xref: /freebsd/sys/kern/kern_proc.c (revision 6356dba0b403daa023dec24559ab1f8e602e4f14)
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_ddb.h"
36 #include "opt_kdtrace.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/kernel.h>
44 #include <sys/lock.h>
45 #include <sys/malloc.h>
46 #include <sys/mount.h>
47 #include <sys/mutex.h>
48 #include <sys/proc.h>
49 #include <sys/refcount.h>
50 #include <sys/sbuf.h>
51 #include <sys/sysent.h>
52 #include <sys/sched.h>
53 #include <sys/smp.h>
54 #include <sys/stack.h>
55 #include <sys/sysctl.h>
56 #include <sys/filedesc.h>
57 #include <sys/tty.h>
58 #include <sys/signalvar.h>
59 #include <sys/sdt.h>
60 #include <sys/sx.h>
61 #include <sys/user.h>
62 #include <sys/jail.h>
63 #include <sys/vnode.h>
64 #include <sys/eventhandler.h>
65 #ifdef KTRACE
66 #include <sys/uio.h>
67 #include <sys/ktrace.h>
68 #endif
69 
70 #ifdef DDB
71 #include <ddb/ddb.h>
72 #endif
73 
74 #include <vm/vm.h>
75 #include <vm/vm_extern.h>
76 #include <vm/pmap.h>
77 #include <vm/vm_map.h>
78 #include <vm/vm_object.h>
79 #include <vm/uma.h>
80 
81 SDT_PROVIDER_DEFINE(proc);
82 SDT_PROBE_DEFINE(proc, kernel, ctor, entry);
83 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 0, "struct proc *");
84 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 1, "int");
85 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 2, "void *");
86 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 3, "int");
87 SDT_PROBE_DEFINE(proc, kernel, ctor, return);
88 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 0, "struct proc *");
89 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 1, "int");
90 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 2, "void *");
91 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 3, "int");
92 SDT_PROBE_DEFINE(proc, kernel, dtor, entry);
93 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 0, "struct proc *");
94 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 1, "int");
95 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 2, "void *");
96 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 3, "struct thread *");
97 SDT_PROBE_DEFINE(proc, kernel, dtor, return);
98 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 0, "struct proc *");
99 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 1, "int");
100 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 2, "void *");
101 SDT_PROBE_DEFINE(proc, kernel, init, entry);
102 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 0, "struct proc *");
103 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 1, "int");
104 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 2, "int");
105 SDT_PROBE_DEFINE(proc, kernel, init, return);
106 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 0, "struct proc *");
107 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 1, "int");
108 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 2, "int");
109 
110 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
111 MALLOC_DEFINE(M_SESSION, "session", "session header");
112 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
113 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
114 
115 static void doenterpgrp(struct proc *, struct pgrp *);
116 static void orphanpg(struct pgrp *pg);
117 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
118 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
119     int preferthread);
120 static void pgadjustjobc(struct pgrp *pgrp, int entering);
121 static void pgdelete(struct pgrp *);
122 static int proc_ctor(void *mem, int size, void *arg, int flags);
123 static void proc_dtor(void *mem, int size, void *arg);
124 static int proc_init(void *mem, int size, int flags);
125 static void proc_fini(void *mem, int size);
126 static void pargs_free(struct pargs *pa);
127 
128 /*
129  * Other process lists
130  */
131 struct pidhashhead *pidhashtbl;
132 u_long pidhash;
133 struct pgrphashhead *pgrphashtbl;
134 u_long pgrphash;
135 struct proclist allproc;
136 struct proclist zombproc;
137 struct sx allproc_lock;
138 struct sx proctree_lock;
139 struct mtx ppeers_lock;
140 uma_zone_t proc_zone;
141 uma_zone_t ithread_zone;
142 
143 int kstack_pages = KSTACK_PAGES;
144 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0, "");
145 
146 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
147 
148 /*
149  * Initialize global process hashing structures.
150  */
151 void
152 procinit()
153 {
154 
155 	sx_init(&allproc_lock, "allproc");
156 	sx_init(&proctree_lock, "proctree");
157 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
158 	LIST_INIT(&allproc);
159 	LIST_INIT(&zombproc);
160 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
161 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
162 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
163 	    proc_ctor, proc_dtor, proc_init, proc_fini,
164 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
165 	uihashinit();
166 }
167 
168 /*
169  * Prepare a proc for use.
170  */
171 static int
172 proc_ctor(void *mem, int size, void *arg, int flags)
173 {
174 	struct proc *p;
175 
176 	p = (struct proc *)mem;
177 	SDT_PROBE(proc, kernel, ctor , entry, p, size, arg, flags, 0);
178 	EVENTHANDLER_INVOKE(process_ctor, p);
179 	SDT_PROBE(proc, kernel, ctor , return, p, size, arg, flags, 0);
180 	return (0);
181 }
182 
183 /*
184  * Reclaim a proc after use.
185  */
186 static void
187 proc_dtor(void *mem, int size, void *arg)
188 {
189 	struct proc *p;
190 	struct thread *td;
191 
192 	/* INVARIANTS checks go here */
193 	p = (struct proc *)mem;
194 	td = FIRST_THREAD_IN_PROC(p);
195 	SDT_PROBE(proc, kernel, dtor, entry, p, size, arg, td, 0);
196 	if (td != NULL) {
197 #ifdef INVARIANTS
198 		KASSERT((p->p_numthreads == 1),
199 		    ("bad number of threads in exiting process"));
200 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
201 #endif
202 		/* Dispose of an alternate kstack, if it exists.
203 		 * XXX What if there are more than one thread in the proc?
204 		 *     The first thread in the proc is special and not
205 		 *     freed, so you gotta do this here.
206 		 */
207 		if (((p->p_flag & P_KTHREAD) != 0) && (td->td_altkstack != 0))
208 			vm_thread_dispose_altkstack(td);
209 	}
210 	EVENTHANDLER_INVOKE(process_dtor, p);
211 	if (p->p_ksi != NULL)
212 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
213 	SDT_PROBE(proc, kernel, dtor, return, p, size, arg, 0, 0);
214 }
215 
216 /*
217  * Initialize type-stable parts of a proc (when newly created).
218  */
219 static int
220 proc_init(void *mem, int size, int flags)
221 {
222 	struct proc *p;
223 
224 	p = (struct proc *)mem;
225 	SDT_PROBE(proc, kernel, init, entry, p, size, flags, 0, 0);
226 	p->p_sched = (struct p_sched *)&p[1];
227 	bzero(&p->p_mtx, sizeof(struct mtx));
228 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK);
229 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_RECURSE);
230 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
231 	EVENTHANDLER_INVOKE(process_init, p);
232 	p->p_stats = pstats_alloc();
233 	SDT_PROBE(proc, kernel, init, return, p, size, flags, 0, 0);
234 	return (0);
235 }
236 
237 /*
238  * UMA should ensure that this function is never called.
239  * Freeing a proc structure would violate type stability.
240  */
241 static void
242 proc_fini(void *mem, int size)
243 {
244 #ifdef notnow
245 	struct proc *p;
246 
247 	p = (struct proc *)mem;
248 	EVENTHANDLER_INVOKE(process_fini, p);
249 	pstats_free(p->p_stats);
250 	thread_free(FIRST_THREAD_IN_PROC(p));
251 	mtx_destroy(&p->p_mtx);
252 	if (p->p_ksi != NULL)
253 		ksiginfo_free(p->p_ksi);
254 #else
255 	panic("proc reclaimed");
256 #endif
257 }
258 
259 /*
260  * Is p an inferior of the current process?
261  */
262 int
263 inferior(p)
264 	register struct proc *p;
265 {
266 
267 	sx_assert(&proctree_lock, SX_LOCKED);
268 	for (; p != curproc; p = p->p_pptr)
269 		if (p->p_pid == 0)
270 			return (0);
271 	return (1);
272 }
273 
274 /*
275  * Locate a process by number; return only "live" processes -- i.e., neither
276  * zombies nor newly born but incompletely initialized processes.  By not
277  * returning processes in the PRS_NEW state, we allow callers to avoid
278  * testing for that condition to avoid dereferencing p_ucred, et al.
279  */
280 struct proc *
281 pfind(pid)
282 	register pid_t pid;
283 {
284 	register struct proc *p;
285 
286 	sx_slock(&allproc_lock);
287 	LIST_FOREACH(p, PIDHASH(pid), p_hash)
288 		if (p->p_pid == pid) {
289 			if (p->p_state == PRS_NEW) {
290 				p = NULL;
291 				break;
292 			}
293 			PROC_LOCK(p);
294 			break;
295 		}
296 	sx_sunlock(&allproc_lock);
297 	return (p);
298 }
299 
300 /*
301  * Locate a process group by number.
302  * The caller must hold proctree_lock.
303  */
304 struct pgrp *
305 pgfind(pgid)
306 	register pid_t pgid;
307 {
308 	register struct pgrp *pgrp;
309 
310 	sx_assert(&proctree_lock, SX_LOCKED);
311 
312 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
313 		if (pgrp->pg_id == pgid) {
314 			PGRP_LOCK(pgrp);
315 			return (pgrp);
316 		}
317 	}
318 	return (NULL);
319 }
320 
321 /*
322  * Create a new process group.
323  * pgid must be equal to the pid of p.
324  * Begin a new session if required.
325  */
326 int
327 enterpgrp(p, pgid, pgrp, sess)
328 	register struct proc *p;
329 	pid_t pgid;
330 	struct pgrp *pgrp;
331 	struct session *sess;
332 {
333 	struct pgrp *pgrp2;
334 
335 	sx_assert(&proctree_lock, SX_XLOCKED);
336 
337 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
338 	KASSERT(p->p_pid == pgid,
339 	    ("enterpgrp: new pgrp and pid != pgid"));
340 
341 	pgrp2 = pgfind(pgid);
342 
343 	KASSERT(pgrp2 == NULL,
344 	    ("enterpgrp: pgrp with pgid exists"));
345 	KASSERT(!SESS_LEADER(p),
346 	    ("enterpgrp: session leader attempted setpgrp"));
347 
348 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
349 
350 	if (sess != NULL) {
351 		/*
352 		 * new session
353 		 */
354 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
355 		PROC_LOCK(p);
356 		p->p_flag &= ~P_CONTROLT;
357 		PROC_UNLOCK(p);
358 		PGRP_LOCK(pgrp);
359 		sess->s_leader = p;
360 		sess->s_sid = p->p_pid;
361 		refcount_init(&sess->s_count, 1);
362 		sess->s_ttyvp = NULL;
363 		sess->s_ttyp = NULL;
364 		bcopy(p->p_session->s_login, sess->s_login,
365 			    sizeof(sess->s_login));
366 		pgrp->pg_session = sess;
367 		KASSERT(p == curproc,
368 		    ("enterpgrp: mksession and p != curproc"));
369 	} else {
370 		pgrp->pg_session = p->p_session;
371 		sess_hold(pgrp->pg_session);
372 		PGRP_LOCK(pgrp);
373 	}
374 	pgrp->pg_id = pgid;
375 	LIST_INIT(&pgrp->pg_members);
376 
377 	/*
378 	 * As we have an exclusive lock of proctree_lock,
379 	 * this should not deadlock.
380 	 */
381 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
382 	pgrp->pg_jobc = 0;
383 	SLIST_INIT(&pgrp->pg_sigiolst);
384 	PGRP_UNLOCK(pgrp);
385 
386 	doenterpgrp(p, pgrp);
387 
388 	return (0);
389 }
390 
391 /*
392  * Move p to an existing process group
393  */
394 int
395 enterthispgrp(p, pgrp)
396 	register struct proc *p;
397 	struct pgrp *pgrp;
398 {
399 
400 	sx_assert(&proctree_lock, SX_XLOCKED);
401 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
402 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
403 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
404 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
405 	KASSERT(pgrp->pg_session == p->p_session,
406 		("%s: pgrp's session %p, p->p_session %p.\n",
407 		__func__,
408 		pgrp->pg_session,
409 		p->p_session));
410 	KASSERT(pgrp != p->p_pgrp,
411 		("%s: p belongs to pgrp.", __func__));
412 
413 	doenterpgrp(p, pgrp);
414 
415 	return (0);
416 }
417 
418 /*
419  * Move p to a process group
420  */
421 static void
422 doenterpgrp(p, pgrp)
423 	struct proc *p;
424 	struct pgrp *pgrp;
425 {
426 	struct pgrp *savepgrp;
427 
428 	sx_assert(&proctree_lock, SX_XLOCKED);
429 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
430 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
431 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
432 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
433 
434 	savepgrp = p->p_pgrp;
435 
436 	/*
437 	 * Adjust eligibility of affected pgrps to participate in job control.
438 	 * Increment eligibility counts before decrementing, otherwise we
439 	 * could reach 0 spuriously during the first call.
440 	 */
441 	fixjobc(p, pgrp, 1);
442 	fixjobc(p, p->p_pgrp, 0);
443 
444 	PGRP_LOCK(pgrp);
445 	PGRP_LOCK(savepgrp);
446 	PROC_LOCK(p);
447 	LIST_REMOVE(p, p_pglist);
448 	p->p_pgrp = pgrp;
449 	PROC_UNLOCK(p);
450 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
451 	PGRP_UNLOCK(savepgrp);
452 	PGRP_UNLOCK(pgrp);
453 	if (LIST_EMPTY(&savepgrp->pg_members))
454 		pgdelete(savepgrp);
455 }
456 
457 /*
458  * remove process from process group
459  */
460 int
461 leavepgrp(p)
462 	register struct proc *p;
463 {
464 	struct pgrp *savepgrp;
465 
466 	sx_assert(&proctree_lock, SX_XLOCKED);
467 	savepgrp = p->p_pgrp;
468 	PGRP_LOCK(savepgrp);
469 	PROC_LOCK(p);
470 	LIST_REMOVE(p, p_pglist);
471 	p->p_pgrp = NULL;
472 	PROC_UNLOCK(p);
473 	PGRP_UNLOCK(savepgrp);
474 	if (LIST_EMPTY(&savepgrp->pg_members))
475 		pgdelete(savepgrp);
476 	return (0);
477 }
478 
479 /*
480  * delete a process group
481  */
482 static void
483 pgdelete(pgrp)
484 	register struct pgrp *pgrp;
485 {
486 	struct session *savesess;
487 	struct tty *tp;
488 
489 	sx_assert(&proctree_lock, SX_XLOCKED);
490 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
491 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
492 
493 	/*
494 	 * Reset any sigio structures pointing to us as a result of
495 	 * F_SETOWN with our pgid.
496 	 */
497 	funsetownlst(&pgrp->pg_sigiolst);
498 
499 	PGRP_LOCK(pgrp);
500 	tp = pgrp->pg_session->s_ttyp;
501 	LIST_REMOVE(pgrp, pg_hash);
502 	savesess = pgrp->pg_session;
503 	PGRP_UNLOCK(pgrp);
504 
505 	/* Remove the reference to the pgrp before deallocating it. */
506 	if (tp != NULL) {
507 		tty_lock(tp);
508 		tty_rel_pgrp(tp, pgrp);
509 		tty_unlock(tp);
510 	}
511 
512 	mtx_destroy(&pgrp->pg_mtx);
513 	FREE(pgrp, M_PGRP);
514 	sess_release(savesess);
515 }
516 
517 static void
518 pgadjustjobc(pgrp, entering)
519 	struct pgrp *pgrp;
520 	int entering;
521 {
522 
523 	PGRP_LOCK(pgrp);
524 	if (entering)
525 		pgrp->pg_jobc++;
526 	else {
527 		--pgrp->pg_jobc;
528 		if (pgrp->pg_jobc == 0)
529 			orphanpg(pgrp);
530 	}
531 	PGRP_UNLOCK(pgrp);
532 }
533 
534 /*
535  * Adjust pgrp jobc counters when specified process changes process group.
536  * We count the number of processes in each process group that "qualify"
537  * the group for terminal job control (those with a parent in a different
538  * process group of the same session).  If that count reaches zero, the
539  * process group becomes orphaned.  Check both the specified process'
540  * process group and that of its children.
541  * entering == 0 => p is leaving specified group.
542  * entering == 1 => p is entering specified group.
543  */
544 void
545 fixjobc(p, pgrp, entering)
546 	register struct proc *p;
547 	register struct pgrp *pgrp;
548 	int entering;
549 {
550 	register struct pgrp *hispgrp;
551 	register struct session *mysession;
552 
553 	sx_assert(&proctree_lock, SX_LOCKED);
554 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
555 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
556 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
557 
558 	/*
559 	 * Check p's parent to see whether p qualifies its own process
560 	 * group; if so, adjust count for p's process group.
561 	 */
562 	mysession = pgrp->pg_session;
563 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
564 	    hispgrp->pg_session == mysession)
565 		pgadjustjobc(pgrp, entering);
566 
567 	/*
568 	 * Check this process' children to see whether they qualify
569 	 * their process groups; if so, adjust counts for children's
570 	 * process groups.
571 	 */
572 	LIST_FOREACH(p, &p->p_children, p_sibling) {
573 		hispgrp = p->p_pgrp;
574 		if (hispgrp == pgrp ||
575 		    hispgrp->pg_session != mysession)
576 			continue;
577 		PROC_LOCK(p);
578 		if (p->p_state == PRS_ZOMBIE) {
579 			PROC_UNLOCK(p);
580 			continue;
581 		}
582 		PROC_UNLOCK(p);
583 		pgadjustjobc(hispgrp, entering);
584 	}
585 }
586 
587 /*
588  * A process group has become orphaned;
589  * if there are any stopped processes in the group,
590  * hang-up all process in that group.
591  */
592 static void
593 orphanpg(pg)
594 	struct pgrp *pg;
595 {
596 	register struct proc *p;
597 
598 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
599 
600 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
601 		PROC_LOCK(p);
602 		if (P_SHOULDSTOP(p)) {
603 			PROC_UNLOCK(p);
604 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
605 				PROC_LOCK(p);
606 				psignal(p, SIGHUP);
607 				psignal(p, SIGCONT);
608 				PROC_UNLOCK(p);
609 			}
610 			return;
611 		}
612 		PROC_UNLOCK(p);
613 	}
614 }
615 
616 void
617 sess_hold(struct session *s)
618 {
619 
620 	refcount_acquire(&s->s_count);
621 }
622 
623 void
624 sess_release(struct session *s)
625 {
626 
627 	if (refcount_release(&s->s_count)) {
628 		if (s->s_ttyp != NULL) {
629 			tty_lock(s->s_ttyp);
630 			tty_rel_sess(s->s_ttyp, s);
631 		}
632 		mtx_destroy(&s->s_mtx);
633 		FREE(s, M_SESSION);
634 	}
635 }
636 
637 #include "opt_ddb.h"
638 #ifdef DDB
639 #include <ddb/ddb.h>
640 
641 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
642 {
643 	register struct pgrp *pgrp;
644 	register struct proc *p;
645 	register int i;
646 
647 	for (i = 0; i <= pgrphash; i++) {
648 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
649 			printf("\tindx %d\n", i);
650 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
651 				printf(
652 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
653 				    (void *)pgrp, (long)pgrp->pg_id,
654 				    (void *)pgrp->pg_session,
655 				    pgrp->pg_session->s_count,
656 				    (void *)LIST_FIRST(&pgrp->pg_members));
657 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
658 					printf("\t\tpid %ld addr %p pgrp %p\n",
659 					    (long)p->p_pid, (void *)p,
660 					    (void *)p->p_pgrp);
661 				}
662 			}
663 		}
664 	}
665 }
666 #endif /* DDB */
667 
668 /*
669  * Clear kinfo_proc and fill in any information that is common
670  * to all threads in the process.
671  * Must be called with the target process locked.
672  */
673 static void
674 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
675 {
676 	struct thread *td0;
677 	struct tty *tp;
678 	struct session *sp;
679 	struct ucred *cred;
680 	struct sigacts *ps;
681 
682 	PROC_LOCK_ASSERT(p, MA_OWNED);
683 	bzero(kp, sizeof(*kp));
684 
685 	kp->ki_structsize = sizeof(*kp);
686 	kp->ki_paddr = p;
687 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
688 	kp->ki_args = p->p_args;
689 	kp->ki_textvp = p->p_textvp;
690 #ifdef KTRACE
691 	kp->ki_tracep = p->p_tracevp;
692 	mtx_lock(&ktrace_mtx);
693 	kp->ki_traceflag = p->p_traceflag;
694 	mtx_unlock(&ktrace_mtx);
695 #endif
696 	kp->ki_fd = p->p_fd;
697 	kp->ki_vmspace = p->p_vmspace;
698 	kp->ki_flag = p->p_flag;
699 	cred = p->p_ucred;
700 	if (cred) {
701 		kp->ki_uid = cred->cr_uid;
702 		kp->ki_ruid = cred->cr_ruid;
703 		kp->ki_svuid = cred->cr_svuid;
704 		/* XXX bde doesn't like KI_NGROUPS */
705 		kp->ki_ngroups = min(cred->cr_ngroups, KI_NGROUPS);
706 		bcopy(cred->cr_groups, kp->ki_groups,
707 		    kp->ki_ngroups * sizeof(gid_t));
708 		kp->ki_rgid = cred->cr_rgid;
709 		kp->ki_svgid = cred->cr_svgid;
710 		/* If jailed(cred), emulate the old P_JAILED flag. */
711 		if (jailed(cred)) {
712 			kp->ki_flag |= P_JAILED;
713 			/* If inside a jail, use 0 as a jail ID. */
714 			if (!jailed(curthread->td_ucred))
715 				kp->ki_jid = cred->cr_prison->pr_id;
716 		}
717 	}
718 	ps = p->p_sigacts;
719 	if (ps) {
720 		mtx_lock(&ps->ps_mtx);
721 		kp->ki_sigignore = ps->ps_sigignore;
722 		kp->ki_sigcatch = ps->ps_sigcatch;
723 		mtx_unlock(&ps->ps_mtx);
724 	}
725 	PROC_SLOCK(p);
726 	if (p->p_state != PRS_NEW &&
727 	    p->p_state != PRS_ZOMBIE &&
728 	    p->p_vmspace != NULL) {
729 		struct vmspace *vm = p->p_vmspace;
730 
731 		kp->ki_size = vm->vm_map.size;
732 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
733 		FOREACH_THREAD_IN_PROC(p, td0) {
734 			if (!TD_IS_SWAPPED(td0))
735 				kp->ki_rssize += td0->td_kstack_pages;
736 			if (td0->td_altkstack_obj != NULL)
737 				kp->ki_rssize += td0->td_altkstack_pages;
738 		}
739 		kp->ki_swrss = vm->vm_swrss;
740 		kp->ki_tsize = vm->vm_tsize;
741 		kp->ki_dsize = vm->vm_dsize;
742 		kp->ki_ssize = vm->vm_ssize;
743 	} else if (p->p_state == PRS_ZOMBIE)
744 		kp->ki_stat = SZOMB;
745 	if (kp->ki_flag & P_INMEM)
746 		kp->ki_sflag = PS_INMEM;
747 	else
748 		kp->ki_sflag = 0;
749 	/* Calculate legacy swtime as seconds since 'swtick'. */
750 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
751 	kp->ki_pid = p->p_pid;
752 	kp->ki_nice = p->p_nice;
753 	rufetch(p, &kp->ki_rusage);
754 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
755 	PROC_SUNLOCK(p);
756 	if ((p->p_flag & P_INMEM) && p->p_stats != NULL) {
757 		kp->ki_start = p->p_stats->p_start;
758 		timevaladd(&kp->ki_start, &boottime);
759 		PROC_SLOCK(p);
760 		calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
761 		PROC_SUNLOCK(p);
762 		calccru(p, &kp->ki_childutime, &kp->ki_childstime);
763 
764 		/* Some callers want child-times in a single value */
765 		kp->ki_childtime = kp->ki_childstime;
766 		timevaladd(&kp->ki_childtime, &kp->ki_childutime);
767 	}
768 	tp = NULL;
769 	if (p->p_pgrp) {
770 		kp->ki_pgid = p->p_pgrp->pg_id;
771 		kp->ki_jobc = p->p_pgrp->pg_jobc;
772 		sp = p->p_pgrp->pg_session;
773 
774 		if (sp != NULL) {
775 			kp->ki_sid = sp->s_sid;
776 			SESS_LOCK(sp);
777 			strlcpy(kp->ki_login, sp->s_login,
778 			    sizeof(kp->ki_login));
779 			if (sp->s_ttyvp)
780 				kp->ki_kiflag |= KI_CTTY;
781 			if (SESS_LEADER(p))
782 				kp->ki_kiflag |= KI_SLEADER;
783 			/* XXX proctree_lock */
784 			tp = sp->s_ttyp;
785 			SESS_UNLOCK(sp);
786 		}
787 	}
788 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
789 		kp->ki_tdev = tty_udev(tp);
790 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
791 		if (tp->t_session)
792 			kp->ki_tsid = tp->t_session->s_sid;
793 	} else
794 		kp->ki_tdev = NODEV;
795 	if (p->p_comm[0] != '\0')
796 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
797 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
798 	    p->p_sysent->sv_name[0] != '\0')
799 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
800 	kp->ki_siglist = p->p_siglist;
801 	kp->ki_xstat = p->p_xstat;
802 	kp->ki_acflag = p->p_acflag;
803 	kp->ki_lock = p->p_lock;
804 	if (p->p_pptr)
805 		kp->ki_ppid = p->p_pptr->p_pid;
806 }
807 
808 /*
809  * Fill in information that is thread specific.  Must be called with p_slock
810  * locked.  If 'preferthread' is set, overwrite certain process-related
811  * fields that are maintained for both threads and processes.
812  */
813 static void
814 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
815 {
816 	struct proc *p;
817 
818 	p = td->td_proc;
819 	PROC_LOCK_ASSERT(p, MA_OWNED);
820 
821 	thread_lock(td);
822 	if (td->td_wmesg != NULL)
823 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
824 	else
825 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
826 	if (td->td_name[0] != '\0')
827 		strlcpy(kp->ki_ocomm, td->td_name, sizeof(kp->ki_ocomm));
828 	if (TD_ON_LOCK(td)) {
829 		kp->ki_kiflag |= KI_LOCKBLOCK;
830 		strlcpy(kp->ki_lockname, td->td_lockname,
831 		    sizeof(kp->ki_lockname));
832 	} else {
833 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
834 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
835 	}
836 
837 	if (p->p_state == PRS_NORMAL) { /* approximate. */
838 		if (TD_ON_RUNQ(td) ||
839 		    TD_CAN_RUN(td) ||
840 		    TD_IS_RUNNING(td)) {
841 			kp->ki_stat = SRUN;
842 		} else if (P_SHOULDSTOP(p)) {
843 			kp->ki_stat = SSTOP;
844 		} else if (TD_IS_SLEEPING(td)) {
845 			kp->ki_stat = SSLEEP;
846 		} else if (TD_ON_LOCK(td)) {
847 			kp->ki_stat = SLOCK;
848 		} else {
849 			kp->ki_stat = SWAIT;
850 		}
851 	} else if (p->p_state == PRS_ZOMBIE) {
852 		kp->ki_stat = SZOMB;
853 	} else {
854 		kp->ki_stat = SIDL;
855 	}
856 
857 	/* Things in the thread */
858 	kp->ki_wchan = td->td_wchan;
859 	kp->ki_pri.pri_level = td->td_priority;
860 	kp->ki_pri.pri_native = td->td_base_pri;
861 	kp->ki_lastcpu = td->td_lastcpu;
862 	kp->ki_oncpu = td->td_oncpu;
863 	kp->ki_tdflags = td->td_flags;
864 	kp->ki_tid = td->td_tid;
865 	kp->ki_numthreads = p->p_numthreads;
866 	kp->ki_pcb = td->td_pcb;
867 	kp->ki_kstack = (void *)td->td_kstack;
868 	kp->ki_pctcpu = sched_pctcpu(td);
869 	kp->ki_estcpu = td->td_estcpu;
870 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
871 	kp->ki_pri.pri_class = td->td_pri_class;
872 	kp->ki_pri.pri_user = td->td_user_pri;
873 
874 	if (preferthread)
875 		kp->ki_runtime = cputick2usec(td->td_runtime);
876 
877 	/* We can't get this anymore but ps etc never used it anyway. */
878 	kp->ki_rqindex = 0;
879 
880 	SIGSETOR(kp->ki_siglist, td->td_siglist);
881 	kp->ki_sigmask = td->td_sigmask;
882 	thread_unlock(td);
883 }
884 
885 /*
886  * Fill in a kinfo_proc structure for the specified process.
887  * Must be called with the target process locked.
888  */
889 void
890 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
891 {
892 
893 	fill_kinfo_proc_only(p, kp);
894 	if (FIRST_THREAD_IN_PROC(p) != NULL)
895 		fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
896 }
897 
898 struct pstats *
899 pstats_alloc(void)
900 {
901 
902 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
903 }
904 
905 /*
906  * Copy parts of p_stats; zero the rest of p_stats (statistics).
907  */
908 void
909 pstats_fork(struct pstats *src, struct pstats *dst)
910 {
911 
912 	bzero(&dst->pstat_startzero,
913 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
914 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
915 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
916 }
917 
918 void
919 pstats_free(struct pstats *ps)
920 {
921 
922 	free(ps, M_SUBPROC);
923 }
924 
925 /*
926  * Locate a zombie process by number
927  */
928 struct proc *
929 zpfind(pid_t pid)
930 {
931 	struct proc *p;
932 
933 	sx_slock(&allproc_lock);
934 	LIST_FOREACH(p, &zombproc, p_list)
935 		if (p->p_pid == pid) {
936 			PROC_LOCK(p);
937 			break;
938 		}
939 	sx_sunlock(&allproc_lock);
940 	return (p);
941 }
942 
943 #define KERN_PROC_ZOMBMASK	0x3
944 #define KERN_PROC_NOTHREADS	0x4
945 
946 /*
947  * Must be called with the process locked and will return with it unlocked.
948  */
949 static int
950 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
951 {
952 	struct thread *td;
953 	struct kinfo_proc kinfo_proc;
954 	int error = 0;
955 	struct proc *np;
956 	pid_t pid = p->p_pid;
957 
958 	PROC_LOCK_ASSERT(p, MA_OWNED);
959 
960 	fill_kinfo_proc_only(p, &kinfo_proc);
961 	if (flags & KERN_PROC_NOTHREADS) {
962 		if (FIRST_THREAD_IN_PROC(p) != NULL)
963 			fill_kinfo_thread(FIRST_THREAD_IN_PROC(p),
964 			    &kinfo_proc, 0);
965 		error = SYSCTL_OUT(req, (caddr_t)&kinfo_proc,
966 				   sizeof(kinfo_proc));
967 	} else {
968 		if (FIRST_THREAD_IN_PROC(p) != NULL)
969 			FOREACH_THREAD_IN_PROC(p, td) {
970 				fill_kinfo_thread(td, &kinfo_proc, 1);
971 				error = SYSCTL_OUT(req, (caddr_t)&kinfo_proc,
972 						   sizeof(kinfo_proc));
973 				if (error)
974 					break;
975 			}
976 		else
977 			error = SYSCTL_OUT(req, (caddr_t)&kinfo_proc,
978 					   sizeof(kinfo_proc));
979 	}
980 	PROC_UNLOCK(p);
981 	if (error)
982 		return (error);
983 	if (flags & KERN_PROC_ZOMBMASK)
984 		np = zpfind(pid);
985 	else {
986 		if (pid == 0)
987 			return (0);
988 		np = pfind(pid);
989 	}
990 	if (np == NULL)
991 		return EAGAIN;
992 	if (np != p) {
993 		PROC_UNLOCK(np);
994 		return EAGAIN;
995 	}
996 	PROC_UNLOCK(np);
997 	return (0);
998 }
999 
1000 static int
1001 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1002 {
1003 	int *name = (int*) arg1;
1004 	u_int namelen = arg2;
1005 	struct proc *p;
1006 	int flags, doingzomb, oid_number;
1007 	int error = 0;
1008 
1009 	oid_number = oidp->oid_number;
1010 	if (oid_number != KERN_PROC_ALL &&
1011 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1012 		flags = KERN_PROC_NOTHREADS;
1013 	else {
1014 		flags = 0;
1015 		oid_number &= ~KERN_PROC_INC_THREAD;
1016 	}
1017 	if (oid_number == KERN_PROC_PID) {
1018 		if (namelen != 1)
1019 			return (EINVAL);
1020 		error = sysctl_wire_old_buffer(req, 0);
1021 		if (error)
1022 			return (error);
1023 		p = pfind((pid_t)name[0]);
1024 		if (!p)
1025 			return (ESRCH);
1026 		if ((error = p_cansee(curthread, p))) {
1027 			PROC_UNLOCK(p);
1028 			return (error);
1029 		}
1030 		error = sysctl_out_proc(p, req, flags);
1031 		return (error);
1032 	}
1033 
1034 	switch (oid_number) {
1035 	case KERN_PROC_ALL:
1036 		if (namelen != 0)
1037 			return (EINVAL);
1038 		break;
1039 	case KERN_PROC_PROC:
1040 		if (namelen != 0 && namelen != 1)
1041 			return (EINVAL);
1042 		break;
1043 	default:
1044 		if (namelen != 1)
1045 			return (EINVAL);
1046 		break;
1047 	}
1048 
1049 	if (!req->oldptr) {
1050 		/* overestimate by 5 procs */
1051 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1052 		if (error)
1053 			return (error);
1054 	}
1055 	error = sysctl_wire_old_buffer(req, 0);
1056 	if (error != 0)
1057 		return (error);
1058 	sx_slock(&allproc_lock);
1059 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1060 		if (!doingzomb)
1061 			p = LIST_FIRST(&allproc);
1062 		else
1063 			p = LIST_FIRST(&zombproc);
1064 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
1065 			/*
1066 			 * Skip embryonic processes.
1067 			 */
1068 			PROC_SLOCK(p);
1069 			if (p->p_state == PRS_NEW) {
1070 				PROC_SUNLOCK(p);
1071 				continue;
1072 			}
1073 			PROC_SUNLOCK(p);
1074 			PROC_LOCK(p);
1075 			KASSERT(p->p_ucred != NULL,
1076 			    ("process credential is NULL for non-NEW proc"));
1077 			/*
1078 			 * Show a user only appropriate processes.
1079 			 */
1080 			if (p_cansee(curthread, p)) {
1081 				PROC_UNLOCK(p);
1082 				continue;
1083 			}
1084 			/*
1085 			 * TODO - make more efficient (see notes below).
1086 			 * do by session.
1087 			 */
1088 			switch (oid_number) {
1089 
1090 			case KERN_PROC_GID:
1091 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1092 					PROC_UNLOCK(p);
1093 					continue;
1094 				}
1095 				break;
1096 
1097 			case KERN_PROC_PGRP:
1098 				/* could do this by traversing pgrp */
1099 				if (p->p_pgrp == NULL ||
1100 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1101 					PROC_UNLOCK(p);
1102 					continue;
1103 				}
1104 				break;
1105 
1106 			case KERN_PROC_RGID:
1107 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1108 					PROC_UNLOCK(p);
1109 					continue;
1110 				}
1111 				break;
1112 
1113 			case KERN_PROC_SESSION:
1114 				if (p->p_session == NULL ||
1115 				    p->p_session->s_sid != (pid_t)name[0]) {
1116 					PROC_UNLOCK(p);
1117 					continue;
1118 				}
1119 				break;
1120 
1121 			case KERN_PROC_TTY:
1122 				if ((p->p_flag & P_CONTROLT) == 0 ||
1123 				    p->p_session == NULL) {
1124 					PROC_UNLOCK(p);
1125 					continue;
1126 				}
1127 				/* XXX proctree_lock */
1128 				SESS_LOCK(p->p_session);
1129 				if (p->p_session->s_ttyp == NULL ||
1130 				    tty_udev(p->p_session->s_ttyp) !=
1131 				    (dev_t)name[0]) {
1132 					SESS_UNLOCK(p->p_session);
1133 					PROC_UNLOCK(p);
1134 					continue;
1135 				}
1136 				SESS_UNLOCK(p->p_session);
1137 				break;
1138 
1139 			case KERN_PROC_UID:
1140 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1141 					PROC_UNLOCK(p);
1142 					continue;
1143 				}
1144 				break;
1145 
1146 			case KERN_PROC_RUID:
1147 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1148 					PROC_UNLOCK(p);
1149 					continue;
1150 				}
1151 				break;
1152 
1153 			case KERN_PROC_PROC:
1154 				break;
1155 
1156 			default:
1157 				break;
1158 
1159 			}
1160 
1161 			error = sysctl_out_proc(p, req, flags | doingzomb);
1162 			if (error) {
1163 				sx_sunlock(&allproc_lock);
1164 				return (error);
1165 			}
1166 		}
1167 	}
1168 	sx_sunlock(&allproc_lock);
1169 	return (0);
1170 }
1171 
1172 struct pargs *
1173 pargs_alloc(int len)
1174 {
1175 	struct pargs *pa;
1176 
1177 	MALLOC(pa, struct pargs *, sizeof(struct pargs) + len, M_PARGS,
1178 		M_WAITOK);
1179 	refcount_init(&pa->ar_ref, 1);
1180 	pa->ar_length = len;
1181 	return (pa);
1182 }
1183 
1184 static void
1185 pargs_free(struct pargs *pa)
1186 {
1187 
1188 	FREE(pa, M_PARGS);
1189 }
1190 
1191 void
1192 pargs_hold(struct pargs *pa)
1193 {
1194 
1195 	if (pa == NULL)
1196 		return;
1197 	refcount_acquire(&pa->ar_ref);
1198 }
1199 
1200 void
1201 pargs_drop(struct pargs *pa)
1202 {
1203 
1204 	if (pa == NULL)
1205 		return;
1206 	if (refcount_release(&pa->ar_ref))
1207 		pargs_free(pa);
1208 }
1209 
1210 /*
1211  * This sysctl allows a process to retrieve the argument list or process
1212  * title for another process without groping around in the address space
1213  * of the other process.  It also allow a process to set its own "process
1214  * title to a string of its own choice.
1215  */
1216 static int
1217 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1218 {
1219 	int *name = (int*) arg1;
1220 	u_int namelen = arg2;
1221 	struct pargs *newpa, *pa;
1222 	struct proc *p;
1223 	int error = 0;
1224 
1225 	if (namelen != 1)
1226 		return (EINVAL);
1227 
1228 	p = pfind((pid_t)name[0]);
1229 	if (!p)
1230 		return (ESRCH);
1231 
1232 	if ((error = p_cansee(curthread, p)) != 0) {
1233 		PROC_UNLOCK(p);
1234 		return (error);
1235 	}
1236 
1237 	if (req->newptr && curproc != p) {
1238 		PROC_UNLOCK(p);
1239 		return (EPERM);
1240 	}
1241 
1242 	pa = p->p_args;
1243 	pargs_hold(pa);
1244 	PROC_UNLOCK(p);
1245 	if (req->oldptr != NULL && pa != NULL)
1246 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1247 	pargs_drop(pa);
1248 	if (error != 0 || req->newptr == NULL)
1249 		return (error);
1250 
1251 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1252 		return (ENOMEM);
1253 	newpa = pargs_alloc(req->newlen);
1254 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1255 	if (error != 0) {
1256 		pargs_free(newpa);
1257 		return (error);
1258 	}
1259 	PROC_LOCK(p);
1260 	pa = p->p_args;
1261 	p->p_args = newpa;
1262 	PROC_UNLOCK(p);
1263 	pargs_drop(pa);
1264 	return (0);
1265 }
1266 
1267 /*
1268  * This sysctl allows a process to retrieve the path of the executable for
1269  * itself or another process.
1270  */
1271 static int
1272 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1273 {
1274 	pid_t *pidp = (pid_t *)arg1;
1275 	unsigned int arglen = arg2;
1276 	struct proc *p;
1277 	struct vnode *vp;
1278 	char *retbuf, *freebuf;
1279 	int error;
1280 
1281 	if (arglen != 1)
1282 		return (EINVAL);
1283 	if (*pidp == -1) {	/* -1 means this process */
1284 		p = req->td->td_proc;
1285 	} else {
1286 		p = pfind(*pidp);
1287 		if (p == NULL)
1288 			return (ESRCH);
1289 		if ((error = p_cansee(curthread, p)) != 0) {
1290 			PROC_UNLOCK(p);
1291 			return (error);
1292 		}
1293 	}
1294 
1295 	vp = p->p_textvp;
1296 	if (vp == NULL) {
1297 		if (*pidp != -1)
1298 			PROC_UNLOCK(p);
1299 		return (0);
1300 	}
1301 	vref(vp);
1302 	if (*pidp != -1)
1303 		PROC_UNLOCK(p);
1304 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1305 	vrele(vp);
1306 	if (error)
1307 		return (error);
1308 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1309 	free(freebuf, M_TEMP);
1310 	return (error);
1311 }
1312 
1313 static int
1314 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1315 {
1316 	struct proc *p;
1317 	char *sv_name;
1318 	int *name;
1319 	int namelen;
1320 	int error;
1321 
1322 	namelen = arg2;
1323 	if (namelen != 1)
1324 		return (EINVAL);
1325 
1326 	name = (int *)arg1;
1327 	if ((p = pfind((pid_t)name[0])) == NULL)
1328 		return (ESRCH);
1329 	if ((error = p_cansee(curthread, p))) {
1330 		PROC_UNLOCK(p);
1331 		return (error);
1332 	}
1333 	sv_name = p->p_sysent->sv_name;
1334 	PROC_UNLOCK(p);
1335 	return (sysctl_handle_string(oidp, sv_name, 0, req));
1336 }
1337 
1338 static int
1339 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
1340 {
1341 	vm_map_entry_t entry, tmp_entry;
1342 	unsigned int last_timestamp;
1343 	char *fullpath, *freepath;
1344 	struct kinfo_vmentry *kve;
1345 	int error, *name;
1346 	struct vnode *vp;
1347 	struct proc *p;
1348 	vm_map_t map;
1349 
1350 	name = (int *)arg1;
1351 	if ((p = pfind((pid_t)name[0])) == NULL)
1352 		return (ESRCH);
1353 	if (p->p_flag & P_WEXIT) {
1354 		PROC_UNLOCK(p);
1355 		return (ESRCH);
1356 	}
1357 	if ((error = p_candebug(curthread, p))) {
1358 		PROC_UNLOCK(p);
1359 		return (error);
1360 	}
1361 	_PHOLD(p);
1362 	PROC_UNLOCK(p);
1363 
1364 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1365 
1366 	map = &p->p_vmspace->vm_map;	/* XXXRW: More locking required? */
1367 	vm_map_lock_read(map);
1368 	for (entry = map->header.next; entry != &map->header;
1369 	    entry = entry->next) {
1370 		vm_object_t obj, tobj, lobj;
1371 		vm_offset_t addr;
1372 		int vfslocked;
1373 
1374 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
1375 			continue;
1376 
1377 		bzero(kve, sizeof(*kve));
1378 		kve->kve_structsize = sizeof(*kve);
1379 
1380 		kve->kve_private_resident = 0;
1381 		obj = entry->object.vm_object;
1382 		if (obj != NULL) {
1383 			VM_OBJECT_LOCK(obj);
1384 			if (obj->shadow_count == 1)
1385 				kve->kve_private_resident =
1386 				    obj->resident_page_count;
1387 		}
1388 		kve->kve_resident = 0;
1389 		addr = entry->start;
1390 		while (addr < entry->end) {
1391 			if (pmap_extract(map->pmap, addr))
1392 				kve->kve_resident++;
1393 			addr += PAGE_SIZE;
1394 		}
1395 
1396 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
1397 			if (tobj != obj)
1398 				VM_OBJECT_LOCK(tobj);
1399 			if (lobj != obj)
1400 				VM_OBJECT_UNLOCK(lobj);
1401 			lobj = tobj;
1402 		}
1403 
1404 		freepath = NULL;
1405 		fullpath = "";
1406 		if (lobj) {
1407 			vp = NULL;
1408 			switch(lobj->type) {
1409 			case OBJT_DEFAULT:
1410 				kve->kve_type = KVME_TYPE_DEFAULT;
1411 				break;
1412 			case OBJT_VNODE:
1413 				kve->kve_type = KVME_TYPE_VNODE;
1414 				vp = lobj->handle;
1415 				vref(vp);
1416 				break;
1417 			case OBJT_SWAP:
1418 				kve->kve_type = KVME_TYPE_SWAP;
1419 				break;
1420 			case OBJT_DEVICE:
1421 				kve->kve_type = KVME_TYPE_DEVICE;
1422 				break;
1423 			case OBJT_PHYS:
1424 				kve->kve_type = KVME_TYPE_PHYS;
1425 				break;
1426 			case OBJT_DEAD:
1427 				kve->kve_type = KVME_TYPE_DEAD;
1428 				break;
1429 			default:
1430 				kve->kve_type = KVME_TYPE_UNKNOWN;
1431 				break;
1432 			}
1433 			if (lobj != obj)
1434 				VM_OBJECT_UNLOCK(lobj);
1435 
1436 			kve->kve_ref_count = obj->ref_count;
1437 			kve->kve_shadow_count = obj->shadow_count;
1438 			VM_OBJECT_UNLOCK(obj);
1439 			if (vp != NULL) {
1440 				vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1441 				vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1442 				vn_fullpath(curthread, vp, &fullpath,
1443 				    &freepath);
1444 				vput(vp);
1445 				VFS_UNLOCK_GIANT(vfslocked);
1446 			}
1447 		} else {
1448 			kve->kve_type = KVME_TYPE_NONE;
1449 			kve->kve_ref_count = 0;
1450 			kve->kve_shadow_count = 0;
1451 		}
1452 
1453 		kve->kve_start = (void*)entry->start;
1454 		kve->kve_end = (void*)entry->end;
1455 
1456 		if (entry->protection & VM_PROT_READ)
1457 			kve->kve_protection |= KVME_PROT_READ;
1458 		if (entry->protection & VM_PROT_WRITE)
1459 			kve->kve_protection |= KVME_PROT_WRITE;
1460 		if (entry->protection & VM_PROT_EXECUTE)
1461 			kve->kve_protection |= KVME_PROT_EXEC;
1462 
1463 		if (entry->eflags & MAP_ENTRY_COW)
1464 			kve->kve_flags |= KVME_FLAG_COW;
1465 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
1466 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
1467 
1468 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
1469 		if (freepath != NULL)
1470 			free(freepath, M_TEMP);
1471 
1472 		last_timestamp = map->timestamp;
1473 		vm_map_unlock_read(map);
1474 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
1475 		vm_map_lock_read(map);
1476 		if (error)
1477 			break;
1478 		if (last_timestamp + 1 != map->timestamp) {
1479 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
1480 			entry = tmp_entry;
1481 		}
1482 	}
1483 	vm_map_unlock_read(map);
1484 	PRELE(p);
1485 	free(kve, M_TEMP);
1486 	return (error);
1487 }
1488 
1489 #if defined(STACK) || defined(DDB)
1490 static int
1491 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
1492 {
1493 	struct kinfo_kstack *kkstp;
1494 	int error, i, *name, numthreads;
1495 	lwpid_t *lwpidarray;
1496 	struct thread *td;
1497 	struct stack *st;
1498 	struct sbuf sb;
1499 	struct proc *p;
1500 
1501 	name = (int *)arg1;
1502 	if ((p = pfind((pid_t)name[0])) == NULL)
1503 		return (ESRCH);
1504 	/* XXXRW: Not clear ESRCH is the right error during proc execve(). */
1505 	if (p->p_flag & P_WEXIT || p->p_flag & P_INEXEC) {
1506 		PROC_UNLOCK(p);
1507 		return (ESRCH);
1508 	}
1509 	if ((error = p_candebug(curthread, p))) {
1510 		PROC_UNLOCK(p);
1511 		return (error);
1512 	}
1513 	_PHOLD(p);
1514 	PROC_UNLOCK(p);
1515 
1516 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
1517 	st = stack_create();
1518 
1519 	lwpidarray = NULL;
1520 	numthreads = 0;
1521 	PROC_LOCK(p);
1522 repeat:
1523 	if (numthreads < p->p_numthreads) {
1524 		if (lwpidarray != NULL) {
1525 			free(lwpidarray, M_TEMP);
1526 			lwpidarray = NULL;
1527 		}
1528 		numthreads = p->p_numthreads;
1529 		PROC_UNLOCK(p);
1530 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
1531 		    M_WAITOK | M_ZERO);
1532 		PROC_LOCK(p);
1533 		goto repeat;
1534 	}
1535 	i = 0;
1536 
1537 	/*
1538 	 * XXXRW: During the below loop, execve(2) and countless other sorts
1539 	 * of changes could have taken place.  Should we check to see if the
1540 	 * vmspace has been replaced, or the like, in order to prevent
1541 	 * giving a snapshot that spans, say, execve(2), with some threads
1542 	 * before and some after?  Among other things, the credentials could
1543 	 * have changed, in which case the right to extract debug info might
1544 	 * no longer be assured.
1545 	 */
1546 	FOREACH_THREAD_IN_PROC(p, td) {
1547 		KASSERT(i < numthreads,
1548 		    ("sysctl_kern_proc_kstack: numthreads"));
1549 		lwpidarray[i] = td->td_tid;
1550 		i++;
1551 	}
1552 	numthreads = i;
1553 	for (i = 0; i < numthreads; i++) {
1554 		td = thread_find(p, lwpidarray[i]);
1555 		if (td == NULL) {
1556 			continue;
1557 		}
1558 		bzero(kkstp, sizeof(*kkstp));
1559 		(void)sbuf_new(&sb, kkstp->kkst_trace,
1560 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
1561 		thread_lock(td);
1562 		kkstp->kkst_tid = td->td_tid;
1563 		if (TD_IS_SWAPPED(td))
1564 			kkstp->kkst_state = KKST_STATE_SWAPPED;
1565 		else if (TD_IS_RUNNING(td))
1566 			kkstp->kkst_state = KKST_STATE_RUNNING;
1567 		else {
1568 			kkstp->kkst_state = KKST_STATE_STACKOK;
1569 			stack_save_td(st, td);
1570 		}
1571 		thread_unlock(td);
1572 		PROC_UNLOCK(p);
1573 		stack_sbuf_print(&sb, st);
1574 		sbuf_finish(&sb);
1575 		sbuf_delete(&sb);
1576 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
1577 		PROC_LOCK(p);
1578 		if (error)
1579 			break;
1580 	}
1581 	_PRELE(p);
1582 	PROC_UNLOCK(p);
1583 	if (lwpidarray != NULL)
1584 		free(lwpidarray, M_TEMP);
1585 	stack_destroy(st);
1586 	free(kkstp, M_TEMP);
1587 	return (error);
1588 }
1589 #endif
1590 
1591 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
1592 
1593 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT,
1594 	0, 0, sysctl_kern_proc, "S,proc", "Return entire process table");
1595 
1596 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD,
1597 	sysctl_kern_proc, "Process table");
1598 
1599 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD,
1600 	sysctl_kern_proc, "Process table");
1601 
1602 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD,
1603 	sysctl_kern_proc, "Process table");
1604 
1605 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD,
1606 	sysctl_kern_proc, "Process table");
1607 
1608 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD,
1609 	sysctl_kern_proc, "Process table");
1610 
1611 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD,
1612 	sysctl_kern_proc, "Process table");
1613 
1614 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD,
1615 	sysctl_kern_proc, "Process table");
1616 
1617 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD,
1618 	sysctl_kern_proc, "Process table");
1619 
1620 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD,
1621 	sysctl_kern_proc, "Return process table, no threads");
1622 
1623 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
1624 	CTLFLAG_RW | CTLFLAG_ANYBODY,
1625 	sysctl_kern_proc_args, "Process argument list");
1626 
1627 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD,
1628 	sysctl_kern_proc_pathname, "Process executable path");
1629 
1630 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD,
1631 	sysctl_kern_proc_sv_name, "Process syscall vector name (ABI type)");
1632 
1633 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
1634 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1635 
1636 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
1637 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1638 
1639 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
1640 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1641 
1642 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
1643 	sid_td, CTLFLAG_RD, sysctl_kern_proc, "Process table");
1644 
1645 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
1646 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1647 
1648 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
1649 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1650 
1651 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
1652 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1653 
1654 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
1655 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1656 
1657 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
1658 	CTLFLAG_RD, sysctl_kern_proc, "Return process table, no threads");
1659 
1660 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD,
1661 	sysctl_kern_proc_vmmap, "Process vm map entries");
1662 
1663 #if defined(STACK) || defined(DDB)
1664 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD,
1665 	sysctl_kern_proc_kstack, "Process kernel stacks");
1666 #endif
1667