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     struct in6_addr ipv6_dst;
  42     ovs_be32 ip_src;
  43     struct in6_addr ipv6_src;
  44     ovs_be64 tun_id;
  45     uint16_t flags;
  46     uint8_t ip_tos;
  47     uint8_t ip_ttl;
  48     ovs_be16 tp_src;
  49     ovs_be16 tp_dst;
  50     ovs_be16 gbp_id;
  51     uint8_t  gbp_flags;
  52     uint8_t  pad1[5];        /* Pad to 64 bits. */
  53     struct tun_metadata metadata;
  54 };
  55
  56 /* Some flags are exposed through OpenFlow while others are used only
  57  * internally. */
  58
  59 /* Public flags */
  60 #define FLOW_TNL_F_OAM (1 << 0)
  61
  62 #define FLOW_TNL_PUB_F_MASK ((1 << 1) - 1)
  63
  64 /* Private flags */
  65 #define FLOW_TNL_F_DONT_FRAGMENT (1 << 1)
  66 #define FLOW_TNL_F_CSUM (1 << 2)
  67 #define FLOW_TNL_F_KEY (1 << 3)
  68
  69 #define FLOW_TNL_F_MASK ((1 << 4) - 1)
  70
  71 /* Purely internal to OVS userspace. These flags should never be exposed to
  72  * the outside world and so aren't included in the flags mask. */
  73
  74 /* Tunnel information is in userspace datapath format. */
  75 #define FLOW_TNL_F_UDPIF (1 << 4)
  76
  77 static inline bool ipv6_addr_is_set(const struct in6_addr *addr);
  78
  79 static inline bool
  80 flow_tnl_dst_is_set(const struct flow_tnl *tnl)
  81 {
  82     return tnl->ip_dst || ipv6_addr_is_set(&tnl->ipv6_dst);
  83 }
  84
  85 struct in6_addr flow_tnl_dst(const struct flow_tnl *tnl);
  86 struct in6_addr flow_tnl_src(const struct flow_tnl *tnl);
  87
  88 /* Returns an offset to 'src' covering all the meaningful fields in 'src'. */
  89 static inline size_t
  90 flow_tnl_size(const struct flow_tnl *src)
  91 {
  92     if (!flow_tnl_dst_is_set(src)) {
  93         /* Covers ip_dst and ipv6_dst only. */
  94         return offsetof(struct flow_tnl, ip_src);
  95     }
  96     if (src->flags & FLOW_TNL_F_UDPIF) {
  97         /* Datapath format, cover all options we have. */
  98         return offsetof(struct flow_tnl, metadata.opts)
  99             + src->metadata.present.len;
 100     }
 101     if (!src->metadata.present.map) {
 102         /* No TLVs, opts is irrelevant. */
 103         return offsetof(struct flow_tnl, metadata.opts);
 104     }
 105     /* Have decoded TLVs, opts is relevant. */
 106     return sizeof *src;
 107 }
 108
 109 /* Copy flow_tnl, but avoid copying unused portions of tun_metadata.  Unused
 110  * data in 'dst' is NOT cleared, so this must not be used in cases where the
 111  * uninitialized portion may be hashed over. */
 112 static inline void
 113 flow_tnl_copy__(struct flow_tnl *dst, const struct flow_tnl *src)
 114 {
 115     memcpy(dst, src, flow_tnl_size(src));
 116 }
 117
 118 static inline bool
 119 flow_tnl_equal(const struct flow_tnl *a, const struct flow_tnl *b)
 120 {
 121     size_t a_size = flow_tnl_size(a);
 122
 123     return a_size == flow_tnl_size(b) && !memcmp(a, b, a_size);
 124 }
 125
 126 /* Unfortunately, a "struct flow" sometimes has to handle OpenFlow port
 127  * numbers and other times datapath (dpif) port numbers.  This union allows
 128  * access to both. */
 129 union flow_in_port {
 130     odp_port_t odp_port;
 131     ofp_port_t ofp_port;
 132 };
 133
 134 /* Datapath packet metadata */
 135 struct pkt_metadata {
 136     uint32_t recirc_id;         /* Recirculation id carried with the
 137                                    recirculating packets. 0 for packets
 138                                    received from the wire. */
 139     uint32_t dp_hash;           /* hash value computed by the recirculation
 140                                    action. */
 141     uint32_t skb_priority;      /* Packet priority for QoS. */
 142     uint32_t pkt_mark;          /* Packet mark. */
 143     uint16_t ct_state;          /* Connection state. */
 144     uint16_t ct_zone;           /* Connection zone. */
 145     uint32_t ct_mark;           /* Connection mark. */
 146     ovs_u128 ct_label;          /* Connection label. */
 147     union flow_in_port in_port; /* Input port. */
 148     struct flow_tnl tunnel;     /* Encapsulating tunnel parameters. Note that
 149                                  * if 'ip_dst' == 0, the rest of the fields may
 150                                  * be uninitialized. */
 151 };
 152
 153 static inline void
 154 pkt_metadata_init(struct pkt_metadata *md, odp_port_t port)
 155 {
 156     /* It can be expensive to zero out all of the tunnel metadata. However,
 157      * we can just zero out ip_dst and the rest of the data will never be
 158      * looked at. */
 159     memset(md, 0, offsetof(struct pkt_metadata, in_port));
 160     md->tunnel.ip_dst = 0;
 161     md->tunnel.ipv6_dst = in6addr_any;
 162
 163     md->in_port.odp_port = port;
 164 }
 165
 166 bool dpid_from_string(const char *s, uint64_t *dpidp);
 167
 168 #define ETH_ADDR_LEN           6
 169
 170 static const struct eth_addr eth_addr_broadcast OVS_UNUSED
 171     = { { { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff } } };
 172
 173 static const struct eth_addr eth_addr_exact OVS_UNUSED
 174     = { { { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff } } };
 175
 176 static const struct eth_addr eth_addr_zero OVS_UNUSED
 177     = { { { 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 } } };
 178
 179 static const struct eth_addr eth_addr_stp OVS_UNUSED
 180     = { { { 0x01, 0x80, 0xC2, 0x00, 0x00, 0x00 } } };
 181
 182 static const struct eth_addr eth_addr_lacp OVS_UNUSED
 183     = { { { 0x01, 0x80, 0xC2, 0x00, 0x00, 0x02 } } };
 184
 185 static const struct eth_addr eth_addr_bfd OVS_UNUSED
 186     = { { { 0x00, 0x23, 0x20, 0x00, 0x00, 0x01 } } };
 187
 188 static inline bool eth_addr_is_broadcast(const struct eth_addr a)
 189 {
 190     return (a.be16[0] & a.be16[1] & a.be16[2]) == htons(0xffff);
 191 }
 192
 193 static inline bool eth_addr_is_multicast(const struct eth_addr a)
 194 {
 195     return a.ea[0] & 1;
 196 }
 197
 198 static inline bool eth_addr_is_local(const struct eth_addr a)
 199 {
 200     /* Local if it is either a locally administered address or a Nicira random
 201      * address. */
 202     return a.ea[0] & 2
 203         || (a.be16[0] == htons(0x0023)
 204             && (a.be16[1] & htons(0xff80)) == htons(0x2080));
 205 }
 206 static inline bool eth_addr_is_zero(const struct eth_addr a)
 207 {
 208     return !(a.be16[0] | a.be16[1] | a.be16[2]);
 209 }
 210
 211 static inline int eth_mask_is_exact(const struct eth_addr a)
 212 {
 213     return (a.be16[0] & a.be16[1] & a.be16[2]) == htons(0xffff);
 214 }
 215
 216 static inline int eth_addr_compare_3way(const struct eth_addr a,
 217                                         const struct eth_addr b)
 218 {
 219     return memcmp(&a, &b, sizeof a);
 220 }
 221
 222 static inline bool eth_addr_equals(const struct eth_addr a,
 223                                    const struct eth_addr b)
 224 {
 225     return !eth_addr_compare_3way(a, b);
 226 }
 227
 228 static inline bool eth_addr_equal_except(const struct eth_addr a,
 229                                          const struct eth_addr b,
 230                                          const struct eth_addr mask)
 231 {
 232     return !(((a.be16[0] ^ b.be16[0]) & mask.be16[0])
 233              || ((a.be16[1] ^ b.be16[1]) & mask.be16[1])
 234              || ((a.be16[2] ^ b.be16[2]) & mask.be16[2]));
 235 }
 236
 237 static inline uint64_t eth_addr_to_uint64(const struct eth_addr ea)
 238 {
 239     return (((uint64_t) ntohs(ea.be16[0]) << 32)
 240             | ((uint64_t) ntohs(ea.be16[1]) << 16)
 241             | ntohs(ea.be16[2]));
 242 }
 243
 244 static inline uint64_t eth_addr_vlan_to_uint64(const struct eth_addr ea,
 245                                                uint16_t vlan)
 246 {
 247     return (((uint64_t)vlan << 48) | eth_addr_to_uint64(ea));
 248 }
 249
 250 static inline void eth_addr_from_uint64(uint64_t x, struct eth_addr *ea)
 251 {
 252     ea->be16[0] = htons(x >> 32);
 253     ea->be16[1] = htons((x & 0xFFFF0000) >> 16);
 254     ea->be16[2] = htons(x & 0xFFFF);
 255 }
 256
 257 static inline struct eth_addr eth_addr_invert(const struct eth_addr src)
 258 {
 259     struct eth_addr dst;
 260
 261     for (int i = 0; i < ARRAY_SIZE(src.be16); i++) {
 262         dst.be16[i] = ~src.be16[i];
 263     }
 264
 265     return dst;
 266 }
 267
 268 static inline void eth_addr_mark_random(struct eth_addr *ea)
 269 {
 270     ea->ea[0] &= ~1;                /* Unicast. */
 271     ea->ea[0] |= 2;                 /* Private. */
 272 }
 273
 274 static inline void eth_addr_random(struct eth_addr *ea)
 275 {
 276     random_bytes((uint8_t *)ea, sizeof *ea);
 277     eth_addr_mark_random(ea);
 278 }
 279
 280 static inline void eth_addr_nicira_random(struct eth_addr *ea)
 281 {
 282     eth_addr_random(ea);
 283
 284     /* Set the OUI to the Nicira one. */
 285     ea->ea[0] = 0x00;
 286     ea->ea[1] = 0x23;
 287     ea->ea[2] = 0x20;
 288
 289     /* Set the top bit to indicate random Nicira address. */
 290     ea->ea[3] |= 0x80;
 291 }
 292 static inline uint32_t hash_mac(const struct eth_addr ea,
 293                                 const uint16_t vlan, const uint32_t basis)
 294 {
 295     return hash_uint64_basis(eth_addr_vlan_to_uint64(ea, vlan), basis);
 296 }
 297
 298 bool eth_addr_is_reserved(const struct eth_addr);
 299 bool eth_addr_from_string(const char *, struct eth_addr *);
 300
 301 void compose_rarp(struct dp_packet *, const struct eth_addr);
 302
 303 void eth_push_vlan(struct dp_packet *, ovs_be16 tpid, ovs_be16 tci);
 304 void eth_pop_vlan(struct dp_packet *);
 305
 306 const char *eth_from_hex(const char *hex, struct dp_packet **packetp);
 307 void eth_format_masked(const struct eth_addr ea,
 308                        const struct eth_addr *mask, struct ds *s);
 309
 310 void set_mpls_lse(struct dp_packet *, ovs_be32 label);
 311 void push_mpls(struct dp_packet *packet, ovs_be16 ethtype, ovs_be32 lse);
 312 void pop_mpls(struct dp_packet *, ovs_be16 ethtype);
 313
 314 void set_mpls_lse_ttl(ovs_be32 *lse, uint8_t ttl);
 315 void set_mpls_lse_tc(ovs_be32 *lse, uint8_t tc);
 316 void set_mpls_lse_label(ovs_be32 *lse, ovs_be32 label);
 317 void set_mpls_lse_bos(ovs_be32 *lse, uint8_t bos);
 318 ovs_be32 set_mpls_lse_values(uint8_t ttl, uint8_t tc, uint8_t bos,
 319                              ovs_be32 label);
 320
 321 /* Example:
 322  *
 323  * struct eth_addr mac;
 324  *    [...]
 325  * printf("The Ethernet address is "ETH_ADDR_FMT"\n", ETH_ADDR_ARGS(mac));
 326  *
 327  */
 328 #define ETH_ADDR_FMT                                                    \
 329     "%02"PRIx8":%02"PRIx8":%02"PRIx8":%02"PRIx8":%02"PRIx8":%02"PRIx8
 330 #define ETH_ADDR_ARGS(EA) ETH_ADDR_BYTES_ARGS((EA).ea)
 331 #define ETH_ADDR_BYTES_ARGS(EAB) \
 332          (EAB)[0], (EAB)[1], (EAB)[2], (EAB)[3], (EAB)[4], (EAB)[5]
 333
 334 /* Example:
 335  *
 336  * char *string = "1 00:11:22:33:44:55 2";
 337  * struct eth_addr mac;
 338  * int a, b;
 339  *
 340  * if (ovs_scan(string, "%d"ETH_ADDR_SCAN_FMT"%d",
 341  *              &a, ETH_ADDR_SCAN_ARGS(mac), &b)) {
 342  *     ...
 343  * }
 344  */
 345 #define ETH_ADDR_SCAN_FMT "%"SCNx8":%"SCNx8":%"SCNx8":%"SCNx8":%"SCNx8":%"SCNx8
 346 #define ETH_ADDR_SCAN_ARGS(EA) \
 347     &(EA).ea[0], &(EA).ea[1], &(EA).ea[2], &(EA).ea[3], &(EA).ea[4], &(EA).ea[5]
 348
 349 #define ETH_TYPE_IP            0x0800
 350 #define ETH_TYPE_ARP           0x0806
 351 #define ETH_TYPE_TEB           0x6558
 352 #define ETH_TYPE_VLAN_8021Q    0x8100
 353 #define ETH_TYPE_VLAN          ETH_TYPE_VLAN_8021Q
 354 #define ETH_TYPE_VLAN_8021AD   0x88a8
 355 #define ETH_TYPE_IPV6          0x86dd
 356 #define ETH_TYPE_LACP          0x8809
 357 #define ETH_TYPE_RARP          0x8035
 358 #define ETH_TYPE_MPLS          0x8847
 359 #define ETH_TYPE_MPLS_MCAST    0x8848
 360
 361 static inline bool eth_type_mpls(ovs_be16 eth_type)
 362 {
 363     return eth_type == htons(ETH_TYPE_MPLS) ||
 364         eth_type == htons(ETH_TYPE_MPLS_MCAST);
 365 }
 366
 367 static inline bool eth_type_vlan(ovs_be16 eth_type)
 368 {
 369     return eth_type == htons(ETH_TYPE_VLAN_8021Q) ||
 370         eth_type == htons(ETH_TYPE_VLAN_8021AD);
 371 }
 372
 373
 374 /* Minimum value for an Ethernet type.  Values below this are IEEE 802.2 frame
 375  * lengths. */
 376 #define ETH_TYPE_MIN           0x600
 377
 378 #define ETH_HEADER_LEN 14
 379 #define ETH_PAYLOAD_MIN 46
 380 #define ETH_PAYLOAD_MAX 1500
 381 #define ETH_TOTAL_MIN (ETH_HEADER_LEN + ETH_PAYLOAD_MIN)
 382 #define ETH_TOTAL_MAX (ETH_HEADER_LEN + ETH_PAYLOAD_MAX)
 383 #define ETH_VLAN_TOTAL_MAX (ETH_HEADER_LEN + VLAN_HEADER_LEN + ETH_PAYLOAD_MAX)
 384 OVS_PACKED(
 385 struct eth_header {
 386     struct eth_addr eth_dst;
 387     struct eth_addr eth_src;
 388     ovs_be16 eth_type;
 389 });
 390 BUILD_ASSERT_DECL(ETH_HEADER_LEN == sizeof(struct eth_header));
 391
 392 #define LLC_DSAP_SNAP 0xaa
 393 #define LLC_SSAP_SNAP 0xaa
 394 #define LLC_CNTL_SNAP 3
 395
 396 #define LLC_HEADER_LEN 3
 397 OVS_PACKED(
 398 struct llc_header {
 399     uint8_t llc_dsap;
 400     uint8_t llc_ssap;
 401     uint8_t llc_cntl;
 402 });
 403 BUILD_ASSERT_DECL(LLC_HEADER_LEN == sizeof(struct llc_header));
 404
 405 /* LLC field values used for STP frames. */
 406 #define STP_LLC_SSAP 0x42
 407 #define STP_LLC_DSAP 0x42
 408 #define STP_LLC_CNTL 0x03
 409
 410 #define SNAP_ORG_ETHERNET "\0\0" /* The compiler adds a null byte, so
 411                                     sizeof(SNAP_ORG_ETHERNET) == 3. */
 412 #define SNAP_HEADER_LEN 5
 413 OVS_PACKED(
 414 struct snap_header {
 415     uint8_t snap_org[3];
 416     ovs_be16 snap_type;
 417 });
 418 BUILD_ASSERT_DECL(SNAP_HEADER_LEN == sizeof(struct snap_header));
 419
 420 #define LLC_SNAP_HEADER_LEN (LLC_HEADER_LEN + SNAP_HEADER_LEN)
 421 OVS_PACKED(
 422 struct llc_snap_header {
 423     struct llc_header llc;
 424     struct snap_header snap;
 425 });
 426 BUILD_ASSERT_DECL(LLC_SNAP_HEADER_LEN == sizeof(struct llc_snap_header));
 427
 428 #define VLAN_VID_MASK 0x0fff
 429 #define VLAN_VID_SHIFT 0
 430
 431 #define VLAN_PCP_MASK 0xe000
 432 #define VLAN_PCP_SHIFT 13
 433
 434 #define VLAN_CFI 0x1000
 435 #define VLAN_CFI_SHIFT 12
 436
 437 /* Given the vlan_tci field from an 802.1Q header, in network byte order,
 438  * returns the VLAN ID in host byte order. */
 439 static inline uint16_t
 440 vlan_tci_to_vid(ovs_be16 vlan_tci)
 441 {
 442     return (ntohs(vlan_tci) & VLAN_VID_MASK) >> VLAN_VID_SHIFT;
 443 }
 444
 445 /* Given the vlan_tci field from an 802.1Q header, in network byte order,
 446  * returns the priority code point (PCP) in host byte order. */
 447 static inline int
 448 vlan_tci_to_pcp(ovs_be16 vlan_tci)
 449 {
 450     return (ntohs(vlan_tci) & VLAN_PCP_MASK) >> VLAN_PCP_SHIFT;
 451 }
 452
 453 /* Given the vlan_tci field from an 802.1Q header, in network byte order,
 454  * returns the Canonical Format Indicator (CFI). */
 455 static inline int
 456 vlan_tci_to_cfi(ovs_be16 vlan_tci)
 457 {
 458     return (vlan_tci & htons(VLAN_CFI)) != 0;
 459 }
 460
 461 #define VLAN_HEADER_LEN 4
 462 struct vlan_header {
 463     ovs_be16 vlan_tci;          /* Lowest 12 bits are VLAN ID. */
 464     ovs_be16 vlan_next_type;
 465 };
 466 BUILD_ASSERT_DECL(VLAN_HEADER_LEN == sizeof(struct vlan_header));
 467
 468 #define VLAN_ETH_HEADER_LEN (ETH_HEADER_LEN + VLAN_HEADER_LEN)
 469 OVS_PACKED(
 470 struct vlan_eth_header {
 471     struct eth_addr veth_dst;
 472     struct eth_addr veth_src;
 473     ovs_be16 veth_type;         /* Always htons(ETH_TYPE_VLAN). */
 474     ovs_be16 veth_tci;          /* Lowest 12 bits are VLAN ID. */
 475     ovs_be16 veth_next_type;
 476 });
 477 BUILD_ASSERT_DECL(VLAN_ETH_HEADER_LEN == sizeof(struct vlan_eth_header));
 478
 479 /* MPLS related definitions */
 480 #define MPLS_TTL_MASK       0x000000ff
 481 #define MPLS_TTL_SHIFT      0
 482
 483 #define MPLS_BOS_MASK       0x00000100
 484 #define MPLS_BOS_SHIFT      8
 485
 486 #define MPLS_TC_MASK        0x00000e00
 487 #define MPLS_TC_SHIFT       9
 488
 489 #define MPLS_LABEL_MASK     0xfffff000
 490 #define MPLS_LABEL_SHIFT    12
 491
 492 #define MPLS_HLEN           4
 493
 494 struct mpls_hdr {
 495     ovs_16aligned_be32 mpls_lse;
 496 };
 497 BUILD_ASSERT_DECL(MPLS_HLEN == sizeof(struct mpls_hdr));
 498
 499 /* Given a mpls label stack entry in network byte order
 500  * return mpls label in host byte order */
 501 static inline uint32_t
 502 mpls_lse_to_label(ovs_be32 mpls_lse)
 503 {
 504     return (ntohl(mpls_lse) & MPLS_LABEL_MASK) >> MPLS_LABEL_SHIFT;
 505 }
 506
 507 /* Given a mpls label stack entry in network byte order
 508  * return mpls tc */
 509 static inline uint8_t
 510 mpls_lse_to_tc(ovs_be32 mpls_lse)
 511 {
 512     return (ntohl(mpls_lse) & MPLS_TC_MASK) >> MPLS_TC_SHIFT;
 513 }
 514
 515 /* Given a mpls label stack entry in network byte order
 516  * return mpls ttl */
 517 static inline uint8_t
 518 mpls_lse_to_ttl(ovs_be32 mpls_lse)
 519 {
 520     return (ntohl(mpls_lse) & MPLS_TTL_MASK) >> MPLS_TTL_SHIFT;
 521 }
 522
 523 /* Set TTL in mpls lse. */
 524 static inline void
 525 flow_set_mpls_lse_ttl(ovs_be32 *mpls_lse, uint8_t ttl)
 526 {
 527     *mpls_lse &= ~htonl(MPLS_TTL_MASK);
 528     *mpls_lse |= htonl(ttl << MPLS_TTL_SHIFT);
 529 }
 530
 531 /* Given a mpls label stack entry in network byte order
 532  * return mpls BoS bit  */
 533 static inline uint8_t
 534 mpls_lse_to_bos(ovs_be32 mpls_lse)
 535 {
 536     return (mpls_lse & htonl(MPLS_BOS_MASK)) != 0;
 537 }
 538
 539 #define IP_FMT "%"PRIu32".%"PRIu32".%"PRIu32".%"PRIu32
 540 #define IP_ARGS(ip)                             \
 541     ntohl(ip) >> 24,                            \
 542     (ntohl(ip) >> 16) & 0xff,                   \
 543     (ntohl(ip) >> 8) & 0xff,                    \
 544     ntohl(ip) & 0xff
 545
 546 /* Example:
 547  *
 548  * char *string = "1 33.44.55.66 2";
 549  * ovs_be32 ip;
 550  * int a, b;
 551  *
 552  * if (ovs_scan(string, "%d"IP_SCAN_FMT"%d", &a, IP_SCAN_ARGS(&ip), &b)) {
 553  *     ...
 554  * }
 555  */
 556 #define IP_SCAN_FMT "%"SCNu8".%"SCNu8".%"SCNu8".%"SCNu8
 557 #define IP_SCAN_ARGS(ip)                                    \
 558         ((void) (ovs_be32) *(ip), &((uint8_t *) ip)[0]),    \
 559         &((uint8_t *) ip)[1],                               \
 560         &((uint8_t *) ip)[2],                               \
 561         &((uint8_t *) ip)[3]
 562
 563 /* Returns true if 'netmask' is a CIDR netmask, that is, if it consists of N
 564  * high-order 1-bits and 32-N low-order 0-bits. */
 565 static inline bool
 566 ip_is_cidr(ovs_be32 netmask)
 567 {
 568     uint32_t x = ~ntohl(netmask);
 569     return !(x & (x + 1));
 570 }
 571 static inline bool
 572 ip_is_multicast(ovs_be32 ip)
 573 {
 574     return (ip & htonl(0xf0000000)) == htonl(0xe0000000);
 575 }
 576 static inline bool
 577 ip_is_local_multicast(ovs_be32 ip)
 578 {
 579     return (ip & htonl(0xffffff00)) == htonl(0xe0000000);
 580 }
 581 int ip_count_cidr_bits(ovs_be32 netmask);
 582 void ip_format_masked(ovs_be32 ip, ovs_be32 mask, struct ds *);
 583 char *ip_parse_masked(const char *s, ovs_be32 *ip, ovs_be32 *mask)
 584     OVS_WARN_UNUSED_RESULT;
 585
 586 #define IP_VER(ip_ihl_ver) ((ip_ihl_ver) >> 4)
 587 #define IP_IHL(ip_ihl_ver) ((ip_ihl_ver) & 15)
 588 #define IP_IHL_VER(ihl, ver) (((ver) << 4) | (ihl))
 589
 590 #ifndef IPPROTO_SCTP
 591 #define IPPROTO_SCTP 132
 592 #endif
 593
 594 /* TOS fields. */
 595 #define IP_ECN_NOT_ECT 0x0
 596 #define IP_ECN_ECT_1 0x01
 597 #define IP_ECN_ECT_0 0x02
 598 #define IP_ECN_CE 0x03
 599 #define IP_ECN_MASK 0x03
 600 #define IP_DSCP_MASK 0xfc
 601
 602 #define IP_VERSION 4
 603
 604 #define IP_DONT_FRAGMENT  0x4000 /* Don't fragment. */
 605 #define IP_MORE_FRAGMENTS 0x2000 /* More fragments. */
 606 #define IP_FRAG_OFF_MASK  0x1fff /* Fragment offset. */
 607 #define IP_IS_FRAGMENT(ip_frag_off) \
 608         ((ip_frag_off) & htons(IP_MORE_FRAGMENTS | IP_FRAG_OFF_MASK))
 609
 610 #define IP_HEADER_LEN 20
 611 struct ip_header {
 612     uint8_t ip_ihl_ver;
 613     uint8_t ip_tos;
 614     ovs_be16 ip_tot_len;
 615     ovs_be16 ip_id;
 616     ovs_be16 ip_frag_off;
 617     uint8_t ip_ttl;
 618     uint8_t ip_proto;
 619     ovs_be16 ip_csum;
 620     ovs_16aligned_be32 ip_src;
 621     ovs_16aligned_be32 ip_dst;
 622 };
 623
 624 BUILD_ASSERT_DECL(IP_HEADER_LEN == sizeof(struct ip_header));
 625
 626 #define ICMP_HEADER_LEN 8
 627 struct icmp_header {
 628     uint8_t icmp_type;
 629     uint8_t icmp_code;
 630     ovs_be16 icmp_csum;
 631     union {
 632         struct {
 633             ovs_be16 id;
 634             ovs_be16 seq;
 635         } echo;
 636         struct {
 637             ovs_be16 empty;
 638             ovs_be16 mtu;
 639         } frag;
 640         ovs_16aligned_be32 gateway;
 641     } icmp_fields;
 642 };
 643 BUILD_ASSERT_DECL(ICMP_HEADER_LEN == sizeof(struct icmp_header));
 644
 645 #define IGMP_HEADER_LEN 8
 646 struct igmp_header {
 647     uint8_t igmp_type;
 648     uint8_t igmp_code;
 649     ovs_be16 igmp_csum;
 650     ovs_16aligned_be32 group;
 651 };
 652 BUILD_ASSERT_DECL(IGMP_HEADER_LEN == sizeof(struct igmp_header));
 653
 654 #define IGMPV3_HEADER_LEN 8
 655 struct igmpv3_header {
 656     uint8_t type;
 657     uint8_t rsvr1;
 658     ovs_be16 csum;
 659     ovs_be16 rsvr2;
 660     ovs_be16 ngrp;
 661 };
 662 BUILD_ASSERT_DECL(IGMPV3_HEADER_LEN == sizeof(struct igmpv3_header));
 663
 664 #define IGMPV3_RECORD_LEN 8
 665 struct igmpv3_record {
 666     uint8_t type;
 667     uint8_t aux_len;
 668     ovs_be16 nsrcs;
 669     ovs_16aligned_be32 maddr;
 670 };
 671 BUILD_ASSERT_DECL(IGMPV3_RECORD_LEN == sizeof(struct igmpv3_record));
 672
 673 #define IGMP_HOST_MEMBERSHIP_QUERY    0x11 /* From RFC1112 */
 674 #define IGMP_HOST_MEMBERSHIP_REPORT   0x12 /* Ditto */
 675 #define IGMPV2_HOST_MEMBERSHIP_REPORT 0x16 /* V2 version of 0x12 */
 676 #define IGMP_HOST_LEAVE_MESSAGE       0x17
 677 #define IGMPV3_HOST_MEMBERSHIP_REPORT 0x22 /* V3 version of 0x12 */
 678
 679 /*
 680  * IGMPv3 and MLDv2 use the same codes.
 681  */
 682 #define IGMPV3_MODE_IS_INCLUDE 1
 683 #define IGMPV3_MODE_IS_EXCLUDE 2
 684 #define IGMPV3_CHANGE_TO_INCLUDE_MODE 3
 685 #define IGMPV3_CHANGE_TO_EXCLUDE_MODE 4
 686 #define IGMPV3_ALLOW_NEW_SOURCES 5
 687 #define IGMPV3_BLOCK_OLD_SOURCES 6
 688
 689 #define SCTP_HEADER_LEN 12
 690 struct sctp_header {
 691     ovs_be16 sctp_src;
 692     ovs_be16 sctp_dst;
 693     ovs_16aligned_be32 sctp_vtag;
 694     ovs_16aligned_be32 sctp_csum;
 695 };
 696 BUILD_ASSERT_DECL(SCTP_HEADER_LEN == sizeof(struct sctp_header));
 697
 698 #define UDP_HEADER_LEN 8
 699 struct udp_header {
 700     ovs_be16 udp_src;
 701     ovs_be16 udp_dst;
 702     ovs_be16 udp_len;
 703     ovs_be16 udp_csum;
 704 };
 705 BUILD_ASSERT_DECL(UDP_HEADER_LEN == sizeof(struct udp_header));
 706
 707 #define TCP_FIN 0x001
 708 #define TCP_SYN 0x002
 709 #define TCP_RST 0x004
 710 #define TCP_PSH 0x008
 711 #define TCP_ACK 0x010
 712 #define TCP_URG 0x020
 713 #define TCP_ECE 0x040
 714 #define TCP_CWR 0x080
 715 #define TCP_NS  0x100
 716
 717 #define TCP_CTL(flags, offset) (htons((flags) | ((offset) << 12)))
 718 #define TCP_FLAGS(tcp_ctl) (ntohs(tcp_ctl) & 0x0fff)
 719 #define TCP_FLAGS_BE16(tcp_ctl) ((tcp_ctl) & htons(0x0fff))
 720 #define TCP_OFFSET(tcp_ctl) (ntohs(tcp_ctl) >> 12)
 721
 722 #define TCP_HEADER_LEN 20
 723 struct tcp_header {
 724     ovs_be16 tcp_src;
 725     ovs_be16 tcp_dst;
 726     ovs_16aligned_be32 tcp_seq;
 727     ovs_16aligned_be32 tcp_ack;
 728     ovs_be16 tcp_ctl;
 729     ovs_be16 tcp_winsz;
 730     ovs_be16 tcp_csum;
 731     ovs_be16 tcp_urg;
 732 };
 733 BUILD_ASSERT_DECL(TCP_HEADER_LEN == sizeof(struct tcp_header));
 734
 735 /* Connection states */
 736 enum {
 737     CS_NEW_BIT =         0,
 738     CS_ESTABLISHED_BIT = 1,
 739     CS_RELATED_BIT =     2,
 740     CS_REPLY_DIR_BIT =   3,
 741     CS_INVALID_BIT =     4,
 742     CS_TRACKED_BIT =     5,
 743     CS_SRC_NAT_BIT =     6,
 744     CS_DST_NAT_BIT =     7,
 745 };
 746
 747 enum {
 748     CS_NEW =         (1 << CS_NEW_BIT),
 749     CS_ESTABLISHED = (1 << CS_ESTABLISHED_BIT),
 750     CS_RELATED =     (1 << CS_RELATED_BIT),
 751     CS_REPLY_DIR =   (1 << CS_REPLY_DIR_BIT),
 752     CS_INVALID =     (1 << CS_INVALID_BIT),
 753     CS_TRACKED =     (1 << CS_TRACKED_BIT),
 754     CS_SRC_NAT =     (1 << CS_SRC_NAT_BIT),
 755     CS_DST_NAT =     (1 << CS_DST_NAT_BIT),
 756 };
 757
 758 /* Undefined connection state bits. */
 759 #define CS_SUPPORTED_MASK    (CS_NEW | CS_ESTABLISHED | CS_RELATED \
 760                               | CS_INVALID | CS_REPLY_DIR | CS_TRACKED \
 761                               | CS_SRC_NAT | CS_DST_NAT)
 762 #define CS_UNSUPPORTED_MASK  (~(uint32_t)CS_SUPPORTED_MASK)
 763
 764 #define ARP_HRD_ETHERNET 1
 765 #define ARP_PRO_IP 0x0800
 766 #define ARP_OP_REQUEST 1
 767 #define ARP_OP_REPLY 2
 768 #define ARP_OP_RARP 3
 769
 770 #define ARP_ETH_HEADER_LEN 28
 771 struct arp_eth_header {
 772     /* Generic members. */
 773     ovs_be16 ar_hrd;           /* Hardware type. */
 774     ovs_be16 ar_pro;           /* Protocol type. */
 775     uint8_t ar_hln;            /* Hardware address length. */
 776     uint8_t ar_pln;            /* Protocol address length. */
 777     ovs_be16 ar_op;            /* Opcode. */
 778
 779     /* Ethernet+IPv4 specific members. */
 780     struct eth_addr ar_sha;     /* Sender hardware address. */
 781     ovs_16aligned_be32 ar_spa;  /* Sender protocol address. */
 782     struct eth_addr ar_tha;     /* Target hardware address. */
 783     ovs_16aligned_be32 ar_tpa;  /* Target protocol address. */
 784 };
 785 BUILD_ASSERT_DECL(ARP_ETH_HEADER_LEN == sizeof(struct arp_eth_header));
 786
 787 #define IPV6_HEADER_LEN 40
 788
 789 /* Like struct in6_addr, but whereas that struct requires 32-bit alignment on
 790  * most implementations, this one only requires 16-bit alignment. */
 791 union ovs_16aligned_in6_addr {
 792     ovs_be16 be16[8];
 793     ovs_16aligned_be32 be32[4];
 794 };
 795
 796 /* Like struct in6_hdr, but whereas that struct requires 32-bit alignment, this
 797  * one only requires 16-bit alignment. */
 798 struct ovs_16aligned_ip6_hdr {
 799     union {
 800         struct ovs_16aligned_ip6_hdrctl {
 801             ovs_16aligned_be32 ip6_un1_flow;
 802             ovs_be16 ip6_un1_plen;
 803             uint8_t ip6_un1_nxt;
 804             uint8_t ip6_un1_hlim;
 805         } ip6_un1;
 806         uint8_t ip6_un2_vfc;
 807     } ip6_ctlun;
 808     union ovs_16aligned_in6_addr ip6_src;
 809     union ovs_16aligned_in6_addr ip6_dst;
 810 };
 811
 812 /* Like struct in6_frag, but whereas that struct requires 32-bit alignment,
 813  * this one only requires 16-bit alignment. */
 814 struct ovs_16aligned_ip6_frag {
 815     uint8_t ip6f_nxt;
 816     uint8_t ip6f_reserved;
 817     ovs_be16 ip6f_offlg;
 818     ovs_16aligned_be32 ip6f_ident;
 819 };
 820
 821 #define ICMP6_HEADER_LEN 4
 822 struct icmp6_header {
 823     uint8_t icmp6_type;
 824     uint8_t icmp6_code;
 825     ovs_be16 icmp6_cksum;
 826 };
 827 BUILD_ASSERT_DECL(ICMP6_HEADER_LEN == sizeof(struct icmp6_header));
 828
 829 uint32_t packet_csum_pseudoheader6(const struct ovs_16aligned_ip6_hdr *);
 830
 831 /* Neighbor Discovery option field.
 832  * ND options are always a multiple of 8 bytes in size. */
 833 #define ND_OPT_LEN 8
 834 struct ovs_nd_opt {
 835     uint8_t  nd_opt_type;      /* Values defined in icmp6.h */
 836     uint8_t  nd_opt_len;       /* in units of 8 octets (the size of this struct) */
 837     struct eth_addr nd_opt_mac;   /* Ethernet address in the case of SLL or TLL options */
 838 };
 839 BUILD_ASSERT_DECL(ND_OPT_LEN == sizeof(struct ovs_nd_opt));
 840
 841 /* Like struct nd_msg (from ndisc.h), but whereas that struct requires 32-bit
 842  * alignment, this one only requires 16-bit alignment. */
 843 #define ND_MSG_LEN 24
 844 struct ovs_nd_msg {
 845     struct icmp6_header icmph;
 846     ovs_16aligned_be32 rco_flags;
 847     union ovs_16aligned_in6_addr target;
 848     struct ovs_nd_opt options[0];
 849 };
 850 BUILD_ASSERT_DECL(ND_MSG_LEN == sizeof(struct ovs_nd_msg));
 851
 852 /*
 853  * Use the same struct for MLD and MLD2, naming members as the defined fields in
 854  * in the corresponding version of the protocol, though they are reserved in the
 855  * other one.
 856  */
 857 #define MLD_HEADER_LEN 8
 858 struct mld_header {
 859     uint8_t type;
 860     uint8_t code;
 861     ovs_be16 csum;
 862     ovs_be16 mrd;
 863     ovs_be16 ngrp;
 864 };
 865 BUILD_ASSERT_DECL(MLD_HEADER_LEN == sizeof(struct mld_header));
 866
 867 #define MLD2_RECORD_LEN 20
 868 struct mld2_record {
 869     uint8_t type;
 870     uint8_t aux_len;
 871     ovs_be16 nsrcs;
 872     union ovs_16aligned_in6_addr maddr;
 873 };
 874 BUILD_ASSERT_DECL(MLD2_RECORD_LEN == sizeof(struct mld2_record));
 875
 876 #define MLD_QUERY 130
 877 #define MLD_REPORT 131
 878 #define MLD_DONE 132
 879 #define MLD2_REPORT 143
 880
 881 /* The IPv6 flow label is in the lower 20 bits of the first 32-bit word. */
 882 #define IPV6_LABEL_MASK 0x000fffff
 883
 884 /* Example:
 885  *
 886  * char *string = "1 ::1 2";
 887  * char ipv6_s[IPV6_SCAN_LEN + 1];
 888  * struct in6_addr ipv6;
 889  *
 890  * if (ovs_scan(string, "%d"IPV6_SCAN_FMT"%d", &a, ipv6_s, &b)
 891  *     && inet_pton(AF_INET6, ipv6_s, &ipv6) == 1) {
 892  *     ...
 893  * }
 894  */
 895 #define IPV6_SCAN_FMT "%46[0123456789abcdefABCDEF:.]"
 896 #define IPV6_SCAN_LEN 46
 897
 898 extern const struct in6_addr in6addr_exact;
 899 #define IN6ADDR_EXACT_INIT { { { 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, \
 900                                  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff } } }
 901
 902 extern const struct in6_addr in6addr_all_hosts;
 903 #define IN6ADDR_ALL_HOSTS_INIT { { { 0xff,0x02,0x00,0x00,0x00,0x00,0x00,0x00, \
 904                                      0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x01 } } }
 905
 906 static inline bool ipv6_addr_equals(const struct in6_addr *a,
 907                                     const struct in6_addr *b)
 908 {
 909 #ifdef IN6_ARE_ADDR_EQUAL
 910     return IN6_ARE_ADDR_EQUAL(a, b);
 911 #else
 912     return !memcmp(a, b, sizeof(*a));
 913 #endif
 914 }
 915
 916 static inline bool ipv6_mask_is_any(const struct in6_addr *mask) {
 917     return ipv6_addr_equals(mask, &in6addr_any);
 918 }
 919
 920 static inline bool ipv6_mask_is_exact(const struct in6_addr *mask) {
 921     return ipv6_addr_equals(mask, &in6addr_exact);
 922 }
 923
 924 static inline bool ipv6_is_all_hosts(const struct in6_addr *addr) {
 925     return ipv6_addr_equals(addr, &in6addr_all_hosts);
 926 }
 927
 928 static inline bool ipv6_addr_is_set(const struct in6_addr *addr) {
 929     return !ipv6_addr_equals(addr, &in6addr_any);
 930 }
 931
 932 static inline bool ipv6_addr_is_multicast(const struct in6_addr *ip) {
 933     return ip->s6_addr[0] == 0xff;
 934 }
 935
 936 static inline struct in6_addr
 937 in6_addr_mapped_ipv4(ovs_be32 ip4)
 938 {
 939     struct in6_addr ip6 = { .s6_addr = { [10] = 0xff, [11] = 0xff } };
 940     memcpy(&ip6.s6_addr[12], &ip4, 4);
 941     return ip6;
 942 }
 943
 944 static inline void
 945 in6_addr_set_mapped_ipv4(struct in6_addr *ip6, ovs_be32 ip4)
 946 {
 947     *ip6 = in6_addr_mapped_ipv4(ip4);
 948 }
 949
 950 static inline ovs_be32
 951 in6_addr_get_mapped_ipv4(const struct in6_addr *addr)
 952 {
 953     union ovs_16aligned_in6_addr *taddr = (void *) addr;
 954     if (IN6_IS_ADDR_V4MAPPED(addr)) {
 955         return get_16aligned_be32(&taddr->be32[3]);
 956     } else {
 957         return INADDR_ANY;
 958     }
 959 }
 960
 961 static inline void
 962 in6_addr_solicited_node(struct in6_addr *addr, const struct in6_addr *ip6)
 963 {
 964     union ovs_16aligned_in6_addr *taddr = (void *) addr;
 965     memset(taddr->be16, 0, sizeof(taddr->be16));
 966     taddr->be16[0] = htons(0xff02);
 967     taddr->be16[5] = htons(0x1);
 968     taddr->be16[6] = htons(0xff00);
 969     memcpy(&addr->s6_addr[13], &ip6->s6_addr[13], 3);
 970 }
 971
 972 static inline void
 973 ipv6_multicast_to_ethernet(struct eth_addr *eth, const struct in6_addr *ip6)
 974 {
 975     eth->ea[0] = 0x33;
 976     eth->ea[1] = 0x33;
 977     eth->ea[2] = ip6->s6_addr[12];
 978     eth->ea[3] = ip6->s6_addr[13];
 979     eth->ea[4] = ip6->s6_addr[14];
 980     eth->ea[5] = ip6->s6_addr[15];
 981 }
 982
 983 static inline bool dl_type_is_ip_any(ovs_be16 dl_type)
 984 {
 985     return dl_type == htons(ETH_TYPE_IP)
 986         || dl_type == htons(ETH_TYPE_IPV6);
 987 }
 988
 989 /* Tunnel header */
 990
 991 /* GRE protocol header */
 992 struct gre_base_hdr {
 993     ovs_be16 flags;
 994     ovs_be16 protocol;
 995 };
 996
 997 #define GRE_CSUM        0x8000
 998 #define GRE_ROUTING     0x4000
 999 #define GRE_KEY         0x2000
1000 #define GRE_SEQ         0x1000
1001 #define GRE_STRICT      0x0800
1002 #define GRE_REC         0x0700
1003 #define GRE_FLAGS       0x00F8
1004 #define GRE_VERSION     0x0007
1005
1006 /* VXLAN protocol header */
1007 struct vxlanhdr {
1008     ovs_16aligned_be32 vx_flags;
1009     ovs_16aligned_be32 vx_vni;
1010 };
1011
1012 #define VXLAN_FLAGS 0x08000000  /* struct vxlanhdr.vx_flags required value. */
1013
1014 void ipv6_format_addr(const struct in6_addr *addr, struct ds *);
1015 void ipv6_format_addr_bracket(const struct in6_addr *addr, struct ds *,
1016                               bool bracket);
1017 void ipv6_format_mapped(const struct in6_addr *addr, struct ds *);
1018 void ipv6_format_masked(const struct in6_addr *addr,
1019                         const struct in6_addr *mask, struct ds *);
1020 const char * ipv6_string_mapped(char *addr_str, const struct in6_addr *addr);
1021 struct in6_addr ipv6_addr_bitand(const struct in6_addr *src,
1022                                  const struct in6_addr *mask);
1023 struct in6_addr ipv6_create_mask(int mask);
1024 int ipv6_count_cidr_bits(const struct in6_addr *netmask);
1025 bool ipv6_is_cidr(const struct in6_addr *netmask);
1026 char *ipv6_parse_masked(const char *s, struct in6_addr *ipv6,
1027                         struct in6_addr *mask);
1028
1029 void *eth_compose(struct dp_packet *, const struct eth_addr eth_dst,
1030                   const struct eth_addr eth_src, uint16_t eth_type,
1031                   size_t size);
1032 void *snap_compose(struct dp_packet *, const struct eth_addr eth_dst,
1033                    const struct eth_addr eth_src,
1034                    unsigned int oui, uint16_t snap_type, size_t size);
1035 void packet_set_ipv4(struct dp_packet *, ovs_be32 src, ovs_be32 dst, uint8_t tos,
1036                      uint8_t ttl);
1037 void packet_set_ipv6(struct dp_packet *, uint8_t proto, const ovs_be32 src[4],
1038                      const ovs_be32 dst[4], uint8_t tc,
1039                      ovs_be32 fl, uint8_t hlmit);
1040 void packet_set_tcp_port(struct dp_packet *, ovs_be16 src, ovs_be16 dst);
1041 void packet_set_udp_port(struct dp_packet *, ovs_be16 src, ovs_be16 dst);
1042 void packet_set_sctp_port(struct dp_packet *, ovs_be16 src, ovs_be16 dst);
1043 void packet_set_icmp(struct dp_packet *, uint8_t type, uint8_t code);
1044 void packet_set_nd(struct dp_packet *, const ovs_be32 target[4],
1045                    const struct eth_addr sll, const struct eth_addr tll);
1046
1047 void packet_format_tcp_flags(struct ds *, uint16_t);
1048 const char *packet_tcp_flag_to_string(uint32_t flag);
1049 void compose_arp(struct dp_packet *, uint16_t arp_op,
1050                  const struct eth_addr arp_sha,
1051                  const struct eth_addr arp_tha, bool broadcast,
1052                  ovs_be32 arp_spa, ovs_be32 arp_tpa);
1053 void compose_nd(struct dp_packet *, const struct eth_addr eth_src,
1054                 struct in6_addr *, struct in6_addr *);
1055 uint32_t packet_csum_pseudoheader(const struct ip_header *);
1056
1057 #endif /* packets.h */