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