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)
47 eth_addr_from_string(const char *s, uint8_t ea[ETH_ADDR_LEN])
49 if (sscanf(s, ETH_ADDR_SCAN_FMT, ETH_ADDR_SCAN_ARGS(ea))
50 == ETH_ADDR_SCAN_COUNT) {
53 memset(ea, 0, ETH_ADDR_LEN);
58 /* Fills 'b' with an 802.2 SNAP packet with Ethernet source address 'eth_src',
59 * the Nicira OUI as SNAP organization and 'snap_type' as SNAP type. The text
60 * string in 'tag' is enclosed as the packet payload.
62 * This function is used by Open vSwitch to compose packets in cases where
63 * context is important but content doesn't (or shouldn't) matter. For this
64 * purpose, 'snap_type' should be a random number and 'tag' should be an
65 * English phrase that explains the purpose of the packet. (The English phrase
66 * gives hapless admins running Wireshark the opportunity to figure out what's
69 compose_benign_packet(struct ofpbuf *b, const char *tag, uint16_t snap_type,
70 const uint8_t eth_src[ETH_ADDR_LEN])
72 size_t tag_size = strlen(tag) + 1;
75 payload = snap_compose(b, eth_addr_broadcast, eth_src, 0x002320, snap_type,
76 tag_size + ETH_ADDR_LEN);
77 memcpy(payload, tag, tag_size);
78 memcpy(payload + tag_size, eth_src, ETH_ADDR_LEN);
81 /* Insert VLAN header according to given TCI. Packet passed must be Ethernet
82 * packet. Ignores the CFI bit of 'tci' using 0 instead.
84 * Also sets 'packet->l2' to point to the new Ethernet header. */
86 eth_push_vlan(struct ofpbuf *packet, ovs_be16 tci)
88 struct eth_header *eh = packet->data;
89 struct vlan_eth_header *veh;
91 /* Insert new 802.1Q header. */
92 struct vlan_eth_header tmp;
93 memcpy(tmp.veth_dst, eh->eth_dst, ETH_ADDR_LEN);
94 memcpy(tmp.veth_src, eh->eth_src, ETH_ADDR_LEN);
95 tmp.veth_type = htons(ETH_TYPE_VLAN);
96 tmp.veth_tci = tci & htons(~VLAN_CFI);
97 tmp.veth_next_type = eh->eth_type;
99 veh = ofpbuf_push_uninit(packet, VLAN_HEADER_LEN);
100 memcpy(veh, &tmp, sizeof tmp);
102 packet->l2 = packet->data;
105 /* Removes outermost VLAN header (if any is present) from 'packet'.
107 * 'packet->l2' must initially point to 'packet''s Ethernet header. */
109 eth_pop_vlan(struct ofpbuf *packet)
111 struct vlan_eth_header *veh = packet->l2;
112 if (packet->size >= sizeof *veh
113 && veh->veth_type == htons(ETH_TYPE_VLAN)) {
114 struct eth_header tmp;
116 memcpy(tmp.eth_dst, veh->veth_dst, ETH_ADDR_LEN);
117 memcpy(tmp.eth_src, veh->veth_src, ETH_ADDR_LEN);
118 tmp.eth_type = veh->veth_next_type;
120 ofpbuf_pull(packet, VLAN_HEADER_LEN);
121 packet->l2 = (char*)packet->l2 + VLAN_HEADER_LEN;
122 memcpy(packet->data, &tmp, sizeof tmp);
126 /* Converts hex digits in 'hex' to an Ethernet packet in '*packetp'. The
127 * caller must free '*packetp'. On success, returns NULL. On failure, returns
128 * an error message and stores NULL in '*packetp'. */
130 eth_from_hex(const char *hex, struct ofpbuf **packetp)
132 struct ofpbuf *packet;
134 packet = *packetp = ofpbuf_new(strlen(hex) / 2);
136 if (ofpbuf_put_hex(packet, hex, NULL)[0] != '\0') {
137 ofpbuf_delete(packet);
139 return "Trailing garbage in packet data";
142 if (packet->size < ETH_HEADER_LEN) {
143 ofpbuf_delete(packet);
145 return "Packet data too short for Ethernet";
152 eth_format_masked(const uint8_t eth[ETH_ADDR_LEN],
153 const uint8_t mask[ETH_ADDR_LEN], struct ds *s)
155 ds_put_format(s, ETH_ADDR_FMT, ETH_ADDR_ARGS(eth));
157 ds_put_format(s, "/"ETH_ADDR_FMT, ETH_ADDR_ARGS(mask));
162 eth_addr_bitand(const uint8_t src[ETH_ADDR_LEN],
163 const uint8_t mask[ETH_ADDR_LEN],
164 uint8_t dst[ETH_ADDR_LEN])
168 for (i = 0; i < ETH_ADDR_LEN; i++) {
169 dst[i] = src[i] & mask[i];
173 /* Given the IP netmask 'netmask', returns the number of bits of the IP address
174 * that it specifies, that is, the number of 1-bits in 'netmask'. 'netmask'
175 * must be a CIDR netmask (see ip_is_cidr()). */
177 ip_count_cidr_bits(ovs_be32 netmask)
179 assert(ip_is_cidr(netmask));
180 return 32 - ctz(ntohl(netmask));
184 ip_format_masked(ovs_be32 ip, ovs_be32 mask, struct ds *s)
186 ds_put_format(s, IP_FMT, IP_ARGS(&ip));
187 if (mask != htonl(UINT32_MAX)) {
188 if (ip_is_cidr(mask)) {
189 ds_put_format(s, "/%d", ip_count_cidr_bits(mask));
191 ds_put_format(s, "/"IP_FMT, IP_ARGS(&mask));
197 /* Stores the string representation of the IPv6 address 'addr' into the
198 * character array 'addr_str', which must be at least INET6_ADDRSTRLEN
201 format_ipv6_addr(char *addr_str, const struct in6_addr *addr)
203 inet_ntop(AF_INET6, addr, addr_str, INET6_ADDRSTRLEN);
207 print_ipv6_addr(struct ds *string, const struct in6_addr *addr)
211 ds_reserve(string, string->length + INET6_ADDRSTRLEN);
213 dst = string->string + string->length;
214 format_ipv6_addr(dst, addr);
215 string->length += strlen(dst);
219 print_ipv6_masked(struct ds *s, const struct in6_addr *addr,
220 const struct in6_addr *mask)
222 print_ipv6_addr(s, addr);
223 if (mask && !ipv6_mask_is_exact(mask)) {
224 if (ipv6_is_cidr(mask)) {
225 int cidr_bits = ipv6_count_cidr_bits(mask);
226 ds_put_format(s, "/%d", cidr_bits);
229 print_ipv6_addr(s, mask);
234 struct in6_addr ipv6_addr_bitand(const struct in6_addr *a,
235 const struct in6_addr *b)
241 for (i=0; i<4; i++) {
242 dst.s6_addr32[i] = a->s6_addr32[i] & b->s6_addr32[i];
245 for (i=0; i<16; i++) {
246 dst.s6_addr[i] = a->s6_addr[i] & b->s6_addr[i];
253 /* Returns an in6_addr consisting of 'mask' high-order 1-bits and 128-N
254 * low-order 0-bits. */
256 ipv6_create_mask(int mask)
258 struct in6_addr netmask;
259 uint8_t *netmaskp = &netmask.s6_addr[0];
261 memset(&netmask, 0, sizeof netmask);
269 *netmaskp = 0xff << (8 - mask);
275 /* Given the IPv6 netmask 'netmask', returns the number of bits of the IPv6
276 * address that it specifies, that is, the number of 1-bits in 'netmask'.
277 * 'netmask' must be a CIDR netmask (see ipv6_is_cidr()). */
279 ipv6_count_cidr_bits(const struct in6_addr *netmask)
283 const uint8_t *netmaskp = &netmask->s6_addr[0];
285 assert(ipv6_is_cidr(netmask));
287 for (i=0; i<16; i++) {
288 if (netmaskp[i] == 0xff) {
293 for(nm = netmaskp[i]; nm; nm <<= 1) {
304 /* Returns true if 'netmask' is a CIDR netmask, that is, if it consists of N
305 * high-order 1-bits and 128-N low-order 0-bits. */
307 ipv6_is_cidr(const struct in6_addr *netmask)
309 const uint8_t *netmaskp = &netmask->s6_addr[0];
312 for (i=0; i<16; i++) {
313 if (netmaskp[i] != 0xff) {
314 uint8_t x = ~netmaskp[i];
329 /* Populates 'b' with an Ethernet II packet headed with the given 'eth_dst',
330 * 'eth_src' and 'eth_type' parameters. A payload of 'size' bytes is allocated
331 * in 'b' and returned. This payload may be populated with appropriate
332 * information by the caller. Sets 'b''s 'l2' and 'l3' pointers to the
333 * Ethernet header and payload respectively.
335 * The returned packet has enough headroom to insert an 802.1Q VLAN header if
338 eth_compose(struct ofpbuf *b, const uint8_t eth_dst[ETH_ADDR_LEN],
339 const uint8_t eth_src[ETH_ADDR_LEN], uint16_t eth_type,
343 struct eth_header *eth;
347 ofpbuf_prealloc_tailroom(b, ETH_HEADER_LEN + VLAN_HEADER_LEN + size);
348 ofpbuf_reserve(b, VLAN_HEADER_LEN);
349 eth = ofpbuf_put_uninit(b, ETH_HEADER_LEN);
350 data = ofpbuf_put_uninit(b, size);
352 memcpy(eth->eth_dst, eth_dst, ETH_ADDR_LEN);
353 memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN);
354 eth->eth_type = htons(eth_type);
362 /* Populates 'b' with an Ethernet LLC+SNAP packet headed with the given
363 * 'eth_dst', 'eth_src', 'snap_org', and 'snap_type'. A payload of 'size'
364 * bytes is allocated in 'b' and returned. This payload may be populated with
365 * appropriate information by the caller.
367 * The returned packet has enough headroom to insert an 802.1Q VLAN header if
370 snap_compose(struct ofpbuf *b, const uint8_t eth_dst[ETH_ADDR_LEN],
371 const uint8_t eth_src[ETH_ADDR_LEN],
372 unsigned int oui, uint16_t snap_type, size_t size)
374 struct eth_header *eth;
375 struct llc_snap_header *llc_snap;
378 /* Compose basic packet structure. (We need the payload size to stick into
379 * the 802.2 header.) */
381 ofpbuf_prealloc_tailroom(b, ETH_HEADER_LEN + VLAN_HEADER_LEN
382 + LLC_SNAP_HEADER_LEN + size);
383 ofpbuf_reserve(b, VLAN_HEADER_LEN);
384 eth = ofpbuf_put_zeros(b, ETH_HEADER_LEN);
385 llc_snap = ofpbuf_put_zeros(b, LLC_SNAP_HEADER_LEN);
386 payload = ofpbuf_put_uninit(b, size);
388 /* Compose 802.2 header. */
389 memcpy(eth->eth_dst, eth_dst, ETH_ADDR_LEN);
390 memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN);
391 eth->eth_type = htons(b->size - ETH_HEADER_LEN);
393 /* Compose LLC, SNAP headers. */
394 llc_snap->llc.llc_dsap = LLC_DSAP_SNAP;
395 llc_snap->llc.llc_ssap = LLC_SSAP_SNAP;
396 llc_snap->llc.llc_cntl = LLC_CNTL_SNAP;
397 llc_snap->snap.snap_org[0] = oui >> 16;
398 llc_snap->snap.snap_org[1] = oui >> 8;
399 llc_snap->snap.snap_org[2] = oui;
400 llc_snap->snap.snap_type = htons(snap_type);
406 packet_set_ipv4_addr(struct ofpbuf *packet, ovs_be32 *addr, ovs_be32 new_addr)
408 struct ip_header *nh = packet->l3;
410 if (nh->ip_proto == IPPROTO_TCP && packet->l7) {
411 struct tcp_header *th = packet->l4;
413 th->tcp_csum = recalc_csum32(th->tcp_csum, *addr, new_addr);
414 } else if (nh->ip_proto == IPPROTO_UDP && packet->l7) {
415 struct udp_header *uh = packet->l4;
418 uh->udp_csum = recalc_csum32(uh->udp_csum, *addr, new_addr);
420 uh->udp_csum = htons(0xffff);
424 nh->ip_csum = recalc_csum32(nh->ip_csum, *addr, new_addr);
428 /* Modifies the IPv4 header fields of 'packet' to be consistent with 'src',
429 * 'dst', 'tos', and 'ttl'. Updates 'packet''s L4 checksums as appropriate.
430 * 'packet' must contain a valid IPv4 packet with correctly populated l[347]
433 packet_set_ipv4(struct ofpbuf *packet, ovs_be32 src, ovs_be32 dst,
434 uint8_t tos, uint8_t ttl)
436 struct ip_header *nh = packet->l3;
438 if (nh->ip_src != src) {
439 packet_set_ipv4_addr(packet, &nh->ip_src, src);
442 if (nh->ip_dst != dst) {
443 packet_set_ipv4_addr(packet, &nh->ip_dst, dst);
446 if (nh->ip_tos != tos) {
447 uint8_t *field = &nh->ip_tos;
449 nh->ip_csum = recalc_csum16(nh->ip_csum, htons((uint16_t) *field),
450 htons((uint16_t) tos));
454 if (nh->ip_ttl != ttl) {
455 uint8_t *field = &nh->ip_ttl;
457 nh->ip_csum = recalc_csum16(nh->ip_csum, htons(*field << 8),
464 packet_set_port(ovs_be16 *port, ovs_be16 new_port, ovs_be16 *csum)
466 if (*port != new_port) {
467 *csum = recalc_csum16(*csum, *port, new_port);
472 /* Sets the TCP source and destination port ('src' and 'dst' respectively) of
473 * the TCP header contained in 'packet'. 'packet' must be a valid TCP packet
474 * with its l4 marker properly populated. */
476 packet_set_tcp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst)
478 struct tcp_header *th = packet->l4;
480 packet_set_port(&th->tcp_src, src, &th->tcp_csum);
481 packet_set_port(&th->tcp_dst, dst, &th->tcp_csum);
484 /* Sets the UDP source and destination port ('src' and 'dst' respectively) of
485 * the UDP header contained in 'packet'. 'packet' must be a valid UDP packet
486 * with its l4 marker properly populated. */
488 packet_set_udp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst)
490 struct udp_header *uh = packet->l4;
493 packet_set_port(&uh->udp_src, src, &uh->udp_csum);
494 packet_set_port(&uh->udp_dst, dst, &uh->udp_csum);
497 uh->udp_csum = htons(0xffff);
505 /* If 'packet' is a TCP packet, returns the TCP flags. Otherwise, returns 0.
507 * 'flow' must be the flow corresponding to 'packet' and 'packet''s header
508 * pointers must be properly initialized (e.g. with flow_extract()). */
510 packet_get_tcp_flags(const struct ofpbuf *packet, const struct flow *flow)
512 if ((flow->dl_type == htons(ETH_TYPE_IP) ||
513 flow->dl_type == htons(ETH_TYPE_IPV6)) &&
514 flow->nw_proto == IPPROTO_TCP && packet->l7) {
515 const struct tcp_header *tcp = packet->l4;
516 return TCP_FLAGS(tcp->tcp_ctl);
522 /* Appends a string representation of the TCP flags value 'tcp_flags'
523 * (e.g. obtained via packet_get_tcp_flags() or TCP_FLAGS) to 's', in the
524 * format used by tcpdump. */
526 packet_format_tcp_flags(struct ds *s, uint8_t tcp_flags)
529 ds_put_cstr(s, "none");
533 if (tcp_flags & TCP_SYN) {
536 if (tcp_flags & TCP_FIN) {
539 if (tcp_flags & TCP_PSH) {
542 if (tcp_flags & TCP_RST) {
545 if (tcp_flags & TCP_URG) {
548 if (tcp_flags & TCP_ACK) {
551 if (tcp_flags & 0x40) {
552 ds_put_cstr(s, "[40]");
554 if (tcp_flags & 0x80) {
555 ds_put_cstr(s, "[80]");