1 #include <linux/kernel.h> 2 #include <linux/skbuff.h> 3 #include <linux/export.h> 4 #include <linux/ip.h> 5 #include <linux/ipv6.h> 6 #include <linux/if_vlan.h> 7 #include <net/dsa.h> 8 #include <net/dst_metadata.h> 9 #include <net/ip.h> 10 #include <net/ipv6.h> 11 #include <net/gre.h> 12 #include <net/pptp.h> 13 #include <net/tipc.h> 14 #include <linux/igmp.h> 15 #include <linux/icmp.h> 16 #include <linux/sctp.h> 17 #include <linux/dccp.h> 18 #include <linux/if_tunnel.h> 19 #include <linux/if_pppox.h> 20 #include <linux/ppp_defs.h> 21 #include <linux/stddef.h> 22 #include <linux/if_ether.h> 23 #include <linux/mpls.h> 24 #include <linux/tcp.h> 25 #include <net/flow_dissector.h> 26 #include <scsi/fc/fc_fcoe.h> 27 #include <uapi/linux/batadv_packet.h> 28 #include <linux/bpf.h> 29 30 static DEFINE_MUTEX(flow_dissector_mutex); 31 32 static void dissector_set_key(struct flow_dissector *flow_dissector, 33 enum flow_dissector_key_id key_id) 34 { 35 flow_dissector->used_keys |= (1 << key_id); 36 } 37 38 void skb_flow_dissector_init(struct flow_dissector *flow_dissector, 39 const struct flow_dissector_key *key, 40 unsigned int key_count) 41 { 42 unsigned int i; 43 44 memset(flow_dissector, 0, sizeof(*flow_dissector)); 45 46 for (i = 0; i < key_count; i++, key++) { 47 /* User should make sure that every key target offset is withing 48 * boundaries of unsigned short. 49 */ 50 BUG_ON(key->offset > USHRT_MAX); 51 BUG_ON(dissector_uses_key(flow_dissector, 52 key->key_id)); 53 54 dissector_set_key(flow_dissector, key->key_id); 55 flow_dissector->offset[key->key_id] = key->offset; 56 } 57 58 /* Ensure that the dissector always includes control and basic key. 59 * That way we are able to avoid handling lack of these in fast path. 60 */ 61 BUG_ON(!dissector_uses_key(flow_dissector, 62 FLOW_DISSECTOR_KEY_CONTROL)); 63 BUG_ON(!dissector_uses_key(flow_dissector, 64 FLOW_DISSECTOR_KEY_BASIC)); 65 } 66 EXPORT_SYMBOL(skb_flow_dissector_init); 67 68 int skb_flow_dissector_prog_query(const union bpf_attr *attr, 69 union bpf_attr __user *uattr) 70 { 71 __u32 __user *prog_ids = u64_to_user_ptr(attr->query.prog_ids); 72 u32 prog_id, prog_cnt = 0, flags = 0; 73 struct bpf_prog *attached; 74 struct net *net; 75 76 if (attr->query.query_flags) 77 return -EINVAL; 78 79 net = get_net_ns_by_fd(attr->query.target_fd); 80 if (IS_ERR(net)) 81 return PTR_ERR(net); 82 83 rcu_read_lock(); 84 attached = rcu_dereference(net->flow_dissector_prog); 85 if (attached) { 86 prog_cnt = 1; 87 prog_id = attached->aux->id; 88 } 89 rcu_read_unlock(); 90 91 put_net(net); 92 93 if (copy_to_user(&uattr->query.attach_flags, &flags, sizeof(flags))) 94 return -EFAULT; 95 if (copy_to_user(&uattr->query.prog_cnt, &prog_cnt, sizeof(prog_cnt))) 96 return -EFAULT; 97 98 if (!attr->query.prog_cnt || !prog_ids || !prog_cnt) 99 return 0; 100 101 if (copy_to_user(prog_ids, &prog_id, sizeof(u32))) 102 return -EFAULT; 103 104 return 0; 105 } 106 107 int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr, 108 struct bpf_prog *prog) 109 { 110 struct bpf_prog *attached; 111 struct net *net; 112 113 net = current->nsproxy->net_ns; 114 mutex_lock(&flow_dissector_mutex); 115 attached = rcu_dereference_protected(net->flow_dissector_prog, 116 lockdep_is_held(&flow_dissector_mutex)); 117 if (attached) { 118 /* Only one BPF program can be attached at a time */ 119 mutex_unlock(&flow_dissector_mutex); 120 return -EEXIST; 121 } 122 rcu_assign_pointer(net->flow_dissector_prog, prog); 123 mutex_unlock(&flow_dissector_mutex); 124 return 0; 125 } 126 127 int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr) 128 { 129 struct bpf_prog *attached; 130 struct net *net; 131 132 net = current->nsproxy->net_ns; 133 mutex_lock(&flow_dissector_mutex); 134 attached = rcu_dereference_protected(net->flow_dissector_prog, 135 lockdep_is_held(&flow_dissector_mutex)); 136 if (!attached) { 137 mutex_unlock(&flow_dissector_mutex); 138 return -ENOENT; 139 } 140 bpf_prog_put(attached); 141 RCU_INIT_POINTER(net->flow_dissector_prog, NULL); 142 mutex_unlock(&flow_dissector_mutex); 143 return 0; 144 } 145 /** 146 * skb_flow_get_be16 - extract be16 entity 147 * @skb: sk_buff to extract from 148 * @poff: offset to extract at 149 * @data: raw buffer pointer to the packet 150 * @hlen: packet header length 151 * 152 * The function will try to retrieve a be32 entity at 153 * offset poff 154 */ 155 static __be16 skb_flow_get_be16(const struct sk_buff *skb, int poff, 156 void *data, int hlen) 157 { 158 __be16 *u, _u; 159 160 u = __skb_header_pointer(skb, poff, sizeof(_u), data, hlen, &_u); 161 if (u) 162 return *u; 163 164 return 0; 165 } 166 167 /** 168 * __skb_flow_get_ports - extract the upper layer ports and return them 169 * @skb: sk_buff to extract the ports from 170 * @thoff: transport header offset 171 * @ip_proto: protocol for which to get port offset 172 * @data: raw buffer pointer to the packet, if NULL use skb->data 173 * @hlen: packet header length, if @data is NULL use skb_headlen(skb) 174 * 175 * The function will try to retrieve the ports at offset thoff + poff where poff 176 * is the protocol port offset returned from proto_ports_offset 177 */ 178 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto, 179 void *data, int hlen) 180 { 181 int poff = proto_ports_offset(ip_proto); 182 183 if (!data) { 184 data = skb->data; 185 hlen = skb_headlen(skb); 186 } 187 188 if (poff >= 0) { 189 __be32 *ports, _ports; 190 191 ports = __skb_header_pointer(skb, thoff + poff, 192 sizeof(_ports), data, hlen, &_ports); 193 if (ports) 194 return *ports; 195 } 196 197 return 0; 198 } 199 EXPORT_SYMBOL(__skb_flow_get_ports); 200 201 static void 202 skb_flow_dissect_set_enc_addr_type(enum flow_dissector_key_id type, 203 struct flow_dissector *flow_dissector, 204 void *target_container) 205 { 206 struct flow_dissector_key_control *ctrl; 207 208 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_CONTROL)) 209 return; 210 211 ctrl = skb_flow_dissector_target(flow_dissector, 212 FLOW_DISSECTOR_KEY_ENC_CONTROL, 213 target_container); 214 ctrl->addr_type = type; 215 } 216 217 void 218 skb_flow_dissect_tunnel_info(const struct sk_buff *skb, 219 struct flow_dissector *flow_dissector, 220 void *target_container) 221 { 222 struct ip_tunnel_info *info; 223 struct ip_tunnel_key *key; 224 225 /* A quick check to see if there might be something to do. */ 226 if (!dissector_uses_key(flow_dissector, 227 FLOW_DISSECTOR_KEY_ENC_KEYID) && 228 !dissector_uses_key(flow_dissector, 229 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) && 230 !dissector_uses_key(flow_dissector, 231 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) && 232 !dissector_uses_key(flow_dissector, 233 FLOW_DISSECTOR_KEY_ENC_CONTROL) && 234 !dissector_uses_key(flow_dissector, 235 FLOW_DISSECTOR_KEY_ENC_PORTS) && 236 !dissector_uses_key(flow_dissector, 237 FLOW_DISSECTOR_KEY_ENC_IP) && 238 !dissector_uses_key(flow_dissector, 239 FLOW_DISSECTOR_KEY_ENC_OPTS)) 240 return; 241 242 info = skb_tunnel_info(skb); 243 if (!info) 244 return; 245 246 key = &info->key; 247 248 switch (ip_tunnel_info_af(info)) { 249 case AF_INET: 250 skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV4_ADDRS, 251 flow_dissector, 252 target_container); 253 if (dissector_uses_key(flow_dissector, 254 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) { 255 struct flow_dissector_key_ipv4_addrs *ipv4; 256 257 ipv4 = skb_flow_dissector_target(flow_dissector, 258 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS, 259 target_container); 260 ipv4->src = key->u.ipv4.src; 261 ipv4->dst = key->u.ipv4.dst; 262 } 263 break; 264 case AF_INET6: 265 skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV6_ADDRS, 266 flow_dissector, 267 target_container); 268 if (dissector_uses_key(flow_dissector, 269 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) { 270 struct flow_dissector_key_ipv6_addrs *ipv6; 271 272 ipv6 = skb_flow_dissector_target(flow_dissector, 273 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS, 274 target_container); 275 ipv6->src = key->u.ipv6.src; 276 ipv6->dst = key->u.ipv6.dst; 277 } 278 break; 279 } 280 281 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) { 282 struct flow_dissector_key_keyid *keyid; 283 284 keyid = skb_flow_dissector_target(flow_dissector, 285 FLOW_DISSECTOR_KEY_ENC_KEYID, 286 target_container); 287 keyid->keyid = tunnel_id_to_key32(key->tun_id); 288 } 289 290 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) { 291 struct flow_dissector_key_ports *tp; 292 293 tp = skb_flow_dissector_target(flow_dissector, 294 FLOW_DISSECTOR_KEY_ENC_PORTS, 295 target_container); 296 tp->src = key->tp_src; 297 tp->dst = key->tp_dst; 298 } 299 300 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_IP)) { 301 struct flow_dissector_key_ip *ip; 302 303 ip = skb_flow_dissector_target(flow_dissector, 304 FLOW_DISSECTOR_KEY_ENC_IP, 305 target_container); 306 ip->tos = key->tos; 307 ip->ttl = key->ttl; 308 } 309 310 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_OPTS)) { 311 struct flow_dissector_key_enc_opts *enc_opt; 312 313 enc_opt = skb_flow_dissector_target(flow_dissector, 314 FLOW_DISSECTOR_KEY_ENC_OPTS, 315 target_container); 316 317 if (info->options_len) { 318 enc_opt->len = info->options_len; 319 ip_tunnel_info_opts_get(enc_opt->data, info); 320 enc_opt->dst_opt_type = info->key.tun_flags & 321 TUNNEL_OPTIONS_PRESENT; 322 } 323 } 324 } 325 EXPORT_SYMBOL(skb_flow_dissect_tunnel_info); 326 327 static enum flow_dissect_ret 328 __skb_flow_dissect_mpls(const struct sk_buff *skb, 329 struct flow_dissector *flow_dissector, 330 void *target_container, void *data, int nhoff, int hlen) 331 { 332 struct flow_dissector_key_keyid *key_keyid; 333 struct mpls_label *hdr, _hdr[2]; 334 u32 entry, label; 335 336 if (!dissector_uses_key(flow_dissector, 337 FLOW_DISSECTOR_KEY_MPLS_ENTROPY) && 338 !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) 339 return FLOW_DISSECT_RET_OUT_GOOD; 340 341 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, 342 hlen, &_hdr); 343 if (!hdr) 344 return FLOW_DISSECT_RET_OUT_BAD; 345 346 entry = ntohl(hdr[0].entry); 347 label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT; 348 349 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) { 350 struct flow_dissector_key_mpls *key_mpls; 351 352 key_mpls = skb_flow_dissector_target(flow_dissector, 353 FLOW_DISSECTOR_KEY_MPLS, 354 target_container); 355 key_mpls->mpls_label = label; 356 key_mpls->mpls_ttl = (entry & MPLS_LS_TTL_MASK) 357 >> MPLS_LS_TTL_SHIFT; 358 key_mpls->mpls_tc = (entry & MPLS_LS_TC_MASK) 359 >> MPLS_LS_TC_SHIFT; 360 key_mpls->mpls_bos = (entry & MPLS_LS_S_MASK) 361 >> MPLS_LS_S_SHIFT; 362 } 363 364 if (label == MPLS_LABEL_ENTROPY) { 365 key_keyid = skb_flow_dissector_target(flow_dissector, 366 FLOW_DISSECTOR_KEY_MPLS_ENTROPY, 367 target_container); 368 key_keyid->keyid = hdr[1].entry & htonl(MPLS_LS_LABEL_MASK); 369 } 370 return FLOW_DISSECT_RET_OUT_GOOD; 371 } 372 373 static enum flow_dissect_ret 374 __skb_flow_dissect_arp(const struct sk_buff *skb, 375 struct flow_dissector *flow_dissector, 376 void *target_container, void *data, int nhoff, int hlen) 377 { 378 struct flow_dissector_key_arp *key_arp; 379 struct { 380 unsigned char ar_sha[ETH_ALEN]; 381 unsigned char ar_sip[4]; 382 unsigned char ar_tha[ETH_ALEN]; 383 unsigned char ar_tip[4]; 384 } *arp_eth, _arp_eth; 385 const struct arphdr *arp; 386 struct arphdr _arp; 387 388 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP)) 389 return FLOW_DISSECT_RET_OUT_GOOD; 390 391 arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data, 392 hlen, &_arp); 393 if (!arp) 394 return FLOW_DISSECT_RET_OUT_BAD; 395 396 if (arp->ar_hrd != htons(ARPHRD_ETHER) || 397 arp->ar_pro != htons(ETH_P_IP) || 398 arp->ar_hln != ETH_ALEN || 399 arp->ar_pln != 4 || 400 (arp->ar_op != htons(ARPOP_REPLY) && 401 arp->ar_op != htons(ARPOP_REQUEST))) 402 return FLOW_DISSECT_RET_OUT_BAD; 403 404 arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp), 405 sizeof(_arp_eth), data, 406 hlen, &_arp_eth); 407 if (!arp_eth) 408 return FLOW_DISSECT_RET_OUT_BAD; 409 410 key_arp = skb_flow_dissector_target(flow_dissector, 411 FLOW_DISSECTOR_KEY_ARP, 412 target_container); 413 414 memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip)); 415 memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip)); 416 417 /* Only store the lower byte of the opcode; 418 * this covers ARPOP_REPLY and ARPOP_REQUEST. 419 */ 420 key_arp->op = ntohs(arp->ar_op) & 0xff; 421 422 ether_addr_copy(key_arp->sha, arp_eth->ar_sha); 423 ether_addr_copy(key_arp->tha, arp_eth->ar_tha); 424 425 return FLOW_DISSECT_RET_OUT_GOOD; 426 } 427 428 static enum flow_dissect_ret 429 __skb_flow_dissect_gre(const struct sk_buff *skb, 430 struct flow_dissector_key_control *key_control, 431 struct flow_dissector *flow_dissector, 432 void *target_container, void *data, 433 __be16 *p_proto, int *p_nhoff, int *p_hlen, 434 unsigned int flags) 435 { 436 struct flow_dissector_key_keyid *key_keyid; 437 struct gre_base_hdr *hdr, _hdr; 438 int offset = 0; 439 u16 gre_ver; 440 441 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), 442 data, *p_hlen, &_hdr); 443 if (!hdr) 444 return FLOW_DISSECT_RET_OUT_BAD; 445 446 /* Only look inside GRE without routing */ 447 if (hdr->flags & GRE_ROUTING) 448 return FLOW_DISSECT_RET_OUT_GOOD; 449 450 /* Only look inside GRE for version 0 and 1 */ 451 gre_ver = ntohs(hdr->flags & GRE_VERSION); 452 if (gre_ver > 1) 453 return FLOW_DISSECT_RET_OUT_GOOD; 454 455 *p_proto = hdr->protocol; 456 if (gre_ver) { 457 /* Version1 must be PPTP, and check the flags */ 458 if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY))) 459 return FLOW_DISSECT_RET_OUT_GOOD; 460 } 461 462 offset += sizeof(struct gre_base_hdr); 463 464 if (hdr->flags & GRE_CSUM) 465 offset += FIELD_SIZEOF(struct gre_full_hdr, csum) + 466 FIELD_SIZEOF(struct gre_full_hdr, reserved1); 467 468 if (hdr->flags & GRE_KEY) { 469 const __be32 *keyid; 470 __be32 _keyid; 471 472 keyid = __skb_header_pointer(skb, *p_nhoff + offset, 473 sizeof(_keyid), 474 data, *p_hlen, &_keyid); 475 if (!keyid) 476 return FLOW_DISSECT_RET_OUT_BAD; 477 478 if (dissector_uses_key(flow_dissector, 479 FLOW_DISSECTOR_KEY_GRE_KEYID)) { 480 key_keyid = skb_flow_dissector_target(flow_dissector, 481 FLOW_DISSECTOR_KEY_GRE_KEYID, 482 target_container); 483 if (gre_ver == 0) 484 key_keyid->keyid = *keyid; 485 else 486 key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK; 487 } 488 offset += FIELD_SIZEOF(struct gre_full_hdr, key); 489 } 490 491 if (hdr->flags & GRE_SEQ) 492 offset += FIELD_SIZEOF(struct pptp_gre_header, seq); 493 494 if (gre_ver == 0) { 495 if (*p_proto == htons(ETH_P_TEB)) { 496 const struct ethhdr *eth; 497 struct ethhdr _eth; 498 499 eth = __skb_header_pointer(skb, *p_nhoff + offset, 500 sizeof(_eth), 501 data, *p_hlen, &_eth); 502 if (!eth) 503 return FLOW_DISSECT_RET_OUT_BAD; 504 *p_proto = eth->h_proto; 505 offset += sizeof(*eth); 506 507 /* Cap headers that we access via pointers at the 508 * end of the Ethernet header as our maximum alignment 509 * at that point is only 2 bytes. 510 */ 511 if (NET_IP_ALIGN) 512 *p_hlen = *p_nhoff + offset; 513 } 514 } else { /* version 1, must be PPTP */ 515 u8 _ppp_hdr[PPP_HDRLEN]; 516 u8 *ppp_hdr; 517 518 if (hdr->flags & GRE_ACK) 519 offset += FIELD_SIZEOF(struct pptp_gre_header, ack); 520 521 ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset, 522 sizeof(_ppp_hdr), 523 data, *p_hlen, _ppp_hdr); 524 if (!ppp_hdr) 525 return FLOW_DISSECT_RET_OUT_BAD; 526 527 switch (PPP_PROTOCOL(ppp_hdr)) { 528 case PPP_IP: 529 *p_proto = htons(ETH_P_IP); 530 break; 531 case PPP_IPV6: 532 *p_proto = htons(ETH_P_IPV6); 533 break; 534 default: 535 /* Could probably catch some more like MPLS */ 536 break; 537 } 538 539 offset += PPP_HDRLEN; 540 } 541 542 *p_nhoff += offset; 543 key_control->flags |= FLOW_DIS_ENCAPSULATION; 544 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) 545 return FLOW_DISSECT_RET_OUT_GOOD; 546 547 return FLOW_DISSECT_RET_PROTO_AGAIN; 548 } 549 550 /** 551 * __skb_flow_dissect_batadv() - dissect batman-adv header 552 * @skb: sk_buff to with the batman-adv header 553 * @key_control: flow dissectors control key 554 * @data: raw buffer pointer to the packet, if NULL use skb->data 555 * @p_proto: pointer used to update the protocol to process next 556 * @p_nhoff: pointer used to update inner network header offset 557 * @hlen: packet header length 558 * @flags: any combination of FLOW_DISSECTOR_F_* 559 * 560 * ETH_P_BATMAN packets are tried to be dissected. Only 561 * &struct batadv_unicast packets are actually processed because they contain an 562 * inner ethernet header and are usually followed by actual network header. This 563 * allows the flow dissector to continue processing the packet. 564 * 565 * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found, 566 * FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation, 567 * otherwise FLOW_DISSECT_RET_OUT_BAD 568 */ 569 static enum flow_dissect_ret 570 __skb_flow_dissect_batadv(const struct sk_buff *skb, 571 struct flow_dissector_key_control *key_control, 572 void *data, __be16 *p_proto, int *p_nhoff, int hlen, 573 unsigned int flags) 574 { 575 struct { 576 struct batadv_unicast_packet batadv_unicast; 577 struct ethhdr eth; 578 } *hdr, _hdr; 579 580 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen, 581 &_hdr); 582 if (!hdr) 583 return FLOW_DISSECT_RET_OUT_BAD; 584 585 if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION) 586 return FLOW_DISSECT_RET_OUT_BAD; 587 588 if (hdr->batadv_unicast.packet_type != BATADV_UNICAST) 589 return FLOW_DISSECT_RET_OUT_BAD; 590 591 *p_proto = hdr->eth.h_proto; 592 *p_nhoff += sizeof(*hdr); 593 594 key_control->flags |= FLOW_DIS_ENCAPSULATION; 595 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) 596 return FLOW_DISSECT_RET_OUT_GOOD; 597 598 return FLOW_DISSECT_RET_PROTO_AGAIN; 599 } 600 601 static void 602 __skb_flow_dissect_tcp(const struct sk_buff *skb, 603 struct flow_dissector *flow_dissector, 604 void *target_container, void *data, int thoff, int hlen) 605 { 606 struct flow_dissector_key_tcp *key_tcp; 607 struct tcphdr *th, _th; 608 609 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP)) 610 return; 611 612 th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th); 613 if (!th) 614 return; 615 616 if (unlikely(__tcp_hdrlen(th) < sizeof(_th))) 617 return; 618 619 key_tcp = skb_flow_dissector_target(flow_dissector, 620 FLOW_DISSECTOR_KEY_TCP, 621 target_container); 622 key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF)); 623 } 624 625 static void 626 __skb_flow_dissect_ipv4(const struct sk_buff *skb, 627 struct flow_dissector *flow_dissector, 628 void *target_container, void *data, const struct iphdr *iph) 629 { 630 struct flow_dissector_key_ip *key_ip; 631 632 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP)) 633 return; 634 635 key_ip = skb_flow_dissector_target(flow_dissector, 636 FLOW_DISSECTOR_KEY_IP, 637 target_container); 638 key_ip->tos = iph->tos; 639 key_ip->ttl = iph->ttl; 640 } 641 642 static void 643 __skb_flow_dissect_ipv6(const struct sk_buff *skb, 644 struct flow_dissector *flow_dissector, 645 void *target_container, void *data, const struct ipv6hdr *iph) 646 { 647 struct flow_dissector_key_ip *key_ip; 648 649 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP)) 650 return; 651 652 key_ip = skb_flow_dissector_target(flow_dissector, 653 FLOW_DISSECTOR_KEY_IP, 654 target_container); 655 key_ip->tos = ipv6_get_dsfield(iph); 656 key_ip->ttl = iph->hop_limit; 657 } 658 659 /* Maximum number of protocol headers that can be parsed in 660 * __skb_flow_dissect 661 */ 662 #define MAX_FLOW_DISSECT_HDRS 15 663 664 static bool skb_flow_dissect_allowed(int *num_hdrs) 665 { 666 ++*num_hdrs; 667 668 return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS); 669 } 670 671 static void __skb_flow_bpf_to_target(const struct bpf_flow_keys *flow_keys, 672 struct flow_dissector *flow_dissector, 673 void *target_container) 674 { 675 struct flow_dissector_key_control *key_control; 676 struct flow_dissector_key_basic *key_basic; 677 struct flow_dissector_key_addrs *key_addrs; 678 struct flow_dissector_key_ports *key_ports; 679 680 key_control = skb_flow_dissector_target(flow_dissector, 681 FLOW_DISSECTOR_KEY_CONTROL, 682 target_container); 683 key_control->thoff = flow_keys->thoff; 684 if (flow_keys->is_frag) 685 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 686 if (flow_keys->is_first_frag) 687 key_control->flags |= FLOW_DIS_FIRST_FRAG; 688 if (flow_keys->is_encap) 689 key_control->flags |= FLOW_DIS_ENCAPSULATION; 690 691 key_basic = skb_flow_dissector_target(flow_dissector, 692 FLOW_DISSECTOR_KEY_BASIC, 693 target_container); 694 key_basic->n_proto = flow_keys->n_proto; 695 key_basic->ip_proto = flow_keys->ip_proto; 696 697 if (flow_keys->addr_proto == ETH_P_IP && 698 dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) { 699 key_addrs = skb_flow_dissector_target(flow_dissector, 700 FLOW_DISSECTOR_KEY_IPV4_ADDRS, 701 target_container); 702 key_addrs->v4addrs.src = flow_keys->ipv4_src; 703 key_addrs->v4addrs.dst = flow_keys->ipv4_dst; 704 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; 705 } else if (flow_keys->addr_proto == ETH_P_IPV6 && 706 dissector_uses_key(flow_dissector, 707 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) { 708 key_addrs = skb_flow_dissector_target(flow_dissector, 709 FLOW_DISSECTOR_KEY_IPV6_ADDRS, 710 target_container); 711 memcpy(&key_addrs->v6addrs, &flow_keys->ipv6_src, 712 sizeof(key_addrs->v6addrs)); 713 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 714 } 715 716 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS)) { 717 key_ports = skb_flow_dissector_target(flow_dissector, 718 FLOW_DISSECTOR_KEY_PORTS, 719 target_container); 720 key_ports->src = flow_keys->sport; 721 key_ports->dst = flow_keys->dport; 722 } 723 } 724 725 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx, 726 __be16 proto, int nhoff, int hlen) 727 { 728 struct bpf_flow_keys *flow_keys = ctx->flow_keys; 729 u32 result; 730 731 /* Pass parameters to the BPF program */ 732 memset(flow_keys, 0, sizeof(*flow_keys)); 733 flow_keys->n_proto = proto; 734 flow_keys->nhoff = nhoff; 735 flow_keys->thoff = flow_keys->nhoff; 736 737 preempt_disable(); 738 result = BPF_PROG_RUN(prog, ctx); 739 preempt_enable(); 740 741 flow_keys->nhoff = clamp_t(u16, flow_keys->nhoff, nhoff, hlen); 742 flow_keys->thoff = clamp_t(u16, flow_keys->thoff, 743 flow_keys->nhoff, hlen); 744 745 return result == BPF_OK; 746 } 747 748 /** 749 * __skb_flow_dissect - extract the flow_keys struct and return it 750 * @net: associated network namespace, derived from @skb if NULL 751 * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified 752 * @flow_dissector: list of keys to dissect 753 * @target_container: target structure to put dissected values into 754 * @data: raw buffer pointer to the packet, if NULL use skb->data 755 * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol 756 * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb) 757 * @hlen: packet header length, if @data is NULL use skb_headlen(skb) 758 * @flags: flags that control the dissection process, e.g. 759 * FLOW_DISSECTOR_F_STOP_AT_L3. 760 * 761 * The function will try to retrieve individual keys into target specified 762 * by flow_dissector from either the skbuff or a raw buffer specified by the 763 * rest parameters. 764 * 765 * Caller must take care of zeroing target container memory. 766 */ 767 bool __skb_flow_dissect(const struct net *net, 768 const struct sk_buff *skb, 769 struct flow_dissector *flow_dissector, 770 void *target_container, 771 void *data, __be16 proto, int nhoff, int hlen, 772 unsigned int flags) 773 { 774 struct flow_dissector_key_control *key_control; 775 struct flow_dissector_key_basic *key_basic; 776 struct flow_dissector_key_addrs *key_addrs; 777 struct flow_dissector_key_ports *key_ports; 778 struct flow_dissector_key_icmp *key_icmp; 779 struct flow_dissector_key_tags *key_tags; 780 struct flow_dissector_key_vlan *key_vlan; 781 struct bpf_prog *attached = NULL; 782 enum flow_dissect_ret fdret; 783 enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX; 784 int num_hdrs = 0; 785 u8 ip_proto = 0; 786 bool ret; 787 788 if (!data) { 789 data = skb->data; 790 proto = skb_vlan_tag_present(skb) ? 791 skb->vlan_proto : skb->protocol; 792 nhoff = skb_network_offset(skb); 793 hlen = skb_headlen(skb); 794 #if IS_ENABLED(CONFIG_NET_DSA) 795 if (unlikely(skb->dev && netdev_uses_dsa(skb->dev))) { 796 const struct dsa_device_ops *ops; 797 int offset; 798 799 ops = skb->dev->dsa_ptr->tag_ops; 800 if (ops->flow_dissect && 801 !ops->flow_dissect(skb, &proto, &offset)) { 802 hlen -= offset; 803 nhoff += offset; 804 } 805 } 806 #endif 807 } 808 809 /* It is ensured by skb_flow_dissector_init() that control key will 810 * be always present. 811 */ 812 key_control = skb_flow_dissector_target(flow_dissector, 813 FLOW_DISSECTOR_KEY_CONTROL, 814 target_container); 815 816 /* It is ensured by skb_flow_dissector_init() that basic key will 817 * be always present. 818 */ 819 key_basic = skb_flow_dissector_target(flow_dissector, 820 FLOW_DISSECTOR_KEY_BASIC, 821 target_container); 822 823 if (skb) { 824 if (!net) { 825 if (skb->dev) 826 net = dev_net(skb->dev); 827 else if (skb->sk) 828 net = sock_net(skb->sk); 829 } 830 } 831 832 WARN_ON_ONCE(!net); 833 if (net) { 834 rcu_read_lock(); 835 attached = rcu_dereference(net->flow_dissector_prog); 836 837 if (attached) { 838 struct bpf_flow_keys flow_keys; 839 struct bpf_flow_dissector ctx = { 840 .flow_keys = &flow_keys, 841 .data = data, 842 .data_end = data + hlen, 843 }; 844 __be16 n_proto = proto; 845 846 if (skb) { 847 ctx.skb = skb; 848 /* we can't use 'proto' in the skb case 849 * because it might be set to skb->vlan_proto 850 * which has been pulled from the data 851 */ 852 n_proto = skb->protocol; 853 } 854 855 ret = bpf_flow_dissect(attached, &ctx, n_proto, nhoff, 856 hlen); 857 __skb_flow_bpf_to_target(&flow_keys, flow_dissector, 858 target_container); 859 rcu_read_unlock(); 860 return ret; 861 } 862 rcu_read_unlock(); 863 } 864 865 if (dissector_uses_key(flow_dissector, 866 FLOW_DISSECTOR_KEY_ETH_ADDRS)) { 867 struct ethhdr *eth = eth_hdr(skb); 868 struct flow_dissector_key_eth_addrs *key_eth_addrs; 869 870 key_eth_addrs = skb_flow_dissector_target(flow_dissector, 871 FLOW_DISSECTOR_KEY_ETH_ADDRS, 872 target_container); 873 memcpy(key_eth_addrs, ð->h_dest, sizeof(*key_eth_addrs)); 874 } 875 876 proto_again: 877 fdret = FLOW_DISSECT_RET_CONTINUE; 878 879 switch (proto) { 880 case htons(ETH_P_IP): { 881 const struct iphdr *iph; 882 struct iphdr _iph; 883 884 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph); 885 if (!iph || iph->ihl < 5) { 886 fdret = FLOW_DISSECT_RET_OUT_BAD; 887 break; 888 } 889 890 nhoff += iph->ihl * 4; 891 892 ip_proto = iph->protocol; 893 894 if (dissector_uses_key(flow_dissector, 895 FLOW_DISSECTOR_KEY_IPV4_ADDRS)) { 896 key_addrs = skb_flow_dissector_target(flow_dissector, 897 FLOW_DISSECTOR_KEY_IPV4_ADDRS, 898 target_container); 899 900 memcpy(&key_addrs->v4addrs, &iph->saddr, 901 sizeof(key_addrs->v4addrs)); 902 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; 903 } 904 905 if (ip_is_fragment(iph)) { 906 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 907 908 if (iph->frag_off & htons(IP_OFFSET)) { 909 fdret = FLOW_DISSECT_RET_OUT_GOOD; 910 break; 911 } else { 912 key_control->flags |= FLOW_DIS_FIRST_FRAG; 913 if (!(flags & 914 FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) { 915 fdret = FLOW_DISSECT_RET_OUT_GOOD; 916 break; 917 } 918 } 919 } 920 921 __skb_flow_dissect_ipv4(skb, flow_dissector, 922 target_container, data, iph); 923 924 if (flags & FLOW_DISSECTOR_F_STOP_AT_L3) { 925 fdret = FLOW_DISSECT_RET_OUT_GOOD; 926 break; 927 } 928 929 break; 930 } 931 case htons(ETH_P_IPV6): { 932 const struct ipv6hdr *iph; 933 struct ipv6hdr _iph; 934 935 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph); 936 if (!iph) { 937 fdret = FLOW_DISSECT_RET_OUT_BAD; 938 break; 939 } 940 941 ip_proto = iph->nexthdr; 942 nhoff += sizeof(struct ipv6hdr); 943 944 if (dissector_uses_key(flow_dissector, 945 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) { 946 key_addrs = skb_flow_dissector_target(flow_dissector, 947 FLOW_DISSECTOR_KEY_IPV6_ADDRS, 948 target_container); 949 950 memcpy(&key_addrs->v6addrs, &iph->saddr, 951 sizeof(key_addrs->v6addrs)); 952 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 953 } 954 955 if ((dissector_uses_key(flow_dissector, 956 FLOW_DISSECTOR_KEY_FLOW_LABEL) || 957 (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) && 958 ip6_flowlabel(iph)) { 959 __be32 flow_label = ip6_flowlabel(iph); 960 961 if (dissector_uses_key(flow_dissector, 962 FLOW_DISSECTOR_KEY_FLOW_LABEL)) { 963 key_tags = skb_flow_dissector_target(flow_dissector, 964 FLOW_DISSECTOR_KEY_FLOW_LABEL, 965 target_container); 966 key_tags->flow_label = ntohl(flow_label); 967 } 968 if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) { 969 fdret = FLOW_DISSECT_RET_OUT_GOOD; 970 break; 971 } 972 } 973 974 __skb_flow_dissect_ipv6(skb, flow_dissector, 975 target_container, data, iph); 976 977 if (flags & FLOW_DISSECTOR_F_STOP_AT_L3) 978 fdret = FLOW_DISSECT_RET_OUT_GOOD; 979 980 break; 981 } 982 case htons(ETH_P_8021AD): 983 case htons(ETH_P_8021Q): { 984 const struct vlan_hdr *vlan = NULL; 985 struct vlan_hdr _vlan; 986 __be16 saved_vlan_tpid = proto; 987 988 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX && 989 skb && skb_vlan_tag_present(skb)) { 990 proto = skb->protocol; 991 } else { 992 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan), 993 data, hlen, &_vlan); 994 if (!vlan) { 995 fdret = FLOW_DISSECT_RET_OUT_BAD; 996 break; 997 } 998 999 proto = vlan->h_vlan_encapsulated_proto; 1000 nhoff += sizeof(*vlan); 1001 } 1002 1003 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) { 1004 dissector_vlan = FLOW_DISSECTOR_KEY_VLAN; 1005 } else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) { 1006 dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN; 1007 } else { 1008 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1009 break; 1010 } 1011 1012 if (dissector_uses_key(flow_dissector, dissector_vlan)) { 1013 key_vlan = skb_flow_dissector_target(flow_dissector, 1014 dissector_vlan, 1015 target_container); 1016 1017 if (!vlan) { 1018 key_vlan->vlan_id = skb_vlan_tag_get_id(skb); 1019 key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb); 1020 } else { 1021 key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) & 1022 VLAN_VID_MASK; 1023 key_vlan->vlan_priority = 1024 (ntohs(vlan->h_vlan_TCI) & 1025 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT; 1026 } 1027 key_vlan->vlan_tpid = saved_vlan_tpid; 1028 } 1029 1030 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1031 break; 1032 } 1033 case htons(ETH_P_PPP_SES): { 1034 struct { 1035 struct pppoe_hdr hdr; 1036 __be16 proto; 1037 } *hdr, _hdr; 1038 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr); 1039 if (!hdr) { 1040 fdret = FLOW_DISSECT_RET_OUT_BAD; 1041 break; 1042 } 1043 1044 proto = hdr->proto; 1045 nhoff += PPPOE_SES_HLEN; 1046 switch (proto) { 1047 case htons(PPP_IP): 1048 proto = htons(ETH_P_IP); 1049 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1050 break; 1051 case htons(PPP_IPV6): 1052 proto = htons(ETH_P_IPV6); 1053 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1054 break; 1055 default: 1056 fdret = FLOW_DISSECT_RET_OUT_BAD; 1057 break; 1058 } 1059 break; 1060 } 1061 case htons(ETH_P_TIPC): { 1062 struct tipc_basic_hdr *hdr, _hdr; 1063 1064 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), 1065 data, hlen, &_hdr); 1066 if (!hdr) { 1067 fdret = FLOW_DISSECT_RET_OUT_BAD; 1068 break; 1069 } 1070 1071 if (dissector_uses_key(flow_dissector, 1072 FLOW_DISSECTOR_KEY_TIPC)) { 1073 key_addrs = skb_flow_dissector_target(flow_dissector, 1074 FLOW_DISSECTOR_KEY_TIPC, 1075 target_container); 1076 key_addrs->tipckey.key = tipc_hdr_rps_key(hdr); 1077 key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC; 1078 } 1079 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1080 break; 1081 } 1082 1083 case htons(ETH_P_MPLS_UC): 1084 case htons(ETH_P_MPLS_MC): 1085 fdret = __skb_flow_dissect_mpls(skb, flow_dissector, 1086 target_container, data, 1087 nhoff, hlen); 1088 break; 1089 case htons(ETH_P_FCOE): 1090 if ((hlen - nhoff) < FCOE_HEADER_LEN) { 1091 fdret = FLOW_DISSECT_RET_OUT_BAD; 1092 break; 1093 } 1094 1095 nhoff += FCOE_HEADER_LEN; 1096 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1097 break; 1098 1099 case htons(ETH_P_ARP): 1100 case htons(ETH_P_RARP): 1101 fdret = __skb_flow_dissect_arp(skb, flow_dissector, 1102 target_container, data, 1103 nhoff, hlen); 1104 break; 1105 1106 case htons(ETH_P_BATMAN): 1107 fdret = __skb_flow_dissect_batadv(skb, key_control, data, 1108 &proto, &nhoff, hlen, flags); 1109 break; 1110 1111 default: 1112 fdret = FLOW_DISSECT_RET_OUT_BAD; 1113 break; 1114 } 1115 1116 /* Process result of proto processing */ 1117 switch (fdret) { 1118 case FLOW_DISSECT_RET_OUT_GOOD: 1119 goto out_good; 1120 case FLOW_DISSECT_RET_PROTO_AGAIN: 1121 if (skb_flow_dissect_allowed(&num_hdrs)) 1122 goto proto_again; 1123 goto out_good; 1124 case FLOW_DISSECT_RET_CONTINUE: 1125 case FLOW_DISSECT_RET_IPPROTO_AGAIN: 1126 break; 1127 case FLOW_DISSECT_RET_OUT_BAD: 1128 default: 1129 goto out_bad; 1130 } 1131 1132 ip_proto_again: 1133 fdret = FLOW_DISSECT_RET_CONTINUE; 1134 1135 switch (ip_proto) { 1136 case IPPROTO_GRE: 1137 fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector, 1138 target_container, data, 1139 &proto, &nhoff, &hlen, flags); 1140 break; 1141 1142 case NEXTHDR_HOP: 1143 case NEXTHDR_ROUTING: 1144 case NEXTHDR_DEST: { 1145 u8 _opthdr[2], *opthdr; 1146 1147 if (proto != htons(ETH_P_IPV6)) 1148 break; 1149 1150 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr), 1151 data, hlen, &_opthdr); 1152 if (!opthdr) { 1153 fdret = FLOW_DISSECT_RET_OUT_BAD; 1154 break; 1155 } 1156 1157 ip_proto = opthdr[0]; 1158 nhoff += (opthdr[1] + 1) << 3; 1159 1160 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN; 1161 break; 1162 } 1163 case NEXTHDR_FRAGMENT: { 1164 struct frag_hdr _fh, *fh; 1165 1166 if (proto != htons(ETH_P_IPV6)) 1167 break; 1168 1169 fh = __skb_header_pointer(skb, nhoff, sizeof(_fh), 1170 data, hlen, &_fh); 1171 1172 if (!fh) { 1173 fdret = FLOW_DISSECT_RET_OUT_BAD; 1174 break; 1175 } 1176 1177 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 1178 1179 nhoff += sizeof(_fh); 1180 ip_proto = fh->nexthdr; 1181 1182 if (!(fh->frag_off & htons(IP6_OFFSET))) { 1183 key_control->flags |= FLOW_DIS_FIRST_FRAG; 1184 if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) { 1185 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN; 1186 break; 1187 } 1188 } 1189 1190 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1191 break; 1192 } 1193 case IPPROTO_IPIP: 1194 proto = htons(ETH_P_IP); 1195 1196 key_control->flags |= FLOW_DIS_ENCAPSULATION; 1197 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) { 1198 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1199 break; 1200 } 1201 1202 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1203 break; 1204 1205 case IPPROTO_IPV6: 1206 proto = htons(ETH_P_IPV6); 1207 1208 key_control->flags |= FLOW_DIS_ENCAPSULATION; 1209 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) { 1210 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1211 break; 1212 } 1213 1214 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1215 break; 1216 1217 1218 case IPPROTO_MPLS: 1219 proto = htons(ETH_P_MPLS_UC); 1220 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1221 break; 1222 1223 case IPPROTO_TCP: 1224 __skb_flow_dissect_tcp(skb, flow_dissector, target_container, 1225 data, nhoff, hlen); 1226 break; 1227 1228 default: 1229 break; 1230 } 1231 1232 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS) && 1233 !(key_control->flags & FLOW_DIS_IS_FRAGMENT)) { 1234 key_ports = skb_flow_dissector_target(flow_dissector, 1235 FLOW_DISSECTOR_KEY_PORTS, 1236 target_container); 1237 key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto, 1238 data, hlen); 1239 } 1240 1241 if (dissector_uses_key(flow_dissector, 1242 FLOW_DISSECTOR_KEY_ICMP)) { 1243 key_icmp = skb_flow_dissector_target(flow_dissector, 1244 FLOW_DISSECTOR_KEY_ICMP, 1245 target_container); 1246 key_icmp->icmp = skb_flow_get_be16(skb, nhoff, data, hlen); 1247 } 1248 1249 /* Process result of IP proto processing */ 1250 switch (fdret) { 1251 case FLOW_DISSECT_RET_PROTO_AGAIN: 1252 if (skb_flow_dissect_allowed(&num_hdrs)) 1253 goto proto_again; 1254 break; 1255 case FLOW_DISSECT_RET_IPPROTO_AGAIN: 1256 if (skb_flow_dissect_allowed(&num_hdrs)) 1257 goto ip_proto_again; 1258 break; 1259 case FLOW_DISSECT_RET_OUT_GOOD: 1260 case FLOW_DISSECT_RET_CONTINUE: 1261 break; 1262 case FLOW_DISSECT_RET_OUT_BAD: 1263 default: 1264 goto out_bad; 1265 } 1266 1267 out_good: 1268 ret = true; 1269 1270 out: 1271 key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen); 1272 key_basic->n_proto = proto; 1273 key_basic->ip_proto = ip_proto; 1274 1275 return ret; 1276 1277 out_bad: 1278 ret = false; 1279 goto out; 1280 } 1281 EXPORT_SYMBOL(__skb_flow_dissect); 1282 1283 static u32 hashrnd __read_mostly; 1284 static __always_inline void __flow_hash_secret_init(void) 1285 { 1286 net_get_random_once(&hashrnd, sizeof(hashrnd)); 1287 } 1288 1289 static __always_inline u32 __flow_hash_words(const u32 *words, u32 length, 1290 u32 keyval) 1291 { 1292 return jhash2(words, length, keyval); 1293 } 1294 1295 static inline const u32 *flow_keys_hash_start(const struct flow_keys *flow) 1296 { 1297 const void *p = flow; 1298 1299 BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % sizeof(u32)); 1300 return (const u32 *)(p + FLOW_KEYS_HASH_OFFSET); 1301 } 1302 1303 static inline size_t flow_keys_hash_length(const struct flow_keys *flow) 1304 { 1305 size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs); 1306 BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32)); 1307 BUILD_BUG_ON(offsetof(typeof(*flow), addrs) != 1308 sizeof(*flow) - sizeof(flow->addrs)); 1309 1310 switch (flow->control.addr_type) { 1311 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1312 diff -= sizeof(flow->addrs.v4addrs); 1313 break; 1314 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1315 diff -= sizeof(flow->addrs.v6addrs); 1316 break; 1317 case FLOW_DISSECTOR_KEY_TIPC: 1318 diff -= sizeof(flow->addrs.tipckey); 1319 break; 1320 } 1321 return (sizeof(*flow) - diff) / sizeof(u32); 1322 } 1323 1324 __be32 flow_get_u32_src(const struct flow_keys *flow) 1325 { 1326 switch (flow->control.addr_type) { 1327 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1328 return flow->addrs.v4addrs.src; 1329 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1330 return (__force __be32)ipv6_addr_hash( 1331 &flow->addrs.v6addrs.src); 1332 case FLOW_DISSECTOR_KEY_TIPC: 1333 return flow->addrs.tipckey.key; 1334 default: 1335 return 0; 1336 } 1337 } 1338 EXPORT_SYMBOL(flow_get_u32_src); 1339 1340 __be32 flow_get_u32_dst(const struct flow_keys *flow) 1341 { 1342 switch (flow->control.addr_type) { 1343 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1344 return flow->addrs.v4addrs.dst; 1345 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1346 return (__force __be32)ipv6_addr_hash( 1347 &flow->addrs.v6addrs.dst); 1348 default: 1349 return 0; 1350 } 1351 } 1352 EXPORT_SYMBOL(flow_get_u32_dst); 1353 1354 static inline void __flow_hash_consistentify(struct flow_keys *keys) 1355 { 1356 int addr_diff, i; 1357 1358 switch (keys->control.addr_type) { 1359 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1360 addr_diff = (__force u32)keys->addrs.v4addrs.dst - 1361 (__force u32)keys->addrs.v4addrs.src; 1362 if ((addr_diff < 0) || 1363 (addr_diff == 0 && 1364 ((__force u16)keys->ports.dst < 1365 (__force u16)keys->ports.src))) { 1366 swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst); 1367 swap(keys->ports.src, keys->ports.dst); 1368 } 1369 break; 1370 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1371 addr_diff = memcmp(&keys->addrs.v6addrs.dst, 1372 &keys->addrs.v6addrs.src, 1373 sizeof(keys->addrs.v6addrs.dst)); 1374 if ((addr_diff < 0) || 1375 (addr_diff == 0 && 1376 ((__force u16)keys->ports.dst < 1377 (__force u16)keys->ports.src))) { 1378 for (i = 0; i < 4; i++) 1379 swap(keys->addrs.v6addrs.src.s6_addr32[i], 1380 keys->addrs.v6addrs.dst.s6_addr32[i]); 1381 swap(keys->ports.src, keys->ports.dst); 1382 } 1383 break; 1384 } 1385 } 1386 1387 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval) 1388 { 1389 u32 hash; 1390 1391 __flow_hash_consistentify(keys); 1392 1393 hash = __flow_hash_words(flow_keys_hash_start(keys), 1394 flow_keys_hash_length(keys), keyval); 1395 if (!hash) 1396 hash = 1; 1397 1398 return hash; 1399 } 1400 1401 u32 flow_hash_from_keys(struct flow_keys *keys) 1402 { 1403 __flow_hash_secret_init(); 1404 return __flow_hash_from_keys(keys, hashrnd); 1405 } 1406 EXPORT_SYMBOL(flow_hash_from_keys); 1407 1408 static inline u32 ___skb_get_hash(const struct sk_buff *skb, 1409 struct flow_keys *keys, u32 keyval) 1410 { 1411 skb_flow_dissect_flow_keys(skb, keys, 1412 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 1413 1414 return __flow_hash_from_keys(keys, keyval); 1415 } 1416 1417 struct _flow_keys_digest_data { 1418 __be16 n_proto; 1419 u8 ip_proto; 1420 u8 padding; 1421 __be32 ports; 1422 __be32 src; 1423 __be32 dst; 1424 }; 1425 1426 void make_flow_keys_digest(struct flow_keys_digest *digest, 1427 const struct flow_keys *flow) 1428 { 1429 struct _flow_keys_digest_data *data = 1430 (struct _flow_keys_digest_data *)digest; 1431 1432 BUILD_BUG_ON(sizeof(*data) > sizeof(*digest)); 1433 1434 memset(digest, 0, sizeof(*digest)); 1435 1436 data->n_proto = flow->basic.n_proto; 1437 data->ip_proto = flow->basic.ip_proto; 1438 data->ports = flow->ports.ports; 1439 data->src = flow->addrs.v4addrs.src; 1440 data->dst = flow->addrs.v4addrs.dst; 1441 } 1442 EXPORT_SYMBOL(make_flow_keys_digest); 1443 1444 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly; 1445 1446 u32 __skb_get_hash_symmetric(const struct sk_buff *skb) 1447 { 1448 struct flow_keys keys; 1449 1450 __flow_hash_secret_init(); 1451 1452 memset(&keys, 0, sizeof(keys)); 1453 __skb_flow_dissect(NULL, skb, &flow_keys_dissector_symmetric, 1454 &keys, NULL, 0, 0, 0, 1455 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 1456 1457 return __flow_hash_from_keys(&keys, hashrnd); 1458 } 1459 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric); 1460 1461 /** 1462 * __skb_get_hash: calculate a flow hash 1463 * @skb: sk_buff to calculate flow hash from 1464 * 1465 * This function calculates a flow hash based on src/dst addresses 1466 * and src/dst port numbers. Sets hash in skb to non-zero hash value 1467 * on success, zero indicates no valid hash. Also, sets l4_hash in skb 1468 * if hash is a canonical 4-tuple hash over transport ports. 1469 */ 1470 void __skb_get_hash(struct sk_buff *skb) 1471 { 1472 struct flow_keys keys; 1473 u32 hash; 1474 1475 __flow_hash_secret_init(); 1476 1477 hash = ___skb_get_hash(skb, &keys, hashrnd); 1478 1479 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); 1480 } 1481 EXPORT_SYMBOL(__skb_get_hash); 1482 1483 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb) 1484 { 1485 struct flow_keys keys; 1486 1487 return ___skb_get_hash(skb, &keys, perturb); 1488 } 1489 EXPORT_SYMBOL(skb_get_hash_perturb); 1490 1491 u32 __skb_get_poff(const struct sk_buff *skb, void *data, 1492 const struct flow_keys_basic *keys, int hlen) 1493 { 1494 u32 poff = keys->control.thoff; 1495 1496 /* skip L4 headers for fragments after the first */ 1497 if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) && 1498 !(keys->control.flags & FLOW_DIS_FIRST_FRAG)) 1499 return poff; 1500 1501 switch (keys->basic.ip_proto) { 1502 case IPPROTO_TCP: { 1503 /* access doff as u8 to avoid unaligned access */ 1504 const u8 *doff; 1505 u8 _doff; 1506 1507 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff), 1508 data, hlen, &_doff); 1509 if (!doff) 1510 return poff; 1511 1512 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2); 1513 break; 1514 } 1515 case IPPROTO_UDP: 1516 case IPPROTO_UDPLITE: 1517 poff += sizeof(struct udphdr); 1518 break; 1519 /* For the rest, we do not really care about header 1520 * extensions at this point for now. 1521 */ 1522 case IPPROTO_ICMP: 1523 poff += sizeof(struct icmphdr); 1524 break; 1525 case IPPROTO_ICMPV6: 1526 poff += sizeof(struct icmp6hdr); 1527 break; 1528 case IPPROTO_IGMP: 1529 poff += sizeof(struct igmphdr); 1530 break; 1531 case IPPROTO_DCCP: 1532 poff += sizeof(struct dccp_hdr); 1533 break; 1534 case IPPROTO_SCTP: 1535 poff += sizeof(struct sctphdr); 1536 break; 1537 } 1538 1539 return poff; 1540 } 1541 1542 /** 1543 * skb_get_poff - get the offset to the payload 1544 * @skb: sk_buff to get the payload offset from 1545 * 1546 * The function will get the offset to the payload as far as it could 1547 * be dissected. The main user is currently BPF, so that we can dynamically 1548 * truncate packets without needing to push actual payload to the user 1549 * space and can analyze headers only, instead. 1550 */ 1551 u32 skb_get_poff(const struct sk_buff *skb) 1552 { 1553 struct flow_keys_basic keys; 1554 1555 if (!skb_flow_dissect_flow_keys_basic(NULL, skb, &keys, 1556 NULL, 0, 0, 0, 0)) 1557 return 0; 1558 1559 return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb)); 1560 } 1561 1562 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys) 1563 { 1564 memset(keys, 0, sizeof(*keys)); 1565 1566 memcpy(&keys->addrs.v6addrs.src, &fl6->saddr, 1567 sizeof(keys->addrs.v6addrs.src)); 1568 memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr, 1569 sizeof(keys->addrs.v6addrs.dst)); 1570 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 1571 keys->ports.src = fl6->fl6_sport; 1572 keys->ports.dst = fl6->fl6_dport; 1573 keys->keyid.keyid = fl6->fl6_gre_key; 1574 keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6); 1575 keys->basic.ip_proto = fl6->flowi6_proto; 1576 1577 return flow_hash_from_keys(keys); 1578 } 1579 EXPORT_SYMBOL(__get_hash_from_flowi6); 1580 1581 static const struct flow_dissector_key flow_keys_dissector_keys[] = { 1582 { 1583 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 1584 .offset = offsetof(struct flow_keys, control), 1585 }, 1586 { 1587 .key_id = FLOW_DISSECTOR_KEY_BASIC, 1588 .offset = offsetof(struct flow_keys, basic), 1589 }, 1590 { 1591 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS, 1592 .offset = offsetof(struct flow_keys, addrs.v4addrs), 1593 }, 1594 { 1595 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS, 1596 .offset = offsetof(struct flow_keys, addrs.v6addrs), 1597 }, 1598 { 1599 .key_id = FLOW_DISSECTOR_KEY_TIPC, 1600 .offset = offsetof(struct flow_keys, addrs.tipckey), 1601 }, 1602 { 1603 .key_id = FLOW_DISSECTOR_KEY_PORTS, 1604 .offset = offsetof(struct flow_keys, ports), 1605 }, 1606 { 1607 .key_id = FLOW_DISSECTOR_KEY_VLAN, 1608 .offset = offsetof(struct flow_keys, vlan), 1609 }, 1610 { 1611 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL, 1612 .offset = offsetof(struct flow_keys, tags), 1613 }, 1614 { 1615 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID, 1616 .offset = offsetof(struct flow_keys, keyid), 1617 }, 1618 }; 1619 1620 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = { 1621 { 1622 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 1623 .offset = offsetof(struct flow_keys, control), 1624 }, 1625 { 1626 .key_id = FLOW_DISSECTOR_KEY_BASIC, 1627 .offset = offsetof(struct flow_keys, basic), 1628 }, 1629 { 1630 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS, 1631 .offset = offsetof(struct flow_keys, addrs.v4addrs), 1632 }, 1633 { 1634 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS, 1635 .offset = offsetof(struct flow_keys, addrs.v6addrs), 1636 }, 1637 { 1638 .key_id = FLOW_DISSECTOR_KEY_PORTS, 1639 .offset = offsetof(struct flow_keys, ports), 1640 }, 1641 }; 1642 1643 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = { 1644 { 1645 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 1646 .offset = offsetof(struct flow_keys, control), 1647 }, 1648 { 1649 .key_id = FLOW_DISSECTOR_KEY_BASIC, 1650 .offset = offsetof(struct flow_keys, basic), 1651 }, 1652 }; 1653 1654 struct flow_dissector flow_keys_dissector __read_mostly; 1655 EXPORT_SYMBOL(flow_keys_dissector); 1656 1657 struct flow_dissector flow_keys_basic_dissector __read_mostly; 1658 EXPORT_SYMBOL(flow_keys_basic_dissector); 1659 1660 static int __init init_default_flow_dissectors(void) 1661 { 1662 skb_flow_dissector_init(&flow_keys_dissector, 1663 flow_keys_dissector_keys, 1664 ARRAY_SIZE(flow_keys_dissector_keys)); 1665 skb_flow_dissector_init(&flow_keys_dissector_symmetric, 1666 flow_keys_dissector_symmetric_keys, 1667 ARRAY_SIZE(flow_keys_dissector_symmetric_keys)); 1668 skb_flow_dissector_init(&flow_keys_basic_dissector, 1669 flow_keys_basic_dissector_keys, 1670 ARRAY_SIZE(flow_keys_basic_dissector_keys)); 1671 return 0; 1672 } 1673 1674 core_initcall(init_default_flow_dissectors); 1675