xref: /freebsd/sys/kern/uipc_domain.c (revision 716fd348e01c5f2ba125f878a634a753436c2994)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
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	8.2 (Berkeley) 10/18/93
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include <sys/param.h>
38 #include <sys/socket.h>
39 #include <sys/protosw.h>
40 #include <sys/domain.h>
41 #include <sys/eventhandler.h>
42 #include <sys/epoch.h>
43 #include <sys/mbuf.h>
44 #include <sys/kernel.h>
45 #include <sys/lock.h>
46 #include <sys/mutex.h>
47 #include <sys/rmlock.h>
48 #include <sys/socketvar.h>
49 #include <sys/systm.h>
50 
51 #include <machine/atomic.h>
52 
53 #include <net/vnet.h>
54 
55 /*
56  * System initialization
57  *
58  * Note: domain initialization takes place on a per domain basis
59  * as a result of traversing a SYSINIT linker set.  Most likely,
60  * each domain would want to call DOMAIN_SET(9) itself, which
61  * would cause the domain to be added just after domaininit()
62  * is called during startup.
63  *
64  * See DOMAIN_SET(9) for details on its use.
65  */
66 
67 static void domaininit(void *);
68 SYSINIT(domain, SI_SUB_PROTO_DOMAININIT, SI_ORDER_ANY, domaininit, NULL);
69 
70 static void domainfinalize(void *);
71 SYSINIT(domainfin, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST, domainfinalize,
72     NULL);
73 
74 static struct callout pffast_callout;
75 static struct callout pfslow_callout;
76 
77 static void	pffasttimo(void *);
78 static void	pfslowtimo(void *);
79 
80 static struct rmlock pftimo_lock;
81 RM_SYSINIT(pftimo_lock, &pftimo_lock, "pftimo");
82 
83 static LIST_HEAD(, protosw) pffast_list =
84     LIST_HEAD_INITIALIZER(pffast_list);
85 static LIST_HEAD(, protosw) pfslow_list =
86     LIST_HEAD_INITIALIZER(pfslow_list);
87 
88 struct domain *domains;		/* registered protocol domains */
89 int domain_init_status = 0;
90 static struct mtx dom_mtx;		/* domain list lock */
91 MTX_SYSINIT(domain, &dom_mtx, "domain list", MTX_DEF);
92 
93 /*
94  * Dummy protocol specific user requests function pointer array.
95  * All functions return EOPNOTSUPP.
96  */
97 struct pr_usrreqs nousrreqs = {
98 	.pru_accept =		pru_accept_notsupp,
99 	.pru_attach =		pru_attach_notsupp,
100 	.pru_bind =		pru_bind_notsupp,
101 	.pru_connect =		pru_connect_notsupp,
102 	.pru_connect2 =		pru_connect2_notsupp,
103 	.pru_control =		pru_control_notsupp,
104 	.pru_disconnect	=	pru_disconnect_notsupp,
105 	.pru_listen =		pru_listen_notsupp,
106 	.pru_peeraddr =		pru_peeraddr_notsupp,
107 	.pru_rcvd =		pru_rcvd_notsupp,
108 	.pru_rcvoob =		pru_rcvoob_notsupp,
109 	.pru_send =		pru_send_notsupp,
110 	.pru_sense =		pru_sense_null,
111 	.pru_shutdown =		pru_shutdown_notsupp,
112 	.pru_sockaddr =		pru_sockaddr_notsupp,
113 	.pru_sosend =		pru_sosend_notsupp,
114 	.pru_soreceive =	pru_soreceive_notsupp,
115 	.pru_sopoll =		pru_sopoll_notsupp,
116 };
117 
118 static void
119 pr_usrreqs_init(struct protosw *pr)
120 {
121 	struct pr_usrreqs *pu;
122 
123 	pu = pr->pr_usrreqs;
124 	KASSERT(pu != NULL, ("%s: %ssw[%d] has no usrreqs!", __func__,
125 	    pr->pr_domain->dom_name,
126 	    (int)(pr - pr->pr_domain->dom_protosw)));
127 
128 	/*
129 	 * Protocol switch methods fall into three categories: mandatory,
130 	 * mandatory but protosw_init() provides a default, and optional.
131 	 *
132 	 * For true protocols (i.e., pru_attach != NULL), KASSERT truly
133 	 * mandatory methods with no defaults, and initialize defaults for
134 	 * other mandatory methods if the protocol hasn't defined an
135 	 * implementation (NULL function pointer).
136 	 */
137 #if 0
138 	if (pu->pru_attach != NULL) {
139 		KASSERT(pu->pru_abort != NULL,
140 		    ("protosw_init: %ssw[%d] pru_abort NULL",
141 		    pr->pr_domain->dom_name,
142 		    (int)(pr - pr->pr_domain->dom_protosw)));
143 		KASSERT(pu->pru_send != NULL,
144 		    ("protosw_init: %ssw[%d] pru_send NULL",
145 		    pr->pr_domain->dom_name,
146 		    (int)(pr - pr->pr_domain->dom_protosw)));
147 	}
148 #endif
149 
150 #define DEFAULT(foo, bar)	if ((foo) == NULL)  (foo) = (bar)
151 	DEFAULT(pu->pru_accept, pru_accept_notsupp);
152 	DEFAULT(pu->pru_aio_queue, pru_aio_queue_notsupp);
153 	DEFAULT(pu->pru_bind, pru_bind_notsupp);
154 	DEFAULT(pu->pru_bindat, pru_bindat_notsupp);
155 	DEFAULT(pu->pru_connect, pru_connect_notsupp);
156 	DEFAULT(pu->pru_connect2, pru_connect2_notsupp);
157 	DEFAULT(pu->pru_connectat, pru_connectat_notsupp);
158 	DEFAULT(pu->pru_control, pru_control_notsupp);
159 	DEFAULT(pu->pru_disconnect, pru_disconnect_notsupp);
160 	DEFAULT(pu->pru_listen, pru_listen_notsupp);
161 	DEFAULT(pu->pru_peeraddr, pru_peeraddr_notsupp);
162 	DEFAULT(pu->pru_rcvd, pru_rcvd_notsupp);
163 	DEFAULT(pu->pru_rcvoob, pru_rcvoob_notsupp);
164 	DEFAULT(pu->pru_sense, pru_sense_null);
165 	DEFAULT(pu->pru_shutdown, pru_shutdown_notsupp);
166 	DEFAULT(pu->pru_sockaddr, pru_sockaddr_notsupp);
167 	DEFAULT(pu->pru_sosend, sosend_generic);
168 	DEFAULT(pu->pru_soreceive, soreceive_generic);
169 	DEFAULT(pu->pru_sopoll, sopoll_generic);
170 	DEFAULT(pu->pru_ready, pru_ready_notsupp);
171 #undef DEFAULT
172 }
173 
174 /*
175  * Add a new protocol domain to the list of supported domains
176  * Note: you cant unload it again because a socket may be using it.
177  * XXX can't fail at this time.
178  */
179 void
180 domain_init(void *arg)
181 {
182 	struct domain *dp = arg;
183 	struct protosw *pr;
184 	int flags;
185 
186 	MPASS(IS_DEFAULT_VNET(curvnet));
187 
188 	flags = atomic_load_acq_int(&dp->dom_flags);
189 	if ((flags & DOMF_SUPPORTED) == 0)
190 		return;
191 	MPASS((flags & DOMF_INITED) == 0);
192 
193 	for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
194 		pr_usrreqs_init(pr);
195 		rm_wlock(&pftimo_lock);
196 		if (pr->pr_fasttimo != NULL)
197 			LIST_INSERT_HEAD(&pffast_list, pr, pr_fasttimos);
198 		if (pr->pr_slowtimo != NULL)
199 			LIST_INSERT_HEAD(&pfslow_list, pr, pr_slowtimos);
200 		rm_wunlock(&pftimo_lock);
201 	}
202 
203 	/*
204 	 * update global information about maximums
205 	 */
206 	max_hdr = max_linkhdr + max_protohdr;
207 	max_datalen = MHLEN - max_hdr;
208 	if (max_datalen < 1)
209 		panic("%s: max_datalen < 1", __func__);
210 	atomic_set_rel_int(&dp->dom_flags, DOMF_INITED);
211 }
212 
213 /*
214  * Add a new protocol domain to the list of supported domains
215  * Note: you cant unload it again because a socket may be using it.
216  * XXX can't fail at this time.
217  */
218 void
219 domain_add(void *data)
220 {
221 	struct domain *dp;
222 
223 	dp = (struct domain *)data;
224 	if (dp->dom_probe != NULL && (*dp->dom_probe)() != 0)
225 		return;
226 	atomic_set_rel_int(&dp->dom_flags, DOMF_SUPPORTED);
227 	mtx_lock(&dom_mtx);
228 	dp->dom_next = domains;
229 	domains = dp;
230 
231 	KASSERT(domain_init_status >= 1,
232 	    ("attempt to domain_add(%s) before domaininit()",
233 	    dp->dom_name));
234 #ifndef INVARIANTS
235 	if (domain_init_status < 1)
236 		printf("WARNING: attempt to domain_add(%s) before "
237 		    "domaininit()\n", dp->dom_name);
238 #endif
239 	mtx_unlock(&dom_mtx);
240 }
241 
242 /* ARGSUSED*/
243 static void
244 domaininit(void *dummy)
245 {
246 
247 	if (max_linkhdr < 16)		/* XXX */
248 		max_linkhdr = 16;
249 
250 	callout_init(&pffast_callout, 1);
251 	callout_init(&pfslow_callout, 1);
252 
253 	mtx_lock(&dom_mtx);
254 	KASSERT(domain_init_status == 0, ("domaininit called too late!"));
255 	domain_init_status = 1;
256 	mtx_unlock(&dom_mtx);
257 }
258 
259 /* ARGSUSED*/
260 static void
261 domainfinalize(void *dummy)
262 {
263 
264 	mtx_lock(&dom_mtx);
265 	KASSERT(domain_init_status == 1, ("domainfinalize called too late!"));
266 	domain_init_status = 2;
267 	mtx_unlock(&dom_mtx);
268 
269 	callout_reset(&pffast_callout, 1, pffasttimo, NULL);
270 	callout_reset(&pfslow_callout, 1, pfslowtimo, NULL);
271 }
272 
273 struct domain *
274 pffinddomain(int family)
275 {
276 	struct domain *dp;
277 
278 	for (dp = domains; dp != NULL; dp = dp->dom_next)
279 		if (dp->dom_family == family)
280 			return (dp);
281 	return (NULL);
282 }
283 
284 struct protosw *
285 pffindtype(int family, int type)
286 {
287 	struct domain *dp;
288 	struct protosw *pr;
289 
290 	dp = pffinddomain(family);
291 	if (dp == NULL)
292 		return (NULL);
293 
294 	for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++)
295 		if (pr->pr_type && pr->pr_type == type)
296 			return (pr);
297 	return (NULL);
298 }
299 
300 struct protosw *
301 pffindproto(int family, int protocol, int type)
302 {
303 	struct domain *dp;
304 	struct protosw *pr;
305 	struct protosw *maybe;
306 
307 	maybe = NULL;
308 	if (family == 0)
309 		return (NULL);
310 
311 	dp = pffinddomain(family);
312 	if (dp == NULL)
313 		return (NULL);
314 
315 	for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
316 		if ((pr->pr_protocol == protocol) && (pr->pr_type == type))
317 			return (pr);
318 
319 		if (type == SOCK_RAW && pr->pr_type == SOCK_RAW &&
320 		    pr->pr_protocol == 0 && maybe == NULL)
321 			maybe = pr;
322 	}
323 	return (maybe);
324 }
325 
326 /*
327  * The caller must make sure that the new protocol is fully set up and ready to
328  * accept requests before it is registered.
329  */
330 int
331 pf_proto_register(int family, struct protosw *npr)
332 {
333 	struct domain *dp;
334 	struct protosw *pr, *fpr;
335 
336 	/* Sanity checks. */
337 	if (family == 0)
338 		return (EPFNOSUPPORT);
339 	if (npr->pr_type == 0)
340 		return (EPROTOTYPE);
341 	if (npr->pr_protocol == 0)
342 		return (EPROTONOSUPPORT);
343 	if (npr->pr_usrreqs == NULL)
344 		return (ENXIO);
345 
346 	/* Try to find the specified domain based on the family. */
347 	dp = pffinddomain(family);
348 	if (dp == NULL)
349 		return (EPFNOSUPPORT);
350 
351 	/* Initialize backpointer to struct domain. */
352 	npr->pr_domain = dp;
353 	fpr = NULL;
354 
355 	/*
356 	 * Protect us against races when two protocol registrations for
357 	 * the same protocol happen at the same time.
358 	 */
359 	mtx_lock(&dom_mtx);
360 
361 	/* The new protocol must not yet exist. */
362 	for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
363 		if ((pr->pr_type == npr->pr_type) &&
364 		    (pr->pr_protocol == npr->pr_protocol)) {
365 			mtx_unlock(&dom_mtx);
366 			return (EEXIST);	/* XXX: Check only protocol? */
367 		}
368 		/* While here, remember the first free spacer. */
369 		if ((fpr == NULL) && (pr->pr_protocol == PROTO_SPACER))
370 			fpr = pr;
371 	}
372 
373 	/* If no free spacer is found we can't add the new protocol. */
374 	if (fpr == NULL) {
375 		mtx_unlock(&dom_mtx);
376 		return (ENOMEM);
377 	}
378 
379 	/* Copy the new struct protosw over the spacer. */
380 	bcopy(npr, fpr, sizeof(*fpr));
381 
382 	pr_usrreqs_init(fpr);
383 	rm_wlock(&pftimo_lock);
384 	if (fpr->pr_fasttimo != NULL)
385 		LIST_INSERT_HEAD(&pffast_list, fpr, pr_fasttimos);
386 	if (fpr->pr_slowtimo != NULL)
387 		LIST_INSERT_HEAD(&pfslow_list, fpr, pr_slowtimos);
388 	rm_wunlock(&pftimo_lock);
389 
390 	/* Job is done, no more protection required. */
391 	mtx_unlock(&dom_mtx);
392 
393 	return (0);
394 }
395 
396 /*
397  * The caller must make sure the protocol and its functions correctly shut down
398  * all sockets and release all locks and memory references.
399  */
400 int
401 pf_proto_unregister(int family, int protocol, int type)
402 {
403 	struct domain *dp;
404 	struct protosw *pr, *dpr;
405 
406 	/* Sanity checks. */
407 	if (family == 0)
408 		return (EPFNOSUPPORT);
409 	if (protocol == 0)
410 		return (EPROTONOSUPPORT);
411 	if (type == 0)
412 		return (EPROTOTYPE);
413 
414 	/* Try to find the specified domain based on the family type. */
415 	dp = pffinddomain(family);
416 	if (dp == NULL)
417 		return (EPFNOSUPPORT);
418 
419 	dpr = NULL;
420 
421 	/* Lock out everyone else while we are manipulating the protosw. */
422 	mtx_lock(&dom_mtx);
423 
424 	/* The protocol must exist and only once. */
425 	for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
426 		if ((pr->pr_type == type) && (pr->pr_protocol == protocol)) {
427 			if (dpr != NULL) {
428 				mtx_unlock(&dom_mtx);
429 				return (EMLINK);   /* Should not happen! */
430 			} else
431 				dpr = pr;
432 		}
433 	}
434 
435 	/* Protocol does not exist. */
436 	if (dpr == NULL) {
437 		mtx_unlock(&dom_mtx);
438 		return (EPROTONOSUPPORT);
439 	}
440 
441 	rm_wlock(&pftimo_lock);
442 	if (dpr->pr_fasttimo != NULL)
443 		LIST_REMOVE(dpr, pr_fasttimos);
444 	if (dpr->pr_slowtimo != NULL)
445 		LIST_REMOVE(dpr, pr_slowtimos);
446 	rm_wunlock(&pftimo_lock);
447 
448 	/* De-orbit the protocol and make the slot available again. */
449 	dpr->pr_type = 0;
450 	dpr->pr_domain = dp;
451 	dpr->pr_protocol = PROTO_SPACER;
452 	dpr->pr_flags = 0;
453 	dpr->pr_input = NULL;
454 	dpr->pr_output = NULL;
455 	dpr->pr_ctlinput = NULL;
456 	dpr->pr_ctloutput = NULL;
457 	dpr->pr_fasttimo = NULL;
458 	dpr->pr_slowtimo = NULL;
459 	dpr->pr_drain = NULL;
460 	dpr->pr_usrreqs = &nousrreqs;
461 
462 	/* Job is done, not more protection required. */
463 	mtx_unlock(&dom_mtx);
464 
465 	return (0);
466 }
467 
468 void
469 pfctlinput(int cmd, struct sockaddr *sa)
470 {
471 	struct domain *dp;
472 	struct protosw *pr;
473 
474 	NET_EPOCH_ASSERT();
475 
476 	for (dp = domains; dp; dp = dp->dom_next)
477 		for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++)
478 			if (pr->pr_ctlinput)
479 				(*pr->pr_ctlinput)(cmd, sa, (void *)0);
480 }
481 
482 static void
483 pfslowtimo(void *arg)
484 {
485 	struct rm_priotracker tracker;
486 	struct epoch_tracker et;
487 	struct protosw *pr;
488 
489 	rm_rlock(&pftimo_lock, &tracker);
490 	NET_EPOCH_ENTER(et);
491 	LIST_FOREACH(pr, &pfslow_list, pr_slowtimos) {
492 		(*pr->pr_slowtimo)();
493 	}
494 	NET_EPOCH_EXIT(et);
495 	rm_runlock(&pftimo_lock, &tracker);
496 	callout_reset(&pfslow_callout, hz/2, pfslowtimo, NULL);
497 }
498 
499 static void
500 pffasttimo(void *arg)
501 {
502 	struct rm_priotracker tracker;
503 	struct epoch_tracker et;
504 	struct protosw *pr;
505 
506 	rm_rlock(&pftimo_lock, &tracker);
507 	NET_EPOCH_ENTER(et);
508 	LIST_FOREACH(pr, &pffast_list, pr_fasttimos) {
509 		(*pr->pr_fasttimo)();
510 	}
511 	NET_EPOCH_EXIT(et);
512 	rm_runlock(&pftimo_lock, &tracker);
513 	callout_reset(&pffast_callout, hz/5, pffasttimo, NULL);
514 }
515