1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 2021 Ng Peng Nam Sean 5 * Copyright (c) 2022 Alexander V. Chernikov <melifaro@FreeBSD.org> 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 26 * SUCH DAMAGE. 27 */ 28 29 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 31 #include "opt_inet.h" 32 #include "opt_inet6.h" 33 #include "opt_route.h" 34 #include <sys/types.h> 35 #include <sys/malloc.h> 36 #include <sys/rmlock.h> 37 #include <sys/socket.h> 38 39 #include <net/if.h> 40 #include <net/route.h> 41 #include <net/route/nhop.h> 42 #include <net/route/route_ctl.h> 43 #include <net/route/route_var.h> 44 #include <netlink/netlink.h> 45 #include <netlink/netlink_ctl.h> 46 #include <netlink/netlink_route.h> 47 #include <netlink/route/route_var.h> 48 49 #define DEBUG_MOD_NAME nl_route 50 #define DEBUG_MAX_LEVEL LOG_DEBUG3 51 #include <netlink/netlink_debug.h> 52 _DECLARE_DEBUG(LOG_DEBUG); 53 54 static unsigned char 55 get_rtm_type(const struct nhop_object *nh) 56 { 57 int nh_flags = nh->nh_flags; 58 59 /* Use the fact that nhg runtime flags are only NHF_MULTIPATH */ 60 if (nh_flags & NHF_BLACKHOLE) 61 return (RTN_BLACKHOLE); 62 else if (nh_flags & NHF_REJECT) 63 return (RTN_PROHIBIT); 64 return (RTN_UNICAST); 65 } 66 67 static uint8_t 68 nl_get_rtm_protocol(const struct nhop_object *nh) 69 { 70 #ifdef ROUTE_MPATH 71 if (NH_IS_NHGRP(nh)) { 72 const struct nhgrp_object *nhg = (const struct nhgrp_object *)nh; 73 uint8_t origin = nhgrp_get_origin(nhg); 74 if (origin != RTPROT_UNSPEC) 75 return (origin); 76 nh = nhg->nhops[0]; 77 } 78 #endif 79 uint8_t origin = nhop_get_origin(nh); 80 if (origin != RTPROT_UNSPEC) 81 return (origin); 82 /* TODO: remove guesswork once all kernel users fill in origin */ 83 int rt_flags = nhop_get_rtflags(nh); 84 if (rt_flags & RTF_PROTO1) 85 return (RTPROT_ZEBRA); 86 if (rt_flags & RTF_STATIC) 87 return (RTPROT_STATIC); 88 return (RTPROT_KERNEL); 89 } 90 91 static int 92 get_rtmsg_type_from_rtsock(int cmd) 93 { 94 switch (cmd) { 95 case RTM_ADD: 96 case RTM_CHANGE: 97 case RTM_GET: 98 return NL_RTM_NEWROUTE; 99 case RTM_DELETE: 100 return NL_RTM_DELROUTE; 101 } 102 103 return (0); 104 } 105 106 /* 107 * fibnum heuristics 108 * 109 * if (dump && rtm_table == 0 && !rta_table) RT_ALL_FIBS 110 * msg rtm_table RTA_TABLE result 111 * RTM_GETROUTE/dump 0 - RT_ALL_FIBS 112 * RTM_GETROUTE/dump 1 - 1 113 * RTM_GETROUTE/get 0 - 0 114 * 115 */ 116 117 static struct nhop_object * 118 rc_get_nhop(const struct rib_cmd_info *rc) 119 { 120 return ((rc->rc_cmd == RTM_DELETE) ? rc->rc_nh_old : rc->rc_nh_new); 121 } 122 123 static void 124 dump_rc_nhop_gw(struct nl_writer *nw, const struct nhop_object *nh) 125 { 126 int upper_family; 127 128 switch (nhop_get_neigh_family(nh)) { 129 case AF_LINK: 130 /* onlink prefix, skip */ 131 break; 132 case AF_INET: 133 nlattr_add(nw, NL_RTA_GATEWAY, 4, &nh->gw4_sa.sin_addr); 134 break; 135 case AF_INET6: 136 upper_family = nhop_get_upper_family(nh); 137 if (upper_family == AF_INET6) { 138 nlattr_add(nw, NL_RTA_GATEWAY, 16, &nh->gw6_sa.sin6_addr); 139 } else if (upper_family == AF_INET) { 140 /* IPv4 over IPv6 */ 141 char buf[20]; 142 struct rtvia *via = (struct rtvia *)&buf[0]; 143 via->rtvia_family = AF_INET6; 144 memcpy(via->rtvia_addr, &nh->gw6_sa.sin6_addr, 16); 145 nlattr_add(nw, NL_RTA_VIA, 17, via); 146 } 147 break; 148 } 149 } 150 151 static void 152 dump_rc_nhop_mtu(struct nl_writer *nw, const struct nhop_object *nh) 153 { 154 int nla_len = sizeof(struct nlattr) * 2 + sizeof(uint32_t); 155 struct nlattr *nla = nlmsg_reserve_data(nw, nla_len, struct nlattr); 156 157 if (nla == NULL) 158 return; 159 nla->nla_type = NL_RTA_METRICS; 160 nla->nla_len = nla_len; 161 nla++; 162 nla->nla_type = NL_RTAX_MTU; 163 nla->nla_len = sizeof(struct nlattr) + sizeof(uint32_t); 164 *((uint32_t *)(nla + 1)) = nh->nh_mtu; 165 } 166 167 #ifdef ROUTE_MPATH 168 static void 169 dump_rc_nhg(struct nl_writer *nw, const struct nhgrp_object *nhg, struct rtmsg *rtm) 170 { 171 uint32_t uidx = nhgrp_get_uidx(nhg); 172 uint32_t num_nhops; 173 const struct weightened_nhop *wn = nhgrp_get_nhops(nhg, &num_nhops); 174 uint32_t base_rtflags = nhop_get_rtflags(wn[0].nh); 175 176 if (uidx != 0) 177 nlattr_add_u32(nw, NL_RTA_NH_ID, uidx); 178 nlattr_add_u32(nw, NL_RTA_KNH_ID, nhgrp_get_idx(nhg)); 179 180 nlattr_add_u32(nw, NL_RTA_RTFLAGS, base_rtflags); 181 int off = nlattr_add_nested(nw, NL_RTA_MULTIPATH); 182 if (off == 0) 183 return; 184 185 for (int i = 0; i < num_nhops; i++) { 186 int nh_off = nlattr_save_offset(nw); 187 struct rtnexthop *rtnh = nlmsg_reserve_object(nw, struct rtnexthop); 188 if (rtnh == NULL) 189 return; 190 rtnh->rtnh_flags = 0; 191 rtnh->rtnh_ifindex = wn[i].nh->nh_ifp->if_index; 192 rtnh->rtnh_hops = wn[i].weight; 193 dump_rc_nhop_gw(nw, wn[i].nh); 194 uint32_t rtflags = nhop_get_rtflags(wn[i].nh); 195 if (rtflags != base_rtflags) 196 nlattr_add_u32(nw, NL_RTA_RTFLAGS, rtflags); 197 if (rtflags & RTF_FIXEDMTU) 198 dump_rc_nhop_mtu(nw, wn[i].nh); 199 rtnh = nlattr_restore_offset(nw, nh_off, struct rtnexthop); 200 /* 201 * nlattr_add() allocates 4-byte aligned storage, no need to aligh 202 * length here 203 * */ 204 rtnh->rtnh_len = nlattr_save_offset(nw) - nh_off; 205 } 206 nlattr_set_len(nw, off); 207 } 208 #endif 209 210 static void 211 dump_rc_nhop(struct nl_writer *nw, const struct nhop_object *nh, struct rtmsg *rtm) 212 { 213 #ifdef ROUTE_MPATH 214 if (NH_IS_NHGRP(nh)) { 215 dump_rc_nhg(nw, (const struct nhgrp_object *)nh, rtm); 216 return; 217 } 218 #endif 219 uint32_t rtflags = nhop_get_rtflags(nh); 220 221 /* 222 * IPv4 over IPv6 223 * ('RTA_VIA', {'family': 10, 'addr': 'fe80::20c:29ff:fe67:2dd'}), ('RTA_OIF', 2), 224 * IPv4 w/ gw 225 * ('RTA_GATEWAY', '172.16.107.131'), ('RTA_OIF', 2)], 226 * Direct route: 227 * ('RTA_OIF', 2) 228 */ 229 if (nh->nh_flags & NHF_GATEWAY) 230 dump_rc_nhop_gw(nw, nh); 231 232 uint32_t uidx = nhop_get_uidx(nh); 233 if (uidx != 0) 234 nlattr_add_u32(nw, NL_RTA_NH_ID, uidx); 235 nlattr_add_u32(nw, NL_RTA_KNH_ID, nhop_get_idx(nh)); 236 nlattr_add_u32(nw, NL_RTA_RTFLAGS, rtflags); 237 238 if (rtflags & RTF_FIXEDMTU) 239 dump_rc_nhop_mtu(nw, nh); 240 uint32_t nh_expire = nhop_get_expire(nh); 241 if (nh_expire > 0) 242 nlattr_add_u32(nw, NL_RTA_EXPIRES, nh_expire - time_uptime); 243 244 /* In any case, fill outgoing interface */ 245 nlattr_add_u32(nw, NL_RTA_OIF, nh->nh_ifp->if_index); 246 } 247 248 /* 249 * Dumps output from a rib command into an rtmsg 250 */ 251 252 static int 253 dump_px(uint32_t fibnum, const struct nlmsghdr *hdr, 254 const struct rtentry *rt, struct route_nhop_data *rnd, 255 struct nl_writer *nw) 256 { 257 struct rtmsg *rtm; 258 int error = 0; 259 260 NET_EPOCH_ASSERT(); 261 262 if (!nlmsg_reply(nw, hdr, sizeof(struct rtmsg))) 263 goto enomem; 264 265 int family = rt_get_family(rt); 266 int rtm_off = nlattr_save_offset(nw); 267 rtm = nlmsg_reserve_object(nw, struct rtmsg); 268 rtm->rtm_family = family; 269 rtm->rtm_dst_len = 0; 270 rtm->rtm_src_len = 0; 271 rtm->rtm_tos = 0; 272 if (fibnum < 255) 273 rtm->rtm_table = (unsigned char)fibnum; 274 rtm->rtm_scope = RT_SCOPE_UNIVERSE; 275 rtm->rtm_protocol = nl_get_rtm_protocol(rnd->rnd_nhop); 276 rtm->rtm_type = get_rtm_type(rnd->rnd_nhop); 277 278 nlattr_add_u32(nw, NL_RTA_TABLE, fibnum); 279 280 int plen = 0; 281 #if defined(INET) || defined(INET6) 282 uint32_t scopeid; 283 #endif 284 switch (family) { 285 #ifdef INET 286 case AF_INET: 287 { 288 struct in_addr addr; 289 rt_get_inet_prefix_plen(rt, &addr, &plen, &scopeid); 290 nlattr_add(nw, NL_RTA_DST, 4, &addr); 291 break; 292 } 293 #endif 294 #ifdef INET6 295 case AF_INET6: 296 { 297 struct in6_addr addr; 298 rt_get_inet6_prefix_plen(rt, &addr, &plen, &scopeid); 299 nlattr_add(nw, NL_RTA_DST, 16, &addr); 300 break; 301 } 302 #endif 303 default: 304 FIB_LOG(LOG_NOTICE, fibnum, family, "unsupported rt family: %d", family); 305 error = EAFNOSUPPORT; 306 goto flush; 307 } 308 309 rtm = nlattr_restore_offset(nw, rtm_off, struct rtmsg); 310 if (plen > 0) 311 rtm->rtm_dst_len = plen; 312 dump_rc_nhop(nw, rnd->rnd_nhop, rtm); 313 314 if (nlmsg_end(nw)) 315 return (0); 316 enomem: 317 error = ENOMEM; 318 flush: 319 nlmsg_abort(nw); 320 return (error); 321 } 322 323 static int 324 family_to_group(int family) 325 { 326 switch (family) { 327 case AF_INET: 328 return (RTNLGRP_IPV4_ROUTE); 329 case AF_INET6: 330 return (RTNLGRP_IPV6_ROUTE); 331 } 332 return (0); 333 } 334 335 336 static void 337 report_operation(uint32_t fibnum, struct rib_cmd_info *rc, 338 struct nlpcb *nlp, struct nlmsghdr *hdr) 339 { 340 struct nl_writer nw = {}; 341 uint32_t group_id = family_to_group(rt_get_family(rc->rc_rt)); 342 343 if (nlmsg_get_group_writer(&nw, NLMSG_SMALL, NETLINK_ROUTE, group_id)) { 344 struct route_nhop_data rnd = { 345 .rnd_nhop = rc_get_nhop(rc), 346 .rnd_weight = rc->rc_nh_weight, 347 }; 348 hdr->nlmsg_flags &= ~(NLM_F_REPLACE | NLM_F_CREATE); 349 hdr->nlmsg_flags &= ~(NLM_F_EXCL | NLM_F_APPEND); 350 switch (rc->rc_cmd) { 351 case RTM_ADD: 352 hdr->nlmsg_type = NL_RTM_NEWROUTE; 353 hdr->nlmsg_flags |= NLM_F_CREATE | NLM_F_EXCL; 354 break; 355 case RTM_CHANGE: 356 hdr->nlmsg_type = NL_RTM_NEWROUTE; 357 hdr->nlmsg_flags |= NLM_F_REPLACE; 358 break; 359 case RTM_DELETE: 360 hdr->nlmsg_type = NL_RTM_DELROUTE; 361 break; 362 } 363 dump_px(fibnum, hdr, rc->rc_rt, &rnd, &nw); 364 nlmsg_flush(&nw); 365 } 366 367 rtsock_callback_p->route_f(fibnum, rc); 368 } 369 370 struct rta_mpath_nh { 371 struct sockaddr *gw; 372 struct ifnet *ifp; 373 uint8_t rtnh_flags; 374 uint8_t rtnh_weight; 375 }; 376 377 #define _IN(_field) offsetof(struct rtnexthop, _field) 378 #define _OUT(_field) offsetof(struct rta_mpath_nh, _field) 379 const static struct nlattr_parser nla_p_rtnh[] = { 380 { .type = NL_RTA_GATEWAY, .off = _OUT(gw), .cb = nlattr_get_ip }, 381 { .type = NL_RTA_VIA, .off = _OUT(gw), .cb = nlattr_get_ipvia }, 382 }; 383 const static struct nlfield_parser nlf_p_rtnh[] = { 384 { .off_in = _IN(rtnh_flags), .off_out = _OUT(rtnh_flags), .cb = nlf_get_u8 }, 385 { .off_in = _IN(rtnh_hops), .off_out = _OUT(rtnh_weight), .cb = nlf_get_u8 }, 386 { .off_in = _IN(rtnh_ifindex), .off_out = _OUT(ifp), .cb = nlf_get_ifpz }, 387 }; 388 #undef _IN 389 #undef _OUT 390 NL_DECLARE_PARSER(mpath_parser, struct rtnexthop, nlf_p_rtnh, nla_p_rtnh); 391 392 struct rta_mpath { 393 int num_nhops; 394 struct rta_mpath_nh nhops[0]; 395 }; 396 397 static int 398 nlattr_get_multipath(struct nlattr *nla, struct nl_pstate *npt, const void *arg, void *target) 399 { 400 int data_len = nla->nla_len - sizeof(struct nlattr); 401 struct rtnexthop *rtnh; 402 403 int max_nhops = data_len / sizeof(struct rtnexthop); 404 405 struct rta_mpath *mp = npt_alloc(npt, (max_nhops + 2) * sizeof(struct rta_mpath_nh)); 406 mp->num_nhops = 0; 407 408 for (rtnh = (struct rtnexthop *)(nla + 1); data_len > 0; ) { 409 struct rta_mpath_nh *mpnh = &mp->nhops[mp->num_nhops++]; 410 411 int error = nl_parse_header(rtnh, rtnh->rtnh_len, &mpath_parser, 412 npt, mpnh); 413 if (error != 0) { 414 NLMSG_REPORT_ERR_MSG(npt, "RTA_MULTIPATH: nexhop %d: parse failed", 415 mp->num_nhops - 1); 416 return (error); 417 } 418 419 int len = NL_ITEM_ALIGN(rtnh->rtnh_len); 420 data_len -= len; 421 rtnh = (struct rtnexthop *)((char *)rtnh + len); 422 } 423 if (data_len != 0 || mp->num_nhops == 0) { 424 NLMSG_REPORT_ERR_MSG(npt, "invalid RTA_MULTIPATH attr"); 425 return (EINVAL); 426 } 427 428 *((struct rta_mpath **)target) = mp; 429 return (0); 430 } 431 432 433 struct nl_parsed_route { 434 struct sockaddr *rta_dst; 435 struct sockaddr *rta_gw; 436 struct ifnet *rta_oif; 437 struct rta_mpath *rta_multipath; 438 uint32_t rta_table; 439 uint32_t rta_rtflags; 440 uint32_t rta_nh_id; 441 uint32_t rtax_mtu; 442 uint8_t rtm_family; 443 uint8_t rtm_dst_len; 444 uint8_t rtm_protocol; 445 }; 446 447 #define _IN(_field) offsetof(struct rtmsg, _field) 448 #define _OUT(_field) offsetof(struct nl_parsed_route, _field) 449 static struct nlattr_parser nla_p_rtmetrics[] = { 450 { .type = NL_RTAX_MTU, .off = _OUT(rtax_mtu), .cb = nlattr_get_uint32 }, 451 }; 452 NL_DECLARE_ATTR_PARSER(metrics_parser, nla_p_rtmetrics); 453 454 static const struct nlattr_parser nla_p_rtmsg[] = { 455 { .type = NL_RTA_DST, .off = _OUT(rta_dst), .cb = nlattr_get_ip }, 456 { .type = NL_RTA_OIF, .off = _OUT(rta_oif), .cb = nlattr_get_ifp }, 457 { .type = NL_RTA_GATEWAY, .off = _OUT(rta_gw), .cb = nlattr_get_ip }, 458 { .type = NL_RTA_METRICS, .arg = &metrics_parser, .cb = nlattr_get_nested }, 459 { .type = NL_RTA_MULTIPATH, .off = _OUT(rta_multipath), .cb = nlattr_get_multipath }, 460 { .type = NL_RTA_RTFLAGS, .off = _OUT(rta_rtflags), .cb = nlattr_get_uint32 }, 461 { .type = NL_RTA_TABLE, .off = _OUT(rta_table), .cb = nlattr_get_uint32 }, 462 { .type = NL_RTA_VIA, .off = _OUT(rta_gw), .cb = nlattr_get_ipvia }, 463 { .type = NL_RTA_NH_ID, .off = _OUT(rta_nh_id), .cb = nlattr_get_uint32 }, 464 }; 465 466 static const struct nlfield_parser nlf_p_rtmsg[] = { 467 {.off_in = _IN(rtm_family), .off_out = _OUT(rtm_family), .cb = nlf_get_u8 }, 468 {.off_in = _IN(rtm_dst_len), .off_out = _OUT(rtm_dst_len), .cb = nlf_get_u8 }, 469 {.off_in = _IN(rtm_protocol), .off_out = _OUT(rtm_protocol), .cb = nlf_get_u8 }, 470 }; 471 #undef _IN 472 #undef _OUT 473 NL_DECLARE_PARSER(rtm_parser, struct rtmsg, nlf_p_rtmsg, nla_p_rtmsg); 474 475 struct netlink_walkargs { 476 struct nl_writer *nw; 477 struct route_nhop_data rnd; 478 struct nlmsghdr hdr; 479 struct nlpcb *nlp; 480 uint32_t fibnum; 481 int family; 482 int error; 483 int count; 484 int dumped; 485 int dumped_tables; 486 }; 487 488 static int 489 dump_rtentry(struct rtentry *rt, void *_arg) 490 { 491 struct netlink_walkargs *wa = (struct netlink_walkargs *)_arg; 492 int error; 493 494 wa->count++; 495 if (wa->error != 0) 496 return (0); 497 wa->dumped++; 498 499 rt_get_rnd(rt, &wa->rnd); 500 501 error = dump_px(wa->fibnum, &wa->hdr, rt, &wa->rnd, wa->nw); 502 503 IF_DEBUG_LEVEL(LOG_DEBUG3) { 504 char rtbuf[INET6_ADDRSTRLEN + 5]; 505 FIB_LOG(LOG_DEBUG3, wa->fibnum, wa->family, 506 "Dump %s, offset %u, error %d", 507 rt_print_buf(rt, rtbuf, sizeof(rtbuf)), 508 wa->nw->offset, error); 509 } 510 wa->error = error; 511 512 return (0); 513 } 514 515 static void 516 dump_rtable_one(struct netlink_walkargs *wa, uint32_t fibnum, int family) 517 { 518 FIB_LOG(LOG_DEBUG2, fibnum, family, "Start dump"); 519 wa->count = 0; 520 wa->dumped = 0; 521 522 rib_walk(fibnum, family, false, dump_rtentry, wa); 523 524 wa->dumped_tables++; 525 526 FIB_LOG(LOG_DEBUG2, fibnum, family, "End dump, iterated %d dumped %d", 527 wa->count, wa->dumped); 528 NL_LOG(LOG_DEBUG2, "Current offset: %d", wa->nw->offset); 529 } 530 531 static int 532 dump_rtable_fib(struct netlink_walkargs *wa, uint32_t fibnum, int family) 533 { 534 wa->fibnum = fibnum; 535 536 if (family == AF_UNSPEC) { 537 for (int i = 0; i < AF_MAX; i++) { 538 if (rt_tables_get_rnh(fibnum, i) != 0) { 539 wa->family = i; 540 dump_rtable_one(wa, fibnum, i); 541 if (wa->error != 0) 542 break; 543 } 544 } 545 } else { 546 if (rt_tables_get_rnh(fibnum, family) != 0) { 547 wa->family = family; 548 dump_rtable_one(wa, fibnum, family); 549 } 550 } 551 552 return (wa->error); 553 } 554 555 static int 556 handle_rtm_getroute(struct nlpcb *nlp, struct nl_parsed_route *attrs, 557 struct nlmsghdr *hdr, struct nl_pstate *npt) 558 { 559 RIB_RLOCK_TRACKER; 560 struct rib_head *rnh; 561 struct rtentry *rt; 562 uint32_t fibnum = attrs->rta_table; 563 sa_family_t family = attrs->rtm_family; 564 565 if (attrs->rta_dst == NULL) { 566 NLMSG_REPORT_ERR_MSG(npt, "No RTA_DST supplied"); 567 return (EINVAL); 568 } 569 570 FIB_LOG(LOG_DEBUG, fibnum, family, "getroute called"); 571 572 rnh = rt_tables_get_rnh(fibnum, family); 573 if (rnh == NULL) 574 return (EAFNOSUPPORT); 575 576 RIB_RLOCK(rnh); 577 578 rt = (struct rtentry *)rnh->rnh_matchaddr(attrs->rta_dst, &rnh->head); 579 if (rt == NULL) { 580 RIB_RUNLOCK(rnh); 581 return (ESRCH); 582 } 583 584 struct route_nhop_data rnd; 585 rt_get_rnd(rt, &rnd); 586 rnd.rnd_nhop = nhop_select_func(rnd.rnd_nhop, 0); 587 588 RIB_RUNLOCK(rnh); 589 590 IF_DEBUG_LEVEL(LOG_DEBUG2) { 591 char rtbuf[NHOP_PRINT_BUFSIZE] __unused, nhbuf[NHOP_PRINT_BUFSIZE] __unused; 592 FIB_LOG(LOG_DEBUG2, fibnum, family, "getroute completed: got %s for %s", 593 nhop_print_buf_any(rnd.rnd_nhop, nhbuf, sizeof(nhbuf)), 594 rt_print_buf(rt, rtbuf, sizeof(rtbuf))); 595 } 596 597 hdr->nlmsg_type = NL_RTM_NEWROUTE; 598 dump_px(fibnum, hdr, rt, &rnd, npt->nw); 599 600 return (0); 601 } 602 603 static int 604 handle_rtm_dump(struct nlpcb *nlp, uint32_t fibnum, int family, 605 struct nlmsghdr *hdr, struct nl_writer *nw) 606 { 607 struct netlink_walkargs wa = { 608 .nlp = nlp, 609 .nw = nw, 610 .hdr.nlmsg_pid = hdr->nlmsg_pid, 611 .hdr.nlmsg_seq = hdr->nlmsg_seq, 612 .hdr.nlmsg_type = NL_RTM_NEWROUTE, 613 .hdr.nlmsg_flags = hdr->nlmsg_flags | NLM_F_MULTI, 614 }; 615 616 if (fibnum == RT_TABLE_UNSPEC) { 617 for (int i = 0; i < V_rt_numfibs; i++) { 618 dump_rtable_fib(&wa, fibnum, family); 619 if (wa.error != 0) 620 break; 621 } 622 } else 623 dump_rtable_fib(&wa, fibnum, family); 624 625 if (wa.error == 0 && wa.dumped_tables == 0) { 626 FIB_LOG(LOG_DEBUG, fibnum, family, "incorrect fibnum/family"); 627 wa.error = ESRCH; 628 // How do we propagate it? 629 } 630 631 if (!nlmsg_end_dump(wa.nw, wa.error, &wa.hdr)) { 632 NL_LOG(LOG_DEBUG, "Unable to finalize the dump"); 633 return (ENOMEM); 634 } 635 636 return (wa.error); 637 } 638 639 static struct nhop_object * 640 finalize_nhop(struct nhop_object *nh, int *perror) 641 { 642 /* 643 * The following MUST be filled: 644 * nh_ifp, nh_ifa, nh_gw 645 */ 646 if (nh->gw_sa.sa_family == 0) { 647 /* 648 * Empty gateway. Can be direct route with RTA_OIF set. 649 */ 650 if (nh->nh_ifp != NULL) 651 nhop_set_direct_gw(nh, nh->nh_ifp); 652 else { 653 NL_LOG(LOG_DEBUG, "empty gateway and interface, skipping"); 654 *perror = EINVAL; 655 return (NULL); 656 } 657 /* Both nh_ifp and gateway are set */ 658 } else { 659 /* Gateway is set up, we can derive ifp if not set */ 660 if (nh->nh_ifp == NULL) { 661 struct ifaddr *ifa = ifa_ifwithnet(&nh->gw_sa, 1, nhop_get_fibnum(nh)); 662 if (ifa == NULL) { 663 NL_LOG(LOG_DEBUG, "Unable to determine ifp, skipping"); 664 *perror = EINVAL; 665 return (NULL); 666 } 667 nhop_set_transmit_ifp(nh, ifa->ifa_ifp); 668 } 669 } 670 /* Both nh_ifp and gateway are set */ 671 if (nh->nh_ifa == NULL) { 672 struct ifaddr *ifa = ifaof_ifpforaddr(&nh->gw_sa, nh->nh_ifp); 673 if (ifa == NULL) { 674 NL_LOG(LOG_DEBUG, "Unable to determine ifa, skipping"); 675 *perror = EINVAL; 676 return (NULL); 677 } 678 nhop_set_src(nh, ifa); 679 } 680 681 return (nhop_get_nhop(nh, perror)); 682 } 683 684 static int 685 get_pxflag(const struct nl_parsed_route *attrs) 686 { 687 int pxflag = 0; 688 switch (attrs->rtm_family) { 689 case AF_INET: 690 if (attrs->rtm_dst_len == 32) 691 pxflag = NHF_HOST; 692 else if (attrs->rtm_dst_len == 0) 693 pxflag = NHF_DEFAULT; 694 break; 695 case AF_INET6: 696 if (attrs->rtm_dst_len == 32) 697 pxflag = NHF_HOST; 698 else if (attrs->rtm_dst_len == 0) 699 pxflag = NHF_DEFAULT; 700 break; 701 } 702 703 return (pxflag); 704 } 705 706 static int 707 get_op_flags(int nlm_flags) 708 { 709 int op_flags = 0; 710 711 op_flags |= (nlm_flags & NLM_F_REPLACE) ? RTM_F_REPLACE : 0; 712 op_flags |= (nlm_flags & NLM_F_EXCL) ? RTM_F_EXCL : 0; 713 op_flags |= (nlm_flags & NLM_F_CREATE) ? RTM_F_CREATE : 0; 714 op_flags |= (nlm_flags & NLM_F_APPEND) ? RTM_F_APPEND : 0; 715 716 return (op_flags); 717 } 718 719 #ifdef ROUTE_MPATH 720 static int 721 create_nexthop_one(struct nl_parsed_route *attrs, struct rta_mpath_nh *mpnh, 722 struct nl_pstate *npt, struct nhop_object **pnh) 723 { 724 int error; 725 726 if (mpnh->gw == NULL) 727 return (EINVAL); 728 729 struct nhop_object *nh = nhop_alloc(attrs->rta_table, attrs->rtm_family); 730 if (nh == NULL) 731 return (ENOMEM); 732 733 nhop_set_gw(nh, mpnh->gw, true); 734 if (mpnh->ifp != NULL) 735 nhop_set_transmit_ifp(nh, mpnh->ifp); 736 nhop_set_rtflags(nh, attrs->rta_rtflags); 737 if (attrs->rtm_protocol > RTPROT_STATIC) 738 nhop_set_origin(nh, attrs->rtm_protocol); 739 740 *pnh = finalize_nhop(nh, &error); 741 742 return (error); 743 } 744 #endif 745 746 static struct nhop_object * 747 create_nexthop_from_attrs(struct nl_parsed_route *attrs, 748 struct nl_pstate *npt, int *perror) 749 { 750 struct nhop_object *nh = NULL; 751 int error = 0; 752 753 if (attrs->rta_multipath != NULL) { 754 #ifdef ROUTE_MPATH 755 /* Multipath w/o explicit nexthops */ 756 int num_nhops = attrs->rta_multipath->num_nhops; 757 struct weightened_nhop *wn = npt_alloc(npt, sizeof(*wn) * num_nhops); 758 759 for (int i = 0; i < num_nhops; i++) { 760 struct rta_mpath_nh *mpnh = &attrs->rta_multipath->nhops[i]; 761 762 error = create_nexthop_one(attrs, mpnh, npt, &wn[i].nh); 763 if (error != 0) { 764 for (int j = 0; j < i; j++) 765 nhop_free(wn[j].nh); 766 break; 767 } 768 wn[i].weight = mpnh->rtnh_weight > 0 ? mpnh->rtnh_weight : 1; 769 } 770 if (error == 0) { 771 struct rib_head *rh = nhop_get_rh(wn[0].nh); 772 struct nhgrp_object *nhg; 773 774 nhg = nhgrp_alloc(rh->rib_fibnum, rh->rib_family, 775 wn, num_nhops, perror); 776 if (nhg != NULL) { 777 if (attrs->rtm_protocol > RTPROT_STATIC) 778 nhgrp_set_origin(nhg, attrs->rtm_protocol); 779 nhg = nhgrp_get_nhgrp(nhg, perror); 780 } 781 for (int i = 0; i < num_nhops; i++) 782 nhop_free(wn[i].nh); 783 if (nhg != NULL) 784 return ((struct nhop_object *)nhg); 785 error = *perror; 786 } 787 #else 788 error = ENOTSUP; 789 #endif 790 *perror = error; 791 } else { 792 nh = nhop_alloc(attrs->rta_table, attrs->rtm_family); 793 if (nh == NULL) { 794 *perror = ENOMEM; 795 return (NULL); 796 } 797 if (attrs->rta_gw != NULL) 798 nhop_set_gw(nh, attrs->rta_gw, true); 799 if (attrs->rta_oif != NULL) 800 nhop_set_transmit_ifp(nh, attrs->rta_oif); 801 if (attrs->rtax_mtu != 0) 802 nhop_set_mtu(nh, attrs->rtax_mtu, true); 803 if (attrs->rta_rtflags & RTF_BROADCAST) 804 nhop_set_broadcast(nh, true); 805 if (attrs->rta_rtflags & RTF_BLACKHOLE) 806 nhop_set_blackhole(nh, NHF_BLACKHOLE); 807 if (attrs->rta_rtflags & RTF_REJECT) 808 nhop_set_blackhole(nh, NHF_REJECT); 809 nhop_set_rtflags(nh, attrs->rta_rtflags); 810 if (attrs->rtm_protocol > RTPROT_STATIC) 811 nhop_set_origin(nh, attrs->rtm_protocol); 812 nh = finalize_nhop(nh, perror); 813 } 814 815 return (nh); 816 } 817 818 static int 819 rtnl_handle_newroute(struct nlmsghdr *hdr, struct nlpcb *nlp, 820 struct nl_pstate *npt) 821 { 822 struct rib_cmd_info rc = {}; 823 struct nhop_object *nh = NULL; 824 int error; 825 826 struct nl_parsed_route attrs = {}; 827 error = nl_parse_nlmsg(hdr, &rtm_parser, npt, &attrs); 828 if (error != 0) 829 return (error); 830 831 /* Check if we have enough data */ 832 if (attrs.rta_dst == NULL) { 833 NL_LOG(LOG_DEBUG, "missing RTA_DST"); 834 return (EINVAL); 835 } 836 837 if (attrs.rta_nh_id != 0) { 838 /* Referenced uindex */ 839 int pxflag = get_pxflag(&attrs); 840 nh = nl_find_nhop(attrs.rta_table, attrs.rtm_family, attrs.rta_nh_id, 841 pxflag, &error); 842 if (error != 0) 843 return (error); 844 } else { 845 nh = create_nexthop_from_attrs(&attrs, npt, &error); 846 if (error != 0) { 847 NL_LOG(LOG_DEBUG, "Error creating nexthop"); 848 return (error); 849 } 850 } 851 852 int weight = NH_IS_NHGRP(nh) ? 0 : RT_DEFAULT_WEIGHT; 853 struct route_nhop_data rnd = { .rnd_nhop = nh, .rnd_weight = weight }; 854 int op_flags = get_op_flags(hdr->nlmsg_flags); 855 856 error = rib_add_route_px(attrs.rta_table, attrs.rta_dst, attrs.rtm_dst_len, 857 &rnd, op_flags, &rc); 858 if (error == 0) 859 report_operation(attrs.rta_table, &rc, nlp, hdr); 860 return (error); 861 } 862 863 static int 864 path_match_func(const struct rtentry *rt, const struct nhop_object *nh, void *_data) 865 { 866 struct nl_parsed_route *attrs = (struct nl_parsed_route *)_data; 867 868 if ((attrs->rta_gw != NULL) && !rib_match_gw(rt, nh, attrs->rta_gw)) 869 return (0); 870 871 if ((attrs->rta_oif != NULL) && (attrs->rta_oif != nh->nh_ifp)) 872 return (0); 873 874 return (1); 875 } 876 877 static int 878 rtnl_handle_delroute(struct nlmsghdr *hdr, struct nlpcb *nlp, 879 struct nl_pstate *npt) 880 { 881 struct rib_cmd_info rc; 882 int error; 883 884 struct nl_parsed_route attrs = {}; 885 error = nl_parse_nlmsg(hdr, &rtm_parser, npt, &attrs); 886 if (error != 0) 887 return (error); 888 889 if (attrs.rta_dst == NULL) { 890 NLMSG_REPORT_ERR_MSG(npt, "RTA_DST is not set"); 891 return (ESRCH); 892 } 893 894 error = rib_del_route_px(attrs.rta_table, attrs.rta_dst, 895 attrs.rtm_dst_len, path_match_func, &attrs, 0, &rc); 896 if (error == 0) 897 report_operation(attrs.rta_table, &rc, nlp, hdr); 898 return (error); 899 } 900 901 static int 902 rtnl_handle_getroute(struct nlmsghdr *hdr, struct nlpcb *nlp, struct nl_pstate *npt) 903 { 904 int error; 905 906 struct nl_parsed_route attrs = {}; 907 error = nl_parse_nlmsg(hdr, &rtm_parser, npt, &attrs); 908 if (error != 0) 909 return (error); 910 911 if (hdr->nlmsg_flags & NLM_F_DUMP) 912 error = handle_rtm_dump(nlp, attrs.rta_table, attrs.rtm_family, hdr, npt->nw); 913 else 914 error = handle_rtm_getroute(nlp, &attrs, hdr, npt); 915 916 return (error); 917 } 918 919 void 920 rtnl_handle_route_event(uint32_t fibnum, const struct rib_cmd_info *rc) 921 { 922 struct nl_writer nw = {}; 923 int family, nlm_flags = 0; 924 925 family = rt_get_family(rc->rc_rt); 926 927 /* XXX: check if there are active listeners first */ 928 929 /* TODO: consider passing PID/type/seq */ 930 switch (rc->rc_cmd) { 931 case RTM_ADD: 932 nlm_flags = NLM_F_EXCL | NLM_F_CREATE; 933 break; 934 case RTM_CHANGE: 935 nlm_flags = NLM_F_REPLACE; 936 break; 937 case RTM_DELETE: 938 nlm_flags = 0; 939 break; 940 } 941 IF_DEBUG_LEVEL(LOG_DEBUG2) { 942 char rtbuf[NHOP_PRINT_BUFSIZE] __unused; 943 FIB_LOG(LOG_DEBUG2, fibnum, family, 944 "received event %s for %s / nlm_flags=%X", 945 rib_print_cmd(rc->rc_cmd), 946 rt_print_buf(rc->rc_rt, rtbuf, sizeof(rtbuf)), 947 nlm_flags); 948 } 949 950 struct nlmsghdr hdr = { 951 .nlmsg_flags = nlm_flags, 952 .nlmsg_type = get_rtmsg_type_from_rtsock(rc->rc_cmd), 953 }; 954 955 struct route_nhop_data rnd = { 956 .rnd_nhop = rc_get_nhop(rc), 957 .rnd_weight = rc->rc_nh_weight, 958 }; 959 960 uint32_t group_id = family_to_group(family); 961 if (!nlmsg_get_group_writer(&nw, NLMSG_SMALL, NETLINK_ROUTE, group_id)) { 962 NL_LOG(LOG_DEBUG, "error allocating event buffer"); 963 return; 964 } 965 966 dump_px(fibnum, &hdr, rc->rc_rt, &rnd, &nw); 967 nlmsg_flush(&nw); 968 } 969 970 static const struct rtnl_cmd_handler cmd_handlers[] = { 971 { 972 .cmd = NL_RTM_GETROUTE, 973 .name = "RTM_GETROUTE", 974 .cb = &rtnl_handle_getroute, 975 }, 976 { 977 .cmd = NL_RTM_DELROUTE, 978 .name = "RTM_DELROUTE", 979 .cb = &rtnl_handle_delroute, 980 .priv = PRIV_NET_ROUTE, 981 }, 982 { 983 .cmd = NL_RTM_NEWROUTE, 984 .name = "RTM_NEWROUTE", 985 .cb = &rtnl_handle_newroute, 986 .priv = PRIV_NET_ROUTE, 987 } 988 }; 989 990 static const struct nlhdr_parser *all_parsers[] = {&mpath_parser, &metrics_parser, &rtm_parser}; 991 992 void 993 rtnl_routes_init(void) 994 { 995 NL_VERIFY_PARSERS(all_parsers); 996 rtnl_register_messages(cmd_handlers, NL_ARRAY_LEN(cmd_handlers)); 997 } 998