2 * Copyright (c) 2008, 2009, 2010, 2011 Nicira Networks.
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.
17 #include <sys/types.h>
21 #include <netinet/in.h>
22 #include <netinet/icmp6.h>
23 #include <netinet/ip6.h>
26 #include "byte-order.h"
29 #include "dynamic-string.h"
32 #include "openflow/openflow.h"
33 #include "openvswitch/datapath-protocol.h"
35 #include "unaligned.h"
38 VLOG_DEFINE_THIS_MODULE(flow);
40 COVERAGE_DEFINE(flow_extract);
42 static struct arp_eth_header *
43 pull_arp(struct ofpbuf *packet)
45 return ofpbuf_try_pull(packet, ARP_ETH_HEADER_LEN);
48 static struct ip_header *
49 pull_ip(struct ofpbuf *packet)
51 if (packet->size >= IP_HEADER_LEN) {
52 struct ip_header *ip = packet->data;
53 int ip_len = IP_IHL(ip->ip_ihl_ver) * 4;
54 if (ip_len >= IP_HEADER_LEN && packet->size >= ip_len) {
55 return ofpbuf_pull(packet, ip_len);
61 static struct tcp_header *
62 pull_tcp(struct ofpbuf *packet)
64 if (packet->size >= TCP_HEADER_LEN) {
65 struct tcp_header *tcp = packet->data;
66 int tcp_len = TCP_OFFSET(tcp->tcp_ctl) * 4;
67 if (tcp_len >= TCP_HEADER_LEN && packet->size >= tcp_len) {
68 return ofpbuf_pull(packet, tcp_len);
74 static struct udp_header *
75 pull_udp(struct ofpbuf *packet)
77 return ofpbuf_try_pull(packet, UDP_HEADER_LEN);
80 static struct icmp_header *
81 pull_icmp(struct ofpbuf *packet)
83 return ofpbuf_try_pull(packet, ICMP_HEADER_LEN);
86 static struct icmp6_hdr *
87 pull_icmpv6(struct ofpbuf *packet)
89 return ofpbuf_try_pull(packet, sizeof(struct icmp6_hdr));
93 parse_vlan(struct ofpbuf *b, struct flow *flow)
96 ovs_be16 eth_type; /* ETH_TYPE_VLAN */
100 if (b->size >= sizeof(struct qtag_prefix) + sizeof(ovs_be16)) {
101 struct qtag_prefix *qp = ofpbuf_pull(b, sizeof *qp);
102 flow->vlan_tci = qp->tci | htons(VLAN_CFI);
107 parse_ethertype(struct ofpbuf *b)
109 struct llc_snap_header *llc;
112 proto = *(ovs_be16 *) ofpbuf_pull(b, sizeof proto);
113 if (ntohs(proto) >= ETH_TYPE_MIN) {
117 if (b->size < sizeof *llc) {
118 return htons(FLOW_DL_TYPE_NONE);
122 if (llc->llc.llc_dsap != LLC_DSAP_SNAP
123 || llc->llc.llc_ssap != LLC_SSAP_SNAP
124 || llc->llc.llc_cntl != LLC_CNTL_SNAP
125 || memcmp(llc->snap.snap_org, SNAP_ORG_ETHERNET,
126 sizeof llc->snap.snap_org)) {
127 return htons(FLOW_DL_TYPE_NONE);
130 ofpbuf_pull(b, sizeof *llc);
131 return llc->snap.snap_type;
135 parse_ipv6(struct ofpbuf *packet, struct flow *flow)
138 int nh_len = sizeof(struct ip6_hdr);
143 if (packet->size < sizeof *nh) {
148 nexthdr = nh->ip6_nxt;
149 payload_len = ntohs(nh->ip6_plen);
151 flow->ipv6_src = nh->ip6_src;
152 flow->ipv6_dst = nh->ip6_dst;
154 tc_flow = get_unaligned_be32(&nh->ip6_flow);
155 flow->nw_tos = (ntohl(tc_flow) >> 4) & IP_DSCP_MASK;
156 flow->nw_proto = IPPROTO_NONE;
158 /* We don't process jumbograms. */
163 if (packet->size < sizeof *nh + payload_len) {
168 if ((nexthdr != IPPROTO_HOPOPTS)
169 && (nexthdr != IPPROTO_ROUTING)
170 && (nexthdr != IPPROTO_DSTOPTS)
171 && (nexthdr != IPPROTO_AH)
172 && (nexthdr != IPPROTO_FRAGMENT)) {
173 /* It's either a terminal header (e.g., TCP, UDP) or one we
174 * don't understand. In either case, we're done with the
175 * packet, so use it to fill in 'nw_proto'. */
179 /* We only verify that at least 8 bytes of the next header are
180 * available, but many of these headers are longer. Ensure that
181 * accesses within the extension header are within those first 8
183 if (packet->size < nh_len + 8) {
187 if ((nexthdr == IPPROTO_HOPOPTS)
188 || (nexthdr == IPPROTO_ROUTING)
189 || (nexthdr == IPPROTO_DSTOPTS)) {
190 /* These headers, while different, have the fields we care about
191 * in the same location and with the same interpretation. */
192 struct ip6_ext *ext_hdr;
194 ext_hdr = (struct ip6_ext *)((char *)packet->data + nh_len);
195 nexthdr = ext_hdr->ip6e_nxt;
196 nh_len += (ext_hdr->ip6e_len + 1) * 8;
197 } else if (nexthdr == IPPROTO_AH) {
198 /* A standard AH definition isn't available, but the fields
199 * we care about are in the same location as the generic
200 * option header--only the header length is calculated
202 struct ip6_ext *ext_hdr;
204 ext_hdr = (struct ip6_ext *)((char *)packet->data + nh_len);
205 nexthdr = ext_hdr->ip6e_nxt;
206 nh_len += (ext_hdr->ip6e_len + 2) * 4;
207 } else if (nexthdr == IPPROTO_FRAGMENT) {
208 struct ip6_frag *frag_hdr;
210 frag_hdr = (struct ip6_frag *)((char *)packet->data + nh_len);
212 nexthdr = frag_hdr->ip6f_nxt;
213 nh_len += sizeof *frag_hdr;
215 /* We only process the first fragment. */
216 if ((frag_hdr->ip6f_offlg & IP6F_OFF_MASK) != htons(0)) {
217 nexthdr = IPPROTO_FRAGMENT;
223 /* The payload length claims to be smaller than the size of the
224 * headers we've already processed. */
225 if (payload_len < nh_len - sizeof *nh) {
229 flow->nw_proto = nexthdr;
233 /* Initializes 'flow' members from 'packet', 'tun_id', and 'in_port.
234 * Initializes 'packet' header pointers as follows:
236 * - packet->l2 to the start of the Ethernet header.
238 * - packet->l3 to just past the Ethernet header, or just past the
239 * vlan_header if one is present, to the first byte of the payload of the
242 * - packet->l4 to just past the IPv4 header, if one is present and has a
243 * correct length, and otherwise NULL.
245 * - packet->l7 to just past the TCP or UDP or ICMP header, if one is
246 * present and has a correct length, and otherwise NULL.
249 flow_extract(struct ofpbuf *packet, ovs_be64 tun_id, uint16_t in_port,
252 struct ofpbuf b = *packet;
253 struct eth_header *eth;
256 COVERAGE_INC(flow_extract);
258 memset(flow, 0, sizeof *flow);
259 flow->tun_id = tun_id;
260 flow->in_port = in_port;
267 if (b.size < sizeof *eth) {
273 memcpy(flow->dl_src, eth->eth_src, ETH_ADDR_LEN);
274 memcpy(flow->dl_dst, eth->eth_dst, ETH_ADDR_LEN);
276 /* dl_type, vlan_tci. */
277 ofpbuf_pull(&b, ETH_ADDR_LEN * 2);
278 if (eth->eth_type == htons(ETH_TYPE_VLAN)) {
279 parse_vlan(&b, flow);
281 flow->dl_type = parse_ethertype(&b);
285 if (flow->dl_type == htons(ETH_TYPE_IP)) {
286 const struct ip_header *nh = pull_ip(&b);
288 flow->nw_src = get_unaligned_be32(&nh->ip_src);
289 flow->nw_dst = get_unaligned_be32(&nh->ip_dst);
290 flow->nw_tos = nh->ip_tos & IP_DSCP_MASK;
291 flow->nw_proto = nh->ip_proto;
293 if (!IP_IS_FRAGMENT(nh->ip_frag_off)) {
294 if (flow->nw_proto == IPPROTO_TCP) {
295 const struct tcp_header *tcp = pull_tcp(&b);
297 flow->tp_src = tcp->tcp_src;
298 flow->tp_dst = tcp->tcp_dst;
301 } else if (flow->nw_proto == IPPROTO_UDP) {
302 const struct udp_header *udp = pull_udp(&b);
304 flow->tp_src = udp->udp_src;
305 flow->tp_dst = udp->udp_dst;
308 } else if (flow->nw_proto == IPPROTO_ICMP) {
309 const struct icmp_header *icmp = pull_icmp(&b);
311 flow->icmp_type = htons(icmp->icmp_type);
312 flow->icmp_code = htons(icmp->icmp_code);
320 } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) {
322 const struct ip6_hdr *nh;
324 nh_len = parse_ipv6(&b, flow);
329 nh = ofpbuf_pull(&b, nh_len);
332 if (flow->nw_proto == IPPROTO_TCP) {
333 const struct tcp_header *tcp = pull_tcp(&b);
335 flow->tp_src = tcp->tcp_src;
336 flow->tp_dst = tcp->tcp_dst;
339 } else if (flow->nw_proto == IPPROTO_UDP) {
340 const struct udp_header *udp = pull_udp(&b);
342 flow->tp_src = udp->udp_src;
343 flow->tp_dst = udp->udp_dst;
346 } else if (flow->nw_proto == IPPROTO_ICMPV6) {
347 const struct icmp6_hdr *icmp = pull_icmpv6(&b);
349 flow->icmp_type = htons(icmp->icmp6_type);
350 flow->icmp_code = htons(icmp->icmp6_code);
355 } else if (flow->dl_type == htons(ETH_TYPE_ARP)) {
356 const struct arp_eth_header *arp = pull_arp(&b);
357 if (arp && arp->ar_hrd == htons(1)
358 && arp->ar_pro == htons(ETH_TYPE_IP)
359 && arp->ar_hln == ETH_ADDR_LEN
360 && arp->ar_pln == 4) {
361 /* We only match on the lower 8 bits of the opcode. */
362 if (ntohs(arp->ar_op) <= 0xff) {
363 flow->nw_proto = ntohs(arp->ar_op);
366 if ((flow->nw_proto == ARP_OP_REQUEST)
367 || (flow->nw_proto == ARP_OP_REPLY)) {
368 flow->nw_src = arp->ar_spa;
369 flow->nw_dst = arp->ar_tpa;
370 memcpy(flow->arp_sha, arp->ar_sha, ETH_ADDR_LEN);
371 memcpy(flow->arp_tha, arp->ar_tha, ETH_ADDR_LEN);
379 /* Extracts the flow stats for a packet. The 'flow' and 'packet'
380 * arguments must have been initialized through a call to flow_extract().
383 flow_extract_stats(const struct flow *flow, struct ofpbuf *packet,
384 struct dpif_flow_stats *stats)
386 memset(stats, 0, sizeof(*stats));
388 if ((flow->dl_type == htons(ETH_TYPE_IP)) && packet->l4) {
389 if ((flow->nw_proto == IPPROTO_TCP) && packet->l7) {
390 struct tcp_header *tcp = packet->l4;
391 stats->tcp_flags = TCP_FLAGS(tcp->tcp_ctl);
395 stats->n_bytes = packet->size;
396 stats->n_packets = 1;
400 flow_to_string(const struct flow *flow)
402 struct ds ds = DS_EMPTY_INITIALIZER;
403 flow_format(&ds, flow);
408 flow_format(struct ds *ds, const struct flow *flow)
410 ds_put_format(ds, "tunnel%#"PRIx64":in_port%04"PRIx16":tci(",
411 flow->tun_id, flow->in_port);
412 if (flow->vlan_tci) {
413 ds_put_format(ds, "vlan%"PRIu16",pcp%d",
414 vlan_tci_to_vid(flow->vlan_tci),
415 vlan_tci_to_pcp(flow->vlan_tci));
417 ds_put_char(ds, '0');
419 ds_put_format(ds, ") mac"ETH_ADDR_FMT"->"ETH_ADDR_FMT
421 ETH_ADDR_ARGS(flow->dl_src),
422 ETH_ADDR_ARGS(flow->dl_dst),
423 ntohs(flow->dl_type));
425 if (flow->dl_type == htons(ETH_TYPE_IPV6)) {
426 ds_put_format(ds, " proto%"PRIu8" tos%"PRIu8" ipv6",
427 flow->nw_proto, flow->nw_tos);
428 print_ipv6_addr(ds, &flow->ipv6_src);
429 ds_put_cstr(ds, "->");
430 print_ipv6_addr(ds, &flow->ipv6_dst);
433 ds_put_format(ds, " proto%"PRIu8
435 " ip"IP_FMT"->"IP_FMT,
438 IP_ARGS(&flow->nw_src),
439 IP_ARGS(&flow->nw_dst));
441 if (flow->tp_src || flow->tp_dst) {
442 ds_put_format(ds, " port%"PRIu16"->%"PRIu16,
443 ntohs(flow->tp_src), ntohs(flow->tp_dst));
445 if (!eth_addr_is_zero(flow->arp_sha) || !eth_addr_is_zero(flow->arp_tha)) {
446 ds_put_format(ds, " arp_ha"ETH_ADDR_FMT"->"ETH_ADDR_FMT,
447 ETH_ADDR_ARGS(flow->arp_sha),
448 ETH_ADDR_ARGS(flow->arp_tha));
453 flow_print(FILE *stream, const struct flow *flow)
455 char *s = flow_to_string(flow);
460 /* flow_wildcards functions. */
462 /* Initializes 'wc' as a set of wildcards that matches every packet. */
464 flow_wildcards_init_catchall(struct flow_wildcards *wc)
466 wc->wildcards = FWW_ALL;
467 wc->tun_id_mask = htonll(0);
468 wc->nw_src_mask = htonl(0);
469 wc->nw_dst_mask = htonl(0);
470 wc->ipv6_src_mask = in6addr_any;
471 wc->ipv6_dst_mask = in6addr_any;
472 memset(wc->reg_masks, 0, sizeof wc->reg_masks);
473 wc->vlan_tci_mask = htons(0);
477 /* Initializes 'wc' as an exact-match set of wildcards; that is, 'wc' does not
478 * wildcard any bits or fields. */
480 flow_wildcards_init_exact(struct flow_wildcards *wc)
483 wc->tun_id_mask = htonll(UINT64_MAX);
484 wc->nw_src_mask = htonl(UINT32_MAX);
485 wc->nw_dst_mask = htonl(UINT32_MAX);
486 wc->ipv6_src_mask = in6addr_exact;
487 wc->ipv6_dst_mask = in6addr_exact;
488 memset(wc->reg_masks, 0xff, sizeof wc->reg_masks);
489 wc->vlan_tci_mask = htons(UINT16_MAX);
493 /* Returns true if 'wc' is exact-match, false if 'wc' wildcards any bits or
496 flow_wildcards_is_exact(const struct flow_wildcards *wc)
501 || wc->tun_id_mask != htonll(UINT64_MAX)
502 || wc->nw_src_mask != htonl(UINT32_MAX)
503 || wc->nw_dst_mask != htonl(UINT32_MAX)
504 || wc->vlan_tci_mask != htons(UINT16_MAX)
505 || !ipv6_mask_is_exact(&wc->ipv6_src_mask)
506 || !ipv6_mask_is_exact(&wc->ipv6_dst_mask)) {
510 for (i = 0; i < FLOW_N_REGS; i++) {
511 if (wc->reg_masks[i] != htonl(UINT32_MAX)) {
519 /* Initializes 'dst' as the combination of wildcards in 'src1' and 'src2'.
520 * That is, a bit or a field is wildcarded in 'dst' if it is wildcarded in
521 * 'src1' or 'src2' or both. */
523 flow_wildcards_combine(struct flow_wildcards *dst,
524 const struct flow_wildcards *src1,
525 const struct flow_wildcards *src2)
529 dst->wildcards = src1->wildcards | src2->wildcards;
530 dst->tun_id_mask = src1->tun_id_mask & src2->tun_id_mask;
531 dst->nw_src_mask = src1->nw_src_mask & src2->nw_src_mask;
532 dst->nw_dst_mask = src1->nw_dst_mask & src2->nw_dst_mask;
533 dst->ipv6_src_mask = ipv6_addr_bitand(&src1->ipv6_src_mask,
534 &src2->ipv6_src_mask);
535 dst->ipv6_dst_mask = ipv6_addr_bitand(&src1->ipv6_dst_mask,
536 &src2->ipv6_dst_mask);
537 for (i = 0; i < FLOW_N_REGS; i++) {
538 dst->reg_masks[i] = src1->reg_masks[i] & src2->reg_masks[i];
540 dst->vlan_tci_mask = src1->vlan_tci_mask & src2->vlan_tci_mask;
543 /* Returns a hash of the wildcards in 'wc'. */
545 flow_wildcards_hash(const struct flow_wildcards *wc)
547 /* If you change struct flow_wildcards and thereby trigger this
548 * assertion, please check that the new struct flow_wildcards has no holes
549 * in it before you update the assertion. */
550 BUILD_ASSERT_DECL(sizeof *wc == 56 + FLOW_N_REGS * 4);
551 return hash_bytes(wc, sizeof *wc, 0);
554 /* Returns true if 'a' and 'b' represent the same wildcards, false if they are
557 flow_wildcards_equal(const struct flow_wildcards *a,
558 const struct flow_wildcards *b)
562 if (a->wildcards != b->wildcards
563 || a->tun_id_mask != b->tun_id_mask
564 || a->nw_src_mask != b->nw_src_mask
565 || a->nw_dst_mask != b->nw_dst_mask
566 || a->vlan_tci_mask != b->vlan_tci_mask
567 || !ipv6_addr_equals(&a->ipv6_src_mask, &b->ipv6_src_mask)
568 || !ipv6_addr_equals(&a->ipv6_dst_mask, &b->ipv6_dst_mask)) {
572 for (i = 0; i < FLOW_N_REGS; i++) {
573 if (a->reg_masks[i] != b->reg_masks[i]) {
581 /* Returns true if at least one bit or field is wildcarded in 'a' but not in
582 * 'b', false otherwise. */
584 flow_wildcards_has_extra(const struct flow_wildcards *a,
585 const struct flow_wildcards *b)
588 struct in6_addr ipv6_masked;
590 for (i = 0; i < FLOW_N_REGS; i++) {
591 if ((a->reg_masks[i] & b->reg_masks[i]) != b->reg_masks[i]) {
596 ipv6_masked = ipv6_addr_bitand(&a->ipv6_src_mask, &b->ipv6_src_mask);
597 if (!ipv6_addr_equals(&ipv6_masked, &b->ipv6_src_mask)) {
601 ipv6_masked = ipv6_addr_bitand(&a->ipv6_dst_mask, &b->ipv6_dst_mask);
602 if (!ipv6_addr_equals(&ipv6_masked, &b->ipv6_dst_mask)) {
606 return (a->wildcards & ~b->wildcards
607 || (a->tun_id_mask & b->tun_id_mask) != b->tun_id_mask
608 || (a->nw_src_mask & b->nw_src_mask) != b->nw_src_mask
609 || (a->nw_dst_mask & b->nw_dst_mask) != b->nw_dst_mask
610 || (a->vlan_tci_mask & b->vlan_tci_mask) != b->vlan_tci_mask);
614 set_nw_mask(ovs_be32 *maskp, ovs_be32 mask)
616 if (ip_is_cidr(mask)) {
624 /* Sets the IP (or ARP) source wildcard mask to CIDR 'mask' (consisting of N
625 * high-order 1-bit and 32-N low-order 0-bits). Returns true if successful,
626 * false if 'mask' is not a CIDR mask. */
628 flow_wildcards_set_nw_src_mask(struct flow_wildcards *wc, ovs_be32 mask)
630 return set_nw_mask(&wc->nw_src_mask, mask);
633 /* Sets the IP (or ARP) destination wildcard mask to CIDR 'mask' (consisting of
634 * N high-order 1-bit and 32-N low-order 0-bits). Returns true if successful,
635 * false if 'mask' is not a CIDR mask. */
637 flow_wildcards_set_nw_dst_mask(struct flow_wildcards *wc, ovs_be32 mask)
639 return set_nw_mask(&wc->nw_dst_mask, mask);
643 set_ipv6_mask(struct in6_addr *maskp, const struct in6_addr *mask)
645 if (ipv6_is_cidr(mask)) {
653 /* Sets the IPv6 source wildcard mask to CIDR 'mask' (consisting of N
654 * high-order 1-bit and 128-N low-order 0-bits). Returns true if successful,
655 * false if 'mask' is not a CIDR mask. */
657 flow_wildcards_set_ipv6_src_mask(struct flow_wildcards *wc,
658 const struct in6_addr *mask)
660 return set_ipv6_mask(&wc->ipv6_src_mask, mask);
663 /* Sets the IPv6 destination wildcard mask to CIDR 'mask' (consisting of
664 * N high-order 1-bit and 128-N low-order 0-bits). Returns true if
665 * successful, false if 'mask' is not a CIDR mask. */
667 flow_wildcards_set_ipv6_dst_mask(struct flow_wildcards *wc,
668 const struct in6_addr *mask)
670 return set_ipv6_mask(&wc->ipv6_dst_mask, mask);
673 /* Sets the wildcard mask for register 'idx' in 'wc' to 'mask'.
674 * (A 0-bit indicates a wildcard bit.) */
676 flow_wildcards_set_reg_mask(struct flow_wildcards *wc, int idx, uint32_t mask)
678 wc->reg_masks[idx] = mask;
681 /* Hashes 'flow' based on its L2 through L4 protocol information. */
683 flow_hash_symmetric_l4(const struct flow *flow, uint32_t basis)
688 struct in6_addr ipv6_addr;
693 uint8_t eth_addr[ETH_ADDR_LEN];
699 memset(&fields, 0, sizeof fields);
700 for (i = 0; i < ETH_ADDR_LEN; i++) {
701 fields.eth_addr[i] = flow->dl_src[i] ^ flow->dl_dst[i];
703 fields.vlan_tci = flow->vlan_tci & htons(VLAN_VID_MASK);
704 fields.eth_type = flow->dl_type;
705 if (fields.eth_type == htons(ETH_TYPE_IP)) {
706 fields.ipv4_addr = flow->nw_src ^ flow->nw_dst;
707 fields.ip_proto = flow->nw_proto;
708 if (fields.ip_proto == IPPROTO_TCP || fields.ip_proto == IPPROTO_UDP) {
709 fields.tp_addr = flow->tp_src ^ flow->tp_dst;
711 } else if (fields.eth_type == htons(ETH_TYPE_IPV6)) {
712 const uint8_t *a = &flow->ipv6_src.s6_addr[0];
713 const uint8_t *b = &flow->ipv6_dst.s6_addr[0];
714 uint8_t *ipv6_addr = &fields.ipv6_addr.s6_addr[0];
716 for (i=0; i<16; i++) {
717 ipv6_addr[i] = a[i] ^ b[i];
719 fields.ip_proto = flow->nw_proto;
720 if (fields.ip_proto == IPPROTO_TCP || fields.ip_proto == IPPROTO_UDP) {
721 fields.tp_addr = flow->tp_src ^ flow->tp_dst;
724 return hash_bytes(&fields, sizeof fields, basis);