xref: /dflybsd-src/sys/kern/kern_proc.c (revision f0a5c1029a4d9e5497fe457e38db8db7a021dc86)
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 1.43 2008/05/18 20:02:02 nth 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 MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
62 MALLOC_DEFINE(M_LWP, "lwp", "lwp structures");
63 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
64 
65 int ps_showallprocs = 1;
66 static int ps_showallthreads = 1;
67 SYSCTL_INT(_security, OID_AUTO, ps_showallprocs, CTLFLAG_RW,
68     &ps_showallprocs, 0,
69     "Unprivileged processes can see proccesses with different UID/GID");
70 SYSCTL_INT(_security, OID_AUTO, ps_showallthreads, CTLFLAG_RW,
71     &ps_showallthreads, 0,
72     "Unprivileged processes can see kernel threads");
73 
74 static void pgdelete(struct pgrp *);
75 static void orphanpg(struct pgrp *pg);
76 static pid_t proc_getnewpid_locked(int random_offset);
77 
78 /*
79  * Other process lists
80  */
81 struct pidhashhead *pidhashtbl;
82 u_long pidhash;
83 struct pgrphashhead *pgrphashtbl;
84 u_long pgrphash;
85 struct proclist allproc;
86 struct proclist zombproc;
87 struct spinlock allproc_spin;
88 vm_zone_t thread_zone;
89 
90 /*
91  * Random component to nextpid generation.  We mix in a random factor to make
92  * it a little harder to predict.  We sanity check the modulus value to avoid
93  * doing it in critical paths.  Don't let it be too small or we pointlessly
94  * waste randomness entropy, and don't let it be impossibly large.  Using a
95  * modulus that is too big causes a LOT more process table scans and slows
96  * down fork processing as the pidchecked caching is defeated.
97  */
98 static int randompid = 0;
99 
100 static int
101 sysctl_kern_randompid(SYSCTL_HANDLER_ARGS)
102 {
103 	int error, pid;
104 
105 	pid = randompid;
106 	error = sysctl_handle_int(oidp, &pid, 0, req);
107 	if (error || !req->newptr)
108 		return (error);
109 	if (pid < 0 || pid > PID_MAX - 100)     /* out of range */
110 		pid = PID_MAX - 100;
111 	else if (pid < 2)                       /* NOP */
112 		pid = 0;
113 	else if (pid < 100)                     /* Make it reasonable */
114 		pid = 100;
115 	randompid = pid;
116 	return (error);
117 }
118 
119 SYSCTL_PROC(_kern, OID_AUTO, randompid, CTLTYPE_INT|CTLFLAG_RW,
120 	    0, 0, sysctl_kern_randompid, "I", "Random PID modulus");
121 
122 /*
123  * Initialize global process hashing structures.
124  */
125 void
126 procinit(void)
127 {
128 	LIST_INIT(&allproc);
129 	LIST_INIT(&zombproc);
130 	spin_init(&allproc_spin);
131 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
132 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
133 	thread_zone = zinit("THREAD", sizeof (struct thread), 0, 0, 5);
134 	uihashinit();
135 }
136 
137 /*
138  * Is p an inferior of the current process?
139  */
140 int
141 inferior(struct proc *p)
142 {
143 	for (; p != curproc; p = p->p_pptr)
144 		if (p->p_pid == 0)
145 			return (0);
146 	return (1);
147 }
148 
149 /*
150  * Locate a process by number
151  */
152 struct proc *
153 pfind(pid_t pid)
154 {
155 	struct proc *p;
156 
157 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
158 		if (p->p_pid == pid)
159 			return (p);
160 	}
161 	return (NULL);
162 }
163 
164 /*
165  * Locate a process group by number
166  */
167 struct pgrp *
168 pgfind(pid_t pgid)
169 {
170 	struct pgrp *pgrp;
171 
172 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
173 		if (pgrp->pg_id == pgid)
174 			return (pgrp);
175 	}
176 	return (NULL);
177 }
178 
179 /*
180  * Move p to a new or existing process group (and session)
181  */
182 int
183 enterpgrp(struct proc *p, pid_t pgid, int mksess)
184 {
185 	struct pgrp *pgrp = pgfind(pgid);
186 
187 	KASSERT(pgrp == NULL || !mksess,
188 	    ("enterpgrp: setsid into non-empty pgrp"));
189 	KASSERT(!SESS_LEADER(p),
190 	    ("enterpgrp: session leader attempted setpgrp"));
191 
192 	if (pgrp == NULL) {
193 		pid_t savepid = p->p_pid;
194 		struct proc *np;
195 		/*
196 		 * new process group
197 		 */
198 		KASSERT(p->p_pid == pgid,
199 		    ("enterpgrp: new pgrp and pid != pgid"));
200 		if ((np = pfind(savepid)) == NULL || np != p)
201 			return (ESRCH);
202 		MALLOC(pgrp, struct pgrp *, sizeof(struct pgrp), M_PGRP,
203 		    M_WAITOK);
204 		if (mksess) {
205 			struct session *sess;
206 
207 			/*
208 			 * new session
209 			 */
210 			MALLOC(sess, struct session *, sizeof(struct session),
211 			    M_SESSION, M_WAITOK);
212 			sess->s_leader = p;
213 			sess->s_sid = p->p_pid;
214 			sess->s_count = 1;
215 			sess->s_ttyvp = NULL;
216 			sess->s_ttyp = NULL;
217 			bcopy(p->p_session->s_login, sess->s_login,
218 			    sizeof(sess->s_login));
219 			p->p_flag &= ~P_CONTROLT;
220 			pgrp->pg_session = sess;
221 			KASSERT(p == curproc,
222 			    ("enterpgrp: mksession and p != curproc"));
223 		} else {
224 			pgrp->pg_session = p->p_session;
225 			sess_hold(pgrp->pg_session);
226 		}
227 		pgrp->pg_id = pgid;
228 		LIST_INIT(&pgrp->pg_members);
229 		LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
230 		pgrp->pg_jobc = 0;
231 		SLIST_INIT(&pgrp->pg_sigiolst);
232 		lockinit(&pgrp->pg_lock, "pgwt", 0, 0);
233 	} else if (pgrp == p->p_pgrp)
234 		return (0);
235 
236 	/*
237 	 * Adjust eligibility of affected pgrps to participate in job control.
238 	 * Increment eligibility counts before decrementing, otherwise we
239 	 * could reach 0 spuriously during the first call.
240 	 */
241 	fixjobc(p, pgrp, 1);
242 	fixjobc(p, p->p_pgrp, 0);
243 
244 	LIST_REMOVE(p, p_pglist);
245 	if (LIST_EMPTY(&p->p_pgrp->pg_members))
246 		pgdelete(p->p_pgrp);
247 	p->p_pgrp = pgrp;
248 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
249 	return (0);
250 }
251 
252 /*
253  * remove process from process group
254  */
255 int
256 leavepgrp(struct proc *p)
257 {
258 
259 	LIST_REMOVE(p, p_pglist);
260 	if (LIST_EMPTY(&p->p_pgrp->pg_members))
261 		pgdelete(p->p_pgrp);
262 	p->p_pgrp = 0;
263 	return (0);
264 }
265 
266 /*
267  * delete a process group
268  */
269 static void
270 pgdelete(struct pgrp *pgrp)
271 {
272 
273 	/*
274 	 * Reset any sigio structures pointing to us as a result of
275 	 * F_SETOWN with our pgid.
276 	 */
277 	funsetownlst(&pgrp->pg_sigiolst);
278 
279 	if (pgrp->pg_session->s_ttyp != NULL &&
280 	    pgrp->pg_session->s_ttyp->t_pgrp == pgrp)
281 		pgrp->pg_session->s_ttyp->t_pgrp = NULL;
282 	LIST_REMOVE(pgrp, pg_hash);
283 	sess_rele(pgrp->pg_session);
284 	kfree(pgrp, M_PGRP);
285 }
286 
287 /*
288  * Adjust the ref count on a session structure.  When the ref count falls to
289  * zero the tty is disassociated from the session and the session structure
290  * is freed.  Note that tty assocation is not itself ref-counted.
291  */
292 void
293 sess_hold(struct session *sp)
294 {
295 	++sp->s_count;
296 }
297 
298 void
299 sess_rele(struct session *sp)
300 {
301 	KKASSERT(sp->s_count > 0);
302 	if (--sp->s_count == 0) {
303 		if (sp->s_ttyp && sp->s_ttyp->t_session) {
304 #ifdef TTY_DO_FULL_CLOSE
305 			/* FULL CLOSE, see ttyclearsession() */
306 			KKASSERT(sp->s_ttyp->t_session == sp);
307 			sp->s_ttyp->t_session = NULL;
308 #else
309 			/* HALF CLOSE, see ttyclearsession() */
310 			if (sp->s_ttyp->t_session == sp)
311 				sp->s_ttyp->t_session = NULL;
312 #endif
313 		}
314 		kfree(sp, M_SESSION);
315 	}
316 }
317 
318 /*
319  * Adjust pgrp jobc counters when specified process changes process group.
320  * We count the number of processes in each process group that "qualify"
321  * the group for terminal job control (those with a parent in a different
322  * process group of the same session).  If that count reaches zero, the
323  * process group becomes orphaned.  Check both the specified process'
324  * process group and that of its children.
325  * entering == 0 => p is leaving specified group.
326  * entering == 1 => p is entering specified group.
327  */
328 void
329 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
330 {
331 	struct pgrp *hispgrp;
332 	struct session *mysession = pgrp->pg_session;
333 
334 	/*
335 	 * Check p's parent to see whether p qualifies its own process
336 	 * group; if so, adjust count for p's process group.
337 	 */
338 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
339 	    hispgrp->pg_session == mysession) {
340 		if (entering)
341 			pgrp->pg_jobc++;
342 		else if (--pgrp->pg_jobc == 0)
343 			orphanpg(pgrp);
344 	}
345 
346 	/*
347 	 * Check this process' children to see whether they qualify
348 	 * their process groups; if so, adjust counts for children's
349 	 * process groups.
350 	 */
351 	LIST_FOREACH(p, &p->p_children, p_sibling)
352 		if ((hispgrp = p->p_pgrp) != pgrp &&
353 		    hispgrp->pg_session == mysession &&
354 		    p->p_stat != SZOMB) {
355 			if (entering)
356 				hispgrp->pg_jobc++;
357 			else if (--hispgrp->pg_jobc == 0)
358 				orphanpg(hispgrp);
359 		}
360 }
361 
362 /*
363  * A process group has become orphaned;
364  * if there are any stopped processes in the group,
365  * hang-up all process in that group.
366  */
367 static void
368 orphanpg(struct pgrp *pg)
369 {
370 	struct proc *p;
371 
372 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
373 		if (p->p_stat == SSTOP) {
374 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
375 				ksignal(p, SIGHUP);
376 				ksignal(p, SIGCONT);
377 			}
378 			return;
379 		}
380 	}
381 }
382 
383 /*
384  * Add a new process to the allproc list and the PID hash.  This
385  * also assigns a pid to the new process.
386  *
387  * MPALMOSTSAFE - acquires mplock for karc4random() call
388  */
389 void
390 proc_add_allproc(struct proc *p)
391 {
392 	int random_offset;
393 
394 	if ((random_offset = randompid) != 0) {
395 		get_mplock();
396 		random_offset = karc4random() % random_offset;
397 		rel_mplock();
398 	}
399 
400 	spin_lock_wr(&allproc_spin);
401 	p->p_pid = proc_getnewpid_locked(random_offset);
402 	LIST_INSERT_HEAD(&allproc, p, p_list);
403 	LIST_INSERT_HEAD(PIDHASH(p->p_pid), p, p_hash);
404 	spin_unlock_wr(&allproc_spin);
405 }
406 
407 /*
408  * Calculate a new process pid.  This function is integrated into
409  * proc_add_allproc() to guarentee that the new pid is not reused before
410  * the new process can be added to the allproc list.
411  *
412  * MPSAFE - must be called with allproc_spin held.
413  */
414 static
415 pid_t
416 proc_getnewpid_locked(int random_offset)
417 {
418 	static pid_t nextpid;
419 	static pid_t pidchecked;
420 	struct proc *p;
421 
422 	/*
423 	 * Find an unused process ID.  We remember a range of unused IDs
424 	 * ready to use (from nextpid+1 through pidchecked-1).
425 	 */
426 	nextpid = nextpid + 1 + random_offset;
427 retry:
428 	/*
429 	 * If the process ID prototype has wrapped around,
430 	 * restart somewhat above 0, as the low-numbered procs
431 	 * tend to include daemons that don't exit.
432 	 */
433 	if (nextpid >= PID_MAX) {
434 		nextpid = nextpid % PID_MAX;
435 		if (nextpid < 100)
436 			nextpid += 100;
437 		pidchecked = 0;
438 	}
439 	if (nextpid >= pidchecked) {
440 		int doingzomb = 0;
441 
442 		pidchecked = PID_MAX;
443 		/*
444 		 * Scan the active and zombie procs to check whether this pid
445 		 * is in use.  Remember the lowest pid that's greater
446 		 * than nextpid, so we can avoid checking for a while.
447 		 */
448 		p = LIST_FIRST(&allproc);
449 again:
450 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
451 			while (p->p_pid == nextpid ||
452 			    p->p_pgrp->pg_id == nextpid ||
453 			    p->p_session->s_sid == nextpid) {
454 				nextpid++;
455 				if (nextpid >= pidchecked)
456 					goto retry;
457 			}
458 			if (p->p_pid > nextpid && pidchecked > p->p_pid)
459 				pidchecked = p->p_pid;
460 			if (p->p_pgrp->pg_id > nextpid &&
461 			    pidchecked > p->p_pgrp->pg_id)
462 				pidchecked = p->p_pgrp->pg_id;
463 			if (p->p_session->s_sid > nextpid &&
464 			    pidchecked > p->p_session->s_sid)
465 				pidchecked = p->p_session->s_sid;
466 		}
467 		if (!doingzomb) {
468 			doingzomb = 1;
469 			p = LIST_FIRST(&zombproc);
470 			goto again;
471 		}
472 	}
473 	return(nextpid);
474 }
475 
476 /*
477  * Called from exit1 to remove a process from the allproc
478  * list and move it to the zombie list.
479  *
480  * MPSAFE
481  */
482 void
483 proc_move_allproc_zombie(struct proc *p)
484 {
485 	spin_lock_wr(&allproc_spin);
486 	while (p->p_lock) {
487 		spin_unlock_wr(&allproc_spin);
488 		tsleep(p, 0, "reap1", hz / 10);
489 		spin_lock_wr(&allproc_spin);
490 	}
491 	LIST_REMOVE(p, p_list);
492 	LIST_INSERT_HEAD(&zombproc, p, p_list);
493 	LIST_REMOVE(p, p_hash);
494 	p->p_stat = SZOMB;
495 	spin_unlock_wr(&allproc_spin);
496 }
497 
498 /*
499  * This routine is called from kern_wait() and will remove the process
500  * from the zombie list and the sibling list.  This routine will block
501  * if someone has a lock on the proces (p_lock).
502  *
503  * MPSAFE
504  */
505 void
506 proc_remove_zombie(struct proc *p)
507 {
508 	spin_lock_wr(&allproc_spin);
509 	while (p->p_lock) {
510 		spin_unlock_wr(&allproc_spin);
511 		tsleep(p, 0, "reap1", hz / 10);
512 		spin_lock_wr(&allproc_spin);
513 	}
514 	LIST_REMOVE(p, p_list); /* off zombproc */
515 	LIST_REMOVE(p, p_sibling);
516 	spin_unlock_wr(&allproc_spin);
517 }
518 
519 /*
520  * Scan all processes on the allproc list.  The process is automatically
521  * held for the callback.  A return value of -1 terminates the loop.
522  *
523  * MPSAFE
524  */
525 void
526 allproc_scan(int (*callback)(struct proc *, void *), void *data)
527 {
528 	struct proc *p;
529 	int r;
530 
531 	spin_lock_rd(&allproc_spin);
532 	LIST_FOREACH(p, &allproc, p_list) {
533 		PHOLD(p);
534 		spin_unlock_rd(&allproc_spin);
535 		r = callback(p, data);
536 		spin_lock_rd(&allproc_spin);
537 		PRELE(p);
538 		if (r < 0)
539 			break;
540 	}
541 	spin_unlock_rd(&allproc_spin);
542 }
543 
544 /*
545  * Scan all lwps of processes on the allproc list.  The lwp is automatically
546  * held for the callback.  A return value of -1 terminates the loop.
547  *
548  * possibly not MPSAFE, needs to access foreingn proc structures
549  */
550 void
551 alllwp_scan(int (*callback)(struct lwp *, void *), void *data)
552 {
553 	struct proc *p;
554 	struct lwp *lp;
555 	int r = 0;
556 
557 	spin_lock_rd(&allproc_spin);
558 	LIST_FOREACH(p, &allproc, p_list) {
559 		PHOLD(p);
560 		spin_unlock_rd(&allproc_spin);
561 		FOREACH_LWP_IN_PROC(lp, p) {
562 			LWPHOLD(lp);
563 			r = callback(lp, data);
564 			LWPRELE(lp);
565 		}
566 		spin_lock_rd(&allproc_spin);
567 		PRELE(p);
568 		if (r < 0)
569 			break;
570 	}
571 	spin_unlock_rd(&allproc_spin);
572 }
573 
574 /*
575  * Scan all processes on the zombproc list.  The process is automatically
576  * held for the callback.  A return value of -1 terminates the loop.
577  *
578  * MPSAFE
579  */
580 void
581 zombproc_scan(int (*callback)(struct proc *, void *), void *data)
582 {
583 	struct proc *p;
584 	int r;
585 
586 	spin_lock_rd(&allproc_spin);
587 	LIST_FOREACH(p, &zombproc, p_list) {
588 		PHOLD(p);
589 		spin_unlock_rd(&allproc_spin);
590 		r = callback(p, data);
591 		spin_lock_rd(&allproc_spin);
592 		PRELE(p);
593 		if (r < 0)
594 			break;
595 	}
596 	spin_unlock_rd(&allproc_spin);
597 }
598 
599 #include "opt_ddb.h"
600 #ifdef DDB
601 #include <ddb/ddb.h>
602 
603 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
604 {
605 	struct pgrp *pgrp;
606 	struct proc *p;
607 	int i;
608 
609 	for (i = 0; i <= pgrphash; i++) {
610 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
611 			kprintf("\tindx %d\n", i);
612 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
613 				kprintf(
614 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
615 				    (void *)pgrp, (long)pgrp->pg_id,
616 				    (void *)pgrp->pg_session,
617 				    pgrp->pg_session->s_count,
618 				    (void *)LIST_FIRST(&pgrp->pg_members));
619 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
620 					kprintf("\t\tpid %ld addr %p pgrp %p\n",
621 					    (long)p->p_pid, (void *)p,
622 					    (void *)p->p_pgrp);
623 				}
624 			}
625 		}
626 	}
627 }
628 #endif /* DDB */
629 
630 /*
631  * Locate a process on the zombie list.  Return a held process or NULL.
632  */
633 struct proc *
634 zpfind(pid_t pid)
635 {
636 	struct proc *p;
637 
638 	LIST_FOREACH(p, &zombproc, p_list)
639 		if (p->p_pid == pid)
640 			return (p);
641 	return (NULL);
642 }
643 
644 static int
645 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
646 {
647 	struct kinfo_proc ki;
648 	struct lwp *lp;
649 	int skp = 0, had_output = 0;
650 	int error;
651 
652 	fill_kinfo_proc(p, &ki);
653 	if ((flags & KERN_PROC_FLAG_LWP) == 0)
654 		skp = 1;
655 	FOREACH_LWP_IN_PROC(lp, p) {
656 		fill_kinfo_lwp(lp, &ki.kp_lwp);
657 output:
658 		had_output = 1;
659 		error = SYSCTL_OUT(req, &ki, sizeof(ki));
660 		if (error)
661 			return error;
662 		if (skp)
663 			break;
664 	}
665 	/* We need to output at least the proc, even if there is no lwp. */
666 	if (!had_output)
667 		goto output;
668 #if 0
669 	if (!doingzomb && pid && (pfind(pid) != p))
670 		return EAGAIN;
671 	if (doingzomb && zpfind(pid) != p)
672 		return EAGAIN;
673 #endif
674 	return (0);
675 }
676 
677 static int
678 sysctl_out_proc_kthread(struct thread *td, struct sysctl_req *req, int flags)
679 {
680 	struct kinfo_proc ki;
681 	int error;
682 
683 	fill_kinfo_proc_kthread(td, &ki);
684 	error = SYSCTL_OUT(req, &ki, sizeof(ki));
685 	if (error)
686 		return error;
687 	return(0);
688 }
689 
690 static int
691 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
692 {
693 	int *name = (int*) arg1;
694 	int oid = oidp->oid_number;
695 	u_int namelen = arg2;
696 	struct proc *p, *np;
697 	struct proclist *plist;
698 	struct thread *td;
699 	int doingzomb, flags = 0;
700 	int error = 0;
701 	int n;
702 	int origcpu;
703 	struct ucred *cr1 = curproc->p_ucred;
704 
705 	flags = oid & KERN_PROC_FLAGMASK;
706 	oid &= ~KERN_PROC_FLAGMASK;
707 
708 	if ((oid == KERN_PROC_ALL && namelen != 0) ||
709 	    (oid != KERN_PROC_ALL && namelen != 1))
710 		return (EINVAL);
711 
712 	if (oid == KERN_PROC_PID) {
713 		p = pfind((pid_t)name[0]);
714 		if (!p)
715 			return (0);
716 		if (!PRISON_CHECK(cr1, p->p_ucred))
717 			return (0);
718 		PHOLD(p);
719 		error = sysctl_out_proc(p, req, flags);
720 		PRELE(p);
721 		return (error);
722 	}
723 
724 	if (!req->oldptr) {
725 		/* overestimate by 5 procs */
726 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
727 		if (error)
728 			return (error);
729 	}
730 	for (doingzomb = 0; doingzomb <= 1; doingzomb++) {
731 		if (doingzomb)
732 			plist = &zombproc;
733 		else
734 			plist = &allproc;
735 		LIST_FOREACH_MUTABLE(p, plist, p_list, np) {
736 			/*
737 			 * Show a user only their processes.
738 			 */
739 			if ((!ps_showallprocs) && p_trespass(cr1, p->p_ucred))
740 				continue;
741 			/*
742 			 * Skip embryonic processes.
743 			 */
744 			if (p->p_stat == SIDL)
745 				continue;
746 			/*
747 			 * TODO - make more efficient (see notes below).
748 			 * do by session.
749 			 */
750 			switch (oid) {
751 			case KERN_PROC_PGRP:
752 				/* could do this by traversing pgrp */
753 				if (p->p_pgrp == NULL ||
754 				    p->p_pgrp->pg_id != (pid_t)name[0])
755 					continue;
756 				break;
757 
758 			case KERN_PROC_TTY:
759 				if ((p->p_flag & P_CONTROLT) == 0 ||
760 				    p->p_session == NULL ||
761 				    p->p_session->s_ttyp == NULL ||
762 				    dev2udev(p->p_session->s_ttyp->t_dev) !=
763 					(udev_t)name[0])
764 					continue;
765 				break;
766 
767 			case KERN_PROC_UID:
768 				if (p->p_ucred == NULL ||
769 				    p->p_ucred->cr_uid != (uid_t)name[0])
770 					continue;
771 				break;
772 
773 			case KERN_PROC_RUID:
774 				if (p->p_ucred == NULL ||
775 				    p->p_ucred->cr_ruid != (uid_t)name[0])
776 					continue;
777 				break;
778 			}
779 
780 			if (!PRISON_CHECK(cr1, p->p_ucred))
781 				continue;
782 			PHOLD(p);
783 			error = sysctl_out_proc(p, req, flags);
784 			PRELE(p);
785 			if (error)
786 				return (error);
787 		}
788 	}
789 
790 	/*
791 	 * Iterate over all active cpus and scan their thread list.  Start
792 	 * with the next logical cpu and end with our original cpu.  We
793 	 * migrate our own thread to each target cpu in order to safely scan
794 	 * its thread list.  In the last loop we migrate back to our original
795 	 * cpu.
796 	 */
797 	origcpu = mycpu->gd_cpuid;
798 	if (!ps_showallthreads || jailed(cr1))
799 		goto post_threads;
800 	for (n = 1; n <= ncpus; ++n) {
801 		globaldata_t rgd;
802 		int nid;
803 
804 		nid = (origcpu + n) % ncpus;
805 		if ((smp_active_mask & (1 << nid)) == 0)
806 			continue;
807 		rgd = globaldata_find(nid);
808 		lwkt_setcpu_self(rgd);
809 
810 		TAILQ_FOREACH(td, &mycpu->gd_tdallq, td_allq) {
811 			if (td->td_proc)
812 				continue;
813 			switch (oid) {
814 			case KERN_PROC_PGRP:
815 			case KERN_PROC_TTY:
816 			case KERN_PROC_UID:
817 			case KERN_PROC_RUID:
818 				continue;
819 			default:
820 				break;
821 			}
822 			lwkt_hold(td);
823 			error = sysctl_out_proc_kthread(td, req, doingzomb);
824 			lwkt_rele(td);
825 			if (error)
826 				return (error);
827 		}
828 	}
829 post_threads:
830 	return (0);
831 }
832 
833 /*
834  * This sysctl allows a process to retrieve the argument list or process
835  * title for another process without groping around in the address space
836  * of the other process.  It also allow a process to set its own "process
837  * title to a string of its own choice.
838  */
839 static int
840 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
841 {
842 	int *name = (int*) arg1;
843 	u_int namelen = arg2;
844 	struct proc *p;
845 	struct pargs *pa;
846 	int error = 0;
847 	struct ucred *cr1 = curproc->p_ucred;
848 
849 	if (namelen != 1)
850 		return (EINVAL);
851 
852 	p = pfind((pid_t)name[0]);
853 	if (!p)
854 		return (0);
855 
856 	if ((!ps_argsopen) && p_trespass(cr1, p->p_ucred))
857 		return (0);
858 
859 	if (req->newptr && curproc != p)
860 		return (EPERM);
861 
862 	if (req->oldptr && p->p_args != NULL)
863 		error = SYSCTL_OUT(req, p->p_args->ar_args, p->p_args->ar_length);
864 	if (req->newptr == NULL)
865 		return (error);
866 
867 	if (p->p_args && --p->p_args->ar_ref == 0)
868 		FREE(p->p_args, M_PARGS);
869 	p->p_args = NULL;
870 
871 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
872 		return (error);
873 
874 	MALLOC(pa, struct pargs *, sizeof(struct pargs) + req->newlen,
875 	    M_PARGS, M_WAITOK);
876 	pa->ar_ref = 1;
877 	pa->ar_length = req->newlen;
878 	error = SYSCTL_IN(req, pa->ar_args, req->newlen);
879 	if (!error)
880 		p->p_args = pa;
881 	else
882 		FREE(pa, M_PARGS);
883 	return (error);
884 }
885 
886 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
887 
888 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT,
889 	0, 0, sysctl_kern_proc, "S,proc", "Return entire process table");
890 
891 SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD,
892 	sysctl_kern_proc, "Process table");
893 
894 SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD,
895 	sysctl_kern_proc, "Process table");
896 
897 SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD,
898 	sysctl_kern_proc, "Process table");
899 
900 SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD,
901 	sysctl_kern_proc, "Process table");
902 
903 SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD,
904 	sysctl_kern_proc, "Process table");
905 
906 SYSCTL_NODE(_kern_proc, (KERN_PROC_ALL | KERN_PROC_FLAG_LWP), all_lwp, CTLFLAG_RD,
907 	sysctl_kern_proc, "Process table");
908 
909 SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_FLAG_LWP), pgrp_lwp, CTLFLAG_RD,
910 	sysctl_kern_proc, "Process table");
911 
912 SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_FLAG_LWP), tty_lwp, CTLFLAG_RD,
913 	sysctl_kern_proc, "Process table");
914 
915 SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_FLAG_LWP), uid_lwp, CTLFLAG_RD,
916 	sysctl_kern_proc, "Process table");
917 
918 SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_FLAG_LWP), ruid_lwp, CTLFLAG_RD,
919 	sysctl_kern_proc, "Process table");
920 
921 SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_FLAG_LWP), pid_lwp, CTLFLAG_RD,
922 	sysctl_kern_proc, "Process table");
923 
924 SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, CTLFLAG_RW | CTLFLAG_ANYBODY,
925 	sysctl_kern_proc_args, "Process argument list");
926