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 <net/vnet.h>
52
53 /*
54 * System initialization
55 *
56 * Note: domain initialization takes place on a per domain basis
57 * as a result of traversing a SYSINIT linker set. Most likely,
58 * each domain would want to call DOMAIN_SET(9) itself, which
59 * would cause the domain to be added just after domaininit()
60 * is called during startup.
61 *
62 * See DOMAIN_SET(9) for details on its use.
63 */
64
65 static void domaininit(void *);
66 SYSINIT(domain, SI_SUB_PROTO_DOMAININIT, SI_ORDER_ANY, domaininit, NULL);
67
68 static void domainfinalize(void *);
69 SYSINIT(domainfin, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST, domainfinalize,
70 NULL);
71
72 static struct callout pffast_callout;
73 static struct callout pfslow_callout;
74
75 static void pffasttimo(void *);
76 static void pfslowtimo(void *);
77
78 static struct rmlock pftimo_lock;
79 RM_SYSINIT(pftimo_lock, &pftimo_lock, "pftimo");
80
81 static LIST_HEAD(, protosw) pffast_list =
82 LIST_HEAD_INITIALIZER(pffast_list);
83 static LIST_HEAD(, protosw) pfslow_list =
84 LIST_HEAD_INITIALIZER(pfslow_list);
85
86 struct domain *domains; /* registered protocol domains */
87 int domain_init_status = 0;
88 static struct mtx dom_mtx; /* domain list lock */
89 MTX_SYSINIT(domain, &dom_mtx, "domain list", MTX_DEF);
90
91 /*
92 * Dummy protocol specific user requests function pointer array.
93 * All functions return EOPNOTSUPP.
94 */
95 struct pr_usrreqs nousrreqs = {
96 .pru_accept = pru_accept_notsupp,
97 .pru_attach = pru_attach_notsupp,
98 .pru_bind = pru_bind_notsupp,
99 .pru_connect = pru_connect_notsupp,
100 .pru_connect2 = pru_connect2_notsupp,
101 .pru_control = pru_control_notsupp,
102 .pru_disconnect = pru_disconnect_notsupp,
103 .pru_listen = pru_listen_notsupp,
104 .pru_peeraddr = pru_peeraddr_notsupp,
105 .pru_rcvd = pru_rcvd_notsupp,
106 .pru_rcvoob = pru_rcvoob_notsupp,
107 .pru_send = pru_send_notsupp,
108 .pru_sense = pru_sense_null,
109 .pru_shutdown = pru_shutdown_notsupp,
110 .pru_sockaddr = pru_sockaddr_notsupp,
111 .pru_sosend = pru_sosend_notsupp,
112 .pru_soreceive = pru_soreceive_notsupp,
113 .pru_sopoll = pru_sopoll_notsupp,
114 };
115
116 static void
117 protosw_init(struct protosw *pr)
118 {
119 struct pr_usrreqs *pu;
120
121 pu = pr->pr_usrreqs;
122 KASSERT(pu != NULL, ("protosw_init: %ssw[%d] has no usrreqs!",
123 pr->pr_domain->dom_name,
124 (int)(pr - pr->pr_domain->dom_protosw)));
125
126 /*
127 * Protocol switch methods fall into three categories: mandatory,
128 * mandatory but protosw_init() provides a default, and optional.
129 *
130 * For true protocols (i.e., pru_attach != NULL), KASSERT truly
131 * mandatory methods with no defaults, and initialize defaults for
132 * other mandatory methods if the protocol hasn't defined an
133 * implementation (NULL function pointer).
134 */
135 #if 0
136 if (pu->pru_attach != NULL) {
137 KASSERT(pu->pru_abort != NULL,
138 ("protosw_init: %ssw[%d] pru_abort NULL",
139 pr->pr_domain->dom_name,
140 (int)(pr - pr->pr_domain->dom_protosw)));
141 KASSERT(pu->pru_send != NULL,
142 ("protosw_init: %ssw[%d] pru_send NULL",
143 pr->pr_domain->dom_name,
144 (int)(pr - pr->pr_domain->dom_protosw)));
145 }
146 #endif
147
148 #define DEFAULT(foo, bar) if ((foo) == NULL) (foo) = (bar)
149 DEFAULT(pu->pru_accept, pru_accept_notsupp);
150 DEFAULT(pu->pru_aio_queue, pru_aio_queue_notsupp);
151 DEFAULT(pu->pru_bind, pru_bind_notsupp);
152 DEFAULT(pu->pru_bindat, pru_bindat_notsupp);
153 DEFAULT(pu->pru_connect, pru_connect_notsupp);
154 DEFAULT(pu->pru_connect2, pru_connect2_notsupp);
155 DEFAULT(pu->pru_connectat, pru_connectat_notsupp);
156 DEFAULT(pu->pru_control, pru_control_notsupp);
157 DEFAULT(pu->pru_disconnect, pru_disconnect_notsupp);
158 DEFAULT(pu->pru_listen, pru_listen_notsupp);
159 DEFAULT(pu->pru_peeraddr, pru_peeraddr_notsupp);
160 DEFAULT(pu->pru_rcvd, pru_rcvd_notsupp);
161 DEFAULT(pu->pru_rcvoob, pru_rcvoob_notsupp);
162 DEFAULT(pu->pru_sense, pru_sense_null);
163 DEFAULT(pu->pru_shutdown, pru_shutdown_notsupp);
164 DEFAULT(pu->pru_sockaddr, pru_sockaddr_notsupp);
165 DEFAULT(pu->pru_sosend, sosend_generic);
166 DEFAULT(pu->pru_soreceive, soreceive_generic);
167 DEFAULT(pu->pru_sopoll, sopoll_generic);
168 DEFAULT(pu->pru_ready, pru_ready_notsupp);
169 #undef DEFAULT
170 if (pr->pr_init)
171 (*pr->pr_init)();
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
185 if (dp->dom_init)
186 (*dp->dom_init)();
187 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
188 protosw_init(pr);
189
190 /*
191 * Note that with VIMAGE enabled, domain_init() will be
192 * re-invoked for each new vnet that's created. The below lists
193 * are intended to be system-wide, so avoid altering global
194 * state for non-default vnets.
195 */
196 if (IS_DEFAULT_VNET(curvnet)) {
197 rm_wlock(&pftimo_lock);
198 if (pr->pr_fasttimo != NULL)
199 LIST_INSERT_HEAD(&pffast_list, pr,
200 pr_fasttimos);
201 if (pr->pr_slowtimo != NULL)
202 LIST_INSERT_HEAD(&pfslow_list, pr,
203 pr_slowtimos);
204 rm_wunlock(&pftimo_lock);
205 }
206 }
207
208 /*
209 * update global information about maximums
210 */
211 max_hdr = max_linkhdr + max_protohdr;
212 max_datalen = MHLEN - max_hdr;
213 if (max_datalen < 1)
214 panic("%s: max_datalen < 1", __func__);
215 }
216
217 #ifdef VIMAGE
218 void
219 vnet_domain_init(void *arg)
220 {
221
222 /* Virtualized case is no different -- call init functions. */
223 domain_init(arg);
224 }
225
226 void
227 vnet_domain_uninit(void *arg)
228 {
229 struct domain *dp = arg;
230
231 if (dp->dom_destroy)
232 (*dp->dom_destroy)();
233 }
234 #endif
235
236 /*
237 * Add a new protocol domain to the list of supported domains
238 * Note: you cant unload it again because a socket may be using it.
239 * XXX can't fail at this time.
240 */
241 void
242 domain_add(void *data)
243 {
244 struct domain *dp;
245
246 dp = (struct domain *)data;
247 mtx_lock(&dom_mtx);
248 dp->dom_next = domains;
249 domains = dp;
250
251 KASSERT(domain_init_status >= 1,
252 ("attempt to domain_add(%s) before domaininit()",
253 dp->dom_name));
254 #ifndef INVARIANTS
255 if (domain_init_status < 1)
256 printf("WARNING: attempt to domain_add(%s) before "
257 "domaininit()\n", dp->dom_name);
258 #endif
259 #ifdef notyet
260 KASSERT(domain_init_status < 2,
261 ("attempt to domain_add(%s) after domainfinalize()",
262 dp->dom_name));
263 #else
264 if (domain_init_status >= 2)
265 printf("WARNING: attempt to domain_add(%s) after "
266 "domainfinalize()\n", dp->dom_name);
267 #endif
268 mtx_unlock(&dom_mtx);
269 }
270
271 /* ARGSUSED*/
272 static void
273 domaininit(void *dummy)
274 {
275
276 if (max_linkhdr < 16) /* XXX */
277 max_linkhdr = 16;
278
279 callout_init(&pffast_callout, 1);
280 callout_init(&pfslow_callout, 1);
281
282 mtx_lock(&dom_mtx);
283 KASSERT(domain_init_status == 0, ("domaininit called too late!"));
284 domain_init_status = 1;
285 mtx_unlock(&dom_mtx);
286 }
287
288 /* ARGSUSED*/
289 static void
290 domainfinalize(void *dummy)
291 {
292
293 mtx_lock(&dom_mtx);
294 KASSERT(domain_init_status == 1, ("domainfinalize called too late!"));
295 domain_init_status = 2;
296 mtx_unlock(&dom_mtx);
297
298 callout_reset(&pffast_callout, 1, pffasttimo, NULL);
299 callout_reset(&pfslow_callout, 1, pfslowtimo, NULL);
300 }
301
302 struct domain *
303 pffinddomain(int family)
304 {
305 struct domain *dp;
306
307 for (dp = domains; dp != NULL; dp = dp->dom_next)
308 if (dp->dom_family == family)
309 return (dp);
310 return (NULL);
311 }
312
313 struct protosw *
314 pffindtype(int family, int type)
315 {
316 struct domain *dp;
317 struct protosw *pr;
318
319 dp = pffinddomain(family);
320 if (dp == NULL)
321 return (NULL);
322
323 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++)
324 if (pr->pr_type && pr->pr_type == type)
325 return (pr);
326 return (NULL);
327 }
328
329 struct protosw *
330 pffindproto(int family, int protocol, int type)
331 {
332 struct domain *dp;
333 struct protosw *pr;
334 struct protosw *maybe;
335
336 maybe = NULL;
337 if (family == 0)
338 return (NULL);
339
340 dp = pffinddomain(family);
341 if (dp == NULL)
342 return (NULL);
343
344 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
345 if ((pr->pr_protocol == protocol) && (pr->pr_type == type))
346 return (pr);
347
348 if (type == SOCK_RAW && pr->pr_type == SOCK_RAW &&
349 pr->pr_protocol == 0 && maybe == NULL)
350 maybe = pr;
351 }
352 return (maybe);
353 }
354
355 /*
356 * The caller must make sure that the new protocol is fully set up and ready to
357 * accept requests before it is registered.
358 */
359 int
360 pf_proto_register(int family, struct protosw *npr)
361 {
362 VNET_ITERATOR_DECL(vnet_iter);
363 struct domain *dp;
364 struct protosw *pr, *fpr;
365
366 /* Sanity checks. */
367 if (family == 0)
368 return (EPFNOSUPPORT);
369 if (npr->pr_type == 0)
370 return (EPROTOTYPE);
371 if (npr->pr_protocol == 0)
372 return (EPROTONOSUPPORT);
373 if (npr->pr_usrreqs == NULL)
374 return (ENXIO);
375
376 /* Try to find the specified domain based on the family. */
377 dp = pffinddomain(family);
378 if (dp == NULL)
379 return (EPFNOSUPPORT);
380
381 /* Initialize backpointer to struct domain. */
382 npr->pr_domain = dp;
383 fpr = NULL;
384
385 /*
386 * Protect us against races when two protocol registrations for
387 * the same protocol happen at the same time.
388 */
389 mtx_lock(&dom_mtx);
390
391 /* The new protocol must not yet exist. */
392 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
393 if ((pr->pr_type == npr->pr_type) &&
394 (pr->pr_protocol == npr->pr_protocol)) {
395 mtx_unlock(&dom_mtx);
396 return (EEXIST); /* XXX: Check only protocol? */
397 }
398 /* While here, remember the first free spacer. */
399 if ((fpr == NULL) && (pr->pr_protocol == PROTO_SPACER))
400 fpr = pr;
401 }
402
403 /* If no free spacer is found we can't add the new protocol. */
404 if (fpr == NULL) {
405 mtx_unlock(&dom_mtx);
406 return (ENOMEM);
407 }
408
409 /* Copy the new struct protosw over the spacer. */
410 bcopy(npr, fpr, sizeof(*fpr));
411
412 rm_wlock(&pftimo_lock);
413 if (fpr->pr_fasttimo != NULL)
414 LIST_INSERT_HEAD(&pffast_list, fpr, pr_fasttimos);
415 if (fpr->pr_slowtimo != NULL)
416 LIST_INSERT_HEAD(&pfslow_list, fpr, pr_slowtimos);
417 rm_wunlock(&pftimo_lock);
418
419 /* Job is done, no more protection required. */
420 mtx_unlock(&dom_mtx);
421
422 /* Initialize and activate the protocol. */
423 VNET_LIST_RLOCK();
424 VNET_FOREACH(vnet_iter) {
425 CURVNET_SET_QUIET(vnet_iter);
426 protosw_init(fpr);
427 CURVNET_RESTORE();
428 }
429 VNET_LIST_RUNLOCK();
430
431 return (0);
432 }
433
434 /*
435 * The caller must make sure the protocol and its functions correctly shut down
436 * all sockets and release all locks and memory references.
437 */
438 int
439 pf_proto_unregister(int family, int protocol, int type)
440 {
441 struct domain *dp;
442 struct protosw *pr, *dpr;
443
444 /* Sanity checks. */
445 if (family == 0)
446 return (EPFNOSUPPORT);
447 if (protocol == 0)
448 return (EPROTONOSUPPORT);
449 if (type == 0)
450 return (EPROTOTYPE);
451
452 /* Try to find the specified domain based on the family type. */
453 dp = pffinddomain(family);
454 if (dp == NULL)
455 return (EPFNOSUPPORT);
456
457 dpr = NULL;
458
459 /* Lock out everyone else while we are manipulating the protosw. */
460 mtx_lock(&dom_mtx);
461
462 /* The protocol must exist and only once. */
463 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
464 if ((pr->pr_type == type) && (pr->pr_protocol == protocol)) {
465 if (dpr != NULL) {
466 mtx_unlock(&dom_mtx);
467 return (EMLINK); /* Should not happen! */
468 } else
469 dpr = pr;
470 }
471 }
472
473 /* Protocol does not exist. */
474 if (dpr == NULL) {
475 mtx_unlock(&dom_mtx);
476 return (EPROTONOSUPPORT);
477 }
478
479 rm_wlock(&pftimo_lock);
480 if (dpr->pr_fasttimo != NULL)
481 LIST_REMOVE(dpr, pr_fasttimos);
482 if (dpr->pr_slowtimo != NULL)
483 LIST_REMOVE(dpr, pr_slowtimos);
484 rm_wunlock(&pftimo_lock);
485
486 /* De-orbit the protocol and make the slot available again. */
487 dpr->pr_type = 0;
488 dpr->pr_domain = dp;
489 dpr->pr_protocol = PROTO_SPACER;
490 dpr->pr_flags = 0;
491 dpr->pr_input = NULL;
492 dpr->pr_output = NULL;
493 dpr->pr_ctlinput = NULL;
494 dpr->pr_ctloutput = NULL;
495 dpr->pr_init = NULL;
496 dpr->pr_fasttimo = NULL;
497 dpr->pr_slowtimo = NULL;
498 dpr->pr_drain = NULL;
499 dpr->pr_usrreqs = &nousrreqs;
500
501 /* Job is done, not more protection required. */
502 mtx_unlock(&dom_mtx);
503
504 return (0);
505 }
506
507 void
508 pfctlinput(int cmd, struct sockaddr *sa)
509 {
510 struct domain *dp;
511 struct protosw *pr;
512
513 NET_EPOCH_ASSERT();
514
515 for (dp = domains; dp; dp = dp->dom_next)
516 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++)
517 if (pr->pr_ctlinput)
518 (*pr->pr_ctlinput)(cmd, sa, (void *)0);
519 }
520
521 static void
522 pfslowtimo(void *arg)
523 {
524 struct rm_priotracker tracker;
525 struct epoch_tracker et;
526 struct protosw *pr;
527
528 rm_rlock(&pftimo_lock, &tracker);
529 NET_EPOCH_ENTER(et);
530 LIST_FOREACH(pr, &pfslow_list, pr_slowtimos) {
531 (*pr->pr_slowtimo)();
532 }
533 NET_EPOCH_EXIT(et);
534 rm_runlock(&pftimo_lock, &tracker);
535 callout_reset(&pfslow_callout, hz/2, pfslowtimo, NULL);
536 }
537
538 static void
539 pffasttimo(void *arg)
540 {
541 struct rm_priotracker tracker;
542 struct epoch_tracker et;
543 struct protosw *pr;
544
545 rm_rlock(&pftimo_lock, &tracker);
546 NET_EPOCH_ENTER(et);
547 LIST_FOREACH(pr, &pffast_list, pr_fasttimos) {
548 (*pr->pr_fasttimo)();
549 }
550 NET_EPOCH_EXIT(et);
551 rm_runlock(&pftimo_lock, &tracker);
552 callout_reset(&pffast_callout, hz/5, pffasttimo, NULL);
553 }
Cache object: e25ebf8becf79e570d376407c3fb4c27
|