xref: /freebsd/sys/kern/kern_proc.c (revision 5bd73b51076b5cb5a2c9810f76c1d7ed20c4460e)
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 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1212 				error = ENOMEM;
1213 		} else
1214 #endif
1215 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1216 				error = ENOMEM;
1217 	} else {
1218 		FOREACH_THREAD_IN_PROC(p, td) {
1219 			fill_kinfo_thread(td, &ki, 1);
1220 #ifdef COMPAT_FREEBSD32
1221 			if ((flags & KERN_PROC_MASK32) != 0) {
1222 				freebsd32_kinfo_proc_out(&ki, &ki32);
1223 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1224 					error = ENOMEM;
1225 			} else
1226 #endif
1227 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1228 					error = ENOMEM;
1229 			if (error != 0)
1230 				break;
1231 		}
1232 	}
1233 	PROC_UNLOCK(p);
1234 	return (error);
1235 }
1236 
1237 static int
1238 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags,
1239     int doingzomb)
1240 {
1241 	struct sbuf sb;
1242 	struct kinfo_proc ki;
1243 	struct proc *np;
1244 	int error, error2;
1245 	pid_t pid;
1246 
1247 	pid = p->p_pid;
1248 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1249 	error = kern_proc_out(p, &sb, flags);
1250 	error2 = sbuf_finish(&sb);
1251 	sbuf_delete(&sb);
1252 	if (error != 0)
1253 		return (error);
1254 	else if (error2 != 0)
1255 		return (error2);
1256 	if (doingzomb)
1257 		np = zpfind(pid);
1258 	else {
1259 		if (pid == 0)
1260 			return (0);
1261 		np = pfind(pid);
1262 	}
1263 	if (np == NULL)
1264 		return (ESRCH);
1265 	if (np != p) {
1266 		PROC_UNLOCK(np);
1267 		return (ESRCH);
1268 	}
1269 	PROC_UNLOCK(np);
1270 	return (0);
1271 }
1272 
1273 static int
1274 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1275 {
1276 	int *name = (int *)arg1;
1277 	u_int namelen = arg2;
1278 	struct proc *p;
1279 	int flags, doingzomb, oid_number;
1280 	int error = 0;
1281 
1282 	oid_number = oidp->oid_number;
1283 	if (oid_number != KERN_PROC_ALL &&
1284 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1285 		flags = KERN_PROC_NOTHREADS;
1286 	else {
1287 		flags = 0;
1288 		oid_number &= ~KERN_PROC_INC_THREAD;
1289 	}
1290 #ifdef COMPAT_FREEBSD32
1291 	if (req->flags & SCTL_MASK32)
1292 		flags |= KERN_PROC_MASK32;
1293 #endif
1294 	if (oid_number == KERN_PROC_PID) {
1295 		if (namelen != 1)
1296 			return (EINVAL);
1297 		error = sysctl_wire_old_buffer(req, 0);
1298 		if (error)
1299 			return (error);
1300 		sx_slock(&proctree_lock);
1301 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1302 		if (error == 0)
1303 			error = sysctl_out_proc(p, req, flags, 0);
1304 		sx_sunlock(&proctree_lock);
1305 		return (error);
1306 	}
1307 
1308 	switch (oid_number) {
1309 	case KERN_PROC_ALL:
1310 		if (namelen != 0)
1311 			return (EINVAL);
1312 		break;
1313 	case KERN_PROC_PROC:
1314 		if (namelen != 0 && namelen != 1)
1315 			return (EINVAL);
1316 		break;
1317 	default:
1318 		if (namelen != 1)
1319 			return (EINVAL);
1320 		break;
1321 	}
1322 
1323 	if (!req->oldptr) {
1324 		/* overestimate by 5 procs */
1325 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1326 		if (error)
1327 			return (error);
1328 	}
1329 	error = sysctl_wire_old_buffer(req, 0);
1330 	if (error != 0)
1331 		return (error);
1332 	sx_slock(&proctree_lock);
1333 	sx_slock(&allproc_lock);
1334 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1335 		if (!doingzomb)
1336 			p = LIST_FIRST(&allproc);
1337 		else
1338 			p = LIST_FIRST(&zombproc);
1339 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
1340 			/*
1341 			 * Skip embryonic processes.
1342 			 */
1343 			PROC_LOCK(p);
1344 			if (p->p_state == PRS_NEW) {
1345 				PROC_UNLOCK(p);
1346 				continue;
1347 			}
1348 			KASSERT(p->p_ucred != NULL,
1349 			    ("process credential is NULL for non-NEW proc"));
1350 			/*
1351 			 * Show a user only appropriate processes.
1352 			 */
1353 			if (p_cansee(curthread, p)) {
1354 				PROC_UNLOCK(p);
1355 				continue;
1356 			}
1357 			/*
1358 			 * TODO - make more efficient (see notes below).
1359 			 * do by session.
1360 			 */
1361 			switch (oid_number) {
1362 
1363 			case KERN_PROC_GID:
1364 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1365 					PROC_UNLOCK(p);
1366 					continue;
1367 				}
1368 				break;
1369 
1370 			case KERN_PROC_PGRP:
1371 				/* could do this by traversing pgrp */
1372 				if (p->p_pgrp == NULL ||
1373 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1374 					PROC_UNLOCK(p);
1375 					continue;
1376 				}
1377 				break;
1378 
1379 			case KERN_PROC_RGID:
1380 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1381 					PROC_UNLOCK(p);
1382 					continue;
1383 				}
1384 				break;
1385 
1386 			case KERN_PROC_SESSION:
1387 				if (p->p_session == NULL ||
1388 				    p->p_session->s_sid != (pid_t)name[0]) {
1389 					PROC_UNLOCK(p);
1390 					continue;
1391 				}
1392 				break;
1393 
1394 			case KERN_PROC_TTY:
1395 				if ((p->p_flag & P_CONTROLT) == 0 ||
1396 				    p->p_session == NULL) {
1397 					PROC_UNLOCK(p);
1398 					continue;
1399 				}
1400 				/* XXX proctree_lock */
1401 				SESS_LOCK(p->p_session);
1402 				if (p->p_session->s_ttyp == NULL ||
1403 				    tty_udev(p->p_session->s_ttyp) !=
1404 				    (dev_t)name[0]) {
1405 					SESS_UNLOCK(p->p_session);
1406 					PROC_UNLOCK(p);
1407 					continue;
1408 				}
1409 				SESS_UNLOCK(p->p_session);
1410 				break;
1411 
1412 			case KERN_PROC_UID:
1413 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1414 					PROC_UNLOCK(p);
1415 					continue;
1416 				}
1417 				break;
1418 
1419 			case KERN_PROC_RUID:
1420 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1421 					PROC_UNLOCK(p);
1422 					continue;
1423 				}
1424 				break;
1425 
1426 			case KERN_PROC_PROC:
1427 				break;
1428 
1429 			default:
1430 				break;
1431 
1432 			}
1433 
1434 			error = sysctl_out_proc(p, req, flags, doingzomb);
1435 			if (error) {
1436 				sx_sunlock(&allproc_lock);
1437 				sx_sunlock(&proctree_lock);
1438 				return (error);
1439 			}
1440 		}
1441 	}
1442 	sx_sunlock(&allproc_lock);
1443 	sx_sunlock(&proctree_lock);
1444 	return (0);
1445 }
1446 
1447 struct pargs *
1448 pargs_alloc(int len)
1449 {
1450 	struct pargs *pa;
1451 
1452 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1453 		M_WAITOK);
1454 	refcount_init(&pa->ar_ref, 1);
1455 	pa->ar_length = len;
1456 	return (pa);
1457 }
1458 
1459 static void
1460 pargs_free(struct pargs *pa)
1461 {
1462 
1463 	free(pa, M_PARGS);
1464 }
1465 
1466 void
1467 pargs_hold(struct pargs *pa)
1468 {
1469 
1470 	if (pa == NULL)
1471 		return;
1472 	refcount_acquire(&pa->ar_ref);
1473 }
1474 
1475 void
1476 pargs_drop(struct pargs *pa)
1477 {
1478 
1479 	if (pa == NULL)
1480 		return;
1481 	if (refcount_release(&pa->ar_ref))
1482 		pargs_free(pa);
1483 }
1484 
1485 static int
1486 proc_read_mem(struct thread *td, struct proc *p, vm_offset_t offset, void* buf,
1487     size_t len)
1488 {
1489 	struct iovec iov;
1490 	struct uio uio;
1491 
1492 	iov.iov_base = (caddr_t)buf;
1493 	iov.iov_len = len;
1494 	uio.uio_iov = &iov;
1495 	uio.uio_iovcnt = 1;
1496 	uio.uio_offset = offset;
1497 	uio.uio_resid = (ssize_t)len;
1498 	uio.uio_segflg = UIO_SYSSPACE;
1499 	uio.uio_rw = UIO_READ;
1500 	uio.uio_td = td;
1501 
1502 	return (proc_rwmem(p, &uio));
1503 }
1504 
1505 static int
1506 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1507     size_t len)
1508 {
1509 	size_t i;
1510 	int error;
1511 
1512 	error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, len);
1513 	/*
1514 	 * Reading the chunk may validly return EFAULT if the string is shorter
1515 	 * than the chunk and is aligned at the end of the page, assuming the
1516 	 * next page is not mapped.  So if EFAULT is returned do a fallback to
1517 	 * one byte read loop.
1518 	 */
1519 	if (error == EFAULT) {
1520 		for (i = 0; i < len; i++, buf++, sptr++) {
1521 			error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, 1);
1522 			if (error != 0)
1523 				return (error);
1524 			if (*buf == '\0')
1525 				break;
1526 		}
1527 		error = 0;
1528 	}
1529 	return (error);
1530 }
1531 
1532 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1533 
1534 enum proc_vector_type {
1535 	PROC_ARG,
1536 	PROC_ENV,
1537 	PROC_AUX,
1538 };
1539 
1540 #ifdef COMPAT_FREEBSD32
1541 static int
1542 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1543     size_t *vsizep, enum proc_vector_type type)
1544 {
1545 	struct freebsd32_ps_strings pss;
1546 	Elf32_Auxinfo aux;
1547 	vm_offset_t vptr, ptr;
1548 	uint32_t *proc_vector32;
1549 	char **proc_vector;
1550 	size_t vsize, size;
1551 	int i, error;
1552 
1553 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1554 	    &pss, sizeof(pss));
1555 	if (error != 0)
1556 		return (error);
1557 	switch (type) {
1558 	case PROC_ARG:
1559 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1560 		vsize = pss.ps_nargvstr;
1561 		if (vsize > ARG_MAX)
1562 			return (ENOEXEC);
1563 		size = vsize * sizeof(int32_t);
1564 		break;
1565 	case PROC_ENV:
1566 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1567 		vsize = pss.ps_nenvstr;
1568 		if (vsize > ARG_MAX)
1569 			return (ENOEXEC);
1570 		size = vsize * sizeof(int32_t);
1571 		break;
1572 	case PROC_AUX:
1573 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1574 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1575 		if (vptr % 4 != 0)
1576 			return (ENOEXEC);
1577 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1578 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1579 			if (error != 0)
1580 				return (error);
1581 			if (aux.a_type == AT_NULL)
1582 				break;
1583 			ptr += sizeof(aux);
1584 		}
1585 		if (aux.a_type != AT_NULL)
1586 			return (ENOEXEC);
1587 		vsize = i + 1;
1588 		size = vsize * sizeof(aux);
1589 		break;
1590 	default:
1591 		KASSERT(0, ("Wrong proc vector type: %d", type));
1592 		return (EINVAL);
1593 	}
1594 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1595 	error = proc_read_mem(td, p, vptr, proc_vector32, size);
1596 	if (error != 0)
1597 		goto done;
1598 	if (type == PROC_AUX) {
1599 		*proc_vectorp = (char **)proc_vector32;
1600 		*vsizep = vsize;
1601 		return (0);
1602 	}
1603 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1604 	for (i = 0; i < (int)vsize; i++)
1605 		proc_vector[i] = PTRIN(proc_vector32[i]);
1606 	*proc_vectorp = proc_vector;
1607 	*vsizep = vsize;
1608 done:
1609 	free(proc_vector32, M_TEMP);
1610 	return (error);
1611 }
1612 #endif
1613 
1614 static int
1615 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1616     size_t *vsizep, enum proc_vector_type type)
1617 {
1618 	struct ps_strings pss;
1619 	Elf_Auxinfo aux;
1620 	vm_offset_t vptr, ptr;
1621 	char **proc_vector;
1622 	size_t vsize, size;
1623 	int error, i;
1624 
1625 #ifdef COMPAT_FREEBSD32
1626 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1627 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1628 #endif
1629 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1630 	    &pss, sizeof(pss));
1631 	if (error != 0)
1632 		return (error);
1633 	switch (type) {
1634 	case PROC_ARG:
1635 		vptr = (vm_offset_t)pss.ps_argvstr;
1636 		vsize = pss.ps_nargvstr;
1637 		if (vsize > ARG_MAX)
1638 			return (ENOEXEC);
1639 		size = vsize * sizeof(char *);
1640 		break;
1641 	case PROC_ENV:
1642 		vptr = (vm_offset_t)pss.ps_envstr;
1643 		vsize = pss.ps_nenvstr;
1644 		if (vsize > ARG_MAX)
1645 			return (ENOEXEC);
1646 		size = vsize * sizeof(char *);
1647 		break;
1648 	case PROC_AUX:
1649 		/*
1650 		 * The aux array is just above env array on the stack. Check
1651 		 * that the address is naturally aligned.
1652 		 */
1653 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1654 		    * sizeof(char *);
1655 #if __ELF_WORD_SIZE == 64
1656 		if (vptr % sizeof(uint64_t) != 0)
1657 #else
1658 		if (vptr % sizeof(uint32_t) != 0)
1659 #endif
1660 			return (ENOEXEC);
1661 		/*
1662 		 * We count the array size reading the aux vectors from the
1663 		 * stack until AT_NULL vector is returned.  So (to keep the code
1664 		 * simple) we read the process stack twice: the first time here
1665 		 * to find the size and the second time when copying the vectors
1666 		 * to the allocated proc_vector.
1667 		 */
1668 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1669 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1670 			if (error != 0)
1671 				return (error);
1672 			if (aux.a_type == AT_NULL)
1673 				break;
1674 			ptr += sizeof(aux);
1675 		}
1676 		/*
1677 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1678 		 * not reached AT_NULL, it is most likely we are reading wrong
1679 		 * data: either the process doesn't have auxv array or data has
1680 		 * been modified. Return the error in this case.
1681 		 */
1682 		if (aux.a_type != AT_NULL)
1683 			return (ENOEXEC);
1684 		vsize = i + 1;
1685 		size = vsize * sizeof(aux);
1686 		break;
1687 	default:
1688 		KASSERT(0, ("Wrong proc vector type: %d", type));
1689 		return (EINVAL); /* In case we are built without INVARIANTS. */
1690 	}
1691 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1692 	if (proc_vector == NULL)
1693 		return (ENOMEM);
1694 	error = proc_read_mem(td, p, vptr, proc_vector, size);
1695 	if (error != 0) {
1696 		free(proc_vector, M_TEMP);
1697 		return (error);
1698 	}
1699 	*proc_vectorp = proc_vector;
1700 	*vsizep = vsize;
1701 
1702 	return (0);
1703 }
1704 
1705 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1706 
1707 static int
1708 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1709     enum proc_vector_type type)
1710 {
1711 	size_t done, len, nchr, vsize;
1712 	int error, i;
1713 	char **proc_vector, *sptr;
1714 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1715 
1716 	PROC_ASSERT_HELD(p);
1717 
1718 	/*
1719 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1720 	 */
1721 	nchr = 2 * (PATH_MAX + ARG_MAX);
1722 
1723 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1724 	if (error != 0)
1725 		return (error);
1726 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1727 		/*
1728 		 * The program may have scribbled into its argv array, e.g. to
1729 		 * remove some arguments.  If that has happened, break out
1730 		 * before trying to read from NULL.
1731 		 */
1732 		if (proc_vector[i] == NULL)
1733 			break;
1734 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1735 			error = proc_read_string(td, p, sptr, pss_string,
1736 			    sizeof(pss_string));
1737 			if (error != 0)
1738 				goto done;
1739 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1740 			if (done + len >= nchr)
1741 				len = nchr - done - 1;
1742 			sbuf_bcat(sb, pss_string, len);
1743 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1744 				break;
1745 			done += GET_PS_STRINGS_CHUNK_SZ;
1746 		}
1747 		sbuf_bcat(sb, "", 1);
1748 		done += len + 1;
1749 	}
1750 done:
1751 	free(proc_vector, M_TEMP);
1752 	return (error);
1753 }
1754 
1755 int
1756 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1757 {
1758 
1759 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1760 }
1761 
1762 int
1763 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1764 {
1765 
1766 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1767 }
1768 
1769 int
1770 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1771 {
1772 	size_t vsize, size;
1773 	char **auxv;
1774 	int error;
1775 
1776 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1777 	if (error == 0) {
1778 #ifdef COMPAT_FREEBSD32
1779 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1780 			size = vsize * sizeof(Elf32_Auxinfo);
1781 		else
1782 #endif
1783 			size = vsize * sizeof(Elf_Auxinfo);
1784 		if (sbuf_bcat(sb, auxv, size) != 0)
1785 			error = ENOMEM;
1786 		free(auxv, M_TEMP);
1787 	}
1788 	return (error);
1789 }
1790 
1791 /*
1792  * This sysctl allows a process to retrieve the argument list or process
1793  * title for another process without groping around in the address space
1794  * of the other process.  It also allow a process to set its own "process
1795  * title to a string of its own choice.
1796  */
1797 static int
1798 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1799 {
1800 	int *name = (int *)arg1;
1801 	u_int namelen = arg2;
1802 	struct pargs *newpa, *pa;
1803 	struct proc *p;
1804 	struct sbuf sb;
1805 	int flags, error = 0, error2;
1806 
1807 	if (namelen != 1)
1808 		return (EINVAL);
1809 
1810 	flags = PGET_CANSEE;
1811 	if (req->newptr != NULL)
1812 		flags |= PGET_ISCURRENT;
1813 	error = pget((pid_t)name[0], flags, &p);
1814 	if (error)
1815 		return (error);
1816 
1817 	pa = p->p_args;
1818 	if (pa != NULL) {
1819 		pargs_hold(pa);
1820 		PROC_UNLOCK(p);
1821 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1822 		pargs_drop(pa);
1823 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1824 		_PHOLD(p);
1825 		PROC_UNLOCK(p);
1826 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1827 		error = proc_getargv(curthread, p, &sb);
1828 		error2 = sbuf_finish(&sb);
1829 		PRELE(p);
1830 		sbuf_delete(&sb);
1831 		if (error == 0 && error2 != 0)
1832 			error = error2;
1833 	} else {
1834 		PROC_UNLOCK(p);
1835 	}
1836 	if (error != 0 || req->newptr == NULL)
1837 		return (error);
1838 
1839 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1840 		return (ENOMEM);
1841 	newpa = pargs_alloc(req->newlen);
1842 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1843 	if (error != 0) {
1844 		pargs_free(newpa);
1845 		return (error);
1846 	}
1847 	PROC_LOCK(p);
1848 	pa = p->p_args;
1849 	p->p_args = newpa;
1850 	PROC_UNLOCK(p);
1851 	pargs_drop(pa);
1852 	return (0);
1853 }
1854 
1855 /*
1856  * This sysctl allows a process to retrieve environment of another process.
1857  */
1858 static int
1859 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
1860 {
1861 	int *name = (int *)arg1;
1862 	u_int namelen = arg2;
1863 	struct proc *p;
1864 	struct sbuf sb;
1865 	int error, error2;
1866 
1867 	if (namelen != 1)
1868 		return (EINVAL);
1869 
1870 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1871 	if (error != 0)
1872 		return (error);
1873 	if ((p->p_flag & P_SYSTEM) != 0) {
1874 		PRELE(p);
1875 		return (0);
1876 	}
1877 
1878 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1879 	error = proc_getenvv(curthread, p, &sb);
1880 	error2 = sbuf_finish(&sb);
1881 	PRELE(p);
1882 	sbuf_delete(&sb);
1883 	return (error != 0 ? error : error2);
1884 }
1885 
1886 /*
1887  * This sysctl allows a process to retrieve ELF auxiliary vector of
1888  * another process.
1889  */
1890 static int
1891 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
1892 {
1893 	int *name = (int *)arg1;
1894 	u_int namelen = arg2;
1895 	struct proc *p;
1896 	struct sbuf sb;
1897 	int error, error2;
1898 
1899 	if (namelen != 1)
1900 		return (EINVAL);
1901 
1902 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1903 	if (error != 0)
1904 		return (error);
1905 	if ((p->p_flag & P_SYSTEM) != 0) {
1906 		PRELE(p);
1907 		return (0);
1908 	}
1909 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1910 	error = proc_getauxv(curthread, p, &sb);
1911 	error2 = sbuf_finish(&sb);
1912 	PRELE(p);
1913 	sbuf_delete(&sb);
1914 	return (error != 0 ? error : error2);
1915 }
1916 
1917 /*
1918  * This sysctl allows a process to retrieve the path of the executable for
1919  * itself or another process.
1920  */
1921 static int
1922 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1923 {
1924 	pid_t *pidp = (pid_t *)arg1;
1925 	unsigned int arglen = arg2;
1926 	struct proc *p;
1927 	struct vnode *vp;
1928 	char *retbuf, *freebuf;
1929 	int error;
1930 
1931 	if (arglen != 1)
1932 		return (EINVAL);
1933 	if (*pidp == -1) {	/* -1 means this process */
1934 		p = req->td->td_proc;
1935 	} else {
1936 		error = pget(*pidp, PGET_CANSEE, &p);
1937 		if (error != 0)
1938 			return (error);
1939 	}
1940 
1941 	vp = p->p_textvp;
1942 	if (vp == NULL) {
1943 		if (*pidp != -1)
1944 			PROC_UNLOCK(p);
1945 		return (0);
1946 	}
1947 	vref(vp);
1948 	if (*pidp != -1)
1949 		PROC_UNLOCK(p);
1950 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1951 	vrele(vp);
1952 	if (error)
1953 		return (error);
1954 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1955 	free(freebuf, M_TEMP);
1956 	return (error);
1957 }
1958 
1959 static int
1960 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1961 {
1962 	struct proc *p;
1963 	char *sv_name;
1964 	int *name;
1965 	int namelen;
1966 	int error;
1967 
1968 	namelen = arg2;
1969 	if (namelen != 1)
1970 		return (EINVAL);
1971 
1972 	name = (int *)arg1;
1973 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
1974 	if (error != 0)
1975 		return (error);
1976 	sv_name = p->p_sysent->sv_name;
1977 	PROC_UNLOCK(p);
1978 	return (sysctl_handle_string(oidp, sv_name, 0, req));
1979 }
1980 
1981 #ifdef KINFO_OVMENTRY_SIZE
1982 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
1983 #endif
1984 
1985 #ifdef COMPAT_FREEBSD7
1986 static int
1987 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
1988 {
1989 	vm_map_entry_t entry, tmp_entry;
1990 	unsigned int last_timestamp;
1991 	char *fullpath, *freepath;
1992 	struct kinfo_ovmentry *kve;
1993 	struct vattr va;
1994 	struct ucred *cred;
1995 	int error, *name;
1996 	struct vnode *vp;
1997 	struct proc *p;
1998 	vm_map_t map;
1999 	struct vmspace *vm;
2000 
2001 	name = (int *)arg1;
2002 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2003 	if (error != 0)
2004 		return (error);
2005 	vm = vmspace_acquire_ref(p);
2006 	if (vm == NULL) {
2007 		PRELE(p);
2008 		return (ESRCH);
2009 	}
2010 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2011 
2012 	map = &vm->vm_map;
2013 	vm_map_lock_read(map);
2014 	for (entry = map->header.next; entry != &map->header;
2015 	    entry = entry->next) {
2016 		vm_object_t obj, tobj, lobj;
2017 		vm_offset_t addr;
2018 
2019 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2020 			continue;
2021 
2022 		bzero(kve, sizeof(*kve));
2023 		kve->kve_structsize = sizeof(*kve);
2024 
2025 		kve->kve_private_resident = 0;
2026 		obj = entry->object.vm_object;
2027 		if (obj != NULL) {
2028 			VM_OBJECT_RLOCK(obj);
2029 			if (obj->shadow_count == 1)
2030 				kve->kve_private_resident =
2031 				    obj->resident_page_count;
2032 		}
2033 		kve->kve_resident = 0;
2034 		addr = entry->start;
2035 		while (addr < entry->end) {
2036 			if (pmap_extract(map->pmap, addr))
2037 				kve->kve_resident++;
2038 			addr += PAGE_SIZE;
2039 		}
2040 
2041 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2042 			if (tobj != obj)
2043 				VM_OBJECT_RLOCK(tobj);
2044 			if (lobj != obj)
2045 				VM_OBJECT_RUNLOCK(lobj);
2046 			lobj = tobj;
2047 		}
2048 
2049 		kve->kve_start = (void*)entry->start;
2050 		kve->kve_end = (void*)entry->end;
2051 		kve->kve_offset = (off_t)entry->offset;
2052 
2053 		if (entry->protection & VM_PROT_READ)
2054 			kve->kve_protection |= KVME_PROT_READ;
2055 		if (entry->protection & VM_PROT_WRITE)
2056 			kve->kve_protection |= KVME_PROT_WRITE;
2057 		if (entry->protection & VM_PROT_EXECUTE)
2058 			kve->kve_protection |= KVME_PROT_EXEC;
2059 
2060 		if (entry->eflags & MAP_ENTRY_COW)
2061 			kve->kve_flags |= KVME_FLAG_COW;
2062 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2063 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2064 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2065 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2066 
2067 		last_timestamp = map->timestamp;
2068 		vm_map_unlock_read(map);
2069 
2070 		kve->kve_fileid = 0;
2071 		kve->kve_fsid = 0;
2072 		freepath = NULL;
2073 		fullpath = "";
2074 		if (lobj) {
2075 			vp = NULL;
2076 			switch (lobj->type) {
2077 			case OBJT_DEFAULT:
2078 				kve->kve_type = KVME_TYPE_DEFAULT;
2079 				break;
2080 			case OBJT_VNODE:
2081 				kve->kve_type = KVME_TYPE_VNODE;
2082 				vp = lobj->handle;
2083 				vref(vp);
2084 				break;
2085 			case OBJT_SWAP:
2086 				kve->kve_type = KVME_TYPE_SWAP;
2087 				break;
2088 			case OBJT_DEVICE:
2089 				kve->kve_type = KVME_TYPE_DEVICE;
2090 				break;
2091 			case OBJT_PHYS:
2092 				kve->kve_type = KVME_TYPE_PHYS;
2093 				break;
2094 			case OBJT_DEAD:
2095 				kve->kve_type = KVME_TYPE_DEAD;
2096 				break;
2097 			case OBJT_SG:
2098 				kve->kve_type = KVME_TYPE_SG;
2099 				break;
2100 			default:
2101 				kve->kve_type = KVME_TYPE_UNKNOWN;
2102 				break;
2103 			}
2104 			if (lobj != obj)
2105 				VM_OBJECT_RUNLOCK(lobj);
2106 
2107 			kve->kve_ref_count = obj->ref_count;
2108 			kve->kve_shadow_count = obj->shadow_count;
2109 			VM_OBJECT_RUNLOCK(obj);
2110 			if (vp != NULL) {
2111 				vn_fullpath(curthread, vp, &fullpath,
2112 				    &freepath);
2113 				cred = curthread->td_ucred;
2114 				vn_lock(vp, LK_SHARED | LK_RETRY);
2115 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2116 					kve->kve_fileid = va.va_fileid;
2117 					kve->kve_fsid = va.va_fsid;
2118 				}
2119 				vput(vp);
2120 			}
2121 		} else {
2122 			kve->kve_type = KVME_TYPE_NONE;
2123 			kve->kve_ref_count = 0;
2124 			kve->kve_shadow_count = 0;
2125 		}
2126 
2127 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2128 		if (freepath != NULL)
2129 			free(freepath, M_TEMP);
2130 
2131 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2132 		vm_map_lock_read(map);
2133 		if (error)
2134 			break;
2135 		if (last_timestamp != map->timestamp) {
2136 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2137 			entry = tmp_entry;
2138 		}
2139 	}
2140 	vm_map_unlock_read(map);
2141 	vmspace_free(vm);
2142 	PRELE(p);
2143 	free(kve, M_TEMP);
2144 	return (error);
2145 }
2146 #endif	/* COMPAT_FREEBSD7 */
2147 
2148 #ifdef KINFO_VMENTRY_SIZE
2149 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2150 #endif
2151 
2152 static void
2153 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2154     struct kinfo_vmentry *kve)
2155 {
2156 	vm_object_t obj, tobj;
2157 	vm_page_t m, m_adv;
2158 	vm_offset_t addr;
2159 	vm_paddr_t locked_pa;
2160 	vm_pindex_t pi, pi_adv, pindex;
2161 
2162 	locked_pa = 0;
2163 	obj = entry->object.vm_object;
2164 	addr = entry->start;
2165 	m_adv = NULL;
2166 	pi = OFF_TO_IDX(entry->offset);
2167 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2168 		if (m_adv != NULL) {
2169 			m = m_adv;
2170 		} else {
2171 			pi_adv = OFF_TO_IDX(entry->end - addr);
2172 			pindex = pi;
2173 			for (tobj = obj;; tobj = tobj->backing_object) {
2174 				m = vm_page_find_least(tobj, pindex);
2175 				if (m != NULL) {
2176 					if (m->pindex == pindex)
2177 						break;
2178 					if (pi_adv > m->pindex - pindex) {
2179 						pi_adv = m->pindex - pindex;
2180 						m_adv = m;
2181 					}
2182 				}
2183 				if (tobj->backing_object == NULL)
2184 					goto next;
2185 				pindex += OFF_TO_IDX(tobj->
2186 				    backing_object_offset);
2187 			}
2188 		}
2189 		m_adv = NULL;
2190 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2191 		    (addr & (pagesizes[1] - 1)) == 0 &&
2192 		    (pmap_mincore(map->pmap, addr, &locked_pa) &
2193 		    MINCORE_SUPER) != 0) {
2194 			kve->kve_flags |= KVME_FLAG_SUPER;
2195 			pi_adv = OFF_TO_IDX(pagesizes[1]);
2196 		} else {
2197 			/*
2198 			 * We do not test the found page on validity.
2199 			 * Either the page is busy and being paged in,
2200 			 * or it was invalidated.  The first case
2201 			 * should be counted as resident, the second
2202 			 * is not so clear; we do account both.
2203 			 */
2204 			pi_adv = 1;
2205 		}
2206 		kve->kve_resident += pi_adv;
2207 next:;
2208 	}
2209 	PA_UNLOCK_COND(locked_pa);
2210 }
2211 
2212 /*
2213  * Must be called with the process locked and will return unlocked.
2214  */
2215 int
2216 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb)
2217 {
2218 	vm_map_entry_t entry, tmp_entry;
2219 	struct vattr va;
2220 	vm_map_t map;
2221 	vm_object_t obj, tobj, lobj;
2222 	char *fullpath, *freepath;
2223 	struct kinfo_vmentry *kve;
2224 	struct ucred *cred;
2225 	struct vnode *vp;
2226 	struct vmspace *vm;
2227 	vm_offset_t addr;
2228 	unsigned int last_timestamp;
2229 	int error;
2230 
2231 	PROC_LOCK_ASSERT(p, MA_OWNED);
2232 
2233 	_PHOLD(p);
2234 	PROC_UNLOCK(p);
2235 	vm = vmspace_acquire_ref(p);
2236 	if (vm == NULL) {
2237 		PRELE(p);
2238 		return (ESRCH);
2239 	}
2240 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2241 
2242 	error = 0;
2243 	map = &vm->vm_map;
2244 	vm_map_lock_read(map);
2245 	for (entry = map->header.next; entry != &map->header;
2246 	    entry = entry->next) {
2247 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2248 			continue;
2249 
2250 		addr = entry->end;
2251 		bzero(kve, sizeof(*kve));
2252 		obj = entry->object.vm_object;
2253 		if (obj != NULL) {
2254 			for (tobj = obj; tobj != NULL;
2255 			    tobj = tobj->backing_object) {
2256 				VM_OBJECT_RLOCK(tobj);
2257 				lobj = tobj;
2258 			}
2259 			if (obj->backing_object == NULL)
2260 				kve->kve_private_resident =
2261 				    obj->resident_page_count;
2262 			if (!vmmap_skip_res_cnt)
2263 				kern_proc_vmmap_resident(map, entry, kve);
2264 			for (tobj = obj; tobj != NULL;
2265 			    tobj = tobj->backing_object) {
2266 				if (tobj != obj && tobj != lobj)
2267 					VM_OBJECT_RUNLOCK(tobj);
2268 			}
2269 		} else {
2270 			lobj = NULL;
2271 		}
2272 
2273 		kve->kve_start = entry->start;
2274 		kve->kve_end = entry->end;
2275 		kve->kve_offset = entry->offset;
2276 
2277 		if (entry->protection & VM_PROT_READ)
2278 			kve->kve_protection |= KVME_PROT_READ;
2279 		if (entry->protection & VM_PROT_WRITE)
2280 			kve->kve_protection |= KVME_PROT_WRITE;
2281 		if (entry->protection & VM_PROT_EXECUTE)
2282 			kve->kve_protection |= KVME_PROT_EXEC;
2283 
2284 		if (entry->eflags & MAP_ENTRY_COW)
2285 			kve->kve_flags |= KVME_FLAG_COW;
2286 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2287 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2288 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2289 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2290 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2291 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2292 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2293 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2294 
2295 		last_timestamp = map->timestamp;
2296 		vm_map_unlock_read(map);
2297 
2298 		freepath = NULL;
2299 		fullpath = "";
2300 		if (lobj != NULL) {
2301 			vp = NULL;
2302 			switch (lobj->type) {
2303 			case OBJT_DEFAULT:
2304 				kve->kve_type = KVME_TYPE_DEFAULT;
2305 				break;
2306 			case OBJT_VNODE:
2307 				kve->kve_type = KVME_TYPE_VNODE;
2308 				vp = lobj->handle;
2309 				vref(vp);
2310 				break;
2311 			case OBJT_SWAP:
2312 				kve->kve_type = KVME_TYPE_SWAP;
2313 				break;
2314 			case OBJT_DEVICE:
2315 				kve->kve_type = KVME_TYPE_DEVICE;
2316 				break;
2317 			case OBJT_PHYS:
2318 				kve->kve_type = KVME_TYPE_PHYS;
2319 				break;
2320 			case OBJT_DEAD:
2321 				kve->kve_type = KVME_TYPE_DEAD;
2322 				break;
2323 			case OBJT_SG:
2324 				kve->kve_type = KVME_TYPE_SG;
2325 				break;
2326 			case OBJT_MGTDEVICE:
2327 				kve->kve_type = KVME_TYPE_MGTDEVICE;
2328 				break;
2329 			default:
2330 				kve->kve_type = KVME_TYPE_UNKNOWN;
2331 				break;
2332 			}
2333 			if (lobj != obj)
2334 				VM_OBJECT_RUNLOCK(lobj);
2335 
2336 			kve->kve_ref_count = obj->ref_count;
2337 			kve->kve_shadow_count = obj->shadow_count;
2338 			VM_OBJECT_RUNLOCK(obj);
2339 			if (vp != NULL) {
2340 				vn_fullpath(curthread, vp, &fullpath,
2341 				    &freepath);
2342 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2343 				cred = curthread->td_ucred;
2344 				vn_lock(vp, LK_SHARED | LK_RETRY);
2345 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2346 					kve->kve_vn_fileid = va.va_fileid;
2347 					kve->kve_vn_fsid = va.va_fsid;
2348 					kve->kve_vn_mode =
2349 					    MAKEIMODE(va.va_type, va.va_mode);
2350 					kve->kve_vn_size = va.va_size;
2351 					kve->kve_vn_rdev = va.va_rdev;
2352 					kve->kve_status = KF_ATTR_VALID;
2353 				}
2354 				vput(vp);
2355 			}
2356 		} else {
2357 			kve->kve_type = KVME_TYPE_NONE;
2358 			kve->kve_ref_count = 0;
2359 			kve->kve_shadow_count = 0;
2360 		}
2361 
2362 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2363 		if (freepath != NULL)
2364 			free(freepath, M_TEMP);
2365 
2366 		/* Pack record size down */
2367 		kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) +
2368 		    strlen(kve->kve_path) + 1;
2369 		kve->kve_structsize = roundup(kve->kve_structsize,
2370 		    sizeof(uint64_t));
2371 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2372 			error = ENOMEM;
2373 		vm_map_lock_read(map);
2374 		if (error != 0)
2375 			break;
2376 		if (last_timestamp != map->timestamp) {
2377 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2378 			entry = tmp_entry;
2379 		}
2380 	}
2381 	vm_map_unlock_read(map);
2382 	vmspace_free(vm);
2383 	PRELE(p);
2384 	free(kve, M_TEMP);
2385 	return (error);
2386 }
2387 
2388 static int
2389 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2390 {
2391 	struct proc *p;
2392 	struct sbuf sb;
2393 	int error, error2, *name;
2394 
2395 	name = (int *)arg1;
2396 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2397 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2398 	if (error != 0) {
2399 		sbuf_delete(&sb);
2400 		return (error);
2401 	}
2402 	error = kern_proc_vmmap_out(p, &sb);
2403 	error2 = sbuf_finish(&sb);
2404 	sbuf_delete(&sb);
2405 	return (error != 0 ? error : error2);
2406 }
2407 
2408 #if defined(STACK) || defined(DDB)
2409 static int
2410 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2411 {
2412 	struct kinfo_kstack *kkstp;
2413 	int error, i, *name, numthreads;
2414 	lwpid_t *lwpidarray;
2415 	struct thread *td;
2416 	struct stack *st;
2417 	struct sbuf sb;
2418 	struct proc *p;
2419 
2420 	name = (int *)arg1;
2421 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2422 	if (error != 0)
2423 		return (error);
2424 
2425 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2426 	st = stack_create();
2427 
2428 	lwpidarray = NULL;
2429 	numthreads = 0;
2430 	PROC_LOCK(p);
2431 repeat:
2432 	if (numthreads < p->p_numthreads) {
2433 		if (lwpidarray != NULL) {
2434 			free(lwpidarray, M_TEMP);
2435 			lwpidarray = NULL;
2436 		}
2437 		numthreads = p->p_numthreads;
2438 		PROC_UNLOCK(p);
2439 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2440 		    M_WAITOK | M_ZERO);
2441 		PROC_LOCK(p);
2442 		goto repeat;
2443 	}
2444 	i = 0;
2445 
2446 	/*
2447 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2448 	 * of changes could have taken place.  Should we check to see if the
2449 	 * vmspace has been replaced, or the like, in order to prevent
2450 	 * giving a snapshot that spans, say, execve(2), with some threads
2451 	 * before and some after?  Among other things, the credentials could
2452 	 * have changed, in which case the right to extract debug info might
2453 	 * no longer be assured.
2454 	 */
2455 	FOREACH_THREAD_IN_PROC(p, td) {
2456 		KASSERT(i < numthreads,
2457 		    ("sysctl_kern_proc_kstack: numthreads"));
2458 		lwpidarray[i] = td->td_tid;
2459 		i++;
2460 	}
2461 	numthreads = i;
2462 	for (i = 0; i < numthreads; i++) {
2463 		td = thread_find(p, lwpidarray[i]);
2464 		if (td == NULL) {
2465 			continue;
2466 		}
2467 		bzero(kkstp, sizeof(*kkstp));
2468 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2469 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2470 		thread_lock(td);
2471 		kkstp->kkst_tid = td->td_tid;
2472 		if (TD_IS_SWAPPED(td))
2473 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2474 		else if (TD_IS_RUNNING(td))
2475 			kkstp->kkst_state = KKST_STATE_RUNNING;
2476 		else {
2477 			kkstp->kkst_state = KKST_STATE_STACKOK;
2478 			stack_save_td(st, td);
2479 		}
2480 		thread_unlock(td);
2481 		PROC_UNLOCK(p);
2482 		stack_sbuf_print(&sb, st);
2483 		sbuf_finish(&sb);
2484 		sbuf_delete(&sb);
2485 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2486 		PROC_LOCK(p);
2487 		if (error)
2488 			break;
2489 	}
2490 	_PRELE(p);
2491 	PROC_UNLOCK(p);
2492 	if (lwpidarray != NULL)
2493 		free(lwpidarray, M_TEMP);
2494 	stack_destroy(st);
2495 	free(kkstp, M_TEMP);
2496 	return (error);
2497 }
2498 #endif
2499 
2500 /*
2501  * This sysctl allows a process to retrieve the full list of groups from
2502  * itself or another process.
2503  */
2504 static int
2505 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2506 {
2507 	pid_t *pidp = (pid_t *)arg1;
2508 	unsigned int arglen = arg2;
2509 	struct proc *p;
2510 	struct ucred *cred;
2511 	int error;
2512 
2513 	if (arglen != 1)
2514 		return (EINVAL);
2515 	if (*pidp == -1) {	/* -1 means this process */
2516 		p = req->td->td_proc;
2517 		PROC_LOCK(p);
2518 	} else {
2519 		error = pget(*pidp, PGET_CANSEE, &p);
2520 		if (error != 0)
2521 			return (error);
2522 	}
2523 
2524 	cred = crhold(p->p_ucred);
2525 	PROC_UNLOCK(p);
2526 
2527 	error = SYSCTL_OUT(req, cred->cr_groups,
2528 	    cred->cr_ngroups * sizeof(gid_t));
2529 	crfree(cred);
2530 	return (error);
2531 }
2532 
2533 /*
2534  * This sysctl allows a process to retrieve or/and set the resource limit for
2535  * another process.
2536  */
2537 static int
2538 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2539 {
2540 	int *name = (int *)arg1;
2541 	u_int namelen = arg2;
2542 	struct rlimit rlim;
2543 	struct proc *p;
2544 	u_int which;
2545 	int flags, error;
2546 
2547 	if (namelen != 2)
2548 		return (EINVAL);
2549 
2550 	which = (u_int)name[1];
2551 	if (which >= RLIM_NLIMITS)
2552 		return (EINVAL);
2553 
2554 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2555 		return (EINVAL);
2556 
2557 	flags = PGET_HOLD | PGET_NOTWEXIT;
2558 	if (req->newptr != NULL)
2559 		flags |= PGET_CANDEBUG;
2560 	else
2561 		flags |= PGET_CANSEE;
2562 	error = pget((pid_t)name[0], flags, &p);
2563 	if (error != 0)
2564 		return (error);
2565 
2566 	/*
2567 	 * Retrieve limit.
2568 	 */
2569 	if (req->oldptr != NULL) {
2570 		PROC_LOCK(p);
2571 		lim_rlimit(p, which, &rlim);
2572 		PROC_UNLOCK(p);
2573 	}
2574 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2575 	if (error != 0)
2576 		goto errout;
2577 
2578 	/*
2579 	 * Set limit.
2580 	 */
2581 	if (req->newptr != NULL) {
2582 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2583 		if (error == 0)
2584 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2585 	}
2586 
2587 errout:
2588 	PRELE(p);
2589 	return (error);
2590 }
2591 
2592 /*
2593  * This sysctl allows a process to retrieve ps_strings structure location of
2594  * another process.
2595  */
2596 static int
2597 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2598 {
2599 	int *name = (int *)arg1;
2600 	u_int namelen = arg2;
2601 	struct proc *p;
2602 	vm_offset_t ps_strings;
2603 	int error;
2604 #ifdef COMPAT_FREEBSD32
2605 	uint32_t ps_strings32;
2606 #endif
2607 
2608 	if (namelen != 1)
2609 		return (EINVAL);
2610 
2611 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2612 	if (error != 0)
2613 		return (error);
2614 #ifdef COMPAT_FREEBSD32
2615 	if ((req->flags & SCTL_MASK32) != 0) {
2616 		/*
2617 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2618 		 * process.
2619 		 */
2620 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2621 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2622 		PROC_UNLOCK(p);
2623 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2624 		return (error);
2625 	}
2626 #endif
2627 	ps_strings = p->p_sysent->sv_psstrings;
2628 	PROC_UNLOCK(p);
2629 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2630 	return (error);
2631 }
2632 
2633 /*
2634  * This sysctl allows a process to retrieve umask of another process.
2635  */
2636 static int
2637 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2638 {
2639 	int *name = (int *)arg1;
2640 	u_int namelen = arg2;
2641 	struct proc *p;
2642 	int error;
2643 	u_short fd_cmask;
2644 
2645 	if (namelen != 1)
2646 		return (EINVAL);
2647 
2648 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2649 	if (error != 0)
2650 		return (error);
2651 
2652 	FILEDESC_SLOCK(p->p_fd);
2653 	fd_cmask = p->p_fd->fd_cmask;
2654 	FILEDESC_SUNLOCK(p->p_fd);
2655 	PRELE(p);
2656 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2657 	return (error);
2658 }
2659 
2660 /*
2661  * This sysctl allows a process to set and retrieve binary osreldate of
2662  * another process.
2663  */
2664 static int
2665 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2666 {
2667 	int *name = (int *)arg1;
2668 	u_int namelen = arg2;
2669 	struct proc *p;
2670 	int flags, error, osrel;
2671 
2672 	if (namelen != 1)
2673 		return (EINVAL);
2674 
2675 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2676 		return (EINVAL);
2677 
2678 	flags = PGET_HOLD | PGET_NOTWEXIT;
2679 	if (req->newptr != NULL)
2680 		flags |= PGET_CANDEBUG;
2681 	else
2682 		flags |= PGET_CANSEE;
2683 	error = pget((pid_t)name[0], flags, &p);
2684 	if (error != 0)
2685 		return (error);
2686 
2687 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2688 	if (error != 0)
2689 		goto errout;
2690 
2691 	if (req->newptr != NULL) {
2692 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2693 		if (error != 0)
2694 			goto errout;
2695 		if (osrel < 0) {
2696 			error = EINVAL;
2697 			goto errout;
2698 		}
2699 		p->p_osrel = osrel;
2700 	}
2701 errout:
2702 	PRELE(p);
2703 	return (error);
2704 }
2705 
2706 static int
2707 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2708 {
2709 	int *name = (int *)arg1;
2710 	u_int namelen = arg2;
2711 	struct proc *p;
2712 	struct kinfo_sigtramp kst;
2713 	const struct sysentvec *sv;
2714 	int error;
2715 #ifdef COMPAT_FREEBSD32
2716 	struct kinfo_sigtramp32 kst32;
2717 #endif
2718 
2719 	if (namelen != 1)
2720 		return (EINVAL);
2721 
2722 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2723 	if (error != 0)
2724 		return (error);
2725 	sv = p->p_sysent;
2726 #ifdef COMPAT_FREEBSD32
2727 	if ((req->flags & SCTL_MASK32) != 0) {
2728 		bzero(&kst32, sizeof(kst32));
2729 		if (SV_PROC_FLAG(p, SV_ILP32)) {
2730 			if (sv->sv_sigcode_base != 0) {
2731 				kst32.ksigtramp_start = sv->sv_sigcode_base;
2732 				kst32.ksigtramp_end = sv->sv_sigcode_base +
2733 				    *sv->sv_szsigcode;
2734 			} else {
2735 				kst32.ksigtramp_start = sv->sv_psstrings -
2736 				    *sv->sv_szsigcode;
2737 				kst32.ksigtramp_end = sv->sv_psstrings;
2738 			}
2739 		}
2740 		PROC_UNLOCK(p);
2741 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2742 		return (error);
2743 	}
2744 #endif
2745 	bzero(&kst, sizeof(kst));
2746 	if (sv->sv_sigcode_base != 0) {
2747 		kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2748 		kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2749 		    *sv->sv_szsigcode;
2750 	} else {
2751 		kst.ksigtramp_start = (char *)sv->sv_psstrings -
2752 		    *sv->sv_szsigcode;
2753 		kst.ksigtramp_end = (char *)sv->sv_psstrings;
2754 	}
2755 	PROC_UNLOCK(p);
2756 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
2757 	return (error);
2758 }
2759 
2760 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2761 
2762 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2763 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2764 	"Return entire process table");
2765 
2766 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2767 	sysctl_kern_proc, "Process table");
2768 
2769 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2770 	sysctl_kern_proc, "Process table");
2771 
2772 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2773 	sysctl_kern_proc, "Process table");
2774 
2775 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2776 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2777 
2778 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2779 	sysctl_kern_proc, "Process table");
2780 
2781 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2782 	sysctl_kern_proc, "Process table");
2783 
2784 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2785 	sysctl_kern_proc, "Process table");
2786 
2787 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2788 	sysctl_kern_proc, "Process table");
2789 
2790 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2791 	sysctl_kern_proc, "Return process table, no threads");
2792 
2793 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2794 	CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2795 	sysctl_kern_proc_args, "Process argument list");
2796 
2797 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
2798 	sysctl_kern_proc_env, "Process environment");
2799 
2800 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
2801 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
2802 
2803 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2804 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2805 
2806 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2807 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2808 	"Process syscall vector name (ABI type)");
2809 
2810 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2811 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2812 
2813 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2814 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2815 
2816 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2817 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2818 
2819 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2820 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2821 
2822 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2823 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2824 
2825 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2826 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2827 
2828 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2829 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2830 
2831 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2832 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2833 
2834 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2835 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2836 	"Return process table, no threads");
2837 
2838 #ifdef COMPAT_FREEBSD7
2839 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2840 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2841 #endif
2842 
2843 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2844 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2845 
2846 #if defined(STACK) || defined(DDB)
2847 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2848 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2849 #endif
2850 
2851 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2852 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
2853 
2854 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
2855 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
2856 	"Process resource limits");
2857 
2858 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
2859 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
2860 	"Process ps_strings location");
2861 
2862 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
2863 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
2864 
2865 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
2866 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
2867 	"Process binary osreldate");
2868 
2869 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
2870 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
2871 	"Process signal trampoline location");
2872