xref: /freebsd/sys/kern/kern_proc.c (revision 7562eaabc01a48e6b11d5b558c41e3b92dae5c2d)
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  * $FreeBSD$
31  */
32 
33 #include <sys/cdefs.h>
34 __FBSDID("$FreeBSD$");
35 
36 #include "opt_ktrace.h"
37 #include "opt_kstack_pages.h"
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/kernel.h>
42 #include <sys/lock.h>
43 #include <sys/malloc.h>
44 #include <sys/mutex.h>
45 #include <sys/proc.h>
46 #include <sys/sysent.h>
47 #include <sys/sched.h>
48 #include <sys/smp.h>
49 #include <sys/sysctl.h>
50 #include <sys/filedesc.h>
51 #include <sys/tty.h>
52 #include <sys/signalvar.h>
53 #include <sys/sx.h>
54 #include <sys/user.h>
55 #include <sys/jail.h>
56 #ifdef KTRACE
57 #include <sys/uio.h>
58 #include <sys/ktrace.h>
59 #endif
60 
61 #include <vm/vm.h>
62 #include <vm/vm_extern.h>
63 #include <vm/pmap.h>
64 #include <vm/vm_map.h>
65 #include <vm/uma.h>
66 #include <machine/critical.h>
67 
68 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
69 MALLOC_DEFINE(M_SESSION, "session", "session header");
70 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
71 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
72 
73 static void doenterpgrp(struct proc *, struct pgrp *);
74 static void orphanpg(struct pgrp *pg);
75 static void pgadjustjobc(struct pgrp *pgrp, int entering);
76 static void pgdelete(struct pgrp *);
77 static int proc_ctor(void *mem, int size, void *arg, int flags);
78 static void proc_dtor(void *mem, int size, void *arg);
79 static int proc_init(void *mem, int size, int flags);
80 static void proc_fini(void *mem, int size);
81 
82 /*
83  * Other process lists
84  */
85 struct pidhashhead *pidhashtbl;
86 u_long pidhash;
87 struct pgrphashhead *pgrphashtbl;
88 u_long pgrphash;
89 struct proclist allproc;
90 struct proclist zombproc;
91 struct sx allproc_lock;
92 struct sx proctree_lock;
93 struct mtx pargs_ref_lock;
94 struct mtx ppeers_lock;
95 uma_zone_t proc_zone;
96 uma_zone_t ithread_zone;
97 
98 int kstack_pages = KSTACK_PAGES;
99 int uarea_pages = UAREA_PAGES;
100 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0, "");
101 SYSCTL_INT(_kern, OID_AUTO, uarea_pages, CTLFLAG_RD, &uarea_pages, 0, "");
102 
103 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
104 
105 /*
106  * Initialize global process hashing structures.
107  */
108 void
109 procinit()
110 {
111 
112 	sx_init(&allproc_lock, "allproc");
113 	sx_init(&proctree_lock, "proctree");
114 	mtx_init(&pargs_ref_lock, "struct pargs.ref", NULL, MTX_DEF);
115 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
116 	LIST_INIT(&allproc);
117 	LIST_INIT(&zombproc);
118 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
119 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
120 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
121 	    proc_ctor, proc_dtor, proc_init, proc_fini,
122 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
123 	uihashinit();
124 }
125 
126 /*
127  * Prepare a proc for use.
128  */
129 static int
130 proc_ctor(void *mem, int size, void *arg, int flags)
131 {
132 	struct proc *p;
133 
134 	p = (struct proc *)mem;
135 	return (0);
136 }
137 
138 /*
139  * Reclaim a proc after use.
140  */
141 static void
142 proc_dtor(void *mem, int size, void *arg)
143 {
144 	struct proc *p;
145 	struct thread *td;
146 #ifdef INVARIANTS
147 	struct ksegrp *kg;
148 #endif
149 
150 	/* INVARIANTS checks go here */
151 	p = (struct proc *)mem;
152         td = FIRST_THREAD_IN_PROC(p);
153 #ifdef INVARIANTS
154 	KASSERT((p->p_numthreads == 1),
155 	    ("bad number of threads in exiting process"));
156 	KASSERT((p->p_numksegrps == 1), ("free proc with > 1 ksegrp"));
157 	KASSERT((td != NULL), ("proc_dtor: bad thread pointer"));
158         kg = FIRST_KSEGRP_IN_PROC(p);
159 	KASSERT((kg != NULL), ("proc_dtor: bad kg pointer"));
160 #endif
161 
162 	/* Dispose of an alternate kstack, if it exists.
163 	 * XXX What if there are more than one thread in the proc?
164 	 *     The first thread in the proc is special and not
165 	 *     freed, so you gotta do this here.
166 	 */
167 	if (((p->p_flag & P_KTHREAD) != 0) && (td->td_altkstack != 0))
168 		vm_thread_dispose_altkstack(td);
169 }
170 
171 /*
172  * Initialize type-stable parts of a proc (when newly created).
173  */
174 static int
175 proc_init(void *mem, int size, int flags)
176 {
177 	struct proc *p;
178 	struct thread *td;
179 	struct ksegrp *kg;
180 
181 	p = (struct proc *)mem;
182 	p->p_sched = (struct p_sched *)&p[1];
183 	vm_proc_new(p);
184 	td = thread_alloc();
185 	kg = ksegrp_alloc();
186 	bzero(&p->p_mtx, sizeof(struct mtx));
187 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK);
188 	proc_linkup(p, kg, td);
189 	sched_newproc(p, kg, td);
190 	return (0);
191 }
192 
193 /*
194  * UMA should ensure that this function is never called.
195  * Freeing a proc structure would violate type stability.
196  */
197 static void
198 proc_fini(void *mem, int size)
199 {
200 
201 	panic("proc reclaimed");
202 }
203 
204 /*
205  * Is p an inferior of the current process?
206  */
207 int
208 inferior(p)
209 	register struct proc *p;
210 {
211 
212 	sx_assert(&proctree_lock, SX_LOCKED);
213 	for (; p != curproc; p = p->p_pptr)
214 		if (p->p_pid == 0)
215 			return (0);
216 	return (1);
217 }
218 
219 /*
220  * Locate a process by number; return only "live" processes -- i.e., neither
221  * zombies nor newly born but incompletely initialized processes.  By not
222  * returning processes in the PRS_NEW state, we allow callers to avoid
223  * testing for that condition to avoid dereferencing p_ucred, et al.
224  */
225 struct proc *
226 pfind(pid)
227 	register pid_t pid;
228 {
229 	register struct proc *p;
230 
231 	sx_slock(&allproc_lock);
232 	LIST_FOREACH(p, PIDHASH(pid), p_hash)
233 		if (p->p_pid == pid) {
234 			if (p->p_state == PRS_NEW) {
235 				p = NULL;
236 				break;
237 			}
238 			PROC_LOCK(p);
239 			break;
240 		}
241 	sx_sunlock(&allproc_lock);
242 	return (p);
243 }
244 
245 /*
246  * Locate a process group by number.
247  * The caller must hold proctree_lock.
248  */
249 struct pgrp *
250 pgfind(pgid)
251 	register pid_t pgid;
252 {
253 	register struct pgrp *pgrp;
254 
255 	sx_assert(&proctree_lock, SX_LOCKED);
256 
257 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
258 		if (pgrp->pg_id == pgid) {
259 			PGRP_LOCK(pgrp);
260 			return (pgrp);
261 		}
262 	}
263 	return (NULL);
264 }
265 
266 /*
267  * Create a new process group.
268  * pgid must be equal to the pid of p.
269  * Begin a new session if required.
270  */
271 int
272 enterpgrp(p, pgid, pgrp, sess)
273 	register struct proc *p;
274 	pid_t pgid;
275 	struct pgrp *pgrp;
276 	struct session *sess;
277 {
278 	struct pgrp *pgrp2;
279 
280 	sx_assert(&proctree_lock, SX_XLOCKED);
281 
282 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
283 	KASSERT(p->p_pid == pgid,
284 	    ("enterpgrp: new pgrp and pid != pgid"));
285 
286 	pgrp2 = pgfind(pgid);
287 
288 	KASSERT(pgrp2 == NULL,
289 	    ("enterpgrp: pgrp with pgid exists"));
290 	KASSERT(!SESS_LEADER(p),
291 	    ("enterpgrp: session leader attempted setpgrp"));
292 
293 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
294 
295 	if (sess != NULL) {
296 		/*
297 		 * new session
298 		 */
299 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
300 		PROC_LOCK(p);
301 		p->p_flag &= ~P_CONTROLT;
302 		PROC_UNLOCK(p);
303 		PGRP_LOCK(pgrp);
304 		sess->s_leader = p;
305 		sess->s_sid = p->p_pid;
306 		sess->s_count = 1;
307 		sess->s_ttyvp = NULL;
308 		sess->s_ttyp = NULL;
309 		bcopy(p->p_session->s_login, sess->s_login,
310 			    sizeof(sess->s_login));
311 		pgrp->pg_session = sess;
312 		KASSERT(p == curproc,
313 		    ("enterpgrp: mksession and p != curproc"));
314 	} else {
315 		pgrp->pg_session = p->p_session;
316 		SESS_LOCK(pgrp->pg_session);
317 		pgrp->pg_session->s_count++;
318 		SESS_UNLOCK(pgrp->pg_session);
319 		PGRP_LOCK(pgrp);
320 	}
321 	pgrp->pg_id = pgid;
322 	LIST_INIT(&pgrp->pg_members);
323 
324 	/*
325 	 * As we have an exclusive lock of proctree_lock,
326 	 * this should not deadlock.
327 	 */
328 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
329 	pgrp->pg_jobc = 0;
330 	SLIST_INIT(&pgrp->pg_sigiolst);
331 	PGRP_UNLOCK(pgrp);
332 
333 	doenterpgrp(p, pgrp);
334 
335 	return (0);
336 }
337 
338 /*
339  * Move p to an existing process group
340  */
341 int
342 enterthispgrp(p, pgrp)
343 	register struct proc *p;
344 	struct pgrp *pgrp;
345 {
346 
347 	sx_assert(&proctree_lock, SX_XLOCKED);
348 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
349 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
350 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
351 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
352 	KASSERT(pgrp->pg_session == p->p_session,
353 		("%s: pgrp's session %p, p->p_session %p.\n",
354 		__func__,
355 		pgrp->pg_session,
356 		p->p_session));
357 	KASSERT(pgrp != p->p_pgrp,
358 		("%s: p belongs to pgrp.", __func__));
359 
360 	doenterpgrp(p, pgrp);
361 
362 	return (0);
363 }
364 
365 /*
366  * Move p to a process group
367  */
368 static void
369 doenterpgrp(p, pgrp)
370 	struct proc *p;
371 	struct pgrp *pgrp;
372 {
373 	struct pgrp *savepgrp;
374 
375 	sx_assert(&proctree_lock, SX_XLOCKED);
376 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
377 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
378 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
379 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
380 
381 	savepgrp = p->p_pgrp;
382 
383 	/*
384 	 * Adjust eligibility of affected pgrps to participate in job control.
385 	 * Increment eligibility counts before decrementing, otherwise we
386 	 * could reach 0 spuriously during the first call.
387 	 */
388 	fixjobc(p, pgrp, 1);
389 	fixjobc(p, p->p_pgrp, 0);
390 
391 	PGRP_LOCK(pgrp);
392 	PGRP_LOCK(savepgrp);
393 	PROC_LOCK(p);
394 	LIST_REMOVE(p, p_pglist);
395 	p->p_pgrp = pgrp;
396 	PROC_UNLOCK(p);
397 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
398 	PGRP_UNLOCK(savepgrp);
399 	PGRP_UNLOCK(pgrp);
400 	if (LIST_EMPTY(&savepgrp->pg_members))
401 		pgdelete(savepgrp);
402 }
403 
404 /*
405  * remove process from process group
406  */
407 int
408 leavepgrp(p)
409 	register struct proc *p;
410 {
411 	struct pgrp *savepgrp;
412 
413 	sx_assert(&proctree_lock, SX_XLOCKED);
414 	savepgrp = p->p_pgrp;
415 	PGRP_LOCK(savepgrp);
416 	PROC_LOCK(p);
417 	LIST_REMOVE(p, p_pglist);
418 	p->p_pgrp = NULL;
419 	PROC_UNLOCK(p);
420 	PGRP_UNLOCK(savepgrp);
421 	if (LIST_EMPTY(&savepgrp->pg_members))
422 		pgdelete(savepgrp);
423 	return (0);
424 }
425 
426 /*
427  * delete a process group
428  */
429 static void
430 pgdelete(pgrp)
431 	register struct pgrp *pgrp;
432 {
433 	struct session *savesess;
434 	int i;
435 
436 	sx_assert(&proctree_lock, SX_XLOCKED);
437 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
438 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
439 
440 	/*
441 	 * Reset any sigio structures pointing to us as a result of
442 	 * F_SETOWN with our pgid.
443 	 */
444 	funsetownlst(&pgrp->pg_sigiolst);
445 
446 	PGRP_LOCK(pgrp);
447 	if (pgrp->pg_session->s_ttyp != NULL &&
448 	    pgrp->pg_session->s_ttyp->t_pgrp == pgrp)
449 		pgrp->pg_session->s_ttyp->t_pgrp = NULL;
450 	LIST_REMOVE(pgrp, pg_hash);
451 	savesess = pgrp->pg_session;
452 	SESS_LOCK(savesess);
453 	i = --savesess->s_count;
454 	SESS_UNLOCK(savesess);
455 	PGRP_UNLOCK(pgrp);
456 	if (i == 0) {
457 		if (savesess->s_ttyp != NULL)
458 			ttyrel(savesess->s_ttyp);
459 		mtx_destroy(&savesess->s_mtx);
460 		FREE(savesess, M_SESSION);
461 	}
462 	mtx_destroy(&pgrp->pg_mtx);
463 	FREE(pgrp, M_PGRP);
464 }
465 
466 static void
467 pgadjustjobc(pgrp, entering)
468 	struct pgrp *pgrp;
469 	int entering;
470 {
471 
472 	PGRP_LOCK(pgrp);
473 	if (entering)
474 		pgrp->pg_jobc++;
475 	else {
476 		--pgrp->pg_jobc;
477 		if (pgrp->pg_jobc == 0)
478 			orphanpg(pgrp);
479 	}
480 	PGRP_UNLOCK(pgrp);
481 }
482 
483 /*
484  * Adjust pgrp jobc counters when specified process changes process group.
485  * We count the number of processes in each process group that "qualify"
486  * the group for terminal job control (those with a parent in a different
487  * process group of the same session).  If that count reaches zero, the
488  * process group becomes orphaned.  Check both the specified process'
489  * process group and that of its children.
490  * entering == 0 => p is leaving specified group.
491  * entering == 1 => p is entering specified group.
492  */
493 void
494 fixjobc(p, pgrp, entering)
495 	register struct proc *p;
496 	register struct pgrp *pgrp;
497 	int entering;
498 {
499 	register struct pgrp *hispgrp;
500 	register struct session *mysession;
501 
502 	sx_assert(&proctree_lock, SX_LOCKED);
503 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
504 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
505 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
506 
507 	/*
508 	 * Check p's parent to see whether p qualifies its own process
509 	 * group; if so, adjust count for p's process group.
510 	 */
511 	mysession = pgrp->pg_session;
512 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
513 	    hispgrp->pg_session == mysession)
514 		pgadjustjobc(pgrp, entering);
515 
516 	/*
517 	 * Check this process' children to see whether they qualify
518 	 * their process groups; if so, adjust counts for children's
519 	 * process groups.
520 	 */
521 	LIST_FOREACH(p, &p->p_children, p_sibling) {
522 		hispgrp = p->p_pgrp;
523 		if (hispgrp == pgrp ||
524 		    hispgrp->pg_session != mysession)
525 			continue;
526 		PROC_LOCK(p);
527 		if (p->p_state == PRS_ZOMBIE) {
528 			PROC_UNLOCK(p);
529 			continue;
530 		}
531 		PROC_UNLOCK(p);
532 		pgadjustjobc(hispgrp, entering);
533 	}
534 }
535 
536 /*
537  * A process group has become orphaned;
538  * if there are any stopped processes in the group,
539  * hang-up all process in that group.
540  */
541 static void
542 orphanpg(pg)
543 	struct pgrp *pg;
544 {
545 	register struct proc *p;
546 
547 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
548 
549 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
550 		PROC_LOCK(p);
551 		if (P_SHOULDSTOP(p)) {
552 			PROC_UNLOCK(p);
553 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
554 				PROC_LOCK(p);
555 				psignal(p, SIGHUP);
556 				psignal(p, SIGCONT);
557 				PROC_UNLOCK(p);
558 			}
559 			return;
560 		}
561 		PROC_UNLOCK(p);
562 	}
563 }
564 
565 #include "opt_ddb.h"
566 #ifdef DDB
567 #include <ddb/ddb.h>
568 
569 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
570 {
571 	register struct pgrp *pgrp;
572 	register struct proc *p;
573 	register int i;
574 
575 	for (i = 0; i <= pgrphash; i++) {
576 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
577 			printf("\tindx %d\n", i);
578 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
579 				printf(
580 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
581 				    (void *)pgrp, (long)pgrp->pg_id,
582 				    (void *)pgrp->pg_session,
583 				    pgrp->pg_session->s_count,
584 				    (void *)LIST_FIRST(&pgrp->pg_members));
585 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
586 					printf("\t\tpid %ld addr %p pgrp %p\n",
587 					    (long)p->p_pid, (void *)p,
588 					    (void *)p->p_pgrp);
589 				}
590 			}
591 		}
592 	}
593 }
594 #endif /* DDB */
595 void
596 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp);
597 
598 /*
599  * Fill in a kinfo_proc structure for the specified process.
600  * Must be called with the target process locked.
601  */
602 void
603 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
604 {
605 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp);
606 }
607 
608 void
609 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp)
610 {
611 	struct proc *p;
612 	struct thread *td0;
613 	struct ksegrp *kg;
614 	struct tty *tp;
615 	struct session *sp;
616 	struct timeval tv;
617 	struct sigacts *ps;
618 
619 	p = td->td_proc;
620 
621 	bzero(kp, sizeof(*kp));
622 
623 	kp->ki_structsize = sizeof(*kp);
624 	kp->ki_paddr = p;
625 	PROC_LOCK_ASSERT(p, MA_OWNED);
626 	kp->ki_addr =/* p->p_addr; */0; /* XXXKSE */
627 	kp->ki_args = p->p_args;
628 	kp->ki_textvp = p->p_textvp;
629 #ifdef KTRACE
630 	kp->ki_tracep = p->p_tracevp;
631 	mtx_lock(&ktrace_mtx);
632 	kp->ki_traceflag = p->p_traceflag;
633 	mtx_unlock(&ktrace_mtx);
634 #endif
635 	kp->ki_fd = p->p_fd;
636 	kp->ki_vmspace = p->p_vmspace;
637 	if (p->p_ucred) {
638 		kp->ki_uid = p->p_ucred->cr_uid;
639 		kp->ki_ruid = p->p_ucred->cr_ruid;
640 		kp->ki_svuid = p->p_ucred->cr_svuid;
641 		/* XXX bde doesn't like KI_NGROUPS */
642 		kp->ki_ngroups = min(p->p_ucred->cr_ngroups, KI_NGROUPS);
643 		bcopy(p->p_ucred->cr_groups, kp->ki_groups,
644 		    kp->ki_ngroups * sizeof(gid_t));
645 		kp->ki_rgid = p->p_ucred->cr_rgid;
646 		kp->ki_svgid = p->p_ucred->cr_svgid;
647 	}
648 	if (p->p_sigacts) {
649 		ps = p->p_sigacts;
650 		mtx_lock(&ps->ps_mtx);
651 		kp->ki_sigignore = ps->ps_sigignore;
652 		kp->ki_sigcatch = ps->ps_sigcatch;
653 		mtx_unlock(&ps->ps_mtx);
654 	}
655 	mtx_lock_spin(&sched_lock);
656 	if (p->p_state != PRS_NEW &&
657 	    p->p_state != PRS_ZOMBIE &&
658 	    p->p_vmspace != NULL) {
659 		struct vmspace *vm = p->p_vmspace;
660 
661 		kp->ki_size = vm->vm_map.size;
662 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
663 		if (p->p_sflag & PS_INMEM)
664 			kp->ki_rssize += UAREA_PAGES;
665 		FOREACH_THREAD_IN_PROC(p, td0) {
666 			if (!TD_IS_SWAPPED(td0))
667 				kp->ki_rssize += td0->td_kstack_pages;
668 			if (td0->td_altkstack_obj != NULL)
669 				kp->ki_rssize += td0->td_altkstack_pages;
670 		}
671 		kp->ki_swrss = vm->vm_swrss;
672 		kp->ki_tsize = vm->vm_tsize;
673 		kp->ki_dsize = vm->vm_dsize;
674 		kp->ki_ssize = vm->vm_ssize;
675 	}
676 	kp->ki_sflag = p->p_sflag;
677 	kp->ki_swtime = p->p_swtime;
678 	kp->ki_pid = p->p_pid;
679 	kp->ki_nice = p->p_nice;
680 	bintime2timeval(&p->p_rux.rux_runtime, &tv);
681 	kp->ki_runtime = tv.tv_sec * (u_int64_t)1000000 + tv.tv_usec;
682 	if (p->p_state != PRS_ZOMBIE) {
683 #if 0
684 		if (td == NULL) {
685 			/* XXXKSE: This should never happen. */
686 			printf("fill_kinfo_proc(): pid %d has no threads!\n",
687 			    p->p_pid);
688 			mtx_unlock_spin(&sched_lock);
689 			return;
690 		}
691 #endif
692 		if (td->td_wmesg != NULL) {
693 			strlcpy(kp->ki_wmesg, td->td_wmesg,
694 			    sizeof(kp->ki_wmesg));
695 		}
696 		if (TD_ON_LOCK(td)) {
697 			kp->ki_kiflag |= KI_LOCKBLOCK;
698 			strlcpy(kp->ki_lockname, td->td_lockname,
699 			    sizeof(kp->ki_lockname));
700 		}
701 
702 		if (p->p_state == PRS_NORMAL) { /*  XXXKSE very approximate */
703 			if (TD_ON_RUNQ(td) ||
704 			    TD_CAN_RUN(td) ||
705 			    TD_IS_RUNNING(td)) {
706 				kp->ki_stat = SRUN;
707 			} else if (P_SHOULDSTOP(p)) {
708 				kp->ki_stat = SSTOP;
709 			} else if (TD_IS_SLEEPING(td)) {
710 				kp->ki_stat = SSLEEP;
711 			} else if (TD_ON_LOCK(td)) {
712 				kp->ki_stat = SLOCK;
713 			} else {
714 				kp->ki_stat = SWAIT;
715 			}
716 		} else {
717 			kp->ki_stat = SIDL;
718 		}
719 
720 		kg = td->td_ksegrp;
721 
722 		/* things in the KSE GROUP */
723 		kp->ki_estcpu = kg->kg_estcpu;
724 		kp->ki_slptime = kg->kg_slptime;
725 		kp->ki_pri.pri_user = kg->kg_user_pri;
726 		kp->ki_pri.pri_class = kg->kg_pri_class;
727 
728 		/* Things in the thread */
729 		kp->ki_wchan = td->td_wchan;
730 		kp->ki_pri.pri_level = td->td_priority;
731 		kp->ki_pri.pri_native = td->td_base_pri;
732 		kp->ki_lastcpu = td->td_lastcpu;
733 		kp->ki_oncpu = td->td_oncpu;
734 		kp->ki_tdflags = td->td_flags;
735 		kp->ki_tid = td->td_tid;
736 		kp->ki_numthreads = p->p_numthreads;
737 		kp->ki_pcb = td->td_pcb;
738 		kp->ki_kstack = (void *)td->td_kstack;
739 		kp->ki_pctcpu = sched_pctcpu(td);
740 
741 		/* We can't get this anymore but ps etc never used it anyway. */
742 		kp->ki_rqindex = 0;
743 
744 	} else {
745 		kp->ki_stat = SZOMB;
746 	}
747 	mtx_unlock_spin(&sched_lock);
748 	if ((p->p_sflag & PS_INMEM) && p->p_stats != NULL) {
749 		kp->ki_start = p->p_stats->p_start;
750 		timevaladd(&kp->ki_start, &boottime);
751 		kp->ki_rusage = p->p_stats->p_ru;
752 		calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
753 		calccru(p, &kp->ki_childutime, &kp->ki_childstime);
754 
755 		/* Some callers want child-times in a single value */
756 		kp->ki_childtime = kp->ki_childstime;
757 		timevaladd(&kp->ki_childtime, &kp->ki_childutime);
758 	}
759 	sp = NULL;
760 	tp = NULL;
761 	if (p->p_pgrp) {
762 		kp->ki_pgid = p->p_pgrp->pg_id;
763 		kp->ki_jobc = p->p_pgrp->pg_jobc;
764 		sp = p->p_pgrp->pg_session;
765 
766 		if (sp != NULL) {
767 			kp->ki_sid = sp->s_sid;
768 			SESS_LOCK(sp);
769 			strlcpy(kp->ki_login, sp->s_login,
770 			    sizeof(kp->ki_login));
771 			if (sp->s_ttyvp)
772 				kp->ki_kiflag |= KI_CTTY;
773 			if (SESS_LEADER(p))
774 				kp->ki_kiflag |= KI_SLEADER;
775 			tp = sp->s_ttyp;
776 			SESS_UNLOCK(sp);
777 		}
778 	}
779 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
780 		kp->ki_tdev = dev2udev(tp->t_dev);
781 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
782 		if (tp->t_session)
783 			kp->ki_tsid = tp->t_session->s_sid;
784 	} else
785 		kp->ki_tdev = NODEV;
786 	if (p->p_comm[0] != '\0') {
787 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
788 		strlcpy(kp->ki_ocomm, p->p_comm, sizeof(kp->ki_ocomm));
789 	}
790 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
791 	    p->p_sysent->sv_name[0] != '\0')
792 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
793 	kp->ki_siglist = p->p_siglist;
794         SIGSETOR(kp->ki_siglist, td->td_siglist);
795 	kp->ki_sigmask = td->td_sigmask;
796 	kp->ki_xstat = p->p_xstat;
797 	kp->ki_acflag = p->p_acflag;
798 	kp->ki_flag = p->p_flag;
799 	/* If jailed(p->p_ucred), emulate the old P_JAILED flag. */
800 	if (jailed(p->p_ucred))
801 		kp->ki_flag |= P_JAILED;
802 	kp->ki_lock = p->p_lock;
803 	if (p->p_pptr)
804 		kp->ki_ppid = p->p_pptr->p_pid;
805 }
806 
807 /*
808  * Locate a zombie process by number
809  */
810 struct proc *
811 zpfind(pid_t pid)
812 {
813 	struct proc *p;
814 
815 	sx_slock(&allproc_lock);
816 	LIST_FOREACH(p, &zombproc, p_list)
817 		if (p->p_pid == pid) {
818 			PROC_LOCK(p);
819 			break;
820 		}
821 	sx_sunlock(&allproc_lock);
822 	return (p);
823 }
824 
825 #define KERN_PROC_ZOMBMASK	0x3
826 #define KERN_PROC_NOTHREADS	0x4
827 
828 /*
829  * Must be called with the process locked and will return with it unlocked.
830  */
831 static int
832 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
833 {
834 	struct thread *td;
835 	struct kinfo_proc kinfo_proc;
836 	int error = 0;
837 	struct proc *np;
838 	pid_t pid = p->p_pid;
839 
840 	PROC_LOCK_ASSERT(p, MA_OWNED);
841 
842 	if (flags & KERN_PROC_NOTHREADS) {
843 		fill_kinfo_proc(p, &kinfo_proc);
844 		PROC_UNLOCK(p);
845 		error = SYSCTL_OUT(req, (caddr_t)&kinfo_proc,
846 				   sizeof(kinfo_proc));
847 		PROC_LOCK(p);
848 	} else {
849 		_PHOLD(p);
850 		FOREACH_THREAD_IN_PROC(p, td) {
851 			fill_kinfo_thread(td, &kinfo_proc);
852 			PROC_UNLOCK(p);
853 			error = SYSCTL_OUT(req, (caddr_t)&kinfo_proc,
854 					   sizeof(kinfo_proc));
855 			PROC_LOCK(p);
856 			if (error)
857 				break;
858 		}
859 		_PRELE(p);
860 	}
861 	PROC_UNLOCK(p);
862 	if (error)
863 		return (error);
864 	if (flags & KERN_PROC_ZOMBMASK)
865 		np = zpfind(pid);
866 	else {
867 		if (pid == 0)
868 			return (0);
869 		np = pfind(pid);
870 	}
871 	if (np == NULL)
872 		return EAGAIN;
873 	if (np != p) {
874 		PROC_UNLOCK(np);
875 		return EAGAIN;
876 	}
877 	PROC_UNLOCK(np);
878 	return (0);
879 }
880 
881 static int
882 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
883 {
884 	int *name = (int*) arg1;
885 	u_int namelen = arg2;
886 	struct proc *p;
887 	int flags, doingzomb, oid_number;
888 	int error = 0;
889 
890 	oid_number = oidp->oid_number;
891 	if (oid_number != KERN_PROC_ALL &&
892 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
893 		flags = KERN_PROC_NOTHREADS;
894 	else {
895 		flags = 0;
896 		oid_number &= ~KERN_PROC_INC_THREAD;
897 	}
898 	if (oid_number == KERN_PROC_PID) {
899 		if (namelen != 1)
900 			return (EINVAL);
901 		p = pfind((pid_t)name[0]);
902 		if (!p)
903 			return (ESRCH);
904 		if ((error = p_cansee(curthread, p))) {
905 			PROC_UNLOCK(p);
906 			return (error);
907 		}
908 		error = sysctl_out_proc(p, req, flags);
909 		return (error);
910 	}
911 
912 	switch (oid_number) {
913 	case KERN_PROC_ALL:
914 		if (namelen != 0)
915 			return (EINVAL);
916 		break;
917 	case KERN_PROC_PROC:
918 		if (namelen != 0 && namelen != 1)
919 			return (EINVAL);
920 		break;
921 	default:
922 		if (namelen != 1)
923 			return (EINVAL);
924 		break;
925 	}
926 
927 	if (!req->oldptr) {
928 		/* overestimate by 5 procs */
929 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
930 		if (error)
931 			return (error);
932 	}
933 	error = sysctl_wire_old_buffer(req, 0);
934 	if (error != 0)
935 		return (error);
936 	sx_slock(&allproc_lock);
937 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
938 		if (!doingzomb)
939 			p = LIST_FIRST(&allproc);
940 		else
941 			p = LIST_FIRST(&zombproc);
942 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
943 			/*
944 			 * Skip embryonic processes.
945 			 */
946 			mtx_lock_spin(&sched_lock);
947 			if (p->p_state == PRS_NEW) {
948 				mtx_unlock_spin(&sched_lock);
949 				continue;
950 			}
951 			mtx_unlock_spin(&sched_lock);
952 			PROC_LOCK(p);
953 			/*
954 			 * Show a user only appropriate processes.
955 			 */
956 			if (p_cansee(curthread, p)) {
957 				PROC_UNLOCK(p);
958 				continue;
959 			}
960 			/*
961 			 * TODO - make more efficient (see notes below).
962 			 * do by session.
963 			 */
964 			switch (oid_number) {
965 
966 			case KERN_PROC_GID:
967 				if (p->p_ucred == NULL ||
968 				    p->p_ucred->cr_gid != (gid_t)name[0]) {
969 					PROC_UNLOCK(p);
970 					continue;
971 				}
972 				break;
973 
974 			case KERN_PROC_PGRP:
975 				/* could do this by traversing pgrp */
976 				if (p->p_pgrp == NULL ||
977 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
978 					PROC_UNLOCK(p);
979 					continue;
980 				}
981 				break;
982 
983 			case KERN_PROC_RGID:
984 				if (p->p_ucred == NULL ||
985 				    p->p_ucred->cr_rgid != (gid_t)name[0]) {
986 					PROC_UNLOCK(p);
987 					continue;
988 				}
989 				break;
990 
991 			case KERN_PROC_SESSION:
992 				if (p->p_session == NULL ||
993 				    p->p_session->s_sid != (pid_t)name[0]) {
994 					PROC_UNLOCK(p);
995 					continue;
996 				}
997 				break;
998 
999 			case KERN_PROC_TTY:
1000 				if ((p->p_flag & P_CONTROLT) == 0 ||
1001 				    p->p_session == NULL) {
1002 					PROC_UNLOCK(p);
1003 					continue;
1004 				}
1005 				SESS_LOCK(p->p_session);
1006 				if (p->p_session->s_ttyp == NULL ||
1007 				    dev2udev(p->p_session->s_ttyp->t_dev) !=
1008 				    (dev_t)name[0]) {
1009 					SESS_UNLOCK(p->p_session);
1010 					PROC_UNLOCK(p);
1011 					continue;
1012 				}
1013 				SESS_UNLOCK(p->p_session);
1014 				break;
1015 
1016 			case KERN_PROC_UID:
1017 				if (p->p_ucred == NULL ||
1018 				    p->p_ucred->cr_uid != (uid_t)name[0]) {
1019 					PROC_UNLOCK(p);
1020 					continue;
1021 				}
1022 				break;
1023 
1024 			case KERN_PROC_RUID:
1025 				if (p->p_ucred == NULL ||
1026 				    p->p_ucred->cr_ruid != (uid_t)name[0]) {
1027 					PROC_UNLOCK(p);
1028 					continue;
1029 				}
1030 				break;
1031 
1032 			case KERN_PROC_PROC:
1033 				break;
1034 
1035 			default:
1036 				break;
1037 
1038 			}
1039 
1040 			error = sysctl_out_proc(p, req, flags | doingzomb);
1041 			if (error) {
1042 				sx_sunlock(&allproc_lock);
1043 				return (error);
1044 			}
1045 		}
1046 	}
1047 	sx_sunlock(&allproc_lock);
1048 	return (0);
1049 }
1050 
1051 struct pargs *
1052 pargs_alloc(int len)
1053 {
1054 	struct pargs *pa;
1055 
1056 	MALLOC(pa, struct pargs *, sizeof(struct pargs) + len, M_PARGS,
1057 		M_WAITOK);
1058 	pa->ar_ref = 1;
1059 	pa->ar_length = len;
1060 	return (pa);
1061 }
1062 
1063 void
1064 pargs_free(struct pargs *pa)
1065 {
1066 
1067 	FREE(pa, M_PARGS);
1068 }
1069 
1070 void
1071 pargs_hold(struct pargs *pa)
1072 {
1073 
1074 	if (pa == NULL)
1075 		return;
1076 	PARGS_LOCK(pa);
1077 	pa->ar_ref++;
1078 	PARGS_UNLOCK(pa);
1079 }
1080 
1081 void
1082 pargs_drop(struct pargs *pa)
1083 {
1084 
1085 	if (pa == NULL)
1086 		return;
1087 	PARGS_LOCK(pa);
1088 	if (--pa->ar_ref == 0) {
1089 		PARGS_UNLOCK(pa);
1090 		pargs_free(pa);
1091 	} else
1092 		PARGS_UNLOCK(pa);
1093 }
1094 
1095 /*
1096  * This sysctl allows a process to retrieve the argument list or process
1097  * title for another process without groping around in the address space
1098  * of the other process.  It also allow a process to set its own "process
1099  * title to a string of its own choice.
1100  */
1101 static int
1102 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1103 {
1104 	int *name = (int*) arg1;
1105 	u_int namelen = arg2;
1106 	struct pargs *newpa, *pa;
1107 	struct proc *p;
1108 	int error = 0;
1109 
1110 	if (namelen != 1)
1111 		return (EINVAL);
1112 
1113 	p = pfind((pid_t)name[0]);
1114 	if (!p)
1115 		return (ESRCH);
1116 
1117 	if ((error = p_cansee(curthread, p)) != 0) {
1118 		PROC_UNLOCK(p);
1119 		return (error);
1120 	}
1121 
1122 	if (req->newptr && curproc != p) {
1123 		PROC_UNLOCK(p);
1124 		return (EPERM);
1125 	}
1126 
1127 	pa = p->p_args;
1128 	pargs_hold(pa);
1129 	PROC_UNLOCK(p);
1130 	if (req->oldptr != NULL && pa != NULL)
1131 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1132 	pargs_drop(pa);
1133 	if (error != 0 || req->newptr == NULL)
1134 		return (error);
1135 
1136 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1137 		return (ENOMEM);
1138 	newpa = pargs_alloc(req->newlen);
1139 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1140 	if (error != 0) {
1141 		pargs_free(newpa);
1142 		return (error);
1143 	}
1144 	PROC_LOCK(p);
1145 	pa = p->p_args;
1146 	p->p_args = newpa;
1147 	PROC_UNLOCK(p);
1148 	pargs_drop(pa);
1149 	return (0);
1150 }
1151 
1152 static int
1153 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1154 {
1155 	struct proc *p;
1156 	char *sv_name;
1157 	int *name;
1158 	int namelen;
1159 	int error;
1160 
1161 	namelen = arg2;
1162 	if (namelen != 1)
1163 		return (EINVAL);
1164 
1165 	name = (int *)arg1;
1166 	if ((p = pfind((pid_t)name[0])) == NULL)
1167 		return (ESRCH);
1168 	if ((error = p_cansee(curthread, p))) {
1169 		PROC_UNLOCK(p);
1170 		return (error);
1171 	}
1172 	sv_name = p->p_sysent->sv_name;
1173 	PROC_UNLOCK(p);
1174 	return (sysctl_handle_string(oidp, sv_name, 0, req));
1175 }
1176 
1177 
1178 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
1179 
1180 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT,
1181 	0, 0, sysctl_kern_proc, "S,proc", "Return entire process table");
1182 
1183 SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD,
1184 	sysctl_kern_proc, "Process table");
1185 
1186 SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD,
1187 	sysctl_kern_proc, "Process table");
1188 
1189 SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD,
1190 	sysctl_kern_proc, "Process table");
1191 
1192 SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD,
1193 	sysctl_kern_proc, "Process table");
1194 
1195 SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD,
1196 	sysctl_kern_proc, "Process table");
1197 
1198 SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD,
1199 	sysctl_kern_proc, "Process table");
1200 
1201 SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD,
1202 	sysctl_kern_proc, "Process table");
1203 
1204 SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD,
1205 	sysctl_kern_proc, "Process table");
1206 
1207 SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD,
1208 	sysctl_kern_proc, "Return process table, no threads");
1209 
1210 SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, CTLFLAG_RW | CTLFLAG_ANYBODY,
1211 	sysctl_kern_proc_args, "Process argument list");
1212 
1213 SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD,
1214 	sysctl_kern_proc_sv_name, "Process syscall vector name (ABI type)");
1215 
1216 SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
1217 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1218 
1219 SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
1220 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1221 
1222 SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
1223 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1224 
1225 SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD), sid_td,
1226 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1227 
1228 SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
1229 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1230 
1231 SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
1232 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1233 
1234 SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
1235 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1236 
1237 SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
1238 	CTLFLAG_RD, sysctl_kern_proc, "Process table");
1239 
1240 SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
1241 	CTLFLAG_RD, sysctl_kern_proc, "Return process table, no threads");
1242