fib_frontend.c revision b854272b3c732316676e9128f7b9e6f1e1ff88b0
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 * Version: $Id: fib_frontend.c,v 1.26 2001/10/31 21:55:54 davem Exp $ 9 * 10 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> 11 * 12 * This program is free software; you can redistribute it and/or 13 * modify it under the terms of the GNU General Public License 14 * as published by the Free Software Foundation; either version 15 * 2 of the License, or (at your option) any later version. 16 */ 17 18#include <linux/module.h> 19#include <asm/uaccess.h> 20#include <asm/system.h> 21#include <linux/bitops.h> 22#include <linux/capability.h> 23#include <linux/types.h> 24#include <linux/kernel.h> 25#include <linux/mm.h> 26#include <linux/string.h> 27#include <linux/socket.h> 28#include <linux/sockios.h> 29#include <linux/errno.h> 30#include <linux/in.h> 31#include <linux/inet.h> 32#include <linux/inetdevice.h> 33#include <linux/netdevice.h> 34#include <linux/if_addr.h> 35#include <linux/if_arp.h> 36#include <linux/skbuff.h> 37#include <linux/init.h> 38#include <linux/list.h> 39 40#include <net/ip.h> 41#include <net/protocol.h> 42#include <net/route.h> 43#include <net/tcp.h> 44#include <net/sock.h> 45#include <net/icmp.h> 46#include <net/arp.h> 47#include <net/ip_fib.h> 48#include <net/rtnetlink.h> 49 50#define FFprint(a...) printk(KERN_DEBUG a) 51 52static struct sock *fibnl; 53 54#ifndef CONFIG_IP_MULTIPLE_TABLES 55 56struct fib_table *ip_fib_local_table; 57struct fib_table *ip_fib_main_table; 58 59#define FIB_TABLE_HASHSZ 1 60static struct hlist_head fib_table_hash[FIB_TABLE_HASHSZ]; 61 62static void __init fib4_rules_init(void) 63{ 64 ip_fib_local_table = fib_hash_init(RT_TABLE_LOCAL); 65 hlist_add_head_rcu(&ip_fib_local_table->tb_hlist, &fib_table_hash[0]); 66 ip_fib_main_table = fib_hash_init(RT_TABLE_MAIN); 67 hlist_add_head_rcu(&ip_fib_main_table->tb_hlist, &fib_table_hash[0]); 68} 69#else 70 71#define FIB_TABLE_HASHSZ 256 72static struct hlist_head fib_table_hash[FIB_TABLE_HASHSZ]; 73 74struct fib_table *fib_new_table(u32 id) 75{ 76 struct fib_table *tb; 77 unsigned int h; 78 79 if (id == 0) 80 id = RT_TABLE_MAIN; 81 tb = fib_get_table(id); 82 if (tb) 83 return tb; 84 tb = fib_hash_init(id); 85 if (!tb) 86 return NULL; 87 h = id & (FIB_TABLE_HASHSZ - 1); 88 hlist_add_head_rcu(&tb->tb_hlist, &fib_table_hash[h]); 89 return tb; 90} 91 92struct fib_table *fib_get_table(u32 id) 93{ 94 struct fib_table *tb; 95 struct hlist_node *node; 96 unsigned int h; 97 98 if (id == 0) 99 id = RT_TABLE_MAIN; 100 h = id & (FIB_TABLE_HASHSZ - 1); 101 rcu_read_lock(); 102 hlist_for_each_entry_rcu(tb, node, &fib_table_hash[h], tb_hlist) { 103 if (tb->tb_id == id) { 104 rcu_read_unlock(); 105 return tb; 106 } 107 } 108 rcu_read_unlock(); 109 return NULL; 110} 111#endif /* CONFIG_IP_MULTIPLE_TABLES */ 112 113static void fib_flush(void) 114{ 115 int flushed = 0; 116 struct fib_table *tb; 117 struct hlist_node *node; 118 unsigned int h; 119 120 for (h = 0; h < FIB_TABLE_HASHSZ; h++) { 121 hlist_for_each_entry(tb, node, &fib_table_hash[h], tb_hlist) 122 flushed += tb->tb_flush(tb); 123 } 124 125 if (flushed) 126 rt_cache_flush(-1); 127} 128 129/* 130 * Find the first device with a given source address. 131 */ 132 133struct net_device * ip_dev_find(__be32 addr) 134{ 135 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = addr } } }; 136 struct fib_result res; 137 struct net_device *dev = NULL; 138 struct fib_table *local_table; 139 140#ifdef CONFIG_IP_MULTIPLE_TABLES 141 res.r = NULL; 142#endif 143 144 local_table = fib_get_table(RT_TABLE_LOCAL); 145 if (!local_table || local_table->tb_lookup(local_table, &fl, &res)) 146 return NULL; 147 if (res.type != RTN_LOCAL) 148 goto out; 149 dev = FIB_RES_DEV(res); 150 151 if (dev) 152 dev_hold(dev); 153out: 154 fib_res_put(&res); 155 return dev; 156} 157 158unsigned inet_addr_type(__be32 addr) 159{ 160 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = addr } } }; 161 struct fib_result res; 162 unsigned ret = RTN_BROADCAST; 163 struct fib_table *local_table; 164 165 if (ZERONET(addr) || BADCLASS(addr)) 166 return RTN_BROADCAST; 167 if (MULTICAST(addr)) 168 return RTN_MULTICAST; 169 170#ifdef CONFIG_IP_MULTIPLE_TABLES 171 res.r = NULL; 172#endif 173 174 local_table = fib_get_table(RT_TABLE_LOCAL); 175 if (local_table) { 176 ret = RTN_UNICAST; 177 if (!local_table->tb_lookup(local_table, &fl, &res)) { 178 ret = res.type; 179 fib_res_put(&res); 180 } 181 } 182 return ret; 183} 184 185/* Given (packet source, input interface) and optional (dst, oif, tos): 186 - (main) check, that source is valid i.e. not broadcast or our local 187 address. 188 - figure out what "logical" interface this packet arrived 189 and calculate "specific destination" address. 190 - check, that packet arrived from expected physical interface. 191 */ 192 193int fib_validate_source(__be32 src, __be32 dst, u8 tos, int oif, 194 struct net_device *dev, __be32 *spec_dst, u32 *itag) 195{ 196 struct in_device *in_dev; 197 struct flowi fl = { .nl_u = { .ip4_u = 198 { .daddr = src, 199 .saddr = dst, 200 .tos = tos } }, 201 .iif = oif }; 202 struct fib_result res; 203 int no_addr, rpf; 204 int ret; 205 206 no_addr = rpf = 0; 207 rcu_read_lock(); 208 in_dev = __in_dev_get_rcu(dev); 209 if (in_dev) { 210 no_addr = in_dev->ifa_list == NULL; 211 rpf = IN_DEV_RPFILTER(in_dev); 212 } 213 rcu_read_unlock(); 214 215 if (in_dev == NULL) 216 goto e_inval; 217 218 if (fib_lookup(&fl, &res)) 219 goto last_resort; 220 if (res.type != RTN_UNICAST) 221 goto e_inval_res; 222 *spec_dst = FIB_RES_PREFSRC(res); 223 fib_combine_itag(itag, &res); 224#ifdef CONFIG_IP_ROUTE_MULTIPATH 225 if (FIB_RES_DEV(res) == dev || res.fi->fib_nhs > 1) 226#else 227 if (FIB_RES_DEV(res) == dev) 228#endif 229 { 230 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST; 231 fib_res_put(&res); 232 return ret; 233 } 234 fib_res_put(&res); 235 if (no_addr) 236 goto last_resort; 237 if (rpf) 238 goto e_inval; 239 fl.oif = dev->ifindex; 240 241 ret = 0; 242 if (fib_lookup(&fl, &res) == 0) { 243 if (res.type == RTN_UNICAST) { 244 *spec_dst = FIB_RES_PREFSRC(res); 245 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST; 246 } 247 fib_res_put(&res); 248 } 249 return ret; 250 251last_resort: 252 if (rpf) 253 goto e_inval; 254 *spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE); 255 *itag = 0; 256 return 0; 257 258e_inval_res: 259 fib_res_put(&res); 260e_inval: 261 return -EINVAL; 262} 263 264static inline __be32 sk_extract_addr(struct sockaddr *addr) 265{ 266 return ((struct sockaddr_in *) addr)->sin_addr.s_addr; 267} 268 269static int put_rtax(struct nlattr *mx, int len, int type, u32 value) 270{ 271 struct nlattr *nla; 272 273 nla = (struct nlattr *) ((char *) mx + len); 274 nla->nla_type = type; 275 nla->nla_len = nla_attr_size(4); 276 *(u32 *) nla_data(nla) = value; 277 278 return len + nla_total_size(4); 279} 280 281static int rtentry_to_fib_config(int cmd, struct rtentry *rt, 282 struct fib_config *cfg) 283{ 284 __be32 addr; 285 int plen; 286 287 memset(cfg, 0, sizeof(*cfg)); 288 289 if (rt->rt_dst.sa_family != AF_INET) 290 return -EAFNOSUPPORT; 291 292 /* 293 * Check mask for validity: 294 * a) it must be contiguous. 295 * b) destination must have all host bits clear. 296 * c) if application forgot to set correct family (AF_INET), 297 * reject request unless it is absolutely clear i.e. 298 * both family and mask are zero. 299 */ 300 plen = 32; 301 addr = sk_extract_addr(&rt->rt_dst); 302 if (!(rt->rt_flags & RTF_HOST)) { 303 __be32 mask = sk_extract_addr(&rt->rt_genmask); 304 305 if (rt->rt_genmask.sa_family != AF_INET) { 306 if (mask || rt->rt_genmask.sa_family) 307 return -EAFNOSUPPORT; 308 } 309 310 if (bad_mask(mask, addr)) 311 return -EINVAL; 312 313 plen = inet_mask_len(mask); 314 } 315 316 cfg->fc_dst_len = plen; 317 cfg->fc_dst = addr; 318 319 if (cmd != SIOCDELRT) { 320 cfg->fc_nlflags = NLM_F_CREATE; 321 cfg->fc_protocol = RTPROT_BOOT; 322 } 323 324 if (rt->rt_metric) 325 cfg->fc_priority = rt->rt_metric - 1; 326 327 if (rt->rt_flags & RTF_REJECT) { 328 cfg->fc_scope = RT_SCOPE_HOST; 329 cfg->fc_type = RTN_UNREACHABLE; 330 return 0; 331 } 332 333 cfg->fc_scope = RT_SCOPE_NOWHERE; 334 cfg->fc_type = RTN_UNICAST; 335 336 if (rt->rt_dev) { 337 char *colon; 338 struct net_device *dev; 339 char devname[IFNAMSIZ]; 340 341 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1)) 342 return -EFAULT; 343 344 devname[IFNAMSIZ-1] = 0; 345 colon = strchr(devname, ':'); 346 if (colon) 347 *colon = 0; 348 dev = __dev_get_by_name(&init_net, devname); 349 if (!dev) 350 return -ENODEV; 351 cfg->fc_oif = dev->ifindex; 352 if (colon) { 353 struct in_ifaddr *ifa; 354 struct in_device *in_dev = __in_dev_get_rtnl(dev); 355 if (!in_dev) 356 return -ENODEV; 357 *colon = ':'; 358 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) 359 if (strcmp(ifa->ifa_label, devname) == 0) 360 break; 361 if (ifa == NULL) 362 return -ENODEV; 363 cfg->fc_prefsrc = ifa->ifa_local; 364 } 365 } 366 367 addr = sk_extract_addr(&rt->rt_gateway); 368 if (rt->rt_gateway.sa_family == AF_INET && addr) { 369 cfg->fc_gw = addr; 370 if (rt->rt_flags & RTF_GATEWAY && 371 inet_addr_type(addr) == RTN_UNICAST) 372 cfg->fc_scope = RT_SCOPE_UNIVERSE; 373 } 374 375 if (cmd == SIOCDELRT) 376 return 0; 377 378 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw) 379 return -EINVAL; 380 381 if (cfg->fc_scope == RT_SCOPE_NOWHERE) 382 cfg->fc_scope = RT_SCOPE_LINK; 383 384 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) { 385 struct nlattr *mx; 386 int len = 0; 387 388 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL); 389 if (mx == NULL) 390 return -ENOMEM; 391 392 if (rt->rt_flags & RTF_MTU) 393 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40); 394 395 if (rt->rt_flags & RTF_WINDOW) 396 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window); 397 398 if (rt->rt_flags & RTF_IRTT) 399 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3); 400 401 cfg->fc_mx = mx; 402 cfg->fc_mx_len = len; 403 } 404 405 return 0; 406} 407 408/* 409 * Handle IP routing ioctl calls. These are used to manipulate the routing tables 410 */ 411 412int ip_rt_ioctl(unsigned int cmd, void __user *arg) 413{ 414 struct fib_config cfg; 415 struct rtentry rt; 416 int err; 417 418 switch (cmd) { 419 case SIOCADDRT: /* Add a route */ 420 case SIOCDELRT: /* Delete a route */ 421 if (!capable(CAP_NET_ADMIN)) 422 return -EPERM; 423 424 if (copy_from_user(&rt, arg, sizeof(rt))) 425 return -EFAULT; 426 427 rtnl_lock(); 428 err = rtentry_to_fib_config(cmd, &rt, &cfg); 429 if (err == 0) { 430 struct fib_table *tb; 431 432 if (cmd == SIOCDELRT) { 433 tb = fib_get_table(cfg.fc_table); 434 if (tb) 435 err = tb->tb_delete(tb, &cfg); 436 else 437 err = -ESRCH; 438 } else { 439 tb = fib_new_table(cfg.fc_table); 440 if (tb) 441 err = tb->tb_insert(tb, &cfg); 442 else 443 err = -ENOBUFS; 444 } 445 446 /* allocated by rtentry_to_fib_config() */ 447 kfree(cfg.fc_mx); 448 } 449 rtnl_unlock(); 450 return err; 451 } 452 return -EINVAL; 453} 454 455const struct nla_policy rtm_ipv4_policy[RTA_MAX+1] = { 456 [RTA_DST] = { .type = NLA_U32 }, 457 [RTA_SRC] = { .type = NLA_U32 }, 458 [RTA_IIF] = { .type = NLA_U32 }, 459 [RTA_OIF] = { .type = NLA_U32 }, 460 [RTA_GATEWAY] = { .type = NLA_U32 }, 461 [RTA_PRIORITY] = { .type = NLA_U32 }, 462 [RTA_PREFSRC] = { .type = NLA_U32 }, 463 [RTA_METRICS] = { .type = NLA_NESTED }, 464 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) }, 465 [RTA_PROTOINFO] = { .type = NLA_U32 }, 466 [RTA_FLOW] = { .type = NLA_U32 }, 467}; 468 469static int rtm_to_fib_config(struct sk_buff *skb, struct nlmsghdr *nlh, 470 struct fib_config *cfg) 471{ 472 struct nlattr *attr; 473 int err, remaining; 474 struct rtmsg *rtm; 475 476 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy); 477 if (err < 0) 478 goto errout; 479 480 memset(cfg, 0, sizeof(*cfg)); 481 482 rtm = nlmsg_data(nlh); 483 cfg->fc_dst_len = rtm->rtm_dst_len; 484 cfg->fc_tos = rtm->rtm_tos; 485 cfg->fc_table = rtm->rtm_table; 486 cfg->fc_protocol = rtm->rtm_protocol; 487 cfg->fc_scope = rtm->rtm_scope; 488 cfg->fc_type = rtm->rtm_type; 489 cfg->fc_flags = rtm->rtm_flags; 490 cfg->fc_nlflags = nlh->nlmsg_flags; 491 492 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid; 493 cfg->fc_nlinfo.nlh = nlh; 494 495 if (cfg->fc_type > RTN_MAX) { 496 err = -EINVAL; 497 goto errout; 498 } 499 500 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) { 501 switch (nla_type(attr)) { 502 case RTA_DST: 503 cfg->fc_dst = nla_get_be32(attr); 504 break; 505 case RTA_OIF: 506 cfg->fc_oif = nla_get_u32(attr); 507 break; 508 case RTA_GATEWAY: 509 cfg->fc_gw = nla_get_be32(attr); 510 break; 511 case RTA_PRIORITY: 512 cfg->fc_priority = nla_get_u32(attr); 513 break; 514 case RTA_PREFSRC: 515 cfg->fc_prefsrc = nla_get_be32(attr); 516 break; 517 case RTA_METRICS: 518 cfg->fc_mx = nla_data(attr); 519 cfg->fc_mx_len = nla_len(attr); 520 break; 521 case RTA_MULTIPATH: 522 cfg->fc_mp = nla_data(attr); 523 cfg->fc_mp_len = nla_len(attr); 524 break; 525 case RTA_FLOW: 526 cfg->fc_flow = nla_get_u32(attr); 527 break; 528 case RTA_TABLE: 529 cfg->fc_table = nla_get_u32(attr); 530 break; 531 } 532 } 533 534 return 0; 535errout: 536 return err; 537} 538 539static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg) 540{ 541 struct net *net = skb->sk->sk_net; 542 struct fib_config cfg; 543 struct fib_table *tb; 544 int err; 545 546 if (net != &init_net) 547 return -EINVAL; 548 549 err = rtm_to_fib_config(skb, nlh, &cfg); 550 if (err < 0) 551 goto errout; 552 553 tb = fib_get_table(cfg.fc_table); 554 if (tb == NULL) { 555 err = -ESRCH; 556 goto errout; 557 } 558 559 err = tb->tb_delete(tb, &cfg); 560errout: 561 return err; 562} 563 564static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg) 565{ 566 struct net *net = skb->sk->sk_net; 567 struct fib_config cfg; 568 struct fib_table *tb; 569 int err; 570 571 if (net != &init_net) 572 return -EINVAL; 573 574 err = rtm_to_fib_config(skb, nlh, &cfg); 575 if (err < 0) 576 goto errout; 577 578 tb = fib_new_table(cfg.fc_table); 579 if (tb == NULL) { 580 err = -ENOBUFS; 581 goto errout; 582 } 583 584 err = tb->tb_insert(tb, &cfg); 585errout: 586 return err; 587} 588 589static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb) 590{ 591 struct net *net = skb->sk->sk_net; 592 unsigned int h, s_h; 593 unsigned int e = 0, s_e; 594 struct fib_table *tb; 595 struct hlist_node *node; 596 int dumped = 0; 597 598 if (net != &init_net) 599 return 0; 600 601 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) && 602 ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED) 603 return ip_rt_dump(skb, cb); 604 605 s_h = cb->args[0]; 606 s_e = cb->args[1]; 607 608 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) { 609 e = 0; 610 hlist_for_each_entry(tb, node, &fib_table_hash[h], tb_hlist) { 611 if (e < s_e) 612 goto next; 613 if (dumped) 614 memset(&cb->args[2], 0, sizeof(cb->args) - 615 2 * sizeof(cb->args[0])); 616 if (tb->tb_dump(tb, skb, cb) < 0) 617 goto out; 618 dumped = 1; 619next: 620 e++; 621 } 622 } 623out: 624 cb->args[1] = e; 625 cb->args[0] = h; 626 627 return skb->len; 628} 629 630/* Prepare and feed intra-kernel routing request. 631 Really, it should be netlink message, but :-( netlink 632 can be not configured, so that we feed it directly 633 to fib engine. It is legal, because all events occur 634 only when netlink is already locked. 635 */ 636 637static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa) 638{ 639 struct fib_table *tb; 640 struct fib_config cfg = { 641 .fc_protocol = RTPROT_KERNEL, 642 .fc_type = type, 643 .fc_dst = dst, 644 .fc_dst_len = dst_len, 645 .fc_prefsrc = ifa->ifa_local, 646 .fc_oif = ifa->ifa_dev->dev->ifindex, 647 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND, 648 }; 649 650 if (type == RTN_UNICAST) 651 tb = fib_new_table(RT_TABLE_MAIN); 652 else 653 tb = fib_new_table(RT_TABLE_LOCAL); 654 655 if (tb == NULL) 656 return; 657 658 cfg.fc_table = tb->tb_id; 659 660 if (type != RTN_LOCAL) 661 cfg.fc_scope = RT_SCOPE_LINK; 662 else 663 cfg.fc_scope = RT_SCOPE_HOST; 664 665 if (cmd == RTM_NEWROUTE) 666 tb->tb_insert(tb, &cfg); 667 else 668 tb->tb_delete(tb, &cfg); 669} 670 671void fib_add_ifaddr(struct in_ifaddr *ifa) 672{ 673 struct in_device *in_dev = ifa->ifa_dev; 674 struct net_device *dev = in_dev->dev; 675 struct in_ifaddr *prim = ifa; 676 __be32 mask = ifa->ifa_mask; 677 __be32 addr = ifa->ifa_local; 678 __be32 prefix = ifa->ifa_address&mask; 679 680 if (ifa->ifa_flags&IFA_F_SECONDARY) { 681 prim = inet_ifa_byprefix(in_dev, prefix, mask); 682 if (prim == NULL) { 683 printk(KERN_DEBUG "fib_add_ifaddr: bug: prim == NULL\n"); 684 return; 685 } 686 } 687 688 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim); 689 690 if (!(dev->flags&IFF_UP)) 691 return; 692 693 /* Add broadcast address, if it is explicitly assigned. */ 694 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF)) 695 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim); 696 697 if (!ZERONET(prefix) && !(ifa->ifa_flags&IFA_F_SECONDARY) && 698 (prefix != addr || ifa->ifa_prefixlen < 32)) { 699 fib_magic(RTM_NEWROUTE, dev->flags&IFF_LOOPBACK ? RTN_LOCAL : 700 RTN_UNICAST, prefix, ifa->ifa_prefixlen, prim); 701 702 /* Add network specific broadcasts, when it takes a sense */ 703 if (ifa->ifa_prefixlen < 31) { 704 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim); 705 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix|~mask, 32, prim); 706 } 707 } 708} 709 710static void fib_del_ifaddr(struct in_ifaddr *ifa) 711{ 712 struct in_device *in_dev = ifa->ifa_dev; 713 struct net_device *dev = in_dev->dev; 714 struct in_ifaddr *ifa1; 715 struct in_ifaddr *prim = ifa; 716 __be32 brd = ifa->ifa_address|~ifa->ifa_mask; 717 __be32 any = ifa->ifa_address&ifa->ifa_mask; 718#define LOCAL_OK 1 719#define BRD_OK 2 720#define BRD0_OK 4 721#define BRD1_OK 8 722 unsigned ok = 0; 723 724 if (!(ifa->ifa_flags&IFA_F_SECONDARY)) 725 fib_magic(RTM_DELROUTE, dev->flags&IFF_LOOPBACK ? RTN_LOCAL : 726 RTN_UNICAST, any, ifa->ifa_prefixlen, prim); 727 else { 728 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask); 729 if (prim == NULL) { 730 printk(KERN_DEBUG "fib_del_ifaddr: bug: prim == NULL\n"); 731 return; 732 } 733 } 734 735 /* Deletion is more complicated than add. 736 We should take care of not to delete too much :-) 737 738 Scan address list to be sure that addresses are really gone. 739 */ 740 741 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) { 742 if (ifa->ifa_local == ifa1->ifa_local) 743 ok |= LOCAL_OK; 744 if (ifa->ifa_broadcast == ifa1->ifa_broadcast) 745 ok |= BRD_OK; 746 if (brd == ifa1->ifa_broadcast) 747 ok |= BRD1_OK; 748 if (any == ifa1->ifa_broadcast) 749 ok |= BRD0_OK; 750 } 751 752 if (!(ok&BRD_OK)) 753 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim); 754 if (!(ok&BRD1_OK)) 755 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim); 756 if (!(ok&BRD0_OK)) 757 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim); 758 if (!(ok&LOCAL_OK)) { 759 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim); 760 761 /* Check, that this local address finally disappeared. */ 762 if (inet_addr_type(ifa->ifa_local) != RTN_LOCAL) { 763 /* And the last, but not the least thing. 764 We must flush stray FIB entries. 765 766 First of all, we scan fib_info list searching 767 for stray nexthop entries, then ignite fib_flush. 768 */ 769 if (fib_sync_down(ifa->ifa_local, NULL, 0)) 770 fib_flush(); 771 } 772 } 773#undef LOCAL_OK 774#undef BRD_OK 775#undef BRD0_OK 776#undef BRD1_OK 777} 778 779static void nl_fib_lookup(struct fib_result_nl *frn, struct fib_table *tb ) 780{ 781 782 struct fib_result res; 783 struct flowi fl = { .mark = frn->fl_mark, 784 .nl_u = { .ip4_u = { .daddr = frn->fl_addr, 785 .tos = frn->fl_tos, 786 .scope = frn->fl_scope } } }; 787 788#ifdef CONFIG_IP_MULTIPLE_TABLES 789 res.r = NULL; 790#endif 791 792 frn->err = -ENOENT; 793 if (tb) { 794 local_bh_disable(); 795 796 frn->tb_id = tb->tb_id; 797 frn->err = tb->tb_lookup(tb, &fl, &res); 798 799 if (!frn->err) { 800 frn->prefixlen = res.prefixlen; 801 frn->nh_sel = res.nh_sel; 802 frn->type = res.type; 803 frn->scope = res.scope; 804 fib_res_put(&res); 805 } 806 local_bh_enable(); 807 } 808} 809 810static void nl_fib_input(struct sk_buff *skb) 811{ 812 struct fib_result_nl *frn; 813 struct nlmsghdr *nlh; 814 struct fib_table *tb; 815 u32 pid; 816 817 nlh = nlmsg_hdr(skb); 818 if (skb->len < NLMSG_SPACE(0) || skb->len < nlh->nlmsg_len || 819 nlh->nlmsg_len < NLMSG_LENGTH(sizeof(*frn))) 820 return; 821 822 skb = skb_clone(skb, GFP_KERNEL); 823 if (skb == NULL) 824 return; 825 nlh = nlmsg_hdr(skb); 826 827 frn = (struct fib_result_nl *) NLMSG_DATA(nlh); 828 tb = fib_get_table(frn->tb_id_in); 829 830 nl_fib_lookup(frn, tb); 831 832 pid = NETLINK_CB(skb).pid; /* pid of sending process */ 833 NETLINK_CB(skb).pid = 0; /* from kernel */ 834 NETLINK_CB(skb).dst_group = 0; /* unicast */ 835 netlink_unicast(fibnl, skb, pid, MSG_DONTWAIT); 836} 837 838static void nl_fib_lookup_init(void) 839{ 840 fibnl = netlink_kernel_create(&init_net, NETLINK_FIB_LOOKUP, 0, 841 nl_fib_input, NULL, THIS_MODULE); 842} 843 844static void fib_disable_ip(struct net_device *dev, int force) 845{ 846 if (fib_sync_down(0, dev, force)) 847 fib_flush(); 848 rt_cache_flush(0); 849 arp_ifdown(dev); 850} 851 852static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr) 853{ 854 struct in_ifaddr *ifa = (struct in_ifaddr*)ptr; 855 856 switch (event) { 857 case NETDEV_UP: 858 fib_add_ifaddr(ifa); 859#ifdef CONFIG_IP_ROUTE_MULTIPATH 860 fib_sync_up(ifa->ifa_dev->dev); 861#endif 862 rt_cache_flush(-1); 863 break; 864 case NETDEV_DOWN: 865 fib_del_ifaddr(ifa); 866 if (ifa->ifa_dev->ifa_list == NULL) { 867 /* Last address was deleted from this interface. 868 Disable IP. 869 */ 870 fib_disable_ip(ifa->ifa_dev->dev, 1); 871 } else { 872 rt_cache_flush(-1); 873 } 874 break; 875 } 876 return NOTIFY_DONE; 877} 878 879static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr) 880{ 881 struct net_device *dev = ptr; 882 struct in_device *in_dev = __in_dev_get_rtnl(dev); 883 884 if (dev->nd_net != &init_net) 885 return NOTIFY_DONE; 886 887 if (event == NETDEV_UNREGISTER) { 888 fib_disable_ip(dev, 2); 889 return NOTIFY_DONE; 890 } 891 892 if (!in_dev) 893 return NOTIFY_DONE; 894 895 switch (event) { 896 case NETDEV_UP: 897 for_ifa(in_dev) { 898 fib_add_ifaddr(ifa); 899 } endfor_ifa(in_dev); 900#ifdef CONFIG_IP_ROUTE_MULTIPATH 901 fib_sync_up(dev); 902#endif 903 rt_cache_flush(-1); 904 break; 905 case NETDEV_DOWN: 906 fib_disable_ip(dev, 0); 907 break; 908 case NETDEV_CHANGEMTU: 909 case NETDEV_CHANGE: 910 rt_cache_flush(0); 911 break; 912 } 913 return NOTIFY_DONE; 914} 915 916static struct notifier_block fib_inetaddr_notifier = { 917 .notifier_call =fib_inetaddr_event, 918}; 919 920static struct notifier_block fib_netdev_notifier = { 921 .notifier_call =fib_netdev_event, 922}; 923 924void __init ip_fib_init(void) 925{ 926 unsigned int i; 927 928 for (i = 0; i < FIB_TABLE_HASHSZ; i++) 929 INIT_HLIST_HEAD(&fib_table_hash[i]); 930 931 fib4_rules_init(); 932 933 register_netdevice_notifier(&fib_netdev_notifier); 934 register_inetaddr_notifier(&fib_inetaddr_notifier); 935 nl_fib_lookup_init(); 936 937 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL); 938 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL); 939 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib); 940} 941 942EXPORT_SYMBOL(inet_addr_type); 943EXPORT_SYMBOL(ip_dev_find); 944