1 // SPDX-License-Identifier: GPL-2.0-only 2 #include <linux/kernel.h> 3 #include <linux/skbuff.h> 4 #include <linux/export.h> 5 #include <linux/ip.h> 6 #include <linux/ipv6.h> 7 #include <linux/if_vlan.h> 8 #include <linux/filter.h> 9 #include <net/dsa.h> 10 #include <net/dst_metadata.h> 11 #include <net/ip.h> 12 #include <net/ipv6.h> 13 #include <net/gre.h> 14 #include <net/pptp.h> 15 #include <net/tipc.h> 16 #include <linux/igmp.h> 17 #include <linux/icmp.h> 18 #include <linux/sctp.h> 19 #include <linux/dccp.h> 20 #include <linux/if_tunnel.h> 21 #include <linux/if_pppox.h> 22 #include <linux/ppp_defs.h> 23 #include <linux/stddef.h> 24 #include <linux/if_ether.h> 25 #include <linux/if_hsr.h> 26 #include <linux/mpls.h> 27 #include <linux/tcp.h> 28 #include <linux/ptp_classify.h> 29 #include <net/flow_dissector.h> 30 #include <net/pkt_cls.h> 31 #include <scsi/fc/fc_fcoe.h> 32 #include <uapi/linux/batadv_packet.h> 33 #include <linux/bpf.h> 34 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 35 #include <net/netfilter/nf_conntrack_core.h> 36 #include <net/netfilter/nf_conntrack_labels.h> 37 #endif 38 #include <linux/bpf-netns.h> 39 40 static void dissector_set_key(struct flow_dissector *flow_dissector, 41 enum flow_dissector_key_id key_id) 42 { 43 flow_dissector->used_keys |= (1ULL << key_id); 44 } 45 46 void skb_flow_dissector_init(struct flow_dissector *flow_dissector, 47 const struct flow_dissector_key *key, 48 unsigned int key_count) 49 { 50 unsigned int i; 51 52 memset(flow_dissector, 0, sizeof(*flow_dissector)); 53 54 for (i = 0; i < key_count; i++, key++) { 55 /* User should make sure that every key target offset is within 56 * boundaries of unsigned short. 57 */ 58 BUG_ON(key->offset > USHRT_MAX); 59 BUG_ON(dissector_uses_key(flow_dissector, 60 key->key_id)); 61 62 dissector_set_key(flow_dissector, key->key_id); 63 flow_dissector->offset[key->key_id] = key->offset; 64 } 65 66 /* Ensure that the dissector always includes control and basic key. 67 * That way we are able to avoid handling lack of these in fast path. 68 */ 69 BUG_ON(!dissector_uses_key(flow_dissector, 70 FLOW_DISSECTOR_KEY_CONTROL)); 71 BUG_ON(!dissector_uses_key(flow_dissector, 72 FLOW_DISSECTOR_KEY_BASIC)); 73 } 74 EXPORT_SYMBOL(skb_flow_dissector_init); 75 76 #ifdef CONFIG_BPF_SYSCALL 77 int flow_dissector_bpf_prog_attach_check(struct net *net, 78 struct bpf_prog *prog) 79 { 80 enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR; 81 82 if (net == &init_net) { 83 /* BPF flow dissector in the root namespace overrides 84 * any per-net-namespace one. When attaching to root, 85 * make sure we don't have any BPF program attached 86 * to the non-root namespaces. 87 */ 88 struct net *ns; 89 90 for_each_net(ns) { 91 if (ns == &init_net) 92 continue; 93 if (rcu_access_pointer(ns->bpf.run_array[type])) 94 return -EEXIST; 95 } 96 } else { 97 /* Make sure root flow dissector is not attached 98 * when attaching to the non-root namespace. 99 */ 100 if (rcu_access_pointer(init_net.bpf.run_array[type])) 101 return -EEXIST; 102 } 103 104 return 0; 105 } 106 #endif /* CONFIG_BPF_SYSCALL */ 107 108 /** 109 * skb_flow_get_ports - extract the upper layer ports and return them 110 * @skb: sk_buff to extract the ports from 111 * @thoff: transport header offset 112 * @ip_proto: protocol for which to get port offset 113 * @data: raw buffer pointer to the packet, if NULL use skb->data 114 * @hlen: packet header length, if @data is NULL use skb_headlen(skb) 115 * 116 * The function will try to retrieve the ports at offset thoff + poff where poff 117 * is the protocol port offset returned from proto_ports_offset 118 */ 119 __be32 skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto, 120 const void *data, int hlen) 121 { 122 int poff = proto_ports_offset(ip_proto); 123 124 if (!data) { 125 data = skb->data; 126 hlen = skb_headlen(skb); 127 } 128 129 if (poff >= 0) { 130 __be32 *ports, _ports; 131 132 ports = __skb_header_pointer(skb, thoff + poff, 133 sizeof(_ports), data, hlen, &_ports); 134 if (ports) 135 return *ports; 136 } 137 138 return 0; 139 } 140 EXPORT_SYMBOL(skb_flow_get_ports); 141 142 static bool icmp_has_id(u8 type) 143 { 144 switch (type) { 145 case ICMP_ECHO: 146 case ICMP_ECHOREPLY: 147 case ICMP_TIMESTAMP: 148 case ICMP_TIMESTAMPREPLY: 149 case ICMPV6_ECHO_REQUEST: 150 case ICMPV6_ECHO_REPLY: 151 return true; 152 } 153 154 return false; 155 } 156 157 /** 158 * skb_flow_get_icmp_tci - extract ICMP(6) Type, Code and Identifier fields 159 * @skb: sk_buff to extract from 160 * @key_icmp: struct flow_dissector_key_icmp to fill 161 * @data: raw buffer pointer to the packet 162 * @thoff: offset to extract at 163 * @hlen: packet header length 164 */ 165 void skb_flow_get_icmp_tci(const struct sk_buff *skb, 166 struct flow_dissector_key_icmp *key_icmp, 167 const void *data, int thoff, int hlen) 168 { 169 struct icmphdr *ih, _ih; 170 171 ih = __skb_header_pointer(skb, thoff, sizeof(_ih), data, hlen, &_ih); 172 if (!ih) 173 return; 174 175 key_icmp->type = ih->type; 176 key_icmp->code = ih->code; 177 178 /* As we use 0 to signal that the Id field is not present, 179 * avoid confusion with packets without such field 180 */ 181 if (icmp_has_id(ih->type)) 182 key_icmp->id = ih->un.echo.id ? ntohs(ih->un.echo.id) : 1; 183 else 184 key_icmp->id = 0; 185 } 186 EXPORT_SYMBOL(skb_flow_get_icmp_tci); 187 188 /* If FLOW_DISSECTOR_KEY_ICMP is set, dissect an ICMP packet 189 * using skb_flow_get_icmp_tci(). 190 */ 191 static void __skb_flow_dissect_icmp(const struct sk_buff *skb, 192 struct flow_dissector *flow_dissector, 193 void *target_container, const void *data, 194 int thoff, int hlen) 195 { 196 struct flow_dissector_key_icmp *key_icmp; 197 198 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ICMP)) 199 return; 200 201 key_icmp = skb_flow_dissector_target(flow_dissector, 202 FLOW_DISSECTOR_KEY_ICMP, 203 target_container); 204 205 skb_flow_get_icmp_tci(skb, key_icmp, data, thoff, hlen); 206 } 207 208 static void __skb_flow_dissect_ah(const struct sk_buff *skb, 209 struct flow_dissector *flow_dissector, 210 void *target_container, const void *data, 211 int nhoff, int hlen) 212 { 213 struct flow_dissector_key_ipsec *key_ah; 214 struct ip_auth_hdr _hdr, *hdr; 215 216 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPSEC)) 217 return; 218 219 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr); 220 if (!hdr) 221 return; 222 223 key_ah = skb_flow_dissector_target(flow_dissector, 224 FLOW_DISSECTOR_KEY_IPSEC, 225 target_container); 226 227 key_ah->spi = hdr->spi; 228 } 229 230 static void __skb_flow_dissect_esp(const struct sk_buff *skb, 231 struct flow_dissector *flow_dissector, 232 void *target_container, const void *data, 233 int nhoff, int hlen) 234 { 235 struct flow_dissector_key_ipsec *key_esp; 236 struct ip_esp_hdr _hdr, *hdr; 237 238 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPSEC)) 239 return; 240 241 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr); 242 if (!hdr) 243 return; 244 245 key_esp = skb_flow_dissector_target(flow_dissector, 246 FLOW_DISSECTOR_KEY_IPSEC, 247 target_container); 248 249 key_esp->spi = hdr->spi; 250 } 251 252 static void __skb_flow_dissect_l2tpv3(const struct sk_buff *skb, 253 struct flow_dissector *flow_dissector, 254 void *target_container, const void *data, 255 int nhoff, int hlen) 256 { 257 struct flow_dissector_key_l2tpv3 *key_l2tpv3; 258 struct { 259 __be32 session_id; 260 } *hdr, _hdr; 261 262 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_L2TPV3)) 263 return; 264 265 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr); 266 if (!hdr) 267 return; 268 269 key_l2tpv3 = skb_flow_dissector_target(flow_dissector, 270 FLOW_DISSECTOR_KEY_L2TPV3, 271 target_container); 272 273 key_l2tpv3->session_id = hdr->session_id; 274 } 275 276 void skb_flow_dissect_meta(const struct sk_buff *skb, 277 struct flow_dissector *flow_dissector, 278 void *target_container) 279 { 280 struct flow_dissector_key_meta *meta; 281 282 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_META)) 283 return; 284 285 meta = skb_flow_dissector_target(flow_dissector, 286 FLOW_DISSECTOR_KEY_META, 287 target_container); 288 meta->ingress_ifindex = skb->skb_iif; 289 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) 290 if (tc_skb_ext_tc_enabled()) { 291 struct tc_skb_ext *ext; 292 293 ext = skb_ext_find(skb, TC_SKB_EXT); 294 if (ext) 295 meta->l2_miss = ext->l2_miss; 296 } 297 #endif 298 } 299 EXPORT_SYMBOL(skb_flow_dissect_meta); 300 301 static void 302 skb_flow_dissect_set_enc_control(enum flow_dissector_key_id type, 303 u32 ctrl_flags, 304 struct flow_dissector *flow_dissector, 305 void *target_container) 306 { 307 struct flow_dissector_key_control *ctrl; 308 309 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_CONTROL)) 310 return; 311 312 ctrl = skb_flow_dissector_target(flow_dissector, 313 FLOW_DISSECTOR_KEY_ENC_CONTROL, 314 target_container); 315 ctrl->addr_type = type; 316 ctrl->flags = ctrl_flags; 317 } 318 319 void 320 skb_flow_dissect_ct(const struct sk_buff *skb, 321 struct flow_dissector *flow_dissector, 322 void *target_container, u16 *ctinfo_map, 323 size_t mapsize, bool post_ct, u16 zone) 324 { 325 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 326 struct flow_dissector_key_ct *key; 327 enum ip_conntrack_info ctinfo; 328 struct nf_conn_labels *cl; 329 struct nf_conn *ct; 330 331 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_CT)) 332 return; 333 334 ct = nf_ct_get(skb, &ctinfo); 335 if (!ct && !post_ct) 336 return; 337 338 key = skb_flow_dissector_target(flow_dissector, 339 FLOW_DISSECTOR_KEY_CT, 340 target_container); 341 342 if (!ct) { 343 key->ct_state = TCA_FLOWER_KEY_CT_FLAGS_TRACKED | 344 TCA_FLOWER_KEY_CT_FLAGS_INVALID; 345 key->ct_zone = zone; 346 return; 347 } 348 349 if (ctinfo < mapsize) 350 key->ct_state = ctinfo_map[ctinfo]; 351 #if IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) 352 key->ct_zone = ct->zone.id; 353 #endif 354 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) 355 key->ct_mark = READ_ONCE(ct->mark); 356 #endif 357 358 cl = nf_ct_labels_find(ct); 359 if (cl) 360 memcpy(key->ct_labels, cl->bits, sizeof(key->ct_labels)); 361 #endif /* CONFIG_NF_CONNTRACK */ 362 } 363 EXPORT_SYMBOL(skb_flow_dissect_ct); 364 365 void 366 skb_flow_dissect_tunnel_info(const struct sk_buff *skb, 367 struct flow_dissector *flow_dissector, 368 void *target_container) 369 { 370 struct ip_tunnel_info *info; 371 struct ip_tunnel_key *key; 372 u32 ctrl_flags = 0; 373 374 /* A quick check to see if there might be something to do. */ 375 if (!dissector_uses_key(flow_dissector, 376 FLOW_DISSECTOR_KEY_ENC_KEYID) && 377 !dissector_uses_key(flow_dissector, 378 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) && 379 !dissector_uses_key(flow_dissector, 380 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) && 381 !dissector_uses_key(flow_dissector, 382 FLOW_DISSECTOR_KEY_ENC_CONTROL) && 383 !dissector_uses_key(flow_dissector, 384 FLOW_DISSECTOR_KEY_ENC_PORTS) && 385 !dissector_uses_key(flow_dissector, 386 FLOW_DISSECTOR_KEY_ENC_IP) && 387 !dissector_uses_key(flow_dissector, 388 FLOW_DISSECTOR_KEY_ENC_OPTS)) 389 return; 390 391 info = skb_tunnel_info(skb); 392 if (!info) 393 return; 394 395 key = &info->key; 396 397 if (test_bit(IP_TUNNEL_CSUM_BIT, key->tun_flags)) 398 ctrl_flags |= FLOW_DIS_F_TUNNEL_CSUM; 399 if (test_bit(IP_TUNNEL_DONT_FRAGMENT_BIT, key->tun_flags)) 400 ctrl_flags |= FLOW_DIS_F_TUNNEL_DONT_FRAGMENT; 401 if (test_bit(IP_TUNNEL_OAM_BIT, key->tun_flags)) 402 ctrl_flags |= FLOW_DIS_F_TUNNEL_OAM; 403 if (test_bit(IP_TUNNEL_CRIT_OPT_BIT, key->tun_flags)) 404 ctrl_flags |= FLOW_DIS_F_TUNNEL_CRIT_OPT; 405 406 switch (ip_tunnel_info_af(info)) { 407 case AF_INET: 408 skb_flow_dissect_set_enc_control(FLOW_DISSECTOR_KEY_IPV4_ADDRS, 409 ctrl_flags, flow_dissector, 410 target_container); 411 if (dissector_uses_key(flow_dissector, 412 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) { 413 struct flow_dissector_key_ipv4_addrs *ipv4; 414 415 ipv4 = skb_flow_dissector_target(flow_dissector, 416 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS, 417 target_container); 418 ipv4->src = key->u.ipv4.src; 419 ipv4->dst = key->u.ipv4.dst; 420 } 421 break; 422 case AF_INET6: 423 skb_flow_dissect_set_enc_control(FLOW_DISSECTOR_KEY_IPV6_ADDRS, 424 ctrl_flags, flow_dissector, 425 target_container); 426 if (dissector_uses_key(flow_dissector, 427 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) { 428 struct flow_dissector_key_ipv6_addrs *ipv6; 429 430 ipv6 = skb_flow_dissector_target(flow_dissector, 431 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS, 432 target_container); 433 ipv6->src = key->u.ipv6.src; 434 ipv6->dst = key->u.ipv6.dst; 435 } 436 break; 437 default: 438 skb_flow_dissect_set_enc_control(0, ctrl_flags, flow_dissector, 439 target_container); 440 break; 441 } 442 443 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) { 444 struct flow_dissector_key_keyid *keyid; 445 446 keyid = skb_flow_dissector_target(flow_dissector, 447 FLOW_DISSECTOR_KEY_ENC_KEYID, 448 target_container); 449 keyid->keyid = tunnel_id_to_key32(key->tun_id); 450 } 451 452 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) { 453 struct flow_dissector_key_ports *tp; 454 455 tp = skb_flow_dissector_target(flow_dissector, 456 FLOW_DISSECTOR_KEY_ENC_PORTS, 457 target_container); 458 tp->src = key->tp_src; 459 tp->dst = key->tp_dst; 460 } 461 462 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_IP)) { 463 struct flow_dissector_key_ip *ip; 464 465 ip = skb_flow_dissector_target(flow_dissector, 466 FLOW_DISSECTOR_KEY_ENC_IP, 467 target_container); 468 ip->tos = key->tos; 469 ip->ttl = key->ttl; 470 } 471 472 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_OPTS)) { 473 struct flow_dissector_key_enc_opts *enc_opt; 474 IP_TUNNEL_DECLARE_FLAGS(flags) = { }; 475 u32 val; 476 477 enc_opt = skb_flow_dissector_target(flow_dissector, 478 FLOW_DISSECTOR_KEY_ENC_OPTS, 479 target_container); 480 481 if (!info->options_len) 482 return; 483 484 enc_opt->len = info->options_len; 485 ip_tunnel_info_opts_get(enc_opt->data, info); 486 487 ip_tunnel_set_options_present(flags); 488 ip_tunnel_flags_and(flags, info->key.tun_flags, flags); 489 490 val = find_next_bit(flags, __IP_TUNNEL_FLAG_NUM, 491 IP_TUNNEL_GENEVE_OPT_BIT); 492 enc_opt->dst_opt_type = val < __IP_TUNNEL_FLAG_NUM ? val : 0; 493 } 494 } 495 EXPORT_SYMBOL(skb_flow_dissect_tunnel_info); 496 497 void skb_flow_dissect_hash(const struct sk_buff *skb, 498 struct flow_dissector *flow_dissector, 499 void *target_container) 500 { 501 struct flow_dissector_key_hash *key; 502 503 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_HASH)) 504 return; 505 506 key = skb_flow_dissector_target(flow_dissector, 507 FLOW_DISSECTOR_KEY_HASH, 508 target_container); 509 510 key->hash = skb_get_hash_raw(skb); 511 } 512 EXPORT_SYMBOL(skb_flow_dissect_hash); 513 514 static enum flow_dissect_ret 515 __skb_flow_dissect_mpls(const struct sk_buff *skb, 516 struct flow_dissector *flow_dissector, 517 void *target_container, const void *data, int nhoff, 518 int hlen, int lse_index, bool *entropy_label) 519 { 520 struct mpls_label *hdr, _hdr; 521 u32 entry, label, bos; 522 523 if (!dissector_uses_key(flow_dissector, 524 FLOW_DISSECTOR_KEY_MPLS_ENTROPY) && 525 !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) 526 return FLOW_DISSECT_RET_OUT_GOOD; 527 528 if (lse_index >= FLOW_DIS_MPLS_MAX) 529 return FLOW_DISSECT_RET_OUT_GOOD; 530 531 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, 532 hlen, &_hdr); 533 if (!hdr) 534 return FLOW_DISSECT_RET_OUT_BAD; 535 536 entry = ntohl(hdr->entry); 537 label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT; 538 bos = (entry & MPLS_LS_S_MASK) >> MPLS_LS_S_SHIFT; 539 540 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) { 541 struct flow_dissector_key_mpls *key_mpls; 542 struct flow_dissector_mpls_lse *lse; 543 544 key_mpls = skb_flow_dissector_target(flow_dissector, 545 FLOW_DISSECTOR_KEY_MPLS, 546 target_container); 547 lse = &key_mpls->ls[lse_index]; 548 549 lse->mpls_ttl = (entry & MPLS_LS_TTL_MASK) >> MPLS_LS_TTL_SHIFT; 550 lse->mpls_bos = bos; 551 lse->mpls_tc = (entry & MPLS_LS_TC_MASK) >> MPLS_LS_TC_SHIFT; 552 lse->mpls_label = label; 553 dissector_set_mpls_lse(key_mpls, lse_index); 554 } 555 556 if (*entropy_label && 557 dissector_uses_key(flow_dissector, 558 FLOW_DISSECTOR_KEY_MPLS_ENTROPY)) { 559 struct flow_dissector_key_keyid *key_keyid; 560 561 key_keyid = skb_flow_dissector_target(flow_dissector, 562 FLOW_DISSECTOR_KEY_MPLS_ENTROPY, 563 target_container); 564 key_keyid->keyid = cpu_to_be32(label); 565 } 566 567 *entropy_label = label == MPLS_LABEL_ENTROPY; 568 569 return bos ? FLOW_DISSECT_RET_OUT_GOOD : FLOW_DISSECT_RET_PROTO_AGAIN; 570 } 571 572 static enum flow_dissect_ret 573 __skb_flow_dissect_arp(const struct sk_buff *skb, 574 struct flow_dissector *flow_dissector, 575 void *target_container, const void *data, 576 int nhoff, int hlen) 577 { 578 struct flow_dissector_key_arp *key_arp; 579 struct { 580 unsigned char ar_sha[ETH_ALEN]; 581 unsigned char ar_sip[4]; 582 unsigned char ar_tha[ETH_ALEN]; 583 unsigned char ar_tip[4]; 584 } *arp_eth, _arp_eth; 585 const struct arphdr *arp; 586 struct arphdr _arp; 587 588 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP)) 589 return FLOW_DISSECT_RET_OUT_GOOD; 590 591 arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data, 592 hlen, &_arp); 593 if (!arp) 594 return FLOW_DISSECT_RET_OUT_BAD; 595 596 if (arp->ar_hrd != htons(ARPHRD_ETHER) || 597 arp->ar_pro != htons(ETH_P_IP) || 598 arp->ar_hln != ETH_ALEN || 599 arp->ar_pln != 4 || 600 (arp->ar_op != htons(ARPOP_REPLY) && 601 arp->ar_op != htons(ARPOP_REQUEST))) 602 return FLOW_DISSECT_RET_OUT_BAD; 603 604 arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp), 605 sizeof(_arp_eth), data, 606 hlen, &_arp_eth); 607 if (!arp_eth) 608 return FLOW_DISSECT_RET_OUT_BAD; 609 610 key_arp = skb_flow_dissector_target(flow_dissector, 611 FLOW_DISSECTOR_KEY_ARP, 612 target_container); 613 614 memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip)); 615 memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip)); 616 617 /* Only store the lower byte of the opcode; 618 * this covers ARPOP_REPLY and ARPOP_REQUEST. 619 */ 620 key_arp->op = ntohs(arp->ar_op) & 0xff; 621 622 ether_addr_copy(key_arp->sha, arp_eth->ar_sha); 623 ether_addr_copy(key_arp->tha, arp_eth->ar_tha); 624 625 return FLOW_DISSECT_RET_OUT_GOOD; 626 } 627 628 static enum flow_dissect_ret 629 __skb_flow_dissect_cfm(const struct sk_buff *skb, 630 struct flow_dissector *flow_dissector, 631 void *target_container, const void *data, 632 int nhoff, int hlen) 633 { 634 struct flow_dissector_key_cfm *key, *hdr, _hdr; 635 636 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_CFM)) 637 return FLOW_DISSECT_RET_OUT_GOOD; 638 639 hdr = __skb_header_pointer(skb, nhoff, sizeof(*key), data, hlen, &_hdr); 640 if (!hdr) 641 return FLOW_DISSECT_RET_OUT_BAD; 642 643 key = skb_flow_dissector_target(flow_dissector, FLOW_DISSECTOR_KEY_CFM, 644 target_container); 645 646 key->mdl_ver = hdr->mdl_ver; 647 key->opcode = hdr->opcode; 648 649 return FLOW_DISSECT_RET_OUT_GOOD; 650 } 651 652 static enum flow_dissect_ret 653 __skb_flow_dissect_gre(const struct sk_buff *skb, 654 struct flow_dissector_key_control *key_control, 655 struct flow_dissector *flow_dissector, 656 void *target_container, const void *data, 657 __be16 *p_proto, int *p_nhoff, int *p_hlen, 658 unsigned int flags) 659 { 660 struct flow_dissector_key_keyid *key_keyid; 661 struct gre_base_hdr *hdr, _hdr; 662 int offset = 0; 663 u16 gre_ver; 664 665 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), 666 data, *p_hlen, &_hdr); 667 if (!hdr) 668 return FLOW_DISSECT_RET_OUT_BAD; 669 670 /* Only look inside GRE without routing */ 671 if (hdr->flags & GRE_ROUTING) 672 return FLOW_DISSECT_RET_OUT_GOOD; 673 674 /* Only look inside GRE for version 0 and 1 */ 675 gre_ver = ntohs(hdr->flags & GRE_VERSION); 676 if (gre_ver > 1) 677 return FLOW_DISSECT_RET_OUT_GOOD; 678 679 *p_proto = hdr->protocol; 680 if (gre_ver) { 681 /* Version1 must be PPTP, and check the flags */ 682 if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY))) 683 return FLOW_DISSECT_RET_OUT_GOOD; 684 } 685 686 offset += sizeof(struct gre_base_hdr); 687 688 if (hdr->flags & GRE_CSUM) 689 offset += sizeof_field(struct gre_full_hdr, csum) + 690 sizeof_field(struct gre_full_hdr, reserved1); 691 692 if (hdr->flags & GRE_KEY) { 693 const __be32 *keyid; 694 __be32 _keyid; 695 696 keyid = __skb_header_pointer(skb, *p_nhoff + offset, 697 sizeof(_keyid), 698 data, *p_hlen, &_keyid); 699 if (!keyid) 700 return FLOW_DISSECT_RET_OUT_BAD; 701 702 if (dissector_uses_key(flow_dissector, 703 FLOW_DISSECTOR_KEY_GRE_KEYID)) { 704 key_keyid = skb_flow_dissector_target(flow_dissector, 705 FLOW_DISSECTOR_KEY_GRE_KEYID, 706 target_container); 707 if (gre_ver == 0) 708 key_keyid->keyid = *keyid; 709 else 710 key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK; 711 } 712 offset += sizeof_field(struct gre_full_hdr, key); 713 } 714 715 if (hdr->flags & GRE_SEQ) 716 offset += sizeof_field(struct pptp_gre_header, seq); 717 718 if (gre_ver == 0) { 719 if (*p_proto == htons(ETH_P_TEB)) { 720 const struct ethhdr *eth; 721 struct ethhdr _eth; 722 723 eth = __skb_header_pointer(skb, *p_nhoff + offset, 724 sizeof(_eth), 725 data, *p_hlen, &_eth); 726 if (!eth) 727 return FLOW_DISSECT_RET_OUT_BAD; 728 *p_proto = eth->h_proto; 729 offset += sizeof(*eth); 730 731 /* Cap headers that we access via pointers at the 732 * end of the Ethernet header as our maximum alignment 733 * at that point is only 2 bytes. 734 */ 735 if (NET_IP_ALIGN) 736 *p_hlen = *p_nhoff + offset; 737 } 738 } else { /* version 1, must be PPTP */ 739 u8 _ppp_hdr[PPP_HDRLEN]; 740 u8 *ppp_hdr; 741 742 if (hdr->flags & GRE_ACK) 743 offset += sizeof_field(struct pptp_gre_header, ack); 744 745 ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset, 746 sizeof(_ppp_hdr), 747 data, *p_hlen, _ppp_hdr); 748 if (!ppp_hdr) 749 return FLOW_DISSECT_RET_OUT_BAD; 750 751 switch (PPP_PROTOCOL(ppp_hdr)) { 752 case PPP_IP: 753 *p_proto = htons(ETH_P_IP); 754 break; 755 case PPP_IPV6: 756 *p_proto = htons(ETH_P_IPV6); 757 break; 758 default: 759 /* Could probably catch some more like MPLS */ 760 break; 761 } 762 763 offset += PPP_HDRLEN; 764 } 765 766 *p_nhoff += offset; 767 key_control->flags |= FLOW_DIS_ENCAPSULATION; 768 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) 769 return FLOW_DISSECT_RET_OUT_GOOD; 770 771 return FLOW_DISSECT_RET_PROTO_AGAIN; 772 } 773 774 /** 775 * __skb_flow_dissect_batadv() - dissect batman-adv header 776 * @skb: sk_buff to with the batman-adv header 777 * @key_control: flow dissectors control key 778 * @data: raw buffer pointer to the packet, if NULL use skb->data 779 * @p_proto: pointer used to update the protocol to process next 780 * @p_nhoff: pointer used to update inner network header offset 781 * @hlen: packet header length 782 * @flags: any combination of FLOW_DISSECTOR_F_* 783 * 784 * ETH_P_BATMAN packets are tried to be dissected. Only 785 * &struct batadv_unicast packets are actually processed because they contain an 786 * inner ethernet header and are usually followed by actual network header. This 787 * allows the flow dissector to continue processing the packet. 788 * 789 * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found, 790 * FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation, 791 * otherwise FLOW_DISSECT_RET_OUT_BAD 792 */ 793 static enum flow_dissect_ret 794 __skb_flow_dissect_batadv(const struct sk_buff *skb, 795 struct flow_dissector_key_control *key_control, 796 const void *data, __be16 *p_proto, int *p_nhoff, 797 int hlen, unsigned int flags) 798 { 799 struct { 800 struct batadv_unicast_packet batadv_unicast; 801 struct ethhdr eth; 802 } *hdr, _hdr; 803 804 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen, 805 &_hdr); 806 if (!hdr) 807 return FLOW_DISSECT_RET_OUT_BAD; 808 809 if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION) 810 return FLOW_DISSECT_RET_OUT_BAD; 811 812 if (hdr->batadv_unicast.packet_type != BATADV_UNICAST) 813 return FLOW_DISSECT_RET_OUT_BAD; 814 815 *p_proto = hdr->eth.h_proto; 816 *p_nhoff += sizeof(*hdr); 817 818 key_control->flags |= FLOW_DIS_ENCAPSULATION; 819 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) 820 return FLOW_DISSECT_RET_OUT_GOOD; 821 822 return FLOW_DISSECT_RET_PROTO_AGAIN; 823 } 824 825 static void 826 __skb_flow_dissect_tcp(const struct sk_buff *skb, 827 struct flow_dissector *flow_dissector, 828 void *target_container, const void *data, 829 int thoff, int hlen) 830 { 831 struct flow_dissector_key_tcp *key_tcp; 832 struct tcphdr *th, _th; 833 834 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP)) 835 return; 836 837 th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th); 838 if (!th) 839 return; 840 841 if (unlikely(__tcp_hdrlen(th) < sizeof(_th))) 842 return; 843 844 key_tcp = skb_flow_dissector_target(flow_dissector, 845 FLOW_DISSECTOR_KEY_TCP, 846 target_container); 847 key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF)); 848 } 849 850 static void 851 __skb_flow_dissect_ports(const struct sk_buff *skb, 852 struct flow_dissector *flow_dissector, 853 void *target_container, const void *data, 854 int nhoff, u8 ip_proto, int hlen) 855 { 856 struct flow_dissector_key_ports_range *key_ports_range = NULL; 857 struct flow_dissector_key_ports *key_ports = NULL; 858 __be32 ports; 859 860 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS)) 861 key_ports = skb_flow_dissector_target(flow_dissector, 862 FLOW_DISSECTOR_KEY_PORTS, 863 target_container); 864 865 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS_RANGE)) 866 key_ports_range = skb_flow_dissector_target(flow_dissector, 867 FLOW_DISSECTOR_KEY_PORTS_RANGE, 868 target_container); 869 870 if (!key_ports && !key_ports_range) 871 return; 872 873 ports = skb_flow_get_ports(skb, nhoff, ip_proto, data, hlen); 874 875 if (key_ports) 876 key_ports->ports = ports; 877 878 if (key_ports_range) 879 key_ports_range->tp.ports = ports; 880 } 881 882 static void 883 __skb_flow_dissect_ipv4(const struct sk_buff *skb, 884 struct flow_dissector *flow_dissector, 885 void *target_container, const void *data, 886 const struct iphdr *iph) 887 { 888 struct flow_dissector_key_ip *key_ip; 889 890 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP)) 891 return; 892 893 key_ip = skb_flow_dissector_target(flow_dissector, 894 FLOW_DISSECTOR_KEY_IP, 895 target_container); 896 key_ip->tos = iph->tos; 897 key_ip->ttl = iph->ttl; 898 } 899 900 static void 901 __skb_flow_dissect_ipv6(const struct sk_buff *skb, 902 struct flow_dissector *flow_dissector, 903 void *target_container, const void *data, 904 const struct ipv6hdr *iph) 905 { 906 struct flow_dissector_key_ip *key_ip; 907 908 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP)) 909 return; 910 911 key_ip = skb_flow_dissector_target(flow_dissector, 912 FLOW_DISSECTOR_KEY_IP, 913 target_container); 914 key_ip->tos = ipv6_get_dsfield(iph); 915 key_ip->ttl = iph->hop_limit; 916 } 917 918 /* Maximum number of protocol headers that can be parsed in 919 * __skb_flow_dissect 920 */ 921 #define MAX_FLOW_DISSECT_HDRS 15 922 923 static bool skb_flow_dissect_allowed(int *num_hdrs) 924 { 925 ++*num_hdrs; 926 927 return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS); 928 } 929 930 static void __skb_flow_bpf_to_target(const struct bpf_flow_keys *flow_keys, 931 struct flow_dissector *flow_dissector, 932 void *target_container) 933 { 934 struct flow_dissector_key_ports_range *key_ports_range = NULL; 935 struct flow_dissector_key_ports *key_ports = NULL; 936 struct flow_dissector_key_control *key_control; 937 struct flow_dissector_key_basic *key_basic; 938 struct flow_dissector_key_addrs *key_addrs; 939 struct flow_dissector_key_tags *key_tags; 940 941 key_control = skb_flow_dissector_target(flow_dissector, 942 FLOW_DISSECTOR_KEY_CONTROL, 943 target_container); 944 key_control->thoff = flow_keys->thoff; 945 if (flow_keys->is_frag) 946 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 947 if (flow_keys->is_first_frag) 948 key_control->flags |= FLOW_DIS_FIRST_FRAG; 949 if (flow_keys->is_encap) 950 key_control->flags |= FLOW_DIS_ENCAPSULATION; 951 952 key_basic = skb_flow_dissector_target(flow_dissector, 953 FLOW_DISSECTOR_KEY_BASIC, 954 target_container); 955 key_basic->n_proto = flow_keys->n_proto; 956 key_basic->ip_proto = flow_keys->ip_proto; 957 958 if (flow_keys->addr_proto == ETH_P_IP && 959 dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) { 960 key_addrs = skb_flow_dissector_target(flow_dissector, 961 FLOW_DISSECTOR_KEY_IPV4_ADDRS, 962 target_container); 963 key_addrs->v4addrs.src = flow_keys->ipv4_src; 964 key_addrs->v4addrs.dst = flow_keys->ipv4_dst; 965 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; 966 } else if (flow_keys->addr_proto == ETH_P_IPV6 && 967 dissector_uses_key(flow_dissector, 968 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) { 969 key_addrs = skb_flow_dissector_target(flow_dissector, 970 FLOW_DISSECTOR_KEY_IPV6_ADDRS, 971 target_container); 972 memcpy(&key_addrs->v6addrs.src, &flow_keys->ipv6_src, 973 sizeof(key_addrs->v6addrs.src)); 974 memcpy(&key_addrs->v6addrs.dst, &flow_keys->ipv6_dst, 975 sizeof(key_addrs->v6addrs.dst)); 976 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 977 } 978 979 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS)) { 980 key_ports = skb_flow_dissector_target(flow_dissector, 981 FLOW_DISSECTOR_KEY_PORTS, 982 target_container); 983 key_ports->src = flow_keys->sport; 984 key_ports->dst = flow_keys->dport; 985 } 986 if (dissector_uses_key(flow_dissector, 987 FLOW_DISSECTOR_KEY_PORTS_RANGE)) { 988 key_ports_range = skb_flow_dissector_target(flow_dissector, 989 FLOW_DISSECTOR_KEY_PORTS_RANGE, 990 target_container); 991 key_ports_range->tp.src = flow_keys->sport; 992 key_ports_range->tp.dst = flow_keys->dport; 993 } 994 995 if (dissector_uses_key(flow_dissector, 996 FLOW_DISSECTOR_KEY_FLOW_LABEL)) { 997 key_tags = skb_flow_dissector_target(flow_dissector, 998 FLOW_DISSECTOR_KEY_FLOW_LABEL, 999 target_container); 1000 key_tags->flow_label = ntohl(flow_keys->flow_label); 1001 } 1002 } 1003 1004 u32 bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx, 1005 __be16 proto, int nhoff, int hlen, unsigned int flags) 1006 { 1007 struct bpf_flow_keys *flow_keys = ctx->flow_keys; 1008 u32 result; 1009 1010 /* Pass parameters to the BPF program */ 1011 memset(flow_keys, 0, sizeof(*flow_keys)); 1012 flow_keys->n_proto = proto; 1013 flow_keys->nhoff = nhoff; 1014 flow_keys->thoff = flow_keys->nhoff; 1015 1016 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG != 1017 (int)FLOW_DISSECTOR_F_PARSE_1ST_FRAG); 1018 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL != 1019 (int)FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 1020 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP != 1021 (int)FLOW_DISSECTOR_F_STOP_AT_ENCAP); 1022 flow_keys->flags = flags; 1023 1024 result = bpf_prog_run_pin_on_cpu(prog, ctx); 1025 1026 flow_keys->nhoff = clamp_t(u16, flow_keys->nhoff, nhoff, hlen); 1027 flow_keys->thoff = clamp_t(u16, flow_keys->thoff, 1028 flow_keys->nhoff, hlen); 1029 1030 return result; 1031 } 1032 1033 static bool is_pppoe_ses_hdr_valid(const struct pppoe_hdr *hdr) 1034 { 1035 return hdr->ver == 1 && hdr->type == 1 && hdr->code == 0; 1036 } 1037 1038 /** 1039 * __skb_flow_dissect - extract the flow_keys struct and return it 1040 * @net: associated network namespace, derived from @skb if NULL 1041 * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified 1042 * @flow_dissector: list of keys to dissect 1043 * @target_container: target structure to put dissected values into 1044 * @data: raw buffer pointer to the packet, if NULL use skb->data 1045 * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol 1046 * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb) 1047 * @hlen: packet header length, if @data is NULL use skb_headlen(skb) 1048 * @flags: flags that control the dissection process, e.g. 1049 * FLOW_DISSECTOR_F_STOP_AT_ENCAP. 1050 * 1051 * The function will try to retrieve individual keys into target specified 1052 * by flow_dissector from either the skbuff or a raw buffer specified by the 1053 * rest parameters. 1054 * 1055 * Caller must take care of zeroing target container memory. 1056 */ 1057 bool __skb_flow_dissect(const struct net *net, 1058 const struct sk_buff *skb, 1059 struct flow_dissector *flow_dissector, 1060 void *target_container, const void *data, 1061 __be16 proto, int nhoff, int hlen, unsigned int flags) 1062 { 1063 struct flow_dissector_key_control *key_control; 1064 struct flow_dissector_key_basic *key_basic; 1065 struct flow_dissector_key_addrs *key_addrs; 1066 struct flow_dissector_key_tags *key_tags; 1067 struct flow_dissector_key_vlan *key_vlan; 1068 enum flow_dissect_ret fdret; 1069 enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX; 1070 bool mpls_el = false; 1071 int mpls_lse = 0; 1072 int num_hdrs = 0; 1073 u8 ip_proto = 0; 1074 bool ret; 1075 1076 if (!data) { 1077 data = skb->data; 1078 proto = skb_vlan_tag_present(skb) ? 1079 skb->vlan_proto : skb->protocol; 1080 nhoff = skb_network_offset(skb); 1081 hlen = skb_headlen(skb); 1082 #if IS_ENABLED(CONFIG_NET_DSA) 1083 if (unlikely(skb->dev && netdev_uses_dsa(skb->dev) && 1084 proto == htons(ETH_P_XDSA))) { 1085 struct metadata_dst *md_dst = skb_metadata_dst(skb); 1086 const struct dsa_device_ops *ops; 1087 int offset = 0; 1088 1089 ops = skb->dev->dsa_ptr->tag_ops; 1090 /* Only DSA header taggers break flow dissection */ 1091 if (ops->needed_headroom && 1092 (!md_dst || md_dst->type != METADATA_HW_PORT_MUX)) { 1093 if (ops->flow_dissect) 1094 ops->flow_dissect(skb, &proto, &offset); 1095 else 1096 dsa_tag_generic_flow_dissect(skb, 1097 &proto, 1098 &offset); 1099 hlen -= offset; 1100 nhoff += offset; 1101 } 1102 } 1103 #endif 1104 } 1105 1106 /* It is ensured by skb_flow_dissector_init() that control key will 1107 * be always present. 1108 */ 1109 key_control = skb_flow_dissector_target(flow_dissector, 1110 FLOW_DISSECTOR_KEY_CONTROL, 1111 target_container); 1112 1113 /* It is ensured by skb_flow_dissector_init() that basic key will 1114 * be always present. 1115 */ 1116 key_basic = skb_flow_dissector_target(flow_dissector, 1117 FLOW_DISSECTOR_KEY_BASIC, 1118 target_container); 1119 1120 rcu_read_lock(); 1121 1122 if (skb) { 1123 if (!net) { 1124 if (skb->dev) 1125 net = dev_net_rcu(skb->dev); 1126 else if (skb->sk) 1127 net = sock_net(skb->sk); 1128 } 1129 } 1130 1131 DEBUG_NET_WARN_ON_ONCE(!net); 1132 if (net) { 1133 enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR; 1134 struct bpf_prog_array *run_array; 1135 1136 run_array = rcu_dereference(init_net.bpf.run_array[type]); 1137 if (!run_array) 1138 run_array = rcu_dereference(net->bpf.run_array[type]); 1139 1140 if (run_array) { 1141 struct bpf_flow_keys flow_keys; 1142 struct bpf_flow_dissector ctx = { 1143 .flow_keys = &flow_keys, 1144 .data = data, 1145 .data_end = data + hlen, 1146 }; 1147 __be16 n_proto = proto; 1148 struct bpf_prog *prog; 1149 u32 result; 1150 1151 if (skb) { 1152 ctx.skb = skb; 1153 /* we can't use 'proto' in the skb case 1154 * because it might be set to skb->vlan_proto 1155 * which has been pulled from the data 1156 */ 1157 n_proto = skb->protocol; 1158 } 1159 1160 prog = READ_ONCE(run_array->items[0].prog); 1161 result = bpf_flow_dissect(prog, &ctx, n_proto, nhoff, 1162 hlen, flags); 1163 if (result != BPF_FLOW_DISSECTOR_CONTINUE) { 1164 __skb_flow_bpf_to_target(&flow_keys, flow_dissector, 1165 target_container); 1166 rcu_read_unlock(); 1167 return result == BPF_OK; 1168 } 1169 } 1170 } 1171 1172 rcu_read_unlock(); 1173 1174 if (dissector_uses_key(flow_dissector, 1175 FLOW_DISSECTOR_KEY_ETH_ADDRS)) { 1176 struct flow_dissector_key_eth_addrs *key_eth_addrs; 1177 1178 key_eth_addrs = skb_flow_dissector_target(flow_dissector, 1179 FLOW_DISSECTOR_KEY_ETH_ADDRS, 1180 target_container); 1181 /* TC filter blocks can be shared across devices with 1182 * different link types, so we cannot validate this 1183 * when the filter is installed -- check at dissect time. 1184 */ 1185 if (skb && skb->dev && 1186 skb->dev->type == ARPHRD_ETHER && 1187 skb_mac_header_was_set(skb)) 1188 memcpy(key_eth_addrs, eth_hdr(skb), sizeof(*key_eth_addrs)); 1189 else 1190 memset(key_eth_addrs, 0, sizeof(*key_eth_addrs)); 1191 } 1192 1193 if (dissector_uses_key(flow_dissector, 1194 FLOW_DISSECTOR_KEY_NUM_OF_VLANS)) { 1195 struct flow_dissector_key_num_of_vlans *key_num_of_vlans; 1196 1197 key_num_of_vlans = skb_flow_dissector_target(flow_dissector, 1198 FLOW_DISSECTOR_KEY_NUM_OF_VLANS, 1199 target_container); 1200 key_num_of_vlans->num_of_vlans = 0; 1201 } 1202 1203 proto_again: 1204 fdret = FLOW_DISSECT_RET_CONTINUE; 1205 1206 switch (proto) { 1207 case htons(ETH_P_IP): { 1208 const struct iphdr *iph; 1209 struct iphdr _iph; 1210 1211 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph); 1212 if (!iph || iph->ihl < 5) { 1213 fdret = FLOW_DISSECT_RET_OUT_BAD; 1214 break; 1215 } 1216 1217 nhoff += iph->ihl * 4; 1218 1219 ip_proto = iph->protocol; 1220 1221 if (dissector_uses_key(flow_dissector, 1222 FLOW_DISSECTOR_KEY_IPV4_ADDRS)) { 1223 key_addrs = skb_flow_dissector_target(flow_dissector, 1224 FLOW_DISSECTOR_KEY_IPV4_ADDRS, 1225 target_container); 1226 1227 memcpy(&key_addrs->v4addrs.src, &iph->saddr, 1228 sizeof(key_addrs->v4addrs.src)); 1229 memcpy(&key_addrs->v4addrs.dst, &iph->daddr, 1230 sizeof(key_addrs->v4addrs.dst)); 1231 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; 1232 } 1233 1234 __skb_flow_dissect_ipv4(skb, flow_dissector, 1235 target_container, data, iph); 1236 1237 if (ip_is_fragment(iph)) { 1238 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 1239 1240 if (iph->frag_off & htons(IP_OFFSET)) { 1241 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1242 break; 1243 } else { 1244 key_control->flags |= FLOW_DIS_FIRST_FRAG; 1245 if (!(flags & 1246 FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) { 1247 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1248 break; 1249 } 1250 } 1251 } 1252 1253 break; 1254 } 1255 case htons(ETH_P_IPV6): { 1256 const struct ipv6hdr *iph; 1257 struct ipv6hdr _iph; 1258 1259 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph); 1260 if (!iph) { 1261 fdret = FLOW_DISSECT_RET_OUT_BAD; 1262 break; 1263 } 1264 1265 ip_proto = iph->nexthdr; 1266 nhoff += sizeof(struct ipv6hdr); 1267 1268 if (dissector_uses_key(flow_dissector, 1269 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) { 1270 key_addrs = skb_flow_dissector_target(flow_dissector, 1271 FLOW_DISSECTOR_KEY_IPV6_ADDRS, 1272 target_container); 1273 1274 memcpy(&key_addrs->v6addrs.src, &iph->saddr, 1275 sizeof(key_addrs->v6addrs.src)); 1276 memcpy(&key_addrs->v6addrs.dst, &iph->daddr, 1277 sizeof(key_addrs->v6addrs.dst)); 1278 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 1279 } 1280 1281 if ((dissector_uses_key(flow_dissector, 1282 FLOW_DISSECTOR_KEY_FLOW_LABEL) || 1283 (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) && 1284 ip6_flowlabel(iph)) { 1285 __be32 flow_label = ip6_flowlabel(iph); 1286 1287 if (dissector_uses_key(flow_dissector, 1288 FLOW_DISSECTOR_KEY_FLOW_LABEL)) { 1289 key_tags = skb_flow_dissector_target(flow_dissector, 1290 FLOW_DISSECTOR_KEY_FLOW_LABEL, 1291 target_container); 1292 key_tags->flow_label = ntohl(flow_label); 1293 } 1294 if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) { 1295 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1296 break; 1297 } 1298 } 1299 1300 __skb_flow_dissect_ipv6(skb, flow_dissector, 1301 target_container, data, iph); 1302 1303 break; 1304 } 1305 case htons(ETH_P_8021AD): 1306 case htons(ETH_P_8021Q): { 1307 const struct vlan_hdr *vlan = NULL; 1308 struct vlan_hdr _vlan; 1309 __be16 saved_vlan_tpid = proto; 1310 1311 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX && 1312 skb && skb_vlan_tag_present(skb)) { 1313 proto = skb->protocol; 1314 } else { 1315 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan), 1316 data, hlen, &_vlan); 1317 if (!vlan) { 1318 fdret = FLOW_DISSECT_RET_OUT_BAD; 1319 break; 1320 } 1321 1322 proto = vlan->h_vlan_encapsulated_proto; 1323 nhoff += sizeof(*vlan); 1324 } 1325 1326 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_NUM_OF_VLANS) && 1327 !(key_control->flags & FLOW_DIS_ENCAPSULATION)) { 1328 struct flow_dissector_key_num_of_vlans *key_nvs; 1329 1330 key_nvs = skb_flow_dissector_target(flow_dissector, 1331 FLOW_DISSECTOR_KEY_NUM_OF_VLANS, 1332 target_container); 1333 key_nvs->num_of_vlans++; 1334 } 1335 1336 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) { 1337 dissector_vlan = FLOW_DISSECTOR_KEY_VLAN; 1338 } else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) { 1339 dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN; 1340 } else { 1341 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1342 break; 1343 } 1344 1345 if (dissector_uses_key(flow_dissector, dissector_vlan)) { 1346 key_vlan = skb_flow_dissector_target(flow_dissector, 1347 dissector_vlan, 1348 target_container); 1349 1350 if (!vlan) { 1351 key_vlan->vlan_id = skb_vlan_tag_get_id(skb); 1352 key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb); 1353 } else { 1354 key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) & 1355 VLAN_VID_MASK; 1356 key_vlan->vlan_priority = 1357 (ntohs(vlan->h_vlan_TCI) & 1358 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT; 1359 } 1360 key_vlan->vlan_tpid = saved_vlan_tpid; 1361 key_vlan->vlan_eth_type = proto; 1362 } 1363 1364 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1365 break; 1366 } 1367 case htons(ETH_P_PPP_SES): { 1368 struct { 1369 struct pppoe_hdr hdr; 1370 __be16 proto; 1371 } *hdr, _hdr; 1372 u16 ppp_proto; 1373 1374 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr); 1375 if (!hdr) { 1376 fdret = FLOW_DISSECT_RET_OUT_BAD; 1377 break; 1378 } 1379 1380 if (!is_pppoe_ses_hdr_valid(&hdr->hdr)) { 1381 fdret = FLOW_DISSECT_RET_OUT_BAD; 1382 break; 1383 } 1384 1385 /* PFC (compressed 1-byte protocol) frames are not processed. 1386 * A compressed protocol field has the least significant bit of 1387 * the most significant octet set, which will fail the following 1388 * ppp_proto_is_valid(), returning FLOW_DISSECT_RET_OUT_BAD. 1389 */ 1390 ppp_proto = ntohs(hdr->proto); 1391 nhoff += PPPOE_SES_HLEN; 1392 1393 if (ppp_proto == PPP_IP) { 1394 proto = htons(ETH_P_IP); 1395 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1396 } else if (ppp_proto == PPP_IPV6) { 1397 proto = htons(ETH_P_IPV6); 1398 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1399 } else if (ppp_proto == PPP_MPLS_UC) { 1400 proto = htons(ETH_P_MPLS_UC); 1401 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1402 } else if (ppp_proto == PPP_MPLS_MC) { 1403 proto = htons(ETH_P_MPLS_MC); 1404 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1405 } else if (ppp_proto_is_valid(ppp_proto)) { 1406 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1407 } else { 1408 fdret = FLOW_DISSECT_RET_OUT_BAD; 1409 break; 1410 } 1411 1412 if (dissector_uses_key(flow_dissector, 1413 FLOW_DISSECTOR_KEY_PPPOE)) { 1414 struct flow_dissector_key_pppoe *key_pppoe; 1415 1416 key_pppoe = skb_flow_dissector_target(flow_dissector, 1417 FLOW_DISSECTOR_KEY_PPPOE, 1418 target_container); 1419 key_pppoe->session_id = hdr->hdr.sid; 1420 key_pppoe->ppp_proto = htons(ppp_proto); 1421 key_pppoe->type = htons(ETH_P_PPP_SES); 1422 } 1423 break; 1424 } 1425 case htons(ETH_P_TIPC): { 1426 struct tipc_basic_hdr *hdr, _hdr; 1427 1428 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), 1429 data, hlen, &_hdr); 1430 if (!hdr) { 1431 fdret = FLOW_DISSECT_RET_OUT_BAD; 1432 break; 1433 } 1434 1435 if (dissector_uses_key(flow_dissector, 1436 FLOW_DISSECTOR_KEY_TIPC)) { 1437 key_addrs = skb_flow_dissector_target(flow_dissector, 1438 FLOW_DISSECTOR_KEY_TIPC, 1439 target_container); 1440 key_addrs->tipckey.key = tipc_hdr_rps_key(hdr); 1441 key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC; 1442 } 1443 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1444 break; 1445 } 1446 1447 case htons(ETH_P_MPLS_UC): 1448 case htons(ETH_P_MPLS_MC): 1449 fdret = __skb_flow_dissect_mpls(skb, flow_dissector, 1450 target_container, data, 1451 nhoff, hlen, mpls_lse, 1452 &mpls_el); 1453 nhoff += sizeof(struct mpls_label); 1454 mpls_lse++; 1455 break; 1456 case htons(ETH_P_FCOE): 1457 if ((hlen - nhoff) < FCOE_HEADER_LEN) { 1458 fdret = FLOW_DISSECT_RET_OUT_BAD; 1459 break; 1460 } 1461 1462 nhoff += FCOE_HEADER_LEN; 1463 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1464 break; 1465 1466 case htons(ETH_P_ARP): 1467 case htons(ETH_P_RARP): 1468 fdret = __skb_flow_dissect_arp(skb, flow_dissector, 1469 target_container, data, 1470 nhoff, hlen); 1471 break; 1472 1473 case htons(ETH_P_BATMAN): 1474 fdret = __skb_flow_dissect_batadv(skb, key_control, data, 1475 &proto, &nhoff, hlen, flags); 1476 break; 1477 1478 case htons(ETH_P_1588): { 1479 struct ptp_header *hdr, _hdr; 1480 1481 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, 1482 hlen, &_hdr); 1483 if (!hdr) { 1484 fdret = FLOW_DISSECT_RET_OUT_BAD; 1485 break; 1486 } 1487 1488 nhoff += sizeof(struct ptp_header); 1489 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1490 break; 1491 } 1492 1493 case htons(ETH_P_PRP): 1494 case htons(ETH_P_HSR): { 1495 struct hsr_tag *hdr, _hdr; 1496 1497 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, 1498 &_hdr); 1499 if (!hdr) { 1500 fdret = FLOW_DISSECT_RET_OUT_BAD; 1501 break; 1502 } 1503 1504 proto = hdr->encap_proto; 1505 nhoff += HSR_HLEN; 1506 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1507 break; 1508 } 1509 1510 case htons(ETH_P_CFM): 1511 fdret = __skb_flow_dissect_cfm(skb, flow_dissector, 1512 target_container, data, 1513 nhoff, hlen); 1514 break; 1515 1516 default: 1517 fdret = FLOW_DISSECT_RET_OUT_BAD; 1518 break; 1519 } 1520 1521 /* Process result of proto processing */ 1522 switch (fdret) { 1523 case FLOW_DISSECT_RET_OUT_GOOD: 1524 goto out_good; 1525 case FLOW_DISSECT_RET_PROTO_AGAIN: 1526 if (skb_flow_dissect_allowed(&num_hdrs)) 1527 goto proto_again; 1528 goto out_good; 1529 case FLOW_DISSECT_RET_CONTINUE: 1530 case FLOW_DISSECT_RET_IPPROTO_AGAIN: 1531 break; 1532 case FLOW_DISSECT_RET_OUT_BAD: 1533 default: 1534 goto out_bad; 1535 } 1536 1537 ip_proto_again: 1538 fdret = FLOW_DISSECT_RET_CONTINUE; 1539 1540 switch (ip_proto) { 1541 case IPPROTO_GRE: 1542 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) { 1543 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1544 break; 1545 } 1546 1547 fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector, 1548 target_container, data, 1549 &proto, &nhoff, &hlen, flags); 1550 break; 1551 1552 case NEXTHDR_HOP: 1553 case NEXTHDR_ROUTING: 1554 case NEXTHDR_DEST: { 1555 u8 _opthdr[2], *opthdr; 1556 1557 if (proto != htons(ETH_P_IPV6)) 1558 break; 1559 1560 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr), 1561 data, hlen, &_opthdr); 1562 if (!opthdr) { 1563 fdret = FLOW_DISSECT_RET_OUT_BAD; 1564 break; 1565 } 1566 1567 ip_proto = opthdr[0]; 1568 nhoff += (opthdr[1] + 1) << 3; 1569 1570 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN; 1571 break; 1572 } 1573 case NEXTHDR_FRAGMENT: { 1574 struct frag_hdr _fh, *fh; 1575 1576 if (proto != htons(ETH_P_IPV6)) 1577 break; 1578 1579 fh = __skb_header_pointer(skb, nhoff, sizeof(_fh), 1580 data, hlen, &_fh); 1581 1582 if (!fh) { 1583 fdret = FLOW_DISSECT_RET_OUT_BAD; 1584 break; 1585 } 1586 1587 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 1588 1589 nhoff += sizeof(_fh); 1590 ip_proto = fh->nexthdr; 1591 1592 if (!(fh->frag_off & htons(IP6_OFFSET))) { 1593 key_control->flags |= FLOW_DIS_FIRST_FRAG; 1594 if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) { 1595 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN; 1596 break; 1597 } 1598 } 1599 1600 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1601 break; 1602 } 1603 case IPPROTO_IPIP: 1604 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) { 1605 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1606 break; 1607 } 1608 1609 proto = htons(ETH_P_IP); 1610 1611 key_control->flags |= FLOW_DIS_ENCAPSULATION; 1612 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) { 1613 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1614 break; 1615 } 1616 1617 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1618 break; 1619 1620 case IPPROTO_IPV6: 1621 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) { 1622 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1623 break; 1624 } 1625 1626 proto = htons(ETH_P_IPV6); 1627 1628 key_control->flags |= FLOW_DIS_ENCAPSULATION; 1629 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) { 1630 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1631 break; 1632 } 1633 1634 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1635 break; 1636 1637 1638 case IPPROTO_MPLS: 1639 proto = htons(ETH_P_MPLS_UC); 1640 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1641 break; 1642 1643 case IPPROTO_TCP: 1644 __skb_flow_dissect_tcp(skb, flow_dissector, target_container, 1645 data, nhoff, hlen); 1646 break; 1647 1648 case IPPROTO_ICMP: 1649 case IPPROTO_ICMPV6: 1650 __skb_flow_dissect_icmp(skb, flow_dissector, target_container, 1651 data, nhoff, hlen); 1652 break; 1653 case IPPROTO_L2TP: 1654 __skb_flow_dissect_l2tpv3(skb, flow_dissector, target_container, 1655 data, nhoff, hlen); 1656 break; 1657 case IPPROTO_ESP: 1658 __skb_flow_dissect_esp(skb, flow_dissector, target_container, 1659 data, nhoff, hlen); 1660 break; 1661 case IPPROTO_AH: 1662 __skb_flow_dissect_ah(skb, flow_dissector, target_container, 1663 data, nhoff, hlen); 1664 break; 1665 default: 1666 break; 1667 } 1668 1669 if (!(key_control->flags & FLOW_DIS_IS_FRAGMENT)) 1670 __skb_flow_dissect_ports(skb, flow_dissector, target_container, 1671 data, nhoff, ip_proto, hlen); 1672 1673 /* Process result of IP proto processing */ 1674 switch (fdret) { 1675 case FLOW_DISSECT_RET_PROTO_AGAIN: 1676 if (skb_flow_dissect_allowed(&num_hdrs)) 1677 goto proto_again; 1678 break; 1679 case FLOW_DISSECT_RET_IPPROTO_AGAIN: 1680 if (skb_flow_dissect_allowed(&num_hdrs)) 1681 goto ip_proto_again; 1682 break; 1683 case FLOW_DISSECT_RET_OUT_GOOD: 1684 case FLOW_DISSECT_RET_CONTINUE: 1685 break; 1686 case FLOW_DISSECT_RET_OUT_BAD: 1687 default: 1688 goto out_bad; 1689 } 1690 1691 out_good: 1692 ret = true; 1693 1694 out: 1695 key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen); 1696 key_basic->n_proto = proto; 1697 key_basic->ip_proto = ip_proto; 1698 1699 return ret; 1700 1701 out_bad: 1702 ret = false; 1703 goto out; 1704 } 1705 EXPORT_SYMBOL(__skb_flow_dissect); 1706 1707 static siphash_aligned_key_t hashrnd; 1708 static __always_inline void __flow_hash_secret_init(void) 1709 { 1710 net_get_random_once(&hashrnd, sizeof(hashrnd)); 1711 } 1712 1713 static const void *flow_keys_hash_start(const struct flow_keys *flow) 1714 { 1715 BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % SIPHASH_ALIGNMENT); 1716 return &flow->FLOW_KEYS_HASH_START_FIELD; 1717 } 1718 1719 static inline size_t flow_keys_hash_length(const struct flow_keys *flow) 1720 { 1721 size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs); 1722 1723 BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32)); 1724 1725 switch (flow->control.addr_type) { 1726 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1727 diff -= sizeof(flow->addrs.v4addrs); 1728 break; 1729 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1730 diff -= sizeof(flow->addrs.v6addrs); 1731 break; 1732 case FLOW_DISSECTOR_KEY_TIPC: 1733 diff -= sizeof(flow->addrs.tipckey); 1734 break; 1735 } 1736 return sizeof(*flow) - diff; 1737 } 1738 1739 __be32 flow_get_u32_src(const struct flow_keys *flow) 1740 { 1741 switch (flow->control.addr_type) { 1742 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1743 return flow->addrs.v4addrs.src; 1744 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1745 return (__force __be32)ipv6_addr_hash( 1746 &flow->addrs.v6addrs.src); 1747 case FLOW_DISSECTOR_KEY_TIPC: 1748 return flow->addrs.tipckey.key; 1749 default: 1750 return 0; 1751 } 1752 } 1753 EXPORT_SYMBOL(flow_get_u32_src); 1754 1755 __be32 flow_get_u32_dst(const struct flow_keys *flow) 1756 { 1757 switch (flow->control.addr_type) { 1758 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1759 return flow->addrs.v4addrs.dst; 1760 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1761 return (__force __be32)ipv6_addr_hash( 1762 &flow->addrs.v6addrs.dst); 1763 default: 1764 return 0; 1765 } 1766 } 1767 EXPORT_SYMBOL(flow_get_u32_dst); 1768 1769 /* Sort the source and destination IP and the ports, 1770 * to have consistent hash within the two directions 1771 */ 1772 static inline void __flow_hash_consistentify(struct flow_keys *keys) 1773 { 1774 int addr_diff, i; 1775 1776 switch (keys->control.addr_type) { 1777 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1778 if ((__force u32)keys->addrs.v4addrs.dst < 1779 (__force u32)keys->addrs.v4addrs.src) 1780 swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst); 1781 1782 if ((__force u16)keys->ports.dst < 1783 (__force u16)keys->ports.src) { 1784 swap(keys->ports.src, keys->ports.dst); 1785 } 1786 break; 1787 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1788 addr_diff = memcmp(&keys->addrs.v6addrs.dst, 1789 &keys->addrs.v6addrs.src, 1790 sizeof(keys->addrs.v6addrs.dst)); 1791 if (addr_diff < 0) { 1792 for (i = 0; i < 4; i++) 1793 swap(keys->addrs.v6addrs.src.s6_addr32[i], 1794 keys->addrs.v6addrs.dst.s6_addr32[i]); 1795 } 1796 if ((__force u16)keys->ports.dst < 1797 (__force u16)keys->ports.src) { 1798 swap(keys->ports.src, keys->ports.dst); 1799 } 1800 break; 1801 } 1802 } 1803 1804 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, 1805 const siphash_key_t *keyval) 1806 { 1807 u32 hash; 1808 1809 __flow_hash_consistentify(keys); 1810 1811 hash = siphash(flow_keys_hash_start(keys), 1812 flow_keys_hash_length(keys), keyval); 1813 if (!hash) 1814 hash = 1; 1815 1816 return hash; 1817 } 1818 1819 u32 flow_hash_from_keys(struct flow_keys *keys) 1820 { 1821 __flow_hash_secret_init(); 1822 return __flow_hash_from_keys(keys, &hashrnd); 1823 } 1824 EXPORT_SYMBOL(flow_hash_from_keys); 1825 1826 u32 flow_hash_from_keys_seed(struct flow_keys *keys, 1827 const siphash_key_t *keyval) 1828 { 1829 return __flow_hash_from_keys(keys, keyval); 1830 } 1831 EXPORT_SYMBOL(flow_hash_from_keys_seed); 1832 1833 static inline u32 ___skb_get_hash(const struct sk_buff *skb, 1834 struct flow_keys *keys, 1835 const siphash_key_t *keyval) 1836 { 1837 skb_flow_dissect_flow_keys(skb, keys, 1838 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 1839 1840 return __flow_hash_from_keys(keys, keyval); 1841 } 1842 1843 struct _flow_keys_digest_data { 1844 __be16 n_proto; 1845 u8 ip_proto; 1846 u8 padding; 1847 __be32 ports; 1848 __be32 src; 1849 __be32 dst; 1850 }; 1851 1852 void make_flow_keys_digest(struct flow_keys_digest *digest, 1853 const struct flow_keys *flow) 1854 { 1855 struct _flow_keys_digest_data *data = 1856 (struct _flow_keys_digest_data *)digest; 1857 1858 BUILD_BUG_ON(sizeof(*data) > sizeof(*digest)); 1859 1860 memset(digest, 0, sizeof(*digest)); 1861 1862 data->n_proto = flow->basic.n_proto; 1863 data->ip_proto = flow->basic.ip_proto; 1864 data->ports = flow->ports.ports; 1865 data->src = flow->addrs.v4addrs.src; 1866 data->dst = flow->addrs.v4addrs.dst; 1867 } 1868 EXPORT_SYMBOL(make_flow_keys_digest); 1869 1870 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly; 1871 1872 u32 __skb_get_hash_symmetric_net(const struct net *net, const struct sk_buff *skb) 1873 { 1874 struct flow_keys keys; 1875 1876 __flow_hash_secret_init(); 1877 1878 memset(&keys, 0, sizeof(keys)); 1879 __skb_flow_dissect(net, skb, &flow_keys_dissector_symmetric, 1880 &keys, NULL, 0, 0, 0, 0); 1881 1882 return __flow_hash_from_keys(&keys, &hashrnd); 1883 } 1884 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric_net); 1885 1886 /** 1887 * __skb_get_hash_net: calculate a flow hash 1888 * @net: associated network namespace, derived from @skb if NULL 1889 * @skb: sk_buff to calculate flow hash from 1890 * 1891 * This function calculates a flow hash based on src/dst addresses 1892 * and src/dst port numbers. Sets hash in skb to non-zero hash value 1893 * on success, zero indicates no valid hash. Also, sets l4_hash in skb 1894 * if hash is a canonical 4-tuple hash over transport ports. 1895 */ 1896 void __skb_get_hash_net(const struct net *net, struct sk_buff *skb) 1897 { 1898 struct flow_keys keys; 1899 u32 hash; 1900 1901 memset(&keys, 0, sizeof(keys)); 1902 1903 __skb_flow_dissect(net, skb, &flow_keys_dissector, 1904 &keys, NULL, 0, 0, 0, 1905 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 1906 1907 __flow_hash_secret_init(); 1908 1909 hash = __flow_hash_from_keys(&keys, &hashrnd); 1910 1911 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); 1912 } 1913 EXPORT_SYMBOL(__skb_get_hash_net); 1914 1915 __u32 skb_get_hash_perturb(const struct sk_buff *skb, 1916 const siphash_key_t *perturb) 1917 { 1918 struct flow_keys keys; 1919 1920 return ___skb_get_hash(skb, &keys, perturb); 1921 } 1922 EXPORT_SYMBOL(skb_get_hash_perturb); 1923 1924 u32 __skb_get_poff(const struct sk_buff *skb, const void *data, 1925 const struct flow_keys_basic *keys, int hlen) 1926 { 1927 u32 poff = keys->control.thoff; 1928 1929 /* skip L4 headers for fragments after the first */ 1930 if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) && 1931 !(keys->control.flags & FLOW_DIS_FIRST_FRAG)) 1932 return poff; 1933 1934 switch (keys->basic.ip_proto) { 1935 case IPPROTO_TCP: { 1936 /* access doff as u8 to avoid unaligned access */ 1937 const u8 *doff; 1938 u8 _doff; 1939 1940 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff), 1941 data, hlen, &_doff); 1942 if (!doff) 1943 return poff; 1944 1945 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2); 1946 break; 1947 } 1948 case IPPROTO_UDP: 1949 case IPPROTO_UDPLITE: 1950 poff += sizeof(struct udphdr); 1951 break; 1952 /* For the rest, we do not really care about header 1953 * extensions at this point for now. 1954 */ 1955 case IPPROTO_ICMP: 1956 poff += sizeof(struct icmphdr); 1957 break; 1958 case IPPROTO_ICMPV6: 1959 poff += sizeof(struct icmp6hdr); 1960 break; 1961 case IPPROTO_IGMP: 1962 poff += sizeof(struct igmphdr); 1963 break; 1964 case IPPROTO_DCCP: 1965 poff += sizeof(struct dccp_hdr); 1966 break; 1967 case IPPROTO_SCTP: 1968 poff += sizeof(struct sctphdr); 1969 break; 1970 } 1971 1972 return poff; 1973 } 1974 1975 /** 1976 * skb_get_poff - get the offset to the payload 1977 * @skb: sk_buff to get the payload offset from 1978 * 1979 * The function will get the offset to the payload as far as it could 1980 * be dissected. The main user is currently BPF, so that we can dynamically 1981 * truncate packets without needing to push actual payload to the user 1982 * space and can analyze headers only, instead. 1983 */ 1984 u32 skb_get_poff(const struct sk_buff *skb) 1985 { 1986 struct flow_keys_basic keys; 1987 1988 if (!skb_flow_dissect_flow_keys_basic(NULL, skb, &keys, 1989 NULL, 0, 0, 0, 0)) 1990 return 0; 1991 1992 return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb)); 1993 } 1994 1995 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys) 1996 { 1997 memset(keys, 0, sizeof(*keys)); 1998 1999 memcpy(&keys->addrs.v6addrs.src, &fl6->saddr, 2000 sizeof(keys->addrs.v6addrs.src)); 2001 memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr, 2002 sizeof(keys->addrs.v6addrs.dst)); 2003 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 2004 keys->ports.src = fl6->fl6_sport; 2005 keys->ports.dst = fl6->fl6_dport; 2006 keys->keyid.keyid = fl6->fl6_gre_key; 2007 keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6); 2008 keys->basic.ip_proto = fl6->flowi6_proto; 2009 2010 return flow_hash_from_keys(keys); 2011 } 2012 EXPORT_SYMBOL(__get_hash_from_flowi6); 2013 2014 static const struct flow_dissector_key flow_keys_dissector_keys[] = { 2015 { 2016 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 2017 .offset = offsetof(struct flow_keys, control), 2018 }, 2019 { 2020 .key_id = FLOW_DISSECTOR_KEY_BASIC, 2021 .offset = offsetof(struct flow_keys, basic), 2022 }, 2023 { 2024 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS, 2025 .offset = offsetof(struct flow_keys, addrs.v4addrs), 2026 }, 2027 { 2028 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS, 2029 .offset = offsetof(struct flow_keys, addrs.v6addrs), 2030 }, 2031 { 2032 .key_id = FLOW_DISSECTOR_KEY_TIPC, 2033 .offset = offsetof(struct flow_keys, addrs.tipckey), 2034 }, 2035 { 2036 .key_id = FLOW_DISSECTOR_KEY_PORTS, 2037 .offset = offsetof(struct flow_keys, ports), 2038 }, 2039 { 2040 .key_id = FLOW_DISSECTOR_KEY_VLAN, 2041 .offset = offsetof(struct flow_keys, vlan), 2042 }, 2043 { 2044 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL, 2045 .offset = offsetof(struct flow_keys, tags), 2046 }, 2047 { 2048 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID, 2049 .offset = offsetof(struct flow_keys, keyid), 2050 }, 2051 }; 2052 2053 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = { 2054 { 2055 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 2056 .offset = offsetof(struct flow_keys, control), 2057 }, 2058 { 2059 .key_id = FLOW_DISSECTOR_KEY_BASIC, 2060 .offset = offsetof(struct flow_keys, basic), 2061 }, 2062 { 2063 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS, 2064 .offset = offsetof(struct flow_keys, addrs.v4addrs), 2065 }, 2066 { 2067 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS, 2068 .offset = offsetof(struct flow_keys, addrs.v6addrs), 2069 }, 2070 { 2071 .key_id = FLOW_DISSECTOR_KEY_PORTS, 2072 .offset = offsetof(struct flow_keys, ports), 2073 }, 2074 }; 2075 2076 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = { 2077 { 2078 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 2079 .offset = offsetof(struct flow_keys, control), 2080 }, 2081 { 2082 .key_id = FLOW_DISSECTOR_KEY_BASIC, 2083 .offset = offsetof(struct flow_keys, basic), 2084 }, 2085 }; 2086 2087 struct flow_dissector flow_keys_dissector __read_mostly; 2088 EXPORT_SYMBOL(flow_keys_dissector); 2089 2090 struct flow_dissector flow_keys_basic_dissector __read_mostly; 2091 EXPORT_SYMBOL(flow_keys_basic_dissector); 2092 2093 static int __init init_default_flow_dissectors(void) 2094 { 2095 skb_flow_dissector_init(&flow_keys_dissector, 2096 flow_keys_dissector_keys, 2097 ARRAY_SIZE(flow_keys_dissector_keys)); 2098 skb_flow_dissector_init(&flow_keys_dissector_symmetric, 2099 flow_keys_dissector_symmetric_keys, 2100 ARRAY_SIZE(flow_keys_dissector_symmetric_keys)); 2101 skb_flow_dissector_init(&flow_keys_basic_dissector, 2102 flow_keys_basic_dissector_keys, 2103 ARRAY_SIZE(flow_keys_basic_dissector_keys)); 2104 return 0; 2105 } 2106 core_initcall(init_default_flow_dissectors); 2107