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