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