1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * IPv4 Forwarding Information Base: FIB frontend. 7 * 8 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> 9 * 10 * This program is free software; you can redistribute it and/or 11 * modify it under the terms of the GNU General Public License 12 * as published by the Free Software Foundation; either version 13 * 2 of the License, or (at your option) any later version. 14 */ 15 16 #include <linux/module.h> 17 #include <linux/uaccess.h> 18 #include <linux/bitops.h> 19 #include <linux/capability.h> 20 #include <linux/types.h> 21 #include <linux/kernel.h> 22 #include <linux/mm.h> 23 #include <linux/string.h> 24 #include <linux/socket.h> 25 #include <linux/sockios.h> 26 #include <linux/errno.h> 27 #include <linux/in.h> 28 #include <linux/inet.h> 29 #include <linux/inetdevice.h> 30 #include <linux/netdevice.h> 31 #include <linux/if_addr.h> 32 #include <linux/if_arp.h> 33 #include <linux/skbuff.h> 34 #include <linux/cache.h> 35 #include <linux/init.h> 36 #include <linux/list.h> 37 #include <linux/slab.h> 38 39 #include <net/ip.h> 40 #include <net/protocol.h> 41 #include <net/route.h> 42 #include <net/tcp.h> 43 #include <net/sock.h> 44 #include <net/arp.h> 45 #include <net/ip_fib.h> 46 #include <net/rtnetlink.h> 47 #include <net/xfrm.h> 48 #include <net/l3mdev.h> 49 #include <net/lwtunnel.h> 50 #include <trace/events/fib.h> 51 52 #ifndef CONFIG_IP_MULTIPLE_TABLES 53 54 static int __net_init fib4_rules_init(struct net *net) 55 { 56 struct fib_table *local_table, *main_table; 57 58 main_table = fib_trie_table(RT_TABLE_MAIN, NULL); 59 if (!main_table) 60 return -ENOMEM; 61 62 local_table = fib_trie_table(RT_TABLE_LOCAL, main_table); 63 if (!local_table) 64 goto fail; 65 66 hlist_add_head_rcu(&local_table->tb_hlist, 67 &net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX]); 68 hlist_add_head_rcu(&main_table->tb_hlist, 69 &net->ipv4.fib_table_hash[TABLE_MAIN_INDEX]); 70 return 0; 71 72 fail: 73 fib_free_table(main_table); 74 return -ENOMEM; 75 } 76 77 static bool fib4_has_custom_rules(struct net *net) 78 { 79 return false; 80 } 81 #else 82 83 struct fib_table *fib_new_table(struct net *net, u32 id) 84 { 85 struct fib_table *tb, *alias = NULL; 86 unsigned int h; 87 88 if (id == 0) 89 id = RT_TABLE_MAIN; 90 tb = fib_get_table(net, id); 91 if (tb) 92 return tb; 93 94 if (id == RT_TABLE_LOCAL && !net->ipv4.fib_has_custom_rules) 95 alias = fib_new_table(net, RT_TABLE_MAIN); 96 97 tb = fib_trie_table(id, alias); 98 if (!tb) 99 return NULL; 100 101 switch (id) { 102 case RT_TABLE_MAIN: 103 rcu_assign_pointer(net->ipv4.fib_main, tb); 104 break; 105 case RT_TABLE_DEFAULT: 106 rcu_assign_pointer(net->ipv4.fib_default, tb); 107 break; 108 default: 109 break; 110 } 111 112 h = id & (FIB_TABLE_HASHSZ - 1); 113 hlist_add_head_rcu(&tb->tb_hlist, &net->ipv4.fib_table_hash[h]); 114 return tb; 115 } 116 EXPORT_SYMBOL_GPL(fib_new_table); 117 118 /* caller must hold either rtnl or rcu read lock */ 119 struct fib_table *fib_get_table(struct net *net, u32 id) 120 { 121 struct fib_table *tb; 122 struct hlist_head *head; 123 unsigned int h; 124 125 if (id == 0) 126 id = RT_TABLE_MAIN; 127 h = id & (FIB_TABLE_HASHSZ - 1); 128 129 head = &net->ipv4.fib_table_hash[h]; 130 hlist_for_each_entry_rcu(tb, head, tb_hlist) { 131 if (tb->tb_id == id) 132 return tb; 133 } 134 return NULL; 135 } 136 137 static bool fib4_has_custom_rules(struct net *net) 138 { 139 return net->ipv4.fib_has_custom_rules; 140 } 141 #endif /* CONFIG_IP_MULTIPLE_TABLES */ 142 143 static void fib_replace_table(struct net *net, struct fib_table *old, 144 struct fib_table *new) 145 { 146 #ifdef CONFIG_IP_MULTIPLE_TABLES 147 switch (new->tb_id) { 148 case RT_TABLE_MAIN: 149 rcu_assign_pointer(net->ipv4.fib_main, new); 150 break; 151 case RT_TABLE_DEFAULT: 152 rcu_assign_pointer(net->ipv4.fib_default, new); 153 break; 154 default: 155 break; 156 } 157 158 #endif 159 /* replace the old table in the hlist */ 160 hlist_replace_rcu(&old->tb_hlist, &new->tb_hlist); 161 } 162 163 int fib_unmerge(struct net *net) 164 { 165 struct fib_table *old, *new, *main_table; 166 167 /* attempt to fetch local table if it has been allocated */ 168 old = fib_get_table(net, RT_TABLE_LOCAL); 169 if (!old) 170 return 0; 171 172 new = fib_trie_unmerge(old); 173 if (!new) 174 return -ENOMEM; 175 176 /* table is already unmerged */ 177 if (new == old) 178 return 0; 179 180 /* replace merged table with clean table */ 181 fib_replace_table(net, old, new); 182 fib_free_table(old); 183 184 /* attempt to fetch main table if it has been allocated */ 185 main_table = fib_get_table(net, RT_TABLE_MAIN); 186 if (!main_table) 187 return 0; 188 189 /* flush local entries from main table */ 190 fib_table_flush_external(main_table); 191 192 return 0; 193 } 194 195 void fib_flush(struct net *net) 196 { 197 int flushed = 0; 198 unsigned int h; 199 200 for (h = 0; h < FIB_TABLE_HASHSZ; h++) { 201 struct hlist_head *head = &net->ipv4.fib_table_hash[h]; 202 struct hlist_node *tmp; 203 struct fib_table *tb; 204 205 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) 206 flushed += fib_table_flush(net, tb, false); 207 } 208 209 if (flushed) 210 rt_cache_flush(net); 211 } 212 213 /* 214 * Find address type as if only "dev" was present in the system. If 215 * on_dev is NULL then all interfaces are taken into consideration. 216 */ 217 static inline unsigned int __inet_dev_addr_type(struct net *net, 218 const struct net_device *dev, 219 __be32 addr, u32 tb_id) 220 { 221 struct flowi4 fl4 = { .daddr = addr }; 222 struct fib_result res; 223 unsigned int ret = RTN_BROADCAST; 224 struct fib_table *table; 225 226 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr)) 227 return RTN_BROADCAST; 228 if (ipv4_is_multicast(addr)) 229 return RTN_MULTICAST; 230 231 rcu_read_lock(); 232 233 table = fib_get_table(net, tb_id); 234 if (table) { 235 ret = RTN_UNICAST; 236 if (!fib_table_lookup(table, &fl4, &res, FIB_LOOKUP_NOREF)) { 237 if (!dev || dev == res.fi->fib_dev) 238 ret = res.type; 239 } 240 } 241 242 rcu_read_unlock(); 243 return ret; 244 } 245 246 unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id) 247 { 248 return __inet_dev_addr_type(net, NULL, addr, tb_id); 249 } 250 EXPORT_SYMBOL(inet_addr_type_table); 251 252 unsigned int inet_addr_type(struct net *net, __be32 addr) 253 { 254 return __inet_dev_addr_type(net, NULL, addr, RT_TABLE_LOCAL); 255 } 256 EXPORT_SYMBOL(inet_addr_type); 257 258 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev, 259 __be32 addr) 260 { 261 u32 rt_table = l3mdev_fib_table(dev) ? : RT_TABLE_LOCAL; 262 263 return __inet_dev_addr_type(net, dev, addr, rt_table); 264 } 265 EXPORT_SYMBOL(inet_dev_addr_type); 266 267 /* inet_addr_type with dev == NULL but using the table from a dev 268 * if one is associated 269 */ 270 unsigned int inet_addr_type_dev_table(struct net *net, 271 const struct net_device *dev, 272 __be32 addr) 273 { 274 u32 rt_table = l3mdev_fib_table(dev) ? : RT_TABLE_LOCAL; 275 276 return __inet_dev_addr_type(net, NULL, addr, rt_table); 277 } 278 EXPORT_SYMBOL(inet_addr_type_dev_table); 279 280 __be32 fib_compute_spec_dst(struct sk_buff *skb) 281 { 282 struct net_device *dev = skb->dev; 283 struct in_device *in_dev; 284 struct fib_result res; 285 struct rtable *rt; 286 struct net *net; 287 int scope; 288 289 rt = skb_rtable(skb); 290 if ((rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | RTCF_LOCAL)) == 291 RTCF_LOCAL) 292 return ip_hdr(skb)->daddr; 293 294 in_dev = __in_dev_get_rcu(dev); 295 296 net = dev_net(dev); 297 298 scope = RT_SCOPE_UNIVERSE; 299 if (!ipv4_is_zeronet(ip_hdr(skb)->saddr)) { 300 bool vmark = in_dev && IN_DEV_SRC_VMARK(in_dev); 301 struct flowi4 fl4 = { 302 .flowi4_iif = LOOPBACK_IFINDEX, 303 .flowi4_oif = l3mdev_master_ifindex_rcu(dev), 304 .daddr = ip_hdr(skb)->saddr, 305 .flowi4_tos = RT_TOS(ip_hdr(skb)->tos), 306 .flowi4_scope = scope, 307 .flowi4_mark = vmark ? skb->mark : 0, 308 }; 309 if (!fib_lookup(net, &fl4, &res, 0)) 310 return fib_result_prefsrc(net, &res); 311 } else { 312 scope = RT_SCOPE_LINK; 313 } 314 315 return inet_select_addr(dev, ip_hdr(skb)->saddr, scope); 316 } 317 318 bool fib_info_nh_uses_dev(struct fib_info *fi, const struct net_device *dev) 319 { 320 bool dev_match = false; 321 #ifdef CONFIG_IP_ROUTE_MULTIPATH 322 int ret; 323 324 for (ret = 0; ret < fi->fib_nhs; ret++) { 325 struct fib_nh *nh = &fi->fib_nh[ret]; 326 327 if (nh->fib_nh_dev == dev) { 328 dev_match = true; 329 break; 330 } else if (l3mdev_master_ifindex_rcu(nh->fib_nh_dev) == dev->ifindex) { 331 dev_match = true; 332 break; 333 } 334 } 335 #else 336 if (fi->fib_nh[0].fib_nh_dev == dev) 337 dev_match = true; 338 #endif 339 340 return dev_match; 341 } 342 EXPORT_SYMBOL_GPL(fib_info_nh_uses_dev); 343 344 /* Given (packet source, input interface) and optional (dst, oif, tos): 345 * - (main) check, that source is valid i.e. not broadcast or our local 346 * address. 347 * - figure out what "logical" interface this packet arrived 348 * and calculate "specific destination" address. 349 * - check, that packet arrived from expected physical interface. 350 * called with rcu_read_lock() 351 */ 352 static int __fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst, 353 u8 tos, int oif, struct net_device *dev, 354 int rpf, struct in_device *idev, u32 *itag) 355 { 356 struct net *net = dev_net(dev); 357 struct flow_keys flkeys; 358 int ret, no_addr; 359 struct fib_result res; 360 struct flowi4 fl4; 361 bool dev_match; 362 363 fl4.flowi4_oif = 0; 364 fl4.flowi4_iif = l3mdev_master_ifindex_rcu(dev); 365 if (!fl4.flowi4_iif) 366 fl4.flowi4_iif = oif ? : LOOPBACK_IFINDEX; 367 fl4.daddr = src; 368 fl4.saddr = dst; 369 fl4.flowi4_tos = tos; 370 fl4.flowi4_scope = RT_SCOPE_UNIVERSE; 371 fl4.flowi4_tun_key.tun_id = 0; 372 fl4.flowi4_flags = 0; 373 fl4.flowi4_uid = sock_net_uid(net, NULL); 374 375 no_addr = idev->ifa_list == NULL; 376 377 fl4.flowi4_mark = IN_DEV_SRC_VMARK(idev) ? skb->mark : 0; 378 if (!fib4_rules_early_flow_dissect(net, skb, &fl4, &flkeys)) { 379 fl4.flowi4_proto = 0; 380 fl4.fl4_sport = 0; 381 fl4.fl4_dport = 0; 382 } 383 384 if (fib_lookup(net, &fl4, &res, 0)) 385 goto last_resort; 386 if (res.type != RTN_UNICAST && 387 (res.type != RTN_LOCAL || !IN_DEV_ACCEPT_LOCAL(idev))) 388 goto e_inval; 389 fib_combine_itag(itag, &res); 390 391 dev_match = fib_info_nh_uses_dev(res.fi, dev); 392 if (dev_match) { 393 ret = FIB_RES_NHC(res)->nhc_scope >= RT_SCOPE_HOST; 394 return ret; 395 } 396 if (no_addr) 397 goto last_resort; 398 if (rpf == 1) 399 goto e_rpf; 400 fl4.flowi4_oif = dev->ifindex; 401 402 ret = 0; 403 if (fib_lookup(net, &fl4, &res, FIB_LOOKUP_IGNORE_LINKSTATE) == 0) { 404 if (res.type == RTN_UNICAST) 405 ret = FIB_RES_NHC(res)->nhc_scope >= RT_SCOPE_HOST; 406 } 407 return ret; 408 409 last_resort: 410 if (rpf) 411 goto e_rpf; 412 *itag = 0; 413 return 0; 414 415 e_inval: 416 return -EINVAL; 417 e_rpf: 418 return -EXDEV; 419 } 420 421 /* Ignore rp_filter for packets protected by IPsec. */ 422 int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst, 423 u8 tos, int oif, struct net_device *dev, 424 struct in_device *idev, u32 *itag) 425 { 426 int r = secpath_exists(skb) ? 0 : IN_DEV_RPFILTER(idev); 427 struct net *net = dev_net(dev); 428 429 if (!r && !fib_num_tclassid_users(net) && 430 (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev))) { 431 if (IN_DEV_ACCEPT_LOCAL(idev)) 432 goto ok; 433 /* with custom local routes in place, checking local addresses 434 * only will be too optimistic, with custom rules, checking 435 * local addresses only can be too strict, e.g. due to vrf 436 */ 437 if (net->ipv4.fib_has_custom_local_routes || 438 fib4_has_custom_rules(net)) 439 goto full_check; 440 if (inet_lookup_ifaddr_rcu(net, src)) 441 return -EINVAL; 442 443 ok: 444 *itag = 0; 445 return 0; 446 } 447 448 full_check: 449 return __fib_validate_source(skb, src, dst, tos, oif, dev, r, idev, itag); 450 } 451 452 static inline __be32 sk_extract_addr(struct sockaddr *addr) 453 { 454 return ((struct sockaddr_in *) addr)->sin_addr.s_addr; 455 } 456 457 static int put_rtax(struct nlattr *mx, int len, int type, u32 value) 458 { 459 struct nlattr *nla; 460 461 nla = (struct nlattr *) ((char *) mx + len); 462 nla->nla_type = type; 463 nla->nla_len = nla_attr_size(4); 464 *(u32 *) nla_data(nla) = value; 465 466 return len + nla_total_size(4); 467 } 468 469 static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt, 470 struct fib_config *cfg) 471 { 472 __be32 addr; 473 int plen; 474 475 memset(cfg, 0, sizeof(*cfg)); 476 cfg->fc_nlinfo.nl_net = net; 477 478 if (rt->rt_dst.sa_family != AF_INET) 479 return -EAFNOSUPPORT; 480 481 /* 482 * Check mask for validity: 483 * a) it must be contiguous. 484 * b) destination must have all host bits clear. 485 * c) if application forgot to set correct family (AF_INET), 486 * reject request unless it is absolutely clear i.e. 487 * both family and mask are zero. 488 */ 489 plen = 32; 490 addr = sk_extract_addr(&rt->rt_dst); 491 if (!(rt->rt_flags & RTF_HOST)) { 492 __be32 mask = sk_extract_addr(&rt->rt_genmask); 493 494 if (rt->rt_genmask.sa_family != AF_INET) { 495 if (mask || rt->rt_genmask.sa_family) 496 return -EAFNOSUPPORT; 497 } 498 499 if (bad_mask(mask, addr)) 500 return -EINVAL; 501 502 plen = inet_mask_len(mask); 503 } 504 505 cfg->fc_dst_len = plen; 506 cfg->fc_dst = addr; 507 508 if (cmd != SIOCDELRT) { 509 cfg->fc_nlflags = NLM_F_CREATE; 510 cfg->fc_protocol = RTPROT_BOOT; 511 } 512 513 if (rt->rt_metric) 514 cfg->fc_priority = rt->rt_metric - 1; 515 516 if (rt->rt_flags & RTF_REJECT) { 517 cfg->fc_scope = RT_SCOPE_HOST; 518 cfg->fc_type = RTN_UNREACHABLE; 519 return 0; 520 } 521 522 cfg->fc_scope = RT_SCOPE_NOWHERE; 523 cfg->fc_type = RTN_UNICAST; 524 525 if (rt->rt_dev) { 526 char *colon; 527 struct net_device *dev; 528 char devname[IFNAMSIZ]; 529 530 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1)) 531 return -EFAULT; 532 533 devname[IFNAMSIZ-1] = 0; 534 colon = strchr(devname, ':'); 535 if (colon) 536 *colon = 0; 537 dev = __dev_get_by_name(net, devname); 538 if (!dev) 539 return -ENODEV; 540 cfg->fc_oif = dev->ifindex; 541 cfg->fc_table = l3mdev_fib_table(dev); 542 if (colon) { 543 struct in_ifaddr *ifa; 544 struct in_device *in_dev = __in_dev_get_rtnl(dev); 545 if (!in_dev) 546 return -ENODEV; 547 *colon = ':'; 548 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) 549 if (strcmp(ifa->ifa_label, devname) == 0) 550 break; 551 if (!ifa) 552 return -ENODEV; 553 cfg->fc_prefsrc = ifa->ifa_local; 554 } 555 } 556 557 addr = sk_extract_addr(&rt->rt_gateway); 558 if (rt->rt_gateway.sa_family == AF_INET && addr) { 559 unsigned int addr_type; 560 561 cfg->fc_gw4 = addr; 562 cfg->fc_gw_family = AF_INET; 563 addr_type = inet_addr_type_table(net, addr, cfg->fc_table); 564 if (rt->rt_flags & RTF_GATEWAY && 565 addr_type == RTN_UNICAST) 566 cfg->fc_scope = RT_SCOPE_UNIVERSE; 567 } 568 569 if (cmd == SIOCDELRT) 570 return 0; 571 572 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw_family) 573 return -EINVAL; 574 575 if (cfg->fc_scope == RT_SCOPE_NOWHERE) 576 cfg->fc_scope = RT_SCOPE_LINK; 577 578 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) { 579 struct nlattr *mx; 580 int len = 0; 581 582 mx = kcalloc(3, nla_total_size(4), GFP_KERNEL); 583 if (!mx) 584 return -ENOMEM; 585 586 if (rt->rt_flags & RTF_MTU) 587 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40); 588 589 if (rt->rt_flags & RTF_WINDOW) 590 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window); 591 592 if (rt->rt_flags & RTF_IRTT) 593 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3); 594 595 cfg->fc_mx = mx; 596 cfg->fc_mx_len = len; 597 } 598 599 return 0; 600 } 601 602 /* 603 * Handle IP routing ioctl calls. 604 * These are used to manipulate the routing tables 605 */ 606 int ip_rt_ioctl(struct net *net, unsigned int cmd, struct rtentry *rt) 607 { 608 struct fib_config cfg; 609 int err; 610 611 switch (cmd) { 612 case SIOCADDRT: /* Add a route */ 613 case SIOCDELRT: /* Delete a route */ 614 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 615 return -EPERM; 616 617 rtnl_lock(); 618 err = rtentry_to_fib_config(net, cmd, rt, &cfg); 619 if (err == 0) { 620 struct fib_table *tb; 621 622 if (cmd == SIOCDELRT) { 623 tb = fib_get_table(net, cfg.fc_table); 624 if (tb) 625 err = fib_table_delete(net, tb, &cfg, 626 NULL); 627 else 628 err = -ESRCH; 629 } else { 630 tb = fib_new_table(net, cfg.fc_table); 631 if (tb) 632 err = fib_table_insert(net, tb, 633 &cfg, NULL); 634 else 635 err = -ENOBUFS; 636 } 637 638 /* allocated by rtentry_to_fib_config() */ 639 kfree(cfg.fc_mx); 640 } 641 rtnl_unlock(); 642 return err; 643 } 644 return -EINVAL; 645 } 646 647 const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = { 648 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 }, 649 [RTA_DST] = { .type = NLA_U32 }, 650 [RTA_SRC] = { .type = NLA_U32 }, 651 [RTA_IIF] = { .type = NLA_U32 }, 652 [RTA_OIF] = { .type = NLA_U32 }, 653 [RTA_GATEWAY] = { .type = NLA_U32 }, 654 [RTA_PRIORITY] = { .type = NLA_U32 }, 655 [RTA_PREFSRC] = { .type = NLA_U32 }, 656 [RTA_METRICS] = { .type = NLA_NESTED }, 657 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) }, 658 [RTA_FLOW] = { .type = NLA_U32 }, 659 [RTA_ENCAP_TYPE] = { .type = NLA_U16 }, 660 [RTA_ENCAP] = { .type = NLA_NESTED }, 661 [RTA_UID] = { .type = NLA_U32 }, 662 [RTA_MARK] = { .type = NLA_U32 }, 663 [RTA_TABLE] = { .type = NLA_U32 }, 664 [RTA_IP_PROTO] = { .type = NLA_U8 }, 665 [RTA_SPORT] = { .type = NLA_U16 }, 666 [RTA_DPORT] = { .type = NLA_U16 }, 667 }; 668 669 int fib_gw_from_via(struct fib_config *cfg, struct nlattr *nla, 670 struct netlink_ext_ack *extack) 671 { 672 struct rtvia *via; 673 int alen; 674 675 if (nla_len(nla) < offsetof(struct rtvia, rtvia_addr)) { 676 NL_SET_ERR_MSG(extack, "Invalid attribute length for RTA_VIA"); 677 return -EINVAL; 678 } 679 680 via = nla_data(nla); 681 alen = nla_len(nla) - offsetof(struct rtvia, rtvia_addr); 682 683 switch (via->rtvia_family) { 684 case AF_INET: 685 if (alen != sizeof(__be32)) { 686 NL_SET_ERR_MSG(extack, "Invalid IPv4 address in RTA_VIA"); 687 return -EINVAL; 688 } 689 cfg->fc_gw_family = AF_INET; 690 cfg->fc_gw4 = *((__be32 *)via->rtvia_addr); 691 break; 692 case AF_INET6: 693 #ifdef CONFIG_IPV6 694 if (alen != sizeof(struct in6_addr)) { 695 NL_SET_ERR_MSG(extack, "Invalid IPv6 address in RTA_VIA"); 696 return -EINVAL; 697 } 698 cfg->fc_gw_family = AF_INET6; 699 cfg->fc_gw6 = *((struct in6_addr *)via->rtvia_addr); 700 #else 701 NL_SET_ERR_MSG(extack, "IPv6 support not enabled in kernel"); 702 return -EINVAL; 703 #endif 704 break; 705 default: 706 NL_SET_ERR_MSG(extack, "Unsupported address family in RTA_VIA"); 707 return -EINVAL; 708 } 709 710 return 0; 711 } 712 713 static int rtm_to_fib_config(struct net *net, struct sk_buff *skb, 714 struct nlmsghdr *nlh, struct fib_config *cfg, 715 struct netlink_ext_ack *extack) 716 { 717 bool has_gw = false, has_via = false; 718 struct nlattr *attr; 719 int err, remaining; 720 struct rtmsg *rtm; 721 722 err = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX, 723 rtm_ipv4_policy, extack); 724 if (err < 0) 725 goto errout; 726 727 memset(cfg, 0, sizeof(*cfg)); 728 729 rtm = nlmsg_data(nlh); 730 cfg->fc_dst_len = rtm->rtm_dst_len; 731 cfg->fc_tos = rtm->rtm_tos; 732 cfg->fc_table = rtm->rtm_table; 733 cfg->fc_protocol = rtm->rtm_protocol; 734 cfg->fc_scope = rtm->rtm_scope; 735 cfg->fc_type = rtm->rtm_type; 736 cfg->fc_flags = rtm->rtm_flags; 737 cfg->fc_nlflags = nlh->nlmsg_flags; 738 739 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid; 740 cfg->fc_nlinfo.nlh = nlh; 741 cfg->fc_nlinfo.nl_net = net; 742 743 if (cfg->fc_type > RTN_MAX) { 744 NL_SET_ERR_MSG(extack, "Invalid route type"); 745 err = -EINVAL; 746 goto errout; 747 } 748 749 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) { 750 switch (nla_type(attr)) { 751 case RTA_DST: 752 cfg->fc_dst = nla_get_be32(attr); 753 break; 754 case RTA_OIF: 755 cfg->fc_oif = nla_get_u32(attr); 756 break; 757 case RTA_GATEWAY: 758 has_gw = true; 759 cfg->fc_gw4 = nla_get_be32(attr); 760 if (cfg->fc_gw4) 761 cfg->fc_gw_family = AF_INET; 762 break; 763 case RTA_VIA: 764 has_via = true; 765 err = fib_gw_from_via(cfg, attr, extack); 766 if (err) 767 goto errout; 768 break; 769 case RTA_PRIORITY: 770 cfg->fc_priority = nla_get_u32(attr); 771 break; 772 case RTA_PREFSRC: 773 cfg->fc_prefsrc = nla_get_be32(attr); 774 break; 775 case RTA_METRICS: 776 cfg->fc_mx = nla_data(attr); 777 cfg->fc_mx_len = nla_len(attr); 778 break; 779 case RTA_MULTIPATH: 780 err = lwtunnel_valid_encap_type_attr(nla_data(attr), 781 nla_len(attr), 782 extack); 783 if (err < 0) 784 goto errout; 785 cfg->fc_mp = nla_data(attr); 786 cfg->fc_mp_len = nla_len(attr); 787 break; 788 case RTA_FLOW: 789 cfg->fc_flow = nla_get_u32(attr); 790 break; 791 case RTA_TABLE: 792 cfg->fc_table = nla_get_u32(attr); 793 break; 794 case RTA_ENCAP: 795 cfg->fc_encap = attr; 796 break; 797 case RTA_ENCAP_TYPE: 798 cfg->fc_encap_type = nla_get_u16(attr); 799 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, 800 extack); 801 if (err < 0) 802 goto errout; 803 break; 804 } 805 } 806 807 if (has_gw && has_via) { 808 NL_SET_ERR_MSG(extack, 809 "Nexthop configuration can not contain both GATEWAY and VIA"); 810 goto errout; 811 } 812 813 return 0; 814 errout: 815 return err; 816 } 817 818 static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh, 819 struct netlink_ext_ack *extack) 820 { 821 struct net *net = sock_net(skb->sk); 822 struct fib_config cfg; 823 struct fib_table *tb; 824 int err; 825 826 err = rtm_to_fib_config(net, skb, nlh, &cfg, extack); 827 if (err < 0) 828 goto errout; 829 830 tb = fib_get_table(net, cfg.fc_table); 831 if (!tb) { 832 NL_SET_ERR_MSG(extack, "FIB table does not exist"); 833 err = -ESRCH; 834 goto errout; 835 } 836 837 err = fib_table_delete(net, tb, &cfg, extack); 838 errout: 839 return err; 840 } 841 842 static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh, 843 struct netlink_ext_ack *extack) 844 { 845 struct net *net = sock_net(skb->sk); 846 struct fib_config cfg; 847 struct fib_table *tb; 848 int err; 849 850 err = rtm_to_fib_config(net, skb, nlh, &cfg, extack); 851 if (err < 0) 852 goto errout; 853 854 tb = fib_new_table(net, cfg.fc_table); 855 if (!tb) { 856 err = -ENOBUFS; 857 goto errout; 858 } 859 860 err = fib_table_insert(net, tb, &cfg, extack); 861 if (!err && cfg.fc_type == RTN_LOCAL) 862 net->ipv4.fib_has_custom_local_routes = true; 863 errout: 864 return err; 865 } 866 867 int ip_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh, 868 struct fib_dump_filter *filter, 869 struct netlink_callback *cb) 870 { 871 struct netlink_ext_ack *extack = cb->extack; 872 struct nlattr *tb[RTA_MAX + 1]; 873 struct rtmsg *rtm; 874 int err, i; 875 876 ASSERT_RTNL(); 877 878 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) { 879 NL_SET_ERR_MSG(extack, "Invalid header for FIB dump request"); 880 return -EINVAL; 881 } 882 883 rtm = nlmsg_data(nlh); 884 if (rtm->rtm_dst_len || rtm->rtm_src_len || rtm->rtm_tos || 885 rtm->rtm_scope) { 886 NL_SET_ERR_MSG(extack, "Invalid values in header for FIB dump request"); 887 return -EINVAL; 888 } 889 if (rtm->rtm_flags & ~(RTM_F_CLONED | RTM_F_PREFIX)) { 890 NL_SET_ERR_MSG(extack, "Invalid flags for FIB dump request"); 891 return -EINVAL; 892 } 893 894 filter->dump_all_families = (rtm->rtm_family == AF_UNSPEC); 895 filter->flags = rtm->rtm_flags; 896 filter->protocol = rtm->rtm_protocol; 897 filter->rt_type = rtm->rtm_type; 898 filter->table_id = rtm->rtm_table; 899 900 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX, 901 rtm_ipv4_policy, extack); 902 if (err < 0) 903 return err; 904 905 for (i = 0; i <= RTA_MAX; ++i) { 906 int ifindex; 907 908 if (!tb[i]) 909 continue; 910 911 switch (i) { 912 case RTA_TABLE: 913 filter->table_id = nla_get_u32(tb[i]); 914 break; 915 case RTA_OIF: 916 ifindex = nla_get_u32(tb[i]); 917 filter->dev = __dev_get_by_index(net, ifindex); 918 if (!filter->dev) 919 return -ENODEV; 920 break; 921 default: 922 NL_SET_ERR_MSG(extack, "Unsupported attribute in dump request"); 923 return -EINVAL; 924 } 925 } 926 927 if (filter->flags || filter->protocol || filter->rt_type || 928 filter->table_id || filter->dev) { 929 filter->filter_set = 1; 930 cb->answer_flags = NLM_F_DUMP_FILTERED; 931 } 932 933 return 0; 934 } 935 EXPORT_SYMBOL_GPL(ip_valid_fib_dump_req); 936 937 static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb) 938 { 939 const struct nlmsghdr *nlh = cb->nlh; 940 struct net *net = sock_net(skb->sk); 941 struct fib_dump_filter filter = {}; 942 unsigned int h, s_h; 943 unsigned int e = 0, s_e; 944 struct fib_table *tb; 945 struct hlist_head *head; 946 int dumped = 0, err; 947 948 if (cb->strict_check) { 949 err = ip_valid_fib_dump_req(net, nlh, &filter, cb); 950 if (err < 0) 951 return err; 952 } else if (nlmsg_len(nlh) >= sizeof(struct rtmsg)) { 953 struct rtmsg *rtm = nlmsg_data(nlh); 954 955 filter.flags = rtm->rtm_flags & (RTM_F_PREFIX | RTM_F_CLONED); 956 } 957 958 /* fib entries are never clones and ipv4 does not use prefix flag */ 959 if (filter.flags & (RTM_F_PREFIX | RTM_F_CLONED)) 960 return skb->len; 961 962 if (filter.table_id) { 963 tb = fib_get_table(net, filter.table_id); 964 if (!tb) { 965 if (filter.dump_all_families) 966 return skb->len; 967 968 NL_SET_ERR_MSG(cb->extack, "ipv4: FIB table does not exist"); 969 return -ENOENT; 970 } 971 972 err = fib_table_dump(tb, skb, cb, &filter); 973 return skb->len ? : err; 974 } 975 976 s_h = cb->args[0]; 977 s_e = cb->args[1]; 978 979 rcu_read_lock(); 980 981 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) { 982 e = 0; 983 head = &net->ipv4.fib_table_hash[h]; 984 hlist_for_each_entry_rcu(tb, head, tb_hlist) { 985 if (e < s_e) 986 goto next; 987 if (dumped) 988 memset(&cb->args[2], 0, sizeof(cb->args) - 989 2 * sizeof(cb->args[0])); 990 err = fib_table_dump(tb, skb, cb, &filter); 991 if (err < 0) { 992 if (likely(skb->len)) 993 goto out; 994 995 goto out_err; 996 } 997 dumped = 1; 998 next: 999 e++; 1000 } 1001 } 1002 out: 1003 err = skb->len; 1004 out_err: 1005 rcu_read_unlock(); 1006 1007 cb->args[1] = e; 1008 cb->args[0] = h; 1009 1010 return err; 1011 } 1012 1013 /* Prepare and feed intra-kernel routing request. 1014 * Really, it should be netlink message, but :-( netlink 1015 * can be not configured, so that we feed it directly 1016 * to fib engine. It is legal, because all events occur 1017 * only when netlink is already locked. 1018 */ 1019 static void fib_magic(int cmd, int type, __be32 dst, int dst_len, 1020 struct in_ifaddr *ifa, u32 rt_priority) 1021 { 1022 struct net *net = dev_net(ifa->ifa_dev->dev); 1023 u32 tb_id = l3mdev_fib_table(ifa->ifa_dev->dev); 1024 struct fib_table *tb; 1025 struct fib_config cfg = { 1026 .fc_protocol = RTPROT_KERNEL, 1027 .fc_type = type, 1028 .fc_dst = dst, 1029 .fc_dst_len = dst_len, 1030 .fc_priority = rt_priority, 1031 .fc_prefsrc = ifa->ifa_local, 1032 .fc_oif = ifa->ifa_dev->dev->ifindex, 1033 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND, 1034 .fc_nlinfo = { 1035 .nl_net = net, 1036 }, 1037 }; 1038 1039 if (!tb_id) 1040 tb_id = (type == RTN_UNICAST) ? RT_TABLE_MAIN : RT_TABLE_LOCAL; 1041 1042 tb = fib_new_table(net, tb_id); 1043 if (!tb) 1044 return; 1045 1046 cfg.fc_table = tb->tb_id; 1047 1048 if (type != RTN_LOCAL) 1049 cfg.fc_scope = RT_SCOPE_LINK; 1050 else 1051 cfg.fc_scope = RT_SCOPE_HOST; 1052 1053 if (cmd == RTM_NEWROUTE) 1054 fib_table_insert(net, tb, &cfg, NULL); 1055 else 1056 fib_table_delete(net, tb, &cfg, NULL); 1057 } 1058 1059 void fib_add_ifaddr(struct in_ifaddr *ifa) 1060 { 1061 struct in_device *in_dev = ifa->ifa_dev; 1062 struct net_device *dev = in_dev->dev; 1063 struct in_ifaddr *prim = ifa; 1064 __be32 mask = ifa->ifa_mask; 1065 __be32 addr = ifa->ifa_local; 1066 __be32 prefix = ifa->ifa_address & mask; 1067 1068 if (ifa->ifa_flags & IFA_F_SECONDARY) { 1069 prim = inet_ifa_byprefix(in_dev, prefix, mask); 1070 if (!prim) { 1071 pr_warn("%s: bug: prim == NULL\n", __func__); 1072 return; 1073 } 1074 } 1075 1076 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim, 0); 1077 1078 if (!(dev->flags & IFF_UP)) 1079 return; 1080 1081 /* Add broadcast address, if it is explicitly assigned. */ 1082 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF)) 1083 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, 1084 prim, 0); 1085 1086 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) && 1087 (prefix != addr || ifa->ifa_prefixlen < 32)) { 1088 if (!(ifa->ifa_flags & IFA_F_NOPREFIXROUTE)) 1089 fib_magic(RTM_NEWROUTE, 1090 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST, 1091 prefix, ifa->ifa_prefixlen, prim, 1092 ifa->ifa_rt_priority); 1093 1094 /* Add network specific broadcasts, when it takes a sense */ 1095 if (ifa->ifa_prefixlen < 31) { 1096 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, 1097 prim, 0); 1098 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask, 1099 32, prim, 0); 1100 } 1101 } 1102 } 1103 1104 void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric) 1105 { 1106 __be32 prefix = ifa->ifa_address & ifa->ifa_mask; 1107 struct in_device *in_dev = ifa->ifa_dev; 1108 struct net_device *dev = in_dev->dev; 1109 1110 if (!(dev->flags & IFF_UP) || 1111 ifa->ifa_flags & (IFA_F_SECONDARY | IFA_F_NOPREFIXROUTE) || 1112 ipv4_is_zeronet(prefix) || 1113 prefix == ifa->ifa_local || ifa->ifa_prefixlen == 32) 1114 return; 1115 1116 /* add the new */ 1117 fib_magic(RTM_NEWROUTE, 1118 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST, 1119 prefix, ifa->ifa_prefixlen, ifa, new_metric); 1120 1121 /* delete the old */ 1122 fib_magic(RTM_DELROUTE, 1123 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST, 1124 prefix, ifa->ifa_prefixlen, ifa, ifa->ifa_rt_priority); 1125 } 1126 1127 /* Delete primary or secondary address. 1128 * Optionally, on secondary address promotion consider the addresses 1129 * from subnet iprim as deleted, even if they are in device list. 1130 * In this case the secondary ifa can be in device list. 1131 */ 1132 void fib_del_ifaddr(struct in_ifaddr *ifa, struct in_ifaddr *iprim) 1133 { 1134 struct in_device *in_dev = ifa->ifa_dev; 1135 struct net_device *dev = in_dev->dev; 1136 struct in_ifaddr *ifa1; 1137 struct in_ifaddr *prim = ifa, *prim1 = NULL; 1138 __be32 brd = ifa->ifa_address | ~ifa->ifa_mask; 1139 __be32 any = ifa->ifa_address & ifa->ifa_mask; 1140 #define LOCAL_OK 1 1141 #define BRD_OK 2 1142 #define BRD0_OK 4 1143 #define BRD1_OK 8 1144 unsigned int ok = 0; 1145 int subnet = 0; /* Primary network */ 1146 int gone = 1; /* Address is missing */ 1147 int same_prefsrc = 0; /* Another primary with same IP */ 1148 1149 if (ifa->ifa_flags & IFA_F_SECONDARY) { 1150 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask); 1151 if (!prim) { 1152 /* if the device has been deleted, we don't perform 1153 * address promotion 1154 */ 1155 if (!in_dev->dead) 1156 pr_warn("%s: bug: prim == NULL\n", __func__); 1157 return; 1158 } 1159 if (iprim && iprim != prim) { 1160 pr_warn("%s: bug: iprim != prim\n", __func__); 1161 return; 1162 } 1163 } else if (!ipv4_is_zeronet(any) && 1164 (any != ifa->ifa_local || ifa->ifa_prefixlen < 32)) { 1165 if (!(ifa->ifa_flags & IFA_F_NOPREFIXROUTE)) 1166 fib_magic(RTM_DELROUTE, 1167 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST, 1168 any, ifa->ifa_prefixlen, prim, 0); 1169 subnet = 1; 1170 } 1171 1172 if (in_dev->dead) 1173 goto no_promotions; 1174 1175 /* Deletion is more complicated than add. 1176 * We should take care of not to delete too much :-) 1177 * 1178 * Scan address list to be sure that addresses are really gone. 1179 */ 1180 1181 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) { 1182 if (ifa1 == ifa) { 1183 /* promotion, keep the IP */ 1184 gone = 0; 1185 continue; 1186 } 1187 /* Ignore IFAs from our subnet */ 1188 if (iprim && ifa1->ifa_mask == iprim->ifa_mask && 1189 inet_ifa_match(ifa1->ifa_address, iprim)) 1190 continue; 1191 1192 /* Ignore ifa1 if it uses different primary IP (prefsrc) */ 1193 if (ifa1->ifa_flags & IFA_F_SECONDARY) { 1194 /* Another address from our subnet? */ 1195 if (ifa1->ifa_mask == prim->ifa_mask && 1196 inet_ifa_match(ifa1->ifa_address, prim)) 1197 prim1 = prim; 1198 else { 1199 /* We reached the secondaries, so 1200 * same_prefsrc should be determined. 1201 */ 1202 if (!same_prefsrc) 1203 continue; 1204 /* Search new prim1 if ifa1 is not 1205 * using the current prim1 1206 */ 1207 if (!prim1 || 1208 ifa1->ifa_mask != prim1->ifa_mask || 1209 !inet_ifa_match(ifa1->ifa_address, prim1)) 1210 prim1 = inet_ifa_byprefix(in_dev, 1211 ifa1->ifa_address, 1212 ifa1->ifa_mask); 1213 if (!prim1) 1214 continue; 1215 if (prim1->ifa_local != prim->ifa_local) 1216 continue; 1217 } 1218 } else { 1219 if (prim->ifa_local != ifa1->ifa_local) 1220 continue; 1221 prim1 = ifa1; 1222 if (prim != prim1) 1223 same_prefsrc = 1; 1224 } 1225 if (ifa->ifa_local == ifa1->ifa_local) 1226 ok |= LOCAL_OK; 1227 if (ifa->ifa_broadcast == ifa1->ifa_broadcast) 1228 ok |= BRD_OK; 1229 if (brd == ifa1->ifa_broadcast) 1230 ok |= BRD1_OK; 1231 if (any == ifa1->ifa_broadcast) 1232 ok |= BRD0_OK; 1233 /* primary has network specific broadcasts */ 1234 if (prim1 == ifa1 && ifa1->ifa_prefixlen < 31) { 1235 __be32 brd1 = ifa1->ifa_address | ~ifa1->ifa_mask; 1236 __be32 any1 = ifa1->ifa_address & ifa1->ifa_mask; 1237 1238 if (!ipv4_is_zeronet(any1)) { 1239 if (ifa->ifa_broadcast == brd1 || 1240 ifa->ifa_broadcast == any1) 1241 ok |= BRD_OK; 1242 if (brd == brd1 || brd == any1) 1243 ok |= BRD1_OK; 1244 if (any == brd1 || any == any1) 1245 ok |= BRD0_OK; 1246 } 1247 } 1248 } 1249 1250 no_promotions: 1251 if (!(ok & BRD_OK)) 1252 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, 1253 prim, 0); 1254 if (subnet && ifa->ifa_prefixlen < 31) { 1255 if (!(ok & BRD1_OK)) 1256 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, 1257 prim, 0); 1258 if (!(ok & BRD0_OK)) 1259 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, 1260 prim, 0); 1261 } 1262 if (!(ok & LOCAL_OK)) { 1263 unsigned int addr_type; 1264 1265 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim, 0); 1266 1267 /* Check, that this local address finally disappeared. */ 1268 addr_type = inet_addr_type_dev_table(dev_net(dev), dev, 1269 ifa->ifa_local); 1270 if (gone && addr_type != RTN_LOCAL) { 1271 /* And the last, but not the least thing. 1272 * We must flush stray FIB entries. 1273 * 1274 * First of all, we scan fib_info list searching 1275 * for stray nexthop entries, then ignite fib_flush. 1276 */ 1277 if (fib_sync_down_addr(dev, ifa->ifa_local)) 1278 fib_flush(dev_net(dev)); 1279 } 1280 } 1281 #undef LOCAL_OK 1282 #undef BRD_OK 1283 #undef BRD0_OK 1284 #undef BRD1_OK 1285 } 1286 1287 static void nl_fib_lookup(struct net *net, struct fib_result_nl *frn) 1288 { 1289 1290 struct fib_result res; 1291 struct flowi4 fl4 = { 1292 .flowi4_mark = frn->fl_mark, 1293 .daddr = frn->fl_addr, 1294 .flowi4_tos = frn->fl_tos, 1295 .flowi4_scope = frn->fl_scope, 1296 }; 1297 struct fib_table *tb; 1298 1299 rcu_read_lock(); 1300 1301 tb = fib_get_table(net, frn->tb_id_in); 1302 1303 frn->err = -ENOENT; 1304 if (tb) { 1305 local_bh_disable(); 1306 1307 frn->tb_id = tb->tb_id; 1308 frn->err = fib_table_lookup(tb, &fl4, &res, FIB_LOOKUP_NOREF); 1309 1310 if (!frn->err) { 1311 frn->prefixlen = res.prefixlen; 1312 frn->nh_sel = res.nh_sel; 1313 frn->type = res.type; 1314 frn->scope = res.scope; 1315 } 1316 local_bh_enable(); 1317 } 1318 1319 rcu_read_unlock(); 1320 } 1321 1322 static void nl_fib_input(struct sk_buff *skb) 1323 { 1324 struct net *net; 1325 struct fib_result_nl *frn; 1326 struct nlmsghdr *nlh; 1327 u32 portid; 1328 1329 net = sock_net(skb->sk); 1330 nlh = nlmsg_hdr(skb); 1331 if (skb->len < nlmsg_total_size(sizeof(*frn)) || 1332 skb->len < nlh->nlmsg_len || 1333 nlmsg_len(nlh) < sizeof(*frn)) 1334 return; 1335 1336 skb = netlink_skb_clone(skb, GFP_KERNEL); 1337 if (!skb) 1338 return; 1339 nlh = nlmsg_hdr(skb); 1340 1341 frn = (struct fib_result_nl *) nlmsg_data(nlh); 1342 nl_fib_lookup(net, frn); 1343 1344 portid = NETLINK_CB(skb).portid; /* netlink portid */ 1345 NETLINK_CB(skb).portid = 0; /* from kernel */ 1346 NETLINK_CB(skb).dst_group = 0; /* unicast */ 1347 netlink_unicast(net->ipv4.fibnl, skb, portid, MSG_DONTWAIT); 1348 } 1349 1350 static int __net_init nl_fib_lookup_init(struct net *net) 1351 { 1352 struct sock *sk; 1353 struct netlink_kernel_cfg cfg = { 1354 .input = nl_fib_input, 1355 }; 1356 1357 sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, &cfg); 1358 if (!sk) 1359 return -EAFNOSUPPORT; 1360 net->ipv4.fibnl = sk; 1361 return 0; 1362 } 1363 1364 static void nl_fib_lookup_exit(struct net *net) 1365 { 1366 netlink_kernel_release(net->ipv4.fibnl); 1367 net->ipv4.fibnl = NULL; 1368 } 1369 1370 static void fib_disable_ip(struct net_device *dev, unsigned long event, 1371 bool force) 1372 { 1373 if (fib_sync_down_dev(dev, event, force)) 1374 fib_flush(dev_net(dev)); 1375 else 1376 rt_cache_flush(dev_net(dev)); 1377 arp_ifdown(dev); 1378 } 1379 1380 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr) 1381 { 1382 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr; 1383 struct net_device *dev = ifa->ifa_dev->dev; 1384 struct net *net = dev_net(dev); 1385 1386 switch (event) { 1387 case NETDEV_UP: 1388 fib_add_ifaddr(ifa); 1389 #ifdef CONFIG_IP_ROUTE_MULTIPATH 1390 fib_sync_up(dev, RTNH_F_DEAD); 1391 #endif 1392 atomic_inc(&net->ipv4.dev_addr_genid); 1393 rt_cache_flush(dev_net(dev)); 1394 break; 1395 case NETDEV_DOWN: 1396 fib_del_ifaddr(ifa, NULL); 1397 atomic_inc(&net->ipv4.dev_addr_genid); 1398 if (!ifa->ifa_dev->ifa_list) { 1399 /* Last address was deleted from this interface. 1400 * Disable IP. 1401 */ 1402 fib_disable_ip(dev, event, true); 1403 } else { 1404 rt_cache_flush(dev_net(dev)); 1405 } 1406 break; 1407 } 1408 return NOTIFY_DONE; 1409 } 1410 1411 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr) 1412 { 1413 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1414 struct netdev_notifier_changeupper_info *upper_info = ptr; 1415 struct netdev_notifier_info_ext *info_ext = ptr; 1416 struct in_device *in_dev; 1417 struct net *net = dev_net(dev); 1418 unsigned int flags; 1419 1420 if (event == NETDEV_UNREGISTER) { 1421 fib_disable_ip(dev, event, true); 1422 rt_flush_dev(dev); 1423 return NOTIFY_DONE; 1424 } 1425 1426 in_dev = __in_dev_get_rtnl(dev); 1427 if (!in_dev) 1428 return NOTIFY_DONE; 1429 1430 switch (event) { 1431 case NETDEV_UP: 1432 for_ifa(in_dev) { 1433 fib_add_ifaddr(ifa); 1434 } endfor_ifa(in_dev); 1435 #ifdef CONFIG_IP_ROUTE_MULTIPATH 1436 fib_sync_up(dev, RTNH_F_DEAD); 1437 #endif 1438 atomic_inc(&net->ipv4.dev_addr_genid); 1439 rt_cache_flush(net); 1440 break; 1441 case NETDEV_DOWN: 1442 fib_disable_ip(dev, event, false); 1443 break; 1444 case NETDEV_CHANGE: 1445 flags = dev_get_flags(dev); 1446 if (flags & (IFF_RUNNING | IFF_LOWER_UP)) 1447 fib_sync_up(dev, RTNH_F_LINKDOWN); 1448 else 1449 fib_sync_down_dev(dev, event, false); 1450 rt_cache_flush(net); 1451 break; 1452 case NETDEV_CHANGEMTU: 1453 fib_sync_mtu(dev, info_ext->ext.mtu); 1454 rt_cache_flush(net); 1455 break; 1456 case NETDEV_CHANGEUPPER: 1457 upper_info = ptr; 1458 /* flush all routes if dev is linked to or unlinked from 1459 * an L3 master device (e.g., VRF) 1460 */ 1461 if (upper_info->upper_dev && 1462 netif_is_l3_master(upper_info->upper_dev)) 1463 fib_disable_ip(dev, NETDEV_DOWN, true); 1464 break; 1465 } 1466 return NOTIFY_DONE; 1467 } 1468 1469 static struct notifier_block fib_inetaddr_notifier = { 1470 .notifier_call = fib_inetaddr_event, 1471 }; 1472 1473 static struct notifier_block fib_netdev_notifier = { 1474 .notifier_call = fib_netdev_event, 1475 }; 1476 1477 static int __net_init ip_fib_net_init(struct net *net) 1478 { 1479 int err; 1480 size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ; 1481 1482 err = fib4_notifier_init(net); 1483 if (err) 1484 return err; 1485 1486 /* Avoid false sharing : Use at least a full cache line */ 1487 size = max_t(size_t, size, L1_CACHE_BYTES); 1488 1489 net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL); 1490 if (!net->ipv4.fib_table_hash) { 1491 err = -ENOMEM; 1492 goto err_table_hash_alloc; 1493 } 1494 1495 err = fib4_rules_init(net); 1496 if (err < 0) 1497 goto err_rules_init; 1498 return 0; 1499 1500 err_rules_init: 1501 kfree(net->ipv4.fib_table_hash); 1502 err_table_hash_alloc: 1503 fib4_notifier_exit(net); 1504 return err; 1505 } 1506 1507 static void ip_fib_net_exit(struct net *net) 1508 { 1509 int i; 1510 1511 rtnl_lock(); 1512 #ifdef CONFIG_IP_MULTIPLE_TABLES 1513 RCU_INIT_POINTER(net->ipv4.fib_main, NULL); 1514 RCU_INIT_POINTER(net->ipv4.fib_default, NULL); 1515 #endif 1516 /* Destroy the tables in reverse order to guarantee that the 1517 * local table, ID 255, is destroyed before the main table, ID 1518 * 254. This is necessary as the local table may contain 1519 * references to data contained in the main table. 1520 */ 1521 for (i = FIB_TABLE_HASHSZ - 1; i >= 0; i--) { 1522 struct hlist_head *head = &net->ipv4.fib_table_hash[i]; 1523 struct hlist_node *tmp; 1524 struct fib_table *tb; 1525 1526 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) { 1527 hlist_del(&tb->tb_hlist); 1528 fib_table_flush(net, tb, true); 1529 fib_free_table(tb); 1530 } 1531 } 1532 1533 #ifdef CONFIG_IP_MULTIPLE_TABLES 1534 fib4_rules_exit(net); 1535 #endif 1536 rtnl_unlock(); 1537 kfree(net->ipv4.fib_table_hash); 1538 fib4_notifier_exit(net); 1539 } 1540 1541 static int __net_init fib_net_init(struct net *net) 1542 { 1543 int error; 1544 1545 #ifdef CONFIG_IP_ROUTE_CLASSID 1546 net->ipv4.fib_num_tclassid_users = 0; 1547 #endif 1548 error = ip_fib_net_init(net); 1549 if (error < 0) 1550 goto out; 1551 error = nl_fib_lookup_init(net); 1552 if (error < 0) 1553 goto out_nlfl; 1554 error = fib_proc_init(net); 1555 if (error < 0) 1556 goto out_proc; 1557 out: 1558 return error; 1559 1560 out_proc: 1561 nl_fib_lookup_exit(net); 1562 out_nlfl: 1563 ip_fib_net_exit(net); 1564 goto out; 1565 } 1566 1567 static void __net_exit fib_net_exit(struct net *net) 1568 { 1569 fib_proc_exit(net); 1570 nl_fib_lookup_exit(net); 1571 ip_fib_net_exit(net); 1572 } 1573 1574 static struct pernet_operations fib_net_ops = { 1575 .init = fib_net_init, 1576 .exit = fib_net_exit, 1577 }; 1578 1579 void __init ip_fib_init(void) 1580 { 1581 fib_trie_init(); 1582 1583 register_pernet_subsys(&fib_net_ops); 1584 1585 register_netdevice_notifier(&fib_netdev_notifier); 1586 register_inetaddr_notifier(&fib_inetaddr_notifier); 1587 1588 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL, 0); 1589 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL, 0); 1590 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib, 0); 1591 } 1592