xref: /dflybsd-src/sys/kern/kern_proc.c (revision 978400d3b04daf8f91ba8bb2dcc382a37ef632f4)
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 1.41 2008/01/04 12:16:19 matthias 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(_security, OID_AUTO, ps_showallprocs, CTLFLAG_RW,
67     &ps_showallprocs, 0,
68     "Unprivileged processes can see proccesses with different UID/GID");
69 SYSCTL_INT(_security, OID_AUTO, ps_showallthreads, CTLFLAG_RW,
70     &ps_showallthreads, 0,
71     "Unprivileged processes can see kernel threads");
72 
73 static void pgdelete(struct pgrp *);
74 static void orphanpg(struct pgrp *pg);
75 static pid_t proc_getnewpid_locked(int random_offset);
76 
77 /*
78  * Other process lists
79  */
80 struct pidhashhead *pidhashtbl;
81 u_long pidhash;
82 struct pgrphashhead *pgrphashtbl;
83 u_long pgrphash;
84 struct proclist allproc;
85 struct proclist zombproc;
86 struct spinlock allproc_spin;
87 vm_zone_t proc_zone;
88 vm_zone_t lwp_zone;
89 vm_zone_t thread_zone;
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 static int
102 sysctl_kern_randompid(SYSCTL_HANDLER_ARGS)
103 {
104 	int error, pid;
105 
106 	pid = randompid;
107 	error = sysctl_handle_int(oidp, &pid, 0, req);
108 	if (error || !req->newptr)
109 		return (error);
110 	if (pid < 0 || pid > PID_MAX - 100)     /* out of range */
111 		pid = PID_MAX - 100;
112 	else if (pid < 2)                       /* NOP */
113 		pid = 0;
114 	else if (pid < 100)                     /* Make it reasonable */
115 		pid = 100;
116 	randompid = pid;
117 	return (error);
118 }
119 
120 SYSCTL_PROC(_kern, OID_AUTO, randompid, CTLTYPE_INT|CTLFLAG_RW,
121 	    0, 0, sysctl_kern_randompid, "I", "Random PID modulus");
122 
123 /*
124  * Initialize global process hashing structures.
125  */
126 void
127 procinit(void)
128 {
129 	LIST_INIT(&allproc);
130 	LIST_INIT(&zombproc);
131 	spin_init(&allproc_spin);
132 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
133 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
134 	proc_zone = zinit("PROC", sizeof (struct proc), 0, 0, 5);
135 	lwp_zone = zinit("LWP", sizeof (struct lwp), 0, 0, 5);
136 	thread_zone = zinit("THREAD", sizeof (struct thread), 0, 0, 5);
137 	uihashinit();
138 }
139 
140 /*
141  * Is p an inferior of the current process?
142  */
143 int
144 inferior(struct proc *p)
145 {
146 	for (; p != curproc; p = p->p_pptr)
147 		if (p->p_pid == 0)
148 			return (0);
149 	return (1);
150 }
151 
152 /*
153  * Locate a process by number
154  */
155 struct proc *
156 pfind(pid_t pid)
157 {
158 	struct proc *p;
159 
160 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
161 		if (p->p_pid == pid)
162 			return (p);
163 	}
164 	return (NULL);
165 }
166 
167 /*
168  * Locate a process group by number
169  */
170 struct pgrp *
171 pgfind(pid_t pgid)
172 {
173 	struct pgrp *pgrp;
174 
175 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
176 		if (pgrp->pg_id == pgid)
177 			return (pgrp);
178 	}
179 	return (NULL);
180 }
181 
182 /*
183  * Move p to a new or existing process group (and session)
184  */
185 int
186 enterpgrp(struct proc *p, pid_t pgid, int mksess)
187 {
188 	struct pgrp *pgrp = pgfind(pgid);
189 
190 	KASSERT(pgrp == NULL || !mksess,
191 	    ("enterpgrp: setsid into non-empty pgrp"));
192 	KASSERT(!SESS_LEADER(p),
193 	    ("enterpgrp: session leader attempted setpgrp"));
194 
195 	if (pgrp == NULL) {
196 		pid_t savepid = p->p_pid;
197 		struct proc *np;
198 		/*
199 		 * new process group
200 		 */
201 		KASSERT(p->p_pid == pgid,
202 		    ("enterpgrp: new pgrp and pid != pgid"));
203 		if ((np = pfind(savepid)) == NULL || np != p)
204 			return (ESRCH);
205 		MALLOC(pgrp, struct pgrp *, sizeof(struct pgrp), M_PGRP,
206 		    M_WAITOK);
207 		if (mksess) {
208 			struct session *sess;
209 
210 			/*
211 			 * new session
212 			 */
213 			MALLOC(sess, struct session *, sizeof(struct session),
214 			    M_SESSION, M_WAITOK);
215 			sess->s_leader = p;
216 			sess->s_sid = p->p_pid;
217 			sess->s_count = 1;
218 			sess->s_ttyvp = NULL;
219 			sess->s_ttyp = NULL;
220 			bcopy(p->p_session->s_login, sess->s_login,
221 			    sizeof(sess->s_login));
222 			p->p_flag &= ~P_CONTROLT;
223 			pgrp->pg_session = sess;
224 			KASSERT(p == curproc,
225 			    ("enterpgrp: mksession and p != curproc"));
226 		} else {
227 			pgrp->pg_session = p->p_session;
228 			sess_hold(pgrp->pg_session);
229 		}
230 		pgrp->pg_id = pgid;
231 		LIST_INIT(&pgrp->pg_members);
232 		LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
233 		pgrp->pg_jobc = 0;
234 		SLIST_INIT(&pgrp->pg_sigiolst);
235 		lockinit(&pgrp->pg_lock, "pgwt", 0, 0);
236 	} else if (pgrp == p->p_pgrp)
237 		return (0);
238 
239 	/*
240 	 * Adjust eligibility of affected pgrps to participate in job control.
241 	 * Increment eligibility counts before decrementing, otherwise we
242 	 * could reach 0 spuriously during the first call.
243 	 */
244 	fixjobc(p, pgrp, 1);
245 	fixjobc(p, p->p_pgrp, 0);
246 
247 	LIST_REMOVE(p, p_pglist);
248 	if (LIST_EMPTY(&p->p_pgrp->pg_members))
249 		pgdelete(p->p_pgrp);
250 	p->p_pgrp = pgrp;
251 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
252 	return (0);
253 }
254 
255 /*
256  * remove process from process group
257  */
258 int
259 leavepgrp(struct proc *p)
260 {
261 
262 	LIST_REMOVE(p, p_pglist);
263 	if (LIST_EMPTY(&p->p_pgrp->pg_members))
264 		pgdelete(p->p_pgrp);
265 	p->p_pgrp = 0;
266 	return (0);
267 }
268 
269 /*
270  * delete a process group
271  */
272 static void
273 pgdelete(struct pgrp *pgrp)
274 {
275 
276 	/*
277 	 * Reset any sigio structures pointing to us as a result of
278 	 * F_SETOWN with our pgid.
279 	 */
280 	funsetownlst(&pgrp->pg_sigiolst);
281 
282 	if (pgrp->pg_session->s_ttyp != NULL &&
283 	    pgrp->pg_session->s_ttyp->t_pgrp == pgrp)
284 		pgrp->pg_session->s_ttyp->t_pgrp = NULL;
285 	LIST_REMOVE(pgrp, pg_hash);
286 	sess_rele(pgrp->pg_session);
287 	kfree(pgrp, M_PGRP);
288 }
289 
290 /*
291  * Adjust the ref count on a session structure.  When the ref count falls to
292  * zero the tty is disassociated from the session and the session structure
293  * is freed.  Note that tty assocation is not itself ref-counted.
294  */
295 void
296 sess_hold(struct session *sp)
297 {
298 	++sp->s_count;
299 }
300 
301 void
302 sess_rele(struct session *sp)
303 {
304 	KKASSERT(sp->s_count > 0);
305 	if (--sp->s_count == 0) {
306 		if (sp->s_ttyp && sp->s_ttyp->t_session) {
307 #ifdef TTY_DO_FULL_CLOSE
308 			/* FULL CLOSE, see ttyclearsession() */
309 			KKASSERT(sp->s_ttyp->t_session == sp);
310 			sp->s_ttyp->t_session = NULL;
311 #else
312 			/* HALF CLOSE, see ttyclearsession() */
313 			if (sp->s_ttyp->t_session == sp)
314 				sp->s_ttyp->t_session = NULL;
315 #endif
316 		}
317 		kfree(sp, M_SESSION);
318 	}
319 }
320 
321 /*
322  * Adjust pgrp jobc counters when specified process changes process group.
323  * We count the number of processes in each process group that "qualify"
324  * the group for terminal job control (those with a parent in a different
325  * process group of the same session).  If that count reaches zero, the
326  * process group becomes orphaned.  Check both the specified process'
327  * process group and that of its children.
328  * entering == 0 => p is leaving specified group.
329  * entering == 1 => p is entering specified group.
330  */
331 void
332 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
333 {
334 	struct pgrp *hispgrp;
335 	struct session *mysession = pgrp->pg_session;
336 
337 	/*
338 	 * Check p's parent to see whether p qualifies its own process
339 	 * group; if so, adjust count for p's process group.
340 	 */
341 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
342 	    hispgrp->pg_session == mysession) {
343 		if (entering)
344 			pgrp->pg_jobc++;
345 		else if (--pgrp->pg_jobc == 0)
346 			orphanpg(pgrp);
347 	}
348 
349 	/*
350 	 * Check this process' children to see whether they qualify
351 	 * their process groups; if so, adjust counts for children's
352 	 * process groups.
353 	 */
354 	LIST_FOREACH(p, &p->p_children, p_sibling)
355 		if ((hispgrp = p->p_pgrp) != pgrp &&
356 		    hispgrp->pg_session == mysession &&
357 		    p->p_stat != SZOMB) {
358 			if (entering)
359 				hispgrp->pg_jobc++;
360 			else if (--hispgrp->pg_jobc == 0)
361 				orphanpg(hispgrp);
362 		}
363 }
364 
365 /*
366  * A process group has become orphaned;
367  * if there are any stopped processes in the group,
368  * hang-up all process in that group.
369  */
370 static void
371 orphanpg(struct pgrp *pg)
372 {
373 	struct proc *p;
374 
375 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
376 		if (p->p_stat == SSTOP) {
377 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
378 				ksignal(p, SIGHUP);
379 				ksignal(p, SIGCONT);
380 			}
381 			return;
382 		}
383 	}
384 }
385 
386 /*
387  * Add a new process to the allproc list and the PID hash.  This
388  * also assigns a pid to the new process.
389  *
390  * MPALMOSTSAFE - acquires mplock for karc4random() call
391  */
392 void
393 proc_add_allproc(struct proc *p)
394 {
395 	int random_offset;
396 
397 	if ((random_offset = randompid) != 0) {
398 		get_mplock();
399 		random_offset = karc4random() % random_offset;
400 		rel_mplock();
401 	}
402 
403 	spin_lock_wr(&allproc_spin);
404 	p->p_pid = proc_getnewpid_locked(random_offset);
405 	LIST_INSERT_HEAD(&allproc, p, p_list);
406 	LIST_INSERT_HEAD(PIDHASH(p->p_pid), p, p_hash);
407 	spin_unlock_wr(&allproc_spin);
408 }
409 
410 /*
411  * Calculate a new process pid.  This function is integrated into
412  * proc_add_allproc() to guarentee that the new pid is not reused before
413  * the new process can be added to the allproc list.
414  *
415  * MPSAFE - must be called with allproc_spin held.
416  */
417 static
418 pid_t
419 proc_getnewpid_locked(int random_offset)
420 {
421 	static pid_t nextpid;
422 	static pid_t pidchecked;
423 	struct proc *p;
424 
425 	/*
426 	 * Find an unused process ID.  We remember a range of unused IDs
427 	 * ready to use (from nextpid+1 through pidchecked-1).
428 	 */
429 	nextpid = nextpid + 1 + random_offset;
430 retry:
431 	/*
432 	 * If the process ID prototype has wrapped around,
433 	 * restart somewhat above 0, as the low-numbered procs
434 	 * tend to include daemons that don't exit.
435 	 */
436 	if (nextpid >= PID_MAX) {
437 		nextpid = nextpid % PID_MAX;
438 		if (nextpid < 100)
439 			nextpid += 100;
440 		pidchecked = 0;
441 	}
442 	if (nextpid >= pidchecked) {
443 		int doingzomb = 0;
444 
445 		pidchecked = PID_MAX;
446 		/*
447 		 * Scan the active and zombie procs to check whether this pid
448 		 * is in use.  Remember the lowest pid that's greater
449 		 * than nextpid, so we can avoid checking for a while.
450 		 */
451 		p = LIST_FIRST(&allproc);
452 again:
453 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
454 			while (p->p_pid == nextpid ||
455 			    p->p_pgrp->pg_id == nextpid ||
456 			    p->p_session->s_sid == nextpid) {
457 				nextpid++;
458 				if (nextpid >= pidchecked)
459 					goto retry;
460 			}
461 			if (p->p_pid > nextpid && pidchecked > p->p_pid)
462 				pidchecked = p->p_pid;
463 			if (p->p_pgrp->pg_id > nextpid &&
464 			    pidchecked > p->p_pgrp->pg_id)
465 				pidchecked = p->p_pgrp->pg_id;
466 			if (p->p_session->s_sid > nextpid &&
467 			    pidchecked > p->p_session->s_sid)
468 				pidchecked = p->p_session->s_sid;
469 		}
470 		if (!doingzomb) {
471 			doingzomb = 1;
472 			p = LIST_FIRST(&zombproc);
473 			goto again;
474 		}
475 	}
476 	return(nextpid);
477 }
478 
479 /*
480  * Called from exit1 to remove a process from the allproc
481  * list and move it to the zombie list.
482  *
483  * MPSAFE
484  */
485 void
486 proc_move_allproc_zombie(struct proc *p)
487 {
488 	spin_lock_wr(&allproc_spin);
489 	while (p->p_lock) {
490 		spin_unlock_wr(&allproc_spin);
491 		tsleep(p, 0, "reap1", hz / 10);
492 		spin_lock_wr(&allproc_spin);
493 	}
494 	LIST_REMOVE(p, p_list);
495 	LIST_INSERT_HEAD(&zombproc, p, p_list);
496 	LIST_REMOVE(p, p_hash);
497 	p->p_stat = SZOMB;
498 	spin_unlock_wr(&allproc_spin);
499 }
500 
501 /*
502  * This routine is called from kern_wait() and will remove the process
503  * from the zombie list and the sibling list.  This routine will block
504  * if someone has a lock on the proces (p_lock).
505  *
506  * MPSAFE
507  */
508 void
509 proc_remove_zombie(struct proc *p)
510 {
511 	spin_lock_wr(&allproc_spin);
512 	while (p->p_lock) {
513 		spin_unlock_wr(&allproc_spin);
514 		tsleep(p, 0, "reap1", hz / 10);
515 		spin_lock_wr(&allproc_spin);
516 	}
517 	LIST_REMOVE(p, p_list); /* off zombproc */
518 	LIST_REMOVE(p, p_sibling);
519 	spin_unlock_wr(&allproc_spin);
520 }
521 
522 /*
523  * Scan all processes on the allproc list.  The process is automatically
524  * held for the callback.  A return value of -1 terminates the loop.
525  *
526  * MPSAFE
527  */
528 void
529 allproc_scan(int (*callback)(struct proc *, void *), void *data)
530 {
531 	struct proc *p;
532 	int r;
533 
534 	spin_lock_rd(&allproc_spin);
535 	LIST_FOREACH(p, &allproc, p_list) {
536 		PHOLD(p);
537 		spin_unlock_rd(&allproc_spin);
538 		r = callback(p, data);
539 		spin_lock_rd(&allproc_spin);
540 		PRELE(p);
541 		if (r < 0)
542 			break;
543 	}
544 	spin_unlock_rd(&allproc_spin);
545 }
546 
547 /*
548  * Scan all lwps of processes on the allproc list.  The lwp is automatically
549  * held for the callback.  A return value of -1 terminates the loop.
550  *
551  * possibly not MPSAFE, needs to access foreingn proc structures
552  */
553 void
554 alllwp_scan(int (*callback)(struct lwp *, void *), void *data)
555 {
556 	struct proc *p;
557 	struct lwp *lp;
558 	int r = 0;
559 
560 	spin_lock_rd(&allproc_spin);
561 	LIST_FOREACH(p, &allproc, p_list) {
562 		PHOLD(p);
563 		spin_unlock_rd(&allproc_spin);
564 		FOREACH_LWP_IN_PROC(lp, p) {
565 			LWPHOLD(lp);
566 			r = callback(lp, data);
567 			LWPRELE(lp);
568 		}
569 		spin_lock_rd(&allproc_spin);
570 		PRELE(p);
571 		if (r < 0)
572 			break;
573 	}
574 	spin_unlock_rd(&allproc_spin);
575 }
576 
577 /*
578  * Scan all processes on the zombproc list.  The process is automatically
579  * held for the callback.  A return value of -1 terminates the loop.
580  *
581  * MPSAFE
582  */
583 void
584 zombproc_scan(int (*callback)(struct proc *, void *), void *data)
585 {
586 	struct proc *p;
587 	int r;
588 
589 	spin_lock_rd(&allproc_spin);
590 	LIST_FOREACH(p, &zombproc, p_list) {
591 		PHOLD(p);
592 		spin_unlock_rd(&allproc_spin);
593 		r = callback(p, data);
594 		spin_lock_rd(&allproc_spin);
595 		PRELE(p);
596 		if (r < 0)
597 			break;
598 	}
599 	spin_unlock_rd(&allproc_spin);
600 }
601 
602 #include "opt_ddb.h"
603 #ifdef DDB
604 #include <ddb/ddb.h>
605 
606 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
607 {
608 	struct pgrp *pgrp;
609 	struct proc *p;
610 	int i;
611 
612 	for (i = 0; i <= pgrphash; i++) {
613 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
614 			kprintf("\tindx %d\n", i);
615 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
616 				kprintf(
617 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
618 				    (void *)pgrp, (long)pgrp->pg_id,
619 				    (void *)pgrp->pg_session,
620 				    pgrp->pg_session->s_count,
621 				    (void *)LIST_FIRST(&pgrp->pg_members));
622 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
623 					kprintf("\t\tpid %ld addr %p pgrp %p\n",
624 					    (long)p->p_pid, (void *)p,
625 					    (void *)p->p_pgrp);
626 				}
627 			}
628 		}
629 	}
630 }
631 #endif /* DDB */
632 
633 /*
634  * Locate a process on the zombie list.  Return a held process or NULL.
635  */
636 struct proc *
637 zpfind(pid_t pid)
638 {
639 	struct proc *p;
640 
641 	LIST_FOREACH(p, &zombproc, p_list)
642 		if (p->p_pid == pid)
643 			return (p);
644 	return (NULL);
645 }
646 
647 static int
648 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
649 {
650 	struct kinfo_proc ki;
651 	struct lwp *lp;
652 	int skp = 0, had_output = 0;
653 	int error;
654 
655 	fill_kinfo_proc(p, &ki);
656 	if ((flags & KERN_PROC_FLAG_LWP) == 0)
657 		skp = 1;
658 	FOREACH_LWP_IN_PROC(lp, p) {
659 		fill_kinfo_lwp(lp, &ki.kp_lwp);
660 output:
661 		had_output = 1;
662 		error = SYSCTL_OUT(req, &ki, sizeof(ki));
663 		if (error)
664 			return error;
665 		if (skp)
666 			break;
667 	}
668 	/* We need to output at least the proc, even if there is no lwp. */
669 	if (!had_output)
670 		goto output;
671 #if 0
672 	if (!doingzomb && pid && (pfind(pid) != p))
673 		return EAGAIN;
674 	if (doingzomb && zpfind(pid) != p)
675 		return EAGAIN;
676 #endif
677 	return (0);
678 }
679 
680 static int
681 sysctl_out_proc_kthread(struct thread *td, struct sysctl_req *req, int flags)
682 {
683 	struct kinfo_proc ki;
684 	int error;
685 
686 	fill_kinfo_proc_kthread(td, &ki);
687 	error = SYSCTL_OUT(req, &ki, sizeof(ki));
688 	if (error)
689 		return error;
690 	return(0);
691 }
692 
693 static int
694 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
695 {
696 	int *name = (int*) arg1;
697 	int oid = oidp->oid_number;
698 	u_int namelen = arg2;
699 	struct proc *p, *np;
700 	struct proclist *plist;
701 	struct thread *td;
702 	int doingzomb, flags = 0;
703 	int error = 0;
704 	int n;
705 	int origcpu;
706 	struct ucred *cr1 = curproc->p_ucred;
707 
708 	flags = oid & KERN_PROC_FLAGMASK;
709 	oid &= ~KERN_PROC_FLAGMASK;
710 
711 	if ((oid == KERN_PROC_ALL && namelen != 0) ||
712 	    (oid != KERN_PROC_ALL && namelen != 1))
713 		return (EINVAL);
714 
715 	if (oid == KERN_PROC_PID) {
716 		p = pfind((pid_t)name[0]);
717 		if (!p)
718 			return (0);
719 		if (!PRISON_CHECK(cr1, p->p_ucred))
720 			return (0);
721 		PHOLD(p);
722 		error = sysctl_out_proc(p, req, flags);
723 		PRELE(p);
724 		return (error);
725 	}
726 
727 	if (!req->oldptr) {
728 		/* overestimate by 5 procs */
729 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
730 		if (error)
731 			return (error);
732 	}
733 	for (doingzomb = 0; doingzomb <= 1; doingzomb++) {
734 		if (doingzomb)
735 			plist = &zombproc;
736 		else
737 			plist = &allproc;
738 		LIST_FOREACH_MUTABLE(p, plist, p_list, np) {
739 			/*
740 			 * Show a user only their processes.
741 			 */
742 			if ((!ps_showallprocs) && p_trespass(cr1, p->p_ucred))
743 				continue;
744 			/*
745 			 * Skip embryonic processes.
746 			 */
747 			if (p->p_stat == SIDL)
748 				continue;
749 			/*
750 			 * TODO - make more efficient (see notes below).
751 			 * do by session.
752 			 */
753 			switch (oid) {
754 			case KERN_PROC_PGRP:
755 				/* could do this by traversing pgrp */
756 				if (p->p_pgrp == NULL ||
757 				    p->p_pgrp->pg_id != (pid_t)name[0])
758 					continue;
759 				break;
760 
761 			case KERN_PROC_TTY:
762 				if ((p->p_flag & P_CONTROLT) == 0 ||
763 				    p->p_session == NULL ||
764 				    p->p_session->s_ttyp == NULL ||
765 				    dev2udev(p->p_session->s_ttyp->t_dev) !=
766 					(udev_t)name[0])
767 					continue;
768 				break;
769 
770 			case KERN_PROC_UID:
771 				if (p->p_ucred == NULL ||
772 				    p->p_ucred->cr_uid != (uid_t)name[0])
773 					continue;
774 				break;
775 
776 			case KERN_PROC_RUID:
777 				if (p->p_ucred == NULL ||
778 				    p->p_ucred->cr_ruid != (uid_t)name[0])
779 					continue;
780 				break;
781 			}
782 
783 			if (!PRISON_CHECK(cr1, p->p_ucred))
784 				continue;
785 			PHOLD(p);
786 			error = sysctl_out_proc(p, req, flags);
787 			PRELE(p);
788 			if (error)
789 				return (error);
790 		}
791 	}
792 
793 	/*
794 	 * Iterate over all active cpus and scan their thread list.  Start
795 	 * with the next logical cpu and end with our original cpu.  We
796 	 * migrate our own thread to each target cpu in order to safely scan
797 	 * its thread list.  In the last loop we migrate back to our original
798 	 * cpu.
799 	 */
800 	origcpu = mycpu->gd_cpuid;
801 	if (!ps_showallthreads || jailed(cr1))
802 		goto post_threads;
803 	for (n = 1; n <= ncpus; ++n) {
804 		globaldata_t rgd;
805 		int nid;
806 
807 		nid = (origcpu + n) % ncpus;
808 		if ((smp_active_mask & (1 << nid)) == 0)
809 			continue;
810 		rgd = globaldata_find(nid);
811 		lwkt_setcpu_self(rgd);
812 
813 		TAILQ_FOREACH(td, &mycpu->gd_tdallq, td_allq) {
814 			if (td->td_proc)
815 				continue;
816 			switch (oid) {
817 			case KERN_PROC_PGRP:
818 			case KERN_PROC_TTY:
819 			case KERN_PROC_UID:
820 			case KERN_PROC_RUID:
821 				continue;
822 			default:
823 				break;
824 			}
825 			lwkt_hold(td);
826 			error = sysctl_out_proc_kthread(td, req, doingzomb);
827 			lwkt_rele(td);
828 			if (error)
829 				return (error);
830 		}
831 	}
832 post_threads:
833 	return (0);
834 }
835 
836 /*
837  * This sysctl allows a process to retrieve the argument list or process
838  * title for another process without groping around in the address space
839  * of the other process.  It also allow a process to set its own "process
840  * title to a string of its own choice.
841  */
842 static int
843 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
844 {
845 	int *name = (int*) arg1;
846 	u_int namelen = arg2;
847 	struct proc *p;
848 	struct pargs *pa;
849 	int error = 0;
850 	struct ucred *cr1 = curproc->p_ucred;
851 
852 	if (namelen != 1)
853 		return (EINVAL);
854 
855 	p = pfind((pid_t)name[0]);
856 	if (!p)
857 		return (0);
858 
859 	if ((!ps_argsopen) && p_trespass(cr1, p->p_ucred))
860 		return (0);
861 
862 	if (req->newptr && curproc != p)
863 		return (EPERM);
864 
865 	if (req->oldptr && p->p_args != NULL)
866 		error = SYSCTL_OUT(req, p->p_args->ar_args, p->p_args->ar_length);
867 	if (req->newptr == NULL)
868 		return (error);
869 
870 	if (p->p_args && --p->p_args->ar_ref == 0)
871 		FREE(p->p_args, M_PARGS);
872 	p->p_args = NULL;
873 
874 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
875 		return (error);
876 
877 	MALLOC(pa, struct pargs *, sizeof(struct pargs) + req->newlen,
878 	    M_PARGS, M_WAITOK);
879 	pa->ar_ref = 1;
880 	pa->ar_length = req->newlen;
881 	error = SYSCTL_IN(req, pa->ar_args, req->newlen);
882 	if (!error)
883 		p->p_args = pa;
884 	else
885 		FREE(pa, M_PARGS);
886 	return (error);
887 }
888 
889 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
890 
891 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT,
892 	0, 0, sysctl_kern_proc, "S,proc", "Return entire process table");
893 
894 SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD,
895 	sysctl_kern_proc, "Process table");
896 
897 SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD,
898 	sysctl_kern_proc, "Process table");
899 
900 SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD,
901 	sysctl_kern_proc, "Process table");
902 
903 SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD,
904 	sysctl_kern_proc, "Process table");
905 
906 SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD,
907 	sysctl_kern_proc, "Process table");
908 
909 SYSCTL_NODE(_kern_proc, (KERN_PROC_ALL | KERN_PROC_FLAG_LWP), all_lwp, CTLFLAG_RD,
910 	sysctl_kern_proc, "Process table");
911 
912 SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_FLAG_LWP), pgrp_lwp, CTLFLAG_RD,
913 	sysctl_kern_proc, "Process table");
914 
915 SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_FLAG_LWP), tty_lwp, CTLFLAG_RD,
916 	sysctl_kern_proc, "Process table");
917 
918 SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_FLAG_LWP), uid_lwp, CTLFLAG_RD,
919 	sysctl_kern_proc, "Process table");
920 
921 SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_FLAG_LWP), ruid_lwp, CTLFLAG_RD,
922 	sysctl_kern_proc, "Process table");
923 
924 SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_FLAG_LWP), pid_lwp, CTLFLAG_RD,
925 	sysctl_kern_proc, "Process table");
926 
927 SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, CTLFLAG_RW | CTLFLAG_ANYBODY,
928 	sysctl_kern_proc_args, "Process argument list");
929