1 /*- 2 * Copyright (c) 2002 Michael Shalayeff. 3 * Copyright (c) 2003 Ryan McBride. 4 * Copyright (c) 2011 Gleb Smirnoff <glebius@FreeBSD.org> 5 * 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 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 17 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 18 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 19 * IN NO EVENT SHALL THE AUTHOR OR HIS RELATIVES BE LIABLE FOR ANY DIRECT, 20 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES 21 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 22 * SERVICES; LOSS OF MIND, USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, 24 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING 25 * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF 26 * THE POSSIBILITY OF SUCH DAMAGE. 27 */ 28 29 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 31 32 #include "opt_bpf.h" 33 #include "opt_inet.h" 34 #include "opt_inet6.h" 35 36 #include <sys/param.h> 37 #include <sys/systm.h> 38 #include <sys/bus.h> 39 #include <sys/jail.h> 40 #include <sys/kernel.h> 41 #include <sys/limits.h> 42 #include <sys/malloc.h> 43 #include <sys/mbuf.h> 44 #include <sys/module.h> 45 #include <sys/priv.h> 46 #include <sys/proc.h> 47 #include <sys/protosw.h> 48 #include <sys/socket.h> 49 #include <sys/sockio.h> 50 #include <sys/sysctl.h> 51 #include <sys/syslog.h> 52 #include <sys/taskqueue.h> 53 #include <sys/counter.h> 54 55 #include <net/ethernet.h> 56 #include <net/fddi.h> 57 #include <net/if.h> 58 #include <net/if_var.h> 59 #include <net/if_dl.h> 60 #include <net/if_llatbl.h> 61 #include <net/if_types.h> 62 #include <net/iso88025.h> 63 #include <net/route.h> 64 #include <net/vnet.h> 65 66 #if defined(INET) || defined(INET6) 67 #include <netinet/in.h> 68 #include <netinet/in_var.h> 69 #include <netinet/ip_carp.h> 70 #include <netinet/ip.h> 71 #include <machine/in_cksum.h> 72 #endif 73 #ifdef INET 74 #include <netinet/ip_var.h> 75 #include <netinet/if_ether.h> 76 #endif 77 78 #ifdef INET6 79 #include <netinet/icmp6.h> 80 #include <netinet/ip6.h> 81 #include <netinet6/ip6protosw.h> 82 #include <netinet6/in6_var.h> 83 #include <netinet6/ip6_var.h> 84 #include <netinet6/scope6_var.h> 85 #include <netinet6/nd6.h> 86 #endif 87 88 #include <crypto/sha1.h> 89 90 static MALLOC_DEFINE(M_CARP, "CARP", "CARP addresses"); 91 92 struct carp_softc { 93 struct ifnet *sc_carpdev; /* Pointer to parent ifnet. */ 94 struct ifaddr **sc_ifas; /* Our ifaddrs. */ 95 struct sockaddr_dl sc_addr; /* Our link level address. */ 96 struct callout sc_ad_tmo; /* Advertising timeout. */ 97 #ifdef INET 98 struct callout sc_md_tmo; /* Master down timeout. */ 99 #endif 100 #ifdef INET6 101 struct callout sc_md6_tmo; /* XXX: Master down timeout. */ 102 #endif 103 struct mtx sc_mtx; 104 105 int sc_vhid; 106 int sc_advskew; 107 int sc_advbase; 108 109 int sc_naddrs; 110 int sc_naddrs6; 111 int sc_ifasiz; 112 enum { INIT = 0, BACKUP, MASTER } sc_state; 113 int sc_suppress; 114 int sc_sendad_errors; 115 #define CARP_SENDAD_MAX_ERRORS 3 116 int sc_sendad_success; 117 #define CARP_SENDAD_MIN_SUCCESS 3 118 119 int sc_init_counter; 120 uint64_t sc_counter; 121 122 /* authentication */ 123 #define CARP_HMAC_PAD 64 124 unsigned char sc_key[CARP_KEY_LEN]; 125 unsigned char sc_pad[CARP_HMAC_PAD]; 126 SHA1_CTX sc_sha1; 127 128 TAILQ_ENTRY(carp_softc) sc_list; /* On the carp_if list. */ 129 LIST_ENTRY(carp_softc) sc_next; /* On the global list. */ 130 }; 131 132 struct carp_if { 133 #ifdef INET 134 int cif_naddrs; 135 #endif 136 #ifdef INET6 137 int cif_naddrs6; 138 #endif 139 TAILQ_HEAD(, carp_softc) cif_vrs; 140 #ifdef INET 141 struct ip_moptions cif_imo; 142 #endif 143 #ifdef INET6 144 struct ip6_moptions cif_im6o; 145 #endif 146 struct ifnet *cif_ifp; 147 struct mtx cif_mtx; 148 uint32_t cif_flags; 149 #define CIF_PROMISC 0x00000001 150 }; 151 152 #define CARP_INET 0 153 #define CARP_INET6 1 154 static int proto_reg[] = {-1, -1}; 155 156 /* 157 * Brief design of carp(4). 158 * 159 * Any carp-capable ifnet may have a list of carp softcs hanging off 160 * its ifp->if_carp pointer. Each softc represents one unique virtual 161 * host id, or vhid. The softc has a back pointer to the ifnet. All 162 * softcs are joined in a global list, which has quite limited use. 163 * 164 * Any interface address that takes part in CARP negotiation has a 165 * pointer to the softc of its vhid, ifa->ifa_carp. That could be either 166 * AF_INET or AF_INET6 address. 167 * 168 * Although, one can get the softc's backpointer to ifnet and traverse 169 * through its ifp->if_addrhead queue to find all interface addresses 170 * involved in CARP, we keep a growable array of ifaddr pointers. This 171 * allows us to avoid grabbing the IF_ADDR_LOCK() in many traversals that 172 * do calls into the network stack, thus avoiding LORs. 173 * 174 * Locking: 175 * 176 * Each softc has a lock sc_mtx. It is used to synchronise carp_input_c(), 177 * callout-driven events and ioctl()s. 178 * 179 * To traverse the list of softcs on an ifnet we use CIF_LOCK(), to 180 * traverse the global list we use the mutex carp_mtx. 181 * 182 * Known issues with locking: 183 * 184 * - There is no protection for races between two ioctl() requests, 185 * neither SIOCSVH, nor SIOCAIFADDR & SIOCAIFADDR_IN6. I think that all 186 * interface ioctl()s should be serialized right in net/if.c. 187 * - Sending ad, we put the pointer to the softc in an mtag, and no reference 188 * counting is done on the softc. 189 * - On module unload we may race (?) with packet processing thread 190 * dereferencing our function pointers. 191 */ 192 193 /* Accept incoming CARP packets. */ 194 static VNET_DEFINE(int, carp_allow) = 1; 195 #define V_carp_allow VNET(carp_allow) 196 197 /* Preempt slower nodes. */ 198 static VNET_DEFINE(int, carp_preempt) = 0; 199 #define V_carp_preempt VNET(carp_preempt) 200 201 /* Log level. */ 202 static VNET_DEFINE(int, carp_log) = 1; 203 #define V_carp_log VNET(carp_log) 204 205 /* Global advskew demotion. */ 206 static VNET_DEFINE(int, carp_demotion) = 0; 207 #define V_carp_demotion VNET(carp_demotion) 208 209 /* Send error demotion factor. */ 210 static VNET_DEFINE(int, carp_senderr_adj) = CARP_MAXSKEW; 211 #define V_carp_senderr_adj VNET(carp_senderr_adj) 212 213 /* Iface down demotion factor. */ 214 static VNET_DEFINE(int, carp_ifdown_adj) = CARP_MAXSKEW; 215 #define V_carp_ifdown_adj VNET(carp_ifdown_adj) 216 217 static int carp_demote_adj_sysctl(SYSCTL_HANDLER_ARGS); 218 219 SYSCTL_NODE(_net_inet, IPPROTO_CARP, carp, CTLFLAG_RW, 0, "CARP"); 220 SYSCTL_VNET_INT(_net_inet_carp, OID_AUTO, allow, CTLFLAG_RW, 221 &VNET_NAME(carp_allow), 0, "Accept incoming CARP packets"); 222 SYSCTL_VNET_INT(_net_inet_carp, OID_AUTO, preempt, CTLFLAG_RW, 223 &VNET_NAME(carp_preempt), 0, "High-priority backup preemption mode"); 224 SYSCTL_VNET_INT(_net_inet_carp, OID_AUTO, log, CTLFLAG_RW, 225 &VNET_NAME(carp_log), 0, "CARP log level"); 226 SYSCTL_VNET_PROC(_net_inet_carp, OID_AUTO, demotion, CTLTYPE_INT|CTLFLAG_RW, 227 0, 0, carp_demote_adj_sysctl, "I", 228 "Adjust demotion factor (skew of advskew)"); 229 SYSCTL_VNET_INT(_net_inet_carp, OID_AUTO, senderr_demotion_factor, CTLFLAG_RW, 230 &VNET_NAME(carp_senderr_adj), 0, "Send error demotion factor adjustment"); 231 SYSCTL_VNET_INT(_net_inet_carp, OID_AUTO, ifdown_demotion_factor, CTLFLAG_RW, 232 &VNET_NAME(carp_ifdown_adj), 0, 233 "Interface down demotion factor adjustment"); 234 235 VNET_PCPUSTAT_DEFINE(struct carpstats, carpstats); 236 VNET_PCPUSTAT_SYSINIT(carpstats); 237 VNET_PCPUSTAT_SYSUNINIT(carpstats); 238 239 #define CARPSTATS_ADD(name, val) \ 240 counter_u64_add(VNET(carpstats)[offsetof(struct carpstats, name) / \ 241 sizeof(uint64_t)], (val)) 242 #define CARPSTATS_INC(name) CARPSTATS_ADD(name, 1) 243 244 SYSCTL_VNET_PCPUSTAT(_net_inet_carp, OID_AUTO, stats, struct carpstats, 245 carpstats, "CARP statistics (struct carpstats, netinet/ip_carp.h)"); 246 247 #define CARP_LOCK_INIT(sc) mtx_init(&(sc)->sc_mtx, "carp_softc", \ 248 NULL, MTX_DEF) 249 #define CARP_LOCK_DESTROY(sc) mtx_destroy(&(sc)->sc_mtx) 250 #define CARP_LOCK_ASSERT(sc) mtx_assert(&(sc)->sc_mtx, MA_OWNED) 251 #define CARP_LOCK(sc) mtx_lock(&(sc)->sc_mtx) 252 #define CARP_UNLOCK(sc) mtx_unlock(&(sc)->sc_mtx) 253 #define CIF_LOCK_INIT(cif) mtx_init(&(cif)->cif_mtx, "carp_if", \ 254 NULL, MTX_DEF) 255 #define CIF_LOCK_DESTROY(cif) mtx_destroy(&(cif)->cif_mtx) 256 #define CIF_LOCK_ASSERT(cif) mtx_assert(&(cif)->cif_mtx, MA_OWNED) 257 #define CIF_LOCK(cif) mtx_lock(&(cif)->cif_mtx) 258 #define CIF_UNLOCK(cif) mtx_unlock(&(cif)->cif_mtx) 259 #define CIF_FREE(cif) do { \ 260 CIF_LOCK_ASSERT(cif); \ 261 if (TAILQ_EMPTY(&(cif)->cif_vrs)) \ 262 carp_free_if(cif); \ 263 else \ 264 CIF_UNLOCK(cif); \ 265 } while (0) 266 267 #define CARP_LOG(...) do { \ 268 if (V_carp_log > 0) \ 269 log(LOG_INFO, "carp: " __VA_ARGS__); \ 270 } while (0) 271 272 #define CARP_DEBUG(...) do { \ 273 if (V_carp_log > 1) \ 274 log(LOG_DEBUG, __VA_ARGS__); \ 275 } while (0) 276 277 #define IFNET_FOREACH_IFA(ifp, ifa) \ 278 IF_ADDR_LOCK_ASSERT(ifp); \ 279 TAILQ_FOREACH((ifa), &(ifp)->if_addrhead, ifa_link) \ 280 if ((ifa)->ifa_carp != NULL) 281 282 #define CARP_FOREACH_IFA(sc, ifa) \ 283 CARP_LOCK_ASSERT(sc); \ 284 for (int _i = 0; \ 285 _i < (sc)->sc_naddrs + (sc)->sc_naddrs6 && \ 286 ((ifa) = sc->sc_ifas[_i]) != NULL; \ 287 ++_i) 288 289 #define IFNET_FOREACH_CARP(ifp, sc) \ 290 CIF_LOCK_ASSERT(ifp->if_carp); \ 291 TAILQ_FOREACH((sc), &(ifp)->if_carp->cif_vrs, sc_list) 292 293 #define DEMOTE_ADVSKEW(sc) \ 294 (((sc)->sc_advskew + V_carp_demotion > CARP_MAXSKEW) ? \ 295 CARP_MAXSKEW : ((sc)->sc_advskew + V_carp_demotion)) 296 297 static void carp_input_c(struct mbuf *, struct carp_header *, sa_family_t); 298 static struct carp_softc 299 *carp_alloc(struct ifnet *); 300 static void carp_detach_locked(struct ifaddr *); 301 static void carp_destroy(struct carp_softc *); 302 static struct carp_if 303 *carp_alloc_if(struct ifnet *); 304 static void carp_free_if(struct carp_if *); 305 static void carp_set_state(struct carp_softc *, int); 306 static void carp_sc_state(struct carp_softc *); 307 static void carp_setrun(struct carp_softc *, sa_family_t); 308 static void carp_master_down(void *); 309 static void carp_master_down_locked(struct carp_softc *); 310 static void carp_send_ad(void *); 311 static void carp_send_ad_locked(struct carp_softc *); 312 static void carp_addroute(struct carp_softc *); 313 static void carp_ifa_addroute(struct ifaddr *); 314 static void carp_delroute(struct carp_softc *); 315 static void carp_ifa_delroute(struct ifaddr *); 316 static void carp_send_ad_all(void *, int); 317 static void carp_demote_adj(int, char *); 318 319 static LIST_HEAD(, carp_softc) carp_list; 320 static struct mtx carp_mtx; 321 static struct task carp_sendall_task = 322 TASK_INITIALIZER(0, carp_send_ad_all, NULL); 323 324 static void 325 carp_hmac_prepare(struct carp_softc *sc) 326 { 327 uint8_t version = CARP_VERSION, type = CARP_ADVERTISEMENT; 328 uint8_t vhid = sc->sc_vhid & 0xff; 329 struct ifaddr *ifa; 330 int i, found; 331 #ifdef INET 332 struct in_addr last, cur, in; 333 #endif 334 #ifdef INET6 335 struct in6_addr last6, cur6, in6; 336 #endif 337 338 CARP_LOCK_ASSERT(sc); 339 340 /* Compute ipad from key. */ 341 bzero(sc->sc_pad, sizeof(sc->sc_pad)); 342 bcopy(sc->sc_key, sc->sc_pad, sizeof(sc->sc_key)); 343 for (i = 0; i < sizeof(sc->sc_pad); i++) 344 sc->sc_pad[i] ^= 0x36; 345 346 /* Precompute first part of inner hash. */ 347 SHA1Init(&sc->sc_sha1); 348 SHA1Update(&sc->sc_sha1, sc->sc_pad, sizeof(sc->sc_pad)); 349 SHA1Update(&sc->sc_sha1, (void *)&version, sizeof(version)); 350 SHA1Update(&sc->sc_sha1, (void *)&type, sizeof(type)); 351 SHA1Update(&sc->sc_sha1, (void *)&vhid, sizeof(vhid)); 352 #ifdef INET 353 cur.s_addr = 0; 354 do { 355 found = 0; 356 last = cur; 357 cur.s_addr = 0xffffffff; 358 CARP_FOREACH_IFA(sc, ifa) { 359 in.s_addr = ifatoia(ifa)->ia_addr.sin_addr.s_addr; 360 if (ifa->ifa_addr->sa_family == AF_INET && 361 ntohl(in.s_addr) > ntohl(last.s_addr) && 362 ntohl(in.s_addr) < ntohl(cur.s_addr)) { 363 cur.s_addr = in.s_addr; 364 found++; 365 } 366 } 367 if (found) 368 SHA1Update(&sc->sc_sha1, (void *)&cur, sizeof(cur)); 369 } while (found); 370 #endif /* INET */ 371 #ifdef INET6 372 memset(&cur6, 0, sizeof(cur6)); 373 do { 374 found = 0; 375 last6 = cur6; 376 memset(&cur6, 0xff, sizeof(cur6)); 377 CARP_FOREACH_IFA(sc, ifa) { 378 in6 = ifatoia6(ifa)->ia_addr.sin6_addr; 379 if (IN6_IS_SCOPE_EMBED(&in6)) 380 in6.s6_addr16[1] = 0; 381 if (ifa->ifa_addr->sa_family == AF_INET6 && 382 memcmp(&in6, &last6, sizeof(in6)) > 0 && 383 memcmp(&in6, &cur6, sizeof(in6)) < 0) { 384 cur6 = in6; 385 found++; 386 } 387 } 388 if (found) 389 SHA1Update(&sc->sc_sha1, (void *)&cur6, sizeof(cur6)); 390 } while (found); 391 #endif /* INET6 */ 392 393 /* convert ipad to opad */ 394 for (i = 0; i < sizeof(sc->sc_pad); i++) 395 sc->sc_pad[i] ^= 0x36 ^ 0x5c; 396 } 397 398 static void 399 carp_hmac_generate(struct carp_softc *sc, uint32_t counter[2], 400 unsigned char md[20]) 401 { 402 SHA1_CTX sha1ctx; 403 404 CARP_LOCK_ASSERT(sc); 405 406 /* fetch first half of inner hash */ 407 bcopy(&sc->sc_sha1, &sha1ctx, sizeof(sha1ctx)); 408 409 SHA1Update(&sha1ctx, (void *)counter, sizeof(sc->sc_counter)); 410 SHA1Final(md, &sha1ctx); 411 412 /* outer hash */ 413 SHA1Init(&sha1ctx); 414 SHA1Update(&sha1ctx, sc->sc_pad, sizeof(sc->sc_pad)); 415 SHA1Update(&sha1ctx, md, 20); 416 SHA1Final(md, &sha1ctx); 417 } 418 419 static int 420 carp_hmac_verify(struct carp_softc *sc, uint32_t counter[2], 421 unsigned char md[20]) 422 { 423 unsigned char md2[20]; 424 425 CARP_LOCK_ASSERT(sc); 426 427 carp_hmac_generate(sc, counter, md2); 428 429 return (bcmp(md, md2, sizeof(md2))); 430 } 431 432 /* 433 * process input packet. 434 * we have rearranged checks order compared to the rfc, 435 * but it seems more efficient this way or not possible otherwise. 436 */ 437 #ifdef INET 438 void 439 carp_input(struct mbuf *m, int hlen) 440 { 441 struct ip *ip = mtod(m, struct ip *); 442 struct carp_header *ch; 443 int iplen, len; 444 445 CARPSTATS_INC(carps_ipackets); 446 447 if (!V_carp_allow) { 448 m_freem(m); 449 return; 450 } 451 452 /* verify that the IP TTL is 255. */ 453 if (ip->ip_ttl != CARP_DFLTTL) { 454 CARPSTATS_INC(carps_badttl); 455 CARP_DEBUG("%s: received ttl %d != 255 on %s\n", __func__, 456 ip->ip_ttl, 457 m->m_pkthdr.rcvif->if_xname); 458 m_freem(m); 459 return; 460 } 461 462 iplen = ip->ip_hl << 2; 463 464 if (m->m_pkthdr.len < iplen + sizeof(*ch)) { 465 CARPSTATS_INC(carps_badlen); 466 CARP_DEBUG("%s: received len %zd < sizeof(struct carp_header) " 467 "on %s\n", __func__, m->m_len - sizeof(struct ip), 468 m->m_pkthdr.rcvif->if_xname); 469 m_freem(m); 470 return; 471 } 472 473 if (iplen + sizeof(*ch) < m->m_len) { 474 if ((m = m_pullup(m, iplen + sizeof(*ch))) == NULL) { 475 CARPSTATS_INC(carps_hdrops); 476 CARP_DEBUG("%s: pullup failed\n", __func__); 477 return; 478 } 479 ip = mtod(m, struct ip *); 480 } 481 ch = (struct carp_header *)((char *)ip + iplen); 482 483 /* 484 * verify that the received packet length is 485 * equal to the CARP header 486 */ 487 len = iplen + sizeof(*ch); 488 if (len > m->m_pkthdr.len) { 489 CARPSTATS_INC(carps_badlen); 490 CARP_DEBUG("%s: packet too short %d on %s\n", __func__, 491 m->m_pkthdr.len, 492 m->m_pkthdr.rcvif->if_xname); 493 m_freem(m); 494 return; 495 } 496 497 if ((m = m_pullup(m, len)) == NULL) { 498 CARPSTATS_INC(carps_hdrops); 499 return; 500 } 501 ip = mtod(m, struct ip *); 502 ch = (struct carp_header *)((char *)ip + iplen); 503 504 /* verify the CARP checksum */ 505 m->m_data += iplen; 506 if (in_cksum(m, len - iplen)) { 507 CARPSTATS_INC(carps_badsum); 508 CARP_DEBUG("%s: checksum failed on %s\n", __func__, 509 m->m_pkthdr.rcvif->if_xname); 510 m_freem(m); 511 return; 512 } 513 m->m_data -= iplen; 514 515 carp_input_c(m, ch, AF_INET); 516 } 517 #endif 518 519 #ifdef INET6 520 int 521 carp6_input(struct mbuf **mp, int *offp, int proto) 522 { 523 struct mbuf *m = *mp; 524 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 525 struct carp_header *ch; 526 u_int len; 527 528 CARPSTATS_INC(carps_ipackets6); 529 530 if (!V_carp_allow) { 531 m_freem(m); 532 return (IPPROTO_DONE); 533 } 534 535 /* check if received on a valid carp interface */ 536 if (m->m_pkthdr.rcvif->if_carp == NULL) { 537 CARPSTATS_INC(carps_badif); 538 CARP_DEBUG("%s: packet received on non-carp interface: %s\n", 539 __func__, m->m_pkthdr.rcvif->if_xname); 540 m_freem(m); 541 return (IPPROTO_DONE); 542 } 543 544 /* verify that the IP TTL is 255 */ 545 if (ip6->ip6_hlim != CARP_DFLTTL) { 546 CARPSTATS_INC(carps_badttl); 547 CARP_DEBUG("%s: received ttl %d != 255 on %s\n", __func__, 548 ip6->ip6_hlim, m->m_pkthdr.rcvif->if_xname); 549 m_freem(m); 550 return (IPPROTO_DONE); 551 } 552 553 /* verify that we have a complete carp packet */ 554 len = m->m_len; 555 IP6_EXTHDR_GET(ch, struct carp_header *, m, *offp, sizeof(*ch)); 556 if (ch == NULL) { 557 CARPSTATS_INC(carps_badlen); 558 CARP_DEBUG("%s: packet size %u too small\n", __func__, len); 559 return (IPPROTO_DONE); 560 } 561 562 563 /* verify the CARP checksum */ 564 m->m_data += *offp; 565 if (in_cksum(m, sizeof(*ch))) { 566 CARPSTATS_INC(carps_badsum); 567 CARP_DEBUG("%s: checksum failed, on %s\n", __func__, 568 m->m_pkthdr.rcvif->if_xname); 569 m_freem(m); 570 return (IPPROTO_DONE); 571 } 572 m->m_data -= *offp; 573 574 carp_input_c(m, ch, AF_INET6); 575 return (IPPROTO_DONE); 576 } 577 #endif /* INET6 */ 578 579 static void 580 carp_input_c(struct mbuf *m, struct carp_header *ch, sa_family_t af) 581 { 582 struct ifnet *ifp = m->m_pkthdr.rcvif; 583 struct ifaddr *ifa; 584 struct carp_softc *sc; 585 uint64_t tmp_counter; 586 struct timeval sc_tv, ch_tv; 587 588 /* verify that the VHID is valid on the receiving interface */ 589 IF_ADDR_RLOCK(ifp); 590 IFNET_FOREACH_IFA(ifp, ifa) 591 if (ifa->ifa_addr->sa_family == af && 592 ifa->ifa_carp->sc_vhid == ch->carp_vhid) { 593 ifa_ref(ifa); 594 break; 595 } 596 IF_ADDR_RUNLOCK(ifp); 597 598 if (ifa == NULL) { 599 CARPSTATS_INC(carps_badvhid); 600 m_freem(m); 601 return; 602 } 603 604 /* verify the CARP version. */ 605 if (ch->carp_version != CARP_VERSION) { 606 CARPSTATS_INC(carps_badver); 607 CARP_DEBUG("%s: invalid version %d\n", ifp->if_xname, 608 ch->carp_version); 609 ifa_free(ifa); 610 m_freem(m); 611 return; 612 } 613 614 sc = ifa->ifa_carp; 615 CARP_LOCK(sc); 616 ifa_free(ifa); 617 618 if (carp_hmac_verify(sc, ch->carp_counter, ch->carp_md)) { 619 CARPSTATS_INC(carps_badauth); 620 CARP_DEBUG("%s: incorrect hash for VHID %u@%s\n", __func__, 621 sc->sc_vhid, ifp->if_xname); 622 goto out; 623 } 624 625 tmp_counter = ntohl(ch->carp_counter[0]); 626 tmp_counter = tmp_counter<<32; 627 tmp_counter += ntohl(ch->carp_counter[1]); 628 629 /* XXX Replay protection goes here */ 630 631 sc->sc_init_counter = 0; 632 sc->sc_counter = tmp_counter; 633 634 sc_tv.tv_sec = sc->sc_advbase; 635 sc_tv.tv_usec = DEMOTE_ADVSKEW(sc) * 1000000 / 256; 636 ch_tv.tv_sec = ch->carp_advbase; 637 ch_tv.tv_usec = ch->carp_advskew * 1000000 / 256; 638 639 switch (sc->sc_state) { 640 case INIT: 641 break; 642 case MASTER: 643 /* 644 * If we receive an advertisement from a master who's going to 645 * be more frequent than us, go into BACKUP state. 646 */ 647 if (timevalcmp(&sc_tv, &ch_tv, >) || 648 timevalcmp(&sc_tv, &ch_tv, ==)) { 649 callout_stop(&sc->sc_ad_tmo); 650 CARP_LOG("VHID %u@%s: MASTER -> BACKUP " 651 "(more frequent advertisement received)\n", 652 sc->sc_vhid, 653 sc->sc_carpdev->if_xname); 654 carp_set_state(sc, BACKUP); 655 carp_setrun(sc, 0); 656 carp_delroute(sc); 657 } 658 break; 659 case BACKUP: 660 /* 661 * If we're pre-empting masters who advertise slower than us, 662 * and this one claims to be slower, treat him as down. 663 */ 664 if (V_carp_preempt && timevalcmp(&sc_tv, &ch_tv, <)) { 665 CARP_LOG("VHID %u@%s: BACKUP -> MASTER " 666 "(preempting a slower master)\n", 667 sc->sc_vhid, 668 sc->sc_carpdev->if_xname); 669 carp_master_down_locked(sc); 670 break; 671 } 672 673 /* 674 * If the master is going to advertise at such a low frequency 675 * that he's guaranteed to time out, we'd might as well just 676 * treat him as timed out now. 677 */ 678 sc_tv.tv_sec = sc->sc_advbase * 3; 679 if (timevalcmp(&sc_tv, &ch_tv, <)) { 680 CARP_LOG("VHID %u@%s: BACKUP -> MASTER " 681 "(master timed out)\n", 682 sc->sc_vhid, 683 sc->sc_carpdev->if_xname); 684 carp_master_down_locked(sc); 685 break; 686 } 687 688 /* 689 * Otherwise, we reset the counter and wait for the next 690 * advertisement. 691 */ 692 carp_setrun(sc, af); 693 break; 694 } 695 696 out: 697 CARP_UNLOCK(sc); 698 m_freem(m); 699 } 700 701 static int 702 carp_prepare_ad(struct mbuf *m, struct carp_softc *sc, struct carp_header *ch) 703 { 704 struct m_tag *mtag; 705 706 if (sc->sc_init_counter) { 707 /* this could also be seconds since unix epoch */ 708 sc->sc_counter = arc4random(); 709 sc->sc_counter = sc->sc_counter << 32; 710 sc->sc_counter += arc4random(); 711 } else 712 sc->sc_counter++; 713 714 ch->carp_counter[0] = htonl((sc->sc_counter>>32)&0xffffffff); 715 ch->carp_counter[1] = htonl(sc->sc_counter&0xffffffff); 716 717 carp_hmac_generate(sc, ch->carp_counter, ch->carp_md); 718 719 /* Tag packet for carp_output */ 720 if ((mtag = m_tag_get(PACKET_TAG_CARP, sizeof(struct carp_softc *), 721 M_NOWAIT)) == NULL) { 722 m_freem(m); 723 CARPSTATS_INC(carps_onomem); 724 return (ENOMEM); 725 } 726 bcopy(&sc, mtag + 1, sizeof(sc)); 727 m_tag_prepend(m, mtag); 728 729 return (0); 730 } 731 732 /* 733 * To avoid LORs and possible recursions this function shouldn't 734 * be called directly, but scheduled via taskqueue. 735 */ 736 static void 737 carp_send_ad_all(void *ctx __unused, int pending __unused) 738 { 739 struct carp_softc *sc; 740 741 mtx_lock(&carp_mtx); 742 LIST_FOREACH(sc, &carp_list, sc_next) 743 if (sc->sc_state == MASTER) { 744 CARP_LOCK(sc); 745 CURVNET_SET(sc->sc_carpdev->if_vnet); 746 carp_send_ad_locked(sc); 747 CURVNET_RESTORE(); 748 CARP_UNLOCK(sc); 749 } 750 mtx_unlock(&carp_mtx); 751 } 752 753 /* Send a periodic advertisement, executed in callout context. */ 754 static void 755 carp_send_ad(void *v) 756 { 757 struct carp_softc *sc = v; 758 759 CARP_LOCK_ASSERT(sc); 760 CURVNET_SET(sc->sc_carpdev->if_vnet); 761 carp_send_ad_locked(sc); 762 CURVNET_RESTORE(); 763 CARP_UNLOCK(sc); 764 } 765 766 static void 767 carp_send_ad_locked(struct carp_softc *sc) 768 { 769 struct carp_header ch; 770 struct timeval tv; 771 struct sockaddr sa; 772 struct ifaddr *ifa; 773 struct carp_header *ch_ptr; 774 struct mbuf *m; 775 int len, advskew; 776 777 CARP_LOCK_ASSERT(sc); 778 779 advskew = DEMOTE_ADVSKEW(sc); 780 tv.tv_sec = sc->sc_advbase; 781 tv.tv_usec = advskew * 1000000 / 256; 782 783 ch.carp_version = CARP_VERSION; 784 ch.carp_type = CARP_ADVERTISEMENT; 785 ch.carp_vhid = sc->sc_vhid; 786 ch.carp_advbase = sc->sc_advbase; 787 ch.carp_advskew = advskew; 788 ch.carp_authlen = 7; /* XXX DEFINE */ 789 ch.carp_pad1 = 0; /* must be zero */ 790 ch.carp_cksum = 0; 791 792 /* XXXGL: OpenBSD picks first ifaddr with needed family. */ 793 794 #ifdef INET 795 if (sc->sc_naddrs) { 796 struct ip *ip; 797 798 m = m_gethdr(M_NOWAIT, MT_DATA); 799 if (m == NULL) { 800 CARPSTATS_INC(carps_onomem); 801 goto resched; 802 } 803 len = sizeof(*ip) + sizeof(ch); 804 m->m_pkthdr.len = len; 805 m->m_pkthdr.rcvif = NULL; 806 m->m_len = len; 807 MH_ALIGN(m, m->m_len); 808 m->m_flags |= M_MCAST; 809 ip = mtod(m, struct ip *); 810 ip->ip_v = IPVERSION; 811 ip->ip_hl = sizeof(*ip) >> 2; 812 ip->ip_tos = IPTOS_LOWDELAY; 813 ip->ip_len = htons(len); 814 ip->ip_id = ip_newid(); 815 ip->ip_off = htons(IP_DF); 816 ip->ip_ttl = CARP_DFLTTL; 817 ip->ip_p = IPPROTO_CARP; 818 ip->ip_sum = 0; 819 820 bzero(&sa, sizeof(sa)); 821 sa.sa_family = AF_INET; 822 ifa = ifaof_ifpforaddr(&sa, sc->sc_carpdev); 823 if (ifa != NULL) { 824 ip->ip_src.s_addr = 825 ifatoia(ifa)->ia_addr.sin_addr.s_addr; 826 ifa_free(ifa); 827 } else 828 ip->ip_src.s_addr = 0; 829 ip->ip_dst.s_addr = htonl(INADDR_CARP_GROUP); 830 831 ch_ptr = (struct carp_header *)(&ip[1]); 832 bcopy(&ch, ch_ptr, sizeof(ch)); 833 if (carp_prepare_ad(m, sc, ch_ptr)) 834 goto resched; 835 836 m->m_data += sizeof(*ip); 837 ch_ptr->carp_cksum = in_cksum(m, len - sizeof(*ip)); 838 m->m_data -= sizeof(*ip); 839 840 CARPSTATS_INC(carps_opackets); 841 842 if (ip_output(m, NULL, NULL, IP_RAWOUTPUT, 843 &sc->sc_carpdev->if_carp->cif_imo, NULL)) { 844 if (sc->sc_sendad_errors < INT_MAX) 845 sc->sc_sendad_errors++; 846 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) 847 carp_demote_adj(V_carp_senderr_adj, 848 "send error"); 849 sc->sc_sendad_success = 0; 850 } else { 851 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) { 852 if (++sc->sc_sendad_success >= 853 CARP_SENDAD_MIN_SUCCESS) { 854 carp_demote_adj(-V_carp_senderr_adj, 855 "send ok"); 856 sc->sc_sendad_errors = 0; 857 } 858 } else 859 sc->sc_sendad_errors = 0; 860 } 861 } 862 #endif /* INET */ 863 #ifdef INET6 864 if (sc->sc_naddrs6) { 865 struct ip6_hdr *ip6; 866 867 m = m_gethdr(M_NOWAIT, MT_DATA); 868 if (m == NULL) { 869 CARPSTATS_INC(carps_onomem); 870 goto resched; 871 } 872 len = sizeof(*ip6) + sizeof(ch); 873 m->m_pkthdr.len = len; 874 m->m_pkthdr.rcvif = NULL; 875 m->m_len = len; 876 MH_ALIGN(m, m->m_len); 877 m->m_flags |= M_MCAST; 878 ip6 = mtod(m, struct ip6_hdr *); 879 bzero(ip6, sizeof(*ip6)); 880 ip6->ip6_vfc |= IPV6_VERSION; 881 ip6->ip6_hlim = CARP_DFLTTL; 882 ip6->ip6_nxt = IPPROTO_CARP; 883 bzero(&sa, sizeof(sa)); 884 885 /* set the source address */ 886 sa.sa_family = AF_INET6; 887 ifa = ifaof_ifpforaddr(&sa, sc->sc_carpdev); 888 if (ifa != NULL) { 889 bcopy(IFA_IN6(ifa), &ip6->ip6_src, 890 sizeof(struct in6_addr)); 891 ifa_free(ifa); 892 } else 893 /* This should never happen with IPv6. */ 894 bzero(&ip6->ip6_src, sizeof(struct in6_addr)); 895 896 /* Set the multicast destination. */ 897 ip6->ip6_dst.s6_addr16[0] = htons(0xff02); 898 ip6->ip6_dst.s6_addr8[15] = 0x12; 899 if (in6_setscope(&ip6->ip6_dst, sc->sc_carpdev, NULL) != 0) { 900 m_freem(m); 901 CARP_DEBUG("%s: in6_setscope failed\n", __func__); 902 goto resched; 903 } 904 905 ch_ptr = (struct carp_header *)(&ip6[1]); 906 bcopy(&ch, ch_ptr, sizeof(ch)); 907 if (carp_prepare_ad(m, sc, ch_ptr)) 908 goto resched; 909 910 m->m_data += sizeof(*ip6); 911 ch_ptr->carp_cksum = in_cksum(m, len - sizeof(*ip6)); 912 m->m_data -= sizeof(*ip6); 913 914 CARPSTATS_INC(carps_opackets6); 915 916 if (ip6_output(m, NULL, NULL, 0, 917 &sc->sc_carpdev->if_carp->cif_im6o, NULL, NULL)) { 918 if (sc->sc_sendad_errors < INT_MAX) 919 sc->sc_sendad_errors++; 920 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) 921 carp_demote_adj(V_carp_senderr_adj, 922 "send6 error"); 923 sc->sc_sendad_success = 0; 924 } else { 925 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) { 926 if (++sc->sc_sendad_success >= 927 CARP_SENDAD_MIN_SUCCESS) { 928 carp_demote_adj(-V_carp_senderr_adj, 929 "send6 ok"); 930 sc->sc_sendad_errors = 0; 931 } 932 } else 933 sc->sc_sendad_errors = 0; 934 } 935 } 936 #endif /* INET6 */ 937 938 resched: 939 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv), carp_send_ad, sc); 940 } 941 942 static void 943 carp_addroute(struct carp_softc *sc) 944 { 945 struct ifaddr *ifa; 946 947 CARP_FOREACH_IFA(sc, ifa) 948 carp_ifa_addroute(ifa); 949 } 950 951 static void 952 carp_ifa_addroute(struct ifaddr *ifa) 953 { 954 955 switch (ifa->ifa_addr->sa_family) { 956 #ifdef INET 957 case AF_INET: 958 in_addprefix(ifatoia(ifa), RTF_UP); 959 ifa_add_loopback_route(ifa, 960 (struct sockaddr *)&ifatoia(ifa)->ia_addr); 961 break; 962 #endif 963 #ifdef INET6 964 case AF_INET6: 965 ifa_add_loopback_route(ifa, 966 (struct sockaddr *)&ifatoia6(ifa)->ia_addr); 967 nd6_add_ifa_lle(ifatoia6(ifa)); 968 break; 969 #endif 970 } 971 } 972 973 static void 974 carp_delroute(struct carp_softc *sc) 975 { 976 struct ifaddr *ifa; 977 978 CARP_FOREACH_IFA(sc, ifa) 979 carp_ifa_delroute(ifa); 980 } 981 982 static void 983 carp_ifa_delroute(struct ifaddr *ifa) 984 { 985 986 switch (ifa->ifa_addr->sa_family) { 987 #ifdef INET 988 case AF_INET: 989 ifa_del_loopback_route(ifa, 990 (struct sockaddr *)&ifatoia(ifa)->ia_addr); 991 in_scrubprefix(ifatoia(ifa), LLE_STATIC); 992 break; 993 #endif 994 #ifdef INET6 995 case AF_INET6: 996 ifa_del_loopback_route(ifa, 997 (struct sockaddr *)&ifatoia6(ifa)->ia_addr); 998 nd6_rem_ifa_lle(ifatoia6(ifa)); 999 break; 1000 #endif 1001 } 1002 } 1003 1004 int 1005 carp_master(struct ifaddr *ifa) 1006 { 1007 struct carp_softc *sc = ifa->ifa_carp; 1008 1009 return (sc->sc_state == MASTER); 1010 } 1011 1012 #ifdef INET 1013 /* 1014 * Broadcast a gratuitous ARP request containing 1015 * the virtual router MAC address for each IP address 1016 * associated with the virtual router. 1017 */ 1018 static void 1019 carp_send_arp(struct carp_softc *sc) 1020 { 1021 struct ifaddr *ifa; 1022 1023 CARP_FOREACH_IFA(sc, ifa) 1024 if (ifa->ifa_addr->sa_family == AF_INET) 1025 arp_ifinit2(sc->sc_carpdev, ifa, LLADDR(&sc->sc_addr)); 1026 } 1027 1028 int 1029 carp_iamatch(struct ifaddr *ifa, uint8_t **enaddr) 1030 { 1031 struct carp_softc *sc = ifa->ifa_carp; 1032 1033 if (sc->sc_state == MASTER) { 1034 *enaddr = LLADDR(&sc->sc_addr); 1035 return (1); 1036 } 1037 1038 return (0); 1039 } 1040 #endif 1041 1042 #ifdef INET6 1043 static void 1044 carp_send_na(struct carp_softc *sc) 1045 { 1046 static struct in6_addr mcast = IN6ADDR_LINKLOCAL_ALLNODES_INIT; 1047 struct ifaddr *ifa; 1048 struct in6_addr *in6; 1049 1050 CARP_FOREACH_IFA(sc, ifa) { 1051 if (ifa->ifa_addr->sa_family != AF_INET6) 1052 continue; 1053 1054 in6 = IFA_IN6(ifa); 1055 nd6_na_output(sc->sc_carpdev, &mcast, in6, 1056 ND_NA_FLAG_OVERRIDE, 1, NULL); 1057 DELAY(1000); /* XXX */ 1058 } 1059 } 1060 1061 /* 1062 * Returns ifa in case it's a carp address and it is MASTER, or if the address 1063 * matches and is not a carp address. Returns NULL otherwise. 1064 */ 1065 struct ifaddr * 1066 carp_iamatch6(struct ifnet *ifp, struct in6_addr *taddr) 1067 { 1068 struct ifaddr *ifa; 1069 1070 ifa = NULL; 1071 IF_ADDR_RLOCK(ifp); 1072 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1073 if (ifa->ifa_addr->sa_family != AF_INET6) 1074 continue; 1075 if (!IN6_ARE_ADDR_EQUAL(taddr, IFA_IN6(ifa))) 1076 continue; 1077 if (ifa->ifa_carp && ifa->ifa_carp->sc_state != MASTER) 1078 ifa = NULL; 1079 else 1080 ifa_ref(ifa); 1081 break; 1082 } 1083 IF_ADDR_RUNLOCK(ifp); 1084 1085 return (ifa); 1086 } 1087 1088 caddr_t 1089 carp_macmatch6(struct ifnet *ifp, struct mbuf *m, const struct in6_addr *taddr) 1090 { 1091 struct ifaddr *ifa; 1092 1093 IF_ADDR_RLOCK(ifp); 1094 IFNET_FOREACH_IFA(ifp, ifa) 1095 if (ifa->ifa_addr->sa_family == AF_INET6 && 1096 IN6_ARE_ADDR_EQUAL(taddr, IFA_IN6(ifa))) { 1097 struct carp_softc *sc = ifa->ifa_carp; 1098 struct m_tag *mtag; 1099 1100 IF_ADDR_RUNLOCK(ifp); 1101 1102 mtag = m_tag_get(PACKET_TAG_CARP, 1103 sizeof(struct carp_softc *), M_NOWAIT); 1104 if (mtag == NULL) 1105 /* Better a bit than nothing. */ 1106 return (LLADDR(&sc->sc_addr)); 1107 1108 bcopy(&sc, mtag + 1, sizeof(sc)); 1109 m_tag_prepend(m, mtag); 1110 1111 return (LLADDR(&sc->sc_addr)); 1112 } 1113 IF_ADDR_RUNLOCK(ifp); 1114 1115 return (NULL); 1116 } 1117 #endif /* INET6 */ 1118 1119 int 1120 carp_forus(struct ifnet *ifp, u_char *dhost) 1121 { 1122 struct carp_softc *sc; 1123 uint8_t *ena = dhost; 1124 1125 if (ena[0] || ena[1] || ena[2] != 0x5e || ena[3] || ena[4] != 1) 1126 return (0); 1127 1128 CIF_LOCK(ifp->if_carp); 1129 IFNET_FOREACH_CARP(ifp, sc) { 1130 CARP_LOCK(sc); 1131 if (sc->sc_state == MASTER && !bcmp(dhost, LLADDR(&sc->sc_addr), 1132 ETHER_ADDR_LEN)) { 1133 CARP_UNLOCK(sc); 1134 CIF_UNLOCK(ifp->if_carp); 1135 return (1); 1136 } 1137 CARP_UNLOCK(sc); 1138 } 1139 CIF_UNLOCK(ifp->if_carp); 1140 1141 return (0); 1142 } 1143 1144 /* Master down timeout event, executed in callout context. */ 1145 static void 1146 carp_master_down(void *v) 1147 { 1148 struct carp_softc *sc = v; 1149 1150 CARP_LOCK_ASSERT(sc); 1151 1152 CURVNET_SET(sc->sc_carpdev->if_vnet); 1153 if (sc->sc_state == BACKUP) { 1154 CARP_LOG("VHID %u@%s: BACKUP -> MASTER (master down)\n", 1155 sc->sc_vhid, 1156 sc->sc_carpdev->if_xname); 1157 carp_master_down_locked(sc); 1158 } 1159 CURVNET_RESTORE(); 1160 1161 CARP_UNLOCK(sc); 1162 } 1163 1164 static void 1165 carp_master_down_locked(struct carp_softc *sc) 1166 { 1167 1168 CARP_LOCK_ASSERT(sc); 1169 1170 switch (sc->sc_state) { 1171 case BACKUP: 1172 carp_set_state(sc, MASTER); 1173 carp_send_ad_locked(sc); 1174 #ifdef INET 1175 carp_send_arp(sc); 1176 #endif 1177 #ifdef INET6 1178 carp_send_na(sc); 1179 #endif 1180 carp_setrun(sc, 0); 1181 carp_addroute(sc); 1182 break; 1183 case INIT: 1184 case MASTER: 1185 #ifdef INVARIANTS 1186 panic("carp: VHID %u@%s: master_down event in %s state\n", 1187 sc->sc_vhid, 1188 sc->sc_carpdev->if_xname, 1189 sc->sc_state ? "MASTER" : "INIT"); 1190 #endif 1191 break; 1192 } 1193 } 1194 1195 /* 1196 * When in backup state, af indicates whether to reset the master down timer 1197 * for v4 or v6. If it's set to zero, reset the ones which are already pending. 1198 */ 1199 static void 1200 carp_setrun(struct carp_softc *sc, sa_family_t af) 1201 { 1202 struct timeval tv; 1203 1204 CARP_LOCK_ASSERT(sc); 1205 1206 if ((sc->sc_carpdev->if_flags & IFF_UP) == 0 || 1207 sc->sc_carpdev->if_link_state != LINK_STATE_UP || 1208 (sc->sc_naddrs == 0 && sc->sc_naddrs6 == 0)) 1209 return; 1210 1211 switch (sc->sc_state) { 1212 case INIT: 1213 CARP_LOG("VHID %u@%s: INIT -> BACKUP\n", 1214 sc->sc_vhid, 1215 sc->sc_carpdev->if_xname); 1216 carp_set_state(sc, BACKUP); 1217 carp_setrun(sc, 0); 1218 break; 1219 case BACKUP: 1220 callout_stop(&sc->sc_ad_tmo); 1221 tv.tv_sec = 3 * sc->sc_advbase; 1222 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1223 switch (af) { 1224 #ifdef INET 1225 case AF_INET: 1226 callout_reset(&sc->sc_md_tmo, tvtohz(&tv), 1227 carp_master_down, sc); 1228 break; 1229 #endif 1230 #ifdef INET6 1231 case AF_INET6: 1232 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv), 1233 carp_master_down, sc); 1234 break; 1235 #endif 1236 default: 1237 #ifdef INET 1238 if (sc->sc_naddrs) 1239 callout_reset(&sc->sc_md_tmo, tvtohz(&tv), 1240 carp_master_down, sc); 1241 #endif 1242 #ifdef INET6 1243 if (sc->sc_naddrs6) 1244 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv), 1245 carp_master_down, sc); 1246 #endif 1247 break; 1248 } 1249 break; 1250 case MASTER: 1251 tv.tv_sec = sc->sc_advbase; 1252 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1253 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv), 1254 carp_send_ad, sc); 1255 break; 1256 } 1257 } 1258 1259 /* 1260 * Setup multicast structures. 1261 */ 1262 static int 1263 carp_multicast_setup(struct carp_if *cif, sa_family_t sa) 1264 { 1265 struct ifnet *ifp = cif->cif_ifp; 1266 int error = 0; 1267 1268 CIF_LOCK_ASSERT(cif); 1269 1270 switch (sa) { 1271 #ifdef INET 1272 case AF_INET: 1273 { 1274 struct ip_moptions *imo = &cif->cif_imo; 1275 struct in_addr addr; 1276 1277 if (imo->imo_membership) 1278 return (0); 1279 1280 imo->imo_membership = (struct in_multi **)malloc( 1281 (sizeof(struct in_multi *) * IP_MIN_MEMBERSHIPS), M_CARP, 1282 M_NOWAIT); 1283 if (imo->imo_membership == NULL) 1284 return (ENOMEM); 1285 imo->imo_mfilters = NULL; 1286 imo->imo_max_memberships = IP_MIN_MEMBERSHIPS; 1287 imo->imo_multicast_vif = -1; 1288 1289 addr.s_addr = htonl(INADDR_CARP_GROUP); 1290 if ((error = in_joingroup(ifp, &addr, NULL, 1291 &imo->imo_membership[0])) != 0) { 1292 free(imo->imo_membership, M_CARP); 1293 break; 1294 } 1295 imo->imo_num_memberships++; 1296 imo->imo_multicast_ifp = ifp; 1297 imo->imo_multicast_ttl = CARP_DFLTTL; 1298 imo->imo_multicast_loop = 0; 1299 break; 1300 } 1301 #endif 1302 #ifdef INET6 1303 case AF_INET6: 1304 { 1305 struct ip6_moptions *im6o = &cif->cif_im6o; 1306 struct in6_addr in6; 1307 struct in6_multi *in6m; 1308 1309 if (im6o->im6o_membership) 1310 return (0); 1311 1312 im6o->im6o_membership = (struct in6_multi **)malloc( 1313 (sizeof(struct in6_multi *) * IPV6_MIN_MEMBERSHIPS), M_CARP, 1314 M_ZERO | M_NOWAIT); 1315 if (im6o->im6o_membership == NULL) 1316 return (ENOMEM); 1317 im6o->im6o_mfilters = NULL; 1318 im6o->im6o_max_memberships = IPV6_MIN_MEMBERSHIPS; 1319 im6o->im6o_multicast_hlim = CARP_DFLTTL; 1320 im6o->im6o_multicast_ifp = ifp; 1321 1322 /* Join IPv6 CARP multicast group. */ 1323 bzero(&in6, sizeof(in6)); 1324 in6.s6_addr16[0] = htons(0xff02); 1325 in6.s6_addr8[15] = 0x12; 1326 if ((error = in6_setscope(&in6, ifp, NULL)) != 0) { 1327 free(im6o->im6o_membership, M_CARP); 1328 break; 1329 } 1330 in6m = NULL; 1331 if ((error = in6_mc_join(ifp, &in6, NULL, &in6m, 0)) != 0) { 1332 free(im6o->im6o_membership, M_CARP); 1333 break; 1334 } 1335 im6o->im6o_membership[0] = in6m; 1336 im6o->im6o_num_memberships++; 1337 1338 /* Join solicited multicast address. */ 1339 bzero(&in6, sizeof(in6)); 1340 in6.s6_addr16[0] = htons(0xff02); 1341 in6.s6_addr32[1] = 0; 1342 in6.s6_addr32[2] = htonl(1); 1343 in6.s6_addr32[3] = 0; 1344 in6.s6_addr8[12] = 0xff; 1345 if ((error = in6_setscope(&in6, ifp, NULL)) != 0) { 1346 in6_mc_leave(im6o->im6o_membership[0], NULL); 1347 free(im6o->im6o_membership, M_CARP); 1348 break; 1349 } 1350 in6m = NULL; 1351 if ((error = in6_mc_join(ifp, &in6, NULL, &in6m, 0)) != 0) { 1352 in6_mc_leave(im6o->im6o_membership[0], NULL); 1353 free(im6o->im6o_membership, M_CARP); 1354 break; 1355 } 1356 im6o->im6o_membership[1] = in6m; 1357 im6o->im6o_num_memberships++; 1358 break; 1359 } 1360 #endif 1361 } 1362 1363 return (error); 1364 } 1365 1366 /* 1367 * Free multicast structures. 1368 */ 1369 static void 1370 carp_multicast_cleanup(struct carp_if *cif, sa_family_t sa) 1371 { 1372 1373 CIF_LOCK_ASSERT(cif); 1374 switch (sa) { 1375 #ifdef INET 1376 case AF_INET: 1377 if (cif->cif_naddrs == 0) { 1378 struct ip_moptions *imo = &cif->cif_imo; 1379 1380 in_leavegroup(imo->imo_membership[0], NULL); 1381 KASSERT(imo->imo_mfilters == NULL, 1382 ("%s: imo_mfilters != NULL", __func__)); 1383 free(imo->imo_membership, M_CARP); 1384 imo->imo_membership = NULL; 1385 1386 } 1387 break; 1388 #endif 1389 #ifdef INET6 1390 case AF_INET6: 1391 if (cif->cif_naddrs6 == 0) { 1392 struct ip6_moptions *im6o = &cif->cif_im6o; 1393 1394 in6_mc_leave(im6o->im6o_membership[0], NULL); 1395 in6_mc_leave(im6o->im6o_membership[1], NULL); 1396 KASSERT(im6o->im6o_mfilters == NULL, 1397 ("%s: im6o_mfilters != NULL", __func__)); 1398 free(im6o->im6o_membership, M_CARP); 1399 im6o->im6o_membership = NULL; 1400 } 1401 break; 1402 #endif 1403 } 1404 } 1405 1406 int 1407 carp_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa) 1408 { 1409 struct m_tag *mtag; 1410 struct carp_softc *sc; 1411 1412 if (!sa) 1413 return (0); 1414 1415 switch (sa->sa_family) { 1416 #ifdef INET 1417 case AF_INET: 1418 break; 1419 #endif 1420 #ifdef INET6 1421 case AF_INET6: 1422 break; 1423 #endif 1424 default: 1425 return (0); 1426 } 1427 1428 mtag = m_tag_find(m, PACKET_TAG_CARP, NULL); 1429 if (mtag == NULL) 1430 return (0); 1431 1432 bcopy(mtag + 1, &sc, sizeof(sc)); 1433 1434 /* Set the source MAC address to the Virtual Router MAC Address. */ 1435 switch (ifp->if_type) { 1436 case IFT_ETHER: 1437 case IFT_BRIDGE: 1438 case IFT_L2VLAN: { 1439 struct ether_header *eh; 1440 1441 eh = mtod(m, struct ether_header *); 1442 eh->ether_shost[0] = 0; 1443 eh->ether_shost[1] = 0; 1444 eh->ether_shost[2] = 0x5e; 1445 eh->ether_shost[3] = 0; 1446 eh->ether_shost[4] = 1; 1447 eh->ether_shost[5] = sc->sc_vhid; 1448 } 1449 break; 1450 case IFT_FDDI: { 1451 struct fddi_header *fh; 1452 1453 fh = mtod(m, struct fddi_header *); 1454 fh->fddi_shost[0] = 0; 1455 fh->fddi_shost[1] = 0; 1456 fh->fddi_shost[2] = 0x5e; 1457 fh->fddi_shost[3] = 0; 1458 fh->fddi_shost[4] = 1; 1459 fh->fddi_shost[5] = sc->sc_vhid; 1460 } 1461 break; 1462 case IFT_ISO88025: { 1463 struct iso88025_header *th; 1464 th = mtod(m, struct iso88025_header *); 1465 th->iso88025_shost[0] = 3; 1466 th->iso88025_shost[1] = 0; 1467 th->iso88025_shost[2] = 0x40 >> (sc->sc_vhid - 1); 1468 th->iso88025_shost[3] = 0x40000 >> (sc->sc_vhid - 1); 1469 th->iso88025_shost[4] = 0; 1470 th->iso88025_shost[5] = 0; 1471 } 1472 break; 1473 default: 1474 printf("%s: carp is not supported for the %d interface type\n", 1475 ifp->if_xname, ifp->if_type); 1476 return (EOPNOTSUPP); 1477 } 1478 1479 return (0); 1480 } 1481 1482 static struct carp_softc* 1483 carp_alloc(struct ifnet *ifp) 1484 { 1485 struct carp_softc *sc; 1486 struct carp_if *cif; 1487 1488 if ((cif = ifp->if_carp) == NULL) 1489 cif = carp_alloc_if(ifp); 1490 1491 sc = malloc(sizeof(*sc), M_CARP, M_WAITOK|M_ZERO); 1492 1493 sc->sc_advbase = CARP_DFLTINTV; 1494 sc->sc_vhid = -1; /* required setting */ 1495 sc->sc_init_counter = 1; 1496 sc->sc_state = INIT; 1497 1498 sc->sc_ifasiz = sizeof(struct ifaddr *); 1499 sc->sc_ifas = malloc(sc->sc_ifasiz, M_CARP, M_WAITOK|M_ZERO); 1500 sc->sc_carpdev = ifp; 1501 1502 CARP_LOCK_INIT(sc); 1503 #ifdef INET 1504 callout_init_mtx(&sc->sc_md_tmo, &sc->sc_mtx, CALLOUT_RETURNUNLOCKED); 1505 #endif 1506 #ifdef INET6 1507 callout_init_mtx(&sc->sc_md6_tmo, &sc->sc_mtx, CALLOUT_RETURNUNLOCKED); 1508 #endif 1509 callout_init_mtx(&sc->sc_ad_tmo, &sc->sc_mtx, CALLOUT_RETURNUNLOCKED); 1510 1511 CIF_LOCK(cif); 1512 TAILQ_INSERT_TAIL(&cif->cif_vrs, sc, sc_list); 1513 CIF_UNLOCK(cif); 1514 1515 mtx_lock(&carp_mtx); 1516 LIST_INSERT_HEAD(&carp_list, sc, sc_next); 1517 mtx_unlock(&carp_mtx); 1518 1519 return (sc); 1520 } 1521 1522 static int 1523 carp_grow_ifas(struct carp_softc *sc) 1524 { 1525 struct ifaddr **new; 1526 1527 CARP_LOCK_ASSERT(sc); 1528 1529 new = malloc(sc->sc_ifasiz * 2, M_CARP, M_NOWAIT|M_ZERO); 1530 if (new == NULL) 1531 return (ENOMEM); 1532 bcopy(sc->sc_ifas, new, sc->sc_ifasiz); 1533 free(sc->sc_ifas, M_CARP); 1534 sc->sc_ifas = new; 1535 sc->sc_ifasiz *= 2; 1536 1537 return (0); 1538 } 1539 1540 static void 1541 carp_destroy(struct carp_softc *sc) 1542 { 1543 struct ifnet *ifp = sc->sc_carpdev; 1544 struct carp_if *cif = ifp->if_carp; 1545 1546 CIF_LOCK_ASSERT(cif); 1547 1548 TAILQ_REMOVE(&cif->cif_vrs, sc, sc_list); 1549 1550 mtx_lock(&carp_mtx); 1551 LIST_REMOVE(sc, sc_next); 1552 mtx_unlock(&carp_mtx); 1553 1554 CARP_LOCK(sc); 1555 if (sc->sc_suppress) 1556 carp_demote_adj(-V_carp_ifdown_adj, "vhid removed"); 1557 callout_drain(&sc->sc_ad_tmo); 1558 #ifdef INET 1559 callout_drain(&sc->sc_md_tmo); 1560 #endif 1561 #ifdef INET6 1562 callout_drain(&sc->sc_md6_tmo); 1563 #endif 1564 CARP_LOCK_DESTROY(sc); 1565 1566 free(sc->sc_ifas, M_CARP); 1567 free(sc, M_CARP); 1568 } 1569 1570 static struct carp_if* 1571 carp_alloc_if(struct ifnet *ifp) 1572 { 1573 struct carp_if *cif; 1574 int error; 1575 1576 cif = malloc(sizeof(*cif), M_CARP, M_WAITOK|M_ZERO); 1577 1578 if ((error = ifpromisc(ifp, 1)) != 0) 1579 printf("%s: ifpromisc(%s) failed: %d\n", 1580 __func__, ifp->if_xname, error); 1581 else 1582 cif->cif_flags |= CIF_PROMISC; 1583 1584 CIF_LOCK_INIT(cif); 1585 cif->cif_ifp = ifp; 1586 TAILQ_INIT(&cif->cif_vrs); 1587 1588 IF_ADDR_WLOCK(ifp); 1589 ifp->if_carp = cif; 1590 if_ref(ifp); 1591 IF_ADDR_WUNLOCK(ifp); 1592 1593 return (cif); 1594 } 1595 1596 static void 1597 carp_free_if(struct carp_if *cif) 1598 { 1599 struct ifnet *ifp = cif->cif_ifp; 1600 1601 CIF_LOCK_ASSERT(cif); 1602 KASSERT(TAILQ_EMPTY(&cif->cif_vrs), ("%s: softc list not empty", 1603 __func__)); 1604 1605 IF_ADDR_WLOCK(ifp); 1606 ifp->if_carp = NULL; 1607 IF_ADDR_WUNLOCK(ifp); 1608 1609 CIF_LOCK_DESTROY(cif); 1610 1611 if (cif->cif_flags & CIF_PROMISC) 1612 ifpromisc(ifp, 0); 1613 if_rele(ifp); 1614 1615 free(cif, M_CARP); 1616 } 1617 1618 static void 1619 carp_carprcp(struct carpreq *carpr, struct carp_softc *sc, int priv) 1620 { 1621 1622 CARP_LOCK(sc); 1623 carpr->carpr_state = sc->sc_state; 1624 carpr->carpr_vhid = sc->sc_vhid; 1625 carpr->carpr_advbase = sc->sc_advbase; 1626 carpr->carpr_advskew = sc->sc_advskew; 1627 if (priv) 1628 bcopy(sc->sc_key, carpr->carpr_key, sizeof(carpr->carpr_key)); 1629 else 1630 bzero(carpr->carpr_key, sizeof(carpr->carpr_key)); 1631 CARP_UNLOCK(sc); 1632 } 1633 1634 int 1635 carp_ioctl(struct ifreq *ifr, u_long cmd, struct thread *td) 1636 { 1637 struct carpreq carpr; 1638 struct ifnet *ifp; 1639 struct carp_softc *sc = NULL; 1640 int error = 0, locked = 0; 1641 1642 if ((error = copyin(ifr->ifr_data, &carpr, sizeof carpr))) 1643 return (error); 1644 1645 ifp = ifunit_ref(ifr->ifr_name); 1646 if (ifp == NULL) 1647 return (ENXIO); 1648 1649 switch (ifp->if_type) { 1650 case IFT_ETHER: 1651 case IFT_L2VLAN: 1652 case IFT_BRIDGE: 1653 case IFT_FDDI: 1654 case IFT_ISO88025: 1655 break; 1656 default: 1657 error = EOPNOTSUPP; 1658 goto out; 1659 } 1660 1661 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 1662 error = EADDRNOTAVAIL; 1663 goto out; 1664 } 1665 1666 switch (cmd) { 1667 case SIOCSVH: 1668 if ((error = priv_check(td, PRIV_NETINET_CARP))) 1669 break; 1670 if (carpr.carpr_vhid <= 0 || carpr.carpr_vhid > CARP_MAXVHID || 1671 carpr.carpr_advbase < 0 || carpr.carpr_advskew < 0) { 1672 error = EINVAL; 1673 break; 1674 } 1675 1676 if (ifp->if_carp) { 1677 CIF_LOCK(ifp->if_carp); 1678 IFNET_FOREACH_CARP(ifp, sc) 1679 if (sc->sc_vhid == carpr.carpr_vhid) 1680 break; 1681 CIF_UNLOCK(ifp->if_carp); 1682 } 1683 if (sc == NULL) { 1684 sc = carp_alloc(ifp); 1685 CARP_LOCK(sc); 1686 sc->sc_vhid = carpr.carpr_vhid; 1687 LLADDR(&sc->sc_addr)[0] = 0; 1688 LLADDR(&sc->sc_addr)[1] = 0; 1689 LLADDR(&sc->sc_addr)[2] = 0x5e; 1690 LLADDR(&sc->sc_addr)[3] = 0; 1691 LLADDR(&sc->sc_addr)[4] = 1; 1692 LLADDR(&sc->sc_addr)[5] = sc->sc_vhid; 1693 } else 1694 CARP_LOCK(sc); 1695 locked = 1; 1696 if (carpr.carpr_advbase > 0) { 1697 if (carpr.carpr_advbase > 255 || 1698 carpr.carpr_advbase < CARP_DFLTINTV) { 1699 error = EINVAL; 1700 break; 1701 } 1702 sc->sc_advbase = carpr.carpr_advbase; 1703 } 1704 if (carpr.carpr_advskew > 0) { 1705 if (carpr.carpr_advskew >= 255) { 1706 error = EINVAL; 1707 break; 1708 } 1709 sc->sc_advskew = carpr.carpr_advskew; 1710 } 1711 if (carpr.carpr_key[0] != '\0') { 1712 bcopy(carpr.carpr_key, sc->sc_key, sizeof(sc->sc_key)); 1713 carp_hmac_prepare(sc); 1714 } 1715 if (sc->sc_state != INIT && 1716 carpr.carpr_state != sc->sc_state) { 1717 switch (carpr.carpr_state) { 1718 case BACKUP: 1719 callout_stop(&sc->sc_ad_tmo); 1720 carp_set_state(sc, BACKUP); 1721 carp_setrun(sc, 0); 1722 carp_delroute(sc); 1723 break; 1724 case MASTER: 1725 carp_master_down_locked(sc); 1726 break; 1727 default: 1728 break; 1729 } 1730 } 1731 break; 1732 1733 case SIOCGVH: 1734 { 1735 int priveleged; 1736 1737 if (carpr.carpr_vhid < 0 || carpr.carpr_vhid > CARP_MAXVHID) { 1738 error = EINVAL; 1739 break; 1740 } 1741 if (carpr.carpr_count < 1) { 1742 error = EMSGSIZE; 1743 break; 1744 } 1745 if (ifp->if_carp == NULL) { 1746 error = ENOENT; 1747 break; 1748 } 1749 1750 priveleged = (priv_check(td, PRIV_NETINET_CARP) == 0); 1751 if (carpr.carpr_vhid != 0) { 1752 CIF_LOCK(ifp->if_carp); 1753 IFNET_FOREACH_CARP(ifp, sc) 1754 if (sc->sc_vhid == carpr.carpr_vhid) 1755 break; 1756 CIF_UNLOCK(ifp->if_carp); 1757 if (sc == NULL) { 1758 error = ENOENT; 1759 break; 1760 } 1761 carp_carprcp(&carpr, sc, priveleged); 1762 error = copyout(&carpr, ifr->ifr_data, sizeof(carpr)); 1763 } else { 1764 int i, count; 1765 1766 count = 0; 1767 CIF_LOCK(ifp->if_carp); 1768 IFNET_FOREACH_CARP(ifp, sc) 1769 count++; 1770 1771 if (count > carpr.carpr_count) { 1772 CIF_UNLOCK(ifp->if_carp); 1773 error = EMSGSIZE; 1774 break; 1775 } 1776 1777 i = 0; 1778 IFNET_FOREACH_CARP(ifp, sc) { 1779 carp_carprcp(&carpr, sc, priveleged); 1780 carpr.carpr_count = count; 1781 error = copyout(&carpr, ifr->ifr_data + 1782 (i * sizeof(carpr)), sizeof(carpr)); 1783 if (error) { 1784 CIF_UNLOCK(ifp->if_carp); 1785 break; 1786 } 1787 i++; 1788 } 1789 CIF_UNLOCK(ifp->if_carp); 1790 } 1791 break; 1792 } 1793 default: 1794 error = EINVAL; 1795 } 1796 1797 out: 1798 if (locked) 1799 CARP_UNLOCK(sc); 1800 if_rele(ifp); 1801 1802 return (error); 1803 } 1804 1805 static int 1806 carp_get_vhid(struct ifaddr *ifa) 1807 { 1808 1809 if (ifa == NULL || ifa->ifa_carp == NULL) 1810 return (0); 1811 1812 return (ifa->ifa_carp->sc_vhid); 1813 } 1814 1815 int 1816 carp_attach(struct ifaddr *ifa, int vhid) 1817 { 1818 struct ifnet *ifp = ifa->ifa_ifp; 1819 struct carp_if *cif = ifp->if_carp; 1820 struct carp_softc *sc; 1821 int index, error; 1822 1823 if (ifp->if_carp == NULL) 1824 return (ENOPROTOOPT); 1825 1826 switch (ifa->ifa_addr->sa_family) { 1827 #ifdef INET 1828 case AF_INET: 1829 #endif 1830 #ifdef INET6 1831 case AF_INET6: 1832 #endif 1833 break; 1834 default: 1835 return (EPROTOTYPE); 1836 } 1837 1838 CIF_LOCK(cif); 1839 IFNET_FOREACH_CARP(ifp, sc) 1840 if (sc->sc_vhid == vhid) 1841 break; 1842 if (sc == NULL) { 1843 CIF_UNLOCK(cif); 1844 return (ENOENT); 1845 } 1846 1847 if (ifa->ifa_carp) { 1848 if (ifa->ifa_carp->sc_vhid != vhid) 1849 carp_detach_locked(ifa); 1850 else { 1851 CIF_UNLOCK(cif); 1852 return (0); 1853 } 1854 } 1855 1856 error = carp_multicast_setup(cif, ifa->ifa_addr->sa_family); 1857 if (error) { 1858 CIF_FREE(cif); 1859 return (error); 1860 } 1861 1862 CARP_LOCK(sc); 1863 index = sc->sc_naddrs + sc->sc_naddrs6 + 1; 1864 if (index > sc->sc_ifasiz / sizeof(struct ifaddr *)) 1865 if ((error = carp_grow_ifas(sc)) != 0) { 1866 carp_multicast_cleanup(cif, 1867 ifa->ifa_addr->sa_family); 1868 CARP_UNLOCK(sc); 1869 CIF_FREE(cif); 1870 return (error); 1871 } 1872 1873 switch (ifa->ifa_addr->sa_family) { 1874 #ifdef INET 1875 case AF_INET: 1876 cif->cif_naddrs++; 1877 sc->sc_naddrs++; 1878 break; 1879 #endif 1880 #ifdef INET6 1881 case AF_INET6: 1882 cif->cif_naddrs6++; 1883 sc->sc_naddrs6++; 1884 break; 1885 #endif 1886 } 1887 1888 ifa_ref(ifa); 1889 sc->sc_ifas[index - 1] = ifa; 1890 ifa->ifa_carp = sc; 1891 1892 carp_hmac_prepare(sc); 1893 carp_sc_state(sc); 1894 1895 CARP_UNLOCK(sc); 1896 CIF_UNLOCK(cif); 1897 1898 return (0); 1899 } 1900 1901 void 1902 carp_detach(struct ifaddr *ifa) 1903 { 1904 struct ifnet *ifp = ifa->ifa_ifp; 1905 struct carp_if *cif = ifp->if_carp; 1906 1907 CIF_LOCK(cif); 1908 carp_detach_locked(ifa); 1909 CIF_FREE(cif); 1910 } 1911 1912 static void 1913 carp_detach_locked(struct ifaddr *ifa) 1914 { 1915 struct ifnet *ifp = ifa->ifa_ifp; 1916 struct carp_if *cif = ifp->if_carp; 1917 struct carp_softc *sc = ifa->ifa_carp; 1918 int i, index; 1919 1920 KASSERT(sc != NULL, ("%s: %p not attached", __func__, ifa)); 1921 1922 CIF_LOCK_ASSERT(cif); 1923 CARP_LOCK(sc); 1924 1925 /* Shift array. */ 1926 index = sc->sc_naddrs + sc->sc_naddrs6; 1927 for (i = 0; i < index; i++) 1928 if (sc->sc_ifas[i] == ifa) 1929 break; 1930 KASSERT(i < index, ("%s: %p no backref", __func__, ifa)); 1931 for (; i < index - 1; i++) 1932 sc->sc_ifas[i] = sc->sc_ifas[i+1]; 1933 sc->sc_ifas[index - 1] = NULL; 1934 1935 switch (ifa->ifa_addr->sa_family) { 1936 #ifdef INET 1937 case AF_INET: 1938 cif->cif_naddrs--; 1939 sc->sc_naddrs--; 1940 break; 1941 #endif 1942 #ifdef INET6 1943 case AF_INET6: 1944 cif->cif_naddrs6--; 1945 sc->sc_naddrs6--; 1946 break; 1947 #endif 1948 } 1949 1950 carp_ifa_delroute(ifa); 1951 carp_multicast_cleanup(cif, ifa->ifa_addr->sa_family); 1952 1953 ifa->ifa_carp = NULL; 1954 ifa_free(ifa); 1955 1956 carp_hmac_prepare(sc); 1957 carp_sc_state(sc); 1958 1959 if (sc->sc_naddrs == 0 && sc->sc_naddrs6 == 0) { 1960 CARP_UNLOCK(sc); 1961 carp_destroy(sc); 1962 } else 1963 CARP_UNLOCK(sc); 1964 } 1965 1966 static void 1967 carp_set_state(struct carp_softc *sc, int state) 1968 { 1969 1970 CARP_LOCK_ASSERT(sc); 1971 1972 if (sc->sc_state != state) { 1973 const char *carp_states[] = { CARP_STATES }; 1974 char subsys[IFNAMSIZ+5]; 1975 1976 sc->sc_state = state; 1977 1978 snprintf(subsys, IFNAMSIZ+5, "%u@%s", sc->sc_vhid, 1979 sc->sc_carpdev->if_xname); 1980 devctl_notify("CARP", subsys, carp_states[state], NULL); 1981 } 1982 } 1983 1984 static void 1985 carp_linkstate(struct ifnet *ifp) 1986 { 1987 struct carp_softc *sc; 1988 1989 CIF_LOCK(ifp->if_carp); 1990 IFNET_FOREACH_CARP(ifp, sc) { 1991 CARP_LOCK(sc); 1992 carp_sc_state(sc); 1993 CARP_UNLOCK(sc); 1994 } 1995 CIF_UNLOCK(ifp->if_carp); 1996 } 1997 1998 static void 1999 carp_sc_state(struct carp_softc *sc) 2000 { 2001 2002 CARP_LOCK_ASSERT(sc); 2003 2004 if (sc->sc_carpdev->if_link_state != LINK_STATE_UP || 2005 !(sc->sc_carpdev->if_flags & IFF_UP)) { 2006 callout_stop(&sc->sc_ad_tmo); 2007 #ifdef INET 2008 callout_stop(&sc->sc_md_tmo); 2009 #endif 2010 #ifdef INET6 2011 callout_stop(&sc->sc_md6_tmo); 2012 #endif 2013 carp_set_state(sc, INIT); 2014 carp_setrun(sc, 0); 2015 if (!sc->sc_suppress) 2016 carp_demote_adj(V_carp_ifdown_adj, "interface down"); 2017 sc->sc_suppress = 1; 2018 } else { 2019 carp_set_state(sc, INIT); 2020 carp_setrun(sc, 0); 2021 if (sc->sc_suppress) 2022 carp_demote_adj(-V_carp_ifdown_adj, "interface up"); 2023 sc->sc_suppress = 0; 2024 } 2025 } 2026 2027 static void 2028 carp_demote_adj(int adj, char *reason) 2029 { 2030 atomic_add_int(&V_carp_demotion, adj); 2031 CARP_LOG("demoted by %d to %d (%s)\n", adj, V_carp_demotion, reason); 2032 taskqueue_enqueue(taskqueue_swi, &carp_sendall_task); 2033 } 2034 2035 static int 2036 carp_demote_adj_sysctl(SYSCTL_HANDLER_ARGS) 2037 { 2038 int new, error; 2039 2040 new = V_carp_demotion; 2041 error = sysctl_handle_int(oidp, &new, 0, req); 2042 if (error || !req->newptr) 2043 return (error); 2044 2045 carp_demote_adj(new, "sysctl"); 2046 2047 return (0); 2048 } 2049 2050 #ifdef INET 2051 extern struct domain inetdomain; 2052 static struct protosw in_carp_protosw = { 2053 .pr_type = SOCK_RAW, 2054 .pr_domain = &inetdomain, 2055 .pr_protocol = IPPROTO_CARP, 2056 .pr_flags = PR_ATOMIC|PR_ADDR, 2057 .pr_input = carp_input, 2058 .pr_output = (pr_output_t *)rip_output, 2059 .pr_ctloutput = rip_ctloutput, 2060 .pr_usrreqs = &rip_usrreqs 2061 }; 2062 #endif 2063 2064 #ifdef INET6 2065 extern struct domain inet6domain; 2066 static struct ip6protosw in6_carp_protosw = { 2067 .pr_type = SOCK_RAW, 2068 .pr_domain = &inet6domain, 2069 .pr_protocol = IPPROTO_CARP, 2070 .pr_flags = PR_ATOMIC|PR_ADDR, 2071 .pr_input = carp6_input, 2072 .pr_output = rip6_output, 2073 .pr_ctloutput = rip6_ctloutput, 2074 .pr_usrreqs = &rip6_usrreqs 2075 }; 2076 #endif 2077 2078 static void 2079 carp_mod_cleanup(void) 2080 { 2081 2082 #ifdef INET 2083 if (proto_reg[CARP_INET] == 0) { 2084 (void)ipproto_unregister(IPPROTO_CARP); 2085 pf_proto_unregister(PF_INET, IPPROTO_CARP, SOCK_RAW); 2086 proto_reg[CARP_INET] = -1; 2087 } 2088 carp_iamatch_p = NULL; 2089 #endif 2090 #ifdef INET6 2091 if (proto_reg[CARP_INET6] == 0) { 2092 (void)ip6proto_unregister(IPPROTO_CARP); 2093 pf_proto_unregister(PF_INET6, IPPROTO_CARP, SOCK_RAW); 2094 proto_reg[CARP_INET6] = -1; 2095 } 2096 carp_iamatch6_p = NULL; 2097 carp_macmatch6_p = NULL; 2098 #endif 2099 carp_ioctl_p = NULL; 2100 carp_attach_p = NULL; 2101 carp_detach_p = NULL; 2102 carp_get_vhid_p = NULL; 2103 carp_linkstate_p = NULL; 2104 carp_forus_p = NULL; 2105 carp_output_p = NULL; 2106 carp_demote_adj_p = NULL; 2107 carp_master_p = NULL; 2108 mtx_unlock(&carp_mtx); 2109 taskqueue_drain(taskqueue_swi, &carp_sendall_task); 2110 mtx_destroy(&carp_mtx); 2111 } 2112 2113 static int 2114 carp_mod_load(void) 2115 { 2116 int err; 2117 2118 mtx_init(&carp_mtx, "carp_mtx", NULL, MTX_DEF); 2119 LIST_INIT(&carp_list); 2120 carp_get_vhid_p = carp_get_vhid; 2121 carp_forus_p = carp_forus; 2122 carp_output_p = carp_output; 2123 carp_linkstate_p = carp_linkstate; 2124 carp_ioctl_p = carp_ioctl; 2125 carp_attach_p = carp_attach; 2126 carp_detach_p = carp_detach; 2127 carp_demote_adj_p = carp_demote_adj; 2128 carp_master_p = carp_master; 2129 #ifdef INET6 2130 carp_iamatch6_p = carp_iamatch6; 2131 carp_macmatch6_p = carp_macmatch6; 2132 proto_reg[CARP_INET6] = pf_proto_register(PF_INET6, 2133 (struct protosw *)&in6_carp_protosw); 2134 if (proto_reg[CARP_INET6]) { 2135 printf("carp: error %d attaching to PF_INET6\n", 2136 proto_reg[CARP_INET6]); 2137 carp_mod_cleanup(); 2138 return (proto_reg[CARP_INET6]); 2139 } 2140 err = ip6proto_register(IPPROTO_CARP); 2141 if (err) { 2142 printf("carp: error %d registering with INET6\n", err); 2143 carp_mod_cleanup(); 2144 return (err); 2145 } 2146 #endif 2147 #ifdef INET 2148 carp_iamatch_p = carp_iamatch; 2149 proto_reg[CARP_INET] = pf_proto_register(PF_INET, &in_carp_protosw); 2150 if (proto_reg[CARP_INET]) { 2151 printf("carp: error %d attaching to PF_INET\n", 2152 proto_reg[CARP_INET]); 2153 carp_mod_cleanup(); 2154 return (proto_reg[CARP_INET]); 2155 } 2156 err = ipproto_register(IPPROTO_CARP); 2157 if (err) { 2158 printf("carp: error %d registering with INET\n", err); 2159 carp_mod_cleanup(); 2160 return (err); 2161 } 2162 #endif 2163 return (0); 2164 } 2165 2166 static int 2167 carp_modevent(module_t mod, int type, void *data) 2168 { 2169 switch (type) { 2170 case MOD_LOAD: 2171 return carp_mod_load(); 2172 /* NOTREACHED */ 2173 case MOD_UNLOAD: 2174 mtx_lock(&carp_mtx); 2175 if (LIST_EMPTY(&carp_list)) 2176 carp_mod_cleanup(); 2177 else { 2178 mtx_unlock(&carp_mtx); 2179 return (EBUSY); 2180 } 2181 break; 2182 2183 default: 2184 return (EINVAL); 2185 } 2186 2187 return (0); 2188 } 2189 2190 static moduledata_t carp_mod = { 2191 "carp", 2192 carp_modevent, 2193 0 2194 }; 2195 2196 DECLARE_MODULE(carp, carp_mod, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY); 2197