1 // SPDX-License-Identifier: GPL-2.0-only 2 #include <linux/kernel.h> 3 #include <linux/skbuff.h> 4 #include <linux/export.h> 5 #include <linux/ip.h> 6 #include <linux/ipv6.h> 7 #include <linux/if_vlan.h> 8 #include <net/dsa.h> 9 #include <net/dst_metadata.h> 10 #include <net/ip.h> 11 #include <net/ipv6.h> 12 #include <net/gre.h> 13 #include <net/pptp.h> 14 #include <net/tipc.h> 15 #include <linux/igmp.h> 16 #include <linux/icmp.h> 17 #include <linux/sctp.h> 18 #include <linux/dccp.h> 19 #include <linux/if_tunnel.h> 20 #include <linux/if_pppox.h> 21 #include <linux/ppp_defs.h> 22 #include <linux/stddef.h> 23 #include <linux/if_ether.h> 24 #include <linux/mpls.h> 25 #include <linux/tcp.h> 26 #include <net/flow_dissector.h> 27 #include <scsi/fc/fc_fcoe.h> 28 #include <uapi/linux/batadv_packet.h> 29 #include <linux/bpf.h> 30 31 static DEFINE_MUTEX(flow_dissector_mutex); 32 33 static void dissector_set_key(struct flow_dissector *flow_dissector, 34 enum flow_dissector_key_id key_id) 35 { 36 flow_dissector->used_keys |= (1 << key_id); 37 } 38 39 void skb_flow_dissector_init(struct flow_dissector *flow_dissector, 40 const struct flow_dissector_key *key, 41 unsigned int key_count) 42 { 43 unsigned int i; 44 45 memset(flow_dissector, 0, sizeof(*flow_dissector)); 46 47 for (i = 0; i < key_count; i++, key++) { 48 /* User should make sure that every key target offset is withing 49 * boundaries of unsigned short. 50 */ 51 BUG_ON(key->offset > USHRT_MAX); 52 BUG_ON(dissector_uses_key(flow_dissector, 53 key->key_id)); 54 55 dissector_set_key(flow_dissector, key->key_id); 56 flow_dissector->offset[key->key_id] = key->offset; 57 } 58 59 /* Ensure that the dissector always includes control and basic key. 60 * That way we are able to avoid handling lack of these in fast path. 61 */ 62 BUG_ON(!dissector_uses_key(flow_dissector, 63 FLOW_DISSECTOR_KEY_CONTROL)); 64 BUG_ON(!dissector_uses_key(flow_dissector, 65 FLOW_DISSECTOR_KEY_BASIC)); 66 } 67 EXPORT_SYMBOL(skb_flow_dissector_init); 68 69 int skb_flow_dissector_prog_query(const union bpf_attr *attr, 70 union bpf_attr __user *uattr) 71 { 72 __u32 __user *prog_ids = u64_to_user_ptr(attr->query.prog_ids); 73 u32 prog_id, prog_cnt = 0, flags = 0; 74 struct bpf_prog *attached; 75 struct net *net; 76 77 if (attr->query.query_flags) 78 return -EINVAL; 79 80 net = get_net_ns_by_fd(attr->query.target_fd); 81 if (IS_ERR(net)) 82 return PTR_ERR(net); 83 84 rcu_read_lock(); 85 attached = rcu_dereference(net->flow_dissector_prog); 86 if (attached) { 87 prog_cnt = 1; 88 prog_id = attached->aux->id; 89 } 90 rcu_read_unlock(); 91 92 put_net(net); 93 94 if (copy_to_user(&uattr->query.attach_flags, &flags, sizeof(flags))) 95 return -EFAULT; 96 if (copy_to_user(&uattr->query.prog_cnt, &prog_cnt, sizeof(prog_cnt))) 97 return -EFAULT; 98 99 if (!attr->query.prog_cnt || !prog_ids || !prog_cnt) 100 return 0; 101 102 if (copy_to_user(prog_ids, &prog_id, sizeof(u32))) 103 return -EFAULT; 104 105 return 0; 106 } 107 108 int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr, 109 struct bpf_prog *prog) 110 { 111 struct bpf_prog *attached; 112 struct net *net; 113 114 net = current->nsproxy->net_ns; 115 mutex_lock(&flow_dissector_mutex); 116 attached = rcu_dereference_protected(net->flow_dissector_prog, 117 lockdep_is_held(&flow_dissector_mutex)); 118 if (attached) { 119 /* Only one BPF program can be attached at a time */ 120 mutex_unlock(&flow_dissector_mutex); 121 return -EEXIST; 122 } 123 rcu_assign_pointer(net->flow_dissector_prog, prog); 124 mutex_unlock(&flow_dissector_mutex); 125 return 0; 126 } 127 128 int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr) 129 { 130 struct bpf_prog *attached; 131 struct net *net; 132 133 net = current->nsproxy->net_ns; 134 mutex_lock(&flow_dissector_mutex); 135 attached = rcu_dereference_protected(net->flow_dissector_prog, 136 lockdep_is_held(&flow_dissector_mutex)); 137 if (!attached) { 138 mutex_unlock(&flow_dissector_mutex); 139 return -ENOENT; 140 } 141 bpf_prog_put(attached); 142 RCU_INIT_POINTER(net->flow_dissector_prog, NULL); 143 mutex_unlock(&flow_dissector_mutex); 144 return 0; 145 } 146 /** 147 * skb_flow_get_be16 - extract be16 entity 148 * @skb: sk_buff to extract from 149 * @poff: offset to extract at 150 * @data: raw buffer pointer to the packet 151 * @hlen: packet header length 152 * 153 * The function will try to retrieve a be32 entity at 154 * offset poff 155 */ 156 static __be16 skb_flow_get_be16(const struct sk_buff *skb, int poff, 157 void *data, int hlen) 158 { 159 __be16 *u, _u; 160 161 u = __skb_header_pointer(skb, poff, sizeof(_u), data, hlen, &_u); 162 if (u) 163 return *u; 164 165 return 0; 166 } 167 168 /** 169 * __skb_flow_get_ports - extract the upper layer ports and return them 170 * @skb: sk_buff to extract the ports from 171 * @thoff: transport header offset 172 * @ip_proto: protocol for which to get port offset 173 * @data: raw buffer pointer to the packet, if NULL use skb->data 174 * @hlen: packet header length, if @data is NULL use skb_headlen(skb) 175 * 176 * The function will try to retrieve the ports at offset thoff + poff where poff 177 * is the protocol port offset returned from proto_ports_offset 178 */ 179 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto, 180 void *data, int hlen) 181 { 182 int poff = proto_ports_offset(ip_proto); 183 184 if (!data) { 185 data = skb->data; 186 hlen = skb_headlen(skb); 187 } 188 189 if (poff >= 0) { 190 __be32 *ports, _ports; 191 192 ports = __skb_header_pointer(skb, thoff + poff, 193 sizeof(_ports), data, hlen, &_ports); 194 if (ports) 195 return *ports; 196 } 197 198 return 0; 199 } 200 EXPORT_SYMBOL(__skb_flow_get_ports); 201 202 void skb_flow_dissect_meta(const struct sk_buff *skb, 203 struct flow_dissector *flow_dissector, 204 void *target_container) 205 { 206 struct flow_dissector_key_meta *meta; 207 208 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_META)) 209 return; 210 211 meta = skb_flow_dissector_target(flow_dissector, 212 FLOW_DISSECTOR_KEY_META, 213 target_container); 214 meta->ingress_ifindex = skb->skb_iif; 215 } 216 EXPORT_SYMBOL(skb_flow_dissect_meta); 217 218 static void 219 skb_flow_dissect_set_enc_addr_type(enum flow_dissector_key_id type, 220 struct flow_dissector *flow_dissector, 221 void *target_container) 222 { 223 struct flow_dissector_key_control *ctrl; 224 225 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_CONTROL)) 226 return; 227 228 ctrl = skb_flow_dissector_target(flow_dissector, 229 FLOW_DISSECTOR_KEY_ENC_CONTROL, 230 target_container); 231 ctrl->addr_type = type; 232 } 233 234 void 235 skb_flow_dissect_tunnel_info(const struct sk_buff *skb, 236 struct flow_dissector *flow_dissector, 237 void *target_container) 238 { 239 struct ip_tunnel_info *info; 240 struct ip_tunnel_key *key; 241 242 /* A quick check to see if there might be something to do. */ 243 if (!dissector_uses_key(flow_dissector, 244 FLOW_DISSECTOR_KEY_ENC_KEYID) && 245 !dissector_uses_key(flow_dissector, 246 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) && 247 !dissector_uses_key(flow_dissector, 248 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) && 249 !dissector_uses_key(flow_dissector, 250 FLOW_DISSECTOR_KEY_ENC_CONTROL) && 251 !dissector_uses_key(flow_dissector, 252 FLOW_DISSECTOR_KEY_ENC_PORTS) && 253 !dissector_uses_key(flow_dissector, 254 FLOW_DISSECTOR_KEY_ENC_IP) && 255 !dissector_uses_key(flow_dissector, 256 FLOW_DISSECTOR_KEY_ENC_OPTS)) 257 return; 258 259 info = skb_tunnel_info(skb); 260 if (!info) 261 return; 262 263 key = &info->key; 264 265 switch (ip_tunnel_info_af(info)) { 266 case AF_INET: 267 skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV4_ADDRS, 268 flow_dissector, 269 target_container); 270 if (dissector_uses_key(flow_dissector, 271 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) { 272 struct flow_dissector_key_ipv4_addrs *ipv4; 273 274 ipv4 = skb_flow_dissector_target(flow_dissector, 275 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS, 276 target_container); 277 ipv4->src = key->u.ipv4.src; 278 ipv4->dst = key->u.ipv4.dst; 279 } 280 break; 281 case AF_INET6: 282 skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV6_ADDRS, 283 flow_dissector, 284 target_container); 285 if (dissector_uses_key(flow_dissector, 286 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) { 287 struct flow_dissector_key_ipv6_addrs *ipv6; 288 289 ipv6 = skb_flow_dissector_target(flow_dissector, 290 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS, 291 target_container); 292 ipv6->src = key->u.ipv6.src; 293 ipv6->dst = key->u.ipv6.dst; 294 } 295 break; 296 } 297 298 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) { 299 struct flow_dissector_key_keyid *keyid; 300 301 keyid = skb_flow_dissector_target(flow_dissector, 302 FLOW_DISSECTOR_KEY_ENC_KEYID, 303 target_container); 304 keyid->keyid = tunnel_id_to_key32(key->tun_id); 305 } 306 307 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) { 308 struct flow_dissector_key_ports *tp; 309 310 tp = skb_flow_dissector_target(flow_dissector, 311 FLOW_DISSECTOR_KEY_ENC_PORTS, 312 target_container); 313 tp->src = key->tp_src; 314 tp->dst = key->tp_dst; 315 } 316 317 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_IP)) { 318 struct flow_dissector_key_ip *ip; 319 320 ip = skb_flow_dissector_target(flow_dissector, 321 FLOW_DISSECTOR_KEY_ENC_IP, 322 target_container); 323 ip->tos = key->tos; 324 ip->ttl = key->ttl; 325 } 326 327 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_OPTS)) { 328 struct flow_dissector_key_enc_opts *enc_opt; 329 330 enc_opt = skb_flow_dissector_target(flow_dissector, 331 FLOW_DISSECTOR_KEY_ENC_OPTS, 332 target_container); 333 334 if (info->options_len) { 335 enc_opt->len = info->options_len; 336 ip_tunnel_info_opts_get(enc_opt->data, info); 337 enc_opt->dst_opt_type = info->key.tun_flags & 338 TUNNEL_OPTIONS_PRESENT; 339 } 340 } 341 } 342 EXPORT_SYMBOL(skb_flow_dissect_tunnel_info); 343 344 static enum flow_dissect_ret 345 __skb_flow_dissect_mpls(const struct sk_buff *skb, 346 struct flow_dissector *flow_dissector, 347 void *target_container, void *data, int nhoff, int hlen) 348 { 349 struct flow_dissector_key_keyid *key_keyid; 350 struct mpls_label *hdr, _hdr[2]; 351 u32 entry, label; 352 353 if (!dissector_uses_key(flow_dissector, 354 FLOW_DISSECTOR_KEY_MPLS_ENTROPY) && 355 !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) 356 return FLOW_DISSECT_RET_OUT_GOOD; 357 358 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, 359 hlen, &_hdr); 360 if (!hdr) 361 return FLOW_DISSECT_RET_OUT_BAD; 362 363 entry = ntohl(hdr[0].entry); 364 label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT; 365 366 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) { 367 struct flow_dissector_key_mpls *key_mpls; 368 369 key_mpls = skb_flow_dissector_target(flow_dissector, 370 FLOW_DISSECTOR_KEY_MPLS, 371 target_container); 372 key_mpls->mpls_label = label; 373 key_mpls->mpls_ttl = (entry & MPLS_LS_TTL_MASK) 374 >> MPLS_LS_TTL_SHIFT; 375 key_mpls->mpls_tc = (entry & MPLS_LS_TC_MASK) 376 >> MPLS_LS_TC_SHIFT; 377 key_mpls->mpls_bos = (entry & MPLS_LS_S_MASK) 378 >> MPLS_LS_S_SHIFT; 379 } 380 381 if (label == MPLS_LABEL_ENTROPY) { 382 key_keyid = skb_flow_dissector_target(flow_dissector, 383 FLOW_DISSECTOR_KEY_MPLS_ENTROPY, 384 target_container); 385 key_keyid->keyid = hdr[1].entry & htonl(MPLS_LS_LABEL_MASK); 386 } 387 return FLOW_DISSECT_RET_OUT_GOOD; 388 } 389 390 static enum flow_dissect_ret 391 __skb_flow_dissect_arp(const struct sk_buff *skb, 392 struct flow_dissector *flow_dissector, 393 void *target_container, void *data, int nhoff, int hlen) 394 { 395 struct flow_dissector_key_arp *key_arp; 396 struct { 397 unsigned char ar_sha[ETH_ALEN]; 398 unsigned char ar_sip[4]; 399 unsigned char ar_tha[ETH_ALEN]; 400 unsigned char ar_tip[4]; 401 } *arp_eth, _arp_eth; 402 const struct arphdr *arp; 403 struct arphdr _arp; 404 405 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP)) 406 return FLOW_DISSECT_RET_OUT_GOOD; 407 408 arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data, 409 hlen, &_arp); 410 if (!arp) 411 return FLOW_DISSECT_RET_OUT_BAD; 412 413 if (arp->ar_hrd != htons(ARPHRD_ETHER) || 414 arp->ar_pro != htons(ETH_P_IP) || 415 arp->ar_hln != ETH_ALEN || 416 arp->ar_pln != 4 || 417 (arp->ar_op != htons(ARPOP_REPLY) && 418 arp->ar_op != htons(ARPOP_REQUEST))) 419 return FLOW_DISSECT_RET_OUT_BAD; 420 421 arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp), 422 sizeof(_arp_eth), data, 423 hlen, &_arp_eth); 424 if (!arp_eth) 425 return FLOW_DISSECT_RET_OUT_BAD; 426 427 key_arp = skb_flow_dissector_target(flow_dissector, 428 FLOW_DISSECTOR_KEY_ARP, 429 target_container); 430 431 memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip)); 432 memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip)); 433 434 /* Only store the lower byte of the opcode; 435 * this covers ARPOP_REPLY and ARPOP_REQUEST. 436 */ 437 key_arp->op = ntohs(arp->ar_op) & 0xff; 438 439 ether_addr_copy(key_arp->sha, arp_eth->ar_sha); 440 ether_addr_copy(key_arp->tha, arp_eth->ar_tha); 441 442 return FLOW_DISSECT_RET_OUT_GOOD; 443 } 444 445 static enum flow_dissect_ret 446 __skb_flow_dissect_gre(const struct sk_buff *skb, 447 struct flow_dissector_key_control *key_control, 448 struct flow_dissector *flow_dissector, 449 void *target_container, void *data, 450 __be16 *p_proto, int *p_nhoff, int *p_hlen, 451 unsigned int flags) 452 { 453 struct flow_dissector_key_keyid *key_keyid; 454 struct gre_base_hdr *hdr, _hdr; 455 int offset = 0; 456 u16 gre_ver; 457 458 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), 459 data, *p_hlen, &_hdr); 460 if (!hdr) 461 return FLOW_DISSECT_RET_OUT_BAD; 462 463 /* Only look inside GRE without routing */ 464 if (hdr->flags & GRE_ROUTING) 465 return FLOW_DISSECT_RET_OUT_GOOD; 466 467 /* Only look inside GRE for version 0 and 1 */ 468 gre_ver = ntohs(hdr->flags & GRE_VERSION); 469 if (gre_ver > 1) 470 return FLOW_DISSECT_RET_OUT_GOOD; 471 472 *p_proto = hdr->protocol; 473 if (gre_ver) { 474 /* Version1 must be PPTP, and check the flags */ 475 if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY))) 476 return FLOW_DISSECT_RET_OUT_GOOD; 477 } 478 479 offset += sizeof(struct gre_base_hdr); 480 481 if (hdr->flags & GRE_CSUM) 482 offset += FIELD_SIZEOF(struct gre_full_hdr, csum) + 483 FIELD_SIZEOF(struct gre_full_hdr, reserved1); 484 485 if (hdr->flags & GRE_KEY) { 486 const __be32 *keyid; 487 __be32 _keyid; 488 489 keyid = __skb_header_pointer(skb, *p_nhoff + offset, 490 sizeof(_keyid), 491 data, *p_hlen, &_keyid); 492 if (!keyid) 493 return FLOW_DISSECT_RET_OUT_BAD; 494 495 if (dissector_uses_key(flow_dissector, 496 FLOW_DISSECTOR_KEY_GRE_KEYID)) { 497 key_keyid = skb_flow_dissector_target(flow_dissector, 498 FLOW_DISSECTOR_KEY_GRE_KEYID, 499 target_container); 500 if (gre_ver == 0) 501 key_keyid->keyid = *keyid; 502 else 503 key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK; 504 } 505 offset += FIELD_SIZEOF(struct gre_full_hdr, key); 506 } 507 508 if (hdr->flags & GRE_SEQ) 509 offset += FIELD_SIZEOF(struct pptp_gre_header, seq); 510 511 if (gre_ver == 0) { 512 if (*p_proto == htons(ETH_P_TEB)) { 513 const struct ethhdr *eth; 514 struct ethhdr _eth; 515 516 eth = __skb_header_pointer(skb, *p_nhoff + offset, 517 sizeof(_eth), 518 data, *p_hlen, &_eth); 519 if (!eth) 520 return FLOW_DISSECT_RET_OUT_BAD; 521 *p_proto = eth->h_proto; 522 offset += sizeof(*eth); 523 524 /* Cap headers that we access via pointers at the 525 * end of the Ethernet header as our maximum alignment 526 * at that point is only 2 bytes. 527 */ 528 if (NET_IP_ALIGN) 529 *p_hlen = *p_nhoff + offset; 530 } 531 } else { /* version 1, must be PPTP */ 532 u8 _ppp_hdr[PPP_HDRLEN]; 533 u8 *ppp_hdr; 534 535 if (hdr->flags & GRE_ACK) 536 offset += FIELD_SIZEOF(struct pptp_gre_header, ack); 537 538 ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset, 539 sizeof(_ppp_hdr), 540 data, *p_hlen, _ppp_hdr); 541 if (!ppp_hdr) 542 return FLOW_DISSECT_RET_OUT_BAD; 543 544 switch (PPP_PROTOCOL(ppp_hdr)) { 545 case PPP_IP: 546 *p_proto = htons(ETH_P_IP); 547 break; 548 case PPP_IPV6: 549 *p_proto = htons(ETH_P_IPV6); 550 break; 551 default: 552 /* Could probably catch some more like MPLS */ 553 break; 554 } 555 556 offset += PPP_HDRLEN; 557 } 558 559 *p_nhoff += offset; 560 key_control->flags |= FLOW_DIS_ENCAPSULATION; 561 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) 562 return FLOW_DISSECT_RET_OUT_GOOD; 563 564 return FLOW_DISSECT_RET_PROTO_AGAIN; 565 } 566 567 /** 568 * __skb_flow_dissect_batadv() - dissect batman-adv header 569 * @skb: sk_buff to with the batman-adv header 570 * @key_control: flow dissectors control key 571 * @data: raw buffer pointer to the packet, if NULL use skb->data 572 * @p_proto: pointer used to update the protocol to process next 573 * @p_nhoff: pointer used to update inner network header offset 574 * @hlen: packet header length 575 * @flags: any combination of FLOW_DISSECTOR_F_* 576 * 577 * ETH_P_BATMAN packets are tried to be dissected. Only 578 * &struct batadv_unicast packets are actually processed because they contain an 579 * inner ethernet header and are usually followed by actual network header. This 580 * allows the flow dissector to continue processing the packet. 581 * 582 * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found, 583 * FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation, 584 * otherwise FLOW_DISSECT_RET_OUT_BAD 585 */ 586 static enum flow_dissect_ret 587 __skb_flow_dissect_batadv(const struct sk_buff *skb, 588 struct flow_dissector_key_control *key_control, 589 void *data, __be16 *p_proto, int *p_nhoff, int hlen, 590 unsigned int flags) 591 { 592 struct { 593 struct batadv_unicast_packet batadv_unicast; 594 struct ethhdr eth; 595 } *hdr, _hdr; 596 597 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen, 598 &_hdr); 599 if (!hdr) 600 return FLOW_DISSECT_RET_OUT_BAD; 601 602 if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION) 603 return FLOW_DISSECT_RET_OUT_BAD; 604 605 if (hdr->batadv_unicast.packet_type != BATADV_UNICAST) 606 return FLOW_DISSECT_RET_OUT_BAD; 607 608 *p_proto = hdr->eth.h_proto; 609 *p_nhoff += sizeof(*hdr); 610 611 key_control->flags |= FLOW_DIS_ENCAPSULATION; 612 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) 613 return FLOW_DISSECT_RET_OUT_GOOD; 614 615 return FLOW_DISSECT_RET_PROTO_AGAIN; 616 } 617 618 static void 619 __skb_flow_dissect_tcp(const struct sk_buff *skb, 620 struct flow_dissector *flow_dissector, 621 void *target_container, void *data, int thoff, int hlen) 622 { 623 struct flow_dissector_key_tcp *key_tcp; 624 struct tcphdr *th, _th; 625 626 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP)) 627 return; 628 629 th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th); 630 if (!th) 631 return; 632 633 if (unlikely(__tcp_hdrlen(th) < sizeof(_th))) 634 return; 635 636 key_tcp = skb_flow_dissector_target(flow_dissector, 637 FLOW_DISSECTOR_KEY_TCP, 638 target_container); 639 key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF)); 640 } 641 642 static void 643 __skb_flow_dissect_ipv4(const struct sk_buff *skb, 644 struct flow_dissector *flow_dissector, 645 void *target_container, void *data, const struct iphdr *iph) 646 { 647 struct flow_dissector_key_ip *key_ip; 648 649 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP)) 650 return; 651 652 key_ip = skb_flow_dissector_target(flow_dissector, 653 FLOW_DISSECTOR_KEY_IP, 654 target_container); 655 key_ip->tos = iph->tos; 656 key_ip->ttl = iph->ttl; 657 } 658 659 static void 660 __skb_flow_dissect_ipv6(const struct sk_buff *skb, 661 struct flow_dissector *flow_dissector, 662 void *target_container, void *data, const struct ipv6hdr *iph) 663 { 664 struct flow_dissector_key_ip *key_ip; 665 666 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP)) 667 return; 668 669 key_ip = skb_flow_dissector_target(flow_dissector, 670 FLOW_DISSECTOR_KEY_IP, 671 target_container); 672 key_ip->tos = ipv6_get_dsfield(iph); 673 key_ip->ttl = iph->hop_limit; 674 } 675 676 /* Maximum number of protocol headers that can be parsed in 677 * __skb_flow_dissect 678 */ 679 #define MAX_FLOW_DISSECT_HDRS 15 680 681 static bool skb_flow_dissect_allowed(int *num_hdrs) 682 { 683 ++*num_hdrs; 684 685 return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS); 686 } 687 688 static void __skb_flow_bpf_to_target(const struct bpf_flow_keys *flow_keys, 689 struct flow_dissector *flow_dissector, 690 void *target_container) 691 { 692 struct flow_dissector_key_control *key_control; 693 struct flow_dissector_key_basic *key_basic; 694 struct flow_dissector_key_addrs *key_addrs; 695 struct flow_dissector_key_ports *key_ports; 696 697 key_control = skb_flow_dissector_target(flow_dissector, 698 FLOW_DISSECTOR_KEY_CONTROL, 699 target_container); 700 key_control->thoff = flow_keys->thoff; 701 if (flow_keys->is_frag) 702 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 703 if (flow_keys->is_first_frag) 704 key_control->flags |= FLOW_DIS_FIRST_FRAG; 705 if (flow_keys->is_encap) 706 key_control->flags |= FLOW_DIS_ENCAPSULATION; 707 708 key_basic = skb_flow_dissector_target(flow_dissector, 709 FLOW_DISSECTOR_KEY_BASIC, 710 target_container); 711 key_basic->n_proto = flow_keys->n_proto; 712 key_basic->ip_proto = flow_keys->ip_proto; 713 714 if (flow_keys->addr_proto == ETH_P_IP && 715 dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) { 716 key_addrs = skb_flow_dissector_target(flow_dissector, 717 FLOW_DISSECTOR_KEY_IPV4_ADDRS, 718 target_container); 719 key_addrs->v4addrs.src = flow_keys->ipv4_src; 720 key_addrs->v4addrs.dst = flow_keys->ipv4_dst; 721 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; 722 } else if (flow_keys->addr_proto == ETH_P_IPV6 && 723 dissector_uses_key(flow_dissector, 724 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) { 725 key_addrs = skb_flow_dissector_target(flow_dissector, 726 FLOW_DISSECTOR_KEY_IPV6_ADDRS, 727 target_container); 728 memcpy(&key_addrs->v6addrs, &flow_keys->ipv6_src, 729 sizeof(key_addrs->v6addrs)); 730 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 731 } 732 733 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS)) { 734 key_ports = skb_flow_dissector_target(flow_dissector, 735 FLOW_DISSECTOR_KEY_PORTS, 736 target_container); 737 key_ports->src = flow_keys->sport; 738 key_ports->dst = flow_keys->dport; 739 } 740 } 741 742 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx, 743 __be16 proto, int nhoff, int hlen) 744 { 745 struct bpf_flow_keys *flow_keys = ctx->flow_keys; 746 u32 result; 747 748 /* Pass parameters to the BPF program */ 749 memset(flow_keys, 0, sizeof(*flow_keys)); 750 flow_keys->n_proto = proto; 751 flow_keys->nhoff = nhoff; 752 flow_keys->thoff = flow_keys->nhoff; 753 754 preempt_disable(); 755 result = BPF_PROG_RUN(prog, ctx); 756 preempt_enable(); 757 758 flow_keys->nhoff = clamp_t(u16, flow_keys->nhoff, nhoff, hlen); 759 flow_keys->thoff = clamp_t(u16, flow_keys->thoff, 760 flow_keys->nhoff, hlen); 761 762 return result == BPF_OK; 763 } 764 765 /** 766 * __skb_flow_dissect - extract the flow_keys struct and return it 767 * @net: associated network namespace, derived from @skb if NULL 768 * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified 769 * @flow_dissector: list of keys to dissect 770 * @target_container: target structure to put dissected values into 771 * @data: raw buffer pointer to the packet, if NULL use skb->data 772 * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol 773 * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb) 774 * @hlen: packet header length, if @data is NULL use skb_headlen(skb) 775 * @flags: flags that control the dissection process, e.g. 776 * FLOW_DISSECTOR_F_STOP_AT_ENCAP. 777 * 778 * The function will try to retrieve individual keys into target specified 779 * by flow_dissector from either the skbuff or a raw buffer specified by the 780 * rest parameters. 781 * 782 * Caller must take care of zeroing target container memory. 783 */ 784 bool __skb_flow_dissect(const struct net *net, 785 const struct sk_buff *skb, 786 struct flow_dissector *flow_dissector, 787 void *target_container, 788 void *data, __be16 proto, int nhoff, int hlen, 789 unsigned int flags) 790 { 791 struct flow_dissector_key_control *key_control; 792 struct flow_dissector_key_basic *key_basic; 793 struct flow_dissector_key_addrs *key_addrs; 794 struct flow_dissector_key_ports *key_ports; 795 struct flow_dissector_key_icmp *key_icmp; 796 struct flow_dissector_key_tags *key_tags; 797 struct flow_dissector_key_vlan *key_vlan; 798 struct bpf_prog *attached = NULL; 799 enum flow_dissect_ret fdret; 800 enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX; 801 int num_hdrs = 0; 802 u8 ip_proto = 0; 803 bool ret; 804 805 if (!data) { 806 data = skb->data; 807 proto = skb_vlan_tag_present(skb) ? 808 skb->vlan_proto : skb->protocol; 809 nhoff = skb_network_offset(skb); 810 hlen = skb_headlen(skb); 811 #if IS_ENABLED(CONFIG_NET_DSA) 812 if (unlikely(skb->dev && netdev_uses_dsa(skb->dev))) { 813 const struct dsa_device_ops *ops; 814 int offset; 815 816 ops = skb->dev->dsa_ptr->tag_ops; 817 if (ops->flow_dissect && 818 !ops->flow_dissect(skb, &proto, &offset)) { 819 hlen -= offset; 820 nhoff += offset; 821 } 822 } 823 #endif 824 } 825 826 /* It is ensured by skb_flow_dissector_init() that control key will 827 * be always present. 828 */ 829 key_control = skb_flow_dissector_target(flow_dissector, 830 FLOW_DISSECTOR_KEY_CONTROL, 831 target_container); 832 833 /* It is ensured by skb_flow_dissector_init() that basic key will 834 * be always present. 835 */ 836 key_basic = skb_flow_dissector_target(flow_dissector, 837 FLOW_DISSECTOR_KEY_BASIC, 838 target_container); 839 840 if (skb) { 841 if (!net) { 842 if (skb->dev) 843 net = dev_net(skb->dev); 844 else if (skb->sk) 845 net = sock_net(skb->sk); 846 } 847 } 848 849 WARN_ON_ONCE(!net); 850 if (net) { 851 rcu_read_lock(); 852 attached = rcu_dereference(net->flow_dissector_prog); 853 854 if (attached) { 855 struct bpf_flow_keys flow_keys; 856 struct bpf_flow_dissector ctx = { 857 .flow_keys = &flow_keys, 858 .data = data, 859 .data_end = data + hlen, 860 }; 861 __be16 n_proto = proto; 862 863 if (skb) { 864 ctx.skb = skb; 865 /* we can't use 'proto' in the skb case 866 * because it might be set to skb->vlan_proto 867 * which has been pulled from the data 868 */ 869 n_proto = skb->protocol; 870 } 871 872 ret = bpf_flow_dissect(attached, &ctx, n_proto, nhoff, 873 hlen); 874 __skb_flow_bpf_to_target(&flow_keys, flow_dissector, 875 target_container); 876 rcu_read_unlock(); 877 return ret; 878 } 879 rcu_read_unlock(); 880 } 881 882 if (dissector_uses_key(flow_dissector, 883 FLOW_DISSECTOR_KEY_ETH_ADDRS)) { 884 struct ethhdr *eth = eth_hdr(skb); 885 struct flow_dissector_key_eth_addrs *key_eth_addrs; 886 887 key_eth_addrs = skb_flow_dissector_target(flow_dissector, 888 FLOW_DISSECTOR_KEY_ETH_ADDRS, 889 target_container); 890 memcpy(key_eth_addrs, ð->h_dest, sizeof(*key_eth_addrs)); 891 } 892 893 proto_again: 894 fdret = FLOW_DISSECT_RET_CONTINUE; 895 896 switch (proto) { 897 case htons(ETH_P_IP): { 898 const struct iphdr *iph; 899 struct iphdr _iph; 900 901 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph); 902 if (!iph || iph->ihl < 5) { 903 fdret = FLOW_DISSECT_RET_OUT_BAD; 904 break; 905 } 906 907 nhoff += iph->ihl * 4; 908 909 ip_proto = iph->protocol; 910 911 if (dissector_uses_key(flow_dissector, 912 FLOW_DISSECTOR_KEY_IPV4_ADDRS)) { 913 key_addrs = skb_flow_dissector_target(flow_dissector, 914 FLOW_DISSECTOR_KEY_IPV4_ADDRS, 915 target_container); 916 917 memcpy(&key_addrs->v4addrs, &iph->saddr, 918 sizeof(key_addrs->v4addrs)); 919 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; 920 } 921 922 if (ip_is_fragment(iph)) { 923 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 924 925 if (iph->frag_off & htons(IP_OFFSET)) { 926 fdret = FLOW_DISSECT_RET_OUT_GOOD; 927 break; 928 } else { 929 key_control->flags |= FLOW_DIS_FIRST_FRAG; 930 if (!(flags & 931 FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) { 932 fdret = FLOW_DISSECT_RET_OUT_GOOD; 933 break; 934 } 935 } 936 } 937 938 __skb_flow_dissect_ipv4(skb, flow_dissector, 939 target_container, data, iph); 940 941 break; 942 } 943 case htons(ETH_P_IPV6): { 944 const struct ipv6hdr *iph; 945 struct ipv6hdr _iph; 946 947 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph); 948 if (!iph) { 949 fdret = FLOW_DISSECT_RET_OUT_BAD; 950 break; 951 } 952 953 ip_proto = iph->nexthdr; 954 nhoff += sizeof(struct ipv6hdr); 955 956 if (dissector_uses_key(flow_dissector, 957 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) { 958 key_addrs = skb_flow_dissector_target(flow_dissector, 959 FLOW_DISSECTOR_KEY_IPV6_ADDRS, 960 target_container); 961 962 memcpy(&key_addrs->v6addrs, &iph->saddr, 963 sizeof(key_addrs->v6addrs)); 964 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 965 } 966 967 if ((dissector_uses_key(flow_dissector, 968 FLOW_DISSECTOR_KEY_FLOW_LABEL) || 969 (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) && 970 ip6_flowlabel(iph)) { 971 __be32 flow_label = ip6_flowlabel(iph); 972 973 if (dissector_uses_key(flow_dissector, 974 FLOW_DISSECTOR_KEY_FLOW_LABEL)) { 975 key_tags = skb_flow_dissector_target(flow_dissector, 976 FLOW_DISSECTOR_KEY_FLOW_LABEL, 977 target_container); 978 key_tags->flow_label = ntohl(flow_label); 979 } 980 if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) { 981 fdret = FLOW_DISSECT_RET_OUT_GOOD; 982 break; 983 } 984 } 985 986 __skb_flow_dissect_ipv6(skb, flow_dissector, 987 target_container, data, iph); 988 989 break; 990 } 991 case htons(ETH_P_8021AD): 992 case htons(ETH_P_8021Q): { 993 const struct vlan_hdr *vlan = NULL; 994 struct vlan_hdr _vlan; 995 __be16 saved_vlan_tpid = proto; 996 997 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX && 998 skb && skb_vlan_tag_present(skb)) { 999 proto = skb->protocol; 1000 } else { 1001 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan), 1002 data, hlen, &_vlan); 1003 if (!vlan) { 1004 fdret = FLOW_DISSECT_RET_OUT_BAD; 1005 break; 1006 } 1007 1008 proto = vlan->h_vlan_encapsulated_proto; 1009 nhoff += sizeof(*vlan); 1010 } 1011 1012 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) { 1013 dissector_vlan = FLOW_DISSECTOR_KEY_VLAN; 1014 } else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) { 1015 dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN; 1016 } else { 1017 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1018 break; 1019 } 1020 1021 if (dissector_uses_key(flow_dissector, dissector_vlan)) { 1022 key_vlan = skb_flow_dissector_target(flow_dissector, 1023 dissector_vlan, 1024 target_container); 1025 1026 if (!vlan) { 1027 key_vlan->vlan_id = skb_vlan_tag_get_id(skb); 1028 key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb); 1029 } else { 1030 key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) & 1031 VLAN_VID_MASK; 1032 key_vlan->vlan_priority = 1033 (ntohs(vlan->h_vlan_TCI) & 1034 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT; 1035 } 1036 key_vlan->vlan_tpid = saved_vlan_tpid; 1037 } 1038 1039 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1040 break; 1041 } 1042 case htons(ETH_P_PPP_SES): { 1043 struct { 1044 struct pppoe_hdr hdr; 1045 __be16 proto; 1046 } *hdr, _hdr; 1047 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr); 1048 if (!hdr) { 1049 fdret = FLOW_DISSECT_RET_OUT_BAD; 1050 break; 1051 } 1052 1053 proto = hdr->proto; 1054 nhoff += PPPOE_SES_HLEN; 1055 switch (proto) { 1056 case htons(PPP_IP): 1057 proto = htons(ETH_P_IP); 1058 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1059 break; 1060 case htons(PPP_IPV6): 1061 proto = htons(ETH_P_IPV6); 1062 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1063 break; 1064 default: 1065 fdret = FLOW_DISSECT_RET_OUT_BAD; 1066 break; 1067 } 1068 break; 1069 } 1070 case htons(ETH_P_TIPC): { 1071 struct tipc_basic_hdr *hdr, _hdr; 1072 1073 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), 1074 data, hlen, &_hdr); 1075 if (!hdr) { 1076 fdret = FLOW_DISSECT_RET_OUT_BAD; 1077 break; 1078 } 1079 1080 if (dissector_uses_key(flow_dissector, 1081 FLOW_DISSECTOR_KEY_TIPC)) { 1082 key_addrs = skb_flow_dissector_target(flow_dissector, 1083 FLOW_DISSECTOR_KEY_TIPC, 1084 target_container); 1085 key_addrs->tipckey.key = tipc_hdr_rps_key(hdr); 1086 key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC; 1087 } 1088 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1089 break; 1090 } 1091 1092 case htons(ETH_P_MPLS_UC): 1093 case htons(ETH_P_MPLS_MC): 1094 fdret = __skb_flow_dissect_mpls(skb, flow_dissector, 1095 target_container, data, 1096 nhoff, hlen); 1097 break; 1098 case htons(ETH_P_FCOE): 1099 if ((hlen - nhoff) < FCOE_HEADER_LEN) { 1100 fdret = FLOW_DISSECT_RET_OUT_BAD; 1101 break; 1102 } 1103 1104 nhoff += FCOE_HEADER_LEN; 1105 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1106 break; 1107 1108 case htons(ETH_P_ARP): 1109 case htons(ETH_P_RARP): 1110 fdret = __skb_flow_dissect_arp(skb, flow_dissector, 1111 target_container, data, 1112 nhoff, hlen); 1113 break; 1114 1115 case htons(ETH_P_BATMAN): 1116 fdret = __skb_flow_dissect_batadv(skb, key_control, data, 1117 &proto, &nhoff, hlen, flags); 1118 break; 1119 1120 default: 1121 fdret = FLOW_DISSECT_RET_OUT_BAD; 1122 break; 1123 } 1124 1125 /* Process result of proto processing */ 1126 switch (fdret) { 1127 case FLOW_DISSECT_RET_OUT_GOOD: 1128 goto out_good; 1129 case FLOW_DISSECT_RET_PROTO_AGAIN: 1130 if (skb_flow_dissect_allowed(&num_hdrs)) 1131 goto proto_again; 1132 goto out_good; 1133 case FLOW_DISSECT_RET_CONTINUE: 1134 case FLOW_DISSECT_RET_IPPROTO_AGAIN: 1135 break; 1136 case FLOW_DISSECT_RET_OUT_BAD: 1137 default: 1138 goto out_bad; 1139 } 1140 1141 ip_proto_again: 1142 fdret = FLOW_DISSECT_RET_CONTINUE; 1143 1144 switch (ip_proto) { 1145 case IPPROTO_GRE: 1146 fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector, 1147 target_container, data, 1148 &proto, &nhoff, &hlen, flags); 1149 break; 1150 1151 case NEXTHDR_HOP: 1152 case NEXTHDR_ROUTING: 1153 case NEXTHDR_DEST: { 1154 u8 _opthdr[2], *opthdr; 1155 1156 if (proto != htons(ETH_P_IPV6)) 1157 break; 1158 1159 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr), 1160 data, hlen, &_opthdr); 1161 if (!opthdr) { 1162 fdret = FLOW_DISSECT_RET_OUT_BAD; 1163 break; 1164 } 1165 1166 ip_proto = opthdr[0]; 1167 nhoff += (opthdr[1] + 1) << 3; 1168 1169 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN; 1170 break; 1171 } 1172 case NEXTHDR_FRAGMENT: { 1173 struct frag_hdr _fh, *fh; 1174 1175 if (proto != htons(ETH_P_IPV6)) 1176 break; 1177 1178 fh = __skb_header_pointer(skb, nhoff, sizeof(_fh), 1179 data, hlen, &_fh); 1180 1181 if (!fh) { 1182 fdret = FLOW_DISSECT_RET_OUT_BAD; 1183 break; 1184 } 1185 1186 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 1187 1188 nhoff += sizeof(_fh); 1189 ip_proto = fh->nexthdr; 1190 1191 if (!(fh->frag_off & htons(IP6_OFFSET))) { 1192 key_control->flags |= FLOW_DIS_FIRST_FRAG; 1193 if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) { 1194 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN; 1195 break; 1196 } 1197 } 1198 1199 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1200 break; 1201 } 1202 case IPPROTO_IPIP: 1203 proto = htons(ETH_P_IP); 1204 1205 key_control->flags |= FLOW_DIS_ENCAPSULATION; 1206 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) { 1207 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1208 break; 1209 } 1210 1211 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1212 break; 1213 1214 case IPPROTO_IPV6: 1215 proto = htons(ETH_P_IPV6); 1216 1217 key_control->flags |= FLOW_DIS_ENCAPSULATION; 1218 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) { 1219 fdret = FLOW_DISSECT_RET_OUT_GOOD; 1220 break; 1221 } 1222 1223 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1224 break; 1225 1226 1227 case IPPROTO_MPLS: 1228 proto = htons(ETH_P_MPLS_UC); 1229 fdret = FLOW_DISSECT_RET_PROTO_AGAIN; 1230 break; 1231 1232 case IPPROTO_TCP: 1233 __skb_flow_dissect_tcp(skb, flow_dissector, target_container, 1234 data, nhoff, hlen); 1235 break; 1236 1237 default: 1238 break; 1239 } 1240 1241 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS) && 1242 !(key_control->flags & FLOW_DIS_IS_FRAGMENT)) { 1243 key_ports = skb_flow_dissector_target(flow_dissector, 1244 FLOW_DISSECTOR_KEY_PORTS, 1245 target_container); 1246 key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto, 1247 data, hlen); 1248 } 1249 1250 if (dissector_uses_key(flow_dissector, 1251 FLOW_DISSECTOR_KEY_ICMP)) { 1252 key_icmp = skb_flow_dissector_target(flow_dissector, 1253 FLOW_DISSECTOR_KEY_ICMP, 1254 target_container); 1255 key_icmp->icmp = skb_flow_get_be16(skb, nhoff, data, hlen); 1256 } 1257 1258 /* Process result of IP proto processing */ 1259 switch (fdret) { 1260 case FLOW_DISSECT_RET_PROTO_AGAIN: 1261 if (skb_flow_dissect_allowed(&num_hdrs)) 1262 goto proto_again; 1263 break; 1264 case FLOW_DISSECT_RET_IPPROTO_AGAIN: 1265 if (skb_flow_dissect_allowed(&num_hdrs)) 1266 goto ip_proto_again; 1267 break; 1268 case FLOW_DISSECT_RET_OUT_GOOD: 1269 case FLOW_DISSECT_RET_CONTINUE: 1270 break; 1271 case FLOW_DISSECT_RET_OUT_BAD: 1272 default: 1273 goto out_bad; 1274 } 1275 1276 out_good: 1277 ret = true; 1278 1279 out: 1280 key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen); 1281 key_basic->n_proto = proto; 1282 key_basic->ip_proto = ip_proto; 1283 1284 return ret; 1285 1286 out_bad: 1287 ret = false; 1288 goto out; 1289 } 1290 EXPORT_SYMBOL(__skb_flow_dissect); 1291 1292 static u32 hashrnd __read_mostly; 1293 static __always_inline void __flow_hash_secret_init(void) 1294 { 1295 net_get_random_once(&hashrnd, sizeof(hashrnd)); 1296 } 1297 1298 static __always_inline u32 __flow_hash_words(const u32 *words, u32 length, 1299 u32 keyval) 1300 { 1301 return jhash2(words, length, keyval); 1302 } 1303 1304 static inline const u32 *flow_keys_hash_start(const struct flow_keys *flow) 1305 { 1306 const void *p = flow; 1307 1308 BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % sizeof(u32)); 1309 return (const u32 *)(p + FLOW_KEYS_HASH_OFFSET); 1310 } 1311 1312 static inline size_t flow_keys_hash_length(const struct flow_keys *flow) 1313 { 1314 size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs); 1315 BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32)); 1316 BUILD_BUG_ON(offsetof(typeof(*flow), addrs) != 1317 sizeof(*flow) - sizeof(flow->addrs)); 1318 1319 switch (flow->control.addr_type) { 1320 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1321 diff -= sizeof(flow->addrs.v4addrs); 1322 break; 1323 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1324 diff -= sizeof(flow->addrs.v6addrs); 1325 break; 1326 case FLOW_DISSECTOR_KEY_TIPC: 1327 diff -= sizeof(flow->addrs.tipckey); 1328 break; 1329 } 1330 return (sizeof(*flow) - diff) / sizeof(u32); 1331 } 1332 1333 __be32 flow_get_u32_src(const struct flow_keys *flow) 1334 { 1335 switch (flow->control.addr_type) { 1336 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1337 return flow->addrs.v4addrs.src; 1338 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1339 return (__force __be32)ipv6_addr_hash( 1340 &flow->addrs.v6addrs.src); 1341 case FLOW_DISSECTOR_KEY_TIPC: 1342 return flow->addrs.tipckey.key; 1343 default: 1344 return 0; 1345 } 1346 } 1347 EXPORT_SYMBOL(flow_get_u32_src); 1348 1349 __be32 flow_get_u32_dst(const struct flow_keys *flow) 1350 { 1351 switch (flow->control.addr_type) { 1352 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1353 return flow->addrs.v4addrs.dst; 1354 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1355 return (__force __be32)ipv6_addr_hash( 1356 &flow->addrs.v6addrs.dst); 1357 default: 1358 return 0; 1359 } 1360 } 1361 EXPORT_SYMBOL(flow_get_u32_dst); 1362 1363 static inline void __flow_hash_consistentify(struct flow_keys *keys) 1364 { 1365 int addr_diff, i; 1366 1367 switch (keys->control.addr_type) { 1368 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 1369 addr_diff = (__force u32)keys->addrs.v4addrs.dst - 1370 (__force u32)keys->addrs.v4addrs.src; 1371 if ((addr_diff < 0) || 1372 (addr_diff == 0 && 1373 ((__force u16)keys->ports.dst < 1374 (__force u16)keys->ports.src))) { 1375 swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst); 1376 swap(keys->ports.src, keys->ports.dst); 1377 } 1378 break; 1379 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 1380 addr_diff = memcmp(&keys->addrs.v6addrs.dst, 1381 &keys->addrs.v6addrs.src, 1382 sizeof(keys->addrs.v6addrs.dst)); 1383 if ((addr_diff < 0) || 1384 (addr_diff == 0 && 1385 ((__force u16)keys->ports.dst < 1386 (__force u16)keys->ports.src))) { 1387 for (i = 0; i < 4; i++) 1388 swap(keys->addrs.v6addrs.src.s6_addr32[i], 1389 keys->addrs.v6addrs.dst.s6_addr32[i]); 1390 swap(keys->ports.src, keys->ports.dst); 1391 } 1392 break; 1393 } 1394 } 1395 1396 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval) 1397 { 1398 u32 hash; 1399 1400 __flow_hash_consistentify(keys); 1401 1402 hash = __flow_hash_words(flow_keys_hash_start(keys), 1403 flow_keys_hash_length(keys), keyval); 1404 if (!hash) 1405 hash = 1; 1406 1407 return hash; 1408 } 1409 1410 u32 flow_hash_from_keys(struct flow_keys *keys) 1411 { 1412 __flow_hash_secret_init(); 1413 return __flow_hash_from_keys(keys, hashrnd); 1414 } 1415 EXPORT_SYMBOL(flow_hash_from_keys); 1416 1417 static inline u32 ___skb_get_hash(const struct sk_buff *skb, 1418 struct flow_keys *keys, u32 keyval) 1419 { 1420 skb_flow_dissect_flow_keys(skb, keys, 1421 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 1422 1423 return __flow_hash_from_keys(keys, keyval); 1424 } 1425 1426 struct _flow_keys_digest_data { 1427 __be16 n_proto; 1428 u8 ip_proto; 1429 u8 padding; 1430 __be32 ports; 1431 __be32 src; 1432 __be32 dst; 1433 }; 1434 1435 void make_flow_keys_digest(struct flow_keys_digest *digest, 1436 const struct flow_keys *flow) 1437 { 1438 struct _flow_keys_digest_data *data = 1439 (struct _flow_keys_digest_data *)digest; 1440 1441 BUILD_BUG_ON(sizeof(*data) > sizeof(*digest)); 1442 1443 memset(digest, 0, sizeof(*digest)); 1444 1445 data->n_proto = flow->basic.n_proto; 1446 data->ip_proto = flow->basic.ip_proto; 1447 data->ports = flow->ports.ports; 1448 data->src = flow->addrs.v4addrs.src; 1449 data->dst = flow->addrs.v4addrs.dst; 1450 } 1451 EXPORT_SYMBOL(make_flow_keys_digest); 1452 1453 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly; 1454 1455 u32 __skb_get_hash_symmetric(const struct sk_buff *skb) 1456 { 1457 struct flow_keys keys; 1458 1459 __flow_hash_secret_init(); 1460 1461 memset(&keys, 0, sizeof(keys)); 1462 __skb_flow_dissect(NULL, skb, &flow_keys_dissector_symmetric, 1463 &keys, NULL, 0, 0, 0, 1464 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 1465 1466 return __flow_hash_from_keys(&keys, hashrnd); 1467 } 1468 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric); 1469 1470 /** 1471 * __skb_get_hash: calculate a flow hash 1472 * @skb: sk_buff to calculate flow hash from 1473 * 1474 * This function calculates a flow hash based on src/dst addresses 1475 * and src/dst port numbers. Sets hash in skb to non-zero hash value 1476 * on success, zero indicates no valid hash. Also, sets l4_hash in skb 1477 * if hash is a canonical 4-tuple hash over transport ports. 1478 */ 1479 void __skb_get_hash(struct sk_buff *skb) 1480 { 1481 struct flow_keys keys; 1482 u32 hash; 1483 1484 __flow_hash_secret_init(); 1485 1486 hash = ___skb_get_hash(skb, &keys, hashrnd); 1487 1488 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); 1489 } 1490 EXPORT_SYMBOL(__skb_get_hash); 1491 1492 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb) 1493 { 1494 struct flow_keys keys; 1495 1496 return ___skb_get_hash(skb, &keys, perturb); 1497 } 1498 EXPORT_SYMBOL(skb_get_hash_perturb); 1499 1500 u32 __skb_get_poff(const struct sk_buff *skb, void *data, 1501 const struct flow_keys_basic *keys, int hlen) 1502 { 1503 u32 poff = keys->control.thoff; 1504 1505 /* skip L4 headers for fragments after the first */ 1506 if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) && 1507 !(keys->control.flags & FLOW_DIS_FIRST_FRAG)) 1508 return poff; 1509 1510 switch (keys->basic.ip_proto) { 1511 case IPPROTO_TCP: { 1512 /* access doff as u8 to avoid unaligned access */ 1513 const u8 *doff; 1514 u8 _doff; 1515 1516 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff), 1517 data, hlen, &_doff); 1518 if (!doff) 1519 return poff; 1520 1521 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2); 1522 break; 1523 } 1524 case IPPROTO_UDP: 1525 case IPPROTO_UDPLITE: 1526 poff += sizeof(struct udphdr); 1527 break; 1528 /* For the rest, we do not really care about header 1529 * extensions at this point for now. 1530 */ 1531 case IPPROTO_ICMP: 1532 poff += sizeof(struct icmphdr); 1533 break; 1534 case IPPROTO_ICMPV6: 1535 poff += sizeof(struct icmp6hdr); 1536 break; 1537 case IPPROTO_IGMP: 1538 poff += sizeof(struct igmphdr); 1539 break; 1540 case IPPROTO_DCCP: 1541 poff += sizeof(struct dccp_hdr); 1542 break; 1543 case IPPROTO_SCTP: 1544 poff += sizeof(struct sctphdr); 1545 break; 1546 } 1547 1548 return poff; 1549 } 1550 1551 /** 1552 * skb_get_poff - get the offset to the payload 1553 * @skb: sk_buff to get the payload offset from 1554 * 1555 * The function will get the offset to the payload as far as it could 1556 * be dissected. The main user is currently BPF, so that we can dynamically 1557 * truncate packets without needing to push actual payload to the user 1558 * space and can analyze headers only, instead. 1559 */ 1560 u32 skb_get_poff(const struct sk_buff *skb) 1561 { 1562 struct flow_keys_basic keys; 1563 1564 if (!skb_flow_dissect_flow_keys_basic(NULL, skb, &keys, 1565 NULL, 0, 0, 0, 0)) 1566 return 0; 1567 1568 return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb)); 1569 } 1570 1571 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys) 1572 { 1573 memset(keys, 0, sizeof(*keys)); 1574 1575 memcpy(&keys->addrs.v6addrs.src, &fl6->saddr, 1576 sizeof(keys->addrs.v6addrs.src)); 1577 memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr, 1578 sizeof(keys->addrs.v6addrs.dst)); 1579 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 1580 keys->ports.src = fl6->fl6_sport; 1581 keys->ports.dst = fl6->fl6_dport; 1582 keys->keyid.keyid = fl6->fl6_gre_key; 1583 keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6); 1584 keys->basic.ip_proto = fl6->flowi6_proto; 1585 1586 return flow_hash_from_keys(keys); 1587 } 1588 EXPORT_SYMBOL(__get_hash_from_flowi6); 1589 1590 static const struct flow_dissector_key flow_keys_dissector_keys[] = { 1591 { 1592 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 1593 .offset = offsetof(struct flow_keys, control), 1594 }, 1595 { 1596 .key_id = FLOW_DISSECTOR_KEY_BASIC, 1597 .offset = offsetof(struct flow_keys, basic), 1598 }, 1599 { 1600 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS, 1601 .offset = offsetof(struct flow_keys, addrs.v4addrs), 1602 }, 1603 { 1604 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS, 1605 .offset = offsetof(struct flow_keys, addrs.v6addrs), 1606 }, 1607 { 1608 .key_id = FLOW_DISSECTOR_KEY_TIPC, 1609 .offset = offsetof(struct flow_keys, addrs.tipckey), 1610 }, 1611 { 1612 .key_id = FLOW_DISSECTOR_KEY_PORTS, 1613 .offset = offsetof(struct flow_keys, ports), 1614 }, 1615 { 1616 .key_id = FLOW_DISSECTOR_KEY_VLAN, 1617 .offset = offsetof(struct flow_keys, vlan), 1618 }, 1619 { 1620 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL, 1621 .offset = offsetof(struct flow_keys, tags), 1622 }, 1623 { 1624 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID, 1625 .offset = offsetof(struct flow_keys, keyid), 1626 }, 1627 }; 1628 1629 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = { 1630 { 1631 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 1632 .offset = offsetof(struct flow_keys, control), 1633 }, 1634 { 1635 .key_id = FLOW_DISSECTOR_KEY_BASIC, 1636 .offset = offsetof(struct flow_keys, basic), 1637 }, 1638 { 1639 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS, 1640 .offset = offsetof(struct flow_keys, addrs.v4addrs), 1641 }, 1642 { 1643 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS, 1644 .offset = offsetof(struct flow_keys, addrs.v6addrs), 1645 }, 1646 { 1647 .key_id = FLOW_DISSECTOR_KEY_PORTS, 1648 .offset = offsetof(struct flow_keys, ports), 1649 }, 1650 }; 1651 1652 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = { 1653 { 1654 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 1655 .offset = offsetof(struct flow_keys, control), 1656 }, 1657 { 1658 .key_id = FLOW_DISSECTOR_KEY_BASIC, 1659 .offset = offsetof(struct flow_keys, basic), 1660 }, 1661 }; 1662 1663 struct flow_dissector flow_keys_dissector __read_mostly; 1664 EXPORT_SYMBOL(flow_keys_dissector); 1665 1666 struct flow_dissector flow_keys_basic_dissector __read_mostly; 1667 EXPORT_SYMBOL(flow_keys_basic_dissector); 1668 1669 static int __init init_default_flow_dissectors(void) 1670 { 1671 skb_flow_dissector_init(&flow_keys_dissector, 1672 flow_keys_dissector_keys, 1673 ARRAY_SIZE(flow_keys_dissector_keys)); 1674 skb_flow_dissector_init(&flow_keys_dissector_symmetric, 1675 flow_keys_dissector_symmetric_keys, 1676 ARRAY_SIZE(flow_keys_dissector_symmetric_keys)); 1677 skb_flow_dissector_init(&flow_keys_basic_dissector, 1678 flow_keys_basic_dissector_keys, 1679 ARRAY_SIZE(flow_keys_basic_dissector_keys)); 1680 return 0; 1681 } 1682 1683 core_initcall(init_default_flow_dissectors); 1684