1 /*- 2 * Copyright (c) 1982, 1986, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)uipc_domain.c 8.2 (Berkeley) 10/18/93 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD$"); 34 35 #include <sys/param.h> 36 #include <sys/socket.h> 37 #include <sys/protosw.h> 38 #include <sys/domain.h> 39 #include <sys/eventhandler.h> 40 #include <sys/mbuf.h> 41 #include <sys/kernel.h> 42 #include <sys/lock.h> 43 #include <sys/mutex.h> 44 #include <sys/socketvar.h> 45 #include <sys/systm.h> 46 #include <sys/vimage.h> 47 #include <vm/uma.h> 48 49 /* 50 * System initialization 51 * 52 * Note: domain initialization takes place on a per domain basis 53 * as a result of traversing a SYSINIT linker set. Most likely, 54 * each domain would want to call DOMAIN_SET(9) itself, which 55 * would cause the domain to be added just after domaininit() 56 * is called during startup. 57 * 58 * See DOMAIN_SET(9) for details on its use. 59 */ 60 61 static void domaininit(void *); 62 SYSINIT(domain, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST, domaininit, NULL); 63 64 static void domainfinalize(void *); 65 SYSINIT(domainfin, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST, domainfinalize, 66 NULL); 67 68 static vnet_attach_fn net_init_domain; 69 #ifdef VIMAGE 70 static vnet_detach_fn net_detach_domain; 71 #endif 72 73 static struct callout pffast_callout; 74 static struct callout pfslow_callout; 75 76 static void pffasttimo(void *); 77 static void pfslowtimo(void *); 78 79 struct domain *domains; /* registered protocol domains */ 80 int domain_init_status = 0; 81 static struct mtx dom_mtx; /* domain list lock */ 82 MTX_SYSINIT(domain, &dom_mtx, "domain list", MTX_DEF); 83 84 /* 85 * Dummy protocol specific user requests function pointer array. 86 * All functions return EOPNOTSUPP. 87 */ 88 struct pr_usrreqs nousrreqs = { 89 .pru_accept = pru_accept_notsupp, 90 .pru_attach = pru_attach_notsupp, 91 .pru_bind = pru_bind_notsupp, 92 .pru_connect = pru_connect_notsupp, 93 .pru_connect2 = pru_connect2_notsupp, 94 .pru_control = pru_control_notsupp, 95 .pru_disconnect = pru_disconnect_notsupp, 96 .pru_listen = pru_listen_notsupp, 97 .pru_peeraddr = pru_peeraddr_notsupp, 98 .pru_rcvd = pru_rcvd_notsupp, 99 .pru_rcvoob = pru_rcvoob_notsupp, 100 .pru_send = pru_send_notsupp, 101 .pru_sense = pru_sense_null, 102 .pru_shutdown = pru_shutdown_notsupp, 103 .pru_sockaddr = pru_sockaddr_notsupp, 104 .pru_sosend = pru_sosend_notsupp, 105 .pru_soreceive = pru_soreceive_notsupp, 106 .pru_sopoll = pru_sopoll_notsupp, 107 }; 108 109 #ifndef VIMAGE_GLOBALS 110 vnet_modinfo_t vnet_domain_modinfo = { 111 .vmi_id = VNET_MOD_DOMAIN, 112 .vmi_name = "domain", 113 .vmi_iattach = net_init_domain, 114 #ifdef VIMAGE 115 .vmi_idetach = net_detach_domain, 116 #endif 117 }; 118 #endif 119 120 static void 121 protosw_init(struct protosw *pr) 122 { 123 struct pr_usrreqs *pu; 124 125 pu = pr->pr_usrreqs; 126 KASSERT(pu != NULL, ("protosw_init: %ssw[%d] has no usrreqs!", 127 pr->pr_domain->dom_name, 128 (int)(pr - pr->pr_domain->dom_protosw))); 129 130 /* 131 * Protocol switch methods fall into three categories: mandatory, 132 * mandatory but protosw_init() provides a default, and optional. 133 * 134 * For true protocols (i.e., pru_attach != NULL), KASSERT truly 135 * mandatory methods with no defaults, and initialize defaults for 136 * other mandatory methods if the protocol hasn't defined an 137 * implementation (NULL function pointer). 138 */ 139 #if 0 140 if (pu->pru_attach != NULL) { 141 KASSERT(pu->pru_abort != NULL, 142 ("protosw_init: %ssw[%d] pru_abort NULL", 143 pr->pr_domain->dom_name, 144 (int)(pr - pr->pr_domain->dom_protosw))); 145 KASSERT(pu->pru_send != NULL, 146 ("protosw_init: %ssw[%d] pru_send NULL", 147 pr->pr_domain->dom_name, 148 (int)(pr - pr->pr_domain->dom_protosw))); 149 } 150 #endif 151 152 #define DEFAULT(foo, bar) if ((foo) == NULL) (foo) = (bar) 153 DEFAULT(pu->pru_accept, pru_accept_notsupp); 154 DEFAULT(pu->pru_bind, pru_bind_notsupp); 155 DEFAULT(pu->pru_connect, pru_connect_notsupp); 156 DEFAULT(pu->pru_connect2, pru_connect2_notsupp); 157 DEFAULT(pu->pru_control, pru_control_notsupp); 158 DEFAULT(pu->pru_disconnect, pru_disconnect_notsupp); 159 DEFAULT(pu->pru_listen, pru_listen_notsupp); 160 DEFAULT(pu->pru_peeraddr, pru_peeraddr_notsupp); 161 DEFAULT(pu->pru_rcvd, pru_rcvd_notsupp); 162 DEFAULT(pu->pru_rcvoob, pru_rcvoob_notsupp); 163 DEFAULT(pu->pru_sense, pru_sense_null); 164 DEFAULT(pu->pru_shutdown, pru_shutdown_notsupp); 165 DEFAULT(pu->pru_sockaddr, pru_sockaddr_notsupp); 166 DEFAULT(pu->pru_sosend, sosend_generic); 167 DEFAULT(pu->pru_soreceive, soreceive_generic); 168 DEFAULT(pu->pru_sopoll, sopoll_generic); 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 static int 180 net_init_domain(const void *arg) 181 { 182 const 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 * update global information about maximums 191 */ 192 max_hdr = max_linkhdr + max_protohdr; 193 max_datalen = MHLEN - max_hdr; 194 if (max_datalen < 1) 195 panic("%s: max_datalen < 1", __func__); 196 return (0); 197 } 198 199 #ifdef VIMAGE 200 /* 201 * Detach / free a domain instance. 202 */ 203 static int 204 net_detach_domain(const void *arg) 205 { 206 const struct domain *dp = arg; 207 struct protosw *pr; 208 209 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) 210 if (pr->pr_destroy) 211 (*pr->pr_destroy)(); 212 if (dp->dom_destroy) 213 (*dp->dom_destroy)(); 214 215 return (0); 216 } 217 #endif 218 219 /* 220 * Add a new protocol domain to the list of supported domains 221 * Note: you cant unload it again because a socket may be using it. 222 * XXX can't fail at this time. 223 */ 224 void 225 net_add_domain(void *data) 226 { 227 struct domain *dp; 228 229 dp = (struct domain *)data; 230 mtx_lock(&dom_mtx); 231 dp->dom_next = domains; 232 domains = dp; 233 234 KASSERT(domain_init_status >= 1, 235 ("attempt to net_add_domain(%s) before domaininit()", 236 dp->dom_name)); 237 #ifndef INVARIANTS 238 if (domain_init_status < 1) 239 printf("WARNING: attempt to net_add_domain(%s) before " 240 "domaininit()\n", dp->dom_name); 241 #endif 242 #ifdef notyet 243 KASSERT(domain_init_status < 2, 244 ("attempt to net_add_domain(%s) after domainfinalize()", 245 dp->dom_name)); 246 #else 247 if (domain_init_status >= 2) 248 printf("WARNING: attempt to net_add_domain(%s) after " 249 "domainfinalize()\n", dp->dom_name); 250 #endif 251 mtx_unlock(&dom_mtx); 252 #ifndef VIMAGE_GLOBALS 253 vnet_mod_register_multi(&vnet_domain_modinfo, dp, dp->dom_name); 254 #else 255 net_init_domain(dp); 256 #endif 257 } 258 259 static void 260 socket_zone_change(void *tag) 261 { 262 263 uma_zone_set_max(socket_zone, maxsockets); 264 } 265 266 /* ARGSUSED*/ 267 static void 268 domaininit(void *dummy) 269 { 270 271 /* 272 * Before we do any setup, make sure to initialize the 273 * zone allocator we get struct sockets from. 274 */ 275 socket_zone = uma_zcreate("socket", sizeof(struct socket), NULL, NULL, 276 NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 277 uma_zone_set_max(socket_zone, maxsockets); 278 EVENTHANDLER_REGISTER(maxsockets_change, socket_zone_change, NULL, 279 EVENTHANDLER_PRI_FIRST); 280 281 if (max_linkhdr < 16) /* XXX */ 282 max_linkhdr = 16; 283 284 callout_init(&pffast_callout, CALLOUT_MPSAFE); 285 callout_init(&pfslow_callout, CALLOUT_MPSAFE); 286 287 mtx_lock(&dom_mtx); 288 KASSERT(domain_init_status == 0, ("domaininit called too late!")); 289 domain_init_status = 1; 290 mtx_unlock(&dom_mtx); 291 } 292 293 /* ARGSUSED*/ 294 static void 295 domainfinalize(void *dummy) 296 { 297 298 mtx_lock(&dom_mtx); 299 KASSERT(domain_init_status == 1, ("domainfinalize called too late!")); 300 domain_init_status = 2; 301 mtx_unlock(&dom_mtx); 302 303 callout_reset(&pffast_callout, 1, pffasttimo, NULL); 304 callout_reset(&pfslow_callout, 1, pfslowtimo, NULL); 305 } 306 307 struct protosw * 308 pffindtype(int family, int type) 309 { 310 struct domain *dp; 311 struct protosw *pr; 312 313 for (dp = domains; dp; dp = dp->dom_next) 314 if (dp->dom_family == family) 315 goto found; 316 return (0); 317 found: 318 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) 319 if (pr->pr_type && pr->pr_type == type) 320 return (pr); 321 return (0); 322 } 323 324 struct protosw * 325 pffindproto(int family, int protocol, int type) 326 { 327 struct domain *dp; 328 struct protosw *pr; 329 struct protosw *maybe = 0; 330 331 if (family == 0) 332 return (0); 333 for (dp = domains; dp; dp = dp->dom_next) 334 if (dp->dom_family == family) 335 goto found; 336 return (0); 337 found: 338 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 339 if ((pr->pr_protocol == protocol) && (pr->pr_type == type)) 340 return (pr); 341 342 if (type == SOCK_RAW && pr->pr_type == SOCK_RAW && 343 pr->pr_protocol == 0 && maybe == (struct protosw *)0) 344 maybe = pr; 345 } 346 return (maybe); 347 } 348 349 /* 350 * The caller must make sure that the new protocol is fully set up and ready to 351 * accept requests before it is registered. 352 */ 353 int 354 pf_proto_register(int family, struct protosw *npr) 355 { 356 struct domain *dp; 357 struct protosw *pr, *fpr; 358 359 /* Sanity checks. */ 360 if (family == 0) 361 return (EPFNOSUPPORT); 362 if (npr->pr_type == 0) 363 return (EPROTOTYPE); 364 if (npr->pr_protocol == 0) 365 return (EPROTONOSUPPORT); 366 if (npr->pr_usrreqs == NULL) 367 return (ENXIO); 368 369 /* Try to find the specified domain based on the family. */ 370 for (dp = domains; dp; dp = dp->dom_next) 371 if (dp->dom_family == family) 372 goto found; 373 return (EPFNOSUPPORT); 374 375 found: 376 /* Initialize backpointer to struct domain. */ 377 npr->pr_domain = dp; 378 fpr = NULL; 379 380 /* 381 * Protect us against races when two protocol registrations for 382 * the same protocol happen at the same time. 383 */ 384 mtx_lock(&dom_mtx); 385 386 /* The new protocol must not yet exist. */ 387 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 388 if ((pr->pr_type == npr->pr_type) && 389 (pr->pr_protocol == npr->pr_protocol)) { 390 mtx_unlock(&dom_mtx); 391 return (EEXIST); /* XXX: Check only protocol? */ 392 } 393 /* While here, remember the first free spacer. */ 394 if ((fpr == NULL) && (pr->pr_protocol == PROTO_SPACER)) 395 fpr = pr; 396 } 397 398 /* If no free spacer is found we can't add the new protocol. */ 399 if (fpr == NULL) { 400 mtx_unlock(&dom_mtx); 401 return (ENOMEM); 402 } 403 404 /* Copy the new struct protosw over the spacer. */ 405 bcopy(npr, fpr, sizeof(*fpr)); 406 407 /* Job is done, no more protection required. */ 408 mtx_unlock(&dom_mtx); 409 410 /* Initialize and activate the protocol. */ 411 protosw_init(fpr); 412 413 return (0); 414 } 415 416 /* 417 * The caller must make sure the protocol and its functions correctly shut down 418 * all sockets and release all locks and memory references. 419 */ 420 int 421 pf_proto_unregister(int family, int protocol, int type) 422 { 423 struct domain *dp; 424 struct protosw *pr, *dpr; 425 426 /* Sanity checks. */ 427 if (family == 0) 428 return (EPFNOSUPPORT); 429 if (protocol == 0) 430 return (EPROTONOSUPPORT); 431 if (type == 0) 432 return (EPROTOTYPE); 433 434 /* Try to find the specified domain based on the family type. */ 435 for (dp = domains; dp; dp = dp->dom_next) 436 if (dp->dom_family == family) 437 goto found; 438 return (EPFNOSUPPORT); 439 440 found: 441 dpr = NULL; 442 443 /* Lock out everyone else while we are manipulating the protosw. */ 444 mtx_lock(&dom_mtx); 445 446 /* The protocol must exist and only once. */ 447 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 448 if ((pr->pr_type == type) && (pr->pr_protocol == protocol)) { 449 if (dpr != NULL) { 450 mtx_unlock(&dom_mtx); 451 return (EMLINK); /* Should not happen! */ 452 } else 453 dpr = pr; 454 } 455 } 456 457 /* Protocol does not exist. */ 458 if (dpr == NULL) { 459 mtx_unlock(&dom_mtx); 460 return (EPROTONOSUPPORT); 461 } 462 463 /* De-orbit the protocol and make the slot available again. */ 464 dpr->pr_type = 0; 465 dpr->pr_domain = dp; 466 dpr->pr_protocol = PROTO_SPACER; 467 dpr->pr_flags = 0; 468 dpr->pr_input = NULL; 469 dpr->pr_output = NULL; 470 dpr->pr_ctlinput = NULL; 471 dpr->pr_ctloutput = NULL; 472 dpr->pr_init = NULL; 473 dpr->pr_fasttimo = NULL; 474 dpr->pr_slowtimo = NULL; 475 dpr->pr_drain = NULL; 476 dpr->pr_usrreqs = &nousrreqs; 477 478 /* Job is done, not more protection required. */ 479 mtx_unlock(&dom_mtx); 480 481 return (0); 482 } 483 484 void 485 pfctlinput(int cmd, struct sockaddr *sa) 486 { 487 struct domain *dp; 488 struct protosw *pr; 489 490 for (dp = domains; dp; dp = dp->dom_next) 491 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) 492 if (pr->pr_ctlinput) 493 (*pr->pr_ctlinput)(cmd, sa, (void *)0); 494 } 495 496 void 497 pfctlinput2(int cmd, struct sockaddr *sa, void *ctlparam) 498 { 499 struct domain *dp; 500 struct protosw *pr; 501 502 if (!sa) 503 return; 504 for (dp = domains; dp; dp = dp->dom_next) { 505 /* 506 * the check must be made by xx_ctlinput() anyways, to 507 * make sure we use data item pointed to by ctlparam in 508 * correct way. the following check is made just for safety. 509 */ 510 if (dp->dom_family != sa->sa_family) 511 continue; 512 513 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) 514 if (pr->pr_ctlinput) 515 (*pr->pr_ctlinput)(cmd, sa, ctlparam); 516 } 517 } 518 519 static void 520 pfslowtimo(void *arg) 521 { 522 struct domain *dp; 523 struct protosw *pr; 524 525 for (dp = domains; dp; dp = dp->dom_next) 526 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) 527 if (pr->pr_slowtimo) 528 (*pr->pr_slowtimo)(); 529 callout_reset(&pfslow_callout, hz/2, pfslowtimo, NULL); 530 } 531 532 static void 533 pffasttimo(void *arg) 534 { 535 struct domain *dp; 536 struct protosw *pr; 537 538 for (dp = domains; dp; dp = dp->dom_next) 539 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) 540 if (pr->pr_fasttimo) 541 (*pr->pr_fasttimo)(); 542 callout_reset(&pffast_callout, hz/5, pffasttimo, NULL); 543 } 544