xref: /freebsd/sys/kern/kern_proc.c (revision 7660b554bc59a07be0431c17e0e33815818baa69)
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  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)kern_proc.c	8.7 (Berkeley) 2/14/95
34  * $FreeBSD$
35  */
36 
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include "opt_ktrace.h"
41 #include "opt_kstack_pages.h"
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/kernel.h>
46 #include <sys/lock.h>
47 #include <sys/malloc.h>
48 #include <sys/mutex.h>
49 #include <sys/proc.h>
50 #include <sys/sysent.h>
51 #include <sys/sched.h>
52 #include <sys/smp.h>
53 #include <sys/sysctl.h>
54 #include <sys/filedesc.h>
55 #include <sys/tty.h>
56 #include <sys/signalvar.h>
57 #include <sys/sx.h>
58 #include <sys/user.h>
59 #include <sys/jail.h>
60 #ifdef KTRACE
61 #include <sys/uio.h>
62 #include <sys/ktrace.h>
63 #endif
64 
65 #include <vm/vm.h>
66 #include <vm/vm_extern.h>
67 #include <vm/pmap.h>
68 #include <vm/vm_map.h>
69 #include <vm/uma.h>
70 #include <machine/critical.h>
71 
72 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
73 MALLOC_DEFINE(M_SESSION, "session", "session header");
74 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
75 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
76 
77 static void doenterpgrp(struct proc *, struct pgrp *);
78 static void orphanpg(struct pgrp *pg);
79 static void pgadjustjobc(struct pgrp *pgrp, int entering);
80 static void pgdelete(struct pgrp *);
81 static void proc_ctor(void *mem, int size, void *arg);
82 static void proc_dtor(void *mem, int size, void *arg);
83 static void proc_init(void *mem, int size);
84 static void proc_fini(void *mem, int size);
85 
86 /*
87  * Other process lists
88  */
89 struct pidhashhead *pidhashtbl;
90 u_long pidhash;
91 struct pgrphashhead *pgrphashtbl;
92 u_long pgrphash;
93 struct proclist allproc;
94 struct proclist zombproc;
95 struct sx allproc_lock;
96 struct sx proctree_lock;
97 struct mtx pargs_ref_lock;
98 struct mtx ppeers_lock;
99 uma_zone_t proc_zone;
100 uma_zone_t ithread_zone;
101 
102 int kstack_pages = KSTACK_PAGES;
103 int uarea_pages = UAREA_PAGES;
104 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0, "");
105 SYSCTL_INT(_kern, OID_AUTO, uarea_pages, CTLFLAG_RD, &uarea_pages, 0, "");
106 
107 #define RANGEOF(type, start, end) (offsetof(type, end) - offsetof(type, start))
108 
109 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
110 
111 /*
112  * Initialize global process hashing structures.
113  */
114 void
115 procinit()
116 {
117 
118 	sx_init(&allproc_lock, "allproc");
119 	sx_init(&proctree_lock, "proctree");
120 	mtx_init(&pargs_ref_lock, "struct pargs.ref", NULL, MTX_DEF);
121 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
122 	LIST_INIT(&allproc);
123 	LIST_INIT(&zombproc);
124 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
125 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
126 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
127 	    proc_ctor, proc_dtor, proc_init, proc_fini,
128 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
129 	uihashinit();
130 }
131 
132 /*
133  * Prepare a proc for use.
134  */
135 static void
136 proc_ctor(void *mem, int size, void *arg)
137 {
138 	struct proc *p;
139 
140 	p = (struct proc *)mem;
141 }
142 
143 /*
144  * Reclaim a proc after use.
145  */
146 static void
147 proc_dtor(void *mem, int size, void *arg)
148 {
149 	struct proc *p;
150 	struct thread *td;
151 	struct ksegrp *kg;
152 	struct kse *ke;
153 
154 	/* INVARIANTS checks go here */
155 	p = (struct proc *)mem;
156 	KASSERT((p->p_numthreads == 1),
157 	    ("bad number of threads in exiting process"));
158         td = FIRST_THREAD_IN_PROC(p);
159 	KASSERT((td != NULL), ("proc_dtor: bad thread pointer"));
160         kg = FIRST_KSEGRP_IN_PROC(p);
161 	KASSERT((kg != NULL), ("proc_dtor: bad kg pointer"));
162         ke = FIRST_KSE_IN_KSEGRP(kg);
163 	KASSERT((ke != NULL), ("proc_dtor: bad ke pointer"));
164 
165 	/* Dispose of an alternate kstack, if it exists.
166 	 * XXX What if there are more than one thread in the proc?
167 	 *     The first thread in the proc is special and not
168 	 *     freed, so you gotta do this here.
169 	 */
170 	if (((p->p_flag & P_KTHREAD) != 0) && (td->td_altkstack != 0))
171 		vm_thread_dispose_altkstack(td);
172 
173 	/*
174 	 * We want to make sure we know the initial linkages.
175 	 * so for now tear them down and remake them.
176 	 * This is probably un-needed as we can probably rely
177 	 * on the state coming in here from wait4().
178 	 */
179 	proc_linkup(p, kg, ke, td);
180 }
181 
182 /*
183  * Initialize type-stable parts of a proc (when newly created).
184  */
185 static void
186 proc_init(void *mem, int size)
187 {
188 	struct proc *p;
189 	struct thread *td;
190 	struct ksegrp *kg;
191 	struct kse *ke;
192 
193 	p = (struct proc *)mem;
194 	p->p_sched = (struct p_sched *)&p[1];
195 	vm_proc_new(p);
196 	td = thread_alloc();
197 	ke = kse_alloc();
198 	kg = ksegrp_alloc();
199 	proc_linkup(p, kg, ke, td);
200 	bzero(&p->p_mtx, sizeof(struct mtx));
201 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK);
202 }
203 
204 /*
205  * Tear down type-stable parts of a proc (just before being discarded)
206  */
207 static void
208 proc_fini(void *mem, int size)
209 {
210 	struct proc *p;
211 	struct thread *td;
212 	struct ksegrp *kg;
213 	struct kse *ke;
214 
215 	p = (struct proc *)mem;
216 	KASSERT((p->p_numthreads == 1),
217 	    ("bad number of threads in freeing process"));
218         td = FIRST_THREAD_IN_PROC(p);
219 	KASSERT((td != NULL), ("proc_dtor: bad thread pointer"));
220         kg = FIRST_KSEGRP_IN_PROC(p);
221 	KASSERT((kg != NULL), ("proc_dtor: bad kg pointer"));
222         ke = FIRST_KSE_IN_KSEGRP(kg);
223 	KASSERT((ke != NULL), ("proc_dtor: bad ke pointer"));
224 	vm_proc_dispose(p);
225 	thread_free(td);
226 	ksegrp_free(kg);
227 	kse_free(ke);
228 	mtx_destroy(&p->p_mtx);
229 }
230 
231 /*
232  * Is p an inferior of the current process?
233  */
234 int
235 inferior(p)
236 	register struct proc *p;
237 {
238 
239 	sx_assert(&proctree_lock, SX_LOCKED);
240 	for (; p != curproc; p = p->p_pptr)
241 		if (p->p_pid == 0)
242 			return (0);
243 	return (1);
244 }
245 
246 /*
247  * Locate a process by number
248  */
249 struct proc *
250 pfind(pid)
251 	register pid_t pid;
252 {
253 	register struct proc *p;
254 
255 	sx_slock(&allproc_lock);
256 	LIST_FOREACH(p, PIDHASH(pid), p_hash)
257 		if (p->p_pid == pid) {
258 			PROC_LOCK(p);
259 			break;
260 		}
261 	sx_sunlock(&allproc_lock);
262 	return (p);
263 }
264 
265 /*
266  * Locate a process group by number.
267  * The caller must hold proctree_lock.
268  */
269 struct pgrp *
270 pgfind(pgid)
271 	register pid_t pgid;
272 {
273 	register struct pgrp *pgrp;
274 
275 	sx_assert(&proctree_lock, SX_LOCKED);
276 
277 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
278 		if (pgrp->pg_id == pgid) {
279 			PGRP_LOCK(pgrp);
280 			return (pgrp);
281 		}
282 	}
283 	return (NULL);
284 }
285 
286 /*
287  * Create a new process group.
288  * pgid must be equal to the pid of p.
289  * Begin a new session if required.
290  */
291 int
292 enterpgrp(p, pgid, pgrp, sess)
293 	register struct proc *p;
294 	pid_t pgid;
295 	struct pgrp *pgrp;
296 	struct session *sess;
297 {
298 	struct pgrp *pgrp2;
299 
300 	sx_assert(&proctree_lock, SX_XLOCKED);
301 
302 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
303 	KASSERT(p->p_pid == pgid,
304 	    ("enterpgrp: new pgrp and pid != pgid"));
305 
306 	pgrp2 = pgfind(pgid);
307 
308 	KASSERT(pgrp2 == NULL,
309 	    ("enterpgrp: pgrp with pgid exists"));
310 	KASSERT(!SESS_LEADER(p),
311 	    ("enterpgrp: session leader attempted setpgrp"));
312 
313 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
314 
315 	if (sess != NULL) {
316 		/*
317 		 * new session
318 		 */
319 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
320 		PROC_LOCK(p);
321 		p->p_flag &= ~P_CONTROLT;
322 		PROC_UNLOCK(p);
323 		PGRP_LOCK(pgrp);
324 		sess->s_leader = p;
325 		sess->s_sid = p->p_pid;
326 		sess->s_count = 1;
327 		sess->s_ttyvp = NULL;
328 		sess->s_ttyp = NULL;
329 		bcopy(p->p_session->s_login, sess->s_login,
330 			    sizeof(sess->s_login));
331 		pgrp->pg_session = sess;
332 		KASSERT(p == curproc,
333 		    ("enterpgrp: mksession and p != curproc"));
334 	} else {
335 		pgrp->pg_session = p->p_session;
336 		SESS_LOCK(pgrp->pg_session);
337 		pgrp->pg_session->s_count++;
338 		SESS_UNLOCK(pgrp->pg_session);
339 		PGRP_LOCK(pgrp);
340 	}
341 	pgrp->pg_id = pgid;
342 	LIST_INIT(&pgrp->pg_members);
343 
344 	/*
345 	 * As we have an exclusive lock of proctree_lock,
346 	 * this should not deadlock.
347 	 */
348 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
349 	pgrp->pg_jobc = 0;
350 	SLIST_INIT(&pgrp->pg_sigiolst);
351 	PGRP_UNLOCK(pgrp);
352 
353 	doenterpgrp(p, pgrp);
354 
355 	return (0);
356 }
357 
358 /*
359  * Move p to an existing process group
360  */
361 int
362 enterthispgrp(p, pgrp)
363 	register struct proc *p;
364 	struct pgrp *pgrp;
365 {
366 
367 	sx_assert(&proctree_lock, SX_XLOCKED);
368 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
369 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
370 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
371 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
372 	KASSERT(pgrp->pg_session == p->p_session,
373 		("%s: pgrp's session %p, p->p_session %p.\n",
374 		__func__,
375 		pgrp->pg_session,
376 		p->p_session));
377 	KASSERT(pgrp != p->p_pgrp,
378 		("%s: p belongs to pgrp.", __func__));
379 
380 	doenterpgrp(p, pgrp);
381 
382 	return (0);
383 }
384 
385 /*
386  * Move p to a process group
387  */
388 static void
389 doenterpgrp(p, pgrp)
390 	struct proc *p;
391 	struct pgrp *pgrp;
392 {
393 	struct pgrp *savepgrp;
394 
395 	sx_assert(&proctree_lock, SX_XLOCKED);
396 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
397 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
398 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
399 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
400 
401 	savepgrp = p->p_pgrp;
402 
403 	/*
404 	 * Adjust eligibility of affected pgrps to participate in job control.
405 	 * Increment eligibility counts before decrementing, otherwise we
406 	 * could reach 0 spuriously during the first call.
407 	 */
408 	fixjobc(p, pgrp, 1);
409 	fixjobc(p, p->p_pgrp, 0);
410 
411 	PGRP_LOCK(pgrp);
412 	PGRP_LOCK(savepgrp);
413 	PROC_LOCK(p);
414 	LIST_REMOVE(p, p_pglist);
415 	p->p_pgrp = pgrp;
416 	PROC_UNLOCK(p);
417 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
418 	PGRP_UNLOCK(savepgrp);
419 	PGRP_UNLOCK(pgrp);
420 	if (LIST_EMPTY(&savepgrp->pg_members))
421 		pgdelete(savepgrp);
422 }
423 
424 /*
425  * remove process from process group
426  */
427 int
428 leavepgrp(p)
429 	register struct proc *p;
430 {
431 	struct pgrp *savepgrp;
432 
433 	sx_assert(&proctree_lock, SX_XLOCKED);
434 	savepgrp = p->p_pgrp;
435 	PGRP_LOCK(savepgrp);
436 	PROC_LOCK(p);
437 	LIST_REMOVE(p, p_pglist);
438 	p->p_pgrp = NULL;
439 	PROC_UNLOCK(p);
440 	PGRP_UNLOCK(savepgrp);
441 	if (LIST_EMPTY(&savepgrp->pg_members))
442 		pgdelete(savepgrp);
443 	return (0);
444 }
445 
446 /*
447  * delete a process group
448  */
449 static void
450 pgdelete(pgrp)
451 	register struct pgrp *pgrp;
452 {
453 	struct session *savesess;
454 
455 	sx_assert(&proctree_lock, SX_XLOCKED);
456 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
457 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
458 
459 	/*
460 	 * Reset any sigio structures pointing to us as a result of
461 	 * F_SETOWN with our pgid.
462 	 */
463 	funsetownlst(&pgrp->pg_sigiolst);
464 
465 	PGRP_LOCK(pgrp);
466 	if (pgrp->pg_session->s_ttyp != NULL &&
467 	    pgrp->pg_session->s_ttyp->t_pgrp == pgrp)
468 		pgrp->pg_session->s_ttyp->t_pgrp = NULL;
469 	LIST_REMOVE(pgrp, pg_hash);
470 	savesess = pgrp->pg_session;
471 	SESS_LOCK(savesess);
472 	savesess->s_count--;
473 	SESS_UNLOCK(savesess);
474 	PGRP_UNLOCK(pgrp);
475 	if (savesess->s_count == 0) {
476 		mtx_destroy(&savesess->s_mtx);
477 		FREE(pgrp->pg_session, M_SESSION);
478 	}
479 	mtx_destroy(&pgrp->pg_mtx);
480 	FREE(pgrp, M_PGRP);
481 }
482 
483 static void
484 pgadjustjobc(pgrp, entering)
485 	struct pgrp *pgrp;
486 	int entering;
487 {
488 
489 	PGRP_LOCK(pgrp);
490 	if (entering)
491 		pgrp->pg_jobc++;
492 	else {
493 		--pgrp->pg_jobc;
494 		if (pgrp->pg_jobc == 0)
495 			orphanpg(pgrp);
496 	}
497 	PGRP_UNLOCK(pgrp);
498 }
499 
500 /*
501  * Adjust pgrp jobc counters when specified process changes process group.
502  * We count the number of processes in each process group that "qualify"
503  * the group for terminal job control (those with a parent in a different
504  * process group of the same session).  If that count reaches zero, the
505  * process group becomes orphaned.  Check both the specified process'
506  * process group and that of its children.
507  * entering == 0 => p is leaving specified group.
508  * entering == 1 => p is entering specified group.
509  */
510 void
511 fixjobc(p, pgrp, entering)
512 	register struct proc *p;
513 	register struct pgrp *pgrp;
514 	int entering;
515 {
516 	register struct pgrp *hispgrp;
517 	register struct session *mysession;
518 
519 	sx_assert(&proctree_lock, SX_LOCKED);
520 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
521 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
522 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
523 
524 	/*
525 	 * Check p's parent to see whether p qualifies its own process
526 	 * group; if so, adjust count for p's process group.
527 	 */
528 	mysession = pgrp->pg_session;
529 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
530 	    hispgrp->pg_session == mysession)
531 		pgadjustjobc(pgrp, entering);
532 
533 	/*
534 	 * Check this process' children to see whether they qualify
535 	 * their process groups; if so, adjust counts for children's
536 	 * process groups.
537 	 */
538 	LIST_FOREACH(p, &p->p_children, p_sibling) {
539 		hispgrp = p->p_pgrp;
540 		if (hispgrp == pgrp ||
541 		    hispgrp->pg_session != mysession)
542 			continue;
543 		PROC_LOCK(p);
544 		if (p->p_state == PRS_ZOMBIE) {
545 			PROC_UNLOCK(p);
546 			continue;
547 		}
548 		PROC_UNLOCK(p);
549 		pgadjustjobc(hispgrp, entering);
550 	}
551 }
552 
553 /*
554  * A process group has become orphaned;
555  * if there are any stopped processes in the group,
556  * hang-up all process in that group.
557  */
558 static void
559 orphanpg(pg)
560 	struct pgrp *pg;
561 {
562 	register struct proc *p;
563 
564 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
565 
566 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
567 		PROC_LOCK(p);
568 		if (P_SHOULDSTOP(p)) {
569 			PROC_UNLOCK(p);
570 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
571 				PROC_LOCK(p);
572 				psignal(p, SIGHUP);
573 				psignal(p, SIGCONT);
574 				PROC_UNLOCK(p);
575 			}
576 			return;
577 		}
578 		PROC_UNLOCK(p);
579 	}
580 }
581 
582 #include "opt_ddb.h"
583 #ifdef DDB
584 #include <ddb/ddb.h>
585 
586 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
587 {
588 	register struct pgrp *pgrp;
589 	register struct proc *p;
590 	register int i;
591 
592 	for (i = 0; i <= pgrphash; i++) {
593 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
594 			printf("\tindx %d\n", i);
595 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
596 				printf(
597 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
598 				    (void *)pgrp, (long)pgrp->pg_id,
599 				    (void *)pgrp->pg_session,
600 				    pgrp->pg_session->s_count,
601 				    (void *)LIST_FIRST(&pgrp->pg_members));
602 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
603 					printf("\t\tpid %ld addr %p pgrp %p\n",
604 					    (long)p->p_pid, (void *)p,
605 					    (void *)p->p_pgrp);
606 				}
607 			}
608 		}
609 	}
610 }
611 #endif /* DDB */
612 void
613 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp);
614 
615 /*
616  * Fill in a kinfo_proc structure for the specified process.
617  * Must be called with the target process locked.
618  */
619 void
620 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
621 {
622 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp);
623 }
624 
625 void
626 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp)
627 {
628 	struct proc *p;
629 	struct thread *td0;
630 	struct kse *ke;
631 	struct ksegrp *kg;
632 	struct tty *tp;
633 	struct session *sp;
634 	struct timeval tv;
635 	struct sigacts *ps;
636 
637 	p = td->td_proc;
638 
639 	bzero(kp, sizeof(*kp));
640 
641 	kp->ki_structsize = sizeof(*kp);
642 	kp->ki_paddr = p;
643 	PROC_LOCK_ASSERT(p, MA_OWNED);
644 	kp->ki_addr =/* p->p_addr; */0; /* XXXKSE */
645 	kp->ki_args = p->p_args;
646 	kp->ki_textvp = p->p_textvp;
647 #ifdef KTRACE
648 	kp->ki_tracep = p->p_tracevp;
649 	mtx_lock(&ktrace_mtx);
650 	kp->ki_traceflag = p->p_traceflag;
651 	mtx_unlock(&ktrace_mtx);
652 #endif
653 	kp->ki_fd = p->p_fd;
654 	kp->ki_vmspace = p->p_vmspace;
655 	if (p->p_ucred) {
656 		kp->ki_uid = p->p_ucred->cr_uid;
657 		kp->ki_ruid = p->p_ucred->cr_ruid;
658 		kp->ki_svuid = p->p_ucred->cr_svuid;
659 		/* XXX bde doesn't like KI_NGROUPS */
660 		kp->ki_ngroups = min(p->p_ucred->cr_ngroups, KI_NGROUPS);
661 		bcopy(p->p_ucred->cr_groups, kp->ki_groups,
662 		    kp->ki_ngroups * sizeof(gid_t));
663 		kp->ki_rgid = p->p_ucred->cr_rgid;
664 		kp->ki_svgid = p->p_ucred->cr_svgid;
665 	}
666 	if (p->p_sigacts) {
667 		ps = p->p_sigacts;
668 		mtx_lock(&ps->ps_mtx);
669 		kp->ki_sigignore = ps->ps_sigignore;
670 		kp->ki_sigcatch = ps->ps_sigcatch;
671 		mtx_unlock(&ps->ps_mtx);
672 	}
673 	mtx_lock_spin(&sched_lock);
674 	if (p->p_state != PRS_NEW &&
675 	    p->p_state != PRS_ZOMBIE &&
676 	    p->p_vmspace != NULL) {
677 		struct vmspace *vm = p->p_vmspace;
678 
679 		kp->ki_size = vm->vm_map.size;
680 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
681 		if (p->p_sflag & PS_INMEM)
682 			kp->ki_rssize += UAREA_PAGES;
683 		FOREACH_THREAD_IN_PROC(p, td0) {
684 			if (!TD_IS_SWAPPED(td0))
685 				kp->ki_rssize += td0->td_kstack_pages;
686 			if (td0->td_altkstack_obj != NULL)
687 				kp->ki_rssize += td0->td_altkstack_pages;
688 		}
689 		kp->ki_swrss = vm->vm_swrss;
690 		kp->ki_tsize = vm->vm_tsize;
691 		kp->ki_dsize = vm->vm_dsize;
692 		kp->ki_ssize = vm->vm_ssize;
693 	}
694 	if ((p->p_sflag & PS_INMEM) && p->p_stats) {
695 		kp->ki_start = p->p_stats->p_start;
696 		timevaladd(&kp->ki_start, &boottime);
697 		kp->ki_rusage = p->p_stats->p_ru;
698 		kp->ki_childtime.tv_sec = p->p_stats->p_cru.ru_utime.tv_sec +
699 		    p->p_stats->p_cru.ru_stime.tv_sec;
700 		kp->ki_childtime.tv_usec = p->p_stats->p_cru.ru_utime.tv_usec +
701 		    p->p_stats->p_cru.ru_stime.tv_usec;
702 	}
703 	if (p->p_state != PRS_ZOMBIE) {
704 #if 0
705 		if (td == NULL) {
706 			/* XXXKSE: This should never happen. */
707 			printf("fill_kinfo_proc(): pid %d has no threads!\n",
708 			    p->p_pid);
709 			mtx_unlock_spin(&sched_lock);
710 			return;
711 		}
712 #endif
713 		if (td->td_wmesg != NULL) {
714 			strlcpy(kp->ki_wmesg, td->td_wmesg,
715 			    sizeof(kp->ki_wmesg));
716 		}
717 		if (TD_ON_LOCK(td)) {
718 			kp->ki_kiflag |= KI_LOCKBLOCK;
719 			strlcpy(kp->ki_lockname, td->td_lockname,
720 			    sizeof(kp->ki_lockname));
721 		}
722 
723 		if (p->p_state == PRS_NORMAL) { /*  XXXKSE very approximate */
724 			if (TD_ON_RUNQ(td) ||
725 			    TD_CAN_RUN(td) ||
726 			    TD_IS_RUNNING(td)) {
727 				kp->ki_stat = SRUN;
728 			} else if (P_SHOULDSTOP(p)) {
729 				kp->ki_stat = SSTOP;
730 			} else if (TD_IS_SLEEPING(td)) {
731 				kp->ki_stat = SSLEEP;
732 			} else if (TD_ON_LOCK(td)) {
733 				kp->ki_stat = SLOCK;
734 			} else {
735 				kp->ki_stat = SWAIT;
736 			}
737 		} else {
738 			kp->ki_stat = SIDL;
739 		}
740 
741 		kp->ki_sflag = p->p_sflag;
742 		kp->ki_swtime = p->p_swtime;
743 		kp->ki_pid = p->p_pid;
744 		kg = td->td_ksegrp;
745 		ke = td->td_kse;
746 		bintime2timeval(&p->p_runtime, &tv);
747 		kp->ki_runtime =
748 		    tv.tv_sec * (u_int64_t)1000000 + tv.tv_usec;
749 
750 		/* things in the KSE GROUP */
751 		kp->ki_estcpu = kg->kg_estcpu;
752 		kp->ki_slptime = kg->kg_slptime;
753 		kp->ki_pri.pri_user = kg->kg_user_pri;
754 		kp->ki_pri.pri_class = kg->kg_pri_class;
755 		kp->ki_nice = kg->kg_nice;
756 
757 		/* Things in the thread */
758 		kp->ki_wchan = td->td_wchan;
759 		kp->ki_pri.pri_level = td->td_priority;
760 		kp->ki_pri.pri_native = td->td_base_pri;
761 		kp->ki_lastcpu = td->td_lastcpu;
762 		kp->ki_oncpu = td->td_oncpu;
763 		kp->ki_tdflags = td->td_flags;
764 		kp->ki_pcb = td->td_pcb;
765 		kp->ki_kstack = (void *)td->td_kstack;
766 
767 		/* Things in the kse */
768 
769 		if (ke) {
770 			kp->ki_rqindex = ke->ke_rqindex;
771 			kp->ki_pctcpu = sched_pctcpu(ke);
772 		} else {
773 			kp->ki_rqindex = 0;
774 			kp->ki_pctcpu = 0;
775 		}
776 
777 	} else {
778 		kp->ki_stat = SZOMB;
779 	}
780 	mtx_unlock_spin(&sched_lock);
781 	sp = NULL;
782 	tp = NULL;
783 	if (p->p_pgrp) {
784 		kp->ki_pgid = p->p_pgrp->pg_id;
785 		kp->ki_jobc = p->p_pgrp->pg_jobc;
786 		sp = p->p_pgrp->pg_session;
787 
788 		if (sp != NULL) {
789 			kp->ki_sid = sp->s_sid;
790 			SESS_LOCK(sp);
791 			strlcpy(kp->ki_login, sp->s_login,
792 			    sizeof(kp->ki_login));
793 			if (sp->s_ttyvp)
794 				kp->ki_kiflag |= KI_CTTY;
795 			if (SESS_LEADER(p))
796 				kp->ki_kiflag |= KI_SLEADER;
797 			tp = sp->s_ttyp;
798 			SESS_UNLOCK(sp);
799 		}
800 	}
801 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
802 		kp->ki_tdev = dev2udev(tp->t_dev);
803 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
804 		if (tp->t_session)
805 			kp->ki_tsid = tp->t_session->s_sid;
806 	} else
807 		kp->ki_tdev = NOUDEV;
808 	if (p->p_comm[0] != '\0') {
809 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
810 		strlcpy(kp->ki_ocomm, p->p_comm, sizeof(kp->ki_ocomm));
811 	}
812 	kp->ki_siglist = p->p_siglist;
813         SIGSETOR(kp->ki_siglist, td->td_siglist);
814 	kp->ki_sigmask = td->td_sigmask;
815 	kp->ki_xstat = p->p_xstat;
816 	kp->ki_acflag = p->p_acflag;
817 	kp->ki_flag = p->p_flag;
818 	/* If jailed(p->p_ucred), emulate the old P_JAILED flag. */
819 	if (jailed(p->p_ucred))
820 		kp->ki_flag |= P_JAILED;
821 	kp->ki_lock = p->p_lock;
822 	if (p->p_pptr)
823 		kp->ki_ppid = p->p_pptr->p_pid;
824 }
825 
826 /*
827  * Locate a zombie process by number
828  */
829 struct proc *
830 zpfind(pid_t pid)
831 {
832 	struct proc *p;
833 
834 	sx_slock(&allproc_lock);
835 	LIST_FOREACH(p, &zombproc, p_list)
836 		if (p->p_pid == pid) {
837 			PROC_LOCK(p);
838 			break;
839 		}
840 	sx_sunlock(&allproc_lock);
841 	return (p);
842 }
843 
844 #define KERN_PROC_ZOMBMASK	0x3
845 #define KERN_PROC_NOTHREADS	0x4
846 
847 /*
848  * Must be called with the process locked and will return with it unlocked.
849  */
850 static int
851 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
852 {
853 	struct thread *td;
854 	struct kinfo_proc kinfo_proc;
855 	int error = 0;
856 	struct proc *np;
857 	pid_t pid = p->p_pid;
858 
859 	PROC_LOCK_ASSERT(p, MA_OWNED);
860 
861 	if (flags & KERN_PROC_NOTHREADS) {
862 		fill_kinfo_proc(p, &kinfo_proc);
863 		PROC_UNLOCK(p);
864 		error = SYSCTL_OUT(req, (caddr_t)&kinfo_proc,
865 				   sizeof(kinfo_proc));
866 		PROC_LOCK(p);
867 	} else {
868 		_PHOLD(p);
869 		FOREACH_THREAD_IN_PROC(p, td) {
870 			fill_kinfo_thread(td, &kinfo_proc);
871 			PROC_UNLOCK(p);
872 			error = SYSCTL_OUT(req, (caddr_t)&kinfo_proc,
873 					   sizeof(kinfo_proc));
874 			PROC_LOCK(p);
875 			if (error)
876 				break;
877 		}
878 		_PRELE(p);
879 	}
880 	PROC_UNLOCK(p);
881 	if (error)
882 		return (error);
883 	if (flags & KERN_PROC_ZOMBMASK)
884 		np = zpfind(pid);
885 	else {
886 		if (pid == 0)
887 			return (0);
888 		np = pfind(pid);
889 	}
890 	if (np == NULL)
891 		return EAGAIN;
892 	if (np != p) {
893 		PROC_UNLOCK(np);
894 		return EAGAIN;
895 	}
896 	PROC_UNLOCK(np);
897 	return (0);
898 }
899 
900 static int
901 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
902 {
903 	int *name = (int*) arg1;
904 	u_int namelen = arg2;
905 	struct proc *p;
906 	int flags, doingzomb;
907 	int error = 0;
908 
909 	if (oidp->oid_number == KERN_PROC_PID) {
910 		if (namelen != 1)
911 			return (EINVAL);
912 		p = pfind((pid_t)name[0]);
913 		if (!p)
914 			return (ESRCH);
915 		if ((error = p_cansee(curthread, p))) {
916 			PROC_UNLOCK(p);
917 			return (error);
918 		}
919 		error = sysctl_out_proc(p, req, KERN_PROC_NOTHREADS);
920 		return (error);
921 	}
922 
923 	switch (oidp->oid_number) {
924 	case KERN_PROC_ALL:
925 	case KERN_PROC_PROC:
926 		/*
927 		 * XXX Temporarily disabled for compat with old userland
928 		 * that passes an extra unused argument to KERN_PROC_PROC.
929 		 */
930 #if 0
931 		if (namelen != 0)
932 			return (EINVAL);
933 #endif
934 		break;
935 	default:
936 		if (namelen != 1)
937 			return (EINVAL);
938 		break;
939 	}
940 
941 	if (!req->oldptr) {
942 		/* overestimate by 5 procs */
943 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
944 		if (error)
945 			return (error);
946 	}
947 	sysctl_wire_old_buffer(req, 0);
948 	sx_slock(&allproc_lock);
949 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
950 		if (!doingzomb)
951 			p = LIST_FIRST(&allproc);
952 		else
953 			p = LIST_FIRST(&zombproc);
954 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
955 			/*
956 			 * Skip embryonic processes.
957 			 */
958 			mtx_lock_spin(&sched_lock);
959 			if (p->p_state == PRS_NEW) {
960 				mtx_unlock_spin(&sched_lock);
961 				continue;
962 			}
963 			mtx_unlock_spin(&sched_lock);
964 			PROC_LOCK(p);
965 			/*
966 			 * Show a user only appropriate processes.
967 			 */
968 			if (p_cansee(curthread, p)) {
969 				PROC_UNLOCK(p);
970 				continue;
971 			}
972 			flags = 0;
973 			/*
974 			 * TODO - make more efficient (see notes below).
975 			 * do by session.
976 			 */
977 			switch (oidp->oid_number) {
978 
979 			case KERN_PROC_PGRP:
980 				/* could do this by traversing pgrp */
981 				if (p->p_pgrp == NULL ||
982 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
983 					PROC_UNLOCK(p);
984 					continue;
985 				}
986 				break;
987 
988 			case KERN_PROC_TTY:
989 				if ((p->p_flag & P_CONTROLT) == 0 ||
990 				    p->p_session == NULL) {
991 					PROC_UNLOCK(p);
992 					continue;
993 				}
994 				SESS_LOCK(p->p_session);
995 				if (p->p_session->s_ttyp == NULL ||
996 				    dev2udev(p->p_session->s_ttyp->t_dev) !=
997 				    (udev_t)name[0]) {
998 					SESS_UNLOCK(p->p_session);
999 					PROC_UNLOCK(p);
1000 					continue;
1001 				}
1002 				SESS_UNLOCK(p->p_session);
1003 				break;
1004 
1005 			case KERN_PROC_UID:
1006 				if (p->p_ucred == NULL ||
1007 				    p->p_ucred->cr_uid != (uid_t)name[0]) {
1008 					PROC_UNLOCK(p);
1009 					continue;
1010 				}
1011 				break;
1012 
1013 			case KERN_PROC_RUID:
1014 				if (p->p_ucred == NULL ||
1015 				    p->p_ucred->cr_ruid != (uid_t)name[0]) {
1016 					PROC_UNLOCK(p);
1017 					continue;
1018 				}
1019 				break;
1020 
1021 			case KERN_PROC_PROC:
1022 				flags |= KERN_PROC_NOTHREADS;
1023 				break;
1024 
1025 			default:
1026 				break;
1027 
1028 			}
1029 
1030 			error = sysctl_out_proc(p, req, flags | doingzomb);
1031 			if (error) {
1032 				sx_sunlock(&allproc_lock);
1033 				return (error);
1034 			}
1035 		}
1036 	}
1037 	sx_sunlock(&allproc_lock);
1038 	return (0);
1039 }
1040 
1041 struct pargs *
1042 pargs_alloc(int len)
1043 {
1044 	struct pargs *pa;
1045 
1046 	MALLOC(pa, struct pargs *, sizeof(struct pargs) + len, M_PARGS,
1047 		M_WAITOK);
1048 	pa->ar_ref = 1;
1049 	pa->ar_length = len;
1050 	return (pa);
1051 }
1052 
1053 void
1054 pargs_free(struct pargs *pa)
1055 {
1056 
1057 	FREE(pa, M_PARGS);
1058 }
1059 
1060 void
1061 pargs_hold(struct pargs *pa)
1062 {
1063 
1064 	if (pa == NULL)
1065 		return;
1066 	PARGS_LOCK(pa);
1067 	pa->ar_ref++;
1068 	PARGS_UNLOCK(pa);
1069 }
1070 
1071 void
1072 pargs_drop(struct pargs *pa)
1073 {
1074 
1075 	if (pa == NULL)
1076 		return;
1077 	PARGS_LOCK(pa);
1078 	if (--pa->ar_ref == 0) {
1079 		PARGS_UNLOCK(pa);
1080 		pargs_free(pa);
1081 	} else
1082 		PARGS_UNLOCK(pa);
1083 }
1084 
1085 /*
1086  * This sysctl allows a process to retrieve the argument list or process
1087  * title for another process without groping around in the address space
1088  * of the other process.  It also allow a process to set its own "process
1089  * title to a string of its own choice.
1090  */
1091 static int
1092 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1093 {
1094 	int *name = (int*) arg1;
1095 	u_int namelen = arg2;
1096 	struct pargs *newpa, *pa;
1097 	struct proc *p;
1098 	int error = 0;
1099 
1100 	if (namelen != 1)
1101 		return (EINVAL);
1102 
1103 	p = pfind((pid_t)name[0]);
1104 	if (!p)
1105 		return (ESRCH);
1106 
1107 	if ((!ps_argsopen) && (error = p_cansee(curthread, p))) {
1108 		PROC_UNLOCK(p);
1109 		return (error);
1110 	}
1111 
1112 	if (req->newptr && curproc != p) {
1113 		PROC_UNLOCK(p);
1114 		return (EPERM);
1115 	}
1116 
1117 	pa = p->p_args;
1118 	pargs_hold(pa);
1119 	PROC_UNLOCK(p);
1120 	if (req->oldptr != NULL && pa != NULL)
1121 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1122 	pargs_drop(pa);
1123 	if (error != 0 || req->newptr == NULL)
1124 		return (error);
1125 
1126 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1127 		return (ENOMEM);
1128 	newpa = pargs_alloc(req->newlen);
1129 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1130 	if (error != 0) {
1131 		pargs_free(newpa);
1132 		return (error);
1133 	}
1134 	PROC_LOCK(p);
1135 	pa = p->p_args;
1136 	p->p_args = newpa;
1137 	PROC_UNLOCK(p);
1138 	pargs_drop(pa);
1139 	return (0);
1140 }
1141 
1142 static int
1143 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1144 {
1145 	struct proc *p;
1146 	char *sv_name;
1147 	int *name;
1148 	int namelen;
1149 	int error;
1150 
1151 	namelen = arg2;
1152 	if (namelen != 1)
1153 		return (EINVAL);
1154 
1155 	name = (int *)arg1;
1156 	if ((p = pfind((pid_t)name[0])) == NULL)
1157 		return (ESRCH);
1158 	if ((error = p_cansee(curthread, p))) {
1159 		PROC_UNLOCK(p);
1160 		return (error);
1161 	}
1162 	sv_name = p->p_sysent->sv_name;
1163 	PROC_UNLOCK(p);
1164 	return (sysctl_handle_string(oidp, sv_name, 0, req));
1165 }
1166 
1167 
1168 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
1169 
1170 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT,
1171 	0, 0, sysctl_kern_proc, "S,proc", "Return entire process table");
1172 
1173 SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD,
1174 	sysctl_kern_proc, "Process table");
1175 
1176 SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD,
1177 	sysctl_kern_proc, "Process table");
1178 
1179 SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD,
1180 	sysctl_kern_proc, "Process table");
1181 
1182 SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD,
1183 	sysctl_kern_proc, "Process table");
1184 
1185 SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD,
1186 	sysctl_kern_proc, "Process table");
1187 
1188 SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD,
1189 	sysctl_kern_proc, "Return process table, no threads");
1190 
1191 SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, CTLFLAG_RW | CTLFLAG_ANYBODY,
1192 	sysctl_kern_proc_args, "Process argument list");
1193 
1194 SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD,
1195 	sysctl_kern_proc_sv_name, "Process syscall vector name (ABI type)");
1196