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"
33 #include "unaligned.h"
35 const struct in6_addr in6addr_exact = IN6ADDR_EXACT_INIT;
37 /* Parses 's' as a 16-digit hexadecimal number representing a datapath ID. On
38 * success stores the dpid into '*dpidp' and returns true, on failure stores 0
39 * into '*dpidp' and returns false.
41 * Rejects an all-zeros dpid as invalid. */
43 dpid_from_string(const char *s, uint64_t *dpidp)
45 *dpidp = (strlen(s) == 16 && strspn(s, "0123456789abcdefABCDEF") == 16
46 ? strtoull(s, NULL, 16)
51 /* Returns true if 'ea' is a reserved address, that a bridge must never
52 * forward, false otherwise.
54 * If you change this function's behavior, please update corresponding
55 * documentation in vswitch.xml at the same time. */
57 eth_addr_is_reserved(const uint8_t ea[ETH_ADDR_LEN])
59 struct eth_addr_node {
60 struct hmap_node hmap_node;
64 static struct eth_addr_node nodes[] = {
65 /* STP, IEEE pause frames, and other reserved protocols. */
66 { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000000ULL },
67 { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000001ULL },
68 { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000002ULL },
69 { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000003ULL },
70 { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000004ULL },
71 { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000005ULL },
72 { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000006ULL },
73 { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000007ULL },
74 { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000008ULL },
75 { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000009ULL },
76 { HMAP_NODE_NULL_INITIALIZER, 0x0180c200000aULL },
77 { HMAP_NODE_NULL_INITIALIZER, 0x0180c200000bULL },
78 { HMAP_NODE_NULL_INITIALIZER, 0x0180c200000cULL },
79 { HMAP_NODE_NULL_INITIALIZER, 0x0180c200000dULL },
80 { HMAP_NODE_NULL_INITIALIZER, 0x0180c200000eULL },
81 { HMAP_NODE_NULL_INITIALIZER, 0x0180c200000fULL },
83 /* Extreme protocols. */
84 { HMAP_NODE_NULL_INITIALIZER, 0x00e02b000000ULL }, /* EDP. */
85 { HMAP_NODE_NULL_INITIALIZER, 0x00e02b000004ULL }, /* EAPS. */
86 { HMAP_NODE_NULL_INITIALIZER, 0x00e02b000006ULL }, /* EAPS. */
88 /* Cisco protocols. */
89 { HMAP_NODE_NULL_INITIALIZER, 0x01000c000000ULL }, /* ISL. */
90 { HMAP_NODE_NULL_INITIALIZER, 0x01000cccccccULL }, /* PAgP, UDLD, CDP,
92 { HMAP_NODE_NULL_INITIALIZER, 0x01000ccccccdULL }, /* PVST+. */
93 { HMAP_NODE_NULL_INITIALIZER, 0x01000ccdcdcdULL }, /* STP Uplink Fast,
97 { HMAP_NODE_NULL_INITIALIZER, 0x01000cccccc0ULL },
98 { HMAP_NODE_NULL_INITIALIZER, 0x01000cccccc1ULL },
99 { HMAP_NODE_NULL_INITIALIZER, 0x01000cccccc2ULL },
100 { HMAP_NODE_NULL_INITIALIZER, 0x01000cccccc3ULL },
101 { HMAP_NODE_NULL_INITIALIZER, 0x01000cccccc4ULL },
102 { HMAP_NODE_NULL_INITIALIZER, 0x01000cccccc5ULL },
103 { HMAP_NODE_NULL_INITIALIZER, 0x01000cccccc6ULL },
104 { HMAP_NODE_NULL_INITIALIZER, 0x01000cccccc7ULL },
107 static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
108 struct eth_addr_node *node;
109 static struct hmap addrs;
112 if (ovsthread_once_start(&once)) {
114 for (node = nodes; node < &nodes[ARRAY_SIZE(nodes)]; node++) {
115 hmap_insert(&addrs, &node->hmap_node,
116 hash_2words(node->ea64, node->ea64 >> 32));
118 ovsthread_once_done(&once);
121 ea64 = eth_addr_to_uint64(ea);
122 HMAP_FOR_EACH_IN_BUCKET (node, hmap_node, hash_2words(ea64, ea64 >> 32),
124 if (node->ea64 == ea64) {
132 eth_addr_from_string(const char *s, uint8_t ea[ETH_ADDR_LEN])
134 if (ovs_scan(s, ETH_ADDR_SCAN_FMT, ETH_ADDR_SCAN_ARGS(ea))) {
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 tpid, 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 = tpid;
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 /* Set ethertype of the packet. */
223 set_ethertype(struct ofpbuf *packet, ovs_be16 eth_type)
225 struct eth_header *eh = packet->data;
227 if (eh->eth_type == htons(ETH_TYPE_VLAN)) {
229 p = ALIGNED_CAST(ovs_be16 *,
230 (char *)(packet->l2_5 ? packet->l2_5 : packet->l3) - 2);
233 eh->eth_type = eth_type;
237 static bool is_mpls(struct ofpbuf *packet)
239 return packet->l2_5 != NULL;
242 /* Set time to live (TTL) of an MPLS label stack entry (LSE). */
244 set_mpls_lse_ttl(ovs_be32 *lse, uint8_t ttl)
246 *lse &= ~htonl(MPLS_TTL_MASK);
247 *lse |= htonl((ttl << MPLS_TTL_SHIFT) & MPLS_TTL_MASK);
250 /* Set traffic class (TC) of an MPLS label stack entry (LSE). */
252 set_mpls_lse_tc(ovs_be32 *lse, uint8_t tc)
254 *lse &= ~htonl(MPLS_TC_MASK);
255 *lse |= htonl((tc << MPLS_TC_SHIFT) & MPLS_TC_MASK);
258 /* Set label of an MPLS label stack entry (LSE). */
260 set_mpls_lse_label(ovs_be32 *lse, ovs_be32 label)
262 *lse &= ~htonl(MPLS_LABEL_MASK);
263 *lse |= htonl((ntohl(label) << MPLS_LABEL_SHIFT) & MPLS_LABEL_MASK);
266 /* Set bottom of stack (BoS) bit of an MPLS label stack entry (LSE). */
268 set_mpls_lse_bos(ovs_be32 *lse, uint8_t bos)
270 *lse &= ~htonl(MPLS_BOS_MASK);
271 *lse |= htonl((bos << MPLS_BOS_SHIFT) & MPLS_BOS_MASK);
274 /* Compose an MPLS label stack entry (LSE) from its components:
275 * label, traffic class (TC), time to live (TTL) and
276 * bottom of stack (BoS) bit. */
278 set_mpls_lse_values(uint8_t ttl, uint8_t tc, uint8_t bos, ovs_be32 label)
280 ovs_be32 lse = htonl(0);
281 set_mpls_lse_ttl(&lse, ttl);
282 set_mpls_lse_tc(&lse, tc);
283 set_mpls_lse_bos(&lse, bos);
284 set_mpls_lse_label(&lse, label);
288 /* Push an new MPLS stack entry onto the MPLS stack and adjust 'packet->l2' and
289 * 'packet->l2_5' accordingly. The new entry will be the outermost entry on
292 * Previous to calling this function, 'packet->l2_5' must be set; if the MPLS
293 * label to be pushed will be the first label in 'packet', then it should be
294 * the same as 'packet->l3'. */
296 push_mpls_lse(struct ofpbuf *packet, struct mpls_hdr *mh)
300 header = ofpbuf_push_uninit(packet, MPLS_HLEN);
301 len = (char *)packet->l2_5 - (char *)packet->l2;
302 memmove(header, packet->l2, len);
303 memcpy(header + len, mh, sizeof *mh);
304 packet->l2 = (char*)packet->l2 - MPLS_HLEN;
305 packet->l2_5 = (char*)packet->l2_5 - MPLS_HLEN;
308 /* Set MPLS label stack entry to outermost MPLS header.*/
310 set_mpls_lse(struct ofpbuf *packet, ovs_be32 mpls_lse)
312 struct mpls_hdr *mh = packet->l2_5;
314 /* Packet type should be MPLS to set label stack entry. */
315 if (is_mpls(packet)) {
316 /* Update mpls label stack entry. */
317 mh->mpls_lse = mpls_lse;
321 /* Push MPLS label stack entry 'lse' onto 'packet' as the the outermost MPLS
322 * header. If 'packet' does not already have any MPLS labels, then its
323 * Ethertype is changed to 'ethtype' (which must be an MPLS Ethertype). */
325 push_mpls(struct ofpbuf *packet, ovs_be16 ethtype, ovs_be32 lse)
329 if (!eth_type_mpls(ethtype)) {
333 set_ethertype(packet, ethtype);
335 if (!is_mpls(packet)) {
336 /* Set MPLS label stack entry. */
337 packet->l2_5 = packet->l3;
340 /* Push new MPLS shim header onto packet. */
342 push_mpls_lse(packet, &mh);
345 /* If 'packet' is an MPLS packet, removes its outermost MPLS label stack entry.
346 * If the label that was removed was the only MPLS label, changes 'packet''s
347 * Ethertype to 'ethtype' (which ordinarily should not be an MPLS
350 pop_mpls(struct ofpbuf *packet, ovs_be16 ethtype)
352 struct mpls_hdr *mh = NULL;
354 if (is_mpls(packet)) {
357 len = (char*)packet->l2_5 - (char*)packet->l2;
358 set_ethertype(packet, ethtype);
359 if (mh->mpls_lse & htonl(MPLS_BOS_MASK)) {
362 packet->l2_5 = (char*)packet->l2_5 + MPLS_HLEN;
364 /* Shift the l2 header forward. */
365 memmove((char*)packet->data + MPLS_HLEN, packet->data, len);
366 packet->size -= MPLS_HLEN;
367 packet->data = (char*)packet->data + MPLS_HLEN;
368 packet->l2 = (char*)packet->l2 + MPLS_HLEN;
372 /* Converts hex digits in 'hex' to an Ethernet packet in '*packetp'. The
373 * caller must free '*packetp'. On success, returns NULL. On failure, returns
374 * an error message and stores NULL in '*packetp'.
376 * Aligns the L3 header of '*packetp' on a 32-bit boundary. */
378 eth_from_hex(const char *hex, struct ofpbuf **packetp)
380 struct ofpbuf *packet;
382 /* Use 2 bytes of headroom to 32-bit align the L3 header. */
383 packet = *packetp = ofpbuf_new_with_headroom(strlen(hex) / 2, 2);
385 if (ofpbuf_put_hex(packet, hex, NULL)[0] != '\0') {
386 ofpbuf_delete(packet);
388 return "Trailing garbage in packet data";
391 if (packet->size < ETH_HEADER_LEN) {
392 ofpbuf_delete(packet);
394 return "Packet data too short for Ethernet";
401 eth_format_masked(const uint8_t eth[ETH_ADDR_LEN],
402 const uint8_t mask[ETH_ADDR_LEN], struct ds *s)
404 ds_put_format(s, ETH_ADDR_FMT, ETH_ADDR_ARGS(eth));
405 if (mask && !eth_mask_is_exact(mask)) {
406 ds_put_format(s, "/"ETH_ADDR_FMT, ETH_ADDR_ARGS(mask));
411 eth_addr_bitand(const uint8_t src[ETH_ADDR_LEN],
412 const uint8_t mask[ETH_ADDR_LEN],
413 uint8_t dst[ETH_ADDR_LEN])
417 for (i = 0; i < ETH_ADDR_LEN; i++) {
418 dst[i] = src[i] & mask[i];
422 /* Given the IP netmask 'netmask', returns the number of bits of the IP address
423 * that it specifies, that is, the number of 1-bits in 'netmask'.
425 * If 'netmask' is not a CIDR netmask (see ip_is_cidr()), the return value will
426 * still be in the valid range but isn't otherwise meaningful. */
428 ip_count_cidr_bits(ovs_be32 netmask)
430 return 32 - ctz32(ntohl(netmask));
434 ip_format_masked(ovs_be32 ip, ovs_be32 mask, struct ds *s)
436 ds_put_format(s, IP_FMT, IP_ARGS(ip));
437 if (mask != OVS_BE32_MAX) {
438 if (ip_is_cidr(mask)) {
439 ds_put_format(s, "/%d", ip_count_cidr_bits(mask));
441 ds_put_format(s, "/"IP_FMT, IP_ARGS(mask));
447 /* Stores the string representation of the IPv6 address 'addr' into the
448 * character array 'addr_str', which must be at least INET6_ADDRSTRLEN
451 format_ipv6_addr(char *addr_str, const struct in6_addr *addr)
453 inet_ntop(AF_INET6, addr, addr_str, INET6_ADDRSTRLEN);
457 print_ipv6_addr(struct ds *string, const struct in6_addr *addr)
461 ds_reserve(string, string->length + INET6_ADDRSTRLEN);
463 dst = string->string + string->length;
464 format_ipv6_addr(dst, addr);
465 string->length += strlen(dst);
469 print_ipv6_masked(struct ds *s, const struct in6_addr *addr,
470 const struct in6_addr *mask)
472 print_ipv6_addr(s, addr);
473 if (mask && !ipv6_mask_is_exact(mask)) {
474 if (ipv6_is_cidr(mask)) {
475 int cidr_bits = ipv6_count_cidr_bits(mask);
476 ds_put_format(s, "/%d", cidr_bits);
479 print_ipv6_addr(s, mask);
484 struct in6_addr ipv6_addr_bitand(const struct in6_addr *a,
485 const struct in6_addr *b)
491 for (i=0; i<4; i++) {
492 dst.s6_addr32[i] = a->s6_addr32[i] & b->s6_addr32[i];
495 for (i=0; i<16; i++) {
496 dst.s6_addr[i] = a->s6_addr[i] & b->s6_addr[i];
503 /* Returns an in6_addr consisting of 'mask' high-order 1-bits and 128-N
504 * low-order 0-bits. */
506 ipv6_create_mask(int mask)
508 struct in6_addr netmask;
509 uint8_t *netmaskp = &netmask.s6_addr[0];
511 memset(&netmask, 0, sizeof netmask);
519 *netmaskp = 0xff << (8 - mask);
525 /* Given the IPv6 netmask 'netmask', returns the number of bits of the IPv6
526 * address that it specifies, that is, the number of 1-bits in 'netmask'.
527 * 'netmask' must be a CIDR netmask (see ipv6_is_cidr()).
529 * If 'netmask' is not a CIDR netmask (see ipv6_is_cidr()), the return value
530 * will still be in the valid range but isn't otherwise meaningful. */
532 ipv6_count_cidr_bits(const struct in6_addr *netmask)
536 const uint8_t *netmaskp = &netmask->s6_addr[0];
538 for (i=0; i<16; i++) {
539 if (netmaskp[i] == 0xff) {
544 for(nm = netmaskp[i]; nm; nm <<= 1) {
555 /* Returns true if 'netmask' is a CIDR netmask, that is, if it consists of N
556 * high-order 1-bits and 128-N low-order 0-bits. */
558 ipv6_is_cidr(const struct in6_addr *netmask)
560 const uint8_t *netmaskp = &netmask->s6_addr[0];
563 for (i=0; i<16; i++) {
564 if (netmaskp[i] != 0xff) {
565 uint8_t x = ~netmaskp[i];
580 /* Populates 'b' with an Ethernet II packet headed with the given 'eth_dst',
581 * 'eth_src' and 'eth_type' parameters. A payload of 'size' bytes is allocated
582 * in 'b' and returned. This payload may be populated with appropriate
583 * information by the caller. Sets 'b''s 'l2' and 'l3' pointers to the
584 * Ethernet header and payload respectively. Aligns b->l3 on a 32-bit
587 * The returned packet has enough headroom to insert an 802.1Q VLAN header if
590 eth_compose(struct ofpbuf *b, const uint8_t eth_dst[ETH_ADDR_LEN],
591 const uint8_t eth_src[ETH_ADDR_LEN], uint16_t eth_type,
595 struct eth_header *eth;
599 /* The magic 2 here ensures that the L3 header (when it is added later)
600 * will be 32-bit aligned. */
601 ofpbuf_prealloc_tailroom(b, 2 + ETH_HEADER_LEN + VLAN_HEADER_LEN + size);
602 ofpbuf_reserve(b, 2 + VLAN_HEADER_LEN);
603 eth = ofpbuf_put_uninit(b, ETH_HEADER_LEN);
604 data = ofpbuf_put_uninit(b, size);
606 memcpy(eth->eth_dst, eth_dst, ETH_ADDR_LEN);
607 memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN);
608 eth->eth_type = htons(eth_type);
617 packet_set_ipv4_addr(struct ofpbuf *packet,
618 ovs_16aligned_be32 *addr, ovs_be32 new_addr)
620 struct ip_header *nh = packet->l3;
621 ovs_be32 old_addr = get_16aligned_be32(addr);
623 if (nh->ip_proto == IPPROTO_TCP && packet->l7) {
624 struct tcp_header *th = packet->l4;
626 th->tcp_csum = recalc_csum32(th->tcp_csum, old_addr, new_addr);
627 } else if (nh->ip_proto == IPPROTO_UDP && packet->l7) {
628 struct udp_header *uh = packet->l4;
631 uh->udp_csum = recalc_csum32(uh->udp_csum, old_addr, new_addr);
633 uh->udp_csum = htons(0xffff);
637 nh->ip_csum = recalc_csum32(nh->ip_csum, old_addr, new_addr);
638 put_16aligned_be32(addr, new_addr);
641 /* Returns true, if packet contains at least one routing header where
642 * segements_left > 0.
644 * This function assumes that L3 and L4 markers are set in the packet. */
646 packet_rh_present(struct ofpbuf *packet)
648 const struct ovs_16aligned_ip6_hdr *nh;
652 uint8_t *data = packet->l3;
654 remaining = (uint8_t *)packet->l4 - (uint8_t *)packet->l3;
656 if (remaining < sizeof *nh) {
659 nh = ALIGNED_CAST(struct ovs_16aligned_ip6_hdr *, data);
661 remaining -= sizeof *nh;
662 nexthdr = nh->ip6_nxt;
665 if ((nexthdr != IPPROTO_HOPOPTS)
666 && (nexthdr != IPPROTO_ROUTING)
667 && (nexthdr != IPPROTO_DSTOPTS)
668 && (nexthdr != IPPROTO_AH)
669 && (nexthdr != IPPROTO_FRAGMENT)) {
670 /* It's either a terminal header (e.g., TCP, UDP) or one we
671 * don't understand. In either case, we're done with the
672 * packet, so use it to fill in 'nw_proto'. */
676 /* We only verify that at least 8 bytes of the next header are
677 * available, but many of these headers are longer. Ensure that
678 * accesses within the extension header are within those first 8
679 * bytes. All extension headers are required to be at least 8
685 if (nexthdr == IPPROTO_AH) {
686 /* A standard AH definition isn't available, but the fields
687 * we care about are in the same location as the generic
688 * option header--only the header length is calculated
690 const struct ip6_ext *ext_hdr = (struct ip6_ext *)data;
692 nexthdr = ext_hdr->ip6e_nxt;
693 len = (ext_hdr->ip6e_len + 2) * 4;
694 } else if (nexthdr == IPPROTO_FRAGMENT) {
695 const struct ovs_16aligned_ip6_frag *frag_hdr
696 = ALIGNED_CAST(struct ovs_16aligned_ip6_frag *, data);
698 nexthdr = frag_hdr->ip6f_nxt;
699 len = sizeof *frag_hdr;
700 } else if (nexthdr == IPPROTO_ROUTING) {
701 const struct ip6_rthdr *rh = (struct ip6_rthdr *)data;
703 if (rh->ip6r_segleft > 0) {
707 nexthdr = rh->ip6r_nxt;
708 len = (rh->ip6r_len + 1) * 8;
710 const struct ip6_ext *ext_hdr = (struct ip6_ext *)data;
712 nexthdr = ext_hdr->ip6e_nxt;
713 len = (ext_hdr->ip6e_len + 1) * 8;
716 if (remaining < len) {
727 packet_update_csum128(struct ofpbuf *packet, uint8_t proto,
728 ovs_16aligned_be32 addr[4], const ovs_be32 new_addr[4])
730 if (proto == IPPROTO_TCP && packet->l7) {
731 struct tcp_header *th = packet->l4;
733 th->tcp_csum = recalc_csum128(th->tcp_csum, addr, new_addr);
734 } else if (proto == IPPROTO_UDP && packet->l7) {
735 struct udp_header *uh = packet->l4;
738 uh->udp_csum = recalc_csum128(uh->udp_csum, addr, new_addr);
740 uh->udp_csum = htons(0xffff);
747 packet_set_ipv6_addr(struct ofpbuf *packet, uint8_t proto,
748 ovs_16aligned_be32 addr[4], const ovs_be32 new_addr[4],
749 bool recalculate_csum)
751 if (recalculate_csum) {
752 packet_update_csum128(packet, proto, addr, new_addr);
754 memcpy(addr, new_addr, sizeof(ovs_be32[4]));
758 packet_set_ipv6_flow_label(ovs_16aligned_be32 *flow_label, ovs_be32 flow_key)
760 ovs_be32 old_label = get_16aligned_be32(flow_label);
761 ovs_be32 new_label = (old_label & htonl(~IPV6_LABEL_MASK)) | flow_key;
762 put_16aligned_be32(flow_label, new_label);
766 packet_set_ipv6_tc(ovs_16aligned_be32 *flow_label, uint8_t tc)
768 ovs_be32 old_label = get_16aligned_be32(flow_label);
769 ovs_be32 new_label = (old_label & htonl(0xF00FFFFF)) | htonl(tc << 20);
770 put_16aligned_be32(flow_label, new_label);
773 /* Modifies the IPv4 header fields of 'packet' to be consistent with 'src',
774 * 'dst', 'tos', and 'ttl'. Updates 'packet''s L4 checksums as appropriate.
775 * 'packet' must contain a valid IPv4 packet with correctly populated l[347]
778 packet_set_ipv4(struct ofpbuf *packet, ovs_be32 src, ovs_be32 dst,
779 uint8_t tos, uint8_t ttl)
781 struct ip_header *nh = packet->l3;
783 if (get_16aligned_be32(&nh->ip_src) != src) {
784 packet_set_ipv4_addr(packet, &nh->ip_src, src);
787 if (get_16aligned_be32(&nh->ip_dst) != dst) {
788 packet_set_ipv4_addr(packet, &nh->ip_dst, dst);
791 if (nh->ip_tos != tos) {
792 uint8_t *field = &nh->ip_tos;
794 nh->ip_csum = recalc_csum16(nh->ip_csum, htons((uint16_t) *field),
795 htons((uint16_t) tos));
799 if (nh->ip_ttl != ttl) {
800 uint8_t *field = &nh->ip_ttl;
802 nh->ip_csum = recalc_csum16(nh->ip_csum, htons(*field << 8),
808 /* Modifies the IPv6 header fields of 'packet' to be consistent with 'src',
809 * 'dst', 'traffic class', and 'next hop'. Updates 'packet''s L4 checksums as
810 * appropriate. 'packet' must contain a valid IPv6 packet with correctly
811 * populated l[347] markers. */
813 packet_set_ipv6(struct ofpbuf *packet, uint8_t proto, const ovs_be32 src[4],
814 const ovs_be32 dst[4], uint8_t key_tc, ovs_be32 key_fl,
817 struct ovs_16aligned_ip6_hdr *nh = packet->l3;
819 if (memcmp(&nh->ip6_src, src, sizeof(ovs_be32[4]))) {
820 packet_set_ipv6_addr(packet, proto, nh->ip6_src.be32, src, true);
823 if (memcmp(&nh->ip6_dst, dst, sizeof(ovs_be32[4]))) {
824 packet_set_ipv6_addr(packet, proto, nh->ip6_dst.be32, dst,
825 !packet_rh_present(packet));
828 packet_set_ipv6_tc(&nh->ip6_flow, key_tc);
830 packet_set_ipv6_flow_label(&nh->ip6_flow, key_fl);
832 nh->ip6_hlim = key_hl;
836 packet_set_port(ovs_be16 *port, ovs_be16 new_port, ovs_be16 *csum)
838 if (*port != new_port) {
839 *csum = recalc_csum16(*csum, *port, new_port);
844 /* Sets the TCP source and destination port ('src' and 'dst' respectively) of
845 * the TCP header contained in 'packet'. 'packet' must be a valid TCP packet
846 * with its l4 marker properly populated. */
848 packet_set_tcp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst)
850 struct tcp_header *th = packet->l4;
852 packet_set_port(&th->tcp_src, src, &th->tcp_csum);
853 packet_set_port(&th->tcp_dst, dst, &th->tcp_csum);
856 /* Sets the UDP source and destination port ('src' and 'dst' respectively) of
857 * the UDP header contained in 'packet'. 'packet' must be a valid UDP packet
858 * with its l4 marker properly populated. */
860 packet_set_udp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst)
862 struct udp_header *uh = packet->l4;
865 packet_set_port(&uh->udp_src, src, &uh->udp_csum);
866 packet_set_port(&uh->udp_dst, dst, &uh->udp_csum);
869 uh->udp_csum = htons(0xffff);
877 /* Sets the SCTP source and destination port ('src' and 'dst' respectively) of
878 * the SCTP header contained in 'packet'. 'packet' must be a valid SCTP packet
879 * with its l4 marker properly populated. */
881 packet_set_sctp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst)
883 struct sctp_header *sh = packet->l4;
884 ovs_be32 old_csum, old_correct_csum, new_csum;
885 uint16_t tp_len = packet->size - ((uint8_t*)sh - (uint8_t*)packet->data);
887 old_csum = sh->sctp_csum;
889 old_correct_csum = crc32c(packet->l4, tp_len);
894 new_csum = crc32c(packet->l4, tp_len);
895 sh->sctp_csum = old_csum ^ old_correct_csum ^ new_csum;
898 /* If 'packet' is a TCP packet, returns the TCP flags. Otherwise, returns 0.
900 * 'flow' must be the flow corresponding to 'packet' and 'packet''s header
901 * pointers must be properly initialized (e.g. with flow_extract()). */
903 packet_get_tcp_flags(const struct ofpbuf *packet, const struct flow *flow)
905 if (dl_type_is_ip_any(flow->dl_type) &&
906 flow->nw_proto == IPPROTO_TCP && packet->l7) {
907 const struct tcp_header *tcp = packet->l4;
908 return TCP_FLAGS(tcp->tcp_ctl);
915 packet_tcp_flag_to_string(uint32_t flag)
947 /* Appends a string representation of the TCP flags value 'tcp_flags'
948 * (e.g. obtained via packet_get_tcp_flags() or TCP_FLAGS) to 's', in the
949 * format used by tcpdump. */
951 packet_format_tcp_flags(struct ds *s, uint16_t tcp_flags)
954 ds_put_cstr(s, "none");
958 if (tcp_flags & TCP_SYN) {
961 if (tcp_flags & TCP_FIN) {
964 if (tcp_flags & TCP_PSH) {
967 if (tcp_flags & TCP_RST) {
970 if (tcp_flags & TCP_URG) {
973 if (tcp_flags & TCP_ACK) {
976 if (tcp_flags & TCP_ECE) {
979 if (tcp_flags & TCP_CWR) {
982 if (tcp_flags & TCP_NS) {
985 if (tcp_flags & 0x200) {
986 ds_put_cstr(s, "[200]");
988 if (tcp_flags & 0x400) {
989 ds_put_cstr(s, "[400]");
991 if (tcp_flags & 0x800) {
992 ds_put_cstr(s, "[800]");
996 void pkt_metadata_init(struct pkt_metadata *md, const struct flow_tnl *tnl,
997 const uint32_t skb_priority,
998 const uint32_t pkt_mark,
999 const union flow_in_port *in_port)
1002 tnl ? memcpy(&md->tunnel, tnl, sizeof(md->tunnel))
1003 : memset(&md->tunnel, 0, sizeof(md->tunnel));
1005 md->skb_priority = skb_priority;
1006 md->pkt_mark = pkt_mark;
1007 md->in_port.odp_port = in_port ? in_port->odp_port : ODPP_NONE;
1010 void pkt_metadata_from_flow(struct pkt_metadata *md, const struct flow *flow)
1012 pkt_metadata_init(md, &flow->tunnel, flow->skb_priority,
1013 flow->pkt_mark, &flow->in_port);