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 m = m_pullup(m, *offp + sizeof(*ch)); 571 if (m == NULL) { 572 CARPSTATS_INC(carps_badlen); 573 CARP_DEBUG("%s: packet size %u too small\n", __func__, len); 574 return (IPPROTO_DONE); 575 } 576 ch = (struct carp_header *)(mtod(m, caddr_t) + *offp); 577 578 579 /* verify the CARP checksum */ 580 m->m_data += *offp; 581 if (in_cksum(m, sizeof(*ch))) { 582 CARPSTATS_INC(carps_badsum); 583 CARP_DEBUG("%s: checksum failed, on %s\n", __func__, 584 m->m_pkthdr.rcvif->if_xname); 585 m_freem(m); 586 return (IPPROTO_DONE); 587 } 588 m->m_data -= *offp; 589 590 carp_input_c(m, ch, AF_INET6); 591 return (IPPROTO_DONE); 592 } 593 #endif /* INET6 */ 594 595 /* 596 * This routine should not be necessary at all, but some switches 597 * (VMWare ESX vswitches) can echo our own packets back at us, 598 * and we must ignore them or they will cause us to drop out of 599 * MASTER mode. 600 * 601 * We cannot catch all cases of network loops. Instead, what we 602 * do here is catch any packet that arrives with a carp header 603 * with a VHID of 0, that comes from an address that is our own. 604 * These packets are by definition "from us" (even if they are from 605 * a misconfigured host that is pretending to be us). 606 * 607 * The VHID test is outside this mini-function. 608 */ 609 static int 610 carp_source_is_self(struct mbuf *m, struct ifaddr *ifa, sa_family_t af) 611 { 612 #ifdef INET 613 struct ip *ip4; 614 struct in_addr in4; 615 #endif 616 #ifdef INET6 617 struct ip6_hdr *ip6; 618 struct in6_addr in6; 619 #endif 620 621 switch (af) { 622 #ifdef INET 623 case AF_INET: 624 ip4 = mtod(m, struct ip *); 625 in4 = ifatoia(ifa)->ia_addr.sin_addr; 626 return (in4.s_addr == ip4->ip_src.s_addr); 627 #endif 628 #ifdef INET6 629 case AF_INET6: 630 ip6 = mtod(m, struct ip6_hdr *); 631 in6 = ifatoia6(ifa)->ia_addr.sin6_addr; 632 return (memcmp(&in6, &ip6->ip6_src, sizeof(in6)) == 0); 633 #endif 634 default: 635 break; 636 } 637 return (0); 638 } 639 640 static void 641 carp_input_c(struct mbuf *m, struct carp_header *ch, sa_family_t af) 642 { 643 struct ifnet *ifp = m->m_pkthdr.rcvif; 644 struct ifaddr *ifa, *match; 645 struct carp_softc *sc; 646 uint64_t tmp_counter; 647 struct timeval sc_tv, ch_tv; 648 int error; 649 650 NET_EPOCH_ASSERT(); 651 652 /* 653 * Verify that the VHID is valid on the receiving interface. 654 * 655 * There should be just one match. If there are none 656 * the VHID is not valid and we drop the packet. If 657 * there are multiple VHID matches, take just the first 658 * one, for compatibility with previous code. While we're 659 * scanning, check for obvious loops in the network topology 660 * (these should never happen, and as noted above, we may 661 * miss real loops; this is just a double-check). 662 */ 663 error = 0; 664 match = NULL; 665 IFNET_FOREACH_IFA(ifp, ifa) { 666 if (match == NULL && ifa->ifa_carp != NULL && 667 ifa->ifa_addr->sa_family == af && 668 ifa->ifa_carp->sc_vhid == ch->carp_vhid) 669 match = ifa; 670 if (ch->carp_vhid == 0 && carp_source_is_self(m, ifa, af)) 671 error = ELOOP; 672 } 673 ifa = error ? NULL : match; 674 if (ifa != NULL) 675 ifa_ref(ifa); 676 677 if (ifa == NULL) { 678 if (error == ELOOP) { 679 CARP_DEBUG("dropping looped packet on interface %s\n", 680 ifp->if_xname); 681 CARPSTATS_INC(carps_badif); /* ??? */ 682 } else { 683 CARPSTATS_INC(carps_badvhid); 684 } 685 m_freem(m); 686 return; 687 } 688 689 /* verify the CARP version. */ 690 if (ch->carp_version != CARP_VERSION) { 691 CARPSTATS_INC(carps_badver); 692 CARP_DEBUG("%s: invalid version %d\n", ifp->if_xname, 693 ch->carp_version); 694 ifa_free(ifa); 695 m_freem(m); 696 return; 697 } 698 699 sc = ifa->ifa_carp; 700 CARP_LOCK(sc); 701 ifa_free(ifa); 702 703 if (carp_hmac_verify(sc, ch->carp_counter, ch->carp_md)) { 704 CARPSTATS_INC(carps_badauth); 705 CARP_DEBUG("%s: incorrect hash for VHID %u@%s\n", __func__, 706 sc->sc_vhid, ifp->if_xname); 707 goto out; 708 } 709 710 tmp_counter = ntohl(ch->carp_counter[0]); 711 tmp_counter = tmp_counter<<32; 712 tmp_counter += ntohl(ch->carp_counter[1]); 713 714 /* XXX Replay protection goes here */ 715 716 sc->sc_init_counter = 0; 717 sc->sc_counter = tmp_counter; 718 719 sc_tv.tv_sec = sc->sc_advbase; 720 sc_tv.tv_usec = DEMOTE_ADVSKEW(sc) * 1000000 / 256; 721 ch_tv.tv_sec = ch->carp_advbase; 722 ch_tv.tv_usec = ch->carp_advskew * 1000000 / 256; 723 724 switch (sc->sc_state) { 725 case INIT: 726 break; 727 case MASTER: 728 /* 729 * If we receive an advertisement from a master who's going to 730 * be more frequent than us, go into BACKUP state. 731 */ 732 if (timevalcmp(&sc_tv, &ch_tv, >) || 733 timevalcmp(&sc_tv, &ch_tv, ==)) { 734 callout_stop(&sc->sc_ad_tmo); 735 carp_set_state(sc, BACKUP, 736 "more frequent advertisement received"); 737 carp_setrun(sc, 0); 738 carp_delroute(sc); 739 } 740 break; 741 case BACKUP: 742 /* 743 * If we're pre-empting masters who advertise slower than us, 744 * and this one claims to be slower, treat him as down. 745 */ 746 if (V_carp_preempt && timevalcmp(&sc_tv, &ch_tv, <)) { 747 carp_master_down_locked(sc, 748 "preempting a slower master"); 749 break; 750 } 751 752 /* 753 * If the master is going to advertise at such a low frequency 754 * that he's guaranteed to time out, we'd might as well just 755 * treat him as timed out now. 756 */ 757 sc_tv.tv_sec = sc->sc_advbase * 3; 758 if (timevalcmp(&sc_tv, &ch_tv, <)) { 759 carp_master_down_locked(sc, "master will time out"); 760 break; 761 } 762 763 /* 764 * Otherwise, we reset the counter and wait for the next 765 * advertisement. 766 */ 767 carp_setrun(sc, af); 768 break; 769 } 770 771 out: 772 CARP_UNLOCK(sc); 773 m_freem(m); 774 } 775 776 static int 777 carp_prepare_ad(struct mbuf *m, struct carp_softc *sc, struct carp_header *ch) 778 { 779 struct m_tag *mtag; 780 781 if (sc->sc_init_counter) { 782 /* this could also be seconds since unix epoch */ 783 sc->sc_counter = arc4random(); 784 sc->sc_counter = sc->sc_counter << 32; 785 sc->sc_counter += arc4random(); 786 } else 787 sc->sc_counter++; 788 789 ch->carp_counter[0] = htonl((sc->sc_counter>>32)&0xffffffff); 790 ch->carp_counter[1] = htonl(sc->sc_counter&0xffffffff); 791 792 carp_hmac_generate(sc, ch->carp_counter, ch->carp_md); 793 794 /* Tag packet for carp_output */ 795 if ((mtag = m_tag_get(PACKET_TAG_CARP, sizeof(struct carp_softc *), 796 M_NOWAIT)) == NULL) { 797 m_freem(m); 798 CARPSTATS_INC(carps_onomem); 799 return (ENOMEM); 800 } 801 bcopy(&sc, mtag + 1, sizeof(sc)); 802 m_tag_prepend(m, mtag); 803 804 return (0); 805 } 806 807 /* 808 * To avoid LORs and possible recursions this function shouldn't 809 * be called directly, but scheduled via taskqueue. 810 */ 811 static void 812 carp_send_ad_all(void *ctx __unused, int pending __unused) 813 { 814 struct carp_softc *sc; 815 816 mtx_lock(&carp_mtx); 817 LIST_FOREACH(sc, &carp_list, sc_next) 818 if (sc->sc_state == MASTER) { 819 CARP_LOCK(sc); 820 CURVNET_SET(sc->sc_carpdev->if_vnet); 821 carp_send_ad_locked(sc); 822 CURVNET_RESTORE(); 823 CARP_UNLOCK(sc); 824 } 825 mtx_unlock(&carp_mtx); 826 } 827 828 /* Send a periodic advertisement, executed in callout context. */ 829 static void 830 carp_send_ad(void *v) 831 { 832 struct carp_softc *sc = v; 833 834 CARP_LOCK_ASSERT(sc); 835 CURVNET_SET(sc->sc_carpdev->if_vnet); 836 carp_send_ad_locked(sc); 837 CURVNET_RESTORE(); 838 CARP_UNLOCK(sc); 839 } 840 841 static void 842 carp_send_ad_error(struct carp_softc *sc, int error) 843 { 844 845 if (error) { 846 if (sc->sc_sendad_errors < INT_MAX) 847 sc->sc_sendad_errors++; 848 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) { 849 static const char fmt[] = "send error %d on %s"; 850 char msg[sizeof(fmt) + IFNAMSIZ]; 851 852 sprintf(msg, fmt, error, sc->sc_carpdev->if_xname); 853 carp_demote_adj(V_carp_senderr_adj, msg); 854 } 855 sc->sc_sendad_success = 0; 856 } else { 857 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS && 858 ++sc->sc_sendad_success >= CARP_SENDAD_MIN_SUCCESS) { 859 static const char fmt[] = "send ok on %s"; 860 char msg[sizeof(fmt) + IFNAMSIZ]; 861 862 sprintf(msg, fmt, sc->sc_carpdev->if_xname); 863 carp_demote_adj(-V_carp_senderr_adj, msg); 864 sc->sc_sendad_errors = 0; 865 } else 866 sc->sc_sendad_errors = 0; 867 } 868 } 869 870 /* 871 * Pick the best ifaddr on the given ifp for sending CARP 872 * advertisements. 873 * 874 * "Best" here is defined by ifa_preferred(). This function is much 875 * much like ifaof_ifpforaddr() except that we just use ifa_preferred(). 876 * 877 * (This could be simplified to return the actual address, except that 878 * it has a different format in AF_INET and AF_INET6.) 879 */ 880 static struct ifaddr * 881 carp_best_ifa(int af, struct ifnet *ifp) 882 { 883 struct ifaddr *ifa, *best; 884 885 NET_EPOCH_ASSERT(); 886 887 if (af >= AF_MAX) 888 return (NULL); 889 best = NULL; 890 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 891 if (ifa->ifa_addr->sa_family == af && 892 (best == NULL || ifa_preferred(best, ifa))) 893 best = ifa; 894 } 895 if (best != NULL) 896 ifa_ref(best); 897 return (best); 898 } 899 900 static void 901 carp_send_ad_locked(struct carp_softc *sc) 902 { 903 struct carp_header ch; 904 struct timeval tv; 905 struct ifaddr *ifa; 906 struct carp_header *ch_ptr; 907 struct mbuf *m; 908 int len, advskew; 909 910 CARP_LOCK_ASSERT(sc); 911 912 advskew = DEMOTE_ADVSKEW(sc); 913 tv.tv_sec = sc->sc_advbase; 914 tv.tv_usec = advskew * 1000000 / 256; 915 916 ch.carp_version = CARP_VERSION; 917 ch.carp_type = CARP_ADVERTISEMENT; 918 ch.carp_vhid = sc->sc_vhid; 919 ch.carp_advbase = sc->sc_advbase; 920 ch.carp_advskew = advskew; 921 ch.carp_authlen = 7; /* XXX DEFINE */ 922 ch.carp_pad1 = 0; /* must be zero */ 923 ch.carp_cksum = 0; 924 925 /* XXXGL: OpenBSD picks first ifaddr with needed family. */ 926 927 #ifdef INET 928 if (sc->sc_naddrs) { 929 struct ip *ip; 930 931 m = m_gethdr(M_NOWAIT, MT_DATA); 932 if (m == NULL) { 933 CARPSTATS_INC(carps_onomem); 934 goto resched; 935 } 936 len = sizeof(*ip) + sizeof(ch); 937 m->m_pkthdr.len = len; 938 m->m_pkthdr.rcvif = NULL; 939 m->m_len = len; 940 M_ALIGN(m, m->m_len); 941 m->m_flags |= M_MCAST; 942 ip = mtod(m, struct ip *); 943 ip->ip_v = IPVERSION; 944 ip->ip_hl = sizeof(*ip) >> 2; 945 ip->ip_tos = V_carp_dscp << IPTOS_DSCP_OFFSET; 946 ip->ip_len = htons(len); 947 ip->ip_off = htons(IP_DF); 948 ip->ip_ttl = CARP_DFLTTL; 949 ip->ip_p = IPPROTO_CARP; 950 ip->ip_sum = 0; 951 ip_fillid(ip); 952 953 ifa = carp_best_ifa(AF_INET, sc->sc_carpdev); 954 if (ifa != NULL) { 955 ip->ip_src.s_addr = 956 ifatoia(ifa)->ia_addr.sin_addr.s_addr; 957 ifa_free(ifa); 958 } else 959 ip->ip_src.s_addr = 0; 960 ip->ip_dst.s_addr = htonl(INADDR_CARP_GROUP); 961 962 ch_ptr = (struct carp_header *)(&ip[1]); 963 bcopy(&ch, ch_ptr, sizeof(ch)); 964 if (carp_prepare_ad(m, sc, ch_ptr)) 965 goto resched; 966 967 m->m_data += sizeof(*ip); 968 ch_ptr->carp_cksum = in_cksum(m, len - sizeof(*ip)); 969 m->m_data -= sizeof(*ip); 970 971 CARPSTATS_INC(carps_opackets); 972 973 carp_send_ad_error(sc, ip_output(m, NULL, NULL, IP_RAWOUTPUT, 974 &sc->sc_carpdev->if_carp->cif_imo, NULL)); 975 } 976 #endif /* INET */ 977 #ifdef INET6 978 if (sc->sc_naddrs6) { 979 struct ip6_hdr *ip6; 980 981 m = m_gethdr(M_NOWAIT, MT_DATA); 982 if (m == NULL) { 983 CARPSTATS_INC(carps_onomem); 984 goto resched; 985 } 986 len = sizeof(*ip6) + sizeof(ch); 987 m->m_pkthdr.len = len; 988 m->m_pkthdr.rcvif = NULL; 989 m->m_len = len; 990 M_ALIGN(m, m->m_len); 991 m->m_flags |= M_MCAST; 992 ip6 = mtod(m, struct ip6_hdr *); 993 bzero(ip6, sizeof(*ip6)); 994 ip6->ip6_vfc |= IPV6_VERSION; 995 /* Traffic class isn't defined in ip6 struct instead 996 * it gets offset into flowid field */ 997 ip6->ip6_flow |= htonl(V_carp_dscp << (IPV6_FLOWLABEL_LEN + 998 IPTOS_DSCP_OFFSET)); 999 ip6->ip6_hlim = CARP_DFLTTL; 1000 ip6->ip6_nxt = IPPROTO_CARP; 1001 1002 /* set the source address */ 1003 ifa = carp_best_ifa(AF_INET6, sc->sc_carpdev); 1004 if (ifa != NULL) { 1005 bcopy(IFA_IN6(ifa), &ip6->ip6_src, 1006 sizeof(struct in6_addr)); 1007 ifa_free(ifa); 1008 } else 1009 /* This should never happen with IPv6. */ 1010 bzero(&ip6->ip6_src, sizeof(struct in6_addr)); 1011 1012 /* Set the multicast destination. */ 1013 ip6->ip6_dst.s6_addr16[0] = htons(0xff02); 1014 ip6->ip6_dst.s6_addr8[15] = 0x12; 1015 if (in6_setscope(&ip6->ip6_dst, sc->sc_carpdev, NULL) != 0) { 1016 m_freem(m); 1017 CARP_DEBUG("%s: in6_setscope failed\n", __func__); 1018 goto resched; 1019 } 1020 1021 ch_ptr = (struct carp_header *)(&ip6[1]); 1022 bcopy(&ch, ch_ptr, sizeof(ch)); 1023 if (carp_prepare_ad(m, sc, ch_ptr)) 1024 goto resched; 1025 1026 m->m_data += sizeof(*ip6); 1027 ch_ptr->carp_cksum = in_cksum(m, len - sizeof(*ip6)); 1028 m->m_data -= sizeof(*ip6); 1029 1030 CARPSTATS_INC(carps_opackets6); 1031 1032 carp_send_ad_error(sc, ip6_output(m, NULL, NULL, 0, 1033 &sc->sc_carpdev->if_carp->cif_im6o, NULL, NULL)); 1034 } 1035 #endif /* INET6 */ 1036 1037 resched: 1038 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv), carp_send_ad, sc); 1039 } 1040 1041 static void 1042 carp_addroute(struct carp_softc *sc) 1043 { 1044 struct ifaddr *ifa; 1045 1046 CARP_FOREACH_IFA(sc, ifa) 1047 carp_ifa_addroute(ifa); 1048 } 1049 1050 static void 1051 carp_ifa_addroute(struct ifaddr *ifa) 1052 { 1053 1054 switch (ifa->ifa_addr->sa_family) { 1055 #ifdef INET 1056 case AF_INET: 1057 in_addprefix(ifatoia(ifa), RTF_UP); 1058 ifa_add_loopback_route(ifa, 1059 (struct sockaddr *)&ifatoia(ifa)->ia_addr); 1060 break; 1061 #endif 1062 #ifdef INET6 1063 case AF_INET6: 1064 ifa_add_loopback_route(ifa, 1065 (struct sockaddr *)&ifatoia6(ifa)->ia_addr); 1066 nd6_add_ifa_lle(ifatoia6(ifa)); 1067 break; 1068 #endif 1069 } 1070 } 1071 1072 static void 1073 carp_delroute(struct carp_softc *sc) 1074 { 1075 struct ifaddr *ifa; 1076 1077 CARP_FOREACH_IFA(sc, ifa) 1078 carp_ifa_delroute(ifa); 1079 } 1080 1081 static void 1082 carp_ifa_delroute(struct ifaddr *ifa) 1083 { 1084 1085 switch (ifa->ifa_addr->sa_family) { 1086 #ifdef INET 1087 case AF_INET: 1088 ifa_del_loopback_route(ifa, 1089 (struct sockaddr *)&ifatoia(ifa)->ia_addr); 1090 in_scrubprefix(ifatoia(ifa), LLE_STATIC); 1091 break; 1092 #endif 1093 #ifdef INET6 1094 case AF_INET6: 1095 ifa_del_loopback_route(ifa, 1096 (struct sockaddr *)&ifatoia6(ifa)->ia_addr); 1097 nd6_rem_ifa_lle(ifatoia6(ifa), 1); 1098 break; 1099 #endif 1100 } 1101 } 1102 1103 int 1104 carp_master(struct ifaddr *ifa) 1105 { 1106 struct carp_softc *sc = ifa->ifa_carp; 1107 1108 return (sc->sc_state == MASTER); 1109 } 1110 1111 #ifdef INET 1112 /* 1113 * Broadcast a gratuitous ARP request containing 1114 * the virtual router MAC address for each IP address 1115 * associated with the virtual router. 1116 */ 1117 static void 1118 carp_send_arp(struct carp_softc *sc) 1119 { 1120 struct ifaddr *ifa; 1121 struct in_addr addr; 1122 1123 CARP_FOREACH_IFA(sc, ifa) { 1124 if (ifa->ifa_addr->sa_family != AF_INET) 1125 continue; 1126 addr = ((struct sockaddr_in *)ifa->ifa_addr)->sin_addr; 1127 arp_announce_ifaddr(sc->sc_carpdev, addr, LLADDR(&sc->sc_addr)); 1128 } 1129 } 1130 1131 int 1132 carp_iamatch(struct ifaddr *ifa, uint8_t **enaddr) 1133 { 1134 struct carp_softc *sc = ifa->ifa_carp; 1135 1136 if (sc->sc_state == MASTER) { 1137 *enaddr = LLADDR(&sc->sc_addr); 1138 return (1); 1139 } 1140 1141 return (0); 1142 } 1143 #endif 1144 1145 #ifdef INET6 1146 static void 1147 carp_send_na(struct carp_softc *sc) 1148 { 1149 static struct in6_addr mcast = IN6ADDR_LINKLOCAL_ALLNODES_INIT; 1150 struct ifaddr *ifa; 1151 struct in6_addr *in6; 1152 1153 CARP_FOREACH_IFA(sc, ifa) { 1154 if (ifa->ifa_addr->sa_family != AF_INET6) 1155 continue; 1156 1157 in6 = IFA_IN6(ifa); 1158 nd6_na_output(sc->sc_carpdev, &mcast, in6, 1159 ND_NA_FLAG_OVERRIDE, 1, NULL); 1160 DELAY(1000); /* XXX */ 1161 } 1162 } 1163 1164 /* 1165 * Returns ifa in case it's a carp address and it is MASTER, or if the address 1166 * matches and is not a carp address. Returns NULL otherwise. 1167 */ 1168 struct ifaddr * 1169 carp_iamatch6(struct ifnet *ifp, struct in6_addr *taddr) 1170 { 1171 struct ifaddr *ifa; 1172 1173 NET_EPOCH_ASSERT(); 1174 1175 ifa = NULL; 1176 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1177 if (ifa->ifa_addr->sa_family != AF_INET6) 1178 continue; 1179 if (!IN6_ARE_ADDR_EQUAL(taddr, IFA_IN6(ifa))) 1180 continue; 1181 if (ifa->ifa_carp && ifa->ifa_carp->sc_state != MASTER) 1182 ifa = NULL; 1183 else 1184 ifa_ref(ifa); 1185 break; 1186 } 1187 1188 return (ifa); 1189 } 1190 1191 caddr_t 1192 carp_macmatch6(struct ifnet *ifp, struct mbuf *m, const struct in6_addr *taddr) 1193 { 1194 struct ifaddr *ifa; 1195 1196 NET_EPOCH_ASSERT(); 1197 1198 IFNET_FOREACH_IFA(ifp, ifa) 1199 if (ifa->ifa_addr->sa_family == AF_INET6 && 1200 IN6_ARE_ADDR_EQUAL(taddr, IFA_IN6(ifa))) { 1201 struct carp_softc *sc = ifa->ifa_carp; 1202 struct m_tag *mtag; 1203 1204 mtag = m_tag_get(PACKET_TAG_CARP, 1205 sizeof(struct carp_softc *), M_NOWAIT); 1206 if (mtag == NULL) 1207 /* Better a bit than nothing. */ 1208 return (LLADDR(&sc->sc_addr)); 1209 1210 bcopy(&sc, mtag + 1, sizeof(sc)); 1211 m_tag_prepend(m, mtag); 1212 1213 return (LLADDR(&sc->sc_addr)); 1214 } 1215 1216 return (NULL); 1217 } 1218 #endif /* INET6 */ 1219 1220 int 1221 carp_forus(struct ifnet *ifp, u_char *dhost) 1222 { 1223 struct carp_softc *sc; 1224 uint8_t *ena = dhost; 1225 1226 if (ena[0] || ena[1] || ena[2] != 0x5e || ena[3] || ena[4] != 1) 1227 return (0); 1228 1229 CIF_LOCK(ifp->if_carp); 1230 IFNET_FOREACH_CARP(ifp, sc) { 1231 CARP_LOCK(sc); 1232 if (sc->sc_state == MASTER && !bcmp(dhost, LLADDR(&sc->sc_addr), 1233 ETHER_ADDR_LEN)) { 1234 CARP_UNLOCK(sc); 1235 CIF_UNLOCK(ifp->if_carp); 1236 return (1); 1237 } 1238 CARP_UNLOCK(sc); 1239 } 1240 CIF_UNLOCK(ifp->if_carp); 1241 1242 return (0); 1243 } 1244 1245 /* Master down timeout event, executed in callout context. */ 1246 static void 1247 carp_master_down(void *v) 1248 { 1249 struct carp_softc *sc = v; 1250 1251 CARP_LOCK_ASSERT(sc); 1252 1253 CURVNET_SET(sc->sc_carpdev->if_vnet); 1254 if (sc->sc_state == BACKUP) { 1255 carp_master_down_locked(sc, "master timed out"); 1256 } 1257 CURVNET_RESTORE(); 1258 1259 CARP_UNLOCK(sc); 1260 } 1261 1262 static void 1263 carp_master_down_locked(struct carp_softc *sc, const char *reason) 1264 { 1265 1266 CARP_LOCK_ASSERT(sc); 1267 1268 switch (sc->sc_state) { 1269 case BACKUP: 1270 carp_set_state(sc, MASTER, reason); 1271 carp_send_ad_locked(sc); 1272 #ifdef INET 1273 carp_send_arp(sc); 1274 #endif 1275 #ifdef INET6 1276 carp_send_na(sc); 1277 #endif 1278 carp_setrun(sc, 0); 1279 carp_addroute(sc); 1280 break; 1281 case INIT: 1282 case MASTER: 1283 #ifdef INVARIANTS 1284 panic("carp: VHID %u@%s: master_down event in %s state\n", 1285 sc->sc_vhid, 1286 sc->sc_carpdev->if_xname, 1287 sc->sc_state ? "MASTER" : "INIT"); 1288 #endif 1289 break; 1290 } 1291 } 1292 1293 /* 1294 * When in backup state, af indicates whether to reset the master down timer 1295 * for v4 or v6. If it's set to zero, reset the ones which are already pending. 1296 */ 1297 static void 1298 carp_setrun(struct carp_softc *sc, sa_family_t af) 1299 { 1300 struct timeval tv; 1301 1302 CARP_LOCK_ASSERT(sc); 1303 1304 if ((sc->sc_carpdev->if_flags & IFF_UP) == 0 || 1305 sc->sc_carpdev->if_link_state != LINK_STATE_UP || 1306 (sc->sc_naddrs == 0 && sc->sc_naddrs6 == 0) || 1307 !V_carp_allow) 1308 return; 1309 1310 switch (sc->sc_state) { 1311 case INIT: 1312 carp_set_state(sc, BACKUP, "initialization complete"); 1313 carp_setrun(sc, 0); 1314 break; 1315 case BACKUP: 1316 callout_stop(&sc->sc_ad_tmo); 1317 tv.tv_sec = 3 * sc->sc_advbase; 1318 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1319 switch (af) { 1320 #ifdef INET 1321 case AF_INET: 1322 callout_reset(&sc->sc_md_tmo, tvtohz(&tv), 1323 carp_master_down, sc); 1324 break; 1325 #endif 1326 #ifdef INET6 1327 case AF_INET6: 1328 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv), 1329 carp_master_down, sc); 1330 break; 1331 #endif 1332 default: 1333 #ifdef INET 1334 if (sc->sc_naddrs) 1335 callout_reset(&sc->sc_md_tmo, tvtohz(&tv), 1336 carp_master_down, sc); 1337 #endif 1338 #ifdef INET6 1339 if (sc->sc_naddrs6) 1340 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv), 1341 carp_master_down, sc); 1342 #endif 1343 break; 1344 } 1345 break; 1346 case MASTER: 1347 tv.tv_sec = sc->sc_advbase; 1348 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1349 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv), 1350 carp_send_ad, sc); 1351 break; 1352 } 1353 } 1354 1355 /* 1356 * Setup multicast structures. 1357 */ 1358 static int 1359 carp_multicast_setup(struct carp_if *cif, sa_family_t sa) 1360 { 1361 struct ifnet *ifp = cif->cif_ifp; 1362 int error = 0; 1363 1364 switch (sa) { 1365 #ifdef INET 1366 case AF_INET: 1367 { 1368 struct ip_moptions *imo = &cif->cif_imo; 1369 struct in_mfilter *imf; 1370 struct in_addr addr; 1371 1372 if (ip_mfilter_first(&imo->imo_head) != NULL) 1373 return (0); 1374 1375 imf = ip_mfilter_alloc(M_WAITOK, 0, 0); 1376 ip_mfilter_init(&imo->imo_head); 1377 imo->imo_multicast_vif = -1; 1378 1379 addr.s_addr = htonl(INADDR_CARP_GROUP); 1380 if ((error = in_joingroup(ifp, &addr, NULL, 1381 &imf->imf_inm)) != 0) { 1382 ip_mfilter_free(imf); 1383 break; 1384 } 1385 1386 ip_mfilter_insert(&imo->imo_head, imf); 1387 imo->imo_multicast_ifp = ifp; 1388 imo->imo_multicast_ttl = CARP_DFLTTL; 1389 imo->imo_multicast_loop = 0; 1390 break; 1391 } 1392 #endif 1393 #ifdef INET6 1394 case AF_INET6: 1395 { 1396 struct ip6_moptions *im6o = &cif->cif_im6o; 1397 struct in6_mfilter *im6f[2]; 1398 struct in6_addr in6; 1399 1400 if (ip6_mfilter_first(&im6o->im6o_head)) 1401 return (0); 1402 1403 im6f[0] = ip6_mfilter_alloc(M_WAITOK, 0, 0); 1404 im6f[1] = ip6_mfilter_alloc(M_WAITOK, 0, 0); 1405 1406 ip6_mfilter_init(&im6o->im6o_head); 1407 im6o->im6o_multicast_hlim = CARP_DFLTTL; 1408 im6o->im6o_multicast_ifp = ifp; 1409 1410 /* Join IPv6 CARP multicast group. */ 1411 bzero(&in6, sizeof(in6)); 1412 in6.s6_addr16[0] = htons(0xff02); 1413 in6.s6_addr8[15] = 0x12; 1414 if ((error = in6_setscope(&in6, ifp, NULL)) != 0) { 1415 ip6_mfilter_free(im6f[0]); 1416 ip6_mfilter_free(im6f[1]); 1417 break; 1418 } 1419 if ((error = in6_joingroup(ifp, &in6, NULL, &im6f[0]->im6f_in6m, 0)) != 0) { 1420 ip6_mfilter_free(im6f[0]); 1421 ip6_mfilter_free(im6f[1]); 1422 break; 1423 } 1424 1425 /* Join solicited multicast address. */ 1426 bzero(&in6, sizeof(in6)); 1427 in6.s6_addr16[0] = htons(0xff02); 1428 in6.s6_addr32[1] = 0; 1429 in6.s6_addr32[2] = htonl(1); 1430 in6.s6_addr32[3] = 0; 1431 in6.s6_addr8[12] = 0xff; 1432 1433 if ((error = in6_setscope(&in6, ifp, NULL)) != 0) { 1434 ip6_mfilter_free(im6f[0]); 1435 ip6_mfilter_free(im6f[1]); 1436 break; 1437 } 1438 1439 if ((error = in6_joingroup(ifp, &in6, NULL, &im6f[1]->im6f_in6m, 0)) != 0) { 1440 in6_leavegroup(im6f[0]->im6f_in6m, NULL); 1441 ip6_mfilter_free(im6f[0]); 1442 ip6_mfilter_free(im6f[1]); 1443 break; 1444 } 1445 ip6_mfilter_insert(&im6o->im6o_head, im6f[0]); 1446 ip6_mfilter_insert(&im6o->im6o_head, im6f[1]); 1447 break; 1448 } 1449 #endif 1450 } 1451 1452 return (error); 1453 } 1454 1455 /* 1456 * Free multicast structures. 1457 */ 1458 static void 1459 carp_multicast_cleanup(struct carp_if *cif, sa_family_t sa) 1460 { 1461 #ifdef INET 1462 struct ip_moptions *imo = &cif->cif_imo; 1463 struct in_mfilter *imf; 1464 #endif 1465 #ifdef INET6 1466 struct ip6_moptions *im6o = &cif->cif_im6o; 1467 struct in6_mfilter *im6f; 1468 #endif 1469 sx_assert(&carp_sx, SA_XLOCKED); 1470 1471 switch (sa) { 1472 #ifdef INET 1473 case AF_INET: 1474 if (cif->cif_naddrs != 0) 1475 break; 1476 1477 while ((imf = ip_mfilter_first(&imo->imo_head)) != NULL) { 1478 ip_mfilter_remove(&imo->imo_head, imf); 1479 in_leavegroup(imf->imf_inm, NULL); 1480 ip_mfilter_free(imf); 1481 } 1482 break; 1483 #endif 1484 #ifdef INET6 1485 case AF_INET6: 1486 if (cif->cif_naddrs6 != 0) 1487 break; 1488 1489 while ((im6f = ip6_mfilter_first(&im6o->im6o_head)) != NULL) { 1490 ip6_mfilter_remove(&im6o->im6o_head, im6f); 1491 in6_leavegroup(im6f->im6f_in6m, NULL); 1492 ip6_mfilter_free(im6f); 1493 } 1494 break; 1495 #endif 1496 } 1497 } 1498 1499 int 1500 carp_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa) 1501 { 1502 struct m_tag *mtag; 1503 struct carp_softc *sc; 1504 1505 if (!sa) 1506 return (0); 1507 1508 switch (sa->sa_family) { 1509 #ifdef INET 1510 case AF_INET: 1511 break; 1512 #endif 1513 #ifdef INET6 1514 case AF_INET6: 1515 break; 1516 #endif 1517 default: 1518 return (0); 1519 } 1520 1521 mtag = m_tag_find(m, PACKET_TAG_CARP, NULL); 1522 if (mtag == NULL) 1523 return (0); 1524 1525 bcopy(mtag + 1, &sc, sizeof(sc)); 1526 1527 /* Set the source MAC address to the Virtual Router MAC Address. */ 1528 switch (ifp->if_type) { 1529 case IFT_ETHER: 1530 case IFT_BRIDGE: 1531 case IFT_L2VLAN: { 1532 struct ether_header *eh; 1533 1534 eh = mtod(m, struct ether_header *); 1535 eh->ether_shost[0] = 0; 1536 eh->ether_shost[1] = 0; 1537 eh->ether_shost[2] = 0x5e; 1538 eh->ether_shost[3] = 0; 1539 eh->ether_shost[4] = 1; 1540 eh->ether_shost[5] = sc->sc_vhid; 1541 } 1542 break; 1543 default: 1544 printf("%s: carp is not supported for the %d interface type\n", 1545 ifp->if_xname, ifp->if_type); 1546 return (EOPNOTSUPP); 1547 } 1548 1549 return (0); 1550 } 1551 1552 static struct carp_softc* 1553 carp_alloc(struct ifnet *ifp) 1554 { 1555 struct carp_softc *sc; 1556 struct carp_if *cif; 1557 1558 sx_assert(&carp_sx, SA_XLOCKED); 1559 1560 if ((cif = ifp->if_carp) == NULL) 1561 cif = carp_alloc_if(ifp); 1562 1563 sc = malloc(sizeof(*sc), M_CARP, M_WAITOK|M_ZERO); 1564 1565 sc->sc_advbase = CARP_DFLTINTV; 1566 sc->sc_vhid = -1; /* required setting */ 1567 sc->sc_init_counter = 1; 1568 sc->sc_state = INIT; 1569 1570 sc->sc_ifasiz = sizeof(struct ifaddr *); 1571 sc->sc_ifas = malloc(sc->sc_ifasiz, M_CARP, M_WAITOK|M_ZERO); 1572 sc->sc_carpdev = ifp; 1573 1574 CARP_LOCK_INIT(sc); 1575 #ifdef INET 1576 callout_init_mtx(&sc->sc_md_tmo, &sc->sc_mtx, CALLOUT_RETURNUNLOCKED); 1577 #endif 1578 #ifdef INET6 1579 callout_init_mtx(&sc->sc_md6_tmo, &sc->sc_mtx, CALLOUT_RETURNUNLOCKED); 1580 #endif 1581 callout_init_mtx(&sc->sc_ad_tmo, &sc->sc_mtx, CALLOUT_RETURNUNLOCKED); 1582 1583 CIF_LOCK(cif); 1584 TAILQ_INSERT_TAIL(&cif->cif_vrs, sc, sc_list); 1585 CIF_UNLOCK(cif); 1586 1587 mtx_lock(&carp_mtx); 1588 LIST_INSERT_HEAD(&carp_list, sc, sc_next); 1589 mtx_unlock(&carp_mtx); 1590 1591 return (sc); 1592 } 1593 1594 static void 1595 carp_grow_ifas(struct carp_softc *sc) 1596 { 1597 struct ifaddr **new; 1598 1599 new = malloc(sc->sc_ifasiz * 2, M_CARP, M_WAITOK | M_ZERO); 1600 CARP_LOCK(sc); 1601 bcopy(sc->sc_ifas, new, sc->sc_ifasiz); 1602 free(sc->sc_ifas, M_CARP); 1603 sc->sc_ifas = new; 1604 sc->sc_ifasiz *= 2; 1605 CARP_UNLOCK(sc); 1606 } 1607 1608 static void 1609 carp_destroy(struct carp_softc *sc) 1610 { 1611 struct ifnet *ifp = sc->sc_carpdev; 1612 struct carp_if *cif = ifp->if_carp; 1613 1614 sx_assert(&carp_sx, SA_XLOCKED); 1615 1616 if (sc->sc_suppress) 1617 carp_demote_adj(-V_carp_ifdown_adj, "vhid removed"); 1618 CARP_UNLOCK(sc); 1619 1620 CIF_LOCK(cif); 1621 TAILQ_REMOVE(&cif->cif_vrs, sc, sc_list); 1622 CIF_UNLOCK(cif); 1623 1624 mtx_lock(&carp_mtx); 1625 LIST_REMOVE(sc, sc_next); 1626 mtx_unlock(&carp_mtx); 1627 1628 callout_drain(&sc->sc_ad_tmo); 1629 #ifdef INET 1630 callout_drain(&sc->sc_md_tmo); 1631 #endif 1632 #ifdef INET6 1633 callout_drain(&sc->sc_md6_tmo); 1634 #endif 1635 CARP_LOCK_DESTROY(sc); 1636 1637 free(sc->sc_ifas, M_CARP); 1638 free(sc, M_CARP); 1639 } 1640 1641 static struct carp_if* 1642 carp_alloc_if(struct ifnet *ifp) 1643 { 1644 struct carp_if *cif; 1645 int error; 1646 1647 cif = malloc(sizeof(*cif), M_CARP, M_WAITOK|M_ZERO); 1648 1649 if ((error = ifpromisc(ifp, 1)) != 0) 1650 printf("%s: ifpromisc(%s) failed: %d\n", 1651 __func__, ifp->if_xname, error); 1652 else 1653 cif->cif_flags |= CIF_PROMISC; 1654 1655 CIF_LOCK_INIT(cif); 1656 cif->cif_ifp = ifp; 1657 TAILQ_INIT(&cif->cif_vrs); 1658 1659 IF_ADDR_WLOCK(ifp); 1660 ifp->if_carp = cif; 1661 if_ref(ifp); 1662 IF_ADDR_WUNLOCK(ifp); 1663 1664 return (cif); 1665 } 1666 1667 static void 1668 carp_free_if(struct carp_if *cif) 1669 { 1670 struct ifnet *ifp = cif->cif_ifp; 1671 1672 CIF_LOCK_ASSERT(cif); 1673 KASSERT(TAILQ_EMPTY(&cif->cif_vrs), ("%s: softc list not empty", 1674 __func__)); 1675 1676 IF_ADDR_WLOCK(ifp); 1677 ifp->if_carp = NULL; 1678 IF_ADDR_WUNLOCK(ifp); 1679 1680 CIF_LOCK_DESTROY(cif); 1681 1682 if (cif->cif_flags & CIF_PROMISC) 1683 ifpromisc(ifp, 0); 1684 if_rele(ifp); 1685 1686 free(cif, M_CARP); 1687 } 1688 1689 static void 1690 carp_carprcp(struct carpreq *carpr, struct carp_softc *sc, int priv) 1691 { 1692 1693 CARP_LOCK(sc); 1694 carpr->carpr_state = sc->sc_state; 1695 carpr->carpr_vhid = sc->sc_vhid; 1696 carpr->carpr_advbase = sc->sc_advbase; 1697 carpr->carpr_advskew = sc->sc_advskew; 1698 if (priv) 1699 bcopy(sc->sc_key, carpr->carpr_key, sizeof(carpr->carpr_key)); 1700 else 1701 bzero(carpr->carpr_key, sizeof(carpr->carpr_key)); 1702 CARP_UNLOCK(sc); 1703 } 1704 1705 int 1706 carp_ioctl(struct ifreq *ifr, u_long cmd, struct thread *td) 1707 { 1708 struct carpreq carpr; 1709 struct ifnet *ifp; 1710 struct carp_softc *sc = NULL; 1711 int error = 0, locked = 0; 1712 1713 if ((error = copyin(ifr_data_get_ptr(ifr), &carpr, sizeof carpr))) 1714 return (error); 1715 1716 ifp = ifunit_ref(ifr->ifr_name); 1717 if (ifp == NULL) 1718 return (ENXIO); 1719 1720 switch (ifp->if_type) { 1721 case IFT_ETHER: 1722 case IFT_L2VLAN: 1723 case IFT_BRIDGE: 1724 break; 1725 default: 1726 error = EOPNOTSUPP; 1727 goto out; 1728 } 1729 1730 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 1731 error = EADDRNOTAVAIL; 1732 goto out; 1733 } 1734 1735 sx_xlock(&carp_sx); 1736 switch (cmd) { 1737 case SIOCSVH: 1738 if ((error = priv_check(td, PRIV_NETINET_CARP))) 1739 break; 1740 if (carpr.carpr_vhid <= 0 || carpr.carpr_vhid > CARP_MAXVHID || 1741 carpr.carpr_advbase < 0 || carpr.carpr_advskew < 0) { 1742 error = EINVAL; 1743 break; 1744 } 1745 1746 if (ifp->if_carp) { 1747 IFNET_FOREACH_CARP(ifp, sc) 1748 if (sc->sc_vhid == carpr.carpr_vhid) 1749 break; 1750 } 1751 if (sc == NULL) { 1752 sc = carp_alloc(ifp); 1753 CARP_LOCK(sc); 1754 sc->sc_vhid = carpr.carpr_vhid; 1755 LLADDR(&sc->sc_addr)[0] = 0; 1756 LLADDR(&sc->sc_addr)[1] = 0; 1757 LLADDR(&sc->sc_addr)[2] = 0x5e; 1758 LLADDR(&sc->sc_addr)[3] = 0; 1759 LLADDR(&sc->sc_addr)[4] = 1; 1760 LLADDR(&sc->sc_addr)[5] = sc->sc_vhid; 1761 } else 1762 CARP_LOCK(sc); 1763 locked = 1; 1764 if (carpr.carpr_advbase > 0) { 1765 if (carpr.carpr_advbase > 255 || 1766 carpr.carpr_advbase < CARP_DFLTINTV) { 1767 error = EINVAL; 1768 break; 1769 } 1770 sc->sc_advbase = carpr.carpr_advbase; 1771 } 1772 if (carpr.carpr_advskew >= 255) { 1773 error = EINVAL; 1774 break; 1775 } 1776 sc->sc_advskew = carpr.carpr_advskew; 1777 if (carpr.carpr_key[0] != '\0') { 1778 bcopy(carpr.carpr_key, sc->sc_key, sizeof(sc->sc_key)); 1779 carp_hmac_prepare(sc); 1780 } 1781 if (sc->sc_state != INIT && 1782 carpr.carpr_state != sc->sc_state) { 1783 switch (carpr.carpr_state) { 1784 case BACKUP: 1785 callout_stop(&sc->sc_ad_tmo); 1786 carp_set_state(sc, BACKUP, 1787 "user requested via ifconfig"); 1788 carp_setrun(sc, 0); 1789 carp_delroute(sc); 1790 break; 1791 case MASTER: 1792 carp_master_down_locked(sc, 1793 "user requested via ifconfig"); 1794 break; 1795 default: 1796 break; 1797 } 1798 } 1799 break; 1800 1801 case SIOCGVH: 1802 { 1803 int priveleged; 1804 1805 if (carpr.carpr_vhid < 0 || carpr.carpr_vhid > CARP_MAXVHID) { 1806 error = EINVAL; 1807 break; 1808 } 1809 if (carpr.carpr_count < 1) { 1810 error = EMSGSIZE; 1811 break; 1812 } 1813 if (ifp->if_carp == NULL) { 1814 error = ENOENT; 1815 break; 1816 } 1817 1818 priveleged = (priv_check(td, PRIV_NETINET_CARP) == 0); 1819 if (carpr.carpr_vhid != 0) { 1820 IFNET_FOREACH_CARP(ifp, sc) 1821 if (sc->sc_vhid == carpr.carpr_vhid) 1822 break; 1823 if (sc == NULL) { 1824 error = ENOENT; 1825 break; 1826 } 1827 carp_carprcp(&carpr, sc, priveleged); 1828 error = copyout(&carpr, ifr_data_get_ptr(ifr), 1829 sizeof(carpr)); 1830 } else { 1831 int i, count; 1832 1833 count = 0; 1834 IFNET_FOREACH_CARP(ifp, sc) 1835 count++; 1836 1837 if (count > carpr.carpr_count) { 1838 CIF_UNLOCK(ifp->if_carp); 1839 error = EMSGSIZE; 1840 break; 1841 } 1842 1843 i = 0; 1844 IFNET_FOREACH_CARP(ifp, sc) { 1845 carp_carprcp(&carpr, sc, priveleged); 1846 carpr.carpr_count = count; 1847 error = copyout(&carpr, 1848 (caddr_t)ifr_data_get_ptr(ifr) + 1849 (i * sizeof(carpr)), sizeof(carpr)); 1850 if (error) { 1851 CIF_UNLOCK(ifp->if_carp); 1852 break; 1853 } 1854 i++; 1855 } 1856 } 1857 break; 1858 } 1859 default: 1860 error = EINVAL; 1861 } 1862 sx_xunlock(&carp_sx); 1863 1864 out: 1865 if (locked) 1866 CARP_UNLOCK(sc); 1867 if_rele(ifp); 1868 1869 return (error); 1870 } 1871 1872 static int 1873 carp_get_vhid(struct ifaddr *ifa) 1874 { 1875 1876 if (ifa == NULL || ifa->ifa_carp == NULL) 1877 return (0); 1878 1879 return (ifa->ifa_carp->sc_vhid); 1880 } 1881 1882 int 1883 carp_attach(struct ifaddr *ifa, int vhid) 1884 { 1885 struct ifnet *ifp = ifa->ifa_ifp; 1886 struct carp_if *cif = ifp->if_carp; 1887 struct carp_softc *sc; 1888 int index, error; 1889 1890 KASSERT(ifa->ifa_carp == NULL, ("%s: ifa %p attached", __func__, ifa)); 1891 1892 switch (ifa->ifa_addr->sa_family) { 1893 #ifdef INET 1894 case AF_INET: 1895 #endif 1896 #ifdef INET6 1897 case AF_INET6: 1898 #endif 1899 break; 1900 default: 1901 return (EPROTOTYPE); 1902 } 1903 1904 sx_xlock(&carp_sx); 1905 if (ifp->if_carp == NULL) { 1906 sx_xunlock(&carp_sx); 1907 return (ENOPROTOOPT); 1908 } 1909 1910 IFNET_FOREACH_CARP(ifp, sc) 1911 if (sc->sc_vhid == vhid) 1912 break; 1913 if (sc == NULL) { 1914 sx_xunlock(&carp_sx); 1915 return (ENOENT); 1916 } 1917 1918 error = carp_multicast_setup(cif, ifa->ifa_addr->sa_family); 1919 if (error) { 1920 CIF_FREE(cif); 1921 sx_xunlock(&carp_sx); 1922 return (error); 1923 } 1924 1925 index = sc->sc_naddrs + sc->sc_naddrs6 + 1; 1926 if (index > sc->sc_ifasiz / sizeof(struct ifaddr *)) 1927 carp_grow_ifas(sc); 1928 1929 switch (ifa->ifa_addr->sa_family) { 1930 #ifdef INET 1931 case AF_INET: 1932 cif->cif_naddrs++; 1933 sc->sc_naddrs++; 1934 break; 1935 #endif 1936 #ifdef INET6 1937 case AF_INET6: 1938 cif->cif_naddrs6++; 1939 sc->sc_naddrs6++; 1940 break; 1941 #endif 1942 } 1943 1944 ifa_ref(ifa); 1945 1946 CARP_LOCK(sc); 1947 sc->sc_ifas[index - 1] = ifa; 1948 ifa->ifa_carp = sc; 1949 carp_hmac_prepare(sc); 1950 carp_sc_state(sc); 1951 CARP_UNLOCK(sc); 1952 1953 sx_xunlock(&carp_sx); 1954 1955 return (0); 1956 } 1957 1958 void 1959 carp_detach(struct ifaddr *ifa, bool keep_cif) 1960 { 1961 struct ifnet *ifp = ifa->ifa_ifp; 1962 struct carp_if *cif = ifp->if_carp; 1963 struct carp_softc *sc = ifa->ifa_carp; 1964 int i, index; 1965 1966 KASSERT(sc != NULL, ("%s: %p not attached", __func__, ifa)); 1967 1968 sx_xlock(&carp_sx); 1969 1970 CARP_LOCK(sc); 1971 /* Shift array. */ 1972 index = sc->sc_naddrs + sc->sc_naddrs6; 1973 for (i = 0; i < index; i++) 1974 if (sc->sc_ifas[i] == ifa) 1975 break; 1976 KASSERT(i < index, ("%s: %p no backref", __func__, ifa)); 1977 for (; i < index - 1; i++) 1978 sc->sc_ifas[i] = sc->sc_ifas[i+1]; 1979 sc->sc_ifas[index - 1] = NULL; 1980 1981 switch (ifa->ifa_addr->sa_family) { 1982 #ifdef INET 1983 case AF_INET: 1984 cif->cif_naddrs--; 1985 sc->sc_naddrs--; 1986 break; 1987 #endif 1988 #ifdef INET6 1989 case AF_INET6: 1990 cif->cif_naddrs6--; 1991 sc->sc_naddrs6--; 1992 break; 1993 #endif 1994 } 1995 1996 carp_ifa_delroute(ifa); 1997 carp_multicast_cleanup(cif, ifa->ifa_addr->sa_family); 1998 1999 ifa->ifa_carp = NULL; 2000 ifa_free(ifa); 2001 2002 carp_hmac_prepare(sc); 2003 carp_sc_state(sc); 2004 2005 if (!keep_cif && sc->sc_naddrs == 0 && sc->sc_naddrs6 == 0) 2006 carp_destroy(sc); 2007 else 2008 CARP_UNLOCK(sc); 2009 2010 if (!keep_cif) 2011 CIF_FREE(cif); 2012 2013 sx_xunlock(&carp_sx); 2014 } 2015 2016 static void 2017 carp_set_state(struct carp_softc *sc, int state, const char *reason) 2018 { 2019 2020 CARP_LOCK_ASSERT(sc); 2021 2022 if (sc->sc_state != state) { 2023 const char *carp_states[] = { CARP_STATES }; 2024 char subsys[IFNAMSIZ+5]; 2025 2026 snprintf(subsys, IFNAMSIZ+5, "%u@%s", sc->sc_vhid, 2027 sc->sc_carpdev->if_xname); 2028 2029 CARP_LOG("%s: %s -> %s (%s)\n", subsys, 2030 carp_states[sc->sc_state], carp_states[state], reason); 2031 2032 sc->sc_state = state; 2033 2034 devctl_notify("CARP", subsys, carp_states[state], NULL); 2035 } 2036 } 2037 2038 static void 2039 carp_linkstate(struct ifnet *ifp) 2040 { 2041 struct carp_softc *sc; 2042 2043 CIF_LOCK(ifp->if_carp); 2044 IFNET_FOREACH_CARP(ifp, sc) { 2045 CARP_LOCK(sc); 2046 carp_sc_state(sc); 2047 CARP_UNLOCK(sc); 2048 } 2049 CIF_UNLOCK(ifp->if_carp); 2050 } 2051 2052 static void 2053 carp_sc_state(struct carp_softc *sc) 2054 { 2055 2056 CARP_LOCK_ASSERT(sc); 2057 2058 if (sc->sc_carpdev->if_link_state != LINK_STATE_UP || 2059 !(sc->sc_carpdev->if_flags & IFF_UP) || 2060 !V_carp_allow) { 2061 callout_stop(&sc->sc_ad_tmo); 2062 #ifdef INET 2063 callout_stop(&sc->sc_md_tmo); 2064 #endif 2065 #ifdef INET6 2066 callout_stop(&sc->sc_md6_tmo); 2067 #endif 2068 carp_set_state(sc, INIT, "hardware interface down"); 2069 carp_setrun(sc, 0); 2070 if (!sc->sc_suppress) 2071 carp_demote_adj(V_carp_ifdown_adj, "interface down"); 2072 sc->sc_suppress = 1; 2073 } else { 2074 carp_set_state(sc, INIT, "hardware interface up"); 2075 carp_setrun(sc, 0); 2076 if (sc->sc_suppress) 2077 carp_demote_adj(-V_carp_ifdown_adj, "interface up"); 2078 sc->sc_suppress = 0; 2079 } 2080 } 2081 2082 static void 2083 carp_demote_adj(int adj, char *reason) 2084 { 2085 atomic_add_int(&V_carp_demotion, adj); 2086 CARP_LOG("demoted by %d to %d (%s)\n", adj, V_carp_demotion, reason); 2087 taskqueue_enqueue(taskqueue_swi, &carp_sendall_task); 2088 } 2089 2090 static int 2091 carp_allow_sysctl(SYSCTL_HANDLER_ARGS) 2092 { 2093 int new, error; 2094 struct carp_softc *sc; 2095 2096 new = V_carp_allow; 2097 error = sysctl_handle_int(oidp, &new, 0, req); 2098 if (error || !req->newptr) 2099 return (error); 2100 2101 if (V_carp_allow != new) { 2102 V_carp_allow = new; 2103 2104 mtx_lock(&carp_mtx); 2105 LIST_FOREACH(sc, &carp_list, sc_next) { 2106 CARP_LOCK(sc); 2107 if (curvnet == sc->sc_carpdev->if_vnet) 2108 carp_sc_state(sc); 2109 CARP_UNLOCK(sc); 2110 } 2111 mtx_unlock(&carp_mtx); 2112 } 2113 2114 return (0); 2115 } 2116 2117 static int 2118 carp_dscp_sysctl(SYSCTL_HANDLER_ARGS) 2119 { 2120 int new, error; 2121 2122 new = V_carp_dscp; 2123 error = sysctl_handle_int(oidp, &new, 0, req); 2124 if (error || !req->newptr) 2125 return (error); 2126 2127 if (new < 0 || new > 63) 2128 return (EINVAL); 2129 2130 V_carp_dscp = new; 2131 2132 return (0); 2133 } 2134 2135 static int 2136 carp_demote_adj_sysctl(SYSCTL_HANDLER_ARGS) 2137 { 2138 int new, error; 2139 2140 new = V_carp_demotion; 2141 error = sysctl_handle_int(oidp, &new, 0, req); 2142 if (error || !req->newptr) 2143 return (error); 2144 2145 carp_demote_adj(new, "sysctl"); 2146 2147 return (0); 2148 } 2149 2150 #ifdef INET 2151 extern struct domain inetdomain; 2152 static struct protosw in_carp_protosw = { 2153 .pr_type = SOCK_RAW, 2154 .pr_domain = &inetdomain, 2155 .pr_protocol = IPPROTO_CARP, 2156 .pr_flags = PR_ATOMIC|PR_ADDR, 2157 .pr_input = carp_input, 2158 .pr_output = rip_output, 2159 .pr_ctloutput = rip_ctloutput, 2160 .pr_usrreqs = &rip_usrreqs 2161 }; 2162 #endif 2163 2164 #ifdef INET6 2165 extern struct domain inet6domain; 2166 static struct protosw in6_carp_protosw = { 2167 .pr_type = SOCK_RAW, 2168 .pr_domain = &inet6domain, 2169 .pr_protocol = IPPROTO_CARP, 2170 .pr_flags = PR_ATOMIC|PR_ADDR, 2171 .pr_input = carp6_input, 2172 .pr_output = rip6_output, 2173 .pr_ctloutput = rip6_ctloutput, 2174 .pr_usrreqs = &rip6_usrreqs 2175 }; 2176 #endif 2177 2178 static void 2179 carp_mod_cleanup(void) 2180 { 2181 2182 #ifdef INET 2183 if (proto_reg[CARP_INET] == 0) { 2184 (void)ipproto_unregister(IPPROTO_CARP); 2185 pf_proto_unregister(PF_INET, IPPROTO_CARP, SOCK_RAW); 2186 proto_reg[CARP_INET] = -1; 2187 } 2188 carp_iamatch_p = NULL; 2189 #endif 2190 #ifdef INET6 2191 if (proto_reg[CARP_INET6] == 0) { 2192 (void)ip6proto_unregister(IPPROTO_CARP); 2193 pf_proto_unregister(PF_INET6, IPPROTO_CARP, SOCK_RAW); 2194 proto_reg[CARP_INET6] = -1; 2195 } 2196 carp_iamatch6_p = NULL; 2197 carp_macmatch6_p = NULL; 2198 #endif 2199 carp_ioctl_p = NULL; 2200 carp_attach_p = NULL; 2201 carp_detach_p = NULL; 2202 carp_get_vhid_p = NULL; 2203 carp_linkstate_p = NULL; 2204 carp_forus_p = NULL; 2205 carp_output_p = NULL; 2206 carp_demote_adj_p = NULL; 2207 carp_master_p = NULL; 2208 mtx_unlock(&carp_mtx); 2209 taskqueue_drain(taskqueue_swi, &carp_sendall_task); 2210 mtx_destroy(&carp_mtx); 2211 sx_destroy(&carp_sx); 2212 } 2213 2214 static int 2215 carp_mod_load(void) 2216 { 2217 int err; 2218 2219 mtx_init(&carp_mtx, "carp_mtx", NULL, MTX_DEF); 2220 sx_init(&carp_sx, "carp_sx"); 2221 LIST_INIT(&carp_list); 2222 carp_get_vhid_p = carp_get_vhid; 2223 carp_forus_p = carp_forus; 2224 carp_output_p = carp_output; 2225 carp_linkstate_p = carp_linkstate; 2226 carp_ioctl_p = carp_ioctl; 2227 carp_attach_p = carp_attach; 2228 carp_detach_p = carp_detach; 2229 carp_demote_adj_p = carp_demote_adj; 2230 carp_master_p = carp_master; 2231 #ifdef INET6 2232 carp_iamatch6_p = carp_iamatch6; 2233 carp_macmatch6_p = carp_macmatch6; 2234 proto_reg[CARP_INET6] = pf_proto_register(PF_INET6, 2235 (struct protosw *)&in6_carp_protosw); 2236 if (proto_reg[CARP_INET6]) { 2237 printf("carp: error %d attaching to PF_INET6\n", 2238 proto_reg[CARP_INET6]); 2239 carp_mod_cleanup(); 2240 return (proto_reg[CARP_INET6]); 2241 } 2242 err = ip6proto_register(IPPROTO_CARP); 2243 if (err) { 2244 printf("carp: error %d registering with INET6\n", err); 2245 carp_mod_cleanup(); 2246 return (err); 2247 } 2248 #endif 2249 #ifdef INET 2250 carp_iamatch_p = carp_iamatch; 2251 proto_reg[CARP_INET] = pf_proto_register(PF_INET, &in_carp_protosw); 2252 if (proto_reg[CARP_INET]) { 2253 printf("carp: error %d attaching to PF_INET\n", 2254 proto_reg[CARP_INET]); 2255 carp_mod_cleanup(); 2256 return (proto_reg[CARP_INET]); 2257 } 2258 err = ipproto_register(IPPROTO_CARP); 2259 if (err) { 2260 printf("carp: error %d registering with INET\n", err); 2261 carp_mod_cleanup(); 2262 return (err); 2263 } 2264 #endif 2265 return (0); 2266 } 2267 2268 static int 2269 carp_modevent(module_t mod, int type, void *data) 2270 { 2271 switch (type) { 2272 case MOD_LOAD: 2273 return carp_mod_load(); 2274 /* NOTREACHED */ 2275 case MOD_UNLOAD: 2276 mtx_lock(&carp_mtx); 2277 if (LIST_EMPTY(&carp_list)) 2278 carp_mod_cleanup(); 2279 else { 2280 mtx_unlock(&carp_mtx); 2281 return (EBUSY); 2282 } 2283 break; 2284 2285 default: 2286 return (EINVAL); 2287 } 2288 2289 return (0); 2290 } 2291 2292 static moduledata_t carp_mod = { 2293 "carp", 2294 carp_modevent, 2295 0 2296 }; 2297 2298 DECLARE_MODULE(carp, carp_mod, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY); 2299