xref: /netbsd-src/sys/kern/kern_proc.c (revision 481fca6e59249d8ffcf24fef7cfbe7b131bfb080)
1 /*	$NetBSD: kern_proc.c,v 1.41 2000/05/27 00:40:45 sommerfeld Exp $	*/
2 
3 /*-
4  * Copyright (c) 1999 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
9  * NASA Ames Research Center.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the NetBSD
22  *	Foundation, Inc. and its contributors.
23  * 4. Neither the name of The NetBSD Foundation nor the names of its
24  *    contributors may be used to endorse or promote products derived
25  *    from this software without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*
41  * Copyright (c) 1982, 1986, 1989, 1991, 1993
42  *	The Regents of the University of California.  All rights reserved.
43  *
44  * Redistribution and use in source and binary forms, with or without
45  * modification, are permitted provided that the following conditions
46  * are met:
47  * 1. Redistributions of source code must retain the above copyright
48  *    notice, this list of conditions and the following disclaimer.
49  * 2. Redistributions in binary form must reproduce the above copyright
50  *    notice, this list of conditions and the following disclaimer in the
51  *    documentation and/or other materials provided with the distribution.
52  * 3. All advertising materials mentioning features or use of this software
53  *    must display the following acknowledgement:
54  *	This product includes software developed by the University of
55  *	California, Berkeley and its contributors.
56  * 4. Neither the name of the University nor the names of its contributors
57  *    may be used to endorse or promote products derived from this software
58  *    without specific prior written permission.
59  *
60  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
61  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
62  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
63  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
64  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
65  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
66  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
67  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
68  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
69  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
70  * SUCH DAMAGE.
71  *
72  *	@(#)kern_proc.c	8.7 (Berkeley) 2/14/95
73  */
74 
75 #include <sys/param.h>
76 #include <sys/systm.h>
77 #include <sys/map.h>
78 #include <sys/kernel.h>
79 #include <sys/proc.h>
80 #include <sys/resourcevar.h>
81 #include <sys/buf.h>
82 #include <sys/acct.h>
83 #include <sys/wait.h>
84 #include <sys/file.h>
85 #include <ufs/ufs/quota.h>
86 #include <sys/uio.h>
87 #include <sys/malloc.h>
88 #include <sys/pool.h>
89 #include <sys/mbuf.h>
90 #include <sys/ioctl.h>
91 #include <sys/tty.h>
92 #include <sys/signalvar.h>
93 
94 /*
95  * Structure associated with user cacheing.
96  */
97 struct uidinfo {
98 	LIST_ENTRY(uidinfo) ui_hash;
99 	uid_t	ui_uid;
100 	long	ui_proccnt;
101 };
102 #define	UIHASH(uid)	(&uihashtbl[(uid) & uihash])
103 LIST_HEAD(uihashhead, uidinfo) *uihashtbl;
104 u_long uihash;		/* size of hash table - 1 */
105 
106 /*
107  * Other process lists
108  */
109 struct pidhashhead *pidhashtbl;
110 u_long pidhash;
111 struct pgrphashhead *pgrphashtbl;
112 u_long pgrphash;
113 
114 struct proclist allproc;
115 struct proclist zombproc;	/* resources have been freed */
116 
117 /*
118  * Process list locking:
119  *
120  * We have two types of locks on the proclists: read locks and write
121  * locks.  Read locks can be used in interrupt context, so while we
122  * hold the write lock, we must also block clock interrupts to
123  * lock out any scheduling changes that may happen in interrupt
124  * context.
125  *
126  * The proclist lock locks the following structures:
127  *
128  *	allproc
129  *	zombproc
130  *	pidhashtbl
131  */
132 struct lock proclist_lock;
133 
134 /*
135  * Locking of this proclist is special; it's accessed in a
136  * critical section of process exit, and thus locking it can't
137  * modify interrupt state.  We use a simple spin lock for this
138  * proclist.  Processes on this proclist are also on zombproc;
139  * we use the p_hash member to linkup to deadproc.
140  */
141 struct simplelock deadproc_slock;
142 struct proclist deadproc;	/* dead, but not yet undead */
143 
144 struct pool proc_pool;
145 struct pool pcred_pool;
146 struct pool plimit_pool;
147 struct pool pgrp_pool;
148 struct pool rusage_pool;
149 
150 /*
151  * The process list descriptors, used during pid allocation and
152  * by sysctl.  No locking on this data structure is needed since
153  * it is completely static.
154  */
155 const struct proclist_desc proclists[] = {
156 	{ &allproc	},
157 	{ &zombproc	},
158 	{ NULL		},
159 };
160 
161 static void orphanpg __P((struct pgrp *));
162 #ifdef DEBUG
163 void pgrpdump __P((void));
164 #endif
165 
166 /*
167  * Initialize global process hashing structures.
168  */
169 void
170 procinit()
171 {
172 	const struct proclist_desc *pd;
173 
174 	for (pd = proclists; pd->pd_list != NULL; pd++)
175 		LIST_INIT(pd->pd_list);
176 
177 	spinlockinit(&proclist_lock, "proclk", 0);
178 
179 	LIST_INIT(&deadproc);
180 	simple_lock_init(&deadproc_slock);
181 
182 	pidhashtbl = hashinit(maxproc / 4, M_PROC, M_WAITOK, &pidhash);
183 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, M_WAITOK, &pgrphash);
184 	uihashtbl = hashinit(maxproc / 16, M_PROC, M_WAITOK, &uihash);
185 
186 	pool_init(&proc_pool, sizeof(struct proc), 0, 0, 0, "procpl",
187 	    0, pool_page_alloc_nointr, pool_page_free_nointr, M_PROC);
188 	pool_init(&pgrp_pool, sizeof(struct pgrp), 0, 0, 0, "pgrppl",
189 	    0, pool_page_alloc_nointr, pool_page_free_nointr, M_PGRP);
190 	pool_init(&pcred_pool, sizeof(struct pcred), 0, 0, 0, "pcredpl",
191 	    0, pool_page_alloc_nointr, pool_page_free_nointr, M_SUBPROC);
192 	pool_init(&plimit_pool, sizeof(struct plimit), 0, 0, 0, "plimitpl",
193 	    0, pool_page_alloc_nointr, pool_page_free_nointr, M_SUBPROC);
194 	pool_init(&rusage_pool, sizeof(struct rusage), 0, 0, 0, "rusgepl",
195 	    0, pool_page_alloc_nointr, pool_page_free_nointr, M_ZOMBIE);
196 }
197 
198 /*
199  * Acquire a read lock on the proclist.
200  */
201 void
202 proclist_lock_read()
203 {
204 	int error, s;
205 
206 	s = splclock();
207 	error = spinlockmgr(&proclist_lock, LK_SHARED, NULL);
208 #ifdef DIAGNOSTIC
209 	if (__predict_false(error != 0))
210 		panic("proclist_lock_read: failed to acquire lock");
211 #endif
212 	splx(s);
213 }
214 
215 /*
216  * Release a read lock on the proclist.
217  */
218 void
219 proclist_unlock_read()
220 {
221 	int s;
222 
223 	s = splclock();
224 	(void) spinlockmgr(&proclist_lock, LK_RELEASE, NULL);
225 	splx(s);
226 }
227 
228 /*
229  * Acquire a write lock on the proclist.
230  */
231 int
232 proclist_lock_write()
233 {
234 	int error, s;
235 
236 	s = splclock();
237 	error = spinlockmgr(&proclist_lock, LK_EXCLUSIVE, NULL);
238 #ifdef DIAGNOSTIC
239 	if (__predict_false(error != 0))
240 		panic("proclist_lock: failed to acquire lock");
241 #endif
242 	return (s);
243 }
244 
245 /*
246  * Release a write lock on the proclist.
247  */
248 void
249 proclist_unlock_write(s)
250 	int s;
251 {
252 
253 	(void) spinlockmgr(&proclist_lock, LK_RELEASE, NULL);
254 	splx(s);
255 }
256 
257 /*
258  * Change the count associated with number of processes
259  * a given user is using.
260  */
261 int
262 chgproccnt(uid, diff)
263 	uid_t	uid;
264 	int	diff;
265 {
266 	struct uidinfo *uip;
267 	struct uihashhead *uipp;
268 
269 	uipp = UIHASH(uid);
270 	for (uip = uipp->lh_first; uip != 0; uip = uip->ui_hash.le_next)
271 		if (uip->ui_uid == uid)
272 			break;
273 	if (uip) {
274 		uip->ui_proccnt += diff;
275 		if (uip->ui_proccnt > 0)
276 			return (uip->ui_proccnt);
277 		if (uip->ui_proccnt < 0)
278 			panic("chgproccnt: procs < 0");
279 		LIST_REMOVE(uip, ui_hash);
280 		FREE(uip, M_PROC);
281 		return (0);
282 	}
283 	if (diff <= 0) {
284 		if (diff == 0)
285 			return(0);
286 		panic("chgproccnt: lost user");
287 	}
288 	MALLOC(uip, struct uidinfo *, sizeof(*uip), M_PROC, M_WAITOK);
289 	LIST_INSERT_HEAD(uipp, uip, ui_hash);
290 	uip->ui_uid = uid;
291 	uip->ui_proccnt = diff;
292 	return (diff);
293 }
294 
295 /*
296  * Is p an inferior of q?
297  */
298 int
299 inferior(p, q)
300 	struct proc *p;
301 	struct proc *q;
302 {
303 
304 	for (; p != q; p = p->p_pptr)
305 		if (p->p_pid == 0)
306 			return (0);
307 	return (1);
308 }
309 
310 /*
311  * Locate a process by number
312  */
313 struct proc *
314 pfind(pid)
315 	pid_t pid;
316 {
317 	struct proc *p;
318 
319 	proclist_lock_read();
320 	for (p = PIDHASH(pid)->lh_first; p != 0; p = p->p_hash.le_next)
321 		if (p->p_pid == pid)
322 			goto out;
323  out:
324 	proclist_unlock_read();
325 	return (p);
326 }
327 
328 /*
329  * Locate a process group by number
330  */
331 struct pgrp *
332 pgfind(pgid)
333 	pid_t pgid;
334 {
335 	struct pgrp *pgrp;
336 
337 	for (pgrp = PGRPHASH(pgid)->lh_first; pgrp != 0; pgrp = pgrp->pg_hash.le_next)
338 		if (pgrp->pg_id == pgid)
339 			return (pgrp);
340 	return (NULL);
341 }
342 
343 /*
344  * Move p to a new or existing process group (and session)
345  */
346 int
347 enterpgrp(p, pgid, mksess)
348 	struct proc *p;
349 	pid_t pgid;
350 	int mksess;
351 {
352 	struct pgrp *pgrp = pgfind(pgid);
353 
354 #ifdef DIAGNOSTIC
355 	if (__predict_false(pgrp != NULL && mksess))	/* firewalls */
356 		panic("enterpgrp: setsid into non-empty pgrp");
357 	if (__predict_false(SESS_LEADER(p)))
358 		panic("enterpgrp: session leader attempted setpgrp");
359 #endif
360 	if (pgrp == NULL) {
361 		pid_t savepid = p->p_pid;
362 		struct proc *np;
363 		/*
364 		 * new process group
365 		 */
366 #ifdef DIAGNOSTIC
367 		if (__predict_false(p->p_pid != pgid))
368 			panic("enterpgrp: new pgrp and pid != pgid");
369 #endif
370 		pgrp = pool_get(&pgrp_pool, PR_WAITOK);
371 		if ((np = pfind(savepid)) == NULL || np != p)
372 			return (ESRCH);
373 		if (mksess) {
374 			struct session *sess;
375 
376 			/*
377 			 * new session
378 			 */
379 			MALLOC(sess, struct session *, sizeof(struct session),
380 			    M_SESSION, M_WAITOK);
381 			sess->s_sid = p->p_pid;
382 			sess->s_leader = p;
383 			sess->s_count = 1;
384 			sess->s_ttyvp = NULL;
385 			sess->s_ttyp = NULL;
386 			memcpy(sess->s_login, p->p_session->s_login,
387 			    sizeof(sess->s_login));
388 			p->p_flag &= ~P_CONTROLT;
389 			pgrp->pg_session = sess;
390 #ifdef DIAGNOSTIC
391 			if (__predict_false(p != curproc))
392 				panic("enterpgrp: mksession and p != curproc");
393 #endif
394 		} else {
395 			pgrp->pg_session = p->p_session;
396 			pgrp->pg_session->s_count++;
397 		}
398 		pgrp->pg_id = pgid;
399 		LIST_INIT(&pgrp->pg_members);
400 		LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
401 		pgrp->pg_jobc = 0;
402 	} else if (pgrp == p->p_pgrp)
403 		return (0);
404 
405 	/*
406 	 * Adjust eligibility of affected pgrps to participate in job control.
407 	 * Increment eligibility counts before decrementing, otherwise we
408 	 * could reach 0 spuriously during the first call.
409 	 */
410 	fixjobc(p, pgrp, 1);
411 	fixjobc(p, p->p_pgrp, 0);
412 
413 	LIST_REMOVE(p, p_pglist);
414 	if (p->p_pgrp->pg_members.lh_first == 0)
415 		pgdelete(p->p_pgrp);
416 	p->p_pgrp = pgrp;
417 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
418 	return (0);
419 }
420 
421 /*
422  * remove process from process group
423  */
424 int
425 leavepgrp(p)
426 	struct proc *p;
427 {
428 
429 	LIST_REMOVE(p, p_pglist);
430 	if (p->p_pgrp->pg_members.lh_first == 0)
431 		pgdelete(p->p_pgrp);
432 	p->p_pgrp = 0;
433 	return (0);
434 }
435 
436 /*
437  * delete a process group
438  */
439 void
440 pgdelete(pgrp)
441 	struct pgrp *pgrp;
442 {
443 
444 	if (pgrp->pg_session->s_ttyp != NULL &&
445 	    pgrp->pg_session->s_ttyp->t_pgrp == pgrp)
446 		pgrp->pg_session->s_ttyp->t_pgrp = NULL;
447 	LIST_REMOVE(pgrp, pg_hash);
448 	if (--pgrp->pg_session->s_count == 0)
449 		FREE(pgrp->pg_session, M_SESSION);
450 	pool_put(&pgrp_pool, pgrp);
451 }
452 
453 /*
454  * Adjust pgrp jobc counters when specified process changes process group.
455  * We count the number of processes in each process group that "qualify"
456  * the group for terminal job control (those with a parent in a different
457  * process group of the same session).  If that count reaches zero, the
458  * process group becomes orphaned.  Check both the specified process'
459  * process group and that of its children.
460  * entering == 0 => p is leaving specified group.
461  * entering == 1 => p is entering specified group.
462  */
463 void
464 fixjobc(p, pgrp, entering)
465 	struct proc *p;
466 	struct pgrp *pgrp;
467 	int entering;
468 {
469 	struct pgrp *hispgrp;
470 	struct session *mysession = pgrp->pg_session;
471 
472 	/*
473 	 * Check p's parent to see whether p qualifies its own process
474 	 * group; if so, adjust count for p's process group.
475 	 */
476 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
477 	    hispgrp->pg_session == mysession) {
478 		if (entering)
479 			pgrp->pg_jobc++;
480 		else if (--pgrp->pg_jobc == 0)
481 			orphanpg(pgrp);
482 	}
483 
484 	/*
485 	 * Check this process' children to see whether they qualify
486 	 * their process groups; if so, adjust counts for children's
487 	 * process groups.
488 	 */
489 	for (p = p->p_children.lh_first; p != 0; p = p->p_sibling.le_next) {
490 		if ((hispgrp = p->p_pgrp) != pgrp &&
491 		    hispgrp->pg_session == mysession &&
492 		    P_ZOMBIE(p) == 0) {
493 			if (entering)
494 				hispgrp->pg_jobc++;
495 			else if (--hispgrp->pg_jobc == 0)
496 				orphanpg(hispgrp);
497 		}
498 	}
499 }
500 
501 /*
502  * A process group has become orphaned;
503  * if there are any stopped processes in the group,
504  * hang-up all process in that group.
505  */
506 static void
507 orphanpg(pg)
508 	struct pgrp *pg;
509 {
510 	struct proc *p;
511 
512 	for (p = pg->pg_members.lh_first; p != 0; p = p->p_pglist.le_next) {
513 		if (p->p_stat == SSTOP) {
514 			for (p = pg->pg_members.lh_first; p != 0;
515 			    p = p->p_pglist.le_next) {
516 				psignal(p, SIGHUP);
517 				psignal(p, SIGCONT);
518 			}
519 			return;
520 		}
521 	}
522 }
523 
524 /* mark process as suid/sgid, reset some values do defaults */
525 void
526 p_sugid(p)
527 	struct proc *p;
528 {
529 	struct plimit *newlim;
530 
531 	p->p_flag |= P_SUGID;
532 	/* reset what needs to be reset in plimit */
533 	if (p->p_limit->pl_corename != defcorename) {
534 		if (p->p_limit->p_refcnt > 1 &&
535 		    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
536 			newlim = limcopy(p->p_limit);
537 			limfree(p->p_limit);
538 			p->p_limit = newlim;
539 		} else {
540 			free(p->p_limit->pl_corename, M_TEMP);
541 		}
542 		p->p_limit->pl_corename = defcorename;
543 	}
544 }
545 
546 
547 #ifdef DEBUG
548 void
549 pgrpdump()
550 {
551 	struct pgrp *pgrp;
552 	struct proc *p;
553 	int i;
554 
555 	for (i = 0; i <= pgrphash; i++) {
556 		if ((pgrp = pgrphashtbl[i].lh_first) != NULL) {
557 			printf("\tindx %d\n", i);
558 			for (; pgrp != 0; pgrp = pgrp->pg_hash.le_next) {
559 				printf("\tpgrp %p, pgid %d, sess %p, sesscnt %d, mem %p\n",
560 				    pgrp, pgrp->pg_id, pgrp->pg_session,
561 				    pgrp->pg_session->s_count,
562 				    pgrp->pg_members.lh_first);
563 				for (p = pgrp->pg_members.lh_first; p != 0;
564 				    p = p->p_pglist.le_next) {
565 					printf("\t\tpid %d addr %p pgrp %p\n",
566 					    p->p_pid, p, p->p_pgrp);
567 				}
568 			}
569 		}
570 	}
571 }
572 #endif /* DEBUG */
573