xref: /freebsd/sys/kern/kern_proc.c (revision 33644623554bb0fc57ed3c7d874193a498679b22)
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 		/* Free all OSD associated to this thread. */
203 		osd_thread_exit(td);
204 
205 		/* Dispose of an alternate kstack, if it exists.
206 		 * XXX What if there are more than one thread in the proc?
207 		 *     The first thread in the proc is special and not
208 		 *     freed, so you gotta do this here.
209 		 */
210 		if (((p->p_flag & P_KTHREAD) != 0) && (td->td_altkstack != 0))
211 			vm_thread_dispose_altkstack(td);
212 	}
213 	EVENTHANDLER_INVOKE(process_dtor, p);
214 	if (p->p_ksi != NULL)
215 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
216 	SDT_PROBE(proc, kernel, dtor, return, p, size, arg, 0, 0);
217 }
218 
219 /*
220  * Initialize type-stable parts of a proc (when newly created).
221  */
222 static int
223 proc_init(void *mem, int size, int flags)
224 {
225 	struct proc *p;
226 
227 	p = (struct proc *)mem;
228 	SDT_PROBE(proc, kernel, init, entry, p, size, flags, 0, 0);
229 	p->p_sched = (struct p_sched *)&p[1];
230 	bzero(&p->p_mtx, sizeof(struct mtx));
231 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK);
232 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_RECURSE);
233 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
234 	EVENTHANDLER_INVOKE(process_init, p);
235 	p->p_stats = pstats_alloc();
236 	SDT_PROBE(proc, kernel, init, return, p, size, flags, 0, 0);
237 	return (0);
238 }
239 
240 /*
241  * UMA should ensure that this function is never called.
242  * Freeing a proc structure would violate type stability.
243  */
244 static void
245 proc_fini(void *mem, int size)
246 {
247 #ifdef notnow
248 	struct proc *p;
249 
250 	p = (struct proc *)mem;
251 	EVENTHANDLER_INVOKE(process_fini, p);
252 	pstats_free(p->p_stats);
253 	thread_free(FIRST_THREAD_IN_PROC(p));
254 	mtx_destroy(&p->p_mtx);
255 	if (p->p_ksi != NULL)
256 		ksiginfo_free(p->p_ksi);
257 #else
258 	panic("proc reclaimed");
259 #endif
260 }
261 
262 /*
263  * Is p an inferior of the current process?
264  */
265 int
266 inferior(p)
267 	register struct proc *p;
268 {
269 
270 	sx_assert(&proctree_lock, SX_LOCKED);
271 	for (; p != curproc; p = p->p_pptr)
272 		if (p->p_pid == 0)
273 			return (0);
274 	return (1);
275 }
276 
277 /*
278  * Locate a process by number; return only "live" processes -- i.e., neither
279  * zombies nor newly born but incompletely initialized processes.  By not
280  * returning processes in the PRS_NEW state, we allow callers to avoid
281  * testing for that condition to avoid dereferencing p_ucred, et al.
282  */
283 struct proc *
284 pfind(pid)
285 	register pid_t pid;
286 {
287 	register struct proc *p;
288 
289 	sx_slock(&allproc_lock);
290 	LIST_FOREACH(p, PIDHASH(pid), p_hash)
291 		if (p->p_pid == pid) {
292 			if (p->p_state == PRS_NEW) {
293 				p = NULL;
294 				break;
295 			}
296 			PROC_LOCK(p);
297 			break;
298 		}
299 	sx_sunlock(&allproc_lock);
300 	return (p);
301 }
302 
303 /*
304  * Locate a process group by number.
305  * The caller must hold proctree_lock.
306  */
307 struct pgrp *
308 pgfind(pgid)
309 	register pid_t pgid;
310 {
311 	register struct pgrp *pgrp;
312 
313 	sx_assert(&proctree_lock, SX_LOCKED);
314 
315 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
316 		if (pgrp->pg_id == pgid) {
317 			PGRP_LOCK(pgrp);
318 			return (pgrp);
319 		}
320 	}
321 	return (NULL);
322 }
323 
324 /*
325  * Create a new process group.
326  * pgid must be equal to the pid of p.
327  * Begin a new session if required.
328  */
329 int
330 enterpgrp(p, pgid, pgrp, sess)
331 	register struct proc *p;
332 	pid_t pgid;
333 	struct pgrp *pgrp;
334 	struct session *sess;
335 {
336 	struct pgrp *pgrp2;
337 
338 	sx_assert(&proctree_lock, SX_XLOCKED);
339 
340 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
341 	KASSERT(p->p_pid == pgid,
342 	    ("enterpgrp: new pgrp and pid != pgid"));
343 
344 	pgrp2 = pgfind(pgid);
345 
346 	KASSERT(pgrp2 == NULL,
347 	    ("enterpgrp: pgrp with pgid exists"));
348 	KASSERT(!SESS_LEADER(p),
349 	    ("enterpgrp: session leader attempted setpgrp"));
350 
351 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
352 
353 	if (sess != NULL) {
354 		/*
355 		 * new session
356 		 */
357 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
358 		PROC_LOCK(p);
359 		p->p_flag &= ~P_CONTROLT;
360 		PROC_UNLOCK(p);
361 		PGRP_LOCK(pgrp);
362 		sess->s_leader = p;
363 		sess->s_sid = p->p_pid;
364 		refcount_init(&sess->s_count, 1);
365 		sess->s_ttyvp = NULL;
366 		sess->s_ttyp = NULL;
367 		bcopy(p->p_session->s_login, sess->s_login,
368 			    sizeof(sess->s_login));
369 		pgrp->pg_session = sess;
370 		KASSERT(p == curproc,
371 		    ("enterpgrp: mksession and p != curproc"));
372 	} else {
373 		pgrp->pg_session = p->p_session;
374 		sess_hold(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 
389 	doenterpgrp(p, pgrp);
390 
391 	return (0);
392 }
393 
394 /*
395  * Move p to an existing process group
396  */
397 int
398 enterthispgrp(p, pgrp)
399 	register struct proc *p;
400 	struct pgrp *pgrp;
401 {
402 
403 	sx_assert(&proctree_lock, SX_XLOCKED);
404 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
405 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
406 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
407 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
408 	KASSERT(pgrp->pg_session == p->p_session,
409 		("%s: pgrp's session %p, p->p_session %p.\n",
410 		__func__,
411 		pgrp->pg_session,
412 		p->p_session));
413 	KASSERT(pgrp != p->p_pgrp,
414 		("%s: p belongs to pgrp.", __func__));
415 
416 	doenterpgrp(p, pgrp);
417 
418 	return (0);
419 }
420 
421 /*
422  * Move p to a process group
423  */
424 static void
425 doenterpgrp(p, pgrp)
426 	struct proc *p;
427 	struct pgrp *pgrp;
428 {
429 	struct pgrp *savepgrp;
430 
431 	sx_assert(&proctree_lock, SX_XLOCKED);
432 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
433 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
434 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
435 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
436 
437 	savepgrp = p->p_pgrp;
438 
439 	/*
440 	 * Adjust eligibility of affected pgrps to participate in job control.
441 	 * Increment eligibility counts before decrementing, otherwise we
442 	 * could reach 0 spuriously during the first call.
443 	 */
444 	fixjobc(p, pgrp, 1);
445 	fixjobc(p, p->p_pgrp, 0);
446 
447 	PGRP_LOCK(pgrp);
448 	PGRP_LOCK(savepgrp);
449 	PROC_LOCK(p);
450 	LIST_REMOVE(p, p_pglist);
451 	p->p_pgrp = pgrp;
452 	PROC_UNLOCK(p);
453 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
454 	PGRP_UNLOCK(savepgrp);
455 	PGRP_UNLOCK(pgrp);
456 	if (LIST_EMPTY(&savepgrp->pg_members))
457 		pgdelete(savepgrp);
458 }
459 
460 /*
461  * remove process from process group
462  */
463 int
464 leavepgrp(p)
465 	register struct proc *p;
466 {
467 	struct pgrp *savepgrp;
468 
469 	sx_assert(&proctree_lock, SX_XLOCKED);
470 	savepgrp = p->p_pgrp;
471 	PGRP_LOCK(savepgrp);
472 	PROC_LOCK(p);
473 	LIST_REMOVE(p, p_pglist);
474 	p->p_pgrp = NULL;
475 	PROC_UNLOCK(p);
476 	PGRP_UNLOCK(savepgrp);
477 	if (LIST_EMPTY(&savepgrp->pg_members))
478 		pgdelete(savepgrp);
479 	return (0);
480 }
481 
482 /*
483  * delete a process group
484  */
485 static void
486 pgdelete(pgrp)
487 	register struct pgrp *pgrp;
488 {
489 	struct session *savesess;
490 	struct tty *tp;
491 
492 	sx_assert(&proctree_lock, SX_XLOCKED);
493 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
494 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
495 
496 	/*
497 	 * Reset any sigio structures pointing to us as a result of
498 	 * F_SETOWN with our pgid.
499 	 */
500 	funsetownlst(&pgrp->pg_sigiolst);
501 
502 	PGRP_LOCK(pgrp);
503 	tp = pgrp->pg_session->s_ttyp;
504 	LIST_REMOVE(pgrp, pg_hash);
505 	savesess = pgrp->pg_session;
506 	PGRP_UNLOCK(pgrp);
507 
508 	/* Remove the reference to the pgrp before deallocating it. */
509 	if (tp != NULL) {
510 		tty_lock(tp);
511 		tty_rel_pgrp(tp, pgrp);
512 	}
513 
514 	mtx_destroy(&pgrp->pg_mtx);
515 	free(pgrp, M_PGRP);
516 	sess_release(savesess);
517 }
518 
519 static void
520 pgadjustjobc(pgrp, entering)
521 	struct pgrp *pgrp;
522 	int entering;
523 {
524 
525 	PGRP_LOCK(pgrp);
526 	if (entering)
527 		pgrp->pg_jobc++;
528 	else {
529 		--pgrp->pg_jobc;
530 		if (pgrp->pg_jobc == 0)
531 			orphanpg(pgrp);
532 	}
533 	PGRP_UNLOCK(pgrp);
534 }
535 
536 /*
537  * Adjust pgrp jobc counters when specified process changes process group.
538  * We count the number of processes in each process group that "qualify"
539  * the group for terminal job control (those with a parent in a different
540  * process group of the same session).  If that count reaches zero, the
541  * process group becomes orphaned.  Check both the specified process'
542  * process group and that of its children.
543  * entering == 0 => p is leaving specified group.
544  * entering == 1 => p is entering specified group.
545  */
546 void
547 fixjobc(p, pgrp, entering)
548 	register struct proc *p;
549 	register struct pgrp *pgrp;
550 	int entering;
551 {
552 	register struct pgrp *hispgrp;
553 	register struct session *mysession;
554 
555 	sx_assert(&proctree_lock, SX_LOCKED);
556 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
557 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
558 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
559 
560 	/*
561 	 * Check p's parent to see whether p qualifies its own process
562 	 * group; if so, adjust count for p's process group.
563 	 */
564 	mysession = pgrp->pg_session;
565 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
566 	    hispgrp->pg_session == mysession)
567 		pgadjustjobc(pgrp, entering);
568 
569 	/*
570 	 * Check this process' children to see whether they qualify
571 	 * their process groups; if so, adjust counts for children's
572 	 * process groups.
573 	 */
574 	LIST_FOREACH(p, &p->p_children, p_sibling) {
575 		hispgrp = p->p_pgrp;
576 		if (hispgrp == pgrp ||
577 		    hispgrp->pg_session != mysession)
578 			continue;
579 		PROC_LOCK(p);
580 		if (p->p_state == PRS_ZOMBIE) {
581 			PROC_UNLOCK(p);
582 			continue;
583 		}
584 		PROC_UNLOCK(p);
585 		pgadjustjobc(hispgrp, entering);
586 	}
587 }
588 
589 /*
590  * A process group has become orphaned;
591  * if there are any stopped processes in the group,
592  * hang-up all process in that group.
593  */
594 static void
595 orphanpg(pg)
596 	struct pgrp *pg;
597 {
598 	register struct proc *p;
599 
600 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
601 
602 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
603 		PROC_LOCK(p);
604 		if (P_SHOULDSTOP(p)) {
605 			PROC_UNLOCK(p);
606 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
607 				PROC_LOCK(p);
608 				psignal(p, SIGHUP);
609 				psignal(p, SIGCONT);
610 				PROC_UNLOCK(p);
611 			}
612 			return;
613 		}
614 		PROC_UNLOCK(p);
615 	}
616 }
617 
618 void
619 sess_hold(struct session *s)
620 {
621 
622 	refcount_acquire(&s->s_count);
623 }
624 
625 void
626 sess_release(struct session *s)
627 {
628 
629 	if (refcount_release(&s->s_count)) {
630 		if (s->s_ttyp != NULL) {
631 			tty_lock(s->s_ttyp);
632 			tty_rel_sess(s->s_ttyp, s);
633 		}
634 		mtx_destroy(&s->s_mtx);
635 		free(s, M_SESSION);
636 	}
637 }
638 
639 #include "opt_ddb.h"
640 #ifdef DDB
641 #include <ddb/ddb.h>
642 
643 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
644 {
645 	register struct pgrp *pgrp;
646 	register struct proc *p;
647 	register int i;
648 
649 	for (i = 0; i <= pgrphash; i++) {
650 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
651 			printf("\tindx %d\n", i);
652 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
653 				printf(
654 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
655 				    (void *)pgrp, (long)pgrp->pg_id,
656 				    (void *)pgrp->pg_session,
657 				    pgrp->pg_session->s_count,
658 				    (void *)LIST_FIRST(&pgrp->pg_members));
659 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
660 					printf("\t\tpid %ld addr %p pgrp %p\n",
661 					    (long)p->p_pid, (void *)p,
662 					    (void *)p->p_pgrp);
663 				}
664 			}
665 		}
666 	}
667 }
668 #endif /* DDB */
669 
670 /*
671  * Clear kinfo_proc and fill in any information that is common
672  * to all threads in the process.
673  * Must be called with the target process locked.
674  */
675 static void
676 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
677 {
678 	struct thread *td0;
679 	struct tty *tp;
680 	struct session *sp;
681 	struct ucred *cred;
682 	struct sigacts *ps;
683 
684 	PROC_LOCK_ASSERT(p, MA_OWNED);
685 	bzero(kp, sizeof(*kp));
686 
687 	kp->ki_structsize = sizeof(*kp);
688 	kp->ki_paddr = p;
689 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
690 	kp->ki_args = p->p_args;
691 	kp->ki_textvp = p->p_textvp;
692 #ifdef KTRACE
693 	kp->ki_tracep = p->p_tracevp;
694 	mtx_lock(&ktrace_mtx);
695 	kp->ki_traceflag = p->p_traceflag;
696 	mtx_unlock(&ktrace_mtx);
697 #endif
698 	kp->ki_fd = p->p_fd;
699 	kp->ki_vmspace = p->p_vmspace;
700 	kp->ki_flag = p->p_flag;
701 	cred = p->p_ucred;
702 	if (cred) {
703 		kp->ki_uid = cred->cr_uid;
704 		kp->ki_ruid = cred->cr_ruid;
705 		kp->ki_svuid = cred->cr_svuid;
706 		/* XXX bde doesn't like KI_NGROUPS */
707 		kp->ki_ngroups = min(cred->cr_ngroups, KI_NGROUPS);
708 		bcopy(cred->cr_groups, kp->ki_groups,
709 		    kp->ki_ngroups * sizeof(gid_t));
710 		kp->ki_rgid = cred->cr_rgid;
711 		kp->ki_svgid = cred->cr_svgid;
712 		/* If jailed(cred), emulate the old P_JAILED flag. */
713 		if (jailed(cred)) {
714 			kp->ki_flag |= P_JAILED;
715 			/* If inside a jail, use 0 as a jail ID. */
716 			if (!jailed(curthread->td_ucred))
717 				kp->ki_jid = cred->cr_prison->pr_id;
718 		}
719 	}
720 	ps = p->p_sigacts;
721 	if (ps) {
722 		mtx_lock(&ps->ps_mtx);
723 		kp->ki_sigignore = ps->ps_sigignore;
724 		kp->ki_sigcatch = ps->ps_sigcatch;
725 		mtx_unlock(&ps->ps_mtx);
726 	}
727 	PROC_SLOCK(p);
728 	if (p->p_state != PRS_NEW &&
729 	    p->p_state != PRS_ZOMBIE &&
730 	    p->p_vmspace != NULL) {
731 		struct vmspace *vm = p->p_vmspace;
732 
733 		kp->ki_size = vm->vm_map.size;
734 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
735 		FOREACH_THREAD_IN_PROC(p, td0) {
736 			if (!TD_IS_SWAPPED(td0))
737 				kp->ki_rssize += td0->td_kstack_pages;
738 			if (td0->td_altkstack_obj != NULL)
739 				kp->ki_rssize += td0->td_altkstack_pages;
740 		}
741 		kp->ki_swrss = vm->vm_swrss;
742 		kp->ki_tsize = vm->vm_tsize;
743 		kp->ki_dsize = vm->vm_dsize;
744 		kp->ki_ssize = vm->vm_ssize;
745 	} else if (p->p_state == PRS_ZOMBIE)
746 		kp->ki_stat = SZOMB;
747 	if (kp->ki_flag & P_INMEM)
748 		kp->ki_sflag = PS_INMEM;
749 	else
750 		kp->ki_sflag = 0;
751 	/* Calculate legacy swtime as seconds since 'swtick'. */
752 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
753 	kp->ki_pid = p->p_pid;
754 	kp->ki_nice = p->p_nice;
755 	rufetch(p, &kp->ki_rusage);
756 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
757 	PROC_SUNLOCK(p);
758 	if ((p->p_flag & P_INMEM) && p->p_stats != NULL) {
759 		kp->ki_start = p->p_stats->p_start;
760 		timevaladd(&kp->ki_start, &boottime);
761 		PROC_SLOCK(p);
762 		calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
763 		PROC_SUNLOCK(p);
764 		calccru(p, &kp->ki_childutime, &kp->ki_childstime);
765 
766 		/* Some callers want child-times in a single value */
767 		kp->ki_childtime = kp->ki_childstime;
768 		timevaladd(&kp->ki_childtime, &kp->ki_childutime);
769 	}
770 	tp = NULL;
771 	if (p->p_pgrp) {
772 		kp->ki_pgid = p->p_pgrp->pg_id;
773 		kp->ki_jobc = p->p_pgrp->pg_jobc;
774 		sp = p->p_pgrp->pg_session;
775 
776 		if (sp != NULL) {
777 			kp->ki_sid = sp->s_sid;
778 			SESS_LOCK(sp);
779 			strlcpy(kp->ki_login, sp->s_login,
780 			    sizeof(kp->ki_login));
781 			if (sp->s_ttyvp)
782 				kp->ki_kiflag |= KI_CTTY;
783 			if (SESS_LEADER(p))
784 				kp->ki_kiflag |= KI_SLEADER;
785 			/* XXX proctree_lock */
786 			tp = sp->s_ttyp;
787 			SESS_UNLOCK(sp);
788 		}
789 	}
790 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
791 		kp->ki_tdev = tty_udev(tp);
792 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
793 		if (tp->t_session)
794 			kp->ki_tsid = tp->t_session->s_sid;
795 	} else
796 		kp->ki_tdev = NODEV;
797 	if (p->p_comm[0] != '\0')
798 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
799 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
800 	    p->p_sysent->sv_name[0] != '\0')
801 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
802 	kp->ki_siglist = p->p_siglist;
803 	kp->ki_xstat = p->p_xstat;
804 	kp->ki_acflag = p->p_acflag;
805 	kp->ki_lock = p->p_lock;
806 	if (p->p_pptr)
807 		kp->ki_ppid = p->p_pptr->p_pid;
808 }
809 
810 /*
811  * Fill in information that is thread specific.  Must be called with p_slock
812  * locked.  If 'preferthread' is set, overwrite certain process-related
813  * fields that are maintained for both threads and processes.
814  */
815 static void
816 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
817 {
818 	struct proc *p;
819 
820 	p = td->td_proc;
821 	PROC_LOCK_ASSERT(p, MA_OWNED);
822 
823 	thread_lock(td);
824 	if (td->td_wmesg != NULL)
825 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
826 	else
827 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
828 	if (td->td_name[0] != '\0')
829 		strlcpy(kp->ki_ocomm, td->td_name, sizeof(kp->ki_ocomm));
830 	if (TD_ON_LOCK(td)) {
831 		kp->ki_kiflag |= KI_LOCKBLOCK;
832 		strlcpy(kp->ki_lockname, td->td_lockname,
833 		    sizeof(kp->ki_lockname));
834 	} else {
835 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
836 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
837 	}
838 
839 	if (p->p_state == PRS_NORMAL) { /* approximate. */
840 		if (TD_ON_RUNQ(td) ||
841 		    TD_CAN_RUN(td) ||
842 		    TD_IS_RUNNING(td)) {
843 			kp->ki_stat = SRUN;
844 		} else if (P_SHOULDSTOP(p)) {
845 			kp->ki_stat = SSTOP;
846 		} else if (TD_IS_SLEEPING(td)) {
847 			kp->ki_stat = SSLEEP;
848 		} else if (TD_ON_LOCK(td)) {
849 			kp->ki_stat = SLOCK;
850 		} else {
851 			kp->ki_stat = SWAIT;
852 		}
853 	} else if (p->p_state == PRS_ZOMBIE) {
854 		kp->ki_stat = SZOMB;
855 	} else {
856 		kp->ki_stat = SIDL;
857 	}
858 
859 	/* Things in the thread */
860 	kp->ki_wchan = td->td_wchan;
861 	kp->ki_pri.pri_level = td->td_priority;
862 	kp->ki_pri.pri_native = td->td_base_pri;
863 	kp->ki_lastcpu = td->td_lastcpu;
864 	kp->ki_oncpu = td->td_oncpu;
865 	kp->ki_tdflags = td->td_flags;
866 	kp->ki_tid = td->td_tid;
867 	kp->ki_numthreads = p->p_numthreads;
868 	kp->ki_pcb = td->td_pcb;
869 	kp->ki_kstack = (void *)td->td_kstack;
870 	kp->ki_pctcpu = sched_pctcpu(td);
871 	kp->ki_estcpu = td->td_estcpu;
872 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
873 	kp->ki_pri.pri_class = td->td_pri_class;
874 	kp->ki_pri.pri_user = td->td_user_pri;
875 
876 	if (preferthread)
877 		kp->ki_runtime = cputick2usec(td->td_runtime);
878 
879 	/* We can't get this anymore but ps etc never used it anyway. */
880 	kp->ki_rqindex = 0;
881 
882 	SIGSETOR(kp->ki_siglist, td->td_siglist);
883 	kp->ki_sigmask = td->td_sigmask;
884 	thread_unlock(td);
885 }
886 
887 /*
888  * Fill in a kinfo_proc structure for the specified process.
889  * Must be called with the target process locked.
890  */
891 void
892 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
893 {
894 
895 	fill_kinfo_proc_only(p, kp);
896 	if (FIRST_THREAD_IN_PROC(p) != NULL)
897 		fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
898 }
899 
900 struct pstats *
901 pstats_alloc(void)
902 {
903 
904 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
905 }
906 
907 /*
908  * Copy parts of p_stats; zero the rest of p_stats (statistics).
909  */
910 void
911 pstats_fork(struct pstats *src, struct pstats *dst)
912 {
913 
914 	bzero(&dst->pstat_startzero,
915 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
916 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
917 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
918 }
919 
920 void
921 pstats_free(struct pstats *ps)
922 {
923 
924 	free(ps, M_SUBPROC);
925 }
926 
927 /*
928  * Locate a zombie process by number
929  */
930 struct proc *
931 zpfind(pid_t pid)
932 {
933 	struct proc *p;
934 
935 	sx_slock(&allproc_lock);
936 	LIST_FOREACH(p, &zombproc, p_list)
937 		if (p->p_pid == pid) {
938 			PROC_LOCK(p);
939 			break;
940 		}
941 	sx_sunlock(&allproc_lock);
942 	return (p);
943 }
944 
945 #define KERN_PROC_ZOMBMASK	0x3
946 #define KERN_PROC_NOTHREADS	0x4
947 
948 /*
949  * Must be called with the process locked and will return with it unlocked.
950  */
951 static int
952 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
953 {
954 	struct thread *td;
955 	struct kinfo_proc kinfo_proc;
956 	int error = 0;
957 	struct proc *np;
958 	pid_t pid = p->p_pid;
959 
960 	PROC_LOCK_ASSERT(p, MA_OWNED);
961 
962 	fill_kinfo_proc_only(p, &kinfo_proc);
963 	if (flags & KERN_PROC_NOTHREADS) {
964 		if (FIRST_THREAD_IN_PROC(p) != NULL)
965 			fill_kinfo_thread(FIRST_THREAD_IN_PROC(p),
966 			    &kinfo_proc, 0);
967 		error = SYSCTL_OUT(req, (caddr_t)&kinfo_proc,
968 				   sizeof(kinfo_proc));
969 	} else {
970 		if (FIRST_THREAD_IN_PROC(p) != NULL)
971 			FOREACH_THREAD_IN_PROC(p, td) {
972 				fill_kinfo_thread(td, &kinfo_proc, 1);
973 				error = SYSCTL_OUT(req, (caddr_t)&kinfo_proc,
974 						   sizeof(kinfo_proc));
975 				if (error)
976 					break;
977 			}
978 		else
979 			error = SYSCTL_OUT(req, (caddr_t)&kinfo_proc,
980 					   sizeof(kinfo_proc));
981 	}
982 	PROC_UNLOCK(p);
983 	if (error)
984 		return (error);
985 	if (flags & KERN_PROC_ZOMBMASK)
986 		np = zpfind(pid);
987 	else {
988 		if (pid == 0)
989 			return (0);
990 		np = pfind(pid);
991 	}
992 	if (np == NULL)
993 		return (ESRCH);
994 	if (np != p) {
995 		PROC_UNLOCK(np);
996 		return (ESRCH);
997 	}
998 	PROC_UNLOCK(np);
999 	return (0);
1000 }
1001 
1002 static int
1003 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1004 {
1005 	int *name = (int*) arg1;
1006 	u_int namelen = arg2;
1007 	struct proc *p;
1008 	int flags, doingzomb, oid_number;
1009 	int error = 0;
1010 
1011 	oid_number = oidp->oid_number;
1012 	if (oid_number != KERN_PROC_ALL &&
1013 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1014 		flags = KERN_PROC_NOTHREADS;
1015 	else {
1016 		flags = 0;
1017 		oid_number &= ~KERN_PROC_INC_THREAD;
1018 	}
1019 	if (oid_number == KERN_PROC_PID) {
1020 		if (namelen != 1)
1021 			return (EINVAL);
1022 		error = sysctl_wire_old_buffer(req, 0);
1023 		if (error)
1024 			return (error);
1025 		p = pfind((pid_t)name[0]);
1026 		if (!p)
1027 			return (ESRCH);
1028 		if ((error = p_cansee(curthread, p))) {
1029 			PROC_UNLOCK(p);
1030 			return (error);
1031 		}
1032 		error = sysctl_out_proc(p, req, flags);
1033 		return (error);
1034 	}
1035 
1036 	switch (oid_number) {
1037 	case KERN_PROC_ALL:
1038 		if (namelen != 0)
1039 			return (EINVAL);
1040 		break;
1041 	case KERN_PROC_PROC:
1042 		if (namelen != 0 && namelen != 1)
1043 			return (EINVAL);
1044 		break;
1045 	default:
1046 		if (namelen != 1)
1047 			return (EINVAL);
1048 		break;
1049 	}
1050 
1051 	if (!req->oldptr) {
1052 		/* overestimate by 5 procs */
1053 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1054 		if (error)
1055 			return (error);
1056 	}
1057 	error = sysctl_wire_old_buffer(req, 0);
1058 	if (error != 0)
1059 		return (error);
1060 	sx_slock(&allproc_lock);
1061 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1062 		if (!doingzomb)
1063 			p = LIST_FIRST(&allproc);
1064 		else
1065 			p = LIST_FIRST(&zombproc);
1066 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
1067 			/*
1068 			 * Skip embryonic processes.
1069 			 */
1070 			PROC_SLOCK(p);
1071 			if (p->p_state == PRS_NEW) {
1072 				PROC_SUNLOCK(p);
1073 				continue;
1074 			}
1075 			PROC_SUNLOCK(p);
1076 			PROC_LOCK(p);
1077 			KASSERT(p->p_ucred != NULL,
1078 			    ("process credential is NULL for non-NEW proc"));
1079 			/*
1080 			 * Show a user only appropriate processes.
1081 			 */
1082 			if (p_cansee(curthread, p)) {
1083 				PROC_UNLOCK(p);
1084 				continue;
1085 			}
1086 			/*
1087 			 * TODO - make more efficient (see notes below).
1088 			 * do by session.
1089 			 */
1090 			switch (oid_number) {
1091 
1092 			case KERN_PROC_GID:
1093 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1094 					PROC_UNLOCK(p);
1095 					continue;
1096 				}
1097 				break;
1098 
1099 			case KERN_PROC_PGRP:
1100 				/* could do this by traversing pgrp */
1101 				if (p->p_pgrp == NULL ||
1102 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1103 					PROC_UNLOCK(p);
1104 					continue;
1105 				}
1106 				break;
1107 
1108 			case KERN_PROC_RGID:
1109 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1110 					PROC_UNLOCK(p);
1111 					continue;
1112 				}
1113 				break;
1114 
1115 			case KERN_PROC_SESSION:
1116 				if (p->p_session == NULL ||
1117 				    p->p_session->s_sid != (pid_t)name[0]) {
1118 					PROC_UNLOCK(p);
1119 					continue;
1120 				}
1121 				break;
1122 
1123 			case KERN_PROC_TTY:
1124 				if ((p->p_flag & P_CONTROLT) == 0 ||
1125 				    p->p_session == NULL) {
1126 					PROC_UNLOCK(p);
1127 					continue;
1128 				}
1129 				/* XXX proctree_lock */
1130 				SESS_LOCK(p->p_session);
1131 				if (p->p_session->s_ttyp == NULL ||
1132 				    tty_udev(p->p_session->s_ttyp) !=
1133 				    (dev_t)name[0]) {
1134 					SESS_UNLOCK(p->p_session);
1135 					PROC_UNLOCK(p);
1136 					continue;
1137 				}
1138 				SESS_UNLOCK(p->p_session);
1139 				break;
1140 
1141 			case KERN_PROC_UID:
1142 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1143 					PROC_UNLOCK(p);
1144 					continue;
1145 				}
1146 				break;
1147 
1148 			case KERN_PROC_RUID:
1149 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1150 					PROC_UNLOCK(p);
1151 					continue;
1152 				}
1153 				break;
1154 
1155 			case KERN_PROC_PROC:
1156 				break;
1157 
1158 			default:
1159 				break;
1160 
1161 			}
1162 
1163 			error = sysctl_out_proc(p, req, flags | doingzomb);
1164 			if (error) {
1165 				sx_sunlock(&allproc_lock);
1166 				return (error);
1167 			}
1168 		}
1169 	}
1170 	sx_sunlock(&allproc_lock);
1171 	return (0);
1172 }
1173 
1174 struct pargs *
1175 pargs_alloc(int len)
1176 {
1177 	struct pargs *pa;
1178 
1179 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1180 		M_WAITOK);
1181 	refcount_init(&pa->ar_ref, 1);
1182 	pa->ar_length = len;
1183 	return (pa);
1184 }
1185 
1186 static void
1187 pargs_free(struct pargs *pa)
1188 {
1189 
1190 	free(pa, M_PARGS);
1191 }
1192 
1193 void
1194 pargs_hold(struct pargs *pa)
1195 {
1196 
1197 	if (pa == NULL)
1198 		return;
1199 	refcount_acquire(&pa->ar_ref);
1200 }
1201 
1202 void
1203 pargs_drop(struct pargs *pa)
1204 {
1205 
1206 	if (pa == NULL)
1207 		return;
1208 	if (refcount_release(&pa->ar_ref))
1209 		pargs_free(pa);
1210 }
1211 
1212 /*
1213  * This sysctl allows a process to retrieve the argument list or process
1214  * title for another process without groping around in the address space
1215  * of the other process.  It also allow a process to set its own "process
1216  * title to a string of its own choice.
1217  */
1218 static int
1219 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1220 {
1221 	int *name = (int*) arg1;
1222 	u_int namelen = arg2;
1223 	struct pargs *newpa, *pa;
1224 	struct proc *p;
1225 	int error = 0;
1226 
1227 	if (namelen != 1)
1228 		return (EINVAL);
1229 
1230 	p = pfind((pid_t)name[0]);
1231 	if (!p)
1232 		return (ESRCH);
1233 
1234 	if ((error = p_cansee(curthread, p)) != 0) {
1235 		PROC_UNLOCK(p);
1236 		return (error);
1237 	}
1238 
1239 	if (req->newptr && curproc != p) {
1240 		PROC_UNLOCK(p);
1241 		return (EPERM);
1242 	}
1243 
1244 	pa = p->p_args;
1245 	pargs_hold(pa);
1246 	PROC_UNLOCK(p);
1247 	if (req->oldptr != NULL && pa != NULL)
1248 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1249 	pargs_drop(pa);
1250 	if (error != 0 || req->newptr == NULL)
1251 		return (error);
1252 
1253 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1254 		return (ENOMEM);
1255 	newpa = pargs_alloc(req->newlen);
1256 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1257 	if (error != 0) {
1258 		pargs_free(newpa);
1259 		return (error);
1260 	}
1261 	PROC_LOCK(p);
1262 	pa = p->p_args;
1263 	p->p_args = newpa;
1264 	PROC_UNLOCK(p);
1265 	pargs_drop(pa);
1266 	return (0);
1267 }
1268 
1269 /*
1270  * This sysctl allows a process to retrieve the path of the executable for
1271  * itself or another process.
1272  */
1273 static int
1274 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1275 {
1276 	pid_t *pidp = (pid_t *)arg1;
1277 	unsigned int arglen = arg2;
1278 	struct proc *p;
1279 	struct vnode *vp;
1280 	char *retbuf, *freebuf;
1281 	int error;
1282 
1283 	if (arglen != 1)
1284 		return (EINVAL);
1285 	if (*pidp == -1) {	/* -1 means this process */
1286 		p = req->td->td_proc;
1287 	} else {
1288 		p = pfind(*pidp);
1289 		if (p == NULL)
1290 			return (ESRCH);
1291 		if ((error = p_cansee(curthread, p)) != 0) {
1292 			PROC_UNLOCK(p);
1293 			return (error);
1294 		}
1295 	}
1296 
1297 	vp = p->p_textvp;
1298 	if (vp == NULL) {
1299 		if (*pidp != -1)
1300 			PROC_UNLOCK(p);
1301 		return (0);
1302 	}
1303 	vref(vp);
1304 	if (*pidp != -1)
1305 		PROC_UNLOCK(p);
1306 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1307 	vrele(vp);
1308 	if (error)
1309 		return (error);
1310 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1311 	free(freebuf, M_TEMP);
1312 	return (error);
1313 }
1314 
1315 static int
1316 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1317 {
1318 	struct proc *p;
1319 	char *sv_name;
1320 	int *name;
1321 	int namelen;
1322 	int error;
1323 
1324 	namelen = arg2;
1325 	if (namelen != 1)
1326 		return (EINVAL);
1327 
1328 	name = (int *)arg1;
1329 	if ((p = pfind((pid_t)name[0])) == NULL)
1330 		return (ESRCH);
1331 	if ((error = p_cansee(curthread, p))) {
1332 		PROC_UNLOCK(p);
1333 		return (error);
1334 	}
1335 	sv_name = p->p_sysent->sv_name;
1336 	PROC_UNLOCK(p);
1337 	return (sysctl_handle_string(oidp, sv_name, 0, req));
1338 }
1339 
1340 static int
1341 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
1342 {
1343 	vm_map_entry_t entry, tmp_entry;
1344 	unsigned int last_timestamp;
1345 	char *fullpath, *freepath;
1346 	struct kinfo_vmentry *kve;
1347 	struct vattr va;
1348 	struct ucred *cred;
1349 	int error, *name;
1350 	struct vnode *vp;
1351 	struct proc *p;
1352 	vm_map_t map;
1353 
1354 	name = (int *)arg1;
1355 	if ((p = pfind((pid_t)name[0])) == NULL)
1356 		return (ESRCH);
1357 	if (p->p_flag & P_WEXIT) {
1358 		PROC_UNLOCK(p);
1359 		return (ESRCH);
1360 	}
1361 	if ((error = p_candebug(curthread, p))) {
1362 		PROC_UNLOCK(p);
1363 		return (error);
1364 	}
1365 	_PHOLD(p);
1366 	PROC_UNLOCK(p);
1367 
1368 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1369 
1370 	map = &p->p_vmspace->vm_map;	/* XXXRW: More locking required? */
1371 	vm_map_lock_read(map);
1372 	for (entry = map->header.next; entry != &map->header;
1373 	    entry = entry->next) {
1374 		vm_object_t obj, tobj, lobj;
1375 		vm_offset_t addr;
1376 		int vfslocked;
1377 
1378 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
1379 			continue;
1380 
1381 		bzero(kve, sizeof(*kve));
1382 		kve->kve_structsize = sizeof(*kve);
1383 
1384 		kve->kve_private_resident = 0;
1385 		obj = entry->object.vm_object;
1386 		if (obj != NULL) {
1387 			VM_OBJECT_LOCK(obj);
1388 			if (obj->shadow_count == 1)
1389 				kve->kve_private_resident =
1390 				    obj->resident_page_count;
1391 		}
1392 		kve->kve_resident = 0;
1393 		addr = entry->start;
1394 		while (addr < entry->end) {
1395 			if (pmap_extract(map->pmap, addr))
1396 				kve->kve_resident++;
1397 			addr += PAGE_SIZE;
1398 		}
1399 
1400 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
1401 			if (tobj != obj)
1402 				VM_OBJECT_LOCK(tobj);
1403 			if (lobj != obj)
1404 				VM_OBJECT_UNLOCK(lobj);
1405 			lobj = tobj;
1406 		}
1407 
1408 		kve->kve_fileid = 0;
1409 		kve->kve_fsid = 0;
1410 		freepath = NULL;
1411 		fullpath = "";
1412 		if (lobj) {
1413 			vp = NULL;
1414 			switch(lobj->type) {
1415 			case OBJT_DEFAULT:
1416 				kve->kve_type = KVME_TYPE_DEFAULT;
1417 				break;
1418 			case OBJT_VNODE:
1419 				kve->kve_type = KVME_TYPE_VNODE;
1420 				vp = lobj->handle;
1421 				vref(vp);
1422 				break;
1423 			case OBJT_SWAP:
1424 				kve->kve_type = KVME_TYPE_SWAP;
1425 				break;
1426 			case OBJT_DEVICE:
1427 				kve->kve_type = KVME_TYPE_DEVICE;
1428 				break;
1429 			case OBJT_PHYS:
1430 				kve->kve_type = KVME_TYPE_PHYS;
1431 				break;
1432 			case OBJT_DEAD:
1433 				kve->kve_type = KVME_TYPE_DEAD;
1434 				break;
1435 			default:
1436 				kve->kve_type = KVME_TYPE_UNKNOWN;
1437 				break;
1438 			}
1439 			if (lobj != obj)
1440 				VM_OBJECT_UNLOCK(lobj);
1441 
1442 			kve->kve_ref_count = obj->ref_count;
1443 			kve->kve_shadow_count = obj->shadow_count;
1444 			VM_OBJECT_UNLOCK(obj);
1445 			if (vp != NULL) {
1446 				vn_fullpath(curthread, vp, &fullpath,
1447 				    &freepath);
1448 				cred = curthread->td_ucred;
1449 				vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1450 				vn_lock(vp, LK_SHARED | LK_RETRY);
1451 				if (VOP_GETATTR(vp, &va, cred) == 0) {
1452 					kve->kve_fileid = va.va_fileid;
1453 					kve->kve_fsid = va.va_fsid;
1454 				}
1455 				vput(vp);
1456 				VFS_UNLOCK_GIANT(vfslocked);
1457 			}
1458 		} else {
1459 			kve->kve_type = KVME_TYPE_NONE;
1460 			kve->kve_ref_count = 0;
1461 			kve->kve_shadow_count = 0;
1462 		}
1463 
1464 		kve->kve_start = (void*)entry->start;
1465 		kve->kve_end = (void*)entry->end;
1466 		kve->kve_offset = (off_t)entry->offset;
1467 
1468 		if (entry->protection & VM_PROT_READ)
1469 			kve->kve_protection |= KVME_PROT_READ;
1470 		if (entry->protection & VM_PROT_WRITE)
1471 			kve->kve_protection |= KVME_PROT_WRITE;
1472 		if (entry->protection & VM_PROT_EXECUTE)
1473 			kve->kve_protection |= KVME_PROT_EXEC;
1474 
1475 		if (entry->eflags & MAP_ENTRY_COW)
1476 			kve->kve_flags |= KVME_FLAG_COW;
1477 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
1478 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
1479 
1480 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
1481 		if (freepath != NULL)
1482 			free(freepath, M_TEMP);
1483 
1484 		last_timestamp = map->timestamp;
1485 		vm_map_unlock_read(map);
1486 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
1487 		vm_map_lock_read(map);
1488 		if (error)
1489 			break;
1490 		if (last_timestamp + 1 != map->timestamp) {
1491 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
1492 			entry = tmp_entry;
1493 		}
1494 	}
1495 	vm_map_unlock_read(map);
1496 	PRELE(p);
1497 	free(kve, M_TEMP);
1498 	return (error);
1499 }
1500 
1501 #if defined(STACK) || defined(DDB)
1502 static int
1503 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
1504 {
1505 	struct kinfo_kstack *kkstp;
1506 	int error, i, *name, numthreads;
1507 	lwpid_t *lwpidarray;
1508 	struct thread *td;
1509 	struct stack *st;
1510 	struct sbuf sb;
1511 	struct proc *p;
1512 
1513 	name = (int *)arg1;
1514 	if ((p = pfind((pid_t)name[0])) == NULL)
1515 		return (ESRCH);
1516 	/* XXXRW: Not clear ESRCH is the right error during proc execve(). */
1517 	if (p->p_flag & P_WEXIT || p->p_flag & P_INEXEC) {
1518 		PROC_UNLOCK(p);
1519 		return (ESRCH);
1520 	}
1521 	if ((error = p_candebug(curthread, p))) {
1522 		PROC_UNLOCK(p);
1523 		return (error);
1524 	}
1525 	_PHOLD(p);
1526 	PROC_UNLOCK(p);
1527 
1528 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
1529 	st = stack_create();
1530 
1531 	lwpidarray = NULL;
1532 	numthreads = 0;
1533 	PROC_LOCK(p);
1534 repeat:
1535 	if (numthreads < p->p_numthreads) {
1536 		if (lwpidarray != NULL) {
1537 			free(lwpidarray, M_TEMP);
1538 			lwpidarray = NULL;
1539 		}
1540 		numthreads = p->p_numthreads;
1541 		PROC_UNLOCK(p);
1542 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
1543 		    M_WAITOK | M_ZERO);
1544 		PROC_LOCK(p);
1545 		goto repeat;
1546 	}
1547 	i = 0;
1548 
1549 	/*
1550 	 * XXXRW: During the below loop, execve(2) and countless other sorts
1551 	 * of changes could have taken place.  Should we check to see if the
1552 	 * vmspace has been replaced, or the like, in order to prevent
1553 	 * giving a snapshot that spans, say, execve(2), with some threads
1554 	 * before and some after?  Among other things, the credentials could
1555 	 * have changed, in which case the right to extract debug info might
1556 	 * no longer be assured.
1557 	 */
1558 	FOREACH_THREAD_IN_PROC(p, td) {
1559 		KASSERT(i < numthreads,
1560 		    ("sysctl_kern_proc_kstack: numthreads"));
1561 		lwpidarray[i] = td->td_tid;
1562 		i++;
1563 	}
1564 	numthreads = i;
1565 	for (i = 0; i < numthreads; i++) {
1566 		td = thread_find(p, lwpidarray[i]);
1567 		if (td == NULL) {
1568 			continue;
1569 		}
1570 		bzero(kkstp, sizeof(*kkstp));
1571 		(void)sbuf_new(&sb, kkstp->kkst_trace,
1572 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
1573 		thread_lock(td);
1574 		kkstp->kkst_tid = td->td_tid;
1575 		if (TD_IS_SWAPPED(td))
1576 			kkstp->kkst_state = KKST_STATE_SWAPPED;
1577 		else if (TD_IS_RUNNING(td))
1578 			kkstp->kkst_state = KKST_STATE_RUNNING;
1579 		else {
1580 			kkstp->kkst_state = KKST_STATE_STACKOK;
1581 			stack_save_td(st, td);
1582 		}
1583 		thread_unlock(td);
1584 		PROC_UNLOCK(p);
1585 		stack_sbuf_print(&sb, st);
1586 		sbuf_finish(&sb);
1587 		sbuf_delete(&sb);
1588 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
1589 		PROC_LOCK(p);
1590 		if (error)
1591 			break;
1592 	}
1593 	_PRELE(p);
1594 	PROC_UNLOCK(p);
1595 	if (lwpidarray != NULL)
1596 		free(lwpidarray, M_TEMP);
1597 	stack_destroy(st);
1598 	free(kkstp, M_TEMP);
1599 	return (error);
1600 }
1601 #endif
1602 
1603 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
1604 
1605 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT,
1606 	0, 0, sysctl_kern_proc, "S,proc", "Return entire process table");
1607 
1608 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD,
1609 	sysctl_kern_proc, "Process table");
1610 
1611 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD,
1612 	sysctl_kern_proc, "Process table");
1613 
1614 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD,
1615 	sysctl_kern_proc, "Process table");
1616 
1617 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD,
1618 	sysctl_kern_proc, "Process table");
1619 
1620 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD,
1621 	sysctl_kern_proc, "Process table");
1622 
1623 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD,
1624 	sysctl_kern_proc, "Process table");
1625 
1626 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD,
1627 	sysctl_kern_proc, "Process table");
1628 
1629 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD,
1630 	sysctl_kern_proc, "Process table");
1631 
1632 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD,
1633 	sysctl_kern_proc, "Return process table, no threads");
1634 
1635 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
1636 	CTLFLAG_RW | CTLFLAG_ANYBODY,
1637 	sysctl_kern_proc_args, "Process argument list");
1638 
1639 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD,
1640 	sysctl_kern_proc_pathname, "Process executable path");
1641 
1642 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD,
1643 	sysctl_kern_proc_sv_name, "Process syscall vector name (ABI type)");
1644 
1645 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
1646 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1647 
1648 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
1649 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1650 
1651 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
1652 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1653 
1654 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
1655 	sid_td, CTLFLAG_RD, sysctl_kern_proc, "Process table");
1656 
1657 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
1658 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1659 
1660 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
1661 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1662 
1663 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
1664 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1665 
1666 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
1667 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1668 
1669 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
1670 	CTLFLAG_RD, sysctl_kern_proc, "Return process table, no threads");
1671 
1672 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD,
1673 	sysctl_kern_proc_vmmap, "Process vm map entries");
1674 
1675 #if defined(STACK) || defined(DDB)
1676 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD,
1677 	sysctl_kern_proc_kstack, "Process kernel stacks");
1678 #endif
1679