1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * SR-IPv6 implementation 4 * 5 * Authors: 6 * David Lebrun <david.lebrun@uclouvain.be> 7 * eBPF support: Mathieu Xhonneux <m.xhonneux@gmail.com> 8 */ 9 10 #include <linux/filter.h> 11 #include <linux/types.h> 12 #include <linux/skbuff.h> 13 #include <linux/net.h> 14 #include <linux/module.h> 15 #include <net/ip.h> 16 #include <net/lwtunnel.h> 17 #include <net/netevent.h> 18 #include <net/netns/generic.h> 19 #include <net/ip6_fib.h> 20 #include <net/route.h> 21 #include <net/seg6.h> 22 #include <linux/seg6.h> 23 #include <linux/seg6_local.h> 24 #include <net/addrconf.h> 25 #include <net/ip6_route.h> 26 #include <net/dst_cache.h> 27 #include <net/ip_tunnels.h> 28 #ifdef CONFIG_IPV6_SEG6_HMAC 29 #include <net/seg6_hmac.h> 30 #endif 31 #include <net/seg6_local.h> 32 #include <linux/etherdevice.h> 33 #include <linux/bpf.h> 34 #include <linux/netfilter.h> 35 36 #define SEG6_F_ATTR(i) BIT(i) 37 38 struct seg6_local_lwt; 39 40 /* callbacks used for customizing the creation and destruction of a behavior */ 41 struct seg6_local_lwtunnel_ops { 42 int (*build_state)(struct seg6_local_lwt *slwt, const void *cfg, 43 struct netlink_ext_ack *extack); 44 void (*destroy_state)(struct seg6_local_lwt *slwt); 45 }; 46 47 struct seg6_action_desc { 48 int action; 49 unsigned long attrs; 50 51 /* The optattrs field is used for specifying all the optional 52 * attributes supported by a specific behavior. 53 * It means that if one of these attributes is not provided in the 54 * netlink message during the behavior creation, no errors will be 55 * returned to the userspace. 56 * 57 * Each attribute can be only of two types (mutually exclusive): 58 * 1) required or 2) optional. 59 * Every user MUST obey to this rule! If you set an attribute as 60 * required the same attribute CANNOT be set as optional and vice 61 * versa. 62 */ 63 unsigned long optattrs; 64 65 int (*input)(struct sk_buff *skb, struct seg6_local_lwt *slwt); 66 int static_headroom; 67 68 struct seg6_local_lwtunnel_ops slwt_ops; 69 }; 70 71 struct bpf_lwt_prog { 72 struct bpf_prog *prog; 73 char *name; 74 }; 75 76 enum seg6_end_dt_mode { 77 DT_INVALID_MODE = -EINVAL, 78 DT_LEGACY_MODE = 0, 79 DT_VRF_MODE = 1, 80 }; 81 82 struct seg6_end_dt_info { 83 enum seg6_end_dt_mode mode; 84 85 struct net *net; 86 /* VRF device associated to the routing table used by the SRv6 87 * End.DT4/DT6 behavior for routing IPv4/IPv6 packets. 88 */ 89 int vrf_ifindex; 90 int vrf_table; 91 92 /* tunneled packet family (IPv4 or IPv6). 93 * Protocol and header length are inferred from family. 94 */ 95 u16 family; 96 }; 97 98 struct pcpu_seg6_local_counters { 99 u64_stats_t packets; 100 u64_stats_t bytes; 101 u64_stats_t errors; 102 103 struct u64_stats_sync syncp; 104 }; 105 106 /* This struct groups all the SRv6 Behavior counters supported so far. 107 * 108 * put_nla_counters() makes use of this data structure to collect all counter 109 * values after the per-CPU counter evaluation has been performed. 110 * Finally, each counter value (in seg6_local_counters) is stored in the 111 * corresponding netlink attribute and sent to user space. 112 * 113 * NB: we don't want to expose this structure to user space! 114 */ 115 struct seg6_local_counters { 116 __u64 packets; 117 __u64 bytes; 118 __u64 errors; 119 }; 120 121 #define seg6_local_alloc_pcpu_counters(__gfp) \ 122 __netdev_alloc_pcpu_stats(struct pcpu_seg6_local_counters, \ 123 ((__gfp) | __GFP_ZERO)) 124 125 #define SEG6_F_LOCAL_COUNTERS SEG6_F_ATTR(SEG6_LOCAL_COUNTERS) 126 127 struct seg6_local_lwt { 128 int action; 129 struct ipv6_sr_hdr *srh; 130 int table; 131 struct in_addr nh4; 132 struct in6_addr nh6; 133 int iif; 134 int oif; 135 struct bpf_lwt_prog bpf; 136 #ifdef CONFIG_NET_L3_MASTER_DEV 137 struct seg6_end_dt_info dt_info; 138 #endif 139 struct pcpu_seg6_local_counters __percpu *pcpu_counters; 140 141 int headroom; 142 struct seg6_action_desc *desc; 143 /* unlike the required attrs, we have to track the optional attributes 144 * that have been effectively parsed. 145 */ 146 unsigned long parsed_optattrs; 147 }; 148 149 static struct seg6_local_lwt *seg6_local_lwtunnel(struct lwtunnel_state *lwt) 150 { 151 return (struct seg6_local_lwt *)lwt->data; 152 } 153 154 static struct ipv6_sr_hdr *get_and_validate_srh(struct sk_buff *skb) 155 { 156 struct ipv6_sr_hdr *srh; 157 158 srh = seg6_get_srh(skb, IP6_FH_F_SKIP_RH); 159 if (!srh) 160 return NULL; 161 162 #ifdef CONFIG_IPV6_SEG6_HMAC 163 if (!seg6_hmac_validate_skb(skb)) 164 return NULL; 165 #endif 166 167 return srh; 168 } 169 170 static bool decap_and_validate(struct sk_buff *skb, int proto) 171 { 172 struct ipv6_sr_hdr *srh; 173 unsigned int off = 0; 174 175 srh = seg6_get_srh(skb, 0); 176 if (srh && srh->segments_left > 0) 177 return false; 178 179 #ifdef CONFIG_IPV6_SEG6_HMAC 180 if (srh && !seg6_hmac_validate_skb(skb)) 181 return false; 182 #endif 183 184 if (ipv6_find_hdr(skb, &off, proto, NULL, NULL) < 0) 185 return false; 186 187 if (!pskb_pull(skb, off)) 188 return false; 189 190 skb_postpull_rcsum(skb, skb_network_header(skb), off); 191 192 skb_reset_network_header(skb); 193 skb_reset_transport_header(skb); 194 if (iptunnel_pull_offloads(skb)) 195 return false; 196 197 return true; 198 } 199 200 static void advance_nextseg(struct ipv6_sr_hdr *srh, struct in6_addr *daddr) 201 { 202 struct in6_addr *addr; 203 204 srh->segments_left--; 205 addr = srh->segments + srh->segments_left; 206 *daddr = *addr; 207 } 208 209 static int 210 seg6_lookup_any_nexthop(struct sk_buff *skb, struct in6_addr *nhaddr, 211 u32 tbl_id, bool local_delivery) 212 { 213 struct net *net = dev_net(skb->dev); 214 struct ipv6hdr *hdr = ipv6_hdr(skb); 215 int flags = RT6_LOOKUP_F_HAS_SADDR; 216 struct dst_entry *dst = NULL; 217 struct rt6_info *rt; 218 struct flowi6 fl6; 219 int dev_flags = 0; 220 221 memset(&fl6, 0, sizeof(fl6)); 222 fl6.flowi6_iif = skb->dev->ifindex; 223 fl6.daddr = nhaddr ? *nhaddr : hdr->daddr; 224 fl6.saddr = hdr->saddr; 225 fl6.flowlabel = ip6_flowinfo(hdr); 226 fl6.flowi6_mark = skb->mark; 227 fl6.flowi6_proto = hdr->nexthdr; 228 229 if (nhaddr) 230 fl6.flowi6_flags = FLOWI_FLAG_KNOWN_NH; 231 232 if (!tbl_id) { 233 dst = ip6_route_input_lookup(net, skb->dev, &fl6, skb, flags); 234 } else { 235 struct fib6_table *table; 236 237 table = fib6_get_table(net, tbl_id); 238 if (!table) 239 goto out; 240 241 rt = ip6_pol_route(net, table, 0, &fl6, skb, flags); 242 dst = &rt->dst; 243 } 244 245 /* we want to discard traffic destined for local packet processing, 246 * if @local_delivery is set to false. 247 */ 248 if (!local_delivery) 249 dev_flags |= IFF_LOOPBACK; 250 251 if (dst && (dst->dev->flags & dev_flags) && !dst->error) { 252 dst_release(dst); 253 dst = NULL; 254 } 255 256 out: 257 if (!dst) { 258 rt = net->ipv6.ip6_blk_hole_entry; 259 dst = &rt->dst; 260 dst_hold(dst); 261 } 262 263 skb_dst_drop(skb); 264 skb_dst_set(skb, dst); 265 return dst->error; 266 } 267 268 int seg6_lookup_nexthop(struct sk_buff *skb, 269 struct in6_addr *nhaddr, u32 tbl_id) 270 { 271 return seg6_lookup_any_nexthop(skb, nhaddr, tbl_id, false); 272 } 273 274 /* regular endpoint function */ 275 static int input_action_end(struct sk_buff *skb, struct seg6_local_lwt *slwt) 276 { 277 struct ipv6_sr_hdr *srh; 278 279 srh = get_and_validate_srh(skb); 280 if (!srh) 281 goto drop; 282 283 advance_nextseg(srh, &ipv6_hdr(skb)->daddr); 284 285 seg6_lookup_nexthop(skb, NULL, 0); 286 287 return dst_input(skb); 288 289 drop: 290 kfree_skb(skb); 291 return -EINVAL; 292 } 293 294 /* regular endpoint, and forward to specified nexthop */ 295 static int input_action_end_x(struct sk_buff *skb, struct seg6_local_lwt *slwt) 296 { 297 struct ipv6_sr_hdr *srh; 298 299 srh = get_and_validate_srh(skb); 300 if (!srh) 301 goto drop; 302 303 advance_nextseg(srh, &ipv6_hdr(skb)->daddr); 304 305 seg6_lookup_nexthop(skb, &slwt->nh6, 0); 306 307 return dst_input(skb); 308 309 drop: 310 kfree_skb(skb); 311 return -EINVAL; 312 } 313 314 static int input_action_end_t(struct sk_buff *skb, struct seg6_local_lwt *slwt) 315 { 316 struct ipv6_sr_hdr *srh; 317 318 srh = get_and_validate_srh(skb); 319 if (!srh) 320 goto drop; 321 322 advance_nextseg(srh, &ipv6_hdr(skb)->daddr); 323 324 seg6_lookup_nexthop(skb, NULL, slwt->table); 325 326 return dst_input(skb); 327 328 drop: 329 kfree_skb(skb); 330 return -EINVAL; 331 } 332 333 /* decapsulate and forward inner L2 frame on specified interface */ 334 static int input_action_end_dx2(struct sk_buff *skb, 335 struct seg6_local_lwt *slwt) 336 { 337 struct net *net = dev_net(skb->dev); 338 struct net_device *odev; 339 struct ethhdr *eth; 340 341 if (!decap_and_validate(skb, IPPROTO_ETHERNET)) 342 goto drop; 343 344 if (!pskb_may_pull(skb, ETH_HLEN)) 345 goto drop; 346 347 skb_reset_mac_header(skb); 348 eth = (struct ethhdr *)skb->data; 349 350 /* To determine the frame's protocol, we assume it is 802.3. This avoids 351 * a call to eth_type_trans(), which is not really relevant for our 352 * use case. 353 */ 354 if (!eth_proto_is_802_3(eth->h_proto)) 355 goto drop; 356 357 odev = dev_get_by_index_rcu(net, slwt->oif); 358 if (!odev) 359 goto drop; 360 361 /* As we accept Ethernet frames, make sure the egress device is of 362 * the correct type. 363 */ 364 if (odev->type != ARPHRD_ETHER) 365 goto drop; 366 367 if (!(odev->flags & IFF_UP) || !netif_carrier_ok(odev)) 368 goto drop; 369 370 skb_orphan(skb); 371 372 if (skb_warn_if_lro(skb)) 373 goto drop; 374 375 skb_forward_csum(skb); 376 377 if (skb->len - ETH_HLEN > odev->mtu) 378 goto drop; 379 380 skb->dev = odev; 381 skb->protocol = eth->h_proto; 382 383 return dev_queue_xmit(skb); 384 385 drop: 386 kfree_skb(skb); 387 return -EINVAL; 388 } 389 390 static int input_action_end_dx6_finish(struct net *net, struct sock *sk, 391 struct sk_buff *skb) 392 { 393 struct dst_entry *orig_dst = skb_dst(skb); 394 struct in6_addr *nhaddr = NULL; 395 struct seg6_local_lwt *slwt; 396 397 slwt = seg6_local_lwtunnel(orig_dst->lwtstate); 398 399 /* The inner packet is not associated to any local interface, 400 * so we do not call netif_rx(). 401 * 402 * If slwt->nh6 is set to ::, then lookup the nexthop for the 403 * inner packet's DA. Otherwise, use the specified nexthop. 404 */ 405 if (!ipv6_addr_any(&slwt->nh6)) 406 nhaddr = &slwt->nh6; 407 408 seg6_lookup_nexthop(skb, nhaddr, 0); 409 410 return dst_input(skb); 411 } 412 413 /* decapsulate and forward to specified nexthop */ 414 static int input_action_end_dx6(struct sk_buff *skb, 415 struct seg6_local_lwt *slwt) 416 { 417 /* this function accepts IPv6 encapsulated packets, with either 418 * an SRH with SL=0, or no SRH. 419 */ 420 421 if (!decap_and_validate(skb, IPPROTO_IPV6)) 422 goto drop; 423 424 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 425 goto drop; 426 427 skb_set_transport_header(skb, sizeof(struct ipv6hdr)); 428 nf_reset_ct(skb); 429 430 if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled)) 431 return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, 432 dev_net(skb->dev), NULL, skb, NULL, 433 skb_dst(skb)->dev, input_action_end_dx6_finish); 434 435 return input_action_end_dx6_finish(dev_net(skb->dev), NULL, skb); 436 drop: 437 kfree_skb(skb); 438 return -EINVAL; 439 } 440 441 static int input_action_end_dx4_finish(struct net *net, struct sock *sk, 442 struct sk_buff *skb) 443 { 444 struct dst_entry *orig_dst = skb_dst(skb); 445 struct seg6_local_lwt *slwt; 446 struct iphdr *iph; 447 __be32 nhaddr; 448 int err; 449 450 slwt = seg6_local_lwtunnel(orig_dst->lwtstate); 451 452 iph = ip_hdr(skb); 453 454 nhaddr = slwt->nh4.s_addr ?: iph->daddr; 455 456 skb_dst_drop(skb); 457 458 err = ip_route_input(skb, nhaddr, iph->saddr, 0, skb->dev); 459 if (err) { 460 kfree_skb(skb); 461 return -EINVAL; 462 } 463 464 return dst_input(skb); 465 } 466 467 static int input_action_end_dx4(struct sk_buff *skb, 468 struct seg6_local_lwt *slwt) 469 { 470 if (!decap_and_validate(skb, IPPROTO_IPIP)) 471 goto drop; 472 473 if (!pskb_may_pull(skb, sizeof(struct iphdr))) 474 goto drop; 475 476 skb->protocol = htons(ETH_P_IP); 477 skb_set_transport_header(skb, sizeof(struct iphdr)); 478 nf_reset_ct(skb); 479 480 if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled)) 481 return NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, 482 dev_net(skb->dev), NULL, skb, NULL, 483 skb_dst(skb)->dev, input_action_end_dx4_finish); 484 485 return input_action_end_dx4_finish(dev_net(skb->dev), NULL, skb); 486 drop: 487 kfree_skb(skb); 488 return -EINVAL; 489 } 490 491 #ifdef CONFIG_NET_L3_MASTER_DEV 492 static struct net *fib6_config_get_net(const struct fib6_config *fib6_cfg) 493 { 494 const struct nl_info *nli = &fib6_cfg->fc_nlinfo; 495 496 return nli->nl_net; 497 } 498 499 static int __seg6_end_dt_vrf_build(struct seg6_local_lwt *slwt, const void *cfg, 500 u16 family, struct netlink_ext_ack *extack) 501 { 502 struct seg6_end_dt_info *info = &slwt->dt_info; 503 int vrf_ifindex; 504 struct net *net; 505 506 net = fib6_config_get_net(cfg); 507 508 /* note that vrf_table was already set by parse_nla_vrftable() */ 509 vrf_ifindex = l3mdev_ifindex_lookup_by_table_id(L3MDEV_TYPE_VRF, net, 510 info->vrf_table); 511 if (vrf_ifindex < 0) { 512 if (vrf_ifindex == -EPERM) { 513 NL_SET_ERR_MSG(extack, 514 "Strict mode for VRF is disabled"); 515 } else if (vrf_ifindex == -ENODEV) { 516 NL_SET_ERR_MSG(extack, 517 "Table has no associated VRF device"); 518 } else { 519 pr_debug("seg6local: SRv6 End.DT* creation error=%d\n", 520 vrf_ifindex); 521 } 522 523 return vrf_ifindex; 524 } 525 526 info->net = net; 527 info->vrf_ifindex = vrf_ifindex; 528 529 info->family = family; 530 info->mode = DT_VRF_MODE; 531 532 return 0; 533 } 534 535 /* The SRv6 End.DT4/DT6 behavior extracts the inner (IPv4/IPv6) packet and 536 * routes the IPv4/IPv6 packet by looking at the configured routing table. 537 * 538 * In the SRv6 End.DT4/DT6 use case, we can receive traffic (IPv6+Segment 539 * Routing Header packets) from several interfaces and the outer IPv6 540 * destination address (DA) is used for retrieving the specific instance of the 541 * End.DT4/DT6 behavior that should process the packets. 542 * 543 * However, the inner IPv4/IPv6 packet is not really bound to any receiving 544 * interface and thus the End.DT4/DT6 sets the VRF (associated with the 545 * corresponding routing table) as the *receiving* interface. 546 * In other words, the End.DT4/DT6 processes a packet as if it has been received 547 * directly by the VRF (and not by one of its slave devices, if any). 548 * In this way, the VRF interface is used for routing the IPv4/IPv6 packet in 549 * according to the routing table configured by the End.DT4/DT6 instance. 550 * 551 * This design allows you to get some interesting features like: 552 * 1) the statistics on rx packets; 553 * 2) the possibility to install a packet sniffer on the receiving interface 554 * (the VRF one) for looking at the incoming packets; 555 * 3) the possibility to leverage the netfilter prerouting hook for the inner 556 * IPv4 packet. 557 * 558 * This function returns: 559 * - the sk_buff* when the VRF rcv handler has processed the packet correctly; 560 * - NULL when the skb is consumed by the VRF rcv handler; 561 * - a pointer which encodes a negative error number in case of error. 562 * Note that in this case, the function takes care of freeing the skb. 563 */ 564 static struct sk_buff *end_dt_vrf_rcv(struct sk_buff *skb, u16 family, 565 struct net_device *dev) 566 { 567 /* based on l3mdev_ip_rcv; we are only interested in the master */ 568 if (unlikely(!netif_is_l3_master(dev) && !netif_has_l3_rx_handler(dev))) 569 goto drop; 570 571 if (unlikely(!dev->l3mdev_ops->l3mdev_l3_rcv)) 572 goto drop; 573 574 /* the decap packet IPv4/IPv6 does not come with any mac header info. 575 * We must unset the mac header to allow the VRF device to rebuild it, 576 * just in case there is a sniffer attached on the device. 577 */ 578 skb_unset_mac_header(skb); 579 580 skb = dev->l3mdev_ops->l3mdev_l3_rcv(dev, skb, family); 581 if (!skb) 582 /* the skb buffer was consumed by the handler */ 583 return NULL; 584 585 /* when a packet is received by a VRF or by one of its slaves, the 586 * master device reference is set into the skb. 587 */ 588 if (unlikely(skb->dev != dev || skb->skb_iif != dev->ifindex)) 589 goto drop; 590 591 return skb; 592 593 drop: 594 kfree_skb(skb); 595 return ERR_PTR(-EINVAL); 596 } 597 598 static struct net_device *end_dt_get_vrf_rcu(struct sk_buff *skb, 599 struct seg6_end_dt_info *info) 600 { 601 int vrf_ifindex = info->vrf_ifindex; 602 struct net *net = info->net; 603 604 if (unlikely(vrf_ifindex < 0)) 605 goto error; 606 607 if (unlikely(!net_eq(dev_net(skb->dev), net))) 608 goto error; 609 610 return dev_get_by_index_rcu(net, vrf_ifindex); 611 612 error: 613 return NULL; 614 } 615 616 static struct sk_buff *end_dt_vrf_core(struct sk_buff *skb, 617 struct seg6_local_lwt *slwt, u16 family) 618 { 619 struct seg6_end_dt_info *info = &slwt->dt_info; 620 struct net_device *vrf; 621 __be16 protocol; 622 int hdrlen; 623 624 vrf = end_dt_get_vrf_rcu(skb, info); 625 if (unlikely(!vrf)) 626 goto drop; 627 628 switch (family) { 629 case AF_INET: 630 protocol = htons(ETH_P_IP); 631 hdrlen = sizeof(struct iphdr); 632 break; 633 case AF_INET6: 634 protocol = htons(ETH_P_IPV6); 635 hdrlen = sizeof(struct ipv6hdr); 636 break; 637 case AF_UNSPEC: 638 fallthrough; 639 default: 640 goto drop; 641 } 642 643 if (unlikely(info->family != AF_UNSPEC && info->family != family)) { 644 pr_warn_once("seg6local: SRv6 End.DT* family mismatch"); 645 goto drop; 646 } 647 648 skb->protocol = protocol; 649 650 skb_dst_drop(skb); 651 652 skb_set_transport_header(skb, hdrlen); 653 nf_reset_ct(skb); 654 655 return end_dt_vrf_rcv(skb, family, vrf); 656 657 drop: 658 kfree_skb(skb); 659 return ERR_PTR(-EINVAL); 660 } 661 662 static int input_action_end_dt4(struct sk_buff *skb, 663 struct seg6_local_lwt *slwt) 664 { 665 struct iphdr *iph; 666 int err; 667 668 if (!decap_and_validate(skb, IPPROTO_IPIP)) 669 goto drop; 670 671 if (!pskb_may_pull(skb, sizeof(struct iphdr))) 672 goto drop; 673 674 skb = end_dt_vrf_core(skb, slwt, AF_INET); 675 if (!skb) 676 /* packet has been processed and consumed by the VRF */ 677 return 0; 678 679 if (IS_ERR(skb)) 680 return PTR_ERR(skb); 681 682 iph = ip_hdr(skb); 683 684 err = ip_route_input(skb, iph->daddr, iph->saddr, 0, skb->dev); 685 if (unlikely(err)) 686 goto drop; 687 688 return dst_input(skb); 689 690 drop: 691 kfree_skb(skb); 692 return -EINVAL; 693 } 694 695 static int seg6_end_dt4_build(struct seg6_local_lwt *slwt, const void *cfg, 696 struct netlink_ext_ack *extack) 697 { 698 return __seg6_end_dt_vrf_build(slwt, cfg, AF_INET, extack); 699 } 700 701 static enum 702 seg6_end_dt_mode seg6_end_dt6_parse_mode(struct seg6_local_lwt *slwt) 703 { 704 unsigned long parsed_optattrs = slwt->parsed_optattrs; 705 bool legacy, vrfmode; 706 707 legacy = !!(parsed_optattrs & SEG6_F_ATTR(SEG6_LOCAL_TABLE)); 708 vrfmode = !!(parsed_optattrs & SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE)); 709 710 if (!(legacy ^ vrfmode)) 711 /* both are absent or present: invalid DT6 mode */ 712 return DT_INVALID_MODE; 713 714 return legacy ? DT_LEGACY_MODE : DT_VRF_MODE; 715 } 716 717 static enum seg6_end_dt_mode seg6_end_dt6_get_mode(struct seg6_local_lwt *slwt) 718 { 719 struct seg6_end_dt_info *info = &slwt->dt_info; 720 721 return info->mode; 722 } 723 724 static int seg6_end_dt6_build(struct seg6_local_lwt *slwt, const void *cfg, 725 struct netlink_ext_ack *extack) 726 { 727 enum seg6_end_dt_mode mode = seg6_end_dt6_parse_mode(slwt); 728 struct seg6_end_dt_info *info = &slwt->dt_info; 729 730 switch (mode) { 731 case DT_LEGACY_MODE: 732 info->mode = DT_LEGACY_MODE; 733 return 0; 734 case DT_VRF_MODE: 735 return __seg6_end_dt_vrf_build(slwt, cfg, AF_INET6, extack); 736 default: 737 NL_SET_ERR_MSG(extack, "table or vrftable must be specified"); 738 return -EINVAL; 739 } 740 } 741 #endif 742 743 static int input_action_end_dt6(struct sk_buff *skb, 744 struct seg6_local_lwt *slwt) 745 { 746 if (!decap_and_validate(skb, IPPROTO_IPV6)) 747 goto drop; 748 749 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 750 goto drop; 751 752 #ifdef CONFIG_NET_L3_MASTER_DEV 753 if (seg6_end_dt6_get_mode(slwt) == DT_LEGACY_MODE) 754 goto legacy_mode; 755 756 /* DT6_VRF_MODE */ 757 skb = end_dt_vrf_core(skb, slwt, AF_INET6); 758 if (!skb) 759 /* packet has been processed and consumed by the VRF */ 760 return 0; 761 762 if (IS_ERR(skb)) 763 return PTR_ERR(skb); 764 765 /* note: this time we do not need to specify the table because the VRF 766 * takes care of selecting the correct table. 767 */ 768 seg6_lookup_any_nexthop(skb, NULL, 0, true); 769 770 return dst_input(skb); 771 772 legacy_mode: 773 #endif 774 skb_set_transport_header(skb, sizeof(struct ipv6hdr)); 775 776 seg6_lookup_any_nexthop(skb, NULL, slwt->table, true); 777 778 return dst_input(skb); 779 780 drop: 781 kfree_skb(skb); 782 return -EINVAL; 783 } 784 785 #ifdef CONFIG_NET_L3_MASTER_DEV 786 static int seg6_end_dt46_build(struct seg6_local_lwt *slwt, const void *cfg, 787 struct netlink_ext_ack *extack) 788 { 789 return __seg6_end_dt_vrf_build(slwt, cfg, AF_UNSPEC, extack); 790 } 791 792 static int input_action_end_dt46(struct sk_buff *skb, 793 struct seg6_local_lwt *slwt) 794 { 795 unsigned int off = 0; 796 int nexthdr; 797 798 nexthdr = ipv6_find_hdr(skb, &off, -1, NULL, NULL); 799 if (unlikely(nexthdr < 0)) 800 goto drop; 801 802 switch (nexthdr) { 803 case IPPROTO_IPIP: 804 return input_action_end_dt4(skb, slwt); 805 case IPPROTO_IPV6: 806 return input_action_end_dt6(skb, slwt); 807 } 808 809 drop: 810 kfree_skb(skb); 811 return -EINVAL; 812 } 813 #endif 814 815 /* push an SRH on top of the current one */ 816 static int input_action_end_b6(struct sk_buff *skb, struct seg6_local_lwt *slwt) 817 { 818 struct ipv6_sr_hdr *srh; 819 int err = -EINVAL; 820 821 srh = get_and_validate_srh(skb); 822 if (!srh) 823 goto drop; 824 825 err = seg6_do_srh_inline(skb, slwt->srh); 826 if (err) 827 goto drop; 828 829 skb_set_transport_header(skb, sizeof(struct ipv6hdr)); 830 831 seg6_lookup_nexthop(skb, NULL, 0); 832 833 return dst_input(skb); 834 835 drop: 836 kfree_skb(skb); 837 return err; 838 } 839 840 /* encapsulate within an outer IPv6 header and a specified SRH */ 841 static int input_action_end_b6_encap(struct sk_buff *skb, 842 struct seg6_local_lwt *slwt) 843 { 844 struct ipv6_sr_hdr *srh; 845 int err = -EINVAL; 846 847 srh = get_and_validate_srh(skb); 848 if (!srh) 849 goto drop; 850 851 advance_nextseg(srh, &ipv6_hdr(skb)->daddr); 852 853 skb_reset_inner_headers(skb); 854 skb->encapsulation = 1; 855 856 err = seg6_do_srh_encap(skb, slwt->srh, IPPROTO_IPV6); 857 if (err) 858 goto drop; 859 860 skb_set_transport_header(skb, sizeof(struct ipv6hdr)); 861 862 seg6_lookup_nexthop(skb, NULL, 0); 863 864 return dst_input(skb); 865 866 drop: 867 kfree_skb(skb); 868 return err; 869 } 870 871 DEFINE_PER_CPU(struct seg6_bpf_srh_state, seg6_bpf_srh_states); 872 873 bool seg6_bpf_has_valid_srh(struct sk_buff *skb) 874 { 875 struct seg6_bpf_srh_state *srh_state = 876 this_cpu_ptr(&seg6_bpf_srh_states); 877 struct ipv6_sr_hdr *srh = srh_state->srh; 878 879 if (unlikely(srh == NULL)) 880 return false; 881 882 if (unlikely(!srh_state->valid)) { 883 if ((srh_state->hdrlen & 7) != 0) 884 return false; 885 886 srh->hdrlen = (u8)(srh_state->hdrlen >> 3); 887 if (!seg6_validate_srh(srh, (srh->hdrlen + 1) << 3, true)) 888 return false; 889 890 srh_state->valid = true; 891 } 892 893 return true; 894 } 895 896 static int input_action_end_bpf(struct sk_buff *skb, 897 struct seg6_local_lwt *slwt) 898 { 899 struct seg6_bpf_srh_state *srh_state = 900 this_cpu_ptr(&seg6_bpf_srh_states); 901 struct ipv6_sr_hdr *srh; 902 int ret; 903 904 srh = get_and_validate_srh(skb); 905 if (!srh) { 906 kfree_skb(skb); 907 return -EINVAL; 908 } 909 advance_nextseg(srh, &ipv6_hdr(skb)->daddr); 910 911 /* preempt_disable is needed to protect the per-CPU buffer srh_state, 912 * which is also accessed by the bpf_lwt_seg6_* helpers 913 */ 914 preempt_disable(); 915 srh_state->srh = srh; 916 srh_state->hdrlen = srh->hdrlen << 3; 917 srh_state->valid = true; 918 919 rcu_read_lock(); 920 bpf_compute_data_pointers(skb); 921 ret = bpf_prog_run_save_cb(slwt->bpf.prog, skb); 922 rcu_read_unlock(); 923 924 switch (ret) { 925 case BPF_OK: 926 case BPF_REDIRECT: 927 break; 928 case BPF_DROP: 929 goto drop; 930 default: 931 pr_warn_once("bpf-seg6local: Illegal return value %u\n", ret); 932 goto drop; 933 } 934 935 if (srh_state->srh && !seg6_bpf_has_valid_srh(skb)) 936 goto drop; 937 938 preempt_enable(); 939 if (ret != BPF_REDIRECT) 940 seg6_lookup_nexthop(skb, NULL, 0); 941 942 return dst_input(skb); 943 944 drop: 945 preempt_enable(); 946 kfree_skb(skb); 947 return -EINVAL; 948 } 949 950 static struct seg6_action_desc seg6_action_table[] = { 951 { 952 .action = SEG6_LOCAL_ACTION_END, 953 .attrs = 0, 954 .optattrs = SEG6_F_LOCAL_COUNTERS, 955 .input = input_action_end, 956 }, 957 { 958 .action = SEG6_LOCAL_ACTION_END_X, 959 .attrs = SEG6_F_ATTR(SEG6_LOCAL_NH6), 960 .optattrs = SEG6_F_LOCAL_COUNTERS, 961 .input = input_action_end_x, 962 }, 963 { 964 .action = SEG6_LOCAL_ACTION_END_T, 965 .attrs = SEG6_F_ATTR(SEG6_LOCAL_TABLE), 966 .optattrs = SEG6_F_LOCAL_COUNTERS, 967 .input = input_action_end_t, 968 }, 969 { 970 .action = SEG6_LOCAL_ACTION_END_DX2, 971 .attrs = SEG6_F_ATTR(SEG6_LOCAL_OIF), 972 .optattrs = SEG6_F_LOCAL_COUNTERS, 973 .input = input_action_end_dx2, 974 }, 975 { 976 .action = SEG6_LOCAL_ACTION_END_DX6, 977 .attrs = SEG6_F_ATTR(SEG6_LOCAL_NH6), 978 .optattrs = SEG6_F_LOCAL_COUNTERS, 979 .input = input_action_end_dx6, 980 }, 981 { 982 .action = SEG6_LOCAL_ACTION_END_DX4, 983 .attrs = SEG6_F_ATTR(SEG6_LOCAL_NH4), 984 .optattrs = SEG6_F_LOCAL_COUNTERS, 985 .input = input_action_end_dx4, 986 }, 987 { 988 .action = SEG6_LOCAL_ACTION_END_DT4, 989 .attrs = SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE), 990 .optattrs = SEG6_F_LOCAL_COUNTERS, 991 #ifdef CONFIG_NET_L3_MASTER_DEV 992 .input = input_action_end_dt4, 993 .slwt_ops = { 994 .build_state = seg6_end_dt4_build, 995 }, 996 #endif 997 }, 998 { 999 .action = SEG6_LOCAL_ACTION_END_DT6, 1000 #ifdef CONFIG_NET_L3_MASTER_DEV 1001 .attrs = 0, 1002 .optattrs = SEG6_F_LOCAL_COUNTERS | 1003 SEG6_F_ATTR(SEG6_LOCAL_TABLE) | 1004 SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE), 1005 .slwt_ops = { 1006 .build_state = seg6_end_dt6_build, 1007 }, 1008 #else 1009 .attrs = SEG6_F_ATTR(SEG6_LOCAL_TABLE), 1010 .optattrs = SEG6_F_LOCAL_COUNTERS, 1011 #endif 1012 .input = input_action_end_dt6, 1013 }, 1014 { 1015 .action = SEG6_LOCAL_ACTION_END_DT46, 1016 .attrs = SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE), 1017 .optattrs = SEG6_F_LOCAL_COUNTERS, 1018 #ifdef CONFIG_NET_L3_MASTER_DEV 1019 .input = input_action_end_dt46, 1020 .slwt_ops = { 1021 .build_state = seg6_end_dt46_build, 1022 }, 1023 #endif 1024 }, 1025 { 1026 .action = SEG6_LOCAL_ACTION_END_B6, 1027 .attrs = SEG6_F_ATTR(SEG6_LOCAL_SRH), 1028 .optattrs = SEG6_F_LOCAL_COUNTERS, 1029 .input = input_action_end_b6, 1030 }, 1031 { 1032 .action = SEG6_LOCAL_ACTION_END_B6_ENCAP, 1033 .attrs = SEG6_F_ATTR(SEG6_LOCAL_SRH), 1034 .optattrs = SEG6_F_LOCAL_COUNTERS, 1035 .input = input_action_end_b6_encap, 1036 .static_headroom = sizeof(struct ipv6hdr), 1037 }, 1038 { 1039 .action = SEG6_LOCAL_ACTION_END_BPF, 1040 .attrs = SEG6_F_ATTR(SEG6_LOCAL_BPF), 1041 .optattrs = SEG6_F_LOCAL_COUNTERS, 1042 .input = input_action_end_bpf, 1043 }, 1044 1045 }; 1046 1047 static struct seg6_action_desc *__get_action_desc(int action) 1048 { 1049 struct seg6_action_desc *desc; 1050 int i, count; 1051 1052 count = ARRAY_SIZE(seg6_action_table); 1053 for (i = 0; i < count; i++) { 1054 desc = &seg6_action_table[i]; 1055 if (desc->action == action) 1056 return desc; 1057 } 1058 1059 return NULL; 1060 } 1061 1062 static bool seg6_lwtunnel_counters_enabled(struct seg6_local_lwt *slwt) 1063 { 1064 return slwt->parsed_optattrs & SEG6_F_LOCAL_COUNTERS; 1065 } 1066 1067 static void seg6_local_update_counters(struct seg6_local_lwt *slwt, 1068 unsigned int len, int err) 1069 { 1070 struct pcpu_seg6_local_counters *pcounters; 1071 1072 pcounters = this_cpu_ptr(slwt->pcpu_counters); 1073 u64_stats_update_begin(&pcounters->syncp); 1074 1075 if (likely(!err)) { 1076 u64_stats_inc(&pcounters->packets); 1077 u64_stats_add(&pcounters->bytes, len); 1078 } else { 1079 u64_stats_inc(&pcounters->errors); 1080 } 1081 1082 u64_stats_update_end(&pcounters->syncp); 1083 } 1084 1085 static int seg6_local_input_core(struct net *net, struct sock *sk, 1086 struct sk_buff *skb) 1087 { 1088 struct dst_entry *orig_dst = skb_dst(skb); 1089 struct seg6_action_desc *desc; 1090 struct seg6_local_lwt *slwt; 1091 unsigned int len = skb->len; 1092 int rc; 1093 1094 slwt = seg6_local_lwtunnel(orig_dst->lwtstate); 1095 desc = slwt->desc; 1096 1097 rc = desc->input(skb, slwt); 1098 1099 if (!seg6_lwtunnel_counters_enabled(slwt)) 1100 return rc; 1101 1102 seg6_local_update_counters(slwt, len, rc); 1103 1104 return rc; 1105 } 1106 1107 static int seg6_local_input(struct sk_buff *skb) 1108 { 1109 if (skb->protocol != htons(ETH_P_IPV6)) { 1110 kfree_skb(skb); 1111 return -EINVAL; 1112 } 1113 1114 if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled)) 1115 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN, 1116 dev_net(skb->dev), NULL, skb, skb->dev, NULL, 1117 seg6_local_input_core); 1118 1119 return seg6_local_input_core(dev_net(skb->dev), NULL, skb); 1120 } 1121 1122 static const struct nla_policy seg6_local_policy[SEG6_LOCAL_MAX + 1] = { 1123 [SEG6_LOCAL_ACTION] = { .type = NLA_U32 }, 1124 [SEG6_LOCAL_SRH] = { .type = NLA_BINARY }, 1125 [SEG6_LOCAL_TABLE] = { .type = NLA_U32 }, 1126 [SEG6_LOCAL_VRFTABLE] = { .type = NLA_U32 }, 1127 [SEG6_LOCAL_NH4] = { .type = NLA_BINARY, 1128 .len = sizeof(struct in_addr) }, 1129 [SEG6_LOCAL_NH6] = { .type = NLA_BINARY, 1130 .len = sizeof(struct in6_addr) }, 1131 [SEG6_LOCAL_IIF] = { .type = NLA_U32 }, 1132 [SEG6_LOCAL_OIF] = { .type = NLA_U32 }, 1133 [SEG6_LOCAL_BPF] = { .type = NLA_NESTED }, 1134 [SEG6_LOCAL_COUNTERS] = { .type = NLA_NESTED }, 1135 }; 1136 1137 static int parse_nla_srh(struct nlattr **attrs, struct seg6_local_lwt *slwt) 1138 { 1139 struct ipv6_sr_hdr *srh; 1140 int len; 1141 1142 srh = nla_data(attrs[SEG6_LOCAL_SRH]); 1143 len = nla_len(attrs[SEG6_LOCAL_SRH]); 1144 1145 /* SRH must contain at least one segment */ 1146 if (len < sizeof(*srh) + sizeof(struct in6_addr)) 1147 return -EINVAL; 1148 1149 if (!seg6_validate_srh(srh, len, false)) 1150 return -EINVAL; 1151 1152 slwt->srh = kmemdup(srh, len, GFP_KERNEL); 1153 if (!slwt->srh) 1154 return -ENOMEM; 1155 1156 slwt->headroom += len; 1157 1158 return 0; 1159 } 1160 1161 static int put_nla_srh(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1162 { 1163 struct ipv6_sr_hdr *srh; 1164 struct nlattr *nla; 1165 int len; 1166 1167 srh = slwt->srh; 1168 len = (srh->hdrlen + 1) << 3; 1169 1170 nla = nla_reserve(skb, SEG6_LOCAL_SRH, len); 1171 if (!nla) 1172 return -EMSGSIZE; 1173 1174 memcpy(nla_data(nla), srh, len); 1175 1176 return 0; 1177 } 1178 1179 static int cmp_nla_srh(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1180 { 1181 int len = (a->srh->hdrlen + 1) << 3; 1182 1183 if (len != ((b->srh->hdrlen + 1) << 3)) 1184 return 1; 1185 1186 return memcmp(a->srh, b->srh, len); 1187 } 1188 1189 static void destroy_attr_srh(struct seg6_local_lwt *slwt) 1190 { 1191 kfree(slwt->srh); 1192 } 1193 1194 static int parse_nla_table(struct nlattr **attrs, struct seg6_local_lwt *slwt) 1195 { 1196 slwt->table = nla_get_u32(attrs[SEG6_LOCAL_TABLE]); 1197 1198 return 0; 1199 } 1200 1201 static int put_nla_table(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1202 { 1203 if (nla_put_u32(skb, SEG6_LOCAL_TABLE, slwt->table)) 1204 return -EMSGSIZE; 1205 1206 return 0; 1207 } 1208 1209 static int cmp_nla_table(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1210 { 1211 if (a->table != b->table) 1212 return 1; 1213 1214 return 0; 1215 } 1216 1217 static struct 1218 seg6_end_dt_info *seg6_possible_end_dt_info(struct seg6_local_lwt *slwt) 1219 { 1220 #ifdef CONFIG_NET_L3_MASTER_DEV 1221 return &slwt->dt_info; 1222 #else 1223 return ERR_PTR(-EOPNOTSUPP); 1224 #endif 1225 } 1226 1227 static int parse_nla_vrftable(struct nlattr **attrs, 1228 struct seg6_local_lwt *slwt) 1229 { 1230 struct seg6_end_dt_info *info = seg6_possible_end_dt_info(slwt); 1231 1232 if (IS_ERR(info)) 1233 return PTR_ERR(info); 1234 1235 info->vrf_table = nla_get_u32(attrs[SEG6_LOCAL_VRFTABLE]); 1236 1237 return 0; 1238 } 1239 1240 static int put_nla_vrftable(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1241 { 1242 struct seg6_end_dt_info *info = seg6_possible_end_dt_info(slwt); 1243 1244 if (IS_ERR(info)) 1245 return PTR_ERR(info); 1246 1247 if (nla_put_u32(skb, SEG6_LOCAL_VRFTABLE, info->vrf_table)) 1248 return -EMSGSIZE; 1249 1250 return 0; 1251 } 1252 1253 static int cmp_nla_vrftable(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1254 { 1255 struct seg6_end_dt_info *info_a = seg6_possible_end_dt_info(a); 1256 struct seg6_end_dt_info *info_b = seg6_possible_end_dt_info(b); 1257 1258 if (info_a->vrf_table != info_b->vrf_table) 1259 return 1; 1260 1261 return 0; 1262 } 1263 1264 static int parse_nla_nh4(struct nlattr **attrs, struct seg6_local_lwt *slwt) 1265 { 1266 memcpy(&slwt->nh4, nla_data(attrs[SEG6_LOCAL_NH4]), 1267 sizeof(struct in_addr)); 1268 1269 return 0; 1270 } 1271 1272 static int put_nla_nh4(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1273 { 1274 struct nlattr *nla; 1275 1276 nla = nla_reserve(skb, SEG6_LOCAL_NH4, sizeof(struct in_addr)); 1277 if (!nla) 1278 return -EMSGSIZE; 1279 1280 memcpy(nla_data(nla), &slwt->nh4, sizeof(struct in_addr)); 1281 1282 return 0; 1283 } 1284 1285 static int cmp_nla_nh4(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1286 { 1287 return memcmp(&a->nh4, &b->nh4, sizeof(struct in_addr)); 1288 } 1289 1290 static int parse_nla_nh6(struct nlattr **attrs, struct seg6_local_lwt *slwt) 1291 { 1292 memcpy(&slwt->nh6, nla_data(attrs[SEG6_LOCAL_NH6]), 1293 sizeof(struct in6_addr)); 1294 1295 return 0; 1296 } 1297 1298 static int put_nla_nh6(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1299 { 1300 struct nlattr *nla; 1301 1302 nla = nla_reserve(skb, SEG6_LOCAL_NH6, sizeof(struct in6_addr)); 1303 if (!nla) 1304 return -EMSGSIZE; 1305 1306 memcpy(nla_data(nla), &slwt->nh6, sizeof(struct in6_addr)); 1307 1308 return 0; 1309 } 1310 1311 static int cmp_nla_nh6(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1312 { 1313 return memcmp(&a->nh6, &b->nh6, sizeof(struct in6_addr)); 1314 } 1315 1316 static int parse_nla_iif(struct nlattr **attrs, struct seg6_local_lwt *slwt) 1317 { 1318 slwt->iif = nla_get_u32(attrs[SEG6_LOCAL_IIF]); 1319 1320 return 0; 1321 } 1322 1323 static int put_nla_iif(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1324 { 1325 if (nla_put_u32(skb, SEG6_LOCAL_IIF, slwt->iif)) 1326 return -EMSGSIZE; 1327 1328 return 0; 1329 } 1330 1331 static int cmp_nla_iif(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1332 { 1333 if (a->iif != b->iif) 1334 return 1; 1335 1336 return 0; 1337 } 1338 1339 static int parse_nla_oif(struct nlattr **attrs, struct seg6_local_lwt *slwt) 1340 { 1341 slwt->oif = nla_get_u32(attrs[SEG6_LOCAL_OIF]); 1342 1343 return 0; 1344 } 1345 1346 static int put_nla_oif(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1347 { 1348 if (nla_put_u32(skb, SEG6_LOCAL_OIF, slwt->oif)) 1349 return -EMSGSIZE; 1350 1351 return 0; 1352 } 1353 1354 static int cmp_nla_oif(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1355 { 1356 if (a->oif != b->oif) 1357 return 1; 1358 1359 return 0; 1360 } 1361 1362 #define MAX_PROG_NAME 256 1363 static const struct nla_policy bpf_prog_policy[SEG6_LOCAL_BPF_PROG_MAX + 1] = { 1364 [SEG6_LOCAL_BPF_PROG] = { .type = NLA_U32, }, 1365 [SEG6_LOCAL_BPF_PROG_NAME] = { .type = NLA_NUL_STRING, 1366 .len = MAX_PROG_NAME }, 1367 }; 1368 1369 static int parse_nla_bpf(struct nlattr **attrs, struct seg6_local_lwt *slwt) 1370 { 1371 struct nlattr *tb[SEG6_LOCAL_BPF_PROG_MAX + 1]; 1372 struct bpf_prog *p; 1373 int ret; 1374 u32 fd; 1375 1376 ret = nla_parse_nested_deprecated(tb, SEG6_LOCAL_BPF_PROG_MAX, 1377 attrs[SEG6_LOCAL_BPF], 1378 bpf_prog_policy, NULL); 1379 if (ret < 0) 1380 return ret; 1381 1382 if (!tb[SEG6_LOCAL_BPF_PROG] || !tb[SEG6_LOCAL_BPF_PROG_NAME]) 1383 return -EINVAL; 1384 1385 slwt->bpf.name = nla_memdup(tb[SEG6_LOCAL_BPF_PROG_NAME], GFP_KERNEL); 1386 if (!slwt->bpf.name) 1387 return -ENOMEM; 1388 1389 fd = nla_get_u32(tb[SEG6_LOCAL_BPF_PROG]); 1390 p = bpf_prog_get_type(fd, BPF_PROG_TYPE_LWT_SEG6LOCAL); 1391 if (IS_ERR(p)) { 1392 kfree(slwt->bpf.name); 1393 return PTR_ERR(p); 1394 } 1395 1396 slwt->bpf.prog = p; 1397 return 0; 1398 } 1399 1400 static int put_nla_bpf(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1401 { 1402 struct nlattr *nest; 1403 1404 if (!slwt->bpf.prog) 1405 return 0; 1406 1407 nest = nla_nest_start_noflag(skb, SEG6_LOCAL_BPF); 1408 if (!nest) 1409 return -EMSGSIZE; 1410 1411 if (nla_put_u32(skb, SEG6_LOCAL_BPF_PROG, slwt->bpf.prog->aux->id)) 1412 return -EMSGSIZE; 1413 1414 if (slwt->bpf.name && 1415 nla_put_string(skb, SEG6_LOCAL_BPF_PROG_NAME, slwt->bpf.name)) 1416 return -EMSGSIZE; 1417 1418 return nla_nest_end(skb, nest); 1419 } 1420 1421 static int cmp_nla_bpf(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1422 { 1423 if (!a->bpf.name && !b->bpf.name) 1424 return 0; 1425 1426 if (!a->bpf.name || !b->bpf.name) 1427 return 1; 1428 1429 return strcmp(a->bpf.name, b->bpf.name); 1430 } 1431 1432 static void destroy_attr_bpf(struct seg6_local_lwt *slwt) 1433 { 1434 kfree(slwt->bpf.name); 1435 if (slwt->bpf.prog) 1436 bpf_prog_put(slwt->bpf.prog); 1437 } 1438 1439 static const struct 1440 nla_policy seg6_local_counters_policy[SEG6_LOCAL_CNT_MAX + 1] = { 1441 [SEG6_LOCAL_CNT_PACKETS] = { .type = NLA_U64 }, 1442 [SEG6_LOCAL_CNT_BYTES] = { .type = NLA_U64 }, 1443 [SEG6_LOCAL_CNT_ERRORS] = { .type = NLA_U64 }, 1444 }; 1445 1446 static int parse_nla_counters(struct nlattr **attrs, 1447 struct seg6_local_lwt *slwt) 1448 { 1449 struct pcpu_seg6_local_counters __percpu *pcounters; 1450 struct nlattr *tb[SEG6_LOCAL_CNT_MAX + 1]; 1451 int ret; 1452 1453 ret = nla_parse_nested_deprecated(tb, SEG6_LOCAL_CNT_MAX, 1454 attrs[SEG6_LOCAL_COUNTERS], 1455 seg6_local_counters_policy, NULL); 1456 if (ret < 0) 1457 return ret; 1458 1459 /* basic support for SRv6 Behavior counters requires at least: 1460 * packets, bytes and errors. 1461 */ 1462 if (!tb[SEG6_LOCAL_CNT_PACKETS] || !tb[SEG6_LOCAL_CNT_BYTES] || 1463 !tb[SEG6_LOCAL_CNT_ERRORS]) 1464 return -EINVAL; 1465 1466 /* counters are always zero initialized */ 1467 pcounters = seg6_local_alloc_pcpu_counters(GFP_KERNEL); 1468 if (!pcounters) 1469 return -ENOMEM; 1470 1471 slwt->pcpu_counters = pcounters; 1472 1473 return 0; 1474 } 1475 1476 static int seg6_local_fill_nla_counters(struct sk_buff *skb, 1477 struct seg6_local_counters *counters) 1478 { 1479 if (nla_put_u64_64bit(skb, SEG6_LOCAL_CNT_PACKETS, counters->packets, 1480 SEG6_LOCAL_CNT_PAD)) 1481 return -EMSGSIZE; 1482 1483 if (nla_put_u64_64bit(skb, SEG6_LOCAL_CNT_BYTES, counters->bytes, 1484 SEG6_LOCAL_CNT_PAD)) 1485 return -EMSGSIZE; 1486 1487 if (nla_put_u64_64bit(skb, SEG6_LOCAL_CNT_ERRORS, counters->errors, 1488 SEG6_LOCAL_CNT_PAD)) 1489 return -EMSGSIZE; 1490 1491 return 0; 1492 } 1493 1494 static int put_nla_counters(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1495 { 1496 struct seg6_local_counters counters = { 0, 0, 0 }; 1497 struct nlattr *nest; 1498 int rc, i; 1499 1500 nest = nla_nest_start(skb, SEG6_LOCAL_COUNTERS); 1501 if (!nest) 1502 return -EMSGSIZE; 1503 1504 for_each_possible_cpu(i) { 1505 struct pcpu_seg6_local_counters *pcounters; 1506 u64 packets, bytes, errors; 1507 unsigned int start; 1508 1509 pcounters = per_cpu_ptr(slwt->pcpu_counters, i); 1510 do { 1511 start = u64_stats_fetch_begin_irq(&pcounters->syncp); 1512 1513 packets = u64_stats_read(&pcounters->packets); 1514 bytes = u64_stats_read(&pcounters->bytes); 1515 errors = u64_stats_read(&pcounters->errors); 1516 1517 } while (u64_stats_fetch_retry_irq(&pcounters->syncp, start)); 1518 1519 counters.packets += packets; 1520 counters.bytes += bytes; 1521 counters.errors += errors; 1522 } 1523 1524 rc = seg6_local_fill_nla_counters(skb, &counters); 1525 if (rc < 0) { 1526 nla_nest_cancel(skb, nest); 1527 return rc; 1528 } 1529 1530 return nla_nest_end(skb, nest); 1531 } 1532 1533 static int cmp_nla_counters(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1534 { 1535 /* a and b are equal if both have pcpu_counters set or not */ 1536 return (!!((unsigned long)a->pcpu_counters)) ^ 1537 (!!((unsigned long)b->pcpu_counters)); 1538 } 1539 1540 static void destroy_attr_counters(struct seg6_local_lwt *slwt) 1541 { 1542 free_percpu(slwt->pcpu_counters); 1543 } 1544 1545 struct seg6_action_param { 1546 int (*parse)(struct nlattr **attrs, struct seg6_local_lwt *slwt); 1547 int (*put)(struct sk_buff *skb, struct seg6_local_lwt *slwt); 1548 int (*cmp)(struct seg6_local_lwt *a, struct seg6_local_lwt *b); 1549 1550 /* optional destroy() callback useful for releasing resources which 1551 * have been previously acquired in the corresponding parse() 1552 * function. 1553 */ 1554 void (*destroy)(struct seg6_local_lwt *slwt); 1555 }; 1556 1557 static struct seg6_action_param seg6_action_params[SEG6_LOCAL_MAX + 1] = { 1558 [SEG6_LOCAL_SRH] = { .parse = parse_nla_srh, 1559 .put = put_nla_srh, 1560 .cmp = cmp_nla_srh, 1561 .destroy = destroy_attr_srh }, 1562 1563 [SEG6_LOCAL_TABLE] = { .parse = parse_nla_table, 1564 .put = put_nla_table, 1565 .cmp = cmp_nla_table }, 1566 1567 [SEG6_LOCAL_NH4] = { .parse = parse_nla_nh4, 1568 .put = put_nla_nh4, 1569 .cmp = cmp_nla_nh4 }, 1570 1571 [SEG6_LOCAL_NH6] = { .parse = parse_nla_nh6, 1572 .put = put_nla_nh6, 1573 .cmp = cmp_nla_nh6 }, 1574 1575 [SEG6_LOCAL_IIF] = { .parse = parse_nla_iif, 1576 .put = put_nla_iif, 1577 .cmp = cmp_nla_iif }, 1578 1579 [SEG6_LOCAL_OIF] = { .parse = parse_nla_oif, 1580 .put = put_nla_oif, 1581 .cmp = cmp_nla_oif }, 1582 1583 [SEG6_LOCAL_BPF] = { .parse = parse_nla_bpf, 1584 .put = put_nla_bpf, 1585 .cmp = cmp_nla_bpf, 1586 .destroy = destroy_attr_bpf }, 1587 1588 [SEG6_LOCAL_VRFTABLE] = { .parse = parse_nla_vrftable, 1589 .put = put_nla_vrftable, 1590 .cmp = cmp_nla_vrftable }, 1591 1592 [SEG6_LOCAL_COUNTERS] = { .parse = parse_nla_counters, 1593 .put = put_nla_counters, 1594 .cmp = cmp_nla_counters, 1595 .destroy = destroy_attr_counters }, 1596 }; 1597 1598 /* call the destroy() callback (if available) for each set attribute in 1599 * @parsed_attrs, starting from the first attribute up to the @max_parsed 1600 * (excluded) attribute. 1601 */ 1602 static void __destroy_attrs(unsigned long parsed_attrs, int max_parsed, 1603 struct seg6_local_lwt *slwt) 1604 { 1605 struct seg6_action_param *param; 1606 int i; 1607 1608 /* Every required seg6local attribute is identified by an ID which is 1609 * encoded as a flag (i.e: 1 << ID) in the 'attrs' bitmask; 1610 * 1611 * We scan the 'parsed_attrs' bitmask, starting from the first attribute 1612 * up to the @max_parsed (excluded) attribute. 1613 * For each set attribute, we retrieve the corresponding destroy() 1614 * callback. If the callback is not available, then we skip to the next 1615 * attribute; otherwise, we call the destroy() callback. 1616 */ 1617 for (i = SEG6_LOCAL_SRH; i < max_parsed; ++i) { 1618 if (!(parsed_attrs & SEG6_F_ATTR(i))) 1619 continue; 1620 1621 param = &seg6_action_params[i]; 1622 1623 if (param->destroy) 1624 param->destroy(slwt); 1625 } 1626 } 1627 1628 /* release all the resources that may have been acquired during parsing 1629 * operations. 1630 */ 1631 static void destroy_attrs(struct seg6_local_lwt *slwt) 1632 { 1633 unsigned long attrs = slwt->desc->attrs | slwt->parsed_optattrs; 1634 1635 __destroy_attrs(attrs, SEG6_LOCAL_MAX + 1, slwt); 1636 } 1637 1638 static int parse_nla_optional_attrs(struct nlattr **attrs, 1639 struct seg6_local_lwt *slwt) 1640 { 1641 struct seg6_action_desc *desc = slwt->desc; 1642 unsigned long parsed_optattrs = 0; 1643 struct seg6_action_param *param; 1644 int err, i; 1645 1646 for (i = SEG6_LOCAL_SRH; i < SEG6_LOCAL_MAX + 1; ++i) { 1647 if (!(desc->optattrs & SEG6_F_ATTR(i)) || !attrs[i]) 1648 continue; 1649 1650 /* once here, the i-th attribute is provided by the 1651 * userspace AND it is identified optional as well. 1652 */ 1653 param = &seg6_action_params[i]; 1654 1655 err = param->parse(attrs, slwt); 1656 if (err < 0) 1657 goto parse_optattrs_err; 1658 1659 /* current attribute has been correctly parsed */ 1660 parsed_optattrs |= SEG6_F_ATTR(i); 1661 } 1662 1663 /* store in the tunnel state all the optional attributed successfully 1664 * parsed. 1665 */ 1666 slwt->parsed_optattrs = parsed_optattrs; 1667 1668 return 0; 1669 1670 parse_optattrs_err: 1671 __destroy_attrs(parsed_optattrs, i, slwt); 1672 1673 return err; 1674 } 1675 1676 /* call the custom constructor of the behavior during its initialization phase 1677 * and after that all its attributes have been parsed successfully. 1678 */ 1679 static int 1680 seg6_local_lwtunnel_build_state(struct seg6_local_lwt *slwt, const void *cfg, 1681 struct netlink_ext_ack *extack) 1682 { 1683 struct seg6_action_desc *desc = slwt->desc; 1684 struct seg6_local_lwtunnel_ops *ops; 1685 1686 ops = &desc->slwt_ops; 1687 if (!ops->build_state) 1688 return 0; 1689 1690 return ops->build_state(slwt, cfg, extack); 1691 } 1692 1693 /* call the custom destructor of the behavior which is invoked before the 1694 * tunnel is going to be destroyed. 1695 */ 1696 static void seg6_local_lwtunnel_destroy_state(struct seg6_local_lwt *slwt) 1697 { 1698 struct seg6_action_desc *desc = slwt->desc; 1699 struct seg6_local_lwtunnel_ops *ops; 1700 1701 ops = &desc->slwt_ops; 1702 if (!ops->destroy_state) 1703 return; 1704 1705 ops->destroy_state(slwt); 1706 } 1707 1708 static int parse_nla_action(struct nlattr **attrs, struct seg6_local_lwt *slwt) 1709 { 1710 struct seg6_action_param *param; 1711 struct seg6_action_desc *desc; 1712 unsigned long invalid_attrs; 1713 int i, err; 1714 1715 desc = __get_action_desc(slwt->action); 1716 if (!desc) 1717 return -EINVAL; 1718 1719 if (!desc->input) 1720 return -EOPNOTSUPP; 1721 1722 slwt->desc = desc; 1723 slwt->headroom += desc->static_headroom; 1724 1725 /* Forcing the desc->optattrs *set* and the desc->attrs *set* to be 1726 * disjoined, this allow us to release acquired resources by optional 1727 * attributes and by required attributes independently from each other 1728 * without any interference. 1729 * In other terms, we are sure that we do not release some the acquired 1730 * resources twice. 1731 * 1732 * Note that if an attribute is configured both as required and as 1733 * optional, it means that the user has messed something up in the 1734 * seg6_action_table. Therefore, this check is required for SRv6 1735 * behaviors to work properly. 1736 */ 1737 invalid_attrs = desc->attrs & desc->optattrs; 1738 if (invalid_attrs) { 1739 WARN_ONCE(1, 1740 "An attribute cannot be both required AND optional"); 1741 return -EINVAL; 1742 } 1743 1744 /* parse the required attributes */ 1745 for (i = SEG6_LOCAL_SRH; i < SEG6_LOCAL_MAX + 1; i++) { 1746 if (desc->attrs & SEG6_F_ATTR(i)) { 1747 if (!attrs[i]) 1748 return -EINVAL; 1749 1750 param = &seg6_action_params[i]; 1751 1752 err = param->parse(attrs, slwt); 1753 if (err < 0) 1754 goto parse_attrs_err; 1755 } 1756 } 1757 1758 /* parse the optional attributes, if any */ 1759 err = parse_nla_optional_attrs(attrs, slwt); 1760 if (err < 0) 1761 goto parse_attrs_err; 1762 1763 return 0; 1764 1765 parse_attrs_err: 1766 /* release any resource that may have been acquired during the i-1 1767 * parse() operations. 1768 */ 1769 __destroy_attrs(desc->attrs, i, slwt); 1770 1771 return err; 1772 } 1773 1774 static int seg6_local_build_state(struct net *net, struct nlattr *nla, 1775 unsigned int family, const void *cfg, 1776 struct lwtunnel_state **ts, 1777 struct netlink_ext_ack *extack) 1778 { 1779 struct nlattr *tb[SEG6_LOCAL_MAX + 1]; 1780 struct lwtunnel_state *newts; 1781 struct seg6_local_lwt *slwt; 1782 int err; 1783 1784 if (family != AF_INET6) 1785 return -EINVAL; 1786 1787 err = nla_parse_nested_deprecated(tb, SEG6_LOCAL_MAX, nla, 1788 seg6_local_policy, extack); 1789 1790 if (err < 0) 1791 return err; 1792 1793 if (!tb[SEG6_LOCAL_ACTION]) 1794 return -EINVAL; 1795 1796 newts = lwtunnel_state_alloc(sizeof(*slwt)); 1797 if (!newts) 1798 return -ENOMEM; 1799 1800 slwt = seg6_local_lwtunnel(newts); 1801 slwt->action = nla_get_u32(tb[SEG6_LOCAL_ACTION]); 1802 1803 err = parse_nla_action(tb, slwt); 1804 if (err < 0) 1805 goto out_free; 1806 1807 err = seg6_local_lwtunnel_build_state(slwt, cfg, extack); 1808 if (err < 0) 1809 goto out_destroy_attrs; 1810 1811 newts->type = LWTUNNEL_ENCAP_SEG6_LOCAL; 1812 newts->flags = LWTUNNEL_STATE_INPUT_REDIRECT; 1813 newts->headroom = slwt->headroom; 1814 1815 *ts = newts; 1816 1817 return 0; 1818 1819 out_destroy_attrs: 1820 destroy_attrs(slwt); 1821 out_free: 1822 kfree(newts); 1823 return err; 1824 } 1825 1826 static void seg6_local_destroy_state(struct lwtunnel_state *lwt) 1827 { 1828 struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt); 1829 1830 seg6_local_lwtunnel_destroy_state(slwt); 1831 1832 destroy_attrs(slwt); 1833 1834 return; 1835 } 1836 1837 static int seg6_local_fill_encap(struct sk_buff *skb, 1838 struct lwtunnel_state *lwt) 1839 { 1840 struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt); 1841 struct seg6_action_param *param; 1842 unsigned long attrs; 1843 int i, err; 1844 1845 if (nla_put_u32(skb, SEG6_LOCAL_ACTION, slwt->action)) 1846 return -EMSGSIZE; 1847 1848 attrs = slwt->desc->attrs | slwt->parsed_optattrs; 1849 1850 for (i = SEG6_LOCAL_SRH; i < SEG6_LOCAL_MAX + 1; i++) { 1851 if (attrs & SEG6_F_ATTR(i)) { 1852 param = &seg6_action_params[i]; 1853 err = param->put(skb, slwt); 1854 if (err < 0) 1855 return err; 1856 } 1857 } 1858 1859 return 0; 1860 } 1861 1862 static int seg6_local_get_encap_size(struct lwtunnel_state *lwt) 1863 { 1864 struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt); 1865 unsigned long attrs; 1866 int nlsize; 1867 1868 nlsize = nla_total_size(4); /* action */ 1869 1870 attrs = slwt->desc->attrs | slwt->parsed_optattrs; 1871 1872 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_SRH)) 1873 nlsize += nla_total_size((slwt->srh->hdrlen + 1) << 3); 1874 1875 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_TABLE)) 1876 nlsize += nla_total_size(4); 1877 1878 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_NH4)) 1879 nlsize += nla_total_size(4); 1880 1881 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_NH6)) 1882 nlsize += nla_total_size(16); 1883 1884 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_IIF)) 1885 nlsize += nla_total_size(4); 1886 1887 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_OIF)) 1888 nlsize += nla_total_size(4); 1889 1890 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_BPF)) 1891 nlsize += nla_total_size(sizeof(struct nlattr)) + 1892 nla_total_size(MAX_PROG_NAME) + 1893 nla_total_size(4); 1894 1895 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE)) 1896 nlsize += nla_total_size(4); 1897 1898 if (attrs & SEG6_F_LOCAL_COUNTERS) 1899 nlsize += nla_total_size(0) + /* nest SEG6_LOCAL_COUNTERS */ 1900 /* SEG6_LOCAL_CNT_PACKETS */ 1901 nla_total_size_64bit(sizeof(__u64)) + 1902 /* SEG6_LOCAL_CNT_BYTES */ 1903 nla_total_size_64bit(sizeof(__u64)) + 1904 /* SEG6_LOCAL_CNT_ERRORS */ 1905 nla_total_size_64bit(sizeof(__u64)); 1906 1907 return nlsize; 1908 } 1909 1910 static int seg6_local_cmp_encap(struct lwtunnel_state *a, 1911 struct lwtunnel_state *b) 1912 { 1913 struct seg6_local_lwt *slwt_a, *slwt_b; 1914 struct seg6_action_param *param; 1915 unsigned long attrs_a, attrs_b; 1916 int i; 1917 1918 slwt_a = seg6_local_lwtunnel(a); 1919 slwt_b = seg6_local_lwtunnel(b); 1920 1921 if (slwt_a->action != slwt_b->action) 1922 return 1; 1923 1924 attrs_a = slwt_a->desc->attrs | slwt_a->parsed_optattrs; 1925 attrs_b = slwt_b->desc->attrs | slwt_b->parsed_optattrs; 1926 1927 if (attrs_a != attrs_b) 1928 return 1; 1929 1930 for (i = SEG6_LOCAL_SRH; i < SEG6_LOCAL_MAX + 1; i++) { 1931 if (attrs_a & SEG6_F_ATTR(i)) { 1932 param = &seg6_action_params[i]; 1933 if (param->cmp(slwt_a, slwt_b)) 1934 return 1; 1935 } 1936 } 1937 1938 return 0; 1939 } 1940 1941 static const struct lwtunnel_encap_ops seg6_local_ops = { 1942 .build_state = seg6_local_build_state, 1943 .destroy_state = seg6_local_destroy_state, 1944 .input = seg6_local_input, 1945 .fill_encap = seg6_local_fill_encap, 1946 .get_encap_size = seg6_local_get_encap_size, 1947 .cmp_encap = seg6_local_cmp_encap, 1948 .owner = THIS_MODULE, 1949 }; 1950 1951 int __init seg6_local_init(void) 1952 { 1953 /* If the max total number of defined attributes is reached, then your 1954 * kernel build stops here. 1955 * 1956 * This check is required to avoid arithmetic overflows when processing 1957 * behavior attributes and the maximum number of defined attributes 1958 * exceeds the allowed value. 1959 */ 1960 BUILD_BUG_ON(SEG6_LOCAL_MAX + 1 > BITS_PER_TYPE(unsigned long)); 1961 1962 return lwtunnel_encap_add_ops(&seg6_local_ops, 1963 LWTUNNEL_ENCAP_SEG6_LOCAL); 1964 } 1965 1966 void seg6_local_exit(void) 1967 { 1968 lwtunnel_encap_del_ops(&seg6_local_ops, LWTUNNEL_ENCAP_SEG6_LOCAL); 1969 } 1970