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