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 /** 687 * __skb_flow_dissect - extract the flow_keys struct and return it 688 * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified 689 * @flow_dissector: list of keys to dissect 690 * @target_container: target structure to put dissected values into 691 * @data: raw buffer pointer to the packet, if NULL use skb->data 692 * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol 693 * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb) 694 * @hlen: packet header length, if @data is NULL use skb_headlen(skb) 695 * 696 * The function will try to retrieve individual keys into target specified 697 * by flow_dissector from either the skbuff or a raw buffer specified by the 698 * rest parameters. 699 * 700 * Caller must take care of zeroing target container memory. 701 */ 702 bool __skb_flow_dissect(const struct sk_buff *skb, 703 struct flow_dissector *flow_dissector, 704 void *target_container, 705 void *data, __be16 proto, int nhoff, int hlen, 706 unsigned int flags) 707 { 708 struct flow_dissector_key_control *key_control; 709 struct flow_dissector_key_basic *key_basic; 710 struct flow_dissector_key_addrs *key_addrs; 711 struct flow_dissector_key_ports *key_ports; 712 struct flow_dissector_key_icmp *key_icmp; 713 struct flow_dissector_key_tags *key_tags; 714 struct flow_dissector_key_vlan *key_vlan; 715 enum flow_dissect_ret fdret; 716 enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX; 717 struct bpf_prog *attached = NULL; 718 int num_hdrs = 0; 719 u8 ip_proto = 0; 720 bool ret; 721 722 if (!data) { 723 data = skb->data; 724 proto = skb_vlan_tag_present(skb) ? 725 skb->vlan_proto : skb->protocol; 726 nhoff = skb_network_offset(skb); 727 hlen = skb_headlen(skb); 728 #if IS_ENABLED(CONFIG_NET_DSA) 729 if (unlikely(skb->dev && netdev_uses_dsa(skb->dev))) { 730 const struct dsa_device_ops *ops; 731 int offset; 732 733 ops = skb->dev->dsa_ptr->tag_ops; 734 if (ops->flow_dissect && 735 !ops->flow_dissect(skb, &proto, &offset)) { 736 hlen -= offset; 737 nhoff += offset; 738 } 739 } 740 #endif 741 } 742 743 /* It is ensured by skb_flow_dissector_init() that control key will 744 * be always present. 745 */ 746 key_control = skb_flow_dissector_target(flow_dissector, 747 FLOW_DISSECTOR_KEY_CONTROL, 748 target_container); 749 750 /* It is ensured by skb_flow_dissector_init() that basic key will 751 * be always present. 752 */ 753 key_basic = skb_flow_dissector_target(flow_dissector, 754 FLOW_DISSECTOR_KEY_BASIC, 755 target_container); 756 757 rcu_read_lock(); 758 if (skb) { 759 if (skb->dev) 760 attached = rcu_dereference(dev_net(skb->dev)->flow_dissector_prog); 761 else if (skb->sk) 762 attached = rcu_dereference(sock_net(skb->sk)->flow_dissector_prog); 763 else 764 WARN_ON_ONCE(1); 765 } 766 if (attached) { 767 /* Note that even though the const qualifier is discarded 768 * throughout the execution of the BPF program, all changes(the 769 * control block) are reverted after the BPF program returns. 770 * Therefore, __skb_flow_dissect does not alter the skb. 771 */ 772 struct bpf_flow_keys flow_keys = {}; 773 struct bpf_skb_data_end cb_saved; 774 struct bpf_skb_data_end *cb; 775 u32 result; 776 777 cb = (struct bpf_skb_data_end *)skb->cb; 778 779 /* Save Control Block */ 780 memcpy(&cb_saved, cb, sizeof(cb_saved)); 781 memset(cb, 0, sizeof(cb_saved)); 782 783 /* Pass parameters to the BPF program */ 784 cb->qdisc_cb.flow_keys = &flow_keys; 785 flow_keys.nhoff = nhoff; 786 flow_keys.thoff = nhoff; 787 788 bpf_compute_data_pointers((struct sk_buff *)skb); 789 result = BPF_PROG_RUN(attached, skb); 790 791 /* Restore state */ 792 memcpy(cb, &cb_saved, sizeof(cb_saved)); 793 794 flow_keys.nhoff = clamp_t(u16, flow_keys.nhoff, 0, skb->len); 795 flow_keys.thoff = clamp_t(u16, flow_keys.thoff, 796 flow_keys.nhoff, skb->len); 797 798 __skb_flow_bpf_to_target(&flow_keys, flow_dissector, 799 target_container); 800 rcu_read_unlock(); 801 return result == BPF_OK; 802 } 803 rcu_read_unlock(); 804 805 if (dissector_uses_key(flow_dissector, 806 FLOW_DISSECTOR_KEY_ETH_ADDRS)) { 807 struct ethhdr *eth = eth_hdr(skb); 808 struct flow_dissector_key_eth_addrs *key_eth_addrs; 809 810 key_eth_addrs = skb_flow_dissector_target(flow_dissector, 811 FLOW_DISSECTOR_KEY_ETH_ADDRS, 812 target_container); 813 memcpy(key_eth_addrs, ð->h_dest, sizeof(*key_eth_addrs)); 814 } 815 816 proto_again: 817 fdret = FLOW_DISSECT_RET_CONTINUE; 818 819 switch (proto) { 820 case htons(ETH_P_IP): { 821 const struct iphdr *iph; 822 struct iphdr _iph; 823 824 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph); 825 if (!iph || iph->ihl < 5) { 826 fdret = FLOW_DISSECT_RET_OUT_BAD; 827 break; 828 } 829 830 nhoff += iph->ihl * 4; 831 832 ip_proto = iph->protocol; 833 834 if (dissector_uses_key(flow_dissector, 835 FLOW_DISSECTOR_KEY_IPV4_ADDRS)) { 836 key_addrs = skb_flow_dissector_target(flow_dissector, 837 FLOW_DISSECTOR_KEY_IPV4_ADDRS, 838 target_container); 839 840 memcpy(&key_addrs->v4addrs, &iph->saddr, 841 sizeof(key_addrs->v4addrs)); 842 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; 843 } 844 845 if (ip_is_fragment(iph)) { 846 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 847 848 if (iph->frag_off & htons(IP_OFFSET)) { 849 fdret = FLOW_DISSECT_RET_OUT_GOOD; 850 break; 851 } else { 852 key_control->flags |= FLOW_DIS_FIRST_FRAG; 853 if (!(flags & 854 FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) { 855 fdret = FLOW_DISSECT_RET_OUT_GOOD; 856 break; 857 } 858 } 859 } 860 861 __skb_flow_dissect_ipv4(skb, flow_dissector, 862 target_container, data, iph); 863 864 if (flags & FLOW_DISSECTOR_F_STOP_AT_L3) { 865 fdret = FLOW_DISSECT_RET_OUT_GOOD; 866 break; 867 } 868 869 break; 870 } 871 case htons(ETH_P_IPV6): { 872 const struct ipv6hdr *iph; 873 struct ipv6hdr _iph; 874 875 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph); 876 if (!iph) { 877 fdret = FLOW_DISSECT_RET_OUT_BAD; 878 break; 879 } 880 881 ip_proto = iph->nexthdr; 882 nhoff += sizeof(struct ipv6hdr); 883 884 if (dissector_uses_key(flow_dissector, 885 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) { 886 key_addrs = skb_flow_dissector_target(flow_dissector, 887 FLOW_DISSECTOR_KEY_IPV6_ADDRS, 888 target_container); 889 890 memcpy(&key_addrs->v6addrs, &iph->saddr, 891 sizeof(key_addrs->v6addrs)); 892 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 893 } 894 895 if ((dissector_uses_key(flow_dissector, 896 FLOW_DISSECTOR_KEY_FLOW_LABEL) || 897 (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) && 898 ip6_flowlabel(iph)) { 899 __be32 flow_label = ip6_flowlabel(iph); 900 901 if (dissector_uses_key(flow_dissector, 902 FLOW_DISSECTOR_KEY_FLOW_LABEL)) { 903 key_tags = skb_flow_dissector_target(flow_dissector, 904 FLOW_DISSECTOR_KEY_FLOW_LABEL, 905 target_container); 906 key_tags->flow_label = ntohl(flow_label); 907 } 908 if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) { 909 fdret = FLOW_DISSECT_RET_OUT_GOOD; 910 break; 911 } 912 } 913 914 __skb_flow_dissect_ipv6(skb, flow_dissector, 915 target_container, data, iph); 916 917 if (flags & FLOW_DISSECTOR_F_STOP_AT_L3) 918 fdret = FLOW_DISSECT_RET_OUT_GOOD; 919 920 break; 921 } 922 case htons(ETH_P_8021AD): 923 case htons(ETH_P_8021Q): { 924 const struct vlan_hdr *vlan = NULL; 925 struct vlan_hdr _vlan; 926 __be16 saved_vlan_tpid = proto; 927 928 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX && 929 skb && skb_vlan_tag_present(skb)) { 930 proto = skb->protocol; 931 } else { 932 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan), 933 data, hlen, &_vlan); 934 if (!vlan) { 935 fdret = FLOW_DISSECT_RET_OUT_BAD; 936 break; 937 } 938 939 proto = vlan->h_vlan_encapsulated_proto; 940 nhoff += sizeof(*vlan); 941 } 942 943 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) { 944 dissector_vlan = FLOW_DISSECTOR_KEY_VLAN; 945 } else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) { 946 dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN; 947 } else { 948 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 949 break; 950 } 951 952 if (dissector_uses_key(flow_dissector, dissector_vlan)) { 953 key_vlan = skb_flow_dissector_target(flow_dissector, 954 dissector_vlan, 955 target_container); 956 957 if (!vlan) { 958 key_vlan->vlan_id = skb_vlan_tag_get_id(skb); 959 key_vlan->vlan_priority = 960 (skb_vlan_tag_get_prio(skb) >> VLAN_PRIO_SHIFT); 961 } else { 962 key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) & 963 VLAN_VID_MASK; 964 key_vlan->vlan_priority = 965 (ntohs(vlan->h_vlan_TCI) & 966 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT; 967 } 968 key_vlan->vlan_tpid = saved_vlan_tpid; 969 } 970 971 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 972 break; 973 } 974 case htons(ETH_P_PPP_SES): { 975 struct { 976 struct pppoe_hdr hdr; 977 __be16 proto; 978 } *hdr, _hdr; 979 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr); 980 if (!hdr) { 981 fdret = FLOW_DISSECT_RET_OUT_BAD; 982 break; 983 } 984 985 proto = hdr->proto; 986 nhoff += PPPOE_SES_HLEN; 987 switch (proto) { 988 case htons(PPP_IP): 989 proto = htons(ETH_P_IP); 990 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 991 break; 992 case htons(PPP_IPV6): 993 proto = htons(ETH_P_IPV6); 994 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 995 break; 996 default: 997 fdret = FLOW_DISSECT_RET_OUT_BAD; 998 break; 999 } 1000 break; 1001 } 1002 case htons(ETH_P_TIPC): { 1003 struct tipc_basic_hdr *hdr, _hdr; 1004 1005 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), 1006 data, hlen, &_hdr); 1007 if (!hdr) { 1008 fdret = FLOW_DISSECT_RET_OUT_BAD; 1009 break; 1010 } 1011 1012 if (dissector_uses_key(flow_dissector, 1013 FLOW_DISSECTOR_KEY_TIPC)) { 1014 key_addrs = skb_flow_dissector_target(flow_dissector, 1015 FLOW_DISSECTOR_KEY_TIPC, 1016 target_container); 1017 key_addrs->tipckey.key = tipc_hdr_rps_key(hdr); 1018 key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC; 1019 } 1020 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1021 break; 1022 } 1023 1024 case htons(ETH_P_MPLS_UC): 1025 case htons(ETH_P_MPLS_MC): 1026 fdret = __skb_flow_dissect_mpls(skb, flow_dissector, 1027 target_container, data, 1028 nhoff, hlen); 1029 break; 1030 case htons(ETH_P_FCOE): 1031 if ((hlen - nhoff) < FCOE_HEADER_LEN) { 1032 fdret = FLOW_DISSECT_RET_OUT_BAD; 1033 break; 1034 } 1035 1036 nhoff += FCOE_HEADER_LEN; 1037 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1038 break; 1039 1040 case htons(ETH_P_ARP): 1041 case htons(ETH_P_RARP): 1042 fdret = __skb_flow_dissect_arp(skb, flow_dissector, 1043 target_container, data, 1044 nhoff, hlen); 1045 break; 1046 1047 case htons(ETH_P_BATMAN): 1048 fdret = __skb_flow_dissect_batadv(skb, key_control, data, 1049 &proto, &nhoff, hlen, flags); 1050 break; 1051 1052 default: 1053 fdret = FLOW_DISSECT_RET_OUT_BAD; 1054 break; 1055 } 1056 1057 /* Process result of proto processing */ 1058 switch (fdret) { 1059 case FLOW_DISSECT_RET_OUT_GOOD: 1060 goto out_good; 1061 case FLOW_DISSECT_RET_PROTO_AGAIN: 1062 if (skb_flow_dissect_allowed(&num_hdrs)) 1063 goto proto_again; 1064 goto out_good; 1065 case FLOW_DISSECT_RET_CONTINUE: 1066 case FLOW_DISSECT_RET_IPPROTO_AGAIN: 1067 break; 1068 case FLOW_DISSECT_RET_OUT_BAD: 1069 default: 1070 goto out_bad; 1071 } 1072 1073 ip_proto_again: 1074 fdret = FLOW_DISSECT_RET_CONTINUE; 1075 1076 switch (ip_proto) { 1077 case IPPROTO_GRE: 1078 fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector, 1079 target_container, data, 1080 &proto, &nhoff, &hlen, flags); 1081 break; 1082 1083 case NEXTHDR_HOP: 1084 case NEXTHDR_ROUTING: 1085 case NEXTHDR_DEST: { 1086 u8 _opthdr[2], *opthdr; 1087 1088 if (proto != htons(ETH_P_IPV6)) 1089 break; 1090 1091 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr), 1092 data, hlen, &_opthdr); 1093 if (!opthdr) { 1094 fdret = FLOW_DISSECT_RET_OUT_BAD; 1095 break; 1096 } 1097 1098 ip_proto = opthdr[0]; 1099 nhoff += (opthdr[1] + 1) << 3; 1100 1101 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN; 1102 break; 1103 } 1104 case NEXTHDR_FRAGMENT: { 1105 struct frag_hdr _fh, *fh; 1106 1107 if (proto != htons(ETH_P_IPV6)) 1108 break; 1109 1110 fh = __skb_header_pointer(skb, nhoff, sizeof(_fh), 1111 data, hlen, &_fh); 1112 1113 if (!fh) { 1114 fdret = FLOW_DISSECT_RET_OUT_BAD; 1115 break; 1116 } 1117 1118 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 1119 1120 nhoff += sizeof(_fh); 1121 ip_proto = fh->nexthdr; 1122 1123 if (!(fh->frag_off & htons(IP6_OFFSET))) { 1124 key_control->flags |= FLOW_DIS_FIRST_FRAG; 1125 if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) { 1126 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN; 1127 break; 1128 } 1129 } 1130 1131 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1132 break; 1133 } 1134 case IPPROTO_IPIP: 1135 proto = htons(ETH_P_IP); 1136 1137 key_control->flags |= FLOW_DIS_ENCAPSULATION; 1138 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) { 1139 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1140 break; 1141 } 1142 1143 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1144 break; 1145 1146 case IPPROTO_IPV6: 1147 proto = htons(ETH_P_IPV6); 1148 1149 key_control->flags |= FLOW_DIS_ENCAPSULATION; 1150 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) { 1151 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1152 break; 1153 } 1154 1155 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1156 break; 1157 1158 1159 case IPPROTO_MPLS: 1160 proto = htons(ETH_P_MPLS_UC); 1161 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1162 break; 1163 1164 case IPPROTO_TCP: 1165 __skb_flow_dissect_tcp(skb, flow_dissector, target_container, 1166 data, nhoff, hlen); 1167 break; 1168 1169 default: 1170 break; 1171 } 1172 1173 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS) && 1174 !(key_control->flags & FLOW_DIS_IS_FRAGMENT)) { 1175 key_ports = skb_flow_dissector_target(flow_dissector, 1176 FLOW_DISSECTOR_KEY_PORTS, 1177 target_container); 1178 key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto, 1179 data, hlen); 1180 } 1181 1182 if (dissector_uses_key(flow_dissector, 1183 FLOW_DISSECTOR_KEY_ICMP)) { 1184 key_icmp = skb_flow_dissector_target(flow_dissector, 1185 FLOW_DISSECTOR_KEY_ICMP, 1186 target_container); 1187 key_icmp->icmp = skb_flow_get_be16(skb, nhoff, data, hlen); 1188 } 1189 1190 /* Process result of IP proto processing */ 1191 switch (fdret) { 1192 case FLOW_DISSECT_RET_PROTO_AGAIN: 1193 if (skb_flow_dissect_allowed(&num_hdrs)) 1194 goto proto_again; 1195 break; 1196 case FLOW_DISSECT_RET_IPPROTO_AGAIN: 1197 if (skb_flow_dissect_allowed(&num_hdrs)) 1198 goto ip_proto_again; 1199 break; 1200 case FLOW_DISSECT_RET_OUT_GOOD: 1201 case FLOW_DISSECT_RET_CONTINUE: 1202 break; 1203 case FLOW_DISSECT_RET_OUT_BAD: 1204 default: 1205 goto out_bad; 1206 } 1207 1208 out_good: 1209 ret = true; 1210 1211 out: 1212 key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen); 1213 key_basic->n_proto = proto; 1214 key_basic->ip_proto = ip_proto; 1215 1216 return ret; 1217 1218 out_bad: 1219 ret = false; 1220 goto out; 1221 } 1222 EXPORT_SYMBOL(__skb_flow_dissect); 1223 1224 static u32 hashrnd __read_mostly; 1225 static __always_inline void __flow_hash_secret_init(void) 1226 { 1227 net_get_random_once(&hashrnd, sizeof(hashrnd)); 1228 } 1229 1230 static __always_inline u32 __flow_hash_words(const u32 *words, u32 length, 1231 u32 keyval) 1232 { 1233 return jhash2(words, length, keyval); 1234 } 1235 1236 static inline const u32 *flow_keys_hash_start(const struct flow_keys *flow) 1237 { 1238 const void *p = flow; 1239 1240 BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % sizeof(u32)); 1241 return (const u32 *)(p + FLOW_KEYS_HASH_OFFSET); 1242 } 1243 1244 static inline size_t flow_keys_hash_length(const struct flow_keys *flow) 1245 { 1246 size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs); 1247 BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32)); 1248 BUILD_BUG_ON(offsetof(typeof(*flow), addrs) != 1249 sizeof(*flow) - sizeof(flow->addrs)); 1250 1251 switch (flow->control.addr_type) { 1252 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1253 diff -= sizeof(flow->addrs.v4addrs); 1254 break; 1255 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1256 diff -= sizeof(flow->addrs.v6addrs); 1257 break; 1258 case FLOW_DISSECTOR_KEY_TIPC: 1259 diff -= sizeof(flow->addrs.tipckey); 1260 break; 1261 } 1262 return (sizeof(*flow) - diff) / sizeof(u32); 1263 } 1264 1265 __be32 flow_get_u32_src(const struct flow_keys *flow) 1266 { 1267 switch (flow->control.addr_type) { 1268 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1269 return flow->addrs.v4addrs.src; 1270 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1271 return (__force __be32)ipv6_addr_hash( 1272 &flow->addrs.v6addrs.src); 1273 case FLOW_DISSECTOR_KEY_TIPC: 1274 return flow->addrs.tipckey.key; 1275 default: 1276 return 0; 1277 } 1278 } 1279 EXPORT_SYMBOL(flow_get_u32_src); 1280 1281 __be32 flow_get_u32_dst(const struct flow_keys *flow) 1282 { 1283 switch (flow->control.addr_type) { 1284 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1285 return flow->addrs.v4addrs.dst; 1286 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1287 return (__force __be32)ipv6_addr_hash( 1288 &flow->addrs.v6addrs.dst); 1289 default: 1290 return 0; 1291 } 1292 } 1293 EXPORT_SYMBOL(flow_get_u32_dst); 1294 1295 static inline void __flow_hash_consistentify(struct flow_keys *keys) 1296 { 1297 int addr_diff, i; 1298 1299 switch (keys->control.addr_type) { 1300 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1301 addr_diff = (__force u32)keys->addrs.v4addrs.dst - 1302 (__force u32)keys->addrs.v4addrs.src; 1303 if ((addr_diff < 0) || 1304 (addr_diff == 0 && 1305 ((__force u16)keys->ports.dst < 1306 (__force u16)keys->ports.src))) { 1307 swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst); 1308 swap(keys->ports.src, keys->ports.dst); 1309 } 1310 break; 1311 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1312 addr_diff = memcmp(&keys->addrs.v6addrs.dst, 1313 &keys->addrs.v6addrs.src, 1314 sizeof(keys->addrs.v6addrs.dst)); 1315 if ((addr_diff < 0) || 1316 (addr_diff == 0 && 1317 ((__force u16)keys->ports.dst < 1318 (__force u16)keys->ports.src))) { 1319 for (i = 0; i < 4; i++) 1320 swap(keys->addrs.v6addrs.src.s6_addr32[i], 1321 keys->addrs.v6addrs.dst.s6_addr32[i]); 1322 swap(keys->ports.src, keys->ports.dst); 1323 } 1324 break; 1325 } 1326 } 1327 1328 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval) 1329 { 1330 u32 hash; 1331 1332 __flow_hash_consistentify(keys); 1333 1334 hash = __flow_hash_words(flow_keys_hash_start(keys), 1335 flow_keys_hash_length(keys), keyval); 1336 if (!hash) 1337 hash = 1; 1338 1339 return hash; 1340 } 1341 1342 u32 flow_hash_from_keys(struct flow_keys *keys) 1343 { 1344 __flow_hash_secret_init(); 1345 return __flow_hash_from_keys(keys, hashrnd); 1346 } 1347 EXPORT_SYMBOL(flow_hash_from_keys); 1348 1349 static inline u32 ___skb_get_hash(const struct sk_buff *skb, 1350 struct flow_keys *keys, u32 keyval) 1351 { 1352 skb_flow_dissect_flow_keys(skb, keys, 1353 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 1354 1355 return __flow_hash_from_keys(keys, keyval); 1356 } 1357 1358 struct _flow_keys_digest_data { 1359 __be16 n_proto; 1360 u8 ip_proto; 1361 u8 padding; 1362 __be32 ports; 1363 __be32 src; 1364 __be32 dst; 1365 }; 1366 1367 void make_flow_keys_digest(struct flow_keys_digest *digest, 1368 const struct flow_keys *flow) 1369 { 1370 struct _flow_keys_digest_data *data = 1371 (struct _flow_keys_digest_data *)digest; 1372 1373 BUILD_BUG_ON(sizeof(*data) > sizeof(*digest)); 1374 1375 memset(digest, 0, sizeof(*digest)); 1376 1377 data->n_proto = flow->basic.n_proto; 1378 data->ip_proto = flow->basic.ip_proto; 1379 data->ports = flow->ports.ports; 1380 data->src = flow->addrs.v4addrs.src; 1381 data->dst = flow->addrs.v4addrs.dst; 1382 } 1383 EXPORT_SYMBOL(make_flow_keys_digest); 1384 1385 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly; 1386 1387 u32 __skb_get_hash_symmetric(const struct sk_buff *skb) 1388 { 1389 struct flow_keys keys; 1390 1391 __flow_hash_secret_init(); 1392 1393 memset(&keys, 0, sizeof(keys)); 1394 __skb_flow_dissect(skb, &flow_keys_dissector_symmetric, &keys, 1395 NULL, 0, 0, 0, 1396 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 1397 1398 return __flow_hash_from_keys(&keys, hashrnd); 1399 } 1400 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric); 1401 1402 /** 1403 * __skb_get_hash: calculate a flow hash 1404 * @skb: sk_buff to calculate flow hash from 1405 * 1406 * This function calculates a flow hash based on src/dst addresses 1407 * and src/dst port numbers. Sets hash in skb to non-zero hash value 1408 * on success, zero indicates no valid hash. Also, sets l4_hash in skb 1409 * if hash is a canonical 4-tuple hash over transport ports. 1410 */ 1411 void __skb_get_hash(struct sk_buff *skb) 1412 { 1413 struct flow_keys keys; 1414 u32 hash; 1415 1416 __flow_hash_secret_init(); 1417 1418 hash = ___skb_get_hash(skb, &keys, hashrnd); 1419 1420 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); 1421 } 1422 EXPORT_SYMBOL(__skb_get_hash); 1423 1424 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb) 1425 { 1426 struct flow_keys keys; 1427 1428 return ___skb_get_hash(skb, &keys, perturb); 1429 } 1430 EXPORT_SYMBOL(skb_get_hash_perturb); 1431 1432 u32 __skb_get_poff(const struct sk_buff *skb, void *data, 1433 const struct flow_keys_basic *keys, int hlen) 1434 { 1435 u32 poff = keys->control.thoff; 1436 1437 /* skip L4 headers for fragments after the first */ 1438 if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) && 1439 !(keys->control.flags & FLOW_DIS_FIRST_FRAG)) 1440 return poff; 1441 1442 switch (keys->basic.ip_proto) { 1443 case IPPROTO_TCP: { 1444 /* access doff as u8 to avoid unaligned access */ 1445 const u8 *doff; 1446 u8 _doff; 1447 1448 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff), 1449 data, hlen, &_doff); 1450 if (!doff) 1451 return poff; 1452 1453 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2); 1454 break; 1455 } 1456 case IPPROTO_UDP: 1457 case IPPROTO_UDPLITE: 1458 poff += sizeof(struct udphdr); 1459 break; 1460 /* For the rest, we do not really care about header 1461 * extensions at this point for now. 1462 */ 1463 case IPPROTO_ICMP: 1464 poff += sizeof(struct icmphdr); 1465 break; 1466 case IPPROTO_ICMPV6: 1467 poff += sizeof(struct icmp6hdr); 1468 break; 1469 case IPPROTO_IGMP: 1470 poff += sizeof(struct igmphdr); 1471 break; 1472 case IPPROTO_DCCP: 1473 poff += sizeof(struct dccp_hdr); 1474 break; 1475 case IPPROTO_SCTP: 1476 poff += sizeof(struct sctphdr); 1477 break; 1478 } 1479 1480 return poff; 1481 } 1482 1483 /** 1484 * skb_get_poff - get the offset to the payload 1485 * @skb: sk_buff to get the payload offset from 1486 * 1487 * The function will get the offset to the payload as far as it could 1488 * be dissected. The main user is currently BPF, so that we can dynamically 1489 * truncate packets without needing to push actual payload to the user 1490 * space and can analyze headers only, instead. 1491 */ 1492 u32 skb_get_poff(const struct sk_buff *skb) 1493 { 1494 struct flow_keys_basic keys; 1495 1496 if (!skb_flow_dissect_flow_keys_basic(skb, &keys, NULL, 0, 0, 0, 0)) 1497 return 0; 1498 1499 return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb)); 1500 } 1501 1502 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys) 1503 { 1504 memset(keys, 0, sizeof(*keys)); 1505 1506 memcpy(&keys->addrs.v6addrs.src, &fl6->saddr, 1507 sizeof(keys->addrs.v6addrs.src)); 1508 memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr, 1509 sizeof(keys->addrs.v6addrs.dst)); 1510 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 1511 keys->ports.src = fl6->fl6_sport; 1512 keys->ports.dst = fl6->fl6_dport; 1513 keys->keyid.keyid = fl6->fl6_gre_key; 1514 keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6); 1515 keys->basic.ip_proto = fl6->flowi6_proto; 1516 1517 return flow_hash_from_keys(keys); 1518 } 1519 EXPORT_SYMBOL(__get_hash_from_flowi6); 1520 1521 static const struct flow_dissector_key flow_keys_dissector_keys[] = { 1522 { 1523 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 1524 .offset = offsetof(struct flow_keys, control), 1525 }, 1526 { 1527 .key_id = FLOW_DISSECTOR_KEY_BASIC, 1528 .offset = offsetof(struct flow_keys, basic), 1529 }, 1530 { 1531 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS, 1532 .offset = offsetof(struct flow_keys, addrs.v4addrs), 1533 }, 1534 { 1535 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS, 1536 .offset = offsetof(struct flow_keys, addrs.v6addrs), 1537 }, 1538 { 1539 .key_id = FLOW_DISSECTOR_KEY_TIPC, 1540 .offset = offsetof(struct flow_keys, addrs.tipckey), 1541 }, 1542 { 1543 .key_id = FLOW_DISSECTOR_KEY_PORTS, 1544 .offset = offsetof(struct flow_keys, ports), 1545 }, 1546 { 1547 .key_id = FLOW_DISSECTOR_KEY_VLAN, 1548 .offset = offsetof(struct flow_keys, vlan), 1549 }, 1550 { 1551 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL, 1552 .offset = offsetof(struct flow_keys, tags), 1553 }, 1554 { 1555 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID, 1556 .offset = offsetof(struct flow_keys, keyid), 1557 }, 1558 }; 1559 1560 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = { 1561 { 1562 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 1563 .offset = offsetof(struct flow_keys, control), 1564 }, 1565 { 1566 .key_id = FLOW_DISSECTOR_KEY_BASIC, 1567 .offset = offsetof(struct flow_keys, basic), 1568 }, 1569 { 1570 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS, 1571 .offset = offsetof(struct flow_keys, addrs.v4addrs), 1572 }, 1573 { 1574 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS, 1575 .offset = offsetof(struct flow_keys, addrs.v6addrs), 1576 }, 1577 { 1578 .key_id = FLOW_DISSECTOR_KEY_PORTS, 1579 .offset = offsetof(struct flow_keys, ports), 1580 }, 1581 }; 1582 1583 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = { 1584 { 1585 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 1586 .offset = offsetof(struct flow_keys, control), 1587 }, 1588 { 1589 .key_id = FLOW_DISSECTOR_KEY_BASIC, 1590 .offset = offsetof(struct flow_keys, basic), 1591 }, 1592 }; 1593 1594 struct flow_dissector flow_keys_dissector __read_mostly; 1595 EXPORT_SYMBOL(flow_keys_dissector); 1596 1597 struct flow_dissector flow_keys_basic_dissector __read_mostly; 1598 EXPORT_SYMBOL(flow_keys_basic_dissector); 1599 1600 static int __init init_default_flow_dissectors(void) 1601 { 1602 skb_flow_dissector_init(&flow_keys_dissector, 1603 flow_keys_dissector_keys, 1604 ARRAY_SIZE(flow_keys_dissector_keys)); 1605 skb_flow_dissector_init(&flow_keys_dissector_symmetric, 1606 flow_keys_dissector_symmetric_keys, 1607 ARRAY_SIZE(flow_keys_dissector_symmetric_keys)); 1608 skb_flow_dissector_init(&flow_keys_basic_dissector, 1609 flow_keys_basic_dissector_keys, 1610 ARRAY_SIZE(flow_keys_basic_dissector_keys)); 1611 return 0; 1612 } 1613 1614 core_initcall(init_default_flow_dissectors); 1615