2 * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016 Nicira, Inc.
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:
8 * http://www.apache.org/licenses/LICENSE-2.0
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.
21 #include <sys/types.h>
25 #include "openvswitch/geneve.h"
26 #include "openvswitch/packets.h"
27 #include "openvswitch/types.h"
28 #include "odp-netlink.h"
31 #include "tun-metadata.h"
32 #include "unaligned.h"
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. */
41 /* Tunnel information is in userspace datapath format. */
42 #define FLOW_TNL_F_UDPIF (1 << 4)
44 static inline bool ipv6_addr_is_set(const struct in6_addr *addr);
47 flow_tnl_dst_is_set(const struct flow_tnl *tnl)
49 return tnl->ip_dst || ipv6_addr_is_set(&tnl->ipv6_dst);
52 struct in6_addr flow_tnl_dst(const struct flow_tnl *tnl);
53 struct in6_addr flow_tnl_src(const struct flow_tnl *tnl);
55 /* Returns an offset to 'src' covering all the meaningful fields in 'src'. */
57 flow_tnl_size(const struct flow_tnl *src)
59 if (!flow_tnl_dst_is_set(src)) {
60 /* Covers ip_dst and ipv6_dst only. */
61 return offsetof(struct flow_tnl, ip_src);
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;
68 if (!src->metadata.present.map) {
69 /* No TLVs, opts is irrelevant. */
70 return offsetof(struct flow_tnl, metadata.opts);
72 /* Have decoded TLVs, opts is relevant. */
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. */
80 flow_tnl_copy__(struct flow_tnl *dst, const struct flow_tnl *src)
82 memcpy(dst, src, flow_tnl_size(src));
86 flow_tnl_equal(const struct flow_tnl *a, const struct flow_tnl *b)
88 size_t a_size = flow_tnl_size(a);
90 return a_size == flow_tnl_size(b) && !memcmp(a, b, a_size);
93 /* Datapath packet 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
100 uint32_t skb_priority; /* Packet priority for QoS. */
101 uint32_t pkt_mark; /* Packet mark. */
102 uint16_t ct_state; /* Connection state. */
103 uint16_t ct_zone; /* Connection zone. */
104 uint32_t ct_mark; /* Connection mark. */
105 ovs_u128 ct_label; /* Connection label. */
106 union flow_in_port in_port; /* Input port. */
107 struct flow_tnl tunnel; /* Encapsulating tunnel parameters. Note that
108 * if 'ip_dst' == 0, the rest of the fields may
109 * be uninitialized. */
113 pkt_metadata_init_tnl(struct pkt_metadata *md)
115 /* Zero up through the tunnel metadata options. The length and table
116 * are before this and as long as they are empty, the options won't
118 memset(md, 0, offsetof(struct pkt_metadata, tunnel.metadata.opts));
122 pkt_metadata_init(struct pkt_metadata *md, odp_port_t port)
124 /* It can be expensive to zero out all of the tunnel metadata. However,
125 * we can just zero out ip_dst and the rest of the data will never be
127 memset(md, 0, offsetof(struct pkt_metadata, in_port));
128 md->tunnel.ip_dst = 0;
129 md->tunnel.ipv6_dst = in6addr_any;
131 md->in_port.odp_port = port;
134 /* This function prefetches the cachelines touched by pkt_metadata_init()
135 * For performance reasons the two functions should be kept in sync. */
137 pkt_metadata_prefetch_init(struct pkt_metadata *md)
139 ovs_prefetch_range(md, offsetof(struct pkt_metadata, tunnel.ip_src));
142 bool dpid_from_string(const char *s, uint64_t *dpidp);
144 #define ETH_ADDR_LEN 6
146 static const struct eth_addr eth_addr_broadcast OVS_UNUSED
147 = { { { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff } } };
149 static const struct eth_addr eth_addr_exact OVS_UNUSED
150 = { { { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff } } };
152 static const struct eth_addr eth_addr_zero OVS_UNUSED
153 = { { { 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 } } };
155 static const struct eth_addr eth_addr_stp OVS_UNUSED
156 = { { { 0x01, 0x80, 0xC2, 0x00, 0x00, 0x00 } } };
158 static const struct eth_addr eth_addr_lacp OVS_UNUSED
159 = { { { 0x01, 0x80, 0xC2, 0x00, 0x00, 0x02 } } };
161 static const struct eth_addr eth_addr_bfd OVS_UNUSED
162 = { { { 0x00, 0x23, 0x20, 0x00, 0x00, 0x01 } } };
164 static inline bool eth_addr_is_broadcast(const struct eth_addr a)
166 return (a.be16[0] & a.be16[1] & a.be16[2]) == htons(0xffff);
169 static inline bool eth_addr_is_multicast(const struct eth_addr a)
174 static inline bool eth_addr_is_local(const struct eth_addr a)
176 /* Local if it is either a locally administered address or a Nicira random
179 || (a.be16[0] == htons(0x0023)
180 && (a.be16[1] & htons(0xff80)) == htons(0x2080));
182 static inline bool eth_addr_is_zero(const struct eth_addr a)
184 return !(a.be16[0] | a.be16[1] | a.be16[2]);
187 static inline int eth_mask_is_exact(const struct eth_addr a)
189 return (a.be16[0] & a.be16[1] & a.be16[2]) == htons(0xffff);
192 static inline int eth_addr_compare_3way(const struct eth_addr a,
193 const struct eth_addr b)
195 return memcmp(&a, &b, sizeof a);
198 static inline bool eth_addr_equals(const struct eth_addr a,
199 const struct eth_addr b)
201 return !eth_addr_compare_3way(a, b);
204 static inline bool eth_addr_equal_except(const struct eth_addr a,
205 const struct eth_addr b,
206 const struct eth_addr mask)
208 return !(((a.be16[0] ^ b.be16[0]) & mask.be16[0])
209 || ((a.be16[1] ^ b.be16[1]) & mask.be16[1])
210 || ((a.be16[2] ^ b.be16[2]) & mask.be16[2]));
213 static inline uint64_t eth_addr_to_uint64(const struct eth_addr ea)
215 return (((uint64_t) ntohs(ea.be16[0]) << 32)
216 | ((uint64_t) ntohs(ea.be16[1]) << 16)
217 | ntohs(ea.be16[2]));
220 static inline uint64_t eth_addr_vlan_to_uint64(const struct eth_addr ea,
223 return (((uint64_t)vlan << 48) | eth_addr_to_uint64(ea));
226 static inline void eth_addr_from_uint64(uint64_t x, struct eth_addr *ea)
228 ea->be16[0] = htons(x >> 32);
229 ea->be16[1] = htons((x & 0xFFFF0000) >> 16);
230 ea->be16[2] = htons(x & 0xFFFF);
233 static inline struct eth_addr eth_addr_invert(const struct eth_addr src)
237 for (int i = 0; i < ARRAY_SIZE(src.be16); i++) {
238 dst.be16[i] = ~src.be16[i];
244 static inline void eth_addr_mark_random(struct eth_addr *ea)
246 ea->ea[0] &= ~1; /* Unicast. */
247 ea->ea[0] |= 2; /* Private. */
250 static inline void eth_addr_random(struct eth_addr *ea)
252 random_bytes((uint8_t *)ea, sizeof *ea);
253 eth_addr_mark_random(ea);
256 static inline void eth_addr_nicira_random(struct eth_addr *ea)
260 /* Set the OUI to the Nicira one. */
265 /* Set the top bit to indicate random Nicira address. */
268 static inline uint32_t hash_mac(const struct eth_addr ea,
269 const uint16_t vlan, const uint32_t basis)
271 return hash_uint64_basis(eth_addr_vlan_to_uint64(ea, vlan), basis);
274 bool eth_addr_is_reserved(const struct eth_addr);
275 bool eth_addr_from_string(const char *, struct eth_addr *);
277 void compose_rarp(struct dp_packet *, const struct eth_addr);
279 void eth_push_vlan(struct dp_packet *, ovs_be16 tpid, ovs_be16 tci);
280 void eth_pop_vlan(struct dp_packet *);
282 const char *eth_from_hex(const char *hex, struct dp_packet **packetp);
283 void eth_format_masked(const struct eth_addr ea,
284 const struct eth_addr *mask, struct ds *s);
286 void set_mpls_lse(struct dp_packet *, ovs_be32 label);
287 void push_mpls(struct dp_packet *packet, ovs_be16 ethtype, ovs_be32 lse);
288 void pop_mpls(struct dp_packet *, ovs_be16 ethtype);
290 void set_mpls_lse_ttl(ovs_be32 *lse, uint8_t ttl);
291 void set_mpls_lse_tc(ovs_be32 *lse, uint8_t tc);
292 void set_mpls_lse_label(ovs_be32 *lse, ovs_be32 label);
293 void set_mpls_lse_bos(ovs_be32 *lse, uint8_t bos);
294 ovs_be32 set_mpls_lse_values(uint8_t ttl, uint8_t tc, uint8_t bos,
299 * struct eth_addr mac;
301 * printf("The Ethernet address is "ETH_ADDR_FMT"\n", ETH_ADDR_ARGS(mac));
304 #define ETH_ADDR_FMT \
305 "%02"PRIx8":%02"PRIx8":%02"PRIx8":%02"PRIx8":%02"PRIx8":%02"PRIx8
306 #define ETH_ADDR_ARGS(EA) ETH_ADDR_BYTES_ARGS((EA).ea)
307 #define ETH_ADDR_BYTES_ARGS(EAB) \
308 (EAB)[0], (EAB)[1], (EAB)[2], (EAB)[3], (EAB)[4], (EAB)[5]
309 #define ETH_ADDR_STRLEN 17
313 * char *string = "1 00:11:22:33:44:55 2";
314 * struct eth_addr mac;
317 * if (ovs_scan(string, "%d"ETH_ADDR_SCAN_FMT"%d",
318 * &a, ETH_ADDR_SCAN_ARGS(mac), &b)) {
322 #define ETH_ADDR_SCAN_FMT "%"SCNx8":%"SCNx8":%"SCNx8":%"SCNx8":%"SCNx8":%"SCNx8
323 #define ETH_ADDR_SCAN_ARGS(EA) \
324 &(EA).ea[0], &(EA).ea[1], &(EA).ea[2], &(EA).ea[3], &(EA).ea[4], &(EA).ea[5]
326 #define ETH_TYPE_IP 0x0800
327 #define ETH_TYPE_ARP 0x0806
328 #define ETH_TYPE_TEB 0x6558
329 #define ETH_TYPE_VLAN_8021Q 0x8100
330 #define ETH_TYPE_VLAN ETH_TYPE_VLAN_8021Q
331 #define ETH_TYPE_VLAN_8021AD 0x88a8
332 #define ETH_TYPE_IPV6 0x86dd
333 #define ETH_TYPE_LACP 0x8809
334 #define ETH_TYPE_RARP 0x8035
335 #define ETH_TYPE_MPLS 0x8847
336 #define ETH_TYPE_MPLS_MCAST 0x8848
338 static inline bool eth_type_mpls(ovs_be16 eth_type)
340 return eth_type == htons(ETH_TYPE_MPLS) ||
341 eth_type == htons(ETH_TYPE_MPLS_MCAST);
344 static inline bool eth_type_vlan(ovs_be16 eth_type)
346 return eth_type == htons(ETH_TYPE_VLAN_8021Q) ||
347 eth_type == htons(ETH_TYPE_VLAN_8021AD);
351 /* Minimum value for an Ethernet type. Values below this are IEEE 802.2 frame
353 #define ETH_TYPE_MIN 0x600
355 #define ETH_HEADER_LEN 14
356 #define ETH_PAYLOAD_MIN 46
357 #define ETH_PAYLOAD_MAX 1500
358 #define ETH_TOTAL_MIN (ETH_HEADER_LEN + ETH_PAYLOAD_MIN)
359 #define ETH_TOTAL_MAX (ETH_HEADER_LEN + ETH_PAYLOAD_MAX)
360 #define ETH_VLAN_TOTAL_MAX (ETH_HEADER_LEN + VLAN_HEADER_LEN + ETH_PAYLOAD_MAX)
363 struct eth_addr eth_dst;
364 struct eth_addr eth_src;
367 BUILD_ASSERT_DECL(ETH_HEADER_LEN == sizeof(struct eth_header));
369 #define LLC_DSAP_SNAP 0xaa
370 #define LLC_SSAP_SNAP 0xaa
371 #define LLC_CNTL_SNAP 3
373 #define LLC_HEADER_LEN 3
380 BUILD_ASSERT_DECL(LLC_HEADER_LEN == sizeof(struct llc_header));
382 /* LLC field values used for STP frames. */
383 #define STP_LLC_SSAP 0x42
384 #define STP_LLC_DSAP 0x42
385 #define STP_LLC_CNTL 0x03
387 #define SNAP_ORG_ETHERNET "\0\0" /* The compiler adds a null byte, so
388 sizeof(SNAP_ORG_ETHERNET) == 3. */
389 #define SNAP_HEADER_LEN 5
395 BUILD_ASSERT_DECL(SNAP_HEADER_LEN == sizeof(struct snap_header));
397 #define LLC_SNAP_HEADER_LEN (LLC_HEADER_LEN + SNAP_HEADER_LEN)
399 struct llc_snap_header {
400 struct llc_header llc;
401 struct snap_header snap;
403 BUILD_ASSERT_DECL(LLC_SNAP_HEADER_LEN == sizeof(struct llc_snap_header));
405 #define VLAN_VID_MASK 0x0fff
406 #define VLAN_VID_SHIFT 0
408 #define VLAN_PCP_MASK 0xe000
409 #define VLAN_PCP_SHIFT 13
411 #define VLAN_CFI 0x1000
412 #define VLAN_CFI_SHIFT 12
414 /* Given the vlan_tci field from an 802.1Q header, in network byte order,
415 * returns the VLAN ID in host byte order. */
416 static inline uint16_t
417 vlan_tci_to_vid(ovs_be16 vlan_tci)
419 return (ntohs(vlan_tci) & VLAN_VID_MASK) >> VLAN_VID_SHIFT;
422 /* Given the vlan_tci field from an 802.1Q header, in network byte order,
423 * returns the priority code point (PCP) in host byte order. */
425 vlan_tci_to_pcp(ovs_be16 vlan_tci)
427 return (ntohs(vlan_tci) & VLAN_PCP_MASK) >> VLAN_PCP_SHIFT;
430 /* Given the vlan_tci field from an 802.1Q header, in network byte order,
431 * returns the Canonical Format Indicator (CFI). */
433 vlan_tci_to_cfi(ovs_be16 vlan_tci)
435 return (vlan_tci & htons(VLAN_CFI)) != 0;
438 #define VLAN_HEADER_LEN 4
440 ovs_be16 vlan_tci; /* Lowest 12 bits are VLAN ID. */
441 ovs_be16 vlan_next_type;
443 BUILD_ASSERT_DECL(VLAN_HEADER_LEN == sizeof(struct vlan_header));
445 #define VLAN_ETH_HEADER_LEN (ETH_HEADER_LEN + VLAN_HEADER_LEN)
447 struct vlan_eth_header {
448 struct eth_addr veth_dst;
449 struct eth_addr veth_src;
450 ovs_be16 veth_type; /* Always htons(ETH_TYPE_VLAN). */
451 ovs_be16 veth_tci; /* Lowest 12 bits are VLAN ID. */
452 ovs_be16 veth_next_type;
454 BUILD_ASSERT_DECL(VLAN_ETH_HEADER_LEN == sizeof(struct vlan_eth_header));
456 /* MPLS related definitions */
457 #define MPLS_TTL_MASK 0x000000ff
458 #define MPLS_TTL_SHIFT 0
460 #define MPLS_BOS_MASK 0x00000100
461 #define MPLS_BOS_SHIFT 8
463 #define MPLS_TC_MASK 0x00000e00
464 #define MPLS_TC_SHIFT 9
466 #define MPLS_LABEL_MASK 0xfffff000
467 #define MPLS_LABEL_SHIFT 12
472 ovs_16aligned_be32 mpls_lse;
474 BUILD_ASSERT_DECL(MPLS_HLEN == sizeof(struct mpls_hdr));
476 /* Given a mpls label stack entry in network byte order
477 * return mpls label in host byte order */
478 static inline uint32_t
479 mpls_lse_to_label(ovs_be32 mpls_lse)
481 return (ntohl(mpls_lse) & MPLS_LABEL_MASK) >> MPLS_LABEL_SHIFT;
484 /* Given a mpls label stack entry in network byte order
486 static inline uint8_t
487 mpls_lse_to_tc(ovs_be32 mpls_lse)
489 return (ntohl(mpls_lse) & MPLS_TC_MASK) >> MPLS_TC_SHIFT;
492 /* Given a mpls label stack entry in network byte order
494 static inline uint8_t
495 mpls_lse_to_ttl(ovs_be32 mpls_lse)
497 return (ntohl(mpls_lse) & MPLS_TTL_MASK) >> MPLS_TTL_SHIFT;
500 /* Set TTL in mpls lse. */
502 flow_set_mpls_lse_ttl(ovs_be32 *mpls_lse, uint8_t ttl)
504 *mpls_lse &= ~htonl(MPLS_TTL_MASK);
505 *mpls_lse |= htonl(ttl << MPLS_TTL_SHIFT);
508 /* Given a mpls label stack entry in network byte order
509 * return mpls BoS bit */
510 static inline uint8_t
511 mpls_lse_to_bos(ovs_be32 mpls_lse)
513 return (mpls_lse & htonl(MPLS_BOS_MASK)) != 0;
516 #define IP_FMT "%"PRIu32".%"PRIu32".%"PRIu32".%"PRIu32
517 #define IP_ARGS(ip) \
519 (ntohl(ip) >> 16) & 0xff, \
520 (ntohl(ip) >> 8) & 0xff, \
525 * char *string = "1 33.44.55.66 2";
529 * if (ovs_scan(string, "%d"IP_SCAN_FMT"%d", &a, IP_SCAN_ARGS(&ip), &b)) {
533 #define IP_SCAN_FMT "%"SCNu8".%"SCNu8".%"SCNu8".%"SCNu8
534 #define IP_SCAN_ARGS(ip) \
535 ((void) (ovs_be32) *(ip), &((uint8_t *) ip)[0]), \
536 &((uint8_t *) ip)[1], \
537 &((uint8_t *) ip)[2], \
540 /* Returns true if 'netmask' is a CIDR netmask, that is, if it consists of N
541 * high-order 1-bits and 32-N low-order 0-bits. */
543 ip_is_cidr(ovs_be32 netmask)
545 uint32_t x = ~ntohl(netmask);
546 return !(x & (x + 1));
549 ip_is_multicast(ovs_be32 ip)
551 return (ip & htonl(0xf0000000)) == htonl(0xe0000000);
554 ip_is_local_multicast(ovs_be32 ip)
556 return (ip & htonl(0xffffff00)) == htonl(0xe0000000);
558 int ip_count_cidr_bits(ovs_be32 netmask);
559 void ip_format_masked(ovs_be32 ip, ovs_be32 mask, struct ds *);
560 bool ip_parse(const char *s, ovs_be32 *ip);
561 char *ip_parse_masked(const char *s, ovs_be32 *ip, ovs_be32 *mask)
562 OVS_WARN_UNUSED_RESULT;
563 char *ip_parse_cidr(const char *s, ovs_be32 *ip, unsigned int *plen)
564 OVS_WARN_UNUSED_RESULT;
565 char *ip_parse_masked_len(const char *s, int *n, ovs_be32 *ip, ovs_be32 *mask)
566 OVS_WARN_UNUSED_RESULT;
567 char *ip_parse_cidr_len(const char *s, int *n, ovs_be32 *ip,
569 OVS_WARN_UNUSED_RESULT;
571 #define IP_VER(ip_ihl_ver) ((ip_ihl_ver) >> 4)
572 #define IP_IHL(ip_ihl_ver) ((ip_ihl_ver) & 15)
573 #define IP_IHL_VER(ihl, ver) (((ver) << 4) | (ihl))
576 #define IPPROTO_SCTP 132
580 #define IP_ECN_NOT_ECT 0x0
581 #define IP_ECN_ECT_1 0x01
582 #define IP_ECN_ECT_0 0x02
583 #define IP_ECN_CE 0x03
584 #define IP_ECN_MASK 0x03
585 #define IP_DSCP_MASK 0xfc
588 IP_ECN_is_ce(uint8_t dsfield)
590 return (dsfield & IP_ECN_MASK) == IP_ECN_CE;
595 #define IP_DONT_FRAGMENT 0x4000 /* Don't fragment. */
596 #define IP_MORE_FRAGMENTS 0x2000 /* More fragments. */
597 #define IP_FRAG_OFF_MASK 0x1fff /* Fragment offset. */
598 #define IP_IS_FRAGMENT(ip_frag_off) \
599 ((ip_frag_off) & htons(IP_MORE_FRAGMENTS | IP_FRAG_OFF_MASK))
601 #define IP_HEADER_LEN 20
607 ovs_be16 ip_frag_off;
611 ovs_16aligned_be32 ip_src;
612 ovs_16aligned_be32 ip_dst;
615 BUILD_ASSERT_DECL(IP_HEADER_LEN == sizeof(struct ip_header));
617 #define ICMP_HEADER_LEN 8
631 ovs_16aligned_be32 gateway;
634 BUILD_ASSERT_DECL(ICMP_HEADER_LEN == sizeof(struct icmp_header));
636 #define IGMP_HEADER_LEN 8
641 ovs_16aligned_be32 group;
643 BUILD_ASSERT_DECL(IGMP_HEADER_LEN == sizeof(struct igmp_header));
645 #define IGMPV3_HEADER_LEN 8
646 struct igmpv3_header {
653 BUILD_ASSERT_DECL(IGMPV3_HEADER_LEN == sizeof(struct igmpv3_header));
655 #define IGMPV3_RECORD_LEN 8
656 struct igmpv3_record {
660 ovs_16aligned_be32 maddr;
662 BUILD_ASSERT_DECL(IGMPV3_RECORD_LEN == sizeof(struct igmpv3_record));
664 #define IGMP_HOST_MEMBERSHIP_QUERY 0x11 /* From RFC1112 */
665 #define IGMP_HOST_MEMBERSHIP_REPORT 0x12 /* Ditto */
666 #define IGMPV2_HOST_MEMBERSHIP_REPORT 0x16 /* V2 version of 0x12 */
667 #define IGMP_HOST_LEAVE_MESSAGE 0x17
668 #define IGMPV3_HOST_MEMBERSHIP_REPORT 0x22 /* V3 version of 0x12 */
671 * IGMPv3 and MLDv2 use the same codes.
673 #define IGMPV3_MODE_IS_INCLUDE 1
674 #define IGMPV3_MODE_IS_EXCLUDE 2
675 #define IGMPV3_CHANGE_TO_INCLUDE_MODE 3
676 #define IGMPV3_CHANGE_TO_EXCLUDE_MODE 4
677 #define IGMPV3_ALLOW_NEW_SOURCES 5
678 #define IGMPV3_BLOCK_OLD_SOURCES 6
680 #define SCTP_HEADER_LEN 12
684 ovs_16aligned_be32 sctp_vtag;
685 ovs_16aligned_be32 sctp_csum;
687 BUILD_ASSERT_DECL(SCTP_HEADER_LEN == sizeof(struct sctp_header));
689 #define UDP_HEADER_LEN 8
696 BUILD_ASSERT_DECL(UDP_HEADER_LEN == sizeof(struct udp_header));
698 #define TCP_FIN 0x001
699 #define TCP_SYN 0x002
700 #define TCP_RST 0x004
701 #define TCP_PSH 0x008
702 #define TCP_ACK 0x010
703 #define TCP_URG 0x020
704 #define TCP_ECE 0x040
705 #define TCP_CWR 0x080
708 #define TCP_CTL(flags, offset) (htons((flags) | ((offset) << 12)))
709 #define TCP_FLAGS(tcp_ctl) (ntohs(tcp_ctl) & 0x0fff)
710 #define TCP_FLAGS_BE16(tcp_ctl) ((tcp_ctl) & htons(0x0fff))
711 #define TCP_OFFSET(tcp_ctl) (ntohs(tcp_ctl) >> 12)
713 #define TCP_HEADER_LEN 20
717 ovs_16aligned_be32 tcp_seq;
718 ovs_16aligned_be32 tcp_ack;
724 BUILD_ASSERT_DECL(TCP_HEADER_LEN == sizeof(struct tcp_header));
726 /* Connection states */
729 CS_ESTABLISHED_BIT = 1,
731 CS_REPLY_DIR_BIT = 3,
739 CS_NEW = (1 << CS_NEW_BIT),
740 CS_ESTABLISHED = (1 << CS_ESTABLISHED_BIT),
741 CS_RELATED = (1 << CS_RELATED_BIT),
742 CS_REPLY_DIR = (1 << CS_REPLY_DIR_BIT),
743 CS_INVALID = (1 << CS_INVALID_BIT),
744 CS_TRACKED = (1 << CS_TRACKED_BIT),
745 CS_SRC_NAT = (1 << CS_SRC_NAT_BIT),
746 CS_DST_NAT = (1 << CS_DST_NAT_BIT),
749 /* Undefined connection state bits. */
750 #define CS_SUPPORTED_MASK (CS_NEW | CS_ESTABLISHED | CS_RELATED \
751 | CS_INVALID | CS_REPLY_DIR | CS_TRACKED \
752 | CS_SRC_NAT | CS_DST_NAT)
753 #define CS_UNSUPPORTED_MASK (~(uint32_t)CS_SUPPORTED_MASK)
755 #define ARP_HRD_ETHERNET 1
756 #define ARP_PRO_IP 0x0800
757 #define ARP_OP_REQUEST 1
758 #define ARP_OP_REPLY 2
759 #define ARP_OP_RARP 3
761 #define ARP_ETH_HEADER_LEN 28
762 struct arp_eth_header {
763 /* Generic members. */
764 ovs_be16 ar_hrd; /* Hardware type. */
765 ovs_be16 ar_pro; /* Protocol type. */
766 uint8_t ar_hln; /* Hardware address length. */
767 uint8_t ar_pln; /* Protocol address length. */
768 ovs_be16 ar_op; /* Opcode. */
770 /* Ethernet+IPv4 specific members. */
771 struct eth_addr ar_sha; /* Sender hardware address. */
772 ovs_16aligned_be32 ar_spa; /* Sender protocol address. */
773 struct eth_addr ar_tha; /* Target hardware address. */
774 ovs_16aligned_be32 ar_tpa; /* Target protocol address. */
776 BUILD_ASSERT_DECL(ARP_ETH_HEADER_LEN == sizeof(struct arp_eth_header));
778 #define IPV6_HEADER_LEN 40
780 /* Like struct in6_addr, but whereas that struct requires 32-bit alignment on
781 * most implementations, this one only requires 16-bit alignment. */
782 union ovs_16aligned_in6_addr {
784 ovs_16aligned_be32 be32[4];
787 /* Like struct in6_hdr, but whereas that struct requires 32-bit alignment, this
788 * one only requires 16-bit alignment. */
789 struct ovs_16aligned_ip6_hdr {
791 struct ovs_16aligned_ip6_hdrctl {
792 ovs_16aligned_be32 ip6_un1_flow;
793 ovs_be16 ip6_un1_plen;
795 uint8_t ip6_un1_hlim;
799 union ovs_16aligned_in6_addr ip6_src;
800 union ovs_16aligned_in6_addr ip6_dst;
803 /* Like struct in6_frag, but whereas that struct requires 32-bit alignment,
804 * this one only requires 16-bit alignment. */
805 struct ovs_16aligned_ip6_frag {
807 uint8_t ip6f_reserved;
809 ovs_16aligned_be32 ip6f_ident;
812 #define ICMP6_HEADER_LEN 4
813 struct icmp6_header {
816 ovs_be16 icmp6_cksum;
818 BUILD_ASSERT_DECL(ICMP6_HEADER_LEN == sizeof(struct icmp6_header));
820 uint32_t packet_csum_pseudoheader6(const struct ovs_16aligned_ip6_hdr *);
822 /* Neighbor Discovery option field.
823 * ND options are always a multiple of 8 bytes in size. */
826 uint8_t nd_opt_type; /* Values defined in icmp6.h */
827 uint8_t nd_opt_len; /* in units of 8 octets (the size of this struct) */
828 struct eth_addr nd_opt_mac; /* Ethernet address in the case of SLL or TLL options */
830 BUILD_ASSERT_DECL(ND_OPT_LEN == sizeof(struct ovs_nd_opt));
832 /* Like struct nd_msg (from ndisc.h), but whereas that struct requires 32-bit
833 * alignment, this one only requires 16-bit alignment. */
834 #define ND_MSG_LEN 24
836 struct icmp6_header icmph;
837 ovs_16aligned_be32 rco_flags;
838 union ovs_16aligned_in6_addr target;
839 struct ovs_nd_opt options[0];
841 BUILD_ASSERT_DECL(ND_MSG_LEN == sizeof(struct ovs_nd_msg));
844 * Use the same struct for MLD and MLD2, naming members as the defined fields in
845 * in the corresponding version of the protocol, though they are reserved in the
848 #define MLD_HEADER_LEN 8
856 BUILD_ASSERT_DECL(MLD_HEADER_LEN == sizeof(struct mld_header));
858 #define MLD2_RECORD_LEN 20
863 union ovs_16aligned_in6_addr maddr;
865 BUILD_ASSERT_DECL(MLD2_RECORD_LEN == sizeof(struct mld2_record));
867 #define MLD_QUERY 130
868 #define MLD_REPORT 131
870 #define MLD2_REPORT 143
872 /* The IPv6 flow label is in the lower 20 bits of the first 32-bit word. */
873 #define IPV6_LABEL_MASK 0x000fffff
877 * char *string = "1 ::1 2";
878 * char ipv6_s[IPV6_SCAN_LEN + 1];
879 * struct in6_addr ipv6;
881 * if (ovs_scan(string, "%d"IPV6_SCAN_FMT"%d", &a, ipv6_s, &b)
882 * && inet_pton(AF_INET6, ipv6_s, &ipv6) == 1) {
886 #define IPV6_SCAN_FMT "%46[0123456789abcdefABCDEF:.]"
887 #define IPV6_SCAN_LEN 46
889 extern const struct in6_addr in6addr_exact;
890 #define IN6ADDR_EXACT_INIT { { { 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, \
891 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff } } }
893 extern const struct in6_addr in6addr_all_hosts;
894 #define IN6ADDR_ALL_HOSTS_INIT { { { 0xff,0x02,0x00,0x00,0x00,0x00,0x00,0x00, \
895 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x01 } } }
897 static inline bool ipv6_addr_equals(const struct in6_addr *a,
898 const struct in6_addr *b)
900 #ifdef IN6_ARE_ADDR_EQUAL
901 return IN6_ARE_ADDR_EQUAL(a, b);
903 return !memcmp(a, b, sizeof(*a));
907 static inline bool ipv6_mask_is_any(const struct in6_addr *mask) {
908 return ipv6_addr_equals(mask, &in6addr_any);
911 static inline bool ipv6_mask_is_exact(const struct in6_addr *mask) {
912 return ipv6_addr_equals(mask, &in6addr_exact);
915 static inline bool ipv6_is_all_hosts(const struct in6_addr *addr) {
916 return ipv6_addr_equals(addr, &in6addr_all_hosts);
919 static inline bool ipv6_addr_is_set(const struct in6_addr *addr) {
920 return !ipv6_addr_equals(addr, &in6addr_any);
923 static inline bool ipv6_addr_is_multicast(const struct in6_addr *ip) {
924 return ip->s6_addr[0] == 0xff;
927 static inline struct in6_addr
928 in6_addr_mapped_ipv4(ovs_be32 ip4)
930 struct in6_addr ip6 = { .s6_addr = { [10] = 0xff, [11] = 0xff } };
931 memcpy(&ip6.s6_addr[12], &ip4, 4);
936 in6_addr_set_mapped_ipv4(struct in6_addr *ip6, ovs_be32 ip4)
938 *ip6 = in6_addr_mapped_ipv4(ip4);
941 static inline ovs_be32
942 in6_addr_get_mapped_ipv4(const struct in6_addr *addr)
944 union ovs_16aligned_in6_addr *taddr = (void *) addr;
945 if (IN6_IS_ADDR_V4MAPPED(addr)) {
946 return get_16aligned_be32(&taddr->be32[3]);
953 in6_addr_solicited_node(struct in6_addr *addr, const struct in6_addr *ip6)
955 union ovs_16aligned_in6_addr *taddr = (void *) addr;
956 memset(taddr->be16, 0, sizeof(taddr->be16));
957 taddr->be16[0] = htons(0xff02);
958 taddr->be16[5] = htons(0x1);
959 taddr->be16[6] = htons(0xff00);
960 memcpy(&addr->s6_addr[13], &ip6->s6_addr[13], 3);
964 * Generates ipv6 link local address from the given eth addr
965 * with prefix 'fe80::/64' and stores it in 'lla'
968 in6_generate_lla(struct eth_addr ea, struct in6_addr *lla)
970 union ovs_16aligned_in6_addr *taddr = (void *) lla;
971 memset(taddr->be16, 0, sizeof(taddr->be16));
972 taddr->be16[0] = htons(0xfe80);
973 taddr->be16[4] = htons(((ea.ea[0] ^ 0x02) << 8) | ea.ea[1]);
974 taddr->be16[5] = htons(ea.ea[2] << 8 | 0x00ff);
975 taddr->be16[6] = htons(0xfe << 8 | ea.ea[3]);
976 taddr->be16[7] = ea.be16[2];
980 ipv6_multicast_to_ethernet(struct eth_addr *eth, const struct in6_addr *ip6)
984 eth->ea[2] = ip6->s6_addr[12];
985 eth->ea[3] = ip6->s6_addr[13];
986 eth->ea[4] = ip6->s6_addr[14];
987 eth->ea[5] = ip6->s6_addr[15];
990 static inline bool dl_type_is_ip_any(ovs_be16 dl_type)
992 return dl_type == htons(ETH_TYPE_IP)
993 || dl_type == htons(ETH_TYPE_IPV6);
998 /* GRE protocol header */
999 struct gre_base_hdr {
1004 #define GRE_CSUM 0x8000
1005 #define GRE_ROUTING 0x4000
1006 #define GRE_KEY 0x2000
1007 #define GRE_SEQ 0x1000
1008 #define GRE_STRICT 0x0800
1009 #define GRE_REC 0x0700
1010 #define GRE_FLAGS 0x00F8
1011 #define GRE_VERSION 0x0007
1013 /* VXLAN protocol header */
1015 ovs_16aligned_be32 vx_flags;
1016 ovs_16aligned_be32 vx_vni;
1019 #define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
1021 void ipv6_format_addr(const struct in6_addr *addr, struct ds *);
1022 void ipv6_format_addr_bracket(const struct in6_addr *addr, struct ds *,
1024 void ipv6_format_mapped(const struct in6_addr *addr, struct ds *);
1025 void ipv6_format_masked(const struct in6_addr *addr,
1026 const struct in6_addr *mask, struct ds *);
1027 const char * ipv6_string_mapped(char *addr_str, const struct in6_addr *addr);
1028 struct in6_addr ipv6_addr_bitand(const struct in6_addr *src,
1029 const struct in6_addr *mask);
1030 struct in6_addr ipv6_create_mask(int mask);
1031 int ipv6_count_cidr_bits(const struct in6_addr *netmask);
1032 bool ipv6_is_cidr(const struct in6_addr *netmask);
1034 bool ipv6_parse(const char *s, struct in6_addr *ip);
1035 char *ipv6_parse_masked(const char *s, struct in6_addr *ipv6,
1036 struct in6_addr *mask);
1037 char *ipv6_parse_cidr(const char *s, struct in6_addr *ip, unsigned int *plen)
1038 OVS_WARN_UNUSED_RESULT;
1039 char *ipv6_parse_masked_len(const char *s, int *n, struct in6_addr *ipv6,
1040 struct in6_addr *mask);
1041 char *ipv6_parse_cidr_len(const char *s, int *n, struct in6_addr *ip,
1043 OVS_WARN_UNUSED_RESULT;
1045 void *eth_compose(struct dp_packet *, const struct eth_addr eth_dst,
1046 const struct eth_addr eth_src, uint16_t eth_type,
1048 void *snap_compose(struct dp_packet *, const struct eth_addr eth_dst,
1049 const struct eth_addr eth_src,
1050 unsigned int oui, uint16_t snap_type, size_t size);
1051 void packet_set_ipv4(struct dp_packet *, ovs_be32 src, ovs_be32 dst, uint8_t tos,
1053 void packet_set_ipv6(struct dp_packet *, const ovs_be32 src[4],
1054 const ovs_be32 dst[4], uint8_t tc,
1055 ovs_be32 fl, uint8_t hlmit);
1056 void packet_set_tcp_port(struct dp_packet *, ovs_be16 src, ovs_be16 dst);
1057 void packet_set_udp_port(struct dp_packet *, ovs_be16 src, ovs_be16 dst);
1058 void packet_set_sctp_port(struct dp_packet *, ovs_be16 src, ovs_be16 dst);
1059 void packet_set_icmp(struct dp_packet *, uint8_t type, uint8_t code);
1060 void packet_set_nd(struct dp_packet *, const ovs_be32 target[4],
1061 const struct eth_addr sll, const struct eth_addr tll);
1063 void packet_format_tcp_flags(struct ds *, uint16_t);
1064 const char *packet_tcp_flag_to_string(uint32_t flag);
1065 void compose_arp__(struct dp_packet *);
1066 void compose_arp(struct dp_packet *, uint16_t arp_op,
1067 const struct eth_addr arp_sha,
1068 const struct eth_addr arp_tha, bool broadcast,
1069 ovs_be32 arp_spa, ovs_be32 arp_tpa);
1070 void compose_nd(struct dp_packet *, const struct eth_addr eth_src,
1071 struct in6_addr *, struct in6_addr *);
1072 void compose_na(struct dp_packet *,
1073 const struct eth_addr eth_src, const struct eth_addr eth_dst,
1074 const ovs_be32 ipv6_src[4], const ovs_be32 ipv6_dst[4],
1075 ovs_be32 rco_flags);
1076 uint32_t packet_csum_pseudoheader(const struct ip_header *);
1077 void IP_ECN_set_ce(struct dp_packet *pkt, bool is_ipv6);
1079 #endif /* packets.h */