xref: /netbsd-src/sys/kern/uipc_domain.c (revision 267197ec1eebfcb9810ea27a89625b6ddf68e3e7)
1 /*	$NetBSD: uipc_domain.c,v 1.74 2008/01/07 16:12:54 ad Exp $	*/
2 
3 /*
4  * Copyright (c) 1982, 1986, 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. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)uipc_domain.c	8.3 (Berkeley) 2/14/95
32  */
33 
34 #include <sys/cdefs.h>
35 __KERNEL_RCSID(0, "$NetBSD: uipc_domain.c,v 1.74 2008/01/07 16:12:54 ad Exp $");
36 
37 #include <sys/param.h>
38 #include <sys/socket.h>
39 #include <sys/socketvar.h>
40 #include <sys/protosw.h>
41 #include <sys/domain.h>
42 #include <sys/mbuf.h>
43 #include <sys/time.h>
44 #include <sys/kernel.h>
45 #include <sys/systm.h>
46 #include <sys/callout.h>
47 #include <sys/queue.h>
48 #include <sys/proc.h>
49 #include <sys/sysctl.h>
50 #include <sys/un.h>
51 #include <sys/unpcb.h>
52 #include <sys/file.h>
53 #include <sys/filedesc.h>
54 #include <sys/kauth.h>
55 
56 MALLOC_DECLARE(M_SOCKADDR);
57 
58 MALLOC_DEFINE(M_SOCKADDR, "sockaddr", "socket endpoints");
59 
60 void	pffasttimo(void *);
61 void	pfslowtimo(void *);
62 
63 struct domainhead domains = STAILQ_HEAD_INITIALIZER(domains);
64 static struct domain *domain_array[AF_MAX];
65 
66 callout_t pffasttimo_ch, pfslowtimo_ch;
67 
68 /*
69  * Current time values for fast and slow timeouts.  We can use u_int
70  * relatively safely.  The fast timer will roll over in 27 years and
71  * the slow timer in 68 years.
72  */
73 u_int	pfslowtimo_now;
74 u_int	pffasttimo_now;
75 
76 void
77 domaininit(void)
78 {
79 	__link_set_decl(domains, struct domain);
80 	struct domain * const * dpp;
81 	struct domain *rt_domain = NULL;
82 
83 	/*
84 	 * Add all of the domains.  Make sure the PF_ROUTE
85 	 * domain is added last.
86 	 */
87 	__link_set_foreach(dpp, domains) {
88 		if ((*dpp)->dom_family == PF_ROUTE)
89 			rt_domain = *dpp;
90 		else
91 			domain_attach(*dpp);
92 	}
93 	if (rt_domain)
94 		domain_attach(rt_domain);
95 
96 	callout_init(&pffasttimo_ch, 0);
97 	callout_init(&pfslowtimo_ch, 0);
98 
99 	callout_reset(&pffasttimo_ch, 1, pffasttimo, NULL);
100 	callout_reset(&pfslowtimo_ch, 1, pfslowtimo, NULL);
101 }
102 
103 void
104 domain_attach(struct domain *dp)
105 {
106 	const struct protosw *pr;
107 
108 	STAILQ_INSERT_TAIL(&domains, dp, dom_link);
109 	if (dp->dom_family < __arraycount(domain_array))
110 		domain_array[dp->dom_family] = dp;
111 
112 	if (dp->dom_init)
113 		(*dp->dom_init)();
114 
115 #ifdef MBUFTRACE
116 	if (dp->dom_mowner.mo_name[0] == '\0') {
117 		strncpy(dp->dom_mowner.mo_name, dp->dom_name,
118 		    sizeof(dp->dom_mowner.mo_name));
119 		MOWNER_ATTACH(&dp->dom_mowner);
120 	}
121 #endif
122 	for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
123 		if (pr->pr_init)
124 			(*pr->pr_init)();
125 	}
126 
127 	if (max_linkhdr < 16)		/* XXX */
128 		max_linkhdr = 16;
129 	max_hdr = max_linkhdr + max_protohdr;
130 	max_datalen = MHLEN - max_hdr;
131 }
132 
133 struct domain *
134 pffinddomain(int family)
135 {
136 	struct domain *dp;
137 
138 	if (family < __arraycount(domain_array) && domain_array[family] != NULL)
139 		return domain_array[family];
140 
141 	DOMAIN_FOREACH(dp)
142 		if (dp->dom_family == family)
143 			return (dp);
144 	return (NULL);
145 }
146 
147 const struct protosw *
148 pffindtype(int family, int type)
149 {
150 	struct domain *dp;
151 	const struct protosw *pr;
152 
153 	dp = pffinddomain(family);
154 	if (dp == NULL)
155 		return (NULL);
156 
157 	for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++)
158 		if (pr->pr_type && pr->pr_type == type)
159 			return (pr);
160 
161 	return (NULL);
162 }
163 
164 const struct protosw *
165 pffindproto(int family, int protocol, int type)
166 {
167 	struct domain *dp;
168 	const struct protosw *pr;
169 	const struct protosw *maybe = NULL;
170 
171 	if (family == 0)
172 		return (NULL);
173 
174 	dp = pffinddomain(family);
175 	if (dp == NULL)
176 		return (NULL);
177 
178 	for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
179 		if ((pr->pr_protocol == protocol) && (pr->pr_type == type))
180 			return (pr);
181 
182 		if (type == SOCK_RAW && pr->pr_type == SOCK_RAW &&
183 		    pr->pr_protocol == 0 && maybe == NULL)
184 			maybe = pr;
185 	}
186 	return (maybe);
187 }
188 
189 void *
190 sockaddr_addr(struct sockaddr *sa, socklen_t *slenp)
191 {
192 	const struct domain *dom;
193 
194 	if ((dom = pffinddomain(sa->sa_family)) == NULL ||
195 	    dom->dom_sockaddr_addr == NULL)
196 		return NULL;
197 
198 	return (*dom->dom_sockaddr_addr)(sa, slenp);
199 }
200 
201 const void *
202 sockaddr_const_addr(const struct sockaddr *sa, socklen_t *slenp)
203 {
204 	const struct domain *dom;
205 
206 	if ((dom = pffinddomain(sa->sa_family)) == NULL ||
207 	    dom->dom_sockaddr_const_addr == NULL)
208 		return NULL;
209 
210 	return (*dom->dom_sockaddr_const_addr)(sa, slenp);
211 }
212 
213 const struct sockaddr *
214 sockaddr_any(const struct sockaddr *sa)
215 {
216 	const struct domain *dom;
217 
218 	if ((dom = pffinddomain(sa->sa_family)) == NULL)
219 		return NULL;
220 
221 	return dom->dom_sa_any;
222 }
223 
224 const void *
225 sockaddr_anyaddr(const struct sockaddr *sa, socklen_t *slenp)
226 {
227 	const struct sockaddr *any;
228 
229 	if ((any = sockaddr_any(sa)) == NULL)
230 		return NULL;
231 
232 	return sockaddr_const_addr(any, slenp);
233 }
234 
235 struct sockaddr *
236 sockaddr_alloc(sa_family_t af, socklen_t socklen, int flags)
237 {
238 	struct sockaddr *sa;
239 	socklen_t reallen = MAX(socklen, offsetof(struct sockaddr, sa_data[0]));
240 
241 	if ((sa = malloc(reallen, M_SOCKADDR, flags)) == NULL)
242 		return NULL;
243 
244 	sa->sa_family = af;
245 	sa->sa_len = reallen;
246 	return sa;
247 }
248 
249 struct sockaddr *
250 sockaddr_copy(struct sockaddr *dst, socklen_t socklen,
251     const struct sockaddr *src)
252 {
253 	if (__predict_false(socklen < src->sa_len)) {
254 		panic("%s: source too long, %d < %d bytes", __func__, socklen,
255 		    src->sa_len);
256 	}
257 	return memcpy(dst, src, src->sa_len);
258 }
259 
260 int
261 sockaddr_cmp(const struct sockaddr *sa1, const struct sockaddr *sa2)
262 {
263 	int len, rc;
264 	struct domain *dom;
265 
266 	if (sa1->sa_family != sa2->sa_family)
267 		return sa1->sa_family - sa2->sa_family;
268 
269 	dom = pffinddomain(sa1->sa_family);
270 
271 	if (dom != NULL && dom->dom_sockaddr_cmp != NULL)
272 		return (*dom->dom_sockaddr_cmp)(sa1, sa2);
273 
274 	len = MIN(sa1->sa_len, sa2->sa_len);
275 
276 	if (dom == NULL || dom->dom_sa_cmplen == 0) {
277 		if ((rc = memcmp(sa1, sa2, len)) != 0)
278 			return rc;
279 		return sa1->sa_len - sa2->sa_len;
280 	}
281 
282 	if ((rc = memcmp((const char *)sa1 + dom->dom_sa_cmpofs,
283 		         (const char *)sa2 + dom->dom_sa_cmpofs,
284 			 MIN(dom->dom_sa_cmplen,
285 			     len - MIN(len, dom->dom_sa_cmpofs)))) != 0)
286 		return rc;
287 
288 	return MIN(dom->dom_sa_cmplen + dom->dom_sa_cmpofs, sa1->sa_len) -
289 	       MIN(dom->dom_sa_cmplen + dom->dom_sa_cmpofs, sa2->sa_len);
290 }
291 
292 struct sockaddr *
293 sockaddr_dup(const struct sockaddr *src, int flags)
294 {
295 	struct sockaddr *dst;
296 
297 	if ((dst = sockaddr_alloc(src->sa_family, src->sa_len, flags)) == NULL)
298 		return NULL;
299 
300 	return sockaddr_copy(dst, dst->sa_len, src);
301 }
302 
303 void
304 sockaddr_free(struct sockaddr *sa)
305 {
306 	free(sa, M_SOCKADDR);
307 }
308 
309 /*
310  * sysctl helper to stuff PF_LOCAL pcbs into sysctl structures
311  */
312 static void
313 sysctl_dounpcb(struct kinfo_pcb *pcb, const struct socket *so)
314 {
315 	struct unpcb *unp = sotounpcb(so);
316 	struct sockaddr_un *un = unp->unp_addr;
317 
318 	memset(pcb, 0, sizeof(*pcb));
319 
320 	pcb->ki_family = so->so_proto->pr_domain->dom_family;
321 	pcb->ki_type = so->so_proto->pr_type;
322 	pcb->ki_protocol = so->so_proto->pr_protocol;
323 	pcb->ki_pflags = unp->unp_flags;
324 
325 	pcb->ki_pcbaddr = PTRTOUINT64(unp);
326 	/* pcb->ki_ppcbaddr = unp has no ppcb... */
327 	pcb->ki_sockaddr = PTRTOUINT64(so);
328 
329 	pcb->ki_sostate = so->so_state;
330 	/* pcb->ki_prstate = unp has no state... */
331 
332 	pcb->ki_rcvq = so->so_rcv.sb_cc;
333 	pcb->ki_sndq = so->so_snd.sb_cc;
334 
335 	un = (struct sockaddr_un *)&pcb->ki_src;
336 	/*
337 	 * local domain sockets may bind without having a local
338 	 * endpoint.  bleah!
339 	 */
340 	if (unp->unp_addr != NULL) {
341 		un->sun_len = unp->unp_addr->sun_len;
342 		un->sun_family = unp->unp_addr->sun_family;
343 		strlcpy(un->sun_path, unp->unp_addr->sun_path,
344 		    sizeof(pcb->ki_s));
345 	}
346 	else {
347 		un->sun_len = offsetof(struct sockaddr_un, sun_path);
348 		un->sun_family = pcb->ki_family;
349 	}
350 	if (unp->unp_conn != NULL) {
351 		un = (struct sockaddr_un *)&pcb->ki_dst;
352 		if (unp->unp_conn->unp_addr != NULL) {
353 			un->sun_len = unp->unp_conn->unp_addr->sun_len;
354 			un->sun_family = unp->unp_conn->unp_addr->sun_family;
355 			un->sun_family = unp->unp_conn->unp_addr->sun_family;
356 			strlcpy(un->sun_path, unp->unp_conn->unp_addr->sun_path,
357 				sizeof(pcb->ki_d));
358 		}
359 		else {
360 			un->sun_len = offsetof(struct sockaddr_un, sun_path);
361 			un->sun_family = pcb->ki_family;
362 		}
363 	}
364 
365 	pcb->ki_inode = unp->unp_ino;
366 	pcb->ki_vnode = PTRTOUINT64(unp->unp_vnode);
367 	pcb->ki_conn = PTRTOUINT64(unp->unp_conn);
368 	pcb->ki_refs = PTRTOUINT64(unp->unp_refs);
369 	pcb->ki_nextref = PTRTOUINT64(unp->unp_nextref);
370 }
371 
372 static int
373 sysctl_unpcblist(SYSCTLFN_ARGS)
374 {
375 	struct file *fp, *dfp, *np;
376 	struct socket *so;
377 	struct kinfo_pcb pcb;
378 	char *dp;
379 	u_int op, arg;
380 	size_t len, needed, elem_size, out_size;
381 	int error, elem_count, pf, type, pf2;
382 
383 	if (namelen == 1 && name[0] == CTL_QUERY)
384 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
385 
386 	if (namelen != 4)
387 		return (EINVAL);
388 
389 	if (oldp != NULL) {
390 		len = *oldlenp;
391 		elem_size = name[2];
392 		elem_count = name[3];
393 		if (elem_size != sizeof(pcb))
394 			return EINVAL;
395 	} else {
396 		len = 0;
397 		elem_size = sizeof(pcb);
398 		elem_count = INT_MAX;
399 	}
400 	error = 0;
401 	dp = oldp;
402 	op = name[0];
403 	arg = name[1];
404 	out_size = elem_size;
405 	needed = 0;
406 
407 	if (name - oname != 4)
408 		return (EINVAL);
409 
410 	pf = oname[1];
411 	type = oname[2];
412 	pf2 = (oldp == NULL) ? 0 : pf;
413 
414 	/*
415 	 * allocate dummy file descriptor to make position in list.
416 	 */
417 	sysctl_unlock();
418 	if ((dfp = fgetdummy()) == NULL) {
419 	 	sysctl_relock();
420 		return ENOMEM;
421 	}
422 
423 	/*
424 	 * there's no "list" of local domain sockets, so we have
425 	 * to walk the file list looking for them.  :-/
426 	 */
427 	mutex_enter(&filelist_lock);
428 	LIST_FOREACH(fp, &filehead, f_list) {
429 	    	np = LIST_NEXT(fp, f_list);
430 		if (fp->f_count == 0 || fp->f_type != DTYPE_SOCKET ||
431 		    fp->f_data == NULL)
432 			continue;
433 		if (kauth_authorize_generic(l->l_cred,
434 		    KAUTH_GENERIC_CANSEE, fp->f_cred) != 0)
435 			continue;
436 		so = (struct socket *)fp->f_data;
437 		if (so->so_type != type)
438 			continue;
439 		if (so->so_proto->pr_domain->dom_family != pf)
440 			continue;
441 		if (len >= elem_size && elem_count > 0) {
442 			FILE_LOCK(fp);
443 			FILE_USE(fp);
444 			LIST_INSERT_AFTER(fp, dfp, f_list);
445 			mutex_exit(&filelist_lock);
446 			sysctl_dounpcb(&pcb, so);
447 			error = copyout(&pcb, dp, out_size);
448 			FILE_UNUSE(fp, NULL);
449 			mutex_enter(&filelist_lock);
450 			np = LIST_NEXT(dfp, f_list);
451 			LIST_REMOVE(dfp, f_list);
452 			if (error)
453 				break;
454 			dp += elem_size;
455 			len -= elem_size;
456 		}
457 		if (elem_count > 0) {
458 			needed += elem_size;
459 			if (elem_count != INT_MAX)
460 				elem_count--;
461 		}
462 	}
463 	mutex_exit(&filelist_lock);
464 	fputdummy(dfp);
465  	*oldlenp = needed;
466 	if (oldp == NULL)
467 		*oldlenp += PCB_SLOP * sizeof(struct kinfo_pcb);
468  	sysctl_relock();
469 
470 	return (error);
471 }
472 
473 SYSCTL_SETUP(sysctl_net_setup, "sysctl net subtree setup")
474 {
475 	sysctl_createv(clog, 0, NULL, NULL,
476 		       CTLFLAG_PERMANENT,
477 		       CTLTYPE_NODE, "net", NULL,
478 		       NULL, 0, NULL, 0,
479 		       CTL_NET, CTL_EOL);
480 	sysctl_createv(clog, 0, NULL, NULL,
481 		       CTLFLAG_PERMANENT,
482 		       CTLTYPE_NODE, "local",
483 		       SYSCTL_DESCR("PF_LOCAL related settings"),
484 		       NULL, 0, NULL, 0,
485 		       CTL_NET, PF_LOCAL, CTL_EOL);
486 	sysctl_createv(clog, 0, NULL, NULL,
487 		       CTLFLAG_PERMANENT,
488 		       CTLTYPE_NODE, "stream",
489 		       SYSCTL_DESCR("SOCK_STREAM settings"),
490 		       NULL, 0, NULL, 0,
491 		       CTL_NET, PF_LOCAL, SOCK_STREAM, CTL_EOL);
492 	sysctl_createv(clog, 0, NULL, NULL,
493 		       CTLFLAG_PERMANENT,
494 		       CTLTYPE_NODE, "dgram",
495 		       SYSCTL_DESCR("SOCK_DGRAM settings"),
496 		       NULL, 0, NULL, 0,
497 		       CTL_NET, PF_LOCAL, SOCK_DGRAM, CTL_EOL);
498 
499 	sysctl_createv(clog, 0, NULL, NULL,
500 		       CTLFLAG_PERMANENT,
501 		       CTLTYPE_STRUCT, "pcblist",
502 		       SYSCTL_DESCR("SOCK_STREAM protocol control block list"),
503 		       sysctl_unpcblist, 0, NULL, 0,
504 		       CTL_NET, PF_LOCAL, SOCK_STREAM, CTL_CREATE, CTL_EOL);
505 	sysctl_createv(clog, 0, NULL, NULL,
506 		       CTLFLAG_PERMANENT,
507 		       CTLTYPE_STRUCT, "pcblist",
508 		       SYSCTL_DESCR("SOCK_DGRAM protocol control block list"),
509 		       sysctl_unpcblist, 0, NULL, 0,
510 		       CTL_NET, PF_LOCAL, SOCK_DGRAM, CTL_CREATE, CTL_EOL);
511 }
512 
513 void
514 pfctlinput(int cmd, const struct sockaddr *sa)
515 {
516 	struct domain *dp;
517 	const struct protosw *pr;
518 
519 	DOMAIN_FOREACH(dp) {
520 		for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
521 			if (pr->pr_ctlinput != NULL)
522 				(*pr->pr_ctlinput)(cmd, sa, NULL);
523 		}
524 	}
525 }
526 
527 void
528 pfctlinput2(int cmd, const struct sockaddr *sa, void *ctlparam)
529 {
530 	struct domain *dp;
531 	const struct protosw *pr;
532 
533 	if (sa == NULL)
534 		return;
535 
536 	DOMAIN_FOREACH(dp) {
537 		/*
538 		 * the check must be made by xx_ctlinput() anyways, to
539 		 * make sure we use data item pointed to by ctlparam in
540 		 * correct way.  the following check is made just for safety.
541 		 */
542 		if (dp->dom_family != sa->sa_family)
543 			continue;
544 
545 		for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
546 			if (pr->pr_ctlinput != NULL)
547 				(*pr->pr_ctlinput)(cmd, sa, ctlparam);
548 		}
549 	}
550 }
551 
552 void
553 pfslowtimo(void *arg)
554 {
555 	struct domain *dp;
556 	const struct protosw *pr;
557 
558 	pfslowtimo_now++;
559 
560 	DOMAIN_FOREACH(dp) {
561 		for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++)
562 			if (pr->pr_slowtimo)
563 				(*pr->pr_slowtimo)();
564 	}
565 	callout_reset(&pfslowtimo_ch, hz / 2, pfslowtimo, NULL);
566 }
567 
568 void
569 pffasttimo(void *arg)
570 {
571 	struct domain *dp;
572 	const struct protosw *pr;
573 
574 	pffasttimo_now++;
575 
576 	DOMAIN_FOREACH(dp) {
577 		for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++)
578 			if (pr->pr_fasttimo)
579 				(*pr->pr_fasttimo)();
580 	}
581 	callout_reset(&pffasttimo_ch, hz / 5, pffasttimo, NULL);
582 }
583