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 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 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, 191 void *data, 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, void *data, int nhoff, int hlen, 413 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, void *data, int nhoff, int hlen) 471 { 472 struct flow_dissector_key_arp *key_arp; 473 struct { 474 unsigned char ar_sha[ETH_ALEN]; 475 unsigned char ar_sip[4]; 476 unsigned char ar_tha[ETH_ALEN]; 477 unsigned char ar_tip[4]; 478 } *arp_eth, _arp_eth; 479 const struct arphdr *arp; 480 struct arphdr _arp; 481 482 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP)) 483 return FLOW_DISSECT_RET_OUT_GOOD; 484 485 arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data, 486 hlen, &_arp); 487 if (!arp) 488 return FLOW_DISSECT_RET_OUT_BAD; 489 490 if (arp->ar_hrd != htons(ARPHRD_ETHER) || 491 arp->ar_pro != htons(ETH_P_IP) || 492 arp->ar_hln != ETH_ALEN || 493 arp->ar_pln != 4 || 494 (arp->ar_op != htons(ARPOP_REPLY) && 495 arp->ar_op != htons(ARPOP_REQUEST))) 496 return FLOW_DISSECT_RET_OUT_BAD; 497 498 arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp), 499 sizeof(_arp_eth), data, 500 hlen, &_arp_eth); 501 if (!arp_eth) 502 return FLOW_DISSECT_RET_OUT_BAD; 503 504 key_arp = skb_flow_dissector_target(flow_dissector, 505 FLOW_DISSECTOR_KEY_ARP, 506 target_container); 507 508 memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip)); 509 memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip)); 510 511 /* Only store the lower byte of the opcode; 512 * this covers ARPOP_REPLY and ARPOP_REQUEST. 513 */ 514 key_arp->op = ntohs(arp->ar_op) & 0xff; 515 516 ether_addr_copy(key_arp->sha, arp_eth->ar_sha); 517 ether_addr_copy(key_arp->tha, arp_eth->ar_tha); 518 519 return FLOW_DISSECT_RET_OUT_GOOD; 520 } 521 522 static enum flow_dissect_ret 523 __skb_flow_dissect_gre(const struct sk_buff *skb, 524 struct flow_dissector_key_control *key_control, 525 struct flow_dissector *flow_dissector, 526 void *target_container, void *data, 527 __be16 *p_proto, int *p_nhoff, int *p_hlen, 528 unsigned int flags) 529 { 530 struct flow_dissector_key_keyid *key_keyid; 531 struct gre_base_hdr *hdr, _hdr; 532 int offset = 0; 533 u16 gre_ver; 534 535 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), 536 data, *p_hlen, &_hdr); 537 if (!hdr) 538 return FLOW_DISSECT_RET_OUT_BAD; 539 540 /* Only look inside GRE without routing */ 541 if (hdr->flags & GRE_ROUTING) 542 return FLOW_DISSECT_RET_OUT_GOOD; 543 544 /* Only look inside GRE for version 0 and 1 */ 545 gre_ver = ntohs(hdr->flags & GRE_VERSION); 546 if (gre_ver > 1) 547 return FLOW_DISSECT_RET_OUT_GOOD; 548 549 *p_proto = hdr->protocol; 550 if (gre_ver) { 551 /* Version1 must be PPTP, and check the flags */ 552 if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY))) 553 return FLOW_DISSECT_RET_OUT_GOOD; 554 } 555 556 offset += sizeof(struct gre_base_hdr); 557 558 if (hdr->flags & GRE_CSUM) 559 offset += sizeof_field(struct gre_full_hdr, csum) + 560 sizeof_field(struct gre_full_hdr, reserved1); 561 562 if (hdr->flags & GRE_KEY) { 563 const __be32 *keyid; 564 __be32 _keyid; 565 566 keyid = __skb_header_pointer(skb, *p_nhoff + offset, 567 sizeof(_keyid), 568 data, *p_hlen, &_keyid); 569 if (!keyid) 570 return FLOW_DISSECT_RET_OUT_BAD; 571 572 if (dissector_uses_key(flow_dissector, 573 FLOW_DISSECTOR_KEY_GRE_KEYID)) { 574 key_keyid = skb_flow_dissector_target(flow_dissector, 575 FLOW_DISSECTOR_KEY_GRE_KEYID, 576 target_container); 577 if (gre_ver == 0) 578 key_keyid->keyid = *keyid; 579 else 580 key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK; 581 } 582 offset += sizeof_field(struct gre_full_hdr, key); 583 } 584 585 if (hdr->flags & GRE_SEQ) 586 offset += sizeof_field(struct pptp_gre_header, seq); 587 588 if (gre_ver == 0) { 589 if (*p_proto == htons(ETH_P_TEB)) { 590 const struct ethhdr *eth; 591 struct ethhdr _eth; 592 593 eth = __skb_header_pointer(skb, *p_nhoff + offset, 594 sizeof(_eth), 595 data, *p_hlen, &_eth); 596 if (!eth) 597 return FLOW_DISSECT_RET_OUT_BAD; 598 *p_proto = eth->h_proto; 599 offset += sizeof(*eth); 600 601 /* Cap headers that we access via pointers at the 602 * end of the Ethernet header as our maximum alignment 603 * at that point is only 2 bytes. 604 */ 605 if (NET_IP_ALIGN) 606 *p_hlen = *p_nhoff + offset; 607 } 608 } else { /* version 1, must be PPTP */ 609 u8 _ppp_hdr[PPP_HDRLEN]; 610 u8 *ppp_hdr; 611 612 if (hdr->flags & GRE_ACK) 613 offset += sizeof_field(struct pptp_gre_header, ack); 614 615 ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset, 616 sizeof(_ppp_hdr), 617 data, *p_hlen, _ppp_hdr); 618 if (!ppp_hdr) 619 return FLOW_DISSECT_RET_OUT_BAD; 620 621 switch (PPP_PROTOCOL(ppp_hdr)) { 622 case PPP_IP: 623 *p_proto = htons(ETH_P_IP); 624 break; 625 case PPP_IPV6: 626 *p_proto = htons(ETH_P_IPV6); 627 break; 628 default: 629 /* Could probably catch some more like MPLS */ 630 break; 631 } 632 633 offset += PPP_HDRLEN; 634 } 635 636 *p_nhoff += offset; 637 key_control->flags |= FLOW_DIS_ENCAPSULATION; 638 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) 639 return FLOW_DISSECT_RET_OUT_GOOD; 640 641 return FLOW_DISSECT_RET_PROTO_AGAIN; 642 } 643 644 /** 645 * __skb_flow_dissect_batadv() - dissect batman-adv header 646 * @skb: sk_buff to with the batman-adv header 647 * @key_control: flow dissectors control key 648 * @data: raw buffer pointer to the packet, if NULL use skb->data 649 * @p_proto: pointer used to update the protocol to process next 650 * @p_nhoff: pointer used to update inner network header offset 651 * @hlen: packet header length 652 * @flags: any combination of FLOW_DISSECTOR_F_* 653 * 654 * ETH_P_BATMAN packets are tried to be dissected. Only 655 * &struct batadv_unicast packets are actually processed because they contain an 656 * inner ethernet header and are usually followed by actual network header. This 657 * allows the flow dissector to continue processing the packet. 658 * 659 * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found, 660 * FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation, 661 * otherwise FLOW_DISSECT_RET_OUT_BAD 662 */ 663 static enum flow_dissect_ret 664 __skb_flow_dissect_batadv(const struct sk_buff *skb, 665 struct flow_dissector_key_control *key_control, 666 void *data, __be16 *p_proto, int *p_nhoff, int hlen, 667 unsigned int flags) 668 { 669 struct { 670 struct batadv_unicast_packet batadv_unicast; 671 struct ethhdr eth; 672 } *hdr, _hdr; 673 674 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen, 675 &_hdr); 676 if (!hdr) 677 return FLOW_DISSECT_RET_OUT_BAD; 678 679 if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION) 680 return FLOW_DISSECT_RET_OUT_BAD; 681 682 if (hdr->batadv_unicast.packet_type != BATADV_UNICAST) 683 return FLOW_DISSECT_RET_OUT_BAD; 684 685 *p_proto = hdr->eth.h_proto; 686 *p_nhoff += sizeof(*hdr); 687 688 key_control->flags |= FLOW_DIS_ENCAPSULATION; 689 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) 690 return FLOW_DISSECT_RET_OUT_GOOD; 691 692 return FLOW_DISSECT_RET_PROTO_AGAIN; 693 } 694 695 static void 696 __skb_flow_dissect_tcp(const struct sk_buff *skb, 697 struct flow_dissector *flow_dissector, 698 void *target_container, void *data, int thoff, int hlen) 699 { 700 struct flow_dissector_key_tcp *key_tcp; 701 struct tcphdr *th, _th; 702 703 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP)) 704 return; 705 706 th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th); 707 if (!th) 708 return; 709 710 if (unlikely(__tcp_hdrlen(th) < sizeof(_th))) 711 return; 712 713 key_tcp = skb_flow_dissector_target(flow_dissector, 714 FLOW_DISSECTOR_KEY_TCP, 715 target_container); 716 key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF)); 717 } 718 719 static void 720 __skb_flow_dissect_ports(const struct sk_buff *skb, 721 struct flow_dissector *flow_dissector, 722 void *target_container, void *data, int nhoff, 723 u8 ip_proto, int hlen) 724 { 725 enum flow_dissector_key_id dissector_ports = FLOW_DISSECTOR_KEY_MAX; 726 struct flow_dissector_key_ports *key_ports; 727 728 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS)) 729 dissector_ports = FLOW_DISSECTOR_KEY_PORTS; 730 else if (dissector_uses_key(flow_dissector, 731 FLOW_DISSECTOR_KEY_PORTS_RANGE)) 732 dissector_ports = FLOW_DISSECTOR_KEY_PORTS_RANGE; 733 734 if (dissector_ports == FLOW_DISSECTOR_KEY_MAX) 735 return; 736 737 key_ports = skb_flow_dissector_target(flow_dissector, 738 dissector_ports, 739 target_container); 740 key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto, 741 data, hlen); 742 } 743 744 static void 745 __skb_flow_dissect_ipv4(const struct sk_buff *skb, 746 struct flow_dissector *flow_dissector, 747 void *target_container, void *data, const struct iphdr *iph) 748 { 749 struct flow_dissector_key_ip *key_ip; 750 751 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP)) 752 return; 753 754 key_ip = skb_flow_dissector_target(flow_dissector, 755 FLOW_DISSECTOR_KEY_IP, 756 target_container); 757 key_ip->tos = iph->tos; 758 key_ip->ttl = iph->ttl; 759 } 760 761 static void 762 __skb_flow_dissect_ipv6(const struct sk_buff *skb, 763 struct flow_dissector *flow_dissector, 764 void *target_container, void *data, const struct ipv6hdr *iph) 765 { 766 struct flow_dissector_key_ip *key_ip; 767 768 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP)) 769 return; 770 771 key_ip = skb_flow_dissector_target(flow_dissector, 772 FLOW_DISSECTOR_KEY_IP, 773 target_container); 774 key_ip->tos = ipv6_get_dsfield(iph); 775 key_ip->ttl = iph->hop_limit; 776 } 777 778 /* Maximum number of protocol headers that can be parsed in 779 * __skb_flow_dissect 780 */ 781 #define MAX_FLOW_DISSECT_HDRS 15 782 783 static bool skb_flow_dissect_allowed(int *num_hdrs) 784 { 785 ++*num_hdrs; 786 787 return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS); 788 } 789 790 static void __skb_flow_bpf_to_target(const struct bpf_flow_keys *flow_keys, 791 struct flow_dissector *flow_dissector, 792 void *target_container) 793 { 794 struct flow_dissector_key_ports *key_ports = NULL; 795 struct flow_dissector_key_control *key_control; 796 struct flow_dissector_key_basic *key_basic; 797 struct flow_dissector_key_addrs *key_addrs; 798 struct flow_dissector_key_tags *key_tags; 799 800 key_control = skb_flow_dissector_target(flow_dissector, 801 FLOW_DISSECTOR_KEY_CONTROL, 802 target_container); 803 key_control->thoff = flow_keys->thoff; 804 if (flow_keys->is_frag) 805 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 806 if (flow_keys->is_first_frag) 807 key_control->flags |= FLOW_DIS_FIRST_FRAG; 808 if (flow_keys->is_encap) 809 key_control->flags |= FLOW_DIS_ENCAPSULATION; 810 811 key_basic = skb_flow_dissector_target(flow_dissector, 812 FLOW_DISSECTOR_KEY_BASIC, 813 target_container); 814 key_basic->n_proto = flow_keys->n_proto; 815 key_basic->ip_proto = flow_keys->ip_proto; 816 817 if (flow_keys->addr_proto == ETH_P_IP && 818 dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) { 819 key_addrs = skb_flow_dissector_target(flow_dissector, 820 FLOW_DISSECTOR_KEY_IPV4_ADDRS, 821 target_container); 822 key_addrs->v4addrs.src = flow_keys->ipv4_src; 823 key_addrs->v4addrs.dst = flow_keys->ipv4_dst; 824 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; 825 } else if (flow_keys->addr_proto == ETH_P_IPV6 && 826 dissector_uses_key(flow_dissector, 827 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) { 828 key_addrs = skb_flow_dissector_target(flow_dissector, 829 FLOW_DISSECTOR_KEY_IPV6_ADDRS, 830 target_container); 831 memcpy(&key_addrs->v6addrs, &flow_keys->ipv6_src, 832 sizeof(key_addrs->v6addrs)); 833 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 834 } 835 836 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS)) 837 key_ports = skb_flow_dissector_target(flow_dissector, 838 FLOW_DISSECTOR_KEY_PORTS, 839 target_container); 840 else if (dissector_uses_key(flow_dissector, 841 FLOW_DISSECTOR_KEY_PORTS_RANGE)) 842 key_ports = skb_flow_dissector_target(flow_dissector, 843 FLOW_DISSECTOR_KEY_PORTS_RANGE, 844 target_container); 845 846 if (key_ports) { 847 key_ports->src = flow_keys->sport; 848 key_ports->dst = flow_keys->dport; 849 } 850 851 if (dissector_uses_key(flow_dissector, 852 FLOW_DISSECTOR_KEY_FLOW_LABEL)) { 853 key_tags = skb_flow_dissector_target(flow_dissector, 854 FLOW_DISSECTOR_KEY_FLOW_LABEL, 855 target_container); 856 key_tags->flow_label = ntohl(flow_keys->flow_label); 857 } 858 } 859 860 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx, 861 __be16 proto, int nhoff, int hlen, unsigned int flags) 862 { 863 struct bpf_flow_keys *flow_keys = ctx->flow_keys; 864 u32 result; 865 866 /* Pass parameters to the BPF program */ 867 memset(flow_keys, 0, sizeof(*flow_keys)); 868 flow_keys->n_proto = proto; 869 flow_keys->nhoff = nhoff; 870 flow_keys->thoff = flow_keys->nhoff; 871 872 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG != 873 (int)FLOW_DISSECTOR_F_PARSE_1ST_FRAG); 874 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL != 875 (int)FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 876 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP != 877 (int)FLOW_DISSECTOR_F_STOP_AT_ENCAP); 878 flow_keys->flags = flags; 879 880 result = bpf_prog_run_pin_on_cpu(prog, ctx); 881 882 flow_keys->nhoff = clamp_t(u16, flow_keys->nhoff, nhoff, hlen); 883 flow_keys->thoff = clamp_t(u16, flow_keys->thoff, 884 flow_keys->nhoff, hlen); 885 886 return result == BPF_OK; 887 } 888 889 /** 890 * __skb_flow_dissect - extract the flow_keys struct and return it 891 * @net: associated network namespace, derived from @skb if NULL 892 * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified 893 * @flow_dissector: list of keys to dissect 894 * @target_container: target structure to put dissected values into 895 * @data: raw buffer pointer to the packet, if NULL use skb->data 896 * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol 897 * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb) 898 * @hlen: packet header length, if @data is NULL use skb_headlen(skb) 899 * @flags: flags that control the dissection process, e.g. 900 * FLOW_DISSECTOR_F_STOP_AT_ENCAP. 901 * 902 * The function will try to retrieve individual keys into target specified 903 * by flow_dissector from either the skbuff or a raw buffer specified by the 904 * rest parameters. 905 * 906 * Caller must take care of zeroing target container memory. 907 */ 908 bool __skb_flow_dissect(const struct net *net, 909 const struct sk_buff *skb, 910 struct flow_dissector *flow_dissector, 911 void *target_container, 912 void *data, __be16 proto, int nhoff, int hlen, 913 unsigned int flags) 914 { 915 struct flow_dissector_key_control *key_control; 916 struct flow_dissector_key_basic *key_basic; 917 struct flow_dissector_key_addrs *key_addrs; 918 struct flow_dissector_key_tags *key_tags; 919 struct flow_dissector_key_vlan *key_vlan; 920 enum flow_dissect_ret fdret; 921 enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX; 922 bool mpls_el = false; 923 int mpls_lse = 0; 924 int num_hdrs = 0; 925 u8 ip_proto = 0; 926 bool ret; 927 928 if (!data) { 929 data = skb->data; 930 proto = skb_vlan_tag_present(skb) ? 931 skb->vlan_proto : skb->protocol; 932 nhoff = skb_network_offset(skb); 933 hlen = skb_headlen(skb); 934 #if IS_ENABLED(CONFIG_NET_DSA) 935 if (unlikely(skb->dev && netdev_uses_dsa(skb->dev) && 936 proto == htons(ETH_P_XDSA))) { 937 const struct dsa_device_ops *ops; 938 int offset = 0; 939 940 ops = skb->dev->dsa_ptr->tag_ops; 941 /* Tail taggers don't break flow dissection */ 942 if (!ops->tail_tag) { 943 if (ops->flow_dissect) 944 ops->flow_dissect(skb, &proto, &offset); 945 else 946 dsa_tag_generic_flow_dissect(skb, 947 &proto, 948 &offset); 949 hlen -= offset; 950 nhoff += offset; 951 } 952 } 953 #endif 954 } 955 956 /* It is ensured by skb_flow_dissector_init() that control key will 957 * be always present. 958 */ 959 key_control = skb_flow_dissector_target(flow_dissector, 960 FLOW_DISSECTOR_KEY_CONTROL, 961 target_container); 962 963 /* It is ensured by skb_flow_dissector_init() that basic key will 964 * be always present. 965 */ 966 key_basic = skb_flow_dissector_target(flow_dissector, 967 FLOW_DISSECTOR_KEY_BASIC, 968 target_container); 969 970 if (skb) { 971 if (!net) { 972 if (skb->dev) 973 net = dev_net(skb->dev); 974 else if (skb->sk) 975 net = sock_net(skb->sk); 976 } 977 } 978 979 WARN_ON_ONCE(!net); 980 if (net) { 981 enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR; 982 struct bpf_prog_array *run_array; 983 984 rcu_read_lock(); 985 run_array = rcu_dereference(init_net.bpf.run_array[type]); 986 if (!run_array) 987 run_array = rcu_dereference(net->bpf.run_array[type]); 988 989 if (run_array) { 990 struct bpf_flow_keys flow_keys; 991 struct bpf_flow_dissector ctx = { 992 .flow_keys = &flow_keys, 993 .data = data, 994 .data_end = data + hlen, 995 }; 996 __be16 n_proto = proto; 997 struct bpf_prog *prog; 998 999 if (skb) { 1000 ctx.skb = skb; 1001 /* we can't use 'proto' in the skb case 1002 * because it might be set to skb->vlan_proto 1003 * which has been pulled from the data 1004 */ 1005 n_proto = skb->protocol; 1006 } 1007 1008 prog = READ_ONCE(run_array->items[0].prog); 1009 ret = bpf_flow_dissect(prog, &ctx, n_proto, nhoff, 1010 hlen, flags); 1011 __skb_flow_bpf_to_target(&flow_keys, flow_dissector, 1012 target_container); 1013 rcu_read_unlock(); 1014 return ret; 1015 } 1016 rcu_read_unlock(); 1017 } 1018 1019 if (dissector_uses_key(flow_dissector, 1020 FLOW_DISSECTOR_KEY_ETH_ADDRS)) { 1021 struct ethhdr *eth = eth_hdr(skb); 1022 struct flow_dissector_key_eth_addrs *key_eth_addrs; 1023 1024 key_eth_addrs = skb_flow_dissector_target(flow_dissector, 1025 FLOW_DISSECTOR_KEY_ETH_ADDRS, 1026 target_container); 1027 memcpy(key_eth_addrs, ð->h_dest, sizeof(*key_eth_addrs)); 1028 } 1029 1030 proto_again: 1031 fdret = FLOW_DISSECT_RET_CONTINUE; 1032 1033 switch (proto) { 1034 case htons(ETH_P_IP): { 1035 const struct iphdr *iph; 1036 struct iphdr _iph; 1037 1038 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph); 1039 if (!iph || iph->ihl < 5) { 1040 fdret = FLOW_DISSECT_RET_OUT_BAD; 1041 break; 1042 } 1043 1044 nhoff += iph->ihl * 4; 1045 1046 ip_proto = iph->protocol; 1047 1048 if (dissector_uses_key(flow_dissector, 1049 FLOW_DISSECTOR_KEY_IPV4_ADDRS)) { 1050 key_addrs = skb_flow_dissector_target(flow_dissector, 1051 FLOW_DISSECTOR_KEY_IPV4_ADDRS, 1052 target_container); 1053 1054 memcpy(&key_addrs->v4addrs, &iph->saddr, 1055 sizeof(key_addrs->v4addrs)); 1056 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; 1057 } 1058 1059 __skb_flow_dissect_ipv4(skb, flow_dissector, 1060 target_container, data, iph); 1061 1062 if (ip_is_fragment(iph)) { 1063 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 1064 1065 if (iph->frag_off & htons(IP_OFFSET)) { 1066 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1067 break; 1068 } else { 1069 key_control->flags |= FLOW_DIS_FIRST_FRAG; 1070 if (!(flags & 1071 FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) { 1072 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1073 break; 1074 } 1075 } 1076 } 1077 1078 break; 1079 } 1080 case htons(ETH_P_IPV6): { 1081 const struct ipv6hdr *iph; 1082 struct ipv6hdr _iph; 1083 1084 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph); 1085 if (!iph) { 1086 fdret = FLOW_DISSECT_RET_OUT_BAD; 1087 break; 1088 } 1089 1090 ip_proto = iph->nexthdr; 1091 nhoff += sizeof(struct ipv6hdr); 1092 1093 if (dissector_uses_key(flow_dissector, 1094 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) { 1095 key_addrs = skb_flow_dissector_target(flow_dissector, 1096 FLOW_DISSECTOR_KEY_IPV6_ADDRS, 1097 target_container); 1098 1099 memcpy(&key_addrs->v6addrs, &iph->saddr, 1100 sizeof(key_addrs->v6addrs)); 1101 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 1102 } 1103 1104 if ((dissector_uses_key(flow_dissector, 1105 FLOW_DISSECTOR_KEY_FLOW_LABEL) || 1106 (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) && 1107 ip6_flowlabel(iph)) { 1108 __be32 flow_label = ip6_flowlabel(iph); 1109 1110 if (dissector_uses_key(flow_dissector, 1111 FLOW_DISSECTOR_KEY_FLOW_LABEL)) { 1112 key_tags = skb_flow_dissector_target(flow_dissector, 1113 FLOW_DISSECTOR_KEY_FLOW_LABEL, 1114 target_container); 1115 key_tags->flow_label = ntohl(flow_label); 1116 } 1117 if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) { 1118 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1119 break; 1120 } 1121 } 1122 1123 __skb_flow_dissect_ipv6(skb, flow_dissector, 1124 target_container, data, iph); 1125 1126 break; 1127 } 1128 case htons(ETH_P_8021AD): 1129 case htons(ETH_P_8021Q): { 1130 const struct vlan_hdr *vlan = NULL; 1131 struct vlan_hdr _vlan; 1132 __be16 saved_vlan_tpid = proto; 1133 1134 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX && 1135 skb && skb_vlan_tag_present(skb)) { 1136 proto = skb->protocol; 1137 } else { 1138 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan), 1139 data, hlen, &_vlan); 1140 if (!vlan) { 1141 fdret = FLOW_DISSECT_RET_OUT_BAD; 1142 break; 1143 } 1144 1145 proto = vlan->h_vlan_encapsulated_proto; 1146 nhoff += sizeof(*vlan); 1147 } 1148 1149 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) { 1150 dissector_vlan = FLOW_DISSECTOR_KEY_VLAN; 1151 } else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) { 1152 dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN; 1153 } else { 1154 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1155 break; 1156 } 1157 1158 if (dissector_uses_key(flow_dissector, dissector_vlan)) { 1159 key_vlan = skb_flow_dissector_target(flow_dissector, 1160 dissector_vlan, 1161 target_container); 1162 1163 if (!vlan) { 1164 key_vlan->vlan_id = skb_vlan_tag_get_id(skb); 1165 key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb); 1166 } else { 1167 key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) & 1168 VLAN_VID_MASK; 1169 key_vlan->vlan_priority = 1170 (ntohs(vlan->h_vlan_TCI) & 1171 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT; 1172 } 1173 key_vlan->vlan_tpid = saved_vlan_tpid; 1174 } 1175 1176 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1177 break; 1178 } 1179 case htons(ETH_P_PPP_SES): { 1180 struct { 1181 struct pppoe_hdr hdr; 1182 __be16 proto; 1183 } *hdr, _hdr; 1184 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr); 1185 if (!hdr) { 1186 fdret = FLOW_DISSECT_RET_OUT_BAD; 1187 break; 1188 } 1189 1190 proto = hdr->proto; 1191 nhoff += PPPOE_SES_HLEN; 1192 switch (proto) { 1193 case htons(PPP_IP): 1194 proto = htons(ETH_P_IP); 1195 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1196 break; 1197 case htons(PPP_IPV6): 1198 proto = htons(ETH_P_IPV6); 1199 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1200 break; 1201 default: 1202 fdret = FLOW_DISSECT_RET_OUT_BAD; 1203 break; 1204 } 1205 break; 1206 } 1207 case htons(ETH_P_TIPC): { 1208 struct tipc_basic_hdr *hdr, _hdr; 1209 1210 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), 1211 data, hlen, &_hdr); 1212 if (!hdr) { 1213 fdret = FLOW_DISSECT_RET_OUT_BAD; 1214 break; 1215 } 1216 1217 if (dissector_uses_key(flow_dissector, 1218 FLOW_DISSECTOR_KEY_TIPC)) { 1219 key_addrs = skb_flow_dissector_target(flow_dissector, 1220 FLOW_DISSECTOR_KEY_TIPC, 1221 target_container); 1222 key_addrs->tipckey.key = tipc_hdr_rps_key(hdr); 1223 key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC; 1224 } 1225 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1226 break; 1227 } 1228 1229 case htons(ETH_P_MPLS_UC): 1230 case htons(ETH_P_MPLS_MC): 1231 fdret = __skb_flow_dissect_mpls(skb, flow_dissector, 1232 target_container, data, 1233 nhoff, hlen, mpls_lse, 1234 &mpls_el); 1235 nhoff += sizeof(struct mpls_label); 1236 mpls_lse++; 1237 break; 1238 case htons(ETH_P_FCOE): 1239 if ((hlen - nhoff) < FCOE_HEADER_LEN) { 1240 fdret = FLOW_DISSECT_RET_OUT_BAD; 1241 break; 1242 } 1243 1244 nhoff += FCOE_HEADER_LEN; 1245 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1246 break; 1247 1248 case htons(ETH_P_ARP): 1249 case htons(ETH_P_RARP): 1250 fdret = __skb_flow_dissect_arp(skb, flow_dissector, 1251 target_container, data, 1252 nhoff, hlen); 1253 break; 1254 1255 case htons(ETH_P_BATMAN): 1256 fdret = __skb_flow_dissect_batadv(skb, key_control, data, 1257 &proto, &nhoff, hlen, flags); 1258 break; 1259 1260 case htons(ETH_P_1588): { 1261 struct ptp_header *hdr, _hdr; 1262 1263 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, 1264 hlen, &_hdr); 1265 if (!hdr) { 1266 fdret = FLOW_DISSECT_RET_OUT_BAD; 1267 break; 1268 } 1269 1270 nhoff += ntohs(hdr->message_length); 1271 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1272 break; 1273 } 1274 1275 default: 1276 fdret = FLOW_DISSECT_RET_OUT_BAD; 1277 break; 1278 } 1279 1280 /* Process result of proto processing */ 1281 switch (fdret) { 1282 case FLOW_DISSECT_RET_OUT_GOOD: 1283 goto out_good; 1284 case FLOW_DISSECT_RET_PROTO_AGAIN: 1285 if (skb_flow_dissect_allowed(&num_hdrs)) 1286 goto proto_again; 1287 goto out_good; 1288 case FLOW_DISSECT_RET_CONTINUE: 1289 case FLOW_DISSECT_RET_IPPROTO_AGAIN: 1290 break; 1291 case FLOW_DISSECT_RET_OUT_BAD: 1292 default: 1293 goto out_bad; 1294 } 1295 1296 ip_proto_again: 1297 fdret = FLOW_DISSECT_RET_CONTINUE; 1298 1299 switch (ip_proto) { 1300 case IPPROTO_GRE: 1301 fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector, 1302 target_container, data, 1303 &proto, &nhoff, &hlen, flags); 1304 break; 1305 1306 case NEXTHDR_HOP: 1307 case NEXTHDR_ROUTING: 1308 case NEXTHDR_DEST: { 1309 u8 _opthdr[2], *opthdr; 1310 1311 if (proto != htons(ETH_P_IPV6)) 1312 break; 1313 1314 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr), 1315 data, hlen, &_opthdr); 1316 if (!opthdr) { 1317 fdret = FLOW_DISSECT_RET_OUT_BAD; 1318 break; 1319 } 1320 1321 ip_proto = opthdr[0]; 1322 nhoff += (opthdr[1] + 1) << 3; 1323 1324 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN; 1325 break; 1326 } 1327 case NEXTHDR_FRAGMENT: { 1328 struct frag_hdr _fh, *fh; 1329 1330 if (proto != htons(ETH_P_IPV6)) 1331 break; 1332 1333 fh = __skb_header_pointer(skb, nhoff, sizeof(_fh), 1334 data, hlen, &_fh); 1335 1336 if (!fh) { 1337 fdret = FLOW_DISSECT_RET_OUT_BAD; 1338 break; 1339 } 1340 1341 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 1342 1343 nhoff += sizeof(_fh); 1344 ip_proto = fh->nexthdr; 1345 1346 if (!(fh->frag_off & htons(IP6_OFFSET))) { 1347 key_control->flags |= FLOW_DIS_FIRST_FRAG; 1348 if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) { 1349 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN; 1350 break; 1351 } 1352 } 1353 1354 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1355 break; 1356 } 1357 case IPPROTO_IPIP: 1358 proto = htons(ETH_P_IP); 1359 1360 key_control->flags |= FLOW_DIS_ENCAPSULATION; 1361 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) { 1362 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1363 break; 1364 } 1365 1366 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1367 break; 1368 1369 case IPPROTO_IPV6: 1370 proto = htons(ETH_P_IPV6); 1371 1372 key_control->flags |= FLOW_DIS_ENCAPSULATION; 1373 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) { 1374 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1375 break; 1376 } 1377 1378 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1379 break; 1380 1381 1382 case IPPROTO_MPLS: 1383 proto = htons(ETH_P_MPLS_UC); 1384 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1385 break; 1386 1387 case IPPROTO_TCP: 1388 __skb_flow_dissect_tcp(skb, flow_dissector, target_container, 1389 data, nhoff, hlen); 1390 break; 1391 1392 case IPPROTO_ICMP: 1393 case IPPROTO_ICMPV6: 1394 __skb_flow_dissect_icmp(skb, flow_dissector, target_container, 1395 data, nhoff, hlen); 1396 break; 1397 1398 default: 1399 break; 1400 } 1401 1402 if (!(key_control->flags & FLOW_DIS_IS_FRAGMENT)) 1403 __skb_flow_dissect_ports(skb, flow_dissector, target_container, 1404 data, nhoff, ip_proto, hlen); 1405 1406 /* Process result of IP proto processing */ 1407 switch (fdret) { 1408 case FLOW_DISSECT_RET_PROTO_AGAIN: 1409 if (skb_flow_dissect_allowed(&num_hdrs)) 1410 goto proto_again; 1411 break; 1412 case FLOW_DISSECT_RET_IPPROTO_AGAIN: 1413 if (skb_flow_dissect_allowed(&num_hdrs)) 1414 goto ip_proto_again; 1415 break; 1416 case FLOW_DISSECT_RET_OUT_GOOD: 1417 case FLOW_DISSECT_RET_CONTINUE: 1418 break; 1419 case FLOW_DISSECT_RET_OUT_BAD: 1420 default: 1421 goto out_bad; 1422 } 1423 1424 out_good: 1425 ret = true; 1426 1427 out: 1428 key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen); 1429 key_basic->n_proto = proto; 1430 key_basic->ip_proto = ip_proto; 1431 1432 return ret; 1433 1434 out_bad: 1435 ret = false; 1436 goto out; 1437 } 1438 EXPORT_SYMBOL(__skb_flow_dissect); 1439 1440 static siphash_key_t hashrnd __read_mostly; 1441 static __always_inline void __flow_hash_secret_init(void) 1442 { 1443 net_get_random_once(&hashrnd, sizeof(hashrnd)); 1444 } 1445 1446 static const void *flow_keys_hash_start(const struct flow_keys *flow) 1447 { 1448 BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % SIPHASH_ALIGNMENT); 1449 return &flow->FLOW_KEYS_HASH_START_FIELD; 1450 } 1451 1452 static inline size_t flow_keys_hash_length(const struct flow_keys *flow) 1453 { 1454 size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs); 1455 1456 BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32)); 1457 1458 switch (flow->control.addr_type) { 1459 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1460 diff -= sizeof(flow->addrs.v4addrs); 1461 break; 1462 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1463 diff -= sizeof(flow->addrs.v6addrs); 1464 break; 1465 case FLOW_DISSECTOR_KEY_TIPC: 1466 diff -= sizeof(flow->addrs.tipckey); 1467 break; 1468 } 1469 return sizeof(*flow) - diff; 1470 } 1471 1472 __be32 flow_get_u32_src(const struct flow_keys *flow) 1473 { 1474 switch (flow->control.addr_type) { 1475 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1476 return flow->addrs.v4addrs.src; 1477 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1478 return (__force __be32)ipv6_addr_hash( 1479 &flow->addrs.v6addrs.src); 1480 case FLOW_DISSECTOR_KEY_TIPC: 1481 return flow->addrs.tipckey.key; 1482 default: 1483 return 0; 1484 } 1485 } 1486 EXPORT_SYMBOL(flow_get_u32_src); 1487 1488 __be32 flow_get_u32_dst(const struct flow_keys *flow) 1489 { 1490 switch (flow->control.addr_type) { 1491 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1492 return flow->addrs.v4addrs.dst; 1493 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1494 return (__force __be32)ipv6_addr_hash( 1495 &flow->addrs.v6addrs.dst); 1496 default: 1497 return 0; 1498 } 1499 } 1500 EXPORT_SYMBOL(flow_get_u32_dst); 1501 1502 /* Sort the source and destination IP (and the ports if the IP are the same), 1503 * to have consistent hash within the two directions 1504 */ 1505 static inline void __flow_hash_consistentify(struct flow_keys *keys) 1506 { 1507 int addr_diff, i; 1508 1509 switch (keys->control.addr_type) { 1510 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1511 addr_diff = (__force u32)keys->addrs.v4addrs.dst - 1512 (__force u32)keys->addrs.v4addrs.src; 1513 if ((addr_diff < 0) || 1514 (addr_diff == 0 && 1515 ((__force u16)keys->ports.dst < 1516 (__force u16)keys->ports.src))) { 1517 swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst); 1518 swap(keys->ports.src, keys->ports.dst); 1519 } 1520 break; 1521 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1522 addr_diff = memcmp(&keys->addrs.v6addrs.dst, 1523 &keys->addrs.v6addrs.src, 1524 sizeof(keys->addrs.v6addrs.dst)); 1525 if ((addr_diff < 0) || 1526 (addr_diff == 0 && 1527 ((__force u16)keys->ports.dst < 1528 (__force u16)keys->ports.src))) { 1529 for (i = 0; i < 4; i++) 1530 swap(keys->addrs.v6addrs.src.s6_addr32[i], 1531 keys->addrs.v6addrs.dst.s6_addr32[i]); 1532 swap(keys->ports.src, keys->ports.dst); 1533 } 1534 break; 1535 } 1536 } 1537 1538 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, 1539 const siphash_key_t *keyval) 1540 { 1541 u32 hash; 1542 1543 __flow_hash_consistentify(keys); 1544 1545 hash = siphash(flow_keys_hash_start(keys), 1546 flow_keys_hash_length(keys), keyval); 1547 if (!hash) 1548 hash = 1; 1549 1550 return hash; 1551 } 1552 1553 u32 flow_hash_from_keys(struct flow_keys *keys) 1554 { 1555 __flow_hash_secret_init(); 1556 return __flow_hash_from_keys(keys, &hashrnd); 1557 } 1558 EXPORT_SYMBOL(flow_hash_from_keys); 1559 1560 static inline u32 ___skb_get_hash(const struct sk_buff *skb, 1561 struct flow_keys *keys, 1562 const siphash_key_t *keyval) 1563 { 1564 skb_flow_dissect_flow_keys(skb, keys, 1565 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 1566 1567 return __flow_hash_from_keys(keys, keyval); 1568 } 1569 1570 struct _flow_keys_digest_data { 1571 __be16 n_proto; 1572 u8 ip_proto; 1573 u8 padding; 1574 __be32 ports; 1575 __be32 src; 1576 __be32 dst; 1577 }; 1578 1579 void make_flow_keys_digest(struct flow_keys_digest *digest, 1580 const struct flow_keys *flow) 1581 { 1582 struct _flow_keys_digest_data *data = 1583 (struct _flow_keys_digest_data *)digest; 1584 1585 BUILD_BUG_ON(sizeof(*data) > sizeof(*digest)); 1586 1587 memset(digest, 0, sizeof(*digest)); 1588 1589 data->n_proto = flow->basic.n_proto; 1590 data->ip_proto = flow->basic.ip_proto; 1591 data->ports = flow->ports.ports; 1592 data->src = flow->addrs.v4addrs.src; 1593 data->dst = flow->addrs.v4addrs.dst; 1594 } 1595 EXPORT_SYMBOL(make_flow_keys_digest); 1596 1597 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly; 1598 1599 u32 __skb_get_hash_symmetric(const struct sk_buff *skb) 1600 { 1601 struct flow_keys keys; 1602 1603 __flow_hash_secret_init(); 1604 1605 memset(&keys, 0, sizeof(keys)); 1606 __skb_flow_dissect(NULL, skb, &flow_keys_dissector_symmetric, 1607 &keys, NULL, 0, 0, 0, 1608 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 1609 1610 return __flow_hash_from_keys(&keys, &hashrnd); 1611 } 1612 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric); 1613 1614 /** 1615 * __skb_get_hash: calculate a flow hash 1616 * @skb: sk_buff to calculate flow hash from 1617 * 1618 * This function calculates a flow hash based on src/dst addresses 1619 * and src/dst port numbers. Sets hash in skb to non-zero hash value 1620 * on success, zero indicates no valid hash. Also, sets l4_hash in skb 1621 * if hash is a canonical 4-tuple hash over transport ports. 1622 */ 1623 void __skb_get_hash(struct sk_buff *skb) 1624 { 1625 struct flow_keys keys; 1626 u32 hash; 1627 1628 __flow_hash_secret_init(); 1629 1630 hash = ___skb_get_hash(skb, &keys, &hashrnd); 1631 1632 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); 1633 } 1634 EXPORT_SYMBOL(__skb_get_hash); 1635 1636 __u32 skb_get_hash_perturb(const struct sk_buff *skb, 1637 const siphash_key_t *perturb) 1638 { 1639 struct flow_keys keys; 1640 1641 return ___skb_get_hash(skb, &keys, perturb); 1642 } 1643 EXPORT_SYMBOL(skb_get_hash_perturb); 1644 1645 u32 __skb_get_poff(const struct sk_buff *skb, void *data, 1646 const struct flow_keys_basic *keys, int hlen) 1647 { 1648 u32 poff = keys->control.thoff; 1649 1650 /* skip L4 headers for fragments after the first */ 1651 if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) && 1652 !(keys->control.flags & FLOW_DIS_FIRST_FRAG)) 1653 return poff; 1654 1655 switch (keys->basic.ip_proto) { 1656 case IPPROTO_TCP: { 1657 /* access doff as u8 to avoid unaligned access */ 1658 const u8 *doff; 1659 u8 _doff; 1660 1661 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff), 1662 data, hlen, &_doff); 1663 if (!doff) 1664 return poff; 1665 1666 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2); 1667 break; 1668 } 1669 case IPPROTO_UDP: 1670 case IPPROTO_UDPLITE: 1671 poff += sizeof(struct udphdr); 1672 break; 1673 /* For the rest, we do not really care about header 1674 * extensions at this point for now. 1675 */ 1676 case IPPROTO_ICMP: 1677 poff += sizeof(struct icmphdr); 1678 break; 1679 case IPPROTO_ICMPV6: 1680 poff += sizeof(struct icmp6hdr); 1681 break; 1682 case IPPROTO_IGMP: 1683 poff += sizeof(struct igmphdr); 1684 break; 1685 case IPPROTO_DCCP: 1686 poff += sizeof(struct dccp_hdr); 1687 break; 1688 case IPPROTO_SCTP: 1689 poff += sizeof(struct sctphdr); 1690 break; 1691 } 1692 1693 return poff; 1694 } 1695 1696 /** 1697 * skb_get_poff - get the offset to the payload 1698 * @skb: sk_buff to get the payload offset from 1699 * 1700 * The function will get the offset to the payload as far as it could 1701 * be dissected. The main user is currently BPF, so that we can dynamically 1702 * truncate packets without needing to push actual payload to the user 1703 * space and can analyze headers only, instead. 1704 */ 1705 u32 skb_get_poff(const struct sk_buff *skb) 1706 { 1707 struct flow_keys_basic keys; 1708 1709 if (!skb_flow_dissect_flow_keys_basic(NULL, skb, &keys, 1710 NULL, 0, 0, 0, 0)) 1711 return 0; 1712 1713 return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb)); 1714 } 1715 1716 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys) 1717 { 1718 memset(keys, 0, sizeof(*keys)); 1719 1720 memcpy(&keys->addrs.v6addrs.src, &fl6->saddr, 1721 sizeof(keys->addrs.v6addrs.src)); 1722 memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr, 1723 sizeof(keys->addrs.v6addrs.dst)); 1724 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 1725 keys->ports.src = fl6->fl6_sport; 1726 keys->ports.dst = fl6->fl6_dport; 1727 keys->keyid.keyid = fl6->fl6_gre_key; 1728 keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6); 1729 keys->basic.ip_proto = fl6->flowi6_proto; 1730 1731 return flow_hash_from_keys(keys); 1732 } 1733 EXPORT_SYMBOL(__get_hash_from_flowi6); 1734 1735 static const struct flow_dissector_key flow_keys_dissector_keys[] = { 1736 { 1737 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 1738 .offset = offsetof(struct flow_keys, control), 1739 }, 1740 { 1741 .key_id = FLOW_DISSECTOR_KEY_BASIC, 1742 .offset = offsetof(struct flow_keys, basic), 1743 }, 1744 { 1745 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS, 1746 .offset = offsetof(struct flow_keys, addrs.v4addrs), 1747 }, 1748 { 1749 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS, 1750 .offset = offsetof(struct flow_keys, addrs.v6addrs), 1751 }, 1752 { 1753 .key_id = FLOW_DISSECTOR_KEY_TIPC, 1754 .offset = offsetof(struct flow_keys, addrs.tipckey), 1755 }, 1756 { 1757 .key_id = FLOW_DISSECTOR_KEY_PORTS, 1758 .offset = offsetof(struct flow_keys, ports), 1759 }, 1760 { 1761 .key_id = FLOW_DISSECTOR_KEY_VLAN, 1762 .offset = offsetof(struct flow_keys, vlan), 1763 }, 1764 { 1765 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL, 1766 .offset = offsetof(struct flow_keys, tags), 1767 }, 1768 { 1769 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID, 1770 .offset = offsetof(struct flow_keys, keyid), 1771 }, 1772 }; 1773 1774 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = { 1775 { 1776 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 1777 .offset = offsetof(struct flow_keys, control), 1778 }, 1779 { 1780 .key_id = FLOW_DISSECTOR_KEY_BASIC, 1781 .offset = offsetof(struct flow_keys, basic), 1782 }, 1783 { 1784 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS, 1785 .offset = offsetof(struct flow_keys, addrs.v4addrs), 1786 }, 1787 { 1788 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS, 1789 .offset = offsetof(struct flow_keys, addrs.v6addrs), 1790 }, 1791 { 1792 .key_id = FLOW_DISSECTOR_KEY_PORTS, 1793 .offset = offsetof(struct flow_keys, ports), 1794 }, 1795 }; 1796 1797 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = { 1798 { 1799 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 1800 .offset = offsetof(struct flow_keys, control), 1801 }, 1802 { 1803 .key_id = FLOW_DISSECTOR_KEY_BASIC, 1804 .offset = offsetof(struct flow_keys, basic), 1805 }, 1806 }; 1807 1808 struct flow_dissector flow_keys_dissector __read_mostly; 1809 EXPORT_SYMBOL(flow_keys_dissector); 1810 1811 struct flow_dissector flow_keys_basic_dissector __read_mostly; 1812 EXPORT_SYMBOL(flow_keys_basic_dissector); 1813 1814 static int __init init_default_flow_dissectors(void) 1815 { 1816 skb_flow_dissector_init(&flow_keys_dissector, 1817 flow_keys_dissector_keys, 1818 ARRAY_SIZE(flow_keys_dissector_keys)); 1819 skb_flow_dissector_init(&flow_keys_dissector_symmetric, 1820 flow_keys_dissector_symmetric_keys, 1821 ARRAY_SIZE(flow_keys_dissector_symmetric_keys)); 1822 skb_flow_dissector_init(&flow_keys_basic_dissector, 1823 flow_keys_basic_dissector_keys, 1824 ARRAY_SIZE(flow_keys_basic_dissector_keys)); 1825 return 0; 1826 } 1827 core_initcall(init_default_flow_dissectors); 1828