lib/packets.h

   1 /*
   2  * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015 Nicira, Inc.
   3  *
   4  * Licensed under the Apache License, Version 2.0 (the "License");
   5  * you may not use this file except in compliance with the License.
   6  * You may obtain a copy of the License at:
   7  *
   8  *     http://www.apache.org/licenses/LICENSE-2.0
   9  *
  10  * Unless required by applicable law or agreed to in writing, software
  11  * distributed under the License is distributed on an "AS IS" BASIS,
  12  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  13  * See the License for the specific language governing permissions and
  14  * limitations under the License.
  15  */
  16
  17 #ifndef PACKETS_H
  18 #define PACKETS_H 1
  19
  20 #include <inttypes.h>
  21 #include <sys/types.h>
  22 #include <netinet/in.h>
  23 #include <stdint.h>
  24 #include <string.h>
  25 #include "compiler.h"
  26 #include "geneve.h"
  27 #include "openvswitch/types.h"
  28 #include "odp-netlink.h"
  29 #include "random.h"
  30 #include "hash.h"
  31 #include "tun-metadata.h"
  32 #include "unaligned.h"
  33 #include "util.h"
  34
  35 struct dp_packet;
  36 struct ds;
  37
  38 /* Tunnel information used in flow key and metadata. */
  39 struct flow_tnl {
  40     ovs_be32 ip_dst;
  41     ovs_be32 ip_src;
  42     ovs_be64 tun_id;
  43     uint16_t flags;
  44     uint8_t ip_tos;
  45     uint8_t ip_ttl;
  46     ovs_be16 tp_src;
  47     ovs_be16 tp_dst;
  48     ovs_be16 gbp_id;
  49     uint8_t  gbp_flags;
  50     uint8_t  pad1[5];        /* Pad to 64 bits. */
  51     struct tun_metadata metadata;
  52 };
  53
  54 /* Some flags are exposed through OpenFlow while others are used only
  55  * internally. */
  56
  57 /* Public flags */
  58 #define FLOW_TNL_F_OAM (1 << 0)
  59
  60 #define FLOW_TNL_PUB_F_MASK ((1 << 1) - 1)
  61
  62 /* Private flags */
  63 #define FLOW_TNL_F_DONT_FRAGMENT (1 << 1)
  64 #define FLOW_TNL_F_CSUM (1 << 2)
  65 #define FLOW_TNL_F_KEY (1 << 3)
  66
  67 #define FLOW_TNL_F_MASK ((1 << 4) - 1)
  68
  69 /* Purely internal to OVS userspace. These flags should never be exposed to
  70  * the outside world and so aren't included in the flags mask. */
  71
  72 /* Tunnel information is in userspace datapath format. */
  73 #define FLOW_TNL_F_UDPIF (1 << 4)
  74
  75 /* Returns an offset to 'src' covering all the meaningful fields in 'src'. */
  76 static inline size_t
  77 flow_tnl_size(const struct flow_tnl *src)
  78 {
  79     if (!src->ip_dst) {
  80         /* Covers ip_dst only. */
  81         return offsetof(struct flow_tnl, ip_src);
  82     }
  83     if (src->flags & FLOW_TNL_F_UDPIF) {
  84         /* Datapath format, cover all options we have. */
  85         return offsetof(struct flow_tnl, metadata.opts)
  86             + src->metadata.present.len;
  87     }
  88     if (!src->metadata.present.map) {
  89         /* No TLVs, opts is irrelevant. */
  90         return offsetof(struct flow_tnl, metadata.opts);
  91     }
  92     /* Have decoded TLVs, opts is relevant. */
  93     return sizeof *src;
  94 }
  95
  96 /* Copy flow_tnl, but avoid copying unused portions of tun_metadata.  Unused
  97  * data in 'dst' is NOT cleared, so this must not be used in cases where the
  98  * uninitialized portion may be hashed over. */
  99 static inline void
 100 flow_tnl_copy__(struct flow_tnl *dst, const struct flow_tnl *src)
 101 {
 102     memcpy(dst, src, flow_tnl_size(src));
 103 }
 104
 105 static inline bool
 106 flow_tnl_equal(const struct flow_tnl *a, const struct flow_tnl *b)
 107 {
 108     size_t a_size = flow_tnl_size(a);
 109
 110     return a_size == flow_tnl_size(b) && !memcmp(a, b, a_size);
 111 }
 112
 113 /* Unfortunately, a "struct flow" sometimes has to handle OpenFlow port
 114  * numbers and other times datapath (dpif) port numbers.  This union allows
 115  * access to both. */
 116 union flow_in_port {
 117     odp_port_t odp_port;
 118     ofp_port_t ofp_port;
 119 };
 120
 121 /* Datapath packet metadata */
 122 struct pkt_metadata {
 123     uint32_t recirc_id;         /* Recirculation id carried with the
 124                                    recirculating packets. 0 for packets
 125                                    received from the wire. */
 126     uint32_t dp_hash;           /* hash value computed by the recirculation
 127                                    action. */
 128     uint32_t skb_priority;      /* Packet priority for QoS. */
 129     uint32_t pkt_mark;          /* Packet mark. */
 130     uint16_t ct_state;          /* Connection state. */
 131     uint16_t ct_zone;           /* Connection zone. */
 132     uint32_t ct_mark;           /* Connection mark. */
 133     ovs_u128 ct_label;          /* Connection label. */
 134     union flow_in_port in_port; /* Input port. */
 135     struct flow_tnl tunnel;     /* Encapsulating tunnel parameters. Note that
 136                                  * if 'ip_dst' == 0, the rest of the fields may
 137                                  * be uninitialized. */
 138 };
 139
 140 static inline void
 141 pkt_metadata_init(struct pkt_metadata *md, odp_port_t port)
 142 {
 143     /* It can be expensive to zero out all of the tunnel metadata. However,
 144      * we can just zero out ip_dst and the rest of the data will never be
 145      * looked at. */
 146     memset(md, 0, offsetof(struct pkt_metadata, in_port));
 147     md->tunnel.ip_dst = 0;
 148
 149     md->in_port.odp_port = port;
 150 }
 151
 152 bool dpid_from_string(const char *s, uint64_t *dpidp);
 153
 154 #define ETH_ADDR_LEN           6
 155
 156 static const struct eth_addr eth_addr_broadcast OVS_UNUSED
 157     = { { { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff } } };
 158
 159 static const struct eth_addr eth_addr_exact OVS_UNUSED
 160     = { { { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff } } };
 161
 162 static const struct eth_addr eth_addr_zero OVS_UNUSED
 163     = { { { 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 } } };
 164
 165 static const struct eth_addr eth_addr_stp OVS_UNUSED
 166     = { { { 0x01, 0x80, 0xC2, 0x00, 0x00, 0x00 } } };
 167
 168 static const struct eth_addr eth_addr_lacp OVS_UNUSED
 169     = { { { 0x01, 0x80, 0xC2, 0x00, 0x00, 0x02 } } };
 170
 171 static const struct eth_addr eth_addr_bfd OVS_UNUSED
 172     = { { { 0x00, 0x23, 0x20, 0x00, 0x00, 0x01 } } };
 173
 174 static inline bool eth_addr_is_broadcast(const struct eth_addr a)
 175 {
 176     return (a.be16[0] & a.be16[1] & a.be16[2]) == htons(0xffff);
 177 }
 178
 179 static inline bool eth_addr_is_multicast(const struct eth_addr a)
 180 {
 181     return a.ea[0] & 1;
 182 }
 183
 184 static inline bool eth_addr_is_local(const struct eth_addr a)
 185 {
 186     /* Local if it is either a locally administered address or a Nicira random
 187      * address. */
 188     return a.ea[0] & 2
 189         || (a.be16[0] == htons(0x0023)
 190             && (a.be16[1] & htons(0xff80)) == htons(0x2080));
 191 }
 192 static inline bool eth_addr_is_zero(const struct eth_addr a)
 193 {
 194     return !(a.be16[0] | a.be16[1] | a.be16[2]);
 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
 208 static inline bool eth_addr_equals(const struct eth_addr a,
 209                                    const struct eth_addr b)
 210 {
 211     return !eth_addr_compare_3way(a, b);
 212 }
 213
 214 static inline bool eth_addr_equal_except(const struct eth_addr a,
 215                                          const struct eth_addr b,
 216                                          const struct eth_addr mask)
 217 {
 218     return !(((a.be16[0] ^ b.be16[0]) & mask.be16[0])
 219              || ((a.be16[1] ^ b.be16[1]) & mask.be16[1])
 220              || ((a.be16[2] ^ b.be16[2]) & mask.be16[2]));
 221 }
 222
 223 static inline uint64_t eth_addr_to_uint64(const struct eth_addr ea)
 224 {
 225     return (((uint64_t) ntohs(ea.be16[0]) << 32)
 226             | ((uint64_t) ntohs(ea.be16[1]) << 16)
 227             | ntohs(ea.be16[2]));
 228 }
 229
 230 static inline uint64_t eth_addr_vlan_to_uint64(const struct eth_addr ea,
 231                                                uint16_t vlan)
 232 {
 233     return (((uint64_t)vlan << 48) | eth_addr_to_uint64(ea));
 234 }
 235
 236 static inline void eth_addr_from_uint64(uint64_t x, struct eth_addr *ea)
 237 {
 238     ea->be16[0] = htons(x >> 32);
 239     ea->be16[1] = htons((x & 0xFFFF0000) >> 16);
 240     ea->be16[2] = htons(x & 0xFFFF);
 241 }
 242
 243 static inline struct eth_addr eth_addr_invert(const struct eth_addr src)
 244 {
 245     struct eth_addr dst;
 246
 247     for (int i = 0; i < ARRAY_SIZE(src.be16); i++) {
 248         dst.be16[i] = ~src.be16[i];
 249     }
 250
 251     return dst;
 252 }
 253
 254 static inline void eth_addr_mark_random(struct eth_addr *ea)
 255 {
 256     ea->ea[0] &= ~1;                /* Unicast. */
 257     ea->ea[0] |= 2;                 /* Private. */
 258 }
 259
 260 static inline void eth_addr_random(struct eth_addr *ea)
 261 {
 262     random_bytes((uint8_t *)ea, sizeof *ea);
 263     eth_addr_mark_random(ea);
 264 }
 265
 266 static inline void eth_addr_nicira_random(struct eth_addr *ea)
 267 {
 268     eth_addr_random(ea);
 269
 270     /* Set the OUI to the Nicira one. */
 271     ea->ea[0] = 0x00;
 272     ea->ea[1] = 0x23;
 273     ea->ea[2] = 0x20;
 274
 275     /* Set the top bit to indicate random Nicira address. */
 276     ea->ea[3] |= 0x80;
 277 }
 278 static inline uint32_t hash_mac(const struct eth_addr ea,
 279                                 const uint16_t vlan, const uint32_t basis)
 280 {
 281     return hash_uint64_basis(eth_addr_vlan_to_uint64(ea, vlan), basis);
 282 }
 283
 284 bool eth_addr_is_reserved(const struct eth_addr);
 285 bool eth_addr_from_string(const char *, struct eth_addr *);
 286
 287 void compose_rarp(struct dp_packet *, const struct eth_addr);
 288
 289 void eth_push_vlan(struct dp_packet *, ovs_be16 tpid, ovs_be16 tci);
 290 void eth_pop_vlan(struct dp_packet *);
 291
 292 const char *eth_from_hex(const char *hex, struct dp_packet **packetp);
 293 void eth_format_masked(const struct eth_addr ea,
 294                        const struct eth_addr *mask, struct ds *s);
 295
 296 void set_mpls_lse(struct dp_packet *, ovs_be32 label);
 297 void push_mpls(struct dp_packet *packet, ovs_be16 ethtype, ovs_be32 lse);
 298 void pop_mpls(struct dp_packet *, ovs_be16 ethtype);
 299
 300 void set_mpls_lse_ttl(ovs_be32 *lse, uint8_t ttl);
 301 void set_mpls_lse_tc(ovs_be32 *lse, uint8_t tc);
 302 void set_mpls_lse_label(ovs_be32 *lse, ovs_be32 label);
 303 void set_mpls_lse_bos(ovs_be32 *lse, uint8_t bos);
 304 ovs_be32 set_mpls_lse_values(uint8_t ttl, uint8_t tc, uint8_t bos,
 305                              ovs_be32 label);
 306
 307 /* Example:
 308  *
 309  * struct eth_addr mac;
 310  *    [...]
 311  * printf("The Ethernet address is "ETH_ADDR_FMT"\n", ETH_ADDR_ARGS(mac));
 312  *
 313  */
 314 #define ETH_ADDR_FMT                                                    \
 315     "%02"PRIx8":%02"PRIx8":%02"PRIx8":%02"PRIx8":%02"PRIx8":%02"PRIx8
 316 #define ETH_ADDR_ARGS(EA) ETH_ADDR_BYTES_ARGS((EA).ea)
 317 #define ETH_ADDR_BYTES_ARGS(EAB) \
 318          (EAB)[0], (EAB)[1], (EAB)[2], (EAB)[3], (EAB)[4], (EAB)[5]
 319
 320 /* Example:
 321  *
 322  * char *string = "1 00:11:22:33:44:55 2";
 323  * struct eth_addr mac;
 324  * int a, b;
 325  *
 326  * if (ovs_scan(string, "%d"ETH_ADDR_SCAN_FMT"%d",
 327  *              &a, ETH_ADDR_SCAN_ARGS(mac), &b)) {
 328  *     ...
 329  * }
 330  */
 331 #define ETH_ADDR_SCAN_FMT "%"SCNx8":%"SCNx8":%"SCNx8":%"SCNx8":%"SCNx8":%"SCNx8
 332 #define ETH_ADDR_SCAN_ARGS(EA) \
 333     &(EA).ea[0], &(EA).ea[1], &(EA).ea[2], &(EA).ea[3], &(EA).ea[4], &(EA).ea[5]
 334
 335 #define ETH_TYPE_IP            0x0800
 336 #define ETH_TYPE_ARP           0x0806
 337 #define ETH_TYPE_TEB           0x6558
 338 #define ETH_TYPE_VLAN_8021Q    0x8100
 339 #define ETH_TYPE_VLAN          ETH_TYPE_VLAN_8021Q
 340 #define ETH_TYPE_VLAN_8021AD   0x88a8
 341 #define ETH_TYPE_IPV6          0x86dd
 342 #define ETH_TYPE_LACP          0x8809
 343 #define ETH_TYPE_RARP          0x8035
 344 #define ETH_TYPE_MPLS          0x8847
 345 #define ETH_TYPE_MPLS_MCAST    0x8848
 346
 347 static inline bool eth_type_mpls(ovs_be16 eth_type)
 348 {
 349     return eth_type == htons(ETH_TYPE_MPLS) ||
 350         eth_type == htons(ETH_TYPE_MPLS_MCAST);
 351 }
 352
 353 static inline bool eth_type_vlan(ovs_be16 eth_type)
 354 {
 355     return eth_type == htons(ETH_TYPE_VLAN_8021Q) ||
 356         eth_type == htons(ETH_TYPE_VLAN_8021AD);
 357 }
 358
 359
 360 /* Minimum value for an Ethernet type.  Values below this are IEEE 802.2 frame
 361  * lengths. */
 362 #define ETH_TYPE_MIN           0x600
 363
 364 #define ETH_HEADER_LEN 14
 365 #define ETH_PAYLOAD_MIN 46
 366 #define ETH_PAYLOAD_MAX 1500
 367 #define ETH_TOTAL_MIN (ETH_HEADER_LEN + ETH_PAYLOAD_MIN)
 368 #define ETH_TOTAL_MAX (ETH_HEADER_LEN + ETH_PAYLOAD_MAX)
 369 #define ETH_VLAN_TOTAL_MAX (ETH_HEADER_LEN + VLAN_HEADER_LEN + ETH_PAYLOAD_MAX)
 370 OVS_PACKED(
 371 struct eth_header {
 372     struct eth_addr eth_dst;
 373     struct eth_addr eth_src;
 374     ovs_be16 eth_type;
 375 });
 376 BUILD_ASSERT_DECL(ETH_HEADER_LEN == sizeof(struct eth_header));
 377
 378 #define LLC_DSAP_SNAP 0xaa
 379 #define LLC_SSAP_SNAP 0xaa
 380 #define LLC_CNTL_SNAP 3
 381
 382 #define LLC_HEADER_LEN 3
 383 OVS_PACKED(
 384 struct llc_header {
 385     uint8_t llc_dsap;
 386     uint8_t llc_ssap;
 387     uint8_t llc_cntl;
 388 });
 389 BUILD_ASSERT_DECL(LLC_HEADER_LEN == sizeof(struct llc_header));
 390
 391 /* LLC field values used for STP frames. */
 392 #define STP_LLC_SSAP 0x42
 393 #define STP_LLC_DSAP 0x42
 394 #define STP_LLC_CNTL 0x03
 395
 396 #define SNAP_ORG_ETHERNET "\0\0" /* The compiler adds a null byte, so
 397                                     sizeof(SNAP_ORG_ETHERNET) == 3. */
 398 #define SNAP_HEADER_LEN 5
 399 OVS_PACKED(
 400 struct snap_header {
 401     uint8_t snap_org[3];
 402     ovs_be16 snap_type;
 403 });
 404 BUILD_ASSERT_DECL(SNAP_HEADER_LEN == sizeof(struct snap_header));
 405
 406 #define LLC_SNAP_HEADER_LEN (LLC_HEADER_LEN + SNAP_HEADER_LEN)
 407 OVS_PACKED(
 408 struct llc_snap_header {
 409     struct llc_header llc;
 410     struct snap_header snap;
 411 });
 412 BUILD_ASSERT_DECL(LLC_SNAP_HEADER_LEN == sizeof(struct llc_snap_header));
 413
 414 #define VLAN_VID_MASK 0x0fff
 415 #define VLAN_VID_SHIFT 0
 416
 417 #define VLAN_PCP_MASK 0xe000
 418 #define VLAN_PCP_SHIFT 13
 419
 420 #define VLAN_CFI 0x1000
 421 #define VLAN_CFI_SHIFT 12
 422
 423 /* Given the vlan_tci field from an 802.1Q header, in network byte order,
 424  * returns the VLAN ID in host byte order. */
 425 static inline uint16_t
 426 vlan_tci_to_vid(ovs_be16 vlan_tci)
 427 {
 428     return (ntohs(vlan_tci) & VLAN_VID_MASK) >> VLAN_VID_SHIFT;
 429 }
 430
 431 /* Given the vlan_tci field from an 802.1Q header, in network byte order,
 432  * returns the priority code point (PCP) in host byte order. */
 433 static inline int
 434 vlan_tci_to_pcp(ovs_be16 vlan_tci)
 435 {
 436     return (ntohs(vlan_tci) & VLAN_PCP_MASK) >> VLAN_PCP_SHIFT;
 437 }
 438
 439 /* Given the vlan_tci field from an 802.1Q header, in network byte order,
 440  * returns the Canonical Format Indicator (CFI). */
 441 static inline int
 442 vlan_tci_to_cfi(ovs_be16 vlan_tci)
 443 {
 444     return (vlan_tci & htons(VLAN_CFI)) != 0;
 445 }
 446
 447 #define VLAN_HEADER_LEN 4
 448 struct vlan_header {
 449     ovs_be16 vlan_tci;          /* Lowest 12 bits are VLAN ID. */
 450     ovs_be16 vlan_next_type;
 451 };
 452 BUILD_ASSERT_DECL(VLAN_HEADER_LEN == sizeof(struct vlan_header));
 453
 454 #define VLAN_ETH_HEADER_LEN (ETH_HEADER_LEN + VLAN_HEADER_LEN)
 455 OVS_PACKED(
 456 struct vlan_eth_header {
 457     struct eth_addr veth_dst;
 458     struct eth_addr veth_src;
 459     ovs_be16 veth_type;         /* Always htons(ETH_TYPE_VLAN). */
 460     ovs_be16 veth_tci;          /* Lowest 12 bits are VLAN ID. */
 461     ovs_be16 veth_next_type;
 462 });
 463 BUILD_ASSERT_DECL(VLAN_ETH_HEADER_LEN == sizeof(struct vlan_eth_header));
 464
 465 /* MPLS related definitions */
 466 #define MPLS_TTL_MASK       0x000000ff
 467 #define MPLS_TTL_SHIFT      0
 468
 469 #define MPLS_BOS_MASK       0x00000100
 470 #define MPLS_BOS_SHIFT      8
 471
 472 #define MPLS_TC_MASK        0x00000e00
 473 #define MPLS_TC_SHIFT       9
 474
 475 #define MPLS_LABEL_MASK     0xfffff000
 476 #define MPLS_LABEL_SHIFT    12
 477
 478 #define MPLS_HLEN           4
 479
 480 struct mpls_hdr {
 481     ovs_16aligned_be32 mpls_lse;
 482 };
 483 BUILD_ASSERT_DECL(MPLS_HLEN == sizeof(struct mpls_hdr));
 484
 485 /* Given a mpls label stack entry in network byte order
 486  * return mpls label in host byte order */
 487 static inline uint32_t
 488 mpls_lse_to_label(ovs_be32 mpls_lse)
 489 {
 490     return (ntohl(mpls_lse) & MPLS_LABEL_MASK) >> MPLS_LABEL_SHIFT;
 491 }
 492
 493 /* Given a mpls label stack entry in network byte order
 494  * return mpls tc */
 495 static inline uint8_t
 496 mpls_lse_to_tc(ovs_be32 mpls_lse)
 497 {
 498     return (ntohl(mpls_lse) & MPLS_TC_MASK) >> MPLS_TC_SHIFT;
 499 }
 500
 501 /* Given a mpls label stack entry in network byte order
 502  * return mpls ttl */
 503 static inline uint8_t
 504 mpls_lse_to_ttl(ovs_be32 mpls_lse)
 505 {
 506     return (ntohl(mpls_lse) & MPLS_TTL_MASK) >> MPLS_TTL_SHIFT;
 507 }
 508
 509 /* Set TTL in mpls lse. */
 510 static inline void
 511 flow_set_mpls_lse_ttl(ovs_be32 *mpls_lse, uint8_t ttl)
 512 {
 513     *mpls_lse &= ~htonl(MPLS_TTL_MASK);
 514     *mpls_lse |= htonl(ttl << MPLS_TTL_SHIFT);
 515 }
 516
 517 /* Given a mpls label stack entry in network byte order
 518  * return mpls BoS bit  */
 519 static inline uint8_t
 520 mpls_lse_to_bos(ovs_be32 mpls_lse)
 521 {
 522     return (mpls_lse & htonl(MPLS_BOS_MASK)) != 0;
 523 }
 524
 525 #define IP_FMT "%"PRIu32".%"PRIu32".%"PRIu32".%"PRIu32
 526 #define IP_ARGS(ip)                             \
 527     ntohl(ip) >> 24,                            \
 528     (ntohl(ip) >> 16) & 0xff,                   \
 529     (ntohl(ip) >> 8) & 0xff,                    \
 530     ntohl(ip) & 0xff
 531
 532 /* Example:
 533  *
 534  * char *string = "1 33.44.55.66 2";
 535  * ovs_be32 ip;
 536  * int a, b;
 537  *
 538  * if (ovs_scan(string, "%d"IP_SCAN_FMT"%d", &a, IP_SCAN_ARGS(&ip), &b)) {
 539  *     ...
 540  * }
 541  */
 542 #define IP_SCAN_FMT "%"SCNu8".%"SCNu8".%"SCNu8".%"SCNu8
 543 #define IP_SCAN_ARGS(ip)                                    \
 544         ((void) (ovs_be32) *(ip), &((uint8_t *) ip)[0]),    \
 545         &((uint8_t *) ip)[1],                               \
 546         &((uint8_t *) ip)[2],                               \
 547         &((uint8_t *) ip)[3]
 548
 549 /* Returns true if 'netmask' is a CIDR netmask, that is, if it consists of N
 550  * high-order 1-bits and 32-N low-order 0-bits. */
 551 static inline bool
 552 ip_is_cidr(ovs_be32 netmask)
 553 {
 554     uint32_t x = ~ntohl(netmask);
 555     return !(x & (x + 1));
 556 }
 557 static inline bool
 558 ip_is_multicast(ovs_be32 ip)
 559 {
 560     return (ip & htonl(0xf0000000)) == htonl(0xe0000000);
 561 }
 562 static inline bool
 563 ip_is_local_multicast(ovs_be32 ip)
 564 {
 565     return (ip & htonl(0xffffff00)) == htonl(0xe0000000);
 566 }
 567 int ip_count_cidr_bits(ovs_be32 netmask);
 568 void ip_format_masked(ovs_be32 ip, ovs_be32 mask, struct ds *);
 569 char *ip_parse_masked(const char *s, ovs_be32 *ip, ovs_be32 *mask)
 570     OVS_WARN_UNUSED_RESULT;
 571
 572 #define IP_VER(ip_ihl_ver) ((ip_ihl_ver) >> 4)
 573 #define IP_IHL(ip_ihl_ver) ((ip_ihl_ver) & 15)
 574 #define IP_IHL_VER(ihl, ver) (((ver) << 4) | (ihl))
 575
 576 #ifndef IPPROTO_SCTP
 577 #define IPPROTO_SCTP 132
 578 #endif
 579
 580 /* TOS fields. */
 581 #define IP_ECN_NOT_ECT 0x0
 582 #define IP_ECN_ECT_1 0x01
 583 #define IP_ECN_ECT_0 0x02
 584 #define IP_ECN_CE 0x03
 585 #define IP_ECN_MASK 0x03
 586 #define IP_DSCP_MASK 0xfc
 587
 588 #define IP_VERSION 4
 589
 590 #define IP_DONT_FRAGMENT  0x4000 /* Don't fragment. */
 591 #define IP_MORE_FRAGMENTS 0x2000 /* More fragments. */
 592 #define IP_FRAG_OFF_MASK  0x1fff /* Fragment offset. */
 593 #define IP_IS_FRAGMENT(ip_frag_off) \
 594         ((ip_frag_off) & htons(IP_MORE_FRAGMENTS | IP_FRAG_OFF_MASK))
 595
 596 #define IP_HEADER_LEN 20
 597 struct ip_header {
 598     uint8_t ip_ihl_ver;
 599     uint8_t ip_tos;
 600     ovs_be16 ip_tot_len;
 601     ovs_be16 ip_id;
 602     ovs_be16 ip_frag_off;
 603     uint8_t ip_ttl;
 604     uint8_t ip_proto;
 605     ovs_be16 ip_csum;
 606     ovs_16aligned_be32 ip_src;
 607     ovs_16aligned_be32 ip_dst;
 608 };
 609
 610 BUILD_ASSERT_DECL(IP_HEADER_LEN == sizeof(struct ip_header));
 611
 612 #define ICMP_HEADER_LEN 8
 613 struct icmp_header {
 614     uint8_t icmp_type;
 615     uint8_t icmp_code;
 616     ovs_be16 icmp_csum;
 617     union {
 618         struct {
 619             ovs_be16 id;
 620             ovs_be16 seq;
 621         } echo;
 622         struct {
 623             ovs_be16 empty;
 624             ovs_be16 mtu;
 625         } frag;
 626         ovs_16aligned_be32 gateway;
 627     } icmp_fields;
 628 };
 629 BUILD_ASSERT_DECL(ICMP_HEADER_LEN == sizeof(struct icmp_header));
 630
 631 #define IGMP_HEADER_LEN 8
 632 struct igmp_header {
 633     uint8_t igmp_type;
 634     uint8_t igmp_code;
 635     ovs_be16 igmp_csum;
 636     ovs_16aligned_be32 group;
 637 };
 638 BUILD_ASSERT_DECL(IGMP_HEADER_LEN == sizeof(struct igmp_header));
 639
 640 #define IGMPV3_HEADER_LEN 8
 641 struct igmpv3_header {
 642     uint8_t type;
 643     uint8_t rsvr1;
 644     ovs_be16 csum;
 645     ovs_be16 rsvr2;
 646     ovs_be16 ngrp;
 647 };
 648 BUILD_ASSERT_DECL(IGMPV3_HEADER_LEN == sizeof(struct igmpv3_header));
 649
 650 #define IGMPV3_RECORD_LEN 8
 651 struct igmpv3_record {
 652     uint8_t type;
 653     uint8_t aux_len;
 654     ovs_be16 nsrcs;
 655     ovs_16aligned_be32 maddr;
 656 };
 657 BUILD_ASSERT_DECL(IGMPV3_RECORD_LEN == sizeof(struct igmpv3_record));
 658
 659 #define IGMP_HOST_MEMBERSHIP_QUERY    0x11 /* From RFC1112 */
 660 #define IGMP_HOST_MEMBERSHIP_REPORT   0x12 /* Ditto */
 661 #define IGMPV2_HOST_MEMBERSHIP_REPORT 0x16 /* V2 version of 0x12 */
 662 #define IGMP_HOST_LEAVE_MESSAGE       0x17
 663 #define IGMPV3_HOST_MEMBERSHIP_REPORT 0x22 /* V3 version of 0x12 */
 664
 665 /*
 666  * IGMPv3 and MLDv2 use the same codes.
 667  */
 668 #define IGMPV3_MODE_IS_INCLUDE 1
 669 #define IGMPV3_MODE_IS_EXCLUDE 2
 670 #define IGMPV3_CHANGE_TO_INCLUDE_MODE 3
 671 #define IGMPV3_CHANGE_TO_EXCLUDE_MODE 4
 672 #define IGMPV3_ALLOW_NEW_SOURCES 5
 673 #define IGMPV3_BLOCK_OLD_SOURCES 6
 674
 675 #define SCTP_HEADER_LEN 12
 676 struct sctp_header {
 677     ovs_be16 sctp_src;
 678     ovs_be16 sctp_dst;
 679     ovs_16aligned_be32 sctp_vtag;
 680     ovs_16aligned_be32 sctp_csum;
 681 };
 682 BUILD_ASSERT_DECL(SCTP_HEADER_LEN == sizeof(struct sctp_header));
 683
 684 #define UDP_HEADER_LEN 8
 685 struct udp_header {
 686     ovs_be16 udp_src;
 687     ovs_be16 udp_dst;
 688     ovs_be16 udp_len;
 689     ovs_be16 udp_csum;
 690 };
 691 BUILD_ASSERT_DECL(UDP_HEADER_LEN == sizeof(struct udp_header));
 692
 693 #define TCP_FIN 0x001
 694 #define TCP_SYN 0x002
 695 #define TCP_RST 0x004
 696 #define TCP_PSH 0x008
 697 #define TCP_ACK 0x010
 698 #define TCP_URG 0x020
 699 #define TCP_ECE 0x040
 700 #define TCP_CWR 0x080
 701 #define TCP_NS  0x100
 702
 703 #define TCP_CTL(flags, offset) (htons((flags) | ((offset) << 12)))
 704 #define TCP_FLAGS(tcp_ctl) (ntohs(tcp_ctl) & 0x0fff)
 705 #define TCP_FLAGS_BE16(tcp_ctl) ((tcp_ctl) & htons(0x0fff))
 706 #define TCP_OFFSET(tcp_ctl) (ntohs(tcp_ctl) >> 12)
 707
 708 #define TCP_HEADER_LEN 20
 709 struct tcp_header {
 710     ovs_be16 tcp_src;
 711     ovs_be16 tcp_dst;
 712     ovs_16aligned_be32 tcp_seq;
 713     ovs_16aligned_be32 tcp_ack;
 714     ovs_be16 tcp_ctl;
 715     ovs_be16 tcp_winsz;
 716     ovs_be16 tcp_csum;
 717     ovs_be16 tcp_urg;
 718 };
 719 BUILD_ASSERT_DECL(TCP_HEADER_LEN == sizeof(struct tcp_header));
 720
 721 /* Connection states */
 722 #define CS_NEW               0x01
 723 #define CS_ESTABLISHED       0x02
 724 #define CS_RELATED           0x04
 725 #define CS_INVALID           0x20
 726 #define CS_REPLY_DIR         0x40
 727 #define CS_TRACKED           0x80
 728
 729 /* Undefined connection state bits. */
 730 #define CS_SUPPORTED_MASK    (CS_NEW | CS_ESTABLISHED | CS_RELATED \
 731                               | CS_INVALID | CS_REPLY_DIR | CS_TRACKED)
 732 #define CS_UNSUPPORTED_MASK  (~(uint32_t)CS_SUPPORTED_MASK)
 733
 734 #define ARP_HRD_ETHERNET 1
 735 #define ARP_PRO_IP 0x0800
 736 #define ARP_OP_REQUEST 1
 737 #define ARP_OP_REPLY 2
 738 #define ARP_OP_RARP 3
 739
 740 #define ARP_ETH_HEADER_LEN 28
 741 struct arp_eth_header {
 742     /* Generic members. */
 743     ovs_be16 ar_hrd;           /* Hardware type. */
 744     ovs_be16 ar_pro;           /* Protocol type. */
 745     uint8_t ar_hln;            /* Hardware address length. */
 746     uint8_t ar_pln;            /* Protocol address length. */
 747     ovs_be16 ar_op;            /* Opcode. */
 748
 749     /* Ethernet+IPv4 specific members. */
 750     struct eth_addr ar_sha;     /* Sender hardware address. */
 751     ovs_16aligned_be32 ar_spa;  /* Sender protocol address. */
 752     struct eth_addr ar_tha;     /* Target hardware address. */
 753     ovs_16aligned_be32 ar_tpa;  /* Target protocol address. */
 754 };
 755 BUILD_ASSERT_DECL(ARP_ETH_HEADER_LEN == sizeof(struct arp_eth_header));
 756
 757 /* Like struct in6_addr, but whereas that struct requires 32-bit alignment on
 758  * most implementations, this one only requires 16-bit alignment. */
 759 union ovs_16aligned_in6_addr {
 760     ovs_be16 be16[8];
 761     ovs_16aligned_be32 be32[4];
 762 };
 763
 764 /* Like struct in6_hdr, but whereas that struct requires 32-bit alignment, this
 765  * one only requires 16-bit alignment. */
 766 struct ovs_16aligned_ip6_hdr {
 767     union {
 768         struct ovs_16aligned_ip6_hdrctl {
 769             ovs_16aligned_be32 ip6_un1_flow;
 770             ovs_be16 ip6_un1_plen;
 771             uint8_t ip6_un1_nxt;
 772             uint8_t ip6_un1_hlim;
 773         } ip6_un1;
 774         uint8_t ip6_un2_vfc;
 775     } ip6_ctlun;
 776     union ovs_16aligned_in6_addr ip6_src;
 777     union ovs_16aligned_in6_addr ip6_dst;
 778 };
 779
 780 /* Like struct in6_frag, but whereas that struct requires 32-bit alignment,
 781  * this one only requires 16-bit alignment. */
 782 struct ovs_16aligned_ip6_frag {
 783     uint8_t ip6f_nxt;
 784     uint8_t ip6f_reserved;
 785     ovs_be16 ip6f_offlg;
 786     ovs_16aligned_be32 ip6f_ident;
 787 };
 788
 789 #define ICMP6_HEADER_LEN 4
 790 struct icmp6_header {
 791     uint8_t icmp6_type;
 792     uint8_t icmp6_code;
 793     ovs_be16 icmp6_cksum;
 794 };
 795 BUILD_ASSERT_DECL(ICMP6_HEADER_LEN == sizeof(struct icmp6_header));
 796
 797 /* Neighbor Discovery option field.
 798  * ND options are always a multiple of 8 bytes in size. */
 799 #define ND_OPT_LEN 8
 800 struct ovs_nd_opt {
 801     uint8_t  nd_opt_type;      /* Values defined in icmp6.h */
 802     uint8_t  nd_opt_len;       /* in units of 8 octets (the size of this struct) */
 803     struct eth_addr nd_opt_mac;   /* Ethernet address in the case of SLL or TLL options */
 804 };
 805 BUILD_ASSERT_DECL(ND_OPT_LEN == sizeof(struct ovs_nd_opt));
 806
 807 /* Like struct nd_msg (from ndisc.h), but whereas that struct requires 32-bit
 808  * alignment, this one only requires 16-bit alignment. */
 809 #define ND_MSG_LEN 24
 810 struct ovs_nd_msg {
 811     struct icmp6_header icmph;
 812     ovs_16aligned_be32 rco_flags;
 813     union ovs_16aligned_in6_addr target;
 814     struct ovs_nd_opt options[0];
 815 };
 816 BUILD_ASSERT_DECL(ND_MSG_LEN == sizeof(struct ovs_nd_msg));
 817
 818 /*
 819  * Use the same struct for MLD and MLD2, naming members as the defined fields in
 820  * in the corresponding version of the protocol, though they are reserved in the
 821  * other one.
 822  */
 823 #define MLD_HEADER_LEN 8
 824 struct mld_header {
 825     uint8_t type;
 826     uint8_t code;
 827     ovs_be16 csum;
 828     ovs_be16 mrd;
 829     ovs_be16 ngrp;
 830 };
 831 BUILD_ASSERT_DECL(MLD_HEADER_LEN == sizeof(struct mld_header));
 832
 833 #define MLD2_RECORD_LEN 20
 834 struct mld2_record {
 835     uint8_t type;
 836     uint8_t aux_len;
 837     ovs_be16 nsrcs;
 838     union ovs_16aligned_in6_addr maddr;
 839 };
 840 BUILD_ASSERT_DECL(MLD2_RECORD_LEN == sizeof(struct mld2_record));
 841
 842 #define MLD_QUERY 130
 843 #define MLD_REPORT 131
 844 #define MLD_DONE 132
 845 #define MLD2_REPORT 143
 846
 847 /* The IPv6 flow label is in the lower 20 bits of the first 32-bit word. */
 848 #define IPV6_LABEL_MASK 0x000fffff
 849
 850 /* Example:
 851  *
 852  * char *string = "1 ::1 2";
 853  * char ipv6_s[IPV6_SCAN_LEN + 1];
 854  * struct in6_addr ipv6;
 855  *
 856  * if (ovs_scan(string, "%d"IPV6_SCAN_FMT"%d", &a, ipv6_s, &b)
 857  *     && inet_pton(AF_INET6, ipv6_s, &ipv6) == 1) {
 858  *     ...
 859  * }
 860  */
 861 #define IPV6_SCAN_FMT "%46[0123456789abcdefABCDEF:.]"
 862 #define IPV6_SCAN_LEN 46
 863
 864 extern const struct in6_addr in6addr_exact;
 865 #define IN6ADDR_EXACT_INIT { { { 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, \
 866                                  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff } } }
 867
 868 extern const struct in6_addr in6addr_all_hosts;
 869 #define IN6ADDR_ALL_HOSTS_INIT { { { 0xff,0x02,0x00,0x00,0x00,0x00,0x00,0x00, \
 870                                      0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x01 } } }
 871
 872 static inline bool ipv6_addr_equals(const struct in6_addr *a,
 873                                     const struct in6_addr *b)
 874 {
 875 #ifdef IN6_ARE_ADDR_EQUAL
 876     return IN6_ARE_ADDR_EQUAL(a, b);
 877 #else
 878     return !memcmp(a, b, sizeof(*a));
 879 #endif
 880 }
 881
 882 static inline bool ipv6_mask_is_any(const struct in6_addr *mask) {
 883     return ipv6_addr_equals(mask, &in6addr_any);
 884 }
 885
 886 static inline bool ipv6_mask_is_exact(const struct in6_addr *mask) {
 887     return ipv6_addr_equals(mask, &in6addr_exact);
 888 }
 889
 890 static inline bool ipv6_is_all_hosts(const struct in6_addr *addr) {
 891     return ipv6_addr_equals(addr, &in6addr_all_hosts);
 892 }
 893
 894 static inline bool ipv6_addr_is_set(const struct in6_addr *addr) {
 895     return !ipv6_addr_equals(addr, &in6addr_any);
 896 }
 897
 898 static inline bool ipv6_addr_is_multicast(const struct in6_addr *ip) {
 899     return ip->s6_addr[0] == 0xff;
 900 }
 901
 902 static inline void
 903 in6_addr_set_mapped_ipv4(struct in6_addr *addr, ovs_be32 ip4)
 904 {
 905     union ovs_16aligned_in6_addr *taddr = (void *) addr;
 906     memset(taddr->be16, 0, sizeof(taddr->be16));
 907     taddr->be16[5] = OVS_BE16_MAX;
 908     put_16aligned_be32(&taddr->be32[3], ip4);
 909 }
 910
 911 static inline ovs_be32
 912 in6_addr_get_mapped_ipv4(const struct in6_addr *addr)
 913 {
 914     union ovs_16aligned_in6_addr *taddr = (void *) addr;
 915     if (IN6_IS_ADDR_V4MAPPED(addr)) {
 916         return get_16aligned_be32(&taddr->be32[3]);
 917     } else {
 918         return INADDR_ANY;
 919     }
 920 }
 921
 922 static inline bool dl_type_is_ip_any(ovs_be16 dl_type)
 923 {
 924     return dl_type == htons(ETH_TYPE_IP)
 925         || dl_type == htons(ETH_TYPE_IPV6);
 926 }
 927
 928 /* Tunnel header */
 929
 930 /* GRE protocol header */
 931 struct gre_base_hdr {
 932     ovs_be16 flags;
 933     ovs_be16 protocol;
 934 };
 935
 936 #define GRE_CSUM        0x8000
 937 #define GRE_ROUTING     0x4000
 938 #define GRE_KEY         0x2000
 939 #define GRE_SEQ         0x1000
 940 #define GRE_STRICT      0x0800
 941 #define GRE_REC         0x0700
 942 #define GRE_FLAGS       0x00F8
 943 #define GRE_VERSION     0x0007
 944
 945 /* VXLAN protocol header */
 946 struct vxlanhdr {
 947     ovs_16aligned_be32 vx_flags;
 948     ovs_16aligned_be32 vx_vni;
 949 };
 950
 951 #define VXLAN_FLAGS 0x08000000  /* struct vxlanhdr.vx_flags required value. */
 952
 953 void format_ipv6_addr(char *addr_str, const struct in6_addr *addr);
 954 void print_ipv6_addr(struct ds *string, const struct in6_addr *addr);
 955 void print_ipv6_mapped(struct ds *string, const struct in6_addr *addr);
 956 void print_ipv6_masked(struct ds *string, const struct in6_addr *addr,
 957                        const struct in6_addr *mask);
 958 struct in6_addr ipv6_addr_bitand(const struct in6_addr *src,
 959                                  const struct in6_addr *mask);
 960 struct in6_addr ipv6_create_mask(int mask);
 961 int ipv6_count_cidr_bits(const struct in6_addr *netmask);
 962 bool ipv6_is_cidr(const struct in6_addr *netmask);
 963
 964 void *eth_compose(struct dp_packet *, const struct eth_addr eth_dst,
 965                   const struct eth_addr eth_src, uint16_t eth_type,
 966                   size_t size);
 967 void *snap_compose(struct dp_packet *, const struct eth_addr eth_dst,
 968                    const struct eth_addr eth_src,
 969                    unsigned int oui, uint16_t snap_type, size_t size);
 970 void packet_set_ipv4(struct dp_packet *, ovs_be32 src, ovs_be32 dst, uint8_t tos,
 971                      uint8_t ttl);
 972 void packet_set_ipv6(struct dp_packet *, uint8_t proto, const ovs_be32 src[4],
 973                      const ovs_be32 dst[4], uint8_t tc,
 974                      ovs_be32 fl, uint8_t hlmit);
 975 void packet_set_tcp_port(struct dp_packet *, ovs_be16 src, ovs_be16 dst);
 976 void packet_set_udp_port(struct dp_packet *, ovs_be16 src, ovs_be16 dst);
 977 void packet_set_sctp_port(struct dp_packet *, ovs_be16 src, ovs_be16 dst);
 978 void packet_set_icmp(struct dp_packet *, uint8_t type, uint8_t code);
 979 void packet_set_nd(struct dp_packet *, const ovs_be32 target[4],
 980                    const struct eth_addr sll, const struct eth_addr tll);
 981
 982 void packet_format_tcp_flags(struct ds *, uint16_t);
 983 const char *packet_tcp_flag_to_string(uint32_t flag);
 984 void compose_arp(struct dp_packet *, uint16_t arp_op,
 985                  const struct eth_addr arp_sha,
 986                  const struct eth_addr arp_tha, bool broadcast,
 987                  ovs_be32 arp_spa, ovs_be32 arp_tpa);
 988 uint32_t packet_csum_pseudoheader(const struct ip_header *);
 989
 990 #endif /* packets.h */