lib/packets.h

   1 /*
   2  * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017 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 <stdint.h>
  23 #include <string.h>
  24 #include "compiler.h"
  25 #include "openvswitch/geneve.h"
  26 #include "openvswitch/packets.h"
  27 #include "openvswitch/types.h"
  28 #include "openvswitch/nsh.h"
  29 #include "odp-netlink.h"
  30 #include "random.h"
  31 #include "hash.h"
  32 #include "tun-metadata.h"
  33 #include "unaligned.h"
  34 #include "util.h"
  35
  36 struct dp_packet;
  37 struct ds;
  38
  39 /* Purely internal to OVS userspace. These flags should never be exposed to
  40  * the outside world and so aren't included in the flags mask. */
  41
  42 /* Tunnel information is in userspace datapath format. */
  43 #define FLOW_TNL_F_UDPIF (1 << 4)
  44
  45 static inline bool ipv6_addr_is_set(const struct in6_addr *addr);
  46
  47 static inline bool
  48 flow_tnl_dst_is_set(const struct flow_tnl *tnl)
  49 {
  50     return tnl->ip_dst || ipv6_addr_is_set(&tnl->ipv6_dst);
  51 }
  52
  53 struct in6_addr flow_tnl_dst(const struct flow_tnl *tnl);
  54 struct in6_addr flow_tnl_src(const struct flow_tnl *tnl);
  55
  56 /* Returns an offset to 'src' covering all the meaningful fields in 'src'. */
  57 static inline size_t
  58 flow_tnl_size(const struct flow_tnl *src)
  59 {
  60     if (!flow_tnl_dst_is_set(src)) {
  61         /* Covers ip_dst and ipv6_dst only. */
  62         return offsetof(struct flow_tnl, ip_src);
  63     }
  64     if (src->flags & FLOW_TNL_F_UDPIF) {
  65         /* Datapath format, cover all options we have. */
  66         return offsetof(struct flow_tnl, metadata.opts)
  67             + src->metadata.present.len;
  68     }
  69     if (!src->metadata.present.map) {
  70         /* No TLVs, opts is irrelevant. */
  71         return offsetof(struct flow_tnl, metadata.opts);
  72     }
  73     /* Have decoded TLVs, opts is relevant. */
  74     return sizeof *src;
  75 }
  76
  77 /* Copy flow_tnl, but avoid copying unused portions of tun_metadata.  Unused
  78  * data in 'dst' is NOT cleared, so this must not be used in cases where the
  79  * uninitialized portion may be hashed over. */
  80 static inline void
  81 flow_tnl_copy__(struct flow_tnl *dst, const struct flow_tnl *src)
  82 {
  83     memcpy(dst, src, flow_tnl_size(src));
  84 }
  85
  86 static inline bool
  87 flow_tnl_equal(const struct flow_tnl *a, const struct flow_tnl *b)
  88 {
  89     size_t a_size = flow_tnl_size(a);
  90
  91     return a_size == flow_tnl_size(b) && !memcmp(a, b, a_size);
  92 }
  93
  94 /* Datapath packet metadata */
  95 struct pkt_metadata {
  96 PADDED_MEMBERS_CACHELINE_MARKER(CACHE_LINE_SIZE, cacheline0,
  97     uint32_t recirc_id;         /* Recirculation id carried with the
  98                                    recirculating packets. 0 for packets
  99                                    received from the wire. */
 100     uint32_t dp_hash;           /* hash value computed by the recirculation
 101                                    action. */
 102     uint32_t skb_priority;      /* Packet priority for QoS. */
 103     uint32_t pkt_mark;          /* Packet mark. */
 104     uint8_t  ct_state;          /* Connection state. */
 105     bool ct_orig_tuple_ipv6;
 106     uint16_t ct_zone;           /* Connection zone. */
 107     uint32_t ct_mark;           /* Connection mark. */
 108     ovs_u128 ct_label;          /* Connection label. */
 109     union flow_in_port in_port; /* Input port. */
 110 );
 111
 112 PADDED_MEMBERS_CACHELINE_MARKER(CACHE_LINE_SIZE, cacheline1,
 113     union {                     /* Populated only for non-zero 'ct_state'. */
 114         struct ovs_key_ct_tuple_ipv4 ipv4;
 115         struct ovs_key_ct_tuple_ipv6 ipv6;   /* Used only if                */
 116     } ct_orig_tuple;                         /* 'ct_orig_tuple_ipv6' is set */
 117 );
 118
 119 PADDED_MEMBERS_CACHELINE_MARKER(CACHE_LINE_SIZE, cacheline2,
 120     struct flow_tnl tunnel;     /* Encapsulating tunnel parameters. Note that
 121                                  * if 'ip_dst' == 0, the rest of the fields may
 122                                  * be uninitialized. */
 123 );
 124 };
 125
 126 BUILD_ASSERT_DECL(offsetof(struct pkt_metadata, cacheline0) == 0);
 127 BUILD_ASSERT_DECL(offsetof(struct pkt_metadata, cacheline1) ==
 128                   CACHE_LINE_SIZE);
 129 BUILD_ASSERT_DECL(offsetof(struct pkt_metadata, cacheline2) ==
 130                   2 * CACHE_LINE_SIZE);
 131
 132 static inline void
 133 pkt_metadata_init_tnl(struct pkt_metadata *md)
 134 {
 135     /* Zero up through the tunnel metadata options. The length and table
 136      * are before this and as long as they are empty, the options won't
 137      * be looked at. */
 138     memset(md, 0, offsetof(struct pkt_metadata, tunnel.metadata.opts));
 139 }
 140
 141 static inline void
 142 pkt_metadata_init(struct pkt_metadata *md, odp_port_t port)
 143 {
 144     /* This is called for every packet in userspace datapath and affects
 145      * performance if all the metadata is initialized. Hence, fields should
 146      * only be zeroed out when necessary.
 147      *
 148      * Initialize only till ct_state. Once the ct_state is zeroed out rest
 149      * of ct fields will not be looked at unless ct_state != 0.
 150      */
 151     memset(md, 0, offsetof(struct pkt_metadata, ct_orig_tuple_ipv6));
 152
 153     /* It can be expensive to zero out all of the tunnel metadata. However,
 154      * we can just zero out ip_dst and the rest of the data will never be
 155      * looked at. */
 156     md->tunnel.ip_dst = 0;
 157     md->tunnel.ipv6_dst = in6addr_any;
 158     md->in_port.odp_port = port;
 159 }
 160
 161 /* This function prefetches the cachelines touched by pkt_metadata_init()
 162  * For performance reasons the two functions should be kept in sync. */
 163 static inline void
 164 pkt_metadata_prefetch_init(struct pkt_metadata *md)
 165 {
 166     /* Prefetch cacheline0 as members till ct_state and odp_port will
 167      * be initialized later in pkt_metadata_init(). */
 168     OVS_PREFETCH(md->cacheline0);
 169
 170     /* Prefetch cachline2 as ip_dst & ipv6_dst fields will be initialized. */
 171     OVS_PREFETCH(md->cacheline2);
 172 }
 173
 174 bool dpid_from_string(const char *s, uint64_t *dpidp);
 175
 176 #define ETH_ADDR_LEN           6
 177
 178 static const struct eth_addr eth_addr_broadcast OVS_UNUSED
 179     = { { { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff } } };
 180
 181 static const struct eth_addr eth_addr_exact OVS_UNUSED
 182     = { { { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff } } };
 183
 184 static const struct eth_addr eth_addr_zero OVS_UNUSED
 185     = { { { 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 } } };
 186 static const struct eth_addr64 eth_addr64_zero OVS_UNUSED
 187     = { { { 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 } } };
 188
 189 static const struct eth_addr eth_addr_stp OVS_UNUSED
 190     = { { { 0x01, 0x80, 0xC2, 0x00, 0x00, 0x00 } } };
 191
 192 static const struct eth_addr eth_addr_lacp OVS_UNUSED
 193     = { { { 0x01, 0x80, 0xC2, 0x00, 0x00, 0x02 } } };
 194
 195 static const struct eth_addr eth_addr_bfd OVS_UNUSED
 196     = { { { 0x00, 0x23, 0x20, 0x00, 0x00, 0x01 } } };
 197
 198 static inline bool eth_addr_is_broadcast(const struct eth_addr a)
 199 {
 200     return (a.be16[0] & a.be16[1] & a.be16[2]) == htons(0xffff);
 201 }
 202
 203 static inline bool eth_addr_is_multicast(const struct eth_addr a)
 204 {
 205     return a.ea[0] & 1;
 206 }
 207
 208 static inline bool eth_addr_is_local(const struct eth_addr a)
 209 {
 210     /* Local if it is either a locally administered address or a Nicira random
 211      * address. */
 212     return a.ea[0] & 2
 213         || (a.be16[0] == htons(0x0023)
 214             && (a.be16[1] & htons(0xff80)) == htons(0x2080));
 215 }
 216 static inline bool eth_addr_is_zero(const struct eth_addr a)
 217 {
 218     return !(a.be16[0] | a.be16[1] | a.be16[2]);
 219 }
 220 static inline bool eth_addr64_is_zero(const struct eth_addr64 a)
 221 {
 222     return !(a.be16[0] | a.be16[1] | a.be16[2] | a.be16[3]);
 223 }
 224
 225 static inline int eth_mask_is_exact(const struct eth_addr a)
 226 {
 227     return (a.be16[0] & a.be16[1] & a.be16[2]) == htons(0xffff);
 228 }
 229
 230 static inline int eth_addr_compare_3way(const struct eth_addr a,
 231                                         const struct eth_addr b)
 232 {
 233     return memcmp(&a, &b, sizeof a);
 234 }
 235 static inline int eth_addr64_compare_3way(const struct eth_addr64 a,
 236                                           const struct eth_addr64 b)
 237 {
 238     return memcmp(&a, &b, sizeof a);
 239 }
 240
 241 static inline bool eth_addr_equals(const struct eth_addr a,
 242                                    const struct eth_addr b)
 243 {
 244     return !eth_addr_compare_3way(a, b);
 245 }
 246 static inline bool eth_addr64_equals(const struct eth_addr64 a,
 247                                      const struct eth_addr64 b)
 248 {
 249     return !eth_addr64_compare_3way(a, b);
 250 }
 251
 252 static inline bool eth_addr_equal_except(const struct eth_addr a,
 253                                          const struct eth_addr b,
 254                                          const struct eth_addr mask)
 255 {
 256     return !(((a.be16[0] ^ b.be16[0]) & mask.be16[0])
 257              || ((a.be16[1] ^ b.be16[1]) & mask.be16[1])
 258              || ((a.be16[2] ^ b.be16[2]) & mask.be16[2]));
 259 }
 260
 261 static inline uint64_t eth_addr_to_uint64(const struct eth_addr ea)
 262 {
 263     return (((uint64_t) ntohs(ea.be16[0]) << 32)
 264             | ((uint64_t) ntohs(ea.be16[1]) << 16)
 265             | ntohs(ea.be16[2]));
 266 }
 267
 268 static inline uint64_t eth_addr_vlan_to_uint64(const struct eth_addr ea,
 269                                                uint16_t vlan)
 270 {
 271     return (((uint64_t)vlan << 48) | eth_addr_to_uint64(ea));
 272 }
 273
 274 static inline void eth_addr_from_uint64(uint64_t x, struct eth_addr *ea)
 275 {
 276     ea->be16[0] = htons(x >> 32);
 277     ea->be16[1] = htons((x & 0xFFFF0000) >> 16);
 278     ea->be16[2] = htons(x & 0xFFFF);
 279 }
 280
 281 static inline struct eth_addr eth_addr_invert(const struct eth_addr src)
 282 {
 283     struct eth_addr dst;
 284
 285     for (int i = 0; i < ARRAY_SIZE(src.be16); i++) {
 286         dst.be16[i] = ~src.be16[i];
 287     }
 288
 289     return dst;
 290 }
 291
 292 static inline void eth_addr_mark_random(struct eth_addr *ea)
 293 {
 294     ea->ea[0] &= ~1;                /* Unicast. */
 295     ea->ea[0] |= 2;                 /* Private. */
 296 }
 297
 298 static inline void eth_addr_random(struct eth_addr *ea)
 299 {
 300     random_bytes((uint8_t *)ea, sizeof *ea);
 301     eth_addr_mark_random(ea);
 302 }
 303
 304 static inline void eth_addr_nicira_random(struct eth_addr *ea)
 305 {
 306     eth_addr_random(ea);
 307
 308     /* Set the OUI to the Nicira one. */
 309     ea->ea[0] = 0x00;
 310     ea->ea[1] = 0x23;
 311     ea->ea[2] = 0x20;
 312
 313     /* Set the top bit to indicate random Nicira address. */
 314     ea->ea[3] |= 0x80;
 315 }
 316 static inline uint32_t hash_mac(const struct eth_addr ea,
 317                                 const uint16_t vlan, const uint32_t basis)
 318 {
 319     return hash_uint64_basis(eth_addr_vlan_to_uint64(ea, vlan), basis);
 320 }
 321
 322 bool eth_addr_is_reserved(const struct eth_addr);
 323 bool eth_addr_from_string(const char *, struct eth_addr *);
 324
 325 void compose_rarp(struct dp_packet *, const struct eth_addr);
 326
 327 void eth_push_vlan(struct dp_packet *, ovs_be16 tpid, ovs_be16 tci);
 328 void eth_pop_vlan(struct dp_packet *);
 329
 330 const char *eth_from_hex(const char *hex, struct dp_packet **packetp);
 331 void eth_format_masked(const struct eth_addr ea,
 332                        const struct eth_addr *mask, struct ds *s);
 333
 334 void set_mpls_lse(struct dp_packet *, ovs_be32 label);
 335 void push_mpls(struct dp_packet *packet, ovs_be16 ethtype, ovs_be32 lse);
 336 void pop_mpls(struct dp_packet *, ovs_be16 ethtype);
 337
 338 void set_mpls_lse_ttl(ovs_be32 *lse, uint8_t ttl);
 339 void set_mpls_lse_tc(ovs_be32 *lse, uint8_t tc);
 340 void set_mpls_lse_label(ovs_be32 *lse, ovs_be32 label);
 341 void set_mpls_lse_bos(ovs_be32 *lse, uint8_t bos);
 342 ovs_be32 set_mpls_lse_values(uint8_t ttl, uint8_t tc, uint8_t bos,
 343                              ovs_be32 label);
 344
 345 /* Example:
 346  *
 347  * struct eth_addr mac;
 348  *    [...]
 349  * printf("The Ethernet address is "ETH_ADDR_FMT"\n", ETH_ADDR_ARGS(mac));
 350  *
 351  */
 352 #define ETH_ADDR_FMT                                                    \
 353     "%02"PRIx8":%02"PRIx8":%02"PRIx8":%02"PRIx8":%02"PRIx8":%02"PRIx8
 354 #define ETH_ADDR_ARGS(EA) ETH_ADDR_BYTES_ARGS((EA).ea)
 355 #define ETH_ADDR_BYTES_ARGS(EAB) \
 356          (EAB)[0], (EAB)[1], (EAB)[2], (EAB)[3], (EAB)[4], (EAB)[5]
 357 #define ETH_ADDR_STRLEN 17
 358
 359 /* Example:
 360  *
 361  * struct eth_addr64 eui64;
 362  *    [...]
 363  * printf("The EUI-64 address is "ETH_ADDR64_FMT"\n", ETH_ADDR64_ARGS(mac));
 364  *
 365  */
 366 #define ETH_ADDR64_FMT \
 367     "%02"PRIx8":%02"PRIx8":%02"PRIx8":%02"PRIx8":" \
 368     "%02"PRIx8":%02"PRIx8":%02"PRIx8":%02"PRIx8
 369 #define ETH_ADDR64_ARGS(EA) ETH_ADDR64_BYTES_ARGS((EA).ea64)
 370 #define ETH_ADDR64_BYTES_ARGS(EAB) \
 371          (EAB)[0], (EAB)[1], (EAB)[2], (EAB)[3], \
 372          (EAB)[4], (EAB)[5], (EAB)[6], (EAB)[7]
 373 #define ETH_ADDR64_STRLEN 23
 374
 375 /* Example:
 376  *
 377  * char *string = "1 00:11:22:33:44:55 2";
 378  * struct eth_addr mac;
 379  * int a, b;
 380  *
 381  * if (ovs_scan(string, "%d"ETH_ADDR_SCAN_FMT"%d",
 382  *              &a, ETH_ADDR_SCAN_ARGS(mac), &b)) {
 383  *     ...
 384  * }
 385  */
 386 #define ETH_ADDR_SCAN_FMT "%"SCNx8":%"SCNx8":%"SCNx8":%"SCNx8":%"SCNx8":%"SCNx8
 387 #define ETH_ADDR_SCAN_ARGS(EA) \
 388     &(EA).ea[0], &(EA).ea[1], &(EA).ea[2], &(EA).ea[3], &(EA).ea[4], &(EA).ea[5]
 389
 390 #define ETH_TYPE_IP            0x0800
 391 #define ETH_TYPE_ARP           0x0806
 392 #define ETH_TYPE_TEB           0x6558
 393 #define ETH_TYPE_VLAN_8021Q    0x8100
 394 #define ETH_TYPE_VLAN          ETH_TYPE_VLAN_8021Q
 395 #define ETH_TYPE_VLAN_8021AD   0x88a8
 396 #define ETH_TYPE_IPV6          0x86dd
 397 #define ETH_TYPE_LACP          0x8809
 398 #define ETH_TYPE_RARP          0x8035
 399 #define ETH_TYPE_MPLS          0x8847
 400 #define ETH_TYPE_MPLS_MCAST    0x8848
 401 #define ETH_TYPE_NSH           0x894f
 402
 403 static inline bool eth_type_mpls(ovs_be16 eth_type)
 404 {
 405     return eth_type == htons(ETH_TYPE_MPLS) ||
 406         eth_type == htons(ETH_TYPE_MPLS_MCAST);
 407 }
 408
 409 static inline bool eth_type_vlan(ovs_be16 eth_type)
 410 {
 411     return eth_type == htons(ETH_TYPE_VLAN_8021Q) ||
 412         eth_type == htons(ETH_TYPE_VLAN_8021AD);
 413 }
 414
 415
 416 /* Minimum value for an Ethernet type.  Values below this are IEEE 802.2 frame
 417  * lengths. */
 418 #define ETH_TYPE_MIN           0x600
 419
 420 #define ETH_HEADER_LEN 14
 421 #define ETH_PAYLOAD_MIN 46
 422 #define ETH_PAYLOAD_MAX 1500
 423 #define ETH_TOTAL_MIN (ETH_HEADER_LEN + ETH_PAYLOAD_MIN)
 424 #define ETH_TOTAL_MAX (ETH_HEADER_LEN + ETH_PAYLOAD_MAX)
 425 #define ETH_VLAN_TOTAL_MAX (ETH_HEADER_LEN + VLAN_HEADER_LEN + ETH_PAYLOAD_MAX)
 426 struct eth_header {
 427     struct eth_addr eth_dst;
 428     struct eth_addr eth_src;
 429     ovs_be16 eth_type;
 430 };
 431 BUILD_ASSERT_DECL(ETH_HEADER_LEN == sizeof(struct eth_header));
 432
 433 void push_eth(struct dp_packet *packet, const struct eth_addr *dst,
 434               const struct eth_addr *src);
 435 void pop_eth(struct dp_packet *packet);
 436
 437 void encap_nsh(struct dp_packet *packet,
 438                const struct ovs_action_encap_nsh *encap_nsh);
 439 bool decap_nsh(struct dp_packet *packet);
 440
 441 #define LLC_DSAP_SNAP 0xaa
 442 #define LLC_SSAP_SNAP 0xaa
 443 #define LLC_CNTL_SNAP 3
 444
 445 #define LLC_HEADER_LEN 3
 446 struct llc_header {
 447     uint8_t llc_dsap;
 448     uint8_t llc_ssap;
 449     uint8_t llc_cntl;
 450 };
 451 BUILD_ASSERT_DECL(LLC_HEADER_LEN == sizeof(struct llc_header));
 452
 453 /* LLC field values used for STP frames. */
 454 #define STP_LLC_SSAP 0x42
 455 #define STP_LLC_DSAP 0x42
 456 #define STP_LLC_CNTL 0x03
 457
 458 #define SNAP_ORG_ETHERNET "\0\0" /* The compiler adds a null byte, so
 459                                     sizeof(SNAP_ORG_ETHERNET) == 3. */
 460 #define SNAP_HEADER_LEN 5
 461 OVS_PACKED(
 462 struct snap_header {
 463     uint8_t snap_org[3];
 464     ovs_be16 snap_type;
 465 });
 466 BUILD_ASSERT_DECL(SNAP_HEADER_LEN == sizeof(struct snap_header));
 467
 468 #define LLC_SNAP_HEADER_LEN (LLC_HEADER_LEN + SNAP_HEADER_LEN)
 469 OVS_PACKED(
 470 struct llc_snap_header {
 471     struct llc_header llc;
 472     struct snap_header snap;
 473 });
 474 BUILD_ASSERT_DECL(LLC_SNAP_HEADER_LEN == sizeof(struct llc_snap_header));
 475
 476 #define VLAN_VID_MASK 0x0fff
 477 #define VLAN_VID_SHIFT 0
 478
 479 #define VLAN_PCP_MASK 0xe000
 480 #define VLAN_PCP_SHIFT 13
 481
 482 #define VLAN_CFI 0x1000
 483 #define VLAN_CFI_SHIFT 12
 484
 485 /* Given the vlan_tci field from an 802.1Q header, in network byte order,
 486  * returns the VLAN ID in host byte order. */
 487 static inline uint16_t
 488 vlan_tci_to_vid(ovs_be16 vlan_tci)
 489 {
 490     return (ntohs(vlan_tci) & VLAN_VID_MASK) >> VLAN_VID_SHIFT;
 491 }
 492
 493 /* Given the vlan_tci field from an 802.1Q header, in network byte order,
 494  * returns the priority code point (PCP) in host byte order. */
 495 static inline int
 496 vlan_tci_to_pcp(ovs_be16 vlan_tci)
 497 {
 498     return (ntohs(vlan_tci) & VLAN_PCP_MASK) >> VLAN_PCP_SHIFT;
 499 }
 500
 501 /* Given the vlan_tci field from an 802.1Q header, in network byte order,
 502  * returns the Canonical Format Indicator (CFI). */
 503 static inline int
 504 vlan_tci_to_cfi(ovs_be16 vlan_tci)
 505 {
 506     return (vlan_tci & htons(VLAN_CFI)) != 0;
 507 }
 508
 509 #define VLAN_HEADER_LEN 4
 510 struct vlan_header {
 511     ovs_be16 vlan_tci;          /* Lowest 12 bits are VLAN ID. */
 512     ovs_be16 vlan_next_type;
 513 };
 514 BUILD_ASSERT_DECL(VLAN_HEADER_LEN == sizeof(struct vlan_header));
 515
 516 #define VLAN_ETH_HEADER_LEN (ETH_HEADER_LEN + VLAN_HEADER_LEN)
 517 struct vlan_eth_header {
 518     struct eth_addr veth_dst;
 519     struct eth_addr veth_src;
 520     ovs_be16 veth_type;         /* Always htons(ETH_TYPE_VLAN). */
 521     ovs_be16 veth_tci;          /* Lowest 12 bits are VLAN ID. */
 522     ovs_be16 veth_next_type;
 523 };
 524 BUILD_ASSERT_DECL(VLAN_ETH_HEADER_LEN == sizeof(struct vlan_eth_header));
 525
 526 /* MPLS related definitions */
 527 #define MPLS_TTL_MASK       0x000000ff
 528 #define MPLS_TTL_SHIFT      0
 529
 530 #define MPLS_BOS_MASK       0x00000100
 531 #define MPLS_BOS_SHIFT      8
 532
 533 #define MPLS_TC_MASK        0x00000e00
 534 #define MPLS_TC_SHIFT       9
 535
 536 #define MPLS_LABEL_MASK     0xfffff000
 537 #define MPLS_LABEL_SHIFT    12
 538
 539 #define MPLS_HLEN           4
 540
 541 struct mpls_hdr {
 542     ovs_16aligned_be32 mpls_lse;
 543 };
 544 BUILD_ASSERT_DECL(MPLS_HLEN == sizeof(struct mpls_hdr));
 545
 546 /* Given a mpls label stack entry in network byte order
 547  * return mpls label in host byte order */
 548 static inline uint32_t
 549 mpls_lse_to_label(ovs_be32 mpls_lse)
 550 {
 551     return (ntohl(mpls_lse) & MPLS_LABEL_MASK) >> MPLS_LABEL_SHIFT;
 552 }
 553
 554 /* Given a mpls label stack entry in network byte order
 555  * return mpls tc */
 556 static inline uint8_t
 557 mpls_lse_to_tc(ovs_be32 mpls_lse)
 558 {
 559     return (ntohl(mpls_lse) & MPLS_TC_MASK) >> MPLS_TC_SHIFT;
 560 }
 561
 562 /* Given a mpls label stack entry in network byte order
 563  * return mpls ttl */
 564 static inline uint8_t
 565 mpls_lse_to_ttl(ovs_be32 mpls_lse)
 566 {
 567     return (ntohl(mpls_lse) & MPLS_TTL_MASK) >> MPLS_TTL_SHIFT;
 568 }
 569
 570 /* Set TTL in mpls lse. */
 571 static inline void
 572 flow_set_mpls_lse_ttl(ovs_be32 *mpls_lse, uint8_t ttl)
 573 {
 574     *mpls_lse &= ~htonl(MPLS_TTL_MASK);
 575     *mpls_lse |= htonl(ttl << MPLS_TTL_SHIFT);
 576 }
 577
 578 /* Given a mpls label stack entry in network byte order
 579  * return mpls BoS bit  */
 580 static inline uint8_t
 581 mpls_lse_to_bos(ovs_be32 mpls_lse)
 582 {
 583     return (mpls_lse & htonl(MPLS_BOS_MASK)) != 0;
 584 }
 585
 586 #define IP_FMT "%"PRIu32".%"PRIu32".%"PRIu32".%"PRIu32
 587 #define IP_ARGS(ip)                             \
 588     ntohl(ip) >> 24,                            \
 589     (ntohl(ip) >> 16) & 0xff,                   \
 590     (ntohl(ip) >> 8) & 0xff,                    \
 591     ntohl(ip) & 0xff
 592
 593 /* Example:
 594  *
 595  * char *string = "1 33.44.55.66 2";
 596  * ovs_be32 ip;
 597  * int a, b;
 598  *
 599  * if (ovs_scan(string, "%d"IP_SCAN_FMT"%d", &a, IP_SCAN_ARGS(&ip), &b)) {
 600  *     ...
 601  * }
 602  */
 603 #define IP_SCAN_FMT "%"SCNu8".%"SCNu8".%"SCNu8".%"SCNu8
 604 #define IP_SCAN_ARGS(ip)                                    \
 605         ((void) (ovs_be32) *(ip), &((uint8_t *) ip)[0]),    \
 606         &((uint8_t *) ip)[1],                               \
 607         &((uint8_t *) ip)[2],                               \
 608         &((uint8_t *) ip)[3]
 609
 610 #define IP_PORT_SCAN_FMT "%"SCNu8".%"SCNu8".%"SCNu8".%"SCNu8":%"SCNu16
 611 #define IP_PORT_SCAN_ARGS(ip, port)                                    \
 612         ((void) (ovs_be32) *(ip), &((uint8_t *) ip)[0]),    \
 613         &((uint8_t *) ip)[1],                               \
 614         &((uint8_t *) ip)[2],                               \
 615         &((uint8_t *) ip)[3],                               \
 616         ((void) (ovs_be16) *(port), (uint16_t *) port)
 617
 618 /* Returns true if 'netmask' is a CIDR netmask, that is, if it consists of N
 619  * high-order 1-bits and 32-N low-order 0-bits. */
 620 static inline bool
 621 ip_is_cidr(ovs_be32 netmask)
 622 {
 623     uint32_t x = ~ntohl(netmask);
 624     return !(x & (x + 1));
 625 }
 626 static inline bool
 627 ip_is_multicast(ovs_be32 ip)
 628 {
 629     return (ip & htonl(0xf0000000)) == htonl(0xe0000000);
 630 }
 631 static inline bool
 632 ip_is_local_multicast(ovs_be32 ip)
 633 {
 634     return (ip & htonl(0xffffff00)) == htonl(0xe0000000);
 635 }
 636 int ip_count_cidr_bits(ovs_be32 netmask);
 637 void ip_format_masked(ovs_be32 ip, ovs_be32 mask, struct ds *);
 638 bool ip_parse(const char *s, ovs_be32 *ip);
 639 char *ip_parse_port(const char *s, ovs_be32 *ip, ovs_be16 *port)
 640     OVS_WARN_UNUSED_RESULT;
 641 char *ip_parse_masked(const char *s, ovs_be32 *ip, ovs_be32 *mask)
 642     OVS_WARN_UNUSED_RESULT;
 643 char *ip_parse_cidr(const char *s, ovs_be32 *ip, unsigned int *plen)
 644     OVS_WARN_UNUSED_RESULT;
 645 char *ip_parse_masked_len(const char *s, int *n, ovs_be32 *ip, ovs_be32 *mask)
 646     OVS_WARN_UNUSED_RESULT;
 647 char *ip_parse_cidr_len(const char *s, int *n, ovs_be32 *ip,
 648                         unsigned int *plen)
 649     OVS_WARN_UNUSED_RESULT;
 650
 651 #define IP_VER(ip_ihl_ver) ((ip_ihl_ver) >> 4)
 652 #define IP_IHL(ip_ihl_ver) ((ip_ihl_ver) & 15)
 653 #define IP_IHL_VER(ihl, ver) (((ver) << 4) | (ihl))
 654
 655 #ifndef IPPROTO_SCTP
 656 #define IPPROTO_SCTP 132
 657 #endif
 658
 659 #ifndef IPPROTO_DCCP
 660 #define IPPROTO_DCCP 33
 661 #endif
 662
 663 #ifndef IPPROTO_IGMP
 664 #define IPPROTO_IGMP 2
 665 #endif
 666
 667 #ifndef IPPROTO_UDPLITE
 668 #define IPPROTO_UDPLITE 136
 669 #endif
 670
 671 /* TOS fields. */
 672 #define IP_ECN_NOT_ECT 0x0
 673 #define IP_ECN_ECT_1 0x01
 674 #define IP_ECN_ECT_0 0x02
 675 #define IP_ECN_CE 0x03
 676 #define IP_ECN_MASK 0x03
 677 #define IP_DSCP_MASK 0xfc
 678
 679 static inline int
 680 IP_ECN_is_ce(uint8_t dsfield)
 681 {
 682     return (dsfield & IP_ECN_MASK) == IP_ECN_CE;
 683 }
 684
 685 #define IP_VERSION 4
 686
 687 #define IP_DONT_FRAGMENT  0x4000 /* Don't fragment. */
 688 #define IP_MORE_FRAGMENTS 0x2000 /* More fragments. */
 689 #define IP_FRAG_OFF_MASK  0x1fff /* Fragment offset. */
 690 #define IP_IS_FRAGMENT(ip_frag_off) \
 691         ((ip_frag_off) & htons(IP_MORE_FRAGMENTS | IP_FRAG_OFF_MASK))
 692
 693 #define IP_HEADER_LEN 20
 694 struct ip_header {
 695     uint8_t ip_ihl_ver;
 696     uint8_t ip_tos;
 697     ovs_be16 ip_tot_len;
 698     ovs_be16 ip_id;
 699     ovs_be16 ip_frag_off;
 700     uint8_t ip_ttl;
 701     uint8_t ip_proto;
 702     ovs_be16 ip_csum;
 703     ovs_16aligned_be32 ip_src;
 704     ovs_16aligned_be32 ip_dst;
 705 };
 706 BUILD_ASSERT_DECL(IP_HEADER_LEN == sizeof(struct ip_header));
 707
 708 /* ICMPv4 types. */
 709 #define ICMP4_ECHO_REPLY 0
 710 #define ICMP4_DST_UNREACH 3
 711 #define ICMP4_SOURCEQUENCH 4
 712 #define ICMP4_REDIRECT 5
 713 #define ICMP4_ECHO_REQUEST 8
 714 #define ICMP4_TIME_EXCEEDED 11
 715 #define ICMP4_PARAM_PROB 12
 716 #define ICMP4_TIMESTAMP 13
 717 #define ICMP4_TIMESTAMPREPLY 14
 718 #define ICMP4_INFOREQUEST 15
 719 #define ICMP4_INFOREPLY 16
 720
 721 #define ICMP_HEADER_LEN 8
 722 struct icmp_header {
 723     uint8_t icmp_type;
 724     uint8_t icmp_code;
 725     ovs_be16 icmp_csum;
 726     union {
 727         struct {
 728             ovs_be16 id;
 729             ovs_be16 seq;
 730         } echo;
 731         struct {
 732             ovs_be16 empty;
 733             ovs_be16 mtu;
 734         } frag;
 735         ovs_16aligned_be32 gateway;
 736     } icmp_fields;
 737 };
 738 BUILD_ASSERT_DECL(ICMP_HEADER_LEN == sizeof(struct icmp_header));
 739
 740 #define IGMP_HEADER_LEN 8
 741 struct igmp_header {
 742     uint8_t igmp_type;
 743     uint8_t igmp_code;
 744     ovs_be16 igmp_csum;
 745     ovs_16aligned_be32 group;
 746 };
 747 BUILD_ASSERT_DECL(IGMP_HEADER_LEN == sizeof(struct igmp_header));
 748
 749 #define IGMPV3_HEADER_LEN 8
 750 struct igmpv3_header {
 751     uint8_t type;
 752     uint8_t rsvr1;
 753     ovs_be16 csum;
 754     ovs_be16 rsvr2;
 755     ovs_be16 ngrp;
 756 };
 757 BUILD_ASSERT_DECL(IGMPV3_HEADER_LEN == sizeof(struct igmpv3_header));
 758
 759 #define IGMPV3_RECORD_LEN 8
 760 struct igmpv3_record {
 761     uint8_t type;
 762     uint8_t aux_len;
 763     ovs_be16 nsrcs;
 764     ovs_16aligned_be32 maddr;
 765 };
 766 BUILD_ASSERT_DECL(IGMPV3_RECORD_LEN == sizeof(struct igmpv3_record));
 767
 768 #define IGMP_HOST_MEMBERSHIP_QUERY    0x11 /* From RFC1112 */
 769 #define IGMP_HOST_MEMBERSHIP_REPORT   0x12 /* Ditto */
 770 #define IGMPV2_HOST_MEMBERSHIP_REPORT 0x16 /* V2 version of 0x12 */
 771 #define IGMP_HOST_LEAVE_MESSAGE       0x17
 772 #define IGMPV3_HOST_MEMBERSHIP_REPORT 0x22 /* V3 version of 0x12 */
 773
 774 /*
 775  * IGMPv3 and MLDv2 use the same codes.
 776  */
 777 #define IGMPV3_MODE_IS_INCLUDE 1
 778 #define IGMPV3_MODE_IS_EXCLUDE 2
 779 #define IGMPV3_CHANGE_TO_INCLUDE_MODE 3
 780 #define IGMPV3_CHANGE_TO_EXCLUDE_MODE 4
 781 #define IGMPV3_ALLOW_NEW_SOURCES 5
 782 #define IGMPV3_BLOCK_OLD_SOURCES 6
 783
 784 #define SCTP_HEADER_LEN 12
 785 struct sctp_header {
 786     ovs_be16 sctp_src;
 787     ovs_be16 sctp_dst;
 788     ovs_16aligned_be32 sctp_vtag;
 789     ovs_16aligned_be32 sctp_csum;
 790 };
 791 BUILD_ASSERT_DECL(SCTP_HEADER_LEN == sizeof(struct sctp_header));
 792
 793 #define UDP_HEADER_LEN 8
 794 struct udp_header {
 795     ovs_be16 udp_src;
 796     ovs_be16 udp_dst;
 797     ovs_be16 udp_len;
 798     ovs_be16 udp_csum;
 799 };
 800 BUILD_ASSERT_DECL(UDP_HEADER_LEN == sizeof(struct udp_header));
 801
 802 #define TCP_FIN 0x001
 803 #define TCP_SYN 0x002
 804 #define TCP_RST 0x004
 805 #define TCP_PSH 0x008
 806 #define TCP_ACK 0x010
 807 #define TCP_URG 0x020
 808 #define TCP_ECE 0x040
 809 #define TCP_CWR 0x080
 810 #define TCP_NS  0x100
 811
 812 #define TCP_CTL(flags, offset) (htons((flags) | ((offset) << 12)))
 813 #define TCP_FLAGS(tcp_ctl) (ntohs(tcp_ctl) & 0x0fff)
 814 #define TCP_FLAGS_BE16(tcp_ctl) ((tcp_ctl) & htons(0x0fff))
 815 #define TCP_OFFSET(tcp_ctl) (ntohs(tcp_ctl) >> 12)
 816
 817 #define TCP_HEADER_LEN 20
 818 struct tcp_header {
 819     ovs_be16 tcp_src;
 820     ovs_be16 tcp_dst;
 821     ovs_16aligned_be32 tcp_seq;
 822     ovs_16aligned_be32 tcp_ack;
 823     ovs_be16 tcp_ctl;
 824     ovs_be16 tcp_winsz;
 825     ovs_be16 tcp_csum;
 826     ovs_be16 tcp_urg;
 827 };
 828 BUILD_ASSERT_DECL(TCP_HEADER_LEN == sizeof(struct tcp_header));
 829
 830 /* Connection states.
 831  *
 832  * Names like CS_RELATED are bit values, e.g. 1 << 2.
 833  * Names like CS_RELATED_BIT are bit indexes, e.g. 2. */
 834 #define CS_STATES                               \
 835     CS_STATE(NEW,         0, "new")             \
 836     CS_STATE(ESTABLISHED, 1, "est")             \
 837     CS_STATE(RELATED,     2, "rel")             \
 838     CS_STATE(REPLY_DIR,   3, "rpl")             \
 839     CS_STATE(INVALID,     4, "inv")             \
 840     CS_STATE(TRACKED,     5, "trk")             \
 841     CS_STATE(SRC_NAT,     6, "snat")            \
 842     CS_STATE(DST_NAT,     7, "dnat")
 843
 844 enum {
 845 #define CS_STATE(ENUM, INDEX, NAME) \
 846     CS_##ENUM = 1 << INDEX, \
 847     CS_##ENUM##_BIT = INDEX,
 848     CS_STATES
 849 #undef CS_STATE
 850 };
 851
 852 /* Undefined connection state bits. */
 853 enum {
 854 #define CS_STATE(ENUM, INDEX, NAME) +CS_##ENUM
 855     CS_SUPPORTED_MASK = CS_STATES
 856 #undef CS_STATE
 857 };
 858 #define CS_UNSUPPORTED_MASK  (~(uint32_t)CS_SUPPORTED_MASK)
 859
 860 #define ARP_HRD_ETHERNET 1
 861 #define ARP_PRO_IP 0x0800
 862 #define ARP_OP_REQUEST 1
 863 #define ARP_OP_REPLY 2
 864 #define ARP_OP_RARP 3
 865
 866 #define ARP_ETH_HEADER_LEN 28
 867 struct arp_eth_header {
 868     /* Generic members. */
 869     ovs_be16 ar_hrd;           /* Hardware type. */
 870     ovs_be16 ar_pro;           /* Protocol type. */
 871     uint8_t ar_hln;            /* Hardware address length. */
 872     uint8_t ar_pln;            /* Protocol address length. */
 873     ovs_be16 ar_op;            /* Opcode. */
 874
 875     /* Ethernet+IPv4 specific members. */
 876     struct eth_addr ar_sha;     /* Sender hardware address. */
 877     ovs_16aligned_be32 ar_spa;  /* Sender protocol address. */
 878     struct eth_addr ar_tha;     /* Target hardware address. */
 879     ovs_16aligned_be32 ar_tpa;  /* Target protocol address. */
 880 };
 881 BUILD_ASSERT_DECL(ARP_ETH_HEADER_LEN == sizeof(struct arp_eth_header));
 882
 883 #define IPV6_HEADER_LEN 40
 884
 885 /* Like struct in6_addr, but whereas that struct requires 32-bit alignment on
 886  * most implementations, this one only requires 16-bit alignment. */
 887 union ovs_16aligned_in6_addr {
 888     ovs_be16 be16[8];
 889     ovs_16aligned_be32 be32[4];
 890 };
 891
 892 /* Like struct in6_hdr, but whereas that struct requires 32-bit alignment, this
 893  * one only requires 16-bit alignment. */
 894 struct ovs_16aligned_ip6_hdr {
 895     union {
 896         struct ovs_16aligned_ip6_hdrctl {
 897             ovs_16aligned_be32 ip6_un1_flow;
 898             ovs_be16 ip6_un1_plen;
 899             uint8_t ip6_un1_nxt;
 900             uint8_t ip6_un1_hlim;
 901         } ip6_un1;
 902         uint8_t ip6_un2_vfc;
 903     } ip6_ctlun;
 904     union ovs_16aligned_in6_addr ip6_src;
 905     union ovs_16aligned_in6_addr ip6_dst;
 906 };
 907
 908 /* Like struct in6_frag, but whereas that struct requires 32-bit alignment,
 909  * this one only requires 16-bit alignment. */
 910 struct ovs_16aligned_ip6_frag {
 911     uint8_t ip6f_nxt;
 912     uint8_t ip6f_reserved;
 913     ovs_be16 ip6f_offlg;
 914     ovs_16aligned_be32 ip6f_ident;
 915 };
 916
 917 #define ICMP6_HEADER_LEN 4
 918 struct icmp6_header {
 919     uint8_t icmp6_type;
 920     uint8_t icmp6_code;
 921     ovs_be16 icmp6_cksum;
 922 };
 923 BUILD_ASSERT_DECL(ICMP6_HEADER_LEN == sizeof(struct icmp6_header));
 924
 925 #define ICMP6_ERROR_HEADER_LEN 8
 926 struct icmp6_error_header {
 927     struct icmp6_header icmp6_base;
 928     ovs_be32 icmp6_error_ext;
 929 };
 930 BUILD_ASSERT_DECL(ICMP6_ERROR_HEADER_LEN == sizeof(struct icmp6_error_header));
 931
 932 uint32_t packet_csum_pseudoheader6(const struct ovs_16aligned_ip6_hdr *);
 933 uint16_t packet_csum_upperlayer6(const struct ovs_16aligned_ip6_hdr *,
 934                                  const void *, uint8_t, uint16_t);
 935
 936 /* Neighbor Discovery option field.
 937  * ND options are always a multiple of 8 bytes in size. */
 938 #define ND_LLA_OPT_LEN 8
 939 struct ovs_nd_lla_opt {
 940     uint8_t type;               /* One of ND_OPT_*_LINKADDR. */
 941     uint8_t len;
 942     struct eth_addr mac;
 943 };
 944 BUILD_ASSERT_DECL(ND_LLA_OPT_LEN == sizeof(struct ovs_nd_lla_opt));
 945
 946 /* Neighbor Discovery option: Prefix Information. */
 947 #define ND_PREFIX_OPT_LEN 32
 948 struct ovs_nd_prefix_opt {
 949     uint8_t type;               /* ND_OPT_PREFIX_INFORMATION. */
 950     uint8_t len;                /* Always 4. */
 951     uint8_t prefix_len;
 952     uint8_t la_flags;           /* ND_PREFIX_* flags. */
 953     ovs_16aligned_be32 valid_lifetime;
 954     ovs_16aligned_be32 preferred_lifetime;
 955     ovs_16aligned_be32 reserved;          /* Always 0. */
 956     union ovs_16aligned_in6_addr prefix;
 957 };
 958 BUILD_ASSERT_DECL(ND_PREFIX_OPT_LEN == sizeof(struct ovs_nd_prefix_opt));
 959
 960 #define ND_PREFIX_ON_LINK            0x80
 961 #define ND_PREFIX_AUTONOMOUS_ADDRESS 0x40
 962
 963 /* Neighbor Discovery option: MTU. */
 964 #define ND_MTU_OPT_LEN 8
 965 struct ovs_nd_mtu_opt {
 966     uint8_t  type;      /* ND_OPT_MTU */
 967     uint8_t  len;       /* Always 1. */
 968     ovs_be16 reserved;  /* Always 0. */
 969     ovs_16aligned_be32 mtu;
 970 };
 971 BUILD_ASSERT_DECL(ND_MTU_OPT_LEN == sizeof(struct ovs_nd_mtu_opt));
 972
 973 /* Like struct nd_msg (from ndisc.h), but whereas that struct requires 32-bit
 974  * alignment, this one only requires 16-bit alignment. */
 975 #define ND_MSG_LEN 24
 976 struct ovs_nd_msg {
 977     struct icmp6_header icmph;
 978     ovs_16aligned_be32 rso_flags;
 979     union ovs_16aligned_in6_addr target;
 980     struct ovs_nd_lla_opt options[0];
 981 };
 982 BUILD_ASSERT_DECL(ND_MSG_LEN == sizeof(struct ovs_nd_msg));
 983
 984 /* Neighbor Discovery packet flags. */
 985 #define ND_RSO_ROUTER    0x80000000
 986 #define ND_RSO_SOLICITED 0x40000000
 987 #define ND_RSO_OVERRIDE  0x20000000
 988
 989 #define RA_MSG_LEN 16
 990 struct ovs_ra_msg {
 991     struct icmp6_header icmph;
 992     uint8_t cur_hop_limit;
 993     uint8_t mo_flags;  /* ND_RA_MANAGED_ADDRESS and ND_RA_OTHER_CONFIG flags. */
 994     ovs_be16 router_lifetime;
 995     ovs_be32 reachable_time;
 996     ovs_be32 retrans_timer;
 997     struct ovs_nd_lla_opt options[0];
 998 };
 999 BUILD_ASSERT_DECL(RA_MSG_LEN == sizeof(struct ovs_ra_msg));
1000
1001 #define ND_RA_MANAGED_ADDRESS 0x80
1002 #define ND_RA_OTHER_CONFIG    0x40
1003
1004 /*
1005  * Use the same struct for MLD and MLD2, naming members as the defined fields in
1006  * in the corresponding version of the protocol, though they are reserved in the
1007  * other one.
1008  */
1009 #define MLD_HEADER_LEN 8
1010 struct mld_header {
1011     uint8_t type;
1012     uint8_t code;
1013     ovs_be16 csum;
1014     ovs_be16 mrd;
1015     ovs_be16 ngrp;
1016 };
1017 BUILD_ASSERT_DECL(MLD_HEADER_LEN == sizeof(struct mld_header));
1018
1019 #define MLD2_RECORD_LEN 20
1020 struct mld2_record {
1021     uint8_t type;
1022     uint8_t aux_len;
1023     ovs_be16 nsrcs;
1024     union ovs_16aligned_in6_addr maddr;
1025 };
1026 BUILD_ASSERT_DECL(MLD2_RECORD_LEN == sizeof(struct mld2_record));
1027
1028 #define MLD_QUERY 130
1029 #define MLD_REPORT 131
1030 #define MLD_DONE 132
1031 #define MLD2_REPORT 143
1032
1033 /* The IPv6 flow label is in the lower 20 bits of the first 32-bit word. */
1034 #define IPV6_LABEL_MASK 0x000fffff
1035
1036 /* Example:
1037  *
1038  * char *string = "1 ::1 2";
1039  * char ipv6_s[IPV6_SCAN_LEN + 1];
1040  * struct in6_addr ipv6;
1041  *
1042  * if (ovs_scan(string, "%d"IPV6_SCAN_FMT"%d", &a, ipv6_s, &b)
1043  *     && inet_pton(AF_INET6, ipv6_s, &ipv6) == 1) {
1044  *     ...
1045  * }
1046  */
1047 #define IPV6_SCAN_FMT "%46[0123456789abcdefABCDEF:.]"
1048 #define IPV6_SCAN_LEN 46
1049
1050 extern const struct in6_addr in6addr_exact;
1051 #define IN6ADDR_EXACT_INIT { { { 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, \
1052                                  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff } } }
1053
1054 extern const struct in6_addr in6addr_all_hosts;
1055 #define IN6ADDR_ALL_HOSTS_INIT { { { 0xff,0x02,0x00,0x00,0x00,0x00,0x00,0x00, \
1056                                      0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x01 } } }
1057
1058 extern const struct in6_addr in6addr_all_routers;
1059 #define IN6ADDR_ALL_ROUTERS_INIT { { { 0xff,0x02,0x00,0x00,0x00,0x00,0x00,0x00, \
1060                                        0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x02 } } }
1061
1062 static inline bool ipv6_addr_equals(const struct in6_addr *a,
1063                                     const struct in6_addr *b)
1064 {
1065 #ifdef IN6_ARE_ADDR_EQUAL
1066     return IN6_ARE_ADDR_EQUAL(a, b);
1067 #else
1068     return !memcmp(a, b, sizeof(*a));
1069 #endif
1070 }
1071
1072 /* Checks the IPv6 address in 'mask' for all zeroes. */
1073 static inline bool ipv6_mask_is_any(const struct in6_addr *mask) {
1074     return ipv6_addr_equals(mask, &in6addr_any);
1075 }
1076
1077 static inline bool ipv6_mask_is_exact(const struct in6_addr *mask) {
1078     return ipv6_addr_equals(mask, &in6addr_exact);
1079 }
1080
1081 static inline bool ipv6_is_all_hosts(const struct in6_addr *addr) {
1082     return ipv6_addr_equals(addr, &in6addr_all_hosts);
1083 }
1084
1085 static inline bool ipv6_addr_is_set(const struct in6_addr *addr) {
1086     return !ipv6_addr_equals(addr, &in6addr_any);
1087 }
1088
1089 static inline bool ipv6_addr_is_multicast(const struct in6_addr *ip) {
1090     return ip->s6_addr[0] == 0xff;
1091 }
1092
1093 static inline struct in6_addr
1094 in6_addr_mapped_ipv4(ovs_be32 ip4)
1095 {
1096     struct in6_addr ip6 = { .s6_addr = { [10] = 0xff, [11] = 0xff } };
1097     memcpy(&ip6.s6_addr[12], &ip4, 4);
1098     return ip6;
1099 }
1100
1101 static inline void
1102 in6_addr_set_mapped_ipv4(struct in6_addr *ip6, ovs_be32 ip4)
1103 {
1104     *ip6 = in6_addr_mapped_ipv4(ip4);
1105 }
1106
1107 static inline ovs_be32
1108 in6_addr_get_mapped_ipv4(const struct in6_addr *addr)
1109 {
1110     union ovs_16aligned_in6_addr *taddr = (void *) addr;
1111     if (IN6_IS_ADDR_V4MAPPED(addr)) {
1112         return get_16aligned_be32(&taddr->be32[3]);
1113     } else {
1114         return INADDR_ANY;
1115     }
1116 }
1117
1118 static inline void
1119 in6_addr_solicited_node(struct in6_addr *addr, const struct in6_addr *ip6)
1120 {
1121     union ovs_16aligned_in6_addr *taddr = (void *) addr;
1122     memset(taddr->be16, 0, sizeof(taddr->be16));
1123     taddr->be16[0] = htons(0xff02);
1124     taddr->be16[5] = htons(0x1);
1125     taddr->be16[6] = htons(0xff00);
1126     memcpy(&addr->s6_addr[13], &ip6->s6_addr[13], 3);
1127 }
1128
1129 /*
1130  * Generates ipv6 EUI64 address from the given eth addr
1131  * and prefix and stores it in 'lla'
1132  */
1133 static inline void
1134 in6_generate_eui64(struct eth_addr ea, struct in6_addr *prefix,
1135                    struct in6_addr *lla)
1136 {
1137     union ovs_16aligned_in6_addr *taddr = (void *) lla;
1138     union ovs_16aligned_in6_addr *prefix_taddr = (void *) prefix;
1139     taddr->be16[0] = prefix_taddr->be16[0];
1140     taddr->be16[1] = prefix_taddr->be16[1];
1141     taddr->be16[2] = prefix_taddr->be16[2];
1142     taddr->be16[3] = prefix_taddr->be16[3];
1143     taddr->be16[4] = htons(((ea.ea[0] ^ 0x02) << 8) | ea.ea[1]);
1144     taddr->be16[5] = htons(ea.ea[2] << 8 | 0x00ff);
1145     taddr->be16[6] = htons(0xfe << 8 | ea.ea[3]);
1146     taddr->be16[7] = ea.be16[2];
1147 }
1148
1149 /*
1150  * Generates ipv6 link local address from the given eth addr
1151  * with prefix 'fe80::/64' and stores it in 'lla'
1152  */
1153 static inline void
1154 in6_generate_lla(struct eth_addr ea, struct in6_addr *lla)
1155 {
1156     union ovs_16aligned_in6_addr *taddr = (void *) lla;
1157     memset(taddr->be16, 0, sizeof(taddr->be16));
1158     taddr->be16[0] = htons(0xfe80);
1159     taddr->be16[4] = htons(((ea.ea[0] ^ 0x02) << 8) | ea.ea[1]);
1160     taddr->be16[5] = htons(ea.ea[2] << 8 | 0x00ff);
1161     taddr->be16[6] = htons(0xfe << 8 | ea.ea[3]);
1162     taddr->be16[7] = ea.be16[2];
1163 }
1164
1165 /* Returns true if 'addr' is a link local address.  Otherwise, false. */
1166 static inline bool
1167 in6_is_lla(struct in6_addr *addr)
1168 {
1169 #ifdef s6_addr32
1170     return addr->s6_addr32[0] == htonl(0xfe800000) && !(addr->s6_addr32[1]);
1171 #else
1172     return addr->s6_addr[0] == 0xfe && addr->s6_addr[1] == 0x80 &&
1173          !(addr->s6_addr[2] | addr->s6_addr[3] | addr->s6_addr[4] |
1174            addr->s6_addr[5] | addr->s6_addr[6] | addr->s6_addr[7]);
1175 #endif
1176 }
1177
1178 static inline void
1179 ipv6_multicast_to_ethernet(struct eth_addr *eth, const struct in6_addr *ip6)
1180 {
1181     eth->ea[0] = 0x33;
1182     eth->ea[1] = 0x33;
1183     eth->ea[2] = ip6->s6_addr[12];
1184     eth->ea[3] = ip6->s6_addr[13];
1185     eth->ea[4] = ip6->s6_addr[14];
1186     eth->ea[5] = ip6->s6_addr[15];
1187 }
1188
1189 static inline bool dl_type_is_ip_any(ovs_be16 dl_type)
1190 {
1191     return dl_type == htons(ETH_TYPE_IP)
1192         || dl_type == htons(ETH_TYPE_IPV6);
1193 }
1194
1195 /* Tunnel header */
1196
1197 /* GRE protocol header */
1198 struct gre_base_hdr {
1199     ovs_be16 flags;
1200     ovs_be16 protocol;
1201 };
1202
1203 #define GRE_CSUM        0x8000
1204 #define GRE_ROUTING     0x4000
1205 #define GRE_KEY         0x2000
1206 #define GRE_SEQ         0x1000
1207 #define GRE_STRICT      0x0800
1208 #define GRE_REC         0x0700
1209 #define GRE_FLAGS       0x00F8
1210 #define GRE_VERSION     0x0007
1211
1212 /* VXLAN protocol header */
1213 struct vxlanhdr {
1214     union {
1215         ovs_16aligned_be32 vx_flags; /* VXLAN flags. */
1216         struct {
1217             uint8_t flags;           /* VXLAN GPE flags. */
1218             uint8_t reserved[2];     /* 16 bits reserved. */
1219             uint8_t next_protocol;   /* Next Protocol field for VXLAN GPE. */
1220         } vx_gpe;
1221     };
1222     ovs_16aligned_be32 vx_vni;
1223 };
1224 BUILD_ASSERT_DECL(sizeof(struct vxlanhdr) == 8);
1225
1226 #define VXLAN_FLAGS 0x08000000  /* struct vxlanhdr.vx_flags required value. */
1227
1228 /*
1229  * VXLAN Generic Protocol Extension (VXLAN_F_GPE):
1230  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1231  * |R|R|Ver|I|P|R|O|       Reserved                |Next Protocol  |
1232  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1233  * |                VXLAN Network Identifier (VNI) |   Reserved    |
1234  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1235  *
1236  * Ver = Version. Indicates VXLAN GPE protocol version.
1237  *
1238  * P = Next Protocol Bit. The P bit is set to indicate that the
1239  *     Next Protocol field is present.
1240  *
1241  * O = OAM Flag Bit. The O bit is set to indicate that the packet
1242  *     is an OAM packet.
1243  *
1244  * Next Protocol = This 8 bit field indicates the protocol header
1245  * immediately following the VXLAN GPE header.
1246  *
1247  * https://tools.ietf.org/html/draft-ietf-nvo3-vxlan-gpe-01
1248  */
1249
1250 /* Fields in struct vxlanhdr.vx_gpe.flags */
1251 #define VXLAN_GPE_FLAGS_VER     0x30    /* Version. */
1252 #define VLXAN_GPE_FLAGS_P       0x04    /* Next Protocol Bit. */
1253 #define VXLAN_GPE_FLAGS_O       0x01    /* OAM Bit. */
1254
1255 /* VXLAN-GPE header flags. */
1256 #define VXLAN_HF_VER   ((1U <<29) | (1U <<28))
1257 #define VXLAN_HF_NP    (1U <<26)
1258 #define VXLAN_HF_OAM   (1U <<24)
1259
1260 #define VXLAN_GPE_USED_BITS (VXLAN_HF_VER | VXLAN_HF_NP | VXLAN_HF_OAM | \
1261                             0xff)
1262
1263 /* VXLAN-GPE header Next Protocol. */
1264 #define VXLAN_GPE_NP_IPV4      0x01
1265 #define VXLAN_GPE_NP_IPV6      0x02
1266 #define VXLAN_GPE_NP_ETHERNET  0x03
1267 #define VXLAN_GPE_NP_NSH       0x04
1268
1269 #define VXLAN_F_GPE  0x4000
1270 #define VXLAN_HF_GPE 0x04000000
1271
1272 /* Input values for PACKET_TYPE macros have to be in host byte order.
1273  * The _BE postfix indicates result is in network byte order. Otherwise result
1274  * is in host byte order. */
1275 #define PACKET_TYPE(NS, NS_TYPE) ((uint32_t) ((NS) << 16 | (NS_TYPE)))
1276 #define PACKET_TYPE_BE(NS, NS_TYPE) (htonl((NS) << 16 | (NS_TYPE)))
1277
1278 /* Returns the host byte ordered namespace of 'packet type'. */
1279 static inline uint16_t
1280 pt_ns(ovs_be32 packet_type)
1281 {
1282     return ntohl(packet_type) >> 16;
1283 }
1284
1285 /* Returns the network byte ordered namespace type of 'packet type'. */
1286 static inline ovs_be16
1287 pt_ns_type_be(ovs_be32 packet_type)
1288 {
1289     return be32_to_be16(packet_type);
1290 }
1291
1292 /* Returns the host byte ordered namespace type of 'packet type'. */
1293 static inline uint16_t
1294 pt_ns_type(ovs_be32 packet_type)
1295 {
1296     return ntohs(pt_ns_type_be(packet_type));
1297 }
1298
1299 /* Well-known packet_type field values. */
1300 enum packet_type {
1301     PT_ETH  = PACKET_TYPE(OFPHTN_ONF, 0x0000),  /* Default PT: Ethernet */
1302     PT_USE_NEXT_PROTO = PACKET_TYPE(OFPHTN_ONF, 0xfffe),  /* Pseudo PT for decap. */
1303     PT_IPV4 = PACKET_TYPE(OFPHTN_ETHERTYPE, ETH_TYPE_IP),
1304     PT_IPV6 = PACKET_TYPE(OFPHTN_ETHERTYPE, ETH_TYPE_IPV6),
1305     PT_MPLS = PACKET_TYPE(OFPHTN_ETHERTYPE, ETH_TYPE_MPLS),
1306     PT_MPLS_MC = PACKET_TYPE(OFPHTN_ETHERTYPE, ETH_TYPE_MPLS_MCAST),
1307     PT_NSH  = PACKET_TYPE(OFPHTN_ETHERTYPE, ETH_TYPE_NSH),
1308     PT_UNKNOWN = PACKET_TYPE(0xffff, 0xffff),   /* Unknown packet type. */
1309 };
1310
1311
1312 void ipv6_format_addr(const struct in6_addr *addr, struct ds *);
1313 void ipv6_format_addr_bracket(const struct in6_addr *addr, struct ds *,
1314                               bool bracket);
1315 void ipv6_format_mapped(const struct in6_addr *addr, struct ds *);
1316 void ipv6_format_masked(const struct in6_addr *addr,
1317                         const struct in6_addr *mask, struct ds *);
1318 const char * ipv6_string_mapped(char *addr_str, const struct in6_addr *addr);
1319 struct in6_addr ipv6_addr_bitand(const struct in6_addr *src,
1320                                  const struct in6_addr *mask);
1321 struct in6_addr ipv6_addr_bitxor(const struct in6_addr *a,
1322                                  const struct in6_addr *b);
1323 bool ipv6_is_zero(const struct in6_addr *a);
1324 struct in6_addr ipv6_create_mask(int mask);
1325 int ipv6_count_cidr_bits(const struct in6_addr *netmask);
1326 bool ipv6_is_cidr(const struct in6_addr *netmask);
1327
1328 bool ipv6_parse(const char *s, struct in6_addr *ip);
1329 char *ipv6_parse_masked(const char *s, struct in6_addr *ipv6,
1330                         struct in6_addr *mask);
1331 char *ipv6_parse_cidr(const char *s, struct in6_addr *ip, unsigned int *plen)
1332     OVS_WARN_UNUSED_RESULT;
1333 char *ipv6_parse_masked_len(const char *s, int *n, struct in6_addr *ipv6,
1334                             struct in6_addr *mask);
1335 char *ipv6_parse_cidr_len(const char *s, int *n, struct in6_addr *ip,
1336                           unsigned int *plen)
1337     OVS_WARN_UNUSED_RESULT;
1338
1339 void *eth_compose(struct dp_packet *, const struct eth_addr eth_dst,
1340                   const struct eth_addr eth_src, uint16_t eth_type,
1341                   size_t size);
1342 void *snap_compose(struct dp_packet *, const struct eth_addr eth_dst,
1343                    const struct eth_addr eth_src,
1344                    unsigned int oui, uint16_t snap_type, size_t size);
1345 void packet_set_ipv4(struct dp_packet *, ovs_be32 src, ovs_be32 dst, uint8_t tos,
1346                      uint8_t ttl);
1347 void packet_set_ipv4_addr(struct dp_packet *packet, ovs_16aligned_be32 *addr,
1348                           ovs_be32 new_addr);
1349 void packet_set_ipv6(struct dp_packet *, const struct in6_addr *src,
1350                      const struct in6_addr *dst, uint8_t tc,
1351                      ovs_be32 fl, uint8_t hlmit);
1352 void packet_set_ipv6_addr(struct dp_packet *packet, uint8_t proto,
1353                           ovs_16aligned_be32 addr[4],
1354                           const struct in6_addr *new_addr,
1355                           bool recalculate_csum);
1356 void packet_set_tcp_port(struct dp_packet *, ovs_be16 src, ovs_be16 dst);
1357 void packet_set_udp_port(struct dp_packet *, ovs_be16 src, ovs_be16 dst);
1358 void packet_set_sctp_port(struct dp_packet *, ovs_be16 src, ovs_be16 dst);
1359 void packet_set_icmp(struct dp_packet *, uint8_t type, uint8_t code);
1360 void packet_set_nd(struct dp_packet *, const struct in6_addr *target,
1361                    const struct eth_addr sll, const struct eth_addr tll);
1362
1363 void packet_format_tcp_flags(struct ds *, uint16_t);
1364 const char *packet_tcp_flag_to_string(uint32_t flag);
1365 void compose_arp__(struct dp_packet *);
1366 void compose_arp(struct dp_packet *, uint16_t arp_op,
1367                  const struct eth_addr arp_sha,
1368                  const struct eth_addr arp_tha, bool broadcast,
1369                  ovs_be32 arp_spa, ovs_be32 arp_tpa);
1370 void compose_nd_ns(struct dp_packet *, const struct eth_addr eth_src,
1371                    const struct in6_addr *ipv6_src,
1372                    const struct in6_addr *ipv6_dst);
1373 void compose_nd_na(struct dp_packet *, const struct eth_addr eth_src,
1374                    const struct eth_addr eth_dst,
1375                    const struct in6_addr *ipv6_src,
1376                    const struct in6_addr *ipv6_dst,
1377                    ovs_be32 rso_flags);
1378 void compose_nd_ra(struct dp_packet *,
1379                    const struct eth_addr eth_src,
1380                    const struct eth_addr eth_dst,
1381                    const struct in6_addr *ipv6_src,
1382                    const struct in6_addr *ipv6_dst,
1383                    uint8_t cur_hop_limit, uint8_t mo_flags,
1384                    ovs_be16 router_lt, ovs_be32 reachable_time,
1385                    ovs_be32 retrans_timer, ovs_be32 mtu);
1386 void packet_put_ra_prefix_opt(struct dp_packet *,
1387                               uint8_t plen, uint8_t la_flags,
1388                               ovs_be32 valid_lifetime,
1389                               ovs_be32 preferred_lifetime,
1390                               const ovs_be128 router_prefix);
1391 uint32_t packet_csum_pseudoheader(const struct ip_header *);
1392 void IP_ECN_set_ce(struct dp_packet *pkt, bool is_ipv6);
1393
1394 #define DNS_HEADER_LEN 12
1395 struct dns_header {
1396     ovs_be16 id;
1397     uint8_t lo_flag; /* QR (1), OPCODE (4), AA (1), TC (1) and RD (1) */
1398     uint8_t hi_flag; /* RA (1), Z (3) and RCODE (4) */
1399     ovs_be16 qdcount; /* Num of entries in the question section. */
1400     ovs_be16 ancount; /* Num of resource records in the answer section. */
1401
1402     /* Num of name server records in the authority record section. */
1403     ovs_be16 nscount;
1404
1405     /* Num of resource records in the additional records section. */
1406     ovs_be16 arcount;
1407 };
1408
1409 BUILD_ASSERT_DECL(DNS_HEADER_LEN == sizeof(struct dns_header));
1410
1411 #define DNS_QUERY_TYPE_A        0x01
1412 #define DNS_QUERY_TYPE_AAAA     0x1c
1413 #define DNS_QUERY_TYPE_ANY      0xff
1414
1415 #define DNS_CLASS_IN            0x01
1416 #define DNS_DEFAULT_RR_TTL      3600
1417
1418 #endif /* packets.h */