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 __be32 flow_label; 212 213 ipv6: 214 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph); 215 if (!iph) 216 goto out_bad; 217 218 ip_proto = iph->nexthdr; 219 nhoff += sizeof(struct ipv6hdr); 220 221 if (dissector_uses_key(flow_dissector, 222 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) { 223 struct flow_dissector_key_ipv6_addrs *key_ipv6_addrs; 224 225 key_ipv6_addrs = skb_flow_dissector_target(flow_dissector, 226 FLOW_DISSECTOR_KEY_IPV6_ADDRS, 227 target_container); 228 229 memcpy(key_ipv6_addrs, &iph->saddr, sizeof(*key_ipv6_addrs)); 230 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 231 } 232 233 flow_label = ip6_flowlabel(iph); 234 if (flow_label) { 235 if (dissector_uses_key(flow_dissector, 236 FLOW_DISSECTOR_KEY_FLOW_LABEL)) { 237 key_tags = skb_flow_dissector_target(flow_dissector, 238 FLOW_DISSECTOR_KEY_FLOW_LABEL, 239 target_container); 240 key_tags->flow_label = ntohl(flow_label); 241 } 242 if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) 243 goto out_good; 244 } 245 246 if (flags & FLOW_DISSECTOR_F_STOP_AT_L3) 247 goto out_good; 248 249 break; 250 } 251 case htons(ETH_P_8021AD): 252 case htons(ETH_P_8021Q): { 253 const struct vlan_hdr *vlan; 254 struct vlan_hdr _vlan; 255 256 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan), data, hlen, &_vlan); 257 if (!vlan) 258 goto out_bad; 259 260 if (dissector_uses_key(flow_dissector, 261 FLOW_DISSECTOR_KEY_VLANID)) { 262 key_tags = skb_flow_dissector_target(flow_dissector, 263 FLOW_DISSECTOR_KEY_VLANID, 264 target_container); 265 266 key_tags->vlan_id = skb_vlan_tag_get_id(skb); 267 } 268 269 proto = vlan->h_vlan_encapsulated_proto; 270 nhoff += sizeof(*vlan); 271 goto again; 272 } 273 case htons(ETH_P_PPP_SES): { 274 struct { 275 struct pppoe_hdr hdr; 276 __be16 proto; 277 } *hdr, _hdr; 278 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr); 279 if (!hdr) 280 goto out_bad; 281 proto = hdr->proto; 282 nhoff += PPPOE_SES_HLEN; 283 switch (proto) { 284 case htons(PPP_IP): 285 goto ip; 286 case htons(PPP_IPV6): 287 goto ipv6; 288 default: 289 goto out_bad; 290 } 291 } 292 case htons(ETH_P_TIPC): { 293 struct { 294 __be32 pre[3]; 295 __be32 srcnode; 296 } *hdr, _hdr; 297 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr); 298 if (!hdr) 299 goto out_bad; 300 301 if (dissector_uses_key(flow_dissector, 302 FLOW_DISSECTOR_KEY_TIPC_ADDRS)) { 303 key_addrs = skb_flow_dissector_target(flow_dissector, 304 FLOW_DISSECTOR_KEY_TIPC_ADDRS, 305 target_container); 306 key_addrs->tipcaddrs.srcnode = hdr->srcnode; 307 key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC_ADDRS; 308 } 309 goto out_good; 310 } 311 312 case htons(ETH_P_MPLS_UC): 313 case htons(ETH_P_MPLS_MC): { 314 struct mpls_label *hdr, _hdr[2]; 315 mpls: 316 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, 317 hlen, &_hdr); 318 if (!hdr) 319 goto out_bad; 320 321 if ((ntohl(hdr[0].entry) & MPLS_LS_LABEL_MASK) >> 322 MPLS_LS_LABEL_SHIFT == MPLS_LABEL_ENTROPY) { 323 if (dissector_uses_key(flow_dissector, 324 FLOW_DISSECTOR_KEY_MPLS_ENTROPY)) { 325 key_keyid = skb_flow_dissector_target(flow_dissector, 326 FLOW_DISSECTOR_KEY_MPLS_ENTROPY, 327 target_container); 328 key_keyid->keyid = hdr[1].entry & 329 htonl(MPLS_LS_LABEL_MASK); 330 } 331 332 goto out_good; 333 } 334 335 goto out_good; 336 } 337 338 case htons(ETH_P_FCOE): 339 key_control->thoff = (u16)(nhoff + FCOE_HEADER_LEN); 340 /* fall through */ 341 default: 342 goto out_bad; 343 } 344 345 ip_proto_again: 346 switch (ip_proto) { 347 case IPPROTO_GRE: { 348 struct gre_hdr { 349 __be16 flags; 350 __be16 proto; 351 } *hdr, _hdr; 352 353 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr); 354 if (!hdr) 355 goto out_bad; 356 /* 357 * Only look inside GRE if version zero and no 358 * routing 359 */ 360 if (hdr->flags & (GRE_VERSION | GRE_ROUTING)) 361 break; 362 363 proto = hdr->proto; 364 nhoff += 4; 365 if (hdr->flags & GRE_CSUM) 366 nhoff += 4; 367 if (hdr->flags & GRE_KEY) { 368 const __be32 *keyid; 369 __be32 _keyid; 370 371 keyid = __skb_header_pointer(skb, nhoff, sizeof(_keyid), 372 data, hlen, &_keyid); 373 374 if (!keyid) 375 goto out_bad; 376 377 if (dissector_uses_key(flow_dissector, 378 FLOW_DISSECTOR_KEY_GRE_KEYID)) { 379 key_keyid = skb_flow_dissector_target(flow_dissector, 380 FLOW_DISSECTOR_KEY_GRE_KEYID, 381 target_container); 382 key_keyid->keyid = *keyid; 383 } 384 nhoff += 4; 385 } 386 if (hdr->flags & GRE_SEQ) 387 nhoff += 4; 388 if (proto == htons(ETH_P_TEB)) { 389 const struct ethhdr *eth; 390 struct ethhdr _eth; 391 392 eth = __skb_header_pointer(skb, nhoff, 393 sizeof(_eth), 394 data, hlen, &_eth); 395 if (!eth) 396 goto out_bad; 397 proto = eth->h_proto; 398 nhoff += sizeof(*eth); 399 } 400 401 key_control->flags |= FLOW_DIS_ENCAPSULATION; 402 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) 403 goto out_good; 404 405 goto again; 406 } 407 case NEXTHDR_HOP: 408 case NEXTHDR_ROUTING: 409 case NEXTHDR_DEST: { 410 u8 _opthdr[2], *opthdr; 411 412 if (proto != htons(ETH_P_IPV6)) 413 break; 414 415 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr), 416 data, hlen, &_opthdr); 417 if (!opthdr) 418 goto out_bad; 419 420 ip_proto = opthdr[0]; 421 nhoff += (opthdr[1] + 1) << 3; 422 423 goto ip_proto_again; 424 } 425 case NEXTHDR_FRAGMENT: { 426 struct frag_hdr _fh, *fh; 427 428 if (proto != htons(ETH_P_IPV6)) 429 break; 430 431 fh = __skb_header_pointer(skb, nhoff, sizeof(_fh), 432 data, hlen, &_fh); 433 434 if (!fh) 435 goto out_bad; 436 437 key_control->flags |= FLOW_DIS_IS_FRAGMENT; 438 439 nhoff += sizeof(_fh); 440 441 if (!(fh->frag_off & htons(IP6_OFFSET))) { 442 key_control->flags |= FLOW_DIS_FIRST_FRAG; 443 if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) { 444 ip_proto = fh->nexthdr; 445 goto ip_proto_again; 446 } 447 } 448 goto out_good; 449 } 450 case IPPROTO_IPIP: 451 proto = htons(ETH_P_IP); 452 453 key_control->flags |= FLOW_DIS_ENCAPSULATION; 454 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) 455 goto out_good; 456 457 goto ip; 458 case IPPROTO_IPV6: 459 proto = htons(ETH_P_IPV6); 460 461 key_control->flags |= FLOW_DIS_ENCAPSULATION; 462 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) 463 goto out_good; 464 465 goto ipv6; 466 case IPPROTO_MPLS: 467 proto = htons(ETH_P_MPLS_UC); 468 goto mpls; 469 default: 470 break; 471 } 472 473 if (dissector_uses_key(flow_dissector, 474 FLOW_DISSECTOR_KEY_PORTS)) { 475 key_ports = skb_flow_dissector_target(flow_dissector, 476 FLOW_DISSECTOR_KEY_PORTS, 477 target_container); 478 key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto, 479 data, hlen); 480 } 481 482 out_good: 483 ret = true; 484 485 out_bad: 486 key_basic->n_proto = proto; 487 key_basic->ip_proto = ip_proto; 488 key_control->thoff = (u16)nhoff; 489 490 return ret; 491 } 492 EXPORT_SYMBOL(__skb_flow_dissect); 493 494 static u32 hashrnd __read_mostly; 495 static __always_inline void __flow_hash_secret_init(void) 496 { 497 net_get_random_once(&hashrnd, sizeof(hashrnd)); 498 } 499 500 static __always_inline u32 __flow_hash_words(const u32 *words, u32 length, 501 u32 keyval) 502 { 503 return jhash2(words, length, keyval); 504 } 505 506 static inline const u32 *flow_keys_hash_start(const struct flow_keys *flow) 507 { 508 const void *p = flow; 509 510 BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % sizeof(u32)); 511 return (const u32 *)(p + FLOW_KEYS_HASH_OFFSET); 512 } 513 514 static inline size_t flow_keys_hash_length(const struct flow_keys *flow) 515 { 516 size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs); 517 BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32)); 518 BUILD_BUG_ON(offsetof(typeof(*flow), addrs) != 519 sizeof(*flow) - sizeof(flow->addrs)); 520 521 switch (flow->control.addr_type) { 522 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 523 diff -= sizeof(flow->addrs.v4addrs); 524 break; 525 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 526 diff -= sizeof(flow->addrs.v6addrs); 527 break; 528 case FLOW_DISSECTOR_KEY_TIPC_ADDRS: 529 diff -= sizeof(flow->addrs.tipcaddrs); 530 break; 531 } 532 return (sizeof(*flow) - diff) / sizeof(u32); 533 } 534 535 __be32 flow_get_u32_src(const struct flow_keys *flow) 536 { 537 switch (flow->control.addr_type) { 538 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 539 return flow->addrs.v4addrs.src; 540 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 541 return (__force __be32)ipv6_addr_hash( 542 &flow->addrs.v6addrs.src); 543 case FLOW_DISSECTOR_KEY_TIPC_ADDRS: 544 return flow->addrs.tipcaddrs.srcnode; 545 default: 546 return 0; 547 } 548 } 549 EXPORT_SYMBOL(flow_get_u32_src); 550 551 __be32 flow_get_u32_dst(const struct flow_keys *flow) 552 { 553 switch (flow->control.addr_type) { 554 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 555 return flow->addrs.v4addrs.dst; 556 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 557 return (__force __be32)ipv6_addr_hash( 558 &flow->addrs.v6addrs.dst); 559 default: 560 return 0; 561 } 562 } 563 EXPORT_SYMBOL(flow_get_u32_dst); 564 565 static inline void __flow_hash_consistentify(struct flow_keys *keys) 566 { 567 int addr_diff, i; 568 569 switch (keys->control.addr_type) { 570 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 571 addr_diff = (__force u32)keys->addrs.v4addrs.dst - 572 (__force u32)keys->addrs.v4addrs.src; 573 if ((addr_diff < 0) || 574 (addr_diff == 0 && 575 ((__force u16)keys->ports.dst < 576 (__force u16)keys->ports.src))) { 577 swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst); 578 swap(keys->ports.src, keys->ports.dst); 579 } 580 break; 581 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 582 addr_diff = memcmp(&keys->addrs.v6addrs.dst, 583 &keys->addrs.v6addrs.src, 584 sizeof(keys->addrs.v6addrs.dst)); 585 if ((addr_diff < 0) || 586 (addr_diff == 0 && 587 ((__force u16)keys->ports.dst < 588 (__force u16)keys->ports.src))) { 589 for (i = 0; i < 4; i++) 590 swap(keys->addrs.v6addrs.src.s6_addr32[i], 591 keys->addrs.v6addrs.dst.s6_addr32[i]); 592 swap(keys->ports.src, keys->ports.dst); 593 } 594 break; 595 } 596 } 597 598 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval) 599 { 600 u32 hash; 601 602 __flow_hash_consistentify(keys); 603 604 hash = __flow_hash_words(flow_keys_hash_start(keys), 605 flow_keys_hash_length(keys), keyval); 606 if (!hash) 607 hash = 1; 608 609 return hash; 610 } 611 612 u32 flow_hash_from_keys(struct flow_keys *keys) 613 { 614 __flow_hash_secret_init(); 615 return __flow_hash_from_keys(keys, hashrnd); 616 } 617 EXPORT_SYMBOL(flow_hash_from_keys); 618 619 static inline u32 ___skb_get_hash(const struct sk_buff *skb, 620 struct flow_keys *keys, u32 keyval) 621 { 622 skb_flow_dissect_flow_keys(skb, keys, 623 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 624 625 return __flow_hash_from_keys(keys, keyval); 626 } 627 628 struct _flow_keys_digest_data { 629 __be16 n_proto; 630 u8 ip_proto; 631 u8 padding; 632 __be32 ports; 633 __be32 src; 634 __be32 dst; 635 }; 636 637 void make_flow_keys_digest(struct flow_keys_digest *digest, 638 const struct flow_keys *flow) 639 { 640 struct _flow_keys_digest_data *data = 641 (struct _flow_keys_digest_data *)digest; 642 643 BUILD_BUG_ON(sizeof(*data) > sizeof(*digest)); 644 645 memset(digest, 0, sizeof(*digest)); 646 647 data->n_proto = flow->basic.n_proto; 648 data->ip_proto = flow->basic.ip_proto; 649 data->ports = flow->ports.ports; 650 data->src = flow->addrs.v4addrs.src; 651 data->dst = flow->addrs.v4addrs.dst; 652 } 653 EXPORT_SYMBOL(make_flow_keys_digest); 654 655 /** 656 * __skb_get_hash: calculate a flow hash 657 * @skb: sk_buff to calculate flow hash from 658 * 659 * This function calculates a flow hash based on src/dst addresses 660 * and src/dst port numbers. Sets hash in skb to non-zero hash value 661 * on success, zero indicates no valid hash. Also, sets l4_hash in skb 662 * if hash is a canonical 4-tuple hash over transport ports. 663 */ 664 void __skb_get_hash(struct sk_buff *skb) 665 { 666 struct flow_keys keys; 667 668 __flow_hash_secret_init(); 669 670 __skb_set_sw_hash(skb, ___skb_get_hash(skb, &keys, hashrnd), 671 flow_keys_have_l4(&keys)); 672 } 673 EXPORT_SYMBOL(__skb_get_hash); 674 675 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb) 676 { 677 struct flow_keys keys; 678 679 return ___skb_get_hash(skb, &keys, perturb); 680 } 681 EXPORT_SYMBOL(skb_get_hash_perturb); 682 683 __u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6) 684 { 685 struct flow_keys keys; 686 687 memset(&keys, 0, sizeof(keys)); 688 689 memcpy(&keys.addrs.v6addrs.src, &fl6->saddr, 690 sizeof(keys.addrs.v6addrs.src)); 691 memcpy(&keys.addrs.v6addrs.dst, &fl6->daddr, 692 sizeof(keys.addrs.v6addrs.dst)); 693 keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 694 keys.ports.src = fl6->fl6_sport; 695 keys.ports.dst = fl6->fl6_dport; 696 keys.keyid.keyid = fl6->fl6_gre_key; 697 keys.tags.flow_label = (__force u32)fl6->flowlabel; 698 keys.basic.ip_proto = fl6->flowi6_proto; 699 700 __skb_set_sw_hash(skb, flow_hash_from_keys(&keys), 701 flow_keys_have_l4(&keys)); 702 703 return skb->hash; 704 } 705 EXPORT_SYMBOL(__skb_get_hash_flowi6); 706 707 __u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4) 708 { 709 struct flow_keys keys; 710 711 memset(&keys, 0, sizeof(keys)); 712 713 keys.addrs.v4addrs.src = fl4->saddr; 714 keys.addrs.v4addrs.dst = fl4->daddr; 715 keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; 716 keys.ports.src = fl4->fl4_sport; 717 keys.ports.dst = fl4->fl4_dport; 718 keys.keyid.keyid = fl4->fl4_gre_key; 719 keys.basic.ip_proto = fl4->flowi4_proto; 720 721 __skb_set_sw_hash(skb, flow_hash_from_keys(&keys), 722 flow_keys_have_l4(&keys)); 723 724 return skb->hash; 725 } 726 EXPORT_SYMBOL(__skb_get_hash_flowi4); 727 728 u32 __skb_get_poff(const struct sk_buff *skb, void *data, 729 const struct flow_keys *keys, int hlen) 730 { 731 u32 poff = keys->control.thoff; 732 733 switch (keys->basic.ip_proto) { 734 case IPPROTO_TCP: { 735 /* access doff as u8 to avoid unaligned access */ 736 const u8 *doff; 737 u8 _doff; 738 739 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff), 740 data, hlen, &_doff); 741 if (!doff) 742 return poff; 743 744 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2); 745 break; 746 } 747 case IPPROTO_UDP: 748 case IPPROTO_UDPLITE: 749 poff += sizeof(struct udphdr); 750 break; 751 /* For the rest, we do not really care about header 752 * extensions at this point for now. 753 */ 754 case IPPROTO_ICMP: 755 poff += sizeof(struct icmphdr); 756 break; 757 case IPPROTO_ICMPV6: 758 poff += sizeof(struct icmp6hdr); 759 break; 760 case IPPROTO_IGMP: 761 poff += sizeof(struct igmphdr); 762 break; 763 case IPPROTO_DCCP: 764 poff += sizeof(struct dccp_hdr); 765 break; 766 case IPPROTO_SCTP: 767 poff += sizeof(struct sctphdr); 768 break; 769 } 770 771 return poff; 772 } 773 774 /** 775 * skb_get_poff - get the offset to the payload 776 * @skb: sk_buff to get the payload offset from 777 * 778 * The function will get the offset to the payload as far as it could 779 * be dissected. The main user is currently BPF, so that we can dynamically 780 * truncate packets without needing to push actual payload to the user 781 * space and can analyze headers only, instead. 782 */ 783 u32 skb_get_poff(const struct sk_buff *skb) 784 { 785 struct flow_keys keys; 786 787 if (!skb_flow_dissect_flow_keys(skb, &keys, 0)) 788 return 0; 789 790 return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb)); 791 } 792 793 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys) 794 { 795 memset(keys, 0, sizeof(*keys)); 796 797 memcpy(&keys->addrs.v6addrs.src, &fl6->saddr, 798 sizeof(keys->addrs.v6addrs.src)); 799 memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr, 800 sizeof(keys->addrs.v6addrs.dst)); 801 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 802 keys->ports.src = fl6->fl6_sport; 803 keys->ports.dst = fl6->fl6_dport; 804 keys->keyid.keyid = fl6->fl6_gre_key; 805 keys->tags.flow_label = (__force u32)fl6->flowlabel; 806 keys->basic.ip_proto = fl6->flowi6_proto; 807 808 return flow_hash_from_keys(keys); 809 } 810 EXPORT_SYMBOL(__get_hash_from_flowi6); 811 812 __u32 __get_hash_from_flowi4(const struct flowi4 *fl4, struct flow_keys *keys) 813 { 814 memset(keys, 0, sizeof(*keys)); 815 816 keys->addrs.v4addrs.src = fl4->saddr; 817 keys->addrs.v4addrs.dst = fl4->daddr; 818 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; 819 keys->ports.src = fl4->fl4_sport; 820 keys->ports.dst = fl4->fl4_dport; 821 keys->keyid.keyid = fl4->fl4_gre_key; 822 keys->basic.ip_proto = fl4->flowi4_proto; 823 824 return flow_hash_from_keys(keys); 825 } 826 EXPORT_SYMBOL(__get_hash_from_flowi4); 827 828 static const struct flow_dissector_key flow_keys_dissector_keys[] = { 829 { 830 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 831 .offset = offsetof(struct flow_keys, control), 832 }, 833 { 834 .key_id = FLOW_DISSECTOR_KEY_BASIC, 835 .offset = offsetof(struct flow_keys, basic), 836 }, 837 { 838 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS, 839 .offset = offsetof(struct flow_keys, addrs.v4addrs), 840 }, 841 { 842 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS, 843 .offset = offsetof(struct flow_keys, addrs.v6addrs), 844 }, 845 { 846 .key_id = FLOW_DISSECTOR_KEY_TIPC_ADDRS, 847 .offset = offsetof(struct flow_keys, addrs.tipcaddrs), 848 }, 849 { 850 .key_id = FLOW_DISSECTOR_KEY_PORTS, 851 .offset = offsetof(struct flow_keys, ports), 852 }, 853 { 854 .key_id = FLOW_DISSECTOR_KEY_VLANID, 855 .offset = offsetof(struct flow_keys, tags), 856 }, 857 { 858 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL, 859 .offset = offsetof(struct flow_keys, tags), 860 }, 861 { 862 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID, 863 .offset = offsetof(struct flow_keys, keyid), 864 }, 865 }; 866 867 static const struct flow_dissector_key flow_keys_buf_dissector_keys[] = { 868 { 869 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 870 .offset = offsetof(struct flow_keys, control), 871 }, 872 { 873 .key_id = FLOW_DISSECTOR_KEY_BASIC, 874 .offset = offsetof(struct flow_keys, basic), 875 }, 876 }; 877 878 struct flow_dissector flow_keys_dissector __read_mostly; 879 EXPORT_SYMBOL(flow_keys_dissector); 880 881 struct flow_dissector flow_keys_buf_dissector __read_mostly; 882 883 static int __init init_default_flow_dissectors(void) 884 { 885 skb_flow_dissector_init(&flow_keys_dissector, 886 flow_keys_dissector_keys, 887 ARRAY_SIZE(flow_keys_dissector_keys)); 888 skb_flow_dissector_init(&flow_keys_buf_dissector, 889 flow_keys_buf_dissector_keys, 890 ARRAY_SIZE(flow_keys_buf_dissector_keys)); 891 return 0; 892 } 893 894 late_initcall_sync(init_default_flow_dissectors); 895