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 <vm/uma.h> 47 48 /* 49 * System initialization 50 * 51 * Note: domain initialization takes place on a per domain basis 52 * as a result of traversing a SYSINIT linker set. Most likely, 53 * each domain would want to call DOMAIN_SET(9) itself, which 54 * would cause the domain to be added just after domaininit() 55 * is called during startup. 56 * 57 * See DOMAIN_SET(9) for details on its use. 58 */ 59 60 static void domaininit(void *); 61 SYSINIT(domain, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST, domaininit, NULL) 62 63 static void domainfinalize(void *); 64 SYSINIT(domainfin, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST, domainfinalize, 65 NULL) 66 67 static struct callout pffast_callout; 68 static struct callout pfslow_callout; 69 70 static void pffasttimo(void *); 71 static void pfslowtimo(void *); 72 73 struct domain *domains; /* registered protocol domains */ 74 int domain_init_status = 0; 75 struct mtx dom_mtx; /* domain list lock */ 76 MTX_SYSINIT(domain, &dom_mtx, "domain list", MTX_DEF); 77 78 /* 79 * Dummy protocol specific user requests function pointer array. 80 * All functions return EOPNOTSUPP. 81 */ 82 struct pr_usrreqs nousrreqs = { 83 .pru_accept = pru_accept_notsupp, 84 .pru_attach = pru_attach_notsupp, 85 .pru_bind = pru_bind_notsupp, 86 .pru_connect = pru_connect_notsupp, 87 .pru_connect2 = pru_connect2_notsupp, 88 .pru_control = pru_control_notsupp, 89 .pru_disconnect = pru_disconnect_notsupp, 90 .pru_listen = pru_listen_notsupp, 91 .pru_peeraddr = pru_peeraddr_notsupp, 92 .pru_rcvd = pru_rcvd_notsupp, 93 .pru_rcvoob = pru_rcvoob_notsupp, 94 .pru_send = pru_send_notsupp, 95 .pru_sense = pru_sense_null, 96 .pru_shutdown = pru_shutdown_notsupp, 97 .pru_sockaddr = pru_sockaddr_notsupp, 98 .pru_sosend = pru_sosend_notsupp, 99 .pru_soreceive = pru_soreceive_notsupp, 100 .pru_sopoll = pru_sopoll_notsupp, 101 }; 102 103 static void 104 protosw_init(struct protosw *pr) 105 { 106 struct pr_usrreqs *pu; 107 108 pu = pr->pr_usrreqs; 109 KASSERT(pu != NULL, ("protosw_init: %ssw[%d] has no usrreqs!", 110 pr->pr_domain->dom_name, 111 (int)(pr - pr->pr_domain->dom_protosw))); 112 113 #define DEFAULT(foo, bar) if ((foo) == NULL) (foo) = (bar) 114 DEFAULT(pu->pru_accept, pru_accept_notsupp); 115 DEFAULT(pu->pru_connect, pru_connect_notsupp); 116 DEFAULT(pu->pru_connect2, pru_connect2_notsupp); 117 DEFAULT(pu->pru_control, pru_control_notsupp); 118 DEFAULT(pu->pru_listen, pru_listen_notsupp); 119 DEFAULT(pu->pru_rcvd, pru_rcvd_notsupp); 120 DEFAULT(pu->pru_rcvoob, pru_rcvoob_notsupp); 121 DEFAULT(pu->pru_sense, pru_sense_null); 122 DEFAULT(pu->pru_sosend, sosend_generic); 123 DEFAULT(pu->pru_soreceive, soreceive_generic); 124 DEFAULT(pu->pru_sopoll, sopoll_generic); 125 #undef DEFAULT 126 if (pr->pr_init) 127 (*pr->pr_init)(); 128 } 129 130 /* 131 * Add a new protocol domain to the list of supported domains 132 * Note: you cant unload it again because a socket may be using it. 133 * XXX can't fail at this time. 134 */ 135 static void 136 net_init_domain(struct domain *dp) 137 { 138 struct protosw *pr; 139 140 if (dp->dom_init) 141 (*dp->dom_init)(); 142 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) 143 protosw_init(pr); 144 /* 145 * update global information about maximums 146 */ 147 max_hdr = max_linkhdr + max_protohdr; 148 max_datalen = MHLEN - max_hdr; 149 if (max_datalen < 1) 150 panic("%s: max_datalen < 1", __func__); 151 } 152 153 /* 154 * Add a new protocol domain to the list of supported domains 155 * Note: you cant unload it again because a socket may be using it. 156 * XXX can't fail at this time. 157 */ 158 void 159 net_add_domain(void *data) 160 { 161 struct domain *dp; 162 163 dp = (struct domain *)data; 164 mtx_lock(&dom_mtx); 165 dp->dom_next = domains; 166 domains = dp; 167 168 KASSERT(domain_init_status >= 1, 169 ("attempt to net_add_domain(%s) before domaininit()", 170 dp->dom_name)); 171 #ifndef INVARIANTS 172 if (domain_init_status < 1) 173 printf("WARNING: attempt to net_add_domain(%s) before " 174 "domaininit()\n", dp->dom_name); 175 #endif 176 #ifdef notyet 177 KASSERT(domain_init_status < 2, 178 ("attempt to net_add_domain(%s) after domainfinalize()", 179 dp->dom_name)); 180 #else 181 if (domain_init_status >= 2) 182 printf("WARNING: attempt to net_add_domain(%s) after " 183 "domainfinalize()\n", dp->dom_name); 184 #endif 185 mtx_unlock(&dom_mtx); 186 net_init_domain(dp); 187 } 188 189 static void 190 socket_zone_change(void *tag) 191 { 192 193 uma_zone_set_max(socket_zone, maxsockets); 194 } 195 196 /* ARGSUSED*/ 197 static void 198 domaininit(void *dummy) 199 { 200 201 /* 202 * Before we do any setup, make sure to initialize the 203 * zone allocator we get struct sockets from. 204 */ 205 socket_zone = uma_zcreate("socket", sizeof(struct socket), NULL, NULL, 206 NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 207 uma_zone_set_max(socket_zone, maxsockets); 208 EVENTHANDLER_REGISTER(maxsockets_change, socket_zone_change, NULL, 209 EVENTHANDLER_PRI_FIRST); 210 211 if (max_linkhdr < 16) /* XXX */ 212 max_linkhdr = 16; 213 214 if (debug_mpsafenet) { 215 callout_init(&pffast_callout, CALLOUT_MPSAFE); 216 callout_init(&pfslow_callout, CALLOUT_MPSAFE); 217 } else { 218 callout_init(&pffast_callout, 0); 219 callout_init(&pfslow_callout, 0); 220 } 221 222 mtx_lock(&dom_mtx); 223 KASSERT(domain_init_status == 0, ("domaininit called too late!")); 224 domain_init_status = 1; 225 mtx_unlock(&dom_mtx); 226 } 227 228 /* ARGSUSED*/ 229 static void 230 domainfinalize(void *dummy) 231 { 232 233 mtx_lock(&dom_mtx); 234 KASSERT(domain_init_status == 1, ("domainfinalize called too late!")); 235 domain_init_status = 2; 236 mtx_unlock(&dom_mtx); 237 238 callout_reset(&pffast_callout, 1, pffasttimo, NULL); 239 callout_reset(&pfslow_callout, 1, pfslowtimo, NULL); 240 } 241 242 struct protosw * 243 pffindtype(int family, int type) 244 { 245 struct domain *dp; 246 struct protosw *pr; 247 248 for (dp = domains; dp; dp = dp->dom_next) 249 if (dp->dom_family == family) 250 goto found; 251 return (0); 252 found: 253 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) 254 if (pr->pr_type && pr->pr_type == type) 255 return (pr); 256 return (0); 257 } 258 259 struct protosw * 260 pffindproto(int family, int protocol, int type) 261 { 262 struct domain *dp; 263 struct protosw *pr; 264 struct protosw *maybe = 0; 265 266 if (family == 0) 267 return (0); 268 for (dp = domains; dp; dp = dp->dom_next) 269 if (dp->dom_family == family) 270 goto found; 271 return (0); 272 found: 273 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 274 if ((pr->pr_protocol == protocol) && (pr->pr_type == type)) 275 return (pr); 276 277 if (type == SOCK_RAW && pr->pr_type == SOCK_RAW && 278 pr->pr_protocol == 0 && maybe == (struct protosw *)0) 279 maybe = pr; 280 } 281 return (maybe); 282 } 283 284 /* 285 * The caller must make sure that the new protocol is fully set up and ready to 286 * accept requests before it is registered. 287 */ 288 int 289 pf_proto_register(int family, struct protosw *npr) 290 { 291 struct domain *dp; 292 struct protosw *pr, *fpr; 293 294 /* Sanity checks. */ 295 if (family == 0) 296 return (EPFNOSUPPORT); 297 if (npr->pr_type == 0) 298 return (EPROTOTYPE); 299 if (npr->pr_protocol == 0) 300 return (EPROTONOSUPPORT); 301 if (npr->pr_usrreqs == NULL) 302 return (ENXIO); 303 304 /* Try to find the specified domain based on the family. */ 305 for (dp = domains; dp; dp = dp->dom_next) 306 if (dp->dom_family == family) 307 goto found; 308 return (EPFNOSUPPORT); 309 310 found: 311 /* Initialize backpointer to struct domain. */ 312 npr->pr_domain = dp; 313 fpr = NULL; 314 315 /* 316 * Protect us against races when two protocol registrations for 317 * the same protocol happen at the same time. 318 */ 319 mtx_lock(&Giant); 320 321 /* The new protocol must not yet exist. */ 322 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 323 if ((pr->pr_type == npr->pr_type) && 324 (pr->pr_protocol == npr->pr_protocol)) { 325 mtx_unlock(&Giant); 326 return (EEXIST); /* XXX: Check only protocol? */ 327 } 328 /* While here, remember the first free spacer. */ 329 if ((fpr == NULL) && (pr->pr_protocol == PROTO_SPACER)) 330 fpr = pr; 331 } 332 333 /* If no free spacer is found we can't add the new protocol. */ 334 if (fpr == NULL) { 335 mtx_unlock(&Giant); 336 return (ENOMEM); 337 } 338 339 /* Copy the new struct protosw over the spacer. */ 340 bcopy(npr, fpr, sizeof(*fpr)); 341 342 /* Job is done, no more protection required. */ 343 mtx_unlock(&Giant); 344 345 /* Initialize and activate the protocol. */ 346 protosw_init(fpr); 347 348 return (0); 349 } 350 351 /* 352 * The caller must make sure the protocol and its functions correctly shut down 353 * all sockets and release all locks and memory references. 354 */ 355 int 356 pf_proto_unregister(int family, int protocol, int type) 357 { 358 struct domain *dp; 359 struct protosw *pr, *dpr; 360 361 /* Sanity checks. */ 362 if (family == 0) 363 return (EPFNOSUPPORT); 364 if (protocol == 0) 365 return (EPROTONOSUPPORT); 366 if (type == 0) 367 return (EPROTOTYPE); 368 369 /* Try to find the specified domain based on the family type. */ 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 dpr = NULL; 377 378 /* Lock out everyone else while we are manipulating the protosw. */ 379 mtx_lock(&Giant); 380 381 /* The protocol must exist and only once. */ 382 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 383 if ((pr->pr_type == type) && (pr->pr_protocol == protocol)) { 384 if (dpr != NULL) { 385 mtx_unlock(&Giant); 386 return (EMLINK); /* Should not happen! */ 387 } else 388 dpr = pr; 389 } 390 } 391 392 /* Protocol does not exist. */ 393 if (dpr == NULL) { 394 mtx_unlock(&Giant); 395 return (EPROTONOSUPPORT); 396 } 397 398 /* De-orbit the protocol and make the slot available again. */ 399 dpr->pr_type = 0; 400 dpr->pr_domain = dp; 401 dpr->pr_protocol = PROTO_SPACER; 402 dpr->pr_flags = 0; 403 dpr->pr_input = NULL; 404 dpr->pr_output = NULL; 405 dpr->pr_ctlinput = NULL; 406 dpr->pr_ctloutput = NULL; 407 dpr->pr_ousrreq = NULL; 408 dpr->pr_init = NULL; 409 dpr->pr_fasttimo = NULL; 410 dpr->pr_slowtimo = NULL; 411 dpr->pr_drain = NULL; 412 dpr->pr_usrreqs = &nousrreqs; 413 414 /* Job is done, not more protection required. */ 415 mtx_unlock(&Giant); 416 417 return (0); 418 } 419 420 void 421 pfctlinput(int cmd, struct sockaddr *sa) 422 { 423 struct domain *dp; 424 struct protosw *pr; 425 426 for (dp = domains; dp; dp = dp->dom_next) 427 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) 428 if (pr->pr_ctlinput) 429 (*pr->pr_ctlinput)(cmd, sa, (void *)0); 430 } 431 432 void 433 pfctlinput2(int cmd, struct sockaddr *sa, void *ctlparam) 434 { 435 struct domain *dp; 436 struct protosw *pr; 437 438 if (!sa) 439 return; 440 for (dp = domains; dp; dp = dp->dom_next) { 441 /* 442 * the check must be made by xx_ctlinput() anyways, to 443 * make sure we use data item pointed to by ctlparam in 444 * correct way. the following check is made just for safety. 445 */ 446 if (dp->dom_family != sa->sa_family) 447 continue; 448 449 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) 450 if (pr->pr_ctlinput) 451 (*pr->pr_ctlinput)(cmd, sa, ctlparam); 452 } 453 } 454 455 static void 456 pfslowtimo(void *arg) 457 { 458 struct domain *dp; 459 struct protosw *pr; 460 461 NET_ASSERT_GIANT(); 462 463 for (dp = domains; dp; dp = dp->dom_next) 464 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) 465 if (pr->pr_slowtimo) 466 (*pr->pr_slowtimo)(); 467 callout_reset(&pfslow_callout, hz/2, pfslowtimo, NULL); 468 } 469 470 static void 471 pffasttimo(void *arg) 472 { 473 struct domain *dp; 474 struct protosw *pr; 475 476 NET_ASSERT_GIANT(); 477 478 for (dp = domains; dp; dp = dp->dom_next) 479 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) 480 if (pr->pr_fasttimo) 481 (*pr->pr_fasttimo)(); 482 callout_reset(&pffast_callout, hz/5, pffasttimo, NULL); 483 } 484