2 * Copyright (c) 2009, 2010, 2011, 2012 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.
20 #include <arpa/inet.h>
21 #include <sys/socket.h>
22 #include <netinet/in.h>
24 #include "byte-order.h"
27 #include "dynamic-string.h"
30 const struct in6_addr in6addr_exact = IN6ADDR_EXACT_INIT;
32 /* Parses 's' as a 16-digit hexadecimal number representing a datapath ID. On
33 * success stores the dpid into '*dpidp' and returns true, on failure stores 0
34 * into '*dpidp' and returns false.
36 * Rejects an all-zeros dpid as invalid. */
38 dpid_from_string(const char *s, uint64_t *dpidp)
40 *dpidp = (strlen(s) == 16 && strspn(s, "0123456789abcdefABCDEF") == 16
41 ? strtoull(s, NULL, 16)
46 /* Returns true if 'ea' is a reserved multicast address, that a bridge must
47 * never forward, false otherwise. Includes some proprietary vendor protocols
48 * that shouldn't be forwarded as well.
50 * If you change this function's behavior, please update corresponding
51 * documentation in vswitch.xml at the same time. */
53 eth_addr_is_reserved(const uint8_t ea[ETH_ADDR_LEN])
55 struct masked_eth_addr {
56 uint8_t ea[ETH_ADDR_LEN];
57 uint8_t mask[ETH_ADDR_LEN];
60 static struct masked_eth_addr mea[] = {
61 { /* STP, IEEE pause frames, and other reserved protocols. */
62 {0x01, 0x08, 0xc2, 0x00, 0x00, 0x00},
63 {0xff, 0xff, 0xff, 0xff, 0xff, 0xf0}},
66 {0x00, 0x00, 0x5e, 0x00, 0x01, 0x00},
67 {0xff, 0xff, 0xff, 0xff, 0xff, 0x00}},
70 {0x00, 0x00, 0x5e, 0x00, 0x02, 0x00},
71 {0xff, 0xff, 0xff, 0xff, 0xff, 0x00}},
74 {0x00, 0x00, 0x0c, 0x07, 0xac, 0x00},
75 {0xff, 0xff, 0xff, 0xff, 0xff, 0x00}},
78 {0x00, 0x00, 0x0c, 0x9f, 0xf0, 0x00},
79 {0xff, 0xff, 0xff, 0xff, 0xf0, 0x00}},
82 {0x00, 0x07, 0xb4, 0x00, 0x00, 0x00},
83 {0xff, 0xff, 0xff, 0x00, 0x00, 0x00}},
85 { /* Extreme Discovery Protocol. */
86 {0x00, 0xE0, 0x2B, 0x00, 0x00, 0x00},
87 {0xff, 0xff, 0xff, 0xff, 0xf0, 0x00}},
89 { /* Cisco Inter Switch Link. */
90 {0x01, 0x00, 0x0c, 0x00, 0x00, 0x00},
91 {0xff, 0xff, 0xff, 0xff, 0xff, 0xff}},
93 { /* Cisco protocols plus others following the same pattern:
95 * CDP, VTP, DTP, PAgP (01-00-0c-cc-cc-cc)
96 * Spanning Tree PVSTP+ (01-00-0c-cc-cc-cd)
97 * STP Uplink Fast (01-00-0c-cd-cd-cd) */
98 {0x01, 0x00, 0x0c, 0xcc, 0xcc, 0xcc},
99 {0xff, 0xff, 0xff, 0xfe, 0xfe, 0xfe}}};
103 for (i = 0; i < ARRAY_SIZE(mea); i++) {
104 if (eth_addr_equal_except(ea, mea[i].ea, mea[i].mask)) {
112 eth_addr_from_string(const char *s, uint8_t ea[ETH_ADDR_LEN])
114 if (sscanf(s, ETH_ADDR_SCAN_FMT, ETH_ADDR_SCAN_ARGS(ea))
115 == ETH_ADDR_SCAN_COUNT) {
118 memset(ea, 0, ETH_ADDR_LEN);
123 /* Fills 'b' with a Reverse ARP packet with Ethernet source address 'eth_src'.
124 * This function is used by Open vSwitch to compose packets in cases where
125 * context is important but content doesn't (or shouldn't) matter.
127 * The returned packet has enough headroom to insert an 802.1Q VLAN header if
130 compose_rarp(struct ofpbuf *b, const uint8_t eth_src[ETH_ADDR_LEN])
132 struct eth_header *eth;
133 struct rarp_header *rarp;
136 ofpbuf_prealloc_tailroom(b, ETH_HEADER_LEN + VLAN_HEADER_LEN
138 ofpbuf_reserve(b, VLAN_HEADER_LEN);
139 eth = ofpbuf_put_uninit(b, sizeof *eth);
140 memcpy(eth->eth_dst, eth_addr_broadcast, ETH_ADDR_LEN);
141 memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN);
142 eth->eth_type = htons(ETH_TYPE_RARP);
144 rarp = ofpbuf_put_uninit(b, sizeof *rarp);
145 rarp->hw_addr_space = htons(ARP_HTYPE_ETH);
146 rarp->proto_addr_space = htons(ETH_TYPE_IP);
147 rarp->hw_addr_length = ETH_ADDR_LEN;
148 rarp->proto_addr_length = sizeof rarp->src_proto_addr;
149 rarp->opcode = htons(RARP_REQUEST_REVERSE);
150 memcpy(rarp->src_hw_addr, eth_src, ETH_ADDR_LEN);
151 rarp->src_proto_addr = htonl(0);
152 memcpy(rarp->target_hw_addr, eth_src, ETH_ADDR_LEN);
153 rarp->target_proto_addr = htonl(0);
156 /* Insert VLAN header according to given TCI. Packet passed must be Ethernet
157 * packet. Ignores the CFI bit of 'tci' using 0 instead.
159 * Also sets 'packet->l2' to point to the new Ethernet header. */
161 eth_push_vlan(struct ofpbuf *packet, ovs_be16 tci)
163 struct eth_header *eh = packet->data;
164 struct vlan_eth_header *veh;
166 /* Insert new 802.1Q header. */
167 struct vlan_eth_header tmp;
168 memcpy(tmp.veth_dst, eh->eth_dst, ETH_ADDR_LEN);
169 memcpy(tmp.veth_src, eh->eth_src, ETH_ADDR_LEN);
170 tmp.veth_type = htons(ETH_TYPE_VLAN);
171 tmp.veth_tci = tci & htons(~VLAN_CFI);
172 tmp.veth_next_type = eh->eth_type;
174 veh = ofpbuf_push_uninit(packet, VLAN_HEADER_LEN);
175 memcpy(veh, &tmp, sizeof tmp);
177 packet->l2 = packet->data;
180 /* Removes outermost VLAN header (if any is present) from 'packet'.
182 * 'packet->l2' must initially point to 'packet''s Ethernet header. */
184 eth_pop_vlan(struct ofpbuf *packet)
186 struct vlan_eth_header *veh = packet->l2;
187 if (packet->size >= sizeof *veh
188 && veh->veth_type == htons(ETH_TYPE_VLAN)) {
189 struct eth_header tmp;
191 memcpy(tmp.eth_dst, veh->veth_dst, ETH_ADDR_LEN);
192 memcpy(tmp.eth_src, veh->veth_src, ETH_ADDR_LEN);
193 tmp.eth_type = veh->veth_next_type;
195 ofpbuf_pull(packet, VLAN_HEADER_LEN);
196 packet->l2 = (char*)packet->l2 + VLAN_HEADER_LEN;
197 memcpy(packet->data, &tmp, sizeof tmp);
201 /* Converts hex digits in 'hex' to an Ethernet packet in '*packetp'. The
202 * caller must free '*packetp'. On success, returns NULL. On failure, returns
203 * an error message and stores NULL in '*packetp'. */
205 eth_from_hex(const char *hex, struct ofpbuf **packetp)
207 struct ofpbuf *packet;
209 packet = *packetp = ofpbuf_new(strlen(hex) / 2);
211 if (ofpbuf_put_hex(packet, hex, NULL)[0] != '\0') {
212 ofpbuf_delete(packet);
214 return "Trailing garbage in packet data";
217 if (packet->size < ETH_HEADER_LEN) {
218 ofpbuf_delete(packet);
220 return "Packet data too short for Ethernet";
227 eth_format_masked(const uint8_t eth[ETH_ADDR_LEN],
228 const uint8_t mask[ETH_ADDR_LEN], struct ds *s)
230 ds_put_format(s, ETH_ADDR_FMT, ETH_ADDR_ARGS(eth));
231 if (mask && !eth_mask_is_exact(mask)) {
232 ds_put_format(s, "/"ETH_ADDR_FMT, ETH_ADDR_ARGS(mask));
237 eth_addr_bitand(const uint8_t src[ETH_ADDR_LEN],
238 const uint8_t mask[ETH_ADDR_LEN],
239 uint8_t dst[ETH_ADDR_LEN])
243 for (i = 0; i < ETH_ADDR_LEN; i++) {
244 dst[i] = src[i] & mask[i];
248 /* Given the IP netmask 'netmask', returns the number of bits of the IP address
249 * that it specifies, that is, the number of 1-bits in 'netmask'.
251 * If 'netmask' is not a CIDR netmask (see ip_is_cidr()), the return value will
252 * still be in the valid range but isn't otherwise meaningful. */
254 ip_count_cidr_bits(ovs_be32 netmask)
256 return 32 - ctz(ntohl(netmask));
260 ip_format_masked(ovs_be32 ip, ovs_be32 mask, struct ds *s)
262 ds_put_format(s, IP_FMT, IP_ARGS(&ip));
263 if (mask != htonl(UINT32_MAX)) {
264 if (ip_is_cidr(mask)) {
265 ds_put_format(s, "/%d", ip_count_cidr_bits(mask));
267 ds_put_format(s, "/"IP_FMT, IP_ARGS(&mask));
273 /* Stores the string representation of the IPv6 address 'addr' into the
274 * character array 'addr_str', which must be at least INET6_ADDRSTRLEN
277 format_ipv6_addr(char *addr_str, const struct in6_addr *addr)
279 inet_ntop(AF_INET6, addr, addr_str, INET6_ADDRSTRLEN);
283 print_ipv6_addr(struct ds *string, const struct in6_addr *addr)
287 ds_reserve(string, string->length + INET6_ADDRSTRLEN);
289 dst = string->string + string->length;
290 format_ipv6_addr(dst, addr);
291 string->length += strlen(dst);
295 print_ipv6_masked(struct ds *s, const struct in6_addr *addr,
296 const struct in6_addr *mask)
298 print_ipv6_addr(s, addr);
299 if (mask && !ipv6_mask_is_exact(mask)) {
300 if (ipv6_is_cidr(mask)) {
301 int cidr_bits = ipv6_count_cidr_bits(mask);
302 ds_put_format(s, "/%d", cidr_bits);
305 print_ipv6_addr(s, mask);
310 struct in6_addr ipv6_addr_bitand(const struct in6_addr *a,
311 const struct in6_addr *b)
317 for (i=0; i<4; i++) {
318 dst.s6_addr32[i] = a->s6_addr32[i] & b->s6_addr32[i];
321 for (i=0; i<16; i++) {
322 dst.s6_addr[i] = a->s6_addr[i] & b->s6_addr[i];
329 /* Returns an in6_addr consisting of 'mask' high-order 1-bits and 128-N
330 * low-order 0-bits. */
332 ipv6_create_mask(int mask)
334 struct in6_addr netmask;
335 uint8_t *netmaskp = &netmask.s6_addr[0];
337 memset(&netmask, 0, sizeof netmask);
345 *netmaskp = 0xff << (8 - mask);
351 /* Given the IPv6 netmask 'netmask', returns the number of bits of the IPv6
352 * address that it specifies, that is, the number of 1-bits in 'netmask'.
353 * 'netmask' must be a CIDR netmask (see ipv6_is_cidr()).
355 * If 'netmask' is not a CIDR netmask (see ipv6_is_cidr()), the return value
356 * will still be in the valid range but isn't otherwise meaningful. */
358 ipv6_count_cidr_bits(const struct in6_addr *netmask)
362 const uint8_t *netmaskp = &netmask->s6_addr[0];
364 for (i=0; i<16; i++) {
365 if (netmaskp[i] == 0xff) {
370 for(nm = netmaskp[i]; nm; nm <<= 1) {
381 /* Returns true if 'netmask' is a CIDR netmask, that is, if it consists of N
382 * high-order 1-bits and 128-N low-order 0-bits. */
384 ipv6_is_cidr(const struct in6_addr *netmask)
386 const uint8_t *netmaskp = &netmask->s6_addr[0];
389 for (i=0; i<16; i++) {
390 if (netmaskp[i] != 0xff) {
391 uint8_t x = ~netmaskp[i];
406 /* Populates 'b' with an Ethernet II packet headed with the given 'eth_dst',
407 * 'eth_src' and 'eth_type' parameters. A payload of 'size' bytes is allocated
408 * in 'b' and returned. This payload may be populated with appropriate
409 * information by the caller. Sets 'b''s 'l2' and 'l3' pointers to the
410 * Ethernet header and payload respectively.
412 * The returned packet has enough headroom to insert an 802.1Q VLAN header if
415 eth_compose(struct ofpbuf *b, const uint8_t eth_dst[ETH_ADDR_LEN],
416 const uint8_t eth_src[ETH_ADDR_LEN], uint16_t eth_type,
420 struct eth_header *eth;
424 ofpbuf_prealloc_tailroom(b, ETH_HEADER_LEN + VLAN_HEADER_LEN + size);
425 ofpbuf_reserve(b, VLAN_HEADER_LEN);
426 eth = ofpbuf_put_uninit(b, ETH_HEADER_LEN);
427 data = ofpbuf_put_uninit(b, size);
429 memcpy(eth->eth_dst, eth_dst, ETH_ADDR_LEN);
430 memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN);
431 eth->eth_type = htons(eth_type);
440 packet_set_ipv4_addr(struct ofpbuf *packet, ovs_be32 *addr, ovs_be32 new_addr)
442 struct ip_header *nh = packet->l3;
444 if (nh->ip_proto == IPPROTO_TCP && packet->l7) {
445 struct tcp_header *th = packet->l4;
447 th->tcp_csum = recalc_csum32(th->tcp_csum, *addr, new_addr);
448 } else if (nh->ip_proto == IPPROTO_UDP && packet->l7) {
449 struct udp_header *uh = packet->l4;
452 uh->udp_csum = recalc_csum32(uh->udp_csum, *addr, new_addr);
454 uh->udp_csum = htons(0xffff);
458 nh->ip_csum = recalc_csum32(nh->ip_csum, *addr, new_addr);
462 /* Modifies the IPv4 header fields of 'packet' to be consistent with 'src',
463 * 'dst', 'tos', and 'ttl'. Updates 'packet''s L4 checksums as appropriate.
464 * 'packet' must contain a valid IPv4 packet with correctly populated l[347]
467 packet_set_ipv4(struct ofpbuf *packet, ovs_be32 src, ovs_be32 dst,
468 uint8_t tos, uint8_t ttl)
470 struct ip_header *nh = packet->l3;
472 if (nh->ip_src != src) {
473 packet_set_ipv4_addr(packet, &nh->ip_src, src);
476 if (nh->ip_dst != dst) {
477 packet_set_ipv4_addr(packet, &nh->ip_dst, dst);
480 if (nh->ip_tos != tos) {
481 uint8_t *field = &nh->ip_tos;
483 nh->ip_csum = recalc_csum16(nh->ip_csum, htons((uint16_t) *field),
484 htons((uint16_t) tos));
488 if (nh->ip_ttl != ttl) {
489 uint8_t *field = &nh->ip_ttl;
491 nh->ip_csum = recalc_csum16(nh->ip_csum, htons(*field << 8),
498 packet_set_port(ovs_be16 *port, ovs_be16 new_port, ovs_be16 *csum)
500 if (*port != new_port) {
501 *csum = recalc_csum16(*csum, *port, new_port);
506 /* Sets the TCP source and destination port ('src' and 'dst' respectively) of
507 * the TCP header contained in 'packet'. 'packet' must be a valid TCP packet
508 * with its l4 marker properly populated. */
510 packet_set_tcp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst)
512 struct tcp_header *th = packet->l4;
514 packet_set_port(&th->tcp_src, src, &th->tcp_csum);
515 packet_set_port(&th->tcp_dst, dst, &th->tcp_csum);
518 /* Sets the UDP source and destination port ('src' and 'dst' respectively) of
519 * the UDP header contained in 'packet'. 'packet' must be a valid UDP packet
520 * with its l4 marker properly populated. */
522 packet_set_udp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst)
524 struct udp_header *uh = packet->l4;
527 packet_set_port(&uh->udp_src, src, &uh->udp_csum);
528 packet_set_port(&uh->udp_dst, dst, &uh->udp_csum);
531 uh->udp_csum = htons(0xffff);
539 /* If 'packet' is a TCP packet, returns the TCP flags. Otherwise, returns 0.
541 * 'flow' must be the flow corresponding to 'packet' and 'packet''s header
542 * pointers must be properly initialized (e.g. with flow_extract()). */
544 packet_get_tcp_flags(const struct ofpbuf *packet, const struct flow *flow)
546 if ((flow->dl_type == htons(ETH_TYPE_IP) ||
547 flow->dl_type == htons(ETH_TYPE_IPV6)) &&
548 flow->nw_proto == IPPROTO_TCP && packet->l7) {
549 const struct tcp_header *tcp = packet->l4;
550 return TCP_FLAGS(tcp->tcp_ctl);
556 /* Appends a string representation of the TCP flags value 'tcp_flags'
557 * (e.g. obtained via packet_get_tcp_flags() or TCP_FLAGS) to 's', in the
558 * format used by tcpdump. */
560 packet_format_tcp_flags(struct ds *s, uint8_t tcp_flags)
563 ds_put_cstr(s, "none");
567 if (tcp_flags & TCP_SYN) {
570 if (tcp_flags & TCP_FIN) {
573 if (tcp_flags & TCP_PSH) {
576 if (tcp_flags & TCP_RST) {
579 if (tcp_flags & TCP_URG) {
582 if (tcp_flags & TCP_ACK) {
585 if (tcp_flags & 0x40) {
586 ds_put_cstr(s, "[40]");
588 if (tcp_flags & 0x80) {
589 ds_put_cstr(s, "[80]");