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