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