
2588  * another process.
2589  */
2590 static int
2591 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2592 {
2593 	int *name = (int *)arg1;
2594 	u_int namelen = arg2;
2595 	struct proc *p;
2596 	vm_offset_t ps_strings;
2597 	int error;
2598 #ifdef COMPAT_FREEBSD32
2599 	uint32_t ps_strings32;
2600 #endif
2601 
2602 	if (namelen != 1)
2603 		return (EINVAL);
2604 
2605 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2606 	if (error != 0)
2607 		return (error);
2608 #ifdef COMPAT_FREEBSD32
2609 	if ((req->flags & SCTL_MASK32) != 0) {
2610 		/*
2611 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2612 		 * process.
2613 		 */
2614 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2615 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2616 		PROC_UNLOCK(p);
2617 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2618 		return (error);
2619 	}
2620 #endif
2621 	ps_strings = p->p_sysent->sv_psstrings;
2622 	PROC_UNLOCK(p);
2623 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2624 	return (error);
2625 }
2626 
2627 /*
2628  * This sysctl allows a process to retrieve umask of another process.
2629  */
2630 static int
2631 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2632 {
2633 	int *name = (int *)arg1;
2634 	u_int namelen = arg2;
2635 	struct proc *p;
2636 	int error;
2637 	u_short fd_cmask;
2638 
2639 	if (namelen != 1)
2640 		return (EINVAL);
2641 
2642 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2643 	if (error != 0)
2644 		return (error);
2645 
2646 	FILEDESC_SLOCK(p->p_fd);
2647 	fd_cmask = p->p_fd->fd_cmask;
2648 	FILEDESC_SUNLOCK(p->p_fd);
2649 	PRELE(p);
2650 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2651 	return (error);
2652 }
2653 
2654 /*
2655  * This sysctl allows a process to set and retrieve binary osreldate of
2656  * another process.
2657  */
2658 static int
2659 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2660 {
2661 	int *name = (int *)arg1;
2662 	u_int namelen = arg2;
2663 	struct proc *p;
2664 	int flags, error, osrel;
2665 
2666 	if (namelen != 1)
2667 		return (EINVAL);
2668 
2669 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2670 		return (EINVAL);
2671 
2672 	flags = PGET_HOLD | PGET_NOTWEXIT;
2673 	if (req->newptr != NULL)
2674 		flags |= PGET_CANDEBUG;
2675 	else
2676 		flags |= PGET_CANSEE;
2677 	error = pget((pid_t)name[0], flags, &p);
2678 	if (error != 0)
2679 		return (error);
2680 
2681 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2682 	if (error != 0)
2683 		goto errout;
2684 
2685 	if (req->newptr != NULL) {
2686 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2687 		if (error != 0)
2688 			goto errout;
2689 		if (osrel < 0) {
2690 			error = EINVAL;
2691 			goto errout;
2692 		}
2693 		p->p_osrel = osrel;
2694 	}
2695 errout:
2696 	PRELE(p);
2697 	return (error);
2698 }
2699 
2700 static int
2701 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2702 {
2703 	int *name = (int *)arg1;
2704 	u_int namelen = arg2;
2705 	struct proc *p;
2706 	struct kinfo_sigtramp kst;
2707 	const struct sysentvec *sv;
2708 	int error;
2709 #ifdef COMPAT_FREEBSD32
2710 	struct kinfo_sigtramp32 kst32;
2711 #endif
2712 
2713 	if (namelen != 1)
2714 		return (EINVAL);
2715 
2716 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2717 	if (error != 0)
2718 		return (error);
2719 	sv = p->p_sysent;
2720 #ifdef COMPAT_FREEBSD32
2721 	if ((req->flags & SCTL_MASK32) != 0) {
2722 		bzero(&kst32, sizeof(kst32));
2723 		if (SV_PROC_FLAG(p, SV_ILP32)) {
2724 			if (sv->sv_sigcode_base != 0) {
2725 				kst32.ksigtramp_start = sv->sv_sigcode_base;
2726 				kst32.ksigtramp_end = sv->sv_sigcode_base +
2727 				    *sv->sv_szsigcode;
2728 			} else {
2729 				kst32.ksigtramp_start = sv->sv_psstrings -
2730 				    *sv->sv_szsigcode;
2731 				kst32.ksigtramp_end = sv->sv_psstrings;
2732 			}
2733 		}
2734 		PROC_UNLOCK(p);
2735 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2736 		return (error);
2737 	}
2738 #endif
2739 	bzero(&kst, sizeof(kst));
2740 	if (sv->sv_sigcode_base != 0) {
2741 		kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2742 		kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2743 		    *sv->sv_szsigcode;
2744 	} else {
2745 		kst.ksigtramp_start = (char *)sv->sv_psstrings -
2746 		    *sv->sv_szsigcode;
2747 		kst.ksigtramp_end = (char *)sv->sv_psstrings;
2748 	}
2749 	PROC_UNLOCK(p);
2750 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
2751 	return (error);
2752 }
2753 
2754 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2755 
2756 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2757 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2758 	"Return entire process table");
2759 
2760 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2761 	sysctl_kern_proc, "Process table");
2762 
2763 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2764 	sysctl_kern_proc, "Process table");
2765 
2766 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2767 	sysctl_kern_proc, "Process table");
2768 
2769 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2770 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2771 
2772 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2773 	sysctl_kern_proc, "Process table");
2774 
2775 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2776 	sysctl_kern_proc, "Process table");
2777 
2778 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2779 	sysctl_kern_proc, "Process table");
2780 
2781 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2782 	sysctl_kern_proc, "Process table");
2783 
2784 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2785 	sysctl_kern_proc, "Return process table, no threads");
2786 
2787 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2788 	CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2789 	sysctl_kern_proc_args, "Process argument list");
2790 
2791 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
2792 	sysctl_kern_proc_env, "Process environment");
2793 
2794 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
2795 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
2796 
2797 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2798 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2799 
2800 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2801 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2802 	"Process syscall vector name (ABI type)");
2803 
2804 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2805 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2806 
2807 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2808 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2809 
2810 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2811 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2812 
2813 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2814 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2815 
2816 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2817 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2818 
2819 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2820 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2821 
2822 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2823 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2824 
2825 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2826 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2827 
2828 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2829 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2830 	"Return process table, no threads");
2831 
2832 #ifdef COMPAT_FREEBSD7
2833 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2834 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2835 #endif
2836 
2837 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2838 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2839 
2840 #if defined(STACK) || defined(DDB)
2841 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2842 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2843 #endif
2844 
2845 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2846 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
2847 
2848 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
2849 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
2850 	"Process resource limits");
2851 
2852 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
2853 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
2854 	"Process ps_strings location");
2855 
2856 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
2857 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
2858 
2859 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
2860 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
2861 	"Process binary osreldate");
2862 
2863 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
2864 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
2865 	"Process signal trampoline location");
2866