1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 2002 Michael Shalayeff. 5 * Copyright (c) 2003 Ryan McBride. 6 * Copyright (c) 2011 Gleb Smirnoff <glebius@FreeBSD.org> 7 * All rights reserved. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 19 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 20 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 21 * IN NO EVENT SHALL THE AUTHOR OR HIS RELATIVES BE LIABLE FOR ANY DIRECT, 22 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES 23 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 24 * SERVICES; LOSS OF MIND, USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, 26 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING 27 * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF 28 * THE POSSIBILITY OF SUCH DAMAGE. 29 */ 30 31 #include <sys/cdefs.h> 32 __FBSDID("$FreeBSD$"); 33 34 #include "opt_bpf.h" 35 #include "opt_inet.h" 36 #include "opt_inet6.h" 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/bus.h> 41 #include <sys/jail.h> 42 #include <sys/kernel.h> 43 #include <sys/limits.h> 44 #include <sys/malloc.h> 45 #include <sys/mbuf.h> 46 #include <sys/module.h> 47 #include <sys/priv.h> 48 #include <sys/proc.h> 49 #include <sys/protosw.h> 50 #include <sys/socket.h> 51 #include <sys/sockio.h> 52 #include <sys/sysctl.h> 53 #include <sys/syslog.h> 54 #include <sys/taskqueue.h> 55 #include <sys/counter.h> 56 57 #include <net/ethernet.h> 58 #include <net/if.h> 59 #include <net/if_var.h> 60 #include <net/if_dl.h> 61 #include <net/if_llatbl.h> 62 #include <net/if_types.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() or carp_sx. 179 * To 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 VNET_DEFINE_STATIC(int, carp_allow) = 1; 191 #define V_carp_allow VNET(carp_allow) 192 193 /* Set DSCP in outgoing CARP packets. */ 194 VNET_DEFINE_STATIC(int, carp_dscp) = 56; 195 #define V_carp_dscp VNET(carp_dscp) 196 197 /* Preempt slower nodes. */ 198 VNET_DEFINE_STATIC(int, carp_preempt) = 0; 199 #define V_carp_preempt VNET(carp_preempt) 200 201 /* Log level. */ 202 VNET_DEFINE_STATIC(int, carp_log) = 1; 203 #define V_carp_log VNET(carp_log) 204 205 /* Global advskew demotion. */ 206 VNET_DEFINE_STATIC(int, carp_demotion) = 0; 207 #define V_carp_demotion VNET(carp_demotion) 208 209 /* Send error demotion factor. */ 210 VNET_DEFINE_STATIC(int, carp_senderr_adj) = CARP_MAXSKEW; 211 #define V_carp_senderr_adj VNET(carp_senderr_adj) 212 213 /* Iface down demotion factor. */ 214 VNET_DEFINE_STATIC(int, carp_ifdown_adj) = CARP_MAXSKEW; 215 #define V_carp_ifdown_adj VNET(carp_ifdown_adj) 216 217 static int carp_allow_sysctl(SYSCTL_HANDLER_ARGS); 218 static int carp_dscp_sysctl(SYSCTL_HANDLER_ARGS); 219 static int carp_demote_adj_sysctl(SYSCTL_HANDLER_ARGS); 220 221 SYSCTL_NODE(_net_inet, IPPROTO_CARP, carp, CTLFLAG_RW, 0, "CARP"); 222 SYSCTL_PROC(_net_inet_carp, OID_AUTO, allow, 223 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, 0, 0, carp_allow_sysctl, "I", 224 "Accept incoming CARP packets"); 225 SYSCTL_PROC(_net_inet_carp, OID_AUTO, dscp, 226 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, 0, 0, carp_dscp_sysctl, "I", 227 "DSCP value for carp packets"); 228 SYSCTL_INT(_net_inet_carp, OID_AUTO, preempt, CTLFLAG_VNET | CTLFLAG_RW, 229 &VNET_NAME(carp_preempt), 0, "High-priority backup preemption mode"); 230 SYSCTL_INT(_net_inet_carp, OID_AUTO, log, CTLFLAG_VNET | CTLFLAG_RW, 231 &VNET_NAME(carp_log), 0, "CARP log level"); 232 SYSCTL_PROC(_net_inet_carp, OID_AUTO, demotion, 233 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, 234 0, 0, carp_demote_adj_sysctl, "I", 235 "Adjust demotion factor (skew of advskew)"); 236 SYSCTL_INT(_net_inet_carp, OID_AUTO, senderr_demotion_factor, 237 CTLFLAG_VNET | CTLFLAG_RW, 238 &VNET_NAME(carp_senderr_adj), 0, "Send error demotion factor adjustment"); 239 SYSCTL_INT(_net_inet_carp, OID_AUTO, ifdown_demotion_factor, 240 CTLFLAG_VNET | CTLFLAG_RW, 241 &VNET_NAME(carp_ifdown_adj), 0, 242 "Interface down demotion factor adjustment"); 243 244 VNET_PCPUSTAT_DEFINE(struct carpstats, carpstats); 245 VNET_PCPUSTAT_SYSINIT(carpstats); 246 VNET_PCPUSTAT_SYSUNINIT(carpstats); 247 248 #define CARPSTATS_ADD(name, val) \ 249 counter_u64_add(VNET(carpstats)[offsetof(struct carpstats, name) / \ 250 sizeof(uint64_t)], (val)) 251 #define CARPSTATS_INC(name) CARPSTATS_ADD(name, 1) 252 253 SYSCTL_VNET_PCPUSTAT(_net_inet_carp, OID_AUTO, stats, struct carpstats, 254 carpstats, "CARP statistics (struct carpstats, netinet/ip_carp.h)"); 255 256 #define CARP_LOCK_INIT(sc) mtx_init(&(sc)->sc_mtx, "carp_softc", \ 257 NULL, MTX_DEF) 258 #define CARP_LOCK_DESTROY(sc) mtx_destroy(&(sc)->sc_mtx) 259 #define CARP_LOCK_ASSERT(sc) mtx_assert(&(sc)->sc_mtx, MA_OWNED) 260 #define CARP_LOCK(sc) mtx_lock(&(sc)->sc_mtx) 261 #define CARP_UNLOCK(sc) mtx_unlock(&(sc)->sc_mtx) 262 #define CIF_LOCK_INIT(cif) mtx_init(&(cif)->cif_mtx, "carp_if", \ 263 NULL, MTX_DEF) 264 #define CIF_LOCK_DESTROY(cif) mtx_destroy(&(cif)->cif_mtx) 265 #define CIF_LOCK_ASSERT(cif) mtx_assert(&(cif)->cif_mtx, MA_OWNED) 266 #define CIF_LOCK(cif) mtx_lock(&(cif)->cif_mtx) 267 #define CIF_UNLOCK(cif) mtx_unlock(&(cif)->cif_mtx) 268 #define CIF_FREE(cif) do { \ 269 CIF_LOCK(cif); \ 270 if (TAILQ_EMPTY(&(cif)->cif_vrs)) \ 271 carp_free_if(cif); \ 272 else \ 273 CIF_UNLOCK(cif); \ 274 } while (0) 275 276 #define CARP_LOG(...) do { \ 277 if (V_carp_log > 0) \ 278 log(LOG_INFO, "carp: " __VA_ARGS__); \ 279 } while (0) 280 281 #define CARP_DEBUG(...) do { \ 282 if (V_carp_log > 1) \ 283 log(LOG_DEBUG, __VA_ARGS__); \ 284 } while (0) 285 286 #define IFNET_FOREACH_IFA(ifp, ifa) \ 287 CK_STAILQ_FOREACH((ifa), &(ifp)->if_addrhead, ifa_link) \ 288 if ((ifa)->ifa_carp != NULL) 289 290 #define CARP_FOREACH_IFA(sc, ifa) \ 291 CARP_LOCK_ASSERT(sc); \ 292 for (int _i = 0; \ 293 _i < (sc)->sc_naddrs + (sc)->sc_naddrs6 && \ 294 ((ifa) = sc->sc_ifas[_i]) != NULL; \ 295 ++_i) 296 297 #define IFNET_FOREACH_CARP(ifp, sc) \ 298 KASSERT(mtx_owned(&ifp->if_carp->cif_mtx) || \ 299 sx_xlocked(&carp_sx), ("cif_vrs not locked")); \ 300 TAILQ_FOREACH((sc), &(ifp)->if_carp->cif_vrs, sc_list) 301 302 #define DEMOTE_ADVSKEW(sc) \ 303 (((sc)->sc_advskew + V_carp_demotion > CARP_MAXSKEW) ? \ 304 CARP_MAXSKEW : ((sc)->sc_advskew + V_carp_demotion)) 305 306 static void carp_input_c(struct mbuf *, struct carp_header *, sa_family_t); 307 static struct carp_softc 308 *carp_alloc(struct ifnet *); 309 static void carp_destroy(struct carp_softc *); 310 static struct carp_if 311 *carp_alloc_if(struct ifnet *); 312 static void carp_free_if(struct carp_if *); 313 static void carp_set_state(struct carp_softc *, int, const char* reason); 314 static void carp_sc_state(struct carp_softc *); 315 static void carp_setrun(struct carp_softc *, sa_family_t); 316 static void carp_master_down(void *); 317 static void carp_master_down_locked(struct carp_softc *, 318 const char* reason); 319 static void carp_send_ad(void *); 320 static void carp_send_ad_locked(struct carp_softc *); 321 static void carp_addroute(struct carp_softc *); 322 static void carp_ifa_addroute(struct ifaddr *); 323 static void carp_delroute(struct carp_softc *); 324 static void carp_ifa_delroute(struct ifaddr *); 325 static void carp_send_ad_all(void *, int); 326 static void carp_demote_adj(int, char *); 327 328 static LIST_HEAD(, carp_softc) carp_list; 329 static struct mtx carp_mtx; 330 static struct sx carp_sx; 331 static struct task carp_sendall_task = 332 TASK_INITIALIZER(0, carp_send_ad_all, NULL); 333 334 static void 335 carp_hmac_prepare(struct carp_softc *sc) 336 { 337 uint8_t version = CARP_VERSION, type = CARP_ADVERTISEMENT; 338 uint8_t vhid = sc->sc_vhid & 0xff; 339 struct ifaddr *ifa; 340 int i, found; 341 #ifdef INET 342 struct in_addr last, cur, in; 343 #endif 344 #ifdef INET6 345 struct in6_addr last6, cur6, in6; 346 #endif 347 348 CARP_LOCK_ASSERT(sc); 349 350 /* Compute ipad from key. */ 351 bzero(sc->sc_pad, sizeof(sc->sc_pad)); 352 bcopy(sc->sc_key, sc->sc_pad, sizeof(sc->sc_key)); 353 for (i = 0; i < sizeof(sc->sc_pad); i++) 354 sc->sc_pad[i] ^= 0x36; 355 356 /* Precompute first part of inner hash. */ 357 SHA1Init(&sc->sc_sha1); 358 SHA1Update(&sc->sc_sha1, sc->sc_pad, sizeof(sc->sc_pad)); 359 SHA1Update(&sc->sc_sha1, (void *)&version, sizeof(version)); 360 SHA1Update(&sc->sc_sha1, (void *)&type, sizeof(type)); 361 SHA1Update(&sc->sc_sha1, (void *)&vhid, sizeof(vhid)); 362 #ifdef INET 363 cur.s_addr = 0; 364 do { 365 found = 0; 366 last = cur; 367 cur.s_addr = 0xffffffff; 368 CARP_FOREACH_IFA(sc, ifa) { 369 in.s_addr = ifatoia(ifa)->ia_addr.sin_addr.s_addr; 370 if (ifa->ifa_addr->sa_family == AF_INET && 371 ntohl(in.s_addr) > ntohl(last.s_addr) && 372 ntohl(in.s_addr) < ntohl(cur.s_addr)) { 373 cur.s_addr = in.s_addr; 374 found++; 375 } 376 } 377 if (found) 378 SHA1Update(&sc->sc_sha1, (void *)&cur, sizeof(cur)); 379 } while (found); 380 #endif /* INET */ 381 #ifdef INET6 382 memset(&cur6, 0, sizeof(cur6)); 383 do { 384 found = 0; 385 last6 = cur6; 386 memset(&cur6, 0xff, sizeof(cur6)); 387 CARP_FOREACH_IFA(sc, ifa) { 388 in6 = ifatoia6(ifa)->ia_addr.sin6_addr; 389 if (IN6_IS_SCOPE_EMBED(&in6)) 390 in6.s6_addr16[1] = 0; 391 if (ifa->ifa_addr->sa_family == AF_INET6 && 392 memcmp(&in6, &last6, sizeof(in6)) > 0 && 393 memcmp(&in6, &cur6, sizeof(in6)) < 0) { 394 cur6 = in6; 395 found++; 396 } 397 } 398 if (found) 399 SHA1Update(&sc->sc_sha1, (void *)&cur6, sizeof(cur6)); 400 } while (found); 401 #endif /* INET6 */ 402 403 /* convert ipad to opad */ 404 for (i = 0; i < sizeof(sc->sc_pad); i++) 405 sc->sc_pad[i] ^= 0x36 ^ 0x5c; 406 } 407 408 static void 409 carp_hmac_generate(struct carp_softc *sc, uint32_t counter[2], 410 unsigned char md[20]) 411 { 412 SHA1_CTX sha1ctx; 413 414 CARP_LOCK_ASSERT(sc); 415 416 /* fetch first half of inner hash */ 417 bcopy(&sc->sc_sha1, &sha1ctx, sizeof(sha1ctx)); 418 419 SHA1Update(&sha1ctx, (void *)counter, sizeof(sc->sc_counter)); 420 SHA1Final(md, &sha1ctx); 421 422 /* outer hash */ 423 SHA1Init(&sha1ctx); 424 SHA1Update(&sha1ctx, sc->sc_pad, sizeof(sc->sc_pad)); 425 SHA1Update(&sha1ctx, md, 20); 426 SHA1Final(md, &sha1ctx); 427 } 428 429 static int 430 carp_hmac_verify(struct carp_softc *sc, uint32_t counter[2], 431 unsigned char md[20]) 432 { 433 unsigned char md2[20]; 434 435 CARP_LOCK_ASSERT(sc); 436 437 carp_hmac_generate(sc, counter, md2); 438 439 return (bcmp(md, md2, sizeof(md2))); 440 } 441 442 /* 443 * process input packet. 444 * we have rearranged checks order compared to the rfc, 445 * but it seems more efficient this way or not possible otherwise. 446 */ 447 #ifdef INET 448 int 449 carp_input(struct mbuf **mp, int *offp, int proto) 450 { 451 struct mbuf *m = *mp; 452 struct ip *ip = mtod(m, struct ip *); 453 struct carp_header *ch; 454 int iplen, len; 455 456 iplen = *offp; 457 *mp = NULL; 458 459 CARPSTATS_INC(carps_ipackets); 460 461 if (!V_carp_allow) { 462 m_freem(m); 463 return (IPPROTO_DONE); 464 } 465 466 /* verify that the IP TTL is 255. */ 467 if (ip->ip_ttl != CARP_DFLTTL) { 468 CARPSTATS_INC(carps_badttl); 469 CARP_DEBUG("%s: received ttl %d != 255 on %s\n", __func__, 470 ip->ip_ttl, 471 m->m_pkthdr.rcvif->if_xname); 472 m_freem(m); 473 return (IPPROTO_DONE); 474 } 475 476 iplen = ip->ip_hl << 2; 477 478 if (m->m_pkthdr.len < iplen + sizeof(*ch)) { 479 CARPSTATS_INC(carps_badlen); 480 CARP_DEBUG("%s: received len %zd < sizeof(struct carp_header) " 481 "on %s\n", __func__, m->m_len - sizeof(struct ip), 482 m->m_pkthdr.rcvif->if_xname); 483 m_freem(m); 484 return (IPPROTO_DONE); 485 } 486 487 if (iplen + sizeof(*ch) < m->m_len) { 488 if ((m = m_pullup(m, iplen + sizeof(*ch))) == NULL) { 489 CARPSTATS_INC(carps_hdrops); 490 CARP_DEBUG("%s: pullup failed\n", __func__); 491 return (IPPROTO_DONE); 492 } 493 ip = mtod(m, struct ip *); 494 } 495 ch = (struct carp_header *)((char *)ip + iplen); 496 497 /* 498 * verify that the received packet length is 499 * equal to the CARP header 500 */ 501 len = iplen + sizeof(*ch); 502 if (len > m->m_pkthdr.len) { 503 CARPSTATS_INC(carps_badlen); 504 CARP_DEBUG("%s: packet too short %d on %s\n", __func__, 505 m->m_pkthdr.len, 506 m->m_pkthdr.rcvif->if_xname); 507 m_freem(m); 508 return (IPPROTO_DONE); 509 } 510 511 if ((m = m_pullup(m, len)) == NULL) { 512 CARPSTATS_INC(carps_hdrops); 513 return (IPPROTO_DONE); 514 } 515 ip = mtod(m, struct ip *); 516 ch = (struct carp_header *)((char *)ip + iplen); 517 518 /* verify the CARP checksum */ 519 m->m_data += iplen; 520 if (in_cksum(m, len - iplen)) { 521 CARPSTATS_INC(carps_badsum); 522 CARP_DEBUG("%s: checksum failed on %s\n", __func__, 523 m->m_pkthdr.rcvif->if_xname); 524 m_freem(m); 525 return (IPPROTO_DONE); 526 } 527 m->m_data -= iplen; 528 529 carp_input_c(m, ch, AF_INET); 530 return (IPPROTO_DONE); 531 } 532 #endif 533 534 #ifdef INET6 535 int 536 carp6_input(struct mbuf **mp, int *offp, int proto) 537 { 538 struct mbuf *m = *mp; 539 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 540 struct carp_header *ch; 541 u_int len; 542 543 CARPSTATS_INC(carps_ipackets6); 544 545 if (!V_carp_allow) { 546 m_freem(m); 547 return (IPPROTO_DONE); 548 } 549 550 /* check if received on a valid carp interface */ 551 if (m->m_pkthdr.rcvif->if_carp == NULL) { 552 CARPSTATS_INC(carps_badif); 553 CARP_DEBUG("%s: packet received on non-carp interface: %s\n", 554 __func__, m->m_pkthdr.rcvif->if_xname); 555 m_freem(m); 556 return (IPPROTO_DONE); 557 } 558 559 /* verify that the IP TTL is 255 */ 560 if (ip6->ip6_hlim != CARP_DFLTTL) { 561 CARPSTATS_INC(carps_badttl); 562 CARP_DEBUG("%s: received ttl %d != 255 on %s\n", __func__, 563 ip6->ip6_hlim, m->m_pkthdr.rcvif->if_xname); 564 m_freem(m); 565 return (IPPROTO_DONE); 566 } 567 568 /* verify that we have a complete carp packet */ 569 len = m->m_len; 570 IP6_EXTHDR_GET(ch, struct carp_header *, m, *offp, sizeof(*ch)); 571 if (ch == NULL) { 572 CARPSTATS_INC(carps_badlen); 573 CARP_DEBUG("%s: packet size %u too small\n", __func__, len); 574 return (IPPROTO_DONE); 575 } 576 577 578 /* verify the CARP checksum */ 579 m->m_data += *offp; 580 if (in_cksum(m, sizeof(*ch))) { 581 CARPSTATS_INC(carps_badsum); 582 CARP_DEBUG("%s: checksum failed, on %s\n", __func__, 583 m->m_pkthdr.rcvif->if_xname); 584 m_freem(m); 585 return (IPPROTO_DONE); 586 } 587 m->m_data -= *offp; 588 589 carp_input_c(m, ch, AF_INET6); 590 return (IPPROTO_DONE); 591 } 592 #endif /* INET6 */ 593 594 /* 595 * This routine should not be necessary at all, but some switches 596 * (VMWare ESX vswitches) can echo our own packets back at us, 597 * and we must ignore them or they will cause us to drop out of 598 * MASTER mode. 599 * 600 * We cannot catch all cases of network loops. Instead, what we 601 * do here is catch any packet that arrives with a carp header 602 * with a VHID of 0, that comes from an address that is our own. 603 * These packets are by definition "from us" (even if they are from 604 * a misconfigured host that is pretending to be us). 605 * 606 * The VHID test is outside this mini-function. 607 */ 608 static int 609 carp_source_is_self(struct mbuf *m, struct ifaddr *ifa, sa_family_t af) 610 { 611 #ifdef INET 612 struct ip *ip4; 613 struct in_addr in4; 614 #endif 615 #ifdef INET6 616 struct ip6_hdr *ip6; 617 struct in6_addr in6; 618 #endif 619 620 switch (af) { 621 #ifdef INET 622 case AF_INET: 623 ip4 = mtod(m, struct ip *); 624 in4 = ifatoia(ifa)->ia_addr.sin_addr; 625 return (in4.s_addr == ip4->ip_src.s_addr); 626 #endif 627 #ifdef INET6 628 case AF_INET6: 629 ip6 = mtod(m, struct ip6_hdr *); 630 in6 = ifatoia6(ifa)->ia_addr.sin6_addr; 631 return (memcmp(&in6, &ip6->ip6_src, sizeof(in6)) == 0); 632 #endif 633 default: 634 break; 635 } 636 return (0); 637 } 638 639 static void 640 carp_input_c(struct mbuf *m, struct carp_header *ch, sa_family_t af) 641 { 642 struct ifnet *ifp = m->m_pkthdr.rcvif; 643 struct ifaddr *ifa, *match; 644 struct carp_softc *sc; 645 uint64_t tmp_counter; 646 struct timeval sc_tv, ch_tv; 647 int error; 648 649 NET_EPOCH_ASSERT(); 650 651 /* 652 * Verify that the VHID is valid on the receiving interface. 653 * 654 * There should be just one match. If there are none 655 * the VHID is not valid and we drop the packet. If 656 * there are multiple VHID matches, take just the first 657 * one, for compatibility with previous code. While we're 658 * scanning, check for obvious loops in the network topology 659 * (these should never happen, and as noted above, we may 660 * miss real loops; this is just a double-check). 661 */ 662 error = 0; 663 match = NULL; 664 IFNET_FOREACH_IFA(ifp, ifa) { 665 if (match == NULL && ifa->ifa_carp != NULL && 666 ifa->ifa_addr->sa_family == af && 667 ifa->ifa_carp->sc_vhid == ch->carp_vhid) 668 match = ifa; 669 if (ch->carp_vhid == 0 && carp_source_is_self(m, ifa, af)) 670 error = ELOOP; 671 } 672 ifa = error ? NULL : match; 673 if (ifa != NULL) 674 ifa_ref(ifa); 675 676 if (ifa == NULL) { 677 if (error == ELOOP) { 678 CARP_DEBUG("dropping looped packet on interface %s\n", 679 ifp->if_xname); 680 CARPSTATS_INC(carps_badif); /* ??? */ 681 } else { 682 CARPSTATS_INC(carps_badvhid); 683 } 684 m_freem(m); 685 return; 686 } 687 688 /* verify the CARP version. */ 689 if (ch->carp_version != CARP_VERSION) { 690 CARPSTATS_INC(carps_badver); 691 CARP_DEBUG("%s: invalid version %d\n", ifp->if_xname, 692 ch->carp_version); 693 ifa_free(ifa); 694 m_freem(m); 695 return; 696 } 697 698 sc = ifa->ifa_carp; 699 CARP_LOCK(sc); 700 ifa_free(ifa); 701 702 if (carp_hmac_verify(sc, ch->carp_counter, ch->carp_md)) { 703 CARPSTATS_INC(carps_badauth); 704 CARP_DEBUG("%s: incorrect hash for VHID %u@%s\n", __func__, 705 sc->sc_vhid, ifp->if_xname); 706 goto out; 707 } 708 709 tmp_counter = ntohl(ch->carp_counter[0]); 710 tmp_counter = tmp_counter<<32; 711 tmp_counter += ntohl(ch->carp_counter[1]); 712 713 /* XXX Replay protection goes here */ 714 715 sc->sc_init_counter = 0; 716 sc->sc_counter = tmp_counter; 717 718 sc_tv.tv_sec = sc->sc_advbase; 719 sc_tv.tv_usec = DEMOTE_ADVSKEW(sc) * 1000000 / 256; 720 ch_tv.tv_sec = ch->carp_advbase; 721 ch_tv.tv_usec = ch->carp_advskew * 1000000 / 256; 722 723 switch (sc->sc_state) { 724 case INIT: 725 break; 726 case MASTER: 727 /* 728 * If we receive an advertisement from a master who's going to 729 * be more frequent than us, go into BACKUP state. 730 */ 731 if (timevalcmp(&sc_tv, &ch_tv, >) || 732 timevalcmp(&sc_tv, &ch_tv, ==)) { 733 callout_stop(&sc->sc_ad_tmo); 734 carp_set_state(sc, BACKUP, 735 "more frequent advertisement received"); 736 carp_setrun(sc, 0); 737 carp_delroute(sc); 738 } 739 break; 740 case BACKUP: 741 /* 742 * If we're pre-empting masters who advertise slower than us, 743 * and this one claims to be slower, treat him as down. 744 */ 745 if (V_carp_preempt && timevalcmp(&sc_tv, &ch_tv, <)) { 746 carp_master_down_locked(sc, 747 "preempting a slower master"); 748 break; 749 } 750 751 /* 752 * If the master is going to advertise at such a low frequency 753 * that he's guaranteed to time out, we'd might as well just 754 * treat him as timed out now. 755 */ 756 sc_tv.tv_sec = sc->sc_advbase * 3; 757 if (timevalcmp(&sc_tv, &ch_tv, <)) { 758 carp_master_down_locked(sc, "master will time out"); 759 break; 760 } 761 762 /* 763 * Otherwise, we reset the counter and wait for the next 764 * advertisement. 765 */ 766 carp_setrun(sc, af); 767 break; 768 } 769 770 out: 771 CARP_UNLOCK(sc); 772 m_freem(m); 773 } 774 775 static int 776 carp_prepare_ad(struct mbuf *m, struct carp_softc *sc, struct carp_header *ch) 777 { 778 struct m_tag *mtag; 779 780 if (sc->sc_init_counter) { 781 /* this could also be seconds since unix epoch */ 782 sc->sc_counter = arc4random(); 783 sc->sc_counter = sc->sc_counter << 32; 784 sc->sc_counter += arc4random(); 785 } else 786 sc->sc_counter++; 787 788 ch->carp_counter[0] = htonl((sc->sc_counter>>32)&0xffffffff); 789 ch->carp_counter[1] = htonl(sc->sc_counter&0xffffffff); 790 791 carp_hmac_generate(sc, ch->carp_counter, ch->carp_md); 792 793 /* Tag packet for carp_output */ 794 if ((mtag = m_tag_get(PACKET_TAG_CARP, sizeof(struct carp_softc *), 795 M_NOWAIT)) == NULL) { 796 m_freem(m); 797 CARPSTATS_INC(carps_onomem); 798 return (ENOMEM); 799 } 800 bcopy(&sc, mtag + 1, sizeof(sc)); 801 m_tag_prepend(m, mtag); 802 803 return (0); 804 } 805 806 /* 807 * To avoid LORs and possible recursions this function shouldn't 808 * be called directly, but scheduled via taskqueue. 809 */ 810 static void 811 carp_send_ad_all(void *ctx __unused, int pending __unused) 812 { 813 struct carp_softc *sc; 814 815 mtx_lock(&carp_mtx); 816 LIST_FOREACH(sc, &carp_list, sc_next) 817 if (sc->sc_state == MASTER) { 818 CARP_LOCK(sc); 819 CURVNET_SET(sc->sc_carpdev->if_vnet); 820 carp_send_ad_locked(sc); 821 CURVNET_RESTORE(); 822 CARP_UNLOCK(sc); 823 } 824 mtx_unlock(&carp_mtx); 825 } 826 827 /* Send a periodic advertisement, executed in callout context. */ 828 static void 829 carp_send_ad(void *v) 830 { 831 struct carp_softc *sc = v; 832 833 CARP_LOCK_ASSERT(sc); 834 CURVNET_SET(sc->sc_carpdev->if_vnet); 835 carp_send_ad_locked(sc); 836 CURVNET_RESTORE(); 837 CARP_UNLOCK(sc); 838 } 839 840 static void 841 carp_send_ad_error(struct carp_softc *sc, int error) 842 { 843 844 if (error) { 845 if (sc->sc_sendad_errors < INT_MAX) 846 sc->sc_sendad_errors++; 847 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) { 848 static const char fmt[] = "send error %d on %s"; 849 char msg[sizeof(fmt) + IFNAMSIZ]; 850 851 sprintf(msg, fmt, error, sc->sc_carpdev->if_xname); 852 carp_demote_adj(V_carp_senderr_adj, msg); 853 } 854 sc->sc_sendad_success = 0; 855 } else { 856 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS && 857 ++sc->sc_sendad_success >= CARP_SENDAD_MIN_SUCCESS) { 858 static const char fmt[] = "send ok on %s"; 859 char msg[sizeof(fmt) + IFNAMSIZ]; 860 861 sprintf(msg, fmt, sc->sc_carpdev->if_xname); 862 carp_demote_adj(-V_carp_senderr_adj, msg); 863 sc->sc_sendad_errors = 0; 864 } else 865 sc->sc_sendad_errors = 0; 866 } 867 } 868 869 /* 870 * Pick the best ifaddr on the given ifp for sending CARP 871 * advertisements. 872 * 873 * "Best" here is defined by ifa_preferred(). This function is much 874 * much like ifaof_ifpforaddr() except that we just use ifa_preferred(). 875 * 876 * (This could be simplified to return the actual address, except that 877 * it has a different format in AF_INET and AF_INET6.) 878 */ 879 static struct ifaddr * 880 carp_best_ifa(int af, struct ifnet *ifp) 881 { 882 struct ifaddr *ifa, *best; 883 884 NET_EPOCH_ASSERT(); 885 886 if (af >= AF_MAX) 887 return (NULL); 888 best = NULL; 889 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 890 if (ifa->ifa_addr->sa_family == af && 891 (best == NULL || ifa_preferred(best, ifa))) 892 best = ifa; 893 } 894 if (best != NULL) 895 ifa_ref(best); 896 return (best); 897 } 898 899 static void 900 carp_send_ad_locked(struct carp_softc *sc) 901 { 902 struct carp_header ch; 903 struct timeval tv; 904 struct ifaddr *ifa; 905 struct carp_header *ch_ptr; 906 struct mbuf *m; 907 int len, advskew; 908 909 CARP_LOCK_ASSERT(sc); 910 911 advskew = DEMOTE_ADVSKEW(sc); 912 tv.tv_sec = sc->sc_advbase; 913 tv.tv_usec = advskew * 1000000 / 256; 914 915 ch.carp_version = CARP_VERSION; 916 ch.carp_type = CARP_ADVERTISEMENT; 917 ch.carp_vhid = sc->sc_vhid; 918 ch.carp_advbase = sc->sc_advbase; 919 ch.carp_advskew = advskew; 920 ch.carp_authlen = 7; /* XXX DEFINE */ 921 ch.carp_pad1 = 0; /* must be zero */ 922 ch.carp_cksum = 0; 923 924 /* XXXGL: OpenBSD picks first ifaddr with needed family. */ 925 926 #ifdef INET 927 if (sc->sc_naddrs) { 928 struct ip *ip; 929 930 m = m_gethdr(M_NOWAIT, MT_DATA); 931 if (m == NULL) { 932 CARPSTATS_INC(carps_onomem); 933 goto resched; 934 } 935 len = sizeof(*ip) + sizeof(ch); 936 m->m_pkthdr.len = len; 937 m->m_pkthdr.rcvif = NULL; 938 m->m_len = len; 939 M_ALIGN(m, m->m_len); 940 m->m_flags |= M_MCAST; 941 ip = mtod(m, struct ip *); 942 ip->ip_v = IPVERSION; 943 ip->ip_hl = sizeof(*ip) >> 2; 944 ip->ip_tos = V_carp_dscp << IPTOS_DSCP_OFFSET; 945 ip->ip_len = htons(len); 946 ip->ip_off = htons(IP_DF); 947 ip->ip_ttl = CARP_DFLTTL; 948 ip->ip_p = IPPROTO_CARP; 949 ip->ip_sum = 0; 950 ip_fillid(ip); 951 952 ifa = carp_best_ifa(AF_INET, sc->sc_carpdev); 953 if (ifa != NULL) { 954 ip->ip_src.s_addr = 955 ifatoia(ifa)->ia_addr.sin_addr.s_addr; 956 ifa_free(ifa); 957 } else 958 ip->ip_src.s_addr = 0; 959 ip->ip_dst.s_addr = htonl(INADDR_CARP_GROUP); 960 961 ch_ptr = (struct carp_header *)(&ip[1]); 962 bcopy(&ch, ch_ptr, sizeof(ch)); 963 if (carp_prepare_ad(m, sc, ch_ptr)) 964 goto resched; 965 966 m->m_data += sizeof(*ip); 967 ch_ptr->carp_cksum = in_cksum(m, len - sizeof(*ip)); 968 m->m_data -= sizeof(*ip); 969 970 CARPSTATS_INC(carps_opackets); 971 972 carp_send_ad_error(sc, ip_output(m, NULL, NULL, IP_RAWOUTPUT, 973 &sc->sc_carpdev->if_carp->cif_imo, NULL)); 974 } 975 #endif /* INET */ 976 #ifdef INET6 977 if (sc->sc_naddrs6) { 978 struct ip6_hdr *ip6; 979 980 m = m_gethdr(M_NOWAIT, MT_DATA); 981 if (m == NULL) { 982 CARPSTATS_INC(carps_onomem); 983 goto resched; 984 } 985 len = sizeof(*ip6) + sizeof(ch); 986 m->m_pkthdr.len = len; 987 m->m_pkthdr.rcvif = NULL; 988 m->m_len = len; 989 M_ALIGN(m, m->m_len); 990 m->m_flags |= M_MCAST; 991 ip6 = mtod(m, struct ip6_hdr *); 992 bzero(ip6, sizeof(*ip6)); 993 ip6->ip6_vfc |= IPV6_VERSION; 994 /* Traffic class isn't defined in ip6 struct instead 995 * it gets offset into flowid field */ 996 ip6->ip6_flow |= htonl(V_carp_dscp << (IPV6_FLOWLABEL_LEN + 997 IPTOS_DSCP_OFFSET)); 998 ip6->ip6_hlim = CARP_DFLTTL; 999 ip6->ip6_nxt = IPPROTO_CARP; 1000 1001 /* set the source address */ 1002 ifa = carp_best_ifa(AF_INET6, sc->sc_carpdev); 1003 if (ifa != NULL) { 1004 bcopy(IFA_IN6(ifa), &ip6->ip6_src, 1005 sizeof(struct in6_addr)); 1006 ifa_free(ifa); 1007 } else 1008 /* This should never happen with IPv6. */ 1009 bzero(&ip6->ip6_src, sizeof(struct in6_addr)); 1010 1011 /* Set the multicast destination. */ 1012 ip6->ip6_dst.s6_addr16[0] = htons(0xff02); 1013 ip6->ip6_dst.s6_addr8[15] = 0x12; 1014 if (in6_setscope(&ip6->ip6_dst, sc->sc_carpdev, NULL) != 0) { 1015 m_freem(m); 1016 CARP_DEBUG("%s: in6_setscope failed\n", __func__); 1017 goto resched; 1018 } 1019 1020 ch_ptr = (struct carp_header *)(&ip6[1]); 1021 bcopy(&ch, ch_ptr, sizeof(ch)); 1022 if (carp_prepare_ad(m, sc, ch_ptr)) 1023 goto resched; 1024 1025 m->m_data += sizeof(*ip6); 1026 ch_ptr->carp_cksum = in_cksum(m, len - sizeof(*ip6)); 1027 m->m_data -= sizeof(*ip6); 1028 1029 CARPSTATS_INC(carps_opackets6); 1030 1031 carp_send_ad_error(sc, ip6_output(m, NULL, NULL, 0, 1032 &sc->sc_carpdev->if_carp->cif_im6o, NULL, NULL)); 1033 } 1034 #endif /* INET6 */ 1035 1036 resched: 1037 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv), carp_send_ad, sc); 1038 } 1039 1040 static void 1041 carp_addroute(struct carp_softc *sc) 1042 { 1043 struct ifaddr *ifa; 1044 1045 CARP_FOREACH_IFA(sc, ifa) 1046 carp_ifa_addroute(ifa); 1047 } 1048 1049 static void 1050 carp_ifa_addroute(struct ifaddr *ifa) 1051 { 1052 1053 switch (ifa->ifa_addr->sa_family) { 1054 #ifdef INET 1055 case AF_INET: 1056 in_addprefix(ifatoia(ifa), RTF_UP); 1057 ifa_add_loopback_route(ifa, 1058 (struct sockaddr *)&ifatoia(ifa)->ia_addr); 1059 break; 1060 #endif 1061 #ifdef INET6 1062 case AF_INET6: 1063 ifa_add_loopback_route(ifa, 1064 (struct sockaddr *)&ifatoia6(ifa)->ia_addr); 1065 nd6_add_ifa_lle(ifatoia6(ifa)); 1066 break; 1067 #endif 1068 } 1069 } 1070 1071 static void 1072 carp_delroute(struct carp_softc *sc) 1073 { 1074 struct ifaddr *ifa; 1075 1076 CARP_FOREACH_IFA(sc, ifa) 1077 carp_ifa_delroute(ifa); 1078 } 1079 1080 static void 1081 carp_ifa_delroute(struct ifaddr *ifa) 1082 { 1083 1084 switch (ifa->ifa_addr->sa_family) { 1085 #ifdef INET 1086 case AF_INET: 1087 ifa_del_loopback_route(ifa, 1088 (struct sockaddr *)&ifatoia(ifa)->ia_addr); 1089 in_scrubprefix(ifatoia(ifa), LLE_STATIC); 1090 break; 1091 #endif 1092 #ifdef INET6 1093 case AF_INET6: 1094 ifa_del_loopback_route(ifa, 1095 (struct sockaddr *)&ifatoia6(ifa)->ia_addr); 1096 nd6_rem_ifa_lle(ifatoia6(ifa), 1); 1097 break; 1098 #endif 1099 } 1100 } 1101 1102 int 1103 carp_master(struct ifaddr *ifa) 1104 { 1105 struct carp_softc *sc = ifa->ifa_carp; 1106 1107 return (sc->sc_state == MASTER); 1108 } 1109 1110 #ifdef INET 1111 /* 1112 * Broadcast a gratuitous ARP request containing 1113 * the virtual router MAC address for each IP address 1114 * associated with the virtual router. 1115 */ 1116 static void 1117 carp_send_arp(struct carp_softc *sc) 1118 { 1119 struct ifaddr *ifa; 1120 struct in_addr addr; 1121 1122 CARP_FOREACH_IFA(sc, ifa) { 1123 if (ifa->ifa_addr->sa_family != AF_INET) 1124 continue; 1125 addr = ((struct sockaddr_in *)ifa->ifa_addr)->sin_addr; 1126 arp_announce_ifaddr(sc->sc_carpdev, addr, LLADDR(&sc->sc_addr)); 1127 } 1128 } 1129 1130 int 1131 carp_iamatch(struct ifaddr *ifa, uint8_t **enaddr) 1132 { 1133 struct carp_softc *sc = ifa->ifa_carp; 1134 1135 if (sc->sc_state == MASTER) { 1136 *enaddr = LLADDR(&sc->sc_addr); 1137 return (1); 1138 } 1139 1140 return (0); 1141 } 1142 #endif 1143 1144 #ifdef INET6 1145 static void 1146 carp_send_na(struct carp_softc *sc) 1147 { 1148 static struct in6_addr mcast = IN6ADDR_LINKLOCAL_ALLNODES_INIT; 1149 struct ifaddr *ifa; 1150 struct in6_addr *in6; 1151 1152 CARP_FOREACH_IFA(sc, ifa) { 1153 if (ifa->ifa_addr->sa_family != AF_INET6) 1154 continue; 1155 1156 in6 = IFA_IN6(ifa); 1157 nd6_na_output(sc->sc_carpdev, &mcast, in6, 1158 ND_NA_FLAG_OVERRIDE, 1, NULL); 1159 DELAY(1000); /* XXX */ 1160 } 1161 } 1162 1163 /* 1164 * Returns ifa in case it's a carp address and it is MASTER, or if the address 1165 * matches and is not a carp address. Returns NULL otherwise. 1166 */ 1167 struct ifaddr * 1168 carp_iamatch6(struct ifnet *ifp, struct in6_addr *taddr) 1169 { 1170 struct ifaddr *ifa; 1171 1172 NET_EPOCH_ASSERT(); 1173 1174 ifa = NULL; 1175 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1176 if (ifa->ifa_addr->sa_family != AF_INET6) 1177 continue; 1178 if (!IN6_ARE_ADDR_EQUAL(taddr, IFA_IN6(ifa))) 1179 continue; 1180 if (ifa->ifa_carp && ifa->ifa_carp->sc_state != MASTER) 1181 ifa = NULL; 1182 else 1183 ifa_ref(ifa); 1184 break; 1185 } 1186 1187 return (ifa); 1188 } 1189 1190 caddr_t 1191 carp_macmatch6(struct ifnet *ifp, struct mbuf *m, const struct in6_addr *taddr) 1192 { 1193 struct ifaddr *ifa; 1194 1195 NET_EPOCH_ASSERT(); 1196 1197 IFNET_FOREACH_IFA(ifp, ifa) 1198 if (ifa->ifa_addr->sa_family == AF_INET6 && 1199 IN6_ARE_ADDR_EQUAL(taddr, IFA_IN6(ifa))) { 1200 struct carp_softc *sc = ifa->ifa_carp; 1201 struct m_tag *mtag; 1202 1203 mtag = m_tag_get(PACKET_TAG_CARP, 1204 sizeof(struct carp_softc *), M_NOWAIT); 1205 if (mtag == NULL) 1206 /* Better a bit than nothing. */ 1207 return (LLADDR(&sc->sc_addr)); 1208 1209 bcopy(&sc, mtag + 1, sizeof(sc)); 1210 m_tag_prepend(m, mtag); 1211 1212 return (LLADDR(&sc->sc_addr)); 1213 } 1214 1215 return (NULL); 1216 } 1217 #endif /* INET6 */ 1218 1219 int 1220 carp_forus(struct ifnet *ifp, u_char *dhost) 1221 { 1222 struct carp_softc *sc; 1223 uint8_t *ena = dhost; 1224 1225 if (ena[0] || ena[1] || ena[2] != 0x5e || ena[3] || ena[4] != 1) 1226 return (0); 1227 1228 CIF_LOCK(ifp->if_carp); 1229 IFNET_FOREACH_CARP(ifp, sc) { 1230 CARP_LOCK(sc); 1231 if (sc->sc_state == MASTER && !bcmp(dhost, LLADDR(&sc->sc_addr), 1232 ETHER_ADDR_LEN)) { 1233 CARP_UNLOCK(sc); 1234 CIF_UNLOCK(ifp->if_carp); 1235 return (1); 1236 } 1237 CARP_UNLOCK(sc); 1238 } 1239 CIF_UNLOCK(ifp->if_carp); 1240 1241 return (0); 1242 } 1243 1244 /* Master down timeout event, executed in callout context. */ 1245 static void 1246 carp_master_down(void *v) 1247 { 1248 struct carp_softc *sc = v; 1249 1250 CARP_LOCK_ASSERT(sc); 1251 1252 CURVNET_SET(sc->sc_carpdev->if_vnet); 1253 if (sc->sc_state == BACKUP) { 1254 carp_master_down_locked(sc, "master timed out"); 1255 } 1256 CURVNET_RESTORE(); 1257 1258 CARP_UNLOCK(sc); 1259 } 1260 1261 static void 1262 carp_master_down_locked(struct carp_softc *sc, const char *reason) 1263 { 1264 1265 CARP_LOCK_ASSERT(sc); 1266 1267 switch (sc->sc_state) { 1268 case BACKUP: 1269 carp_set_state(sc, MASTER, reason); 1270 carp_send_ad_locked(sc); 1271 #ifdef INET 1272 carp_send_arp(sc); 1273 #endif 1274 #ifdef INET6 1275 carp_send_na(sc); 1276 #endif 1277 carp_setrun(sc, 0); 1278 carp_addroute(sc); 1279 break; 1280 case INIT: 1281 case MASTER: 1282 #ifdef INVARIANTS 1283 panic("carp: VHID %u@%s: master_down event in %s state\n", 1284 sc->sc_vhid, 1285 sc->sc_carpdev->if_xname, 1286 sc->sc_state ? "MASTER" : "INIT"); 1287 #endif 1288 break; 1289 } 1290 } 1291 1292 /* 1293 * When in backup state, af indicates whether to reset the master down timer 1294 * for v4 or v6. If it's set to zero, reset the ones which are already pending. 1295 */ 1296 static void 1297 carp_setrun(struct carp_softc *sc, sa_family_t af) 1298 { 1299 struct timeval tv; 1300 1301 CARP_LOCK_ASSERT(sc); 1302 1303 if ((sc->sc_carpdev->if_flags & IFF_UP) == 0 || 1304 sc->sc_carpdev->if_link_state != LINK_STATE_UP || 1305 (sc->sc_naddrs == 0 && sc->sc_naddrs6 == 0) || 1306 !V_carp_allow) 1307 return; 1308 1309 switch (sc->sc_state) { 1310 case INIT: 1311 carp_set_state(sc, BACKUP, "initialization complete"); 1312 carp_setrun(sc, 0); 1313 break; 1314 case BACKUP: 1315 callout_stop(&sc->sc_ad_tmo); 1316 tv.tv_sec = 3 * sc->sc_advbase; 1317 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1318 switch (af) { 1319 #ifdef INET 1320 case AF_INET: 1321 callout_reset(&sc->sc_md_tmo, tvtohz(&tv), 1322 carp_master_down, sc); 1323 break; 1324 #endif 1325 #ifdef INET6 1326 case AF_INET6: 1327 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv), 1328 carp_master_down, sc); 1329 break; 1330 #endif 1331 default: 1332 #ifdef INET 1333 if (sc->sc_naddrs) 1334 callout_reset(&sc->sc_md_tmo, tvtohz(&tv), 1335 carp_master_down, sc); 1336 #endif 1337 #ifdef INET6 1338 if (sc->sc_naddrs6) 1339 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv), 1340 carp_master_down, sc); 1341 #endif 1342 break; 1343 } 1344 break; 1345 case MASTER: 1346 tv.tv_sec = sc->sc_advbase; 1347 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1348 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv), 1349 carp_send_ad, sc); 1350 break; 1351 } 1352 } 1353 1354 /* 1355 * Setup multicast structures. 1356 */ 1357 static int 1358 carp_multicast_setup(struct carp_if *cif, sa_family_t sa) 1359 { 1360 struct ifnet *ifp = cif->cif_ifp; 1361 int error = 0; 1362 1363 switch (sa) { 1364 #ifdef INET 1365 case AF_INET: 1366 { 1367 struct ip_moptions *imo = &cif->cif_imo; 1368 struct in_mfilter *imf; 1369 struct in_addr addr; 1370 1371 if (ip_mfilter_first(&imo->imo_head) != NULL) 1372 return (0); 1373 1374 imf = ip_mfilter_alloc(M_WAITOK, 0, 0); 1375 ip_mfilter_init(&imo->imo_head); 1376 imo->imo_multicast_vif = -1; 1377 1378 addr.s_addr = htonl(INADDR_CARP_GROUP); 1379 if ((error = in_joingroup(ifp, &addr, NULL, 1380 &imf->imf_inm)) != 0) { 1381 ip_mfilter_free(imf); 1382 break; 1383 } 1384 1385 ip_mfilter_insert(&imo->imo_head, imf); 1386 imo->imo_multicast_ifp = ifp; 1387 imo->imo_multicast_ttl = CARP_DFLTTL; 1388 imo->imo_multicast_loop = 0; 1389 break; 1390 } 1391 #endif 1392 #ifdef INET6 1393 case AF_INET6: 1394 { 1395 struct ip6_moptions *im6o = &cif->cif_im6o; 1396 struct in6_mfilter *im6f[2]; 1397 struct in6_addr in6; 1398 1399 if (ip6_mfilter_first(&im6o->im6o_head)) 1400 return (0); 1401 1402 im6f[0] = ip6_mfilter_alloc(M_WAITOK, 0, 0); 1403 im6f[1] = ip6_mfilter_alloc(M_WAITOK, 0, 0); 1404 1405 ip6_mfilter_init(&im6o->im6o_head); 1406 im6o->im6o_multicast_hlim = CARP_DFLTTL; 1407 im6o->im6o_multicast_ifp = ifp; 1408 1409 /* Join IPv6 CARP multicast group. */ 1410 bzero(&in6, sizeof(in6)); 1411 in6.s6_addr16[0] = htons(0xff02); 1412 in6.s6_addr8[15] = 0x12; 1413 if ((error = in6_setscope(&in6, ifp, NULL)) != 0) { 1414 ip6_mfilter_free(im6f[0]); 1415 ip6_mfilter_free(im6f[1]); 1416 break; 1417 } 1418 if ((error = in6_joingroup(ifp, &in6, NULL, &im6f[0]->im6f_in6m, 0)) != 0) { 1419 ip6_mfilter_free(im6f[0]); 1420 ip6_mfilter_free(im6f[1]); 1421 break; 1422 } 1423 1424 /* Join solicited multicast address. */ 1425 bzero(&in6, sizeof(in6)); 1426 in6.s6_addr16[0] = htons(0xff02); 1427 in6.s6_addr32[1] = 0; 1428 in6.s6_addr32[2] = htonl(1); 1429 in6.s6_addr32[3] = 0; 1430 in6.s6_addr8[12] = 0xff; 1431 1432 if ((error = in6_setscope(&in6, ifp, NULL)) != 0) { 1433 ip6_mfilter_free(im6f[0]); 1434 ip6_mfilter_free(im6f[1]); 1435 break; 1436 } 1437 1438 if ((error = in6_joingroup(ifp, &in6, NULL, &im6f[1]->im6f_in6m, 0)) != 0) { 1439 in6_leavegroup(im6f[0]->im6f_in6m, NULL); 1440 ip6_mfilter_free(im6f[0]); 1441 ip6_mfilter_free(im6f[1]); 1442 break; 1443 } 1444 ip6_mfilter_insert(&im6o->im6o_head, im6f[0]); 1445 ip6_mfilter_insert(&im6o->im6o_head, im6f[1]); 1446 break; 1447 } 1448 #endif 1449 } 1450 1451 return (error); 1452 } 1453 1454 /* 1455 * Free multicast structures. 1456 */ 1457 static void 1458 carp_multicast_cleanup(struct carp_if *cif, sa_family_t sa) 1459 { 1460 #ifdef INET 1461 struct ip_moptions *imo = &cif->cif_imo; 1462 struct in_mfilter *imf; 1463 #endif 1464 #ifdef INET6 1465 struct ip6_moptions *im6o = &cif->cif_im6o; 1466 struct in6_mfilter *im6f; 1467 #endif 1468 sx_assert(&carp_sx, SA_XLOCKED); 1469 1470 switch (sa) { 1471 #ifdef INET 1472 case AF_INET: 1473 if (cif->cif_naddrs != 0) 1474 break; 1475 1476 while ((imf = ip_mfilter_first(&imo->imo_head)) != NULL) { 1477 ip_mfilter_remove(&imo->imo_head, imf); 1478 in_leavegroup(imf->imf_inm, NULL); 1479 ip_mfilter_free(imf); 1480 } 1481 break; 1482 #endif 1483 #ifdef INET6 1484 case AF_INET6: 1485 if (cif->cif_naddrs6 != 0) 1486 break; 1487 1488 while ((im6f = ip6_mfilter_first(&im6o->im6o_head)) != NULL) { 1489 ip6_mfilter_remove(&im6o->im6o_head, im6f); 1490 in6_leavegroup(im6f->im6f_in6m, NULL); 1491 ip6_mfilter_free(im6f); 1492 } 1493 break; 1494 #endif 1495 } 1496 } 1497 1498 int 1499 carp_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa) 1500 { 1501 struct m_tag *mtag; 1502 struct carp_softc *sc; 1503 1504 if (!sa) 1505 return (0); 1506 1507 switch (sa->sa_family) { 1508 #ifdef INET 1509 case AF_INET: 1510 break; 1511 #endif 1512 #ifdef INET6 1513 case AF_INET6: 1514 break; 1515 #endif 1516 default: 1517 return (0); 1518 } 1519 1520 mtag = m_tag_find(m, PACKET_TAG_CARP, NULL); 1521 if (mtag == NULL) 1522 return (0); 1523 1524 bcopy(mtag + 1, &sc, sizeof(sc)); 1525 1526 /* Set the source MAC address to the Virtual Router MAC Address. */ 1527 switch (ifp->if_type) { 1528 case IFT_ETHER: 1529 case IFT_BRIDGE: 1530 case IFT_L2VLAN: { 1531 struct ether_header *eh; 1532 1533 eh = mtod(m, struct ether_header *); 1534 eh->ether_shost[0] = 0; 1535 eh->ether_shost[1] = 0; 1536 eh->ether_shost[2] = 0x5e; 1537 eh->ether_shost[3] = 0; 1538 eh->ether_shost[4] = 1; 1539 eh->ether_shost[5] = sc->sc_vhid; 1540 } 1541 break; 1542 default: 1543 printf("%s: carp is not supported for the %d interface type\n", 1544 ifp->if_xname, ifp->if_type); 1545 return (EOPNOTSUPP); 1546 } 1547 1548 return (0); 1549 } 1550 1551 static struct carp_softc* 1552 carp_alloc(struct ifnet *ifp) 1553 { 1554 struct carp_softc *sc; 1555 struct carp_if *cif; 1556 1557 sx_assert(&carp_sx, SA_XLOCKED); 1558 1559 if ((cif = ifp->if_carp) == NULL) 1560 cif = carp_alloc_if(ifp); 1561 1562 sc = malloc(sizeof(*sc), M_CARP, M_WAITOK|M_ZERO); 1563 1564 sc->sc_advbase = CARP_DFLTINTV; 1565 sc->sc_vhid = -1; /* required setting */ 1566 sc->sc_init_counter = 1; 1567 sc->sc_state = INIT; 1568 1569 sc->sc_ifasiz = sizeof(struct ifaddr *); 1570 sc->sc_ifas = malloc(sc->sc_ifasiz, M_CARP, M_WAITOK|M_ZERO); 1571 sc->sc_carpdev = ifp; 1572 1573 CARP_LOCK_INIT(sc); 1574 #ifdef INET 1575 callout_init_mtx(&sc->sc_md_tmo, &sc->sc_mtx, CALLOUT_RETURNUNLOCKED); 1576 #endif 1577 #ifdef INET6 1578 callout_init_mtx(&sc->sc_md6_tmo, &sc->sc_mtx, CALLOUT_RETURNUNLOCKED); 1579 #endif 1580 callout_init_mtx(&sc->sc_ad_tmo, &sc->sc_mtx, CALLOUT_RETURNUNLOCKED); 1581 1582 CIF_LOCK(cif); 1583 TAILQ_INSERT_TAIL(&cif->cif_vrs, sc, sc_list); 1584 CIF_UNLOCK(cif); 1585 1586 mtx_lock(&carp_mtx); 1587 LIST_INSERT_HEAD(&carp_list, sc, sc_next); 1588 mtx_unlock(&carp_mtx); 1589 1590 return (sc); 1591 } 1592 1593 static void 1594 carp_grow_ifas(struct carp_softc *sc) 1595 { 1596 struct ifaddr **new; 1597 1598 new = malloc(sc->sc_ifasiz * 2, M_CARP, M_WAITOK | M_ZERO); 1599 CARP_LOCK(sc); 1600 bcopy(sc->sc_ifas, new, sc->sc_ifasiz); 1601 free(sc->sc_ifas, M_CARP); 1602 sc->sc_ifas = new; 1603 sc->sc_ifasiz *= 2; 1604 CARP_UNLOCK(sc); 1605 } 1606 1607 static void 1608 carp_destroy(struct carp_softc *sc) 1609 { 1610 struct ifnet *ifp = sc->sc_carpdev; 1611 struct carp_if *cif = ifp->if_carp; 1612 1613 sx_assert(&carp_sx, SA_XLOCKED); 1614 1615 if (sc->sc_suppress) 1616 carp_demote_adj(-V_carp_ifdown_adj, "vhid removed"); 1617 CARP_UNLOCK(sc); 1618 1619 CIF_LOCK(cif); 1620 TAILQ_REMOVE(&cif->cif_vrs, sc, sc_list); 1621 CIF_UNLOCK(cif); 1622 1623 mtx_lock(&carp_mtx); 1624 LIST_REMOVE(sc, sc_next); 1625 mtx_unlock(&carp_mtx); 1626 1627 callout_drain(&sc->sc_ad_tmo); 1628 #ifdef INET 1629 callout_drain(&sc->sc_md_tmo); 1630 #endif 1631 #ifdef INET6 1632 callout_drain(&sc->sc_md6_tmo); 1633 #endif 1634 CARP_LOCK_DESTROY(sc); 1635 1636 free(sc->sc_ifas, M_CARP); 1637 free(sc, M_CARP); 1638 } 1639 1640 static struct carp_if* 1641 carp_alloc_if(struct ifnet *ifp) 1642 { 1643 struct carp_if *cif; 1644 int error; 1645 1646 cif = malloc(sizeof(*cif), M_CARP, M_WAITOK|M_ZERO); 1647 1648 if ((error = ifpromisc(ifp, 1)) != 0) 1649 printf("%s: ifpromisc(%s) failed: %d\n", 1650 __func__, ifp->if_xname, error); 1651 else 1652 cif->cif_flags |= CIF_PROMISC; 1653 1654 CIF_LOCK_INIT(cif); 1655 cif->cif_ifp = ifp; 1656 TAILQ_INIT(&cif->cif_vrs); 1657 1658 IF_ADDR_WLOCK(ifp); 1659 ifp->if_carp = cif; 1660 if_ref(ifp); 1661 IF_ADDR_WUNLOCK(ifp); 1662 1663 return (cif); 1664 } 1665 1666 static void 1667 carp_free_if(struct carp_if *cif) 1668 { 1669 struct ifnet *ifp = cif->cif_ifp; 1670 1671 CIF_LOCK_ASSERT(cif); 1672 KASSERT(TAILQ_EMPTY(&cif->cif_vrs), ("%s: softc list not empty", 1673 __func__)); 1674 1675 IF_ADDR_WLOCK(ifp); 1676 ifp->if_carp = NULL; 1677 IF_ADDR_WUNLOCK(ifp); 1678 1679 CIF_LOCK_DESTROY(cif); 1680 1681 if (cif->cif_flags & CIF_PROMISC) 1682 ifpromisc(ifp, 0); 1683 if_rele(ifp); 1684 1685 free(cif, M_CARP); 1686 } 1687 1688 static void 1689 carp_carprcp(struct carpreq *carpr, struct carp_softc *sc, int priv) 1690 { 1691 1692 CARP_LOCK(sc); 1693 carpr->carpr_state = sc->sc_state; 1694 carpr->carpr_vhid = sc->sc_vhid; 1695 carpr->carpr_advbase = sc->sc_advbase; 1696 carpr->carpr_advskew = sc->sc_advskew; 1697 if (priv) 1698 bcopy(sc->sc_key, carpr->carpr_key, sizeof(carpr->carpr_key)); 1699 else 1700 bzero(carpr->carpr_key, sizeof(carpr->carpr_key)); 1701 CARP_UNLOCK(sc); 1702 } 1703 1704 int 1705 carp_ioctl(struct ifreq *ifr, u_long cmd, struct thread *td) 1706 { 1707 struct carpreq carpr; 1708 struct ifnet *ifp; 1709 struct carp_softc *sc = NULL; 1710 int error = 0, locked = 0; 1711 1712 if ((error = copyin(ifr_data_get_ptr(ifr), &carpr, sizeof carpr))) 1713 return (error); 1714 1715 ifp = ifunit_ref(ifr->ifr_name); 1716 if (ifp == NULL) 1717 return (ENXIO); 1718 1719 switch (ifp->if_type) { 1720 case IFT_ETHER: 1721 case IFT_L2VLAN: 1722 case IFT_BRIDGE: 1723 break; 1724 default: 1725 error = EOPNOTSUPP; 1726 goto out; 1727 } 1728 1729 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 1730 error = EADDRNOTAVAIL; 1731 goto out; 1732 } 1733 1734 sx_xlock(&carp_sx); 1735 switch (cmd) { 1736 case SIOCSVH: 1737 if ((error = priv_check(td, PRIV_NETINET_CARP))) 1738 break; 1739 if (carpr.carpr_vhid <= 0 || carpr.carpr_vhid > CARP_MAXVHID || 1740 carpr.carpr_advbase < 0 || carpr.carpr_advskew < 0) { 1741 error = EINVAL; 1742 break; 1743 } 1744 1745 if (ifp->if_carp) { 1746 IFNET_FOREACH_CARP(ifp, sc) 1747 if (sc->sc_vhid == carpr.carpr_vhid) 1748 break; 1749 } 1750 if (sc == NULL) { 1751 sc = carp_alloc(ifp); 1752 CARP_LOCK(sc); 1753 sc->sc_vhid = carpr.carpr_vhid; 1754 LLADDR(&sc->sc_addr)[0] = 0; 1755 LLADDR(&sc->sc_addr)[1] = 0; 1756 LLADDR(&sc->sc_addr)[2] = 0x5e; 1757 LLADDR(&sc->sc_addr)[3] = 0; 1758 LLADDR(&sc->sc_addr)[4] = 1; 1759 LLADDR(&sc->sc_addr)[5] = sc->sc_vhid; 1760 } else 1761 CARP_LOCK(sc); 1762 locked = 1; 1763 if (carpr.carpr_advbase > 0) { 1764 if (carpr.carpr_advbase > 255 || 1765 carpr.carpr_advbase < CARP_DFLTINTV) { 1766 error = EINVAL; 1767 break; 1768 } 1769 sc->sc_advbase = carpr.carpr_advbase; 1770 } 1771 if (carpr.carpr_advskew >= 255) { 1772 error = EINVAL; 1773 break; 1774 } 1775 sc->sc_advskew = carpr.carpr_advskew; 1776 if (carpr.carpr_key[0] != '\0') { 1777 bcopy(carpr.carpr_key, sc->sc_key, sizeof(sc->sc_key)); 1778 carp_hmac_prepare(sc); 1779 } 1780 if (sc->sc_state != INIT && 1781 carpr.carpr_state != sc->sc_state) { 1782 switch (carpr.carpr_state) { 1783 case BACKUP: 1784 callout_stop(&sc->sc_ad_tmo); 1785 carp_set_state(sc, BACKUP, 1786 "user requested via ifconfig"); 1787 carp_setrun(sc, 0); 1788 carp_delroute(sc); 1789 break; 1790 case MASTER: 1791 carp_master_down_locked(sc, 1792 "user requested via ifconfig"); 1793 break; 1794 default: 1795 break; 1796 } 1797 } 1798 break; 1799 1800 case SIOCGVH: 1801 { 1802 int priveleged; 1803 1804 if (carpr.carpr_vhid < 0 || carpr.carpr_vhid > CARP_MAXVHID) { 1805 error = EINVAL; 1806 break; 1807 } 1808 if (carpr.carpr_count < 1) { 1809 error = EMSGSIZE; 1810 break; 1811 } 1812 if (ifp->if_carp == NULL) { 1813 error = ENOENT; 1814 break; 1815 } 1816 1817 priveleged = (priv_check(td, PRIV_NETINET_CARP) == 0); 1818 if (carpr.carpr_vhid != 0) { 1819 IFNET_FOREACH_CARP(ifp, sc) 1820 if (sc->sc_vhid == carpr.carpr_vhid) 1821 break; 1822 if (sc == NULL) { 1823 error = ENOENT; 1824 break; 1825 } 1826 carp_carprcp(&carpr, sc, priveleged); 1827 error = copyout(&carpr, ifr_data_get_ptr(ifr), 1828 sizeof(carpr)); 1829 } else { 1830 int i, count; 1831 1832 count = 0; 1833 IFNET_FOREACH_CARP(ifp, sc) 1834 count++; 1835 1836 if (count > carpr.carpr_count) { 1837 CIF_UNLOCK(ifp->if_carp); 1838 error = EMSGSIZE; 1839 break; 1840 } 1841 1842 i = 0; 1843 IFNET_FOREACH_CARP(ifp, sc) { 1844 carp_carprcp(&carpr, sc, priveleged); 1845 carpr.carpr_count = count; 1846 error = copyout(&carpr, 1847 (caddr_t)ifr_data_get_ptr(ifr) + 1848 (i * sizeof(carpr)), sizeof(carpr)); 1849 if (error) { 1850 CIF_UNLOCK(ifp->if_carp); 1851 break; 1852 } 1853 i++; 1854 } 1855 } 1856 break; 1857 } 1858 default: 1859 error = EINVAL; 1860 } 1861 sx_xunlock(&carp_sx); 1862 1863 out: 1864 if (locked) 1865 CARP_UNLOCK(sc); 1866 if_rele(ifp); 1867 1868 return (error); 1869 } 1870 1871 static int 1872 carp_get_vhid(struct ifaddr *ifa) 1873 { 1874 1875 if (ifa == NULL || ifa->ifa_carp == NULL) 1876 return (0); 1877 1878 return (ifa->ifa_carp->sc_vhid); 1879 } 1880 1881 int 1882 carp_attach(struct ifaddr *ifa, int vhid) 1883 { 1884 struct ifnet *ifp = ifa->ifa_ifp; 1885 struct carp_if *cif = ifp->if_carp; 1886 struct carp_softc *sc; 1887 int index, error; 1888 1889 KASSERT(ifa->ifa_carp == NULL, ("%s: ifa %p attached", __func__, ifa)); 1890 1891 switch (ifa->ifa_addr->sa_family) { 1892 #ifdef INET 1893 case AF_INET: 1894 #endif 1895 #ifdef INET6 1896 case AF_INET6: 1897 #endif 1898 break; 1899 default: 1900 return (EPROTOTYPE); 1901 } 1902 1903 sx_xlock(&carp_sx); 1904 if (ifp->if_carp == NULL) { 1905 sx_xunlock(&carp_sx); 1906 return (ENOPROTOOPT); 1907 } 1908 1909 IFNET_FOREACH_CARP(ifp, sc) 1910 if (sc->sc_vhid == vhid) 1911 break; 1912 if (sc == NULL) { 1913 sx_xunlock(&carp_sx); 1914 return (ENOENT); 1915 } 1916 1917 error = carp_multicast_setup(cif, ifa->ifa_addr->sa_family); 1918 if (error) { 1919 CIF_FREE(cif); 1920 sx_xunlock(&carp_sx); 1921 return (error); 1922 } 1923 1924 index = sc->sc_naddrs + sc->sc_naddrs6 + 1; 1925 if (index > sc->sc_ifasiz / sizeof(struct ifaddr *)) 1926 carp_grow_ifas(sc); 1927 1928 switch (ifa->ifa_addr->sa_family) { 1929 #ifdef INET 1930 case AF_INET: 1931 cif->cif_naddrs++; 1932 sc->sc_naddrs++; 1933 break; 1934 #endif 1935 #ifdef INET6 1936 case AF_INET6: 1937 cif->cif_naddrs6++; 1938 sc->sc_naddrs6++; 1939 break; 1940 #endif 1941 } 1942 1943 ifa_ref(ifa); 1944 1945 CARP_LOCK(sc); 1946 sc->sc_ifas[index - 1] = ifa; 1947 ifa->ifa_carp = sc; 1948 carp_hmac_prepare(sc); 1949 carp_sc_state(sc); 1950 CARP_UNLOCK(sc); 1951 1952 sx_xunlock(&carp_sx); 1953 1954 return (0); 1955 } 1956 1957 void 1958 carp_detach(struct ifaddr *ifa, bool keep_cif) 1959 { 1960 struct ifnet *ifp = ifa->ifa_ifp; 1961 struct carp_if *cif = ifp->if_carp; 1962 struct carp_softc *sc = ifa->ifa_carp; 1963 int i, index; 1964 1965 KASSERT(sc != NULL, ("%s: %p not attached", __func__, ifa)); 1966 1967 sx_xlock(&carp_sx); 1968 1969 CARP_LOCK(sc); 1970 /* Shift array. */ 1971 index = sc->sc_naddrs + sc->sc_naddrs6; 1972 for (i = 0; i < index; i++) 1973 if (sc->sc_ifas[i] == ifa) 1974 break; 1975 KASSERT(i < index, ("%s: %p no backref", __func__, ifa)); 1976 for (; i < index - 1; i++) 1977 sc->sc_ifas[i] = sc->sc_ifas[i+1]; 1978 sc->sc_ifas[index - 1] = NULL; 1979 1980 switch (ifa->ifa_addr->sa_family) { 1981 #ifdef INET 1982 case AF_INET: 1983 cif->cif_naddrs--; 1984 sc->sc_naddrs--; 1985 break; 1986 #endif 1987 #ifdef INET6 1988 case AF_INET6: 1989 cif->cif_naddrs6--; 1990 sc->sc_naddrs6--; 1991 break; 1992 #endif 1993 } 1994 1995 carp_ifa_delroute(ifa); 1996 carp_multicast_cleanup(cif, ifa->ifa_addr->sa_family); 1997 1998 ifa->ifa_carp = NULL; 1999 ifa_free(ifa); 2000 2001 carp_hmac_prepare(sc); 2002 carp_sc_state(sc); 2003 2004 if (!keep_cif && sc->sc_naddrs == 0 && sc->sc_naddrs6 == 0) 2005 carp_destroy(sc); 2006 else 2007 CARP_UNLOCK(sc); 2008 2009 if (!keep_cif) 2010 CIF_FREE(cif); 2011 2012 sx_xunlock(&carp_sx); 2013 } 2014 2015 static void 2016 carp_set_state(struct carp_softc *sc, int state, const char *reason) 2017 { 2018 2019 CARP_LOCK_ASSERT(sc); 2020 2021 if (sc->sc_state != state) { 2022 const char *carp_states[] = { CARP_STATES }; 2023 char subsys[IFNAMSIZ+5]; 2024 2025 snprintf(subsys, IFNAMSIZ+5, "%u@%s", sc->sc_vhid, 2026 sc->sc_carpdev->if_xname); 2027 2028 CARP_LOG("%s: %s -> %s (%s)\n", subsys, 2029 carp_states[sc->sc_state], carp_states[state], reason); 2030 2031 sc->sc_state = state; 2032 2033 devctl_notify("CARP", subsys, carp_states[state], NULL); 2034 } 2035 } 2036 2037 static void 2038 carp_linkstate(struct ifnet *ifp) 2039 { 2040 struct carp_softc *sc; 2041 2042 CIF_LOCK(ifp->if_carp); 2043 IFNET_FOREACH_CARP(ifp, sc) { 2044 CARP_LOCK(sc); 2045 carp_sc_state(sc); 2046 CARP_UNLOCK(sc); 2047 } 2048 CIF_UNLOCK(ifp->if_carp); 2049 } 2050 2051 static void 2052 carp_sc_state(struct carp_softc *sc) 2053 { 2054 2055 CARP_LOCK_ASSERT(sc); 2056 2057 if (sc->sc_carpdev->if_link_state != LINK_STATE_UP || 2058 !(sc->sc_carpdev->if_flags & IFF_UP) || 2059 !V_carp_allow) { 2060 callout_stop(&sc->sc_ad_tmo); 2061 #ifdef INET 2062 callout_stop(&sc->sc_md_tmo); 2063 #endif 2064 #ifdef INET6 2065 callout_stop(&sc->sc_md6_tmo); 2066 #endif 2067 carp_set_state(sc, INIT, "hardware interface down"); 2068 carp_setrun(sc, 0); 2069 if (!sc->sc_suppress) 2070 carp_demote_adj(V_carp_ifdown_adj, "interface down"); 2071 sc->sc_suppress = 1; 2072 } else { 2073 carp_set_state(sc, INIT, "hardware interface up"); 2074 carp_setrun(sc, 0); 2075 if (sc->sc_suppress) 2076 carp_demote_adj(-V_carp_ifdown_adj, "interface up"); 2077 sc->sc_suppress = 0; 2078 } 2079 } 2080 2081 static void 2082 carp_demote_adj(int adj, char *reason) 2083 { 2084 atomic_add_int(&V_carp_demotion, adj); 2085 CARP_LOG("demoted by %d to %d (%s)\n", adj, V_carp_demotion, reason); 2086 taskqueue_enqueue(taskqueue_swi, &carp_sendall_task); 2087 } 2088 2089 static int 2090 carp_allow_sysctl(SYSCTL_HANDLER_ARGS) 2091 { 2092 int new, error; 2093 struct carp_softc *sc; 2094 2095 new = V_carp_allow; 2096 error = sysctl_handle_int(oidp, &new, 0, req); 2097 if (error || !req->newptr) 2098 return (error); 2099 2100 if (V_carp_allow != new) { 2101 V_carp_allow = new; 2102 2103 mtx_lock(&carp_mtx); 2104 LIST_FOREACH(sc, &carp_list, sc_next) { 2105 CARP_LOCK(sc); 2106 if (curvnet == sc->sc_carpdev->if_vnet) 2107 carp_sc_state(sc); 2108 CARP_UNLOCK(sc); 2109 } 2110 mtx_unlock(&carp_mtx); 2111 } 2112 2113 return (0); 2114 } 2115 2116 static int 2117 carp_dscp_sysctl(SYSCTL_HANDLER_ARGS) 2118 { 2119 int new, error; 2120 2121 new = V_carp_dscp; 2122 error = sysctl_handle_int(oidp, &new, 0, req); 2123 if (error || !req->newptr) 2124 return (error); 2125 2126 if (new < 0 || new > 63) 2127 return (EINVAL); 2128 2129 V_carp_dscp = new; 2130 2131 return (0); 2132 } 2133 2134 static int 2135 carp_demote_adj_sysctl(SYSCTL_HANDLER_ARGS) 2136 { 2137 int new, error; 2138 2139 new = V_carp_demotion; 2140 error = sysctl_handle_int(oidp, &new, 0, req); 2141 if (error || !req->newptr) 2142 return (error); 2143 2144 carp_demote_adj(new, "sysctl"); 2145 2146 return (0); 2147 } 2148 2149 #ifdef INET 2150 extern struct domain inetdomain; 2151 static struct protosw in_carp_protosw = { 2152 .pr_type = SOCK_RAW, 2153 .pr_domain = &inetdomain, 2154 .pr_protocol = IPPROTO_CARP, 2155 .pr_flags = PR_ATOMIC|PR_ADDR, 2156 .pr_input = carp_input, 2157 .pr_output = rip_output, 2158 .pr_ctloutput = rip_ctloutput, 2159 .pr_usrreqs = &rip_usrreqs 2160 }; 2161 #endif 2162 2163 #ifdef INET6 2164 extern struct domain inet6domain; 2165 static struct protosw in6_carp_protosw = { 2166 .pr_type = SOCK_RAW, 2167 .pr_domain = &inet6domain, 2168 .pr_protocol = IPPROTO_CARP, 2169 .pr_flags = PR_ATOMIC|PR_ADDR, 2170 .pr_input = carp6_input, 2171 .pr_output = rip6_output, 2172 .pr_ctloutput = rip6_ctloutput, 2173 .pr_usrreqs = &rip6_usrreqs 2174 }; 2175 #endif 2176 2177 static void 2178 carp_mod_cleanup(void) 2179 { 2180 2181 #ifdef INET 2182 if (proto_reg[CARP_INET] == 0) { 2183 (void)ipproto_unregister(IPPROTO_CARP); 2184 pf_proto_unregister(PF_INET, IPPROTO_CARP, SOCK_RAW); 2185 proto_reg[CARP_INET] = -1; 2186 } 2187 carp_iamatch_p = NULL; 2188 #endif 2189 #ifdef INET6 2190 if (proto_reg[CARP_INET6] == 0) { 2191 (void)ip6proto_unregister(IPPROTO_CARP); 2192 pf_proto_unregister(PF_INET6, IPPROTO_CARP, SOCK_RAW); 2193 proto_reg[CARP_INET6] = -1; 2194 } 2195 carp_iamatch6_p = NULL; 2196 carp_macmatch6_p = NULL; 2197 #endif 2198 carp_ioctl_p = NULL; 2199 carp_attach_p = NULL; 2200 carp_detach_p = NULL; 2201 carp_get_vhid_p = NULL; 2202 carp_linkstate_p = NULL; 2203 carp_forus_p = NULL; 2204 carp_output_p = NULL; 2205 carp_demote_adj_p = NULL; 2206 carp_master_p = NULL; 2207 mtx_unlock(&carp_mtx); 2208 taskqueue_drain(taskqueue_swi, &carp_sendall_task); 2209 mtx_destroy(&carp_mtx); 2210 sx_destroy(&carp_sx); 2211 } 2212 2213 static int 2214 carp_mod_load(void) 2215 { 2216 int err; 2217 2218 mtx_init(&carp_mtx, "carp_mtx", NULL, MTX_DEF); 2219 sx_init(&carp_sx, "carp_sx"); 2220 LIST_INIT(&carp_list); 2221 carp_get_vhid_p = carp_get_vhid; 2222 carp_forus_p = carp_forus; 2223 carp_output_p = carp_output; 2224 carp_linkstate_p = carp_linkstate; 2225 carp_ioctl_p = carp_ioctl; 2226 carp_attach_p = carp_attach; 2227 carp_detach_p = carp_detach; 2228 carp_demote_adj_p = carp_demote_adj; 2229 carp_master_p = carp_master; 2230 #ifdef INET6 2231 carp_iamatch6_p = carp_iamatch6; 2232 carp_macmatch6_p = carp_macmatch6; 2233 proto_reg[CARP_INET6] = pf_proto_register(PF_INET6, 2234 (struct protosw *)&in6_carp_protosw); 2235 if (proto_reg[CARP_INET6]) { 2236 printf("carp: error %d attaching to PF_INET6\n", 2237 proto_reg[CARP_INET6]); 2238 carp_mod_cleanup(); 2239 return (proto_reg[CARP_INET6]); 2240 } 2241 err = ip6proto_register(IPPROTO_CARP); 2242 if (err) { 2243 printf("carp: error %d registering with INET6\n", err); 2244 carp_mod_cleanup(); 2245 return (err); 2246 } 2247 #endif 2248 #ifdef INET 2249 carp_iamatch_p = carp_iamatch; 2250 proto_reg[CARP_INET] = pf_proto_register(PF_INET, &in_carp_protosw); 2251 if (proto_reg[CARP_INET]) { 2252 printf("carp: error %d attaching to PF_INET\n", 2253 proto_reg[CARP_INET]); 2254 carp_mod_cleanup(); 2255 return (proto_reg[CARP_INET]); 2256 } 2257 err = ipproto_register(IPPROTO_CARP); 2258 if (err) { 2259 printf("carp: error %d registering with INET\n", err); 2260 carp_mod_cleanup(); 2261 return (err); 2262 } 2263 #endif 2264 return (0); 2265 } 2266 2267 static int 2268 carp_modevent(module_t mod, int type, void *data) 2269 { 2270 switch (type) { 2271 case MOD_LOAD: 2272 return carp_mod_load(); 2273 /* NOTREACHED */ 2274 case MOD_UNLOAD: 2275 mtx_lock(&carp_mtx); 2276 if (LIST_EMPTY(&carp_list)) 2277 carp_mod_cleanup(); 2278 else { 2279 mtx_unlock(&carp_mtx); 2280 return (EBUSY); 2281 } 2282 break; 2283 2284 default: 2285 return (EINVAL); 2286 } 2287 2288 return (0); 2289 } 2290 2291 static moduledata_t carp_mod = { 2292 "carp", 2293 carp_modevent, 2294 0 2295 }; 2296 2297 DECLARE_MODULE(carp, carp_mod, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY); 2298