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"
29 #include "dynamic-string.h"
32 #include "openflow/openflow.h"
34 #include "unaligned.h"
37 VLOG_DEFINE_THIS_MODULE(flow);
39 COVERAGE_DEFINE(flow_extract);
41 static struct arp_eth_header *
42 pull_arp(struct ofpbuf *packet)
44 return ofpbuf_try_pull(packet, ARP_ETH_HEADER_LEN);
47 static struct ip_header *
48 pull_ip(struct ofpbuf *packet)
50 if (packet->size >= IP_HEADER_LEN) {
51 struct ip_header *ip = packet->data;
52 int ip_len = IP_IHL(ip->ip_ihl_ver) * 4;
53 if (ip_len >= IP_HEADER_LEN && packet->size >= ip_len) {
54 return ofpbuf_pull(packet, ip_len);
60 static struct tcp_header *
61 pull_tcp(struct ofpbuf *packet)
63 if (packet->size >= TCP_HEADER_LEN) {
64 struct tcp_header *tcp = packet->data;
65 int tcp_len = TCP_OFFSET(tcp->tcp_ctl) * 4;
66 if (tcp_len >= TCP_HEADER_LEN && packet->size >= tcp_len) {
67 return ofpbuf_pull(packet, tcp_len);
73 static struct udp_header *
74 pull_udp(struct ofpbuf *packet)
76 return ofpbuf_try_pull(packet, UDP_HEADER_LEN);
79 static struct icmp_header *
80 pull_icmp(struct ofpbuf *packet)
82 return ofpbuf_try_pull(packet, ICMP_HEADER_LEN);
85 static struct icmp6_hdr *
86 pull_icmpv6(struct ofpbuf *packet)
88 return ofpbuf_try_pull(packet, sizeof(struct icmp6_hdr));
92 parse_vlan(struct ofpbuf *b, struct flow *flow)
95 ovs_be16 eth_type; /* ETH_TYPE_VLAN */
99 if (b->size >= sizeof(struct qtag_prefix) + sizeof(ovs_be16)) {
100 struct qtag_prefix *qp = ofpbuf_pull(b, sizeof *qp);
101 flow->vlan_tci = qp->tci | htons(VLAN_CFI);
106 parse_ethertype(struct ofpbuf *b)
108 struct llc_snap_header *llc;
111 proto = *(ovs_be16 *) ofpbuf_pull(b, sizeof proto);
112 if (ntohs(proto) >= ETH_TYPE_MIN) {
116 if (b->size < sizeof *llc) {
117 return htons(FLOW_DL_TYPE_NONE);
121 if (llc->llc.llc_dsap != LLC_DSAP_SNAP
122 || llc->llc.llc_ssap != LLC_SSAP_SNAP
123 || llc->llc.llc_cntl != LLC_CNTL_SNAP
124 || memcmp(llc->snap.snap_org, SNAP_ORG_ETHERNET,
125 sizeof llc->snap.snap_org)) {
126 return htons(FLOW_DL_TYPE_NONE);
129 ofpbuf_pull(b, sizeof *llc);
130 return llc->snap.snap_type;
134 parse_ipv6(struct ofpbuf *packet, struct flow *flow)
136 const struct ip6_hdr *nh;
140 nh = ofpbuf_try_pull(packet, sizeof *nh);
145 nexthdr = nh->ip6_nxt;
147 flow->ipv6_src = nh->ip6_src;
148 flow->ipv6_dst = nh->ip6_dst;
150 tc_flow = get_unaligned_be32(&nh->ip6_flow);
151 flow->tos_frag = (ntohl(tc_flow) >> 4) & IP_DSCP_MASK;
152 flow->nw_proto = IPPROTO_NONE;
155 if ((nexthdr != IPPROTO_HOPOPTS)
156 && (nexthdr != IPPROTO_ROUTING)
157 && (nexthdr != IPPROTO_DSTOPTS)
158 && (nexthdr != IPPROTO_AH)
159 && (nexthdr != IPPROTO_FRAGMENT)) {
160 /* It's either a terminal header (e.g., TCP, UDP) or one we
161 * don't understand. In either case, we're done with the
162 * packet, so use it to fill in 'nw_proto'. */
166 /* We only verify that at least 8 bytes of the next header are
167 * available, but many of these headers are longer. Ensure that
168 * accesses within the extension header are within those first 8
169 * bytes. All extension headers are required to be at least 8
171 if (packet->size < 8) {
175 if ((nexthdr == IPPROTO_HOPOPTS)
176 || (nexthdr == IPPROTO_ROUTING)
177 || (nexthdr == IPPROTO_DSTOPTS)) {
178 /* These headers, while different, have the fields we care about
179 * in the same location and with the same interpretation. */
180 const struct ip6_ext *ext_hdr = (struct ip6_ext *)packet->data;
181 nexthdr = ext_hdr->ip6e_nxt;
182 if (!ofpbuf_try_pull(packet, (ext_hdr->ip6e_len + 1) * 8)) {
185 } else if (nexthdr == IPPROTO_AH) {
186 /* A standard AH definition isn't available, but the fields
187 * we care about are in the same location as the generic
188 * option header--only the header length is calculated
190 const struct ip6_ext *ext_hdr = (struct ip6_ext *)packet->data;
191 nexthdr = ext_hdr->ip6e_nxt;
192 if (!ofpbuf_try_pull(packet, (ext_hdr->ip6e_len + 2) * 4)) {
195 } else if (nexthdr == IPPROTO_FRAGMENT) {
196 const struct ip6_frag *frag_hdr = (struct ip6_frag *)packet->data;
198 nexthdr = frag_hdr->ip6f_nxt;
199 if (!ofpbuf_try_pull(packet, sizeof *frag_hdr)) {
203 /* We only process the first fragment. */
204 flow->tos_frag &= ~FLOW_FRAG_MASK;
205 flow->tos_frag |= FLOW_FRAG_ANY;
206 if ((frag_hdr->ip6f_offlg & IP6F_OFF_MASK) != htons(0)) {
207 flow->tos_frag |= FLOW_FRAG_LATER;
208 nexthdr = IPPROTO_FRAGMENT;
214 flow->nw_proto = nexthdr;
219 parse_tcp(struct ofpbuf *packet, struct ofpbuf *b, struct flow *flow)
221 const struct tcp_header *tcp = pull_tcp(b);
223 flow->tp_src = tcp->tcp_src;
224 flow->tp_dst = tcp->tcp_dst;
225 packet->l7 = b->data;
230 parse_udp(struct ofpbuf *packet, struct ofpbuf *b, struct flow *flow)
232 const struct udp_header *udp = pull_udp(b);
234 flow->tp_src = udp->udp_src;
235 flow->tp_dst = udp->udp_dst;
236 packet->l7 = b->data;
241 parse_icmpv6(struct ofpbuf *b, struct flow *flow)
243 const struct icmp6_hdr *icmp = pull_icmpv6(b);
249 /* The ICMPv6 type and code fields use the 16-bit transport port
250 * fields, so we need to store them in 16-bit network byte order. */
251 flow->tp_src = htons(icmp->icmp6_type);
252 flow->tp_dst = htons(icmp->icmp6_code);
254 if (icmp->icmp6_code == 0 &&
255 (icmp->icmp6_type == ND_NEIGHBOR_SOLICIT ||
256 icmp->icmp6_type == ND_NEIGHBOR_ADVERT)) {
257 const struct in6_addr *nd_target;
259 nd_target = ofpbuf_try_pull(b, sizeof *nd_target);
263 flow->nd_target = *nd_target;
265 while (b->size >= 8) {
266 /* The minimum size of an option is 8 bytes, which also is
267 * the size of Ethernet link-layer options. */
268 const struct nd_opt_hdr *nd_opt = b->data;
269 int opt_len = nd_opt->nd_opt_len * 8;
271 if (!opt_len || opt_len > b->size) {
275 /* Store the link layer address if the appropriate option is
276 * provided. It is considered an error if the same link
277 * layer option is specified twice. */
278 if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LINKADDR
280 if (eth_addr_is_zero(flow->arp_sha)) {
281 memcpy(flow->arp_sha, nd_opt + 1, ETH_ADDR_LEN);
285 } else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LINKADDR
287 if (eth_addr_is_zero(flow->arp_tha)) {
288 memcpy(flow->arp_tha, nd_opt + 1, ETH_ADDR_LEN);
294 if (!ofpbuf_try_pull(b, opt_len)) {
303 memset(&flow->nd_target, 0, sizeof(flow->nd_target));
304 memset(flow->arp_sha, 0, sizeof(flow->arp_sha));
305 memset(flow->arp_tha, 0, sizeof(flow->arp_tha));
311 /* Initializes 'flow' members from 'packet', 'tun_id', and 'ofp_in_port'.
312 * Initializes 'packet' header pointers as follows:
314 * - packet->l2 to the start of the Ethernet header.
316 * - packet->l3 to just past the Ethernet header, or just past the
317 * vlan_header if one is present, to the first byte of the payload of the
320 * - packet->l4 to just past the IPv4 header, if one is present and has a
321 * correct length, and otherwise NULL.
323 * - packet->l7 to just past the TCP or UDP or ICMP header, if one is
324 * present and has a correct length, and otherwise NULL.
327 flow_extract(struct ofpbuf *packet, uint32_t priority, ovs_be64 tun_id,
328 uint16_t ofp_in_port, struct flow *flow)
330 struct ofpbuf b = *packet;
331 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;
338 flow->priority = priority;
345 if (b.size < sizeof *eth) {
351 memcpy(flow->dl_src, eth->eth_src, ETH_ADDR_LEN);
352 memcpy(flow->dl_dst, eth->eth_dst, ETH_ADDR_LEN);
354 /* dl_type, vlan_tci. */
355 ofpbuf_pull(&b, ETH_ADDR_LEN * 2);
356 if (eth->eth_type == htons(ETH_TYPE_VLAN)) {
357 parse_vlan(&b, flow);
359 flow->dl_type = parse_ethertype(&b);
363 if (flow->dl_type == htons(ETH_TYPE_IP)) {
364 const struct ip_header *nh = pull_ip(&b);
368 flow->nw_src = get_unaligned_be32(&nh->ip_src);
369 flow->nw_dst = get_unaligned_be32(&nh->ip_dst);
370 flow->nw_proto = nh->ip_proto;
372 flow->tos_frag = nh->ip_tos & IP_DSCP_MASK;
373 if (IP_IS_FRAGMENT(nh->ip_frag_off)) {
374 flow->tos_frag |= FLOW_FRAG_ANY;
375 if (nh->ip_frag_off & htons(IP_FRAG_OFF_MASK)) {
376 flow->tos_frag |= FLOW_FRAG_LATER;
380 if (!(nh->ip_frag_off & htons(IP_FRAG_OFF_MASK))) {
381 if (flow->nw_proto == IPPROTO_TCP) {
382 parse_tcp(packet, &b, flow);
383 } else if (flow->nw_proto == IPPROTO_UDP) {
384 parse_udp(packet, &b, flow);
385 } else if (flow->nw_proto == IPPROTO_ICMP) {
386 const struct icmp_header *icmp = pull_icmp(&b);
388 flow->tp_src = htons(icmp->icmp_type);
389 flow->tp_dst = htons(icmp->icmp_code);
395 } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) {
396 if (parse_ipv6(&b, flow)) {
401 if (flow->nw_proto == IPPROTO_TCP) {
402 parse_tcp(packet, &b, flow);
403 } else if (flow->nw_proto == IPPROTO_UDP) {
404 parse_udp(packet, &b, flow);
405 } else if (flow->nw_proto == IPPROTO_ICMPV6) {
406 if (parse_icmpv6(&b, flow)) {
410 } else if (flow->dl_type == htons(ETH_TYPE_ARP)) {
411 const struct arp_eth_header *arp = pull_arp(&b);
412 if (arp && arp->ar_hrd == htons(1)
413 && arp->ar_pro == htons(ETH_TYPE_IP)
414 && arp->ar_hln == ETH_ADDR_LEN
415 && arp->ar_pln == 4) {
416 /* We only match on the lower 8 bits of the opcode. */
417 if (ntohs(arp->ar_op) <= 0xff) {
418 flow->nw_proto = ntohs(arp->ar_op);
421 if ((flow->nw_proto == ARP_OP_REQUEST)
422 || (flow->nw_proto == ARP_OP_REPLY)) {
423 flow->nw_src = arp->ar_spa;
424 flow->nw_dst = arp->ar_tpa;
425 memcpy(flow->arp_sha, arp->ar_sha, ETH_ADDR_LEN);
426 memcpy(flow->arp_tha, arp->ar_tha, ETH_ADDR_LEN);
432 /* For every bit of a field that is wildcarded in 'wildcards', sets the
433 * corresponding bit in 'flow' to zero. */
435 flow_zero_wildcards(struct flow *flow, const struct flow_wildcards *wildcards)
437 const flow_wildcards_t wc = wildcards->wildcards;
440 BUILD_ASSERT_DECL(FLOW_WC_SEQ == 3);
442 for (i = 0; i < FLOW_N_REGS; i++) {
443 flow->regs[i] &= wildcards->reg_masks[i];
445 flow->tun_id &= wildcards->tun_id_mask;
446 flow->nw_src &= wildcards->nw_src_mask;
447 flow->nw_dst &= wildcards->nw_dst_mask;
448 if (wc & FWW_IN_PORT) {
451 flow->vlan_tci &= wildcards->vlan_tci_mask;
452 if (wc & FWW_DL_TYPE) {
455 if (wc & FWW_TP_SRC) {
458 if (wc & FWW_TP_DST) {
461 if (wc & FWW_DL_SRC) {
462 memset(flow->dl_src, 0, sizeof flow->dl_src);
464 if (wc & FWW_DL_DST) {
465 flow->dl_dst[0] &= 0x01;
466 memset(&flow->dl_dst[1], 0, 5);
468 if (wc & FWW_ETH_MCAST) {
469 flow->dl_dst[0] &= 0xfe;
471 if (wc & FWW_NW_PROTO) {
474 flow->tos_frag &= wildcards->tos_frag_mask;
475 if (wc & FWW_ARP_SHA) {
476 memset(flow->arp_sha, 0, sizeof flow->arp_sha);
478 if (wc & FWW_ARP_THA) {
479 memset(flow->arp_tha, 0, sizeof flow->arp_tha);
481 flow->ipv6_src = ipv6_addr_bitand(&flow->ipv6_src,
482 &wildcards->ipv6_src_mask);
483 flow->ipv6_dst = ipv6_addr_bitand(&flow->ipv6_dst,
484 &wildcards->ipv6_dst_mask);
485 if (wc & FWW_ND_TARGET) {
486 memset(&flow->nd_target, 0, sizeof flow->nd_target);
492 flow_to_string(const struct flow *flow)
494 struct ds ds = DS_EMPTY_INITIALIZER;
495 flow_format(&ds, flow);
500 flow_format(struct ds *ds, const struct flow *flow)
504 ds_put_format(ds, "priority%"PRIu32
508 ntohll(flow->tun_id),
511 ds_put_format(ds, ":tci(");
512 if (flow->vlan_tci) {
513 ds_put_format(ds, "vlan%"PRIu16",pcp%d",
514 vlan_tci_to_vid(flow->vlan_tci),
515 vlan_tci_to_pcp(flow->vlan_tci));
517 ds_put_char(ds, '0');
519 ds_put_format(ds, ") mac"ETH_ADDR_FMT"->"ETH_ADDR_FMT
521 ETH_ADDR_ARGS(flow->dl_src),
522 ETH_ADDR_ARGS(flow->dl_dst),
523 ntohs(flow->dl_type));
525 if (flow->dl_type == htons(ETH_TYPE_IPV6)) {
526 ds_put_format(ds, " proto%"PRIu8" tos%"PRIu8" ipv6",
527 flow->nw_proto, flow->tos_frag & IP_DSCP_MASK);
528 print_ipv6_addr(ds, &flow->ipv6_src);
529 ds_put_cstr(ds, "->");
530 print_ipv6_addr(ds, &flow->ipv6_dst);
533 ds_put_format(ds, " proto%"PRIu8
535 " ip"IP_FMT"->"IP_FMT,
537 flow->tos_frag & IP_DSCP_MASK,
538 IP_ARGS(&flow->nw_src),
539 IP_ARGS(&flow->nw_dst));
541 frag = flow->tos_frag & FLOW_FRAG_MASK;
543 ds_put_format(ds, " frag(%s)",
544 frag == FLOW_FRAG_ANY ? "first"
545 : frag == (FLOW_FRAG_ANY | FLOW_FRAG_LATER) ? "later"
548 if (flow->tp_src || flow->tp_dst) {
549 ds_put_format(ds, " port%"PRIu16"->%"PRIu16,
550 ntohs(flow->tp_src), ntohs(flow->tp_dst));
552 if (!eth_addr_is_zero(flow->arp_sha) || !eth_addr_is_zero(flow->arp_tha)) {
553 ds_put_format(ds, " arp_ha"ETH_ADDR_FMT"->"ETH_ADDR_FMT,
554 ETH_ADDR_ARGS(flow->arp_sha),
555 ETH_ADDR_ARGS(flow->arp_tha));
560 flow_print(FILE *stream, const struct flow *flow)
562 char *s = flow_to_string(flow);
567 /* flow_wildcards functions. */
569 /* Initializes 'wc' as a set of wildcards that matches every packet. */
571 flow_wildcards_init_catchall(struct flow_wildcards *wc)
573 BUILD_ASSERT_DECL(FLOW_WC_SEQ == 3);
575 wc->wildcards = FWW_ALL;
576 wc->tun_id_mask = htonll(0);
577 wc->nw_src_mask = htonl(0);
578 wc->nw_dst_mask = htonl(0);
579 wc->ipv6_src_mask = in6addr_any;
580 wc->ipv6_dst_mask = in6addr_any;
581 memset(wc->reg_masks, 0, sizeof wc->reg_masks);
582 wc->vlan_tci_mask = htons(0);
583 wc->tos_frag_mask = 0;
584 memset(wc->zeros, 0, sizeof wc->zeros);
587 /* Initializes 'wc' as an exact-match set of wildcards; that is, 'wc' does not
588 * wildcard any bits or fields. */
590 flow_wildcards_init_exact(struct flow_wildcards *wc)
592 BUILD_ASSERT_DECL(FLOW_WC_SEQ == 3);
595 wc->tun_id_mask = htonll(UINT64_MAX);
596 wc->nw_src_mask = htonl(UINT32_MAX);
597 wc->nw_dst_mask = htonl(UINT32_MAX);
598 wc->ipv6_src_mask = in6addr_exact;
599 wc->ipv6_dst_mask = in6addr_exact;
600 memset(wc->reg_masks, 0xff, sizeof wc->reg_masks);
601 wc->vlan_tci_mask = htons(UINT16_MAX);
602 wc->tos_frag_mask = UINT8_MAX;
603 memset(wc->zeros, 0, sizeof wc->zeros);
606 /* Returns true if 'wc' is exact-match, false if 'wc' wildcards any bits or
609 flow_wildcards_is_exact(const struct flow_wildcards *wc)
613 BUILD_ASSERT_DECL(FLOW_WC_SEQ == 3);
616 || wc->tun_id_mask != htonll(UINT64_MAX)
617 || wc->nw_src_mask != htonl(UINT32_MAX)
618 || wc->nw_dst_mask != htonl(UINT32_MAX)
619 || wc->vlan_tci_mask != htons(UINT16_MAX)
620 || !ipv6_mask_is_exact(&wc->ipv6_src_mask)
621 || !ipv6_mask_is_exact(&wc->ipv6_dst_mask)
622 || wc->tos_frag_mask != UINT8_MAX) {
626 for (i = 0; i < FLOW_N_REGS; i++) {
627 if (wc->reg_masks[i] != UINT32_MAX) {
635 /* Returns true if 'wc' matches every packet, false if 'wc' fixes any bits or
638 flow_wildcards_is_catchall(const struct flow_wildcards *wc)
642 BUILD_ASSERT_DECL(FLOW_WC_SEQ == 3);
644 if (wc->wildcards != FWW_ALL
645 || wc->tun_id_mask != htonll(0)
646 || wc->nw_src_mask != htonl(0)
647 || wc->nw_dst_mask != htonl(0)
648 || wc->vlan_tci_mask != htons(0)
649 || !ipv6_mask_is_any(&wc->ipv6_src_mask)
650 || !ipv6_mask_is_any(&wc->ipv6_dst_mask)
651 || wc->tos_frag_mask != 0) {
655 for (i = 0; i < FLOW_N_REGS; i++) {
656 if (wc->reg_masks[i] != 0) {
664 /* Initializes 'dst' as the combination of wildcards in 'src1' and 'src2'.
665 * That is, a bit or a field is wildcarded in 'dst' if it is wildcarded in
666 * 'src1' or 'src2' or both. */
668 flow_wildcards_combine(struct flow_wildcards *dst,
669 const struct flow_wildcards *src1,
670 const struct flow_wildcards *src2)
674 dst->wildcards = src1->wildcards | src2->wildcards;
675 dst->tun_id_mask = src1->tun_id_mask & src2->tun_id_mask;
676 dst->nw_src_mask = src1->nw_src_mask & src2->nw_src_mask;
677 dst->nw_dst_mask = src1->nw_dst_mask & src2->nw_dst_mask;
678 dst->ipv6_src_mask = ipv6_addr_bitand(&src1->ipv6_src_mask,
679 &src2->ipv6_src_mask);
680 dst->ipv6_dst_mask = ipv6_addr_bitand(&src1->ipv6_dst_mask,
681 &src2->ipv6_dst_mask);
682 for (i = 0; i < FLOW_N_REGS; i++) {
683 dst->reg_masks[i] = src1->reg_masks[i] & src2->reg_masks[i];
685 dst->vlan_tci_mask = src1->vlan_tci_mask & src2->vlan_tci_mask;
688 /* Returns a hash of the wildcards in 'wc'. */
690 flow_wildcards_hash(const struct flow_wildcards *wc, uint32_t basis)
692 /* If you change struct flow_wildcards and thereby trigger this
693 * assertion, please check that the new struct flow_wildcards has no holes
694 * in it before you update the assertion. */
695 BUILD_ASSERT_DECL(sizeof *wc == 60 + FLOW_N_REGS * 4);
696 return hash_bytes(wc, sizeof *wc, basis);
699 /* Returns true if 'a' and 'b' represent the same wildcards, false if they are
702 flow_wildcards_equal(const struct flow_wildcards *a,
703 const struct flow_wildcards *b)
707 if (a->wildcards != b->wildcards
708 || a->tun_id_mask != b->tun_id_mask
709 || a->nw_src_mask != b->nw_src_mask
710 || a->nw_dst_mask != b->nw_dst_mask
711 || a->vlan_tci_mask != b->vlan_tci_mask
712 || !ipv6_addr_equals(&a->ipv6_src_mask, &b->ipv6_src_mask)
713 || !ipv6_addr_equals(&a->ipv6_dst_mask, &b->ipv6_dst_mask)) {
717 for (i = 0; i < FLOW_N_REGS; i++) {
718 if (a->reg_masks[i] != b->reg_masks[i]) {
726 /* Returns true if at least one bit or field is wildcarded in 'a' but not in
727 * 'b', false otherwise. */
729 flow_wildcards_has_extra(const struct flow_wildcards *a,
730 const struct flow_wildcards *b)
733 struct in6_addr ipv6_masked;
735 for (i = 0; i < FLOW_N_REGS; i++) {
736 if ((a->reg_masks[i] & b->reg_masks[i]) != b->reg_masks[i]) {
741 ipv6_masked = ipv6_addr_bitand(&a->ipv6_src_mask, &b->ipv6_src_mask);
742 if (!ipv6_addr_equals(&ipv6_masked, &b->ipv6_src_mask)) {
746 ipv6_masked = ipv6_addr_bitand(&a->ipv6_dst_mask, &b->ipv6_dst_mask);
747 if (!ipv6_addr_equals(&ipv6_masked, &b->ipv6_dst_mask)) {
751 return (a->wildcards & ~b->wildcards
752 || (a->tun_id_mask & b->tun_id_mask) != b->tun_id_mask
753 || (a->nw_src_mask & b->nw_src_mask) != b->nw_src_mask
754 || (a->nw_dst_mask & b->nw_dst_mask) != b->nw_dst_mask
755 || (a->vlan_tci_mask & b->vlan_tci_mask) != b->vlan_tci_mask);
759 set_nw_mask(ovs_be32 *maskp, ovs_be32 mask)
761 if (ip_is_cidr(mask)) {
769 /* Sets the IP (or ARP) source wildcard mask to CIDR 'mask' (consisting of N
770 * high-order 1-bit and 32-N low-order 0-bits). Returns true if successful,
771 * false if 'mask' is not a CIDR mask. */
773 flow_wildcards_set_nw_src_mask(struct flow_wildcards *wc, ovs_be32 mask)
775 return set_nw_mask(&wc->nw_src_mask, mask);
778 /* Sets the IP (or ARP) destination wildcard mask to CIDR 'mask' (consisting of
779 * N high-order 1-bit and 32-N low-order 0-bits). Returns true if successful,
780 * false if 'mask' is not a CIDR mask. */
782 flow_wildcards_set_nw_dst_mask(struct flow_wildcards *wc, ovs_be32 mask)
784 return set_nw_mask(&wc->nw_dst_mask, mask);
788 set_ipv6_mask(struct in6_addr *maskp, const struct in6_addr *mask)
790 if (ipv6_is_cidr(mask)) {
798 /* Sets the IPv6 source wildcard mask to CIDR 'mask' (consisting of N
799 * high-order 1-bit and 128-N low-order 0-bits). Returns true if successful,
800 * false if 'mask' is not a CIDR mask. */
802 flow_wildcards_set_ipv6_src_mask(struct flow_wildcards *wc,
803 const struct in6_addr *mask)
805 return set_ipv6_mask(&wc->ipv6_src_mask, mask);
808 /* Sets the IPv6 destination wildcard mask to CIDR 'mask' (consisting of
809 * N high-order 1-bit and 128-N low-order 0-bits). Returns true if
810 * successful, false if 'mask' is not a CIDR mask. */
812 flow_wildcards_set_ipv6_dst_mask(struct flow_wildcards *wc,
813 const struct in6_addr *mask)
815 return set_ipv6_mask(&wc->ipv6_dst_mask, mask);
818 /* Sets the wildcard mask for register 'idx' in 'wc' to 'mask'.
819 * (A 0-bit indicates a wildcard bit.) */
821 flow_wildcards_set_reg_mask(struct flow_wildcards *wc, int idx, uint32_t mask)
823 wc->reg_masks[idx] = mask;
826 /* Returns the wildcard bitmask for the Ethernet destination address
827 * that 'wc' specifies. The bitmask has a 0 in each bit that is wildcarded
828 * and a 1 in each bit that must match. */
830 flow_wildcards_to_dl_dst_mask(flow_wildcards_t wc)
832 static const uint8_t no_wild[] = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff};
833 static const uint8_t addr_wild[] = {0x01, 0x00, 0x00, 0x00, 0x00, 0x00};
834 static const uint8_t mcast_wild[] = {0xfe, 0xff, 0xff, 0xff, 0xff, 0xff};
835 static const uint8_t all_wild[] = {0x00, 0x00, 0x00, 0x00, 0x00, 0x00};
837 switch (wc & (FWW_DL_DST | FWW_ETH_MCAST)) {
838 case 0: return no_wild;
839 case FWW_DL_DST: return addr_wild;
840 case FWW_ETH_MCAST: return mcast_wild;
841 case FWW_DL_DST | FWW_ETH_MCAST: return all_wild;
846 /* Returns true if 'mask' is a valid wildcard bitmask for the Ethernet
847 * destination address. Valid bitmasks are either all-bits-0 or all-bits-1,
848 * except that the multicast bit may differ from the rest of the bits. So,
849 * there are four possible valid bitmasks:
851 * - 00:00:00:00:00:00
852 * - 01:00:00:00:00:00
853 * - fe:ff:ff:ff:ff:ff
854 * - ff:ff:ff:ff:ff:ff
856 * All other bitmasks are invalid. */
858 flow_wildcards_is_dl_dst_mask_valid(const uint8_t mask[ETH_ADDR_LEN])
863 return (mask[1] | mask[2] | mask[3] | mask[4] | mask[5]) == 0x00;
867 return (mask[1] & mask[2] & mask[3] & mask[4] & mask[5]) == 0xff;
874 /* Returns 'wc' with the FWW_DL_DST and FWW_ETH_MCAST bits modified
875 * appropriately to match 'mask'.
877 * This function will assert-fail if 'mask' is invalid. Only 'mask' values
878 * accepted by flow_wildcards_is_dl_dst_mask_valid() are allowed. */
880 flow_wildcards_set_dl_dst_mask(flow_wildcards_t wc,
881 const uint8_t mask[ETH_ADDR_LEN])
883 assert(flow_wildcards_is_dl_dst_mask_valid(mask));
887 return wc | FWW_DL_DST | FWW_ETH_MCAST;
890 return (wc | FWW_DL_DST) & ~FWW_ETH_MCAST;
893 return (wc & ~FWW_DL_DST) | FWW_ETH_MCAST;
896 return wc & ~(FWW_DL_DST | FWW_ETH_MCAST);
903 /* Hashes 'flow' based on its L2 through L4 protocol information. */
905 flow_hash_symmetric_l4(const struct flow *flow, uint32_t basis)
910 struct in6_addr ipv6_addr;
915 uint8_t eth_addr[ETH_ADDR_LEN];
921 memset(&fields, 0, sizeof fields);
922 for (i = 0; i < ETH_ADDR_LEN; i++) {
923 fields.eth_addr[i] = flow->dl_src[i] ^ flow->dl_dst[i];
925 fields.vlan_tci = flow->vlan_tci & htons(VLAN_VID_MASK);
926 fields.eth_type = flow->dl_type;
928 /* UDP source and destination port are not taken into account because they
929 * will not necessarily be symmetric in a bidirectional flow. */
930 if (fields.eth_type == htons(ETH_TYPE_IP)) {
931 fields.ipv4_addr = flow->nw_src ^ flow->nw_dst;
932 fields.ip_proto = flow->nw_proto;
933 if (fields.ip_proto == IPPROTO_TCP) {
934 fields.tp_addr = flow->tp_src ^ flow->tp_dst;
936 } else if (fields.eth_type == htons(ETH_TYPE_IPV6)) {
937 const uint8_t *a = &flow->ipv6_src.s6_addr[0];
938 const uint8_t *b = &flow->ipv6_dst.s6_addr[0];
939 uint8_t *ipv6_addr = &fields.ipv6_addr.s6_addr[0];
941 for (i=0; i<16; i++) {
942 ipv6_addr[i] = a[i] ^ b[i];
944 fields.ip_proto = flow->nw_proto;
945 if (fields.ip_proto == IPPROTO_TCP) {
946 fields.tp_addr = flow->tp_src ^ flow->tp_dst;
949 return hash_bytes(&fields, sizeof fields, basis);
952 /* Hashes the portions of 'flow' designated by 'fields'. */
954 flow_hash_fields(const struct flow *flow, enum nx_hash_fields fields,
959 case NX_HASH_FIELDS_ETH_SRC:
960 return hash_bytes(flow->dl_src, sizeof flow->dl_src, basis);
962 case NX_HASH_FIELDS_SYMMETRIC_L4:
963 return flow_hash_symmetric_l4(flow, basis);
969 /* Returns a string representation of 'fields'. */
971 flow_hash_fields_to_str(enum nx_hash_fields fields)
974 case NX_HASH_FIELDS_ETH_SRC: return "eth_src";
975 case NX_HASH_FIELDS_SYMMETRIC_L4: return "symmetric_l4";
976 default: return "<unknown>";
980 /* Returns true if the value of 'fields' is supported. Otherwise false. */
982 flow_hash_fields_valid(enum nx_hash_fields fields)
984 return fields == NX_HASH_FIELDS_ETH_SRC
985 || fields == NX_HASH_FIELDS_SYMMETRIC_L4;
988 /* Puts into 'b' a packet that flow_extract() would parse as having the given
991 * (This is useful only for testing, obviously, and the packet isn't really
992 * valid. It hasn't got any checksums filled in, for one, and lots of fields
993 * are just zeroed.) */
995 flow_compose(struct ofpbuf *b, const struct flow *flow)
997 eth_compose(b, flow->dl_dst, flow->dl_src, ntohs(flow->dl_type), 0);
998 if (flow->dl_type == htons(FLOW_DL_TYPE_NONE)) {
999 struct eth_header *eth = b->l2;
1000 eth->eth_type = htons(b->size);
1004 if (flow->vlan_tci & htons(VLAN_CFI)) {
1005 eth_push_vlan(b, flow->vlan_tci & ~htons(VLAN_CFI));
1008 if (flow->dl_type == htons(ETH_TYPE_IP)) {
1009 struct ip_header *ip;
1011 b->l3 = ip = ofpbuf_put_zeros(b, sizeof *ip);
1012 ip->ip_ihl_ver = IP_IHL_VER(5, 4);
1013 ip->ip_tos = flow->tos_frag & IP_DSCP_MASK;
1014 ip->ip_proto = flow->nw_proto;
1015 ip->ip_src = flow->nw_src;
1016 ip->ip_dst = flow->nw_dst;
1018 if (flow->tos_frag & FLOW_FRAG_ANY) {
1019 ip->ip_frag_off |= htons(IP_MORE_FRAGMENTS);
1020 if (flow->tos_frag & FLOW_FRAG_LATER) {
1021 ip->ip_frag_off |= htons(100);
1024 if (!(flow->tos_frag & FLOW_FRAG_ANY)
1025 || !(flow->tos_frag & FLOW_FRAG_LATER)) {
1026 if (flow->nw_proto == IPPROTO_TCP) {
1027 struct tcp_header *tcp;
1029 b->l4 = tcp = ofpbuf_put_zeros(b, sizeof *tcp);
1030 tcp->tcp_src = flow->tp_src;
1031 tcp->tcp_dst = flow->tp_dst;
1032 } else if (flow->nw_proto == IPPROTO_UDP) {
1033 struct udp_header *udp;
1035 b->l4 = udp = ofpbuf_put_zeros(b, sizeof *udp);
1036 udp->udp_src = flow->tp_src;
1037 udp->udp_dst = flow->tp_dst;
1038 } else if (flow->nw_proto == IPPROTO_ICMP) {
1039 struct icmp_header *icmp;
1041 b->l4 = icmp = ofpbuf_put_zeros(b, sizeof *icmp);
1042 icmp->icmp_type = ntohs(flow->tp_src);
1043 icmp->icmp_code = ntohs(flow->tp_dst);
1046 } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) {
1048 } else if (flow->dl_type == htons(ETH_TYPE_ARP)) {
1049 struct arp_eth_header *arp;
1051 b->l3 = arp = ofpbuf_put_zeros(b, sizeof *arp);
1052 arp->ar_hrd = htons(1);
1053 arp->ar_pro = htons(ETH_TYPE_IP);
1054 arp->ar_hln = ETH_ADDR_LEN;
1056 arp->ar_op = htons(flow->nw_proto);
1058 if (flow->nw_proto == ARP_OP_REQUEST ||
1059 flow->nw_proto == ARP_OP_REPLY) {
1060 arp->ar_spa = flow->nw_src;
1061 arp->ar_tpa = flow->nw_dst;
1062 memcpy(arp->ar_sha, flow->arp_sha, ETH_ADDR_LEN);
1063 memcpy(arp->ar_tha, flow->arp_tha, ETH_ADDR_LEN);