1 // SPDX-License-Identifier: GPL-2.0-only 2 #include <linux/types.h> 3 #include <linux/skbuff.h> 4 #include <linux/socket.h> 5 #include <linux/sysctl.h> 6 #include <linux/net.h> 7 #include <linux/module.h> 8 #include <linux/if_arp.h> 9 #include <linux/ipv6.h> 10 #include <linux/mpls.h> 11 #include <linux/netconf.h> 12 #include <linux/nospec.h> 13 #include <linux/vmalloc.h> 14 #include <linux/percpu.h> 15 #include <net/ip.h> 16 #include <net/dst.h> 17 #include <net/sock.h> 18 #include <net/arp.h> 19 #include <net/ip_fib.h> 20 #include <net/netevent.h> 21 #include <net/ip_tunnels.h> 22 #include <net/netns/generic.h> 23 #if IS_ENABLED(CONFIG_IPV6) 24 #include <net/ipv6.h> 25 #endif 26 #include <net/ipv6_stubs.h> 27 #include <net/rtnh.h> 28 #include "internal.h" 29 30 /* max memory we will use for mpls_route */ 31 #define MAX_MPLS_ROUTE_MEM 4096 32 33 /* Maximum number of labels to look ahead at when selecting a path of 34 * a multipath route 35 */ 36 #define MAX_MP_SELECT_LABELS 4 37 38 #define MPLS_NEIGH_TABLE_UNSPEC (NEIGH_LINK_TABLE + 1) 39 40 static int label_limit = (1 << 20) - 1; 41 static int ttl_max = 255; 42 43 #if IS_ENABLED(CONFIG_NET_IP_TUNNEL) 44 static size_t ipgre_mpls_encap_hlen(struct ip_tunnel_encap *e) 45 { 46 return sizeof(struct mpls_shim_hdr); 47 } 48 49 static const struct ip_tunnel_encap_ops mpls_iptun_ops = { 50 .encap_hlen = ipgre_mpls_encap_hlen, 51 }; 52 53 static int ipgre_tunnel_encap_add_mpls_ops(void) 54 { 55 return ip_tunnel_encap_add_ops(&mpls_iptun_ops, TUNNEL_ENCAP_MPLS); 56 } 57 58 static void ipgre_tunnel_encap_del_mpls_ops(void) 59 { 60 ip_tunnel_encap_del_ops(&mpls_iptun_ops, TUNNEL_ENCAP_MPLS); 61 } 62 #else 63 static int ipgre_tunnel_encap_add_mpls_ops(void) 64 { 65 return 0; 66 } 67 68 static void ipgre_tunnel_encap_del_mpls_ops(void) 69 { 70 } 71 #endif 72 73 static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt, 74 struct nlmsghdr *nlh, struct net *net, u32 portid, 75 unsigned int nlm_flags); 76 77 static struct mpls_route *mpls_route_input_rcu(struct net *net, unsigned index) 78 { 79 struct mpls_route *rt = NULL; 80 81 if (index < net->mpls.platform_labels) { 82 struct mpls_route __rcu **platform_label = 83 rcu_dereference(net->mpls.platform_label); 84 rt = rcu_dereference(platform_label[index]); 85 } 86 return rt; 87 } 88 89 bool mpls_output_possible(const struct net_device *dev) 90 { 91 return dev && (dev->flags & IFF_UP) && netif_carrier_ok(dev); 92 } 93 EXPORT_SYMBOL_GPL(mpls_output_possible); 94 95 static u8 *__mpls_nh_via(struct mpls_route *rt, struct mpls_nh *nh) 96 { 97 return (u8 *)nh + rt->rt_via_offset; 98 } 99 100 static const u8 *mpls_nh_via(const struct mpls_route *rt, 101 const struct mpls_nh *nh) 102 { 103 return __mpls_nh_via((struct mpls_route *)rt, (struct mpls_nh *)nh); 104 } 105 106 static unsigned int mpls_nh_header_size(const struct mpls_nh *nh) 107 { 108 /* The size of the layer 2.5 labels to be added for this route */ 109 return nh->nh_labels * sizeof(struct mpls_shim_hdr); 110 } 111 112 unsigned int mpls_dev_mtu(const struct net_device *dev) 113 { 114 /* The amount of data the layer 2 frame can hold */ 115 return dev->mtu; 116 } 117 EXPORT_SYMBOL_GPL(mpls_dev_mtu); 118 119 bool mpls_pkt_too_big(const struct sk_buff *skb, unsigned int mtu) 120 { 121 if (skb->len <= mtu) 122 return false; 123 124 if (skb_is_gso(skb) && skb_gso_validate_network_len(skb, mtu)) 125 return false; 126 127 return true; 128 } 129 EXPORT_SYMBOL_GPL(mpls_pkt_too_big); 130 131 void mpls_stats_inc_outucastpkts(struct net_device *dev, 132 const struct sk_buff *skb) 133 { 134 struct mpls_dev *mdev; 135 136 if (skb->protocol == htons(ETH_P_MPLS_UC)) { 137 mdev = mpls_dev_get(dev); 138 if (mdev) 139 MPLS_INC_STATS_LEN(mdev, skb->len, 140 tx_packets, 141 tx_bytes); 142 } else if (skb->protocol == htons(ETH_P_IP)) { 143 IP_UPD_PO_STATS(dev_net(dev), IPSTATS_MIB_OUT, skb->len); 144 #if IS_ENABLED(CONFIG_IPV6) 145 } else if (skb->protocol == htons(ETH_P_IPV6)) { 146 struct inet6_dev *in6dev = __in6_dev_get(dev); 147 148 if (in6dev) 149 IP6_UPD_PO_STATS(dev_net(dev), in6dev, 150 IPSTATS_MIB_OUT, skb->len); 151 #endif 152 } 153 } 154 EXPORT_SYMBOL_GPL(mpls_stats_inc_outucastpkts); 155 156 static u32 mpls_multipath_hash(struct mpls_route *rt, struct sk_buff *skb) 157 { 158 struct mpls_entry_decoded dec; 159 unsigned int mpls_hdr_len = 0; 160 struct mpls_shim_hdr *hdr; 161 bool eli_seen = false; 162 int label_index; 163 u32 hash = 0; 164 165 for (label_index = 0; label_index < MAX_MP_SELECT_LABELS; 166 label_index++) { 167 mpls_hdr_len += sizeof(*hdr); 168 if (!pskb_may_pull(skb, mpls_hdr_len)) 169 break; 170 171 /* Read and decode the current label */ 172 hdr = mpls_hdr(skb) + label_index; 173 dec = mpls_entry_decode(hdr); 174 175 /* RFC6790 - reserved labels MUST NOT be used as keys 176 * for the load-balancing function 177 */ 178 if (likely(dec.label >= MPLS_LABEL_FIRST_UNRESERVED)) { 179 hash = jhash_1word(dec.label, hash); 180 181 /* The entropy label follows the entropy label 182 * indicator, so this means that the entropy 183 * label was just added to the hash - no need to 184 * go any deeper either in the label stack or in the 185 * payload 186 */ 187 if (eli_seen) 188 break; 189 } else if (dec.label == MPLS_LABEL_ENTROPY) { 190 eli_seen = true; 191 } 192 193 if (!dec.bos) 194 continue; 195 196 /* found bottom label; does skb have room for a header? */ 197 if (pskb_may_pull(skb, mpls_hdr_len + sizeof(struct iphdr))) { 198 const struct iphdr *v4hdr; 199 200 v4hdr = (const struct iphdr *)(hdr + 1); 201 if (v4hdr->version == 4) { 202 hash = jhash_3words(ntohl(v4hdr->saddr), 203 ntohl(v4hdr->daddr), 204 v4hdr->protocol, hash); 205 } else if (v4hdr->version == 6 && 206 pskb_may_pull(skb, mpls_hdr_len + 207 sizeof(struct ipv6hdr))) { 208 const struct ipv6hdr *v6hdr; 209 210 v6hdr = (const struct ipv6hdr *)(hdr + 1); 211 hash = __ipv6_addr_jhash(&v6hdr->saddr, hash); 212 hash = __ipv6_addr_jhash(&v6hdr->daddr, hash); 213 hash = jhash_1word(v6hdr->nexthdr, hash); 214 } 215 } 216 217 break; 218 } 219 220 return hash; 221 } 222 223 static struct mpls_nh *mpls_get_nexthop(struct mpls_route *rt, u8 index) 224 { 225 return (struct mpls_nh *)((u8 *)rt->rt_nh + index * rt->rt_nh_size); 226 } 227 228 /* number of alive nexthops (rt->rt_nhn_alive) and the flags for 229 * a next hop (nh->nh_flags) are modified by netdev event handlers. 230 * Since those fields can change at any moment, use READ_ONCE to 231 * access both. 232 */ 233 static const struct mpls_nh *mpls_select_multipath(struct mpls_route *rt, 234 struct sk_buff *skb) 235 { 236 u32 hash = 0; 237 int nh_index = 0; 238 int n = 0; 239 u8 alive; 240 241 /* No need to look further into packet if there's only 242 * one path 243 */ 244 if (rt->rt_nhn == 1) 245 return rt->rt_nh; 246 247 alive = READ_ONCE(rt->rt_nhn_alive); 248 if (alive == 0) 249 return NULL; 250 251 hash = mpls_multipath_hash(rt, skb); 252 nh_index = hash % alive; 253 if (alive == rt->rt_nhn) 254 goto out; 255 for_nexthops(rt) { 256 unsigned int nh_flags = READ_ONCE(nh->nh_flags); 257 258 if (nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)) 259 continue; 260 if (n == nh_index) 261 return nh; 262 n++; 263 } endfor_nexthops(rt); 264 265 out: 266 return mpls_get_nexthop(rt, nh_index); 267 } 268 269 static bool mpls_egress(struct net *net, struct mpls_route *rt, 270 struct sk_buff *skb, struct mpls_entry_decoded dec) 271 { 272 enum mpls_payload_type payload_type; 273 bool success = false; 274 275 /* The IPv4 code below accesses through the IPv4 header 276 * checksum, which is 12 bytes into the packet. 277 * The IPv6 code below accesses through the IPv6 hop limit 278 * which is 8 bytes into the packet. 279 * 280 * For all supported cases there should always be at least 12 281 * bytes of packet data present. The IPv4 header is 20 bytes 282 * without options and the IPv6 header is always 40 bytes 283 * long. 284 */ 285 if (!pskb_may_pull(skb, 12)) 286 return false; 287 288 payload_type = rt->rt_payload_type; 289 if (payload_type == MPT_UNSPEC) 290 payload_type = ip_hdr(skb)->version; 291 292 switch (payload_type) { 293 case MPT_IPV4: { 294 struct iphdr *hdr4 = ip_hdr(skb); 295 u8 new_ttl; 296 skb->protocol = htons(ETH_P_IP); 297 298 /* If propagating TTL, take the decremented TTL from 299 * the incoming MPLS header, otherwise decrement the 300 * TTL, but only if not 0 to avoid underflow. 301 */ 302 if (rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED || 303 (rt->rt_ttl_propagate == MPLS_TTL_PROP_DEFAULT && 304 net->mpls.ip_ttl_propagate)) 305 new_ttl = dec.ttl; 306 else 307 new_ttl = hdr4->ttl ? hdr4->ttl - 1 : 0; 308 309 csum_replace2(&hdr4->check, 310 htons(hdr4->ttl << 8), 311 htons(new_ttl << 8)); 312 hdr4->ttl = new_ttl; 313 success = true; 314 break; 315 } 316 case MPT_IPV6: { 317 struct ipv6hdr *hdr6 = ipv6_hdr(skb); 318 skb->protocol = htons(ETH_P_IPV6); 319 320 /* If propagating TTL, take the decremented TTL from 321 * the incoming MPLS header, otherwise decrement the 322 * hop limit, but only if not 0 to avoid underflow. 323 */ 324 if (rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED || 325 (rt->rt_ttl_propagate == MPLS_TTL_PROP_DEFAULT && 326 net->mpls.ip_ttl_propagate)) 327 hdr6->hop_limit = dec.ttl; 328 else if (hdr6->hop_limit) 329 hdr6->hop_limit = hdr6->hop_limit - 1; 330 success = true; 331 break; 332 } 333 case MPT_UNSPEC: 334 /* Should have decided which protocol it is by now */ 335 break; 336 } 337 338 return success; 339 } 340 341 static int mpls_forward(struct sk_buff *skb, struct net_device *dev, 342 struct packet_type *pt, struct net_device *orig_dev) 343 { 344 struct net *net = dev_net(dev); 345 struct mpls_shim_hdr *hdr; 346 const struct mpls_nh *nh; 347 struct mpls_route *rt; 348 struct mpls_entry_decoded dec; 349 struct net_device *out_dev; 350 struct mpls_dev *out_mdev; 351 struct mpls_dev *mdev; 352 unsigned int hh_len; 353 unsigned int new_header_size; 354 unsigned int mtu; 355 int err; 356 357 /* Careful this entire function runs inside of an rcu critical section */ 358 359 mdev = mpls_dev_get(dev); 360 if (!mdev) 361 goto drop; 362 363 MPLS_INC_STATS_LEN(mdev, skb->len, rx_packets, 364 rx_bytes); 365 366 if (!mdev->input_enabled) { 367 MPLS_INC_STATS(mdev, rx_dropped); 368 goto drop; 369 } 370 371 if (skb->pkt_type != PACKET_HOST) 372 goto err; 373 374 if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL) 375 goto err; 376 377 if (!pskb_may_pull(skb, sizeof(*hdr))) 378 goto err; 379 380 skb_dst_drop(skb); 381 382 /* Read and decode the label */ 383 hdr = mpls_hdr(skb); 384 dec = mpls_entry_decode(hdr); 385 386 rt = mpls_route_input_rcu(net, dec.label); 387 if (!rt) { 388 MPLS_INC_STATS(mdev, rx_noroute); 389 goto drop; 390 } 391 392 nh = mpls_select_multipath(rt, skb); 393 if (!nh) 394 goto err; 395 396 /* Pop the label */ 397 skb_pull(skb, sizeof(*hdr)); 398 skb_reset_network_header(skb); 399 400 skb_orphan(skb); 401 402 if (skb_warn_if_lro(skb)) 403 goto err; 404 405 skb_forward_csum(skb); 406 407 /* Verify ttl is valid */ 408 if (dec.ttl <= 1) 409 goto err; 410 411 /* Find the output device */ 412 out_dev = nh->nh_dev; 413 if (!mpls_output_possible(out_dev)) 414 goto tx_err; 415 416 /* Verify the destination can hold the packet */ 417 new_header_size = mpls_nh_header_size(nh); 418 mtu = mpls_dev_mtu(out_dev); 419 if (mpls_pkt_too_big(skb, mtu - new_header_size)) 420 goto tx_err; 421 422 hh_len = LL_RESERVED_SPACE(out_dev); 423 if (!out_dev->header_ops) 424 hh_len = 0; 425 426 /* Ensure there is enough space for the headers in the skb */ 427 if (skb_cow(skb, hh_len + new_header_size)) 428 goto tx_err; 429 430 skb->dev = out_dev; 431 skb->protocol = htons(ETH_P_MPLS_UC); 432 433 dec.ttl -= 1; 434 if (unlikely(!new_header_size && dec.bos)) { 435 /* Penultimate hop popping */ 436 if (!mpls_egress(dev_net(out_dev), rt, skb, dec)) 437 goto err; 438 } else { 439 bool bos; 440 int i; 441 skb_push(skb, new_header_size); 442 skb_reset_network_header(skb); 443 /* Push the new labels */ 444 hdr = mpls_hdr(skb); 445 bos = dec.bos; 446 for (i = nh->nh_labels - 1; i >= 0; i--) { 447 hdr[i] = mpls_entry_encode(nh->nh_label[i], 448 dec.ttl, 0, bos); 449 bos = false; 450 } 451 } 452 453 mpls_stats_inc_outucastpkts(out_dev, skb); 454 455 /* If via wasn't specified then send out using device address */ 456 if (nh->nh_via_table == MPLS_NEIGH_TABLE_UNSPEC) 457 err = neigh_xmit(NEIGH_LINK_TABLE, out_dev, 458 out_dev->dev_addr, skb); 459 else 460 err = neigh_xmit(nh->nh_via_table, out_dev, 461 mpls_nh_via(rt, nh), skb); 462 if (err) 463 net_dbg_ratelimited("%s: packet transmission failed: %d\n", 464 __func__, err); 465 return 0; 466 467 tx_err: 468 out_mdev = out_dev ? mpls_dev_get(out_dev) : NULL; 469 if (out_mdev) 470 MPLS_INC_STATS(out_mdev, tx_errors); 471 goto drop; 472 err: 473 MPLS_INC_STATS(mdev, rx_errors); 474 drop: 475 kfree_skb(skb); 476 return NET_RX_DROP; 477 } 478 479 static struct packet_type mpls_packet_type __read_mostly = { 480 .type = cpu_to_be16(ETH_P_MPLS_UC), 481 .func = mpls_forward, 482 }; 483 484 static const struct nla_policy rtm_mpls_policy[RTA_MAX+1] = { 485 [RTA_DST] = { .type = NLA_U32 }, 486 [RTA_OIF] = { .type = NLA_U32 }, 487 [RTA_TTL_PROPAGATE] = { .type = NLA_U8 }, 488 }; 489 490 struct mpls_route_config { 491 u32 rc_protocol; 492 u32 rc_ifindex; 493 u8 rc_via_table; 494 u8 rc_via_alen; 495 u8 rc_via[MAX_VIA_ALEN]; 496 u32 rc_label; 497 u8 rc_ttl_propagate; 498 u8 rc_output_labels; 499 u32 rc_output_label[MAX_NEW_LABELS]; 500 u32 rc_nlflags; 501 enum mpls_payload_type rc_payload_type; 502 struct nl_info rc_nlinfo; 503 struct rtnexthop *rc_mp; 504 int rc_mp_len; 505 }; 506 507 /* all nexthops within a route have the same size based on max 508 * number of labels and max via length for a hop 509 */ 510 static struct mpls_route *mpls_rt_alloc(u8 num_nh, u8 max_alen, u8 max_labels) 511 { 512 u8 nh_size = MPLS_NH_SIZE(max_labels, max_alen); 513 struct mpls_route *rt; 514 size_t size; 515 516 size = sizeof(*rt) + num_nh * nh_size; 517 if (size > MAX_MPLS_ROUTE_MEM) 518 return ERR_PTR(-EINVAL); 519 520 rt = kzalloc(size, GFP_KERNEL); 521 if (!rt) 522 return ERR_PTR(-ENOMEM); 523 524 rt->rt_nhn = num_nh; 525 rt->rt_nhn_alive = num_nh; 526 rt->rt_nh_size = nh_size; 527 rt->rt_via_offset = MPLS_NH_VIA_OFF(max_labels); 528 529 return rt; 530 } 531 532 static void mpls_rt_free(struct mpls_route *rt) 533 { 534 if (rt) 535 kfree_rcu(rt, rt_rcu); 536 } 537 538 static void mpls_notify_route(struct net *net, unsigned index, 539 struct mpls_route *old, struct mpls_route *new, 540 const struct nl_info *info) 541 { 542 struct nlmsghdr *nlh = info ? info->nlh : NULL; 543 unsigned portid = info ? info->portid : 0; 544 int event = new ? RTM_NEWROUTE : RTM_DELROUTE; 545 struct mpls_route *rt = new ? new : old; 546 unsigned nlm_flags = (old && new) ? NLM_F_REPLACE : 0; 547 /* Ignore reserved labels for now */ 548 if (rt && (index >= MPLS_LABEL_FIRST_UNRESERVED)) 549 rtmsg_lfib(event, index, rt, nlh, net, portid, nlm_flags); 550 } 551 552 static void mpls_route_update(struct net *net, unsigned index, 553 struct mpls_route *new, 554 const struct nl_info *info) 555 { 556 struct mpls_route __rcu **platform_label; 557 struct mpls_route *rt; 558 559 ASSERT_RTNL(); 560 561 platform_label = rtnl_dereference(net->mpls.platform_label); 562 rt = rtnl_dereference(platform_label[index]); 563 rcu_assign_pointer(platform_label[index], new); 564 565 mpls_notify_route(net, index, rt, new, info); 566 567 /* If we removed a route free it now */ 568 mpls_rt_free(rt); 569 } 570 571 static unsigned find_free_label(struct net *net) 572 { 573 struct mpls_route __rcu **platform_label; 574 size_t platform_labels; 575 unsigned index; 576 577 platform_label = rtnl_dereference(net->mpls.platform_label); 578 platform_labels = net->mpls.platform_labels; 579 for (index = MPLS_LABEL_FIRST_UNRESERVED; index < platform_labels; 580 index++) { 581 if (!rtnl_dereference(platform_label[index])) 582 return index; 583 } 584 return LABEL_NOT_SPECIFIED; 585 } 586 587 #if IS_ENABLED(CONFIG_INET) 588 static struct net_device *inet_fib_lookup_dev(struct net *net, 589 const void *addr) 590 { 591 struct net_device *dev; 592 struct rtable *rt; 593 struct in_addr daddr; 594 595 memcpy(&daddr, addr, sizeof(struct in_addr)); 596 rt = ip_route_output(net, daddr.s_addr, 0, 0, 0); 597 if (IS_ERR(rt)) 598 return ERR_CAST(rt); 599 600 dev = rt->dst.dev; 601 dev_hold(dev); 602 603 ip_rt_put(rt); 604 605 return dev; 606 } 607 #else 608 static struct net_device *inet_fib_lookup_dev(struct net *net, 609 const void *addr) 610 { 611 return ERR_PTR(-EAFNOSUPPORT); 612 } 613 #endif 614 615 #if IS_ENABLED(CONFIG_IPV6) 616 static struct net_device *inet6_fib_lookup_dev(struct net *net, 617 const void *addr) 618 { 619 struct net_device *dev; 620 struct dst_entry *dst; 621 struct flowi6 fl6; 622 623 if (!ipv6_stub) 624 return ERR_PTR(-EAFNOSUPPORT); 625 626 memset(&fl6, 0, sizeof(fl6)); 627 memcpy(&fl6.daddr, addr, sizeof(struct in6_addr)); 628 dst = ipv6_stub->ipv6_dst_lookup_flow(net, NULL, &fl6, NULL); 629 if (IS_ERR(dst)) 630 return ERR_CAST(dst); 631 632 dev = dst->dev; 633 dev_hold(dev); 634 dst_release(dst); 635 636 return dev; 637 } 638 #else 639 static struct net_device *inet6_fib_lookup_dev(struct net *net, 640 const void *addr) 641 { 642 return ERR_PTR(-EAFNOSUPPORT); 643 } 644 #endif 645 646 static struct net_device *find_outdev(struct net *net, 647 struct mpls_route *rt, 648 struct mpls_nh *nh, int oif) 649 { 650 struct net_device *dev = NULL; 651 652 if (!oif) { 653 switch (nh->nh_via_table) { 654 case NEIGH_ARP_TABLE: 655 dev = inet_fib_lookup_dev(net, mpls_nh_via(rt, nh)); 656 break; 657 case NEIGH_ND_TABLE: 658 dev = inet6_fib_lookup_dev(net, mpls_nh_via(rt, nh)); 659 break; 660 case NEIGH_LINK_TABLE: 661 break; 662 } 663 } else { 664 dev = dev_get_by_index(net, oif); 665 } 666 667 if (!dev) 668 return ERR_PTR(-ENODEV); 669 670 if (IS_ERR(dev)) 671 return dev; 672 673 /* The caller is holding rtnl anyways, so release the dev reference */ 674 dev_put(dev); 675 676 return dev; 677 } 678 679 static int mpls_nh_assign_dev(struct net *net, struct mpls_route *rt, 680 struct mpls_nh *nh, int oif) 681 { 682 struct net_device *dev = NULL; 683 int err = -ENODEV; 684 685 dev = find_outdev(net, rt, nh, oif); 686 if (IS_ERR(dev)) { 687 err = PTR_ERR(dev); 688 dev = NULL; 689 goto errout; 690 } 691 692 /* Ensure this is a supported device */ 693 err = -EINVAL; 694 if (!mpls_dev_get(dev)) 695 goto errout; 696 697 if ((nh->nh_via_table == NEIGH_LINK_TABLE) && 698 (dev->addr_len != nh->nh_via_alen)) 699 goto errout; 700 701 nh->nh_dev = dev; 702 703 if (!(dev->flags & IFF_UP)) { 704 nh->nh_flags |= RTNH_F_DEAD; 705 } else { 706 unsigned int flags; 707 708 flags = dev_get_flags(dev); 709 if (!(flags & (IFF_RUNNING | IFF_LOWER_UP))) 710 nh->nh_flags |= RTNH_F_LINKDOWN; 711 } 712 713 return 0; 714 715 errout: 716 return err; 717 } 718 719 static int nla_get_via(const struct nlattr *nla, u8 *via_alen, u8 *via_table, 720 u8 via_addr[], struct netlink_ext_ack *extack) 721 { 722 struct rtvia *via = nla_data(nla); 723 int err = -EINVAL; 724 int alen; 725 726 if (nla_len(nla) < offsetof(struct rtvia, rtvia_addr)) { 727 NL_SET_ERR_MSG_ATTR(extack, nla, 728 "Invalid attribute length for RTA_VIA"); 729 goto errout; 730 } 731 alen = nla_len(nla) - 732 offsetof(struct rtvia, rtvia_addr); 733 if (alen > MAX_VIA_ALEN) { 734 NL_SET_ERR_MSG_ATTR(extack, nla, 735 "Invalid address length for RTA_VIA"); 736 goto errout; 737 } 738 739 /* Validate the address family */ 740 switch (via->rtvia_family) { 741 case AF_PACKET: 742 *via_table = NEIGH_LINK_TABLE; 743 break; 744 case AF_INET: 745 *via_table = NEIGH_ARP_TABLE; 746 if (alen != 4) 747 goto errout; 748 break; 749 case AF_INET6: 750 *via_table = NEIGH_ND_TABLE; 751 if (alen != 16) 752 goto errout; 753 break; 754 default: 755 /* Unsupported address family */ 756 goto errout; 757 } 758 759 memcpy(via_addr, via->rtvia_addr, alen); 760 *via_alen = alen; 761 err = 0; 762 763 errout: 764 return err; 765 } 766 767 static int mpls_nh_build_from_cfg(struct mpls_route_config *cfg, 768 struct mpls_route *rt) 769 { 770 struct net *net = cfg->rc_nlinfo.nl_net; 771 struct mpls_nh *nh = rt->rt_nh; 772 int err; 773 int i; 774 775 if (!nh) 776 return -ENOMEM; 777 778 nh->nh_labels = cfg->rc_output_labels; 779 for (i = 0; i < nh->nh_labels; i++) 780 nh->nh_label[i] = cfg->rc_output_label[i]; 781 782 nh->nh_via_table = cfg->rc_via_table; 783 memcpy(__mpls_nh_via(rt, nh), cfg->rc_via, cfg->rc_via_alen); 784 nh->nh_via_alen = cfg->rc_via_alen; 785 786 err = mpls_nh_assign_dev(net, rt, nh, cfg->rc_ifindex); 787 if (err) 788 goto errout; 789 790 if (nh->nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)) 791 rt->rt_nhn_alive--; 792 793 return 0; 794 795 errout: 796 return err; 797 } 798 799 static int mpls_nh_build(struct net *net, struct mpls_route *rt, 800 struct mpls_nh *nh, int oif, struct nlattr *via, 801 struct nlattr *newdst, u8 max_labels, 802 struct netlink_ext_ack *extack) 803 { 804 int err = -ENOMEM; 805 806 if (!nh) 807 goto errout; 808 809 if (newdst) { 810 err = nla_get_labels(newdst, max_labels, &nh->nh_labels, 811 nh->nh_label, extack); 812 if (err) 813 goto errout; 814 } 815 816 if (via) { 817 err = nla_get_via(via, &nh->nh_via_alen, &nh->nh_via_table, 818 __mpls_nh_via(rt, nh), extack); 819 if (err) 820 goto errout; 821 } else { 822 nh->nh_via_table = MPLS_NEIGH_TABLE_UNSPEC; 823 } 824 825 err = mpls_nh_assign_dev(net, rt, nh, oif); 826 if (err) 827 goto errout; 828 829 return 0; 830 831 errout: 832 return err; 833 } 834 835 static u8 mpls_count_nexthops(struct rtnexthop *rtnh, int len, 836 u8 cfg_via_alen, u8 *max_via_alen, 837 u8 *max_labels) 838 { 839 int remaining = len; 840 u8 nhs = 0; 841 842 *max_via_alen = 0; 843 *max_labels = 0; 844 845 while (rtnh_ok(rtnh, remaining)) { 846 struct nlattr *nla, *attrs = rtnh_attrs(rtnh); 847 int attrlen; 848 u8 n_labels = 0; 849 850 attrlen = rtnh_attrlen(rtnh); 851 nla = nla_find(attrs, attrlen, RTA_VIA); 852 if (nla && nla_len(nla) >= 853 offsetof(struct rtvia, rtvia_addr)) { 854 int via_alen = nla_len(nla) - 855 offsetof(struct rtvia, rtvia_addr); 856 857 if (via_alen <= MAX_VIA_ALEN) 858 *max_via_alen = max_t(u16, *max_via_alen, 859 via_alen); 860 } 861 862 nla = nla_find(attrs, attrlen, RTA_NEWDST); 863 if (nla && 864 nla_get_labels(nla, MAX_NEW_LABELS, &n_labels, 865 NULL, NULL) != 0) 866 return 0; 867 868 *max_labels = max_t(u8, *max_labels, n_labels); 869 870 /* number of nexthops is tracked by a u8. 871 * Check for overflow. 872 */ 873 if (nhs == 255) 874 return 0; 875 nhs++; 876 877 rtnh = rtnh_next(rtnh, &remaining); 878 } 879 880 /* leftover implies invalid nexthop configuration, discard it */ 881 return remaining > 0 ? 0 : nhs; 882 } 883 884 static int mpls_nh_build_multi(struct mpls_route_config *cfg, 885 struct mpls_route *rt, u8 max_labels, 886 struct netlink_ext_ack *extack) 887 { 888 struct rtnexthop *rtnh = cfg->rc_mp; 889 struct nlattr *nla_via, *nla_newdst; 890 int remaining = cfg->rc_mp_len; 891 int err = 0; 892 u8 nhs = 0; 893 894 change_nexthops(rt) { 895 int attrlen; 896 897 nla_via = NULL; 898 nla_newdst = NULL; 899 900 err = -EINVAL; 901 if (!rtnh_ok(rtnh, remaining)) 902 goto errout; 903 904 /* neither weighted multipath nor any flags 905 * are supported 906 */ 907 if (rtnh->rtnh_hops || rtnh->rtnh_flags) 908 goto errout; 909 910 attrlen = rtnh_attrlen(rtnh); 911 if (attrlen > 0) { 912 struct nlattr *attrs = rtnh_attrs(rtnh); 913 914 nla_via = nla_find(attrs, attrlen, RTA_VIA); 915 nla_newdst = nla_find(attrs, attrlen, RTA_NEWDST); 916 } 917 918 err = mpls_nh_build(cfg->rc_nlinfo.nl_net, rt, nh, 919 rtnh->rtnh_ifindex, nla_via, nla_newdst, 920 max_labels, extack); 921 if (err) 922 goto errout; 923 924 if (nh->nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)) 925 rt->rt_nhn_alive--; 926 927 rtnh = rtnh_next(rtnh, &remaining); 928 nhs++; 929 } endfor_nexthops(rt); 930 931 rt->rt_nhn = nhs; 932 933 return 0; 934 935 errout: 936 return err; 937 } 938 939 static bool mpls_label_ok(struct net *net, unsigned int *index, 940 struct netlink_ext_ack *extack) 941 { 942 bool is_ok = true; 943 944 /* Reserved labels may not be set */ 945 if (*index < MPLS_LABEL_FIRST_UNRESERVED) { 946 NL_SET_ERR_MSG(extack, 947 "Invalid label - must be MPLS_LABEL_FIRST_UNRESERVED or higher"); 948 is_ok = false; 949 } 950 951 /* The full 20 bit range may not be supported. */ 952 if (is_ok && *index >= net->mpls.platform_labels) { 953 NL_SET_ERR_MSG(extack, 954 "Label >= configured maximum in platform_labels"); 955 is_ok = false; 956 } 957 958 *index = array_index_nospec(*index, net->mpls.platform_labels); 959 return is_ok; 960 } 961 962 static int mpls_route_add(struct mpls_route_config *cfg, 963 struct netlink_ext_ack *extack) 964 { 965 struct mpls_route __rcu **platform_label; 966 struct net *net = cfg->rc_nlinfo.nl_net; 967 struct mpls_route *rt, *old; 968 int err = -EINVAL; 969 u8 max_via_alen; 970 unsigned index; 971 u8 max_labels; 972 u8 nhs; 973 974 index = cfg->rc_label; 975 976 /* If a label was not specified during insert pick one */ 977 if ((index == LABEL_NOT_SPECIFIED) && 978 (cfg->rc_nlflags & NLM_F_CREATE)) { 979 index = find_free_label(net); 980 } 981 982 if (!mpls_label_ok(net, &index, extack)) 983 goto errout; 984 985 /* Append makes no sense with mpls */ 986 err = -EOPNOTSUPP; 987 if (cfg->rc_nlflags & NLM_F_APPEND) { 988 NL_SET_ERR_MSG(extack, "MPLS does not support route append"); 989 goto errout; 990 } 991 992 err = -EEXIST; 993 platform_label = rtnl_dereference(net->mpls.platform_label); 994 old = rtnl_dereference(platform_label[index]); 995 if ((cfg->rc_nlflags & NLM_F_EXCL) && old) 996 goto errout; 997 998 err = -EEXIST; 999 if (!(cfg->rc_nlflags & NLM_F_REPLACE) && old) 1000 goto errout; 1001 1002 err = -ENOENT; 1003 if (!(cfg->rc_nlflags & NLM_F_CREATE) && !old) 1004 goto errout; 1005 1006 err = -EINVAL; 1007 if (cfg->rc_mp) { 1008 nhs = mpls_count_nexthops(cfg->rc_mp, cfg->rc_mp_len, 1009 cfg->rc_via_alen, &max_via_alen, 1010 &max_labels); 1011 } else { 1012 max_via_alen = cfg->rc_via_alen; 1013 max_labels = cfg->rc_output_labels; 1014 nhs = 1; 1015 } 1016 1017 if (nhs == 0) { 1018 NL_SET_ERR_MSG(extack, "Route does not contain a nexthop"); 1019 goto errout; 1020 } 1021 1022 rt = mpls_rt_alloc(nhs, max_via_alen, max_labels); 1023 if (IS_ERR(rt)) { 1024 err = PTR_ERR(rt); 1025 goto errout; 1026 } 1027 1028 rt->rt_protocol = cfg->rc_protocol; 1029 rt->rt_payload_type = cfg->rc_payload_type; 1030 rt->rt_ttl_propagate = cfg->rc_ttl_propagate; 1031 1032 if (cfg->rc_mp) 1033 err = mpls_nh_build_multi(cfg, rt, max_labels, extack); 1034 else 1035 err = mpls_nh_build_from_cfg(cfg, rt); 1036 if (err) 1037 goto freert; 1038 1039 mpls_route_update(net, index, rt, &cfg->rc_nlinfo); 1040 1041 return 0; 1042 1043 freert: 1044 mpls_rt_free(rt); 1045 errout: 1046 return err; 1047 } 1048 1049 static int mpls_route_del(struct mpls_route_config *cfg, 1050 struct netlink_ext_ack *extack) 1051 { 1052 struct net *net = cfg->rc_nlinfo.nl_net; 1053 unsigned index; 1054 int err = -EINVAL; 1055 1056 index = cfg->rc_label; 1057 1058 if (!mpls_label_ok(net, &index, extack)) 1059 goto errout; 1060 1061 mpls_route_update(net, index, NULL, &cfg->rc_nlinfo); 1062 1063 err = 0; 1064 errout: 1065 return err; 1066 } 1067 1068 static void mpls_get_stats(struct mpls_dev *mdev, 1069 struct mpls_link_stats *stats) 1070 { 1071 struct mpls_pcpu_stats *p; 1072 int i; 1073 1074 memset(stats, 0, sizeof(*stats)); 1075 1076 for_each_possible_cpu(i) { 1077 struct mpls_link_stats local; 1078 unsigned int start; 1079 1080 p = per_cpu_ptr(mdev->stats, i); 1081 do { 1082 start = u64_stats_fetch_begin(&p->syncp); 1083 local = p->stats; 1084 } while (u64_stats_fetch_retry(&p->syncp, start)); 1085 1086 stats->rx_packets += local.rx_packets; 1087 stats->rx_bytes += local.rx_bytes; 1088 stats->tx_packets += local.tx_packets; 1089 stats->tx_bytes += local.tx_bytes; 1090 stats->rx_errors += local.rx_errors; 1091 stats->tx_errors += local.tx_errors; 1092 stats->rx_dropped += local.rx_dropped; 1093 stats->tx_dropped += local.tx_dropped; 1094 stats->rx_noroute += local.rx_noroute; 1095 } 1096 } 1097 1098 static int mpls_fill_stats_af(struct sk_buff *skb, 1099 const struct net_device *dev) 1100 { 1101 struct mpls_link_stats *stats; 1102 struct mpls_dev *mdev; 1103 struct nlattr *nla; 1104 1105 mdev = mpls_dev_get(dev); 1106 if (!mdev) 1107 return -ENODATA; 1108 1109 nla = nla_reserve_64bit(skb, MPLS_STATS_LINK, 1110 sizeof(struct mpls_link_stats), 1111 MPLS_STATS_UNSPEC); 1112 if (!nla) 1113 return -EMSGSIZE; 1114 1115 stats = nla_data(nla); 1116 mpls_get_stats(mdev, stats); 1117 1118 return 0; 1119 } 1120 1121 static size_t mpls_get_stats_af_size(const struct net_device *dev) 1122 { 1123 struct mpls_dev *mdev; 1124 1125 mdev = mpls_dev_get(dev); 1126 if (!mdev) 1127 return 0; 1128 1129 return nla_total_size_64bit(sizeof(struct mpls_link_stats)); 1130 } 1131 1132 static int mpls_netconf_fill_devconf(struct sk_buff *skb, struct mpls_dev *mdev, 1133 u32 portid, u32 seq, int event, 1134 unsigned int flags, int type) 1135 { 1136 struct nlmsghdr *nlh; 1137 struct netconfmsg *ncm; 1138 bool all = false; 1139 1140 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg), 1141 flags); 1142 if (!nlh) 1143 return -EMSGSIZE; 1144 1145 if (type == NETCONFA_ALL) 1146 all = true; 1147 1148 ncm = nlmsg_data(nlh); 1149 ncm->ncm_family = AF_MPLS; 1150 1151 if (nla_put_s32(skb, NETCONFA_IFINDEX, mdev->dev->ifindex) < 0) 1152 goto nla_put_failure; 1153 1154 if ((all || type == NETCONFA_INPUT) && 1155 nla_put_s32(skb, NETCONFA_INPUT, 1156 mdev->input_enabled) < 0) 1157 goto nla_put_failure; 1158 1159 nlmsg_end(skb, nlh); 1160 return 0; 1161 1162 nla_put_failure: 1163 nlmsg_cancel(skb, nlh); 1164 return -EMSGSIZE; 1165 } 1166 1167 static int mpls_netconf_msgsize_devconf(int type) 1168 { 1169 int size = NLMSG_ALIGN(sizeof(struct netconfmsg)) 1170 + nla_total_size(4); /* NETCONFA_IFINDEX */ 1171 bool all = false; 1172 1173 if (type == NETCONFA_ALL) 1174 all = true; 1175 1176 if (all || type == NETCONFA_INPUT) 1177 size += nla_total_size(4); 1178 1179 return size; 1180 } 1181 1182 static void mpls_netconf_notify_devconf(struct net *net, int event, 1183 int type, struct mpls_dev *mdev) 1184 { 1185 struct sk_buff *skb; 1186 int err = -ENOBUFS; 1187 1188 skb = nlmsg_new(mpls_netconf_msgsize_devconf(type), GFP_KERNEL); 1189 if (!skb) 1190 goto errout; 1191 1192 err = mpls_netconf_fill_devconf(skb, mdev, 0, 0, event, 0, type); 1193 if (err < 0) { 1194 /* -EMSGSIZE implies BUG in mpls_netconf_msgsize_devconf() */ 1195 WARN_ON(err == -EMSGSIZE); 1196 kfree_skb(skb); 1197 goto errout; 1198 } 1199 1200 rtnl_notify(skb, net, 0, RTNLGRP_MPLS_NETCONF, NULL, GFP_KERNEL); 1201 return; 1202 errout: 1203 if (err < 0) 1204 rtnl_set_sk_err(net, RTNLGRP_MPLS_NETCONF, err); 1205 } 1206 1207 static const struct nla_policy devconf_mpls_policy[NETCONFA_MAX + 1] = { 1208 [NETCONFA_IFINDEX] = { .len = sizeof(int) }, 1209 }; 1210 1211 static int mpls_netconf_valid_get_req(struct sk_buff *skb, 1212 const struct nlmsghdr *nlh, 1213 struct nlattr **tb, 1214 struct netlink_ext_ack *extack) 1215 { 1216 int i, err; 1217 1218 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) { 1219 NL_SET_ERR_MSG_MOD(extack, 1220 "Invalid header for netconf get request"); 1221 return -EINVAL; 1222 } 1223 1224 if (!netlink_strict_get_check(skb)) 1225 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg), 1226 tb, NETCONFA_MAX, 1227 devconf_mpls_policy, extack); 1228 1229 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg), 1230 tb, NETCONFA_MAX, 1231 devconf_mpls_policy, extack); 1232 if (err) 1233 return err; 1234 1235 for (i = 0; i <= NETCONFA_MAX; i++) { 1236 if (!tb[i]) 1237 continue; 1238 1239 switch (i) { 1240 case NETCONFA_IFINDEX: 1241 break; 1242 default: 1243 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request"); 1244 return -EINVAL; 1245 } 1246 } 1247 1248 return 0; 1249 } 1250 1251 static int mpls_netconf_get_devconf(struct sk_buff *in_skb, 1252 struct nlmsghdr *nlh, 1253 struct netlink_ext_ack *extack) 1254 { 1255 struct net *net = sock_net(in_skb->sk); 1256 struct nlattr *tb[NETCONFA_MAX + 1]; 1257 struct net_device *dev; 1258 struct mpls_dev *mdev; 1259 struct sk_buff *skb; 1260 int ifindex; 1261 int err; 1262 1263 err = mpls_netconf_valid_get_req(in_skb, nlh, tb, extack); 1264 if (err < 0) 1265 goto errout; 1266 1267 err = -EINVAL; 1268 if (!tb[NETCONFA_IFINDEX]) 1269 goto errout; 1270 1271 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]); 1272 dev = __dev_get_by_index(net, ifindex); 1273 if (!dev) 1274 goto errout; 1275 1276 mdev = mpls_dev_get(dev); 1277 if (!mdev) 1278 goto errout; 1279 1280 err = -ENOBUFS; 1281 skb = nlmsg_new(mpls_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL); 1282 if (!skb) 1283 goto errout; 1284 1285 err = mpls_netconf_fill_devconf(skb, mdev, 1286 NETLINK_CB(in_skb).portid, 1287 nlh->nlmsg_seq, RTM_NEWNETCONF, 0, 1288 NETCONFA_ALL); 1289 if (err < 0) { 1290 /* -EMSGSIZE implies BUG in mpls_netconf_msgsize_devconf() */ 1291 WARN_ON(err == -EMSGSIZE); 1292 kfree_skb(skb); 1293 goto errout; 1294 } 1295 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid); 1296 errout: 1297 return err; 1298 } 1299 1300 static int mpls_netconf_dump_devconf(struct sk_buff *skb, 1301 struct netlink_callback *cb) 1302 { 1303 const struct nlmsghdr *nlh = cb->nlh; 1304 struct net *net = sock_net(skb->sk); 1305 struct hlist_head *head; 1306 struct net_device *dev; 1307 struct mpls_dev *mdev; 1308 int idx, s_idx; 1309 int h, s_h; 1310 1311 if (cb->strict_check) { 1312 struct netlink_ext_ack *extack = cb->extack; 1313 struct netconfmsg *ncm; 1314 1315 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) { 1316 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request"); 1317 return -EINVAL; 1318 } 1319 1320 if (nlmsg_attrlen(nlh, sizeof(*ncm))) { 1321 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request"); 1322 return -EINVAL; 1323 } 1324 } 1325 1326 s_h = cb->args[0]; 1327 s_idx = idx = cb->args[1]; 1328 1329 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 1330 idx = 0; 1331 head = &net->dev_index_head[h]; 1332 rcu_read_lock(); 1333 cb->seq = net->dev_base_seq; 1334 hlist_for_each_entry_rcu(dev, head, index_hlist) { 1335 if (idx < s_idx) 1336 goto cont; 1337 mdev = mpls_dev_get(dev); 1338 if (!mdev) 1339 goto cont; 1340 if (mpls_netconf_fill_devconf(skb, mdev, 1341 NETLINK_CB(cb->skb).portid, 1342 nlh->nlmsg_seq, 1343 RTM_NEWNETCONF, 1344 NLM_F_MULTI, 1345 NETCONFA_ALL) < 0) { 1346 rcu_read_unlock(); 1347 goto done; 1348 } 1349 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 1350 cont: 1351 idx++; 1352 } 1353 rcu_read_unlock(); 1354 } 1355 done: 1356 cb->args[0] = h; 1357 cb->args[1] = idx; 1358 1359 return skb->len; 1360 } 1361 1362 #define MPLS_PERDEV_SYSCTL_OFFSET(field) \ 1363 (&((struct mpls_dev *)0)->field) 1364 1365 static int mpls_conf_proc(struct ctl_table *ctl, int write, 1366 void *buffer, size_t *lenp, loff_t *ppos) 1367 { 1368 int oval = *(int *)ctl->data; 1369 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos); 1370 1371 if (write) { 1372 struct mpls_dev *mdev = ctl->extra1; 1373 int i = (int *)ctl->data - (int *)mdev; 1374 struct net *net = ctl->extra2; 1375 int val = *(int *)ctl->data; 1376 1377 if (i == offsetof(struct mpls_dev, input_enabled) && 1378 val != oval) { 1379 mpls_netconf_notify_devconf(net, RTM_NEWNETCONF, 1380 NETCONFA_INPUT, mdev); 1381 } 1382 } 1383 1384 return ret; 1385 } 1386 1387 static const struct ctl_table mpls_dev_table[] = { 1388 { 1389 .procname = "input", 1390 .maxlen = sizeof(int), 1391 .mode = 0644, 1392 .proc_handler = mpls_conf_proc, 1393 .data = MPLS_PERDEV_SYSCTL_OFFSET(input_enabled), 1394 }, 1395 { } 1396 }; 1397 1398 static int mpls_dev_sysctl_register(struct net_device *dev, 1399 struct mpls_dev *mdev) 1400 { 1401 char path[sizeof("net/mpls/conf/") + IFNAMSIZ]; 1402 struct net *net = dev_net(dev); 1403 struct ctl_table *table; 1404 int i; 1405 1406 table = kmemdup(&mpls_dev_table, sizeof(mpls_dev_table), GFP_KERNEL); 1407 if (!table) 1408 goto out; 1409 1410 /* Table data contains only offsets relative to the base of 1411 * the mdev at this point, so make them absolute. 1412 */ 1413 for (i = 0; i < ARRAY_SIZE(mpls_dev_table); i++) { 1414 table[i].data = (char *)mdev + (uintptr_t)table[i].data; 1415 table[i].extra1 = mdev; 1416 table[i].extra2 = net; 1417 } 1418 1419 snprintf(path, sizeof(path), "net/mpls/conf/%s", dev->name); 1420 1421 mdev->sysctl = register_net_sysctl(net, path, table); 1422 if (!mdev->sysctl) 1423 goto free; 1424 1425 mpls_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL, mdev); 1426 return 0; 1427 1428 free: 1429 kfree(table); 1430 out: 1431 return -ENOBUFS; 1432 } 1433 1434 static void mpls_dev_sysctl_unregister(struct net_device *dev, 1435 struct mpls_dev *mdev) 1436 { 1437 struct net *net = dev_net(dev); 1438 struct ctl_table *table; 1439 1440 table = mdev->sysctl->ctl_table_arg; 1441 unregister_net_sysctl_table(mdev->sysctl); 1442 kfree(table); 1443 1444 mpls_netconf_notify_devconf(net, RTM_DELNETCONF, 0, mdev); 1445 } 1446 1447 static struct mpls_dev *mpls_add_dev(struct net_device *dev) 1448 { 1449 struct mpls_dev *mdev; 1450 int err = -ENOMEM; 1451 int i; 1452 1453 ASSERT_RTNL(); 1454 1455 mdev = kzalloc(sizeof(*mdev), GFP_KERNEL); 1456 if (!mdev) 1457 return ERR_PTR(err); 1458 1459 mdev->stats = alloc_percpu(struct mpls_pcpu_stats); 1460 if (!mdev->stats) 1461 goto free; 1462 1463 for_each_possible_cpu(i) { 1464 struct mpls_pcpu_stats *mpls_stats; 1465 1466 mpls_stats = per_cpu_ptr(mdev->stats, i); 1467 u64_stats_init(&mpls_stats->syncp); 1468 } 1469 1470 mdev->dev = dev; 1471 1472 err = mpls_dev_sysctl_register(dev, mdev); 1473 if (err) 1474 goto free; 1475 1476 rcu_assign_pointer(dev->mpls_ptr, mdev); 1477 1478 return mdev; 1479 1480 free: 1481 free_percpu(mdev->stats); 1482 kfree(mdev); 1483 return ERR_PTR(err); 1484 } 1485 1486 static void mpls_dev_destroy_rcu(struct rcu_head *head) 1487 { 1488 struct mpls_dev *mdev = container_of(head, struct mpls_dev, rcu); 1489 1490 free_percpu(mdev->stats); 1491 kfree(mdev); 1492 } 1493 1494 static int mpls_ifdown(struct net_device *dev, int event) 1495 { 1496 struct mpls_route __rcu **platform_label; 1497 struct net *net = dev_net(dev); 1498 unsigned index; 1499 1500 platform_label = rtnl_dereference(net->mpls.platform_label); 1501 for (index = 0; index < net->mpls.platform_labels; index++) { 1502 struct mpls_route *rt = rtnl_dereference(platform_label[index]); 1503 bool nh_del = false; 1504 u8 alive = 0; 1505 1506 if (!rt) 1507 continue; 1508 1509 if (event == NETDEV_UNREGISTER) { 1510 u8 deleted = 0; 1511 1512 for_nexthops(rt) { 1513 if (!nh->nh_dev || nh->nh_dev == dev) 1514 deleted++; 1515 if (nh->nh_dev == dev) 1516 nh_del = true; 1517 } endfor_nexthops(rt); 1518 1519 /* if there are no more nexthops, delete the route */ 1520 if (deleted == rt->rt_nhn) { 1521 mpls_route_update(net, index, NULL, NULL); 1522 continue; 1523 } 1524 1525 if (nh_del) { 1526 size_t size = sizeof(*rt) + rt->rt_nhn * 1527 rt->rt_nh_size; 1528 struct mpls_route *orig = rt; 1529 1530 rt = kmalloc(size, GFP_KERNEL); 1531 if (!rt) 1532 return -ENOMEM; 1533 memcpy(rt, orig, size); 1534 } 1535 } 1536 1537 change_nexthops(rt) { 1538 unsigned int nh_flags = nh->nh_flags; 1539 1540 if (nh->nh_dev != dev) 1541 goto next; 1542 1543 switch (event) { 1544 case NETDEV_DOWN: 1545 case NETDEV_UNREGISTER: 1546 nh_flags |= RTNH_F_DEAD; 1547 fallthrough; 1548 case NETDEV_CHANGE: 1549 nh_flags |= RTNH_F_LINKDOWN; 1550 break; 1551 } 1552 if (event == NETDEV_UNREGISTER) 1553 nh->nh_dev = NULL; 1554 1555 if (nh->nh_flags != nh_flags) 1556 WRITE_ONCE(nh->nh_flags, nh_flags); 1557 next: 1558 if (!(nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN))) 1559 alive++; 1560 } endfor_nexthops(rt); 1561 1562 WRITE_ONCE(rt->rt_nhn_alive, alive); 1563 1564 if (nh_del) 1565 mpls_route_update(net, index, rt, NULL); 1566 } 1567 1568 return 0; 1569 } 1570 1571 static void mpls_ifup(struct net_device *dev, unsigned int flags) 1572 { 1573 struct mpls_route __rcu **platform_label; 1574 struct net *net = dev_net(dev); 1575 unsigned index; 1576 u8 alive; 1577 1578 platform_label = rtnl_dereference(net->mpls.platform_label); 1579 for (index = 0; index < net->mpls.platform_labels; index++) { 1580 struct mpls_route *rt = rtnl_dereference(platform_label[index]); 1581 1582 if (!rt) 1583 continue; 1584 1585 alive = 0; 1586 change_nexthops(rt) { 1587 unsigned int nh_flags = nh->nh_flags; 1588 1589 if (!(nh_flags & flags)) { 1590 alive++; 1591 continue; 1592 } 1593 if (nh->nh_dev != dev) 1594 continue; 1595 alive++; 1596 nh_flags &= ~flags; 1597 WRITE_ONCE(nh->nh_flags, nh_flags); 1598 } endfor_nexthops(rt); 1599 1600 WRITE_ONCE(rt->rt_nhn_alive, alive); 1601 } 1602 } 1603 1604 static int mpls_dev_notify(struct notifier_block *this, unsigned long event, 1605 void *ptr) 1606 { 1607 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1608 struct mpls_dev *mdev; 1609 unsigned int flags; 1610 int err; 1611 1612 if (event == NETDEV_REGISTER) { 1613 mdev = mpls_add_dev(dev); 1614 if (IS_ERR(mdev)) 1615 return notifier_from_errno(PTR_ERR(mdev)); 1616 1617 return NOTIFY_OK; 1618 } 1619 1620 mdev = mpls_dev_get(dev); 1621 if (!mdev) 1622 return NOTIFY_OK; 1623 1624 switch (event) { 1625 1626 case NETDEV_DOWN: 1627 err = mpls_ifdown(dev, event); 1628 if (err) 1629 return notifier_from_errno(err); 1630 break; 1631 case NETDEV_UP: 1632 flags = dev_get_flags(dev); 1633 if (flags & (IFF_RUNNING | IFF_LOWER_UP)) 1634 mpls_ifup(dev, RTNH_F_DEAD | RTNH_F_LINKDOWN); 1635 else 1636 mpls_ifup(dev, RTNH_F_DEAD); 1637 break; 1638 case NETDEV_CHANGE: 1639 flags = dev_get_flags(dev); 1640 if (flags & (IFF_RUNNING | IFF_LOWER_UP)) { 1641 mpls_ifup(dev, RTNH_F_DEAD | RTNH_F_LINKDOWN); 1642 } else { 1643 err = mpls_ifdown(dev, event); 1644 if (err) 1645 return notifier_from_errno(err); 1646 } 1647 break; 1648 case NETDEV_UNREGISTER: 1649 err = mpls_ifdown(dev, event); 1650 if (err) 1651 return notifier_from_errno(err); 1652 mdev = mpls_dev_get(dev); 1653 if (mdev) { 1654 mpls_dev_sysctl_unregister(dev, mdev); 1655 RCU_INIT_POINTER(dev->mpls_ptr, NULL); 1656 call_rcu(&mdev->rcu, mpls_dev_destroy_rcu); 1657 } 1658 break; 1659 case NETDEV_CHANGENAME: 1660 mdev = mpls_dev_get(dev); 1661 if (mdev) { 1662 mpls_dev_sysctl_unregister(dev, mdev); 1663 err = mpls_dev_sysctl_register(dev, mdev); 1664 if (err) 1665 return notifier_from_errno(err); 1666 } 1667 break; 1668 } 1669 return NOTIFY_OK; 1670 } 1671 1672 static struct notifier_block mpls_dev_notifier = { 1673 .notifier_call = mpls_dev_notify, 1674 }; 1675 1676 static int nla_put_via(struct sk_buff *skb, 1677 u8 table, const void *addr, int alen) 1678 { 1679 static const int table_to_family[NEIGH_NR_TABLES + 1] = { 1680 AF_INET, AF_INET6, AF_DECnet, AF_PACKET, 1681 }; 1682 struct nlattr *nla; 1683 struct rtvia *via; 1684 int family = AF_UNSPEC; 1685 1686 nla = nla_reserve(skb, RTA_VIA, alen + 2); 1687 if (!nla) 1688 return -EMSGSIZE; 1689 1690 if (table <= NEIGH_NR_TABLES) 1691 family = table_to_family[table]; 1692 1693 via = nla_data(nla); 1694 via->rtvia_family = family; 1695 memcpy(via->rtvia_addr, addr, alen); 1696 return 0; 1697 } 1698 1699 int nla_put_labels(struct sk_buff *skb, int attrtype, 1700 u8 labels, const u32 label[]) 1701 { 1702 struct nlattr *nla; 1703 struct mpls_shim_hdr *nla_label; 1704 bool bos; 1705 int i; 1706 nla = nla_reserve(skb, attrtype, labels*4); 1707 if (!nla) 1708 return -EMSGSIZE; 1709 1710 nla_label = nla_data(nla); 1711 bos = true; 1712 for (i = labels - 1; i >= 0; i--) { 1713 nla_label[i] = mpls_entry_encode(label[i], 0, 0, bos); 1714 bos = false; 1715 } 1716 1717 return 0; 1718 } 1719 EXPORT_SYMBOL_GPL(nla_put_labels); 1720 1721 int nla_get_labels(const struct nlattr *nla, u8 max_labels, u8 *labels, 1722 u32 label[], struct netlink_ext_ack *extack) 1723 { 1724 unsigned len = nla_len(nla); 1725 struct mpls_shim_hdr *nla_label; 1726 u8 nla_labels; 1727 bool bos; 1728 int i; 1729 1730 /* len needs to be an even multiple of 4 (the label size). Number 1731 * of labels is a u8 so check for overflow. 1732 */ 1733 if (len & 3 || len / 4 > 255) { 1734 NL_SET_ERR_MSG_ATTR(extack, nla, 1735 "Invalid length for labels attribute"); 1736 return -EINVAL; 1737 } 1738 1739 /* Limit the number of new labels allowed */ 1740 nla_labels = len/4; 1741 if (nla_labels > max_labels) { 1742 NL_SET_ERR_MSG(extack, "Too many labels"); 1743 return -EINVAL; 1744 } 1745 1746 /* when label == NULL, caller wants number of labels */ 1747 if (!label) 1748 goto out; 1749 1750 nla_label = nla_data(nla); 1751 bos = true; 1752 for (i = nla_labels - 1; i >= 0; i--, bos = false) { 1753 struct mpls_entry_decoded dec; 1754 dec = mpls_entry_decode(nla_label + i); 1755 1756 /* Ensure the bottom of stack flag is properly set 1757 * and ttl and tc are both clear. 1758 */ 1759 if (dec.ttl) { 1760 NL_SET_ERR_MSG_ATTR(extack, nla, 1761 "TTL in label must be 0"); 1762 return -EINVAL; 1763 } 1764 1765 if (dec.tc) { 1766 NL_SET_ERR_MSG_ATTR(extack, nla, 1767 "Traffic class in label must be 0"); 1768 return -EINVAL; 1769 } 1770 1771 if (dec.bos != bos) { 1772 NL_SET_BAD_ATTR(extack, nla); 1773 if (bos) { 1774 NL_SET_ERR_MSG(extack, 1775 "BOS bit must be set in first label"); 1776 } else { 1777 NL_SET_ERR_MSG(extack, 1778 "BOS bit can only be set in first label"); 1779 } 1780 return -EINVAL; 1781 } 1782 1783 switch (dec.label) { 1784 case MPLS_LABEL_IMPLNULL: 1785 /* RFC3032: This is a label that an LSR may 1786 * assign and distribute, but which never 1787 * actually appears in the encapsulation. 1788 */ 1789 NL_SET_ERR_MSG_ATTR(extack, nla, 1790 "Implicit NULL Label (3) can not be used in encapsulation"); 1791 return -EINVAL; 1792 } 1793 1794 label[i] = dec.label; 1795 } 1796 out: 1797 *labels = nla_labels; 1798 return 0; 1799 } 1800 EXPORT_SYMBOL_GPL(nla_get_labels); 1801 1802 static int rtm_to_route_config(struct sk_buff *skb, 1803 struct nlmsghdr *nlh, 1804 struct mpls_route_config *cfg, 1805 struct netlink_ext_ack *extack) 1806 { 1807 struct rtmsg *rtm; 1808 struct nlattr *tb[RTA_MAX+1]; 1809 int index; 1810 int err; 1811 1812 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX, 1813 rtm_mpls_policy, extack); 1814 if (err < 0) 1815 goto errout; 1816 1817 err = -EINVAL; 1818 rtm = nlmsg_data(nlh); 1819 1820 if (rtm->rtm_family != AF_MPLS) { 1821 NL_SET_ERR_MSG(extack, "Invalid address family in rtmsg"); 1822 goto errout; 1823 } 1824 if (rtm->rtm_dst_len != 20) { 1825 NL_SET_ERR_MSG(extack, "rtm_dst_len must be 20 for MPLS"); 1826 goto errout; 1827 } 1828 if (rtm->rtm_src_len != 0) { 1829 NL_SET_ERR_MSG(extack, "rtm_src_len must be 0 for MPLS"); 1830 goto errout; 1831 } 1832 if (rtm->rtm_tos != 0) { 1833 NL_SET_ERR_MSG(extack, "rtm_tos must be 0 for MPLS"); 1834 goto errout; 1835 } 1836 if (rtm->rtm_table != RT_TABLE_MAIN) { 1837 NL_SET_ERR_MSG(extack, 1838 "MPLS only supports the main route table"); 1839 goto errout; 1840 } 1841 /* Any value is acceptable for rtm_protocol */ 1842 1843 /* As mpls uses destination specific addresses 1844 * (or source specific address in the case of multicast) 1845 * all addresses have universal scope. 1846 */ 1847 if (rtm->rtm_scope != RT_SCOPE_UNIVERSE) { 1848 NL_SET_ERR_MSG(extack, 1849 "Invalid route scope - MPLS only supports UNIVERSE"); 1850 goto errout; 1851 } 1852 if (rtm->rtm_type != RTN_UNICAST) { 1853 NL_SET_ERR_MSG(extack, 1854 "Invalid route type - MPLS only supports UNICAST"); 1855 goto errout; 1856 } 1857 if (rtm->rtm_flags != 0) { 1858 NL_SET_ERR_MSG(extack, "rtm_flags must be 0 for MPLS"); 1859 goto errout; 1860 } 1861 1862 cfg->rc_label = LABEL_NOT_SPECIFIED; 1863 cfg->rc_protocol = rtm->rtm_protocol; 1864 cfg->rc_via_table = MPLS_NEIGH_TABLE_UNSPEC; 1865 cfg->rc_ttl_propagate = MPLS_TTL_PROP_DEFAULT; 1866 cfg->rc_nlflags = nlh->nlmsg_flags; 1867 cfg->rc_nlinfo.portid = NETLINK_CB(skb).portid; 1868 cfg->rc_nlinfo.nlh = nlh; 1869 cfg->rc_nlinfo.nl_net = sock_net(skb->sk); 1870 1871 for (index = 0; index <= RTA_MAX; index++) { 1872 struct nlattr *nla = tb[index]; 1873 if (!nla) 1874 continue; 1875 1876 switch (index) { 1877 case RTA_OIF: 1878 cfg->rc_ifindex = nla_get_u32(nla); 1879 break; 1880 case RTA_NEWDST: 1881 if (nla_get_labels(nla, MAX_NEW_LABELS, 1882 &cfg->rc_output_labels, 1883 cfg->rc_output_label, extack)) 1884 goto errout; 1885 break; 1886 case RTA_DST: 1887 { 1888 u8 label_count; 1889 if (nla_get_labels(nla, 1, &label_count, 1890 &cfg->rc_label, extack)) 1891 goto errout; 1892 1893 if (!mpls_label_ok(cfg->rc_nlinfo.nl_net, 1894 &cfg->rc_label, extack)) 1895 goto errout; 1896 break; 1897 } 1898 case RTA_GATEWAY: 1899 NL_SET_ERR_MSG(extack, "MPLS does not support RTA_GATEWAY attribute"); 1900 goto errout; 1901 case RTA_VIA: 1902 { 1903 if (nla_get_via(nla, &cfg->rc_via_alen, 1904 &cfg->rc_via_table, cfg->rc_via, 1905 extack)) 1906 goto errout; 1907 break; 1908 } 1909 case RTA_MULTIPATH: 1910 { 1911 cfg->rc_mp = nla_data(nla); 1912 cfg->rc_mp_len = nla_len(nla); 1913 break; 1914 } 1915 case RTA_TTL_PROPAGATE: 1916 { 1917 u8 ttl_propagate = nla_get_u8(nla); 1918 1919 if (ttl_propagate > 1) { 1920 NL_SET_ERR_MSG_ATTR(extack, nla, 1921 "RTA_TTL_PROPAGATE can only be 0 or 1"); 1922 goto errout; 1923 } 1924 cfg->rc_ttl_propagate = ttl_propagate ? 1925 MPLS_TTL_PROP_ENABLED : 1926 MPLS_TTL_PROP_DISABLED; 1927 break; 1928 } 1929 default: 1930 NL_SET_ERR_MSG_ATTR(extack, nla, "Unknown attribute"); 1931 /* Unsupported attribute */ 1932 goto errout; 1933 } 1934 } 1935 1936 err = 0; 1937 errout: 1938 return err; 1939 } 1940 1941 static int mpls_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh, 1942 struct netlink_ext_ack *extack) 1943 { 1944 struct mpls_route_config *cfg; 1945 int err; 1946 1947 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL); 1948 if (!cfg) 1949 return -ENOMEM; 1950 1951 err = rtm_to_route_config(skb, nlh, cfg, extack); 1952 if (err < 0) 1953 goto out; 1954 1955 err = mpls_route_del(cfg, extack); 1956 out: 1957 kfree(cfg); 1958 1959 return err; 1960 } 1961 1962 1963 static int mpls_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh, 1964 struct netlink_ext_ack *extack) 1965 { 1966 struct mpls_route_config *cfg; 1967 int err; 1968 1969 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL); 1970 if (!cfg) 1971 return -ENOMEM; 1972 1973 err = rtm_to_route_config(skb, nlh, cfg, extack); 1974 if (err < 0) 1975 goto out; 1976 1977 err = mpls_route_add(cfg, extack); 1978 out: 1979 kfree(cfg); 1980 1981 return err; 1982 } 1983 1984 static int mpls_dump_route(struct sk_buff *skb, u32 portid, u32 seq, int event, 1985 u32 label, struct mpls_route *rt, int flags) 1986 { 1987 struct net_device *dev; 1988 struct nlmsghdr *nlh; 1989 struct rtmsg *rtm; 1990 1991 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*rtm), flags); 1992 if (nlh == NULL) 1993 return -EMSGSIZE; 1994 1995 rtm = nlmsg_data(nlh); 1996 rtm->rtm_family = AF_MPLS; 1997 rtm->rtm_dst_len = 20; 1998 rtm->rtm_src_len = 0; 1999 rtm->rtm_tos = 0; 2000 rtm->rtm_table = RT_TABLE_MAIN; 2001 rtm->rtm_protocol = rt->rt_protocol; 2002 rtm->rtm_scope = RT_SCOPE_UNIVERSE; 2003 rtm->rtm_type = RTN_UNICAST; 2004 rtm->rtm_flags = 0; 2005 2006 if (nla_put_labels(skb, RTA_DST, 1, &label)) 2007 goto nla_put_failure; 2008 2009 if (rt->rt_ttl_propagate != MPLS_TTL_PROP_DEFAULT) { 2010 bool ttl_propagate = 2011 rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED; 2012 2013 if (nla_put_u8(skb, RTA_TTL_PROPAGATE, 2014 ttl_propagate)) 2015 goto nla_put_failure; 2016 } 2017 if (rt->rt_nhn == 1) { 2018 const struct mpls_nh *nh = rt->rt_nh; 2019 2020 if (nh->nh_labels && 2021 nla_put_labels(skb, RTA_NEWDST, nh->nh_labels, 2022 nh->nh_label)) 2023 goto nla_put_failure; 2024 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC && 2025 nla_put_via(skb, nh->nh_via_table, mpls_nh_via(rt, nh), 2026 nh->nh_via_alen)) 2027 goto nla_put_failure; 2028 dev = nh->nh_dev; 2029 if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex)) 2030 goto nla_put_failure; 2031 if (nh->nh_flags & RTNH_F_LINKDOWN) 2032 rtm->rtm_flags |= RTNH_F_LINKDOWN; 2033 if (nh->nh_flags & RTNH_F_DEAD) 2034 rtm->rtm_flags |= RTNH_F_DEAD; 2035 } else { 2036 struct rtnexthop *rtnh; 2037 struct nlattr *mp; 2038 u8 linkdown = 0; 2039 u8 dead = 0; 2040 2041 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH); 2042 if (!mp) 2043 goto nla_put_failure; 2044 2045 for_nexthops(rt) { 2046 dev = nh->nh_dev; 2047 if (!dev) 2048 continue; 2049 2050 rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh)); 2051 if (!rtnh) 2052 goto nla_put_failure; 2053 2054 rtnh->rtnh_ifindex = dev->ifindex; 2055 if (nh->nh_flags & RTNH_F_LINKDOWN) { 2056 rtnh->rtnh_flags |= RTNH_F_LINKDOWN; 2057 linkdown++; 2058 } 2059 if (nh->nh_flags & RTNH_F_DEAD) { 2060 rtnh->rtnh_flags |= RTNH_F_DEAD; 2061 dead++; 2062 } 2063 2064 if (nh->nh_labels && nla_put_labels(skb, RTA_NEWDST, 2065 nh->nh_labels, 2066 nh->nh_label)) 2067 goto nla_put_failure; 2068 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC && 2069 nla_put_via(skb, nh->nh_via_table, 2070 mpls_nh_via(rt, nh), 2071 nh->nh_via_alen)) 2072 goto nla_put_failure; 2073 2074 /* length of rtnetlink header + attributes */ 2075 rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh; 2076 } endfor_nexthops(rt); 2077 2078 if (linkdown == rt->rt_nhn) 2079 rtm->rtm_flags |= RTNH_F_LINKDOWN; 2080 if (dead == rt->rt_nhn) 2081 rtm->rtm_flags |= RTNH_F_DEAD; 2082 2083 nla_nest_end(skb, mp); 2084 } 2085 2086 nlmsg_end(skb, nlh); 2087 return 0; 2088 2089 nla_put_failure: 2090 nlmsg_cancel(skb, nlh); 2091 return -EMSGSIZE; 2092 } 2093 2094 #if IS_ENABLED(CONFIG_INET) 2095 static int mpls_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh, 2096 struct fib_dump_filter *filter, 2097 struct netlink_callback *cb) 2098 { 2099 return ip_valid_fib_dump_req(net, nlh, filter, cb); 2100 } 2101 #else 2102 static int mpls_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh, 2103 struct fib_dump_filter *filter, 2104 struct netlink_callback *cb) 2105 { 2106 struct netlink_ext_ack *extack = cb->extack; 2107 struct nlattr *tb[RTA_MAX + 1]; 2108 struct rtmsg *rtm; 2109 int err, i; 2110 2111 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) { 2112 NL_SET_ERR_MSG_MOD(extack, "Invalid header for FIB dump request"); 2113 return -EINVAL; 2114 } 2115 2116 rtm = nlmsg_data(nlh); 2117 if (rtm->rtm_dst_len || rtm->rtm_src_len || rtm->rtm_tos || 2118 rtm->rtm_table || rtm->rtm_scope || rtm->rtm_type || 2119 rtm->rtm_flags) { 2120 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for FIB dump request"); 2121 return -EINVAL; 2122 } 2123 2124 if (rtm->rtm_protocol) { 2125 filter->protocol = rtm->rtm_protocol; 2126 filter->filter_set = 1; 2127 cb->answer_flags = NLM_F_DUMP_FILTERED; 2128 } 2129 2130 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX, 2131 rtm_mpls_policy, extack); 2132 if (err < 0) 2133 return err; 2134 2135 for (i = 0; i <= RTA_MAX; ++i) { 2136 int ifindex; 2137 2138 if (i == RTA_OIF) { 2139 ifindex = nla_get_u32(tb[i]); 2140 filter->dev = __dev_get_by_index(net, ifindex); 2141 if (!filter->dev) 2142 return -ENODEV; 2143 filter->filter_set = 1; 2144 } else if (tb[i]) { 2145 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request"); 2146 return -EINVAL; 2147 } 2148 } 2149 2150 return 0; 2151 } 2152 #endif 2153 2154 static bool mpls_rt_uses_dev(struct mpls_route *rt, 2155 const struct net_device *dev) 2156 { 2157 if (rt->rt_nhn == 1) { 2158 struct mpls_nh *nh = rt->rt_nh; 2159 2160 if (nh->nh_dev == dev) 2161 return true; 2162 } else { 2163 for_nexthops(rt) { 2164 if (nh->nh_dev == dev) 2165 return true; 2166 } endfor_nexthops(rt); 2167 } 2168 2169 return false; 2170 } 2171 2172 static int mpls_dump_routes(struct sk_buff *skb, struct netlink_callback *cb) 2173 { 2174 const struct nlmsghdr *nlh = cb->nlh; 2175 struct net *net = sock_net(skb->sk); 2176 struct mpls_route __rcu **platform_label; 2177 struct fib_dump_filter filter = {}; 2178 unsigned int flags = NLM_F_MULTI; 2179 size_t platform_labels; 2180 unsigned int index; 2181 2182 ASSERT_RTNL(); 2183 2184 if (cb->strict_check) { 2185 int err; 2186 2187 err = mpls_valid_fib_dump_req(net, nlh, &filter, cb); 2188 if (err < 0) 2189 return err; 2190 2191 /* for MPLS, there is only 1 table with fixed type and flags. 2192 * If either are set in the filter then return nothing. 2193 */ 2194 if ((filter.table_id && filter.table_id != RT_TABLE_MAIN) || 2195 (filter.rt_type && filter.rt_type != RTN_UNICAST) || 2196 filter.flags) 2197 return skb->len; 2198 } 2199 2200 index = cb->args[0]; 2201 if (index < MPLS_LABEL_FIRST_UNRESERVED) 2202 index = MPLS_LABEL_FIRST_UNRESERVED; 2203 2204 platform_label = rtnl_dereference(net->mpls.platform_label); 2205 platform_labels = net->mpls.platform_labels; 2206 2207 if (filter.filter_set) 2208 flags |= NLM_F_DUMP_FILTERED; 2209 2210 for (; index < platform_labels; index++) { 2211 struct mpls_route *rt; 2212 2213 rt = rtnl_dereference(platform_label[index]); 2214 if (!rt) 2215 continue; 2216 2217 if ((filter.dev && !mpls_rt_uses_dev(rt, filter.dev)) || 2218 (filter.protocol && rt->rt_protocol != filter.protocol)) 2219 continue; 2220 2221 if (mpls_dump_route(skb, NETLINK_CB(cb->skb).portid, 2222 cb->nlh->nlmsg_seq, RTM_NEWROUTE, 2223 index, rt, flags) < 0) 2224 break; 2225 } 2226 cb->args[0] = index; 2227 2228 return skb->len; 2229 } 2230 2231 static inline size_t lfib_nlmsg_size(struct mpls_route *rt) 2232 { 2233 size_t payload = 2234 NLMSG_ALIGN(sizeof(struct rtmsg)) 2235 + nla_total_size(4) /* RTA_DST */ 2236 + nla_total_size(1); /* RTA_TTL_PROPAGATE */ 2237 2238 if (rt->rt_nhn == 1) { 2239 struct mpls_nh *nh = rt->rt_nh; 2240 2241 if (nh->nh_dev) 2242 payload += nla_total_size(4); /* RTA_OIF */ 2243 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC) /* RTA_VIA */ 2244 payload += nla_total_size(2 + nh->nh_via_alen); 2245 if (nh->nh_labels) /* RTA_NEWDST */ 2246 payload += nla_total_size(nh->nh_labels * 4); 2247 } else { 2248 /* each nexthop is packed in an attribute */ 2249 size_t nhsize = 0; 2250 2251 for_nexthops(rt) { 2252 if (!nh->nh_dev) 2253 continue; 2254 nhsize += nla_total_size(sizeof(struct rtnexthop)); 2255 /* RTA_VIA */ 2256 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC) 2257 nhsize += nla_total_size(2 + nh->nh_via_alen); 2258 if (nh->nh_labels) 2259 nhsize += nla_total_size(nh->nh_labels * 4); 2260 } endfor_nexthops(rt); 2261 /* nested attribute */ 2262 payload += nla_total_size(nhsize); 2263 } 2264 2265 return payload; 2266 } 2267 2268 static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt, 2269 struct nlmsghdr *nlh, struct net *net, u32 portid, 2270 unsigned int nlm_flags) 2271 { 2272 struct sk_buff *skb; 2273 u32 seq = nlh ? nlh->nlmsg_seq : 0; 2274 int err = -ENOBUFS; 2275 2276 skb = nlmsg_new(lfib_nlmsg_size(rt), GFP_KERNEL); 2277 if (skb == NULL) 2278 goto errout; 2279 2280 err = mpls_dump_route(skb, portid, seq, event, label, rt, nlm_flags); 2281 if (err < 0) { 2282 /* -EMSGSIZE implies BUG in lfib_nlmsg_size */ 2283 WARN_ON(err == -EMSGSIZE); 2284 kfree_skb(skb); 2285 goto errout; 2286 } 2287 rtnl_notify(skb, net, portid, RTNLGRP_MPLS_ROUTE, nlh, GFP_KERNEL); 2288 2289 return; 2290 errout: 2291 if (err < 0) 2292 rtnl_set_sk_err(net, RTNLGRP_MPLS_ROUTE, err); 2293 } 2294 2295 static int mpls_valid_getroute_req(struct sk_buff *skb, 2296 const struct nlmsghdr *nlh, 2297 struct nlattr **tb, 2298 struct netlink_ext_ack *extack) 2299 { 2300 struct rtmsg *rtm; 2301 int i, err; 2302 2303 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) { 2304 NL_SET_ERR_MSG_MOD(extack, 2305 "Invalid header for get route request"); 2306 return -EINVAL; 2307 } 2308 2309 if (!netlink_strict_get_check(skb)) 2310 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX, 2311 rtm_mpls_policy, extack); 2312 2313 rtm = nlmsg_data(nlh); 2314 if ((rtm->rtm_dst_len && rtm->rtm_dst_len != 20) || 2315 rtm->rtm_src_len || rtm->rtm_tos || rtm->rtm_table || 2316 rtm->rtm_protocol || rtm->rtm_scope || rtm->rtm_type) { 2317 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request"); 2318 return -EINVAL; 2319 } 2320 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) { 2321 NL_SET_ERR_MSG_MOD(extack, 2322 "Invalid flags for get route request"); 2323 return -EINVAL; 2324 } 2325 2326 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX, 2327 rtm_mpls_policy, extack); 2328 if (err) 2329 return err; 2330 2331 if ((tb[RTA_DST] || tb[RTA_NEWDST]) && !rtm->rtm_dst_len) { 2332 NL_SET_ERR_MSG_MOD(extack, "rtm_dst_len must be 20 for MPLS"); 2333 return -EINVAL; 2334 } 2335 2336 for (i = 0; i <= RTA_MAX; i++) { 2337 if (!tb[i]) 2338 continue; 2339 2340 switch (i) { 2341 case RTA_DST: 2342 case RTA_NEWDST: 2343 break; 2344 default: 2345 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request"); 2346 return -EINVAL; 2347 } 2348 } 2349 2350 return 0; 2351 } 2352 2353 static int mpls_getroute(struct sk_buff *in_skb, struct nlmsghdr *in_nlh, 2354 struct netlink_ext_ack *extack) 2355 { 2356 struct net *net = sock_net(in_skb->sk); 2357 u32 portid = NETLINK_CB(in_skb).portid; 2358 u32 in_label = LABEL_NOT_SPECIFIED; 2359 struct nlattr *tb[RTA_MAX + 1]; 2360 u32 labels[MAX_NEW_LABELS]; 2361 struct mpls_shim_hdr *hdr; 2362 unsigned int hdr_size = 0; 2363 const struct mpls_nh *nh; 2364 struct net_device *dev; 2365 struct mpls_route *rt; 2366 struct rtmsg *rtm, *r; 2367 struct nlmsghdr *nlh; 2368 struct sk_buff *skb; 2369 u8 n_labels; 2370 int err; 2371 2372 err = mpls_valid_getroute_req(in_skb, in_nlh, tb, extack); 2373 if (err < 0) 2374 goto errout; 2375 2376 rtm = nlmsg_data(in_nlh); 2377 2378 if (tb[RTA_DST]) { 2379 u8 label_count; 2380 2381 if (nla_get_labels(tb[RTA_DST], 1, &label_count, 2382 &in_label, extack)) { 2383 err = -EINVAL; 2384 goto errout; 2385 } 2386 2387 if (!mpls_label_ok(net, &in_label, extack)) { 2388 err = -EINVAL; 2389 goto errout; 2390 } 2391 } 2392 2393 rt = mpls_route_input_rcu(net, in_label); 2394 if (!rt) { 2395 err = -ENETUNREACH; 2396 goto errout; 2397 } 2398 2399 if (rtm->rtm_flags & RTM_F_FIB_MATCH) { 2400 skb = nlmsg_new(lfib_nlmsg_size(rt), GFP_KERNEL); 2401 if (!skb) { 2402 err = -ENOBUFS; 2403 goto errout; 2404 } 2405 2406 err = mpls_dump_route(skb, portid, in_nlh->nlmsg_seq, 2407 RTM_NEWROUTE, in_label, rt, 0); 2408 if (err < 0) { 2409 /* -EMSGSIZE implies BUG in lfib_nlmsg_size */ 2410 WARN_ON(err == -EMSGSIZE); 2411 goto errout_free; 2412 } 2413 2414 return rtnl_unicast(skb, net, portid); 2415 } 2416 2417 if (tb[RTA_NEWDST]) { 2418 if (nla_get_labels(tb[RTA_NEWDST], MAX_NEW_LABELS, &n_labels, 2419 labels, extack) != 0) { 2420 err = -EINVAL; 2421 goto errout; 2422 } 2423 2424 hdr_size = n_labels * sizeof(struct mpls_shim_hdr); 2425 } 2426 2427 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL); 2428 if (!skb) { 2429 err = -ENOBUFS; 2430 goto errout; 2431 } 2432 2433 skb->protocol = htons(ETH_P_MPLS_UC); 2434 2435 if (hdr_size) { 2436 bool bos; 2437 int i; 2438 2439 if (skb_cow(skb, hdr_size)) { 2440 err = -ENOBUFS; 2441 goto errout_free; 2442 } 2443 2444 skb_reserve(skb, hdr_size); 2445 skb_push(skb, hdr_size); 2446 skb_reset_network_header(skb); 2447 2448 /* Push new labels */ 2449 hdr = mpls_hdr(skb); 2450 bos = true; 2451 for (i = n_labels - 1; i >= 0; i--) { 2452 hdr[i] = mpls_entry_encode(labels[i], 2453 1, 0, bos); 2454 bos = false; 2455 } 2456 } 2457 2458 nh = mpls_select_multipath(rt, skb); 2459 if (!nh) { 2460 err = -ENETUNREACH; 2461 goto errout_free; 2462 } 2463 2464 if (hdr_size) { 2465 skb_pull(skb, hdr_size); 2466 skb_reset_network_header(skb); 2467 } 2468 2469 nlh = nlmsg_put(skb, portid, in_nlh->nlmsg_seq, 2470 RTM_NEWROUTE, sizeof(*r), 0); 2471 if (!nlh) { 2472 err = -EMSGSIZE; 2473 goto errout_free; 2474 } 2475 2476 r = nlmsg_data(nlh); 2477 r->rtm_family = AF_MPLS; 2478 r->rtm_dst_len = 20; 2479 r->rtm_src_len = 0; 2480 r->rtm_table = RT_TABLE_MAIN; 2481 r->rtm_type = RTN_UNICAST; 2482 r->rtm_scope = RT_SCOPE_UNIVERSE; 2483 r->rtm_protocol = rt->rt_protocol; 2484 r->rtm_flags = 0; 2485 2486 if (nla_put_labels(skb, RTA_DST, 1, &in_label)) 2487 goto nla_put_failure; 2488 2489 if (nh->nh_labels && 2490 nla_put_labels(skb, RTA_NEWDST, nh->nh_labels, 2491 nh->nh_label)) 2492 goto nla_put_failure; 2493 2494 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC && 2495 nla_put_via(skb, nh->nh_via_table, mpls_nh_via(rt, nh), 2496 nh->nh_via_alen)) 2497 goto nla_put_failure; 2498 dev = nh->nh_dev; 2499 if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex)) 2500 goto nla_put_failure; 2501 2502 nlmsg_end(skb, nlh); 2503 2504 err = rtnl_unicast(skb, net, portid); 2505 errout: 2506 return err; 2507 2508 nla_put_failure: 2509 nlmsg_cancel(skb, nlh); 2510 err = -EMSGSIZE; 2511 errout_free: 2512 kfree_skb(skb); 2513 return err; 2514 } 2515 2516 static int resize_platform_label_table(struct net *net, size_t limit) 2517 { 2518 size_t size = sizeof(struct mpls_route *) * limit; 2519 size_t old_limit; 2520 size_t cp_size; 2521 struct mpls_route __rcu **labels = NULL, **old; 2522 struct mpls_route *rt0 = NULL, *rt2 = NULL; 2523 unsigned index; 2524 2525 if (size) { 2526 labels = kvzalloc(size, GFP_KERNEL); 2527 if (!labels) 2528 goto nolabels; 2529 } 2530 2531 /* In case the predefined labels need to be populated */ 2532 if (limit > MPLS_LABEL_IPV4NULL) { 2533 struct net_device *lo = net->loopback_dev; 2534 rt0 = mpls_rt_alloc(1, lo->addr_len, 0); 2535 if (IS_ERR(rt0)) 2536 goto nort0; 2537 rt0->rt_nh->nh_dev = lo; 2538 rt0->rt_protocol = RTPROT_KERNEL; 2539 rt0->rt_payload_type = MPT_IPV4; 2540 rt0->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT; 2541 rt0->rt_nh->nh_via_table = NEIGH_LINK_TABLE; 2542 rt0->rt_nh->nh_via_alen = lo->addr_len; 2543 memcpy(__mpls_nh_via(rt0, rt0->rt_nh), lo->dev_addr, 2544 lo->addr_len); 2545 } 2546 if (limit > MPLS_LABEL_IPV6NULL) { 2547 struct net_device *lo = net->loopback_dev; 2548 rt2 = mpls_rt_alloc(1, lo->addr_len, 0); 2549 if (IS_ERR(rt2)) 2550 goto nort2; 2551 rt2->rt_nh->nh_dev = lo; 2552 rt2->rt_protocol = RTPROT_KERNEL; 2553 rt2->rt_payload_type = MPT_IPV6; 2554 rt2->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT; 2555 rt2->rt_nh->nh_via_table = NEIGH_LINK_TABLE; 2556 rt2->rt_nh->nh_via_alen = lo->addr_len; 2557 memcpy(__mpls_nh_via(rt2, rt2->rt_nh), lo->dev_addr, 2558 lo->addr_len); 2559 } 2560 2561 rtnl_lock(); 2562 /* Remember the original table */ 2563 old = rtnl_dereference(net->mpls.platform_label); 2564 old_limit = net->mpls.platform_labels; 2565 2566 /* Free any labels beyond the new table */ 2567 for (index = limit; index < old_limit; index++) 2568 mpls_route_update(net, index, NULL, NULL); 2569 2570 /* Copy over the old labels */ 2571 cp_size = size; 2572 if (old_limit < limit) 2573 cp_size = old_limit * sizeof(struct mpls_route *); 2574 2575 memcpy(labels, old, cp_size); 2576 2577 /* If needed set the predefined labels */ 2578 if ((old_limit <= MPLS_LABEL_IPV6NULL) && 2579 (limit > MPLS_LABEL_IPV6NULL)) { 2580 RCU_INIT_POINTER(labels[MPLS_LABEL_IPV6NULL], rt2); 2581 rt2 = NULL; 2582 } 2583 2584 if ((old_limit <= MPLS_LABEL_IPV4NULL) && 2585 (limit > MPLS_LABEL_IPV4NULL)) { 2586 RCU_INIT_POINTER(labels[MPLS_LABEL_IPV4NULL], rt0); 2587 rt0 = NULL; 2588 } 2589 2590 /* Update the global pointers */ 2591 net->mpls.platform_labels = limit; 2592 rcu_assign_pointer(net->mpls.platform_label, labels); 2593 2594 rtnl_unlock(); 2595 2596 mpls_rt_free(rt2); 2597 mpls_rt_free(rt0); 2598 2599 if (old) { 2600 synchronize_rcu(); 2601 kvfree(old); 2602 } 2603 return 0; 2604 2605 nort2: 2606 mpls_rt_free(rt0); 2607 nort0: 2608 kvfree(labels); 2609 nolabels: 2610 return -ENOMEM; 2611 } 2612 2613 static int mpls_platform_labels(struct ctl_table *table, int write, 2614 void *buffer, size_t *lenp, loff_t *ppos) 2615 { 2616 struct net *net = table->data; 2617 int platform_labels = net->mpls.platform_labels; 2618 int ret; 2619 struct ctl_table tmp = { 2620 .procname = table->procname, 2621 .data = &platform_labels, 2622 .maxlen = sizeof(int), 2623 .mode = table->mode, 2624 .extra1 = SYSCTL_ZERO, 2625 .extra2 = &label_limit, 2626 }; 2627 2628 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos); 2629 2630 if (write && ret == 0) 2631 ret = resize_platform_label_table(net, platform_labels); 2632 2633 return ret; 2634 } 2635 2636 #define MPLS_NS_SYSCTL_OFFSET(field) \ 2637 (&((struct net *)0)->field) 2638 2639 static const struct ctl_table mpls_table[] = { 2640 { 2641 .procname = "platform_labels", 2642 .data = NULL, 2643 .maxlen = sizeof(int), 2644 .mode = 0644, 2645 .proc_handler = mpls_platform_labels, 2646 }, 2647 { 2648 .procname = "ip_ttl_propagate", 2649 .data = MPLS_NS_SYSCTL_OFFSET(mpls.ip_ttl_propagate), 2650 .maxlen = sizeof(int), 2651 .mode = 0644, 2652 .proc_handler = proc_dointvec_minmax, 2653 .extra1 = SYSCTL_ZERO, 2654 .extra2 = SYSCTL_ONE, 2655 }, 2656 { 2657 .procname = "default_ttl", 2658 .data = MPLS_NS_SYSCTL_OFFSET(mpls.default_ttl), 2659 .maxlen = sizeof(int), 2660 .mode = 0644, 2661 .proc_handler = proc_dointvec_minmax, 2662 .extra1 = SYSCTL_ONE, 2663 .extra2 = &ttl_max, 2664 }, 2665 { } 2666 }; 2667 2668 static int mpls_net_init(struct net *net) 2669 { 2670 struct ctl_table *table; 2671 int i; 2672 2673 net->mpls.platform_labels = 0; 2674 net->mpls.platform_label = NULL; 2675 net->mpls.ip_ttl_propagate = 1; 2676 net->mpls.default_ttl = 255; 2677 2678 table = kmemdup(mpls_table, sizeof(mpls_table), GFP_KERNEL); 2679 if (table == NULL) 2680 return -ENOMEM; 2681 2682 /* Table data contains only offsets relative to the base of 2683 * the mdev at this point, so make them absolute. 2684 */ 2685 for (i = 0; i < ARRAY_SIZE(mpls_table) - 1; i++) 2686 table[i].data = (char *)net + (uintptr_t)table[i].data; 2687 2688 net->mpls.ctl = register_net_sysctl(net, "net/mpls", table); 2689 if (net->mpls.ctl == NULL) { 2690 kfree(table); 2691 return -ENOMEM; 2692 } 2693 2694 return 0; 2695 } 2696 2697 static void mpls_net_exit(struct net *net) 2698 { 2699 struct mpls_route __rcu **platform_label; 2700 size_t platform_labels; 2701 struct ctl_table *table; 2702 unsigned int index; 2703 2704 table = net->mpls.ctl->ctl_table_arg; 2705 unregister_net_sysctl_table(net->mpls.ctl); 2706 kfree(table); 2707 2708 /* An rcu grace period has passed since there was a device in 2709 * the network namespace (and thus the last in flight packet) 2710 * left this network namespace. This is because 2711 * unregister_netdevice_many and netdev_run_todo has completed 2712 * for each network device that was in this network namespace. 2713 * 2714 * As such no additional rcu synchronization is necessary when 2715 * freeing the platform_label table. 2716 */ 2717 rtnl_lock(); 2718 platform_label = rtnl_dereference(net->mpls.platform_label); 2719 platform_labels = net->mpls.platform_labels; 2720 for (index = 0; index < platform_labels; index++) { 2721 struct mpls_route *rt = rtnl_dereference(platform_label[index]); 2722 RCU_INIT_POINTER(platform_label[index], NULL); 2723 mpls_notify_route(net, index, rt, NULL, NULL); 2724 mpls_rt_free(rt); 2725 } 2726 rtnl_unlock(); 2727 2728 kvfree(platform_label); 2729 } 2730 2731 static struct pernet_operations mpls_net_ops = { 2732 .init = mpls_net_init, 2733 .exit = mpls_net_exit, 2734 }; 2735 2736 static struct rtnl_af_ops mpls_af_ops __read_mostly = { 2737 .family = AF_MPLS, 2738 .fill_stats_af = mpls_fill_stats_af, 2739 .get_stats_af_size = mpls_get_stats_af_size, 2740 }; 2741 2742 static int __init mpls_init(void) 2743 { 2744 int err; 2745 2746 BUILD_BUG_ON(sizeof(struct mpls_shim_hdr) != 4); 2747 2748 err = register_pernet_subsys(&mpls_net_ops); 2749 if (err) 2750 goto out; 2751 2752 err = register_netdevice_notifier(&mpls_dev_notifier); 2753 if (err) 2754 goto out_unregister_pernet; 2755 2756 dev_add_pack(&mpls_packet_type); 2757 2758 rtnl_af_register(&mpls_af_ops); 2759 2760 rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_NEWROUTE, 2761 mpls_rtm_newroute, NULL, 0); 2762 rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_DELROUTE, 2763 mpls_rtm_delroute, NULL, 0); 2764 rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_GETROUTE, 2765 mpls_getroute, mpls_dump_routes, 0); 2766 rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_GETNETCONF, 2767 mpls_netconf_get_devconf, 2768 mpls_netconf_dump_devconf, 0); 2769 err = ipgre_tunnel_encap_add_mpls_ops(); 2770 if (err) 2771 pr_err("Can't add mpls over gre tunnel ops\n"); 2772 2773 err = 0; 2774 out: 2775 return err; 2776 2777 out_unregister_pernet: 2778 unregister_pernet_subsys(&mpls_net_ops); 2779 goto out; 2780 } 2781 module_init(mpls_init); 2782 2783 static void __exit mpls_exit(void) 2784 { 2785 rtnl_unregister_all(PF_MPLS); 2786 rtnl_af_unregister(&mpls_af_ops); 2787 dev_remove_pack(&mpls_packet_type); 2788 unregister_netdevice_notifier(&mpls_dev_notifier); 2789 unregister_pernet_subsys(&mpls_net_ops); 2790 ipgre_tunnel_encap_del_mpls_ops(); 2791 } 2792 module_exit(mpls_exit); 2793 2794 MODULE_DESCRIPTION("MultiProtocol Label Switching"); 2795 MODULE_LICENSE("GPL v2"); 2796 MODULE_ALIAS_NETPROTO(PF_MPLS); 2797