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, hash_uint64(node->ea64));
117 ovsthread_once_done(&once);
120 ea64 = eth_addr_to_uint64(ea);
121 HMAP_FOR_EACH_IN_BUCKET (node, hmap_node, hash_uint64(ea64), &addrs) {
122 if (node->ea64 == ea64) {
130 eth_addr_from_string(const char *s, uint8_t ea[ETH_ADDR_LEN])
132 if (ovs_scan(s, ETH_ADDR_SCAN_FMT, ETH_ADDR_SCAN_ARGS(ea))) {
135 memset(ea, 0, ETH_ADDR_LEN);
140 /* Fills 'b' with a Reverse ARP packet with Ethernet source address 'eth_src'.
141 * This function is used by Open vSwitch to compose packets in cases where
142 * context is important but content doesn't (or shouldn't) matter.
144 * The returned packet has enough headroom to insert an 802.1Q VLAN header if
147 compose_rarp(struct ofpbuf *b, const uint8_t eth_src[ETH_ADDR_LEN])
149 struct eth_header *eth;
150 struct arp_eth_header *arp;
153 ofpbuf_prealloc_tailroom(b, 2 + ETH_HEADER_LEN + VLAN_HEADER_LEN
154 + ARP_ETH_HEADER_LEN);
155 ofpbuf_reserve(b, 2 + VLAN_HEADER_LEN);
156 eth = ofpbuf_put_uninit(b, sizeof *eth);
157 memcpy(eth->eth_dst, eth_addr_broadcast, ETH_ADDR_LEN);
158 memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN);
159 eth->eth_type = htons(ETH_TYPE_RARP);
161 arp = ofpbuf_put_uninit(b, sizeof *arp);
162 arp->ar_hrd = htons(ARP_HRD_ETHERNET);
163 arp->ar_pro = htons(ARP_PRO_IP);
164 arp->ar_hln = sizeof arp->ar_sha;
165 arp->ar_pln = sizeof arp->ar_spa;
166 arp->ar_op = htons(ARP_OP_RARP);
167 memcpy(arp->ar_sha, eth_src, ETH_ADDR_LEN);
168 put_16aligned_be32(&arp->ar_spa, htonl(0));
169 memcpy(arp->ar_tha, eth_src, ETH_ADDR_LEN);
170 put_16aligned_be32(&arp->ar_tpa, htonl(0));
173 /* Insert VLAN header according to given TCI. Packet passed must be Ethernet
174 * packet. Ignores the CFI bit of 'tci' using 0 instead.
176 * Also sets 'packet->l2' to point to the new Ethernet header and adjusts
177 * the layer offsets accordingly. */
179 eth_push_vlan(struct ofpbuf *packet, ovs_be16 tpid, ovs_be16 tci)
181 struct vlan_eth_header *veh;
183 /* Insert new 802.1Q header. */
184 veh = ofpbuf_resize_l2(packet, VLAN_HEADER_LEN);
185 memmove(veh, (char *)veh + VLAN_HEADER_LEN, 2 * ETH_ADDR_LEN);
186 veh->veth_type = tpid;
187 veh->veth_tci = tci & htons(~VLAN_CFI);
190 /* Removes outermost VLAN header (if any is present) from 'packet'.
192 * 'packet->l2_5' should initially point to 'packet''s outer-most MPLS header
193 * or may be NULL if there are no MPLS headers. */
195 eth_pop_vlan(struct ofpbuf *packet)
197 struct vlan_eth_header *veh = packet->l2;
199 if (packet->size >= sizeof *veh
200 && veh->veth_type == htons(ETH_TYPE_VLAN)) {
202 memmove((char *)veh + VLAN_HEADER_LEN, veh, 2 * ETH_ADDR_LEN);
203 ofpbuf_resize_l2(packet, -VLAN_HEADER_LEN);
207 /* Set ethertype of the packet. */
209 set_ethertype(struct ofpbuf *packet, ovs_be16 eth_type)
211 struct eth_header *eh = packet->l2;
213 if (eh->eth_type == htons(ETH_TYPE_VLAN)) {
215 char *l2_5 = ofpbuf_get_l2_5(packet);
217 p = ALIGNED_CAST(ovs_be16 *,
218 (l2_5 ? l2_5 : (char *)ofpbuf_get_l3(packet)) - 2);
221 eh->eth_type = eth_type;
225 static bool is_mpls(struct ofpbuf *packet)
227 return packet->l2_5_ofs != UINT16_MAX;
230 /* Set time to live (TTL) of an MPLS label stack entry (LSE). */
232 set_mpls_lse_ttl(ovs_be32 *lse, uint8_t ttl)
234 *lse &= ~htonl(MPLS_TTL_MASK);
235 *lse |= htonl((ttl << MPLS_TTL_SHIFT) & MPLS_TTL_MASK);
238 /* Set traffic class (TC) of an MPLS label stack entry (LSE). */
240 set_mpls_lse_tc(ovs_be32 *lse, uint8_t tc)
242 *lse &= ~htonl(MPLS_TC_MASK);
243 *lse |= htonl((tc << MPLS_TC_SHIFT) & MPLS_TC_MASK);
246 /* Set label of an MPLS label stack entry (LSE). */
248 set_mpls_lse_label(ovs_be32 *lse, ovs_be32 label)
250 *lse &= ~htonl(MPLS_LABEL_MASK);
251 *lse |= htonl((ntohl(label) << MPLS_LABEL_SHIFT) & MPLS_LABEL_MASK);
254 /* Set bottom of stack (BoS) bit of an MPLS label stack entry (LSE). */
256 set_mpls_lse_bos(ovs_be32 *lse, uint8_t bos)
258 *lse &= ~htonl(MPLS_BOS_MASK);
259 *lse |= htonl((bos << MPLS_BOS_SHIFT) & MPLS_BOS_MASK);
262 /* Compose an MPLS label stack entry (LSE) from its components:
263 * label, traffic class (TC), time to live (TTL) and
264 * bottom of stack (BoS) bit. */
266 set_mpls_lse_values(uint8_t ttl, uint8_t tc, uint8_t bos, ovs_be32 label)
268 ovs_be32 lse = htonl(0);
269 set_mpls_lse_ttl(&lse, ttl);
270 set_mpls_lse_tc(&lse, tc);
271 set_mpls_lse_bos(&lse, bos);
272 set_mpls_lse_label(&lse, label);
276 /* Set MPLS label stack entry to outermost MPLS header.*/
278 set_mpls_lse(struct ofpbuf *packet, ovs_be32 mpls_lse)
280 /* Packet type should be MPLS to set label stack entry. */
281 if (is_mpls(packet)) {
282 struct mpls_hdr *mh = ofpbuf_get_l2_5(packet);
284 /* Update mpls label stack entry. */
285 mh->mpls_lse = mpls_lse;
289 /* Push MPLS label stack entry 'lse' onto 'packet' as the the outermost MPLS
290 * header. If 'packet' does not already have any MPLS labels, then its
291 * Ethertype is changed to 'ethtype' (which must be an MPLS Ethertype). */
293 push_mpls(struct ofpbuf *packet, ovs_be16 ethtype, ovs_be32 lse)
298 if (!eth_type_mpls(ethtype)) {
302 if (!is_mpls(packet)) {
303 /* Set MPLS label stack offset. */
304 packet->l2_5_ofs = packet->l3_ofs;
307 set_ethertype(packet, ethtype);
309 /* Push new MPLS shim header onto packet. */
310 len = packet->l2_5_ofs;
311 header = ofpbuf_resize_l2_5(packet, MPLS_HLEN);
312 memmove(header, header + MPLS_HLEN, len);
313 memcpy(header + len, &lse, sizeof lse);
316 /* If 'packet' is an MPLS packet, removes its outermost MPLS label stack entry.
317 * If the label that was removed was the only MPLS label, changes 'packet''s
318 * Ethertype to 'ethtype' (which ordinarily should not be an MPLS
321 pop_mpls(struct ofpbuf *packet, ovs_be16 ethtype)
323 if (is_mpls(packet)) {
324 struct mpls_hdr *mh = ofpbuf_get_l2_5(packet);
325 size_t len = packet->l2_5_ofs;
327 set_ethertype(packet, ethtype);
328 if (mh->mpls_lse & htonl(MPLS_BOS_MASK)) {
329 ofpbuf_set_l2_5(packet, NULL);
331 /* Shift the l2 header forward. */
332 memmove((char*)packet->data + MPLS_HLEN, packet->data, len);
333 ofpbuf_resize_l2_5(packet, -MPLS_HLEN);
337 /* Converts hex digits in 'hex' to an Ethernet packet in '*packetp'. The
338 * caller must free '*packetp'. On success, returns NULL. On failure, returns
339 * an error message and stores NULL in '*packetp'.
341 * Aligns the L3 header of '*packetp' on a 32-bit boundary. */
343 eth_from_hex(const char *hex, struct ofpbuf **packetp)
345 struct ofpbuf *packet;
347 /* Use 2 bytes of headroom to 32-bit align the L3 header. */
348 packet = *packetp = ofpbuf_new_with_headroom(strlen(hex) / 2, 2);
350 if (ofpbuf_put_hex(packet, hex, NULL)[0] != '\0') {
351 ofpbuf_delete(packet);
353 return "Trailing garbage in packet data";
356 if (packet->size < ETH_HEADER_LEN) {
357 ofpbuf_delete(packet);
359 return "Packet data too short for Ethernet";
366 eth_format_masked(const uint8_t eth[ETH_ADDR_LEN],
367 const uint8_t mask[ETH_ADDR_LEN], struct ds *s)
369 ds_put_format(s, ETH_ADDR_FMT, ETH_ADDR_ARGS(eth));
370 if (mask && !eth_mask_is_exact(mask)) {
371 ds_put_format(s, "/"ETH_ADDR_FMT, ETH_ADDR_ARGS(mask));
376 eth_addr_bitand(const uint8_t src[ETH_ADDR_LEN],
377 const uint8_t mask[ETH_ADDR_LEN],
378 uint8_t dst[ETH_ADDR_LEN])
382 for (i = 0; i < ETH_ADDR_LEN; i++) {
383 dst[i] = src[i] & mask[i];
387 /* Given the IP netmask 'netmask', returns the number of bits of the IP address
388 * that it specifies, that is, the number of 1-bits in 'netmask'.
390 * If 'netmask' is not a CIDR netmask (see ip_is_cidr()), the return value will
391 * still be in the valid range but isn't otherwise meaningful. */
393 ip_count_cidr_bits(ovs_be32 netmask)
395 return 32 - ctz32(ntohl(netmask));
399 ip_format_masked(ovs_be32 ip, ovs_be32 mask, struct ds *s)
401 ds_put_format(s, IP_FMT, IP_ARGS(ip));
402 if (mask != OVS_BE32_MAX) {
403 if (ip_is_cidr(mask)) {
404 ds_put_format(s, "/%d", ip_count_cidr_bits(mask));
406 ds_put_format(s, "/"IP_FMT, IP_ARGS(mask));
412 /* Stores the string representation of the IPv6 address 'addr' into the
413 * character array 'addr_str', which must be at least INET6_ADDRSTRLEN
416 format_ipv6_addr(char *addr_str, const struct in6_addr *addr)
418 inet_ntop(AF_INET6, addr, addr_str, INET6_ADDRSTRLEN);
422 print_ipv6_addr(struct ds *string, const struct in6_addr *addr)
426 ds_reserve(string, string->length + INET6_ADDRSTRLEN);
428 dst = string->string + string->length;
429 format_ipv6_addr(dst, addr);
430 string->length += strlen(dst);
434 print_ipv6_masked(struct ds *s, const struct in6_addr *addr,
435 const struct in6_addr *mask)
437 print_ipv6_addr(s, addr);
438 if (mask && !ipv6_mask_is_exact(mask)) {
439 if (ipv6_is_cidr(mask)) {
440 int cidr_bits = ipv6_count_cidr_bits(mask);
441 ds_put_format(s, "/%d", cidr_bits);
444 print_ipv6_addr(s, mask);
449 struct in6_addr ipv6_addr_bitand(const struct in6_addr *a,
450 const struct in6_addr *b)
456 for (i=0; i<4; i++) {
457 dst.s6_addr32[i] = a->s6_addr32[i] & b->s6_addr32[i];
460 for (i=0; i<16; i++) {
461 dst.s6_addr[i] = a->s6_addr[i] & b->s6_addr[i];
468 /* Returns an in6_addr consisting of 'mask' high-order 1-bits and 128-N
469 * low-order 0-bits. */
471 ipv6_create_mask(int mask)
473 struct in6_addr netmask;
474 uint8_t *netmaskp = &netmask.s6_addr[0];
476 memset(&netmask, 0, sizeof netmask);
484 *netmaskp = 0xff << (8 - mask);
490 /* Given the IPv6 netmask 'netmask', returns the number of bits of the IPv6
491 * address that it specifies, that is, the number of 1-bits in 'netmask'.
492 * 'netmask' must be a CIDR netmask (see ipv6_is_cidr()).
494 * If 'netmask' is not a CIDR netmask (see ipv6_is_cidr()), the return value
495 * will still be in the valid range but isn't otherwise meaningful. */
497 ipv6_count_cidr_bits(const struct in6_addr *netmask)
501 const uint8_t *netmaskp = &netmask->s6_addr[0];
503 for (i=0; i<16; i++) {
504 if (netmaskp[i] == 0xff) {
509 for(nm = netmaskp[i]; nm; nm <<= 1) {
520 /* Returns true if 'netmask' is a CIDR netmask, that is, if it consists of N
521 * high-order 1-bits and 128-N low-order 0-bits. */
523 ipv6_is_cidr(const struct in6_addr *netmask)
525 const uint8_t *netmaskp = &netmask->s6_addr[0];
528 for (i=0; i<16; i++) {
529 if (netmaskp[i] != 0xff) {
530 uint8_t x = ~netmaskp[i];
545 /* Populates 'b' with an Ethernet II packet headed with the given 'eth_dst',
546 * 'eth_src' and 'eth_type' parameters. A payload of 'size' bytes is allocated
547 * in 'b' and returned. This payload may be populated with appropriate
548 * information by the caller. Sets 'b''s 'l2' pointer and 'l3' offset to the
549 * Ethernet header and payload respectively. Aligns b->l3 on a 32-bit
552 * The returned packet has enough headroom to insert an 802.1Q VLAN header if
555 eth_compose(struct ofpbuf *b, const uint8_t eth_dst[ETH_ADDR_LEN],
556 const uint8_t eth_src[ETH_ADDR_LEN], uint16_t eth_type,
560 struct eth_header *eth;
564 /* The magic 2 here ensures that the L3 header (when it is added later)
565 * will be 32-bit aligned. */
566 ofpbuf_prealloc_tailroom(b, 2 + ETH_HEADER_LEN + VLAN_HEADER_LEN + size);
567 ofpbuf_reserve(b, 2 + VLAN_HEADER_LEN);
568 eth = ofpbuf_put_uninit(b, ETH_HEADER_LEN);
569 data = ofpbuf_put_uninit(b, size);
571 memcpy(eth->eth_dst, eth_dst, ETH_ADDR_LEN);
572 memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN);
573 eth->eth_type = htons(eth_type);
576 ofpbuf_set_l3(b, data);
582 packet_set_ipv4_addr(struct ofpbuf *packet,
583 ovs_16aligned_be32 *addr, ovs_be32 new_addr)
585 struct ip_header *nh = ofpbuf_get_l3(packet);
586 ovs_be32 old_addr = get_16aligned_be32(addr);
587 size_t l4_size = ofpbuf_get_l4_size(packet);
589 if (nh->ip_proto == IPPROTO_TCP && l4_size >= TCP_HEADER_LEN) {
590 struct tcp_header *th = ofpbuf_get_l4(packet);
592 th->tcp_csum = recalc_csum32(th->tcp_csum, old_addr, new_addr);
593 } else if (nh->ip_proto == IPPROTO_UDP && l4_size >= UDP_HEADER_LEN ) {
594 struct udp_header *uh = ofpbuf_get_l4(packet);
597 uh->udp_csum = recalc_csum32(uh->udp_csum, old_addr, new_addr);
599 uh->udp_csum = htons(0xffff);
603 nh->ip_csum = recalc_csum32(nh->ip_csum, old_addr, new_addr);
604 put_16aligned_be32(addr, new_addr);
607 /* Returns true, if packet contains at least one routing header where
608 * segements_left > 0.
610 * This function assumes that L3 and L4 offsets are set in the packet. */
612 packet_rh_present(struct ofpbuf *packet)
614 const struct ovs_16aligned_ip6_hdr *nh;
618 uint8_t *data = ofpbuf_get_l3(packet);
620 remaining = packet->l4_ofs - packet->l3_ofs;
622 if (remaining < sizeof *nh) {
625 nh = ALIGNED_CAST(struct ovs_16aligned_ip6_hdr *, data);
627 remaining -= sizeof *nh;
628 nexthdr = nh->ip6_nxt;
631 if ((nexthdr != IPPROTO_HOPOPTS)
632 && (nexthdr != IPPROTO_ROUTING)
633 && (nexthdr != IPPROTO_DSTOPTS)
634 && (nexthdr != IPPROTO_AH)
635 && (nexthdr != IPPROTO_FRAGMENT)) {
636 /* It's either a terminal header (e.g., TCP, UDP) or one we
637 * don't understand. In either case, we're done with the
638 * packet, so use it to fill in 'nw_proto'. */
642 /* We only verify that at least 8 bytes of the next header are
643 * available, but many of these headers are longer. Ensure that
644 * accesses within the extension header are within those first 8
645 * bytes. All extension headers are required to be at least 8
651 if (nexthdr == IPPROTO_AH) {
652 /* A standard AH definition isn't available, but the fields
653 * we care about are in the same location as the generic
654 * option header--only the header length is calculated
656 const struct ip6_ext *ext_hdr = (struct ip6_ext *)data;
658 nexthdr = ext_hdr->ip6e_nxt;
659 len = (ext_hdr->ip6e_len + 2) * 4;
660 } else if (nexthdr == IPPROTO_FRAGMENT) {
661 const struct ovs_16aligned_ip6_frag *frag_hdr
662 = ALIGNED_CAST(struct ovs_16aligned_ip6_frag *, data);
664 nexthdr = frag_hdr->ip6f_nxt;
665 len = sizeof *frag_hdr;
666 } else if (nexthdr == IPPROTO_ROUTING) {
667 const struct ip6_rthdr *rh = (struct ip6_rthdr *)data;
669 if (rh->ip6r_segleft > 0) {
673 nexthdr = rh->ip6r_nxt;
674 len = (rh->ip6r_len + 1) * 8;
676 const struct ip6_ext *ext_hdr = (struct ip6_ext *)data;
678 nexthdr = ext_hdr->ip6e_nxt;
679 len = (ext_hdr->ip6e_len + 1) * 8;
682 if (remaining < len) {
693 packet_update_csum128(struct ofpbuf *packet, uint8_t proto,
694 ovs_16aligned_be32 addr[4], const ovs_be32 new_addr[4])
696 size_t l4_size = ofpbuf_get_l4_size(packet);
698 if (proto == IPPROTO_TCP && l4_size >= TCP_HEADER_LEN) {
699 struct tcp_header *th = ofpbuf_get_l4(packet);
701 th->tcp_csum = recalc_csum128(th->tcp_csum, addr, new_addr);
702 } else if (proto == IPPROTO_UDP && l4_size >= UDP_HEADER_LEN) {
703 struct udp_header *uh = ofpbuf_get_l4(packet);
706 uh->udp_csum = recalc_csum128(uh->udp_csum, addr, new_addr);
708 uh->udp_csum = htons(0xffff);
715 packet_set_ipv6_addr(struct ofpbuf *packet, uint8_t proto,
716 ovs_16aligned_be32 addr[4], const ovs_be32 new_addr[4],
717 bool recalculate_csum)
719 if (recalculate_csum) {
720 packet_update_csum128(packet, proto, addr, new_addr);
722 memcpy(addr, new_addr, sizeof(ovs_be32[4]));
726 packet_set_ipv6_flow_label(ovs_16aligned_be32 *flow_label, ovs_be32 flow_key)
728 ovs_be32 old_label = get_16aligned_be32(flow_label);
729 ovs_be32 new_label = (old_label & htonl(~IPV6_LABEL_MASK)) | flow_key;
730 put_16aligned_be32(flow_label, new_label);
734 packet_set_ipv6_tc(ovs_16aligned_be32 *flow_label, uint8_t tc)
736 ovs_be32 old_label = get_16aligned_be32(flow_label);
737 ovs_be32 new_label = (old_label & htonl(0xF00FFFFF)) | htonl(tc << 20);
738 put_16aligned_be32(flow_label, new_label);
741 /* Modifies the IPv4 header fields of 'packet' to be consistent with 'src',
742 * 'dst', 'tos', and 'ttl'. Updates 'packet''s L4 checksums as appropriate.
743 * 'packet' must contain a valid IPv4 packet with correctly populated l[347]
746 packet_set_ipv4(struct ofpbuf *packet, ovs_be32 src, ovs_be32 dst,
747 uint8_t tos, uint8_t ttl)
749 struct ip_header *nh = ofpbuf_get_l3(packet);
751 if (get_16aligned_be32(&nh->ip_src) != src) {
752 packet_set_ipv4_addr(packet, &nh->ip_src, src);
755 if (get_16aligned_be32(&nh->ip_dst) != dst) {
756 packet_set_ipv4_addr(packet, &nh->ip_dst, dst);
759 if (nh->ip_tos != tos) {
760 uint8_t *field = &nh->ip_tos;
762 nh->ip_csum = recalc_csum16(nh->ip_csum, htons((uint16_t) *field),
763 htons((uint16_t) tos));
767 if (nh->ip_ttl != ttl) {
768 uint8_t *field = &nh->ip_ttl;
770 nh->ip_csum = recalc_csum16(nh->ip_csum, htons(*field << 8),
776 /* Modifies the IPv6 header fields of 'packet' to be consistent with 'src',
777 * 'dst', 'traffic class', and 'next hop'. Updates 'packet''s L4 checksums as
778 * appropriate. 'packet' must contain a valid IPv6 packet with correctly
779 * populated l[34] offsets. */
781 packet_set_ipv6(struct ofpbuf *packet, uint8_t proto, const ovs_be32 src[4],
782 const ovs_be32 dst[4], uint8_t key_tc, ovs_be32 key_fl,
785 struct ovs_16aligned_ip6_hdr *nh = ofpbuf_get_l3(packet);
787 if (memcmp(&nh->ip6_src, src, sizeof(ovs_be32[4]))) {
788 packet_set_ipv6_addr(packet, proto, nh->ip6_src.be32, src, true);
791 if (memcmp(&nh->ip6_dst, dst, sizeof(ovs_be32[4]))) {
792 packet_set_ipv6_addr(packet, proto, nh->ip6_dst.be32, dst,
793 !packet_rh_present(packet));
796 packet_set_ipv6_tc(&nh->ip6_flow, key_tc);
798 packet_set_ipv6_flow_label(&nh->ip6_flow, key_fl);
800 nh->ip6_hlim = key_hl;
804 packet_set_port(ovs_be16 *port, ovs_be16 new_port, ovs_be16 *csum)
806 if (*port != new_port) {
807 *csum = recalc_csum16(*csum, *port, new_port);
812 /* Sets the TCP source and destination port ('src' and 'dst' respectively) of
813 * the TCP header contained in 'packet'. 'packet' must be a valid TCP packet
814 * with its l4 offset properly populated. */
816 packet_set_tcp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst)
818 struct tcp_header *th = ofpbuf_get_l4(packet);
820 packet_set_port(&th->tcp_src, src, &th->tcp_csum);
821 packet_set_port(&th->tcp_dst, dst, &th->tcp_csum);
824 /* Sets the UDP source and destination port ('src' and 'dst' respectively) of
825 * the UDP header contained in 'packet'. 'packet' must be a valid UDP packet
826 * with its l4 offset properly populated. */
828 packet_set_udp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst)
830 struct udp_header *uh = ofpbuf_get_l4(packet);
833 packet_set_port(&uh->udp_src, src, &uh->udp_csum);
834 packet_set_port(&uh->udp_dst, dst, &uh->udp_csum);
837 uh->udp_csum = htons(0xffff);
845 /* Sets the SCTP source and destination port ('src' and 'dst' respectively) of
846 * the SCTP header contained in 'packet'. 'packet' must be a valid SCTP packet
847 * with its l4 offset properly populated. */
849 packet_set_sctp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst)
851 struct sctp_header *sh = ofpbuf_get_l4(packet);
852 ovs_be32 old_csum, old_correct_csum, new_csum;
853 uint16_t tp_len = packet->size - ((uint8_t*)sh - (uint8_t*)packet->data);
855 old_csum = sh->sctp_csum;
857 old_correct_csum = crc32c((void *)sh, tp_len);
862 new_csum = crc32c((void *)sh, tp_len);
863 sh->sctp_csum = old_csum ^ old_correct_csum ^ new_csum;
867 packet_tcp_flag_to_string(uint32_t flag)
899 /* Appends a string representation of the TCP flags value 'tcp_flags'
900 * (e.g. from struct flow.tcp_flags or obtained via TCP_FLAGS) to 's', in the
901 * format used by tcpdump. */
903 packet_format_tcp_flags(struct ds *s, uint16_t tcp_flags)
906 ds_put_cstr(s, "none");
910 if (tcp_flags & TCP_SYN) {
913 if (tcp_flags & TCP_FIN) {
916 if (tcp_flags & TCP_PSH) {
919 if (tcp_flags & TCP_RST) {
922 if (tcp_flags & TCP_URG) {
925 if (tcp_flags & TCP_ACK) {
928 if (tcp_flags & TCP_ECE) {
931 if (tcp_flags & TCP_CWR) {
934 if (tcp_flags & TCP_NS) {
937 if (tcp_flags & 0x200) {
938 ds_put_cstr(s, "[200]");
940 if (tcp_flags & 0x400) {
941 ds_put_cstr(s, "[400]");
943 if (tcp_flags & 0x800) {
944 ds_put_cstr(s, "[800]");