xref: /dragonfly/sys/kern/kern_proc.c (revision 10cbe914)
1 /*
2  * (MPSAFE)
3  *
4  * Copyright (c) 1982, 1986, 1989, 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. All advertising materials mentioning features or use of this software
16  *    must display the following acknowledgement:
17  *	This product includes software developed by the University of
18  *	California, Berkeley and its contributors.
19  * 4. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	@(#)kern_proc.c	8.7 (Berkeley) 2/14/95
36  * $FreeBSD: src/sys/kern/kern_proc.c,v 1.63.2.9 2003/05/08 07:47:16 kbyanc Exp $
37  * $DragonFly: src/sys/kern/kern_proc.c,v 1.45 2008/06/12 23:25:02 dillon Exp $
38  */
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/kernel.h>
43 #include <sys/sysctl.h>
44 #include <sys/malloc.h>
45 #include <sys/proc.h>
46 #include <sys/jail.h>
47 #include <sys/filedesc.h>
48 #include <sys/tty.h>
49 #include <sys/dsched.h>
50 #include <sys/signalvar.h>
51 #include <sys/spinlock.h>
52 #include <vm/vm.h>
53 #include <sys/lock.h>
54 #include <vm/pmap.h>
55 #include <vm/vm_map.h>
56 #include <sys/user.h>
57 #include <machine/smp.h>
58 
59 #include <sys/spinlock2.h>
60 #include <sys/mplock2.h>
61 
62 static MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
63 MALLOC_DEFINE(M_SESSION, "session", "session header");
64 MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
65 MALLOC_DEFINE(M_LWP, "lwp", "lwp structures");
66 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
67 
68 int ps_showallprocs = 1;
69 static int ps_showallthreads = 1;
70 SYSCTL_INT(_security, OID_AUTO, ps_showallprocs, CTLFLAG_RW,
71     &ps_showallprocs, 0,
72     "Unprivileged processes can see proccesses with different UID/GID");
73 SYSCTL_INT(_security, OID_AUTO, ps_showallthreads, CTLFLAG_RW,
74     &ps_showallthreads, 0,
75     "Unprivileged processes can see kernel threads");
76 
77 static void pgdelete(struct pgrp *);
78 static void orphanpg(struct pgrp *pg);
79 static pid_t proc_getnewpid_locked(int random_offset);
80 
81 /*
82  * Other process lists
83  */
84 struct pidhashhead *pidhashtbl;
85 u_long pidhash;
86 struct pgrphashhead *pgrphashtbl;
87 u_long pgrphash;
88 struct proclist allproc;
89 struct proclist zombproc;
90 
91 /*
92  * Random component to nextpid generation.  We mix in a random factor to make
93  * it a little harder to predict.  We sanity check the modulus value to avoid
94  * doing it in critical paths.  Don't let it be too small or we pointlessly
95  * waste randomness entropy, and don't let it be impossibly large.  Using a
96  * modulus that is too big causes a LOT more process table scans and slows
97  * down fork processing as the pidchecked caching is defeated.
98  */
99 static int randompid = 0;
100 
101 /*
102  * No requirements.
103  */
104 static int
105 sysctl_kern_randompid(SYSCTL_HANDLER_ARGS)
106 {
107 	int error, pid;
108 
109 	pid = randompid;
110 	error = sysctl_handle_int(oidp, &pid, 0, req);
111 	if (error || !req->newptr)
112 		return (error);
113 	if (pid < 0 || pid > PID_MAX - 100)     /* out of range */
114 		pid = PID_MAX - 100;
115 	else if (pid < 2)                       /* NOP */
116 		pid = 0;
117 	else if (pid < 100)                     /* Make it reasonable */
118 		pid = 100;
119 	randompid = pid;
120 	return (error);
121 }
122 
123 SYSCTL_PROC(_kern, OID_AUTO, randompid, CTLTYPE_INT|CTLFLAG_RW,
124 	    0, 0, sysctl_kern_randompid, "I", "Random PID modulus");
125 
126 /*
127  * Initialize global process hashing structures.
128  *
129  * Called from the low level boot code only.
130  */
131 void
132 procinit(void)
133 {
134 	LIST_INIT(&allproc);
135 	LIST_INIT(&zombproc);
136 	lwkt_init();
137 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
138 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
139 	uihashinit();
140 }
141 
142 /*
143  * Is p an inferior of the current process?
144  *
145  * No requirements.
146  * The caller must hold proc_token if the caller wishes a stable result.
147  */
148 int
149 inferior(struct proc *p)
150 {
151 	lwkt_gettoken(&proc_token);
152 	while (p != curproc) {
153 		if (p->p_pid == 0) {
154 			lwkt_reltoken(&proc_token);
155 			return (0);
156 		}
157 		p = p->p_pptr;
158 	}
159 	lwkt_reltoken(&proc_token);
160 	return (1);
161 }
162 
163 /*
164  * Locate a process by number
165  *
166  * XXX TODO - change API to PHOLD() the returned process ?
167  *
168  * No requirements.
169  * The caller must hold proc_token if the caller wishes a stable result.
170  */
171 struct proc *
172 pfind(pid_t pid)
173 {
174 	struct proc *p;
175 
176 	lwkt_gettoken(&proc_token);
177 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
178 		if (p->p_pid == pid) {
179 			lwkt_reltoken(&proc_token);
180 			return (p);
181 		}
182 	}
183 	lwkt_reltoken(&proc_token);
184 	return (NULL);
185 }
186 
187 /*
188  * Locate a process group by number
189  *
190  * No requirements.
191  * The caller must hold proc_token if the caller wishes a stable result.
192  */
193 struct pgrp *
194 pgfind(pid_t pgid)
195 {
196 	struct pgrp *pgrp;
197 
198 	lwkt_gettoken(&proc_token);
199 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
200 		if (pgrp->pg_id == pgid) {
201 			lwkt_reltoken(&proc_token);
202 			return (pgrp);
203 		}
204 	}
205 	lwkt_reltoken(&proc_token);
206 	return (NULL);
207 }
208 
209 /*
210  * Move p to a new or existing process group (and session)
211  *
212  * No requirements.
213  */
214 int
215 enterpgrp(struct proc *p, pid_t pgid, int mksess)
216 {
217 	struct pgrp *pgrp;
218 	int error;
219 
220 	lwkt_gettoken(&proc_token);
221 	pgrp = pgfind(pgid);
222 
223 	KASSERT(pgrp == NULL || !mksess,
224 		("enterpgrp: setsid into non-empty pgrp"));
225 	KASSERT(!SESS_LEADER(p),
226 		("enterpgrp: session leader attempted setpgrp"));
227 
228 	if (pgrp == NULL) {
229 		pid_t savepid = p->p_pid;
230 		struct proc *np;
231 		/*
232 		 * new process group
233 		 */
234 		KASSERT(p->p_pid == pgid,
235 			("enterpgrp: new pgrp and pid != pgid"));
236 		if ((np = pfind(savepid)) == NULL || np != p) {
237 			error = ESRCH;
238 			goto fatal;
239 		}
240 		MALLOC(pgrp, struct pgrp *, sizeof(struct pgrp),
241 		       M_PGRP, M_WAITOK);
242 		if (mksess) {
243 			struct session *sess;
244 
245 			/*
246 			 * new session
247 			 */
248 			MALLOC(sess, struct session *, sizeof(struct session),
249 			       M_SESSION, M_WAITOK);
250 			sess->s_leader = p;
251 			sess->s_sid = p->p_pid;
252 			sess->s_count = 1;
253 			sess->s_ttyvp = NULL;
254 			sess->s_ttyp = NULL;
255 			bcopy(p->p_session->s_login, sess->s_login,
256 			      sizeof(sess->s_login));
257 			p->p_flag &= ~P_CONTROLT;
258 			pgrp->pg_session = sess;
259 			KASSERT(p == curproc,
260 				("enterpgrp: mksession and p != curproc"));
261 		} else {
262 			pgrp->pg_session = p->p_session;
263 			sess_hold(pgrp->pg_session);
264 		}
265 		pgrp->pg_id = pgid;
266 		LIST_INIT(&pgrp->pg_members);
267 		LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
268 		pgrp->pg_jobc = 0;
269 		SLIST_INIT(&pgrp->pg_sigiolst);
270 		lockinit(&pgrp->pg_lock, "pgwt", 0, 0);
271 	} else if (pgrp == p->p_pgrp) {
272 		goto done;
273 	}
274 
275 	/*
276 	 * Adjust eligibility of affected pgrps to participate in job control.
277 	 * Increment eligibility counts before decrementing, otherwise we
278 	 * could reach 0 spuriously during the first call.
279 	 */
280 	fixjobc(p, pgrp, 1);
281 	fixjobc(p, p->p_pgrp, 0);
282 
283 	LIST_REMOVE(p, p_pglist);
284 	if (LIST_EMPTY(&p->p_pgrp->pg_members))
285 		pgdelete(p->p_pgrp);
286 	p->p_pgrp = pgrp;
287 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
288 done:
289 	error = 0;
290 fatal:
291 	lwkt_reltoken(&proc_token);
292 	return (error);
293 }
294 
295 /*
296  * Remove process from process group
297  *
298  * No requirements.
299  */
300 int
301 leavepgrp(struct proc *p)
302 {
303 	lwkt_gettoken(&proc_token);
304 	LIST_REMOVE(p, p_pglist);
305 	if (LIST_EMPTY(&p->p_pgrp->pg_members))
306 		pgdelete(p->p_pgrp);
307 	p->p_pgrp = NULL;
308 	lwkt_reltoken(&proc_token);
309 	return (0);
310 }
311 
312 /*
313  * Delete a process group
314  *
315  * The caller must hold proc_token.
316  */
317 static void
318 pgdelete(struct pgrp *pgrp)
319 {
320 	/*
321 	 * Reset any sigio structures pointing to us as a result of
322 	 * F_SETOWN with our pgid.
323 	 */
324 	funsetownlst(&pgrp->pg_sigiolst);
325 
326 	if (pgrp->pg_session->s_ttyp != NULL &&
327 	    pgrp->pg_session->s_ttyp->t_pgrp == pgrp)
328 		pgrp->pg_session->s_ttyp->t_pgrp = NULL;
329 	LIST_REMOVE(pgrp, pg_hash);
330 	sess_rele(pgrp->pg_session);
331 	kfree(pgrp, M_PGRP);
332 }
333 
334 /*
335  * Adjust the ref count on a session structure.  When the ref count falls to
336  * zero the tty is disassociated from the session and the session structure
337  * is freed.  Note that tty assocation is not itself ref-counted.
338  *
339  * No requirements.
340  */
341 void
342 sess_hold(struct session *sp)
343 {
344 	lwkt_gettoken(&tty_token);
345 	++sp->s_count;
346 	lwkt_reltoken(&tty_token);
347 }
348 
349 /*
350  * No requirements.
351  */
352 void
353 sess_rele(struct session *sp)
354 {
355 	struct tty *tp;
356 
357 	KKASSERT(sp->s_count > 0);
358 	lwkt_gettoken(&tty_token);
359 	if (--sp->s_count == 0) {
360 		if (sp->s_ttyp && sp->s_ttyp->t_session) {
361 #ifdef TTY_DO_FULL_CLOSE
362 			/* FULL CLOSE, see ttyclearsession() */
363 			KKASSERT(sp->s_ttyp->t_session == sp);
364 			sp->s_ttyp->t_session = NULL;
365 #else
366 			/* HALF CLOSE, see ttyclearsession() */
367 			if (sp->s_ttyp->t_session == sp)
368 				sp->s_ttyp->t_session = NULL;
369 #endif
370 		}
371 		if ((tp = sp->s_ttyp) != NULL) {
372 			sp->s_ttyp = NULL;
373 			ttyunhold(tp);
374 		}
375 		kfree(sp, M_SESSION);
376 	}
377 	lwkt_reltoken(&tty_token);
378 }
379 
380 /*
381  * Adjust pgrp jobc counters when specified process changes process group.
382  * We count the number of processes in each process group that "qualify"
383  * the group for terminal job control (those with a parent in a different
384  * process group of the same session).  If that count reaches zero, the
385  * process group becomes orphaned.  Check both the specified process'
386  * process group and that of its children.
387  * entering == 0 => p is leaving specified group.
388  * entering == 1 => p is entering specified group.
389  *
390  * No requirements.
391  */
392 void
393 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
394 {
395 	struct pgrp *hispgrp;
396 	struct session *mysession;
397 
398 	/*
399 	 * Check p's parent to see whether p qualifies its own process
400 	 * group; if so, adjust count for p's process group.
401 	 */
402 	lwkt_gettoken(&proc_token);
403 	mysession = pgrp->pg_session;
404 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
405 	    hispgrp->pg_session == mysession) {
406 		if (entering)
407 			pgrp->pg_jobc++;
408 		else if (--pgrp->pg_jobc == 0)
409 			orphanpg(pgrp);
410 	}
411 
412 	/*
413 	 * Check this process' children to see whether they qualify
414 	 * their process groups; if so, adjust counts for children's
415 	 * process groups.
416 	 */
417 	LIST_FOREACH(p, &p->p_children, p_sibling) {
418 		if ((hispgrp = p->p_pgrp) != pgrp &&
419 		    hispgrp->pg_session == mysession &&
420 		    p->p_stat != SZOMB) {
421 			if (entering)
422 				hispgrp->pg_jobc++;
423 			else if (--hispgrp->pg_jobc == 0)
424 				orphanpg(hispgrp);
425 		}
426 	}
427 	lwkt_reltoken(&proc_token);
428 }
429 
430 /*
431  * A process group has become orphaned;
432  * if there are any stopped processes in the group,
433  * hang-up all process in that group.
434  *
435  * The caller must hold proc_token.
436  */
437 static void
438 orphanpg(struct pgrp *pg)
439 {
440 	struct proc *p;
441 
442 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
443 		if (p->p_stat == SSTOP) {
444 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
445 				ksignal(p, SIGHUP);
446 				ksignal(p, SIGCONT);
447 			}
448 			return;
449 		}
450 	}
451 }
452 
453 /*
454  * Add a new process to the allproc list and the PID hash.  This
455  * also assigns a pid to the new process.
456  *
457  * No requirements.
458  */
459 void
460 proc_add_allproc(struct proc *p)
461 {
462 	int random_offset;
463 
464 	if ((random_offset = randompid) != 0) {
465 		get_mplock();
466 		random_offset = karc4random() % random_offset;
467 		rel_mplock();
468 	}
469 
470 	lwkt_gettoken(&proc_token);
471 	p->p_pid = proc_getnewpid_locked(random_offset);
472 	LIST_INSERT_HEAD(&allproc, p, p_list);
473 	LIST_INSERT_HEAD(PIDHASH(p->p_pid), p, p_hash);
474 	lwkt_reltoken(&proc_token);
475 }
476 
477 /*
478  * Calculate a new process pid.  This function is integrated into
479  * proc_add_allproc() to guarentee that the new pid is not reused before
480  * the new process can be added to the allproc list.
481  *
482  * The caller must hold proc_token.
483  */
484 static
485 pid_t
486 proc_getnewpid_locked(int random_offset)
487 {
488 	static pid_t nextpid;
489 	static pid_t pidchecked;
490 	struct proc *p;
491 
492 	/*
493 	 * Find an unused process ID.  We remember a range of unused IDs
494 	 * ready to use (from nextpid+1 through pidchecked-1).
495 	 */
496 	nextpid = nextpid + 1 + random_offset;
497 retry:
498 	/*
499 	 * If the process ID prototype has wrapped around,
500 	 * restart somewhat above 0, as the low-numbered procs
501 	 * tend to include daemons that don't exit.
502 	 */
503 	if (nextpid >= PID_MAX) {
504 		nextpid = nextpid % PID_MAX;
505 		if (nextpid < 100)
506 			nextpid += 100;
507 		pidchecked = 0;
508 	}
509 	if (nextpid >= pidchecked) {
510 		int doingzomb = 0;
511 
512 		pidchecked = PID_MAX;
513 		/*
514 		 * Scan the active and zombie procs to check whether this pid
515 		 * is in use.  Remember the lowest pid that's greater
516 		 * than nextpid, so we can avoid checking for a while.
517 		 */
518 		p = LIST_FIRST(&allproc);
519 again:
520 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
521 			while (p->p_pid == nextpid ||
522 			    p->p_pgrp->pg_id == nextpid ||
523 			    p->p_session->s_sid == nextpid) {
524 				nextpid++;
525 				if (nextpid >= pidchecked)
526 					goto retry;
527 			}
528 			if (p->p_pid > nextpid && pidchecked > p->p_pid)
529 				pidchecked = p->p_pid;
530 			if (p->p_pgrp->pg_id > nextpid &&
531 			    pidchecked > p->p_pgrp->pg_id)
532 				pidchecked = p->p_pgrp->pg_id;
533 			if (p->p_session->s_sid > nextpid &&
534 			    pidchecked > p->p_session->s_sid)
535 				pidchecked = p->p_session->s_sid;
536 		}
537 		if (!doingzomb) {
538 			doingzomb = 1;
539 			p = LIST_FIRST(&zombproc);
540 			goto again;
541 		}
542 	}
543 	return(nextpid);
544 }
545 
546 /*
547  * Called from exit1 to remove a process from the allproc
548  * list and move it to the zombie list.
549  *
550  * No requirements.
551  */
552 void
553 proc_move_allproc_zombie(struct proc *p)
554 {
555 	lwkt_gettoken(&proc_token);
556 	while (p->p_lock) {
557 		tsleep(p, 0, "reap1", hz / 10);
558 	}
559 	LIST_REMOVE(p, p_list);
560 	LIST_INSERT_HEAD(&zombproc, p, p_list);
561 	LIST_REMOVE(p, p_hash);
562 	p->p_stat = SZOMB;
563 	lwkt_reltoken(&proc_token);
564 	dsched_exit_proc(p);
565 }
566 
567 /*
568  * This routine is called from kern_wait() and will remove the process
569  * from the zombie list and the sibling list.  This routine will block
570  * if someone has a lock on the proces (p_lock).
571  *
572  * No requirements.
573  */
574 void
575 proc_remove_zombie(struct proc *p)
576 {
577 	lwkt_gettoken(&proc_token);
578 	while (p->p_lock) {
579 		tsleep(p, 0, "reap1", hz / 10);
580 	}
581 	LIST_REMOVE(p, p_list); /* off zombproc */
582 	LIST_REMOVE(p, p_sibling);
583 	lwkt_reltoken(&proc_token);
584 }
585 
586 /*
587  * Scan all processes on the allproc list.  The process is automatically
588  * held for the callback.  A return value of -1 terminates the loop.
589  *
590  * No requirements.
591  * The callback is made with the process held and proc_token held.
592  */
593 void
594 allproc_scan(int (*callback)(struct proc *, void *), void *data)
595 {
596 	struct proc *p;
597 	int r;
598 
599 	lwkt_gettoken(&proc_token);
600 	LIST_FOREACH(p, &allproc, p_list) {
601 		PHOLD(p);
602 		r = callback(p, data);
603 		PRELE(p);
604 		if (r < 0)
605 			break;
606 	}
607 	lwkt_reltoken(&proc_token);
608 }
609 
610 /*
611  * Scan all lwps of processes on the allproc list.  The lwp is automatically
612  * held for the callback.  A return value of -1 terminates the loop.
613  *
614  * No requirements.
615  * The callback is made with the proces and lwp both held, and proc_token held.
616  */
617 void
618 alllwp_scan(int (*callback)(struct lwp *, void *), void *data)
619 {
620 	struct proc *p;
621 	struct lwp *lp;
622 	int r = 0;
623 
624 	lwkt_gettoken(&proc_token);
625 	LIST_FOREACH(p, &allproc, p_list) {
626 		PHOLD(p);
627 		FOREACH_LWP_IN_PROC(lp, p) {
628 			LWPHOLD(lp);
629 			r = callback(lp, data);
630 			LWPRELE(lp);
631 		}
632 		PRELE(p);
633 		if (r < 0)
634 			break;
635 	}
636 	lwkt_reltoken(&proc_token);
637 }
638 
639 /*
640  * Scan all processes on the zombproc list.  The process is automatically
641  * held for the callback.  A return value of -1 terminates the loop.
642  *
643  * No requirements.
644  * The callback is made with the proces held and proc_token held.
645  */
646 void
647 zombproc_scan(int (*callback)(struct proc *, void *), void *data)
648 {
649 	struct proc *p;
650 	int r;
651 
652 	lwkt_gettoken(&proc_token);
653 	LIST_FOREACH(p, &zombproc, p_list) {
654 		PHOLD(p);
655 		r = callback(p, data);
656 		PRELE(p);
657 		if (r < 0)
658 			break;
659 	}
660 	lwkt_reltoken(&proc_token);
661 }
662 
663 #include "opt_ddb.h"
664 #ifdef DDB
665 #include <ddb/ddb.h>
666 
667 /*
668  * Debugging only
669  */
670 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
671 {
672 	struct pgrp *pgrp;
673 	struct proc *p;
674 	int i;
675 
676 	for (i = 0; i <= pgrphash; i++) {
677 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
678 			kprintf("\tindx %d\n", i);
679 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
680 				kprintf(
681 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
682 				    (void *)pgrp, (long)pgrp->pg_id,
683 				    (void *)pgrp->pg_session,
684 				    pgrp->pg_session->s_count,
685 				    (void *)LIST_FIRST(&pgrp->pg_members));
686 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
687 					kprintf("\t\tpid %ld addr %p pgrp %p\n",
688 					    (long)p->p_pid, (void *)p,
689 					    (void *)p->p_pgrp);
690 				}
691 			}
692 		}
693 	}
694 }
695 #endif /* DDB */
696 
697 /*
698  * Locate a process on the zombie list.  Return a process or NULL.
699  *
700  * The caller must hold proc_token if a stable result is desired.
701  * No other requirements.
702  */
703 struct proc *
704 zpfind(pid_t pid)
705 {
706 	struct proc *p;
707 
708 	lwkt_gettoken(&proc_token);
709 	LIST_FOREACH(p, &zombproc, p_list) {
710 		if (p->p_pid == pid) {
711 			lwkt_reltoken(&proc_token);
712 			return (p);
713 		}
714 	}
715 	lwkt_reltoken(&proc_token);
716 	return (NULL);
717 }
718 
719 /*
720  * The caller must hold proc_token.
721  */
722 static int
723 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
724 {
725 	struct kinfo_proc ki;
726 	struct lwp *lp;
727 	int skp = 0, had_output = 0;
728 	int error;
729 
730 	bzero(&ki, sizeof(ki));
731 	fill_kinfo_proc(p, &ki);
732 	if ((flags & KERN_PROC_FLAG_LWP) == 0)
733 		skp = 1;
734 	error = 0;
735 	FOREACH_LWP_IN_PROC(lp, p) {
736 		LWPHOLD(lp);
737 		fill_kinfo_lwp(lp, &ki.kp_lwp);
738 		had_output = 1;
739 		error = SYSCTL_OUT(req, &ki, sizeof(ki));
740 		LWPRELE(lp);
741 		if (error)
742 			break;
743 		if (skp)
744 			break;
745 	}
746 	/* We need to output at least the proc, even if there is no lwp. */
747 	if (had_output == 0) {
748 		error = SYSCTL_OUT(req, &ki, sizeof(ki));
749 	}
750 	return (error);
751 }
752 
753 /*
754  * The caller must hold proc_token.
755  */
756 static int
757 sysctl_out_proc_kthread(struct thread *td, struct sysctl_req *req, int flags)
758 {
759 	struct kinfo_proc ki;
760 	int error;
761 
762 	fill_kinfo_proc_kthread(td, &ki);
763 	error = SYSCTL_OUT(req, &ki, sizeof(ki));
764 	if (error)
765 		return error;
766 	return(0);
767 }
768 
769 /*
770  * No requirements.
771  */
772 static int
773 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
774 {
775 	int *name = (int*) arg1;
776 	int oid = oidp->oid_number;
777 	u_int namelen = arg2;
778 	struct proc *p;
779 	struct proclist *plist;
780 	struct thread *td;
781 	int doingzomb, flags = 0;
782 	int error = 0;
783 	int n;
784 	int origcpu;
785 	struct ucred *cr1 = curproc->p_ucred;
786 
787 	flags = oid & KERN_PROC_FLAGMASK;
788 	oid &= ~KERN_PROC_FLAGMASK;
789 
790 	if ((oid == KERN_PROC_ALL && namelen != 0) ||
791 	    (oid != KERN_PROC_ALL && namelen != 1))
792 		return (EINVAL);
793 
794 	lwkt_gettoken(&proc_token);
795 	if (oid == KERN_PROC_PID) {
796 		p = pfind((pid_t)name[0]);
797 		if (p == NULL)
798 			goto post_threads;
799 		if (!PRISON_CHECK(cr1, p->p_ucred))
800 			goto post_threads;
801 		PHOLD(p);
802 		error = sysctl_out_proc(p, req, flags);
803 		PRELE(p);
804 		goto post_threads;
805 	}
806 
807 	if (!req->oldptr) {
808 		/* overestimate by 5 procs */
809 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
810 		if (error)
811 			goto post_threads;
812 	}
813 	for (doingzomb = 0; doingzomb <= 1; doingzomb++) {
814 		if (doingzomb)
815 			plist = &zombproc;
816 		else
817 			plist = &allproc;
818 		LIST_FOREACH(p, plist, p_list) {
819 			/*
820 			 * Show a user only their processes.
821 			 */
822 			if ((!ps_showallprocs) && p_trespass(cr1, p->p_ucred))
823 				continue;
824 			/*
825 			 * Skip embryonic processes.
826 			 */
827 			if (p->p_stat == SIDL)
828 				continue;
829 			/*
830 			 * TODO - make more efficient (see notes below).
831 			 * do by session.
832 			 */
833 			switch (oid) {
834 			case KERN_PROC_PGRP:
835 				/* could do this by traversing pgrp */
836 				if (p->p_pgrp == NULL ||
837 				    p->p_pgrp->pg_id != (pid_t)name[0])
838 					continue;
839 				break;
840 
841 			case KERN_PROC_TTY:
842 				if ((p->p_flag & P_CONTROLT) == 0 ||
843 				    p->p_session == NULL ||
844 				    p->p_session->s_ttyp == NULL ||
845 				    dev2udev(p->p_session->s_ttyp->t_dev) !=
846 					(udev_t)name[0])
847 					continue;
848 				break;
849 
850 			case KERN_PROC_UID:
851 				if (p->p_ucred == NULL ||
852 				    p->p_ucred->cr_uid != (uid_t)name[0])
853 					continue;
854 				break;
855 
856 			case KERN_PROC_RUID:
857 				if (p->p_ucred == NULL ||
858 				    p->p_ucred->cr_ruid != (uid_t)name[0])
859 					continue;
860 				break;
861 			}
862 
863 			if (!PRISON_CHECK(cr1, p->p_ucred))
864 				continue;
865 			PHOLD(p);
866 			error = sysctl_out_proc(p, req, flags);
867 			PRELE(p);
868 			if (error)
869 				goto post_threads;
870 		}
871 	}
872 
873 	/*
874 	 * Iterate over all active cpus and scan their thread list.  Start
875 	 * with the next logical cpu and end with our original cpu.  We
876 	 * migrate our own thread to each target cpu in order to safely scan
877 	 * its thread list.  In the last loop we migrate back to our original
878 	 * cpu.
879 	 */
880 	origcpu = mycpu->gd_cpuid;
881 	if (!ps_showallthreads || jailed(cr1))
882 		goto post_threads;
883 
884 	for (n = 1; n <= ncpus; ++n) {
885 		globaldata_t rgd;
886 		int nid;
887 
888 		nid = (origcpu + n) % ncpus;
889 		if ((smp_active_mask & CPUMASK(nid)) == 0)
890 			continue;
891 		rgd = globaldata_find(nid);
892 		lwkt_setcpu_self(rgd);
893 
894 		TAILQ_FOREACH(td, &mycpu->gd_tdallq, td_allq) {
895 			if (td->td_proc)
896 				continue;
897 			switch (oid) {
898 			case KERN_PROC_PGRP:
899 			case KERN_PROC_TTY:
900 			case KERN_PROC_UID:
901 			case KERN_PROC_RUID:
902 				continue;
903 			default:
904 				break;
905 			}
906 			lwkt_hold(td);
907 			error = sysctl_out_proc_kthread(td, req, doingzomb);
908 			lwkt_rele(td);
909 			if (error)
910 				goto post_threads;
911 		}
912 	}
913 post_threads:
914 	lwkt_reltoken(&proc_token);
915 	return (error);
916 }
917 
918 /*
919  * This sysctl allows a process to retrieve the argument list or process
920  * title for another process without groping around in the address space
921  * of the other process.  It also allow a process to set its own "process
922  * title to a string of its own choice.
923  *
924  * No requirements.
925  */
926 static int
927 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
928 {
929 	int *name = (int*) arg1;
930 	u_int namelen = arg2;
931 	struct proc *p;
932 	struct pargs *pa;
933 	int error = 0;
934 	struct ucred *cr1 = curproc->p_ucred;
935 
936 	if (namelen != 1)
937 		return (EINVAL);
938 
939 	lwkt_gettoken(&proc_token);
940 	p = pfind((pid_t)name[0]);
941 	if (p == NULL)
942 		goto done;
943 
944 	if ((!ps_argsopen) && p_trespass(cr1, p->p_ucred))
945 		goto done;
946 
947 	if (req->newptr && curproc != p) {
948 		error = EPERM;
949 		goto done;
950 	}
951 
952 	PHOLD(p);
953 	if (req->oldptr && p->p_args != NULL) {
954 		error = SYSCTL_OUT(req, p->p_args->ar_args,
955 				   p->p_args->ar_length);
956 	}
957 	if (req->newptr == NULL) {
958 		PRELE(p);
959 		goto done;
960 	}
961 
962 	if (p->p_args && --p->p_args->ar_ref == 0)
963 		FREE(p->p_args, M_PARGS);
964 	p->p_args = NULL;
965 
966 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit) {
967 		PRELE(p);
968 		goto done;
969 	}
970 
971 	MALLOC(pa, struct pargs *, sizeof(struct pargs) + req->newlen,
972 	       M_PARGS, M_WAITOK);
973 	pa->ar_ref = 1;
974 	pa->ar_length = req->newlen;
975 	error = SYSCTL_IN(req, pa->ar_args, req->newlen);
976 	if (!error)
977 		p->p_args = pa;
978 	else
979 		FREE(pa, M_PARGS);
980 	PRELE(p);
981 done:
982 	lwkt_reltoken(&proc_token);
983 	return (error);
984 }
985 
986 static int
987 sysctl_kern_proc_cwd(SYSCTL_HANDLER_ARGS)
988 {
989 	int *name = (int*) arg1;
990 	u_int namelen = arg2;
991 	struct proc *p;
992 	int error = 0;
993 	char *fullpath, *freepath;
994 	struct ucred *cr1 = curproc->p_ucred;
995 
996 	if (namelen != 1)
997 		return (EINVAL);
998 
999 	lwkt_gettoken(&proc_token);
1000 	p = pfind((pid_t)name[0]);
1001 	if (p == NULL)
1002 		goto done;
1003 
1004 	/*
1005 	 * If we are not allowed to see other args, we certainly shouldn't
1006 	 * get the cwd either. Also check the usual trespassing.
1007 	 */
1008 	if ((!ps_argsopen) && p_trespass(cr1, p->p_ucred))
1009 		goto done;
1010 
1011 	PHOLD(p);
1012 	if (req->oldptr && p->p_fd != NULL) {
1013 		error = cache_fullpath(p, &p->p_fd->fd_ncdir,
1014 		    &fullpath, &freepath, 0);
1015 		if (error)
1016 			goto done;
1017 		error = SYSCTL_OUT(req, fullpath, strlen(fullpath) + 1);
1018 		kfree(freepath, M_TEMP);
1019 	}
1020 
1021 	PRELE(p);
1022 
1023 done:
1024 	lwkt_reltoken(&proc_token);
1025 	return (error);
1026 }
1027 
1028 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
1029 
1030 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT,
1031 	0, 0, sysctl_kern_proc, "S,proc", "Return entire process table");
1032 
1033 SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD,
1034 	sysctl_kern_proc, "Process table");
1035 
1036 SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD,
1037 	sysctl_kern_proc, "Process table");
1038 
1039 SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD,
1040 	sysctl_kern_proc, "Process table");
1041 
1042 SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD,
1043 	sysctl_kern_proc, "Process table");
1044 
1045 SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD,
1046 	sysctl_kern_proc, "Process table");
1047 
1048 SYSCTL_NODE(_kern_proc, (KERN_PROC_ALL | KERN_PROC_FLAG_LWP), all_lwp, CTLFLAG_RD,
1049 	sysctl_kern_proc, "Process table");
1050 
1051 SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_FLAG_LWP), pgrp_lwp, CTLFLAG_RD,
1052 	sysctl_kern_proc, "Process table");
1053 
1054 SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_FLAG_LWP), tty_lwp, CTLFLAG_RD,
1055 	sysctl_kern_proc, "Process table");
1056 
1057 SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_FLAG_LWP), uid_lwp, CTLFLAG_RD,
1058 	sysctl_kern_proc, "Process table");
1059 
1060 SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_FLAG_LWP), ruid_lwp, CTLFLAG_RD,
1061 	sysctl_kern_proc, "Process table");
1062 
1063 SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_FLAG_LWP), pid_lwp, CTLFLAG_RD,
1064 	sysctl_kern_proc, "Process table");
1065 
1066 SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, CTLFLAG_RW | CTLFLAG_ANYBODY,
1067 	sysctl_kern_proc_args, "Process argument list");
1068 
1069 SYSCTL_NODE(_kern_proc, KERN_PROC_CWD, cwd, CTLFLAG_RD | CTLFLAG_ANYBODY,
1070 	sysctl_kern_proc_cwd, "Process argument list");
1071