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 (sscanf(s, ETH_ADDR_SCAN_FMT, ETH_ADDR_SCAN_ARGS(ea))
134 == ETH_ADDR_SCAN_COUNT) {
137 memset(ea, 0, ETH_ADDR_LEN);
142 /* Fills 'b' with a Reverse ARP packet with Ethernet source address 'eth_src'.
143 * This function is used by Open vSwitch to compose packets in cases where
144 * context is important but content doesn't (or shouldn't) matter.
146 * The returned packet has enough headroom to insert an 802.1Q VLAN header if
149 compose_rarp(struct ofpbuf *b, const uint8_t eth_src[ETH_ADDR_LEN])
151 struct eth_header *eth;
152 struct arp_eth_header *arp;
155 ofpbuf_prealloc_tailroom(b, 2 + ETH_HEADER_LEN + VLAN_HEADER_LEN
156 + ARP_ETH_HEADER_LEN);
157 ofpbuf_reserve(b, 2 + VLAN_HEADER_LEN);
158 eth = ofpbuf_put_uninit(b, sizeof *eth);
159 memcpy(eth->eth_dst, eth_addr_broadcast, ETH_ADDR_LEN);
160 memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN);
161 eth->eth_type = htons(ETH_TYPE_RARP);
163 arp = ofpbuf_put_uninit(b, sizeof *arp);
164 arp->ar_hrd = htons(ARP_HRD_ETHERNET);
165 arp->ar_pro = htons(ARP_PRO_IP);
166 arp->ar_hln = sizeof arp->ar_sha;
167 arp->ar_pln = sizeof arp->ar_spa;
168 arp->ar_op = htons(ARP_OP_RARP);
169 memcpy(arp->ar_sha, eth_src, ETH_ADDR_LEN);
170 put_16aligned_be32(&arp->ar_spa, htonl(0));
171 memcpy(arp->ar_tha, eth_src, ETH_ADDR_LEN);
172 put_16aligned_be32(&arp->ar_tpa, htonl(0));
175 /* Insert VLAN header according to given TCI. Packet passed must be Ethernet
176 * packet. Ignores the CFI bit of 'tci' using 0 instead.
178 * Also sets 'packet->l2' to point to the new Ethernet header. */
180 eth_push_vlan(struct ofpbuf *packet, ovs_be16 tci)
182 struct eth_header *eh = packet->data;
183 struct vlan_eth_header *veh;
185 /* Insert new 802.1Q header. */
186 struct vlan_eth_header tmp;
187 memcpy(tmp.veth_dst, eh->eth_dst, ETH_ADDR_LEN);
188 memcpy(tmp.veth_src, eh->eth_src, ETH_ADDR_LEN);
189 tmp.veth_type = htons(ETH_TYPE_VLAN);
190 tmp.veth_tci = tci & htons(~VLAN_CFI);
191 tmp.veth_next_type = eh->eth_type;
193 veh = ofpbuf_push_uninit(packet, VLAN_HEADER_LEN);
194 memcpy(veh, &tmp, sizeof tmp);
196 packet->l2 = packet->data;
199 /* Removes outermost VLAN header (if any is present) from 'packet'.
201 * 'packet->l2_5' should initially point to 'packet''s outer-most MPLS header
202 * or may be NULL if there are no MPLS headers. */
204 eth_pop_vlan(struct ofpbuf *packet)
206 struct vlan_eth_header *veh = packet->l2;
207 if (packet->size >= sizeof *veh
208 && veh->veth_type == htons(ETH_TYPE_VLAN)) {
209 struct eth_header tmp;
211 memcpy(tmp.eth_dst, veh->veth_dst, ETH_ADDR_LEN);
212 memcpy(tmp.eth_src, veh->veth_src, ETH_ADDR_LEN);
213 tmp.eth_type = veh->veth_next_type;
215 ofpbuf_pull(packet, VLAN_HEADER_LEN);
216 packet->l2 = (char*)packet->l2 + VLAN_HEADER_LEN;
217 memcpy(packet->data, &tmp, sizeof tmp);
221 /* Return depth of mpls stack.
223 * 'packet->l2_5' should initially point to 'packet''s outer-most MPLS header
224 * or may be NULL if there are no MPLS headers. */
226 eth_mpls_depth(const struct ofpbuf *packet)
228 struct mpls_hdr *mh = packet->l2_5;
236 while (packet->size >= ((char *)mh - (char *)packet->data) + sizeof *mh) {
238 if (mh->mpls_lse & htonl(MPLS_BOS_MASK)) {
247 /* Set ethertype of the packet. */
249 set_ethertype(struct ofpbuf *packet, ovs_be16 eth_type)
251 struct eth_header *eh = packet->data;
253 if (eh->eth_type == htons(ETH_TYPE_VLAN)) {
255 p = ALIGNED_CAST(ovs_be16 *,
256 (char *)(packet->l2_5 ? packet->l2_5 : packet->l3) - 2);
259 eh->eth_type = eth_type;
263 static bool is_mpls(struct ofpbuf *packet)
265 return packet->l2_5 != NULL;
268 /* Set time to live (TTL) of an MPLS label stack entry (LSE). */
270 set_mpls_lse_ttl(ovs_be32 *lse, uint8_t ttl)
272 *lse &= ~htonl(MPLS_TTL_MASK);
273 *lse |= htonl((ttl << MPLS_TTL_SHIFT) & MPLS_TTL_MASK);
276 /* Set traffic class (TC) of an MPLS label stack entry (LSE). */
278 set_mpls_lse_tc(ovs_be32 *lse, uint8_t tc)
280 *lse &= ~htonl(MPLS_TC_MASK);
281 *lse |= htonl((tc << MPLS_TC_SHIFT) & MPLS_TC_MASK);
284 /* Set label of an MPLS label stack entry (LSE). */
286 set_mpls_lse_label(ovs_be32 *lse, ovs_be32 label)
288 *lse &= ~htonl(MPLS_LABEL_MASK);
289 *lse |= htonl((ntohl(label) << MPLS_LABEL_SHIFT) & MPLS_LABEL_MASK);
292 /* Set bottom of stack (BoS) bit of an MPLS label stack entry (LSE). */
294 set_mpls_lse_bos(ovs_be32 *lse, uint8_t bos)
296 *lse &= ~htonl(MPLS_BOS_MASK);
297 *lse |= htonl((bos << MPLS_BOS_SHIFT) & MPLS_BOS_MASK);
300 /* Compose an MPLS label stack entry (LSE) from its components:
301 * label, traffic class (TC), time to live (TTL) and
302 * bottom of stack (BoS) bit. */
304 set_mpls_lse_values(uint8_t ttl, uint8_t tc, uint8_t bos, ovs_be32 label)
306 ovs_be32 lse = htonl(0);
307 set_mpls_lse_ttl(&lse, ttl);
308 set_mpls_lse_tc(&lse, tc);
309 set_mpls_lse_bos(&lse, bos);
310 set_mpls_lse_label(&lse, label);
314 /* Push an new MPLS stack entry onto the MPLS stack and adjust 'packet->l2' and
315 * 'packet->l2_5' accordingly. The new entry will be the outermost entry on
318 * Previous to calling this function, 'packet->l2_5' must be set; if the MPLS
319 * label to be pushed will be the first label in 'packet', then it should be
320 * the same as 'packet->l3'. */
322 push_mpls_lse(struct ofpbuf *packet, struct mpls_hdr *mh)
326 header = ofpbuf_push_uninit(packet, MPLS_HLEN);
327 len = (char *)packet->l2_5 - (char *)packet->l2;
328 memmove(header, packet->l2, len);
329 memcpy(header + len, mh, sizeof *mh);
330 packet->l2 = (char*)packet->l2 - MPLS_HLEN;
331 packet->l2_5 = (char*)packet->l2_5 - MPLS_HLEN;
334 /* Set MPLS label stack entry to outermost MPLS header.*/
336 set_mpls_lse(struct ofpbuf *packet, ovs_be32 mpls_lse)
338 struct mpls_hdr *mh = packet->l2_5;
340 /* Packet type should be MPLS to set label stack entry. */
341 if (is_mpls(packet)) {
342 /* Update mpls label stack entry. */
343 mh->mpls_lse = mpls_lse;
347 /* Push MPLS label stack entry 'lse' onto 'packet' as the the outermost MPLS
348 * header. If 'packet' does not already have any MPLS labels, then its
349 * Ethertype is changed to 'ethtype' (which must be an MPLS Ethertype). */
351 push_mpls(struct ofpbuf *packet, ovs_be16 ethtype, ovs_be32 lse)
355 if (!eth_type_mpls(ethtype)) {
359 if (!is_mpls(packet)) {
360 /* Set ethtype and MPLS label stack entry. */
361 set_ethertype(packet, ethtype);
362 packet->l2_5 = packet->l3;
365 /* Push new MPLS shim header onto packet. */
367 push_mpls_lse(packet, &mh);
370 /* If 'packet' is an MPLS packet, removes its outermost MPLS label stack entry.
371 * If the label that was removed was the only MPLS label, changes 'packet''s
372 * Ethertype to 'ethtype' (which ordinarily should not be an MPLS
375 pop_mpls(struct ofpbuf *packet, ovs_be16 ethtype)
377 struct mpls_hdr *mh = NULL;
379 if (is_mpls(packet)) {
382 len = (char*)packet->l2_5 - (char*)packet->l2;
383 set_ethertype(packet, ethtype);
384 if (mh->mpls_lse & htonl(MPLS_BOS_MASK)) {
387 packet->l2_5 = (char*)packet->l2_5 + MPLS_HLEN;
389 /* Shift the l2 header forward. */
390 memmove((char*)packet->data + MPLS_HLEN, packet->data, len);
391 packet->size -= MPLS_HLEN;
392 packet->data = (char*)packet->data + MPLS_HLEN;
393 packet->l2 = (char*)packet->l2 + MPLS_HLEN;
397 /* Converts hex digits in 'hex' to an Ethernet packet in '*packetp'. The
398 * caller must free '*packetp'. On success, returns NULL. On failure, returns
399 * an error message and stores NULL in '*packetp'.
401 * Aligns the L3 header of '*packetp' on a 32-bit boundary. */
403 eth_from_hex(const char *hex, struct ofpbuf **packetp)
405 struct ofpbuf *packet;
407 /* Use 2 bytes of headroom to 32-bit align the L3 header. */
408 packet = *packetp = ofpbuf_new_with_headroom(strlen(hex) / 2, 2);
410 if (ofpbuf_put_hex(packet, hex, NULL)[0] != '\0') {
411 ofpbuf_delete(packet);
413 return "Trailing garbage in packet data";
416 if (packet->size < ETH_HEADER_LEN) {
417 ofpbuf_delete(packet);
419 return "Packet data too short for Ethernet";
426 eth_format_masked(const uint8_t eth[ETH_ADDR_LEN],
427 const uint8_t mask[ETH_ADDR_LEN], struct ds *s)
429 ds_put_format(s, ETH_ADDR_FMT, ETH_ADDR_ARGS(eth));
430 if (mask && !eth_mask_is_exact(mask)) {
431 ds_put_format(s, "/"ETH_ADDR_FMT, ETH_ADDR_ARGS(mask));
436 eth_addr_bitand(const uint8_t src[ETH_ADDR_LEN],
437 const uint8_t mask[ETH_ADDR_LEN],
438 uint8_t dst[ETH_ADDR_LEN])
442 for (i = 0; i < ETH_ADDR_LEN; i++) {
443 dst[i] = src[i] & mask[i];
447 /* Given the IP netmask 'netmask', returns the number of bits of the IP address
448 * that it specifies, that is, the number of 1-bits in 'netmask'.
450 * If 'netmask' is not a CIDR netmask (see ip_is_cidr()), the return value will
451 * still be in the valid range but isn't otherwise meaningful. */
453 ip_count_cidr_bits(ovs_be32 netmask)
455 return 32 - ctz(ntohl(netmask));
459 ip_format_masked(ovs_be32 ip, ovs_be32 mask, struct ds *s)
461 ds_put_format(s, IP_FMT, IP_ARGS(ip));
462 if (mask != OVS_BE32_MAX) {
463 if (ip_is_cidr(mask)) {
464 ds_put_format(s, "/%d", ip_count_cidr_bits(mask));
466 ds_put_format(s, "/"IP_FMT, IP_ARGS(mask));
472 /* Stores the string representation of the IPv6 address 'addr' into the
473 * character array 'addr_str', which must be at least INET6_ADDRSTRLEN
476 format_ipv6_addr(char *addr_str, const struct in6_addr *addr)
478 inet_ntop(AF_INET6, addr, addr_str, INET6_ADDRSTRLEN);
482 print_ipv6_addr(struct ds *string, const struct in6_addr *addr)
486 ds_reserve(string, string->length + INET6_ADDRSTRLEN);
488 dst = string->string + string->length;
489 format_ipv6_addr(dst, addr);
490 string->length += strlen(dst);
494 print_ipv6_masked(struct ds *s, const struct in6_addr *addr,
495 const struct in6_addr *mask)
497 print_ipv6_addr(s, addr);
498 if (mask && !ipv6_mask_is_exact(mask)) {
499 if (ipv6_is_cidr(mask)) {
500 int cidr_bits = ipv6_count_cidr_bits(mask);
501 ds_put_format(s, "/%d", cidr_bits);
504 print_ipv6_addr(s, mask);
509 struct in6_addr ipv6_addr_bitand(const struct in6_addr *a,
510 const struct in6_addr *b)
516 for (i=0; i<4; i++) {
517 dst.s6_addr32[i] = a->s6_addr32[i] & b->s6_addr32[i];
520 for (i=0; i<16; i++) {
521 dst.s6_addr[i] = a->s6_addr[i] & b->s6_addr[i];
528 /* Returns an in6_addr consisting of 'mask' high-order 1-bits and 128-N
529 * low-order 0-bits. */
531 ipv6_create_mask(int mask)
533 struct in6_addr netmask;
534 uint8_t *netmaskp = &netmask.s6_addr[0];
536 memset(&netmask, 0, sizeof netmask);
544 *netmaskp = 0xff << (8 - mask);
550 /* Given the IPv6 netmask 'netmask', returns the number of bits of the IPv6
551 * address that it specifies, that is, the number of 1-bits in 'netmask'.
552 * 'netmask' must be a CIDR netmask (see ipv6_is_cidr()).
554 * If 'netmask' is not a CIDR netmask (see ipv6_is_cidr()), the return value
555 * will still be in the valid range but isn't otherwise meaningful. */
557 ipv6_count_cidr_bits(const struct in6_addr *netmask)
561 const uint8_t *netmaskp = &netmask->s6_addr[0];
563 for (i=0; i<16; i++) {
564 if (netmaskp[i] == 0xff) {
569 for(nm = netmaskp[i]; nm; nm <<= 1) {
580 /* Returns true if 'netmask' is a CIDR netmask, that is, if it consists of N
581 * high-order 1-bits and 128-N low-order 0-bits. */
583 ipv6_is_cidr(const struct in6_addr *netmask)
585 const uint8_t *netmaskp = &netmask->s6_addr[0];
588 for (i=0; i<16; i++) {
589 if (netmaskp[i] != 0xff) {
590 uint8_t x = ~netmaskp[i];
605 /* Populates 'b' with an Ethernet II packet headed with the given 'eth_dst',
606 * 'eth_src' and 'eth_type' parameters. A payload of 'size' bytes is allocated
607 * in 'b' and returned. This payload may be populated with appropriate
608 * information by the caller. Sets 'b''s 'l2' and 'l3' pointers to the
609 * Ethernet header and payload respectively. Aligns b->l3 on a 32-bit
612 * The returned packet has enough headroom to insert an 802.1Q VLAN header if
615 eth_compose(struct ofpbuf *b, const uint8_t eth_dst[ETH_ADDR_LEN],
616 const uint8_t eth_src[ETH_ADDR_LEN], uint16_t eth_type,
620 struct eth_header *eth;
624 /* The magic 2 here ensures that the L3 header (when it is added later)
625 * will be 32-bit aligned. */
626 ofpbuf_prealloc_tailroom(b, 2 + ETH_HEADER_LEN + VLAN_HEADER_LEN + size);
627 ofpbuf_reserve(b, 2 + VLAN_HEADER_LEN);
628 eth = ofpbuf_put_uninit(b, ETH_HEADER_LEN);
629 data = ofpbuf_put_uninit(b, size);
631 memcpy(eth->eth_dst, eth_dst, ETH_ADDR_LEN);
632 memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN);
633 eth->eth_type = htons(eth_type);
642 packet_set_ipv4_addr(struct ofpbuf *packet,
643 ovs_16aligned_be32 *addr, ovs_be32 new_addr)
645 struct ip_header *nh = packet->l3;
646 ovs_be32 old_addr = get_16aligned_be32(addr);
648 if (nh->ip_proto == IPPROTO_TCP && packet->l7) {
649 struct tcp_header *th = packet->l4;
651 th->tcp_csum = recalc_csum32(th->tcp_csum, old_addr, new_addr);
652 } else if (nh->ip_proto == IPPROTO_UDP && packet->l7) {
653 struct udp_header *uh = packet->l4;
656 uh->udp_csum = recalc_csum32(uh->udp_csum, old_addr, new_addr);
658 uh->udp_csum = htons(0xffff);
662 nh->ip_csum = recalc_csum32(nh->ip_csum, old_addr, new_addr);
663 put_16aligned_be32(addr, new_addr);
666 /* Returns true, if packet contains at least one routing header where
667 * segements_left > 0.
669 * This function assumes that L3 and L4 markers are set in the packet. */
671 packet_rh_present(struct ofpbuf *packet)
673 const struct ovs_16aligned_ip6_hdr *nh;
677 uint8_t *data = packet->l3;
679 remaining = (uint8_t *)packet->l4 - (uint8_t *)packet->l3;
681 if (remaining < sizeof *nh) {
684 nh = ALIGNED_CAST(struct ovs_16aligned_ip6_hdr *, data);
686 remaining -= sizeof *nh;
687 nexthdr = nh->ip6_nxt;
690 if ((nexthdr != IPPROTO_HOPOPTS)
691 && (nexthdr != IPPROTO_ROUTING)
692 && (nexthdr != IPPROTO_DSTOPTS)
693 && (nexthdr != IPPROTO_AH)
694 && (nexthdr != IPPROTO_FRAGMENT)) {
695 /* It's either a terminal header (e.g., TCP, UDP) or one we
696 * don't understand. In either case, we're done with the
697 * packet, so use it to fill in 'nw_proto'. */
701 /* We only verify that at least 8 bytes of the next header are
702 * available, but many of these headers are longer. Ensure that
703 * accesses within the extension header are within those first 8
704 * bytes. All extension headers are required to be at least 8
710 if (nexthdr == IPPROTO_AH) {
711 /* A standard AH definition isn't available, but the fields
712 * we care about are in the same location as the generic
713 * option header--only the header length is calculated
715 const struct ip6_ext *ext_hdr = (struct ip6_ext *)data;
717 nexthdr = ext_hdr->ip6e_nxt;
718 len = (ext_hdr->ip6e_len + 2) * 4;
719 } else if (nexthdr == IPPROTO_FRAGMENT) {
720 const struct ovs_16aligned_ip6_frag *frag_hdr
721 = ALIGNED_CAST(struct ovs_16aligned_ip6_frag *, data);
723 nexthdr = frag_hdr->ip6f_nxt;
724 len = sizeof *frag_hdr;
725 } else if (nexthdr == IPPROTO_ROUTING) {
726 const struct ip6_rthdr *rh = (struct ip6_rthdr *)data;
728 if (rh->ip6r_segleft > 0) {
732 nexthdr = rh->ip6r_nxt;
733 len = (rh->ip6r_len + 1) * 8;
735 const struct ip6_ext *ext_hdr = (struct ip6_ext *)data;
737 nexthdr = ext_hdr->ip6e_nxt;
738 len = (ext_hdr->ip6e_len + 1) * 8;
741 if (remaining < len) {
752 packet_update_csum128(struct ofpbuf *packet, uint8_t proto,
753 ovs_16aligned_be32 addr[4], const ovs_be32 new_addr[4])
755 if (proto == IPPROTO_TCP && packet->l7) {
756 struct tcp_header *th = packet->l4;
758 th->tcp_csum = recalc_csum128(th->tcp_csum, addr, new_addr);
759 } else if (proto == IPPROTO_UDP && packet->l7) {
760 struct udp_header *uh = packet->l4;
763 uh->udp_csum = recalc_csum128(uh->udp_csum, addr, new_addr);
765 uh->udp_csum = htons(0xffff);
772 packet_set_ipv6_addr(struct ofpbuf *packet, uint8_t proto,
773 ovs_16aligned_be32 *addr, const ovs_be32 new_addr[4],
774 bool recalculate_csum)
776 if (recalculate_csum) {
777 packet_update_csum128(packet, proto, addr, new_addr);
779 memcpy(addr, new_addr, sizeof(*addr));
783 packet_set_ipv6_flow_label(ovs_16aligned_be32 *flow_label, ovs_be32 flow_key)
785 ovs_be32 old_label = get_16aligned_be32(flow_label);
786 ovs_be32 new_label = (old_label & htonl(~IPV6_LABEL_MASK)) | flow_key;
787 put_16aligned_be32(flow_label, new_label);
791 packet_set_ipv6_tc(ovs_16aligned_be32 *flow_label, uint8_t tc)
793 ovs_be32 old_label = get_16aligned_be32(flow_label);
794 ovs_be32 new_label = (old_label & htonl(0xF00FFFFF)) | htonl(tc << 20);
795 put_16aligned_be32(flow_label, new_label);
798 /* Modifies the IPv4 header fields of 'packet' to be consistent with 'src',
799 * 'dst', 'tos', and 'ttl'. Updates 'packet''s L4 checksums as appropriate.
800 * 'packet' must contain a valid IPv4 packet with correctly populated l[347]
803 packet_set_ipv4(struct ofpbuf *packet, ovs_be32 src, ovs_be32 dst,
804 uint8_t tos, uint8_t ttl)
806 struct ip_header *nh = packet->l3;
808 if (get_16aligned_be32(&nh->ip_src) != src) {
809 packet_set_ipv4_addr(packet, &nh->ip_src, src);
812 if (get_16aligned_be32(&nh->ip_dst) != dst) {
813 packet_set_ipv4_addr(packet, &nh->ip_dst, dst);
816 if (nh->ip_tos != tos) {
817 uint8_t *field = &nh->ip_tos;
819 nh->ip_csum = recalc_csum16(nh->ip_csum, htons((uint16_t) *field),
820 htons((uint16_t) tos));
824 if (nh->ip_ttl != ttl) {
825 uint8_t *field = &nh->ip_ttl;
827 nh->ip_csum = recalc_csum16(nh->ip_csum, htons(*field << 8),
833 /* Modifies the IPv6 header fields of 'packet' to be consistent with 'src',
834 * 'dst', 'traffic class', and 'next hop'. Updates 'packet''s L4 checksums as
835 * appropriate. 'packet' must contain a valid IPv6 packet with correctly
836 * populated l[347] markers. */
838 packet_set_ipv6(struct ofpbuf *packet, uint8_t proto, const ovs_be32 src[4],
839 const ovs_be32 dst[4], uint8_t key_tc, ovs_be32 key_fl,
842 struct ovs_16aligned_ip6_hdr *nh = packet->l3;
844 if (memcmp(&nh->ip6_src, src, sizeof(ovs_be32[4]))) {
845 packet_set_ipv6_addr(packet, proto, nh->ip6_src.be32, src, true);
848 if (memcmp(&nh->ip6_dst, dst, sizeof(ovs_be32[4]))) {
849 packet_set_ipv6_addr(packet, proto, nh->ip6_dst.be32, dst,
850 !packet_rh_present(packet));
853 packet_set_ipv6_tc(&nh->ip6_flow, key_tc);
855 packet_set_ipv6_flow_label(&nh->ip6_flow, key_fl);
857 nh->ip6_hlim = key_hl;
861 packet_set_port(ovs_be16 *port, ovs_be16 new_port, ovs_be16 *csum)
863 if (*port != new_port) {
864 *csum = recalc_csum16(*csum, *port, new_port);
869 /* Sets the TCP source and destination port ('src' and 'dst' respectively) of
870 * the TCP header contained in 'packet'. 'packet' must be a valid TCP packet
871 * with its l4 marker properly populated. */
873 packet_set_tcp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst)
875 struct tcp_header *th = packet->l4;
877 packet_set_port(&th->tcp_src, src, &th->tcp_csum);
878 packet_set_port(&th->tcp_dst, dst, &th->tcp_csum);
881 /* Sets the UDP source and destination port ('src' and 'dst' respectively) of
882 * the UDP header contained in 'packet'. 'packet' must be a valid UDP packet
883 * with its l4 marker properly populated. */
885 packet_set_udp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst)
887 struct udp_header *uh = packet->l4;
890 packet_set_port(&uh->udp_src, src, &uh->udp_csum);
891 packet_set_port(&uh->udp_dst, dst, &uh->udp_csum);
894 uh->udp_csum = htons(0xffff);
902 /* Sets the SCTP source and destination port ('src' and 'dst' respectively) of
903 * the SCTP header contained in 'packet'. 'packet' must be a valid SCTP packet
904 * with its l4 marker properly populated. */
906 packet_set_sctp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst)
908 struct sctp_header *sh = packet->l4;
909 ovs_be32 old_csum, old_correct_csum, new_csum;
910 uint16_t tp_len = packet->size - ((uint8_t*)sh - (uint8_t*)packet->data);
912 old_csum = sh->sctp_csum;
914 old_correct_csum = crc32c(packet->l4, tp_len);
919 new_csum = crc32c(packet->l4, tp_len);
920 sh->sctp_csum = old_csum ^ old_correct_csum ^ new_csum;
923 /* If 'packet' is a TCP packet, returns the TCP flags. Otherwise, returns 0.
925 * 'flow' must be the flow corresponding to 'packet' and 'packet''s header
926 * pointers must be properly initialized (e.g. with flow_extract()). */
928 packet_get_tcp_flags(const struct ofpbuf *packet, const struct flow *flow)
930 if (dl_type_is_ip_any(flow->dl_type) &&
931 flow->nw_proto == IPPROTO_TCP && packet->l7) {
932 const struct tcp_header *tcp = packet->l4;
933 return TCP_FLAGS(tcp->tcp_ctl);
939 /* Appends a string representation of the TCP flags value 'tcp_flags'
940 * (e.g. obtained via packet_get_tcp_flags() or TCP_FLAGS) to 's', in the
941 * format used by tcpdump. */
943 packet_format_tcp_flags(struct ds *s, uint8_t tcp_flags)
946 ds_put_cstr(s, "none");
950 if (tcp_flags & TCP_SYN) {
953 if (tcp_flags & TCP_FIN) {
956 if (tcp_flags & TCP_PSH) {
959 if (tcp_flags & TCP_RST) {
962 if (tcp_flags & TCP_URG) {
965 if (tcp_flags & TCP_ACK) {
968 if (tcp_flags & 0x40) {
969 ds_put_cstr(s, "[40]");
971 if (tcp_flags & 0x80) {
972 ds_put_cstr(s, "[80]");