xref: /freebsd/sys/kern/kern_proc.c (revision a9148abd9da5db2f1c682fb17bed791845fc41c9)
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 	}
510 
511 	mtx_destroy(&pgrp->pg_mtx);
512 	free(pgrp, M_PGRP);
513 	sess_release(savesess);
514 }
515 
516 static void
517 pgadjustjobc(pgrp, entering)
518 	struct pgrp *pgrp;
519 	int entering;
520 {
521 
522 	PGRP_LOCK(pgrp);
523 	if (entering)
524 		pgrp->pg_jobc++;
525 	else {
526 		--pgrp->pg_jobc;
527 		if (pgrp->pg_jobc == 0)
528 			orphanpg(pgrp);
529 	}
530 	PGRP_UNLOCK(pgrp);
531 }
532 
533 /*
534  * Adjust pgrp jobc counters when specified process changes process group.
535  * We count the number of processes in each process group that "qualify"
536  * the group for terminal job control (those with a parent in a different
537  * process group of the same session).  If that count reaches zero, the
538  * process group becomes orphaned.  Check both the specified process'
539  * process group and that of its children.
540  * entering == 0 => p is leaving specified group.
541  * entering == 1 => p is entering specified group.
542  */
543 void
544 fixjobc(p, pgrp, entering)
545 	register struct proc *p;
546 	register struct pgrp *pgrp;
547 	int entering;
548 {
549 	register struct pgrp *hispgrp;
550 	register struct session *mysession;
551 
552 	sx_assert(&proctree_lock, SX_LOCKED);
553 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
554 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
555 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
556 
557 	/*
558 	 * Check p's parent to see whether p qualifies its own process
559 	 * group; if so, adjust count for p's process group.
560 	 */
561 	mysession = pgrp->pg_session;
562 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
563 	    hispgrp->pg_session == mysession)
564 		pgadjustjobc(pgrp, entering);
565 
566 	/*
567 	 * Check this process' children to see whether they qualify
568 	 * their process groups; if so, adjust counts for children's
569 	 * process groups.
570 	 */
571 	LIST_FOREACH(p, &p->p_children, p_sibling) {
572 		hispgrp = p->p_pgrp;
573 		if (hispgrp == pgrp ||
574 		    hispgrp->pg_session != mysession)
575 			continue;
576 		PROC_LOCK(p);
577 		if (p->p_state == PRS_ZOMBIE) {
578 			PROC_UNLOCK(p);
579 			continue;
580 		}
581 		PROC_UNLOCK(p);
582 		pgadjustjobc(hispgrp, entering);
583 	}
584 }
585 
586 /*
587  * A process group has become orphaned;
588  * if there are any stopped processes in the group,
589  * hang-up all process in that group.
590  */
591 static void
592 orphanpg(pg)
593 	struct pgrp *pg;
594 {
595 	register struct proc *p;
596 
597 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
598 
599 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
600 		PROC_LOCK(p);
601 		if (P_SHOULDSTOP(p)) {
602 			PROC_UNLOCK(p);
603 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
604 				PROC_LOCK(p);
605 				psignal(p, SIGHUP);
606 				psignal(p, SIGCONT);
607 				PROC_UNLOCK(p);
608 			}
609 			return;
610 		}
611 		PROC_UNLOCK(p);
612 	}
613 }
614 
615 void
616 sess_hold(struct session *s)
617 {
618 
619 	refcount_acquire(&s->s_count);
620 }
621 
622 void
623 sess_release(struct session *s)
624 {
625 
626 	if (refcount_release(&s->s_count)) {
627 		if (s->s_ttyp != NULL) {
628 			tty_lock(s->s_ttyp);
629 			tty_rel_sess(s->s_ttyp, s);
630 		}
631 		mtx_destroy(&s->s_mtx);
632 		free(s, M_SESSION);
633 	}
634 }
635 
636 #include "opt_ddb.h"
637 #ifdef DDB
638 #include <ddb/ddb.h>
639 
640 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
641 {
642 	register struct pgrp *pgrp;
643 	register struct proc *p;
644 	register int i;
645 
646 	for (i = 0; i <= pgrphash; i++) {
647 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
648 			printf("\tindx %d\n", i);
649 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
650 				printf(
651 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
652 				    (void *)pgrp, (long)pgrp->pg_id,
653 				    (void *)pgrp->pg_session,
654 				    pgrp->pg_session->s_count,
655 				    (void *)LIST_FIRST(&pgrp->pg_members));
656 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
657 					printf("\t\tpid %ld addr %p pgrp %p\n",
658 					    (long)p->p_pid, (void *)p,
659 					    (void *)p->p_pgrp);
660 				}
661 			}
662 		}
663 	}
664 }
665 #endif /* DDB */
666 
667 /*
668  * Clear kinfo_proc and fill in any information that is common
669  * to all threads in the process.
670  * Must be called with the target process locked.
671  */
672 static void
673 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
674 {
675 	struct thread *td0;
676 	struct tty *tp;
677 	struct session *sp;
678 	struct ucred *cred;
679 	struct sigacts *ps;
680 
681 	PROC_LOCK_ASSERT(p, MA_OWNED);
682 	bzero(kp, sizeof(*kp));
683 
684 	kp->ki_structsize = sizeof(*kp);
685 	kp->ki_paddr = p;
686 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
687 	kp->ki_args = p->p_args;
688 	kp->ki_textvp = p->p_textvp;
689 #ifdef KTRACE
690 	kp->ki_tracep = p->p_tracevp;
691 	mtx_lock(&ktrace_mtx);
692 	kp->ki_traceflag = p->p_traceflag;
693 	mtx_unlock(&ktrace_mtx);
694 #endif
695 	kp->ki_fd = p->p_fd;
696 	kp->ki_vmspace = p->p_vmspace;
697 	kp->ki_flag = p->p_flag;
698 	cred = p->p_ucred;
699 	if (cred) {
700 		kp->ki_uid = cred->cr_uid;
701 		kp->ki_ruid = cred->cr_ruid;
702 		kp->ki_svuid = cred->cr_svuid;
703 		/* XXX bde doesn't like KI_NGROUPS */
704 		kp->ki_ngroups = min(cred->cr_ngroups, KI_NGROUPS);
705 		bcopy(cred->cr_groups, kp->ki_groups,
706 		    kp->ki_ngroups * sizeof(gid_t));
707 		kp->ki_rgid = cred->cr_rgid;
708 		kp->ki_svgid = cred->cr_svgid;
709 		/* If jailed(cred), emulate the old P_JAILED flag. */
710 		if (jailed(cred)) {
711 			kp->ki_flag |= P_JAILED;
712 			/* If inside a jail, use 0 as a jail ID. */
713 			if (!jailed(curthread->td_ucred))
714 				kp->ki_jid = cred->cr_prison->pr_id;
715 		}
716 	}
717 	ps = p->p_sigacts;
718 	if (ps) {
719 		mtx_lock(&ps->ps_mtx);
720 		kp->ki_sigignore = ps->ps_sigignore;
721 		kp->ki_sigcatch = ps->ps_sigcatch;
722 		mtx_unlock(&ps->ps_mtx);
723 	}
724 	PROC_SLOCK(p);
725 	if (p->p_state != PRS_NEW &&
726 	    p->p_state != PRS_ZOMBIE &&
727 	    p->p_vmspace != NULL) {
728 		struct vmspace *vm = p->p_vmspace;
729 
730 		kp->ki_size = vm->vm_map.size;
731 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
732 		FOREACH_THREAD_IN_PROC(p, td0) {
733 			if (!TD_IS_SWAPPED(td0))
734 				kp->ki_rssize += td0->td_kstack_pages;
735 			if (td0->td_altkstack_obj != NULL)
736 				kp->ki_rssize += td0->td_altkstack_pages;
737 		}
738 		kp->ki_swrss = vm->vm_swrss;
739 		kp->ki_tsize = vm->vm_tsize;
740 		kp->ki_dsize = vm->vm_dsize;
741 		kp->ki_ssize = vm->vm_ssize;
742 	} else if (p->p_state == PRS_ZOMBIE)
743 		kp->ki_stat = SZOMB;
744 	if (kp->ki_flag & P_INMEM)
745 		kp->ki_sflag = PS_INMEM;
746 	else
747 		kp->ki_sflag = 0;
748 	/* Calculate legacy swtime as seconds since 'swtick'. */
749 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
750 	kp->ki_pid = p->p_pid;
751 	kp->ki_nice = p->p_nice;
752 	rufetch(p, &kp->ki_rusage);
753 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
754 	PROC_SUNLOCK(p);
755 	if ((p->p_flag & P_INMEM) && p->p_stats != NULL) {
756 		kp->ki_start = p->p_stats->p_start;
757 		timevaladd(&kp->ki_start, &boottime);
758 		PROC_SLOCK(p);
759 		calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
760 		PROC_SUNLOCK(p);
761 		calccru(p, &kp->ki_childutime, &kp->ki_childstime);
762 
763 		/* Some callers want child-times in a single value */
764 		kp->ki_childtime = kp->ki_childstime;
765 		timevaladd(&kp->ki_childtime, &kp->ki_childutime);
766 	}
767 	tp = NULL;
768 	if (p->p_pgrp) {
769 		kp->ki_pgid = p->p_pgrp->pg_id;
770 		kp->ki_jobc = p->p_pgrp->pg_jobc;
771 		sp = p->p_pgrp->pg_session;
772 
773 		if (sp != NULL) {
774 			kp->ki_sid = sp->s_sid;
775 			SESS_LOCK(sp);
776 			strlcpy(kp->ki_login, sp->s_login,
777 			    sizeof(kp->ki_login));
778 			if (sp->s_ttyvp)
779 				kp->ki_kiflag |= KI_CTTY;
780 			if (SESS_LEADER(p))
781 				kp->ki_kiflag |= KI_SLEADER;
782 			/* XXX proctree_lock */
783 			tp = sp->s_ttyp;
784 			SESS_UNLOCK(sp);
785 		}
786 	}
787 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
788 		kp->ki_tdev = tty_udev(tp);
789 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
790 		if (tp->t_session)
791 			kp->ki_tsid = tp->t_session->s_sid;
792 	} else
793 		kp->ki_tdev = NODEV;
794 	if (p->p_comm[0] != '\0')
795 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
796 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
797 	    p->p_sysent->sv_name[0] != '\0')
798 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
799 	kp->ki_siglist = p->p_siglist;
800 	kp->ki_xstat = p->p_xstat;
801 	kp->ki_acflag = p->p_acflag;
802 	kp->ki_lock = p->p_lock;
803 	if (p->p_pptr)
804 		kp->ki_ppid = p->p_pptr->p_pid;
805 }
806 
807 /*
808  * Fill in information that is thread specific.  Must be called with p_slock
809  * locked.  If 'preferthread' is set, overwrite certain process-related
810  * fields that are maintained for both threads and processes.
811  */
812 static void
813 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
814 {
815 	struct proc *p;
816 
817 	p = td->td_proc;
818 	PROC_LOCK_ASSERT(p, MA_OWNED);
819 
820 	thread_lock(td);
821 	if (td->td_wmesg != NULL)
822 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
823 	else
824 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
825 	if (td->td_name[0] != '\0')
826 		strlcpy(kp->ki_ocomm, td->td_name, sizeof(kp->ki_ocomm));
827 	if (TD_ON_LOCK(td)) {
828 		kp->ki_kiflag |= KI_LOCKBLOCK;
829 		strlcpy(kp->ki_lockname, td->td_lockname,
830 		    sizeof(kp->ki_lockname));
831 	} else {
832 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
833 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
834 	}
835 
836 	if (p->p_state == PRS_NORMAL) { /* approximate. */
837 		if (TD_ON_RUNQ(td) ||
838 		    TD_CAN_RUN(td) ||
839 		    TD_IS_RUNNING(td)) {
840 			kp->ki_stat = SRUN;
841 		} else if (P_SHOULDSTOP(p)) {
842 			kp->ki_stat = SSTOP;
843 		} else if (TD_IS_SLEEPING(td)) {
844 			kp->ki_stat = SSLEEP;
845 		} else if (TD_ON_LOCK(td)) {
846 			kp->ki_stat = SLOCK;
847 		} else {
848 			kp->ki_stat = SWAIT;
849 		}
850 	} else if (p->p_state == PRS_ZOMBIE) {
851 		kp->ki_stat = SZOMB;
852 	} else {
853 		kp->ki_stat = SIDL;
854 	}
855 
856 	/* Things in the thread */
857 	kp->ki_wchan = td->td_wchan;
858 	kp->ki_pri.pri_level = td->td_priority;
859 	kp->ki_pri.pri_native = td->td_base_pri;
860 	kp->ki_lastcpu = td->td_lastcpu;
861 	kp->ki_oncpu = td->td_oncpu;
862 	kp->ki_tdflags = td->td_flags;
863 	kp->ki_tid = td->td_tid;
864 	kp->ki_numthreads = p->p_numthreads;
865 	kp->ki_pcb = td->td_pcb;
866 	kp->ki_kstack = (void *)td->td_kstack;
867 	kp->ki_pctcpu = sched_pctcpu(td);
868 	kp->ki_estcpu = td->td_estcpu;
869 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
870 	kp->ki_pri.pri_class = td->td_pri_class;
871 	kp->ki_pri.pri_user = td->td_user_pri;
872 
873 	if (preferthread)
874 		kp->ki_runtime = cputick2usec(td->td_runtime);
875 
876 	/* We can't get this anymore but ps etc never used it anyway. */
877 	kp->ki_rqindex = 0;
878 
879 	SIGSETOR(kp->ki_siglist, td->td_siglist);
880 	kp->ki_sigmask = td->td_sigmask;
881 	thread_unlock(td);
882 }
883 
884 /*
885  * Fill in a kinfo_proc structure for the specified process.
886  * Must be called with the target process locked.
887  */
888 void
889 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
890 {
891 
892 	fill_kinfo_proc_only(p, kp);
893 	if (FIRST_THREAD_IN_PROC(p) != NULL)
894 		fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
895 }
896 
897 struct pstats *
898 pstats_alloc(void)
899 {
900 
901 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
902 }
903 
904 /*
905  * Copy parts of p_stats; zero the rest of p_stats (statistics).
906  */
907 void
908 pstats_fork(struct pstats *src, struct pstats *dst)
909 {
910 
911 	bzero(&dst->pstat_startzero,
912 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
913 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
914 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
915 }
916 
917 void
918 pstats_free(struct pstats *ps)
919 {
920 
921 	free(ps, M_SUBPROC);
922 }
923 
924 /*
925  * Locate a zombie process by number
926  */
927 struct proc *
928 zpfind(pid_t pid)
929 {
930 	struct proc *p;
931 
932 	sx_slock(&allproc_lock);
933 	LIST_FOREACH(p, &zombproc, p_list)
934 		if (p->p_pid == pid) {
935 			PROC_LOCK(p);
936 			break;
937 		}
938 	sx_sunlock(&allproc_lock);
939 	return (p);
940 }
941 
942 #define KERN_PROC_ZOMBMASK	0x3
943 #define KERN_PROC_NOTHREADS	0x4
944 
945 /*
946  * Must be called with the process locked and will return with it unlocked.
947  */
948 static int
949 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
950 {
951 	struct thread *td;
952 	struct kinfo_proc kinfo_proc;
953 	int error = 0;
954 	struct proc *np;
955 	pid_t pid = p->p_pid;
956 
957 	PROC_LOCK_ASSERT(p, MA_OWNED);
958 
959 	fill_kinfo_proc_only(p, &kinfo_proc);
960 	if (flags & KERN_PROC_NOTHREADS) {
961 		if (FIRST_THREAD_IN_PROC(p) != NULL)
962 			fill_kinfo_thread(FIRST_THREAD_IN_PROC(p),
963 			    &kinfo_proc, 0);
964 		error = SYSCTL_OUT(req, (caddr_t)&kinfo_proc,
965 				   sizeof(kinfo_proc));
966 	} else {
967 		if (FIRST_THREAD_IN_PROC(p) != NULL)
968 			FOREACH_THREAD_IN_PROC(p, td) {
969 				fill_kinfo_thread(td, &kinfo_proc, 1);
970 				error = SYSCTL_OUT(req, (caddr_t)&kinfo_proc,
971 						   sizeof(kinfo_proc));
972 				if (error)
973 					break;
974 			}
975 		else
976 			error = SYSCTL_OUT(req, (caddr_t)&kinfo_proc,
977 					   sizeof(kinfo_proc));
978 	}
979 	PROC_UNLOCK(p);
980 	if (error)
981 		return (error);
982 	if (flags & KERN_PROC_ZOMBMASK)
983 		np = zpfind(pid);
984 	else {
985 		if (pid == 0)
986 			return (0);
987 		np = pfind(pid);
988 	}
989 	if (np == NULL)
990 		return (ESRCH);
991 	if (np != p) {
992 		PROC_UNLOCK(np);
993 		return (ESRCH);
994 	}
995 	PROC_UNLOCK(np);
996 	return (0);
997 }
998 
999 static int
1000 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1001 {
1002 	int *name = (int*) arg1;
1003 	u_int namelen = arg2;
1004 	struct proc *p;
1005 	int flags, doingzomb, oid_number;
1006 	int error = 0;
1007 
1008 	oid_number = oidp->oid_number;
1009 	if (oid_number != KERN_PROC_ALL &&
1010 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1011 		flags = KERN_PROC_NOTHREADS;
1012 	else {
1013 		flags = 0;
1014 		oid_number &= ~KERN_PROC_INC_THREAD;
1015 	}
1016 	if (oid_number == KERN_PROC_PID) {
1017 		if (namelen != 1)
1018 			return (EINVAL);
1019 		error = sysctl_wire_old_buffer(req, 0);
1020 		if (error)
1021 			return (error);
1022 		p = pfind((pid_t)name[0]);
1023 		if (!p)
1024 			return (ESRCH);
1025 		if ((error = p_cansee(curthread, p))) {
1026 			PROC_UNLOCK(p);
1027 			return (error);
1028 		}
1029 		error = sysctl_out_proc(p, req, flags);
1030 		return (error);
1031 	}
1032 
1033 	switch (oid_number) {
1034 	case KERN_PROC_ALL:
1035 		if (namelen != 0)
1036 			return (EINVAL);
1037 		break;
1038 	case KERN_PROC_PROC:
1039 		if (namelen != 0 && namelen != 1)
1040 			return (EINVAL);
1041 		break;
1042 	default:
1043 		if (namelen != 1)
1044 			return (EINVAL);
1045 		break;
1046 	}
1047 
1048 	if (!req->oldptr) {
1049 		/* overestimate by 5 procs */
1050 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1051 		if (error)
1052 			return (error);
1053 	}
1054 	error = sysctl_wire_old_buffer(req, 0);
1055 	if (error != 0)
1056 		return (error);
1057 	sx_slock(&allproc_lock);
1058 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1059 		if (!doingzomb)
1060 			p = LIST_FIRST(&allproc);
1061 		else
1062 			p = LIST_FIRST(&zombproc);
1063 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
1064 			/*
1065 			 * Skip embryonic processes.
1066 			 */
1067 			PROC_SLOCK(p);
1068 			if (p->p_state == PRS_NEW) {
1069 				PROC_SUNLOCK(p);
1070 				continue;
1071 			}
1072 			PROC_SUNLOCK(p);
1073 			PROC_LOCK(p);
1074 			KASSERT(p->p_ucred != NULL,
1075 			    ("process credential is NULL for non-NEW proc"));
1076 			/*
1077 			 * Show a user only appropriate processes.
1078 			 */
1079 			if (p_cansee(curthread, p)) {
1080 				PROC_UNLOCK(p);
1081 				continue;
1082 			}
1083 			/*
1084 			 * TODO - make more efficient (see notes below).
1085 			 * do by session.
1086 			 */
1087 			switch (oid_number) {
1088 
1089 			case KERN_PROC_GID:
1090 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1091 					PROC_UNLOCK(p);
1092 					continue;
1093 				}
1094 				break;
1095 
1096 			case KERN_PROC_PGRP:
1097 				/* could do this by traversing pgrp */
1098 				if (p->p_pgrp == NULL ||
1099 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1100 					PROC_UNLOCK(p);
1101 					continue;
1102 				}
1103 				break;
1104 
1105 			case KERN_PROC_RGID:
1106 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1107 					PROC_UNLOCK(p);
1108 					continue;
1109 				}
1110 				break;
1111 
1112 			case KERN_PROC_SESSION:
1113 				if (p->p_session == NULL ||
1114 				    p->p_session->s_sid != (pid_t)name[0]) {
1115 					PROC_UNLOCK(p);
1116 					continue;
1117 				}
1118 				break;
1119 
1120 			case KERN_PROC_TTY:
1121 				if ((p->p_flag & P_CONTROLT) == 0 ||
1122 				    p->p_session == NULL) {
1123 					PROC_UNLOCK(p);
1124 					continue;
1125 				}
1126 				/* XXX proctree_lock */
1127 				SESS_LOCK(p->p_session);
1128 				if (p->p_session->s_ttyp == NULL ||
1129 				    tty_udev(p->p_session->s_ttyp) !=
1130 				    (dev_t)name[0]) {
1131 					SESS_UNLOCK(p->p_session);
1132 					PROC_UNLOCK(p);
1133 					continue;
1134 				}
1135 				SESS_UNLOCK(p->p_session);
1136 				break;
1137 
1138 			case KERN_PROC_UID:
1139 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1140 					PROC_UNLOCK(p);
1141 					continue;
1142 				}
1143 				break;
1144 
1145 			case KERN_PROC_RUID:
1146 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1147 					PROC_UNLOCK(p);
1148 					continue;
1149 				}
1150 				break;
1151 
1152 			case KERN_PROC_PROC:
1153 				break;
1154 
1155 			default:
1156 				break;
1157 
1158 			}
1159 
1160 			error = sysctl_out_proc(p, req, flags | doingzomb);
1161 			if (error) {
1162 				sx_sunlock(&allproc_lock);
1163 				return (error);
1164 			}
1165 		}
1166 	}
1167 	sx_sunlock(&allproc_lock);
1168 	return (0);
1169 }
1170 
1171 struct pargs *
1172 pargs_alloc(int len)
1173 {
1174 	struct pargs *pa;
1175 
1176 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1177 		M_WAITOK);
1178 	refcount_init(&pa->ar_ref, 1);
1179 	pa->ar_length = len;
1180 	return (pa);
1181 }
1182 
1183 static void
1184 pargs_free(struct pargs *pa)
1185 {
1186 
1187 	free(pa, M_PARGS);
1188 }
1189 
1190 void
1191 pargs_hold(struct pargs *pa)
1192 {
1193 
1194 	if (pa == NULL)
1195 		return;
1196 	refcount_acquire(&pa->ar_ref);
1197 }
1198 
1199 void
1200 pargs_drop(struct pargs *pa)
1201 {
1202 
1203 	if (pa == NULL)
1204 		return;
1205 	if (refcount_release(&pa->ar_ref))
1206 		pargs_free(pa);
1207 }
1208 
1209 /*
1210  * This sysctl allows a process to retrieve the argument list or process
1211  * title for another process without groping around in the address space
1212  * of the other process.  It also allow a process to set its own "process
1213  * title to a string of its own choice.
1214  */
1215 static int
1216 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1217 {
1218 	int *name = (int*) arg1;
1219 	u_int namelen = arg2;
1220 	struct pargs *newpa, *pa;
1221 	struct proc *p;
1222 	int error = 0;
1223 
1224 	if (namelen != 1)
1225 		return (EINVAL);
1226 
1227 	p = pfind((pid_t)name[0]);
1228 	if (!p)
1229 		return (ESRCH);
1230 
1231 	if ((error = p_cansee(curthread, p)) != 0) {
1232 		PROC_UNLOCK(p);
1233 		return (error);
1234 	}
1235 
1236 	if (req->newptr && curproc != p) {
1237 		PROC_UNLOCK(p);
1238 		return (EPERM);
1239 	}
1240 
1241 	pa = p->p_args;
1242 	pargs_hold(pa);
1243 	PROC_UNLOCK(p);
1244 	if (req->oldptr != NULL && pa != NULL)
1245 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1246 	pargs_drop(pa);
1247 	if (error != 0 || req->newptr == NULL)
1248 		return (error);
1249 
1250 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1251 		return (ENOMEM);
1252 	newpa = pargs_alloc(req->newlen);
1253 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1254 	if (error != 0) {
1255 		pargs_free(newpa);
1256 		return (error);
1257 	}
1258 	PROC_LOCK(p);
1259 	pa = p->p_args;
1260 	p->p_args = newpa;
1261 	PROC_UNLOCK(p);
1262 	pargs_drop(pa);
1263 	return (0);
1264 }
1265 
1266 /*
1267  * This sysctl allows a process to retrieve the path of the executable for
1268  * itself or another process.
1269  */
1270 static int
1271 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1272 {
1273 	pid_t *pidp = (pid_t *)arg1;
1274 	unsigned int arglen = arg2;
1275 	struct proc *p;
1276 	struct vnode *vp;
1277 	char *retbuf, *freebuf;
1278 	int error;
1279 
1280 	if (arglen != 1)
1281 		return (EINVAL);
1282 	if (*pidp == -1) {	/* -1 means this process */
1283 		p = req->td->td_proc;
1284 	} else {
1285 		p = pfind(*pidp);
1286 		if (p == NULL)
1287 			return (ESRCH);
1288 		if ((error = p_cansee(curthread, p)) != 0) {
1289 			PROC_UNLOCK(p);
1290 			return (error);
1291 		}
1292 	}
1293 
1294 	vp = p->p_textvp;
1295 	if (vp == NULL) {
1296 		if (*pidp != -1)
1297 			PROC_UNLOCK(p);
1298 		return (0);
1299 	}
1300 	vref(vp);
1301 	if (*pidp != -1)
1302 		PROC_UNLOCK(p);
1303 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1304 	vrele(vp);
1305 	if (error)
1306 		return (error);
1307 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1308 	free(freebuf, M_TEMP);
1309 	return (error);
1310 }
1311 
1312 static int
1313 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1314 {
1315 	struct proc *p;
1316 	char *sv_name;
1317 	int *name;
1318 	int namelen;
1319 	int error;
1320 
1321 	namelen = arg2;
1322 	if (namelen != 1)
1323 		return (EINVAL);
1324 
1325 	name = (int *)arg1;
1326 	if ((p = pfind((pid_t)name[0])) == NULL)
1327 		return (ESRCH);
1328 	if ((error = p_cansee(curthread, p))) {
1329 		PROC_UNLOCK(p);
1330 		return (error);
1331 	}
1332 	sv_name = p->p_sysent->sv_name;
1333 	PROC_UNLOCK(p);
1334 	return (sysctl_handle_string(oidp, sv_name, 0, req));
1335 }
1336 
1337 static int
1338 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
1339 {
1340 	vm_map_entry_t entry, tmp_entry;
1341 	unsigned int last_timestamp;
1342 	char *fullpath, *freepath;
1343 	struct kinfo_vmentry *kve;
1344 	struct vattr va;
1345 	struct ucred *cred;
1346 	int error, *name;
1347 	struct vnode *vp;
1348 	struct proc *p;
1349 	vm_map_t map;
1350 
1351 	name = (int *)arg1;
1352 	if ((p = pfind((pid_t)name[0])) == NULL)
1353 		return (ESRCH);
1354 	if (p->p_flag & P_WEXIT) {
1355 		PROC_UNLOCK(p);
1356 		return (ESRCH);
1357 	}
1358 	if ((error = p_candebug(curthread, p))) {
1359 		PROC_UNLOCK(p);
1360 		return (error);
1361 	}
1362 	_PHOLD(p);
1363 	PROC_UNLOCK(p);
1364 
1365 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1366 
1367 	map = &p->p_vmspace->vm_map;	/* XXXRW: More locking required? */
1368 	vm_map_lock_read(map);
1369 	for (entry = map->header.next; entry != &map->header;
1370 	    entry = entry->next) {
1371 		vm_object_t obj, tobj, lobj;
1372 		vm_offset_t addr;
1373 		int vfslocked;
1374 
1375 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
1376 			continue;
1377 
1378 		bzero(kve, sizeof(*kve));
1379 		kve->kve_structsize = sizeof(*kve);
1380 
1381 		kve->kve_private_resident = 0;
1382 		obj = entry->object.vm_object;
1383 		if (obj != NULL) {
1384 			VM_OBJECT_LOCK(obj);
1385 			if (obj->shadow_count == 1)
1386 				kve->kve_private_resident =
1387 				    obj->resident_page_count;
1388 		}
1389 		kve->kve_resident = 0;
1390 		addr = entry->start;
1391 		while (addr < entry->end) {
1392 			if (pmap_extract(map->pmap, addr))
1393 				kve->kve_resident++;
1394 			addr += PAGE_SIZE;
1395 		}
1396 
1397 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
1398 			if (tobj != obj)
1399 				VM_OBJECT_LOCK(tobj);
1400 			if (lobj != obj)
1401 				VM_OBJECT_UNLOCK(lobj);
1402 			lobj = tobj;
1403 		}
1404 
1405 		kve->kve_fileid = 0;
1406 		kve->kve_fsid = 0;
1407 		freepath = NULL;
1408 		fullpath = "";
1409 		if (lobj) {
1410 			vp = NULL;
1411 			switch(lobj->type) {
1412 			case OBJT_DEFAULT:
1413 				kve->kve_type = KVME_TYPE_DEFAULT;
1414 				break;
1415 			case OBJT_VNODE:
1416 				kve->kve_type = KVME_TYPE_VNODE;
1417 				vp = lobj->handle;
1418 				vref(vp);
1419 				break;
1420 			case OBJT_SWAP:
1421 				kve->kve_type = KVME_TYPE_SWAP;
1422 				break;
1423 			case OBJT_DEVICE:
1424 				kve->kve_type = KVME_TYPE_DEVICE;
1425 				break;
1426 			case OBJT_PHYS:
1427 				kve->kve_type = KVME_TYPE_PHYS;
1428 				break;
1429 			case OBJT_DEAD:
1430 				kve->kve_type = KVME_TYPE_DEAD;
1431 				break;
1432 			default:
1433 				kve->kve_type = KVME_TYPE_UNKNOWN;
1434 				break;
1435 			}
1436 			if (lobj != obj)
1437 				VM_OBJECT_UNLOCK(lobj);
1438 
1439 			kve->kve_ref_count = obj->ref_count;
1440 			kve->kve_shadow_count = obj->shadow_count;
1441 			VM_OBJECT_UNLOCK(obj);
1442 			if (vp != NULL) {
1443 				vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1444 				vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1445 				vn_fullpath(curthread, vp, &fullpath,
1446 				    &freepath);
1447 				cred = curthread->td_ucred;
1448 				if (VOP_GETATTR(vp, &va, cred) == 0) {
1449 					kve->kve_fileid = va.va_fileid;
1450 					kve->kve_fsid = va.va_fsid;
1451 				}
1452 				vput(vp);
1453 				VFS_UNLOCK_GIANT(vfslocked);
1454 			}
1455 		} else {
1456 			kve->kve_type = KVME_TYPE_NONE;
1457 			kve->kve_ref_count = 0;
1458 			kve->kve_shadow_count = 0;
1459 		}
1460 
1461 		kve->kve_start = (void*)entry->start;
1462 		kve->kve_end = (void*)entry->end;
1463 		kve->kve_offset = (off_t)entry->offset;
1464 
1465 		if (entry->protection & VM_PROT_READ)
1466 			kve->kve_protection |= KVME_PROT_READ;
1467 		if (entry->protection & VM_PROT_WRITE)
1468 			kve->kve_protection |= KVME_PROT_WRITE;
1469 		if (entry->protection & VM_PROT_EXECUTE)
1470 			kve->kve_protection |= KVME_PROT_EXEC;
1471 
1472 		if (entry->eflags & MAP_ENTRY_COW)
1473 			kve->kve_flags |= KVME_FLAG_COW;
1474 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
1475 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
1476 
1477 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
1478 		if (freepath != NULL)
1479 			free(freepath, M_TEMP);
1480 
1481 		last_timestamp = map->timestamp;
1482 		vm_map_unlock_read(map);
1483 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
1484 		vm_map_lock_read(map);
1485 		if (error)
1486 			break;
1487 		if (last_timestamp + 1 != map->timestamp) {
1488 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
1489 			entry = tmp_entry;
1490 		}
1491 	}
1492 	vm_map_unlock_read(map);
1493 	PRELE(p);
1494 	free(kve, M_TEMP);
1495 	return (error);
1496 }
1497 
1498 #if defined(STACK) || defined(DDB)
1499 static int
1500 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
1501 {
1502 	struct kinfo_kstack *kkstp;
1503 	int error, i, *name, numthreads;
1504 	lwpid_t *lwpidarray;
1505 	struct thread *td;
1506 	struct stack *st;
1507 	struct sbuf sb;
1508 	struct proc *p;
1509 
1510 	name = (int *)arg1;
1511 	if ((p = pfind((pid_t)name[0])) == NULL)
1512 		return (ESRCH);
1513 	/* XXXRW: Not clear ESRCH is the right error during proc execve(). */
1514 	if (p->p_flag & P_WEXIT || p->p_flag & P_INEXEC) {
1515 		PROC_UNLOCK(p);
1516 		return (ESRCH);
1517 	}
1518 	if ((error = p_candebug(curthread, p))) {
1519 		PROC_UNLOCK(p);
1520 		return (error);
1521 	}
1522 	_PHOLD(p);
1523 	PROC_UNLOCK(p);
1524 
1525 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
1526 	st = stack_create();
1527 
1528 	lwpidarray = NULL;
1529 	numthreads = 0;
1530 	PROC_LOCK(p);
1531 repeat:
1532 	if (numthreads < p->p_numthreads) {
1533 		if (lwpidarray != NULL) {
1534 			free(lwpidarray, M_TEMP);
1535 			lwpidarray = NULL;
1536 		}
1537 		numthreads = p->p_numthreads;
1538 		PROC_UNLOCK(p);
1539 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
1540 		    M_WAITOK | M_ZERO);
1541 		PROC_LOCK(p);
1542 		goto repeat;
1543 	}
1544 	i = 0;
1545 
1546 	/*
1547 	 * XXXRW: During the below loop, execve(2) and countless other sorts
1548 	 * of changes could have taken place.  Should we check to see if the
1549 	 * vmspace has been replaced, or the like, in order to prevent
1550 	 * giving a snapshot that spans, say, execve(2), with some threads
1551 	 * before and some after?  Among other things, the credentials could
1552 	 * have changed, in which case the right to extract debug info might
1553 	 * no longer be assured.
1554 	 */
1555 	FOREACH_THREAD_IN_PROC(p, td) {
1556 		KASSERT(i < numthreads,
1557 		    ("sysctl_kern_proc_kstack: numthreads"));
1558 		lwpidarray[i] = td->td_tid;
1559 		i++;
1560 	}
1561 	numthreads = i;
1562 	for (i = 0; i < numthreads; i++) {
1563 		td = thread_find(p, lwpidarray[i]);
1564 		if (td == NULL) {
1565 			continue;
1566 		}
1567 		bzero(kkstp, sizeof(*kkstp));
1568 		(void)sbuf_new(&sb, kkstp->kkst_trace,
1569 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
1570 		thread_lock(td);
1571 		kkstp->kkst_tid = td->td_tid;
1572 		if (TD_IS_SWAPPED(td))
1573 			kkstp->kkst_state = KKST_STATE_SWAPPED;
1574 		else if (TD_IS_RUNNING(td))
1575 			kkstp->kkst_state = KKST_STATE_RUNNING;
1576 		else {
1577 			kkstp->kkst_state = KKST_STATE_STACKOK;
1578 			stack_save_td(st, td);
1579 		}
1580 		thread_unlock(td);
1581 		PROC_UNLOCK(p);
1582 		stack_sbuf_print(&sb, st);
1583 		sbuf_finish(&sb);
1584 		sbuf_delete(&sb);
1585 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
1586 		PROC_LOCK(p);
1587 		if (error)
1588 			break;
1589 	}
1590 	_PRELE(p);
1591 	PROC_UNLOCK(p);
1592 	if (lwpidarray != NULL)
1593 		free(lwpidarray, M_TEMP);
1594 	stack_destroy(st);
1595 	free(kkstp, M_TEMP);
1596 	return (error);
1597 }
1598 #endif
1599 
1600 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
1601 
1602 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT,
1603 	0, 0, sysctl_kern_proc, "S,proc", "Return entire process table");
1604 
1605 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD,
1606 	sysctl_kern_proc, "Process table");
1607 
1608 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD,
1609 	sysctl_kern_proc, "Process table");
1610 
1611 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD,
1612 	sysctl_kern_proc, "Process table");
1613 
1614 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD,
1615 	sysctl_kern_proc, "Process table");
1616 
1617 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD,
1618 	sysctl_kern_proc, "Process table");
1619 
1620 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD,
1621 	sysctl_kern_proc, "Process table");
1622 
1623 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD,
1624 	sysctl_kern_proc, "Process table");
1625 
1626 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD,
1627 	sysctl_kern_proc, "Process table");
1628 
1629 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD,
1630 	sysctl_kern_proc, "Return process table, no threads");
1631 
1632 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
1633 	CTLFLAG_RW | CTLFLAG_ANYBODY,
1634 	sysctl_kern_proc_args, "Process argument list");
1635 
1636 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD,
1637 	sysctl_kern_proc_pathname, "Process executable path");
1638 
1639 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD,
1640 	sysctl_kern_proc_sv_name, "Process syscall vector name (ABI type)");
1641 
1642 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
1643 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1644 
1645 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
1646 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1647 
1648 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
1649 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1650 
1651 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
1652 	sid_td, CTLFLAG_RD, sysctl_kern_proc, "Process table");
1653 
1654 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
1655 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1656 
1657 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
1658 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1659 
1660 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
1661 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1662 
1663 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
1664 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1665 
1666 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
1667 	CTLFLAG_RD, sysctl_kern_proc, "Return process table, no threads");
1668 
1669 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD,
1670 	sysctl_kern_proc_vmmap, "Process vm map entries");
1671 
1672 #if defined(STACK) || defined(DDB)
1673 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD,
1674 	sysctl_kern_proc_kstack, "Process kernel stacks");
1675 #endif
1676