lib/packets.h

   1 /*
   2  * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015 Nicira, Inc.
   3  *
   4  * Licensed under the Apache License, Version 2.0 (the "License");
   5  * you may not use this file except in compliance with the License.
   6  * You may obtain a copy of the License at:
   7  *
   8  *     http://www.apache.org/licenses/LICENSE-2.0
   9  *
  10  * Unless required by applicable law or agreed to in writing, software
  11  * distributed under the License is distributed on an "AS IS" BASIS,
  12  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  13  * See the License for the specific language governing permissions and
  14  * limitations under the License.
  15  */
  16
  17 #ifndef PACKETS_H
  18 #define PACKETS_H 1
  19
  20 #include <inttypes.h>
  21 #include <sys/types.h>
  22 #include <netinet/in.h>
  23 #include <stdint.h>
  24 #include <string.h>
  25 #include "compiler.h"
  26 #include "geneve.h"
  27 #include "openvswitch/types.h"
  28 #include "odp-netlink.h"
  29 #include "random.h"
  30 #include "hash.h"
  31 #include "tun-metadata.h"
  32 #include "unaligned.h"
  33 #include "util.h"
  34
  35 struct dp_packet;
  36 struct ds;
  37
  38 /* Tunnel information used in flow key and metadata. */
  39 struct flow_tnl {
  40     ovs_be32 ip_dst;
  41     ovs_be32 ip_src;
  42     ovs_be64 tun_id;
  43     uint16_t flags;
  44     uint8_t ip_tos;
  45     uint8_t ip_ttl;
  46     ovs_be16 tp_src;
  47     ovs_be16 tp_dst;
  48     ovs_be16 gbp_id;
  49     uint8_t  gbp_flags;
  50     uint8_t  pad1[5];        /* Pad to 64 bits. */
  51     struct tun_metadata metadata;
  52 };
  53
  54 /* Some flags are exposed through OpenFlow while others are used only
  55  * internally. */
  56
  57 /* Public flags */
  58 #define FLOW_TNL_F_OAM (1 << 0)
  59
  60 #define FLOW_TNL_PUB_F_MASK ((1 << 1) - 1)
  61
  62 /* Private flags */
  63 #define FLOW_TNL_F_DONT_FRAGMENT (1 << 1)
  64 #define FLOW_TNL_F_CSUM (1 << 2)
  65 #define FLOW_TNL_F_KEY (1 << 3)
  66
  67 #define FLOW_TNL_F_MASK ((1 << 4) - 1)
  68
  69 /* Purely internal to OVS userspace. These flags should never be exposed to
  70  * the outside world and so aren't included in the flags mask. */
  71
  72 /* Tunnel information is in userspace datapath format. */
  73 #define FLOW_TNL_F_UDPIF (1 << 4)
  74
  75 /* Returns an offset to 'src' covering all the meaningful fields in 'src'. */
  76 static inline size_t
  77 flow_tnl_size(const struct flow_tnl *src)
  78 {
  79     if (!src->ip_dst) {
  80         /* Covers ip_dst only. */
  81         return offsetof(struct flow_tnl, ip_src);
  82     }
  83     if (src->flags & FLOW_TNL_F_UDPIF) {
  84         /* Datapath format, cover all options we have. */
  85         return offsetof(struct flow_tnl, metadata.opts)
  86             + src->metadata.present.len;
  87     }
  88     if (!src->metadata.present.map) {
  89         /* No TLVs, opts is irrelevant. */
  90         return offsetof(struct flow_tnl, metadata.opts);
  91     }
  92     /* Have decoded TLVs, opts is relevant. */
  93     return sizeof *src;
  94 }
  95
  96 /* Copy flow_tnl, but avoid copying unused portions of tun_metadata.  Unused
  97  * data in 'dst' is NOT cleared, so this must not be used in cases where the
  98  * uninitialized portion may be hashed over. */
  99 static inline void
 100 flow_tnl_copy__(struct flow_tnl *dst, const struct flow_tnl *src)
 101 {
 102     memcpy(dst, src, flow_tnl_size(src));
 103 }
 104
 105 static inline bool
 106 flow_tnl_equal(const struct flow_tnl *a, const struct flow_tnl *b)
 107 {
 108     size_t a_size = flow_tnl_size(a);
 109
 110     return a_size == flow_tnl_size(b) && !memcmp(a, b, a_size);
 111 }
 112
 113 /* Unfortunately, a "struct flow" sometimes has to handle OpenFlow port
 114  * numbers and other times datapath (dpif) port numbers.  This union allows
 115  * access to both. */
 116 union flow_in_port {
 117     odp_port_t odp_port;
 118     ofp_port_t ofp_port;
 119 };
 120
 121 /* Datapath packet metadata */
 122 struct pkt_metadata {
 123     uint32_t recirc_id;         /* Recirculation id carried with the
 124                                    recirculating packets. 0 for packets
 125                                    received from the wire. */
 126     uint32_t dp_hash;           /* hash value computed by the recirculation
 127                                    action. */
 128     uint32_t skb_priority;      /* Packet priority for QoS. */
 129     uint32_t pkt_mark;          /* Packet mark. */
 130     uint16_t ct_state;          /* Connection state. */
 131     uint16_t ct_zone;           /* Connection zone. */
 132     union flow_in_port in_port; /* Input port. */
 133     struct flow_tnl tunnel;     /* Encapsulating tunnel parameters. Note that
 134                                  * if 'ip_dst' == 0, the rest of the fields may
 135                                  * be uninitialized. */
 136 };
 137
 138 static inline void
 139 pkt_metadata_init(struct pkt_metadata *md, odp_port_t port)
 140 {
 141     /* It can be expensive to zero out all of the tunnel metadata. However,
 142      * we can just zero out ip_dst and the rest of the data will never be
 143      * looked at. */
 144     memset(md, 0, offsetof(struct pkt_metadata, in_port));
 145     md->tunnel.ip_dst = 0;
 146
 147     md->in_port.odp_port = port;
 148 }
 149
 150 bool dpid_from_string(const char *s, uint64_t *dpidp);
 151
 152 #define ETH_ADDR_LEN           6
 153
 154 static const struct eth_addr eth_addr_broadcast OVS_UNUSED
 155     = { { { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff } } };
 156
 157 static const struct eth_addr eth_addr_exact OVS_UNUSED
 158     = { { { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff } } };
 159
 160 static const struct eth_addr eth_addr_zero OVS_UNUSED
 161     = { { { 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 } } };
 162
 163 static const struct eth_addr eth_addr_stp OVS_UNUSED
 164     = { { { 0x01, 0x80, 0xC2, 0x00, 0x00, 0x00 } } };
 165
 166 static const struct eth_addr eth_addr_lacp OVS_UNUSED
 167     = { { { 0x01, 0x80, 0xC2, 0x00, 0x00, 0x02 } } };
 168
 169 static const struct eth_addr eth_addr_bfd OVS_UNUSED
 170     = { { { 0x00, 0x23, 0x20, 0x00, 0x00, 0x01 } } };
 171
 172 static inline bool eth_addr_is_broadcast(const struct eth_addr a)
 173 {
 174     return (a.be16[0] & a.be16[1] & a.be16[2]) == htons(0xffff);
 175 }
 176
 177 static inline bool eth_addr_is_multicast(const struct eth_addr a)
 178 {
 179     return a.ea[0] & 1;
 180 }
 181
 182 static inline bool eth_addr_is_local(const struct eth_addr a)
 183 {
 184     /* Local if it is either a locally administered address or a Nicira random
 185      * address. */
 186     return a.ea[0] & 2
 187         || (a.be16[0] == htons(0x0023)
 188             && (a.be16[1] & htons(0xff80)) == htons(0x2080));
 189 }
 190 static inline bool eth_addr_is_zero(const struct eth_addr a)
 191 {
 192     return !(a.be16[0] | a.be16[1] | a.be16[2]);
 193 }
 194
 195 static inline int eth_mask_is_exact(const struct eth_addr a)
 196 {
 197     return (a.be16[0] & a.be16[1] & a.be16[2]) == htons(0xffff);
 198 }
 199
 200 static inline int eth_addr_compare_3way(const struct eth_addr a,
 201                                         const struct eth_addr b)
 202 {
 203     return memcmp(&a, &b, sizeof a);
 204 }
 205
 206 static inline bool eth_addr_equals(const struct eth_addr a,
 207                                    const struct eth_addr b)
 208 {
 209     return !eth_addr_compare_3way(a, b);
 210 }
 211
 212 static inline bool eth_addr_equal_except(const struct eth_addr a,
 213                                          const struct eth_addr b,
 214                                          const struct eth_addr mask)
 215 {
 216     return !(((a.be16[0] ^ b.be16[0]) & mask.be16[0])
 217              || ((a.be16[1] ^ b.be16[1]) & mask.be16[1])
 218              || ((a.be16[2] ^ b.be16[2]) & mask.be16[2]));
 219 }
 220
 221 static inline uint64_t eth_addr_to_uint64(const struct eth_addr ea)
 222 {
 223     return (((uint64_t) ntohs(ea.be16[0]) << 32)
 224             | ((uint64_t) ntohs(ea.be16[1]) << 16)
 225             | ntohs(ea.be16[2]));
 226 }
 227
 228 static inline uint64_t eth_addr_vlan_to_uint64(const struct eth_addr ea,
 229                                                uint16_t vlan)
 230 {
 231     return (((uint64_t)vlan << 48) | eth_addr_to_uint64(ea));
 232 }
 233
 234 static inline void eth_addr_from_uint64(uint64_t x, struct eth_addr *ea)
 235 {
 236     ea->be16[0] = htons(x >> 32);
 237     ea->be16[1] = htons((x & 0xFFFF0000) >> 16);
 238     ea->be16[2] = htons(x & 0xFFFF);
 239 }
 240
 241 static inline struct eth_addr eth_addr_invert(const struct eth_addr src)
 242 {
 243     struct eth_addr dst;
 244
 245     for (int i = 0; i < ARRAY_SIZE(src.be16); i++) {
 246         dst.be16[i] = ~src.be16[i];
 247     }
 248
 249     return dst;
 250 }
 251
 252 static inline void eth_addr_mark_random(struct eth_addr *ea)
 253 {
 254     ea->ea[0] &= ~1;                /* Unicast. */
 255     ea->ea[0] |= 2;                 /* Private. */
 256 }
 257
 258 static inline void eth_addr_random(struct eth_addr *ea)
 259 {
 260     random_bytes((uint8_t *)ea, sizeof *ea);
 261     eth_addr_mark_random(ea);
 262 }
 263
 264 static inline void eth_addr_nicira_random(struct eth_addr *ea)
 265 {
 266     eth_addr_random(ea);
 267
 268     /* Set the OUI to the Nicira one. */
 269     ea->ea[0] = 0x00;
 270     ea->ea[1] = 0x23;
 271     ea->ea[2] = 0x20;
 272
 273     /* Set the top bit to indicate random Nicira address. */
 274     ea->ea[3] |= 0x80;
 275 }
 276 static inline uint32_t hash_mac(const struct eth_addr ea,
 277                                 const uint16_t vlan, const uint32_t basis)
 278 {
 279     return hash_uint64_basis(eth_addr_vlan_to_uint64(ea, vlan), basis);
 280 }
 281
 282 bool eth_addr_is_reserved(const struct eth_addr);
 283 bool eth_addr_from_string(const char *, struct eth_addr *);
 284
 285 void compose_rarp(struct dp_packet *, const struct eth_addr);
 286
 287 void eth_push_vlan(struct dp_packet *, ovs_be16 tpid, ovs_be16 tci);
 288 void eth_pop_vlan(struct dp_packet *);
 289
 290 const char *eth_from_hex(const char *hex, struct dp_packet **packetp);
 291 void eth_format_masked(const struct eth_addr ea,
 292                        const struct eth_addr *mask, struct ds *s);
 293
 294 void set_mpls_lse(struct dp_packet *, ovs_be32 label);
 295 void push_mpls(struct dp_packet *packet, ovs_be16 ethtype, ovs_be32 lse);
 296 void pop_mpls(struct dp_packet *, ovs_be16 ethtype);
 297
 298 void set_mpls_lse_ttl(ovs_be32 *lse, uint8_t ttl);
 299 void set_mpls_lse_tc(ovs_be32 *lse, uint8_t tc);
 300 void set_mpls_lse_label(ovs_be32 *lse, ovs_be32 label);
 301 void set_mpls_lse_bos(ovs_be32 *lse, uint8_t bos);
 302 ovs_be32 set_mpls_lse_values(uint8_t ttl, uint8_t tc, uint8_t bos,
 303                              ovs_be32 label);
 304
 305 /* Example:
 306  *
 307  * struct eth_addr mac;
 308  *    [...]
 309  * printf("The Ethernet address is "ETH_ADDR_FMT"\n", ETH_ADDR_ARGS(mac));
 310  *
 311  */
 312 #define ETH_ADDR_FMT                                                    \
 313     "%02"PRIx8":%02"PRIx8":%02"PRIx8":%02"PRIx8":%02"PRIx8":%02"PRIx8
 314 #define ETH_ADDR_ARGS(EA) ETH_ADDR_BYTES_ARGS((EA).ea)
 315 #define ETH_ADDR_BYTES_ARGS(EAB) \
 316          (EAB)[0], (EAB)[1], (EAB)[2], (EAB)[3], (EAB)[4], (EAB)[5]
 317
 318 /* Example:
 319  *
 320  * char *string = "1 00:11:22:33:44:55 2";
 321  * struct eth_addr mac;
 322  * int a, b;
 323  *
 324  * if (ovs_scan(string, "%d"ETH_ADDR_SCAN_FMT"%d",
 325  *              &a, ETH_ADDR_SCAN_ARGS(mac), &b)) {
 326  *     ...
 327  * }
 328  */
 329 #define ETH_ADDR_SCAN_FMT "%"SCNx8":%"SCNx8":%"SCNx8":%"SCNx8":%"SCNx8":%"SCNx8
 330 #define ETH_ADDR_SCAN_ARGS(EA) \
 331     &(EA).ea[0], &(EA).ea[1], &(EA).ea[2], &(EA).ea[3], &(EA).ea[4], &(EA).ea[5]
 332
 333 #define ETH_TYPE_IP            0x0800
 334 #define ETH_TYPE_ARP           0x0806
 335 #define ETH_TYPE_TEB           0x6558
 336 #define ETH_TYPE_VLAN_8021Q    0x8100
 337 #define ETH_TYPE_VLAN          ETH_TYPE_VLAN_8021Q
 338 #define ETH_TYPE_VLAN_8021AD   0x88a8
 339 #define ETH_TYPE_IPV6          0x86dd
 340 #define ETH_TYPE_LACP          0x8809
 341 #define ETH_TYPE_RARP          0x8035
 342 #define ETH_TYPE_MPLS          0x8847
 343 #define ETH_TYPE_MPLS_MCAST    0x8848
 344
 345 static inline bool eth_type_mpls(ovs_be16 eth_type)
 346 {
 347     return eth_type == htons(ETH_TYPE_MPLS) ||
 348         eth_type == htons(ETH_TYPE_MPLS_MCAST);
 349 }
 350
 351 static inline bool eth_type_vlan(ovs_be16 eth_type)
 352 {
 353     return eth_type == htons(ETH_TYPE_VLAN_8021Q) ||
 354         eth_type == htons(ETH_TYPE_VLAN_8021AD);
 355 }
 356
 357
 358 /* Minimum value for an Ethernet type.  Values below this are IEEE 802.2 frame
 359  * lengths. */
 360 #define ETH_TYPE_MIN           0x600
 361
 362 #define ETH_HEADER_LEN 14
 363 #define ETH_PAYLOAD_MIN 46
 364 #define ETH_PAYLOAD_MAX 1500
 365 #define ETH_TOTAL_MIN (ETH_HEADER_LEN + ETH_PAYLOAD_MIN)
 366 #define ETH_TOTAL_MAX (ETH_HEADER_LEN + ETH_PAYLOAD_MAX)
 367 #define ETH_VLAN_TOTAL_MAX (ETH_HEADER_LEN + VLAN_HEADER_LEN + ETH_PAYLOAD_MAX)
 368 OVS_PACKED(
 369 struct eth_header {
 370     struct eth_addr eth_dst;
 371     struct eth_addr eth_src;
 372     ovs_be16 eth_type;
 373 });
 374 BUILD_ASSERT_DECL(ETH_HEADER_LEN == sizeof(struct eth_header));
 375
 376 #define LLC_DSAP_SNAP 0xaa
 377 #define LLC_SSAP_SNAP 0xaa
 378 #define LLC_CNTL_SNAP 3
 379
 380 #define LLC_HEADER_LEN 3
 381 OVS_PACKED(
 382 struct llc_header {
 383     uint8_t llc_dsap;
 384     uint8_t llc_ssap;
 385     uint8_t llc_cntl;
 386 });
 387 BUILD_ASSERT_DECL(LLC_HEADER_LEN == sizeof(struct llc_header));
 388
 389 /* LLC field values used for STP frames. */
 390 #define STP_LLC_SSAP 0x42
 391 #define STP_LLC_DSAP 0x42
 392 #define STP_LLC_CNTL 0x03
 393
 394 #define SNAP_ORG_ETHERNET "\0\0" /* The compiler adds a null byte, so
 395                                     sizeof(SNAP_ORG_ETHERNET) == 3. */
 396 #define SNAP_HEADER_LEN 5
 397 OVS_PACKED(
 398 struct snap_header {
 399     uint8_t snap_org[3];
 400     ovs_be16 snap_type;
 401 });
 402 BUILD_ASSERT_DECL(SNAP_HEADER_LEN == sizeof(struct snap_header));
 403
 404 #define LLC_SNAP_HEADER_LEN (LLC_HEADER_LEN + SNAP_HEADER_LEN)
 405 OVS_PACKED(
 406 struct llc_snap_header {
 407     struct llc_header llc;
 408     struct snap_header snap;
 409 });
 410 BUILD_ASSERT_DECL(LLC_SNAP_HEADER_LEN == sizeof(struct llc_snap_header));
 411
 412 #define VLAN_VID_MASK 0x0fff
 413 #define VLAN_VID_SHIFT 0
 414
 415 #define VLAN_PCP_MASK 0xe000
 416 #define VLAN_PCP_SHIFT 13
 417
 418 #define VLAN_CFI 0x1000
 419 #define VLAN_CFI_SHIFT 12
 420
 421 /* Given the vlan_tci field from an 802.1Q header, in network byte order,
 422  * returns the VLAN ID in host byte order. */
 423 static inline uint16_t
 424 vlan_tci_to_vid(ovs_be16 vlan_tci)
 425 {
 426     return (ntohs(vlan_tci) & VLAN_VID_MASK) >> VLAN_VID_SHIFT;
 427 }
 428
 429 /* Given the vlan_tci field from an 802.1Q header, in network byte order,
 430  * returns the priority code point (PCP) in host byte order. */
 431 static inline int
 432 vlan_tci_to_pcp(ovs_be16 vlan_tci)
 433 {
 434     return (ntohs(vlan_tci) & VLAN_PCP_MASK) >> VLAN_PCP_SHIFT;
 435 }
 436
 437 /* Given the vlan_tci field from an 802.1Q header, in network byte order,
 438  * returns the Canonical Format Indicator (CFI). */
 439 static inline int
 440 vlan_tci_to_cfi(ovs_be16 vlan_tci)
 441 {
 442     return (vlan_tci & htons(VLAN_CFI)) != 0;
 443 }
 444
 445 #define VLAN_HEADER_LEN 4
 446 struct vlan_header {
 447     ovs_be16 vlan_tci;          /* Lowest 12 bits are VLAN ID. */
 448     ovs_be16 vlan_next_type;
 449 };
 450 BUILD_ASSERT_DECL(VLAN_HEADER_LEN == sizeof(struct vlan_header));
 451
 452 #define VLAN_ETH_HEADER_LEN (ETH_HEADER_LEN + VLAN_HEADER_LEN)
 453 OVS_PACKED(
 454 struct vlan_eth_header {
 455     struct eth_addr veth_dst;
 456     struct eth_addr veth_src;
 457     ovs_be16 veth_type;         /* Always htons(ETH_TYPE_VLAN). */
 458     ovs_be16 veth_tci;          /* Lowest 12 bits are VLAN ID. */
 459     ovs_be16 veth_next_type;
 460 });
 461 BUILD_ASSERT_DECL(VLAN_ETH_HEADER_LEN == sizeof(struct vlan_eth_header));
 462
 463 /* MPLS related definitions */
 464 #define MPLS_TTL_MASK       0x000000ff
 465 #define MPLS_TTL_SHIFT      0
 466
 467 #define MPLS_BOS_MASK       0x00000100
 468 #define MPLS_BOS_SHIFT      8
 469
 470 #define MPLS_TC_MASK        0x00000e00
 471 #define MPLS_TC_SHIFT       9
 472
 473 #define MPLS_LABEL_MASK     0xfffff000
 474 #define MPLS_LABEL_SHIFT    12
 475
 476 #define MPLS_HLEN           4
 477
 478 struct mpls_hdr {
 479     ovs_16aligned_be32 mpls_lse;
 480 };
 481 BUILD_ASSERT_DECL(MPLS_HLEN == sizeof(struct mpls_hdr));
 482
 483 /* Given a mpls label stack entry in network byte order
 484  * return mpls label in host byte order */
 485 static inline uint32_t
 486 mpls_lse_to_label(ovs_be32 mpls_lse)
 487 {
 488     return (ntohl(mpls_lse) & MPLS_LABEL_MASK) >> MPLS_LABEL_SHIFT;
 489 }
 490
 491 /* Given a mpls label stack entry in network byte order
 492  * return mpls tc */
 493 static inline uint8_t
 494 mpls_lse_to_tc(ovs_be32 mpls_lse)
 495 {
 496     return (ntohl(mpls_lse) & MPLS_TC_MASK) >> MPLS_TC_SHIFT;
 497 }
 498
 499 /* Given a mpls label stack entry in network byte order
 500  * return mpls ttl */
 501 static inline uint8_t
 502 mpls_lse_to_ttl(ovs_be32 mpls_lse)
 503 {
 504     return (ntohl(mpls_lse) & MPLS_TTL_MASK) >> MPLS_TTL_SHIFT;
 505 }
 506
 507 /* Set TTL in mpls lse. */
 508 static inline void
 509 flow_set_mpls_lse_ttl(ovs_be32 *mpls_lse, uint8_t ttl)
 510 {
 511     *mpls_lse &= ~htonl(MPLS_TTL_MASK);
 512     *mpls_lse |= htonl(ttl << MPLS_TTL_SHIFT);
 513 }
 514
 515 /* Given a mpls label stack entry in network byte order
 516  * return mpls BoS bit  */
 517 static inline uint8_t
 518 mpls_lse_to_bos(ovs_be32 mpls_lse)
 519 {
 520     return (mpls_lse & htonl(MPLS_BOS_MASK)) != 0;
 521 }
 522
 523 #define IP_FMT "%"PRIu32".%"PRIu32".%"PRIu32".%"PRIu32
 524 #define IP_ARGS(ip)                             \
 525     ntohl(ip) >> 24,                            \
 526     (ntohl(ip) >> 16) & 0xff,                   \
 527     (ntohl(ip) >> 8) & 0xff,                    \
 528     ntohl(ip) & 0xff
 529
 530 /* Example:
 531  *
 532  * char *string = "1 33.44.55.66 2";
 533  * ovs_be32 ip;
 534  * int a, b;
 535  *
 536  * if (ovs_scan(string, "%d"IP_SCAN_FMT"%d", &a, IP_SCAN_ARGS(&ip), &b)) {
 537  *     ...
 538  * }
 539  */
 540 #define IP_SCAN_FMT "%"SCNu8".%"SCNu8".%"SCNu8".%"SCNu8
 541 #define IP_SCAN_ARGS(ip)                                    \
 542         ((void) (ovs_be32) *(ip), &((uint8_t *) ip)[0]),    \
 543         &((uint8_t *) ip)[1],                               \
 544         &((uint8_t *) ip)[2],                               \
 545         &((uint8_t *) ip)[3]
 546
 547 /* Returns true if 'netmask' is a CIDR netmask, that is, if it consists of N
 548  * high-order 1-bits and 32-N low-order 0-bits. */
 549 static inline bool
 550 ip_is_cidr(ovs_be32 netmask)
 551 {
 552     uint32_t x = ~ntohl(netmask);
 553     return !(x & (x + 1));
 554 }
 555 static inline bool
 556 ip_is_multicast(ovs_be32 ip)
 557 {
 558     return (ip & htonl(0xf0000000)) == htonl(0xe0000000);
 559 }
 560 static inline bool
 561 ip_is_local_multicast(ovs_be32 ip)
 562 {
 563     return (ip & htonl(0xffffff00)) == htonl(0xe0000000);
 564 }
 565 int ip_count_cidr_bits(ovs_be32 netmask);
 566 void ip_format_masked(ovs_be32 ip, ovs_be32 mask, struct ds *);
 567
 568 #define IP_VER(ip_ihl_ver) ((ip_ihl_ver) >> 4)
 569 #define IP_IHL(ip_ihl_ver) ((ip_ihl_ver) & 15)
 570 #define IP_IHL_VER(ihl, ver) (((ver) << 4) | (ihl))
 571
 572 #ifndef IPPROTO_SCTP
 573 #define IPPROTO_SCTP 132
 574 #endif
 575
 576 /* TOS fields. */
 577 #define IP_ECN_NOT_ECT 0x0
 578 #define IP_ECN_ECT_1 0x01
 579 #define IP_ECN_ECT_0 0x02
 580 #define IP_ECN_CE 0x03
 581 #define IP_ECN_MASK 0x03
 582 #define IP_DSCP_MASK 0xfc
 583
 584 #define IP_VERSION 4
 585
 586 #define IP_DONT_FRAGMENT  0x4000 /* Don't fragment. */
 587 #define IP_MORE_FRAGMENTS 0x2000 /* More fragments. */
 588 #define IP_FRAG_OFF_MASK  0x1fff /* Fragment offset. */
 589 #define IP_IS_FRAGMENT(ip_frag_off) \
 590         ((ip_frag_off) & htons(IP_MORE_FRAGMENTS | IP_FRAG_OFF_MASK))
 591
 592 #define IP_HEADER_LEN 20
 593 struct ip_header {
 594     uint8_t ip_ihl_ver;
 595     uint8_t ip_tos;
 596     ovs_be16 ip_tot_len;
 597     ovs_be16 ip_id;
 598     ovs_be16 ip_frag_off;
 599     uint8_t ip_ttl;
 600     uint8_t ip_proto;
 601     ovs_be16 ip_csum;
 602     ovs_16aligned_be32 ip_src;
 603     ovs_16aligned_be32 ip_dst;
 604 };
 605
 606 BUILD_ASSERT_DECL(IP_HEADER_LEN == sizeof(struct ip_header));
 607
 608 #define ICMP_HEADER_LEN 8
 609 struct icmp_header {
 610     uint8_t icmp_type;
 611     uint8_t icmp_code;
 612     ovs_be16 icmp_csum;
 613     union {
 614         struct {
 615             ovs_be16 id;
 616             ovs_be16 seq;
 617         } echo;
 618         struct {
 619             ovs_be16 empty;
 620             ovs_be16 mtu;
 621         } frag;
 622         ovs_16aligned_be32 gateway;
 623     } icmp_fields;
 624 };
 625 BUILD_ASSERT_DECL(ICMP_HEADER_LEN == sizeof(struct icmp_header));
 626
 627 #define IGMP_HEADER_LEN 8
 628 struct igmp_header {
 629     uint8_t igmp_type;
 630     uint8_t igmp_code;
 631     ovs_be16 igmp_csum;
 632     ovs_16aligned_be32 group;
 633 };
 634 BUILD_ASSERT_DECL(IGMP_HEADER_LEN == sizeof(struct igmp_header));
 635
 636 #define IGMPV3_HEADER_LEN 8
 637 struct igmpv3_header {
 638     uint8_t type;
 639     uint8_t rsvr1;
 640     ovs_be16 csum;
 641     ovs_be16 rsvr2;
 642     ovs_be16 ngrp;
 643 };
 644 BUILD_ASSERT_DECL(IGMPV3_HEADER_LEN == sizeof(struct igmpv3_header));
 645
 646 #define IGMPV3_RECORD_LEN 8
 647 struct igmpv3_record {
 648     uint8_t type;
 649     uint8_t aux_len;
 650     ovs_be16 nsrcs;
 651     ovs_16aligned_be32 maddr;
 652 };
 653 BUILD_ASSERT_DECL(IGMPV3_RECORD_LEN == sizeof(struct igmpv3_record));
 654
 655 #define IGMP_HOST_MEMBERSHIP_QUERY    0x11 /* From RFC1112 */
 656 #define IGMP_HOST_MEMBERSHIP_REPORT   0x12 /* Ditto */
 657 #define IGMPV2_HOST_MEMBERSHIP_REPORT 0x16 /* V2 version of 0x12 */
 658 #define IGMP_HOST_LEAVE_MESSAGE       0x17
 659 #define IGMPV3_HOST_MEMBERSHIP_REPORT 0x22 /* V3 version of 0x12 */
 660
 661 /*
 662  * IGMPv3 and MLDv2 use the same codes.
 663  */
 664 #define IGMPV3_MODE_IS_INCLUDE 1
 665 #define IGMPV3_MODE_IS_EXCLUDE 2
 666 #define IGMPV3_CHANGE_TO_INCLUDE_MODE 3
 667 #define IGMPV3_CHANGE_TO_EXCLUDE_MODE 4
 668 #define IGMPV3_ALLOW_NEW_SOURCES 5
 669 #define IGMPV3_BLOCK_OLD_SOURCES 6
 670
 671 #define SCTP_HEADER_LEN 12
 672 struct sctp_header {
 673     ovs_be16 sctp_src;
 674     ovs_be16 sctp_dst;
 675     ovs_16aligned_be32 sctp_vtag;
 676     ovs_16aligned_be32 sctp_csum;
 677 };
 678 BUILD_ASSERT_DECL(SCTP_HEADER_LEN == sizeof(struct sctp_header));
 679
 680 #define UDP_HEADER_LEN 8
 681 struct udp_header {
 682     ovs_be16 udp_src;
 683     ovs_be16 udp_dst;
 684     ovs_be16 udp_len;
 685     ovs_be16 udp_csum;
 686 };
 687 BUILD_ASSERT_DECL(UDP_HEADER_LEN == sizeof(struct udp_header));
 688
 689 #define TCP_FIN 0x001
 690 #define TCP_SYN 0x002
 691 #define TCP_RST 0x004
 692 #define TCP_PSH 0x008
 693 #define TCP_ACK 0x010
 694 #define TCP_URG 0x020
 695 #define TCP_ECE 0x040
 696 #define TCP_CWR 0x080
 697 #define TCP_NS  0x100
 698
 699 #define TCP_CTL(flags, offset) (htons((flags) | ((offset) << 12)))
 700 #define TCP_FLAGS(tcp_ctl) (ntohs(tcp_ctl) & 0x0fff)
 701 #define TCP_FLAGS_BE16(tcp_ctl) ((tcp_ctl) & htons(0x0fff))
 702 #define TCP_OFFSET(tcp_ctl) (ntohs(tcp_ctl) >> 12)
 703
 704 #define TCP_HEADER_LEN 20
 705 struct tcp_header {
 706     ovs_be16 tcp_src;
 707     ovs_be16 tcp_dst;
 708     ovs_16aligned_be32 tcp_seq;
 709     ovs_16aligned_be32 tcp_ack;
 710     ovs_be16 tcp_ctl;
 711     ovs_be16 tcp_winsz;
 712     ovs_be16 tcp_csum;
 713     ovs_be16 tcp_urg;
 714 };
 715 BUILD_ASSERT_DECL(TCP_HEADER_LEN == sizeof(struct tcp_header));
 716
 717 /* Connection states */
 718 #define CS_NEW               0x01
 719 #define CS_ESTABLISHED       0x02
 720 #define CS_RELATED           0x04
 721 #define CS_INVALID           0x20
 722 #define CS_REPLY_DIR         0x40
 723 #define CS_TRACKED           0x80
 724
 725 /* Undefined connection state bits. */
 726 #define CS_SUPPORTED_MASK    (CS_NEW | CS_ESTABLISHED | CS_RELATED \
 727                               | CS_INVALID | CS_REPLY_DIR | CS_TRACKED)
 728 #define CS_UNSUPPORTED_MASK  (~(uint32_t)CS_SUPPORTED_MASK)
 729
 730 #define ARP_HRD_ETHERNET 1
 731 #define ARP_PRO_IP 0x0800
 732 #define ARP_OP_REQUEST 1
 733 #define ARP_OP_REPLY 2
 734 #define ARP_OP_RARP 3
 735
 736 #define ARP_ETH_HEADER_LEN 28
 737 struct arp_eth_header {
 738     /* Generic members. */
 739     ovs_be16 ar_hrd;           /* Hardware type. */
 740     ovs_be16 ar_pro;           /* Protocol type. */
 741     uint8_t ar_hln;            /* Hardware address length. */
 742     uint8_t ar_pln;            /* Protocol address length. */
 743     ovs_be16 ar_op;            /* Opcode. */
 744
 745     /* Ethernet+IPv4 specific members. */
 746     struct eth_addr ar_sha;     /* Sender hardware address. */
 747     ovs_16aligned_be32 ar_spa;  /* Sender protocol address. */
 748     struct eth_addr ar_tha;     /* Target hardware address. */
 749     ovs_16aligned_be32 ar_tpa;  /* Target protocol address. */
 750 };
 751 BUILD_ASSERT_DECL(ARP_ETH_HEADER_LEN == sizeof(struct arp_eth_header));
 752
 753 /* Like struct in6_addr, but whereas that struct requires 32-bit alignment on
 754  * most implementations, this one only requires 16-bit alignment. */
 755 union ovs_16aligned_in6_addr {
 756     ovs_be16 be16[8];
 757     ovs_16aligned_be32 be32[4];
 758 };
 759
 760 /* Like struct in6_hdr, but whereas that struct requires 32-bit alignment, this
 761  * one only requires 16-bit alignment. */
 762 struct ovs_16aligned_ip6_hdr {
 763     union {
 764         struct ovs_16aligned_ip6_hdrctl {
 765             ovs_16aligned_be32 ip6_un1_flow;
 766             ovs_be16 ip6_un1_plen;
 767             uint8_t ip6_un1_nxt;
 768             uint8_t ip6_un1_hlim;
 769         } ip6_un1;
 770         uint8_t ip6_un2_vfc;
 771     } ip6_ctlun;
 772     union ovs_16aligned_in6_addr ip6_src;
 773     union ovs_16aligned_in6_addr ip6_dst;
 774 };
 775
 776 /* Like struct in6_frag, but whereas that struct requires 32-bit alignment,
 777  * this one only requires 16-bit alignment. */
 778 struct ovs_16aligned_ip6_frag {
 779     uint8_t ip6f_nxt;
 780     uint8_t ip6f_reserved;
 781     ovs_be16 ip6f_offlg;
 782     ovs_16aligned_be32 ip6f_ident;
 783 };
 784
 785 #define ICMP6_HEADER_LEN 4
 786 struct icmp6_header {
 787     uint8_t icmp6_type;
 788     uint8_t icmp6_code;
 789     ovs_be16 icmp6_cksum;
 790 };
 791 BUILD_ASSERT_DECL(ICMP6_HEADER_LEN == sizeof(struct icmp6_header));
 792
 793 /* Neighbor Discovery option field.
 794  * ND options are always a multiple of 8 bytes in size. */
 795 #define ND_OPT_LEN 8
 796 struct ovs_nd_opt {
 797     uint8_t  nd_opt_type;      /* Values defined in icmp6.h */
 798     uint8_t  nd_opt_len;       /* in units of 8 octets (the size of this struct) */
 799     struct eth_addr nd_opt_mac;   /* Ethernet address in the case of SLL or TLL options */
 800 };
 801 BUILD_ASSERT_DECL(ND_OPT_LEN == sizeof(struct ovs_nd_opt));
 802
 803 /* Like struct nd_msg (from ndisc.h), but whereas that struct requires 32-bit
 804  * alignment, this one only requires 16-bit alignment. */
 805 #define ND_MSG_LEN 24
 806 struct ovs_nd_msg {
 807     struct icmp6_header icmph;
 808     ovs_16aligned_be32 rco_flags;
 809     union ovs_16aligned_in6_addr target;
 810     struct ovs_nd_opt options[0];
 811 };
 812 BUILD_ASSERT_DECL(ND_MSG_LEN == sizeof(struct ovs_nd_msg));
 813
 814 /*
 815  * Use the same struct for MLD and MLD2, naming members as the defined fields in
 816  * in the corresponding version of the protocol, though they are reserved in the
 817  * other one.
 818  */
 819 #define MLD_HEADER_LEN 8
 820 struct mld_header {
 821     uint8_t type;
 822     uint8_t code;
 823     ovs_be16 csum;
 824     ovs_be16 mrd;
 825     ovs_be16 ngrp;
 826 };
 827 BUILD_ASSERT_DECL(MLD_HEADER_LEN == sizeof(struct mld_header));
 828
 829 #define MLD2_RECORD_LEN 20
 830 struct mld2_record {
 831     uint8_t type;
 832     uint8_t aux_len;
 833     ovs_be16 nsrcs;
 834     union ovs_16aligned_in6_addr maddr;
 835 };
 836 BUILD_ASSERT_DECL(MLD2_RECORD_LEN == sizeof(struct mld2_record));
 837
 838 #define MLD_QUERY 130
 839 #define MLD_REPORT 131
 840 #define MLD_DONE 132
 841 #define MLD2_REPORT 143
 842
 843 /* The IPv6 flow label is in the lower 20 bits of the first 32-bit word. */
 844 #define IPV6_LABEL_MASK 0x000fffff
 845
 846 /* Example:
 847  *
 848  * char *string = "1 ::1 2";
 849  * char ipv6_s[IPV6_SCAN_LEN + 1];
 850  * struct in6_addr ipv6;
 851  *
 852  * if (ovs_scan(string, "%d"IPV6_SCAN_FMT"%d", &a, ipv6_s, &b)
 853  *     && inet_pton(AF_INET6, ipv6_s, &ipv6) == 1) {
 854  *     ...
 855  * }
 856  */
 857 #define IPV6_SCAN_FMT "%46[0123456789abcdefABCDEF:.]"
 858 #define IPV6_SCAN_LEN 46
 859
 860 extern const struct in6_addr in6addr_exact;
 861 #define IN6ADDR_EXACT_INIT { { { 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, \
 862                                  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff } } }
 863
 864 extern const struct in6_addr in6addr_all_hosts;
 865 #define IN6ADDR_ALL_HOSTS_INIT { { { 0xff,0x02,0x00,0x00,0x00,0x00,0x00,0x00, \
 866                                      0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x01 } } }
 867
 868 static inline bool ipv6_addr_equals(const struct in6_addr *a,
 869                                     const struct in6_addr *b)
 870 {
 871 #ifdef IN6_ARE_ADDR_EQUAL
 872     return IN6_ARE_ADDR_EQUAL(a, b);
 873 #else
 874     return !memcmp(a, b, sizeof(*a));
 875 #endif
 876 }
 877
 878 static inline bool ipv6_mask_is_any(const struct in6_addr *mask) {
 879     return ipv6_addr_equals(mask, &in6addr_any);
 880 }
 881
 882 static inline bool ipv6_mask_is_exact(const struct in6_addr *mask) {
 883     return ipv6_addr_equals(mask, &in6addr_exact);
 884 }
 885
 886 static inline bool ipv6_is_all_hosts(const struct in6_addr *addr) {
 887     return ipv6_addr_equals(addr, &in6addr_all_hosts);
 888 }
 889
 890 static inline bool ipv6_addr_is_set(const struct in6_addr *addr) {
 891     return !ipv6_addr_equals(addr, &in6addr_any);
 892 }
 893
 894 static inline bool ipv6_addr_is_multicast(const struct in6_addr *ip) {
 895     return ip->s6_addr[0] == 0xff;
 896 }
 897
 898 static inline void
 899 in6_addr_set_mapped_ipv4(struct in6_addr *addr, ovs_be32 ip4)
 900 {
 901     union ovs_16aligned_in6_addr *taddr = (void *) addr;
 902     memset(taddr->be16, 0, sizeof(taddr->be16));
 903     taddr->be16[5] = OVS_BE16_MAX;
 904     put_16aligned_be32(&taddr->be32[3], ip4);
 905 }
 906
 907 static inline ovs_be32
 908 in6_addr_get_mapped_ipv4(const struct in6_addr *addr)
 909 {
 910     union ovs_16aligned_in6_addr *taddr = (void *) addr;
 911     if (IN6_IS_ADDR_V4MAPPED(addr)) {
 912         return get_16aligned_be32(&taddr->be32[3]);
 913     } else {
 914         return INADDR_ANY;
 915     }
 916 }
 917
 918 static inline bool dl_type_is_ip_any(ovs_be16 dl_type)
 919 {
 920     return dl_type == htons(ETH_TYPE_IP)
 921         || dl_type == htons(ETH_TYPE_IPV6);
 922 }
 923
 924 /* Tunnel header */
 925
 926 /* GRE protocol header */
 927 struct gre_base_hdr {
 928     ovs_be16 flags;
 929     ovs_be16 protocol;
 930 };
 931
 932 #define GRE_CSUM        0x8000
 933 #define GRE_ROUTING     0x4000
 934 #define GRE_KEY         0x2000
 935 #define GRE_SEQ         0x1000
 936 #define GRE_STRICT      0x0800
 937 #define GRE_REC         0x0700
 938 #define GRE_FLAGS       0x00F8
 939 #define GRE_VERSION     0x0007
 940
 941 /* VXLAN protocol header */
 942 struct vxlanhdr {
 943     ovs_16aligned_be32 vx_flags;
 944     ovs_16aligned_be32 vx_vni;
 945 };
 946
 947 #define VXLAN_FLAGS 0x08000000  /* struct vxlanhdr.vx_flags required value. */
 948
 949 void format_ipv6_addr(char *addr_str, const struct in6_addr *addr);
 950 void print_ipv6_addr(struct ds *string, const struct in6_addr *addr);
 951 void print_ipv6_mapped(struct ds *string, const struct in6_addr *addr);
 952 void print_ipv6_masked(struct ds *string, const struct in6_addr *addr,
 953                        const struct in6_addr *mask);
 954 struct in6_addr ipv6_addr_bitand(const struct in6_addr *src,
 955                                  const struct in6_addr *mask);
 956 struct in6_addr ipv6_create_mask(int mask);
 957 int ipv6_count_cidr_bits(const struct in6_addr *netmask);
 958 bool ipv6_is_cidr(const struct in6_addr *netmask);
 959
 960 void *eth_compose(struct dp_packet *, const struct eth_addr eth_dst,
 961                   const struct eth_addr eth_src, uint16_t eth_type,
 962                   size_t size);
 963 void *snap_compose(struct dp_packet *, const struct eth_addr eth_dst,
 964                    const struct eth_addr eth_src,
 965                    unsigned int oui, uint16_t snap_type, size_t size);
 966 void packet_set_ipv4(struct dp_packet *, ovs_be32 src, ovs_be32 dst, uint8_t tos,
 967                      uint8_t ttl);
 968 void packet_set_ipv6(struct dp_packet *, uint8_t proto, const ovs_be32 src[4],
 969                      const ovs_be32 dst[4], uint8_t tc,
 970                      ovs_be32 fl, uint8_t hlmit);
 971 void packet_set_tcp_port(struct dp_packet *, ovs_be16 src, ovs_be16 dst);
 972 void packet_set_udp_port(struct dp_packet *, ovs_be16 src, ovs_be16 dst);
 973 void packet_set_sctp_port(struct dp_packet *, ovs_be16 src, ovs_be16 dst);
 974 void packet_set_nd(struct dp_packet *, const ovs_be32 target[4],
 975                    const struct eth_addr sll, const struct eth_addr tll);
 976
 977 void packet_format_tcp_flags(struct ds *, uint16_t);
 978 const char *packet_tcp_flag_to_string(uint32_t flag);
 979 void compose_arp(struct dp_packet *, uint16_t arp_op,
 980                  const struct eth_addr arp_sha,
 981                  const struct eth_addr arp_tha, bool broadcast,
 982                  ovs_be32 arp_spa, ovs_be32 arp_tpa);
 983 uint32_t packet_csum_pseudoheader(const struct ip_header *);
 984
 985 #endif /* packets.h */