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 if (m->m_len < *offp + sizeof(*ch)) { 570 len = m->m_len; 571 m = m_pullup(m, *offp + sizeof(*ch)); 572 if (m == NULL) { 573 CARPSTATS_INC(carps_badlen); 574 CARP_DEBUG("%s: packet size %u too small\n", __func__, len); 575 return (IPPROTO_DONE); 576 } 577 } 578 ch = (struct carp_header *)(mtod(m, char *) + *offp); 579 580 581 /* verify the CARP checksum */ 582 m->m_data += *offp; 583 if (in_cksum(m, sizeof(*ch))) { 584 CARPSTATS_INC(carps_badsum); 585 CARP_DEBUG("%s: checksum failed, on %s\n", __func__, 586 m->m_pkthdr.rcvif->if_xname); 587 m_freem(m); 588 return (IPPROTO_DONE); 589 } 590 m->m_data -= *offp; 591 592 carp_input_c(m, ch, AF_INET6); 593 return (IPPROTO_DONE); 594 } 595 #endif /* INET6 */ 596 597 /* 598 * This routine should not be necessary at all, but some switches 599 * (VMWare ESX vswitches) can echo our own packets back at us, 600 * and we must ignore them or they will cause us to drop out of 601 * MASTER mode. 602 * 603 * We cannot catch all cases of network loops. Instead, what we 604 * do here is catch any packet that arrives with a carp header 605 * with a VHID of 0, that comes from an address that is our own. 606 * These packets are by definition "from us" (even if they are from 607 * a misconfigured host that is pretending to be us). 608 * 609 * The VHID test is outside this mini-function. 610 */ 611 static int 612 carp_source_is_self(struct mbuf *m, struct ifaddr *ifa, sa_family_t af) 613 { 614 #ifdef INET 615 struct ip *ip4; 616 struct in_addr in4; 617 #endif 618 #ifdef INET6 619 struct ip6_hdr *ip6; 620 struct in6_addr in6; 621 #endif 622 623 switch (af) { 624 #ifdef INET 625 case AF_INET: 626 ip4 = mtod(m, struct ip *); 627 in4 = ifatoia(ifa)->ia_addr.sin_addr; 628 return (in4.s_addr == ip4->ip_src.s_addr); 629 #endif 630 #ifdef INET6 631 case AF_INET6: 632 ip6 = mtod(m, struct ip6_hdr *); 633 in6 = ifatoia6(ifa)->ia_addr.sin6_addr; 634 return (memcmp(&in6, &ip6->ip6_src, sizeof(in6)) == 0); 635 #endif 636 default: 637 break; 638 } 639 return (0); 640 } 641 642 static void 643 carp_input_c(struct mbuf *m, struct carp_header *ch, sa_family_t af) 644 { 645 struct ifnet *ifp = m->m_pkthdr.rcvif; 646 struct ifaddr *ifa, *match; 647 struct carp_softc *sc; 648 uint64_t tmp_counter; 649 struct timeval sc_tv, ch_tv; 650 int error; 651 652 NET_EPOCH_ASSERT(); 653 654 /* 655 * Verify that the VHID is valid on the receiving interface. 656 * 657 * There should be just one match. If there are none 658 * the VHID is not valid and we drop the packet. If 659 * there are multiple VHID matches, take just the first 660 * one, for compatibility with previous code. While we're 661 * scanning, check for obvious loops in the network topology 662 * (these should never happen, and as noted above, we may 663 * miss real loops; this is just a double-check). 664 */ 665 error = 0; 666 match = NULL; 667 IFNET_FOREACH_IFA(ifp, ifa) { 668 if (match == NULL && ifa->ifa_carp != NULL && 669 ifa->ifa_addr->sa_family == af && 670 ifa->ifa_carp->sc_vhid == ch->carp_vhid) 671 match = ifa; 672 if (ch->carp_vhid == 0 && carp_source_is_self(m, ifa, af)) 673 error = ELOOP; 674 } 675 ifa = error ? NULL : match; 676 if (ifa != NULL) 677 ifa_ref(ifa); 678 679 if (ifa == NULL) { 680 if (error == ELOOP) { 681 CARP_DEBUG("dropping looped packet on interface %s\n", 682 ifp->if_xname); 683 CARPSTATS_INC(carps_badif); /* ??? */ 684 } else { 685 CARPSTATS_INC(carps_badvhid); 686 } 687 m_freem(m); 688 return; 689 } 690 691 /* verify the CARP version. */ 692 if (ch->carp_version != CARP_VERSION) { 693 CARPSTATS_INC(carps_badver); 694 CARP_DEBUG("%s: invalid version %d\n", ifp->if_xname, 695 ch->carp_version); 696 ifa_free(ifa); 697 m_freem(m); 698 return; 699 } 700 701 sc = ifa->ifa_carp; 702 CARP_LOCK(sc); 703 ifa_free(ifa); 704 705 if (carp_hmac_verify(sc, ch->carp_counter, ch->carp_md)) { 706 CARPSTATS_INC(carps_badauth); 707 CARP_DEBUG("%s: incorrect hash for VHID %u@%s\n", __func__, 708 sc->sc_vhid, ifp->if_xname); 709 goto out; 710 } 711 712 tmp_counter = ntohl(ch->carp_counter[0]); 713 tmp_counter = tmp_counter<<32; 714 tmp_counter += ntohl(ch->carp_counter[1]); 715 716 /* XXX Replay protection goes here */ 717 718 sc->sc_init_counter = 0; 719 sc->sc_counter = tmp_counter; 720 721 sc_tv.tv_sec = sc->sc_advbase; 722 sc_tv.tv_usec = DEMOTE_ADVSKEW(sc) * 1000000 / 256; 723 ch_tv.tv_sec = ch->carp_advbase; 724 ch_tv.tv_usec = ch->carp_advskew * 1000000 / 256; 725 726 switch (sc->sc_state) { 727 case INIT: 728 break; 729 case MASTER: 730 /* 731 * If we receive an advertisement from a master who's going to 732 * be more frequent than us, go into BACKUP state. 733 */ 734 if (timevalcmp(&sc_tv, &ch_tv, >) || 735 timevalcmp(&sc_tv, &ch_tv, ==)) { 736 callout_stop(&sc->sc_ad_tmo); 737 carp_set_state(sc, BACKUP, 738 "more frequent advertisement received"); 739 carp_setrun(sc, 0); 740 carp_delroute(sc); 741 } 742 break; 743 case BACKUP: 744 /* 745 * If we're pre-empting masters who advertise slower than us, 746 * and this one claims to be slower, treat him as down. 747 */ 748 if (V_carp_preempt && timevalcmp(&sc_tv, &ch_tv, <)) { 749 carp_master_down_locked(sc, 750 "preempting a slower master"); 751 break; 752 } 753 754 /* 755 * If the master is going to advertise at such a low frequency 756 * that he's guaranteed to time out, we'd might as well just 757 * treat him as timed out now. 758 */ 759 sc_tv.tv_sec = sc->sc_advbase * 3; 760 if (timevalcmp(&sc_tv, &ch_tv, <)) { 761 carp_master_down_locked(sc, "master will time out"); 762 break; 763 } 764 765 /* 766 * Otherwise, we reset the counter and wait for the next 767 * advertisement. 768 */ 769 carp_setrun(sc, af); 770 break; 771 } 772 773 out: 774 CARP_UNLOCK(sc); 775 m_freem(m); 776 } 777 778 static int 779 carp_prepare_ad(struct mbuf *m, struct carp_softc *sc, struct carp_header *ch) 780 { 781 struct m_tag *mtag; 782 783 if (sc->sc_init_counter) { 784 /* this could also be seconds since unix epoch */ 785 sc->sc_counter = arc4random(); 786 sc->sc_counter = sc->sc_counter << 32; 787 sc->sc_counter += arc4random(); 788 } else 789 sc->sc_counter++; 790 791 ch->carp_counter[0] = htonl((sc->sc_counter>>32)&0xffffffff); 792 ch->carp_counter[1] = htonl(sc->sc_counter&0xffffffff); 793 794 carp_hmac_generate(sc, ch->carp_counter, ch->carp_md); 795 796 /* Tag packet for carp_output */ 797 if ((mtag = m_tag_get(PACKET_TAG_CARP, sizeof(struct carp_softc *), 798 M_NOWAIT)) == NULL) { 799 m_freem(m); 800 CARPSTATS_INC(carps_onomem); 801 return (ENOMEM); 802 } 803 bcopy(&sc, mtag + 1, sizeof(sc)); 804 m_tag_prepend(m, mtag); 805 806 return (0); 807 } 808 809 /* 810 * To avoid LORs and possible recursions this function shouldn't 811 * be called directly, but scheduled via taskqueue. 812 */ 813 static void 814 carp_send_ad_all(void *ctx __unused, int pending __unused) 815 { 816 struct carp_softc *sc; 817 818 mtx_lock(&carp_mtx); 819 LIST_FOREACH(sc, &carp_list, sc_next) 820 if (sc->sc_state == MASTER) { 821 CARP_LOCK(sc); 822 CURVNET_SET(sc->sc_carpdev->if_vnet); 823 carp_send_ad_locked(sc); 824 CURVNET_RESTORE(); 825 CARP_UNLOCK(sc); 826 } 827 mtx_unlock(&carp_mtx); 828 } 829 830 /* Send a periodic advertisement, executed in callout context. */ 831 static void 832 carp_send_ad(void *v) 833 { 834 struct carp_softc *sc = v; 835 836 CARP_LOCK_ASSERT(sc); 837 CURVNET_SET(sc->sc_carpdev->if_vnet); 838 carp_send_ad_locked(sc); 839 CURVNET_RESTORE(); 840 CARP_UNLOCK(sc); 841 } 842 843 static void 844 carp_send_ad_error(struct carp_softc *sc, int error) 845 { 846 847 if (error) { 848 if (sc->sc_sendad_errors < INT_MAX) 849 sc->sc_sendad_errors++; 850 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) { 851 static const char fmt[] = "send error %d on %s"; 852 char msg[sizeof(fmt) + IFNAMSIZ]; 853 854 sprintf(msg, fmt, error, sc->sc_carpdev->if_xname); 855 carp_demote_adj(V_carp_senderr_adj, msg); 856 } 857 sc->sc_sendad_success = 0; 858 } else { 859 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS && 860 ++sc->sc_sendad_success >= CARP_SENDAD_MIN_SUCCESS) { 861 static const char fmt[] = "send ok on %s"; 862 char msg[sizeof(fmt) + IFNAMSIZ]; 863 864 sprintf(msg, fmt, sc->sc_carpdev->if_xname); 865 carp_demote_adj(-V_carp_senderr_adj, msg); 866 sc->sc_sendad_errors = 0; 867 } else 868 sc->sc_sendad_errors = 0; 869 } 870 } 871 872 /* 873 * Pick the best ifaddr on the given ifp for sending CARP 874 * advertisements. 875 * 876 * "Best" here is defined by ifa_preferred(). This function is much 877 * much like ifaof_ifpforaddr() except that we just use ifa_preferred(). 878 * 879 * (This could be simplified to return the actual address, except that 880 * it has a different format in AF_INET and AF_INET6.) 881 */ 882 static struct ifaddr * 883 carp_best_ifa(int af, struct ifnet *ifp) 884 { 885 struct ifaddr *ifa, *best; 886 887 NET_EPOCH_ASSERT(); 888 889 if (af >= AF_MAX) 890 return (NULL); 891 best = NULL; 892 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 893 if (ifa->ifa_addr->sa_family == af && 894 (best == NULL || ifa_preferred(best, ifa))) 895 best = ifa; 896 } 897 if (best != NULL) 898 ifa_ref(best); 899 return (best); 900 } 901 902 static void 903 carp_send_ad_locked(struct carp_softc *sc) 904 { 905 struct carp_header ch; 906 struct timeval tv; 907 struct ifaddr *ifa; 908 struct carp_header *ch_ptr; 909 struct mbuf *m; 910 int len, advskew; 911 912 CARP_LOCK_ASSERT(sc); 913 914 advskew = DEMOTE_ADVSKEW(sc); 915 tv.tv_sec = sc->sc_advbase; 916 tv.tv_usec = advskew * 1000000 / 256; 917 918 ch.carp_version = CARP_VERSION; 919 ch.carp_type = CARP_ADVERTISEMENT; 920 ch.carp_vhid = sc->sc_vhid; 921 ch.carp_advbase = sc->sc_advbase; 922 ch.carp_advskew = advskew; 923 ch.carp_authlen = 7; /* XXX DEFINE */ 924 ch.carp_pad1 = 0; /* must be zero */ 925 ch.carp_cksum = 0; 926 927 /* XXXGL: OpenBSD picks first ifaddr with needed family. */ 928 929 #ifdef INET 930 if (sc->sc_naddrs) { 931 struct ip *ip; 932 933 m = m_gethdr(M_NOWAIT, MT_DATA); 934 if (m == NULL) { 935 CARPSTATS_INC(carps_onomem); 936 goto resched; 937 } 938 len = sizeof(*ip) + sizeof(ch); 939 m->m_pkthdr.len = len; 940 m->m_pkthdr.rcvif = NULL; 941 m->m_len = len; 942 M_ALIGN(m, m->m_len); 943 m->m_flags |= M_MCAST; 944 ip = mtod(m, struct ip *); 945 ip->ip_v = IPVERSION; 946 ip->ip_hl = sizeof(*ip) >> 2; 947 ip->ip_tos = V_carp_dscp << IPTOS_DSCP_OFFSET; 948 ip->ip_len = htons(len); 949 ip->ip_off = htons(IP_DF); 950 ip->ip_ttl = CARP_DFLTTL; 951 ip->ip_p = IPPROTO_CARP; 952 ip->ip_sum = 0; 953 ip_fillid(ip); 954 955 ifa = carp_best_ifa(AF_INET, sc->sc_carpdev); 956 if (ifa != NULL) { 957 ip->ip_src.s_addr = 958 ifatoia(ifa)->ia_addr.sin_addr.s_addr; 959 ifa_free(ifa); 960 } else 961 ip->ip_src.s_addr = 0; 962 ip->ip_dst.s_addr = htonl(INADDR_CARP_GROUP); 963 964 ch_ptr = (struct carp_header *)(&ip[1]); 965 bcopy(&ch, ch_ptr, sizeof(ch)); 966 if (carp_prepare_ad(m, sc, ch_ptr)) 967 goto resched; 968 969 m->m_data += sizeof(*ip); 970 ch_ptr->carp_cksum = in_cksum(m, len - sizeof(*ip)); 971 m->m_data -= sizeof(*ip); 972 973 CARPSTATS_INC(carps_opackets); 974 975 carp_send_ad_error(sc, ip_output(m, NULL, NULL, IP_RAWOUTPUT, 976 &sc->sc_carpdev->if_carp->cif_imo, NULL)); 977 } 978 #endif /* INET */ 979 #ifdef INET6 980 if (sc->sc_naddrs6) { 981 struct ip6_hdr *ip6; 982 983 m = m_gethdr(M_NOWAIT, MT_DATA); 984 if (m == NULL) { 985 CARPSTATS_INC(carps_onomem); 986 goto resched; 987 } 988 len = sizeof(*ip6) + sizeof(ch); 989 m->m_pkthdr.len = len; 990 m->m_pkthdr.rcvif = NULL; 991 m->m_len = len; 992 M_ALIGN(m, m->m_len); 993 m->m_flags |= M_MCAST; 994 ip6 = mtod(m, struct ip6_hdr *); 995 bzero(ip6, sizeof(*ip6)); 996 ip6->ip6_vfc |= IPV6_VERSION; 997 /* Traffic class isn't defined in ip6 struct instead 998 * it gets offset into flowid field */ 999 ip6->ip6_flow |= htonl(V_carp_dscp << (IPV6_FLOWLABEL_LEN + 1000 IPTOS_DSCP_OFFSET)); 1001 ip6->ip6_hlim = CARP_DFLTTL; 1002 ip6->ip6_nxt = IPPROTO_CARP; 1003 1004 /* set the source address */ 1005 ifa = carp_best_ifa(AF_INET6, sc->sc_carpdev); 1006 if (ifa != NULL) { 1007 bcopy(IFA_IN6(ifa), &ip6->ip6_src, 1008 sizeof(struct in6_addr)); 1009 ifa_free(ifa); 1010 } else 1011 /* This should never happen with IPv6. */ 1012 bzero(&ip6->ip6_src, sizeof(struct in6_addr)); 1013 1014 /* Set the multicast destination. */ 1015 ip6->ip6_dst.s6_addr16[0] = htons(0xff02); 1016 ip6->ip6_dst.s6_addr8[15] = 0x12; 1017 if (in6_setscope(&ip6->ip6_dst, sc->sc_carpdev, NULL) != 0) { 1018 m_freem(m); 1019 CARP_DEBUG("%s: in6_setscope failed\n", __func__); 1020 goto resched; 1021 } 1022 1023 ch_ptr = (struct carp_header *)(&ip6[1]); 1024 bcopy(&ch, ch_ptr, sizeof(ch)); 1025 if (carp_prepare_ad(m, sc, ch_ptr)) 1026 goto resched; 1027 1028 m->m_data += sizeof(*ip6); 1029 ch_ptr->carp_cksum = in_cksum(m, len - sizeof(*ip6)); 1030 m->m_data -= sizeof(*ip6); 1031 1032 CARPSTATS_INC(carps_opackets6); 1033 1034 carp_send_ad_error(sc, ip6_output(m, NULL, NULL, 0, 1035 &sc->sc_carpdev->if_carp->cif_im6o, NULL, NULL)); 1036 } 1037 #endif /* INET6 */ 1038 1039 resched: 1040 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv), carp_send_ad, sc); 1041 } 1042 1043 static void 1044 carp_addroute(struct carp_softc *sc) 1045 { 1046 struct ifaddr *ifa; 1047 1048 CARP_FOREACH_IFA(sc, ifa) 1049 carp_ifa_addroute(ifa); 1050 } 1051 1052 static void 1053 carp_ifa_addroute(struct ifaddr *ifa) 1054 { 1055 1056 switch (ifa->ifa_addr->sa_family) { 1057 #ifdef INET 1058 case AF_INET: 1059 in_addprefix(ifatoia(ifa), RTF_UP); 1060 ifa_add_loopback_route(ifa, 1061 (struct sockaddr *)&ifatoia(ifa)->ia_addr); 1062 break; 1063 #endif 1064 #ifdef INET6 1065 case AF_INET6: 1066 ifa_add_loopback_route(ifa, 1067 (struct sockaddr *)&ifatoia6(ifa)->ia_addr); 1068 nd6_add_ifa_lle(ifatoia6(ifa)); 1069 break; 1070 #endif 1071 } 1072 } 1073 1074 static void 1075 carp_delroute(struct carp_softc *sc) 1076 { 1077 struct ifaddr *ifa; 1078 1079 CARP_FOREACH_IFA(sc, ifa) 1080 carp_ifa_delroute(ifa); 1081 } 1082 1083 static void 1084 carp_ifa_delroute(struct ifaddr *ifa) 1085 { 1086 1087 switch (ifa->ifa_addr->sa_family) { 1088 #ifdef INET 1089 case AF_INET: 1090 ifa_del_loopback_route(ifa, 1091 (struct sockaddr *)&ifatoia(ifa)->ia_addr); 1092 in_scrubprefix(ifatoia(ifa), LLE_STATIC); 1093 break; 1094 #endif 1095 #ifdef INET6 1096 case AF_INET6: 1097 ifa_del_loopback_route(ifa, 1098 (struct sockaddr *)&ifatoia6(ifa)->ia_addr); 1099 nd6_rem_ifa_lle(ifatoia6(ifa), 1); 1100 break; 1101 #endif 1102 } 1103 } 1104 1105 int 1106 carp_master(struct ifaddr *ifa) 1107 { 1108 struct carp_softc *sc = ifa->ifa_carp; 1109 1110 return (sc->sc_state == MASTER); 1111 } 1112 1113 #ifdef INET 1114 /* 1115 * Broadcast a gratuitous ARP request containing 1116 * the virtual router MAC address for each IP address 1117 * associated with the virtual router. 1118 */ 1119 static void 1120 carp_send_arp(struct carp_softc *sc) 1121 { 1122 struct ifaddr *ifa; 1123 struct in_addr addr; 1124 1125 CARP_FOREACH_IFA(sc, ifa) { 1126 if (ifa->ifa_addr->sa_family != AF_INET) 1127 continue; 1128 addr = ((struct sockaddr_in *)ifa->ifa_addr)->sin_addr; 1129 arp_announce_ifaddr(sc->sc_carpdev, addr, LLADDR(&sc->sc_addr)); 1130 } 1131 } 1132 1133 int 1134 carp_iamatch(struct ifaddr *ifa, uint8_t **enaddr) 1135 { 1136 struct carp_softc *sc = ifa->ifa_carp; 1137 1138 if (sc->sc_state == MASTER) { 1139 *enaddr = LLADDR(&sc->sc_addr); 1140 return (1); 1141 } 1142 1143 return (0); 1144 } 1145 #endif 1146 1147 #ifdef INET6 1148 static void 1149 carp_send_na(struct carp_softc *sc) 1150 { 1151 static struct in6_addr mcast = IN6ADDR_LINKLOCAL_ALLNODES_INIT; 1152 struct ifaddr *ifa; 1153 struct in6_addr *in6; 1154 1155 CARP_FOREACH_IFA(sc, ifa) { 1156 if (ifa->ifa_addr->sa_family != AF_INET6) 1157 continue; 1158 1159 in6 = IFA_IN6(ifa); 1160 nd6_na_output(sc->sc_carpdev, &mcast, in6, 1161 ND_NA_FLAG_OVERRIDE, 1, NULL); 1162 DELAY(1000); /* XXX */ 1163 } 1164 } 1165 1166 /* 1167 * Returns ifa in case it's a carp address and it is MASTER, or if the address 1168 * matches and is not a carp address. Returns NULL otherwise. 1169 */ 1170 struct ifaddr * 1171 carp_iamatch6(struct ifnet *ifp, struct in6_addr *taddr) 1172 { 1173 struct ifaddr *ifa; 1174 1175 NET_EPOCH_ASSERT(); 1176 1177 ifa = NULL; 1178 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1179 if (ifa->ifa_addr->sa_family != AF_INET6) 1180 continue; 1181 if (!IN6_ARE_ADDR_EQUAL(taddr, IFA_IN6(ifa))) 1182 continue; 1183 if (ifa->ifa_carp && ifa->ifa_carp->sc_state != MASTER) 1184 ifa = NULL; 1185 else 1186 ifa_ref(ifa); 1187 break; 1188 } 1189 1190 return (ifa); 1191 } 1192 1193 char * 1194 carp_macmatch6(struct ifnet *ifp, struct mbuf *m, const struct in6_addr *taddr) 1195 { 1196 struct ifaddr *ifa; 1197 1198 NET_EPOCH_ASSERT(); 1199 1200 IFNET_FOREACH_IFA(ifp, ifa) 1201 if (ifa->ifa_addr->sa_family == AF_INET6 && 1202 IN6_ARE_ADDR_EQUAL(taddr, IFA_IN6(ifa))) { 1203 struct carp_softc *sc = ifa->ifa_carp; 1204 struct m_tag *mtag; 1205 1206 mtag = m_tag_get(PACKET_TAG_CARP, 1207 sizeof(struct carp_softc *), M_NOWAIT); 1208 if (mtag == NULL) 1209 /* Better a bit than nothing. */ 1210 return (LLADDR(&sc->sc_addr)); 1211 1212 bcopy(&sc, mtag + 1, sizeof(sc)); 1213 m_tag_prepend(m, mtag); 1214 1215 return (LLADDR(&sc->sc_addr)); 1216 } 1217 1218 return (NULL); 1219 } 1220 #endif /* INET6 */ 1221 1222 int 1223 carp_forus(struct ifnet *ifp, u_char *dhost) 1224 { 1225 struct carp_softc *sc; 1226 uint8_t *ena = dhost; 1227 1228 if (ena[0] || ena[1] || ena[2] != 0x5e || ena[3] || ena[4] != 1) 1229 return (0); 1230 1231 CIF_LOCK(ifp->if_carp); 1232 IFNET_FOREACH_CARP(ifp, sc) { 1233 /* 1234 * CARP_LOCK() is not here, since would protect nothing, but 1235 * cause deadlock with if_bridge, calling this under its lock. 1236 */ 1237 if (sc->sc_state == MASTER && !bcmp(dhost, LLADDR(&sc->sc_addr), 1238 ETHER_ADDR_LEN)) { 1239 CIF_UNLOCK(ifp->if_carp); 1240 return (1); 1241 } 1242 } 1243 CIF_UNLOCK(ifp->if_carp); 1244 1245 return (0); 1246 } 1247 1248 /* Master down timeout event, executed in callout context. */ 1249 static void 1250 carp_master_down(void *v) 1251 { 1252 struct carp_softc *sc = v; 1253 1254 CARP_LOCK_ASSERT(sc); 1255 1256 CURVNET_SET(sc->sc_carpdev->if_vnet); 1257 if (sc->sc_state == BACKUP) { 1258 carp_master_down_locked(sc, "master timed out"); 1259 } 1260 CURVNET_RESTORE(); 1261 1262 CARP_UNLOCK(sc); 1263 } 1264 1265 static void 1266 carp_master_down_locked(struct carp_softc *sc, const char *reason) 1267 { 1268 1269 CARP_LOCK_ASSERT(sc); 1270 1271 switch (sc->sc_state) { 1272 case BACKUP: 1273 carp_set_state(sc, MASTER, reason); 1274 carp_send_ad_locked(sc); 1275 #ifdef INET 1276 carp_send_arp(sc); 1277 #endif 1278 #ifdef INET6 1279 carp_send_na(sc); 1280 #endif 1281 carp_setrun(sc, 0); 1282 carp_addroute(sc); 1283 break; 1284 case INIT: 1285 case MASTER: 1286 #ifdef INVARIANTS 1287 panic("carp: VHID %u@%s: master_down event in %s state\n", 1288 sc->sc_vhid, 1289 sc->sc_carpdev->if_xname, 1290 sc->sc_state ? "MASTER" : "INIT"); 1291 #endif 1292 break; 1293 } 1294 } 1295 1296 /* 1297 * When in backup state, af indicates whether to reset the master down timer 1298 * for v4 or v6. If it's set to zero, reset the ones which are already pending. 1299 */ 1300 static void 1301 carp_setrun(struct carp_softc *sc, sa_family_t af) 1302 { 1303 struct timeval tv; 1304 1305 CARP_LOCK_ASSERT(sc); 1306 1307 if ((sc->sc_carpdev->if_flags & IFF_UP) == 0 || 1308 sc->sc_carpdev->if_link_state != LINK_STATE_UP || 1309 (sc->sc_naddrs == 0 && sc->sc_naddrs6 == 0) || 1310 !V_carp_allow) 1311 return; 1312 1313 switch (sc->sc_state) { 1314 case INIT: 1315 carp_set_state(sc, BACKUP, "initialization complete"); 1316 carp_setrun(sc, 0); 1317 break; 1318 case BACKUP: 1319 callout_stop(&sc->sc_ad_tmo); 1320 tv.tv_sec = 3 * sc->sc_advbase; 1321 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1322 switch (af) { 1323 #ifdef INET 1324 case AF_INET: 1325 callout_reset(&sc->sc_md_tmo, tvtohz(&tv), 1326 carp_master_down, sc); 1327 break; 1328 #endif 1329 #ifdef INET6 1330 case AF_INET6: 1331 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv), 1332 carp_master_down, sc); 1333 break; 1334 #endif 1335 default: 1336 #ifdef INET 1337 if (sc->sc_naddrs) 1338 callout_reset(&sc->sc_md_tmo, tvtohz(&tv), 1339 carp_master_down, sc); 1340 #endif 1341 #ifdef INET6 1342 if (sc->sc_naddrs6) 1343 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv), 1344 carp_master_down, sc); 1345 #endif 1346 break; 1347 } 1348 break; 1349 case MASTER: 1350 tv.tv_sec = sc->sc_advbase; 1351 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1352 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv), 1353 carp_send_ad, sc); 1354 break; 1355 } 1356 } 1357 1358 /* 1359 * Setup multicast structures. 1360 */ 1361 static int 1362 carp_multicast_setup(struct carp_if *cif, sa_family_t sa) 1363 { 1364 struct ifnet *ifp = cif->cif_ifp; 1365 int error = 0; 1366 1367 switch (sa) { 1368 #ifdef INET 1369 case AF_INET: 1370 { 1371 struct ip_moptions *imo = &cif->cif_imo; 1372 struct in_mfilter *imf; 1373 struct in_addr addr; 1374 1375 if (ip_mfilter_first(&imo->imo_head) != NULL) 1376 return (0); 1377 1378 imf = ip_mfilter_alloc(M_WAITOK, 0, 0); 1379 ip_mfilter_init(&imo->imo_head); 1380 imo->imo_multicast_vif = -1; 1381 1382 addr.s_addr = htonl(INADDR_CARP_GROUP); 1383 if ((error = in_joingroup(ifp, &addr, NULL, 1384 &imf->imf_inm)) != 0) { 1385 ip_mfilter_free(imf); 1386 break; 1387 } 1388 1389 ip_mfilter_insert(&imo->imo_head, imf); 1390 imo->imo_multicast_ifp = ifp; 1391 imo->imo_multicast_ttl = CARP_DFLTTL; 1392 imo->imo_multicast_loop = 0; 1393 break; 1394 } 1395 #endif 1396 #ifdef INET6 1397 case AF_INET6: 1398 { 1399 struct ip6_moptions *im6o = &cif->cif_im6o; 1400 struct in6_mfilter *im6f[2]; 1401 struct in6_addr in6; 1402 1403 if (ip6_mfilter_first(&im6o->im6o_head)) 1404 return (0); 1405 1406 im6f[0] = ip6_mfilter_alloc(M_WAITOK, 0, 0); 1407 im6f[1] = ip6_mfilter_alloc(M_WAITOK, 0, 0); 1408 1409 ip6_mfilter_init(&im6o->im6o_head); 1410 im6o->im6o_multicast_hlim = CARP_DFLTTL; 1411 im6o->im6o_multicast_ifp = ifp; 1412 1413 /* Join IPv6 CARP multicast group. */ 1414 bzero(&in6, sizeof(in6)); 1415 in6.s6_addr16[0] = htons(0xff02); 1416 in6.s6_addr8[15] = 0x12; 1417 if ((error = in6_setscope(&in6, ifp, NULL)) != 0) { 1418 ip6_mfilter_free(im6f[0]); 1419 ip6_mfilter_free(im6f[1]); 1420 break; 1421 } 1422 if ((error = in6_joingroup(ifp, &in6, NULL, &im6f[0]->im6f_in6m, 0)) != 0) { 1423 ip6_mfilter_free(im6f[0]); 1424 ip6_mfilter_free(im6f[1]); 1425 break; 1426 } 1427 1428 /* Join solicited multicast address. */ 1429 bzero(&in6, sizeof(in6)); 1430 in6.s6_addr16[0] = htons(0xff02); 1431 in6.s6_addr32[1] = 0; 1432 in6.s6_addr32[2] = htonl(1); 1433 in6.s6_addr32[3] = 0; 1434 in6.s6_addr8[12] = 0xff; 1435 1436 if ((error = in6_setscope(&in6, ifp, NULL)) != 0) { 1437 ip6_mfilter_free(im6f[0]); 1438 ip6_mfilter_free(im6f[1]); 1439 break; 1440 } 1441 1442 if ((error = in6_joingroup(ifp, &in6, NULL, &im6f[1]->im6f_in6m, 0)) != 0) { 1443 in6_leavegroup(im6f[0]->im6f_in6m, NULL); 1444 ip6_mfilter_free(im6f[0]); 1445 ip6_mfilter_free(im6f[1]); 1446 break; 1447 } 1448 ip6_mfilter_insert(&im6o->im6o_head, im6f[0]); 1449 ip6_mfilter_insert(&im6o->im6o_head, im6f[1]); 1450 break; 1451 } 1452 #endif 1453 } 1454 1455 return (error); 1456 } 1457 1458 /* 1459 * Free multicast structures. 1460 */ 1461 static void 1462 carp_multicast_cleanup(struct carp_if *cif, sa_family_t sa) 1463 { 1464 #ifdef INET 1465 struct ip_moptions *imo = &cif->cif_imo; 1466 struct in_mfilter *imf; 1467 #endif 1468 #ifdef INET6 1469 struct ip6_moptions *im6o = &cif->cif_im6o; 1470 struct in6_mfilter *im6f; 1471 #endif 1472 sx_assert(&carp_sx, SA_XLOCKED); 1473 1474 switch (sa) { 1475 #ifdef INET 1476 case AF_INET: 1477 if (cif->cif_naddrs != 0) 1478 break; 1479 1480 while ((imf = ip_mfilter_first(&imo->imo_head)) != NULL) { 1481 ip_mfilter_remove(&imo->imo_head, imf); 1482 in_leavegroup(imf->imf_inm, NULL); 1483 ip_mfilter_free(imf); 1484 } 1485 break; 1486 #endif 1487 #ifdef INET6 1488 case AF_INET6: 1489 if (cif->cif_naddrs6 != 0) 1490 break; 1491 1492 while ((im6f = ip6_mfilter_first(&im6o->im6o_head)) != NULL) { 1493 ip6_mfilter_remove(&im6o->im6o_head, im6f); 1494 in6_leavegroup(im6f->im6f_in6m, NULL); 1495 ip6_mfilter_free(im6f); 1496 } 1497 break; 1498 #endif 1499 } 1500 } 1501 1502 int 1503 carp_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa) 1504 { 1505 struct m_tag *mtag; 1506 struct carp_softc *sc; 1507 1508 if (!sa) 1509 return (0); 1510 1511 switch (sa->sa_family) { 1512 #ifdef INET 1513 case AF_INET: 1514 break; 1515 #endif 1516 #ifdef INET6 1517 case AF_INET6: 1518 break; 1519 #endif 1520 default: 1521 return (0); 1522 } 1523 1524 mtag = m_tag_find(m, PACKET_TAG_CARP, NULL); 1525 if (mtag == NULL) 1526 return (0); 1527 1528 bcopy(mtag + 1, &sc, sizeof(sc)); 1529 1530 /* Set the source MAC address to the Virtual Router MAC Address. */ 1531 switch (ifp->if_type) { 1532 case IFT_ETHER: 1533 case IFT_BRIDGE: 1534 case IFT_L2VLAN: { 1535 struct ether_header *eh; 1536 1537 eh = mtod(m, struct ether_header *); 1538 eh->ether_shost[0] = 0; 1539 eh->ether_shost[1] = 0; 1540 eh->ether_shost[2] = 0x5e; 1541 eh->ether_shost[3] = 0; 1542 eh->ether_shost[4] = 1; 1543 eh->ether_shost[5] = sc->sc_vhid; 1544 } 1545 break; 1546 default: 1547 printf("%s: carp is not supported for the %d interface type\n", 1548 ifp->if_xname, ifp->if_type); 1549 return (EOPNOTSUPP); 1550 } 1551 1552 return (0); 1553 } 1554 1555 static struct carp_softc* 1556 carp_alloc(struct ifnet *ifp) 1557 { 1558 struct carp_softc *sc; 1559 struct carp_if *cif; 1560 1561 sx_assert(&carp_sx, SA_XLOCKED); 1562 1563 if ((cif = ifp->if_carp) == NULL) 1564 cif = carp_alloc_if(ifp); 1565 1566 sc = malloc(sizeof(*sc), M_CARP, M_WAITOK|M_ZERO); 1567 1568 sc->sc_advbase = CARP_DFLTINTV; 1569 sc->sc_vhid = -1; /* required setting */ 1570 sc->sc_init_counter = 1; 1571 sc->sc_state = INIT; 1572 1573 sc->sc_ifasiz = sizeof(struct ifaddr *); 1574 sc->sc_ifas = malloc(sc->sc_ifasiz, M_CARP, M_WAITOK|M_ZERO); 1575 sc->sc_carpdev = ifp; 1576 1577 CARP_LOCK_INIT(sc); 1578 #ifdef INET 1579 callout_init_mtx(&sc->sc_md_tmo, &sc->sc_mtx, CALLOUT_RETURNUNLOCKED); 1580 #endif 1581 #ifdef INET6 1582 callout_init_mtx(&sc->sc_md6_tmo, &sc->sc_mtx, CALLOUT_RETURNUNLOCKED); 1583 #endif 1584 callout_init_mtx(&sc->sc_ad_tmo, &sc->sc_mtx, CALLOUT_RETURNUNLOCKED); 1585 1586 CIF_LOCK(cif); 1587 TAILQ_INSERT_TAIL(&cif->cif_vrs, sc, sc_list); 1588 CIF_UNLOCK(cif); 1589 1590 mtx_lock(&carp_mtx); 1591 LIST_INSERT_HEAD(&carp_list, sc, sc_next); 1592 mtx_unlock(&carp_mtx); 1593 1594 return (sc); 1595 } 1596 1597 static void 1598 carp_grow_ifas(struct carp_softc *sc) 1599 { 1600 struct ifaddr **new; 1601 1602 new = malloc(sc->sc_ifasiz * 2, M_CARP, M_WAITOK | M_ZERO); 1603 CARP_LOCK(sc); 1604 bcopy(sc->sc_ifas, new, sc->sc_ifasiz); 1605 free(sc->sc_ifas, M_CARP); 1606 sc->sc_ifas = new; 1607 sc->sc_ifasiz *= 2; 1608 CARP_UNLOCK(sc); 1609 } 1610 1611 static void 1612 carp_destroy(struct carp_softc *sc) 1613 { 1614 struct ifnet *ifp = sc->sc_carpdev; 1615 struct carp_if *cif = ifp->if_carp; 1616 1617 sx_assert(&carp_sx, SA_XLOCKED); 1618 1619 if (sc->sc_suppress) 1620 carp_demote_adj(-V_carp_ifdown_adj, "vhid removed"); 1621 CARP_UNLOCK(sc); 1622 1623 CIF_LOCK(cif); 1624 TAILQ_REMOVE(&cif->cif_vrs, sc, sc_list); 1625 CIF_UNLOCK(cif); 1626 1627 mtx_lock(&carp_mtx); 1628 LIST_REMOVE(sc, sc_next); 1629 mtx_unlock(&carp_mtx); 1630 1631 callout_drain(&sc->sc_ad_tmo); 1632 #ifdef INET 1633 callout_drain(&sc->sc_md_tmo); 1634 #endif 1635 #ifdef INET6 1636 callout_drain(&sc->sc_md6_tmo); 1637 #endif 1638 CARP_LOCK_DESTROY(sc); 1639 1640 free(sc->sc_ifas, M_CARP); 1641 free(sc, M_CARP); 1642 } 1643 1644 static struct carp_if* 1645 carp_alloc_if(struct ifnet *ifp) 1646 { 1647 struct carp_if *cif; 1648 int error; 1649 1650 cif = malloc(sizeof(*cif), M_CARP, M_WAITOK|M_ZERO); 1651 1652 if ((error = ifpromisc(ifp, 1)) != 0) 1653 printf("%s: ifpromisc(%s) failed: %d\n", 1654 __func__, ifp->if_xname, error); 1655 else 1656 cif->cif_flags |= CIF_PROMISC; 1657 1658 CIF_LOCK_INIT(cif); 1659 cif->cif_ifp = ifp; 1660 TAILQ_INIT(&cif->cif_vrs); 1661 1662 IF_ADDR_WLOCK(ifp); 1663 ifp->if_carp = cif; 1664 if_ref(ifp); 1665 IF_ADDR_WUNLOCK(ifp); 1666 1667 return (cif); 1668 } 1669 1670 static void 1671 carp_free_if(struct carp_if *cif) 1672 { 1673 struct ifnet *ifp = cif->cif_ifp; 1674 1675 CIF_LOCK_ASSERT(cif); 1676 KASSERT(TAILQ_EMPTY(&cif->cif_vrs), ("%s: softc list not empty", 1677 __func__)); 1678 1679 IF_ADDR_WLOCK(ifp); 1680 ifp->if_carp = NULL; 1681 IF_ADDR_WUNLOCK(ifp); 1682 1683 CIF_LOCK_DESTROY(cif); 1684 1685 if (cif->cif_flags & CIF_PROMISC) 1686 ifpromisc(ifp, 0); 1687 if_rele(ifp); 1688 1689 free(cif, M_CARP); 1690 } 1691 1692 static void 1693 carp_carprcp(struct carpreq *carpr, struct carp_softc *sc, int priv) 1694 { 1695 1696 CARP_LOCK(sc); 1697 carpr->carpr_state = sc->sc_state; 1698 carpr->carpr_vhid = sc->sc_vhid; 1699 carpr->carpr_advbase = sc->sc_advbase; 1700 carpr->carpr_advskew = sc->sc_advskew; 1701 if (priv) 1702 bcopy(sc->sc_key, carpr->carpr_key, sizeof(carpr->carpr_key)); 1703 else 1704 bzero(carpr->carpr_key, sizeof(carpr->carpr_key)); 1705 CARP_UNLOCK(sc); 1706 } 1707 1708 int 1709 carp_ioctl(struct ifreq *ifr, u_long cmd, struct thread *td) 1710 { 1711 struct carpreq carpr; 1712 struct ifnet *ifp; 1713 struct carp_softc *sc = NULL; 1714 int error = 0, locked = 0; 1715 1716 if ((error = copyin(ifr_data_get_ptr(ifr), &carpr, sizeof carpr))) 1717 return (error); 1718 1719 ifp = ifunit_ref(ifr->ifr_name); 1720 if (ifp == NULL) 1721 return (ENXIO); 1722 1723 switch (ifp->if_type) { 1724 case IFT_ETHER: 1725 case IFT_L2VLAN: 1726 case IFT_BRIDGE: 1727 break; 1728 default: 1729 error = EOPNOTSUPP; 1730 goto out; 1731 } 1732 1733 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 1734 error = EADDRNOTAVAIL; 1735 goto out; 1736 } 1737 1738 sx_xlock(&carp_sx); 1739 switch (cmd) { 1740 case SIOCSVH: 1741 if ((error = priv_check(td, PRIV_NETINET_CARP))) 1742 break; 1743 if (carpr.carpr_vhid <= 0 || carpr.carpr_vhid > CARP_MAXVHID || 1744 carpr.carpr_advbase < 0 || carpr.carpr_advskew < 0) { 1745 error = EINVAL; 1746 break; 1747 } 1748 1749 if (ifp->if_carp) { 1750 IFNET_FOREACH_CARP(ifp, sc) 1751 if (sc->sc_vhid == carpr.carpr_vhid) 1752 break; 1753 } 1754 if (sc == NULL) { 1755 sc = carp_alloc(ifp); 1756 CARP_LOCK(sc); 1757 sc->sc_vhid = carpr.carpr_vhid; 1758 LLADDR(&sc->sc_addr)[0] = 0; 1759 LLADDR(&sc->sc_addr)[1] = 0; 1760 LLADDR(&sc->sc_addr)[2] = 0x5e; 1761 LLADDR(&sc->sc_addr)[3] = 0; 1762 LLADDR(&sc->sc_addr)[4] = 1; 1763 LLADDR(&sc->sc_addr)[5] = sc->sc_vhid; 1764 } else 1765 CARP_LOCK(sc); 1766 locked = 1; 1767 if (carpr.carpr_advbase > 0) { 1768 if (carpr.carpr_advbase > 255 || 1769 carpr.carpr_advbase < CARP_DFLTINTV) { 1770 error = EINVAL; 1771 break; 1772 } 1773 sc->sc_advbase = carpr.carpr_advbase; 1774 } 1775 if (carpr.carpr_advskew >= 255) { 1776 error = EINVAL; 1777 break; 1778 } 1779 sc->sc_advskew = carpr.carpr_advskew; 1780 if (carpr.carpr_key[0] != '\0') { 1781 bcopy(carpr.carpr_key, sc->sc_key, sizeof(sc->sc_key)); 1782 carp_hmac_prepare(sc); 1783 } 1784 if (sc->sc_state != INIT && 1785 carpr.carpr_state != sc->sc_state) { 1786 switch (carpr.carpr_state) { 1787 case BACKUP: 1788 callout_stop(&sc->sc_ad_tmo); 1789 carp_set_state(sc, BACKUP, 1790 "user requested via ifconfig"); 1791 carp_setrun(sc, 0); 1792 carp_delroute(sc); 1793 break; 1794 case MASTER: 1795 carp_master_down_locked(sc, 1796 "user requested via ifconfig"); 1797 break; 1798 default: 1799 break; 1800 } 1801 } 1802 break; 1803 1804 case SIOCGVH: 1805 { 1806 int priveleged; 1807 1808 if (carpr.carpr_vhid < 0 || carpr.carpr_vhid > CARP_MAXVHID) { 1809 error = EINVAL; 1810 break; 1811 } 1812 if (carpr.carpr_count < 1) { 1813 error = EMSGSIZE; 1814 break; 1815 } 1816 if (ifp->if_carp == NULL) { 1817 error = ENOENT; 1818 break; 1819 } 1820 1821 priveleged = (priv_check(td, PRIV_NETINET_CARP) == 0); 1822 if (carpr.carpr_vhid != 0) { 1823 IFNET_FOREACH_CARP(ifp, sc) 1824 if (sc->sc_vhid == carpr.carpr_vhid) 1825 break; 1826 if (sc == NULL) { 1827 error = ENOENT; 1828 break; 1829 } 1830 carp_carprcp(&carpr, sc, priveleged); 1831 error = copyout(&carpr, ifr_data_get_ptr(ifr), 1832 sizeof(carpr)); 1833 } else { 1834 int i, count; 1835 1836 count = 0; 1837 IFNET_FOREACH_CARP(ifp, sc) 1838 count++; 1839 1840 if (count > carpr.carpr_count) { 1841 CIF_UNLOCK(ifp->if_carp); 1842 error = EMSGSIZE; 1843 break; 1844 } 1845 1846 i = 0; 1847 IFNET_FOREACH_CARP(ifp, sc) { 1848 carp_carprcp(&carpr, sc, priveleged); 1849 carpr.carpr_count = count; 1850 error = copyout(&carpr, 1851 (char *)ifr_data_get_ptr(ifr) + 1852 (i * sizeof(carpr)), sizeof(carpr)); 1853 if (error) { 1854 CIF_UNLOCK(ifp->if_carp); 1855 break; 1856 } 1857 i++; 1858 } 1859 } 1860 break; 1861 } 1862 default: 1863 error = EINVAL; 1864 } 1865 sx_xunlock(&carp_sx); 1866 1867 out: 1868 if (locked) 1869 CARP_UNLOCK(sc); 1870 if_rele(ifp); 1871 1872 return (error); 1873 } 1874 1875 static int 1876 carp_get_vhid(struct ifaddr *ifa) 1877 { 1878 1879 if (ifa == NULL || ifa->ifa_carp == NULL) 1880 return (0); 1881 1882 return (ifa->ifa_carp->sc_vhid); 1883 } 1884 1885 int 1886 carp_attach(struct ifaddr *ifa, int vhid) 1887 { 1888 struct ifnet *ifp = ifa->ifa_ifp; 1889 struct carp_if *cif = ifp->if_carp; 1890 struct carp_softc *sc; 1891 int index, error; 1892 1893 KASSERT(ifa->ifa_carp == NULL, ("%s: ifa %p attached", __func__, ifa)); 1894 1895 switch (ifa->ifa_addr->sa_family) { 1896 #ifdef INET 1897 case AF_INET: 1898 #endif 1899 #ifdef INET6 1900 case AF_INET6: 1901 #endif 1902 break; 1903 default: 1904 return (EPROTOTYPE); 1905 } 1906 1907 sx_xlock(&carp_sx); 1908 if (ifp->if_carp == NULL) { 1909 sx_xunlock(&carp_sx); 1910 return (ENOPROTOOPT); 1911 } 1912 1913 IFNET_FOREACH_CARP(ifp, sc) 1914 if (sc->sc_vhid == vhid) 1915 break; 1916 if (sc == NULL) { 1917 sx_xunlock(&carp_sx); 1918 return (ENOENT); 1919 } 1920 1921 error = carp_multicast_setup(cif, ifa->ifa_addr->sa_family); 1922 if (error) { 1923 CIF_FREE(cif); 1924 sx_xunlock(&carp_sx); 1925 return (error); 1926 } 1927 1928 index = sc->sc_naddrs + sc->sc_naddrs6 + 1; 1929 if (index > sc->sc_ifasiz / sizeof(struct ifaddr *)) 1930 carp_grow_ifas(sc); 1931 1932 switch (ifa->ifa_addr->sa_family) { 1933 #ifdef INET 1934 case AF_INET: 1935 cif->cif_naddrs++; 1936 sc->sc_naddrs++; 1937 break; 1938 #endif 1939 #ifdef INET6 1940 case AF_INET6: 1941 cif->cif_naddrs6++; 1942 sc->sc_naddrs6++; 1943 break; 1944 #endif 1945 } 1946 1947 ifa_ref(ifa); 1948 1949 CARP_LOCK(sc); 1950 sc->sc_ifas[index - 1] = ifa; 1951 ifa->ifa_carp = sc; 1952 carp_hmac_prepare(sc); 1953 carp_sc_state(sc); 1954 CARP_UNLOCK(sc); 1955 1956 sx_xunlock(&carp_sx); 1957 1958 return (0); 1959 } 1960 1961 void 1962 carp_detach(struct ifaddr *ifa, bool keep_cif) 1963 { 1964 struct ifnet *ifp = ifa->ifa_ifp; 1965 struct carp_if *cif = ifp->if_carp; 1966 struct carp_softc *sc = ifa->ifa_carp; 1967 int i, index; 1968 1969 KASSERT(sc != NULL, ("%s: %p not attached", __func__, ifa)); 1970 1971 sx_xlock(&carp_sx); 1972 1973 CARP_LOCK(sc); 1974 /* Shift array. */ 1975 index = sc->sc_naddrs + sc->sc_naddrs6; 1976 for (i = 0; i < index; i++) 1977 if (sc->sc_ifas[i] == ifa) 1978 break; 1979 KASSERT(i < index, ("%s: %p no backref", __func__, ifa)); 1980 for (; i < index - 1; i++) 1981 sc->sc_ifas[i] = sc->sc_ifas[i+1]; 1982 sc->sc_ifas[index - 1] = NULL; 1983 1984 switch (ifa->ifa_addr->sa_family) { 1985 #ifdef INET 1986 case AF_INET: 1987 cif->cif_naddrs--; 1988 sc->sc_naddrs--; 1989 break; 1990 #endif 1991 #ifdef INET6 1992 case AF_INET6: 1993 cif->cif_naddrs6--; 1994 sc->sc_naddrs6--; 1995 break; 1996 #endif 1997 } 1998 1999 carp_ifa_delroute(ifa); 2000 carp_multicast_cleanup(cif, ifa->ifa_addr->sa_family); 2001 2002 ifa->ifa_carp = NULL; 2003 ifa_free(ifa); 2004 2005 carp_hmac_prepare(sc); 2006 carp_sc_state(sc); 2007 2008 if (!keep_cif && sc->sc_naddrs == 0 && sc->sc_naddrs6 == 0) 2009 carp_destroy(sc); 2010 else 2011 CARP_UNLOCK(sc); 2012 2013 if (!keep_cif) 2014 CIF_FREE(cif); 2015 2016 sx_xunlock(&carp_sx); 2017 } 2018 2019 static void 2020 carp_set_state(struct carp_softc *sc, int state, const char *reason) 2021 { 2022 2023 CARP_LOCK_ASSERT(sc); 2024 2025 if (sc->sc_state != state) { 2026 const char *carp_states[] = { CARP_STATES }; 2027 char subsys[IFNAMSIZ+5]; 2028 2029 snprintf(subsys, IFNAMSIZ+5, "%u@%s", sc->sc_vhid, 2030 sc->sc_carpdev->if_xname); 2031 2032 CARP_LOG("%s: %s -> %s (%s)\n", subsys, 2033 carp_states[sc->sc_state], carp_states[state], reason); 2034 2035 sc->sc_state = state; 2036 2037 devctl_notify("CARP", subsys, carp_states[state], NULL); 2038 } 2039 } 2040 2041 static void 2042 carp_linkstate(struct ifnet *ifp) 2043 { 2044 struct carp_softc *sc; 2045 2046 CIF_LOCK(ifp->if_carp); 2047 IFNET_FOREACH_CARP(ifp, sc) { 2048 CARP_LOCK(sc); 2049 carp_sc_state(sc); 2050 CARP_UNLOCK(sc); 2051 } 2052 CIF_UNLOCK(ifp->if_carp); 2053 } 2054 2055 static void 2056 carp_sc_state(struct carp_softc *sc) 2057 { 2058 2059 CARP_LOCK_ASSERT(sc); 2060 2061 if (sc->sc_carpdev->if_link_state != LINK_STATE_UP || 2062 !(sc->sc_carpdev->if_flags & IFF_UP) || 2063 !V_carp_allow) { 2064 callout_stop(&sc->sc_ad_tmo); 2065 #ifdef INET 2066 callout_stop(&sc->sc_md_tmo); 2067 #endif 2068 #ifdef INET6 2069 callout_stop(&sc->sc_md6_tmo); 2070 #endif 2071 carp_set_state(sc, INIT, "hardware interface down"); 2072 carp_setrun(sc, 0); 2073 if (!sc->sc_suppress) 2074 carp_demote_adj(V_carp_ifdown_adj, "interface down"); 2075 sc->sc_suppress = 1; 2076 } else { 2077 carp_set_state(sc, INIT, "hardware interface up"); 2078 carp_setrun(sc, 0); 2079 if (sc->sc_suppress) 2080 carp_demote_adj(-V_carp_ifdown_adj, "interface up"); 2081 sc->sc_suppress = 0; 2082 } 2083 } 2084 2085 static void 2086 carp_demote_adj(int adj, char *reason) 2087 { 2088 atomic_add_int(&V_carp_demotion, adj); 2089 CARP_LOG("demoted by %d to %d (%s)\n", adj, V_carp_demotion, reason); 2090 taskqueue_enqueue(taskqueue_swi, &carp_sendall_task); 2091 } 2092 2093 static int 2094 carp_allow_sysctl(SYSCTL_HANDLER_ARGS) 2095 { 2096 int new, error; 2097 struct carp_softc *sc; 2098 2099 new = V_carp_allow; 2100 error = sysctl_handle_int(oidp, &new, 0, req); 2101 if (error || !req->newptr) 2102 return (error); 2103 2104 if (V_carp_allow != new) { 2105 V_carp_allow = new; 2106 2107 mtx_lock(&carp_mtx); 2108 LIST_FOREACH(sc, &carp_list, sc_next) { 2109 CARP_LOCK(sc); 2110 if (curvnet == sc->sc_carpdev->if_vnet) 2111 carp_sc_state(sc); 2112 CARP_UNLOCK(sc); 2113 } 2114 mtx_unlock(&carp_mtx); 2115 } 2116 2117 return (0); 2118 } 2119 2120 static int 2121 carp_dscp_sysctl(SYSCTL_HANDLER_ARGS) 2122 { 2123 int new, error; 2124 2125 new = V_carp_dscp; 2126 error = sysctl_handle_int(oidp, &new, 0, req); 2127 if (error || !req->newptr) 2128 return (error); 2129 2130 if (new < 0 || new > 63) 2131 return (EINVAL); 2132 2133 V_carp_dscp = new; 2134 2135 return (0); 2136 } 2137 2138 static int 2139 carp_demote_adj_sysctl(SYSCTL_HANDLER_ARGS) 2140 { 2141 int new, error; 2142 2143 new = V_carp_demotion; 2144 error = sysctl_handle_int(oidp, &new, 0, req); 2145 if (error || !req->newptr) 2146 return (error); 2147 2148 carp_demote_adj(new, "sysctl"); 2149 2150 return (0); 2151 } 2152 2153 #ifdef INET 2154 extern struct domain inetdomain; 2155 static struct protosw in_carp_protosw = { 2156 .pr_type = SOCK_RAW, 2157 .pr_domain = &inetdomain, 2158 .pr_protocol = IPPROTO_CARP, 2159 .pr_flags = PR_ATOMIC|PR_ADDR, 2160 .pr_input = carp_input, 2161 .pr_output = rip_output, 2162 .pr_ctloutput = rip_ctloutput, 2163 .pr_usrreqs = &rip_usrreqs 2164 }; 2165 #endif 2166 2167 #ifdef INET6 2168 extern struct domain inet6domain; 2169 static struct protosw in6_carp_protosw = { 2170 .pr_type = SOCK_RAW, 2171 .pr_domain = &inet6domain, 2172 .pr_protocol = IPPROTO_CARP, 2173 .pr_flags = PR_ATOMIC|PR_ADDR, 2174 .pr_input = carp6_input, 2175 .pr_output = rip6_output, 2176 .pr_ctloutput = rip6_ctloutput, 2177 .pr_usrreqs = &rip6_usrreqs 2178 }; 2179 #endif 2180 2181 static void 2182 carp_mod_cleanup(void) 2183 { 2184 2185 #ifdef INET 2186 if (proto_reg[CARP_INET] == 0) { 2187 (void)ipproto_unregister(IPPROTO_CARP); 2188 pf_proto_unregister(PF_INET, IPPROTO_CARP, SOCK_RAW); 2189 proto_reg[CARP_INET] = -1; 2190 } 2191 carp_iamatch_p = NULL; 2192 #endif 2193 #ifdef INET6 2194 if (proto_reg[CARP_INET6] == 0) { 2195 (void)ip6proto_unregister(IPPROTO_CARP); 2196 pf_proto_unregister(PF_INET6, IPPROTO_CARP, SOCK_RAW); 2197 proto_reg[CARP_INET6] = -1; 2198 } 2199 carp_iamatch6_p = NULL; 2200 carp_macmatch6_p = NULL; 2201 #endif 2202 carp_ioctl_p = NULL; 2203 carp_attach_p = NULL; 2204 carp_detach_p = NULL; 2205 carp_get_vhid_p = NULL; 2206 carp_linkstate_p = NULL; 2207 carp_forus_p = NULL; 2208 carp_output_p = NULL; 2209 carp_demote_adj_p = NULL; 2210 carp_master_p = NULL; 2211 mtx_unlock(&carp_mtx); 2212 taskqueue_drain(taskqueue_swi, &carp_sendall_task); 2213 mtx_destroy(&carp_mtx); 2214 sx_destroy(&carp_sx); 2215 } 2216 2217 static int 2218 carp_mod_load(void) 2219 { 2220 int err; 2221 2222 mtx_init(&carp_mtx, "carp_mtx", NULL, MTX_DEF); 2223 sx_init(&carp_sx, "carp_sx"); 2224 LIST_INIT(&carp_list); 2225 carp_get_vhid_p = carp_get_vhid; 2226 carp_forus_p = carp_forus; 2227 carp_output_p = carp_output; 2228 carp_linkstate_p = carp_linkstate; 2229 carp_ioctl_p = carp_ioctl; 2230 carp_attach_p = carp_attach; 2231 carp_detach_p = carp_detach; 2232 carp_demote_adj_p = carp_demote_adj; 2233 carp_master_p = carp_master; 2234 #ifdef INET6 2235 carp_iamatch6_p = carp_iamatch6; 2236 carp_macmatch6_p = carp_macmatch6; 2237 proto_reg[CARP_INET6] = pf_proto_register(PF_INET6, 2238 (struct protosw *)&in6_carp_protosw); 2239 if (proto_reg[CARP_INET6]) { 2240 printf("carp: error %d attaching to PF_INET6\n", 2241 proto_reg[CARP_INET6]); 2242 carp_mod_cleanup(); 2243 return (proto_reg[CARP_INET6]); 2244 } 2245 err = ip6proto_register(IPPROTO_CARP); 2246 if (err) { 2247 printf("carp: error %d registering with INET6\n", err); 2248 carp_mod_cleanup(); 2249 return (err); 2250 } 2251 #endif 2252 #ifdef INET 2253 carp_iamatch_p = carp_iamatch; 2254 proto_reg[CARP_INET] = pf_proto_register(PF_INET, &in_carp_protosw); 2255 if (proto_reg[CARP_INET]) { 2256 printf("carp: error %d attaching to PF_INET\n", 2257 proto_reg[CARP_INET]); 2258 carp_mod_cleanup(); 2259 return (proto_reg[CARP_INET]); 2260 } 2261 err = ipproto_register(IPPROTO_CARP); 2262 if (err) { 2263 printf("carp: error %d registering with INET\n", err); 2264 carp_mod_cleanup(); 2265 return (err); 2266 } 2267 #endif 2268 return (0); 2269 } 2270 2271 static int 2272 carp_modevent(module_t mod, int type, void *data) 2273 { 2274 switch (type) { 2275 case MOD_LOAD: 2276 return carp_mod_load(); 2277 /* NOTREACHED */ 2278 case MOD_UNLOAD: 2279 mtx_lock(&carp_mtx); 2280 if (LIST_EMPTY(&carp_list)) 2281 carp_mod_cleanup(); 2282 else { 2283 mtx_unlock(&carp_mtx); 2284 return (EBUSY); 2285 } 2286 break; 2287 2288 default: 2289 return (EINVAL); 2290 } 2291 2292 return (0); 2293 } 2294 2295 static moduledata_t carp_mod = { 2296 "carp", 2297 carp_modevent, 2298 0 2299 }; 2300 2301 DECLARE_MODULE(carp, carp_mod, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY); 2302