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/ip.h> 8 #include <net/ipv6.h> 9 #include <linux/igmp.h> 10 #include <linux/icmp.h> 11 #include <linux/sctp.h> 12 #include <linux/dccp.h> 13 #include <linux/if_tunnel.h> 14 #include <linux/if_pppox.h> 15 #include <linux/ppp_defs.h> 16 #include <linux/stddef.h> 17 #include <linux/if_ether.h> 18 #include <linux/mpls.h> 19 #include <net/flow_dissector.h> 20 #include <scsi/fc/fc_fcoe.h> 21 22 static bool dissector_uses_key(const struct flow_dissector *flow_dissector, 23 enum flow_dissector_key_id key_id) 24 { 25 return flow_dissector->used_keys & (1 << key_id); 26 } 27 28 static void dissector_set_key(struct flow_dissector *flow_dissector, 29 enum flow_dissector_key_id key_id) 30 { 31 flow_dissector->used_keys |= (1 << key_id); 32 } 33 34 static void *skb_flow_dissector_target(struct flow_dissector *flow_dissector, 35 enum flow_dissector_key_id key_id, 36 void *target_container) 37 { 38 return ((char *) target_container) + flow_dissector->offset[key_id]; 39 } 40 41 void skb_flow_dissector_init(struct flow_dissector *flow_dissector, 42 const struct flow_dissector_key *key, 43 unsigned int key_count) 44 { 45 unsigned int i; 46 47 memset(flow_dissector, 0, sizeof(*flow_dissector)); 48 49 for (i = 0; i < key_count; i++, key++) { 50 /* User should make sure that every key target offset is withing 51 * boundaries of unsigned short. 52 */ 53 BUG_ON(key->offset > USHRT_MAX); 54 BUG_ON(dissector_uses_key(flow_dissector, 55 key->key_id)); 56 57 dissector_set_key(flow_dissector, key->key_id); 58 flow_dissector->offset[key->key_id] = key->offset; 59 } 60 61 /* Ensure that the dissector always includes control and basic key. 62 * That way we are able to avoid handling lack of these in fast path. 63 */ 64 BUG_ON(!dissector_uses_key(flow_dissector, 65 FLOW_DISSECTOR_KEY_CONTROL)); 66 BUG_ON(!dissector_uses_key(flow_dissector, 67 FLOW_DISSECTOR_KEY_BASIC)); 68 } 69 EXPORT_SYMBOL(skb_flow_dissector_init); 70 71 /** 72 * __skb_flow_get_ports - extract the upper layer ports and return them 73 * @skb: sk_buff to extract the ports from 74 * @thoff: transport header offset 75 * @ip_proto: protocol for which to get port offset 76 * @data: raw buffer pointer to the packet, if NULL use skb->data 77 * @hlen: packet header length, if @data is NULL use skb_headlen(skb) 78 * 79 * The function will try to retrieve the ports at offset thoff + poff where poff 80 * is the protocol port offset returned from proto_ports_offset 81 */ 82 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto, 83 void *data, int hlen) 84 { 85 int poff = proto_ports_offset(ip_proto); 86 87 if (!data) { 88 data = skb->data; 89 hlen = skb_headlen(skb); 90 } 91 92 if (poff >= 0) { 93 __be32 *ports, _ports; 94 95 ports = __skb_header_pointer(skb, thoff + poff, 96 sizeof(_ports), data, hlen, &_ports); 97 if (ports) 98 return *ports; 99 } 100 101 return 0; 102 } 103 EXPORT_SYMBOL(__skb_flow_get_ports); 104 105 /** 106 * __skb_flow_dissect - extract the flow_keys struct and return it 107 * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified 108 * @flow_dissector: list of keys to dissect 109 * @target_container: target structure to put dissected values into 110 * @data: raw buffer pointer to the packet, if NULL use skb->data 111 * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol 112 * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb) 113 * @hlen: packet header length, if @data is NULL use skb_headlen(skb) 114 * 115 * The function will try to retrieve individual keys into target specified 116 * by flow_dissector from either the skbuff or a raw buffer specified by the 117 * rest parameters. 118 * 119 * Caller must take care of zeroing target container memory. 120 */ 121 bool __skb_flow_dissect(const struct sk_buff *skb, 122 struct flow_dissector *flow_dissector, 123 void *target_container, 124 void *data, __be16 proto, int nhoff, int hlen, 125 unsigned int flags) 126 { 127 struct flow_dissector_key_control *key_control; 128 struct flow_dissector_key_basic *key_basic; 129 struct flow_dissector_key_addrs *key_addrs; 130 struct flow_dissector_key_ports *key_ports; 131 struct flow_dissector_key_tags *key_tags; 132 struct flow_dissector_key_keyid *key_keyid; 133 u8 ip_proto = 0; 134 bool ret = false; 135 136 if (!data) { 137 data = skb->data; 138 proto = skb->protocol; 139 nhoff = skb_network_offset(skb); 140 hlen = skb_headlen(skb); 141 } 142 143 /* It is ensured by skb_flow_dissector_init() that control key will 144 * be always present. 145 */ 146 key_control = skb_flow_dissector_target(flow_dissector, 147 FLOW_DISSECTOR_KEY_CONTROL, 148 target_container); 149 150 /* It is ensured by skb_flow_dissector_init() that basic key will 151 * be always present. 152 */ 153 key_basic = skb_flow_dissector_target(flow_dissector, 154 FLOW_DISSECTOR_KEY_BASIC, 155 target_container); 156 157 if (dissector_uses_key(flow_dissector, 158 FLOW_DISSECTOR_KEY_ETH_ADDRS)) { 159 struct ethhdr *eth = eth_hdr(skb); 160 struct flow_dissector_key_eth_addrs *key_eth_addrs; 161 162 key_eth_addrs = skb_flow_dissector_target(flow_dissector, 163 FLOW_DISSECTOR_KEY_ETH_ADDRS, 164 target_container); 165 memcpy(key_eth_addrs, ð->h_dest, sizeof(*key_eth_addrs)); 166 } 167 168 again: 169 switch (proto) { 170 case htons(ETH_P_IP): { 171 const struct iphdr *iph; 172 struct iphdr _iph; 173 ip: 174 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph); 175 if (!iph || iph->ihl < 5) 176 goto out_bad; 177 nhoff += iph->ihl * 4; 178 179 ip_proto = iph->protocol; 180 181 if (!dissector_uses_key(flow_dissector, 182 FLOW_DISSECTOR_KEY_IPV4_ADDRS)) 183 break; 184 185 key_addrs = skb_flow_dissector_target(flow_dissector, 186 FLOW_DISSECTOR_KEY_IPV4_ADDRS, target_container); 187 memcpy(&key_addrs->v4addrs, &iph->saddr, 188 sizeof(key_addrs->v4addrs)); 189 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; 190 191 if (ip_is_fragment(iph)) { 192 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 193 194 if (iph->frag_off & htons(IP_OFFSET)) { 195 goto out_good; 196 } else { 197 key_control->flags |= FLOW_DIS_FIRST_FRAG; 198 if (!(flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) 199 goto out_good; 200 } 201 } 202 203 if (flags & FLOW_DISSECTOR_F_STOP_AT_L3) 204 goto out_good; 205 206 break; 207 } 208 case htons(ETH_P_IPV6): { 209 const struct ipv6hdr *iph; 210 struct ipv6hdr _iph; 211 212 ipv6: 213 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph); 214 if (!iph) 215 goto out_bad; 216 217 ip_proto = iph->nexthdr; 218 nhoff += sizeof(struct ipv6hdr); 219 220 if (dissector_uses_key(flow_dissector, 221 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) { 222 struct flow_dissector_key_ipv6_addrs *key_ipv6_addrs; 223 224 key_ipv6_addrs = skb_flow_dissector_target(flow_dissector, 225 FLOW_DISSECTOR_KEY_IPV6_ADDRS, 226 target_container); 227 228 memcpy(key_ipv6_addrs, &iph->saddr, sizeof(*key_ipv6_addrs)); 229 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 230 } 231 232 if ((dissector_uses_key(flow_dissector, 233 FLOW_DISSECTOR_KEY_FLOW_LABEL) || 234 (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) && 235 ip6_flowlabel(iph)) { 236 __be32 flow_label = ip6_flowlabel(iph); 237 238 if (dissector_uses_key(flow_dissector, 239 FLOW_DISSECTOR_KEY_FLOW_LABEL)) { 240 key_tags = skb_flow_dissector_target(flow_dissector, 241 FLOW_DISSECTOR_KEY_FLOW_LABEL, 242 target_container); 243 key_tags->flow_label = ntohl(flow_label); 244 } 245 if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) 246 goto out_good; 247 } 248 249 if (flags & FLOW_DISSECTOR_F_STOP_AT_L3) 250 goto out_good; 251 252 break; 253 } 254 case htons(ETH_P_8021AD): 255 case htons(ETH_P_8021Q): { 256 const struct vlan_hdr *vlan; 257 struct vlan_hdr _vlan; 258 259 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan), data, hlen, &_vlan); 260 if (!vlan) 261 goto out_bad; 262 263 if (dissector_uses_key(flow_dissector, 264 FLOW_DISSECTOR_KEY_VLANID)) { 265 key_tags = skb_flow_dissector_target(flow_dissector, 266 FLOW_DISSECTOR_KEY_VLANID, 267 target_container); 268 269 key_tags->vlan_id = skb_vlan_tag_get_id(skb); 270 } 271 272 proto = vlan->h_vlan_encapsulated_proto; 273 nhoff += sizeof(*vlan); 274 goto again; 275 } 276 case htons(ETH_P_PPP_SES): { 277 struct { 278 struct pppoe_hdr hdr; 279 __be16 proto; 280 } *hdr, _hdr; 281 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr); 282 if (!hdr) 283 goto out_bad; 284 proto = hdr->proto; 285 nhoff += PPPOE_SES_HLEN; 286 switch (proto) { 287 case htons(PPP_IP): 288 goto ip; 289 case htons(PPP_IPV6): 290 goto ipv6; 291 default: 292 goto out_bad; 293 } 294 } 295 case htons(ETH_P_TIPC): { 296 struct { 297 __be32 pre[3]; 298 __be32 srcnode; 299 } *hdr, _hdr; 300 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr); 301 if (!hdr) 302 goto out_bad; 303 304 if (dissector_uses_key(flow_dissector, 305 FLOW_DISSECTOR_KEY_TIPC_ADDRS)) { 306 key_addrs = skb_flow_dissector_target(flow_dissector, 307 FLOW_DISSECTOR_KEY_TIPC_ADDRS, 308 target_container); 309 key_addrs->tipcaddrs.srcnode = hdr->srcnode; 310 key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC_ADDRS; 311 } 312 goto out_good; 313 } 314 315 case htons(ETH_P_MPLS_UC): 316 case htons(ETH_P_MPLS_MC): { 317 struct mpls_label *hdr, _hdr[2]; 318 mpls: 319 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, 320 hlen, &_hdr); 321 if (!hdr) 322 goto out_bad; 323 324 if ((ntohl(hdr[0].entry) & MPLS_LS_LABEL_MASK) >> 325 MPLS_LS_LABEL_SHIFT == MPLS_LABEL_ENTROPY) { 326 if (dissector_uses_key(flow_dissector, 327 FLOW_DISSECTOR_KEY_MPLS_ENTROPY)) { 328 key_keyid = skb_flow_dissector_target(flow_dissector, 329 FLOW_DISSECTOR_KEY_MPLS_ENTROPY, 330 target_container); 331 key_keyid->keyid = hdr[1].entry & 332 htonl(MPLS_LS_LABEL_MASK); 333 } 334 335 goto out_good; 336 } 337 338 goto out_good; 339 } 340 341 case htons(ETH_P_FCOE): 342 key_control->thoff = (u16)(nhoff + FCOE_HEADER_LEN); 343 /* fall through */ 344 default: 345 goto out_bad; 346 } 347 348 ip_proto_again: 349 switch (ip_proto) { 350 case IPPROTO_GRE: { 351 struct gre_hdr { 352 __be16 flags; 353 __be16 proto; 354 } *hdr, _hdr; 355 356 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr); 357 if (!hdr) 358 goto out_bad; 359 /* 360 * Only look inside GRE if version zero and no 361 * routing 362 */ 363 if (hdr->flags & (GRE_VERSION | GRE_ROUTING)) 364 break; 365 366 proto = hdr->proto; 367 nhoff += 4; 368 if (hdr->flags & GRE_CSUM) 369 nhoff += 4; 370 if (hdr->flags & GRE_KEY) { 371 const __be32 *keyid; 372 __be32 _keyid; 373 374 keyid = __skb_header_pointer(skb, nhoff, sizeof(_keyid), 375 data, hlen, &_keyid); 376 377 if (!keyid) 378 goto out_bad; 379 380 if (dissector_uses_key(flow_dissector, 381 FLOW_DISSECTOR_KEY_GRE_KEYID)) { 382 key_keyid = skb_flow_dissector_target(flow_dissector, 383 FLOW_DISSECTOR_KEY_GRE_KEYID, 384 target_container); 385 key_keyid->keyid = *keyid; 386 } 387 nhoff += 4; 388 } 389 if (hdr->flags & GRE_SEQ) 390 nhoff += 4; 391 if (proto == htons(ETH_P_TEB)) { 392 const struct ethhdr *eth; 393 struct ethhdr _eth; 394 395 eth = __skb_header_pointer(skb, nhoff, 396 sizeof(_eth), 397 data, hlen, &_eth); 398 if (!eth) 399 goto out_bad; 400 proto = eth->h_proto; 401 nhoff += sizeof(*eth); 402 403 /* Cap headers that we access via pointers at the 404 * end of the Ethernet header as our maximum alignment 405 * at that point is only 2 bytes. 406 */ 407 if (NET_IP_ALIGN) 408 hlen = nhoff; 409 } 410 411 key_control->flags |= FLOW_DIS_ENCAPSULATION; 412 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) 413 goto out_good; 414 415 goto again; 416 } 417 case NEXTHDR_HOP: 418 case NEXTHDR_ROUTING: 419 case NEXTHDR_DEST: { 420 u8 _opthdr[2], *opthdr; 421 422 if (proto != htons(ETH_P_IPV6)) 423 break; 424 425 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr), 426 data, hlen, &_opthdr); 427 if (!opthdr) 428 goto out_bad; 429 430 ip_proto = opthdr[0]; 431 nhoff += (opthdr[1] + 1) << 3; 432 433 goto ip_proto_again; 434 } 435 case NEXTHDR_FRAGMENT: { 436 struct frag_hdr _fh, *fh; 437 438 if (proto != htons(ETH_P_IPV6)) 439 break; 440 441 fh = __skb_header_pointer(skb, nhoff, sizeof(_fh), 442 data, hlen, &_fh); 443 444 if (!fh) 445 goto out_bad; 446 447 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 448 449 nhoff += sizeof(_fh); 450 451 if (!(fh->frag_off & htons(IP6_OFFSET))) { 452 key_control->flags |= FLOW_DIS_FIRST_FRAG; 453 if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) { 454 ip_proto = fh->nexthdr; 455 goto ip_proto_again; 456 } 457 } 458 goto out_good; 459 } 460 case IPPROTO_IPIP: 461 proto = htons(ETH_P_IP); 462 463 key_control->flags |= FLOW_DIS_ENCAPSULATION; 464 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) 465 goto out_good; 466 467 goto ip; 468 case IPPROTO_IPV6: 469 proto = htons(ETH_P_IPV6); 470 471 key_control->flags |= FLOW_DIS_ENCAPSULATION; 472 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) 473 goto out_good; 474 475 goto ipv6; 476 case IPPROTO_MPLS: 477 proto = htons(ETH_P_MPLS_UC); 478 goto mpls; 479 default: 480 break; 481 } 482 483 if (dissector_uses_key(flow_dissector, 484 FLOW_DISSECTOR_KEY_PORTS)) { 485 key_ports = skb_flow_dissector_target(flow_dissector, 486 FLOW_DISSECTOR_KEY_PORTS, 487 target_container); 488 key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto, 489 data, hlen); 490 } 491 492 out_good: 493 ret = true; 494 495 out_bad: 496 key_basic->n_proto = proto; 497 key_basic->ip_proto = ip_proto; 498 key_control->thoff = (u16)nhoff; 499 500 return ret; 501 } 502 EXPORT_SYMBOL(__skb_flow_dissect); 503 504 static u32 hashrnd __read_mostly; 505 static __always_inline void __flow_hash_secret_init(void) 506 { 507 net_get_random_once(&hashrnd, sizeof(hashrnd)); 508 } 509 510 static __always_inline u32 __flow_hash_words(const u32 *words, u32 length, 511 u32 keyval) 512 { 513 return jhash2(words, length, keyval); 514 } 515 516 static inline const u32 *flow_keys_hash_start(const struct flow_keys *flow) 517 { 518 const void *p = flow; 519 520 BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % sizeof(u32)); 521 return (const u32 *)(p + FLOW_KEYS_HASH_OFFSET); 522 } 523 524 static inline size_t flow_keys_hash_length(const struct flow_keys *flow) 525 { 526 size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs); 527 BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32)); 528 BUILD_BUG_ON(offsetof(typeof(*flow), addrs) != 529 sizeof(*flow) - sizeof(flow->addrs)); 530 531 switch (flow->control.addr_type) { 532 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 533 diff -= sizeof(flow->addrs.v4addrs); 534 break; 535 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 536 diff -= sizeof(flow->addrs.v6addrs); 537 break; 538 case FLOW_DISSECTOR_KEY_TIPC_ADDRS: 539 diff -= sizeof(flow->addrs.tipcaddrs); 540 break; 541 } 542 return (sizeof(*flow) - diff) / sizeof(u32); 543 } 544 545 __be32 flow_get_u32_src(const struct flow_keys *flow) 546 { 547 switch (flow->control.addr_type) { 548 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 549 return flow->addrs.v4addrs.src; 550 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 551 return (__force __be32)ipv6_addr_hash( 552 &flow->addrs.v6addrs.src); 553 case FLOW_DISSECTOR_KEY_TIPC_ADDRS: 554 return flow->addrs.tipcaddrs.srcnode; 555 default: 556 return 0; 557 } 558 } 559 EXPORT_SYMBOL(flow_get_u32_src); 560 561 __be32 flow_get_u32_dst(const struct flow_keys *flow) 562 { 563 switch (flow->control.addr_type) { 564 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 565 return flow->addrs.v4addrs.dst; 566 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 567 return (__force __be32)ipv6_addr_hash( 568 &flow->addrs.v6addrs.dst); 569 default: 570 return 0; 571 } 572 } 573 EXPORT_SYMBOL(flow_get_u32_dst); 574 575 static inline void __flow_hash_consistentify(struct flow_keys *keys) 576 { 577 int addr_diff, i; 578 579 switch (keys->control.addr_type) { 580 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 581 addr_diff = (__force u32)keys->addrs.v4addrs.dst - 582 (__force u32)keys->addrs.v4addrs.src; 583 if ((addr_diff < 0) || 584 (addr_diff == 0 && 585 ((__force u16)keys->ports.dst < 586 (__force u16)keys->ports.src))) { 587 swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst); 588 swap(keys->ports.src, keys->ports.dst); 589 } 590 break; 591 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 592 addr_diff = memcmp(&keys->addrs.v6addrs.dst, 593 &keys->addrs.v6addrs.src, 594 sizeof(keys->addrs.v6addrs.dst)); 595 if ((addr_diff < 0) || 596 (addr_diff == 0 && 597 ((__force u16)keys->ports.dst < 598 (__force u16)keys->ports.src))) { 599 for (i = 0; i < 4; i++) 600 swap(keys->addrs.v6addrs.src.s6_addr32[i], 601 keys->addrs.v6addrs.dst.s6_addr32[i]); 602 swap(keys->ports.src, keys->ports.dst); 603 } 604 break; 605 } 606 } 607 608 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval) 609 { 610 u32 hash; 611 612 __flow_hash_consistentify(keys); 613 614 hash = __flow_hash_words(flow_keys_hash_start(keys), 615 flow_keys_hash_length(keys), keyval); 616 if (!hash) 617 hash = 1; 618 619 return hash; 620 } 621 622 u32 flow_hash_from_keys(struct flow_keys *keys) 623 { 624 __flow_hash_secret_init(); 625 return __flow_hash_from_keys(keys, hashrnd); 626 } 627 EXPORT_SYMBOL(flow_hash_from_keys); 628 629 static inline u32 ___skb_get_hash(const struct sk_buff *skb, 630 struct flow_keys *keys, u32 keyval) 631 { 632 skb_flow_dissect_flow_keys(skb, keys, 633 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 634 635 return __flow_hash_from_keys(keys, keyval); 636 } 637 638 struct _flow_keys_digest_data { 639 __be16 n_proto; 640 u8 ip_proto; 641 u8 padding; 642 __be32 ports; 643 __be32 src; 644 __be32 dst; 645 }; 646 647 void make_flow_keys_digest(struct flow_keys_digest *digest, 648 const struct flow_keys *flow) 649 { 650 struct _flow_keys_digest_data *data = 651 (struct _flow_keys_digest_data *)digest; 652 653 BUILD_BUG_ON(sizeof(*data) > sizeof(*digest)); 654 655 memset(digest, 0, sizeof(*digest)); 656 657 data->n_proto = flow->basic.n_proto; 658 data->ip_proto = flow->basic.ip_proto; 659 data->ports = flow->ports.ports; 660 data->src = flow->addrs.v4addrs.src; 661 data->dst = flow->addrs.v4addrs.dst; 662 } 663 EXPORT_SYMBOL(make_flow_keys_digest); 664 665 /** 666 * __skb_get_hash: calculate a flow hash 667 * @skb: sk_buff to calculate flow hash from 668 * 669 * This function calculates a flow hash based on src/dst addresses 670 * and src/dst port numbers. Sets hash in skb to non-zero hash value 671 * on success, zero indicates no valid hash. Also, sets l4_hash in skb 672 * if hash is a canonical 4-tuple hash over transport ports. 673 */ 674 void __skb_get_hash(struct sk_buff *skb) 675 { 676 struct flow_keys keys; 677 678 __flow_hash_secret_init(); 679 680 __skb_set_sw_hash(skb, ___skb_get_hash(skb, &keys, hashrnd), 681 flow_keys_have_l4(&keys)); 682 } 683 EXPORT_SYMBOL(__skb_get_hash); 684 685 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb) 686 { 687 struct flow_keys keys; 688 689 return ___skb_get_hash(skb, &keys, perturb); 690 } 691 EXPORT_SYMBOL(skb_get_hash_perturb); 692 693 __u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6) 694 { 695 struct flow_keys keys; 696 697 memset(&keys, 0, sizeof(keys)); 698 699 memcpy(&keys.addrs.v6addrs.src, &fl6->saddr, 700 sizeof(keys.addrs.v6addrs.src)); 701 memcpy(&keys.addrs.v6addrs.dst, &fl6->daddr, 702 sizeof(keys.addrs.v6addrs.dst)); 703 keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 704 keys.ports.src = fl6->fl6_sport; 705 keys.ports.dst = fl6->fl6_dport; 706 keys.keyid.keyid = fl6->fl6_gre_key; 707 keys.tags.flow_label = (__force u32)fl6->flowlabel; 708 keys.basic.ip_proto = fl6->flowi6_proto; 709 710 __skb_set_sw_hash(skb, flow_hash_from_keys(&keys), 711 flow_keys_have_l4(&keys)); 712 713 return skb->hash; 714 } 715 EXPORT_SYMBOL(__skb_get_hash_flowi6); 716 717 __u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4) 718 { 719 struct flow_keys keys; 720 721 memset(&keys, 0, sizeof(keys)); 722 723 keys.addrs.v4addrs.src = fl4->saddr; 724 keys.addrs.v4addrs.dst = fl4->daddr; 725 keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; 726 keys.ports.src = fl4->fl4_sport; 727 keys.ports.dst = fl4->fl4_dport; 728 keys.keyid.keyid = fl4->fl4_gre_key; 729 keys.basic.ip_proto = fl4->flowi4_proto; 730 731 __skb_set_sw_hash(skb, flow_hash_from_keys(&keys), 732 flow_keys_have_l4(&keys)); 733 734 return skb->hash; 735 } 736 EXPORT_SYMBOL(__skb_get_hash_flowi4); 737 738 u32 __skb_get_poff(const struct sk_buff *skb, void *data, 739 const struct flow_keys *keys, int hlen) 740 { 741 u32 poff = keys->control.thoff; 742 743 switch (keys->basic.ip_proto) { 744 case IPPROTO_TCP: { 745 /* access doff as u8 to avoid unaligned access */ 746 const u8 *doff; 747 u8 _doff; 748 749 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff), 750 data, hlen, &_doff); 751 if (!doff) 752 return poff; 753 754 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2); 755 break; 756 } 757 case IPPROTO_UDP: 758 case IPPROTO_UDPLITE: 759 poff += sizeof(struct udphdr); 760 break; 761 /* For the rest, we do not really care about header 762 * extensions at this point for now. 763 */ 764 case IPPROTO_ICMP: 765 poff += sizeof(struct icmphdr); 766 break; 767 case IPPROTO_ICMPV6: 768 poff += sizeof(struct icmp6hdr); 769 break; 770 case IPPROTO_IGMP: 771 poff += sizeof(struct igmphdr); 772 break; 773 case IPPROTO_DCCP: 774 poff += sizeof(struct dccp_hdr); 775 break; 776 case IPPROTO_SCTP: 777 poff += sizeof(struct sctphdr); 778 break; 779 } 780 781 return poff; 782 } 783 784 /** 785 * skb_get_poff - get the offset to the payload 786 * @skb: sk_buff to get the payload offset from 787 * 788 * The function will get the offset to the payload as far as it could 789 * be dissected. The main user is currently BPF, so that we can dynamically 790 * truncate packets without needing to push actual payload to the user 791 * space and can analyze headers only, instead. 792 */ 793 u32 skb_get_poff(const struct sk_buff *skb) 794 { 795 struct flow_keys keys; 796 797 if (!skb_flow_dissect_flow_keys(skb, &keys, 0)) 798 return 0; 799 800 return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb)); 801 } 802 803 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys) 804 { 805 memset(keys, 0, sizeof(*keys)); 806 807 memcpy(&keys->addrs.v6addrs.src, &fl6->saddr, 808 sizeof(keys->addrs.v6addrs.src)); 809 memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr, 810 sizeof(keys->addrs.v6addrs.dst)); 811 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 812 keys->ports.src = fl6->fl6_sport; 813 keys->ports.dst = fl6->fl6_dport; 814 keys->keyid.keyid = fl6->fl6_gre_key; 815 keys->tags.flow_label = (__force u32)fl6->flowlabel; 816 keys->basic.ip_proto = fl6->flowi6_proto; 817 818 return flow_hash_from_keys(keys); 819 } 820 EXPORT_SYMBOL(__get_hash_from_flowi6); 821 822 __u32 __get_hash_from_flowi4(const struct flowi4 *fl4, struct flow_keys *keys) 823 { 824 memset(keys, 0, sizeof(*keys)); 825 826 keys->addrs.v4addrs.src = fl4->saddr; 827 keys->addrs.v4addrs.dst = fl4->daddr; 828 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; 829 keys->ports.src = fl4->fl4_sport; 830 keys->ports.dst = fl4->fl4_dport; 831 keys->keyid.keyid = fl4->fl4_gre_key; 832 keys->basic.ip_proto = fl4->flowi4_proto; 833 834 return flow_hash_from_keys(keys); 835 } 836 EXPORT_SYMBOL(__get_hash_from_flowi4); 837 838 static const struct flow_dissector_key flow_keys_dissector_keys[] = { 839 { 840 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 841 .offset = offsetof(struct flow_keys, control), 842 }, 843 { 844 .key_id = FLOW_DISSECTOR_KEY_BASIC, 845 .offset = offsetof(struct flow_keys, basic), 846 }, 847 { 848 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS, 849 .offset = offsetof(struct flow_keys, addrs.v4addrs), 850 }, 851 { 852 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS, 853 .offset = offsetof(struct flow_keys, addrs.v6addrs), 854 }, 855 { 856 .key_id = FLOW_DISSECTOR_KEY_TIPC_ADDRS, 857 .offset = offsetof(struct flow_keys, addrs.tipcaddrs), 858 }, 859 { 860 .key_id = FLOW_DISSECTOR_KEY_PORTS, 861 .offset = offsetof(struct flow_keys, ports), 862 }, 863 { 864 .key_id = FLOW_DISSECTOR_KEY_VLANID, 865 .offset = offsetof(struct flow_keys, tags), 866 }, 867 { 868 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL, 869 .offset = offsetof(struct flow_keys, tags), 870 }, 871 { 872 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID, 873 .offset = offsetof(struct flow_keys, keyid), 874 }, 875 }; 876 877 static const struct flow_dissector_key flow_keys_buf_dissector_keys[] = { 878 { 879 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 880 .offset = offsetof(struct flow_keys, control), 881 }, 882 { 883 .key_id = FLOW_DISSECTOR_KEY_BASIC, 884 .offset = offsetof(struct flow_keys, basic), 885 }, 886 }; 887 888 struct flow_dissector flow_keys_dissector __read_mostly; 889 EXPORT_SYMBOL(flow_keys_dissector); 890 891 struct flow_dissector flow_keys_buf_dissector __read_mostly; 892 893 static int __init init_default_flow_dissectors(void) 894 { 895 skb_flow_dissector_init(&flow_keys_dissector, 896 flow_keys_dissector_keys, 897 ARRAY_SIZE(flow_keys_dissector_keys)); 898 skb_flow_dissector_init(&flow_keys_buf_dissector, 899 flow_keys_buf_dissector_keys, 900 ARRAY_SIZE(flow_keys_buf_dissector_keys)); 901 return 0; 902 } 903 904 late_initcall_sync(init_default_flow_dissectors); 905