xref: /freebsd/sys/kern/kern_proc.c (revision 721351876cd4d3a8a700f62d2061331fa951a488)
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 
127 /*
128  * Other process lists
129  */
130 struct pidhashhead *pidhashtbl;
131 u_long pidhash;
132 struct pgrphashhead *pgrphashtbl;
133 u_long pgrphash;
134 struct proclist allproc;
135 struct proclist zombproc;
136 struct sx allproc_lock;
137 struct sx proctree_lock;
138 struct mtx ppeers_lock;
139 uma_zone_t proc_zone;
140 uma_zone_t ithread_zone;
141 
142 int kstack_pages = KSTACK_PAGES;
143 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0, "");
144 
145 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
146 
147 /*
148  * Initialize global process hashing structures.
149  */
150 void
151 procinit()
152 {
153 
154 	sx_init(&allproc_lock, "allproc");
155 	sx_init(&proctree_lock, "proctree");
156 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
157 	LIST_INIT(&allproc);
158 	LIST_INIT(&zombproc);
159 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
160 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
161 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
162 	    proc_ctor, proc_dtor, proc_init, proc_fini,
163 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
164 	uihashinit();
165 }
166 
167 /*
168  * Prepare a proc for use.
169  */
170 static int
171 proc_ctor(void *mem, int size, void *arg, int flags)
172 {
173 	struct proc *p;
174 
175 	p = (struct proc *)mem;
176 	SDT_PROBE(proc, kernel, ctor , entry, p, size, arg, flags, 0);
177 	EVENTHANDLER_INVOKE(process_ctor, p);
178 	SDT_PROBE(proc, kernel, ctor , return, p, size, arg, flags, 0);
179 	return (0);
180 }
181 
182 /*
183  * Reclaim a proc after use.
184  */
185 static void
186 proc_dtor(void *mem, int size, void *arg)
187 {
188 	struct proc *p;
189 	struct thread *td;
190 
191 	/* INVARIANTS checks go here */
192 	p = (struct proc *)mem;
193 	td = FIRST_THREAD_IN_PROC(p);
194 	SDT_PROBE(proc, kernel, dtor, entry, p, size, arg, td, 0);
195 	if (td != NULL) {
196 #ifdef INVARIANTS
197 		KASSERT((p->p_numthreads == 1),
198 		    ("bad number of threads in exiting process"));
199 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
200 #endif
201 		/* Dispose of an alternate kstack, if it exists.
202 		 * XXX What if there are more than one thread in the proc?
203 		 *     The first thread in the proc is special and not
204 		 *     freed, so you gotta do this here.
205 		 */
206 		if (((p->p_flag & P_KTHREAD) != 0) && (td->td_altkstack != 0))
207 			vm_thread_dispose_altkstack(td);
208 	}
209 	EVENTHANDLER_INVOKE(process_dtor, p);
210 	if (p->p_ksi != NULL)
211 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
212 	SDT_PROBE(proc, kernel, dtor, return, p, size, arg, 0, 0);
213 }
214 
215 /*
216  * Initialize type-stable parts of a proc (when newly created).
217  */
218 static int
219 proc_init(void *mem, int size, int flags)
220 {
221 	struct proc *p;
222 
223 	p = (struct proc *)mem;
224 	SDT_PROBE(proc, kernel, init, entry, p, size, flags, 0, 0);
225 	p->p_sched = (struct p_sched *)&p[1];
226 	bzero(&p->p_mtx, sizeof(struct mtx));
227 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK);
228 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_RECURSE);
229 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
230 	EVENTHANDLER_INVOKE(process_init, p);
231 	p->p_stats = pstats_alloc();
232 	SDT_PROBE(proc, kernel, init, return, p, size, flags, 0, 0);
233 	return (0);
234 }
235 
236 /*
237  * UMA should ensure that this function is never called.
238  * Freeing a proc structure would violate type stability.
239  */
240 static void
241 proc_fini(void *mem, int size)
242 {
243 #ifdef notnow
244 	struct proc *p;
245 
246 	p = (struct proc *)mem;
247 	EVENTHANDLER_INVOKE(process_fini, p);
248 	pstats_free(p->p_stats);
249 	thread_free(FIRST_THREAD_IN_PROC(p));
250 	mtx_destroy(&p->p_mtx);
251 	if (p->p_ksi != NULL)
252 		ksiginfo_free(p->p_ksi);
253 #else
254 	panic("proc reclaimed");
255 #endif
256 }
257 
258 /*
259  * Is p an inferior of the current process?
260  */
261 int
262 inferior(p)
263 	register struct proc *p;
264 {
265 
266 	sx_assert(&proctree_lock, SX_LOCKED);
267 	for (; p != curproc; p = p->p_pptr)
268 		if (p->p_pid == 0)
269 			return (0);
270 	return (1);
271 }
272 
273 /*
274  * Locate a process by number; return only "live" processes -- i.e., neither
275  * zombies nor newly born but incompletely initialized processes.  By not
276  * returning processes in the PRS_NEW state, we allow callers to avoid
277  * testing for that condition to avoid dereferencing p_ucred, et al.
278  */
279 struct proc *
280 pfind(pid)
281 	register pid_t pid;
282 {
283 	register struct proc *p;
284 
285 	sx_slock(&allproc_lock);
286 	LIST_FOREACH(p, PIDHASH(pid), p_hash)
287 		if (p->p_pid == pid) {
288 			if (p->p_state == PRS_NEW) {
289 				p = NULL;
290 				break;
291 			}
292 			PROC_LOCK(p);
293 			break;
294 		}
295 	sx_sunlock(&allproc_lock);
296 	return (p);
297 }
298 
299 /*
300  * Locate a process group by number.
301  * The caller must hold proctree_lock.
302  */
303 struct pgrp *
304 pgfind(pgid)
305 	register pid_t pgid;
306 {
307 	register struct pgrp *pgrp;
308 
309 	sx_assert(&proctree_lock, SX_LOCKED);
310 
311 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
312 		if (pgrp->pg_id == pgid) {
313 			PGRP_LOCK(pgrp);
314 			return (pgrp);
315 		}
316 	}
317 	return (NULL);
318 }
319 
320 /*
321  * Create a new process group.
322  * pgid must be equal to the pid of p.
323  * Begin a new session if required.
324  */
325 int
326 enterpgrp(p, pgid, pgrp, sess)
327 	register struct proc *p;
328 	pid_t pgid;
329 	struct pgrp *pgrp;
330 	struct session *sess;
331 {
332 	struct pgrp *pgrp2;
333 
334 	sx_assert(&proctree_lock, SX_XLOCKED);
335 
336 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
337 	KASSERT(p->p_pid == pgid,
338 	    ("enterpgrp: new pgrp and pid != pgid"));
339 
340 	pgrp2 = pgfind(pgid);
341 
342 	KASSERT(pgrp2 == NULL,
343 	    ("enterpgrp: pgrp with pgid exists"));
344 	KASSERT(!SESS_LEADER(p),
345 	    ("enterpgrp: session leader attempted setpgrp"));
346 
347 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
348 
349 	if (sess != NULL) {
350 		/*
351 		 * new session
352 		 */
353 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
354 		mtx_lock(&Giant);       /* XXX TTY */
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 		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 		mtx_lock(&Giant);       /* XXX TTY */
371 		pgrp->pg_session = p->p_session;
372 		SESS_LOCK(pgrp->pg_session);
373 		pgrp->pg_session->s_count++;
374 		SESS_UNLOCK(pgrp->pg_session);
375 		PGRP_LOCK(pgrp);
376 	}
377 	pgrp->pg_id = pgid;
378 	LIST_INIT(&pgrp->pg_members);
379 
380 	/*
381 	 * As we have an exclusive lock of proctree_lock,
382 	 * this should not deadlock.
383 	 */
384 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
385 	pgrp->pg_jobc = 0;
386 	SLIST_INIT(&pgrp->pg_sigiolst);
387 	PGRP_UNLOCK(pgrp);
388 	mtx_unlock(&Giant);       /* XXX TTY */
389 
390 	doenterpgrp(p, pgrp);
391 
392 	return (0);
393 }
394 
395 /*
396  * Move p to an existing process group
397  */
398 int
399 enterthispgrp(p, pgrp)
400 	register struct proc *p;
401 	struct pgrp *pgrp;
402 {
403 
404 	sx_assert(&proctree_lock, SX_XLOCKED);
405 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
406 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
407 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
408 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
409 	KASSERT(pgrp->pg_session == p->p_session,
410 		("%s: pgrp's session %p, p->p_session %p.\n",
411 		__func__,
412 		pgrp->pg_session,
413 		p->p_session));
414 	KASSERT(pgrp != p->p_pgrp,
415 		("%s: p belongs to pgrp.", __func__));
416 
417 	doenterpgrp(p, pgrp);
418 
419 	return (0);
420 }
421 
422 /*
423  * Move p to a process group
424  */
425 static void
426 doenterpgrp(p, pgrp)
427 	struct proc *p;
428 	struct pgrp *pgrp;
429 {
430 	struct pgrp *savepgrp;
431 
432 	sx_assert(&proctree_lock, SX_XLOCKED);
433 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
434 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
435 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
436 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
437 
438 	savepgrp = p->p_pgrp;
439 
440 	/*
441 	 * Adjust eligibility of affected pgrps to participate in job control.
442 	 * Increment eligibility counts before decrementing, otherwise we
443 	 * could reach 0 spuriously during the first call.
444 	 */
445 	fixjobc(p, pgrp, 1);
446 	fixjobc(p, p->p_pgrp, 0);
447 
448 	mtx_lock(&Giant);       /* XXX TTY */
449 	PGRP_LOCK(pgrp);
450 	PGRP_LOCK(savepgrp);
451 	PROC_LOCK(p);
452 	LIST_REMOVE(p, p_pglist);
453 	p->p_pgrp = pgrp;
454 	PROC_UNLOCK(p);
455 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
456 	PGRP_UNLOCK(savepgrp);
457 	PGRP_UNLOCK(pgrp);
458 	mtx_unlock(&Giant);     /* XXX TTY */
459 	if (LIST_EMPTY(&savepgrp->pg_members))
460 		pgdelete(savepgrp);
461 }
462 
463 /*
464  * remove process from process group
465  */
466 int
467 leavepgrp(p)
468 	register struct proc *p;
469 {
470 	struct pgrp *savepgrp;
471 
472 	sx_assert(&proctree_lock, SX_XLOCKED);
473 	savepgrp = p->p_pgrp;
474 	mtx_lock(&Giant);	/* XXX TTY */
475 	PGRP_LOCK(savepgrp);
476 	PROC_LOCK(p);
477 	LIST_REMOVE(p, p_pglist);
478 	p->p_pgrp = NULL;
479 	PROC_UNLOCK(p);
480 	PGRP_UNLOCK(savepgrp);
481 	mtx_unlock(&Giant);	/* XXX TTY */
482 	if (LIST_EMPTY(&savepgrp->pg_members))
483 		pgdelete(savepgrp);
484 	return (0);
485 }
486 
487 /*
488  * delete a process group
489  */
490 static void
491 pgdelete(pgrp)
492 	register struct pgrp *pgrp;
493 {
494 	struct session *savesess;
495 
496 	sx_assert(&proctree_lock, SX_XLOCKED);
497 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
498 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
499 
500 	/*
501 	 * Reset any sigio structures pointing to us as a result of
502 	 * F_SETOWN with our pgid.
503 	 */
504 	funsetownlst(&pgrp->pg_sigiolst);
505 
506 	mtx_lock(&Giant);       /* XXX TTY */
507 	PGRP_LOCK(pgrp);
508 	if (pgrp->pg_session->s_ttyp != NULL &&
509 	    pgrp->pg_session->s_ttyp->t_pgrp == pgrp)
510 		pgrp->pg_session->s_ttyp->t_pgrp = NULL;
511 	LIST_REMOVE(pgrp, pg_hash);
512 	savesess = pgrp->pg_session;
513 	SESSRELE(savesess);
514 	PGRP_UNLOCK(pgrp);
515 	mtx_destroy(&pgrp->pg_mtx);
516 	FREE(pgrp, M_PGRP);
517 	mtx_unlock(&Giant);     /* XXX TTY */
518 }
519 
520 static void
521 pgadjustjobc(pgrp, entering)
522 	struct pgrp *pgrp;
523 	int entering;
524 {
525 
526 	PGRP_LOCK(pgrp);
527 	if (entering)
528 		pgrp->pg_jobc++;
529 	else {
530 		--pgrp->pg_jobc;
531 		if (pgrp->pg_jobc == 0)
532 			orphanpg(pgrp);
533 	}
534 	PGRP_UNLOCK(pgrp);
535 }
536 
537 /*
538  * Adjust pgrp jobc counters when specified process changes process group.
539  * We count the number of processes in each process group that "qualify"
540  * the group for terminal job control (those with a parent in a different
541  * process group of the same session).  If that count reaches zero, the
542  * process group becomes orphaned.  Check both the specified process'
543  * process group and that of its children.
544  * entering == 0 => p is leaving specified group.
545  * entering == 1 => p is entering specified group.
546  */
547 void
548 fixjobc(p, pgrp, entering)
549 	register struct proc *p;
550 	register struct pgrp *pgrp;
551 	int entering;
552 {
553 	register struct pgrp *hispgrp;
554 	register struct session *mysession;
555 
556 	sx_assert(&proctree_lock, SX_LOCKED);
557 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
558 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
559 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
560 
561 	/*
562 	 * Check p's parent to see whether p qualifies its own process
563 	 * group; if so, adjust count for p's process group.
564 	 */
565 	mysession = pgrp->pg_session;
566 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
567 	    hispgrp->pg_session == mysession)
568 		pgadjustjobc(pgrp, entering);
569 
570 	/*
571 	 * Check this process' children to see whether they qualify
572 	 * their process groups; if so, adjust counts for children's
573 	 * process groups.
574 	 */
575 	LIST_FOREACH(p, &p->p_children, p_sibling) {
576 		hispgrp = p->p_pgrp;
577 		if (hispgrp == pgrp ||
578 		    hispgrp->pg_session != mysession)
579 			continue;
580 		PROC_LOCK(p);
581 		if (p->p_state == PRS_ZOMBIE) {
582 			PROC_UNLOCK(p);
583 			continue;
584 		}
585 		PROC_UNLOCK(p);
586 		pgadjustjobc(hispgrp, entering);
587 	}
588 }
589 
590 /*
591  * A process group has become orphaned;
592  * if there are any stopped processes in the group,
593  * hang-up all process in that group.
594  */
595 static void
596 orphanpg(pg)
597 	struct pgrp *pg;
598 {
599 	register struct proc *p;
600 
601 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
602 
603 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
604 		PROC_LOCK(p);
605 		if (P_SHOULDSTOP(p)) {
606 			PROC_UNLOCK(p);
607 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
608 				PROC_LOCK(p);
609 				psignal(p, SIGHUP);
610 				psignal(p, SIGCONT);
611 				PROC_UNLOCK(p);
612 			}
613 			return;
614 		}
615 		PROC_UNLOCK(p);
616 	}
617 }
618 
619 void
620 sessrele(struct session *s)
621 {
622 	int i;
623 
624 	SESS_LOCK(s);
625 	i = --s->s_count;
626 	SESS_UNLOCK(s);
627 	if (i == 0) {
628 		if (s->s_ttyp != NULL)
629 			ttyrel(s->s_ttyp);
630 		mtx_destroy(&s->s_mtx);
631 		FREE(s, M_SESSION);
632 	}
633 }
634 
635 #include "opt_ddb.h"
636 #ifdef DDB
637 #include <ddb/ddb.h>
638 
639 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
640 {
641 	register struct pgrp *pgrp;
642 	register struct proc *p;
643 	register int i;
644 
645 	for (i = 0; i <= pgrphash; i++) {
646 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
647 			printf("\tindx %d\n", i);
648 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
649 				printf(
650 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
651 				    (void *)pgrp, (long)pgrp->pg_id,
652 				    (void *)pgrp->pg_session,
653 				    pgrp->pg_session->s_count,
654 				    (void *)LIST_FIRST(&pgrp->pg_members));
655 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
656 					printf("\t\tpid %ld addr %p pgrp %p\n",
657 					    (long)p->p_pid, (void *)p,
658 					    (void *)p->p_pgrp);
659 				}
660 			}
661 		}
662 	}
663 }
664 #endif /* DDB */
665 
666 /*
667  * Clear kinfo_proc and fill in any information that is common
668  * to all threads in the process.
669  * Must be called with the target process locked.
670  */
671 static void
672 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
673 {
674 	struct thread *td0;
675 	struct tty *tp;
676 	struct session *sp;
677 	struct ucred *cred;
678 	struct sigacts *ps;
679 
680 	PROC_LOCK_ASSERT(p, MA_OWNED);
681 	bzero(kp, sizeof(*kp));
682 
683 	kp->ki_structsize = sizeof(*kp);
684 	kp->ki_paddr = p;
685 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
686 	kp->ki_args = p->p_args;
687 	kp->ki_textvp = p->p_textvp;
688 #ifdef KTRACE
689 	kp->ki_tracep = p->p_tracevp;
690 	mtx_lock(&ktrace_mtx);
691 	kp->ki_traceflag = p->p_traceflag;
692 	mtx_unlock(&ktrace_mtx);
693 #endif
694 	kp->ki_fd = p->p_fd;
695 	kp->ki_vmspace = p->p_vmspace;
696 	kp->ki_flag = p->p_flag;
697 	cred = p->p_ucred;
698 	if (cred) {
699 		kp->ki_uid = cred->cr_uid;
700 		kp->ki_ruid = cred->cr_ruid;
701 		kp->ki_svuid = cred->cr_svuid;
702 		/* XXX bde doesn't like KI_NGROUPS */
703 		kp->ki_ngroups = min(cred->cr_ngroups, KI_NGROUPS);
704 		bcopy(cred->cr_groups, kp->ki_groups,
705 		    kp->ki_ngroups * sizeof(gid_t));
706 		kp->ki_rgid = cred->cr_rgid;
707 		kp->ki_svgid = cred->cr_svgid;
708 		/* If jailed(cred), emulate the old P_JAILED flag. */
709 		if (jailed(cred)) {
710 			kp->ki_flag |= P_JAILED;
711 			/* If inside a jail, use 0 as a jail ID. */
712 			if (!jailed(curthread->td_ucred))
713 				kp->ki_jid = cred->cr_prison->pr_id;
714 		}
715 	}
716 	ps = p->p_sigacts;
717 	if (ps) {
718 		mtx_lock(&ps->ps_mtx);
719 		kp->ki_sigignore = ps->ps_sigignore;
720 		kp->ki_sigcatch = ps->ps_sigcatch;
721 		mtx_unlock(&ps->ps_mtx);
722 	}
723 	PROC_SLOCK(p);
724 	if (p->p_state != PRS_NEW &&
725 	    p->p_state != PRS_ZOMBIE &&
726 	    p->p_vmspace != NULL) {
727 		struct vmspace *vm = p->p_vmspace;
728 
729 		kp->ki_size = vm->vm_map.size;
730 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
731 		FOREACH_THREAD_IN_PROC(p, td0) {
732 			if (!TD_IS_SWAPPED(td0))
733 				kp->ki_rssize += td0->td_kstack_pages;
734 			if (td0->td_altkstack_obj != NULL)
735 				kp->ki_rssize += td0->td_altkstack_pages;
736 		}
737 		kp->ki_swrss = vm->vm_swrss;
738 		kp->ki_tsize = vm->vm_tsize;
739 		kp->ki_dsize = vm->vm_dsize;
740 		kp->ki_ssize = vm->vm_ssize;
741 	} else if (p->p_state == PRS_ZOMBIE)
742 		kp->ki_stat = SZOMB;
743 	if (kp->ki_flag & P_INMEM)
744 		kp->ki_sflag = PS_INMEM;
745 	else
746 		kp->ki_sflag = 0;
747 	/* Calculate legacy swtime as seconds since 'swtick'. */
748 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
749 	kp->ki_pid = p->p_pid;
750 	kp->ki_nice = p->p_nice;
751 	rufetch(p, &kp->ki_rusage);
752 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
753 	PROC_SUNLOCK(p);
754 	if ((p->p_flag & P_INMEM) && p->p_stats != NULL) {
755 		kp->ki_start = p->p_stats->p_start;
756 		timevaladd(&kp->ki_start, &boottime);
757 		PROC_SLOCK(p);
758 		calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
759 		PROC_SUNLOCK(p);
760 		calccru(p, &kp->ki_childutime, &kp->ki_childstime);
761 
762 		/* Some callers want child-times in a single value */
763 		kp->ki_childtime = kp->ki_childstime;
764 		timevaladd(&kp->ki_childtime, &kp->ki_childutime);
765 	}
766 	tp = NULL;
767 	if (p->p_pgrp) {
768 		kp->ki_pgid = p->p_pgrp->pg_id;
769 		kp->ki_jobc = p->p_pgrp->pg_jobc;
770 		sp = p->p_pgrp->pg_session;
771 
772 		if (sp != NULL) {
773 			kp->ki_sid = sp->s_sid;
774 			SESS_LOCK(sp);
775 			strlcpy(kp->ki_login, sp->s_login,
776 			    sizeof(kp->ki_login));
777 			if (sp->s_ttyvp)
778 				kp->ki_kiflag |= KI_CTTY;
779 			if (SESS_LEADER(p))
780 				kp->ki_kiflag |= KI_SLEADER;
781 			tp = sp->s_ttyp;
782 			SESS_UNLOCK(sp);
783 		}
784 	}
785 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
786 		kp->ki_tdev = dev2udev(tp->t_dev);
787 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
788 		if (tp->t_session)
789 			kp->ki_tsid = tp->t_session->s_sid;
790 	} else
791 		kp->ki_tdev = NODEV;
792 	if (p->p_comm[0] != '\0')
793 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
794 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
795 	    p->p_sysent->sv_name[0] != '\0')
796 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
797 	kp->ki_siglist = p->p_siglist;
798 	kp->ki_xstat = p->p_xstat;
799 	kp->ki_acflag = p->p_acflag;
800 	kp->ki_lock = p->p_lock;
801 	if (p->p_pptr)
802 		kp->ki_ppid = p->p_pptr->p_pid;
803 }
804 
805 /*
806  * Fill in information that is thread specific.  Must be called with p_slock
807  * locked.  If 'preferthread' is set, overwrite certain process-related
808  * fields that are maintained for both threads and processes.
809  */
810 static void
811 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
812 {
813 	struct proc *p;
814 
815 	p = td->td_proc;
816 	PROC_LOCK_ASSERT(p, MA_OWNED);
817 
818 	thread_lock(td);
819 	if (td->td_wmesg != NULL)
820 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
821 	else
822 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
823 	if (td->td_name[0] != '\0')
824 		strlcpy(kp->ki_ocomm, td->td_name, sizeof(kp->ki_ocomm));
825 	if (TD_ON_LOCK(td)) {
826 		kp->ki_kiflag |= KI_LOCKBLOCK;
827 		strlcpy(kp->ki_lockname, td->td_lockname,
828 		    sizeof(kp->ki_lockname));
829 	} else {
830 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
831 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
832 	}
833 
834 	if (p->p_state == PRS_NORMAL) { /* approximate. */
835 		if (TD_ON_RUNQ(td) ||
836 		    TD_CAN_RUN(td) ||
837 		    TD_IS_RUNNING(td)) {
838 			kp->ki_stat = SRUN;
839 		} else if (P_SHOULDSTOP(p)) {
840 			kp->ki_stat = SSTOP;
841 		} else if (TD_IS_SLEEPING(td)) {
842 			kp->ki_stat = SSLEEP;
843 		} else if (TD_ON_LOCK(td)) {
844 			kp->ki_stat = SLOCK;
845 		} else {
846 			kp->ki_stat = SWAIT;
847 		}
848 	} else if (p->p_state == PRS_ZOMBIE) {
849 		kp->ki_stat = SZOMB;
850 	} else {
851 		kp->ki_stat = SIDL;
852 	}
853 
854 	/* Things in the thread */
855 	kp->ki_wchan = td->td_wchan;
856 	kp->ki_pri.pri_level = td->td_priority;
857 	kp->ki_pri.pri_native = td->td_base_pri;
858 	kp->ki_lastcpu = td->td_lastcpu;
859 	kp->ki_oncpu = td->td_oncpu;
860 	kp->ki_tdflags = td->td_flags;
861 	kp->ki_tid = td->td_tid;
862 	kp->ki_numthreads = p->p_numthreads;
863 	kp->ki_pcb = td->td_pcb;
864 	kp->ki_kstack = (void *)td->td_kstack;
865 	kp->ki_pctcpu = sched_pctcpu(td);
866 	kp->ki_estcpu = td->td_estcpu;
867 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
868 	kp->ki_pri.pri_class = td->td_pri_class;
869 	kp->ki_pri.pri_user = td->td_user_pri;
870 
871 	if (preferthread)
872 		kp->ki_runtime = cputick2usec(td->td_runtime);
873 
874 	/* We can't get this anymore but ps etc never used it anyway. */
875 	kp->ki_rqindex = 0;
876 
877 	SIGSETOR(kp->ki_siglist, td->td_siglist);
878 	kp->ki_sigmask = td->td_sigmask;
879 	thread_unlock(td);
880 }
881 
882 /*
883  * Fill in a kinfo_proc structure for the specified process.
884  * Must be called with the target process locked.
885  */
886 void
887 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
888 {
889 
890 	fill_kinfo_proc_only(p, kp);
891 	if (FIRST_THREAD_IN_PROC(p) != NULL)
892 		fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
893 }
894 
895 struct pstats *
896 pstats_alloc(void)
897 {
898 
899 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
900 }
901 
902 /*
903  * Copy parts of p_stats; zero the rest of p_stats (statistics).
904  */
905 void
906 pstats_fork(struct pstats *src, struct pstats *dst)
907 {
908 
909 	bzero(&dst->pstat_startzero,
910 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
911 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
912 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
913 }
914 
915 void
916 pstats_free(struct pstats *ps)
917 {
918 
919 	free(ps, M_SUBPROC);
920 }
921 
922 /*
923  * Locate a zombie process by number
924  */
925 struct proc *
926 zpfind(pid_t pid)
927 {
928 	struct proc *p;
929 
930 	sx_slock(&allproc_lock);
931 	LIST_FOREACH(p, &zombproc, p_list)
932 		if (p->p_pid == pid) {
933 			PROC_LOCK(p);
934 			break;
935 		}
936 	sx_sunlock(&allproc_lock);
937 	return (p);
938 }
939 
940 #define KERN_PROC_ZOMBMASK	0x3
941 #define KERN_PROC_NOTHREADS	0x4
942 
943 /*
944  * Must be called with the process locked and will return with it unlocked.
945  */
946 static int
947 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
948 {
949 	struct thread *td;
950 	struct kinfo_proc kinfo_proc;
951 	int error = 0;
952 	struct proc *np;
953 	pid_t pid = p->p_pid;
954 
955 	PROC_LOCK_ASSERT(p, MA_OWNED);
956 
957 	fill_kinfo_proc_only(p, &kinfo_proc);
958 	if (flags & KERN_PROC_NOTHREADS) {
959 		if (FIRST_THREAD_IN_PROC(p) != NULL)
960 			fill_kinfo_thread(FIRST_THREAD_IN_PROC(p),
961 			    &kinfo_proc, 0);
962 		error = SYSCTL_OUT(req, (caddr_t)&kinfo_proc,
963 				   sizeof(kinfo_proc));
964 	} else {
965 		if (FIRST_THREAD_IN_PROC(p) != NULL)
966 			FOREACH_THREAD_IN_PROC(p, td) {
967 				fill_kinfo_thread(td, &kinfo_proc, 1);
968 				error = SYSCTL_OUT(req, (caddr_t)&kinfo_proc,
969 						   sizeof(kinfo_proc));
970 				if (error)
971 					break;
972 			}
973 		else
974 			error = SYSCTL_OUT(req, (caddr_t)&kinfo_proc,
975 					   sizeof(kinfo_proc));
976 	}
977 	PROC_UNLOCK(p);
978 	if (error)
979 		return (error);
980 	if (flags & KERN_PROC_ZOMBMASK)
981 		np = zpfind(pid);
982 	else {
983 		if (pid == 0)
984 			return (0);
985 		np = pfind(pid);
986 	}
987 	if (np == NULL)
988 		return EAGAIN;
989 	if (np != p) {
990 		PROC_UNLOCK(np);
991 		return EAGAIN;
992 	}
993 	PROC_UNLOCK(np);
994 	return (0);
995 }
996 
997 static int
998 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
999 {
1000 	int *name = (int*) arg1;
1001 	u_int namelen = arg2;
1002 	struct proc *p;
1003 	int flags, doingzomb, oid_number;
1004 	int error = 0;
1005 
1006 	oid_number = oidp->oid_number;
1007 	if (oid_number != KERN_PROC_ALL &&
1008 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1009 		flags = KERN_PROC_NOTHREADS;
1010 	else {
1011 		flags = 0;
1012 		oid_number &= ~KERN_PROC_INC_THREAD;
1013 	}
1014 	if (oid_number == KERN_PROC_PID) {
1015 		if (namelen != 1)
1016 			return (EINVAL);
1017 		error = sysctl_wire_old_buffer(req, 0);
1018 		if (error)
1019 			return (error);
1020 		p = pfind((pid_t)name[0]);
1021 		if (!p)
1022 			return (ESRCH);
1023 		if ((error = p_cansee(curthread, p))) {
1024 			PROC_UNLOCK(p);
1025 			return (error);
1026 		}
1027 		error = sysctl_out_proc(p, req, flags);
1028 		return (error);
1029 	}
1030 
1031 	switch (oid_number) {
1032 	case KERN_PROC_ALL:
1033 		if (namelen != 0)
1034 			return (EINVAL);
1035 		break;
1036 	case KERN_PROC_PROC:
1037 		if (namelen != 0 && namelen != 1)
1038 			return (EINVAL);
1039 		break;
1040 	default:
1041 		if (namelen != 1)
1042 			return (EINVAL);
1043 		break;
1044 	}
1045 
1046 	if (!req->oldptr) {
1047 		/* overestimate by 5 procs */
1048 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1049 		if (error)
1050 			return (error);
1051 	}
1052 	error = sysctl_wire_old_buffer(req, 0);
1053 	if (error != 0)
1054 		return (error);
1055 	sx_slock(&allproc_lock);
1056 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1057 		if (!doingzomb)
1058 			p = LIST_FIRST(&allproc);
1059 		else
1060 			p = LIST_FIRST(&zombproc);
1061 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
1062 			/*
1063 			 * Skip embryonic processes.
1064 			 */
1065 			PROC_SLOCK(p);
1066 			if (p->p_state == PRS_NEW) {
1067 				PROC_SUNLOCK(p);
1068 				continue;
1069 			}
1070 			PROC_SUNLOCK(p);
1071 			PROC_LOCK(p);
1072 			KASSERT(p->p_ucred != NULL,
1073 			    ("process credential is NULL for non-NEW proc"));
1074 			/*
1075 			 * Show a user only appropriate processes.
1076 			 */
1077 			if (p_cansee(curthread, p)) {
1078 				PROC_UNLOCK(p);
1079 				continue;
1080 			}
1081 			/*
1082 			 * TODO - make more efficient (see notes below).
1083 			 * do by session.
1084 			 */
1085 			switch (oid_number) {
1086 
1087 			case KERN_PROC_GID:
1088 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1089 					PROC_UNLOCK(p);
1090 					continue;
1091 				}
1092 				break;
1093 
1094 			case KERN_PROC_PGRP:
1095 				/* could do this by traversing pgrp */
1096 				if (p->p_pgrp == NULL ||
1097 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1098 					PROC_UNLOCK(p);
1099 					continue;
1100 				}
1101 				break;
1102 
1103 			case KERN_PROC_RGID:
1104 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1105 					PROC_UNLOCK(p);
1106 					continue;
1107 				}
1108 				break;
1109 
1110 			case KERN_PROC_SESSION:
1111 				if (p->p_session == NULL ||
1112 				    p->p_session->s_sid != (pid_t)name[0]) {
1113 					PROC_UNLOCK(p);
1114 					continue;
1115 				}
1116 				break;
1117 
1118 			case KERN_PROC_TTY:
1119 				if ((p->p_flag & P_CONTROLT) == 0 ||
1120 				    p->p_session == NULL) {
1121 					PROC_UNLOCK(p);
1122 					continue;
1123 				}
1124 				SESS_LOCK(p->p_session);
1125 				if (p->p_session->s_ttyp == NULL ||
1126 				    dev2udev(p->p_session->s_ttyp->t_dev) !=
1127 				    (dev_t)name[0]) {
1128 					SESS_UNLOCK(p->p_session);
1129 					PROC_UNLOCK(p);
1130 					continue;
1131 				}
1132 				SESS_UNLOCK(p->p_session);
1133 				break;
1134 
1135 			case KERN_PROC_UID:
1136 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1137 					PROC_UNLOCK(p);
1138 					continue;
1139 				}
1140 				break;
1141 
1142 			case KERN_PROC_RUID:
1143 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1144 					PROC_UNLOCK(p);
1145 					continue;
1146 				}
1147 				break;
1148 
1149 			case KERN_PROC_PROC:
1150 				break;
1151 
1152 			default:
1153 				break;
1154 
1155 			}
1156 
1157 			error = sysctl_out_proc(p, req, flags | doingzomb);
1158 			if (error) {
1159 				sx_sunlock(&allproc_lock);
1160 				return (error);
1161 			}
1162 		}
1163 	}
1164 	sx_sunlock(&allproc_lock);
1165 	return (0);
1166 }
1167 
1168 struct pargs *
1169 pargs_alloc(int len)
1170 {
1171 	struct pargs *pa;
1172 
1173 	MALLOC(pa, struct pargs *, sizeof(struct pargs) + len, M_PARGS,
1174 		M_WAITOK);
1175 	refcount_init(&pa->ar_ref, 1);
1176 	pa->ar_length = len;
1177 	return (pa);
1178 }
1179 
1180 void
1181 pargs_free(struct pargs *pa)
1182 {
1183 
1184 	FREE(pa, M_PARGS);
1185 }
1186 
1187 void
1188 pargs_hold(struct pargs *pa)
1189 {
1190 
1191 	if (pa == NULL)
1192 		return;
1193 	refcount_acquire(&pa->ar_ref);
1194 }
1195 
1196 void
1197 pargs_drop(struct pargs *pa)
1198 {
1199 
1200 	if (pa == NULL)
1201 		return;
1202 	if (refcount_release(&pa->ar_ref))
1203 		pargs_free(pa);
1204 }
1205 
1206 /*
1207  * This sysctl allows a process to retrieve the argument list or process
1208  * title for another process without groping around in the address space
1209  * of the other process.  It also allow a process to set its own "process
1210  * title to a string of its own choice.
1211  */
1212 static int
1213 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1214 {
1215 	int *name = (int*) arg1;
1216 	u_int namelen = arg2;
1217 	struct pargs *newpa, *pa;
1218 	struct proc *p;
1219 	int error = 0;
1220 
1221 	if (namelen != 1)
1222 		return (EINVAL);
1223 
1224 	p = pfind((pid_t)name[0]);
1225 	if (!p)
1226 		return (ESRCH);
1227 
1228 	if ((error = p_cansee(curthread, p)) != 0) {
1229 		PROC_UNLOCK(p);
1230 		return (error);
1231 	}
1232 
1233 	if (req->newptr && curproc != p) {
1234 		PROC_UNLOCK(p);
1235 		return (EPERM);
1236 	}
1237 
1238 	pa = p->p_args;
1239 	pargs_hold(pa);
1240 	PROC_UNLOCK(p);
1241 	if (req->oldptr != NULL && pa != NULL)
1242 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1243 	pargs_drop(pa);
1244 	if (error != 0 || req->newptr == NULL)
1245 		return (error);
1246 
1247 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1248 		return (ENOMEM);
1249 	newpa = pargs_alloc(req->newlen);
1250 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1251 	if (error != 0) {
1252 		pargs_free(newpa);
1253 		return (error);
1254 	}
1255 	PROC_LOCK(p);
1256 	pa = p->p_args;
1257 	p->p_args = newpa;
1258 	PROC_UNLOCK(p);
1259 	pargs_drop(pa);
1260 	return (0);
1261 }
1262 
1263 /*
1264  * This sysctl allows a process to retrieve the path of the executable for
1265  * itself or another process.
1266  */
1267 static int
1268 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1269 {
1270 	pid_t *pidp = (pid_t *)arg1;
1271 	unsigned int arglen = arg2;
1272 	struct proc *p;
1273 	struct vnode *vp;
1274 	char *retbuf, *freebuf;
1275 	int error;
1276 
1277 	if (arglen != 1)
1278 		return (EINVAL);
1279 	if (*pidp == -1) {	/* -1 means this process */
1280 		p = req->td->td_proc;
1281 	} else {
1282 		p = pfind(*pidp);
1283 		if (p == NULL)
1284 			return (ESRCH);
1285 		if ((error = p_cansee(curthread, p)) != 0) {
1286 			PROC_UNLOCK(p);
1287 			return (error);
1288 		}
1289 	}
1290 
1291 	vp = p->p_textvp;
1292 	if (vp == NULL) {
1293 		if (*pidp != -1)
1294 			PROC_UNLOCK(p);
1295 		return (0);
1296 	}
1297 	vref(vp);
1298 	if (*pidp != -1)
1299 		PROC_UNLOCK(p);
1300 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1301 	vrele(vp);
1302 	if (error)
1303 		return (error);
1304 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1305 	free(freebuf, M_TEMP);
1306 	return (error);
1307 }
1308 
1309 static int
1310 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1311 {
1312 	struct proc *p;
1313 	char *sv_name;
1314 	int *name;
1315 	int namelen;
1316 	int error;
1317 
1318 	namelen = arg2;
1319 	if (namelen != 1)
1320 		return (EINVAL);
1321 
1322 	name = (int *)arg1;
1323 	if ((p = pfind((pid_t)name[0])) == NULL)
1324 		return (ESRCH);
1325 	if ((error = p_cansee(curthread, p))) {
1326 		PROC_UNLOCK(p);
1327 		return (error);
1328 	}
1329 	sv_name = p->p_sysent->sv_name;
1330 	PROC_UNLOCK(p);
1331 	return (sysctl_handle_string(oidp, sv_name, 0, req));
1332 }
1333 
1334 static int
1335 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
1336 {
1337 	vm_map_entry_t entry, tmp_entry;
1338 	unsigned int last_timestamp;
1339 	char *fullpath, *freepath;
1340 	struct kinfo_vmentry *kve;
1341 	int error, *name;
1342 	struct vnode *vp;
1343 	struct proc *p;
1344 	vm_map_t map;
1345 
1346 	name = (int *)arg1;
1347 	if ((p = pfind((pid_t)name[0])) == NULL)
1348 		return (ESRCH);
1349 	if (p->p_flag & P_WEXIT) {
1350 		PROC_UNLOCK(p);
1351 		return (ESRCH);
1352 	}
1353 	if ((error = p_candebug(curthread, p))) {
1354 		PROC_UNLOCK(p);
1355 		return (error);
1356 	}
1357 	_PHOLD(p);
1358 	PROC_UNLOCK(p);
1359 
1360 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1361 
1362 	map = &p->p_vmspace->vm_map;	/* XXXRW: More locking required? */
1363 	vm_map_lock_read(map);
1364 	for (entry = map->header.next; entry != &map->header;
1365 	    entry = entry->next) {
1366 		vm_object_t obj, tobj, lobj;
1367 		vm_offset_t addr;
1368 		int vfslocked;
1369 
1370 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
1371 			continue;
1372 
1373 		bzero(kve, sizeof(*kve));
1374 		kve->kve_structsize = sizeof(*kve);
1375 
1376 		kve->kve_private_resident = 0;
1377 		obj = entry->object.vm_object;
1378 		if (obj != NULL) {
1379 			VM_OBJECT_LOCK(obj);
1380 			if (obj->shadow_count == 1)
1381 				kve->kve_private_resident =
1382 				    obj->resident_page_count;
1383 		}
1384 		kve->kve_resident = 0;
1385 		addr = entry->start;
1386 		while (addr < entry->end) {
1387 			if (pmap_extract(map->pmap, addr))
1388 				kve->kve_resident++;
1389 			addr += PAGE_SIZE;
1390 		}
1391 
1392 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
1393 			if (tobj != obj)
1394 				VM_OBJECT_LOCK(tobj);
1395 			if (lobj != obj)
1396 				VM_OBJECT_UNLOCK(lobj);
1397 			lobj = tobj;
1398 		}
1399 
1400 		freepath = NULL;
1401 		fullpath = "";
1402 		if (lobj) {
1403 			vp = NULL;
1404 			switch(lobj->type) {
1405 			case OBJT_DEFAULT:
1406 				kve->kve_type = KVME_TYPE_DEFAULT;
1407 				break;
1408 			case OBJT_VNODE:
1409 				kve->kve_type = KVME_TYPE_VNODE;
1410 				vp = lobj->handle;
1411 				vref(vp);
1412 				break;
1413 			case OBJT_SWAP:
1414 				kve->kve_type = KVME_TYPE_SWAP;
1415 				break;
1416 			case OBJT_DEVICE:
1417 				kve->kve_type = KVME_TYPE_DEVICE;
1418 				break;
1419 			case OBJT_PHYS:
1420 				kve->kve_type = KVME_TYPE_PHYS;
1421 				break;
1422 			case OBJT_DEAD:
1423 				kve->kve_type = KVME_TYPE_DEAD;
1424 				break;
1425 			default:
1426 				kve->kve_type = KVME_TYPE_UNKNOWN;
1427 				break;
1428 			}
1429 			if (lobj != obj)
1430 				VM_OBJECT_UNLOCK(lobj);
1431 
1432 			kve->kve_ref_count = obj->ref_count;
1433 			kve->kve_shadow_count = obj->shadow_count;
1434 			VM_OBJECT_UNLOCK(obj);
1435 			if (vp != NULL) {
1436 				vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1437 				vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1438 				vn_fullpath(curthread, vp, &fullpath,
1439 				    &freepath);
1440 				vput(vp);
1441 				VFS_UNLOCK_GIANT(vfslocked);
1442 			}
1443 		} else {
1444 			kve->kve_type = KVME_TYPE_NONE;
1445 			kve->kve_ref_count = 0;
1446 			kve->kve_shadow_count = 0;
1447 		}
1448 
1449 		kve->kve_start = (void*)entry->start;
1450 		kve->kve_end = (void*)entry->end;
1451 
1452 		if (entry->protection & VM_PROT_READ)
1453 			kve->kve_protection |= KVME_PROT_READ;
1454 		if (entry->protection & VM_PROT_WRITE)
1455 			kve->kve_protection |= KVME_PROT_WRITE;
1456 		if (entry->protection & VM_PROT_EXECUTE)
1457 			kve->kve_protection |= KVME_PROT_EXEC;
1458 
1459 		if (entry->eflags & MAP_ENTRY_COW)
1460 			kve->kve_flags |= KVME_FLAG_COW;
1461 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
1462 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
1463 
1464 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
1465 		if (freepath != NULL)
1466 			free(freepath, M_TEMP);
1467 
1468 		last_timestamp = map->timestamp;
1469 		vm_map_unlock_read(map);
1470 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
1471 		vm_map_lock_read(map);
1472 		if (error)
1473 			break;
1474 		if (last_timestamp + 1 != map->timestamp) {
1475 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
1476 			entry = tmp_entry;
1477 		}
1478 	}
1479 	vm_map_unlock_read(map);
1480 	PRELE(p);
1481 	free(kve, M_TEMP);
1482 	return (error);
1483 }
1484 
1485 #if defined(STACK) || defined(DDB)
1486 static int
1487 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
1488 {
1489 	struct kinfo_kstack *kkstp;
1490 	int error, i, *name, numthreads;
1491 	lwpid_t *lwpidarray;
1492 	struct thread *td;
1493 	struct stack *st;
1494 	struct sbuf sb;
1495 	struct proc *p;
1496 
1497 	name = (int *)arg1;
1498 	if ((p = pfind((pid_t)name[0])) == NULL)
1499 		return (ESRCH);
1500 	/* XXXRW: Not clear ESRCH is the right error during proc execve(). */
1501 	if (p->p_flag & P_WEXIT || p->p_flag & P_INEXEC) {
1502 		PROC_UNLOCK(p);
1503 		return (ESRCH);
1504 	}
1505 	if ((error = p_candebug(curthread, p))) {
1506 		PROC_UNLOCK(p);
1507 		return (error);
1508 	}
1509 	_PHOLD(p);
1510 	PROC_UNLOCK(p);
1511 
1512 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
1513 	st = stack_create();
1514 
1515 	lwpidarray = NULL;
1516 	numthreads = 0;
1517 	PROC_LOCK(p);
1518 repeat:
1519 	if (numthreads < p->p_numthreads) {
1520 		if (lwpidarray != NULL) {
1521 			free(lwpidarray, M_TEMP);
1522 			lwpidarray = NULL;
1523 		}
1524 		numthreads = p->p_numthreads;
1525 		PROC_UNLOCK(p);
1526 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
1527 		    M_WAITOK | M_ZERO);
1528 		PROC_LOCK(p);
1529 		goto repeat;
1530 	}
1531 	i = 0;
1532 
1533 	/*
1534 	 * XXXRW: During the below loop, execve(2) and countless other sorts
1535 	 * of changes could have taken place.  Should we check to see if the
1536 	 * vmspace has been replaced, or the like, in order to prevent
1537 	 * giving a snapshot that spans, say, execve(2), with some threads
1538 	 * before and some after?  Among other things, the credentials could
1539 	 * have changed, in which case the right to extract debug info might
1540 	 * no longer be assured.
1541 	 */
1542 	FOREACH_THREAD_IN_PROC(p, td) {
1543 		KASSERT(i < numthreads,
1544 		    ("sysctl_kern_proc_kstack: numthreads"));
1545 		lwpidarray[i] = td->td_tid;
1546 		i++;
1547 	}
1548 	numthreads = i;
1549 	for (i = 0; i < numthreads; i++) {
1550 		td = thread_find(p, lwpidarray[i]);
1551 		if (td == NULL) {
1552 			continue;
1553 		}
1554 		bzero(kkstp, sizeof(*kkstp));
1555 		(void)sbuf_new(&sb, kkstp->kkst_trace,
1556 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
1557 		thread_lock(td);
1558 		kkstp->kkst_tid = td->td_tid;
1559 		if (TD_IS_SWAPPED(td))
1560 			kkstp->kkst_state = KKST_STATE_SWAPPED;
1561 		else if (TD_IS_RUNNING(td))
1562 			kkstp->kkst_state = KKST_STATE_RUNNING;
1563 		else {
1564 			kkstp->kkst_state = KKST_STATE_STACKOK;
1565 			stack_save_td(st, td);
1566 		}
1567 		thread_unlock(td);
1568 		PROC_UNLOCK(p);
1569 		stack_sbuf_print(&sb, st);
1570 		sbuf_finish(&sb);
1571 		sbuf_delete(&sb);
1572 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
1573 		PROC_LOCK(p);
1574 		if (error)
1575 			break;
1576 	}
1577 	_PRELE(p);
1578 	PROC_UNLOCK(p);
1579 	if (lwpidarray != NULL)
1580 		free(lwpidarray, M_TEMP);
1581 	stack_destroy(st);
1582 	free(kkstp, M_TEMP);
1583 	return (error);
1584 }
1585 #endif
1586 
1587 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
1588 
1589 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT,
1590 	0, 0, sysctl_kern_proc, "S,proc", "Return entire process table");
1591 
1592 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD,
1593 	sysctl_kern_proc, "Process table");
1594 
1595 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD,
1596 	sysctl_kern_proc, "Process table");
1597 
1598 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD,
1599 	sysctl_kern_proc, "Process table");
1600 
1601 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD,
1602 	sysctl_kern_proc, "Process table");
1603 
1604 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD,
1605 	sysctl_kern_proc, "Process table");
1606 
1607 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD,
1608 	sysctl_kern_proc, "Process table");
1609 
1610 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD,
1611 	sysctl_kern_proc, "Process table");
1612 
1613 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD,
1614 	sysctl_kern_proc, "Process table");
1615 
1616 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD,
1617 	sysctl_kern_proc, "Return process table, no threads");
1618 
1619 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
1620 	CTLFLAG_RW | CTLFLAG_ANYBODY,
1621 	sysctl_kern_proc_args, "Process argument list");
1622 
1623 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD,
1624 	sysctl_kern_proc_pathname, "Process executable path");
1625 
1626 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD,
1627 	sysctl_kern_proc_sv_name, "Process syscall vector name (ABI type)");
1628 
1629 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
1630 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1631 
1632 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
1633 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1634 
1635 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
1636 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1637 
1638 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
1639 	sid_td, CTLFLAG_RD, sysctl_kern_proc, "Process table");
1640 
1641 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
1642 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1643 
1644 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
1645 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1646 
1647 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
1648 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1649 
1650 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
1651 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1652 
1653 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
1654 	CTLFLAG_RD, sysctl_kern_proc, "Return process table, no threads");
1655 
1656 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD,
1657 	sysctl_kern_proc_vmmap, "Process vm map entries");
1658 
1659 #if defined(STACK) || defined(DDB)
1660 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD,
1661 	sysctl_kern_proc_kstack, "Process kernel stacks");
1662 #endif
1663