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 protosw_init(struct protosw *pr) 120 { 121 struct pr_usrreqs *pu; 122 123 pu = pr->pr_usrreqs; 124 KASSERT(pu != NULL, ("protosw_init: %ssw[%d] has no usrreqs!", 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 if (pr->pr_init) 173 (*pr->pr_init)(); 174 } 175 176 /* 177 * Add a new protocol domain to the list of supported domains 178 * Note: you cant unload it again because a socket may be using it. 179 * XXX can't fail at this time. 180 */ 181 void 182 domain_init(void *arg) 183 { 184 struct domain *dp = arg; 185 struct protosw *pr; 186 int flags; 187 188 flags = atomic_load_acq_int(&dp->dom_flags); 189 if ((flags & DOMF_SUPPORTED) == 0) 190 return; 191 KASSERT((flags & DOMF_INITED) == 0 || !IS_DEFAULT_VNET(curvnet), 192 ("Premature initialization of domain in non-default vnet")); 193 if (dp->dom_init) 194 (*dp->dom_init)(); 195 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 196 protosw_init(pr); 197 198 /* 199 * Note that with VIMAGE enabled, domain_init() will be 200 * re-invoked for each new vnet that's created. The below lists 201 * are intended to be system-wide, so avoid altering global 202 * state for non-default vnets. 203 */ 204 if (IS_DEFAULT_VNET(curvnet)) { 205 rm_wlock(&pftimo_lock); 206 if (pr->pr_fasttimo != NULL) 207 LIST_INSERT_HEAD(&pffast_list, pr, 208 pr_fasttimos); 209 if (pr->pr_slowtimo != NULL) 210 LIST_INSERT_HEAD(&pfslow_list, pr, 211 pr_slowtimos); 212 rm_wunlock(&pftimo_lock); 213 } 214 } 215 216 /* 217 * update global information about maximums 218 */ 219 max_hdr = max_linkhdr + max_protohdr; 220 max_datalen = MHLEN - max_hdr; 221 if (max_datalen < 1) 222 panic("%s: max_datalen < 1", __func__); 223 if (IS_DEFAULT_VNET(curvnet)) 224 atomic_set_rel_int(&dp->dom_flags, DOMF_INITED); 225 } 226 227 #ifdef VIMAGE 228 void 229 vnet_domain_init(void *arg) 230 { 231 232 /* Virtualized case is no different -- call init functions. */ 233 domain_init(arg); 234 } 235 236 void 237 vnet_domain_uninit(void *arg) 238 { 239 struct domain *dp = arg; 240 241 if ((atomic_load_acq_int(&dp->dom_flags) & DOMF_SUPPORTED) == 0) 242 return; 243 if (dp->dom_destroy) 244 (*dp->dom_destroy)(); 245 } 246 #endif 247 248 /* 249 * Add a new protocol domain to the list of supported domains 250 * Note: you cant unload it again because a socket may be using it. 251 * XXX can't fail at this time. 252 */ 253 void 254 domain_add(void *data) 255 { 256 struct domain *dp; 257 258 dp = (struct domain *)data; 259 if (dp->dom_probe != NULL && (*dp->dom_probe)() != 0) 260 return; 261 atomic_set_rel_int(&dp->dom_flags, DOMF_SUPPORTED); 262 mtx_lock(&dom_mtx); 263 dp->dom_next = domains; 264 domains = dp; 265 266 KASSERT(domain_init_status >= 1, 267 ("attempt to domain_add(%s) before domaininit()", 268 dp->dom_name)); 269 #ifndef INVARIANTS 270 if (domain_init_status < 1) 271 printf("WARNING: attempt to domain_add(%s) before " 272 "domaininit()\n", dp->dom_name); 273 #endif 274 mtx_unlock(&dom_mtx); 275 } 276 277 /* ARGSUSED*/ 278 static void 279 domaininit(void *dummy) 280 { 281 282 if (max_linkhdr < 16) /* XXX */ 283 max_linkhdr = 16; 284 285 callout_init(&pffast_callout, 1); 286 callout_init(&pfslow_callout, 1); 287 288 mtx_lock(&dom_mtx); 289 KASSERT(domain_init_status == 0, ("domaininit called too late!")); 290 domain_init_status = 1; 291 mtx_unlock(&dom_mtx); 292 } 293 294 /* ARGSUSED*/ 295 static void 296 domainfinalize(void *dummy) 297 { 298 299 mtx_lock(&dom_mtx); 300 KASSERT(domain_init_status == 1, ("domainfinalize called too late!")); 301 domain_init_status = 2; 302 mtx_unlock(&dom_mtx); 303 304 callout_reset(&pffast_callout, 1, pffasttimo, NULL); 305 callout_reset(&pfslow_callout, 1, pfslowtimo, NULL); 306 } 307 308 struct domain * 309 pffinddomain(int family) 310 { 311 struct domain *dp; 312 313 for (dp = domains; dp != NULL; dp = dp->dom_next) 314 if (dp->dom_family == family) 315 return (dp); 316 return (NULL); 317 } 318 319 struct protosw * 320 pffindtype(int family, int type) 321 { 322 struct domain *dp; 323 struct protosw *pr; 324 325 dp = pffinddomain(family); 326 if (dp == NULL) 327 return (NULL); 328 329 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) 330 if (pr->pr_type && pr->pr_type == type) 331 return (pr); 332 return (NULL); 333 } 334 335 struct protosw * 336 pffindproto(int family, int protocol, int type) 337 { 338 struct domain *dp; 339 struct protosw *pr; 340 struct protosw *maybe; 341 342 maybe = NULL; 343 if (family == 0) 344 return (NULL); 345 346 dp = pffinddomain(family); 347 if (dp == NULL) 348 return (NULL); 349 350 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 351 if ((pr->pr_protocol == protocol) && (pr->pr_type == type)) 352 return (pr); 353 354 if (type == SOCK_RAW && pr->pr_type == SOCK_RAW && 355 pr->pr_protocol == 0 && maybe == NULL) 356 maybe = pr; 357 } 358 return (maybe); 359 } 360 361 /* 362 * The caller must make sure that the new protocol is fully set up and ready to 363 * accept requests before it is registered. 364 */ 365 int 366 pf_proto_register(int family, struct protosw *npr) 367 { 368 VNET_ITERATOR_DECL(vnet_iter); 369 struct domain *dp; 370 struct protosw *pr, *fpr; 371 372 /* Sanity checks. */ 373 if (family == 0) 374 return (EPFNOSUPPORT); 375 if (npr->pr_type == 0) 376 return (EPROTOTYPE); 377 if (npr->pr_protocol == 0) 378 return (EPROTONOSUPPORT); 379 if (npr->pr_usrreqs == NULL) 380 return (ENXIO); 381 382 /* Try to find the specified domain based on the family. */ 383 dp = pffinddomain(family); 384 if (dp == NULL) 385 return (EPFNOSUPPORT); 386 387 /* Initialize backpointer to struct domain. */ 388 npr->pr_domain = dp; 389 fpr = NULL; 390 391 /* 392 * Protect us against races when two protocol registrations for 393 * the same protocol happen at the same time. 394 */ 395 mtx_lock(&dom_mtx); 396 397 /* The new protocol must not yet exist. */ 398 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 399 if ((pr->pr_type == npr->pr_type) && 400 (pr->pr_protocol == npr->pr_protocol)) { 401 mtx_unlock(&dom_mtx); 402 return (EEXIST); /* XXX: Check only protocol? */ 403 } 404 /* While here, remember the first free spacer. */ 405 if ((fpr == NULL) && (pr->pr_protocol == PROTO_SPACER)) 406 fpr = pr; 407 } 408 409 /* If no free spacer is found we can't add the new protocol. */ 410 if (fpr == NULL) { 411 mtx_unlock(&dom_mtx); 412 return (ENOMEM); 413 } 414 415 /* Copy the new struct protosw over the spacer. */ 416 bcopy(npr, fpr, sizeof(*fpr)); 417 418 rm_wlock(&pftimo_lock); 419 if (fpr->pr_fasttimo != NULL) 420 LIST_INSERT_HEAD(&pffast_list, fpr, pr_fasttimos); 421 if (fpr->pr_slowtimo != NULL) 422 LIST_INSERT_HEAD(&pfslow_list, fpr, pr_slowtimos); 423 rm_wunlock(&pftimo_lock); 424 425 /* Job is done, no more protection required. */ 426 mtx_unlock(&dom_mtx); 427 428 /* Initialize and activate the protocol. */ 429 VNET_LIST_RLOCK(); 430 VNET_FOREACH(vnet_iter) { 431 CURVNET_SET_QUIET(vnet_iter); 432 protosw_init(fpr); 433 CURVNET_RESTORE(); 434 } 435 VNET_LIST_RUNLOCK(); 436 437 return (0); 438 } 439 440 /* 441 * The caller must make sure the protocol and its functions correctly shut down 442 * all sockets and release all locks and memory references. 443 */ 444 int 445 pf_proto_unregister(int family, int protocol, int type) 446 { 447 struct domain *dp; 448 struct protosw *pr, *dpr; 449 450 /* Sanity checks. */ 451 if (family == 0) 452 return (EPFNOSUPPORT); 453 if (protocol == 0) 454 return (EPROTONOSUPPORT); 455 if (type == 0) 456 return (EPROTOTYPE); 457 458 /* Try to find the specified domain based on the family type. */ 459 dp = pffinddomain(family); 460 if (dp == NULL) 461 return (EPFNOSUPPORT); 462 463 dpr = NULL; 464 465 /* Lock out everyone else while we are manipulating the protosw. */ 466 mtx_lock(&dom_mtx); 467 468 /* The protocol must exist and only once. */ 469 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 470 if ((pr->pr_type == type) && (pr->pr_protocol == protocol)) { 471 if (dpr != NULL) { 472 mtx_unlock(&dom_mtx); 473 return (EMLINK); /* Should not happen! */ 474 } else 475 dpr = pr; 476 } 477 } 478 479 /* Protocol does not exist. */ 480 if (dpr == NULL) { 481 mtx_unlock(&dom_mtx); 482 return (EPROTONOSUPPORT); 483 } 484 485 rm_wlock(&pftimo_lock); 486 if (dpr->pr_fasttimo != NULL) 487 LIST_REMOVE(dpr, pr_fasttimos); 488 if (dpr->pr_slowtimo != NULL) 489 LIST_REMOVE(dpr, pr_slowtimos); 490 rm_wunlock(&pftimo_lock); 491 492 /* De-orbit the protocol and make the slot available again. */ 493 dpr->pr_type = 0; 494 dpr->pr_domain = dp; 495 dpr->pr_protocol = PROTO_SPACER; 496 dpr->pr_flags = 0; 497 dpr->pr_input = NULL; 498 dpr->pr_output = NULL; 499 dpr->pr_ctlinput = NULL; 500 dpr->pr_ctloutput = NULL; 501 dpr->pr_init = NULL; 502 dpr->pr_fasttimo = NULL; 503 dpr->pr_slowtimo = NULL; 504 dpr->pr_drain = NULL; 505 dpr->pr_usrreqs = &nousrreqs; 506 507 /* Job is done, not more protection required. */ 508 mtx_unlock(&dom_mtx); 509 510 return (0); 511 } 512 513 void 514 pfctlinput(int cmd, struct sockaddr *sa) 515 { 516 struct domain *dp; 517 struct protosw *pr; 518 519 NET_EPOCH_ASSERT(); 520 521 for (dp = domains; dp; dp = dp->dom_next) 522 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) 523 if (pr->pr_ctlinput) 524 (*pr->pr_ctlinput)(cmd, sa, (void *)0); 525 } 526 527 static void 528 pfslowtimo(void *arg) 529 { 530 struct rm_priotracker tracker; 531 struct epoch_tracker et; 532 struct protosw *pr; 533 534 rm_rlock(&pftimo_lock, &tracker); 535 NET_EPOCH_ENTER(et); 536 LIST_FOREACH(pr, &pfslow_list, pr_slowtimos) { 537 (*pr->pr_slowtimo)(); 538 } 539 NET_EPOCH_EXIT(et); 540 rm_runlock(&pftimo_lock, &tracker); 541 callout_reset(&pfslow_callout, hz/2, pfslowtimo, NULL); 542 } 543 544 static void 545 pffasttimo(void *arg) 546 { 547 struct rm_priotracker tracker; 548 struct epoch_tracker et; 549 struct protosw *pr; 550 551 rm_rlock(&pftimo_lock, &tracker); 552 NET_EPOCH_ENTER(et); 553 LIST_FOREACH(pr, &pffast_list, pr_fasttimos) { 554 (*pr->pr_fasttimo)(); 555 } 556 NET_EPOCH_EXIT(et); 557 rm_runlock(&pftimo_lock, &tracker); 558 callout_reset(&pffast_callout, hz/5, pffasttimo, NULL); 559 } 560