2 * Copyright (c) 2009, 2010, 2011, 2012, 2013 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.
19 #include <arpa/inet.h>
20 #include <sys/socket.h>
21 #include <netinet/in.h>
22 #include <netinet/ip6.h>
24 #include "byte-order.h"
29 #include "dynamic-string.h"
31 #include "ovs-thread.h"
32 #include "unaligned.h"
34 const struct in6_addr in6addr_exact = IN6ADDR_EXACT_INIT;
36 /* Parses 's' as a 16-digit hexadecimal number representing a datapath ID. On
37 * success stores the dpid into '*dpidp' and returns true, on failure stores 0
38 * into '*dpidp' and returns false.
40 * Rejects an all-zeros dpid as invalid. */
42 dpid_from_string(const char *s, uint64_t *dpidp)
44 *dpidp = (strlen(s) == 16 && strspn(s, "0123456789abcdefABCDEF") == 16
45 ? strtoull(s, NULL, 16)
50 /* Returns true if 'ea' is a reserved address, that a bridge must never
51 * forward, false otherwise.
53 * If you change this function's behavior, please update corresponding
54 * documentation in vswitch.xml at the same time. */
56 eth_addr_is_reserved(const uint8_t ea[ETH_ADDR_LEN])
58 struct eth_addr_node {
59 struct hmap_node hmap_node;
63 static struct eth_addr_node nodes[] = {
64 /* STP, IEEE pause frames, and other reserved protocols. */
65 { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000000ULL },
66 { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000001ULL },
67 { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000002ULL },
68 { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000003ULL },
69 { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000004ULL },
70 { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000005ULL },
71 { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000006ULL },
72 { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000007ULL },
73 { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000008ULL },
74 { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000009ULL },
75 { HMAP_NODE_NULL_INITIALIZER, 0x0180c200000aULL },
76 { HMAP_NODE_NULL_INITIALIZER, 0x0180c200000bULL },
77 { HMAP_NODE_NULL_INITIALIZER, 0x0180c200000cULL },
78 { HMAP_NODE_NULL_INITIALIZER, 0x0180c200000dULL },
79 { HMAP_NODE_NULL_INITIALIZER, 0x0180c200000eULL },
80 { HMAP_NODE_NULL_INITIALIZER, 0x0180c200000fULL },
82 /* Extreme protocols. */
83 { HMAP_NODE_NULL_INITIALIZER, 0x00e02b000000ULL }, /* EDP. */
84 { HMAP_NODE_NULL_INITIALIZER, 0x00e02b000004ULL }, /* EAPS. */
85 { HMAP_NODE_NULL_INITIALIZER, 0x00e02b000006ULL }, /* EAPS. */
87 /* Cisco protocols. */
88 { HMAP_NODE_NULL_INITIALIZER, 0x01000c000000ULL }, /* ISL. */
89 { HMAP_NODE_NULL_INITIALIZER, 0x01000cccccccULL }, /* PAgP, UDLD, CDP,
91 { HMAP_NODE_NULL_INITIALIZER, 0x01000ccccccdULL }, /* PVST+. */
92 { HMAP_NODE_NULL_INITIALIZER, 0x01000ccdcdcdULL }, /* STP Uplink Fast,
96 { HMAP_NODE_NULL_INITIALIZER, 0x01000cccccc0ULL },
97 { HMAP_NODE_NULL_INITIALIZER, 0x01000cccccc1ULL },
98 { HMAP_NODE_NULL_INITIALIZER, 0x01000cccccc2ULL },
99 { HMAP_NODE_NULL_INITIALIZER, 0x01000cccccc3ULL },
100 { HMAP_NODE_NULL_INITIALIZER, 0x01000cccccc4ULL },
101 { HMAP_NODE_NULL_INITIALIZER, 0x01000cccccc5ULL },
102 { HMAP_NODE_NULL_INITIALIZER, 0x01000cccccc6ULL },
103 { HMAP_NODE_NULL_INITIALIZER, 0x01000cccccc7ULL },
106 static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
107 struct eth_addr_node *node;
108 static struct hmap addrs;
111 if (ovsthread_once_start(&once)) {
113 for (node = nodes; node < &nodes[ARRAY_SIZE(nodes)]; node++) {
114 hmap_insert(&addrs, &node->hmap_node,
115 hash_2words(node->ea64, node->ea64 >> 32));
117 ovsthread_once_done(&once);
120 ea64 = eth_addr_to_uint64(ea);
121 HMAP_FOR_EACH_IN_BUCKET (node, hmap_node, hash_2words(ea64, ea64 >> 32),
123 if (node->ea64 == ea64) {
131 eth_addr_from_string(const char *s, uint8_t ea[ETH_ADDR_LEN])
133 if (ovs_scan(s, ETH_ADDR_SCAN_FMT, ETH_ADDR_SCAN_ARGS(ea))) {
136 memset(ea, 0, ETH_ADDR_LEN);
141 /* Fills 'b' with a Reverse ARP packet with Ethernet source address 'eth_src'.
142 * This function is used by Open vSwitch to compose packets in cases where
143 * context is important but content doesn't (or shouldn't) matter.
145 * The returned packet has enough headroom to insert an 802.1Q VLAN header if
148 compose_rarp(struct ofpbuf *b, const uint8_t eth_src[ETH_ADDR_LEN])
150 struct eth_header *eth;
151 struct arp_eth_header *arp;
154 ofpbuf_prealloc_tailroom(b, 2 + ETH_HEADER_LEN + VLAN_HEADER_LEN
155 + ARP_ETH_HEADER_LEN);
156 ofpbuf_reserve(b, 2 + VLAN_HEADER_LEN);
157 eth = ofpbuf_put_uninit(b, sizeof *eth);
158 memcpy(eth->eth_dst, eth_addr_broadcast, ETH_ADDR_LEN);
159 memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN);
160 eth->eth_type = htons(ETH_TYPE_RARP);
162 arp = ofpbuf_put_uninit(b, sizeof *arp);
163 arp->ar_hrd = htons(ARP_HRD_ETHERNET);
164 arp->ar_pro = htons(ARP_PRO_IP);
165 arp->ar_hln = sizeof arp->ar_sha;
166 arp->ar_pln = sizeof arp->ar_spa;
167 arp->ar_op = htons(ARP_OP_RARP);
168 memcpy(arp->ar_sha, eth_src, ETH_ADDR_LEN);
169 put_16aligned_be32(&arp->ar_spa, htonl(0));
170 memcpy(arp->ar_tha, eth_src, ETH_ADDR_LEN);
171 put_16aligned_be32(&arp->ar_tpa, htonl(0));
174 /* Insert VLAN header according to given TCI. Packet passed must be Ethernet
175 * packet. Ignores the CFI bit of 'tci' using 0 instead.
177 * Also sets 'packet->l2' to point to the new Ethernet header. */
179 eth_push_vlan(struct ofpbuf *packet, ovs_be16 tpid, ovs_be16 tci)
181 struct eth_header *eh = packet->data;
182 struct vlan_eth_header *veh;
184 /* Insert new 802.1Q header. */
185 struct vlan_eth_header tmp;
186 memcpy(tmp.veth_dst, eh->eth_dst, ETH_ADDR_LEN);
187 memcpy(tmp.veth_src, eh->eth_src, ETH_ADDR_LEN);
188 tmp.veth_type = tpid;
189 tmp.veth_tci = tci & htons(~VLAN_CFI);
190 tmp.veth_next_type = eh->eth_type;
192 veh = ofpbuf_push_uninit(packet, VLAN_HEADER_LEN);
193 memcpy(veh, &tmp, sizeof tmp);
195 packet->l2 = packet->data;
198 /* Removes outermost VLAN header (if any is present) from 'packet'.
200 * 'packet->l2_5' should initially point to 'packet''s outer-most MPLS header
201 * or may be NULL if there are no MPLS headers. */
203 eth_pop_vlan(struct ofpbuf *packet)
205 struct vlan_eth_header *veh = packet->l2;
206 if (packet->size >= sizeof *veh
207 && veh->veth_type == htons(ETH_TYPE_VLAN)) {
208 struct eth_header tmp;
210 memcpy(tmp.eth_dst, veh->veth_dst, ETH_ADDR_LEN);
211 memcpy(tmp.eth_src, veh->veth_src, ETH_ADDR_LEN);
212 tmp.eth_type = veh->veth_next_type;
214 ofpbuf_pull(packet, VLAN_HEADER_LEN);
215 packet->l2 = (char*)packet->l2 + VLAN_HEADER_LEN;
216 memcpy(packet->data, &tmp, sizeof tmp);
220 /* Set ethertype of the packet. */
222 set_ethertype(struct ofpbuf *packet, ovs_be16 eth_type)
224 struct eth_header *eh = packet->data;
226 if (eh->eth_type == htons(ETH_TYPE_VLAN)) {
228 p = ALIGNED_CAST(ovs_be16 *,
229 (char *)(packet->l2_5 ? packet->l2_5 : packet->l3) - 2);
232 eh->eth_type = eth_type;
236 static bool is_mpls(struct ofpbuf *packet)
238 return packet->l2_5 != NULL;
241 /* Set time to live (TTL) of an MPLS label stack entry (LSE). */
243 set_mpls_lse_ttl(ovs_be32 *lse, uint8_t ttl)
245 *lse &= ~htonl(MPLS_TTL_MASK);
246 *lse |= htonl((ttl << MPLS_TTL_SHIFT) & MPLS_TTL_MASK);
249 /* Set traffic class (TC) of an MPLS label stack entry (LSE). */
251 set_mpls_lse_tc(ovs_be32 *lse, uint8_t tc)
253 *lse &= ~htonl(MPLS_TC_MASK);
254 *lse |= htonl((tc << MPLS_TC_SHIFT) & MPLS_TC_MASK);
257 /* Set label of an MPLS label stack entry (LSE). */
259 set_mpls_lse_label(ovs_be32 *lse, ovs_be32 label)
261 *lse &= ~htonl(MPLS_LABEL_MASK);
262 *lse |= htonl((ntohl(label) << MPLS_LABEL_SHIFT) & MPLS_LABEL_MASK);
265 /* Set bottom of stack (BoS) bit of an MPLS label stack entry (LSE). */
267 set_mpls_lse_bos(ovs_be32 *lse, uint8_t bos)
269 *lse &= ~htonl(MPLS_BOS_MASK);
270 *lse |= htonl((bos << MPLS_BOS_SHIFT) & MPLS_BOS_MASK);
273 /* Compose an MPLS label stack entry (LSE) from its components:
274 * label, traffic class (TC), time to live (TTL) and
275 * bottom of stack (BoS) bit. */
277 set_mpls_lse_values(uint8_t ttl, uint8_t tc, uint8_t bos, ovs_be32 label)
279 ovs_be32 lse = htonl(0);
280 set_mpls_lse_ttl(&lse, ttl);
281 set_mpls_lse_tc(&lse, tc);
282 set_mpls_lse_bos(&lse, bos);
283 set_mpls_lse_label(&lse, label);
287 /* Push an new MPLS stack entry onto the MPLS stack and adjust 'packet->l2' and
288 * 'packet->l2_5' accordingly. The new entry will be the outermost entry on
291 * Previous to calling this function, 'packet->l2_5' must be set; if the MPLS
292 * label to be pushed will be the first label in 'packet', then it should be
293 * the same as 'packet->l3'. */
295 push_mpls_lse(struct ofpbuf *packet, struct mpls_hdr *mh)
299 header = ofpbuf_push_uninit(packet, MPLS_HLEN);
300 len = (char *)packet->l2_5 - (char *)packet->l2;
301 memmove(header, packet->l2, len);
302 memcpy(header + len, mh, sizeof *mh);
303 packet->l2 = (char*)packet->l2 - MPLS_HLEN;
304 packet->l2_5 = (char*)packet->l2_5 - MPLS_HLEN;
307 /* Set MPLS label stack entry to outermost MPLS header.*/
309 set_mpls_lse(struct ofpbuf *packet, ovs_be32 mpls_lse)
311 struct mpls_hdr *mh = packet->l2_5;
313 /* Packet type should be MPLS to set label stack entry. */
314 if (is_mpls(packet)) {
315 /* Update mpls label stack entry. */
316 mh->mpls_lse = mpls_lse;
320 /* Push MPLS label stack entry 'lse' onto 'packet' as the the outermost MPLS
321 * header. If 'packet' does not already have any MPLS labels, then its
322 * Ethertype is changed to 'ethtype' (which must be an MPLS Ethertype). */
324 push_mpls(struct ofpbuf *packet, ovs_be16 ethtype, ovs_be32 lse)
328 if (!eth_type_mpls(ethtype)) {
332 if (!is_mpls(packet)) {
333 /* Set ethtype and MPLS label stack entry. */
334 set_ethertype(packet, ethtype);
335 packet->l2_5 = packet->l3;
338 /* Push new MPLS shim header onto packet. */
340 push_mpls_lse(packet, &mh);
343 /* If 'packet' is an MPLS packet, removes its outermost MPLS label stack entry.
344 * If the label that was removed was the only MPLS label, changes 'packet''s
345 * Ethertype to 'ethtype' (which ordinarily should not be an MPLS
348 pop_mpls(struct ofpbuf *packet, ovs_be16 ethtype)
350 struct mpls_hdr *mh = NULL;
352 if (is_mpls(packet)) {
355 len = (char*)packet->l2_5 - (char*)packet->l2;
356 set_ethertype(packet, ethtype);
357 if (mh->mpls_lse & htonl(MPLS_BOS_MASK)) {
360 packet->l2_5 = (char*)packet->l2_5 + MPLS_HLEN;
362 /* Shift the l2 header forward. */
363 memmove((char*)packet->data + MPLS_HLEN, packet->data, len);
364 packet->size -= MPLS_HLEN;
365 packet->data = (char*)packet->data + MPLS_HLEN;
366 packet->l2 = (char*)packet->l2 + MPLS_HLEN;
370 /* Converts hex digits in 'hex' to an Ethernet packet in '*packetp'. The
371 * caller must free '*packetp'. On success, returns NULL. On failure, returns
372 * an error message and stores NULL in '*packetp'.
374 * Aligns the L3 header of '*packetp' on a 32-bit boundary. */
376 eth_from_hex(const char *hex, struct ofpbuf **packetp)
378 struct ofpbuf *packet;
380 /* Use 2 bytes of headroom to 32-bit align the L3 header. */
381 packet = *packetp = ofpbuf_new_with_headroom(strlen(hex) / 2, 2);
383 if (ofpbuf_put_hex(packet, hex, NULL)[0] != '\0') {
384 ofpbuf_delete(packet);
386 return "Trailing garbage in packet data";
389 if (packet->size < ETH_HEADER_LEN) {
390 ofpbuf_delete(packet);
392 return "Packet data too short for Ethernet";
399 eth_format_masked(const uint8_t eth[ETH_ADDR_LEN],
400 const uint8_t mask[ETH_ADDR_LEN], struct ds *s)
402 ds_put_format(s, ETH_ADDR_FMT, ETH_ADDR_ARGS(eth));
403 if (mask && !eth_mask_is_exact(mask)) {
404 ds_put_format(s, "/"ETH_ADDR_FMT, ETH_ADDR_ARGS(mask));
409 eth_addr_bitand(const uint8_t src[ETH_ADDR_LEN],
410 const uint8_t mask[ETH_ADDR_LEN],
411 uint8_t dst[ETH_ADDR_LEN])
415 for (i = 0; i < ETH_ADDR_LEN; i++) {
416 dst[i] = src[i] & mask[i];
420 /* Given the IP netmask 'netmask', returns the number of bits of the IP address
421 * that it specifies, that is, the number of 1-bits in 'netmask'.
423 * If 'netmask' is not a CIDR netmask (see ip_is_cidr()), the return value will
424 * still be in the valid range but isn't otherwise meaningful. */
426 ip_count_cidr_bits(ovs_be32 netmask)
428 return 32 - ctz32(ntohl(netmask));
432 ip_format_masked(ovs_be32 ip, ovs_be32 mask, struct ds *s)
434 ds_put_format(s, IP_FMT, IP_ARGS(ip));
435 if (mask != OVS_BE32_MAX) {
436 if (ip_is_cidr(mask)) {
437 ds_put_format(s, "/%d", ip_count_cidr_bits(mask));
439 ds_put_format(s, "/"IP_FMT, IP_ARGS(mask));
445 /* Stores the string representation of the IPv6 address 'addr' into the
446 * character array 'addr_str', which must be at least INET6_ADDRSTRLEN
449 format_ipv6_addr(char *addr_str, const struct in6_addr *addr)
451 inet_ntop(AF_INET6, addr, addr_str, INET6_ADDRSTRLEN);
455 print_ipv6_addr(struct ds *string, const struct in6_addr *addr)
459 ds_reserve(string, string->length + INET6_ADDRSTRLEN);
461 dst = string->string + string->length;
462 format_ipv6_addr(dst, addr);
463 string->length += strlen(dst);
467 print_ipv6_masked(struct ds *s, const struct in6_addr *addr,
468 const struct in6_addr *mask)
470 print_ipv6_addr(s, addr);
471 if (mask && !ipv6_mask_is_exact(mask)) {
472 if (ipv6_is_cidr(mask)) {
473 int cidr_bits = ipv6_count_cidr_bits(mask);
474 ds_put_format(s, "/%d", cidr_bits);
477 print_ipv6_addr(s, mask);
482 struct in6_addr ipv6_addr_bitand(const struct in6_addr *a,
483 const struct in6_addr *b)
489 for (i=0; i<4; i++) {
490 dst.s6_addr32[i] = a->s6_addr32[i] & b->s6_addr32[i];
493 for (i=0; i<16; i++) {
494 dst.s6_addr[i] = a->s6_addr[i] & b->s6_addr[i];
501 /* Returns an in6_addr consisting of 'mask' high-order 1-bits and 128-N
502 * low-order 0-bits. */
504 ipv6_create_mask(int mask)
506 struct in6_addr netmask;
507 uint8_t *netmaskp = &netmask.s6_addr[0];
509 memset(&netmask, 0, sizeof netmask);
517 *netmaskp = 0xff << (8 - mask);
523 /* Given the IPv6 netmask 'netmask', returns the number of bits of the IPv6
524 * address that it specifies, that is, the number of 1-bits in 'netmask'.
525 * 'netmask' must be a CIDR netmask (see ipv6_is_cidr()).
527 * If 'netmask' is not a CIDR netmask (see ipv6_is_cidr()), the return value
528 * will still be in the valid range but isn't otherwise meaningful. */
530 ipv6_count_cidr_bits(const struct in6_addr *netmask)
534 const uint8_t *netmaskp = &netmask->s6_addr[0];
536 for (i=0; i<16; i++) {
537 if (netmaskp[i] == 0xff) {
542 for(nm = netmaskp[i]; nm; nm <<= 1) {
553 /* Returns true if 'netmask' is a CIDR netmask, that is, if it consists of N
554 * high-order 1-bits and 128-N low-order 0-bits. */
556 ipv6_is_cidr(const struct in6_addr *netmask)
558 const uint8_t *netmaskp = &netmask->s6_addr[0];
561 for (i=0; i<16; i++) {
562 if (netmaskp[i] != 0xff) {
563 uint8_t x = ~netmaskp[i];
578 /* Populates 'b' with an Ethernet II packet headed with the given 'eth_dst',
579 * 'eth_src' and 'eth_type' parameters. A payload of 'size' bytes is allocated
580 * in 'b' and returned. This payload may be populated with appropriate
581 * information by the caller. Sets 'b''s 'l2' and 'l3' pointers to the
582 * Ethernet header and payload respectively. Aligns b->l3 on a 32-bit
585 * The returned packet has enough headroom to insert an 802.1Q VLAN header if
588 eth_compose(struct ofpbuf *b, const uint8_t eth_dst[ETH_ADDR_LEN],
589 const uint8_t eth_src[ETH_ADDR_LEN], uint16_t eth_type,
593 struct eth_header *eth;
597 /* The magic 2 here ensures that the L3 header (when it is added later)
598 * will be 32-bit aligned. */
599 ofpbuf_prealloc_tailroom(b, 2 + ETH_HEADER_LEN + VLAN_HEADER_LEN + size);
600 ofpbuf_reserve(b, 2 + VLAN_HEADER_LEN);
601 eth = ofpbuf_put_uninit(b, ETH_HEADER_LEN);
602 data = ofpbuf_put_uninit(b, size);
604 memcpy(eth->eth_dst, eth_dst, ETH_ADDR_LEN);
605 memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN);
606 eth->eth_type = htons(eth_type);
615 packet_set_ipv4_addr(struct ofpbuf *packet,
616 ovs_16aligned_be32 *addr, ovs_be32 new_addr)
618 struct ip_header *nh = packet->l3;
619 ovs_be32 old_addr = get_16aligned_be32(addr);
621 if (nh->ip_proto == IPPROTO_TCP && packet->l7) {
622 struct tcp_header *th = packet->l4;
624 th->tcp_csum = recalc_csum32(th->tcp_csum, old_addr, new_addr);
625 } else if (nh->ip_proto == IPPROTO_UDP && packet->l7) {
626 struct udp_header *uh = packet->l4;
629 uh->udp_csum = recalc_csum32(uh->udp_csum, old_addr, new_addr);
631 uh->udp_csum = htons(0xffff);
635 nh->ip_csum = recalc_csum32(nh->ip_csum, old_addr, new_addr);
636 put_16aligned_be32(addr, new_addr);
639 /* Returns true, if packet contains at least one routing header where
640 * segements_left > 0.
642 * This function assumes that L3 and L4 markers are set in the packet. */
644 packet_rh_present(struct ofpbuf *packet)
646 const struct ovs_16aligned_ip6_hdr *nh;
650 uint8_t *data = packet->l3;
652 remaining = (uint8_t *)packet->l4 - (uint8_t *)packet->l3;
654 if (remaining < sizeof *nh) {
657 nh = ALIGNED_CAST(struct ovs_16aligned_ip6_hdr *, data);
659 remaining -= sizeof *nh;
660 nexthdr = nh->ip6_nxt;
663 if ((nexthdr != IPPROTO_HOPOPTS)
664 && (nexthdr != IPPROTO_ROUTING)
665 && (nexthdr != IPPROTO_DSTOPTS)
666 && (nexthdr != IPPROTO_AH)
667 && (nexthdr != IPPROTO_FRAGMENT)) {
668 /* It's either a terminal header (e.g., TCP, UDP) or one we
669 * don't understand. In either case, we're done with the
670 * packet, so use it to fill in 'nw_proto'. */
674 /* We only verify that at least 8 bytes of the next header are
675 * available, but many of these headers are longer. Ensure that
676 * accesses within the extension header are within those first 8
677 * bytes. All extension headers are required to be at least 8
683 if (nexthdr == IPPROTO_AH) {
684 /* A standard AH definition isn't available, but the fields
685 * we care about are in the same location as the generic
686 * option header--only the header length is calculated
688 const struct ip6_ext *ext_hdr = (struct ip6_ext *)data;
690 nexthdr = ext_hdr->ip6e_nxt;
691 len = (ext_hdr->ip6e_len + 2) * 4;
692 } else if (nexthdr == IPPROTO_FRAGMENT) {
693 const struct ovs_16aligned_ip6_frag *frag_hdr
694 = ALIGNED_CAST(struct ovs_16aligned_ip6_frag *, data);
696 nexthdr = frag_hdr->ip6f_nxt;
697 len = sizeof *frag_hdr;
698 } else if (nexthdr == IPPROTO_ROUTING) {
699 const struct ip6_rthdr *rh = (struct ip6_rthdr *)data;
701 if (rh->ip6r_segleft > 0) {
705 nexthdr = rh->ip6r_nxt;
706 len = (rh->ip6r_len + 1) * 8;
708 const struct ip6_ext *ext_hdr = (struct ip6_ext *)data;
710 nexthdr = ext_hdr->ip6e_nxt;
711 len = (ext_hdr->ip6e_len + 1) * 8;
714 if (remaining < len) {
725 packet_update_csum128(struct ofpbuf *packet, uint8_t proto,
726 ovs_16aligned_be32 addr[4], const ovs_be32 new_addr[4])
728 if (proto == IPPROTO_TCP && packet->l7) {
729 struct tcp_header *th = packet->l4;
731 th->tcp_csum = recalc_csum128(th->tcp_csum, addr, new_addr);
732 } else if (proto == IPPROTO_UDP && packet->l7) {
733 struct udp_header *uh = packet->l4;
736 uh->udp_csum = recalc_csum128(uh->udp_csum, addr, new_addr);
738 uh->udp_csum = htons(0xffff);
745 packet_set_ipv6_addr(struct ofpbuf *packet, uint8_t proto,
746 ovs_16aligned_be32 addr[4], const ovs_be32 new_addr[4],
747 bool recalculate_csum)
749 if (recalculate_csum) {
750 packet_update_csum128(packet, proto, addr, new_addr);
752 memcpy(addr, new_addr, sizeof(ovs_be32[4]));
756 packet_set_ipv6_flow_label(ovs_16aligned_be32 *flow_label, ovs_be32 flow_key)
758 ovs_be32 old_label = get_16aligned_be32(flow_label);
759 ovs_be32 new_label = (old_label & htonl(~IPV6_LABEL_MASK)) | flow_key;
760 put_16aligned_be32(flow_label, new_label);
764 packet_set_ipv6_tc(ovs_16aligned_be32 *flow_label, uint8_t tc)
766 ovs_be32 old_label = get_16aligned_be32(flow_label);
767 ovs_be32 new_label = (old_label & htonl(0xF00FFFFF)) | htonl(tc << 20);
768 put_16aligned_be32(flow_label, new_label);
771 /* Modifies the IPv4 header fields of 'packet' to be consistent with 'src',
772 * 'dst', 'tos', and 'ttl'. Updates 'packet''s L4 checksums as appropriate.
773 * 'packet' must contain a valid IPv4 packet with correctly populated l[347]
776 packet_set_ipv4(struct ofpbuf *packet, ovs_be32 src, ovs_be32 dst,
777 uint8_t tos, uint8_t ttl)
779 struct ip_header *nh = packet->l3;
781 if (get_16aligned_be32(&nh->ip_src) != src) {
782 packet_set_ipv4_addr(packet, &nh->ip_src, src);
785 if (get_16aligned_be32(&nh->ip_dst) != dst) {
786 packet_set_ipv4_addr(packet, &nh->ip_dst, dst);
789 if (nh->ip_tos != tos) {
790 uint8_t *field = &nh->ip_tos;
792 nh->ip_csum = recalc_csum16(nh->ip_csum, htons((uint16_t) *field),
793 htons((uint16_t) tos));
797 if (nh->ip_ttl != ttl) {
798 uint8_t *field = &nh->ip_ttl;
800 nh->ip_csum = recalc_csum16(nh->ip_csum, htons(*field << 8),
806 /* Modifies the IPv6 header fields of 'packet' to be consistent with 'src',
807 * 'dst', 'traffic class', and 'next hop'. Updates 'packet''s L4 checksums as
808 * appropriate. 'packet' must contain a valid IPv6 packet with correctly
809 * populated l[347] markers. */
811 packet_set_ipv6(struct ofpbuf *packet, uint8_t proto, const ovs_be32 src[4],
812 const ovs_be32 dst[4], uint8_t key_tc, ovs_be32 key_fl,
815 struct ovs_16aligned_ip6_hdr *nh = packet->l3;
817 if (memcmp(&nh->ip6_src, src, sizeof(ovs_be32[4]))) {
818 packet_set_ipv6_addr(packet, proto, nh->ip6_src.be32, src, true);
821 if (memcmp(&nh->ip6_dst, dst, sizeof(ovs_be32[4]))) {
822 packet_set_ipv6_addr(packet, proto, nh->ip6_dst.be32, dst,
823 !packet_rh_present(packet));
826 packet_set_ipv6_tc(&nh->ip6_flow, key_tc);
828 packet_set_ipv6_flow_label(&nh->ip6_flow, key_fl);
830 nh->ip6_hlim = key_hl;
834 packet_set_port(ovs_be16 *port, ovs_be16 new_port, ovs_be16 *csum)
836 if (*port != new_port) {
837 *csum = recalc_csum16(*csum, *port, new_port);
842 /* Sets the TCP source and destination port ('src' and 'dst' respectively) of
843 * the TCP header contained in 'packet'. 'packet' must be a valid TCP packet
844 * with its l4 marker properly populated. */
846 packet_set_tcp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst)
848 struct tcp_header *th = packet->l4;
850 packet_set_port(&th->tcp_src, src, &th->tcp_csum);
851 packet_set_port(&th->tcp_dst, dst, &th->tcp_csum);
854 /* Sets the UDP source and destination port ('src' and 'dst' respectively) of
855 * the UDP header contained in 'packet'. 'packet' must be a valid UDP packet
856 * with its l4 marker properly populated. */
858 packet_set_udp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst)
860 struct udp_header *uh = packet->l4;
863 packet_set_port(&uh->udp_src, src, &uh->udp_csum);
864 packet_set_port(&uh->udp_dst, dst, &uh->udp_csum);
867 uh->udp_csum = htons(0xffff);
875 /* Sets the SCTP source and destination port ('src' and 'dst' respectively) of
876 * the SCTP header contained in 'packet'. 'packet' must be a valid SCTP packet
877 * with its l4 marker properly populated. */
879 packet_set_sctp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst)
881 struct sctp_header *sh = packet->l4;
882 ovs_be32 old_csum, old_correct_csum, new_csum;
883 uint16_t tp_len = packet->size - ((uint8_t*)sh - (uint8_t*)packet->data);
885 old_csum = sh->sctp_csum;
887 old_correct_csum = crc32c(packet->l4, tp_len);
892 new_csum = crc32c(packet->l4, tp_len);
893 sh->sctp_csum = old_csum ^ old_correct_csum ^ new_csum;
896 /* If 'packet' is a TCP packet, returns the TCP flags. Otherwise, returns 0.
898 * 'flow' must be the flow corresponding to 'packet' and 'packet''s header
899 * pointers must be properly initialized (e.g. with flow_extract()). */
901 packet_get_tcp_flags(const struct ofpbuf *packet, const struct flow *flow)
903 if (dl_type_is_ip_any(flow->dl_type) &&
904 flow->nw_proto == IPPROTO_TCP && packet->l7) {
905 const struct tcp_header *tcp = packet->l4;
906 return TCP_FLAGS(tcp->tcp_ctl);
913 packet_tcp_flag_to_string(uint32_t flag)
945 /* Appends a string representation of the TCP flags value 'tcp_flags'
946 * (e.g. obtained via packet_get_tcp_flags() or TCP_FLAGS) to 's', in the
947 * format used by tcpdump. */
949 packet_format_tcp_flags(struct ds *s, uint16_t tcp_flags)
952 ds_put_cstr(s, "none");
956 if (tcp_flags & TCP_SYN) {
959 if (tcp_flags & TCP_FIN) {
962 if (tcp_flags & TCP_PSH) {
965 if (tcp_flags & TCP_RST) {
968 if (tcp_flags & TCP_URG) {
971 if (tcp_flags & TCP_ACK) {
974 if (tcp_flags & TCP_ECE) {
977 if (tcp_flags & TCP_CWR) {
980 if (tcp_flags & TCP_NS) {
983 if (tcp_flags & 0x200) {
984 ds_put_cstr(s, "[200]");
986 if (tcp_flags & 0x400) {
987 ds_put_cstr(s, "[400]");
989 if (tcp_flags & 0x800) {
990 ds_put_cstr(s, "[800]");