xref: /dflybsd-src/sys/kern/kern_proc.c (revision 2e7bf158f373428dba2c765c927f14d9e94f00a4)
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 	KKASSERT(sp->s_count > 0);
356 	lwkt_gettoken(&tty_token);
357 	if (--sp->s_count == 0) {
358 		if (sp->s_ttyp && sp->s_ttyp->t_session) {
359 #ifdef TTY_DO_FULL_CLOSE
360 			/* FULL CLOSE, see ttyclearsession() */
361 			KKASSERT(sp->s_ttyp->t_session == sp);
362 			sp->s_ttyp->t_session = NULL;
363 #else
364 			/* HALF CLOSE, see ttyclearsession() */
365 			if (sp->s_ttyp->t_session == sp)
366 				sp->s_ttyp->t_session = NULL;
367 #endif
368 		}
369 		kfree(sp, M_SESSION);
370 	}
371 	lwkt_reltoken(&tty_token);
372 }
373 
374 /*
375  * Adjust pgrp jobc counters when specified process changes process group.
376  * We count the number of processes in each process group that "qualify"
377  * the group for terminal job control (those with a parent in a different
378  * process group of the same session).  If that count reaches zero, the
379  * process group becomes orphaned.  Check both the specified process'
380  * process group and that of its children.
381  * entering == 0 => p is leaving specified group.
382  * entering == 1 => p is entering specified group.
383  *
384  * No requirements.
385  */
386 void
387 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
388 {
389 	struct pgrp *hispgrp;
390 	struct session *mysession;
391 
392 	/*
393 	 * Check p's parent to see whether p qualifies its own process
394 	 * group; if so, adjust count for p's process group.
395 	 */
396 	lwkt_gettoken(&proc_token);
397 	mysession = pgrp->pg_session;
398 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
399 	    hispgrp->pg_session == mysession) {
400 		if (entering)
401 			pgrp->pg_jobc++;
402 		else if (--pgrp->pg_jobc == 0)
403 			orphanpg(pgrp);
404 	}
405 
406 	/*
407 	 * Check this process' children to see whether they qualify
408 	 * their process groups; if so, adjust counts for children's
409 	 * process groups.
410 	 */
411 	LIST_FOREACH(p, &p->p_children, p_sibling) {
412 		if ((hispgrp = p->p_pgrp) != pgrp &&
413 		    hispgrp->pg_session == mysession &&
414 		    p->p_stat != SZOMB) {
415 			if (entering)
416 				hispgrp->pg_jobc++;
417 			else if (--hispgrp->pg_jobc == 0)
418 				orphanpg(hispgrp);
419 		}
420 	}
421 	lwkt_reltoken(&proc_token);
422 }
423 
424 /*
425  * A process group has become orphaned;
426  * if there are any stopped processes in the group,
427  * hang-up all process in that group.
428  *
429  * The caller must hold proc_token.
430  */
431 static void
432 orphanpg(struct pgrp *pg)
433 {
434 	struct proc *p;
435 
436 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
437 		if (p->p_stat == SSTOP) {
438 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
439 				ksignal(p, SIGHUP);
440 				ksignal(p, SIGCONT);
441 			}
442 			return;
443 		}
444 	}
445 }
446 
447 /*
448  * Add a new process to the allproc list and the PID hash.  This
449  * also assigns a pid to the new process.
450  *
451  * No requirements.
452  */
453 void
454 proc_add_allproc(struct proc *p)
455 {
456 	int random_offset;
457 
458 	if ((random_offset = randompid) != 0) {
459 		get_mplock();
460 		random_offset = karc4random() % random_offset;
461 		rel_mplock();
462 	}
463 
464 	lwkt_gettoken(&proc_token);
465 	p->p_pid = proc_getnewpid_locked(random_offset);
466 	LIST_INSERT_HEAD(&allproc, p, p_list);
467 	LIST_INSERT_HEAD(PIDHASH(p->p_pid), p, p_hash);
468 	lwkt_reltoken(&proc_token);
469 }
470 
471 /*
472  * Calculate a new process pid.  This function is integrated into
473  * proc_add_allproc() to guarentee that the new pid is not reused before
474  * the new process can be added to the allproc list.
475  *
476  * The caller must hold proc_token.
477  */
478 static
479 pid_t
480 proc_getnewpid_locked(int random_offset)
481 {
482 	static pid_t nextpid;
483 	static pid_t pidchecked;
484 	struct proc *p;
485 
486 	/*
487 	 * Find an unused process ID.  We remember a range of unused IDs
488 	 * ready to use (from nextpid+1 through pidchecked-1).
489 	 */
490 	nextpid = nextpid + 1 + random_offset;
491 retry:
492 	/*
493 	 * If the process ID prototype has wrapped around,
494 	 * restart somewhat above 0, as the low-numbered procs
495 	 * tend to include daemons that don't exit.
496 	 */
497 	if (nextpid >= PID_MAX) {
498 		nextpid = nextpid % PID_MAX;
499 		if (nextpid < 100)
500 			nextpid += 100;
501 		pidchecked = 0;
502 	}
503 	if (nextpid >= pidchecked) {
504 		int doingzomb = 0;
505 
506 		pidchecked = PID_MAX;
507 		/*
508 		 * Scan the active and zombie procs to check whether this pid
509 		 * is in use.  Remember the lowest pid that's greater
510 		 * than nextpid, so we can avoid checking for a while.
511 		 */
512 		p = LIST_FIRST(&allproc);
513 again:
514 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
515 			while (p->p_pid == nextpid ||
516 			    p->p_pgrp->pg_id == nextpid ||
517 			    p->p_session->s_sid == nextpid) {
518 				nextpid++;
519 				if (nextpid >= pidchecked)
520 					goto retry;
521 			}
522 			if (p->p_pid > nextpid && pidchecked > p->p_pid)
523 				pidchecked = p->p_pid;
524 			if (p->p_pgrp->pg_id > nextpid &&
525 			    pidchecked > p->p_pgrp->pg_id)
526 				pidchecked = p->p_pgrp->pg_id;
527 			if (p->p_session->s_sid > nextpid &&
528 			    pidchecked > p->p_session->s_sid)
529 				pidchecked = p->p_session->s_sid;
530 		}
531 		if (!doingzomb) {
532 			doingzomb = 1;
533 			p = LIST_FIRST(&zombproc);
534 			goto again;
535 		}
536 	}
537 	return(nextpid);
538 }
539 
540 /*
541  * Called from exit1 to remove a process from the allproc
542  * list and move it to the zombie list.
543  *
544  * No requirements.
545  */
546 void
547 proc_move_allproc_zombie(struct proc *p)
548 {
549 	lwkt_gettoken(&proc_token);
550 	while (p->p_lock) {
551 		tsleep(p, 0, "reap1", hz / 10);
552 	}
553 	LIST_REMOVE(p, p_list);
554 	LIST_INSERT_HEAD(&zombproc, p, p_list);
555 	LIST_REMOVE(p, p_hash);
556 	p->p_stat = SZOMB;
557 	lwkt_reltoken(&proc_token);
558 	dsched_exit_proc(p);
559 }
560 
561 /*
562  * This routine is called from kern_wait() and will remove the process
563  * from the zombie list and the sibling list.  This routine will block
564  * if someone has a lock on the proces (p_lock).
565  *
566  * No requirements.
567  */
568 void
569 proc_remove_zombie(struct proc *p)
570 {
571 	lwkt_gettoken(&proc_token);
572 	while (p->p_lock) {
573 		tsleep(p, 0, "reap1", hz / 10);
574 	}
575 	LIST_REMOVE(p, p_list); /* off zombproc */
576 	LIST_REMOVE(p, p_sibling);
577 	lwkt_reltoken(&proc_token);
578 }
579 
580 /*
581  * Scan all processes on the allproc list.  The process is automatically
582  * held for the callback.  A return value of -1 terminates the loop.
583  *
584  * No requirements.
585  * The callback is made with the process held and proc_token held.
586  */
587 void
588 allproc_scan(int (*callback)(struct proc *, void *), void *data)
589 {
590 	struct proc *p;
591 	int r;
592 
593 	lwkt_gettoken(&proc_token);
594 	LIST_FOREACH(p, &allproc, p_list) {
595 		PHOLD(p);
596 		r = callback(p, data);
597 		PRELE(p);
598 		if (r < 0)
599 			break;
600 	}
601 	lwkt_reltoken(&proc_token);
602 }
603 
604 /*
605  * Scan all lwps of processes on the allproc list.  The lwp is automatically
606  * held for the callback.  A return value of -1 terminates the loop.
607  *
608  * No requirements.
609  * The callback is made with the proces and lwp both held, and proc_token held.
610  */
611 void
612 alllwp_scan(int (*callback)(struct lwp *, void *), void *data)
613 {
614 	struct proc *p;
615 	struct lwp *lp;
616 	int r = 0;
617 
618 	lwkt_gettoken(&proc_token);
619 	LIST_FOREACH(p, &allproc, p_list) {
620 		PHOLD(p);
621 		FOREACH_LWP_IN_PROC(lp, p) {
622 			LWPHOLD(lp);
623 			r = callback(lp, data);
624 			LWPRELE(lp);
625 		}
626 		PRELE(p);
627 		if (r < 0)
628 			break;
629 	}
630 	lwkt_reltoken(&proc_token);
631 }
632 
633 /*
634  * Scan all processes on the zombproc list.  The process is automatically
635  * held for the callback.  A return value of -1 terminates the loop.
636  *
637  * No requirements.
638  * The callback is made with the proces held and proc_token held.
639  */
640 void
641 zombproc_scan(int (*callback)(struct proc *, void *), void *data)
642 {
643 	struct proc *p;
644 	int r;
645 
646 	lwkt_gettoken(&proc_token);
647 	LIST_FOREACH(p, &zombproc, p_list) {
648 		PHOLD(p);
649 		r = callback(p, data);
650 		PRELE(p);
651 		if (r < 0)
652 			break;
653 	}
654 	lwkt_reltoken(&proc_token);
655 }
656 
657 #include "opt_ddb.h"
658 #ifdef DDB
659 #include <ddb/ddb.h>
660 
661 /*
662  * Debugging only
663  */
664 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
665 {
666 	struct pgrp *pgrp;
667 	struct proc *p;
668 	int i;
669 
670 	for (i = 0; i <= pgrphash; i++) {
671 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
672 			kprintf("\tindx %d\n", i);
673 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
674 				kprintf(
675 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
676 				    (void *)pgrp, (long)pgrp->pg_id,
677 				    (void *)pgrp->pg_session,
678 				    pgrp->pg_session->s_count,
679 				    (void *)LIST_FIRST(&pgrp->pg_members));
680 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
681 					kprintf("\t\tpid %ld addr %p pgrp %p\n",
682 					    (long)p->p_pid, (void *)p,
683 					    (void *)p->p_pgrp);
684 				}
685 			}
686 		}
687 	}
688 }
689 #endif /* DDB */
690 
691 /*
692  * Locate a process on the zombie list.  Return a process or NULL.
693  *
694  * The caller must hold proc_token if a stable result is desired.
695  * No other requirements.
696  */
697 struct proc *
698 zpfind(pid_t pid)
699 {
700 	struct proc *p;
701 
702 	lwkt_gettoken(&proc_token);
703 	LIST_FOREACH(p, &zombproc, p_list) {
704 		if (p->p_pid == pid) {
705 			lwkt_reltoken(&proc_token);
706 			return (p);
707 		}
708 	}
709 	lwkt_reltoken(&proc_token);
710 	return (NULL);
711 }
712 
713 /*
714  * The caller must hold proc_token.
715  */
716 static int
717 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
718 {
719 	struct kinfo_proc ki;
720 	struct lwp *lp;
721 	int skp = 0, had_output = 0;
722 	int error;
723 
724 	bzero(&ki, sizeof(ki));
725 	fill_kinfo_proc(p, &ki);
726 	if ((flags & KERN_PROC_FLAG_LWP) == 0)
727 		skp = 1;
728 	error = 0;
729 	FOREACH_LWP_IN_PROC(lp, p) {
730 		LWPHOLD(lp);
731 		fill_kinfo_lwp(lp, &ki.kp_lwp);
732 		had_output = 1;
733 		error = SYSCTL_OUT(req, &ki, sizeof(ki));
734 		LWPRELE(lp);
735 		if (error)
736 			break;
737 		if (skp)
738 			break;
739 	}
740 	/* We need to output at least the proc, even if there is no lwp. */
741 	if (had_output == 0) {
742 		error = SYSCTL_OUT(req, &ki, sizeof(ki));
743 	}
744 	return (error);
745 }
746 
747 /*
748  * The caller must hold proc_token.
749  */
750 static int
751 sysctl_out_proc_kthread(struct thread *td, struct sysctl_req *req, int flags)
752 {
753 	struct kinfo_proc ki;
754 	int error;
755 
756 	fill_kinfo_proc_kthread(td, &ki);
757 	error = SYSCTL_OUT(req, &ki, sizeof(ki));
758 	if (error)
759 		return error;
760 	return(0);
761 }
762 
763 /*
764  * No requirements.
765  */
766 static int
767 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
768 {
769 	int *name = (int*) arg1;
770 	int oid = oidp->oid_number;
771 	u_int namelen = arg2;
772 	struct proc *p;
773 	struct proclist *plist;
774 	struct thread *td;
775 	int doingzomb, flags = 0;
776 	int error = 0;
777 	int n;
778 	int origcpu;
779 	struct ucred *cr1 = curproc->p_ucred;
780 
781 	flags = oid & KERN_PROC_FLAGMASK;
782 	oid &= ~KERN_PROC_FLAGMASK;
783 
784 	if ((oid == KERN_PROC_ALL && namelen != 0) ||
785 	    (oid != KERN_PROC_ALL && namelen != 1))
786 		return (EINVAL);
787 
788 	lwkt_gettoken(&proc_token);
789 	if (oid == KERN_PROC_PID) {
790 		p = pfind((pid_t)name[0]);
791 		if (p == NULL)
792 			goto post_threads;
793 		if (!PRISON_CHECK(cr1, p->p_ucred))
794 			goto post_threads;
795 		PHOLD(p);
796 		error = sysctl_out_proc(p, req, flags);
797 		PRELE(p);
798 		goto post_threads;
799 	}
800 
801 	if (!req->oldptr) {
802 		/* overestimate by 5 procs */
803 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
804 		if (error)
805 			goto post_threads;
806 	}
807 	for (doingzomb = 0; doingzomb <= 1; doingzomb++) {
808 		if (doingzomb)
809 			plist = &zombproc;
810 		else
811 			plist = &allproc;
812 		LIST_FOREACH(p, plist, p_list) {
813 			/*
814 			 * Show a user only their processes.
815 			 */
816 			if ((!ps_showallprocs) && p_trespass(cr1, p->p_ucred))
817 				continue;
818 			/*
819 			 * Skip embryonic processes.
820 			 */
821 			if (p->p_stat == SIDL)
822 				continue;
823 			/*
824 			 * TODO - make more efficient (see notes below).
825 			 * do by session.
826 			 */
827 			switch (oid) {
828 			case KERN_PROC_PGRP:
829 				/* could do this by traversing pgrp */
830 				if (p->p_pgrp == NULL ||
831 				    p->p_pgrp->pg_id != (pid_t)name[0])
832 					continue;
833 				break;
834 
835 			case KERN_PROC_TTY:
836 				if ((p->p_flag & P_CONTROLT) == 0 ||
837 				    p->p_session == NULL ||
838 				    p->p_session->s_ttyp == NULL ||
839 				    dev2udev(p->p_session->s_ttyp->t_dev) !=
840 					(udev_t)name[0])
841 					continue;
842 				break;
843 
844 			case KERN_PROC_UID:
845 				if (p->p_ucred == NULL ||
846 				    p->p_ucred->cr_uid != (uid_t)name[0])
847 					continue;
848 				break;
849 
850 			case KERN_PROC_RUID:
851 				if (p->p_ucred == NULL ||
852 				    p->p_ucred->cr_ruid != (uid_t)name[0])
853 					continue;
854 				break;
855 			}
856 
857 			if (!PRISON_CHECK(cr1, p->p_ucred))
858 				continue;
859 			PHOLD(p);
860 			error = sysctl_out_proc(p, req, flags);
861 			PRELE(p);
862 			if (error)
863 				goto post_threads;
864 		}
865 	}
866 
867 	/*
868 	 * Iterate over all active cpus and scan their thread list.  Start
869 	 * with the next logical cpu and end with our original cpu.  We
870 	 * migrate our own thread to each target cpu in order to safely scan
871 	 * its thread list.  In the last loop we migrate back to our original
872 	 * cpu.
873 	 */
874 	origcpu = mycpu->gd_cpuid;
875 	if (!ps_showallthreads || jailed(cr1))
876 		goto post_threads;
877 
878 	for (n = 1; n <= ncpus; ++n) {
879 		globaldata_t rgd;
880 		int nid;
881 
882 		nid = (origcpu + n) % ncpus;
883 		if ((smp_active_mask & (1 << nid)) == 0)
884 			continue;
885 		rgd = globaldata_find(nid);
886 		lwkt_setcpu_self(rgd);
887 
888 		TAILQ_FOREACH(td, &mycpu->gd_tdallq, td_allq) {
889 			if (td->td_proc)
890 				continue;
891 			switch (oid) {
892 			case KERN_PROC_PGRP:
893 			case KERN_PROC_TTY:
894 			case KERN_PROC_UID:
895 			case KERN_PROC_RUID:
896 				continue;
897 			default:
898 				break;
899 			}
900 			lwkt_hold(td);
901 			error = sysctl_out_proc_kthread(td, req, doingzomb);
902 			lwkt_rele(td);
903 			if (error)
904 				goto post_threads;
905 		}
906 	}
907 post_threads:
908 	lwkt_reltoken(&proc_token);
909 	return (error);
910 }
911 
912 /*
913  * This sysctl allows a process to retrieve the argument list or process
914  * title for another process without groping around in the address space
915  * of the other process.  It also allow a process to set its own "process
916  * title to a string of its own choice.
917  *
918  * No requirements.
919  */
920 static int
921 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
922 {
923 	int *name = (int*) arg1;
924 	u_int namelen = arg2;
925 	struct proc *p;
926 	struct pargs *pa;
927 	int error = 0;
928 	struct ucred *cr1 = curproc->p_ucred;
929 
930 	if (namelen != 1)
931 		return (EINVAL);
932 
933 	lwkt_gettoken(&proc_token);
934 	p = pfind((pid_t)name[0]);
935 	if (p == NULL)
936 		goto done;
937 
938 	if ((!ps_argsopen) && p_trespass(cr1, p->p_ucred))
939 		goto done;
940 
941 	if (req->newptr && curproc != p) {
942 		error = EPERM;
943 		goto done;
944 	}
945 
946 	PHOLD(p);
947 	if (req->oldptr && p->p_args != NULL) {
948 		error = SYSCTL_OUT(req, p->p_args->ar_args,
949 				   p->p_args->ar_length);
950 	}
951 	if (req->newptr == NULL) {
952 		PRELE(p);
953 		goto done;
954 	}
955 
956 	if (p->p_args && --p->p_args->ar_ref == 0)
957 		FREE(p->p_args, M_PARGS);
958 	p->p_args = NULL;
959 
960 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit) {
961 		PRELE(p);
962 		goto done;
963 	}
964 
965 	MALLOC(pa, struct pargs *, sizeof(struct pargs) + req->newlen,
966 	       M_PARGS, M_WAITOK);
967 	pa->ar_ref = 1;
968 	pa->ar_length = req->newlen;
969 	error = SYSCTL_IN(req, pa->ar_args, req->newlen);
970 	if (!error)
971 		p->p_args = pa;
972 	else
973 		FREE(pa, M_PARGS);
974 	PRELE(p);
975 done:
976 	lwkt_reltoken(&proc_token);
977 	return (error);
978 }
979 
980 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
981 
982 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT,
983 	0, 0, sysctl_kern_proc, "S,proc", "Return entire process table");
984 
985 SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD,
986 	sysctl_kern_proc, "Process table");
987 
988 SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD,
989 	sysctl_kern_proc, "Process table");
990 
991 SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD,
992 	sysctl_kern_proc, "Process table");
993 
994 SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD,
995 	sysctl_kern_proc, "Process table");
996 
997 SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD,
998 	sysctl_kern_proc, "Process table");
999 
1000 SYSCTL_NODE(_kern_proc, (KERN_PROC_ALL | KERN_PROC_FLAG_LWP), all_lwp, CTLFLAG_RD,
1001 	sysctl_kern_proc, "Process table");
1002 
1003 SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_FLAG_LWP), pgrp_lwp, CTLFLAG_RD,
1004 	sysctl_kern_proc, "Process table");
1005 
1006 SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_FLAG_LWP), tty_lwp, CTLFLAG_RD,
1007 	sysctl_kern_proc, "Process table");
1008 
1009 SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_FLAG_LWP), uid_lwp, CTLFLAG_RD,
1010 	sysctl_kern_proc, "Process table");
1011 
1012 SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_FLAG_LWP), ruid_lwp, CTLFLAG_RD,
1013 	sysctl_kern_proc, "Process table");
1014 
1015 SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_FLAG_LWP), pid_lwp, CTLFLAG_RD,
1016 	sysctl_kern_proc, "Process table");
1017 
1018 SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, CTLFLAG_RW | CTLFLAG_ANYBODY,
1019 	sysctl_kern_proc_args, "Process argument list");
1020