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>
22 #include <netinet/in.h>
23 #include <netinet/icmp6.h>
24 #include <netinet/ip6.h>
27 #include "byte-order.h"
30 #include "dynamic-string.h"
33 #include "openflow/openflow.h"
34 #include "openvswitch/datapath-protocol.h"
36 #include "unaligned.h"
39 VLOG_DEFINE_THIS_MODULE(flow);
41 COVERAGE_DEFINE(flow_extract);
43 static struct arp_eth_header *
44 pull_arp(struct ofpbuf *packet)
46 return ofpbuf_try_pull(packet, ARP_ETH_HEADER_LEN);
49 static struct ip_header *
50 pull_ip(struct ofpbuf *packet)
52 if (packet->size >= IP_HEADER_LEN) {
53 struct ip_header *ip = packet->data;
54 int ip_len = IP_IHL(ip->ip_ihl_ver) * 4;
55 if (ip_len >= IP_HEADER_LEN && packet->size >= ip_len) {
56 return ofpbuf_pull(packet, ip_len);
62 static struct tcp_header *
63 pull_tcp(struct ofpbuf *packet)
65 if (packet->size >= TCP_HEADER_LEN) {
66 struct tcp_header *tcp = packet->data;
67 int tcp_len = TCP_OFFSET(tcp->tcp_ctl) * 4;
68 if (tcp_len >= TCP_HEADER_LEN && packet->size >= tcp_len) {
69 return ofpbuf_pull(packet, tcp_len);
75 static struct udp_header *
76 pull_udp(struct ofpbuf *packet)
78 return ofpbuf_try_pull(packet, UDP_HEADER_LEN);
81 static struct icmp_header *
82 pull_icmp(struct ofpbuf *packet)
84 return ofpbuf_try_pull(packet, ICMP_HEADER_LEN);
87 static struct icmp6_hdr *
88 pull_icmpv6(struct ofpbuf *packet)
90 return ofpbuf_try_pull(packet, sizeof(struct icmp6_hdr));
94 parse_vlan(struct ofpbuf *b, struct flow *flow)
97 ovs_be16 eth_type; /* ETH_TYPE_VLAN */
101 if (b->size >= sizeof(struct qtag_prefix) + sizeof(ovs_be16)) {
102 struct qtag_prefix *qp = ofpbuf_pull(b, sizeof *qp);
103 flow->vlan_tci = qp->tci | htons(VLAN_CFI);
108 parse_ethertype(struct ofpbuf *b)
110 struct llc_snap_header *llc;
113 proto = *(ovs_be16 *) ofpbuf_pull(b, sizeof proto);
114 if (ntohs(proto) >= ETH_TYPE_MIN) {
118 if (b->size < sizeof *llc) {
119 return htons(FLOW_DL_TYPE_NONE);
123 if (llc->llc.llc_dsap != LLC_DSAP_SNAP
124 || llc->llc.llc_ssap != LLC_SSAP_SNAP
125 || llc->llc.llc_cntl != LLC_CNTL_SNAP
126 || memcmp(llc->snap.snap_org, SNAP_ORG_ETHERNET,
127 sizeof llc->snap.snap_org)) {
128 return htons(FLOW_DL_TYPE_NONE);
131 ofpbuf_pull(b, sizeof *llc);
132 return llc->snap.snap_type;
136 parse_ipv6(struct ofpbuf *packet, struct flow *flow)
138 const struct ip6_hdr *nh;
142 nh = ofpbuf_try_pull(packet, sizeof *nh);
147 nexthdr = nh->ip6_nxt;
149 flow->ipv6_src = nh->ip6_src;
150 flow->ipv6_dst = nh->ip6_dst;
152 tc_flow = get_unaligned_be32(&nh->ip6_flow);
153 flow->nw_tos = (ntohl(tc_flow) >> 4) & IP_DSCP_MASK;
154 flow->nw_proto = IPPROTO_NONE;
157 if ((nexthdr != IPPROTO_HOPOPTS)
158 && (nexthdr != IPPROTO_ROUTING)
159 && (nexthdr != IPPROTO_DSTOPTS)
160 && (nexthdr != IPPROTO_AH)
161 && (nexthdr != IPPROTO_FRAGMENT)) {
162 /* It's either a terminal header (e.g., TCP, UDP) or one we
163 * don't understand. In either case, we're done with the
164 * packet, so use it to fill in 'nw_proto'. */
168 /* We only verify that at least 8 bytes of the next header are
169 * available, but many of these headers are longer. Ensure that
170 * accesses within the extension header are within those first 8
171 * bytes. All extension headers are required to be at least 8
173 if (packet->size < 8) {
177 if ((nexthdr == IPPROTO_HOPOPTS)
178 || (nexthdr == IPPROTO_ROUTING)
179 || (nexthdr == IPPROTO_DSTOPTS)) {
180 /* These headers, while different, have the fields we care about
181 * in the same location and with the same interpretation. */
182 const struct ip6_ext *ext_hdr = (struct ip6_ext *)packet->data;
183 nexthdr = ext_hdr->ip6e_nxt;
184 if (!ofpbuf_try_pull(packet, (ext_hdr->ip6e_len + 1) * 8)) {
187 } else if (nexthdr == IPPROTO_AH) {
188 /* A standard AH definition isn't available, but the fields
189 * we care about are in the same location as the generic
190 * option header--only the header length is calculated
192 const struct ip6_ext *ext_hdr = (struct ip6_ext *)packet->data;
193 nexthdr = ext_hdr->ip6e_nxt;
194 if (!ofpbuf_try_pull(packet, (ext_hdr->ip6e_len + 2) * 4)) {
197 } else if (nexthdr == IPPROTO_FRAGMENT) {
198 const struct ip6_frag *frag_hdr = (struct ip6_frag *)packet->data;
200 nexthdr = frag_hdr->ip6f_nxt;
201 if (!ofpbuf_try_pull(packet, sizeof *frag_hdr)) {
205 /* We only process the first fragment. */
206 if ((frag_hdr->ip6f_offlg & IP6F_OFF_MASK) != htons(0)) {
207 nexthdr = IPPROTO_FRAGMENT;
213 flow->nw_proto = nexthdr;
218 parse_tcp(struct ofpbuf *packet, struct ofpbuf *b, struct flow *flow)
220 const struct tcp_header *tcp = pull_tcp(b);
222 flow->tp_src = tcp->tcp_src;
223 flow->tp_dst = tcp->tcp_dst;
224 packet->l7 = b->data;
229 parse_udp(struct ofpbuf *packet, struct ofpbuf *b, struct flow *flow)
231 const struct udp_header *udp = pull_udp(b);
233 flow->tp_src = udp->udp_src;
234 flow->tp_dst = udp->udp_dst;
235 packet->l7 = b->data;
240 parse_icmpv6(struct ofpbuf *b, struct flow *flow)
242 const struct icmp6_hdr *icmp = pull_icmpv6(b);
248 /* The ICMPv6 type and code fields use the 16-bit transport port
249 * fields, so we need to store them in 16-bit network byte order. */
250 flow->icmp_type = htons(icmp->icmp6_type);
251 flow->icmp_code = htons(icmp->icmp6_code);
253 if (icmp->icmp6_code == 0 &&
254 (icmp->icmp6_type == ND_NEIGHBOR_SOLICIT ||
255 icmp->icmp6_type == ND_NEIGHBOR_ADVERT)) {
256 const struct in6_addr *nd_target;
258 nd_target = ofpbuf_try_pull(b, sizeof *nd_target);
262 flow->nd_target = *nd_target;
264 while (b->size >= 8) {
265 /* The minimum size of an option is 8 bytes, which also is
266 * the size of Ethernet link-layer options. */
267 const struct nd_opt_hdr *nd_opt = b->data;
268 int opt_len = nd_opt->nd_opt_len * 8;
270 if (!opt_len || opt_len > b->size) {
274 /* Store the link layer address if the appropriate option is
275 * provided. It is considered an error if the same link
276 * layer option is specified twice. */
277 if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LINKADDR
279 if (eth_addr_is_zero(flow->arp_sha)) {
280 memcpy(flow->arp_sha, nd_opt + 1, ETH_ADDR_LEN);
284 } else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LINKADDR
286 if (eth_addr_is_zero(flow->arp_tha)) {
287 memcpy(flow->arp_tha, nd_opt + 1, ETH_ADDR_LEN);
293 if (!ofpbuf_try_pull(b, opt_len)) {
302 memset(&flow->nd_target, 0, sizeof(flow->nd_target));
303 memset(flow->arp_sha, 0, sizeof(flow->arp_sha));
304 memset(flow->arp_tha, 0, sizeof(flow->arp_tha));
310 /* Initializes 'flow' members from 'packet', 'tun_id', and 'ofp_in_port'.
311 * Initializes 'packet' header pointers as follows:
313 * - packet->l2 to the start of the Ethernet header.
315 * - packet->l3 to just past the Ethernet header, or just past the
316 * vlan_header if one is present, to the first byte of the payload of the
319 * - packet->l4 to just past the IPv4 header, if one is present and has a
320 * correct length, and otherwise NULL.
322 * - packet->l7 to just past the TCP or UDP or ICMP header, if one is
323 * present and has a correct length, and otherwise NULL.
326 flow_extract(struct ofpbuf *packet, ovs_be64 tun_id, uint16_t ofp_in_port,
329 struct ofpbuf b = *packet;
330 struct eth_header *eth;
333 COVERAGE_INC(flow_extract);
335 memset(flow, 0, sizeof *flow);
336 flow->tun_id = tun_id;
337 flow->in_port = ofp_in_port;
344 if (b.size < sizeof *eth) {
350 memcpy(flow->dl_src, eth->eth_src, ETH_ADDR_LEN);
351 memcpy(flow->dl_dst, eth->eth_dst, ETH_ADDR_LEN);
353 /* dl_type, vlan_tci. */
354 ofpbuf_pull(&b, ETH_ADDR_LEN * 2);
355 if (eth->eth_type == htons(ETH_TYPE_VLAN)) {
356 parse_vlan(&b, flow);
358 flow->dl_type = parse_ethertype(&b);
362 if (flow->dl_type == htons(ETH_TYPE_IP)) {
363 const struct ip_header *nh = pull_ip(&b);
365 flow->nw_src = get_unaligned_be32(&nh->ip_src);
366 flow->nw_dst = get_unaligned_be32(&nh->ip_dst);
367 flow->nw_tos = nh->ip_tos & IP_DSCP_MASK;
368 flow->nw_proto = nh->ip_proto;
370 if (!IP_IS_FRAGMENT(nh->ip_frag_off)) {
371 if (flow->nw_proto == IPPROTO_TCP) {
372 parse_tcp(packet, &b, flow);
373 } else if (flow->nw_proto == IPPROTO_UDP) {
374 parse_udp(packet, &b, flow);
375 } else if (flow->nw_proto == IPPROTO_ICMP) {
376 const struct icmp_header *icmp = pull_icmp(&b);
378 flow->icmp_type = htons(icmp->icmp_type);
379 flow->icmp_code = htons(icmp->icmp_code);
387 } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) {
389 retval = parse_ipv6(&b, flow);
395 if (flow->nw_proto == IPPROTO_TCP) {
396 parse_tcp(packet, &b, flow);
397 } else if (flow->nw_proto == IPPROTO_UDP) {
398 parse_udp(packet, &b, flow);
399 } else if (flow->nw_proto == IPPROTO_ICMPV6) {
400 if (parse_icmpv6(&b, flow)) {
404 } else if (flow->dl_type == htons(ETH_TYPE_ARP)) {
405 const struct arp_eth_header *arp = pull_arp(&b);
406 if (arp && arp->ar_hrd == htons(1)
407 && arp->ar_pro == htons(ETH_TYPE_IP)
408 && arp->ar_hln == ETH_ADDR_LEN
409 && arp->ar_pln == 4) {
410 /* We only match on the lower 8 bits of the opcode. */
411 if (ntohs(arp->ar_op) <= 0xff) {
412 flow->nw_proto = ntohs(arp->ar_op);
415 if ((flow->nw_proto == ARP_OP_REQUEST)
416 || (flow->nw_proto == ARP_OP_REPLY)) {
417 flow->nw_src = arp->ar_spa;
418 flow->nw_dst = arp->ar_tpa;
419 memcpy(flow->arp_sha, arp->ar_sha, ETH_ADDR_LEN);
420 memcpy(flow->arp_tha, arp->ar_tha, ETH_ADDR_LEN);
428 /* Extracts the flow stats for a packet. The 'flow' and 'packet'
429 * arguments must have been initialized through a call to flow_extract().
432 flow_extract_stats(const struct flow *flow, struct ofpbuf *packet,
433 struct dpif_flow_stats *stats)
435 memset(stats, 0, sizeof(*stats));
437 if ((flow->dl_type == htons(ETH_TYPE_IP)) && packet->l4) {
438 if ((flow->nw_proto == IPPROTO_TCP) && packet->l7) {
439 struct tcp_header *tcp = packet->l4;
440 stats->tcp_flags = TCP_FLAGS(tcp->tcp_ctl);
444 stats->n_bytes = packet->size;
445 stats->n_packets = 1;
449 flow_to_string(const struct flow *flow)
451 struct ds ds = DS_EMPTY_INITIALIZER;
452 flow_format(&ds, flow);
457 flow_format(struct ds *ds, const struct flow *flow)
459 ds_put_format(ds, "tunnel%#"PRIx64":in_port%04"PRIx16":tci(",
460 ntohll(flow->tun_id), flow->in_port);
461 if (flow->vlan_tci) {
462 ds_put_format(ds, "vlan%"PRIu16",pcp%d",
463 vlan_tci_to_vid(flow->vlan_tci),
464 vlan_tci_to_pcp(flow->vlan_tci));
466 ds_put_char(ds, '0');
468 ds_put_format(ds, ") mac"ETH_ADDR_FMT"->"ETH_ADDR_FMT
470 ETH_ADDR_ARGS(flow->dl_src),
471 ETH_ADDR_ARGS(flow->dl_dst),
472 ntohs(flow->dl_type));
474 if (flow->dl_type == htons(ETH_TYPE_IPV6)) {
475 ds_put_format(ds, " proto%"PRIu8" tos%"PRIu8" ipv6",
476 flow->nw_proto, flow->nw_tos);
477 print_ipv6_addr(ds, &flow->ipv6_src);
478 ds_put_cstr(ds, "->");
479 print_ipv6_addr(ds, &flow->ipv6_dst);
482 ds_put_format(ds, " proto%"PRIu8
484 " ip"IP_FMT"->"IP_FMT,
487 IP_ARGS(&flow->nw_src),
488 IP_ARGS(&flow->nw_dst));
490 if (flow->tp_src || flow->tp_dst) {
491 ds_put_format(ds, " port%"PRIu16"->%"PRIu16,
492 ntohs(flow->tp_src), ntohs(flow->tp_dst));
494 if (!eth_addr_is_zero(flow->arp_sha) || !eth_addr_is_zero(flow->arp_tha)) {
495 ds_put_format(ds, " arp_ha"ETH_ADDR_FMT"->"ETH_ADDR_FMT,
496 ETH_ADDR_ARGS(flow->arp_sha),
497 ETH_ADDR_ARGS(flow->arp_tha));
502 flow_print(FILE *stream, const struct flow *flow)
504 char *s = flow_to_string(flow);
509 /* flow_wildcards functions. */
511 /* Initializes 'wc' as a set of wildcards that matches every packet. */
513 flow_wildcards_init_catchall(struct flow_wildcards *wc)
515 wc->wildcards = FWW_ALL;
516 wc->tun_id_mask = htonll(0);
517 wc->nw_src_mask = htonl(0);
518 wc->nw_dst_mask = htonl(0);
519 wc->ipv6_src_mask = in6addr_any;
520 wc->ipv6_dst_mask = in6addr_any;
521 memset(wc->reg_masks, 0, sizeof wc->reg_masks);
522 wc->vlan_tci_mask = htons(0);
526 /* Initializes 'wc' as an exact-match set of wildcards; that is, 'wc' does not
527 * wildcard any bits or fields. */
529 flow_wildcards_init_exact(struct flow_wildcards *wc)
532 wc->tun_id_mask = htonll(UINT64_MAX);
533 wc->nw_src_mask = htonl(UINT32_MAX);
534 wc->nw_dst_mask = htonl(UINT32_MAX);
535 wc->ipv6_src_mask = in6addr_exact;
536 wc->ipv6_dst_mask = in6addr_exact;
537 memset(wc->reg_masks, 0xff, sizeof wc->reg_masks);
538 wc->vlan_tci_mask = htons(UINT16_MAX);
542 /* Returns true if 'wc' is exact-match, false if 'wc' wildcards any bits or
545 flow_wildcards_is_exact(const struct flow_wildcards *wc)
550 || wc->tun_id_mask != htonll(UINT64_MAX)
551 || wc->nw_src_mask != htonl(UINT32_MAX)
552 || wc->nw_dst_mask != htonl(UINT32_MAX)
553 || wc->vlan_tci_mask != htons(UINT16_MAX)
554 || !ipv6_mask_is_exact(&wc->ipv6_src_mask)
555 || !ipv6_mask_is_exact(&wc->ipv6_dst_mask)) {
559 for (i = 0; i < FLOW_N_REGS; i++) {
560 if (wc->reg_masks[i] != UINT32_MAX) {
568 /* Initializes 'dst' as the combination of wildcards in 'src1' and 'src2'.
569 * That is, a bit or a field is wildcarded in 'dst' if it is wildcarded in
570 * 'src1' or 'src2' or both. */
572 flow_wildcards_combine(struct flow_wildcards *dst,
573 const struct flow_wildcards *src1,
574 const struct flow_wildcards *src2)
578 dst->wildcards = src1->wildcards | src2->wildcards;
579 dst->tun_id_mask = src1->tun_id_mask & src2->tun_id_mask;
580 dst->nw_src_mask = src1->nw_src_mask & src2->nw_src_mask;
581 dst->nw_dst_mask = src1->nw_dst_mask & src2->nw_dst_mask;
582 dst->ipv6_src_mask = ipv6_addr_bitand(&src1->ipv6_src_mask,
583 &src2->ipv6_src_mask);
584 dst->ipv6_dst_mask = ipv6_addr_bitand(&src1->ipv6_dst_mask,
585 &src2->ipv6_dst_mask);
586 for (i = 0; i < FLOW_N_REGS; i++) {
587 dst->reg_masks[i] = src1->reg_masks[i] & src2->reg_masks[i];
589 dst->vlan_tci_mask = src1->vlan_tci_mask & src2->vlan_tci_mask;
592 /* Returns a hash of the wildcards in 'wc'. */
594 flow_wildcards_hash(const struct flow_wildcards *wc, uint32_t basis)
596 /* If you change struct flow_wildcards and thereby trigger this
597 * assertion, please check that the new struct flow_wildcards has no holes
598 * in it before you update the assertion. */
599 BUILD_ASSERT_DECL(sizeof *wc == 56 + FLOW_N_REGS * 4);
600 return hash_bytes(wc, sizeof *wc, basis);
603 /* Returns true if 'a' and 'b' represent the same wildcards, false if they are
606 flow_wildcards_equal(const struct flow_wildcards *a,
607 const struct flow_wildcards *b)
611 if (a->wildcards != b->wildcards
612 || a->tun_id_mask != b->tun_id_mask
613 || a->nw_src_mask != b->nw_src_mask
614 || a->nw_dst_mask != b->nw_dst_mask
615 || a->vlan_tci_mask != b->vlan_tci_mask
616 || !ipv6_addr_equals(&a->ipv6_src_mask, &b->ipv6_src_mask)
617 || !ipv6_addr_equals(&a->ipv6_dst_mask, &b->ipv6_dst_mask)) {
621 for (i = 0; i < FLOW_N_REGS; i++) {
622 if (a->reg_masks[i] != b->reg_masks[i]) {
630 /* Returns true if at least one bit or field is wildcarded in 'a' but not in
631 * 'b', false otherwise. */
633 flow_wildcards_has_extra(const struct flow_wildcards *a,
634 const struct flow_wildcards *b)
637 struct in6_addr ipv6_masked;
639 for (i = 0; i < FLOW_N_REGS; i++) {
640 if ((a->reg_masks[i] & b->reg_masks[i]) != b->reg_masks[i]) {
645 ipv6_masked = ipv6_addr_bitand(&a->ipv6_src_mask, &b->ipv6_src_mask);
646 if (!ipv6_addr_equals(&ipv6_masked, &b->ipv6_src_mask)) {
650 ipv6_masked = ipv6_addr_bitand(&a->ipv6_dst_mask, &b->ipv6_dst_mask);
651 if (!ipv6_addr_equals(&ipv6_masked, &b->ipv6_dst_mask)) {
655 return (a->wildcards & ~b->wildcards
656 || (a->tun_id_mask & b->tun_id_mask) != b->tun_id_mask
657 || (a->nw_src_mask & b->nw_src_mask) != b->nw_src_mask
658 || (a->nw_dst_mask & b->nw_dst_mask) != b->nw_dst_mask
659 || (a->vlan_tci_mask & b->vlan_tci_mask) != b->vlan_tci_mask);
663 set_nw_mask(ovs_be32 *maskp, ovs_be32 mask)
665 if (ip_is_cidr(mask)) {
673 /* Sets the IP (or ARP) source wildcard mask to CIDR 'mask' (consisting of N
674 * high-order 1-bit and 32-N low-order 0-bits). Returns true if successful,
675 * false if 'mask' is not a CIDR mask. */
677 flow_wildcards_set_nw_src_mask(struct flow_wildcards *wc, ovs_be32 mask)
679 return set_nw_mask(&wc->nw_src_mask, mask);
682 /* Sets the IP (or ARP) destination wildcard mask to CIDR 'mask' (consisting of
683 * N high-order 1-bit and 32-N low-order 0-bits). Returns true if successful,
684 * false if 'mask' is not a CIDR mask. */
686 flow_wildcards_set_nw_dst_mask(struct flow_wildcards *wc, ovs_be32 mask)
688 return set_nw_mask(&wc->nw_dst_mask, mask);
692 set_ipv6_mask(struct in6_addr *maskp, const struct in6_addr *mask)
694 if (ipv6_is_cidr(mask)) {
702 /* Sets the IPv6 source wildcard mask to CIDR 'mask' (consisting of N
703 * high-order 1-bit and 128-N low-order 0-bits). Returns true if successful,
704 * false if 'mask' is not a CIDR mask. */
706 flow_wildcards_set_ipv6_src_mask(struct flow_wildcards *wc,
707 const struct in6_addr *mask)
709 return set_ipv6_mask(&wc->ipv6_src_mask, mask);
712 /* Sets the IPv6 destination wildcard mask to CIDR 'mask' (consisting of
713 * N high-order 1-bit and 128-N low-order 0-bits). Returns true if
714 * successful, false if 'mask' is not a CIDR mask. */
716 flow_wildcards_set_ipv6_dst_mask(struct flow_wildcards *wc,
717 const struct in6_addr *mask)
719 return set_ipv6_mask(&wc->ipv6_dst_mask, mask);
722 /* Sets the wildcard mask for register 'idx' in 'wc' to 'mask'.
723 * (A 0-bit indicates a wildcard bit.) */
725 flow_wildcards_set_reg_mask(struct flow_wildcards *wc, int idx, uint32_t mask)
727 wc->reg_masks[idx] = mask;
730 /* Returns the wildcard bitmask for the Ethernet destination address
731 * that 'wc' specifies. The bitmask has a 0 in each bit that is wildcarded
732 * and a 1 in each bit that must match. */
734 flow_wildcards_to_dl_dst_mask(flow_wildcards_t wc)
736 static const uint8_t no_wild[] = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff};
737 static const uint8_t addr_wild[] = {0x01, 0x00, 0x00, 0x00, 0x00, 0x00};
738 static const uint8_t mcast_wild[] = {0xfe, 0xff, 0xff, 0xff, 0xff, 0xff};
739 static const uint8_t all_wild[] = {0x00, 0x00, 0x00, 0x00, 0x00, 0x00};
741 switch (wc & (FWW_DL_DST | FWW_ETH_MCAST)) {
742 case 0: return no_wild;
743 case FWW_DL_DST: return addr_wild;
744 case FWW_ETH_MCAST: return mcast_wild;
745 case FWW_DL_DST | FWW_ETH_MCAST: return all_wild;
750 /* Returns true if 'mask' is a valid wildcard bitmask for the Ethernet
751 * destination address. Valid bitmasks are either all-bits-0 or all-bits-1,
752 * except that the multicast bit may differ from the rest of the bits. So,
753 * there are four possible valid bitmasks:
755 * - 00:00:00:00:00:00
756 * - 01:00:00:00:00:00
757 * - fe:ff:ff:ff:ff:ff
758 * - ff:ff:ff:ff:ff:ff
760 * All other bitmasks are invalid. */
762 flow_wildcards_is_dl_dst_mask_valid(const uint8_t mask[ETH_ADDR_LEN])
767 return (mask[1] | mask[2] | mask[3] | mask[4] | mask[5]) == 0x00;
771 return (mask[1] & mask[2] & mask[3] & mask[4] & mask[5]) == 0xff;
778 /* Returns 'wc' with the FWW_DL_DST and FWW_ETH_MCAST bits modified
779 * appropriately to match 'mask'.
781 * This function will assert-fail if 'mask' is invalid. Only 'mask' values
782 * accepted by flow_wildcards_is_dl_dst_mask_valid() are allowed. */
784 flow_wildcards_set_dl_dst_mask(flow_wildcards_t wc,
785 const uint8_t mask[ETH_ADDR_LEN])
787 assert(flow_wildcards_is_dl_dst_mask_valid(mask));
791 return wc | FWW_DL_DST | FWW_ETH_MCAST;
794 return (wc | FWW_DL_DST) & ~FWW_ETH_MCAST;
797 return (wc & ~FWW_DL_DST) | FWW_ETH_MCAST;
800 return wc & ~(FWW_DL_DST | FWW_ETH_MCAST);
807 /* Hashes 'flow' based on its L2 through L4 protocol information. */
809 flow_hash_symmetric_l4(const struct flow *flow, uint32_t basis)
814 struct in6_addr ipv6_addr;
819 uint8_t eth_addr[ETH_ADDR_LEN];
825 memset(&fields, 0, sizeof fields);
826 for (i = 0; i < ETH_ADDR_LEN; i++) {
827 fields.eth_addr[i] = flow->dl_src[i] ^ flow->dl_dst[i];
829 fields.vlan_tci = flow->vlan_tci & htons(VLAN_VID_MASK);
830 fields.eth_type = flow->dl_type;
832 /* UDP source and destination port are not taken into account because they
833 * will not necessarily be symmetric in a bidirectional flow. */
834 if (fields.eth_type == htons(ETH_TYPE_IP)) {
835 fields.ipv4_addr = flow->nw_src ^ flow->nw_dst;
836 fields.ip_proto = flow->nw_proto;
837 if (fields.ip_proto == IPPROTO_TCP) {
838 fields.tp_addr = flow->tp_src ^ flow->tp_dst;
840 } else if (fields.eth_type == htons(ETH_TYPE_IPV6)) {
841 const uint8_t *a = &flow->ipv6_src.s6_addr[0];
842 const uint8_t *b = &flow->ipv6_dst.s6_addr[0];
843 uint8_t *ipv6_addr = &fields.ipv6_addr.s6_addr[0];
845 for (i=0; i<16; i++) {
846 ipv6_addr[i] = a[i] ^ b[i];
848 fields.ip_proto = flow->nw_proto;
849 if (fields.ip_proto == IPPROTO_TCP) {
850 fields.tp_addr = flow->tp_src ^ flow->tp_dst;
853 return hash_bytes(&fields, sizeof fields, basis);
856 /* Hashes the portions of 'flow' designated by 'fields'. */
858 flow_hash_fields(const struct flow *flow, enum nx_hash_fields fields,
863 case NX_HASH_FIELDS_ETH_SRC:
864 return hash_bytes(flow->dl_src, sizeof flow->dl_src, basis);
866 case NX_HASH_FIELDS_SYMMETRIC_L4:
867 return flow_hash_symmetric_l4(flow, basis);
873 /* Returns a string representation of 'fields'. */
875 flow_hash_fields_to_str(enum nx_hash_fields fields)
878 case NX_HASH_FIELDS_ETH_SRC: return "eth_src";
879 case NX_HASH_FIELDS_SYMMETRIC_L4: return "symmetric_l4";
880 default: return "<unknown>";
884 /* Returns true if the value of 'fields' is supported. Otherwise false. */
886 flow_hash_fields_valid(enum nx_hash_fields fields)
888 return fields == NX_HASH_FIELDS_ETH_SRC
889 || fields == NX_HASH_FIELDS_SYMMETRIC_L4;
892 /* Puts into 'b' a packet that flow_extract() would parse as having the given
895 * (This is useful only for testing, obviously, and the packet isn't really
896 * valid. It hasn't got any checksums filled in, for one, and lots of fields
897 * are just zeroed.) */
899 flow_compose(struct ofpbuf *b, const struct flow *flow)
901 eth_compose(b, flow->dl_dst, flow->dl_src, ntohs(flow->dl_type), 0);
902 if (flow->dl_type == htons(FLOW_DL_TYPE_NONE)) {
903 struct eth_header *eth = b->l2;
904 eth->eth_type = htons(b->size);
908 if (flow->vlan_tci & htons(VLAN_CFI)) {
909 eth_push_vlan(b, flow->vlan_tci & ~htons(VLAN_CFI));
912 if (flow->dl_type == htons(ETH_TYPE_IP)) {
913 struct ip_header *ip;
915 b->l3 = ip = ofpbuf_put_zeros(b, sizeof *ip);
916 ip->ip_ihl_ver = IP_IHL_VER(5, 4);
917 ip->ip_tos = flow->nw_tos;
918 ip->ip_proto = flow->nw_proto;
919 ip->ip_src = flow->nw_src;
920 ip->ip_dst = flow->nw_dst;
922 if (flow->nw_proto == IPPROTO_TCP) {
923 struct tcp_header *tcp;
925 b->l4 = tcp = ofpbuf_put_zeros(b, sizeof *tcp);
926 tcp->tcp_src = flow->tp_src;
927 tcp->tcp_dst = flow->tp_dst;
928 } else if (flow->nw_proto == IPPROTO_UDP) {
929 struct udp_header *udp;
931 b->l4 = udp = ofpbuf_put_zeros(b, sizeof *udp);
932 udp->udp_src = flow->tp_src;
933 udp->udp_dst = flow->tp_dst;
934 } else if (flow->nw_proto == IPPROTO_ICMP) {
935 struct icmp_header *icmp;
937 b->l4 = icmp = ofpbuf_put_zeros(b, sizeof *icmp);
938 icmp->icmp_type = ntohs(flow->tp_src);
939 icmp->icmp_code = ntohs(flow->tp_dst);
941 } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) {
943 } else if (flow->dl_type == htons(ETH_TYPE_ARP)) {
944 struct arp_eth_header *arp;
946 b->l3 = arp = ofpbuf_put_zeros(b, sizeof *arp);
947 arp->ar_hrd = htons(1);
948 arp->ar_pro = htons(ETH_TYPE_IP);
949 arp->ar_hln = ETH_ADDR_LEN;
951 arp->ar_op = htons(flow->nw_proto);
953 if (flow->nw_proto == ARP_OP_REQUEST ||
954 flow->nw_proto == ARP_OP_REPLY) {
955 arp->ar_spa = flow->nw_src;
956 arp->ar_tpa = flow->nw_dst;
957 memcpy(arp->ar_sha, flow->arp_sha, ETH_ADDR_LEN);
958 memcpy(arp->ar_tha, flow->arp_tha, ETH_ADDR_LEN);