xref: /freebsd/sys/kern/kern_proc.c (revision a3e8fd0b7f663db7eafff527d5c3ca3bcfa8a537)
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 "opt_ktrace.h"
38 #include "opt_kstack_pages.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/kernel.h>
43 #include <sys/lock.h>
44 #include <sys/malloc.h>
45 #include <sys/mutex.h>
46 #include <sys/proc.h>
47 #include <sys/kse.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 struct proc *dopfind(register pid_t);
74 
75 static void doenterpgrp(struct proc *, struct pgrp *);
76 
77 static void pgdelete(struct pgrp *);
78 
79 static void orphanpg(struct pgrp *pg);
80 
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", sizeof (struct 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 	KASSERT((size == sizeof(struct proc)),
141 	    ("size mismatch: %d != %d\n", size, (int)sizeof(struct proc)));
142 	p = (struct proc *)mem;
143 }
144 
145 /*
146  * Reclaim a proc after use.
147  */
148 static void
149 proc_dtor(void *mem, int size, void *arg)
150 {
151 	struct proc *p;
152 	struct thread *td;
153 	struct ksegrp *kg;
154 	struct kse *ke;
155 
156 	/* INVARIANTS checks go here */
157 	KASSERT((size == sizeof(struct proc)),
158 	    ("size mismatch: %d != %d\n", size, (int)sizeof(struct proc)));
159 	p = (struct proc *)mem;
160 	KASSERT((p->p_numthreads == 1),
161 	    ("bad number of threads in exiting process"));
162         td = FIRST_THREAD_IN_PROC(p);
163 	KASSERT((td != NULL), ("proc_dtor: bad thread pointer"));
164         kg = FIRST_KSEGRP_IN_PROC(p);
165 	KASSERT((kg != NULL), ("proc_dtor: bad kg pointer"));
166         ke = FIRST_KSE_IN_KSEGRP(kg);
167 	KASSERT((ke != NULL), ("proc_dtor: bad ke pointer"));
168 
169 	/* Dispose of an alternate kstack, if it exists.
170 	 * XXX What if there are more than one thread in the proc?
171 	 *     The first thread in the proc is special and not
172 	 *     freed, so you gotta do this here.
173 	 */
174 	if (((p->p_flag & P_KTHREAD) != 0) && (td->td_altkstack != 0))
175 		pmap_dispose_altkstack(td);
176 
177 	/*
178 	 * We want to make sure we know the initial linkages.
179 	 * so for now tear them down and remake them.
180 	 * This is probably un-needed as we can probably rely
181 	 * on the state coming in here from wait4().
182 	 */
183 	proc_linkup(p, kg, ke, td);
184 }
185 
186 /*
187  * Initialize type-stable parts of a proc (when newly created).
188  */
189 static void
190 proc_init(void *mem, int size)
191 {
192 	struct proc *p;
193 	struct thread *td;
194 	struct ksegrp *kg;
195 	struct kse *ke;
196 
197 	KASSERT((size == sizeof(struct proc)),
198 	    ("size mismatch: %d != %d\n", size, (int)sizeof(struct proc)));
199 	p = (struct proc *)mem;
200 	vm_proc_new(p);
201 	td = thread_alloc();
202 	ke = kse_alloc();
203 	kg = ksegrp_alloc();
204 	proc_linkup(p, kg, ke, td);
205 }
206 
207 /*
208  * Tear down type-stable parts of a proc (just before being discarded)
209  */
210 static void
211 proc_fini(void *mem, int size)
212 {
213 	struct proc *p;
214 	struct thread *td;
215 	struct ksegrp *kg;
216 	struct kse *ke;
217 
218 	KASSERT((size == sizeof(struct proc)),
219 	    ("size mismatch: %d != %d\n", size, (int)sizeof(struct proc)));
220 	p = (struct proc *)mem;
221 	KASSERT((p->p_numthreads == 1),
222 	    ("bad number of threads in freeing process"));
223         td = FIRST_THREAD_IN_PROC(p);
224 	KASSERT((td != NULL), ("proc_dtor: bad thread pointer"));
225         kg = FIRST_KSEGRP_IN_PROC(p);
226 	KASSERT((kg != NULL), ("proc_dtor: bad kg pointer"));
227         ke = FIRST_KSE_IN_KSEGRP(kg);
228 	KASSERT((ke != NULL), ("proc_dtor: bad ke pointer"));
229 	vm_proc_dispose(p);
230 	thread_free(td);
231 	ksegrp_free(kg);
232 	kse_free(ke);
233 }
234 
235 /*
236  * Is p an inferior of the current process?
237  */
238 int
239 inferior(p)
240 	register struct proc *p;
241 {
242 
243 	sx_assert(&proctree_lock, SX_LOCKED);
244 	for (; p != curproc; p = p->p_pptr)
245 		if (p->p_pid == 0)
246 			return (0);
247 	return (1);
248 }
249 
250 /*
251  * Locate a process by number
252  */
253 struct proc *
254 pfind(pid)
255 	register pid_t pid;
256 {
257 	register struct proc *p;
258 
259 	sx_slock(&allproc_lock);
260 	p = dopfind(pid);
261 	sx_sunlock(&allproc_lock);
262 	return (p);
263 }
264 
265 static struct proc *
266 dopfind(pid)
267 	register pid_t pid;
268 {
269 	register struct proc *p;
270 
271 	sx_assert(&allproc_lock, SX_LOCKED);
272 
273 	LIST_FOREACH(p, PIDHASH(pid), p_hash)
274 		if (p->p_pid == pid) {
275 			PROC_LOCK(p);
276 			break;
277 		}
278 	return (p);
279 }
280 
281 /*
282  * Locate a process group by number.
283  * The caller must hold proctree_lock.
284  */
285 struct pgrp *
286 pgfind(pgid)
287 	register pid_t pgid;
288 {
289 	register struct pgrp *pgrp;
290 
291 	sx_assert(&proctree_lock, SX_LOCKED);
292 
293 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
294 		if (pgrp->pg_id == pgid) {
295 			PGRP_LOCK(pgrp);
296 			return (pgrp);
297 		}
298 	}
299 	return (NULL);
300 }
301 
302 /*
303  * Create a new process group.
304  * pgid must be equal to the pid of p.
305  * Begin a new session if required.
306  */
307 int
308 enterpgrp(p, pgid, pgrp, sess)
309 	register struct proc *p;
310 	pid_t pgid;
311 	struct pgrp *pgrp;
312 	struct session *sess;
313 {
314 	struct pgrp *pgrp2;
315 
316 	sx_assert(&proctree_lock, SX_XLOCKED);
317 
318 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
319 	KASSERT(p->p_pid == pgid,
320 	    ("enterpgrp: new pgrp and pid != pgid"));
321 
322 	pgrp2 = pgfind(pgid);
323 
324 	KASSERT(pgrp2 == NULL,
325 	    ("enterpgrp: pgrp with pgid exists"));
326 	KASSERT(!SESS_LEADER(p),
327 	    ("enterpgrp: session leader attempted setpgrp"));
328 
329 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
330 
331 	if (sess != NULL) {
332 		/*
333 		 * new session
334 		 */
335 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
336 		PROC_LOCK(p);
337 		p->p_flag &= ~P_CONTROLT;
338 		PROC_UNLOCK(p);
339 		PGRP_LOCK(pgrp);
340 		sess->s_leader = p;
341 		sess->s_sid = p->p_pid;
342 		sess->s_count = 1;
343 		sess->s_ttyvp = NULL;
344 		sess->s_ttyp = NULL;
345 		bcopy(p->p_session->s_login, sess->s_login,
346 			    sizeof(sess->s_login));
347 		pgrp->pg_session = sess;
348 		KASSERT(p == curproc,
349 		    ("enterpgrp: mksession and p != curproc"));
350 	} else {
351 		pgrp->pg_session = p->p_session;
352 		SESS_LOCK(pgrp->pg_session);
353 		pgrp->pg_session->s_count++;
354 		SESS_UNLOCK(pgrp->pg_session);
355 		PGRP_LOCK(pgrp);
356 	}
357 	pgrp->pg_id = pgid;
358 	LIST_INIT(&pgrp->pg_members);
359 
360 	/*
361 	 * As we have an exclusive lock of proctree_lock,
362 	 * this should not deadlock.
363 	 */
364 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
365 	pgrp->pg_jobc = 0;
366 	SLIST_INIT(&pgrp->pg_sigiolst);
367 	PGRP_UNLOCK(pgrp);
368 
369 	doenterpgrp(p, pgrp);
370 
371 	return (0);
372 }
373 
374 /*
375  * Move p to an existing process group
376  */
377 int
378 enterthispgrp(p, pgrp)
379 	register struct proc *p;
380 	struct pgrp *pgrp;
381 {
382 
383 	sx_assert(&proctree_lock, SX_XLOCKED);
384 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
385 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
386 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
387 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
388 	KASSERT(pgrp->pg_session == p->p_session,
389 		("%s: pgrp's session %p, p->p_session %p.\n",
390 		__func__,
391 		pgrp->pg_session,
392 		p->p_session));
393 	KASSERT(pgrp != p->p_pgrp,
394 		("%s: p belongs to pgrp.", __func__));
395 
396 	doenterpgrp(p, pgrp);
397 
398 	return (0);
399 }
400 
401 /*
402  * Move p to a process group
403  */
404 static void
405 doenterpgrp(p, pgrp)
406 	struct proc *p;
407 	struct pgrp *pgrp;
408 {
409 	struct pgrp *savepgrp;
410 
411 	sx_assert(&proctree_lock, SX_XLOCKED);
412 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
413 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
414 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
415 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
416 
417 	savepgrp = p->p_pgrp;
418 
419 	/*
420 	 * Adjust eligibility of affected pgrps to participate in job control.
421 	 * Increment eligibility counts before decrementing, otherwise we
422 	 * could reach 0 spuriously during the first call.
423 	 */
424 	fixjobc(p, pgrp, 1);
425 	fixjobc(p, p->p_pgrp, 0);
426 
427 	PGRP_LOCK(pgrp);
428 	PGRP_LOCK(savepgrp);
429 	PROC_LOCK(p);
430 	LIST_REMOVE(p, p_pglist);
431 	p->p_pgrp = pgrp;
432 	PROC_UNLOCK(p);
433 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
434 	PGRP_UNLOCK(savepgrp);
435 	PGRP_UNLOCK(pgrp);
436 	if (LIST_EMPTY(&savepgrp->pg_members))
437 		pgdelete(savepgrp);
438 }
439 
440 /*
441  * remove process from process group
442  */
443 int
444 leavepgrp(p)
445 	register struct proc *p;
446 {
447 	struct pgrp *savepgrp;
448 
449 	sx_assert(&proctree_lock, SX_XLOCKED);
450 	savepgrp = p->p_pgrp;
451 	PGRP_LOCK(savepgrp);
452 	PROC_LOCK(p);
453 	LIST_REMOVE(p, p_pglist);
454 	p->p_pgrp = NULL;
455 	PROC_UNLOCK(p);
456 	PGRP_UNLOCK(savepgrp);
457 	if (LIST_EMPTY(&savepgrp->pg_members))
458 		pgdelete(savepgrp);
459 	return (0);
460 }
461 
462 /*
463  * delete a process group
464  */
465 static void
466 pgdelete(pgrp)
467 	register struct pgrp *pgrp;
468 {
469 	struct session *savesess;
470 
471 	sx_assert(&proctree_lock, SX_XLOCKED);
472 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
473 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
474 
475 	/*
476 	 * Reset any sigio structures pointing to us as a result of
477 	 * F_SETOWN with our pgid.
478 	 */
479 	funsetownlst(&pgrp->pg_sigiolst);
480 
481 	PGRP_LOCK(pgrp);
482 	if (pgrp->pg_session->s_ttyp != NULL &&
483 	    pgrp->pg_session->s_ttyp->t_pgrp == pgrp)
484 		pgrp->pg_session->s_ttyp->t_pgrp = NULL;
485 	LIST_REMOVE(pgrp, pg_hash);
486 	savesess = pgrp->pg_session;
487 	SESS_LOCK(savesess);
488 	savesess->s_count--;
489 	SESS_UNLOCK(savesess);
490 	PGRP_UNLOCK(pgrp);
491 	if (savesess->s_count == 0) {
492 		mtx_destroy(&savesess->s_mtx);
493 		FREE(pgrp->pg_session, M_SESSION);
494 	}
495 	mtx_destroy(&pgrp->pg_mtx);
496 	FREE(pgrp, M_PGRP);
497 }
498 
499 /*
500  * Adjust pgrp jobc counters when specified process changes process group.
501  * We count the number of processes in each process group that "qualify"
502  * the group for terminal job control (those with a parent in a different
503  * process group of the same session).  If that count reaches zero, the
504  * process group becomes orphaned.  Check both the specified process'
505  * process group and that of its children.
506  * entering == 0 => p is leaving specified group.
507  * entering == 1 => p is entering specified group.
508  */
509 void
510 fixjobc(p, pgrp, entering)
511 	register struct proc *p;
512 	register struct pgrp *pgrp;
513 	int entering;
514 {
515 	register struct pgrp *hispgrp;
516 	register struct session *mysession;
517 
518 	sx_assert(&proctree_lock, SX_LOCKED);
519 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
520 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
521 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
522 
523 	/*
524 	 * Check p's parent to see whether p qualifies its own process
525 	 * group; if so, adjust count for p's process group.
526 	 */
527 	mysession = pgrp->pg_session;
528 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
529 	    hispgrp->pg_session == mysession) {
530 		PGRP_LOCK(pgrp);
531 		if (entering)
532 			pgrp->pg_jobc++;
533 		else {
534 			--pgrp->pg_jobc;
535 			if (pgrp->pg_jobc == 0)
536 				orphanpg(pgrp);
537 		}
538 		PGRP_UNLOCK(pgrp);
539 	}
540 
541 	/*
542 	 * Check this process' children to see whether they qualify
543 	 * their process groups; if so, adjust counts for children's
544 	 * process groups.
545 	 */
546 	LIST_FOREACH(p, &p->p_children, p_sibling) {
547 		if ((hispgrp = p->p_pgrp) != pgrp &&
548 		    hispgrp->pg_session == mysession &&
549 		    p->p_state != PRS_ZOMBIE) {
550 			PGRP_LOCK(hispgrp);
551 			if (entering)
552 				hispgrp->pg_jobc++;
553 			else {
554 				--hispgrp->pg_jobc;
555 				if (hispgrp->pg_jobc == 0)
556 					orphanpg(hispgrp);
557 			}
558 			PGRP_UNLOCK(hispgrp);
559 		}
560 	}
561 }
562 
563 /*
564  * A process group has become orphaned;
565  * if there are any stopped processes in the group,
566  * hang-up all process in that group.
567  */
568 static void
569 orphanpg(pg)
570 	struct pgrp *pg;
571 {
572 	register struct proc *p;
573 
574 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
575 
576 	mtx_lock_spin(&sched_lock);
577 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
578 		if (P_SHOULDSTOP(p)) {
579 			mtx_unlock_spin(&sched_lock);
580 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
581 				PROC_LOCK(p);
582 				psignal(p, SIGHUP);
583 				psignal(p, SIGCONT);
584 				PROC_UNLOCK(p);
585 			}
586 			return;
587 		}
588 	}
589 	mtx_unlock_spin(&sched_lock);
590 }
591 
592 #include "opt_ddb.h"
593 #ifdef DDB
594 #include <ddb/ddb.h>
595 
596 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
597 {
598 	register struct pgrp *pgrp;
599 	register struct proc *p;
600 	register int i;
601 
602 	for (i = 0; i <= pgrphash; i++) {
603 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
604 			printf("\tindx %d\n", i);
605 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
606 				printf(
607 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
608 				    (void *)pgrp, (long)pgrp->pg_id,
609 				    (void *)pgrp->pg_session,
610 				    pgrp->pg_session->s_count,
611 				    (void *)LIST_FIRST(&pgrp->pg_members));
612 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
613 					printf("\t\tpid %ld addr %p pgrp %p\n",
614 					    (long)p->p_pid, (void *)p,
615 					    (void *)p->p_pgrp);
616 				}
617 			}
618 		}
619 	}
620 }
621 #endif /* DDB */
622 
623 /*
624  * Fill in an kinfo_proc structure for the specified process.
625  * Must be called with the target process locked.
626  */
627 void
628 fill_kinfo_proc(p, kp)
629 	struct proc *p;
630 	struct kinfo_proc *kp;
631 {
632 	struct thread *td;
633 	struct kse *ke;
634 	struct ksegrp *kg;
635 	struct tty *tp;
636 	struct session *sp;
637 	struct timeval tv;
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_tracep;
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_procsig) {
667 		kp->ki_sigignore = p->p_procsig->ps_sigignore;
668 		kp->ki_sigcatch = p->p_procsig->ps_sigcatch;
669 	}
670 	mtx_lock_spin(&sched_lock);
671 	if (p->p_state != PRS_NEW &&
672 	    p->p_state != PRS_ZOMBIE &&
673 	    p->p_vmspace != NULL) {
674 		struct vmspace *vm = p->p_vmspace;
675 
676 		kp->ki_size = vm->vm_map.size;
677 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
678 		if (p->p_sflag & PS_INMEM)
679 			kp->ki_rssize += UAREA_PAGES;
680 		FOREACH_THREAD_IN_PROC(p, td) /* XXXKSE: thread swapout check */
681 			kp->ki_rssize += KSTACK_PAGES;
682 		kp->ki_swrss = vm->vm_swrss;
683 		kp->ki_tsize = vm->vm_tsize;
684 		kp->ki_dsize = vm->vm_dsize;
685 		kp->ki_ssize = vm->vm_ssize;
686 	}
687 	if ((p->p_sflag & PS_INMEM) && p->p_stats) {
688 		kp->ki_start = p->p_stats->p_start;
689 		kp->ki_rusage = p->p_stats->p_ru;
690 		kp->ki_childtime.tv_sec = p->p_stats->p_cru.ru_utime.tv_sec +
691 		    p->p_stats->p_cru.ru_stime.tv_sec;
692 		kp->ki_childtime.tv_usec = p->p_stats->p_cru.ru_utime.tv_usec +
693 		    p->p_stats->p_cru.ru_stime.tv_usec;
694 	}
695 	if (p->p_state != PRS_ZOMBIE) {
696 		td = FIRST_THREAD_IN_PROC(p);
697 		if (td == NULL) {
698 			/* XXXKSE: This should never happen. */
699 			printf("fill_kinfo_proc(): pid %d has no threads!\n",
700 			    p->p_pid);
701 			mtx_unlock_spin(&sched_lock);
702 			return;
703 		}
704 		if (!(p->p_flag & P_KSES)) {
705 			if (td->td_wmesg != NULL) {
706 				strlcpy(kp->ki_wmesg, td->td_wmesg,
707 				    sizeof(kp->ki_wmesg));
708 			}
709 			if (TD_ON_LOCK(td)) {
710 				kp->ki_kiflag |= KI_LOCKBLOCK;
711 				strlcpy(kp->ki_lockname, td->td_lockname,
712 				    sizeof(kp->ki_lockname));
713 			}
714 		}
715 
716 		if (p->p_state == PRS_NORMAL) { /*  XXXKSE very approximate */
717 			if (TD_ON_RUNQ(td) ||
718 			    TD_CAN_RUN(td) ||
719 			    TD_IS_RUNNING(td)) {
720 				kp->ki_stat = SRUN;
721 			} else if (P_SHOULDSTOP(p)) {
722 				kp->ki_stat = SSTOP;
723 			} else if (TD_IS_SLEEPING(td)) {
724 				kp->ki_stat = SSLEEP;
725 			} else if (TD_ON_LOCK(td)) {
726 				kp->ki_stat = SLOCK;
727 			} else {
728 				kp->ki_stat = SWAIT;
729 			}
730 		} else {
731 			kp->ki_stat = SIDL;
732 		}
733 
734 		kp->ki_sflag = p->p_sflag;
735 		kp->ki_swtime = p->p_swtime;
736 		kp->ki_pid = p->p_pid;
737 		/* vvv XXXKSE */
738 		if (!(p->p_flag & P_KSES)) {
739 			kg = td->td_ksegrp;
740 			ke = td->td_kse;
741 			KASSERT((ke != NULL), ("fill_kinfo_proc: Null KSE"));
742 			bintime2timeval(&p->p_runtime, &tv);
743 			kp->ki_runtime =
744 			    tv.tv_sec * (u_int64_t)1000000 + tv.tv_usec;
745 
746 			/* things in the KSE GROUP */
747 			kp->ki_estcpu = kg->kg_estcpu;
748 			kp->ki_slptime = kg->kg_slptime;
749 			kp->ki_pri.pri_user = kg->kg_user_pri;
750 			kp->ki_pri.pri_class = kg->kg_pri_class;
751 			kp->ki_nice = kg->kg_nice;
752 
753 			/* Things in the thread */
754 			kp->ki_wchan = td->td_wchan;
755 			kp->ki_pri.pri_level = td->td_priority;
756 			kp->ki_pri.pri_native = td->td_base_pri;
757 			kp->ki_lastcpu = td->td_lastcpu;
758 			kp->ki_tdflags = td->td_flags;
759 			kp->ki_pcb = td->td_pcb;
760 			kp->ki_kstack = (void *)td->td_kstack;
761 
762 			/* Things in the kse */
763 			kp->ki_rqindex = ke->ke_rqindex;
764 			kp->ki_oncpu = ke->ke_oncpu;
765 			kp->ki_pctcpu = ke->ke_pctcpu;
766 		} else {
767 			kp->ki_oncpu = -1;
768 			kp->ki_lastcpu = -1;
769 			kp->ki_tdflags = -1;
770 			/* All the rest are 0 for now */
771 		}
772 		/* ^^^ XXXKSE */
773 	} else {
774 		kp->ki_stat = SZOMB;
775 	}
776 	mtx_unlock_spin(&sched_lock);
777 	sp = NULL;
778 	tp = NULL;
779 	if (p->p_pgrp) {
780 		kp->ki_pgid = p->p_pgrp->pg_id;
781 		kp->ki_jobc = p->p_pgrp->pg_jobc;
782 		sp = p->p_pgrp->pg_session;
783 
784 		if (sp != NULL) {
785 			kp->ki_sid = sp->s_sid;
786 			SESS_LOCK(sp);
787 			strlcpy(kp->ki_login, sp->s_login,
788 			    sizeof(kp->ki_login));
789 			if (sp->s_ttyvp)
790 				kp->ki_kiflag |= KI_CTTY;
791 			if (SESS_LEADER(p))
792 				kp->ki_kiflag |= KI_SLEADER;
793 			tp = sp->s_ttyp;
794 			SESS_UNLOCK(sp);
795 		}
796 	}
797 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
798 		kp->ki_tdev = dev2udev(tp->t_dev);
799 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
800 		if (tp->t_session)
801 			kp->ki_tsid = tp->t_session->s_sid;
802 	} else
803 		kp->ki_tdev = NOUDEV;
804 	if (p->p_comm[0] != '\0') {
805 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
806 		strlcpy(kp->ki_ocomm, p->p_comm, sizeof(kp->ki_ocomm));
807 	}
808 	kp->ki_siglist = p->p_siglist;
809 	kp->ki_sigmask = p->p_sigmask;
810 	kp->ki_xstat = p->p_xstat;
811 	kp->ki_acflag = p->p_acflag;
812 	kp->ki_flag = p->p_flag;
813 	/* If jailed(p->p_ucred), emulate the old P_JAILED flag. */
814 	if (jailed(p->p_ucred))
815 		kp->ki_flag |= P_JAILED;
816 	kp->ki_lock = p->p_lock;
817 	if (p->p_pptr)
818 		kp->ki_ppid = p->p_pptr->p_pid;
819 }
820 
821 /*
822  * Locate a zombie process by number
823  */
824 struct proc *
825 zpfind(pid_t pid)
826 {
827 	struct proc *p;
828 
829 	sx_slock(&allproc_lock);
830 	LIST_FOREACH(p, &zombproc, p_list)
831 		if (p->p_pid == pid) {
832 			PROC_LOCK(p);
833 			break;
834 		}
835 	sx_sunlock(&allproc_lock);
836 	return (p);
837 }
838 
839 
840 /*
841  * Must be called with the process locked and will return with it unlocked.
842  */
843 static int
844 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int doingzomb)
845 {
846 	struct kinfo_proc kinfo_proc;
847 	int error;
848 	struct proc *np;
849 	pid_t pid = p->p_pid;
850 
851 	PROC_LOCK_ASSERT(p, MA_OWNED);
852 	fill_kinfo_proc(p, &kinfo_proc);
853 	PROC_UNLOCK(p);
854 	error = SYSCTL_OUT(req, (caddr_t)&kinfo_proc, sizeof(kinfo_proc));
855 	if (error)
856 		return (error);
857 	if (doingzomb)
858 		np = zpfind(pid);
859 	else {
860 		if (pid == 0)
861 			return (0);
862 		np = pfind(pid);
863 	}
864 	if (np == NULL)
865 		return EAGAIN;
866 	if (np != p) {
867 		PROC_UNLOCK(np);
868 		return EAGAIN;
869 	}
870 	PROC_UNLOCK(np);
871 	return (0);
872 }
873 
874 static int
875 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
876 {
877 	int *name = (int*) arg1;
878 	u_int namelen = arg2;
879 	struct proc *p;
880 	int doingzomb;
881 	int error = 0;
882 
883 	if (oidp->oid_number == KERN_PROC_PID) {
884 		if (namelen != 1)
885 			return (EINVAL);
886 		p = pfind((pid_t)name[0]);
887 		if (!p)
888 			return (0);
889 		if (p_cansee(curthread, p)) {
890 			PROC_UNLOCK(p);
891 			return (0);
892 		}
893 		error = sysctl_out_proc(p, req, 0);
894 		return (error);
895 	}
896 	if (oidp->oid_number == KERN_PROC_ALL && !namelen)
897 		;
898 	else if (oidp->oid_number != KERN_PROC_ALL && namelen == 1)
899 		;
900 	else
901 		return (EINVAL);
902 
903 	if (!req->oldptr) {
904 		/* overestimate by 5 procs */
905 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
906 		if (error)
907 			return (error);
908 	}
909 	sysctl_wire_old_buffer(req, 0);
910 	sx_slock(&allproc_lock);
911 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
912 		if (!doingzomb)
913 			p = LIST_FIRST(&allproc);
914 		else
915 			p = LIST_FIRST(&zombproc);
916 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
917 			PROC_LOCK(p);
918 			/*
919 			 * Show a user only appropriate processes.
920 			 */
921 			if (p_cansee(curthread, p)) {
922 				PROC_UNLOCK(p);
923 				continue;
924 			}
925 			/*
926 			 * Skip embryonic processes.
927 			 */
928 			if (p->p_state == PRS_NEW) {
929 				PROC_UNLOCK(p);
930 				continue;
931 			}
932 			/*
933 			 * TODO - make more efficient (see notes below).
934 			 * do by session.
935 			 */
936 			switch (oidp->oid_number) {
937 
938 			case KERN_PROC_PGRP:
939 				/* could do this by traversing pgrp */
940 				if (p->p_pgrp == NULL ||
941 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
942 					PROC_UNLOCK(p);
943 					continue;
944 				}
945 				break;
946 
947 			case KERN_PROC_TTY:
948 				if ((p->p_flag & P_CONTROLT) == 0 ||
949 				    p->p_session == NULL) {
950 					PROC_UNLOCK(p);
951 					continue;
952 				}
953 				SESS_LOCK(p->p_session);
954 				if (p->p_session->s_ttyp == NULL ||
955 				    dev2udev(p->p_session->s_ttyp->t_dev) !=
956 				    (udev_t)name[0]) {
957 					SESS_UNLOCK(p->p_session);
958 					PROC_UNLOCK(p);
959 					continue;
960 				}
961 				SESS_UNLOCK(p->p_session);
962 				break;
963 
964 			case KERN_PROC_UID:
965 				if (p->p_ucred == NULL ||
966 				    p->p_ucred->cr_uid != (uid_t)name[0]) {
967 					PROC_UNLOCK(p);
968 					continue;
969 				}
970 				break;
971 
972 			case KERN_PROC_RUID:
973 				if (p->p_ucred == NULL ||
974 				    p->p_ucred->cr_ruid != (uid_t)name[0]) {
975 					PROC_UNLOCK(p);
976 					continue;
977 				}
978 				break;
979 			}
980 
981 			error = sysctl_out_proc(p, req, doingzomb);
982 			if (error) {
983 				sx_sunlock(&allproc_lock);
984 				return (error);
985 			}
986 		}
987 	}
988 	sx_sunlock(&allproc_lock);
989 	return (0);
990 }
991 
992 struct pargs *
993 pargs_alloc(int len)
994 {
995 	struct pargs *pa;
996 
997 	MALLOC(pa, struct pargs *, sizeof(struct pargs) + len, M_PARGS,
998 		M_WAITOK);
999 	pa->ar_ref = 1;
1000 	pa->ar_length = len;
1001 	return (pa);
1002 }
1003 
1004 void
1005 pargs_free(struct pargs *pa)
1006 {
1007 
1008 	FREE(pa, M_PARGS);
1009 }
1010 
1011 void
1012 pargs_hold(struct pargs *pa)
1013 {
1014 
1015 	if (pa == NULL)
1016 		return;
1017 	PARGS_LOCK(pa);
1018 	pa->ar_ref++;
1019 	PARGS_UNLOCK(pa);
1020 }
1021 
1022 void
1023 pargs_drop(struct pargs *pa)
1024 {
1025 
1026 	if (pa == NULL)
1027 		return;
1028 	PARGS_LOCK(pa);
1029 	if (--pa->ar_ref == 0) {
1030 		PARGS_UNLOCK(pa);
1031 		pargs_free(pa);
1032 	} else
1033 		PARGS_UNLOCK(pa);
1034 }
1035 
1036 /*
1037  * This sysctl allows a process to retrieve the argument list or process
1038  * title for another process without groping around in the address space
1039  * of the other process.  It also allow a process to set its own "process
1040  * title to a string of its own choice.
1041  */
1042 static int
1043 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1044 {
1045 	int *name = (int*) arg1;
1046 	u_int namelen = arg2;
1047 	struct proc *p;
1048 	struct pargs *pa;
1049 	int error = 0;
1050 
1051 	if (namelen != 1)
1052 		return (EINVAL);
1053 
1054 	p = pfind((pid_t)name[0]);
1055 	if (!p)
1056 		return (0);
1057 
1058 	if ((!ps_argsopen) && p_cansee(curthread, p)) {
1059 		PROC_UNLOCK(p);
1060 		return (0);
1061 	}
1062 	PROC_UNLOCK(p);
1063 
1064 	if (req->newptr && curproc != p)
1065 		return (EPERM);
1066 
1067 	PROC_LOCK(p);
1068 	pa = p->p_args;
1069 	pargs_hold(pa);
1070 	PROC_UNLOCK(p);
1071 	if (req->oldptr && pa != NULL) {
1072 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1073 	}
1074 	pargs_drop(pa);
1075 	if (req->newptr == NULL)
1076 		return (error);
1077 
1078 	PROC_LOCK(p);
1079 	pa = p->p_args;
1080 	p->p_args = NULL;
1081 	PROC_UNLOCK(p);
1082 	pargs_drop(pa);
1083 
1084 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1085 		return (error);
1086 
1087 	pa = pargs_alloc(req->newlen);
1088 	error = SYSCTL_IN(req, pa->ar_args, req->newlen);
1089 	if (!error) {
1090 		PROC_LOCK(p);
1091 		p->p_args = pa;
1092 		PROC_UNLOCK(p);
1093 	} else
1094 		pargs_free(pa);
1095 	return (error);
1096 }
1097 
1098 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
1099 
1100 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT,
1101 	0, 0, sysctl_kern_proc, "S,proc", "Return entire process table");
1102 
1103 SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD,
1104 	sysctl_kern_proc, "Process table");
1105 
1106 SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD,
1107 	sysctl_kern_proc, "Process table");
1108 
1109 SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD,
1110 	sysctl_kern_proc, "Process table");
1111 
1112 SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD,
1113 	sysctl_kern_proc, "Process table");
1114 
1115 SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD,
1116 	sysctl_kern_proc, "Process table");
1117 
1118 SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, CTLFLAG_RW | CTLFLAG_ANYBODY,
1119 	sysctl_kern_proc_args, "Process argument list");
1120 
1121