xref: /dflybsd-src/sys/kern/kern_proc.c (revision 07caec20a93f40352fd28eec40bb9b40199edf40)
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: src/sys/kern/kern_proc.c,v 1.63.2.9 2003/05/08 07:47:16 kbyanc Exp $
35  * $DragonFly: src/sys/kern/kern_proc.c,v 1.29 2006/10/10 15:40:46 dillon Exp $
36  */
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/kernel.h>
41 #include <sys/sysctl.h>
42 #include <sys/malloc.h>
43 #include <sys/proc.h>
44 #include <sys/jail.h>
45 #include <sys/filedesc.h>
46 #include <sys/tty.h>
47 #include <sys/signalvar.h>
48 #include <sys/spinlock.h>
49 #include <vm/vm.h>
50 #include <sys/lock.h>
51 #include <vm/pmap.h>
52 #include <vm/vm_map.h>
53 #include <sys/user.h>
54 #include <vm/vm_zone.h>
55 #include <machine/smp.h>
56 
57 #include <sys/spinlock2.h>
58 
59 static MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
60 MALLOC_DEFINE(M_SESSION, "session", "session header");
61 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
62 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
63 
64 int ps_showallprocs = 1;
65 static int ps_showallthreads = 1;
66 SYSCTL_INT(_kern, OID_AUTO, ps_showallprocs, CTLFLAG_RW,
67     &ps_showallprocs, 0, "");
68 SYSCTL_INT(_kern, OID_AUTO, ps_showallthreads, CTLFLAG_RW,
69     &ps_showallthreads, 0, "");
70 
71 static void pgdelete(struct pgrp *);
72 static void orphanpg(struct pgrp *pg);
73 static pid_t proc_getnewpid_locked(int random_offset);
74 
75 /*
76  * Other process lists
77  */
78 struct pidhashhead *pidhashtbl;
79 u_long pidhash;
80 struct pgrphashhead *pgrphashtbl;
81 u_long pgrphash;
82 struct proclist allproc;
83 struct proclist zombproc;
84 struct spinlock allproc_spin;
85 vm_zone_t proc_zone;
86 vm_zone_t thread_zone;
87 
88 /*
89  * Random component to nextpid generation.  We mix in a random factor to make
90  * it a little harder to predict.  We sanity check the modulus value to avoid
91  * doing it in critical paths.  Don't let it be too small or we pointlessly
92  * waste randomness entropy, and don't let it be impossibly large.  Using a
93  * modulus that is too big causes a LOT more process table scans and slows
94  * down fork processing as the pidchecked caching is defeated.
95  */
96 static int randompid = 0;
97 
98 static int
99 sysctl_kern_randompid(SYSCTL_HANDLER_ARGS)
100 {
101 	int error, pid;
102 
103 	pid = randompid;
104 	error = sysctl_handle_int(oidp, &pid, 0, req);
105 	if (error || !req->newptr)
106 		return (error);
107 	if (pid < 0 || pid > PID_MAX - 100)     /* out of range */
108 		pid = PID_MAX - 100;
109 	else if (pid < 2)                       /* NOP */
110 		pid = 0;
111 	else if (pid < 100)                     /* Make it reasonable */
112 		pid = 100;
113 	randompid = pid;
114 	return (error);
115 }
116 
117 SYSCTL_PROC(_kern, OID_AUTO, randompid, CTLTYPE_INT|CTLFLAG_RW,
118 	    0, 0, sysctl_kern_randompid, "I", "Random PID modulus");
119 
120 /*
121  * Initialize global process hashing structures.
122  */
123 void
124 procinit(void)
125 {
126 	LIST_INIT(&allproc);
127 	LIST_INIT(&zombproc);
128 	spin_init(&allproc_spin);
129 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
130 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
131 	proc_zone = zinit("PROC", sizeof (struct proc), 0, 0, 5);
132 	thread_zone = zinit("THREAD", sizeof (struct thread), 0, 0, 5);
133 	uihashinit();
134 }
135 
136 /*
137  * Is p an inferior of the current process?
138  */
139 int
140 inferior(struct proc *p)
141 {
142 	for (; p != curproc; p = p->p_pptr)
143 		if (p->p_pid == 0)
144 			return (0);
145 	return (1);
146 }
147 
148 /*
149  * Locate a process by number
150  */
151 struct proc *
152 pfind(pid_t pid)
153 {
154 	struct proc *p;
155 
156 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
157 		if (p->p_pid == pid)
158 			return (p);
159 	}
160 	return (NULL);
161 }
162 
163 /*
164  * Locate a process group by number
165  */
166 struct pgrp *
167 pgfind(pid_t pgid)
168 {
169 	struct pgrp *pgrp;
170 
171 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
172 		if (pgrp->pg_id == pgid)
173 			return (pgrp);
174 	}
175 	return (NULL);
176 }
177 
178 /*
179  * Move p to a new or existing process group (and session)
180  */
181 int
182 enterpgrp(struct proc *p, pid_t pgid, int mksess)
183 {
184 	struct pgrp *pgrp = pgfind(pgid);
185 
186 	KASSERT(pgrp == NULL || !mksess,
187 	    ("enterpgrp: setsid into non-empty pgrp"));
188 	KASSERT(!SESS_LEADER(p),
189 	    ("enterpgrp: session leader attempted setpgrp"));
190 
191 	if (pgrp == NULL) {
192 		pid_t savepid = p->p_pid;
193 		struct proc *np;
194 		/*
195 		 * new process group
196 		 */
197 		KASSERT(p->p_pid == pgid,
198 		    ("enterpgrp: new pgrp and pid != pgid"));
199 		if ((np = pfind(savepid)) == NULL || np != p)
200 			return (ESRCH);
201 		MALLOC(pgrp, struct pgrp *, sizeof(struct pgrp), M_PGRP,
202 		    M_WAITOK);
203 		if (mksess) {
204 			struct session *sess;
205 
206 			/*
207 			 * new session
208 			 */
209 			MALLOC(sess, struct session *, sizeof(struct session),
210 			    M_SESSION, M_WAITOK);
211 			sess->s_leader = p;
212 			sess->s_sid = p->p_pid;
213 			sess->s_count = 1;
214 			sess->s_ttyvp = NULL;
215 			sess->s_ttyp = NULL;
216 			bcopy(p->p_session->s_login, sess->s_login,
217 			    sizeof(sess->s_login));
218 			p->p_flag &= ~P_CONTROLT;
219 			pgrp->pg_session = sess;
220 			KASSERT(p == curproc,
221 			    ("enterpgrp: mksession and p != curproc"));
222 		} else {
223 			pgrp->pg_session = p->p_session;
224 			sess_hold(pgrp->pg_session);
225 		}
226 		pgrp->pg_id = pgid;
227 		LIST_INIT(&pgrp->pg_members);
228 		LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
229 		pgrp->pg_jobc = 0;
230 		SLIST_INIT(&pgrp->pg_sigiolst);
231 		lockinit(&pgrp->pg_lock, "pgwt", 0, 0);
232 	} else if (pgrp == p->p_pgrp)
233 		return (0);
234 
235 	/*
236 	 * Adjust eligibility of affected pgrps to participate in job control.
237 	 * Increment eligibility counts before decrementing, otherwise we
238 	 * could reach 0 spuriously during the first call.
239 	 */
240 	fixjobc(p, pgrp, 1);
241 	fixjobc(p, p->p_pgrp, 0);
242 
243 	LIST_REMOVE(p, p_pglist);
244 	if (LIST_EMPTY(&p->p_pgrp->pg_members))
245 		pgdelete(p->p_pgrp);
246 	p->p_pgrp = pgrp;
247 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
248 	return (0);
249 }
250 
251 /*
252  * remove process from process group
253  */
254 int
255 leavepgrp(struct proc *p)
256 {
257 
258 	LIST_REMOVE(p, p_pglist);
259 	if (LIST_EMPTY(&p->p_pgrp->pg_members))
260 		pgdelete(p->p_pgrp);
261 	p->p_pgrp = 0;
262 	return (0);
263 }
264 
265 /*
266  * delete a process group
267  */
268 static void
269 pgdelete(struct pgrp *pgrp)
270 {
271 
272 	/*
273 	 * Reset any sigio structures pointing to us as a result of
274 	 * F_SETOWN with our pgid.
275 	 */
276 	funsetownlst(&pgrp->pg_sigiolst);
277 
278 	if (pgrp->pg_session->s_ttyp != NULL &&
279 	    pgrp->pg_session->s_ttyp->t_pgrp == pgrp)
280 		pgrp->pg_session->s_ttyp->t_pgrp = NULL;
281 	LIST_REMOVE(pgrp, pg_hash);
282 	sess_rele(pgrp->pg_session);
283 	kfree(pgrp, M_PGRP);
284 }
285 
286 /*
287  * Adjust the ref count on a session structure.  When the ref count falls to
288  * zero the tty is disassociated from the session and the session structure
289  * is freed.  Note that tty assocation is not itself ref-counted.
290  */
291 void
292 sess_hold(struct session *sp)
293 {
294 	++sp->s_count;
295 }
296 
297 void
298 sess_rele(struct session *sp)
299 {
300 	KKASSERT(sp->s_count > 0);
301 	if (--sp->s_count == 0) {
302 		if (sp->s_ttyp && sp->s_ttyp->t_session) {
303 #ifdef TTY_DO_FULL_CLOSE
304 			/* FULL CLOSE, see ttyclearsession() */
305 			KKASSERT(sp->s_ttyp->t_session == sp);
306 			sp->s_ttyp->t_session = NULL;
307 #else
308 			/* HALF CLOSE, see ttyclearsession() */
309 			if (sp->s_ttyp->t_session == sp)
310 				sp->s_ttyp->t_session = NULL;
311 #endif
312 		}
313 		kfree(sp, M_SESSION);
314 	}
315 }
316 
317 /*
318  * Adjust pgrp jobc counters when specified process changes process group.
319  * We count the number of processes in each process group that "qualify"
320  * the group for terminal job control (those with a parent in a different
321  * process group of the same session).  If that count reaches zero, the
322  * process group becomes orphaned.  Check both the specified process'
323  * process group and that of its children.
324  * entering == 0 => p is leaving specified group.
325  * entering == 1 => p is entering specified group.
326  */
327 void
328 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
329 {
330 	struct pgrp *hispgrp;
331 	struct session *mysession = pgrp->pg_session;
332 
333 	/*
334 	 * Check p's parent to see whether p qualifies its own process
335 	 * group; if so, adjust count for p's process group.
336 	 */
337 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
338 	    hispgrp->pg_session == mysession) {
339 		if (entering)
340 			pgrp->pg_jobc++;
341 		else if (--pgrp->pg_jobc == 0)
342 			orphanpg(pgrp);
343 	}
344 
345 	/*
346 	 * Check this process' children to see whether they qualify
347 	 * their process groups; if so, adjust counts for children's
348 	 * process groups.
349 	 */
350 	LIST_FOREACH(p, &p->p_children, p_sibling)
351 		if ((hispgrp = p->p_pgrp) != pgrp &&
352 		    hispgrp->pg_session == mysession &&
353 		    (p->p_flag & P_ZOMBIE) == 0) {
354 			if (entering)
355 				hispgrp->pg_jobc++;
356 			else if (--hispgrp->pg_jobc == 0)
357 				orphanpg(hispgrp);
358 		}
359 }
360 
361 /*
362  * A process group has become orphaned;
363  * if there are any stopped processes in the group,
364  * hang-up all process in that group.
365  */
366 static void
367 orphanpg(struct pgrp *pg)
368 {
369 	struct proc *p;
370 
371 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
372 		if (p->p_flag & P_STOPPED) {
373 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
374 				ksignal(p, SIGHUP);
375 				ksignal(p, SIGCONT);
376 			}
377 			return;
378 		}
379 	}
380 }
381 
382 /*
383  * Add a new process to the allproc list and the PID hash.  This
384  * also assigns a pid to the new process.
385  *
386  * MPALMOSTSAFE - acquires mplock for karc4random() call
387  */
388 void
389 proc_add_allproc(struct proc *p)
390 {
391 	int random_offset;
392 
393 	if ((random_offset = randompid) != 0) {
394 		get_mplock();
395 		random_offset = karc4random() % random_offset;
396 		rel_mplock();
397 	}
398 
399 	spin_lock_wr(&allproc_spin);
400 	p->p_pid = proc_getnewpid_locked(random_offset);
401 	LIST_INSERT_HEAD(&allproc, p, p_list);
402 	LIST_INSERT_HEAD(PIDHASH(p->p_pid), p, p_hash);
403 	spin_unlock_wr(&allproc_spin);
404 }
405 
406 /*
407  * Calculate a new process pid.  This function is integrated into
408  * proc_add_allproc() to guarentee that the new pid is not reused before
409  * the new process can be added to the allproc list.
410  *
411  * MPSAFE - must be called with allproc_spin held.
412  */
413 static
414 pid_t
415 proc_getnewpid_locked(int random_offset)
416 {
417 	static pid_t nextpid;
418 	static pid_t pidchecked;
419 	struct proc *p;
420 
421 	/*
422 	 * Find an unused process ID.  We remember a range of unused IDs
423 	 * ready to use (from nextpid+1 through pidchecked-1).
424 	 */
425 	nextpid = nextpid + 1 + random_offset;
426 retry:
427 	/*
428 	 * If the process ID prototype has wrapped around,
429 	 * restart somewhat above 0, as the low-numbered procs
430 	 * tend to include daemons that don't exit.
431 	 */
432 	if (nextpid >= PID_MAX) {
433 		nextpid = nextpid % PID_MAX;
434 		if (nextpid < 100)
435 			nextpid += 100;
436 		pidchecked = 0;
437 	}
438 	if (nextpid >= pidchecked) {
439 		int doingzomb = 0;
440 
441 		pidchecked = PID_MAX;
442 		/*
443 		 * Scan the active and zombie procs to check whether this pid
444 		 * is in use.  Remember the lowest pid that's greater
445 		 * than nextpid, so we can avoid checking for a while.
446 		 */
447 		p = LIST_FIRST(&allproc);
448 again:
449 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
450 			while (p->p_pid == nextpid ||
451 			    p->p_pgrp->pg_id == nextpid ||
452 			    p->p_session->s_sid == nextpid) {
453 				nextpid++;
454 				if (nextpid >= pidchecked)
455 					goto retry;
456 			}
457 			if (p->p_pid > nextpid && pidchecked > p->p_pid)
458 				pidchecked = p->p_pid;
459 			if (p->p_pgrp->pg_id > nextpid &&
460 			    pidchecked > p->p_pgrp->pg_id)
461 				pidchecked = p->p_pgrp->pg_id;
462 			if (p->p_session->s_sid > nextpid &&
463 			    pidchecked > p->p_session->s_sid)
464 				pidchecked = p->p_session->s_sid;
465 		}
466 		if (!doingzomb) {
467 			doingzomb = 1;
468 			p = LIST_FIRST(&zombproc);
469 			goto again;
470 		}
471 	}
472 	return(nextpid);
473 }
474 
475 /*
476  * Called from exit1 to remove a process from the allproc
477  * list and move it to the zombie list.
478  *
479  * MPSAFE
480  */
481 void
482 proc_move_allproc_zombie(struct proc *p)
483 {
484 	spin_lock_wr(&allproc_spin);
485 	while (p->p_lock) {
486 		spin_unlock_wr(&allproc_spin);
487 		tsleep(p, 0, "reap1", hz / 10);
488 		spin_lock_wr(&allproc_spin);
489 	}
490 	LIST_REMOVE(p, p_list);
491 	LIST_INSERT_HEAD(&zombproc, p, p_list);
492 	LIST_REMOVE(p, p_hash);
493 	p->p_flag |= P_ZOMBIE;
494 	spin_unlock_wr(&allproc_spin);
495 }
496 
497 /*
498  * This routine is called from kern_wait() and will remove the process
499  * from the zombie list and the sibling list.  This routine will block
500  * if someone has a lock on the proces (p_lock).
501  *
502  * MPSAFE
503  */
504 void
505 proc_remove_zombie(struct proc *p)
506 {
507 	spin_lock_wr(&allproc_spin);
508 	while (p->p_lock) {
509 		spin_unlock_wr(&allproc_spin);
510 		tsleep(p, 0, "reap1", hz / 10);
511 		spin_lock_wr(&allproc_spin);
512 	}
513 	LIST_REMOVE(p, p_list); /* off zombproc */
514 	LIST_REMOVE(p, p_sibling);
515 	spin_unlock_wr(&allproc_spin);
516 }
517 
518 /*
519  * Scan all processes on the allproc list.  The process is automatically
520  * held for the callback.  A return value of -1 terminates the loop.
521  *
522  * MPSAFE
523  */
524 void
525 allproc_scan(int (*callback)(struct proc *, void *), void *data)
526 {
527 	struct proc *p;
528 	int r;
529 
530 	spin_lock_rd(&allproc_spin);
531 	LIST_FOREACH(p, &allproc, p_list) {
532 		PHOLD(p);
533 		spin_unlock_rd(&allproc_spin);
534 		r = callback(p, data);
535 		spin_lock_rd(&allproc_spin);
536 		PRELE(p);
537 		if (r < 0)
538 			break;
539 	}
540 	spin_unlock_rd(&allproc_spin);
541 }
542 
543 /*
544  * Scan all processes on the zombproc list.  The process is automatically
545  * held for the callback.  A return value of -1 terminates the loop.
546  *
547  * MPSAFE
548  */
549 void
550 zombproc_scan(int (*callback)(struct proc *, void *), void *data)
551 {
552 	struct proc *p;
553 	int r;
554 
555 	spin_lock_rd(&allproc_spin);
556 	LIST_FOREACH(p, &zombproc, p_list) {
557 		PHOLD(p);
558 		spin_unlock_rd(&allproc_spin);
559 		r = callback(p, data);
560 		spin_lock_rd(&allproc_spin);
561 		PRELE(p);
562 		if (r < 0)
563 			break;
564 	}
565 	spin_unlock_rd(&allproc_spin);
566 }
567 
568 #include "opt_ddb.h"
569 #ifdef DDB
570 #include <ddb/ddb.h>
571 
572 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
573 {
574 	struct pgrp *pgrp;
575 	struct proc *p;
576 	int i;
577 
578 	for (i = 0; i <= pgrphash; i++) {
579 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
580 			printf("\tindx %d\n", i);
581 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
582 				printf(
583 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
584 				    (void *)pgrp, (long)pgrp->pg_id,
585 				    (void *)pgrp->pg_session,
586 				    pgrp->pg_session->s_count,
587 				    (void *)LIST_FIRST(&pgrp->pg_members));
588 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
589 					printf("\t\tpid %ld addr %p pgrp %p\n",
590 					    (long)p->p_pid, (void *)p,
591 					    (void *)p->p_pgrp);
592 				}
593 			}
594 		}
595 	}
596 }
597 #endif /* DDB */
598 
599 /*
600  * Fill in an eproc structure for the specified thread.
601  */
602 void
603 fill_eproc_td(thread_t td, struct eproc *ep, struct proc *xp)
604 {
605 	bzero(ep, sizeof(*ep));
606 
607 	ep->e_uticks = td->td_uticks;
608 	ep->e_sticks = td->td_sticks;
609 	ep->e_iticks = td->td_iticks;
610 	ep->e_tdev = NOUDEV;
611 	ep->e_cpuid = td->td_gd->gd_cpuid;
612 	if (td->td_wmesg) {
613 		strncpy(ep->e_wmesg, td->td_wmesg, WMESGLEN);
614 		ep->e_wmesg[WMESGLEN] = 0;
615 	}
616 
617 	/*
618 	 * Fake up portions of the proc structure copied out by the sysctl
619 	 * to return useful information.  Note that using td_pri directly
620 	 * is messy because it includes critial section data so we fake
621 	 * up an rtprio.prio for threads.
622 	 */
623 	if (xp) {
624 		*xp = *initproc;
625 		xp->p_rtprio.type = RTP_PRIO_THREAD;
626 		xp->p_rtprio.prio = td->td_pri & TDPRI_MASK;
627 		xp->p_pid = -1;
628 	}
629 }
630 
631 /*
632  * Fill in an eproc structure for the specified process.
633  */
634 void
635 fill_eproc(struct proc *p, struct eproc *ep)
636 {
637 	struct tty *tp;
638 
639 	fill_eproc_td(p->p_thread, ep, NULL);
640 
641 	ep->e_paddr = p;
642 	if (p->p_ucred) {
643 		ep->e_ucred = *p->p_ucred;
644 	}
645 	if (p->p_procsig) {
646 		ep->e_procsig = *p->p_procsig;
647 	}
648 	if (p->p_stat != SIDL && (p->p_flag & P_ZOMBIE) == 0 &&
649 	    p->p_vmspace != NULL) {
650 		struct vmspace *vm = p->p_vmspace;
651 		ep->e_vm = *vm;
652 		ep->e_vm.vm_rssize = vmspace_resident_count(vm); /*XXX*/
653 	}
654 	if ((p->p_flag & P_SWAPPEDOUT) == 0 && p->p_stats)
655 		ep->e_stats = *p->p_stats;
656 	if (p->p_pptr)
657 		ep->e_ppid = p->p_pptr->p_pid;
658 	if (p->p_pgrp) {
659 		ep->e_pgid = p->p_pgrp->pg_id;
660 		ep->e_jobc = p->p_pgrp->pg_jobc;
661 		ep->e_sess = p->p_pgrp->pg_session;
662 
663 		if (ep->e_sess) {
664 			bcopy(ep->e_sess->s_login, ep->e_login, sizeof(ep->e_login));
665 			if (ep->e_sess->s_ttyvp)
666 				ep->e_flag = EPROC_CTTY;
667 			if (p->p_session && SESS_LEADER(p))
668 				ep->e_flag |= EPROC_SLEADER;
669 		}
670 	}
671 	if ((p->p_flag & P_CONTROLT) &&
672 	    (ep->e_sess != NULL) &&
673 	    ((tp = ep->e_sess->s_ttyp) != NULL)) {
674 		ep->e_tdev = dev2udev(tp->t_dev);
675 		ep->e_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
676 		ep->e_tsess = tp->t_session;
677 	} else {
678 		ep->e_tdev = NOUDEV;
679 	}
680 	if (p->p_ucred->cr_prison)
681 		ep->e_jailid = p->p_ucred->cr_prison->pr_id;
682 }
683 
684 /*
685  * Locate a process on the zombie list.  Return a held process or NULL.
686  */
687 struct proc *
688 zpfind(pid_t pid)
689 {
690 	struct proc *p;
691 
692 	LIST_FOREACH(p, &zombproc, p_list)
693 		if (p->p_pid == pid)
694 			return (p);
695 	return (NULL);
696 }
697 
698 static int
699 sysctl_out_proc(struct proc *p, struct thread *td, struct sysctl_req *req, int doingzomb)
700 {
701 	struct eproc eproc;
702 	struct proc xproc;
703 	int error;
704 #if 0
705 	pid_t pid = p->p_pid;
706 #endif
707 
708 	if (p) {
709 		td = p->p_thread;
710 		fill_eproc(p, &eproc);
711 		xproc = *p;
712 
713 		/*
714 		 * p_stat fixup.  If we are in a thread sleep mark p_stat
715 		 * as sleeping if the thread is blocked.
716 		 */
717 		if (p->p_stat == SRUN && td && (td->td_flags & TDF_BLOCKED)) {
718 			xproc.p_stat = SSLEEP;
719 		}
720 		/*
721 		 * If the process is being stopped but is in a normal tsleep,
722 		 * mark it as being SSTOP.
723 		 */
724 		if (p->p_stat == SSLEEP && (p->p_flag & P_STOPPED))
725 			xproc.p_stat = SSTOP;
726 		if (p->p_flag & P_ZOMBIE)
727 			xproc.p_stat = SZOMB;
728 	} else if (td) {
729 		fill_eproc_td(td, &eproc, &xproc);
730 	}
731 	error = SYSCTL_OUT(req,(caddr_t)&xproc, sizeof(struct proc));
732 	if (error)
733 		return (error);
734 	error = SYSCTL_OUT(req,(caddr_t)&eproc, sizeof(eproc));
735 	if (error)
736 		return (error);
737 	error = SYSCTL_OUT(req,(caddr_t)td, sizeof(struct thread));
738 	if (error)
739 		return (error);
740 #if 0
741 	if (!doingzomb && pid && (pfind(pid) != p))
742 		return EAGAIN;
743 	if (doingzomb && zpfind(pid) != p)
744 		return EAGAIN;
745 #endif
746 	return (0);
747 }
748 
749 static int
750 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
751 {
752 	int *name = (int*) arg1;
753 	u_int namelen = arg2;
754 	struct proc *p;
755 	struct thread *td;
756 	int doingzomb;
757 	int error = 0;
758 	int n;
759 	int origcpu;
760 	struct ucred *cr1 = curproc->p_ucred;
761 
762 	if (oidp->oid_number == KERN_PROC_PID) {
763 		if (namelen != 1)
764 			return (EINVAL);
765 		p = pfind((pid_t)name[0]);
766 		if (!p)
767 			return (0);
768 		if (!PRISON_CHECK(cr1, p->p_ucred))
769 			return (0);
770 		error = sysctl_out_proc(p, NULL, req, 0);
771 		return (error);
772 	}
773 	if (oidp->oid_number == KERN_PROC_ALL && !namelen)
774 		;
775 	else if (oidp->oid_number != KERN_PROC_ALL && namelen == 1)
776 		;
777 	else
778 		return (EINVAL);
779 
780 	if (!req->oldptr) {
781 		/* overestimate by 5 procs */
782 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
783 		if (error)
784 			return (error);
785 	}
786 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
787 		if (!doingzomb)
788 			p = LIST_FIRST(&allproc);
789 		else
790 			p = LIST_FIRST(&zombproc);
791 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
792 			/*
793 			 * Show a user only their processes.
794 			 */
795 			if ((!ps_showallprocs) && p_trespass(cr1, p->p_ucred))
796 				continue;
797 			/*
798 			 * Skip embryonic processes.
799 			 */
800 			if (p->p_stat == SIDL)
801 				continue;
802 			/*
803 			 * TODO - make more efficient (see notes below).
804 			 * do by session.
805 			 */
806 			switch (oidp->oid_number) {
807 			case KERN_PROC_PGRP:
808 				/* could do this by traversing pgrp */
809 				if (p->p_pgrp == NULL ||
810 				    p->p_pgrp->pg_id != (pid_t)name[0])
811 					continue;
812 				break;
813 
814 			case KERN_PROC_TTY:
815 				if ((p->p_flag & P_CONTROLT) == 0 ||
816 				    p->p_session == NULL ||
817 				    p->p_session->s_ttyp == NULL ||
818 				    dev2udev(p->p_session->s_ttyp->t_dev) !=
819 					(udev_t)name[0])
820 					continue;
821 				break;
822 
823 			case KERN_PROC_UID:
824 				if (p->p_ucred == NULL ||
825 				    p->p_ucred->cr_uid != (uid_t)name[0])
826 					continue;
827 				break;
828 
829 			case KERN_PROC_RUID:
830 				if (p->p_ucred == NULL ||
831 				    p->p_ucred->cr_ruid != (uid_t)name[0])
832 					continue;
833 				break;
834 			}
835 
836 			if (!PRISON_CHECK(cr1, p->p_ucred))
837 				continue;
838 			PHOLD(p);
839 			error = sysctl_out_proc(p, NULL, req, doingzomb);
840 			PRELE(p);
841 			if (error)
842 				return (error);
843 		}
844 	}
845 
846 	/*
847 	 * Iterate over all active cpus and scan their thread list.  Start
848 	 * with the next logical cpu and end with our original cpu.  We
849 	 * migrate our own thread to each target cpu in order to safely scan
850 	 * its thread list.  In the last loop we migrate back to our original
851 	 * cpu.
852 	 */
853 	origcpu = mycpu->gd_cpuid;
854 	if (!ps_showallthreads || jailed(cr1))
855 		goto post_threads;
856 	for (n = 1; n <= ncpus; ++n) {
857 		globaldata_t rgd;
858 		int nid;
859 
860 		nid = (origcpu + n) % ncpus;
861 		if ((smp_active_mask & (1 << nid)) == 0)
862 			continue;
863 		rgd = globaldata_find(nid);
864 		lwkt_setcpu_self(rgd);
865 
866 		TAILQ_FOREACH(td, &mycpu->gd_tdallq, td_allq) {
867 			if (td->td_proc)
868 				continue;
869 			switch (oidp->oid_number) {
870 			case KERN_PROC_PGRP:
871 			case KERN_PROC_TTY:
872 			case KERN_PROC_UID:
873 			case KERN_PROC_RUID:
874 				continue;
875 			default:
876 				break;
877 			}
878 			lwkt_hold(td);
879 			error = sysctl_out_proc(NULL, td, req, doingzomb);
880 			lwkt_rele(td);
881 			if (error)
882 				return (error);
883 		}
884 	}
885 post_threads:
886 	return (0);
887 }
888 
889 /*
890  * This sysctl allows a process to retrieve the argument list or process
891  * title for another process without groping around in the address space
892  * of the other process.  It also allow a process to set its own "process
893  * title to a string of its own choice.
894  */
895 static int
896 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
897 {
898 	int *name = (int*) arg1;
899 	u_int namelen = arg2;
900 	struct proc *p;
901 	struct pargs *pa;
902 	int error = 0;
903 	struct ucred *cr1 = curproc->p_ucred;
904 
905 	if (namelen != 1)
906 		return (EINVAL);
907 
908 	p = pfind((pid_t)name[0]);
909 	if (!p)
910 		return (0);
911 
912 	if ((!ps_argsopen) && p_trespass(cr1, p->p_ucred))
913 		return (0);
914 
915 	if (req->newptr && curproc != p)
916 		return (EPERM);
917 
918 	if (req->oldptr && p->p_args != NULL)
919 		error = SYSCTL_OUT(req, p->p_args->ar_args, p->p_args->ar_length);
920 	if (req->newptr == NULL)
921 		return (error);
922 
923 	if (p->p_args && --p->p_args->ar_ref == 0)
924 		FREE(p->p_args, M_PARGS);
925 	p->p_args = NULL;
926 
927 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
928 		return (error);
929 
930 	MALLOC(pa, struct pargs *, sizeof(struct pargs) + req->newlen,
931 	    M_PARGS, M_WAITOK);
932 	pa->ar_ref = 1;
933 	pa->ar_length = req->newlen;
934 	error = SYSCTL_IN(req, pa->ar_args, req->newlen);
935 	if (!error)
936 		p->p_args = pa;
937 	else
938 		FREE(pa, M_PARGS);
939 	return (error);
940 }
941 
942 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
943 
944 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT,
945 	0, 0, sysctl_kern_proc, "S,proc", "Return entire process table");
946 
947 SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD,
948 	sysctl_kern_proc, "Process table");
949 
950 SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD,
951 	sysctl_kern_proc, "Process table");
952 
953 SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD,
954 	sysctl_kern_proc, "Process table");
955 
956 SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD,
957 	sysctl_kern_proc, "Process table");
958 
959 SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD,
960 	sysctl_kern_proc, "Process table");
961 
962 SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, CTLFLAG_RW | CTLFLAG_ANYBODY,
963 	sysctl_kern_proc_args, "Process argument list");
964