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}},
65 { /* Cisco Inter Switch Link. */
66 {0x01, 0x00, 0x0c, 0x00, 0x00, 0x00},
67 {0xff, 0xff, 0xff, 0xff, 0xff, 0xff}},
69 { /* Cisco protocols plus others following the same pattern:
71 * CDP, VTP, DTP, PAgP (01-00-0c-cc-cc-cc)
72 * Spanning Tree PVSTP+ (01-00-0c-cc-cc-cd)
73 * STP Uplink Fast (01-00-0c-cd-cd-cd) */
74 {0x01, 0x00, 0x0c, 0xcc, 0xcc, 0xcc},
75 {0xff, 0xff, 0xff, 0xfe, 0xfe, 0xfe}}};
79 for (i = 0; i < ARRAY_SIZE(mea); i++) {
80 if (eth_addr_equal_except(ea, mea[i].ea, mea[i].mask)) {
88 eth_addr_from_string(const char *s, uint8_t ea[ETH_ADDR_LEN])
90 if (sscanf(s, ETH_ADDR_SCAN_FMT, ETH_ADDR_SCAN_ARGS(ea))
91 == ETH_ADDR_SCAN_COUNT) {
94 memset(ea, 0, ETH_ADDR_LEN);
99 /* Fills 'b' with an 802.2 SNAP packet with Ethernet source address 'eth_src',
100 * the Nicira OUI as SNAP organization and 'snap_type' as SNAP type. The text
101 * string in 'tag' is enclosed as the packet payload.
103 * This function is used by Open vSwitch to compose packets in cases where
104 * context is important but content doesn't (or shouldn't) matter. For this
105 * purpose, 'snap_type' should be a random number and 'tag' should be an
106 * English phrase that explains the purpose of the packet. (The English phrase
107 * gives hapless admins running Wireshark the opportunity to figure out what's
110 compose_benign_packet(struct ofpbuf *b, const char *tag, uint16_t snap_type,
111 const uint8_t eth_src[ETH_ADDR_LEN])
113 size_t tag_size = strlen(tag) + 1;
116 payload = snap_compose(b, eth_addr_broadcast, eth_src, 0x002320, snap_type,
117 tag_size + ETH_ADDR_LEN);
118 memcpy(payload, tag, tag_size);
119 memcpy(payload + tag_size, eth_src, ETH_ADDR_LEN);
122 /* Insert VLAN header according to given TCI. Packet passed must be Ethernet
123 * packet. Ignores the CFI bit of 'tci' using 0 instead.
125 * Also sets 'packet->l2' to point to the new Ethernet header. */
127 eth_push_vlan(struct ofpbuf *packet, ovs_be16 tci)
129 struct eth_header *eh = packet->data;
130 struct vlan_eth_header *veh;
132 /* Insert new 802.1Q header. */
133 struct vlan_eth_header tmp;
134 memcpy(tmp.veth_dst, eh->eth_dst, ETH_ADDR_LEN);
135 memcpy(tmp.veth_src, eh->eth_src, ETH_ADDR_LEN);
136 tmp.veth_type = htons(ETH_TYPE_VLAN);
137 tmp.veth_tci = tci & htons(~VLAN_CFI);
138 tmp.veth_next_type = eh->eth_type;
140 veh = ofpbuf_push_uninit(packet, VLAN_HEADER_LEN);
141 memcpy(veh, &tmp, sizeof tmp);
143 packet->l2 = packet->data;
146 /* Removes outermost VLAN header (if any is present) from 'packet'.
148 * 'packet->l2' must initially point to 'packet''s Ethernet header. */
150 eth_pop_vlan(struct ofpbuf *packet)
152 struct vlan_eth_header *veh = packet->l2;
153 if (packet->size >= sizeof *veh
154 && veh->veth_type == htons(ETH_TYPE_VLAN)) {
155 struct eth_header tmp;
157 memcpy(tmp.eth_dst, veh->veth_dst, ETH_ADDR_LEN);
158 memcpy(tmp.eth_src, veh->veth_src, ETH_ADDR_LEN);
159 tmp.eth_type = veh->veth_next_type;
161 ofpbuf_pull(packet, VLAN_HEADER_LEN);
162 packet->l2 = (char*)packet->l2 + VLAN_HEADER_LEN;
163 memcpy(packet->data, &tmp, sizeof tmp);
167 /* Converts hex digits in 'hex' to an Ethernet packet in '*packetp'. The
168 * caller must free '*packetp'. On success, returns NULL. On failure, returns
169 * an error message and stores NULL in '*packetp'. */
171 eth_from_hex(const char *hex, struct ofpbuf **packetp)
173 struct ofpbuf *packet;
175 packet = *packetp = ofpbuf_new(strlen(hex) / 2);
177 if (ofpbuf_put_hex(packet, hex, NULL)[0] != '\0') {
178 ofpbuf_delete(packet);
180 return "Trailing garbage in packet data";
183 if (packet->size < ETH_HEADER_LEN) {
184 ofpbuf_delete(packet);
186 return "Packet data too short for Ethernet";
193 eth_format_masked(const uint8_t eth[ETH_ADDR_LEN],
194 const uint8_t mask[ETH_ADDR_LEN], struct ds *s)
196 ds_put_format(s, ETH_ADDR_FMT, ETH_ADDR_ARGS(eth));
197 if (mask && !eth_mask_is_exact(mask)) {
198 ds_put_format(s, "/"ETH_ADDR_FMT, ETH_ADDR_ARGS(mask));
203 eth_addr_bitand(const uint8_t src[ETH_ADDR_LEN],
204 const uint8_t mask[ETH_ADDR_LEN],
205 uint8_t dst[ETH_ADDR_LEN])
209 for (i = 0; i < ETH_ADDR_LEN; i++) {
210 dst[i] = src[i] & mask[i];
214 /* Given the IP netmask 'netmask', returns the number of bits of the IP address
215 * that it specifies, that is, the number of 1-bits in 'netmask'. 'netmask'
216 * must be a CIDR netmask (see ip_is_cidr()). */
218 ip_count_cidr_bits(ovs_be32 netmask)
220 assert(ip_is_cidr(netmask));
221 return 32 - ctz(ntohl(netmask));
225 ip_format_masked(ovs_be32 ip, ovs_be32 mask, struct ds *s)
227 ds_put_format(s, IP_FMT, IP_ARGS(&ip));
228 if (mask != htonl(UINT32_MAX)) {
229 if (ip_is_cidr(mask)) {
230 ds_put_format(s, "/%d", ip_count_cidr_bits(mask));
232 ds_put_format(s, "/"IP_FMT, IP_ARGS(&mask));
238 /* Stores the string representation of the IPv6 address 'addr' into the
239 * character array 'addr_str', which must be at least INET6_ADDRSTRLEN
242 format_ipv6_addr(char *addr_str, const struct in6_addr *addr)
244 inet_ntop(AF_INET6, addr, addr_str, INET6_ADDRSTRLEN);
248 print_ipv6_addr(struct ds *string, const struct in6_addr *addr)
252 ds_reserve(string, string->length + INET6_ADDRSTRLEN);
254 dst = string->string + string->length;
255 format_ipv6_addr(dst, addr);
256 string->length += strlen(dst);
260 print_ipv6_masked(struct ds *s, const struct in6_addr *addr,
261 const struct in6_addr *mask)
263 print_ipv6_addr(s, addr);
264 if (mask && !ipv6_mask_is_exact(mask)) {
265 if (ipv6_is_cidr(mask)) {
266 int cidr_bits = ipv6_count_cidr_bits(mask);
267 ds_put_format(s, "/%d", cidr_bits);
270 print_ipv6_addr(s, mask);
275 struct in6_addr ipv6_addr_bitand(const struct in6_addr *a,
276 const struct in6_addr *b)
282 for (i=0; i<4; i++) {
283 dst.s6_addr32[i] = a->s6_addr32[i] & b->s6_addr32[i];
286 for (i=0; i<16; i++) {
287 dst.s6_addr[i] = a->s6_addr[i] & b->s6_addr[i];
294 /* Returns an in6_addr consisting of 'mask' high-order 1-bits and 128-N
295 * low-order 0-bits. */
297 ipv6_create_mask(int mask)
299 struct in6_addr netmask;
300 uint8_t *netmaskp = &netmask.s6_addr[0];
302 memset(&netmask, 0, sizeof netmask);
310 *netmaskp = 0xff << (8 - mask);
316 /* Given the IPv6 netmask 'netmask', returns the number of bits of the IPv6
317 * address that it specifies, that is, the number of 1-bits in 'netmask'.
318 * 'netmask' must be a CIDR netmask (see ipv6_is_cidr()). */
320 ipv6_count_cidr_bits(const struct in6_addr *netmask)
324 const uint8_t *netmaskp = &netmask->s6_addr[0];
326 assert(ipv6_is_cidr(netmask));
328 for (i=0; i<16; i++) {
329 if (netmaskp[i] == 0xff) {
334 for(nm = netmaskp[i]; nm; nm <<= 1) {
345 /* Returns true if 'netmask' is a CIDR netmask, that is, if it consists of N
346 * high-order 1-bits and 128-N low-order 0-bits. */
348 ipv6_is_cidr(const struct in6_addr *netmask)
350 const uint8_t *netmaskp = &netmask->s6_addr[0];
353 for (i=0; i<16; i++) {
354 if (netmaskp[i] != 0xff) {
355 uint8_t x = ~netmaskp[i];
370 /* Populates 'b' with an Ethernet II packet headed with the given 'eth_dst',
371 * 'eth_src' and 'eth_type' parameters. A payload of 'size' bytes is allocated
372 * in 'b' and returned. This payload may be populated with appropriate
373 * information by the caller. Sets 'b''s 'l2' and 'l3' pointers to the
374 * Ethernet header and payload respectively.
376 * The returned packet has enough headroom to insert an 802.1Q VLAN header if
379 eth_compose(struct ofpbuf *b, const uint8_t eth_dst[ETH_ADDR_LEN],
380 const uint8_t eth_src[ETH_ADDR_LEN], uint16_t eth_type,
384 struct eth_header *eth;
388 ofpbuf_prealloc_tailroom(b, ETH_HEADER_LEN + VLAN_HEADER_LEN + size);
389 ofpbuf_reserve(b, VLAN_HEADER_LEN);
390 eth = ofpbuf_put_uninit(b, ETH_HEADER_LEN);
391 data = ofpbuf_put_uninit(b, size);
393 memcpy(eth->eth_dst, eth_dst, ETH_ADDR_LEN);
394 memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN);
395 eth->eth_type = htons(eth_type);
403 /* Populates 'b' with an Ethernet LLC+SNAP packet headed with the given
404 * 'eth_dst', 'eth_src', 'snap_org', and 'snap_type'. A payload of 'size'
405 * bytes is allocated in 'b' and returned. This payload may be populated with
406 * appropriate information by the caller.
408 * The returned packet has enough headroom to insert an 802.1Q VLAN header if
411 snap_compose(struct ofpbuf *b, const uint8_t eth_dst[ETH_ADDR_LEN],
412 const uint8_t eth_src[ETH_ADDR_LEN],
413 unsigned int oui, uint16_t snap_type, size_t size)
415 struct eth_header *eth;
416 struct llc_snap_header *llc_snap;
419 /* Compose basic packet structure. (We need the payload size to stick into
420 * the 802.2 header.) */
422 ofpbuf_prealloc_tailroom(b, ETH_HEADER_LEN + VLAN_HEADER_LEN
423 + LLC_SNAP_HEADER_LEN + size);
424 ofpbuf_reserve(b, VLAN_HEADER_LEN);
425 eth = ofpbuf_put_zeros(b, ETH_HEADER_LEN);
426 llc_snap = ofpbuf_put_zeros(b, LLC_SNAP_HEADER_LEN);
427 payload = ofpbuf_put_uninit(b, size);
429 /* Compose 802.2 header. */
430 memcpy(eth->eth_dst, eth_dst, ETH_ADDR_LEN);
431 memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN);
432 eth->eth_type = htons(b->size - ETH_HEADER_LEN);
434 /* Compose LLC, SNAP headers. */
435 llc_snap->llc.llc_dsap = LLC_DSAP_SNAP;
436 llc_snap->llc.llc_ssap = LLC_SSAP_SNAP;
437 llc_snap->llc.llc_cntl = LLC_CNTL_SNAP;
438 llc_snap->snap.snap_org[0] = oui >> 16;
439 llc_snap->snap.snap_org[1] = oui >> 8;
440 llc_snap->snap.snap_org[2] = oui;
441 llc_snap->snap.snap_type = htons(snap_type);
447 packet_set_ipv4_addr(struct ofpbuf *packet, ovs_be32 *addr, ovs_be32 new_addr)
449 struct ip_header *nh = packet->l3;
451 if (nh->ip_proto == IPPROTO_TCP && packet->l7) {
452 struct tcp_header *th = packet->l4;
454 th->tcp_csum = recalc_csum32(th->tcp_csum, *addr, new_addr);
455 } else if (nh->ip_proto == IPPROTO_UDP && packet->l7) {
456 struct udp_header *uh = packet->l4;
459 uh->udp_csum = recalc_csum32(uh->udp_csum, *addr, new_addr);
461 uh->udp_csum = htons(0xffff);
465 nh->ip_csum = recalc_csum32(nh->ip_csum, *addr, new_addr);
469 /* Modifies the IPv4 header fields of 'packet' to be consistent with 'src',
470 * 'dst', 'tos', and 'ttl'. Updates 'packet''s L4 checksums as appropriate.
471 * 'packet' must contain a valid IPv4 packet with correctly populated l[347]
474 packet_set_ipv4(struct ofpbuf *packet, ovs_be32 src, ovs_be32 dst,
475 uint8_t tos, uint8_t ttl)
477 struct ip_header *nh = packet->l3;
479 if (nh->ip_src != src) {
480 packet_set_ipv4_addr(packet, &nh->ip_src, src);
483 if (nh->ip_dst != dst) {
484 packet_set_ipv4_addr(packet, &nh->ip_dst, dst);
487 if (nh->ip_tos != tos) {
488 uint8_t *field = &nh->ip_tos;
490 nh->ip_csum = recalc_csum16(nh->ip_csum, htons((uint16_t) *field),
491 htons((uint16_t) tos));
495 if (nh->ip_ttl != ttl) {
496 uint8_t *field = &nh->ip_ttl;
498 nh->ip_csum = recalc_csum16(nh->ip_csum, htons(*field << 8),
505 packet_set_port(ovs_be16 *port, ovs_be16 new_port, ovs_be16 *csum)
507 if (*port != new_port) {
508 *csum = recalc_csum16(*csum, *port, new_port);
513 /* Sets the TCP source and destination port ('src' and 'dst' respectively) of
514 * the TCP header contained in 'packet'. 'packet' must be a valid TCP packet
515 * with its l4 marker properly populated. */
517 packet_set_tcp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst)
519 struct tcp_header *th = packet->l4;
521 packet_set_port(&th->tcp_src, src, &th->tcp_csum);
522 packet_set_port(&th->tcp_dst, dst, &th->tcp_csum);
525 /* Sets the UDP source and destination port ('src' and 'dst' respectively) of
526 * the UDP header contained in 'packet'. 'packet' must be a valid UDP packet
527 * with its l4 marker properly populated. */
529 packet_set_udp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst)
531 struct udp_header *uh = packet->l4;
534 packet_set_port(&uh->udp_src, src, &uh->udp_csum);
535 packet_set_port(&uh->udp_dst, dst, &uh->udp_csum);
538 uh->udp_csum = htons(0xffff);
546 /* If 'packet' is a TCP packet, returns the TCP flags. Otherwise, returns 0.
548 * 'flow' must be the flow corresponding to 'packet' and 'packet''s header
549 * pointers must be properly initialized (e.g. with flow_extract()). */
551 packet_get_tcp_flags(const struct ofpbuf *packet, const struct flow *flow)
553 if ((flow->dl_type == htons(ETH_TYPE_IP) ||
554 flow->dl_type == htons(ETH_TYPE_IPV6)) &&
555 flow->nw_proto == IPPROTO_TCP && packet->l7) {
556 const struct tcp_header *tcp = packet->l4;
557 return TCP_FLAGS(tcp->tcp_ctl);
563 /* Appends a string representation of the TCP flags value 'tcp_flags'
564 * (e.g. obtained via packet_get_tcp_flags() or TCP_FLAGS) to 's', in the
565 * format used by tcpdump. */
567 packet_format_tcp_flags(struct ds *s, uint8_t tcp_flags)
570 ds_put_cstr(s, "none");
574 if (tcp_flags & TCP_SYN) {
577 if (tcp_flags & TCP_FIN) {
580 if (tcp_flags & TCP_PSH) {
583 if (tcp_flags & TCP_RST) {
586 if (tcp_flags & TCP_URG) {
589 if (tcp_flags & TCP_ACK) {
592 if (tcp_flags & 0x40) {
593 ds_put_cstr(s, "[40]");
595 if (tcp_flags & 0x80) {
596 ds_put_cstr(s, "[80]");