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 enum flow_dissector_key_id dissector_ports = FLOW_DISSECTOR_KEY_MAX; 857 struct flow_dissector_key_ports *key_ports; 858 859 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS)) 860 dissector_ports = FLOW_DISSECTOR_KEY_PORTS; 861 else if (dissector_uses_key(flow_dissector, 862 FLOW_DISSECTOR_KEY_PORTS_RANGE)) 863 dissector_ports = FLOW_DISSECTOR_KEY_PORTS_RANGE; 864 865 if (dissector_ports == FLOW_DISSECTOR_KEY_MAX) 866 return; 867 868 key_ports = skb_flow_dissector_target(flow_dissector, 869 dissector_ports, 870 target_container); 871 key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto, 872 data, hlen); 873 } 874 875 static void 876 __skb_flow_dissect_ipv4(const struct sk_buff *skb, 877 struct flow_dissector *flow_dissector, 878 void *target_container, const void *data, 879 const struct iphdr *iph) 880 { 881 struct flow_dissector_key_ip *key_ip; 882 883 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP)) 884 return; 885 886 key_ip = skb_flow_dissector_target(flow_dissector, 887 FLOW_DISSECTOR_KEY_IP, 888 target_container); 889 key_ip->tos = iph->tos; 890 key_ip->ttl = iph->ttl; 891 } 892 893 static void 894 __skb_flow_dissect_ipv6(const struct sk_buff *skb, 895 struct flow_dissector *flow_dissector, 896 void *target_container, const void *data, 897 const struct ipv6hdr *iph) 898 { 899 struct flow_dissector_key_ip *key_ip; 900 901 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP)) 902 return; 903 904 key_ip = skb_flow_dissector_target(flow_dissector, 905 FLOW_DISSECTOR_KEY_IP, 906 target_container); 907 key_ip->tos = ipv6_get_dsfield(iph); 908 key_ip->ttl = iph->hop_limit; 909 } 910 911 /* Maximum number of protocol headers that can be parsed in 912 * __skb_flow_dissect 913 */ 914 #define MAX_FLOW_DISSECT_HDRS 15 915 916 static bool skb_flow_dissect_allowed(int *num_hdrs) 917 { 918 ++*num_hdrs; 919 920 return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS); 921 } 922 923 static void __skb_flow_bpf_to_target(const struct bpf_flow_keys *flow_keys, 924 struct flow_dissector *flow_dissector, 925 void *target_container) 926 { 927 struct flow_dissector_key_ports *key_ports = NULL; 928 struct flow_dissector_key_control *key_control; 929 struct flow_dissector_key_basic *key_basic; 930 struct flow_dissector_key_addrs *key_addrs; 931 struct flow_dissector_key_tags *key_tags; 932 933 key_control = skb_flow_dissector_target(flow_dissector, 934 FLOW_DISSECTOR_KEY_CONTROL, 935 target_container); 936 key_control->thoff = flow_keys->thoff; 937 if (flow_keys->is_frag) 938 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 939 if (flow_keys->is_first_frag) 940 key_control->flags |= FLOW_DIS_FIRST_FRAG; 941 if (flow_keys->is_encap) 942 key_control->flags |= FLOW_DIS_ENCAPSULATION; 943 944 key_basic = skb_flow_dissector_target(flow_dissector, 945 FLOW_DISSECTOR_KEY_BASIC, 946 target_container); 947 key_basic->n_proto = flow_keys->n_proto; 948 key_basic->ip_proto = flow_keys->ip_proto; 949 950 if (flow_keys->addr_proto == ETH_P_IP && 951 dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) { 952 key_addrs = skb_flow_dissector_target(flow_dissector, 953 FLOW_DISSECTOR_KEY_IPV4_ADDRS, 954 target_container); 955 key_addrs->v4addrs.src = flow_keys->ipv4_src; 956 key_addrs->v4addrs.dst = flow_keys->ipv4_dst; 957 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; 958 } else if (flow_keys->addr_proto == ETH_P_IPV6 && 959 dissector_uses_key(flow_dissector, 960 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) { 961 key_addrs = skb_flow_dissector_target(flow_dissector, 962 FLOW_DISSECTOR_KEY_IPV6_ADDRS, 963 target_container); 964 memcpy(&key_addrs->v6addrs.src, &flow_keys->ipv6_src, 965 sizeof(key_addrs->v6addrs.src)); 966 memcpy(&key_addrs->v6addrs.dst, &flow_keys->ipv6_dst, 967 sizeof(key_addrs->v6addrs.dst)); 968 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 969 } 970 971 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS)) 972 key_ports = skb_flow_dissector_target(flow_dissector, 973 FLOW_DISSECTOR_KEY_PORTS, 974 target_container); 975 else if (dissector_uses_key(flow_dissector, 976 FLOW_DISSECTOR_KEY_PORTS_RANGE)) 977 key_ports = skb_flow_dissector_target(flow_dissector, 978 FLOW_DISSECTOR_KEY_PORTS_RANGE, 979 target_container); 980 981 if (key_ports) { 982 key_ports->src = flow_keys->sport; 983 key_ports->dst = flow_keys->dport; 984 } 985 986 if (dissector_uses_key(flow_dissector, 987 FLOW_DISSECTOR_KEY_FLOW_LABEL)) { 988 key_tags = skb_flow_dissector_target(flow_dissector, 989 FLOW_DISSECTOR_KEY_FLOW_LABEL, 990 target_container); 991 key_tags->flow_label = ntohl(flow_keys->flow_label); 992 } 993 } 994 995 u32 bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx, 996 __be16 proto, int nhoff, int hlen, unsigned int flags) 997 { 998 struct bpf_flow_keys *flow_keys = ctx->flow_keys; 999 u32 result; 1000 1001 /* Pass parameters to the BPF program */ 1002 memset(flow_keys, 0, sizeof(*flow_keys)); 1003 flow_keys->n_proto = proto; 1004 flow_keys->nhoff = nhoff; 1005 flow_keys->thoff = flow_keys->nhoff; 1006 1007 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG != 1008 (int)FLOW_DISSECTOR_F_PARSE_1ST_FRAG); 1009 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL != 1010 (int)FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 1011 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP != 1012 (int)FLOW_DISSECTOR_F_STOP_AT_ENCAP); 1013 flow_keys->flags = flags; 1014 1015 result = bpf_prog_run_pin_on_cpu(prog, ctx); 1016 1017 flow_keys->nhoff = clamp_t(u16, flow_keys->nhoff, nhoff, hlen); 1018 flow_keys->thoff = clamp_t(u16, flow_keys->thoff, 1019 flow_keys->nhoff, hlen); 1020 1021 return result; 1022 } 1023 1024 static bool is_pppoe_ses_hdr_valid(const struct pppoe_hdr *hdr) 1025 { 1026 return hdr->ver == 1 && hdr->type == 1 && hdr->code == 0; 1027 } 1028 1029 /** 1030 * __skb_flow_dissect - extract the flow_keys struct and return it 1031 * @net: associated network namespace, derived from @skb if NULL 1032 * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified 1033 * @flow_dissector: list of keys to dissect 1034 * @target_container: target structure to put dissected values into 1035 * @data: raw buffer pointer to the packet, if NULL use skb->data 1036 * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol 1037 * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb) 1038 * @hlen: packet header length, if @data is NULL use skb_headlen(skb) 1039 * @flags: flags that control the dissection process, e.g. 1040 * FLOW_DISSECTOR_F_STOP_AT_ENCAP. 1041 * 1042 * The function will try to retrieve individual keys into target specified 1043 * by flow_dissector from either the skbuff or a raw buffer specified by the 1044 * rest parameters. 1045 * 1046 * Caller must take care of zeroing target container memory. 1047 */ 1048 bool __skb_flow_dissect(const struct net *net, 1049 const struct sk_buff *skb, 1050 struct flow_dissector *flow_dissector, 1051 void *target_container, const void *data, 1052 __be16 proto, int nhoff, int hlen, unsigned int flags) 1053 { 1054 struct flow_dissector_key_control *key_control; 1055 struct flow_dissector_key_basic *key_basic; 1056 struct flow_dissector_key_addrs *key_addrs; 1057 struct flow_dissector_key_tags *key_tags; 1058 struct flow_dissector_key_vlan *key_vlan; 1059 enum flow_dissect_ret fdret; 1060 enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX; 1061 bool mpls_el = false; 1062 int mpls_lse = 0; 1063 int num_hdrs = 0; 1064 u8 ip_proto = 0; 1065 bool ret; 1066 1067 if (!data) { 1068 data = skb->data; 1069 proto = skb_vlan_tag_present(skb) ? 1070 skb->vlan_proto : skb->protocol; 1071 nhoff = skb_network_offset(skb); 1072 hlen = skb_headlen(skb); 1073 #if IS_ENABLED(CONFIG_NET_DSA) 1074 if (unlikely(skb->dev && netdev_uses_dsa(skb->dev) && 1075 proto == htons(ETH_P_XDSA))) { 1076 struct metadata_dst *md_dst = skb_metadata_dst(skb); 1077 const struct dsa_device_ops *ops; 1078 int offset = 0; 1079 1080 ops = skb->dev->dsa_ptr->tag_ops; 1081 /* Only DSA header taggers break flow dissection */ 1082 if (ops->needed_headroom && 1083 (!md_dst || md_dst->type != METADATA_HW_PORT_MUX)) { 1084 if (ops->flow_dissect) 1085 ops->flow_dissect(skb, &proto, &offset); 1086 else 1087 dsa_tag_generic_flow_dissect(skb, 1088 &proto, 1089 &offset); 1090 hlen -= offset; 1091 nhoff += offset; 1092 } 1093 } 1094 #endif 1095 } 1096 1097 /* It is ensured by skb_flow_dissector_init() that control key will 1098 * be always present. 1099 */ 1100 key_control = skb_flow_dissector_target(flow_dissector, 1101 FLOW_DISSECTOR_KEY_CONTROL, 1102 target_container); 1103 1104 /* It is ensured by skb_flow_dissector_init() that basic key will 1105 * be always present. 1106 */ 1107 key_basic = skb_flow_dissector_target(flow_dissector, 1108 FLOW_DISSECTOR_KEY_BASIC, 1109 target_container); 1110 1111 rcu_read_lock(); 1112 1113 if (skb) { 1114 if (!net) { 1115 if (skb->dev) 1116 net = dev_net_rcu(skb->dev); 1117 else if (skb->sk) 1118 net = sock_net(skb->sk); 1119 } 1120 } 1121 1122 DEBUG_NET_WARN_ON_ONCE(!net); 1123 if (net) { 1124 enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR; 1125 struct bpf_prog_array *run_array; 1126 1127 run_array = rcu_dereference(init_net.bpf.run_array[type]); 1128 if (!run_array) 1129 run_array = rcu_dereference(net->bpf.run_array[type]); 1130 1131 if (run_array) { 1132 struct bpf_flow_keys flow_keys; 1133 struct bpf_flow_dissector ctx = { 1134 .flow_keys = &flow_keys, 1135 .data = data, 1136 .data_end = data + hlen, 1137 }; 1138 __be16 n_proto = proto; 1139 struct bpf_prog *prog; 1140 u32 result; 1141 1142 if (skb) { 1143 ctx.skb = skb; 1144 /* we can't use 'proto' in the skb case 1145 * because it might be set to skb->vlan_proto 1146 * which has been pulled from the data 1147 */ 1148 n_proto = skb->protocol; 1149 } 1150 1151 prog = READ_ONCE(run_array->items[0].prog); 1152 result = bpf_flow_dissect(prog, &ctx, n_proto, nhoff, 1153 hlen, flags); 1154 if (result != BPF_FLOW_DISSECTOR_CONTINUE) { 1155 __skb_flow_bpf_to_target(&flow_keys, flow_dissector, 1156 target_container); 1157 rcu_read_unlock(); 1158 return result == BPF_OK; 1159 } 1160 } 1161 } 1162 1163 rcu_read_unlock(); 1164 1165 if (dissector_uses_key(flow_dissector, 1166 FLOW_DISSECTOR_KEY_ETH_ADDRS)) { 1167 struct ethhdr *eth = eth_hdr(skb); 1168 struct flow_dissector_key_eth_addrs *key_eth_addrs; 1169 1170 key_eth_addrs = skb_flow_dissector_target(flow_dissector, 1171 FLOW_DISSECTOR_KEY_ETH_ADDRS, 1172 target_container); 1173 memcpy(key_eth_addrs, eth, sizeof(*key_eth_addrs)); 1174 } 1175 1176 if (dissector_uses_key(flow_dissector, 1177 FLOW_DISSECTOR_KEY_NUM_OF_VLANS)) { 1178 struct flow_dissector_key_num_of_vlans *key_num_of_vlans; 1179 1180 key_num_of_vlans = skb_flow_dissector_target(flow_dissector, 1181 FLOW_DISSECTOR_KEY_NUM_OF_VLANS, 1182 target_container); 1183 key_num_of_vlans->num_of_vlans = 0; 1184 } 1185 1186 proto_again: 1187 fdret = FLOW_DISSECT_RET_CONTINUE; 1188 1189 switch (proto) { 1190 case htons(ETH_P_IP): { 1191 const struct iphdr *iph; 1192 struct iphdr _iph; 1193 1194 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph); 1195 if (!iph || iph->ihl < 5) { 1196 fdret = FLOW_DISSECT_RET_OUT_BAD; 1197 break; 1198 } 1199 1200 nhoff += iph->ihl * 4; 1201 1202 ip_proto = iph->protocol; 1203 1204 if (dissector_uses_key(flow_dissector, 1205 FLOW_DISSECTOR_KEY_IPV4_ADDRS)) { 1206 key_addrs = skb_flow_dissector_target(flow_dissector, 1207 FLOW_DISSECTOR_KEY_IPV4_ADDRS, 1208 target_container); 1209 1210 memcpy(&key_addrs->v4addrs.src, &iph->saddr, 1211 sizeof(key_addrs->v4addrs.src)); 1212 memcpy(&key_addrs->v4addrs.dst, &iph->daddr, 1213 sizeof(key_addrs->v4addrs.dst)); 1214 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; 1215 } 1216 1217 __skb_flow_dissect_ipv4(skb, flow_dissector, 1218 target_container, data, iph); 1219 1220 if (ip_is_fragment(iph)) { 1221 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 1222 1223 if (iph->frag_off & htons(IP_OFFSET)) { 1224 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1225 break; 1226 } else { 1227 key_control->flags |= FLOW_DIS_FIRST_FRAG; 1228 if (!(flags & 1229 FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) { 1230 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1231 break; 1232 } 1233 } 1234 } 1235 1236 break; 1237 } 1238 case htons(ETH_P_IPV6): { 1239 const struct ipv6hdr *iph; 1240 struct ipv6hdr _iph; 1241 1242 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph); 1243 if (!iph) { 1244 fdret = FLOW_DISSECT_RET_OUT_BAD; 1245 break; 1246 } 1247 1248 ip_proto = iph->nexthdr; 1249 nhoff += sizeof(struct ipv6hdr); 1250 1251 if (dissector_uses_key(flow_dissector, 1252 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) { 1253 key_addrs = skb_flow_dissector_target(flow_dissector, 1254 FLOW_DISSECTOR_KEY_IPV6_ADDRS, 1255 target_container); 1256 1257 memcpy(&key_addrs->v6addrs.src, &iph->saddr, 1258 sizeof(key_addrs->v6addrs.src)); 1259 memcpy(&key_addrs->v6addrs.dst, &iph->daddr, 1260 sizeof(key_addrs->v6addrs.dst)); 1261 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 1262 } 1263 1264 if ((dissector_uses_key(flow_dissector, 1265 FLOW_DISSECTOR_KEY_FLOW_LABEL) || 1266 (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) && 1267 ip6_flowlabel(iph)) { 1268 __be32 flow_label = ip6_flowlabel(iph); 1269 1270 if (dissector_uses_key(flow_dissector, 1271 FLOW_DISSECTOR_KEY_FLOW_LABEL)) { 1272 key_tags = skb_flow_dissector_target(flow_dissector, 1273 FLOW_DISSECTOR_KEY_FLOW_LABEL, 1274 target_container); 1275 key_tags->flow_label = ntohl(flow_label); 1276 } 1277 if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) { 1278 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1279 break; 1280 } 1281 } 1282 1283 __skb_flow_dissect_ipv6(skb, flow_dissector, 1284 target_container, data, iph); 1285 1286 break; 1287 } 1288 case htons(ETH_P_8021AD): 1289 case htons(ETH_P_8021Q): { 1290 const struct vlan_hdr *vlan = NULL; 1291 struct vlan_hdr _vlan; 1292 __be16 saved_vlan_tpid = proto; 1293 1294 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX && 1295 skb && skb_vlan_tag_present(skb)) { 1296 proto = skb->protocol; 1297 } else { 1298 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan), 1299 data, hlen, &_vlan); 1300 if (!vlan) { 1301 fdret = FLOW_DISSECT_RET_OUT_BAD; 1302 break; 1303 } 1304 1305 proto = vlan->h_vlan_encapsulated_proto; 1306 nhoff += sizeof(*vlan); 1307 } 1308 1309 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_NUM_OF_VLANS) && 1310 !(key_control->flags & FLOW_DIS_ENCAPSULATION)) { 1311 struct flow_dissector_key_num_of_vlans *key_nvs; 1312 1313 key_nvs = skb_flow_dissector_target(flow_dissector, 1314 FLOW_DISSECTOR_KEY_NUM_OF_VLANS, 1315 target_container); 1316 key_nvs->num_of_vlans++; 1317 } 1318 1319 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) { 1320 dissector_vlan = FLOW_DISSECTOR_KEY_VLAN; 1321 } else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) { 1322 dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN; 1323 } else { 1324 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1325 break; 1326 } 1327 1328 if (dissector_uses_key(flow_dissector, dissector_vlan)) { 1329 key_vlan = skb_flow_dissector_target(flow_dissector, 1330 dissector_vlan, 1331 target_container); 1332 1333 if (!vlan) { 1334 key_vlan->vlan_id = skb_vlan_tag_get_id(skb); 1335 key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb); 1336 } else { 1337 key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) & 1338 VLAN_VID_MASK; 1339 key_vlan->vlan_priority = 1340 (ntohs(vlan->h_vlan_TCI) & 1341 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT; 1342 } 1343 key_vlan->vlan_tpid = saved_vlan_tpid; 1344 key_vlan->vlan_eth_type = proto; 1345 } 1346 1347 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1348 break; 1349 } 1350 case htons(ETH_P_PPP_SES): { 1351 struct { 1352 struct pppoe_hdr hdr; 1353 __be16 proto; 1354 } *hdr, _hdr; 1355 u16 ppp_proto; 1356 1357 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr); 1358 if (!hdr) { 1359 fdret = FLOW_DISSECT_RET_OUT_BAD; 1360 break; 1361 } 1362 1363 if (!is_pppoe_ses_hdr_valid(&hdr->hdr)) { 1364 fdret = FLOW_DISSECT_RET_OUT_BAD; 1365 break; 1366 } 1367 1368 /* least significant bit of the most significant octet 1369 * indicates if protocol field was compressed 1370 */ 1371 ppp_proto = ntohs(hdr->proto); 1372 if (ppp_proto & 0x0100) { 1373 ppp_proto = ppp_proto >> 8; 1374 nhoff += PPPOE_SES_HLEN - 1; 1375 } else { 1376 nhoff += PPPOE_SES_HLEN; 1377 } 1378 1379 if (ppp_proto == PPP_IP) { 1380 proto = htons(ETH_P_IP); 1381 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1382 } else if (ppp_proto == PPP_IPV6) { 1383 proto = htons(ETH_P_IPV6); 1384 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1385 } else if (ppp_proto == PPP_MPLS_UC) { 1386 proto = htons(ETH_P_MPLS_UC); 1387 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1388 } else if (ppp_proto == PPP_MPLS_MC) { 1389 proto = htons(ETH_P_MPLS_MC); 1390 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1391 } else if (ppp_proto_is_valid(ppp_proto)) { 1392 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1393 } else { 1394 fdret = FLOW_DISSECT_RET_OUT_BAD; 1395 break; 1396 } 1397 1398 if (dissector_uses_key(flow_dissector, 1399 FLOW_DISSECTOR_KEY_PPPOE)) { 1400 struct flow_dissector_key_pppoe *key_pppoe; 1401 1402 key_pppoe = skb_flow_dissector_target(flow_dissector, 1403 FLOW_DISSECTOR_KEY_PPPOE, 1404 target_container); 1405 key_pppoe->session_id = hdr->hdr.sid; 1406 key_pppoe->ppp_proto = htons(ppp_proto); 1407 key_pppoe->type = htons(ETH_P_PPP_SES); 1408 } 1409 break; 1410 } 1411 case htons(ETH_P_TIPC): { 1412 struct tipc_basic_hdr *hdr, _hdr; 1413 1414 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), 1415 data, hlen, &_hdr); 1416 if (!hdr) { 1417 fdret = FLOW_DISSECT_RET_OUT_BAD; 1418 break; 1419 } 1420 1421 if (dissector_uses_key(flow_dissector, 1422 FLOW_DISSECTOR_KEY_TIPC)) { 1423 key_addrs = skb_flow_dissector_target(flow_dissector, 1424 FLOW_DISSECTOR_KEY_TIPC, 1425 target_container); 1426 key_addrs->tipckey.key = tipc_hdr_rps_key(hdr); 1427 key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC; 1428 } 1429 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1430 break; 1431 } 1432 1433 case htons(ETH_P_MPLS_UC): 1434 case htons(ETH_P_MPLS_MC): 1435 fdret = __skb_flow_dissect_mpls(skb, flow_dissector, 1436 target_container, data, 1437 nhoff, hlen, mpls_lse, 1438 &mpls_el); 1439 nhoff += sizeof(struct mpls_label); 1440 mpls_lse++; 1441 break; 1442 case htons(ETH_P_FCOE): 1443 if ((hlen - nhoff) < FCOE_HEADER_LEN) { 1444 fdret = FLOW_DISSECT_RET_OUT_BAD; 1445 break; 1446 } 1447 1448 nhoff += FCOE_HEADER_LEN; 1449 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1450 break; 1451 1452 case htons(ETH_P_ARP): 1453 case htons(ETH_P_RARP): 1454 fdret = __skb_flow_dissect_arp(skb, flow_dissector, 1455 target_container, data, 1456 nhoff, hlen); 1457 break; 1458 1459 case htons(ETH_P_BATMAN): 1460 fdret = __skb_flow_dissect_batadv(skb, key_control, data, 1461 &proto, &nhoff, hlen, flags); 1462 break; 1463 1464 case htons(ETH_P_1588): { 1465 struct ptp_header *hdr, _hdr; 1466 1467 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, 1468 hlen, &_hdr); 1469 if (!hdr) { 1470 fdret = FLOW_DISSECT_RET_OUT_BAD; 1471 break; 1472 } 1473 1474 nhoff += sizeof(struct ptp_header); 1475 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1476 break; 1477 } 1478 1479 case htons(ETH_P_PRP): 1480 case htons(ETH_P_HSR): { 1481 struct hsr_tag *hdr, _hdr; 1482 1483 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, 1484 &_hdr); 1485 if (!hdr) { 1486 fdret = FLOW_DISSECT_RET_OUT_BAD; 1487 break; 1488 } 1489 1490 proto = hdr->encap_proto; 1491 nhoff += HSR_HLEN; 1492 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1493 break; 1494 } 1495 1496 case htons(ETH_P_CFM): 1497 fdret = __skb_flow_dissect_cfm(skb, flow_dissector, 1498 target_container, data, 1499 nhoff, hlen); 1500 break; 1501 1502 default: 1503 fdret = FLOW_DISSECT_RET_OUT_BAD; 1504 break; 1505 } 1506 1507 /* Process result of proto processing */ 1508 switch (fdret) { 1509 case FLOW_DISSECT_RET_OUT_GOOD: 1510 goto out_good; 1511 case FLOW_DISSECT_RET_PROTO_AGAIN: 1512 if (skb_flow_dissect_allowed(&num_hdrs)) 1513 goto proto_again; 1514 goto out_good; 1515 case FLOW_DISSECT_RET_CONTINUE: 1516 case FLOW_DISSECT_RET_IPPROTO_AGAIN: 1517 break; 1518 case FLOW_DISSECT_RET_OUT_BAD: 1519 default: 1520 goto out_bad; 1521 } 1522 1523 ip_proto_again: 1524 fdret = FLOW_DISSECT_RET_CONTINUE; 1525 1526 switch (ip_proto) { 1527 case IPPROTO_GRE: 1528 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) { 1529 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1530 break; 1531 } 1532 1533 fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector, 1534 target_container, data, 1535 &proto, &nhoff, &hlen, flags); 1536 break; 1537 1538 case NEXTHDR_HOP: 1539 case NEXTHDR_ROUTING: 1540 case NEXTHDR_DEST: { 1541 u8 _opthdr[2], *opthdr; 1542 1543 if (proto != htons(ETH_P_IPV6)) 1544 break; 1545 1546 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr), 1547 data, hlen, &_opthdr); 1548 if (!opthdr) { 1549 fdret = FLOW_DISSECT_RET_OUT_BAD; 1550 break; 1551 } 1552 1553 ip_proto = opthdr[0]; 1554 nhoff += (opthdr[1] + 1) << 3; 1555 1556 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN; 1557 break; 1558 } 1559 case NEXTHDR_FRAGMENT: { 1560 struct frag_hdr _fh, *fh; 1561 1562 if (proto != htons(ETH_P_IPV6)) 1563 break; 1564 1565 fh = __skb_header_pointer(skb, nhoff, sizeof(_fh), 1566 data, hlen, &_fh); 1567 1568 if (!fh) { 1569 fdret = FLOW_DISSECT_RET_OUT_BAD; 1570 break; 1571 } 1572 1573 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 1574 1575 nhoff += sizeof(_fh); 1576 ip_proto = fh->nexthdr; 1577 1578 if (!(fh->frag_off & htons(IP6_OFFSET))) { 1579 key_control->flags |= FLOW_DIS_FIRST_FRAG; 1580 if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) { 1581 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN; 1582 break; 1583 } 1584 } 1585 1586 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1587 break; 1588 } 1589 case IPPROTO_IPIP: 1590 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) { 1591 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1592 break; 1593 } 1594 1595 proto = htons(ETH_P_IP); 1596 1597 key_control->flags |= FLOW_DIS_ENCAPSULATION; 1598 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) { 1599 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1600 break; 1601 } 1602 1603 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1604 break; 1605 1606 case IPPROTO_IPV6: 1607 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) { 1608 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1609 break; 1610 } 1611 1612 proto = htons(ETH_P_IPV6); 1613 1614 key_control->flags |= FLOW_DIS_ENCAPSULATION; 1615 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) { 1616 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1617 break; 1618 } 1619 1620 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1621 break; 1622 1623 1624 case IPPROTO_MPLS: 1625 proto = htons(ETH_P_MPLS_UC); 1626 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1627 break; 1628 1629 case IPPROTO_TCP: 1630 __skb_flow_dissect_tcp(skb, flow_dissector, target_container, 1631 data, nhoff, hlen); 1632 break; 1633 1634 case IPPROTO_ICMP: 1635 case IPPROTO_ICMPV6: 1636 __skb_flow_dissect_icmp(skb, flow_dissector, target_container, 1637 data, nhoff, hlen); 1638 break; 1639 case IPPROTO_L2TP: 1640 __skb_flow_dissect_l2tpv3(skb, flow_dissector, target_container, 1641 data, nhoff, hlen); 1642 break; 1643 case IPPROTO_ESP: 1644 __skb_flow_dissect_esp(skb, flow_dissector, target_container, 1645 data, nhoff, hlen); 1646 break; 1647 case IPPROTO_AH: 1648 __skb_flow_dissect_ah(skb, flow_dissector, target_container, 1649 data, nhoff, hlen); 1650 break; 1651 default: 1652 break; 1653 } 1654 1655 if (!(key_control->flags & FLOW_DIS_IS_FRAGMENT)) 1656 __skb_flow_dissect_ports(skb, flow_dissector, target_container, 1657 data, nhoff, ip_proto, hlen); 1658 1659 /* Process result of IP proto processing */ 1660 switch (fdret) { 1661 case FLOW_DISSECT_RET_PROTO_AGAIN: 1662 if (skb_flow_dissect_allowed(&num_hdrs)) 1663 goto proto_again; 1664 break; 1665 case FLOW_DISSECT_RET_IPPROTO_AGAIN: 1666 if (skb_flow_dissect_allowed(&num_hdrs)) 1667 goto ip_proto_again; 1668 break; 1669 case FLOW_DISSECT_RET_OUT_GOOD: 1670 case FLOW_DISSECT_RET_CONTINUE: 1671 break; 1672 case FLOW_DISSECT_RET_OUT_BAD: 1673 default: 1674 goto out_bad; 1675 } 1676 1677 out_good: 1678 ret = true; 1679 1680 out: 1681 key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen); 1682 key_basic->n_proto = proto; 1683 key_basic->ip_proto = ip_proto; 1684 1685 return ret; 1686 1687 out_bad: 1688 ret = false; 1689 goto out; 1690 } 1691 EXPORT_SYMBOL(__skb_flow_dissect); 1692 1693 static siphash_aligned_key_t hashrnd; 1694 static __always_inline void __flow_hash_secret_init(void) 1695 { 1696 net_get_random_once(&hashrnd, sizeof(hashrnd)); 1697 } 1698 1699 static const void *flow_keys_hash_start(const struct flow_keys *flow) 1700 { 1701 BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % SIPHASH_ALIGNMENT); 1702 return &flow->FLOW_KEYS_HASH_START_FIELD; 1703 } 1704 1705 static inline size_t flow_keys_hash_length(const struct flow_keys *flow) 1706 { 1707 size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs); 1708 1709 BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32)); 1710 1711 switch (flow->control.addr_type) { 1712 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1713 diff -= sizeof(flow->addrs.v4addrs); 1714 break; 1715 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1716 diff -= sizeof(flow->addrs.v6addrs); 1717 break; 1718 case FLOW_DISSECTOR_KEY_TIPC: 1719 diff -= sizeof(flow->addrs.tipckey); 1720 break; 1721 } 1722 return sizeof(*flow) - diff; 1723 } 1724 1725 __be32 flow_get_u32_src(const struct flow_keys *flow) 1726 { 1727 switch (flow->control.addr_type) { 1728 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1729 return flow->addrs.v4addrs.src; 1730 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1731 return (__force __be32)ipv6_addr_hash( 1732 &flow->addrs.v6addrs.src); 1733 case FLOW_DISSECTOR_KEY_TIPC: 1734 return flow->addrs.tipckey.key; 1735 default: 1736 return 0; 1737 } 1738 } 1739 EXPORT_SYMBOL(flow_get_u32_src); 1740 1741 __be32 flow_get_u32_dst(const struct flow_keys *flow) 1742 { 1743 switch (flow->control.addr_type) { 1744 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1745 return flow->addrs.v4addrs.dst; 1746 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1747 return (__force __be32)ipv6_addr_hash( 1748 &flow->addrs.v6addrs.dst); 1749 default: 1750 return 0; 1751 } 1752 } 1753 EXPORT_SYMBOL(flow_get_u32_dst); 1754 1755 /* Sort the source and destination IP and the ports, 1756 * to have consistent hash within the two directions 1757 */ 1758 static inline void __flow_hash_consistentify(struct flow_keys *keys) 1759 { 1760 int addr_diff, i; 1761 1762 switch (keys->control.addr_type) { 1763 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1764 if ((__force u32)keys->addrs.v4addrs.dst < 1765 (__force u32)keys->addrs.v4addrs.src) 1766 swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst); 1767 1768 if ((__force u16)keys->ports.dst < 1769 (__force u16)keys->ports.src) { 1770 swap(keys->ports.src, keys->ports.dst); 1771 } 1772 break; 1773 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1774 addr_diff = memcmp(&keys->addrs.v6addrs.dst, 1775 &keys->addrs.v6addrs.src, 1776 sizeof(keys->addrs.v6addrs.dst)); 1777 if (addr_diff < 0) { 1778 for (i = 0; i < 4; i++) 1779 swap(keys->addrs.v6addrs.src.s6_addr32[i], 1780 keys->addrs.v6addrs.dst.s6_addr32[i]); 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 } 1788 } 1789 1790 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, 1791 const siphash_key_t *keyval) 1792 { 1793 u32 hash; 1794 1795 __flow_hash_consistentify(keys); 1796 1797 hash = siphash(flow_keys_hash_start(keys), 1798 flow_keys_hash_length(keys), keyval); 1799 if (!hash) 1800 hash = 1; 1801 1802 return hash; 1803 } 1804 1805 u32 flow_hash_from_keys(struct flow_keys *keys) 1806 { 1807 __flow_hash_secret_init(); 1808 return __flow_hash_from_keys(keys, &hashrnd); 1809 } 1810 EXPORT_SYMBOL(flow_hash_from_keys); 1811 1812 u32 flow_hash_from_keys_seed(struct flow_keys *keys, 1813 const siphash_key_t *keyval) 1814 { 1815 return __flow_hash_from_keys(keys, keyval); 1816 } 1817 EXPORT_SYMBOL(flow_hash_from_keys_seed); 1818 1819 static inline u32 ___skb_get_hash(const struct sk_buff *skb, 1820 struct flow_keys *keys, 1821 const siphash_key_t *keyval) 1822 { 1823 skb_flow_dissect_flow_keys(skb, keys, 1824 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 1825 1826 return __flow_hash_from_keys(keys, keyval); 1827 } 1828 1829 struct _flow_keys_digest_data { 1830 __be16 n_proto; 1831 u8 ip_proto; 1832 u8 padding; 1833 __be32 ports; 1834 __be32 src; 1835 __be32 dst; 1836 }; 1837 1838 void make_flow_keys_digest(struct flow_keys_digest *digest, 1839 const struct flow_keys *flow) 1840 { 1841 struct _flow_keys_digest_data *data = 1842 (struct _flow_keys_digest_data *)digest; 1843 1844 BUILD_BUG_ON(sizeof(*data) > sizeof(*digest)); 1845 1846 memset(digest, 0, sizeof(*digest)); 1847 1848 data->n_proto = flow->basic.n_proto; 1849 data->ip_proto = flow->basic.ip_proto; 1850 data->ports = flow->ports.ports; 1851 data->src = flow->addrs.v4addrs.src; 1852 data->dst = flow->addrs.v4addrs.dst; 1853 } 1854 EXPORT_SYMBOL(make_flow_keys_digest); 1855 1856 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly; 1857 1858 u32 __skb_get_hash_symmetric_net(const struct net *net, const struct sk_buff *skb) 1859 { 1860 struct flow_keys keys; 1861 1862 __flow_hash_secret_init(); 1863 1864 memset(&keys, 0, sizeof(keys)); 1865 __skb_flow_dissect(net, skb, &flow_keys_dissector_symmetric, 1866 &keys, NULL, 0, 0, 0, 0); 1867 1868 return __flow_hash_from_keys(&keys, &hashrnd); 1869 } 1870 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric_net); 1871 1872 /** 1873 * __skb_get_hash_net: calculate a flow hash 1874 * @net: associated network namespace, derived from @skb if NULL 1875 * @skb: sk_buff to calculate flow hash from 1876 * 1877 * This function calculates a flow hash based on src/dst addresses 1878 * and src/dst port numbers. Sets hash in skb to non-zero hash value 1879 * on success, zero indicates no valid hash. Also, sets l4_hash in skb 1880 * if hash is a canonical 4-tuple hash over transport ports. 1881 */ 1882 void __skb_get_hash_net(const struct net *net, struct sk_buff *skb) 1883 { 1884 struct flow_keys keys; 1885 u32 hash; 1886 1887 memset(&keys, 0, sizeof(keys)); 1888 1889 __skb_flow_dissect(net, skb, &flow_keys_dissector, 1890 &keys, NULL, 0, 0, 0, 1891 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 1892 1893 __flow_hash_secret_init(); 1894 1895 hash = __flow_hash_from_keys(&keys, &hashrnd); 1896 1897 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); 1898 } 1899 EXPORT_SYMBOL(__skb_get_hash_net); 1900 1901 __u32 skb_get_hash_perturb(const struct sk_buff *skb, 1902 const siphash_key_t *perturb) 1903 { 1904 struct flow_keys keys; 1905 1906 return ___skb_get_hash(skb, &keys, perturb); 1907 } 1908 EXPORT_SYMBOL(skb_get_hash_perturb); 1909 1910 u32 __skb_get_poff(const struct sk_buff *skb, const void *data, 1911 const struct flow_keys_basic *keys, int hlen) 1912 { 1913 u32 poff = keys->control.thoff; 1914 1915 /* skip L4 headers for fragments after the first */ 1916 if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) && 1917 !(keys->control.flags & FLOW_DIS_FIRST_FRAG)) 1918 return poff; 1919 1920 switch (keys->basic.ip_proto) { 1921 case IPPROTO_TCP: { 1922 /* access doff as u8 to avoid unaligned access */ 1923 const u8 *doff; 1924 u8 _doff; 1925 1926 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff), 1927 data, hlen, &_doff); 1928 if (!doff) 1929 return poff; 1930 1931 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2); 1932 break; 1933 } 1934 case IPPROTO_UDP: 1935 case IPPROTO_UDPLITE: 1936 poff += sizeof(struct udphdr); 1937 break; 1938 /* For the rest, we do not really care about header 1939 * extensions at this point for now. 1940 */ 1941 case IPPROTO_ICMP: 1942 poff += sizeof(struct icmphdr); 1943 break; 1944 case IPPROTO_ICMPV6: 1945 poff += sizeof(struct icmp6hdr); 1946 break; 1947 case IPPROTO_IGMP: 1948 poff += sizeof(struct igmphdr); 1949 break; 1950 case IPPROTO_DCCP: 1951 poff += sizeof(struct dccp_hdr); 1952 break; 1953 case IPPROTO_SCTP: 1954 poff += sizeof(struct sctphdr); 1955 break; 1956 } 1957 1958 return poff; 1959 } 1960 1961 /** 1962 * skb_get_poff - get the offset to the payload 1963 * @skb: sk_buff to get the payload offset from 1964 * 1965 * The function will get the offset to the payload as far as it could 1966 * be dissected. The main user is currently BPF, so that we can dynamically 1967 * truncate packets without needing to push actual payload to the user 1968 * space and can analyze headers only, instead. 1969 */ 1970 u32 skb_get_poff(const struct sk_buff *skb) 1971 { 1972 struct flow_keys_basic keys; 1973 1974 if (!skb_flow_dissect_flow_keys_basic(NULL, skb, &keys, 1975 NULL, 0, 0, 0, 0)) 1976 return 0; 1977 1978 return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb)); 1979 } 1980 1981 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys) 1982 { 1983 memset(keys, 0, sizeof(*keys)); 1984 1985 memcpy(&keys->addrs.v6addrs.src, &fl6->saddr, 1986 sizeof(keys->addrs.v6addrs.src)); 1987 memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr, 1988 sizeof(keys->addrs.v6addrs.dst)); 1989 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 1990 keys->ports.src = fl6->fl6_sport; 1991 keys->ports.dst = fl6->fl6_dport; 1992 keys->keyid.keyid = fl6->fl6_gre_key; 1993 keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6); 1994 keys->basic.ip_proto = fl6->flowi6_proto; 1995 1996 return flow_hash_from_keys(keys); 1997 } 1998 EXPORT_SYMBOL(__get_hash_from_flowi6); 1999 2000 static const struct flow_dissector_key flow_keys_dissector_keys[] = { 2001 { 2002 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 2003 .offset = offsetof(struct flow_keys, control), 2004 }, 2005 { 2006 .key_id = FLOW_DISSECTOR_KEY_BASIC, 2007 .offset = offsetof(struct flow_keys, basic), 2008 }, 2009 { 2010 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS, 2011 .offset = offsetof(struct flow_keys, addrs.v4addrs), 2012 }, 2013 { 2014 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS, 2015 .offset = offsetof(struct flow_keys, addrs.v6addrs), 2016 }, 2017 { 2018 .key_id = FLOW_DISSECTOR_KEY_TIPC, 2019 .offset = offsetof(struct flow_keys, addrs.tipckey), 2020 }, 2021 { 2022 .key_id = FLOW_DISSECTOR_KEY_PORTS, 2023 .offset = offsetof(struct flow_keys, ports), 2024 }, 2025 { 2026 .key_id = FLOW_DISSECTOR_KEY_VLAN, 2027 .offset = offsetof(struct flow_keys, vlan), 2028 }, 2029 { 2030 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL, 2031 .offset = offsetof(struct flow_keys, tags), 2032 }, 2033 { 2034 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID, 2035 .offset = offsetof(struct flow_keys, keyid), 2036 }, 2037 }; 2038 2039 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = { 2040 { 2041 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 2042 .offset = offsetof(struct flow_keys, control), 2043 }, 2044 { 2045 .key_id = FLOW_DISSECTOR_KEY_BASIC, 2046 .offset = offsetof(struct flow_keys, basic), 2047 }, 2048 { 2049 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS, 2050 .offset = offsetof(struct flow_keys, addrs.v4addrs), 2051 }, 2052 { 2053 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS, 2054 .offset = offsetof(struct flow_keys, addrs.v6addrs), 2055 }, 2056 { 2057 .key_id = FLOW_DISSECTOR_KEY_PORTS, 2058 .offset = offsetof(struct flow_keys, ports), 2059 }, 2060 }; 2061 2062 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = { 2063 { 2064 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 2065 .offset = offsetof(struct flow_keys, control), 2066 }, 2067 { 2068 .key_id = FLOW_DISSECTOR_KEY_BASIC, 2069 .offset = offsetof(struct flow_keys, basic), 2070 }, 2071 }; 2072 2073 struct flow_dissector flow_keys_dissector __read_mostly; 2074 EXPORT_SYMBOL(flow_keys_dissector); 2075 2076 struct flow_dissector flow_keys_basic_dissector __read_mostly; 2077 EXPORT_SYMBOL(flow_keys_basic_dissector); 2078 2079 static int __init init_default_flow_dissectors(void) 2080 { 2081 skb_flow_dissector_init(&flow_keys_dissector, 2082 flow_keys_dissector_keys, 2083 ARRAY_SIZE(flow_keys_dissector_keys)); 2084 skb_flow_dissector_init(&flow_keys_dissector_symmetric, 2085 flow_keys_dissector_symmetric_keys, 2086 ARRAY_SIZE(flow_keys_dissector_symmetric_keys)); 2087 skb_flow_dissector_init(&flow_keys_basic_dissector, 2088 flow_keys_basic_dissector_keys, 2089 ARRAY_SIZE(flow_keys_basic_dissector_keys)); 2090 return 0; 2091 } 2092 core_initcall(init_default_flow_dissectors); 2093