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 |= (1 << 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_l2tpv3(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_l2tpv3 *key_l2tpv3; 214 struct { 215 __be32 session_id; 216 } *hdr, _hdr; 217 218 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_L2TPV3)) 219 return; 220 221 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr); 222 if (!hdr) 223 return; 224 225 key_l2tpv3 = skb_flow_dissector_target(flow_dissector, 226 FLOW_DISSECTOR_KEY_L2TPV3, 227 target_container); 228 229 key_l2tpv3->session_id = hdr->session_id; 230 } 231 232 void skb_flow_dissect_meta(const struct sk_buff *skb, 233 struct flow_dissector *flow_dissector, 234 void *target_container) 235 { 236 struct flow_dissector_key_meta *meta; 237 238 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_META)) 239 return; 240 241 meta = skb_flow_dissector_target(flow_dissector, 242 FLOW_DISSECTOR_KEY_META, 243 target_container); 244 meta->ingress_ifindex = skb->skb_iif; 245 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) 246 if (tc_skb_ext_tc_enabled()) { 247 struct tc_skb_ext *ext; 248 249 ext = skb_ext_find(skb, TC_SKB_EXT); 250 if (ext) 251 meta->l2_miss = ext->l2_miss; 252 } 253 #endif 254 } 255 EXPORT_SYMBOL(skb_flow_dissect_meta); 256 257 static void 258 skb_flow_dissect_set_enc_addr_type(enum flow_dissector_key_id type, 259 struct flow_dissector *flow_dissector, 260 void *target_container) 261 { 262 struct flow_dissector_key_control *ctrl; 263 264 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_CONTROL)) 265 return; 266 267 ctrl = skb_flow_dissector_target(flow_dissector, 268 FLOW_DISSECTOR_KEY_ENC_CONTROL, 269 target_container); 270 ctrl->addr_type = type; 271 } 272 273 void 274 skb_flow_dissect_ct(const struct sk_buff *skb, 275 struct flow_dissector *flow_dissector, 276 void *target_container, u16 *ctinfo_map, 277 size_t mapsize, bool post_ct, u16 zone) 278 { 279 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 280 struct flow_dissector_key_ct *key; 281 enum ip_conntrack_info ctinfo; 282 struct nf_conn_labels *cl; 283 struct nf_conn *ct; 284 285 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_CT)) 286 return; 287 288 ct = nf_ct_get(skb, &ctinfo); 289 if (!ct && !post_ct) 290 return; 291 292 key = skb_flow_dissector_target(flow_dissector, 293 FLOW_DISSECTOR_KEY_CT, 294 target_container); 295 296 if (!ct) { 297 key->ct_state = TCA_FLOWER_KEY_CT_FLAGS_TRACKED | 298 TCA_FLOWER_KEY_CT_FLAGS_INVALID; 299 key->ct_zone = zone; 300 return; 301 } 302 303 if (ctinfo < mapsize) 304 key->ct_state = ctinfo_map[ctinfo]; 305 #if IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) 306 key->ct_zone = ct->zone.id; 307 #endif 308 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) 309 key->ct_mark = READ_ONCE(ct->mark); 310 #endif 311 312 cl = nf_ct_labels_find(ct); 313 if (cl) 314 memcpy(key->ct_labels, cl->bits, sizeof(key->ct_labels)); 315 #endif /* CONFIG_NF_CONNTRACK */ 316 } 317 EXPORT_SYMBOL(skb_flow_dissect_ct); 318 319 void 320 skb_flow_dissect_tunnel_info(const struct sk_buff *skb, 321 struct flow_dissector *flow_dissector, 322 void *target_container) 323 { 324 struct ip_tunnel_info *info; 325 struct ip_tunnel_key *key; 326 327 /* A quick check to see if there might be something to do. */ 328 if (!dissector_uses_key(flow_dissector, 329 FLOW_DISSECTOR_KEY_ENC_KEYID) && 330 !dissector_uses_key(flow_dissector, 331 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) && 332 !dissector_uses_key(flow_dissector, 333 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) && 334 !dissector_uses_key(flow_dissector, 335 FLOW_DISSECTOR_KEY_ENC_CONTROL) && 336 !dissector_uses_key(flow_dissector, 337 FLOW_DISSECTOR_KEY_ENC_PORTS) && 338 !dissector_uses_key(flow_dissector, 339 FLOW_DISSECTOR_KEY_ENC_IP) && 340 !dissector_uses_key(flow_dissector, 341 FLOW_DISSECTOR_KEY_ENC_OPTS)) 342 return; 343 344 info = skb_tunnel_info(skb); 345 if (!info) 346 return; 347 348 key = &info->key; 349 350 switch (ip_tunnel_info_af(info)) { 351 case AF_INET: 352 skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV4_ADDRS, 353 flow_dissector, 354 target_container); 355 if (dissector_uses_key(flow_dissector, 356 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) { 357 struct flow_dissector_key_ipv4_addrs *ipv4; 358 359 ipv4 = skb_flow_dissector_target(flow_dissector, 360 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS, 361 target_container); 362 ipv4->src = key->u.ipv4.src; 363 ipv4->dst = key->u.ipv4.dst; 364 } 365 break; 366 case AF_INET6: 367 skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV6_ADDRS, 368 flow_dissector, 369 target_container); 370 if (dissector_uses_key(flow_dissector, 371 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) { 372 struct flow_dissector_key_ipv6_addrs *ipv6; 373 374 ipv6 = skb_flow_dissector_target(flow_dissector, 375 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS, 376 target_container); 377 ipv6->src = key->u.ipv6.src; 378 ipv6->dst = key->u.ipv6.dst; 379 } 380 break; 381 } 382 383 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) { 384 struct flow_dissector_key_keyid *keyid; 385 386 keyid = skb_flow_dissector_target(flow_dissector, 387 FLOW_DISSECTOR_KEY_ENC_KEYID, 388 target_container); 389 keyid->keyid = tunnel_id_to_key32(key->tun_id); 390 } 391 392 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) { 393 struct flow_dissector_key_ports *tp; 394 395 tp = skb_flow_dissector_target(flow_dissector, 396 FLOW_DISSECTOR_KEY_ENC_PORTS, 397 target_container); 398 tp->src = key->tp_src; 399 tp->dst = key->tp_dst; 400 } 401 402 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_IP)) { 403 struct flow_dissector_key_ip *ip; 404 405 ip = skb_flow_dissector_target(flow_dissector, 406 FLOW_DISSECTOR_KEY_ENC_IP, 407 target_container); 408 ip->tos = key->tos; 409 ip->ttl = key->ttl; 410 } 411 412 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_OPTS)) { 413 struct flow_dissector_key_enc_opts *enc_opt; 414 415 enc_opt = skb_flow_dissector_target(flow_dissector, 416 FLOW_DISSECTOR_KEY_ENC_OPTS, 417 target_container); 418 419 if (info->options_len) { 420 enc_opt->len = info->options_len; 421 ip_tunnel_info_opts_get(enc_opt->data, info); 422 enc_opt->dst_opt_type = info->key.tun_flags & 423 TUNNEL_OPTIONS_PRESENT; 424 } 425 } 426 } 427 EXPORT_SYMBOL(skb_flow_dissect_tunnel_info); 428 429 void skb_flow_dissect_hash(const struct sk_buff *skb, 430 struct flow_dissector *flow_dissector, 431 void *target_container) 432 { 433 struct flow_dissector_key_hash *key; 434 435 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_HASH)) 436 return; 437 438 key = skb_flow_dissector_target(flow_dissector, 439 FLOW_DISSECTOR_KEY_HASH, 440 target_container); 441 442 key->hash = skb_get_hash_raw(skb); 443 } 444 EXPORT_SYMBOL(skb_flow_dissect_hash); 445 446 static enum flow_dissect_ret 447 __skb_flow_dissect_mpls(const struct sk_buff *skb, 448 struct flow_dissector *flow_dissector, 449 void *target_container, const void *data, int nhoff, 450 int hlen, int lse_index, bool *entropy_label) 451 { 452 struct mpls_label *hdr, _hdr; 453 u32 entry, label, bos; 454 455 if (!dissector_uses_key(flow_dissector, 456 FLOW_DISSECTOR_KEY_MPLS_ENTROPY) && 457 !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) 458 return FLOW_DISSECT_RET_OUT_GOOD; 459 460 if (lse_index >= FLOW_DIS_MPLS_MAX) 461 return FLOW_DISSECT_RET_OUT_GOOD; 462 463 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, 464 hlen, &_hdr); 465 if (!hdr) 466 return FLOW_DISSECT_RET_OUT_BAD; 467 468 entry = ntohl(hdr->entry); 469 label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT; 470 bos = (entry & MPLS_LS_S_MASK) >> MPLS_LS_S_SHIFT; 471 472 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) { 473 struct flow_dissector_key_mpls *key_mpls; 474 struct flow_dissector_mpls_lse *lse; 475 476 key_mpls = skb_flow_dissector_target(flow_dissector, 477 FLOW_DISSECTOR_KEY_MPLS, 478 target_container); 479 lse = &key_mpls->ls[lse_index]; 480 481 lse->mpls_ttl = (entry & MPLS_LS_TTL_MASK) >> MPLS_LS_TTL_SHIFT; 482 lse->mpls_bos = bos; 483 lse->mpls_tc = (entry & MPLS_LS_TC_MASK) >> MPLS_LS_TC_SHIFT; 484 lse->mpls_label = label; 485 dissector_set_mpls_lse(key_mpls, lse_index); 486 } 487 488 if (*entropy_label && 489 dissector_uses_key(flow_dissector, 490 FLOW_DISSECTOR_KEY_MPLS_ENTROPY)) { 491 struct flow_dissector_key_keyid *key_keyid; 492 493 key_keyid = skb_flow_dissector_target(flow_dissector, 494 FLOW_DISSECTOR_KEY_MPLS_ENTROPY, 495 target_container); 496 key_keyid->keyid = cpu_to_be32(label); 497 } 498 499 *entropy_label = label == MPLS_LABEL_ENTROPY; 500 501 return bos ? FLOW_DISSECT_RET_OUT_GOOD : FLOW_DISSECT_RET_PROTO_AGAIN; 502 } 503 504 static enum flow_dissect_ret 505 __skb_flow_dissect_arp(const struct sk_buff *skb, 506 struct flow_dissector *flow_dissector, 507 void *target_container, const void *data, 508 int nhoff, int hlen) 509 { 510 struct flow_dissector_key_arp *key_arp; 511 struct { 512 unsigned char ar_sha[ETH_ALEN]; 513 unsigned char ar_sip[4]; 514 unsigned char ar_tha[ETH_ALEN]; 515 unsigned char ar_tip[4]; 516 } *arp_eth, _arp_eth; 517 const struct arphdr *arp; 518 struct arphdr _arp; 519 520 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP)) 521 return FLOW_DISSECT_RET_OUT_GOOD; 522 523 arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data, 524 hlen, &_arp); 525 if (!arp) 526 return FLOW_DISSECT_RET_OUT_BAD; 527 528 if (arp->ar_hrd != htons(ARPHRD_ETHER) || 529 arp->ar_pro != htons(ETH_P_IP) || 530 arp->ar_hln != ETH_ALEN || 531 arp->ar_pln != 4 || 532 (arp->ar_op != htons(ARPOP_REPLY) && 533 arp->ar_op != htons(ARPOP_REQUEST))) 534 return FLOW_DISSECT_RET_OUT_BAD; 535 536 arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp), 537 sizeof(_arp_eth), data, 538 hlen, &_arp_eth); 539 if (!arp_eth) 540 return FLOW_DISSECT_RET_OUT_BAD; 541 542 key_arp = skb_flow_dissector_target(flow_dissector, 543 FLOW_DISSECTOR_KEY_ARP, 544 target_container); 545 546 memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip)); 547 memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip)); 548 549 /* Only store the lower byte of the opcode; 550 * this covers ARPOP_REPLY and ARPOP_REQUEST. 551 */ 552 key_arp->op = ntohs(arp->ar_op) & 0xff; 553 554 ether_addr_copy(key_arp->sha, arp_eth->ar_sha); 555 ether_addr_copy(key_arp->tha, arp_eth->ar_tha); 556 557 return FLOW_DISSECT_RET_OUT_GOOD; 558 } 559 560 static enum flow_dissect_ret 561 __skb_flow_dissect_cfm(const struct sk_buff *skb, 562 struct flow_dissector *flow_dissector, 563 void *target_container, const void *data, 564 int nhoff, int hlen) 565 { 566 struct flow_dissector_key_cfm *key, *hdr, _hdr; 567 568 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_CFM)) 569 return FLOW_DISSECT_RET_OUT_GOOD; 570 571 hdr = __skb_header_pointer(skb, nhoff, sizeof(*key), data, hlen, &_hdr); 572 if (!hdr) 573 return FLOW_DISSECT_RET_OUT_BAD; 574 575 key = skb_flow_dissector_target(flow_dissector, FLOW_DISSECTOR_KEY_CFM, 576 target_container); 577 578 key->mdl_ver = hdr->mdl_ver; 579 key->opcode = hdr->opcode; 580 581 return FLOW_DISSECT_RET_OUT_GOOD; 582 } 583 584 static enum flow_dissect_ret 585 __skb_flow_dissect_gre(const struct sk_buff *skb, 586 struct flow_dissector_key_control *key_control, 587 struct flow_dissector *flow_dissector, 588 void *target_container, const void *data, 589 __be16 *p_proto, int *p_nhoff, int *p_hlen, 590 unsigned int flags) 591 { 592 struct flow_dissector_key_keyid *key_keyid; 593 struct gre_base_hdr *hdr, _hdr; 594 int offset = 0; 595 u16 gre_ver; 596 597 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), 598 data, *p_hlen, &_hdr); 599 if (!hdr) 600 return FLOW_DISSECT_RET_OUT_BAD; 601 602 /* Only look inside GRE without routing */ 603 if (hdr->flags & GRE_ROUTING) 604 return FLOW_DISSECT_RET_OUT_GOOD; 605 606 /* Only look inside GRE for version 0 and 1 */ 607 gre_ver = ntohs(hdr->flags & GRE_VERSION); 608 if (gre_ver > 1) 609 return FLOW_DISSECT_RET_OUT_GOOD; 610 611 *p_proto = hdr->protocol; 612 if (gre_ver) { 613 /* Version1 must be PPTP, and check the flags */ 614 if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY))) 615 return FLOW_DISSECT_RET_OUT_GOOD; 616 } 617 618 offset += sizeof(struct gre_base_hdr); 619 620 if (hdr->flags & GRE_CSUM) 621 offset += sizeof_field(struct gre_full_hdr, csum) + 622 sizeof_field(struct gre_full_hdr, reserved1); 623 624 if (hdr->flags & GRE_KEY) { 625 const __be32 *keyid; 626 __be32 _keyid; 627 628 keyid = __skb_header_pointer(skb, *p_nhoff + offset, 629 sizeof(_keyid), 630 data, *p_hlen, &_keyid); 631 if (!keyid) 632 return FLOW_DISSECT_RET_OUT_BAD; 633 634 if (dissector_uses_key(flow_dissector, 635 FLOW_DISSECTOR_KEY_GRE_KEYID)) { 636 key_keyid = skb_flow_dissector_target(flow_dissector, 637 FLOW_DISSECTOR_KEY_GRE_KEYID, 638 target_container); 639 if (gre_ver == 0) 640 key_keyid->keyid = *keyid; 641 else 642 key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK; 643 } 644 offset += sizeof_field(struct gre_full_hdr, key); 645 } 646 647 if (hdr->flags & GRE_SEQ) 648 offset += sizeof_field(struct pptp_gre_header, seq); 649 650 if (gre_ver == 0) { 651 if (*p_proto == htons(ETH_P_TEB)) { 652 const struct ethhdr *eth; 653 struct ethhdr _eth; 654 655 eth = __skb_header_pointer(skb, *p_nhoff + offset, 656 sizeof(_eth), 657 data, *p_hlen, &_eth); 658 if (!eth) 659 return FLOW_DISSECT_RET_OUT_BAD; 660 *p_proto = eth->h_proto; 661 offset += sizeof(*eth); 662 663 /* Cap headers that we access via pointers at the 664 * end of the Ethernet header as our maximum alignment 665 * at that point is only 2 bytes. 666 */ 667 if (NET_IP_ALIGN) 668 *p_hlen = *p_nhoff + offset; 669 } 670 } else { /* version 1, must be PPTP */ 671 u8 _ppp_hdr[PPP_HDRLEN]; 672 u8 *ppp_hdr; 673 674 if (hdr->flags & GRE_ACK) 675 offset += sizeof_field(struct pptp_gre_header, ack); 676 677 ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset, 678 sizeof(_ppp_hdr), 679 data, *p_hlen, _ppp_hdr); 680 if (!ppp_hdr) 681 return FLOW_DISSECT_RET_OUT_BAD; 682 683 switch (PPP_PROTOCOL(ppp_hdr)) { 684 case PPP_IP: 685 *p_proto = htons(ETH_P_IP); 686 break; 687 case PPP_IPV6: 688 *p_proto = htons(ETH_P_IPV6); 689 break; 690 default: 691 /* Could probably catch some more like MPLS */ 692 break; 693 } 694 695 offset += PPP_HDRLEN; 696 } 697 698 *p_nhoff += offset; 699 key_control->flags |= FLOW_DIS_ENCAPSULATION; 700 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) 701 return FLOW_DISSECT_RET_OUT_GOOD; 702 703 return FLOW_DISSECT_RET_PROTO_AGAIN; 704 } 705 706 /** 707 * __skb_flow_dissect_batadv() - dissect batman-adv header 708 * @skb: sk_buff to with the batman-adv header 709 * @key_control: flow dissectors control key 710 * @data: raw buffer pointer to the packet, if NULL use skb->data 711 * @p_proto: pointer used to update the protocol to process next 712 * @p_nhoff: pointer used to update inner network header offset 713 * @hlen: packet header length 714 * @flags: any combination of FLOW_DISSECTOR_F_* 715 * 716 * ETH_P_BATMAN packets are tried to be dissected. Only 717 * &struct batadv_unicast packets are actually processed because they contain an 718 * inner ethernet header and are usually followed by actual network header. This 719 * allows the flow dissector to continue processing the packet. 720 * 721 * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found, 722 * FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation, 723 * otherwise FLOW_DISSECT_RET_OUT_BAD 724 */ 725 static enum flow_dissect_ret 726 __skb_flow_dissect_batadv(const struct sk_buff *skb, 727 struct flow_dissector_key_control *key_control, 728 const void *data, __be16 *p_proto, int *p_nhoff, 729 int hlen, unsigned int flags) 730 { 731 struct { 732 struct batadv_unicast_packet batadv_unicast; 733 struct ethhdr eth; 734 } *hdr, _hdr; 735 736 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen, 737 &_hdr); 738 if (!hdr) 739 return FLOW_DISSECT_RET_OUT_BAD; 740 741 if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION) 742 return FLOW_DISSECT_RET_OUT_BAD; 743 744 if (hdr->batadv_unicast.packet_type != BATADV_UNICAST) 745 return FLOW_DISSECT_RET_OUT_BAD; 746 747 *p_proto = hdr->eth.h_proto; 748 *p_nhoff += sizeof(*hdr); 749 750 key_control->flags |= FLOW_DIS_ENCAPSULATION; 751 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) 752 return FLOW_DISSECT_RET_OUT_GOOD; 753 754 return FLOW_DISSECT_RET_PROTO_AGAIN; 755 } 756 757 static void 758 __skb_flow_dissect_tcp(const struct sk_buff *skb, 759 struct flow_dissector *flow_dissector, 760 void *target_container, const void *data, 761 int thoff, int hlen) 762 { 763 struct flow_dissector_key_tcp *key_tcp; 764 struct tcphdr *th, _th; 765 766 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP)) 767 return; 768 769 th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th); 770 if (!th) 771 return; 772 773 if (unlikely(__tcp_hdrlen(th) < sizeof(_th))) 774 return; 775 776 key_tcp = skb_flow_dissector_target(flow_dissector, 777 FLOW_DISSECTOR_KEY_TCP, 778 target_container); 779 key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF)); 780 } 781 782 static void 783 __skb_flow_dissect_ports(const struct sk_buff *skb, 784 struct flow_dissector *flow_dissector, 785 void *target_container, const void *data, 786 int nhoff, u8 ip_proto, int hlen) 787 { 788 enum flow_dissector_key_id dissector_ports = FLOW_DISSECTOR_KEY_MAX; 789 struct flow_dissector_key_ports *key_ports; 790 791 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS)) 792 dissector_ports = FLOW_DISSECTOR_KEY_PORTS; 793 else if (dissector_uses_key(flow_dissector, 794 FLOW_DISSECTOR_KEY_PORTS_RANGE)) 795 dissector_ports = FLOW_DISSECTOR_KEY_PORTS_RANGE; 796 797 if (dissector_ports == FLOW_DISSECTOR_KEY_MAX) 798 return; 799 800 key_ports = skb_flow_dissector_target(flow_dissector, 801 dissector_ports, 802 target_container); 803 key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto, 804 data, hlen); 805 } 806 807 static void 808 __skb_flow_dissect_ipv4(const struct sk_buff *skb, 809 struct flow_dissector *flow_dissector, 810 void *target_container, const void *data, 811 const struct iphdr *iph) 812 { 813 struct flow_dissector_key_ip *key_ip; 814 815 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP)) 816 return; 817 818 key_ip = skb_flow_dissector_target(flow_dissector, 819 FLOW_DISSECTOR_KEY_IP, 820 target_container); 821 key_ip->tos = iph->tos; 822 key_ip->ttl = iph->ttl; 823 } 824 825 static void 826 __skb_flow_dissect_ipv6(const struct sk_buff *skb, 827 struct flow_dissector *flow_dissector, 828 void *target_container, const void *data, 829 const struct ipv6hdr *iph) 830 { 831 struct flow_dissector_key_ip *key_ip; 832 833 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP)) 834 return; 835 836 key_ip = skb_flow_dissector_target(flow_dissector, 837 FLOW_DISSECTOR_KEY_IP, 838 target_container); 839 key_ip->tos = ipv6_get_dsfield(iph); 840 key_ip->ttl = iph->hop_limit; 841 } 842 843 /* Maximum number of protocol headers that can be parsed in 844 * __skb_flow_dissect 845 */ 846 #define MAX_FLOW_DISSECT_HDRS 15 847 848 static bool skb_flow_dissect_allowed(int *num_hdrs) 849 { 850 ++*num_hdrs; 851 852 return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS); 853 } 854 855 static void __skb_flow_bpf_to_target(const struct bpf_flow_keys *flow_keys, 856 struct flow_dissector *flow_dissector, 857 void *target_container) 858 { 859 struct flow_dissector_key_ports *key_ports = NULL; 860 struct flow_dissector_key_control *key_control; 861 struct flow_dissector_key_basic *key_basic; 862 struct flow_dissector_key_addrs *key_addrs; 863 struct flow_dissector_key_tags *key_tags; 864 865 key_control = skb_flow_dissector_target(flow_dissector, 866 FLOW_DISSECTOR_KEY_CONTROL, 867 target_container); 868 key_control->thoff = flow_keys->thoff; 869 if (flow_keys->is_frag) 870 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 871 if (flow_keys->is_first_frag) 872 key_control->flags |= FLOW_DIS_FIRST_FRAG; 873 if (flow_keys->is_encap) 874 key_control->flags |= FLOW_DIS_ENCAPSULATION; 875 876 key_basic = skb_flow_dissector_target(flow_dissector, 877 FLOW_DISSECTOR_KEY_BASIC, 878 target_container); 879 key_basic->n_proto = flow_keys->n_proto; 880 key_basic->ip_proto = flow_keys->ip_proto; 881 882 if (flow_keys->addr_proto == ETH_P_IP && 883 dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) { 884 key_addrs = skb_flow_dissector_target(flow_dissector, 885 FLOW_DISSECTOR_KEY_IPV4_ADDRS, 886 target_container); 887 key_addrs->v4addrs.src = flow_keys->ipv4_src; 888 key_addrs->v4addrs.dst = flow_keys->ipv4_dst; 889 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; 890 } else if (flow_keys->addr_proto == ETH_P_IPV6 && 891 dissector_uses_key(flow_dissector, 892 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) { 893 key_addrs = skb_flow_dissector_target(flow_dissector, 894 FLOW_DISSECTOR_KEY_IPV6_ADDRS, 895 target_container); 896 memcpy(&key_addrs->v6addrs.src, &flow_keys->ipv6_src, 897 sizeof(key_addrs->v6addrs.src)); 898 memcpy(&key_addrs->v6addrs.dst, &flow_keys->ipv6_dst, 899 sizeof(key_addrs->v6addrs.dst)); 900 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 901 } 902 903 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS)) 904 key_ports = skb_flow_dissector_target(flow_dissector, 905 FLOW_DISSECTOR_KEY_PORTS, 906 target_container); 907 else if (dissector_uses_key(flow_dissector, 908 FLOW_DISSECTOR_KEY_PORTS_RANGE)) 909 key_ports = skb_flow_dissector_target(flow_dissector, 910 FLOW_DISSECTOR_KEY_PORTS_RANGE, 911 target_container); 912 913 if (key_ports) { 914 key_ports->src = flow_keys->sport; 915 key_ports->dst = flow_keys->dport; 916 } 917 918 if (dissector_uses_key(flow_dissector, 919 FLOW_DISSECTOR_KEY_FLOW_LABEL)) { 920 key_tags = skb_flow_dissector_target(flow_dissector, 921 FLOW_DISSECTOR_KEY_FLOW_LABEL, 922 target_container); 923 key_tags->flow_label = ntohl(flow_keys->flow_label); 924 } 925 } 926 927 u32 bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx, 928 __be16 proto, int nhoff, int hlen, unsigned int flags) 929 { 930 struct bpf_flow_keys *flow_keys = ctx->flow_keys; 931 u32 result; 932 933 /* Pass parameters to the BPF program */ 934 memset(flow_keys, 0, sizeof(*flow_keys)); 935 flow_keys->n_proto = proto; 936 flow_keys->nhoff = nhoff; 937 flow_keys->thoff = flow_keys->nhoff; 938 939 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG != 940 (int)FLOW_DISSECTOR_F_PARSE_1ST_FRAG); 941 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL != 942 (int)FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 943 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP != 944 (int)FLOW_DISSECTOR_F_STOP_AT_ENCAP); 945 flow_keys->flags = flags; 946 947 result = bpf_prog_run_pin_on_cpu(prog, ctx); 948 949 flow_keys->nhoff = clamp_t(u16, flow_keys->nhoff, nhoff, hlen); 950 flow_keys->thoff = clamp_t(u16, flow_keys->thoff, 951 flow_keys->nhoff, hlen); 952 953 return result; 954 } 955 956 static bool is_pppoe_ses_hdr_valid(const struct pppoe_hdr *hdr) 957 { 958 return hdr->ver == 1 && hdr->type == 1 && hdr->code == 0; 959 } 960 961 /** 962 * __skb_flow_dissect - extract the flow_keys struct and return it 963 * @net: associated network namespace, derived from @skb if NULL 964 * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified 965 * @flow_dissector: list of keys to dissect 966 * @target_container: target structure to put dissected values into 967 * @data: raw buffer pointer to the packet, if NULL use skb->data 968 * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol 969 * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb) 970 * @hlen: packet header length, if @data is NULL use skb_headlen(skb) 971 * @flags: flags that control the dissection process, e.g. 972 * FLOW_DISSECTOR_F_STOP_AT_ENCAP. 973 * 974 * The function will try to retrieve individual keys into target specified 975 * by flow_dissector from either the skbuff or a raw buffer specified by the 976 * rest parameters. 977 * 978 * Caller must take care of zeroing target container memory. 979 */ 980 bool __skb_flow_dissect(const struct net *net, 981 const struct sk_buff *skb, 982 struct flow_dissector *flow_dissector, 983 void *target_container, const void *data, 984 __be16 proto, int nhoff, int hlen, unsigned int flags) 985 { 986 struct flow_dissector_key_control *key_control; 987 struct flow_dissector_key_basic *key_basic; 988 struct flow_dissector_key_addrs *key_addrs; 989 struct flow_dissector_key_tags *key_tags; 990 struct flow_dissector_key_vlan *key_vlan; 991 enum flow_dissect_ret fdret; 992 enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX; 993 bool mpls_el = false; 994 int mpls_lse = 0; 995 int num_hdrs = 0; 996 u8 ip_proto = 0; 997 bool ret; 998 999 if (!data) { 1000 data = skb->data; 1001 proto = skb_vlan_tag_present(skb) ? 1002 skb->vlan_proto : skb->protocol; 1003 nhoff = skb_network_offset(skb); 1004 hlen = skb_headlen(skb); 1005 #if IS_ENABLED(CONFIG_NET_DSA) 1006 if (unlikely(skb->dev && netdev_uses_dsa(skb->dev) && 1007 proto == htons(ETH_P_XDSA))) { 1008 struct metadata_dst *md_dst = skb_metadata_dst(skb); 1009 const struct dsa_device_ops *ops; 1010 int offset = 0; 1011 1012 ops = skb->dev->dsa_ptr->tag_ops; 1013 /* Only DSA header taggers break flow dissection */ 1014 if (ops->needed_headroom && 1015 (!md_dst || md_dst->type != METADATA_HW_PORT_MUX)) { 1016 if (ops->flow_dissect) 1017 ops->flow_dissect(skb, &proto, &offset); 1018 else 1019 dsa_tag_generic_flow_dissect(skb, 1020 &proto, 1021 &offset); 1022 hlen -= offset; 1023 nhoff += offset; 1024 } 1025 } 1026 #endif 1027 } 1028 1029 /* It is ensured by skb_flow_dissector_init() that control key will 1030 * be always present. 1031 */ 1032 key_control = skb_flow_dissector_target(flow_dissector, 1033 FLOW_DISSECTOR_KEY_CONTROL, 1034 target_container); 1035 1036 /* It is ensured by skb_flow_dissector_init() that basic key will 1037 * be always present. 1038 */ 1039 key_basic = skb_flow_dissector_target(flow_dissector, 1040 FLOW_DISSECTOR_KEY_BASIC, 1041 target_container); 1042 1043 if (skb) { 1044 if (!net) { 1045 if (skb->dev) 1046 net = dev_net(skb->dev); 1047 else if (skb->sk) 1048 net = sock_net(skb->sk); 1049 } 1050 } 1051 1052 WARN_ON_ONCE(!net); 1053 if (net) { 1054 enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR; 1055 struct bpf_prog_array *run_array; 1056 1057 rcu_read_lock(); 1058 run_array = rcu_dereference(init_net.bpf.run_array[type]); 1059 if (!run_array) 1060 run_array = rcu_dereference(net->bpf.run_array[type]); 1061 1062 if (run_array) { 1063 struct bpf_flow_keys flow_keys; 1064 struct bpf_flow_dissector ctx = { 1065 .flow_keys = &flow_keys, 1066 .data = data, 1067 .data_end = data + hlen, 1068 }; 1069 __be16 n_proto = proto; 1070 struct bpf_prog *prog; 1071 u32 result; 1072 1073 if (skb) { 1074 ctx.skb = skb; 1075 /* we can't use 'proto' in the skb case 1076 * because it might be set to skb->vlan_proto 1077 * which has been pulled from the data 1078 */ 1079 n_proto = skb->protocol; 1080 } 1081 1082 prog = READ_ONCE(run_array->items[0].prog); 1083 result = bpf_flow_dissect(prog, &ctx, n_proto, nhoff, 1084 hlen, flags); 1085 if (result == BPF_FLOW_DISSECTOR_CONTINUE) 1086 goto dissect_continue; 1087 __skb_flow_bpf_to_target(&flow_keys, flow_dissector, 1088 target_container); 1089 rcu_read_unlock(); 1090 return result == BPF_OK; 1091 } 1092 dissect_continue: 1093 rcu_read_unlock(); 1094 } 1095 1096 if (dissector_uses_key(flow_dissector, 1097 FLOW_DISSECTOR_KEY_ETH_ADDRS)) { 1098 struct ethhdr *eth = eth_hdr(skb); 1099 struct flow_dissector_key_eth_addrs *key_eth_addrs; 1100 1101 key_eth_addrs = skb_flow_dissector_target(flow_dissector, 1102 FLOW_DISSECTOR_KEY_ETH_ADDRS, 1103 target_container); 1104 memcpy(key_eth_addrs, eth, sizeof(*key_eth_addrs)); 1105 } 1106 1107 if (dissector_uses_key(flow_dissector, 1108 FLOW_DISSECTOR_KEY_NUM_OF_VLANS)) { 1109 struct flow_dissector_key_num_of_vlans *key_num_of_vlans; 1110 1111 key_num_of_vlans = skb_flow_dissector_target(flow_dissector, 1112 FLOW_DISSECTOR_KEY_NUM_OF_VLANS, 1113 target_container); 1114 key_num_of_vlans->num_of_vlans = 0; 1115 } 1116 1117 proto_again: 1118 fdret = FLOW_DISSECT_RET_CONTINUE; 1119 1120 switch (proto) { 1121 case htons(ETH_P_IP): { 1122 const struct iphdr *iph; 1123 struct iphdr _iph; 1124 1125 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph); 1126 if (!iph || iph->ihl < 5) { 1127 fdret = FLOW_DISSECT_RET_OUT_BAD; 1128 break; 1129 } 1130 1131 nhoff += iph->ihl * 4; 1132 1133 ip_proto = iph->protocol; 1134 1135 if (dissector_uses_key(flow_dissector, 1136 FLOW_DISSECTOR_KEY_IPV4_ADDRS)) { 1137 key_addrs = skb_flow_dissector_target(flow_dissector, 1138 FLOW_DISSECTOR_KEY_IPV4_ADDRS, 1139 target_container); 1140 1141 memcpy(&key_addrs->v4addrs.src, &iph->saddr, 1142 sizeof(key_addrs->v4addrs.src)); 1143 memcpy(&key_addrs->v4addrs.dst, &iph->daddr, 1144 sizeof(key_addrs->v4addrs.dst)); 1145 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; 1146 } 1147 1148 __skb_flow_dissect_ipv4(skb, flow_dissector, 1149 target_container, data, iph); 1150 1151 if (ip_is_fragment(iph)) { 1152 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 1153 1154 if (iph->frag_off & htons(IP_OFFSET)) { 1155 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1156 break; 1157 } else { 1158 key_control->flags |= FLOW_DIS_FIRST_FRAG; 1159 if (!(flags & 1160 FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) { 1161 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1162 break; 1163 } 1164 } 1165 } 1166 1167 break; 1168 } 1169 case htons(ETH_P_IPV6): { 1170 const struct ipv6hdr *iph; 1171 struct ipv6hdr _iph; 1172 1173 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph); 1174 if (!iph) { 1175 fdret = FLOW_DISSECT_RET_OUT_BAD; 1176 break; 1177 } 1178 1179 ip_proto = iph->nexthdr; 1180 nhoff += sizeof(struct ipv6hdr); 1181 1182 if (dissector_uses_key(flow_dissector, 1183 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) { 1184 key_addrs = skb_flow_dissector_target(flow_dissector, 1185 FLOW_DISSECTOR_KEY_IPV6_ADDRS, 1186 target_container); 1187 1188 memcpy(&key_addrs->v6addrs.src, &iph->saddr, 1189 sizeof(key_addrs->v6addrs.src)); 1190 memcpy(&key_addrs->v6addrs.dst, &iph->daddr, 1191 sizeof(key_addrs->v6addrs.dst)); 1192 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 1193 } 1194 1195 if ((dissector_uses_key(flow_dissector, 1196 FLOW_DISSECTOR_KEY_FLOW_LABEL) || 1197 (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) && 1198 ip6_flowlabel(iph)) { 1199 __be32 flow_label = ip6_flowlabel(iph); 1200 1201 if (dissector_uses_key(flow_dissector, 1202 FLOW_DISSECTOR_KEY_FLOW_LABEL)) { 1203 key_tags = skb_flow_dissector_target(flow_dissector, 1204 FLOW_DISSECTOR_KEY_FLOW_LABEL, 1205 target_container); 1206 key_tags->flow_label = ntohl(flow_label); 1207 } 1208 if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) { 1209 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1210 break; 1211 } 1212 } 1213 1214 __skb_flow_dissect_ipv6(skb, flow_dissector, 1215 target_container, data, iph); 1216 1217 break; 1218 } 1219 case htons(ETH_P_8021AD): 1220 case htons(ETH_P_8021Q): { 1221 const struct vlan_hdr *vlan = NULL; 1222 struct vlan_hdr _vlan; 1223 __be16 saved_vlan_tpid = proto; 1224 1225 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX && 1226 skb && skb_vlan_tag_present(skb)) { 1227 proto = skb->protocol; 1228 } else { 1229 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan), 1230 data, hlen, &_vlan); 1231 if (!vlan) { 1232 fdret = FLOW_DISSECT_RET_OUT_BAD; 1233 break; 1234 } 1235 1236 proto = vlan->h_vlan_encapsulated_proto; 1237 nhoff += sizeof(*vlan); 1238 } 1239 1240 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_NUM_OF_VLANS) && 1241 !(key_control->flags & FLOW_DIS_ENCAPSULATION)) { 1242 struct flow_dissector_key_num_of_vlans *key_nvs; 1243 1244 key_nvs = skb_flow_dissector_target(flow_dissector, 1245 FLOW_DISSECTOR_KEY_NUM_OF_VLANS, 1246 target_container); 1247 key_nvs->num_of_vlans++; 1248 } 1249 1250 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) { 1251 dissector_vlan = FLOW_DISSECTOR_KEY_VLAN; 1252 } else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) { 1253 dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN; 1254 } else { 1255 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1256 break; 1257 } 1258 1259 if (dissector_uses_key(flow_dissector, dissector_vlan)) { 1260 key_vlan = skb_flow_dissector_target(flow_dissector, 1261 dissector_vlan, 1262 target_container); 1263 1264 if (!vlan) { 1265 key_vlan->vlan_id = skb_vlan_tag_get_id(skb); 1266 key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb); 1267 } else { 1268 key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) & 1269 VLAN_VID_MASK; 1270 key_vlan->vlan_priority = 1271 (ntohs(vlan->h_vlan_TCI) & 1272 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT; 1273 } 1274 key_vlan->vlan_tpid = saved_vlan_tpid; 1275 key_vlan->vlan_eth_type = proto; 1276 } 1277 1278 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1279 break; 1280 } 1281 case htons(ETH_P_PPP_SES): { 1282 struct { 1283 struct pppoe_hdr hdr; 1284 __be16 proto; 1285 } *hdr, _hdr; 1286 u16 ppp_proto; 1287 1288 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr); 1289 if (!hdr) { 1290 fdret = FLOW_DISSECT_RET_OUT_BAD; 1291 break; 1292 } 1293 1294 if (!is_pppoe_ses_hdr_valid(&hdr->hdr)) { 1295 fdret = FLOW_DISSECT_RET_OUT_BAD; 1296 break; 1297 } 1298 1299 /* least significant bit of the most significant octet 1300 * indicates if protocol field was compressed 1301 */ 1302 ppp_proto = ntohs(hdr->proto); 1303 if (ppp_proto & 0x0100) { 1304 ppp_proto = ppp_proto >> 8; 1305 nhoff += PPPOE_SES_HLEN - 1; 1306 } else { 1307 nhoff += PPPOE_SES_HLEN; 1308 } 1309 1310 if (ppp_proto == PPP_IP) { 1311 proto = htons(ETH_P_IP); 1312 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1313 } else if (ppp_proto == PPP_IPV6) { 1314 proto = htons(ETH_P_IPV6); 1315 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1316 } else if (ppp_proto == PPP_MPLS_UC) { 1317 proto = htons(ETH_P_MPLS_UC); 1318 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1319 } else if (ppp_proto == PPP_MPLS_MC) { 1320 proto = htons(ETH_P_MPLS_MC); 1321 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1322 } else if (ppp_proto_is_valid(ppp_proto)) { 1323 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1324 } else { 1325 fdret = FLOW_DISSECT_RET_OUT_BAD; 1326 break; 1327 } 1328 1329 if (dissector_uses_key(flow_dissector, 1330 FLOW_DISSECTOR_KEY_PPPOE)) { 1331 struct flow_dissector_key_pppoe *key_pppoe; 1332 1333 key_pppoe = skb_flow_dissector_target(flow_dissector, 1334 FLOW_DISSECTOR_KEY_PPPOE, 1335 target_container); 1336 key_pppoe->session_id = hdr->hdr.sid; 1337 key_pppoe->ppp_proto = htons(ppp_proto); 1338 key_pppoe->type = htons(ETH_P_PPP_SES); 1339 } 1340 break; 1341 } 1342 case htons(ETH_P_TIPC): { 1343 struct tipc_basic_hdr *hdr, _hdr; 1344 1345 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), 1346 data, hlen, &_hdr); 1347 if (!hdr) { 1348 fdret = FLOW_DISSECT_RET_OUT_BAD; 1349 break; 1350 } 1351 1352 if (dissector_uses_key(flow_dissector, 1353 FLOW_DISSECTOR_KEY_TIPC)) { 1354 key_addrs = skb_flow_dissector_target(flow_dissector, 1355 FLOW_DISSECTOR_KEY_TIPC, 1356 target_container); 1357 key_addrs->tipckey.key = tipc_hdr_rps_key(hdr); 1358 key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC; 1359 } 1360 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1361 break; 1362 } 1363 1364 case htons(ETH_P_MPLS_UC): 1365 case htons(ETH_P_MPLS_MC): 1366 fdret = __skb_flow_dissect_mpls(skb, flow_dissector, 1367 target_container, data, 1368 nhoff, hlen, mpls_lse, 1369 &mpls_el); 1370 nhoff += sizeof(struct mpls_label); 1371 mpls_lse++; 1372 break; 1373 case htons(ETH_P_FCOE): 1374 if ((hlen - nhoff) < FCOE_HEADER_LEN) { 1375 fdret = FLOW_DISSECT_RET_OUT_BAD; 1376 break; 1377 } 1378 1379 nhoff += FCOE_HEADER_LEN; 1380 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1381 break; 1382 1383 case htons(ETH_P_ARP): 1384 case htons(ETH_P_RARP): 1385 fdret = __skb_flow_dissect_arp(skb, flow_dissector, 1386 target_container, data, 1387 nhoff, hlen); 1388 break; 1389 1390 case htons(ETH_P_BATMAN): 1391 fdret = __skb_flow_dissect_batadv(skb, key_control, data, 1392 &proto, &nhoff, hlen, flags); 1393 break; 1394 1395 case htons(ETH_P_1588): { 1396 struct ptp_header *hdr, _hdr; 1397 1398 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, 1399 hlen, &_hdr); 1400 if (!hdr) { 1401 fdret = FLOW_DISSECT_RET_OUT_BAD; 1402 break; 1403 } 1404 1405 nhoff += ntohs(hdr->message_length); 1406 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1407 break; 1408 } 1409 1410 case htons(ETH_P_PRP): 1411 case htons(ETH_P_HSR): { 1412 struct hsr_tag *hdr, _hdr; 1413 1414 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, 1415 &_hdr); 1416 if (!hdr) { 1417 fdret = FLOW_DISSECT_RET_OUT_BAD; 1418 break; 1419 } 1420 1421 proto = hdr->encap_proto; 1422 nhoff += HSR_HLEN; 1423 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1424 break; 1425 } 1426 1427 case htons(ETH_P_CFM): 1428 fdret = __skb_flow_dissect_cfm(skb, flow_dissector, 1429 target_container, data, 1430 nhoff, hlen); 1431 break; 1432 1433 default: 1434 fdret = FLOW_DISSECT_RET_OUT_BAD; 1435 break; 1436 } 1437 1438 /* Process result of proto processing */ 1439 switch (fdret) { 1440 case FLOW_DISSECT_RET_OUT_GOOD: 1441 goto out_good; 1442 case FLOW_DISSECT_RET_PROTO_AGAIN: 1443 if (skb_flow_dissect_allowed(&num_hdrs)) 1444 goto proto_again; 1445 goto out_good; 1446 case FLOW_DISSECT_RET_CONTINUE: 1447 case FLOW_DISSECT_RET_IPPROTO_AGAIN: 1448 break; 1449 case FLOW_DISSECT_RET_OUT_BAD: 1450 default: 1451 goto out_bad; 1452 } 1453 1454 ip_proto_again: 1455 fdret = FLOW_DISSECT_RET_CONTINUE; 1456 1457 switch (ip_proto) { 1458 case IPPROTO_GRE: 1459 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) { 1460 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1461 break; 1462 } 1463 1464 fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector, 1465 target_container, data, 1466 &proto, &nhoff, &hlen, flags); 1467 break; 1468 1469 case NEXTHDR_HOP: 1470 case NEXTHDR_ROUTING: 1471 case NEXTHDR_DEST: { 1472 u8 _opthdr[2], *opthdr; 1473 1474 if (proto != htons(ETH_P_IPV6)) 1475 break; 1476 1477 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr), 1478 data, hlen, &_opthdr); 1479 if (!opthdr) { 1480 fdret = FLOW_DISSECT_RET_OUT_BAD; 1481 break; 1482 } 1483 1484 ip_proto = opthdr[0]; 1485 nhoff += (opthdr[1] + 1) << 3; 1486 1487 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN; 1488 break; 1489 } 1490 case NEXTHDR_FRAGMENT: { 1491 struct frag_hdr _fh, *fh; 1492 1493 if (proto != htons(ETH_P_IPV6)) 1494 break; 1495 1496 fh = __skb_header_pointer(skb, nhoff, sizeof(_fh), 1497 data, hlen, &_fh); 1498 1499 if (!fh) { 1500 fdret = FLOW_DISSECT_RET_OUT_BAD; 1501 break; 1502 } 1503 1504 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 1505 1506 nhoff += sizeof(_fh); 1507 ip_proto = fh->nexthdr; 1508 1509 if (!(fh->frag_off & htons(IP6_OFFSET))) { 1510 key_control->flags |= FLOW_DIS_FIRST_FRAG; 1511 if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) { 1512 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN; 1513 break; 1514 } 1515 } 1516 1517 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1518 break; 1519 } 1520 case IPPROTO_IPIP: 1521 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) { 1522 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1523 break; 1524 } 1525 1526 proto = htons(ETH_P_IP); 1527 1528 key_control->flags |= FLOW_DIS_ENCAPSULATION; 1529 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) { 1530 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1531 break; 1532 } 1533 1534 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1535 break; 1536 1537 case IPPROTO_IPV6: 1538 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) { 1539 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1540 break; 1541 } 1542 1543 proto = htons(ETH_P_IPV6); 1544 1545 key_control->flags |= FLOW_DIS_ENCAPSULATION; 1546 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) { 1547 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1548 break; 1549 } 1550 1551 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1552 break; 1553 1554 1555 case IPPROTO_MPLS: 1556 proto = htons(ETH_P_MPLS_UC); 1557 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1558 break; 1559 1560 case IPPROTO_TCP: 1561 __skb_flow_dissect_tcp(skb, flow_dissector, target_container, 1562 data, nhoff, hlen); 1563 break; 1564 1565 case IPPROTO_ICMP: 1566 case IPPROTO_ICMPV6: 1567 __skb_flow_dissect_icmp(skb, flow_dissector, target_container, 1568 data, nhoff, hlen); 1569 break; 1570 case IPPROTO_L2TP: 1571 __skb_flow_dissect_l2tpv3(skb, flow_dissector, target_container, 1572 data, nhoff, hlen); 1573 break; 1574 1575 default: 1576 break; 1577 } 1578 1579 if (!(key_control->flags & FLOW_DIS_IS_FRAGMENT)) 1580 __skb_flow_dissect_ports(skb, flow_dissector, target_container, 1581 data, nhoff, ip_proto, hlen); 1582 1583 /* Process result of IP proto processing */ 1584 switch (fdret) { 1585 case FLOW_DISSECT_RET_PROTO_AGAIN: 1586 if (skb_flow_dissect_allowed(&num_hdrs)) 1587 goto proto_again; 1588 break; 1589 case FLOW_DISSECT_RET_IPPROTO_AGAIN: 1590 if (skb_flow_dissect_allowed(&num_hdrs)) 1591 goto ip_proto_again; 1592 break; 1593 case FLOW_DISSECT_RET_OUT_GOOD: 1594 case FLOW_DISSECT_RET_CONTINUE: 1595 break; 1596 case FLOW_DISSECT_RET_OUT_BAD: 1597 default: 1598 goto out_bad; 1599 } 1600 1601 out_good: 1602 ret = true; 1603 1604 out: 1605 key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen); 1606 key_basic->n_proto = proto; 1607 key_basic->ip_proto = ip_proto; 1608 1609 return ret; 1610 1611 out_bad: 1612 ret = false; 1613 goto out; 1614 } 1615 EXPORT_SYMBOL(__skb_flow_dissect); 1616 1617 static siphash_aligned_key_t hashrnd; 1618 static __always_inline void __flow_hash_secret_init(void) 1619 { 1620 net_get_random_once(&hashrnd, sizeof(hashrnd)); 1621 } 1622 1623 static const void *flow_keys_hash_start(const struct flow_keys *flow) 1624 { 1625 BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % SIPHASH_ALIGNMENT); 1626 return &flow->FLOW_KEYS_HASH_START_FIELD; 1627 } 1628 1629 static inline size_t flow_keys_hash_length(const struct flow_keys *flow) 1630 { 1631 size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs); 1632 1633 BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32)); 1634 1635 switch (flow->control.addr_type) { 1636 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1637 diff -= sizeof(flow->addrs.v4addrs); 1638 break; 1639 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1640 diff -= sizeof(flow->addrs.v6addrs); 1641 break; 1642 case FLOW_DISSECTOR_KEY_TIPC: 1643 diff -= sizeof(flow->addrs.tipckey); 1644 break; 1645 } 1646 return sizeof(*flow) - diff; 1647 } 1648 1649 __be32 flow_get_u32_src(const struct flow_keys *flow) 1650 { 1651 switch (flow->control.addr_type) { 1652 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1653 return flow->addrs.v4addrs.src; 1654 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1655 return (__force __be32)ipv6_addr_hash( 1656 &flow->addrs.v6addrs.src); 1657 case FLOW_DISSECTOR_KEY_TIPC: 1658 return flow->addrs.tipckey.key; 1659 default: 1660 return 0; 1661 } 1662 } 1663 EXPORT_SYMBOL(flow_get_u32_src); 1664 1665 __be32 flow_get_u32_dst(const struct flow_keys *flow) 1666 { 1667 switch (flow->control.addr_type) { 1668 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1669 return flow->addrs.v4addrs.dst; 1670 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1671 return (__force __be32)ipv6_addr_hash( 1672 &flow->addrs.v6addrs.dst); 1673 default: 1674 return 0; 1675 } 1676 } 1677 EXPORT_SYMBOL(flow_get_u32_dst); 1678 1679 /* Sort the source and destination IP and the ports, 1680 * to have consistent hash within the two directions 1681 */ 1682 static inline void __flow_hash_consistentify(struct flow_keys *keys) 1683 { 1684 int addr_diff, i; 1685 1686 switch (keys->control.addr_type) { 1687 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1688 if ((__force u32)keys->addrs.v4addrs.dst < 1689 (__force u32)keys->addrs.v4addrs.src) 1690 swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst); 1691 1692 if ((__force u16)keys->ports.dst < 1693 (__force u16)keys->ports.src) { 1694 swap(keys->ports.src, keys->ports.dst); 1695 } 1696 break; 1697 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1698 addr_diff = memcmp(&keys->addrs.v6addrs.dst, 1699 &keys->addrs.v6addrs.src, 1700 sizeof(keys->addrs.v6addrs.dst)); 1701 if (addr_diff < 0) { 1702 for (i = 0; i < 4; i++) 1703 swap(keys->addrs.v6addrs.src.s6_addr32[i], 1704 keys->addrs.v6addrs.dst.s6_addr32[i]); 1705 } 1706 if ((__force u16)keys->ports.dst < 1707 (__force u16)keys->ports.src) { 1708 swap(keys->ports.src, keys->ports.dst); 1709 } 1710 break; 1711 } 1712 } 1713 1714 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, 1715 const siphash_key_t *keyval) 1716 { 1717 u32 hash; 1718 1719 __flow_hash_consistentify(keys); 1720 1721 hash = siphash(flow_keys_hash_start(keys), 1722 flow_keys_hash_length(keys), keyval); 1723 if (!hash) 1724 hash = 1; 1725 1726 return hash; 1727 } 1728 1729 u32 flow_hash_from_keys(struct flow_keys *keys) 1730 { 1731 __flow_hash_secret_init(); 1732 return __flow_hash_from_keys(keys, &hashrnd); 1733 } 1734 EXPORT_SYMBOL(flow_hash_from_keys); 1735 1736 static inline u32 ___skb_get_hash(const struct sk_buff *skb, 1737 struct flow_keys *keys, 1738 const siphash_key_t *keyval) 1739 { 1740 skb_flow_dissect_flow_keys(skb, keys, 1741 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 1742 1743 return __flow_hash_from_keys(keys, keyval); 1744 } 1745 1746 struct _flow_keys_digest_data { 1747 __be16 n_proto; 1748 u8 ip_proto; 1749 u8 padding; 1750 __be32 ports; 1751 __be32 src; 1752 __be32 dst; 1753 }; 1754 1755 void make_flow_keys_digest(struct flow_keys_digest *digest, 1756 const struct flow_keys *flow) 1757 { 1758 struct _flow_keys_digest_data *data = 1759 (struct _flow_keys_digest_data *)digest; 1760 1761 BUILD_BUG_ON(sizeof(*data) > sizeof(*digest)); 1762 1763 memset(digest, 0, sizeof(*digest)); 1764 1765 data->n_proto = flow->basic.n_proto; 1766 data->ip_proto = flow->basic.ip_proto; 1767 data->ports = flow->ports.ports; 1768 data->src = flow->addrs.v4addrs.src; 1769 data->dst = flow->addrs.v4addrs.dst; 1770 } 1771 EXPORT_SYMBOL(make_flow_keys_digest); 1772 1773 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly; 1774 1775 u32 __skb_get_hash_symmetric(const struct sk_buff *skb) 1776 { 1777 struct flow_keys keys; 1778 1779 __flow_hash_secret_init(); 1780 1781 memset(&keys, 0, sizeof(keys)); 1782 __skb_flow_dissect(NULL, skb, &flow_keys_dissector_symmetric, 1783 &keys, NULL, 0, 0, 0, 1784 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 1785 1786 return __flow_hash_from_keys(&keys, &hashrnd); 1787 } 1788 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric); 1789 1790 /** 1791 * __skb_get_hash: calculate a flow hash 1792 * @skb: sk_buff to calculate flow hash from 1793 * 1794 * This function calculates a flow hash based on src/dst addresses 1795 * and src/dst port numbers. Sets hash in skb to non-zero hash value 1796 * on success, zero indicates no valid hash. Also, sets l4_hash in skb 1797 * if hash is a canonical 4-tuple hash over transport ports. 1798 */ 1799 void __skb_get_hash(struct sk_buff *skb) 1800 { 1801 struct flow_keys keys; 1802 u32 hash; 1803 1804 __flow_hash_secret_init(); 1805 1806 hash = ___skb_get_hash(skb, &keys, &hashrnd); 1807 1808 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); 1809 } 1810 EXPORT_SYMBOL(__skb_get_hash); 1811 1812 __u32 skb_get_hash_perturb(const struct sk_buff *skb, 1813 const siphash_key_t *perturb) 1814 { 1815 struct flow_keys keys; 1816 1817 return ___skb_get_hash(skb, &keys, perturb); 1818 } 1819 EXPORT_SYMBOL(skb_get_hash_perturb); 1820 1821 u32 __skb_get_poff(const struct sk_buff *skb, const void *data, 1822 const struct flow_keys_basic *keys, int hlen) 1823 { 1824 u32 poff = keys->control.thoff; 1825 1826 /* skip L4 headers for fragments after the first */ 1827 if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) && 1828 !(keys->control.flags & FLOW_DIS_FIRST_FRAG)) 1829 return poff; 1830 1831 switch (keys->basic.ip_proto) { 1832 case IPPROTO_TCP: { 1833 /* access doff as u8 to avoid unaligned access */ 1834 const u8 *doff; 1835 u8 _doff; 1836 1837 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff), 1838 data, hlen, &_doff); 1839 if (!doff) 1840 return poff; 1841 1842 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2); 1843 break; 1844 } 1845 case IPPROTO_UDP: 1846 case IPPROTO_UDPLITE: 1847 poff += sizeof(struct udphdr); 1848 break; 1849 /* For the rest, we do not really care about header 1850 * extensions at this point for now. 1851 */ 1852 case IPPROTO_ICMP: 1853 poff += sizeof(struct icmphdr); 1854 break; 1855 case IPPROTO_ICMPV6: 1856 poff += sizeof(struct icmp6hdr); 1857 break; 1858 case IPPROTO_IGMP: 1859 poff += sizeof(struct igmphdr); 1860 break; 1861 case IPPROTO_DCCP: 1862 poff += sizeof(struct dccp_hdr); 1863 break; 1864 case IPPROTO_SCTP: 1865 poff += sizeof(struct sctphdr); 1866 break; 1867 } 1868 1869 return poff; 1870 } 1871 1872 /** 1873 * skb_get_poff - get the offset to the payload 1874 * @skb: sk_buff to get the payload offset from 1875 * 1876 * The function will get the offset to the payload as far as it could 1877 * be dissected. The main user is currently BPF, so that we can dynamically 1878 * truncate packets without needing to push actual payload to the user 1879 * space and can analyze headers only, instead. 1880 */ 1881 u32 skb_get_poff(const struct sk_buff *skb) 1882 { 1883 struct flow_keys_basic keys; 1884 1885 if (!skb_flow_dissect_flow_keys_basic(NULL, skb, &keys, 1886 NULL, 0, 0, 0, 0)) 1887 return 0; 1888 1889 return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb)); 1890 } 1891 1892 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys) 1893 { 1894 memset(keys, 0, sizeof(*keys)); 1895 1896 memcpy(&keys->addrs.v6addrs.src, &fl6->saddr, 1897 sizeof(keys->addrs.v6addrs.src)); 1898 memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr, 1899 sizeof(keys->addrs.v6addrs.dst)); 1900 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 1901 keys->ports.src = fl6->fl6_sport; 1902 keys->ports.dst = fl6->fl6_dport; 1903 keys->keyid.keyid = fl6->fl6_gre_key; 1904 keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6); 1905 keys->basic.ip_proto = fl6->flowi6_proto; 1906 1907 return flow_hash_from_keys(keys); 1908 } 1909 EXPORT_SYMBOL(__get_hash_from_flowi6); 1910 1911 static const struct flow_dissector_key flow_keys_dissector_keys[] = { 1912 { 1913 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 1914 .offset = offsetof(struct flow_keys, control), 1915 }, 1916 { 1917 .key_id = FLOW_DISSECTOR_KEY_BASIC, 1918 .offset = offsetof(struct flow_keys, basic), 1919 }, 1920 { 1921 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS, 1922 .offset = offsetof(struct flow_keys, addrs.v4addrs), 1923 }, 1924 { 1925 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS, 1926 .offset = offsetof(struct flow_keys, addrs.v6addrs), 1927 }, 1928 { 1929 .key_id = FLOW_DISSECTOR_KEY_TIPC, 1930 .offset = offsetof(struct flow_keys, addrs.tipckey), 1931 }, 1932 { 1933 .key_id = FLOW_DISSECTOR_KEY_PORTS, 1934 .offset = offsetof(struct flow_keys, ports), 1935 }, 1936 { 1937 .key_id = FLOW_DISSECTOR_KEY_VLAN, 1938 .offset = offsetof(struct flow_keys, vlan), 1939 }, 1940 { 1941 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL, 1942 .offset = offsetof(struct flow_keys, tags), 1943 }, 1944 { 1945 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID, 1946 .offset = offsetof(struct flow_keys, keyid), 1947 }, 1948 }; 1949 1950 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = { 1951 { 1952 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 1953 .offset = offsetof(struct flow_keys, control), 1954 }, 1955 { 1956 .key_id = FLOW_DISSECTOR_KEY_BASIC, 1957 .offset = offsetof(struct flow_keys, basic), 1958 }, 1959 { 1960 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS, 1961 .offset = offsetof(struct flow_keys, addrs.v4addrs), 1962 }, 1963 { 1964 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS, 1965 .offset = offsetof(struct flow_keys, addrs.v6addrs), 1966 }, 1967 { 1968 .key_id = FLOW_DISSECTOR_KEY_PORTS, 1969 .offset = offsetof(struct flow_keys, ports), 1970 }, 1971 }; 1972 1973 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = { 1974 { 1975 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 1976 .offset = offsetof(struct flow_keys, control), 1977 }, 1978 { 1979 .key_id = FLOW_DISSECTOR_KEY_BASIC, 1980 .offset = offsetof(struct flow_keys, basic), 1981 }, 1982 }; 1983 1984 struct flow_dissector flow_keys_dissector __read_mostly; 1985 EXPORT_SYMBOL(flow_keys_dissector); 1986 1987 struct flow_dissector flow_keys_basic_dissector __read_mostly; 1988 EXPORT_SYMBOL(flow_keys_basic_dissector); 1989 1990 static int __init init_default_flow_dissectors(void) 1991 { 1992 skb_flow_dissector_init(&flow_keys_dissector, 1993 flow_keys_dissector_keys, 1994 ARRAY_SIZE(flow_keys_dissector_keys)); 1995 skb_flow_dissector_init(&flow_keys_dissector_symmetric, 1996 flow_keys_dissector_symmetric_keys, 1997 ARRAY_SIZE(flow_keys_dissector_symmetric_keys)); 1998 skb_flow_dissector_init(&flow_keys_basic_dissector, 1999 flow_keys_basic_dissector_keys, 2000 ARRAY_SIZE(flow_keys_basic_dissector_keys)); 2001 return 0; 2002 } 2003 core_initcall(init_default_flow_dissectors); 2004